A Reader on the Sanskrit Grammarians

Studies in Linguistics edited by Samuel Jay Keyser

1. A Reader on the Sansk J. F. Staal, editor

A Reader on the Sanskrit Grammarians

Edited by J. F. Staal

The MIT Press

Cambridge, Massachusetts, and London, England

Copyright 1972 by The Massachusetts Institute of Technology This book was designed by the MIT Press Design Department. It was set in Monotype Gill Sans and printed on Finch Textbook Offset by Halliday Lithograph Corporation and bound in Whitman POC Gold by Halliday Lithograph Corporation in the United States of America. All rights reserved. No part of this book may be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without permission in writing from the publisher. Library of Congress Cataloging in Publication Data Staal, Johan Frederik comp. A reader on the Sanskrit grammarians. Bibliography: p. I. Sanskrit philologyCollections. 2. Scholars, SanskritCollections. I. Title. PK402.S8 49l.2'08 77-122265 ISBN 0-262-19078-8 (hardcover)

anantaparam kila sabdasastram svalpam tathyur bahavas ca vighnh/ sram tato grhyam apsya phalgu hamsair yath ksJram ivmbumadhytH Boundless Indeed is the science of language, but life is short and obstacles are numerous. Hence take what is good and leave what is worthless, as geese take milk from the midst of water.

Preface

xi xvi xviii

Notes on the Frontispiece Note on Sanskrit Transliteration The Sanskrit Alphabet Abbreviations xx xix

Table of the Sanskrit Grammarians

xxiv

1. Hsan Tsang (602-664)

4 6 7

A. Records of the Western Countries (Seventh Century)

Early Accounts
1

B. Hwui Li : The Life of Hsan Tsang (Seventh Century) 2. I Tsing (634-713) 11 12

The Method of Learning in the West (691-692) 3. Fa Tsang (643-712) 18

Account of Exploring the Mysteries of the Avatamsaka Stra (Seventh Century) 19 4. Ab Raihnal-Brni (973-1048) 20

An Accurate Description of All Categories of Hindu Thought (1030) 22 5. Trantha (born 1573) 23 24 24

A. Geschichte des Buddhismus in Indien (1608): I B. Geschichtedes Buddhismus in Indien (1608): II

2
The Foundations of Western Scholarship 27

6. Jean Franois Pons (1698-1752)

30

Richesse et nergie de la langue Samskret, et comment et par qui elle a t rduite en Grammaire (1740) 31 7. Henry Thomas Colebrooke (1765-1837) On the Sanskrit and Prcrt Languages (1803) 33 35

3
The Romantic Period 47

8. August Wilhelm von Schlegel (1767-1845)

49

A. Ueber den gegenwrtigen Zustand der indischen Ph ilologie (1819) 51 B. Rflexions sur l'tude des langues asiatiques (1832) C. Controversy with H. H. Wilson 9. Wilhelm von Humboldt (1767-1835) 57 59 52

A. ber die Verschiedenheit des menschlichen Sprachbaues (1836) : I 60 B. ber die Verschiedenheit des menschlichen Sprachbaues (1836): II 61

4
The Golden Days 65

10. RamkrishnaGopal Bhandarkar (1837-1925) A. Review of Goldstcker's Pn/n/ (1864,1877)

70 72

B. On the Date of Patajali and the King in Whose Reign He Lived (1872) 78 C. crya, the Friend of the Student, and the Relations between the Three cr/as (1876) 81 D. Development of Language and of Sanskrit, First Wilson Lecture (1883) 87 E. Relations between Sanskrit, Pali, the Prkrts, and the Modern Vernaculars, Seventh Wilson Lecture (1883) 94 11. Franz Kiel horn (1840-1908) A. Der Grammatiker Pnini (1885) 102 103

B. The Authorities on Grammar Quoted in the Mahbhsya (1887) 106 C. The Text of Pnini's Stras, as Given in the Ksik-Vrtti, Compared with the Text Known to Ktyyanaand Patajali (1887) 115 D. Some Devices of Indian Grammarians (1887)

123

5
The Skeptics and Their Critics
135

12. William Dwight Whitney (1827-1894)

138

A. The Study of Hindu Grammar and the Study of Sanskrit (1884) 142 B. Franz Kielhorn : Book Notice of Whitney's Roots, Verb-forms and Primary Derivatives of the Sanskrit Language (1886) 155 13. Bruno Liebich (1862-1939) 158

A. Prfung der Argumente Prof. Whitney's (1891)

159

l. William Dwight Whitney: On Recent Studies in Hindu Grammar (1893) 165 14. Otto Boehtngk (1815-1904) 185

Whitney's letzte Angriffe auf Pnini (1893) 15. Georg Bhler (1837-1898) 193

186

The Roots of the Dhatupatha Not Found in Literature (1894) 194

6
The Transition 205

16. Bernhard Geiger (1881-1964)

207

Mahabhasya zu P. VI, 4, 22 und 132 nebst Kaiyata's Kommentar (1909) ' 209

7
The

17. Leonard Bloomfield (1887-1949) On Some Rules of Pnini (1927) 18. Barend Faddegon (1874-1955) 266

264

273

Modern
Period
261

The Mnemotechnics of Pnini's Grammar I : The Siva-Stra (1929) 275 19. Kshitish Chandra Chatterji (1896-1961) The Critics of Sanskrit Grammar (1934) 287 286

20. Paul Thieme (born 1905)

298

A. Bhsyazu vrttika 5 zu Pnini 1.1.9 und seine einheimischen Erklrer (1935) 299* B. On the Identity ofthe Vrttikakra (1937-1938) 21. Pierre Boudon 357 332

Une Application du raisonnement par l'absurde dans l'interprtation de Pnini (les jnpakasiddhoparibhs) (1938) 358 22. K. A. Subramania Iyer (born 1896) The Point of View ofthe Vaiykaranas (1948) 23. John Brough (born 1917) 401 392 393

A. Theories of General Linguistics in the Sanskrit Grammarians (1951) 402 B. Some Indian Theories of Meaning (1953) 24. YutakaOjihara (born 1923) 424 425 414

Jti 'genus' et deux dfinitions pr-patajaliennes (1967) 25. Louis Renou (1896-1966) 432

A. Les connexions entre le rituel et lagrammaire en sanskrit (1941-1942) 435 B. Grammaire et Vdanta (1957) 470 478

C. Grammaire et potique en sanskrit (1961)

D. La thorie des temps du verbe d'aprs les grammairiens sanskrits (1960) 500 Bibliography Index of Names 526 539 544

Index of Sanskrit Terms Index of Sutras 550

Plates

Frontispiece: Patajali

xvii 3

I The sages Jaimini, Vyghrapda, and Patajali II A bronze image of Patajali 29

III A page from a manuscript in the India Office, London, of Pnini's AstdhyyJ 68 IV A page from the text of Patajali's commentary on Pnini's grammar, Mahbhsya, with Kaiyata's subcommentary. Bhasyapradpa 137 V A page from a manuscript in the India Office, London, of Patajali's Mahbhsya with Kaiyata's subcommentary
Bhasyapradpa 263

Preface

The study of the Sanskrit grammarians, and in particular of Pnini, has been regarded for almost a century as a hyperspecialist occupation. Only the specialists of vykarana ('grammar') among specialists of Sanskrit philology were considered capable of arriving at an understanding of the difficult texts of the Indian grammatical tradition. Pnini was accordingly treated even by linguists as an object of Indological investigation, not as a deceased colleague of great genius. This view has not always been predominant, however, and today, for reasons that will be sketched here, it is once again beginning to change. It is now generally recognized that Pnini, despite his exclusive preoccupation with Sanskrit, was the greatest linguist of antiquity, if not of all time, and deserves to be treated as such. Accordingly, linguists, dissatisfied with mere lip service, are beginning to turn to him and to the Sanskrit grammarians, just as logicians turn to Aristotle. But the difficulties awaiting them are still formidable. Pnini's insights about language are disguised in his metalinguistic analysis of Sanskrit in Sanskrit, which presupposes some familiarity with the complexities of the Sanskrit language. Also, the arguments and patterns of reasoning of Pnini, Patajali, and many later grammarians are no less compact, precise, and profound than those of Aristotle. In addition, neither is Sanskrit as familiar as Greek, nor are notions in contemporary linguistics derived from Pnini in the way many notions in modern logic are derived from Aristotle. The text of Pnini's grammar, the Astdhyy, is accessible in Boehtlingk's edition and translation of 1887, reprinted in 1964. There are at least two other translations, one by Renou of 19481954 and one by S. C. Vasu of 1891, reprinted in 1962. However, all of these are difficult to use without the help of a competent guide. In addition, there is not available at present an introduction to Pnini that is adequate and up-to-date. The two most important existing monographs are one by Goldstcker of 1861 (reprinted in India and in Germany during recent years) and one by Liebich of 1891 (one chapter from the latter is reproduced in the present volume). Many important articles do exist, but they have often been published in out-of-the-way places and they are difficult to find even in good libraries. Accordingly, it is difficult not only for linguists but also for Sanskritists to acquaint themselves adequately with Pnini and the Indian grammarians. One of the aims of this book is to render the work of the Sanskrit grammarians more easily accessible. In the absence of a helpful introduction to the subject, there is no better way to achieve a first understanding than through studying some of the best articles that have appeared during the last one and a half centuries. However, the following selection may pave the way for, and in due course supplement, a new general introduction to the study of the Sanskrit grammarians, which is presently under preparation (Joshi, Kiparsky, and Staal). The second aim of the book is to throw some light on an interesting line of development within Western linguistics itself. In doing so it may also dispel two widespread but erroneous notions. The first is that the Sanskrit grammarians have in the past been studied by philologists but neglected by linguists. The second, turning the previous supposition into a principle of research, is that the materials have first to be made available and interpreted by philologists before they can be evaluated by linguists. The circularity of such a principle is apparent. We shall see how, on the

x Preface

contrary, the results of philological research have often been colored by underlying linguistic views. In India itself, the grammatical t r a d i t i o n , which is largely but not exclusively concerned w i t h Pnini and his school, has been revived several times and continues t o the present day. Its foremost living authorities are the traditional grammarians or pandits of vykarana, whose knowledge is handed d o w n almost entirely in Sanskrit. Such Indian scholars trained the W e s t e r n Sanskritists w h o first devoted close attention t o the Indian grammarians. W i t h o u t t h e i r assistance the original sources w o u l d have remained unintelligible. The history of the study of the Sanskrit grammarians f r o m outside the t r a d i t i o n falls into several periods, t o which the chapters of this volume correspond. The earliest information was provided by Buddhist pilgrims f r o m China and Tibet, by Muslim travelers f r o m the Near East, and by Christian missionaries f r o m Europe. The foundations of Sanskrit studies in the W e s t was laid by British scholars w h o w o r k e d in India, often as administrators and civil servants, and w h o were assisted by Indian pandits. Colebrooke, foremost among t h e m , provided detailed and accurate information on the Sanskrit grammarians. This was followed on the C o n t i n e n t and especially in Germany by a wave of enthusiasm for ancient Indian c u l t u r e , resulting in linguistic evaluations of Pnini by A. W . von Schlegel and W . von H u m b o l d t . For some t i m e , discussions on the Sanskrit grammarians continued t o be part and parcel of general linguistics. This development, combined w i t h renewed contacts w i t h the living grammatical t r a d i t i o n in India, led t o a deeper understanding of the Sanskrit grammarians, culminating, less than a century ago, in the w o r k of Bhandarkar and especially of K i e l h o r n . A relapse f o l l o w e d , w i t h a nadir in W h i t n e y , resulting in a general decrease of understanding and an accordingly low evaluation of the " n a t i v e " grammarians. But the t r a d i t i o n founded by Kielhorn was not entirely destroyed, and it led eventually t o the modern period of rather specialized philological w o r k t o which W e s t e r n , Indian, and Japanese scholars c o n t r i b u t e . C o n t e m p o r a r y scholarship w i t h i n this philological perspective continues t o produce valuable results that are new t o W e s t e r n scholars though often perfectly well k n o w n t o Indian pandits. A l l the same, there is a recent revival of interest in the Sanskrit grammarians which is partly linked w i t h the e x t r a o r d i n a r y developments that have taken place during the last decade and a half in W e s t e r n linguistics. This new and much w i d e r interest is marked by the participation of linguists as well as Sanskritistsand constitutes in this respect a kind of r e t u r n . But now there are at least t w o totally new reasons for such a revival. First, the activities of the Indian grammarians are the closest parallel in history t o contemporary linguistics. Second, it appears that only contemporary insights can pave the way for adequate interpretations of the w o r k of the Sanskrit grammarians themselves. There are many parallels for such a state of affairs in o t h e r domains where progress has not been linear, for example, in the history of philosophy o r logic. It is indeed not at all surprising that Pnini and his colleagues cannot be properly understood unless they are taken seriously for what they were, that is, linguists. This change of perspective has definite consequences, for it is an illusion t o imagine that it is possible t o do justice t o the Sanskrit grammarians as linguists w i t h o u t considering the linguistic problems w i t h which they t h e m selves were coming t o grips. It is at the same t i m e clear, however,

XIII

Preface

that this contemporary approach should not lead t o what might be called retrospective i n t e r p r e t a t i o n . In particular, it remains i m p o r t a n t that the valuable insights and results of earlier generations of scholars be preserved. This book, t h e n , sets out t o perform part of a task that in the wake of new discoveries seems of some urgency. The articles and fragments reproduced here are selected mainly in view of t h e i r analytic and linguistic relevance, excluding, on the one hand, purely historical and philological articles and, on the o t h e r hand, papers that deal w i t h the i n t e r p r e t a t i o n of particular rules in Panini's grammar w i t h o u t elucidating problems of general interest. These considerations have sometimes been overruled by the desire t o provide selections f r o m a variety of viewpoints. This w o r k seeks t o go beyond a mere collection of articles by providing a connected historical outline, especially w i t h regard t o the earlier periods. The i n t r o d u c t o r y notes preceding the selections include information about related w o r k , contain excerpts f r o m o t h e r sources, and explain particular points of a more technical nature. They sometimes give additional references t o later publications treating the same or similar topics o r exhibiting a similar approach. The selections included here are in English, German, and French. A few passages quoted in Latin and Greek have been provided w i t h an English translation. For Chinese, Tibetan, and Arabic texts, use has been made of existing translations. For Sanskrit terms and passages, a translation has been generally supplied w h e n ever it w o u l d be difficult t o follow the argument otherwise. But Sanskrit expressions treated as elements of the object-language, and quoted on account of certain formal features, have not been translated. Similarly, stras and linguistic rules are often left untranslated when their function in the context is clear and when a mere translation w o u l d be intelligible only when in t u r n accompanied by a fuller analysis. Editor's translations and explications are generally placed w i t h i n square brackets. The transliteration f r o m Sanskrit i n t o Roman script has had many vicissitudes since the beginning of W e s t e r n scholarship, and quotations in the articles here reproduced had t o be modified accordingly. The original transliteration has been retained only where it was of some independent historical interest, as in the w o r k of Pons and Colebrooke. Elsewhere the transliteration has been unified t h r o u g h o u t . For more precise information in this respect, see the editor's N o t e on Sanskrit Transliteration. Unlike most linguistic books, where illustrations are confined t o such things as sound waves or the vocal chords, this volume has been illustrated w i t h a variety of plates, emphasizing that "homo fonticas indicus was no mere cross-sectioned larynx sited under an empty cranium . . . ; on the contrary, the whole man, belly, heart and head, produced v o i c e " (J. E. B. Gray 1959, 520, w i t h textual support f r o m the Mahbhsya and the Pninlyasiks). It is perhaps risky t o publish illustrations that may reinforce the t r e n d t o regard linguistics in India as an exotic g r o w t h , but the articles included provide enough material t o correct such tendencies. O t h e r features of the book are more self-explanatory. The list of abbreviations reflects the idiosyncrasies of a variety of authors, some changing t h e i r method of referring t o texts more than once even w i t h i n a single article. It may be added here that abbreviations consisting of three numerals (in various forms, for example, II, 3.5 or 2.3.5) always refer t o Panini's AstdhyyJ, specifying adhyya.

xv Preface

poda, and stra. The bibliography lists all publications referred to in the introductions and explanatory notes accompanying the selections. To have included the sources referred to in the selections themselves would have led to at least a fivefold increase. The motto that introduces this book occurs in the Pacatantra and in other collections and has here been quoted from Otto Boehtlingk's Indische Sprche, St. Petersburg, 1870-1873 (reprint Osnabrck, 1966), vol. I, p. 45, no. 243. Now there remains the pleasant task of thanking numerous persons and institutions for their help in preparing this book. I am first of all grateful to the authors who have permitted me to reprint their articles: Professors John Brough (Cambridge), Yutaka Ojihara (Kyoto), K. A. Subramania Iyer (Lucknow), and Paul Thieme (Tbingen). The authors of the majority of articles here reproduced are no longer living. But an attempt has been made to obtain permission in all cases for republication, even if the material has long been in the public domain. I am greatly indebted to the following persons, publishers, and institutions for their assistance and for their permission to reprint articles: Acta Orientalia; Akademie der Wissenschaften, Gttingen ; The Asiatic Society of Bombay, Bombay; Bhandarkar Oriental Research Institute, Poona; Calcutta University Press, Calcutta; Clarendon Press, Oxford; ditions E. de Boccard, Paris; H. Haessel Verlag, Frankfurt; Indian Research Institute, Calcutta; International Academy of Indian Culture, New Delhi; Japanese Association of Indian and Buddhist Studies, Tokyo; Johann Ambrosius Barth, Leipzig; The Johns Hopkins Press, Baltimore, Maryland ; Johnson Reprint Corporation, New York \ Journal of the American Oriental Society; Dr. S. M. Katre, M.A., Ph.D. (Lond.), Poona; Kuppuswami Sastri Research Institute, Madras; Lambert Schneider, Heidelberg; Longmans Green & Co. Ltd., Harlow; sterreichische Akademie der Wissenschaften, Vienna; Philological Society, Oxford ; Routledge & Kegan Paul, London; S. Hirzel Verlag, Stuttgart; Socit Asiatique, Paris; Walter de Gruyter & Co. Verlag, Berlin. I owe special gratitude to the following scholars, who have helped me with their encouragement, criticism, and suggestions, or who have answered specific queries: Professors George Cardona (Philadelphia), Murray B. Emeneau (Berkeley), Jean Filliozat (Paris), Morris Halle (Cambridge, Mass.), J. C. Heesterman (Leiden), S. M. Katre (Poona), Paul Kiparsky (Cambridge, Mass.), B. A. van Nooten (Berkeley), Gerhard Oberhammer (Vienna), and John W. M. Verhaar (Djakarta). The Theology Department, University of San Francisco, helped me by providing information on Jean Franois Pons. I am particularly indebted to Professors Kiparsky and van Nooten, who read an earlier version of the entire manuscript and provided me with a wealth of ideas and suggestions. I am also very grateful to those who have assisted me in various ways in connection with the illustrations: the Director General of the Archaeological Survey of India (New Delhi), Mr. J. E. B. Gray (London), Dr. J. C. Harle (Oxford), Professor V. Raghavan (Madras), Dr. R. Ramachandran (Annamalainagar), Dr. S. T. Satyamurti (Madras), the late Mr. J. M. Somasundaram Pillai (Annamalainagar), and Mr. C. Srinivasan (Madras). I wish in particular to thank my secretary, Mrs. Barbara Sinclair, without whose acumen and care this book might not have seen the light of day. In a book of this kind, replete with quotations, references, footnotes, technical terms, and diacritical marks, proofreading presents a formidable task. My gratitude goes to the University

xv Preface

of California, which provided me with research assistance during the proofreading, and to Mr. Tai Shung Paik for his sharp eye and expert assistance at this stage and for preparing the indexes. I also thank the staff of The MIT Press for its assistance at all times. J.F. S. Berkeley, California

N o t e s OH t h e Frontispiece: Patajali

Patajali the grammarian and author of the Mahbhsya is traditionally identified with Patajali the philosopher and yogi, author of the Yogastra. Whether the two Patajali's were really the same person has been a matter of much dispute. The date of the Yogastra, at any rate, is rather uncertain, and the name Patajali is fairly uncommon. J. H. Woods, the American translator of the Yogastra and its main commentaries, compared the treatment of the concept of substance in the Yogastra and in the Mahbhsya and, as a result of this study, doubted the identity of their authors (Woods 1914, xiii-xvii). Surendranath Dasgupta, one of the leading experts on Indian philosophy in general and on the Yoga system In particular, controverted Woods's conclusions and argued that the two were identical (Dasgupta 1922, 230-238). Several other scholars have contributed to the controversy (for a survey see Eliade 1954, 363-364), which, however, remains unresolved. This personage Patajali, to whom the Mahbhsya, the Yogastra, and also sometimes a medical work are attributed, is represented traditionally as a snake from the waist down and sometimes with a large hood composed of five serpents above his head (cf. this volume, pp. 25, 38). There are numerous legends connected with this figure, particularly in South India, where Patajali is regarded as one of the two sages to whom Siva manifested himself as Siva Natarja, namely, in his cosmic dance (the other sage being the more mythical Vyghrapda, "tiger foot"). According to one legend Patajali, who as a yogi did not wish to crush insects with his feet, obtained from Natarja the favor of being partially changed into a serpent. Another legend relates how Patajali fell (pat) into the hollow of the hand (ajali) of Pnini in the shape of a snake (Jouveau-Dubreuil 1937, 51-52). There is at least one shrine in South India especially connected with Patajali, the Anantsvara temple in Cidambaram (Somasundaram Pillai 1955,146). The temple has recently been restored and the image of Patajali installed there is of recent origin. There are images of Patajali in many Siva temples in the South, but rarely published or described (for Tehkasi, see Gopinatha Rao 1916, 255; for Sudndram, see Pilla/ 1953, 386-387). The most important Natarja temple in India is the large temple complex in Cidambaram, where the oldest and most interesting images of Patajali are found. The frontispiece pictures a stone image of the East gopura or temple gateway at Cidambaram (built around the middle of the thirteenth century during the Late Co|a period), where Patajali is represented on two large serpent coils, with fangs and with his hands in ajali pose with a beaded chaplet pressed between his palms (Harle 1963,121 ; illustrated in Krishna Sastri 1916, 85). This illustration is by courtesy of the Archaeological Survey of India.

Note on Sanskrit Transliteration

The selections reproduced in this book come from a great variety of sources, adopting a bewildering variety of transliterations of Sanskrit. Moreover, in many of the originals, especially those published in India, Sanskrit terms and passages were printed in the Ngar script (this applies to the articles by Bhandarkar, Kielhorn, Boehtlingk, Bhler, Chatterji, and Subramania Iyer). The oldest transliterations into Roman script, that is, those by Pons and Colebrooke, are interesting as attempts not only at transliteration but also at phonological analysis ; they have therefore been retained. Elsewhere the Ngar has been transliterated into Roman script, and all transliterations have been unified, adopting the system that is now in common use by Sanskritists all over the world. Sometimes the spelling used for a Sanskrit word has also had to be regularized. All this occasioned quite considerable changes in spelling, especially with regard to the older selections, where, for example, the retroflex fricative was written as sh instead of s, the palatal fricative as instead of s', anusvra as m instead of m, and where the vowel r and the diphthongs ai and au were written as ri and ai and u, respectively. Lengthening signs over vowels used to be the circumflex * instead of the horizontal bar ". Some modern authors use the circumflex only over vowels when they result from sandhi ; this practice has been retained in the one paper where it was adopted, that is, in Ojihara's paper. More idiosyncratic and confusing uses, such as, j for the liquid nowadays written y, had to be modified also. The lengthening signs used by Kielhorn over e and o have been dropped ; though the marked vowels may be phonetically long (as reflected in the practice of South Indian manuscripts written, for example, in Grantha script), the distinction between short and long e and o is not phonemic in Sanskrit. Among the later papers only Brough's transliteration has been changed, involving the replacement of the curly hooks that some British linguists were in the habit of attaching to letters by the customary diacritics. All Roman transliterations now in use go back to William Jones's " A Dissertation on the Orthography of Asiatick Words in Roman Letters" of 1786. As Sir M. Monier Monier-Williams told the Royal Asiatic Society in 1890: "as a result of a kind of natural selection or survival_of the fittest, the practice of all Oriental scholars so far as Aryan languages are concerned is settling down into an acceptance of Sir William Jones' principles of transliteration" (quoted by G. H. Cannon in Sebeok1966,1, 56). The transliteration adopted is given in the following table.

vela

glot

dem
t

pala

Sanskrit Alphabet
Consonants unas pi rated voiceless aspirated stops unas pi rated voiced aspirated

L.

"S

retr ofli

The

&
s.

k kh g gh

P
-q

^r
ch

Z
th

ph d b

~5

i
jh Y

d dh Z
n

nasals

liquids

voiceless fricatives voiced

h h

Vowels short a
;

"

long

Ml

f
e

3T 0

Abbreviations

A
AB

AiGr AKM APr,AthPrt,AVPr pS(S) Asv(GS) sv(SS) AV B(SS), Bau(SS) BAU Bau(GS) BB

Aitareyranyaka Aitareyabrhmana (Wackernagel's) Altindische

Grammatik Abhandlungen zur Kunde des Morgenlandes

Atharvaprtiskhya pastambasrauta(stra) svalyana(grhyastra) svalyana(srautastra) Atharvaveda Baudhyana(srautastra) Brhadranyakopanisad Baudhyana(grhyastra)

Bezzenberger's Beitrge zur Kunde der indogermanischen Sprachen

Bh, Bhs Bh(SS), Bhr($S) Bhg Bhm Bhar Bhsv Bhatt Biblind BM Br BR Brhaddev BSS C(V) ChU Dasak Dhanv Dhtuprad DhS Dra DSS, DrhS Durgh Ganaratn GB GGA GN GS H, Hern, Hemac Hel Hir(SS)

Mahbhsya Bhradvja(srautastra) Bhgavata Bhmaha Bharata Bhsvrtti Bhattoji Diksita

Bibliotheca Indica

Blamanoram Brhmana Boehtlingk-Roth SanskritWrterbuch

Brhaddevat Bombay Sanskrit Series Candra(vrtti) Chndogyopanisad Dasakumracarita Dhanvin Dhtupradipa Dharmastra Dravya(samuddesa) Drhyyanasrautastra Durghatavrtti Ganaratnamahodadhi Gopathabrhmana
Gttingische gelehrte Anzeigen Nachrichten von der Gesellschaft der Wissenschaften zu Gttingen

Gobhilagrhyastra Hemacandra Helrja Hiranyakesi(srautastra)

xxi Abbreviations

IA, IndAnt IHQ IndCu Ind Sprche ISt, IndStud J J, Jaim Ja JA, Jas JAOS JB J.Dep.Letters jap JRAS JS JUB K, Ks K Kai, Kaiy kr Kt, Kty Kaus KvMi

Indian Antiquary Indian Historical Quarterly Indian Culture (Boehtlingk's) Indische Sprche (Weber's) Indische Studien

Jainendra Jaimini, Jaiminya Jti(samuddesa)

Journal asiatique Journal of the American Oriental Society

Jaiminyabrahmana
Journal of the Department of Letters, University of Calcutta

jnpaka
Journal of the Royal Asiatic Society

KB
Kiel KP Kri

KS
Ksratar K$S, KtyS Kt Kuval

Jaimini's Mimmsstra Jaiminyopanisadbrahmana Ksikvrtti Klasamuddesa Kaiyata krik Ktyyana Kausika Kvyamimms Kausitakibrhmana Kielhorn Kvyapraksa Kriy(samuddesa) Kthakasarphit Ksratarangin Ktyyanasrautastra Ktantra Kuvalaynanda
Zeitschrift fr vergleichende Sprachwissenschaft (begrndet

KZ
LSS, LtS LSS, Laghus M, Mbh, Moh M(SG) MS MSL MS(S), MnS(S) N Nais Nr

von A. Kuhn) Ltyyanasrautastra Laghusabdendusekhara Mahbhsya Mimms(stra) Maitryanisamhit

Mmoires de la Socit Linguistique de Paris

Mnavasrauta(stra) Nysa Naisadhya Nrada

xx Abbreviations

Ng, Ng, Ngoj [\||A NidS Nigh Nir NM Nyyaprak Ny(S) P, Pn Padamaj Pacavidh paribh Paribh, Paribhlndus, ParibhS, PS Pat, Pataj PB PGS Pig Pitrme Pr Pr, Prt PrakKaum Pratij praty Pur Pusp Rktan RPr, RkPrt RV, RgV s, su S, Sank S(SS) SA ShDarp Sam SamhUB Smkhyakr Sarvnukram SB SB, SVB SB(B), SB(Bayer) SK,SiddhK,SiddhKaum SK, Sabdakaust SK,SKbh,SarKbhar Skand Skt SrngPrak

Nagojbhatta New Indian Antiquary Nidnastra Nighantu Nirukta Nyyamanjar Nyyapraksa Nyya(stra) Pnini Padamanjar Pacavidhastra paribhs Paribhsendusekhara Patajali Pancavirpsabrhmana Praskaragrhyastra Pngala Pitrmedha Pradpa Prtiskhya Prakriykaumudi Pratijastra pratyhra Purna Puspastra Rktantra Rkprtiskhya Rgveda stra Sakara Shkhyana(srautastra) Snkhyanranyaka Shityadarpana Sambandha (samuddesa) Samhitopanisadbrhmana Smkhyakrik Sarvnukraman Satapathabrhmana Sadvimsabrhmana Sitzungsberichte (der bayerischen Akademie der Wissenschaften) Siddhntakaumudi Sabdakaustubha Sarasvatkanthbharana Skanda(svmin) Sanskrit Srngrapraksa

XXIII

Abbreviations

SS SS sv SV TA, TaittAr TB TPr TPS Tribhsy TS TV Ud Up v V(Pr) v, vrtt Vaidikbhar Vaidyan Vais(S) Vm Vyup Ve(S) Visnudharmott VP, Vk, Vkyapad VS W, Wack Wh WZKM Yjnav Yjnik Yo(SG) YV ZDMG

Sivastra Srautastra svmin Smaveda Taittiriyranyaka Taittiryabrhmana Taittinyaprtiskhya Transactions of the Philological Society Tribhsyaratna Taittiryasamhit Tantravrttika Uddyota Upan sad vrtti Vjasaneyi(prtiskhya) Vaitna(srautastra) vrttika Vaidikbharana Vaidyantha Vaisesika(stra) Vmana Vyupurna Vednta(stra) Visnudharmottara(purna) Vakyapadya Vjasaneyisarnhit Wackernagel Whitney Wiener Zeitschrift fr die Kunde des Morgenlandes Yj naval kya Yjnikadeva Yoga(stra) Yajurveda Zeitschrift der deutschen morgenlndischen Gesellschaft

Table of the Sanskrit

Vedic

Pnnya

Ktantra

Cndra

Jainendra, etc.

Padaptha: SV RV X-VIII AV 1 B.C. TS/VS ) Prtiskhya: Rk \ Atharva (2)] Taittirya i Vjasaneyi (10) Pnini V B.C. Ktyyana 300 B.C. Patajali 150 B.C. Sarvavarman I A.D. Bhartrhari V Ksik VII Jinendrabuddhi Vlil Kaiyata XI Purusottama XII Saranadeva XII Haradatta XIII Rmacandra XV Nryanabhatta Bhattojdksita HEMACANDRA XII JAUMARA Xll-Xlll VOPADEVA XIII SRASVATA XIII? SAUPADMA XIV Durgasirnha VIII Candragomin VI SKATYANA IX Devanandin VI

Vj_|V c

Sma

Nirukta VI-IV B.C.

xvii
Nagojbhatta XVIII

A Reader on the Sanskrit Grammarians

Early Accounts

Plate I

Plate I reproduces a later sculpture from Cidambaram, representing, from left to right, the sages Jaimini (author of the Mlmmsstra: see this volume, pages 286-297), Vyghrapda (with tiger paws, carry-

ing an implement probably used for picking flowers and fruit) and Patajali. I owe this illustration to the kindness of the late Mr. J. M. Somasundaram Pillai. See also p. xxi and Kulke 1970.

I
Hsiian Tsang (602-664)

Throughout the centuries the Sanskrit grammarians occupied a

central place in the complex fabric of Indian civilization. Grammar (vykarana) was the science of sciences, to which scholars in other fields looked for inspiration and technical assistance. It is not surprising, therefore, that foreigners who came in close contact with Indian culture have left accounts of the grammatical tradition. | n addition to the Western scholarly occupation with the Sanskrit grammarians, which itself constitutes an account of this kind, it is of some interest to look at other accounts. Among foreign observers, the one exception in this respect was that of the classical Greeks. The Greeks had quite extensive information about India, and their narrations are invaluable for the study of classical Indian history. But as far as Indian civilization is concerned, they seem to have been mainly struck by the more spectacular aspects of Indian philosophies and religions. There does not appear to have been a single Greek account of the Sanskrit grammarians or, for that matter, of the Sanskrit language. All the Greeks seem to have known is that there were many languages in India and that people did not always understand each other: ean
he 7To\X eOvea 'Ivhcov KOI OVK 6[jL<f)Jva afilai ' t h e r e are many nations

among the Indians, and they don't speak the same language' (Herodotus, Historiae 3.98). The situation is quite different with respect to Asian nations; the Chinese, the Tibetans, and Muslims from different Near Eastern countries have left detailed accounts of Indian civilization, not excluding references to the grammatical tradition. That their interest in India was on a more advanced level than that of the ancient Greeks is also apparent from the fact that some of these foreign scholars knew something of Indian languages, and, in particular, of Sanskrit. The Chinese who left us such accounts had good reasons for being interested in India. They were pious Buddhists who traveled to India as pilgrims to see the country where the founder of their religion had lived and taught. One of the most celebrated of these was Hsan Tsang, who left China for India on foot in A.D. 629 and stayed in India for many years. He described his experiences in a large work entitled Records of the Western Countries (Si-yu-ki). We have additional information about his experiences from his biographer, and pupil, Hwui Li. While many Chinese pilgrims were very superstitious, Hsan Tsang was obviously a good scholar. He was not averse to reporting local legends, but he also provided detailed and reliable information. The first book of the Records describes Central Asia up to the Northwest Indian frontier. The second book contains a description of the first Indian areas that Hsan Tsang visited but starts with a long introduction that provides general information about India. This introduction contains about one paragraph about the language and a brief reference to grammar (Beal 1885,1, 77-78; Julien 1857, 1,71-73): The letters of their alphabet were arranged by the god Brahma, and their forms have been handed down from the first till now. They are forty-seven in number, and are combined so as to form words according to the object, and according to circumstances (of time or place): there are other forms (inflexions) used. This alphabet has spread in different directions and formed diverse branches, according to circumstances; therefore there have been slight modifications in the sounds of the words (spoken language);

5 Hsan Tsang

but in its great features there has been no change. Middle India preserves the original character of the language in its integrity. Here the pronunciation is soft and agreeable, and like the language of the Devas. The pronunciation of the words is clear and pure, and fit as a model for all men. The people of the frontiers have contracted several erroneous modes of pronunciation ; for according to the licentious habits of the people, so also will be the corrupt nature of their language. The most striking feature of this Chinese approach to Sanskrit remains a characteristic of all later Chinese studies. To the Chinese, the alphabet meant everything. They simply assumed that knowledge of the Sanskrit script was as important a feature of Sanskrit as knowledge of characters is of Chinese. Hence they had only a very vague notion of what was involved in knowing the Sanskrit language beyond knowing its alphabet. After the introductory paragraph, previously quoted, Hsan Tsang goes on to say that the young in India are instructed from the age of seven in five sciences (vidyd), of which "The first is called the elucidation of sounds (sabdavidy). This treatise explains and illustrates the agreement (concordance) of words, and it provides an index for derivatives." After this introduction Hsan Tsang describes in his second chapter how he reached the area of Gndhra (on the upper Indus); after relating various further events, he describes how he reaches a town called So-lo-tu-lo (Beal 1885,1,114-117; Julien 1867,1, 125-128). Part of his account agrees with the Indian tradition that the pratyhra-stras (o, /, u, etc.), which occur at the beginning of Pnini's grammar and form an integral part of it, were divinely inspired (they are, therefore, sometimes referred to as Sivastra). Attempts have been made to utilize Hsan Tsang's description to help date Pnini, but this seems impossible. All we are told in this passage is that some five hundred years after the death of the Buddha, Pnini was reborn. It might at most be argued that the original Pnini must have lived after the Buddha, for otherwise it would have been unreasonable to suppose that he had incurred an amount of demerit sufficient to be reborn by not embracing the teachings of the Buddha. I have here reproduced BeaPs translation (with minute changes) mainly because it is in English. But the text was first translated into French, by Stanislas Julien (Julien 1857-1858). Beal claims that his translation is independent of Julien's (Beal 1885, Introduction xxii). But Waley (1952,11) says: " I found that almost everything European writers have said about him [that is, about Hsan Tsang] is taken, directly or indirectly, from an incomplete and very imperfect translation of his biography by Stanislas Julien." This probably applies to the Records as well. These translations pose many problems to Sanskritists, for just as the Chinese must have been greatly puzzled by the structure, and in particular by the inflections of Sanskrit, Western Sinologists may have missed technical references to Sanskrit grammar. The expression in BeaPs translation "there are other forms (inflexions) used" has no parallel in Julien. Similarly, Julien has merely "sens des mots" where Beal translates "agreement (concordance) of words." The Chinese equivalent for sabdavidy (which also occurs in I Tsing: see page 12) might be an equivalent for sabdnussana, which is a well-known Sanskrit expression for grammar.

A. Records Of the Western Countries (Seventh Century) Hsan Tsang

To the north-west of U-to-kia-han-c'ha, 20 I i or so, we come to the town of So-lo-tu-lo (Saltura). This is the place where the Rsi Pnini, who composed the Ching-ming-lun (vykarana), was born. Referring to the most ancient times, letters were very numerous; but when, in the process of ages, the world was destroyed and remained as avoid, the Devas of long life descended spiritually to guide the people. Such was the origin of the ancient letters and composition. From this time and after it the source (of language) spread and passed its (former) bounds. Brahma Deva and Sakra (Devendr) established rules (forms or examples) according to the requirements. Rsis [seers] belonging to different schools each drew up forms of letters. Men in their successive generations put into use what had been delivered to them ; but nevertheless students without ability (religious ability) were unable to make use (of these characters). And now men's lives were reduced to the length of a hundred years, when the Rsi Panini was born ; he was from his birth extensively informed about things (men and things). The times being dull and careless, he wished to reform the vague and false rules (of writing and speaking)to fix the rules and correct improprieties. As he wandered about asking for right ways, he encountered svara Deva, and recounted to him the plan of his undertaking. Isvara Deva said, "Wonderful ! I will assist you in this." The Rsi, having received instruction, retired. He then laboured incessantly and put forth all his power of mind. He collected a multitude of words, and made a book on letters which contained a thousand slokas [Verses'] ; each sloka was of thirty-two syllables. It contained everything known from the first till then, without exception, respecting letters and words. He then closed it and sent it to the king (supreme ruler), who exceedingly prized it, and issued an edict that throughout the kingdom it should be used and taught to others; and he added that whoever should learn it from beginning to end should receive as his reward a thousand pieces of gold. And so from that time masters have received it and handed it down in its completeness for the good of the world. Hence the Brhmans of this town are well grounded in their literary work, and are of high renown for their talents, well informed as to things (men and things), and of a vigorous understanding (memory). In the town of So-lo-tu-lo is astupa. This is the spot where an Arhat [elder] converted a disciple of Pnini. Tathgata [that is, the Buddha] had left the world for some five hundred years, when there was a great Arhat who came to the country of Kasmlr, and went about converting men. Coming to this place, he saw a Brahmacrin occupied in chastising a boy whom he was instructing in letters. Then the Arhat spake to the Brahman thus: " W h y do you cause pain to this child ?" The Brahman replied, " I am teaching him the Shing-ming (Sabdavidy), but he makes no proper progress." The Arhat smiled significantly, on which the Brahman said, "Shamans [that is, Buddhist monks] are of a pitiful and loving disposition, and well disposed to men and creatures generally; why did you smile, honoured sir? Pray let me know! " The Arhat replied, " Light words are not becoming, and I fear to cause in you incredulous thoughts and unbelief. No doubt you have heard of the Rsi Pnini, who compiled the Sabdavidy Sastra, which he has left for the instruction of the world." The Brahman replied, "The children of this town, who are his disciples, revere his eminent qualities, and a statue erected to his memory still exists." The Arhat continued: "This little boy whom you are

7 The Life of Hsan Tsang

instructing was that very (Panini) Rsi. As he devoted his vigorous mind t o investigate w o r l d l y literature, he only produced heretical treatises w i t h o u t any power of t r u e reason in t h e m . His spirit and his wisdom were dispersed, and he has run t h r o u g h the cycles of continued b i r t h f r o m then t i l l now. Thanks t o some remnant of t r u e v i r t u e , he has been now born as your attached child ; but the literature of the w o r l d and these treatises on letters are only cause of useless efforts t o him, and are as nothing compared t o the holy teaching of Tathgata, which, by its mysterious influences, procures both happiness and w i s d o m . "

B. The Life of Hsan Tsang (Seventh Century) Hwui Li

Hwui Li, the biographer of Hsan Tsang, relates in his Life of Hsan Tsang how his teacher (whom he calls thoughout his work " t h e Master of the Law") learned Sanskrit grammar in the Nlanda monastery. He inserts a little treatise on this subject, " no doubt," says Waley (1952, 47), " t o impress us with the difficulties that Tripitaka [that is, Hsan Tsang] overcame in his study of Sanskrit writing and grammar." This treatise is reproduced here in full from Beal's translation (Beal 1911,121-125), though the biography has also been translated by Julien (Julien 1853). There is a recent English translation of Hwui Li's biography by Li Ying-hsi, published under the auspices of the San Shih Buddhist Institute by the Chinese Buddhist Association and printed in the People's Republic of China (Peking 1959); this passage occurs on pages 117-120. The references to Sanskrit have been regularized ; for example, bhavatah has been inserted in the proper place, and the case endings are the correct ones. But vyati, considered an ending (like "underneath the nine inflections" in this context), remains still unexplained, and it is unclear whether the regularizar o n is based upon new manuscript material (in the Foreword by Chao Pu-chu it is said that " t h e Ching Ling Buddhist Text Society at Nanking has completed the task of carving new printing blocks to replace the lost ones for the complete works translated or written by the Ven. Hsan-tsang"). Pnini's grammar consists of just under 4000 stras, but the Chinese writers on India always seem to measure the size of writings in terms of slokas. Asta-dhtu can refer neither to the Dhtuptha (which has ten chapters, corresponding to the ten verb classes), nor to Pnini's grammar itself, the Astdhyyl (eight chapters), as it is distinguished from both by I Tsing (see page 13). But Astadhtu may be an elementary book, such as is referred to under that name in the Durghatavrtti (Renou 1940, Introduction, 60). The terms tianta and subanta for "what ends in tin, i.e., verbal endings" (verbs) and "what ends in sup, i.e., nominal endings" (nouns) are undoubtedly Pninian. The expressions tiantapadam ("a word ending in t/ri ) and subantapadam ("a word ending in sup'1) are generally used in this connection, but vjya could stand for vkya (in general meaning "sentence" but here perhaps " e x pression " ) . Parasmai(pada) and tmane(pada) are also Pninian terms for the Active and Middle endings, respectively. The reference to vyati (or ve, ya> ti) is puzzling (the process here referred to is described differently in Pnini 3.4.78-80)*. The explanation of

* The ending of the first person singular Middle is (e.g., bhave); the element -ya- is attached to the root

before the endings of the Passive (e.g., kriyate); the ending -ti occurs only in the Active.

8 Hsan Tsang

the cases, with the exception of the Genitive, is reminiscent of Pnini's kraka theory. However, Pnini did not consider the Vocative a separate case, but regarded it as a special use (mantrila) of the Nominative. Last, it is apparent that there was considerable confusion with regard to the case endings. For some further discussion see van Gulik (1956,14-18). The Master of the Law whilst he stopped in the convent, heard the explanation of the Yoga-sstra, three times : the NyyAnusra-sstra, once; the Hin-hiang-tui-fa-ming, once; the Hetuvidy-sstra and the Sabdavidy and the tsah Hang sastras, twice ; the Prnyamla sastra-tk, and the Sata-sstra, thrice. The Kosa, Vibhs, and the Satpadbhidharma sastras, he had already heard explained in the different parts of Kasmir ; but when he came to this convent he wished to study them again to satisfy some doubts he had : this done, he also devoted himself to the study of the Brahman books and the work called Vykarana on Indian letters, whose origin is from the most remote date, and whose author is unknown. At the beginning of each Kalpa, Brahma-raja first declares it [the Vykarana], and then transmits it for Devas and men to use. Being thus declared by Brahma-rja, therefore men call it Fan, or Brahma, writing. The words of this book are very extensive, comprising a hundred myriad slokas. It is the same as the old commentary calls the Vykara(na)-sstra, but this pronunciation is not complete. If correct, it would be Vykaranam, which is another name for 'a treatise relating to the record of the science of sounds.' It treats at large, in a mnemonic way, on all the laws of language and illustrates them, hence the name. At the beginning of the Kalpa of perfection (vaivarta kalpa) Brahma-rja first declared this book; it then comprised 100 myriad [thousand] of slokas; afterwards, at the beginning of the Vaivarta-siddha:Kalpa, that is, the kalpa, or period, of establishment, Ti-shih (Sakra-raja) reduced them to ten myriad slokas. After this a Brahman of the town Saltura in Gandhra of North India, whose name was Panini Rsi, reduced them to 8000 slokas. This is the work at present used in India. Lately a Brahman of South India, at the request of a king of South India, reduced them further to 2500 slokas. This work is widely spread, and used throughout all the frontier provinces, but the well-read scholars of India do not follow it as their guide in practice. This then is the fundamental treatise relating to sounds and letters of the Western world, their branch-divisions, distinctions and mutual connections. Again, there is a Vykaranam work (mnemonic treatise) of a short kind having 1000slokas; again, there is one of 300slokas on the roots (bases) of letters (i.e. letter roots or bases); again, there are (treatises on the) two separate kinds of letter-groupings, one named Mandaka in 3000 slokas, the other called Undi in 2500 slokas. These distinguish letter-groupings from letter-roots. Again, there is the treatise called Asta-dhtu (Dhtuvrttil) in 800 slokas ; in this work there is a brief conjunction of letter-bases and letter-groupings. These are all the Vykarana treatises. In distinguishing active and passive expositions (i.e. in expounding the principles of grammar, relating to active and passive verbs), there are these two rules : the first, called Ti-yen-to-shing

9 The Life of HsanTsang

(Tianta-vajyam) having eighteen inflections ; the second Su-manto-shing (Subanta-vjyam), having twenty-four inflections; the Tianta "sounds" are used in elegant compositions, but seldom in light literature. The twenty-four "sounds" are used in all kinds of composition alike. The eighteen inflections of the Tianta "sounds" are of two characters: 1st, Parasmai, 2nd, At mane; each of these has nine inflections, and so together there are eighteen. W i t h respect to the nine which come first : we know that in ordinary discourse everything has three ways of being viewed (/.e.
as one thing, or two things, or many things) ; every other person has t h r e e ways of being considered (i.e. as one other, two other, or many

other); and also "oneself " can be considered in three ways (i.e. as / myself, two of us, or many of us). Thus every single thing may be regarded in these three ways, as one, two of a class, or many; here, then, are three (three persons and three numbers, altogether nine). In both (voices) the root-word is the same, but the (final) sounds are different. So there are two sets of nine. Now, taking the Parasmai sounds : we may speak of a thing as existing or not existing, in all cases. Supposing then we say a thing exists, there are three ways of putting (naming) this fact; we may say " it exists" (bhavati) or, " t w o things exist" (bhavapay or, "they exist" (bhavanti). And so, speaking of another, we may say "thou dost exist" (bhavasi), or, "you two exist" (bhavapa, for bhavathah), or "you all exist" (bhavatha); and so again speaking of oneself we may say " I exist" (bhavmi), or, " w e two exist" (bhavavah), or, " w e all exist" (bhavmah). W i t h regard to the nine case-endings of the Atmane class, they simply take underneath the nine inflections just named the word "vyati,11 (or, the words ve, ya, ti) ; in other respects they are the same as the above. Thus touching these things, we see how askilfu I writer inthis language is saved from ambiguity, and also how his meaning may be expressed in the most elegant manner. W i t h respect to the twenty-four inflections of the Subanta "sound (endings)," it is to be observed that every word has altogether eight inflections (cases), and that each of these cases or inflections is subject to three conditions as to number, viz., when one, or two, or many, are concerned. Hence arise the twenty-four (sound-endings). Then, again, in connection with these twenty-four inflections we have three other terms, viz., the masculine sound ending, the feminine, and the neuter. But regarding the eight inflections, the first exhibits the substance, or basis, of the thing conceived (nominative); the second exhibits the deed done (objective); the third, the means by which, and the doer (instrumental); the fourth, for whom the thing is done (dative); the fifth, what causes the thing (ablative) ; the sixth, whose is the thing (genitive); the seventh, that which determines (localises) the thing (locative); the eighth, the calling, or summoning, the thing (vocative). Now, for example, let us take the masculine ending, as in the word " man," and go through the eight cases named above. The word " man " in Indian speech is purusa. The root-word has three inflexions, viz., purusah, purusau, puruss. The thing done (object) has three, purusam, purusau, purusn ; the instrument by which the thing is done by the doer has also three inflexions purusena, puru(s)bhym, purusbhih or purusaih; " f o r whom the
1

For Bhavatoh.

10 HsanTsang

thing is done" [is rendered by] purusoyo, purusobhyom, purusesu; " t h e cause from which the thing proceeds," [by] purusot, purusbhym, purusesu; "whose is the thing," [by] purusasya, purusbhym, purusnm; " t h e place where," [by] puruse, purusayos, purusnm; " t h e calling case," [by] hi purusa, hi purusau, hi purusdh. From these one or two examples, other cases may be understood; it would be difficult to make a full statement of particulars. The Master of the Law thoroughly investigated the language (words and phrases), and by talking with those men on the subject of the " pure writing," he advanced excellently in his knowledge. Thus he penetrated, and examined completely, all the collection (of Buddhist books), and also studied the sacred books of the Brahmans during five years.

2
I Tsing (634-71 3)

Another Chinese scholar-pilgrim, I Tsing, left China in A.D. 671

by sea for India via Indonesia. He was then thirty-seven years old. Before he entered India proper, he stayed six months in Sumatra to learn Sanskrit. This would hardly be the place to learn that language nowadays, but in the seventh century, Eastern Sumatra w a s t n e center of the Srivijaya empire and, at the same time, a center for Buddhist, and therefore also Sanskrit, learning. After leaving Sumatra, I Tsing had many adventures on the way to India and in India, where he visited mainly the Ganges plain. He returned again via the sea route and Indonesia. Before reaching China in 695, he spent at least another four years in Sumatra. Here he wrote, probably during 691-692, his Record of Buddhist Practices sent home from the Southern Sea. This work, translated by Takakusu, gives a very detailed account of life in India and contains a long chapter (the thirty-fourth) on "The method of Learning in the West." In the beginning of this chapter, I Tsing discusses Sanskrit and Buddhist philosophy. He stresses that the highest truth (paramrtha-satya) is beyond language, but admits that lower truth (samvrti'Satya) " may be explained by words and phrases." He then continues with a detailed account of the Sanskrit grammarians, which, despite its fascination with numbers and enumeration, is full of precious information (Takakusu 1896,168-180, omitting most of the notes). A fair amount of literature has been devoted to these compact pages: Max Mller (1883, 210-213, 281-366; also in Takakusu 1896, ix-xvi); F. Kielhorn (1883,226); B. Liebich (1930,266-270, 281-284). Many difficulties of interpretation remain, and the text ought to be studied anew.

The Method Of Learning in t h e W e s t (691-692) I Tsing

The old translators have seldom told us the rules of the Sanskrit language. Those who lately introduced the Stras to our notice spoke only of the first seven cases. This is not because of ignorance (of grammar), but they have kept silence thinking it useless (to teach the eighth), (i.e., the vocative). I trust that now a thorough study of Sanskrit grammar may clear up many difficulties we encounter whilst engaged in translation. In this hope, I shall, in the following paragraphs, briefly explain some points as an introduction to grammar. (Note by l-tsing): Even in the island of Pulo Condore (in the south) and in the country of Soli (in the north), peopie praise the Sanskrit Sutras, how much more then should people of the Divine Land (China), as well as the Celestial Store House (India), teach the real rules of the language! Thus the people of Indiasaid in praise (of China): "The wise MajusrT is at present in Ping Chou, where the people are greatly blessed by his presence. We ought therefore to respect and admire that country, &c." The whole of their account1 is too long to be produced. Grammatical science is called, in Sanskrit, Sabdavidy, one of the five Vidys; Sabda meaning 'voice,' and Vidy 'science.' The name for the general secular literature in India is Vykarana, of which there are about five works, similar to the Five Classics of the Divine Land (China). I. The Si-t'an-chang (Siddha-composition) f o r Beginners* This is also called Siddhirastu, signifying ' Be there success1 (in Sanskrit, the Chinese text has ' complete be good luck! ') for so named is the first section of this small (book of) learning. There are forty-nine letters (of the alphabet) which are combined with one another and arranged in eighteen sections; the total number of syllables is more than 10,000, or more than 300 slokas. Generally speaking, each sloka contains four feet (pdas), each foot consisting of eight syllables; each sloka has therefore thirty-two syllables. Again there are long and short siokas ; of these it is impossible here to give a minute account. Children learn this book when they are six years old, and finish it in six months. This is said to have been originally taught by Mahesvara-deva (Siva). II. The Sutra The Stra is the foundation of all grammatical science. This name can be translated by 'short aphorism,' and signifies that important principles are expounded in an abridged form. It contains 1000 slokas, and is the work of Pnini, a very learned scholar of old, who is said to have been inspired and assisted by Mahesvara-deva, and endowed with three eyes; this is generally believed by the Indians of today. Children begin to learn the Stra when they are eight years old, and can repeat it in eight months' time.
l-tsing seems to be quoting these passages from a book. [* Kielhorn (1883, 226), quoted by Takakusu, considers this an elementary Siddhnta, "which teaches the
1

letters, their combinations, the organs with which they are pronounced, etc." The words siddhir astu ' let there be success ' may occur at the beginning of almost any book.]

13

HI. The Book on Dhatu

This consists of 1000 slokas, and treats particularly of grammatical roots. It is as useful as the above Stra.

The Method of Learning in the West

IV. The Book on the Three Khilas**

Khila means 'waste land,' so called because this (part of grammar) may be likened to the way in which afarmer prepares his fields for corn. It may be called a book on the three pieces of waste land. (1) Astadhtu consists of 1000 slokas ; (2) Wen-ch'a (Manda or Munda) also consists of 1000 slokas; (3) Undi too consists of 1000 slokas. 1. Astadhtu This treats of the seven cases (Sup) and ten Las, and eighteen finals (Tin, 2 x 9 personal terminations). a. The Seven Cases Every noun has seven cases, and every case has three numbers, i.e. singular (Ekavacana), dual (Dvivacana), and plural (Bahuvacana); so every noun has twenty-one forms altogether. Take the word " man," for instance. If one man is meant it is 'Purusah', two men, 'Purusau', and three (or more) men, ' Purush'. These forms of a noun are also distinguished as heavy and light (probably 'accented and unaccented '), oras pronounced by the open and closed breathings (perhaps ' nouns with an open vowel or those witha closed vowel '). Besides the seven cases there is the eighth,the vocative case (Amantrita), which makes up the eight cases. As the first case has three numbers, so have the remaining ones, the forms of which, being too numerous to be mentioned, are omitted here. A noun is called Subanta, having (3 x 8) twenty-four (inflected) forms. . The Ten Las There are ten signs with L (for the verbal tenses); in conjugating (lit. expressing) a verb, the distinctions of the three times, i.e. past, present, and future, are expressed. y. The Eighteen Finals ( T i n ) . These are the forms of the first, second, and the third person (of the three numbers of a verb), showing the differences of the worthy and unworthy, or this and that.2 Thus every verb (in one tense) has eighteen different forms which are called Tianta. 2. Wen-ch'a (Manda or Munda) This treats of the formation of words by means of combining (a root and a suffix or suffixes). For instance, one of many names for " t r e e " in Sanskrit is vrksa. Thus a name for a thing or a matter is formed by joining (the syllables) together, according to the rules of the Stra, which consist of more than twenty verses.
[* This is obviously the Dhatupatha.] [** The word khila means ' appendix,' but what the Astadhtu, also mentioned by Hwui Li (Liebich 1930, 282: "eine Sammlung von Deklinations- und Konjugationsparadigmen ") and the Manda or Munda are is unclear. Like sup and tin (see page 7), Lisa technical expression in Pnini's grammar. It is the initial sound of the technical names of the ten tenses and moods (i.e., at ' present,' lit ' perfect,' lut 'periphrasticfuture,' In 'future,' let 'subjunctive,' lot 'imperative,' Ian 'imperfect,' lin 'optative,' lu 'aorist,1 and Im 'conditional '). For Undi suffixes see page 62.] 2 We should expect here 'tmanepadaand Parasmaipada.' 'This and that' may be a vague way of expressing the grammatical terms 'tmane' and ' Parasmai,' for Chinese has no grammatical terms for these. Still, 'worthy and unworthy' is very strange.

14 I Tsing

3. The Uad i This is nearly the same as the above (Mand a), with the exception that what is fully explained in the one is only mentioned briefly in the other, and vice versa. Boys begin to learn the book on the three Khilas (or 'three pieces of waste land ') when they are ten years old, and understand them thoroughly after three years' diligent study.

V. The Vrtti-stra ( Ksikvrtti)*

This is a commentary on the foregoing Sutra (i.e. Pnini's Stra). There were many commentaries composed in former times, and this is the best of them. It cites the text of the Stra, and explains minutely its manifold meaning, consisting altogether of 18,000 slokas. It exposes the laws of the universe, and the regulations of gods and men. Boys of fifteen begin to study this commentary, and understand it after five years. If men of China go to Indiafor study, they have first of all to learn this (grammatical) work, then other subjects; if not, their labour will be thrown away. All these books should be learnt by heart. But this, as a rule, applies only to men of high talent, while for those of medium or little ability a different measure (method) must be taken according to their wishes. They should study hard day and night, without letting a moment pass for idle repose. They should be like the Father K'ung (i.e. Confucius), by whose hard study the leather binding of his Yi-king was three times worn away; or imitate Sui-Shih, who used to read a book a hundred times. The hairs of a bull are counted by thousands, but a unicorn has only one horn. 3 The labour or merit of learning the above works is equal to that of proceeding to (the grade of) the Master of Classics (' Ming-king'). This Vrtti-stra is a work of the learned Jayditya. He was a man of great ability; his literary power was very striking. He understood things which he had heard once, not requiring to be taught twice. He revered the Three Honourable Ones (i.e. Triratna), and constantly performed the meritorious actions. It is now nearly thirty years since his death (A.D. 661-662). After having studied this commentary, students begin to learn composition in prose and verse, and devote themselves to logic (Hetuvidy) and metaphysic (Abhidharmakosa). In learning the Nyya-dvra-traka-sstra, they rightly draw inferences (Anumna); and by studying the Jtakaml their powers of comprehension increase. Thus, instructed by their teachers and instructing others, they pass two or three years, generally in the Nlanda monastery in Central India, or in the country of Valabhl (Wal) in

[* The first part of the Kasikavrtti commentary (though Belvalkar 1915, 35, doubts that this text is meant here by the term vrttistra, the regular name being stravrtti; but Liebich, 1930, 283, is not taken aback by this inversion) was written by Jayditya; its second part, by Vmana possibly after I Tsing had left India. The dates 661-662, which are undoubtedly too precise, are based

upon this important passage alone, In a grammatical context a reference to "the Three Honourable Ones" is most readily interpreted as a reference to the munitrayom 'three sages' or ' three cryos ' (see below page 86), viz., Pnini, Ktyyana and Patajali. Moreover, the Buddhist trirotnaBuddha, law, and order consists of at most one person.] 3 That is to say, ' Few are clever.'

15 The Method of Learning in the West

Western India. These two places are like Chin-ma, Shih-ch'u, Lung-mn, and Ch'ue-li in China, and there eminent and accomplished men assemble in crowds, discuss possible and impossible doctrines, and after having been assured of the excellence of their opinions by wise men, become far famed for their wisdom. To try the sharpness of their wit (lit. ' sharp point of the sword '), they proceed to the king's court to lay down before it the sharp weapon (of their abilities); there they present their schemes and show their (political) talent, seeking to be appointed in the practical government. When they are present in the House of Debate, they raise their seat and seek to prove their wonderful cleverness. When they are refuting heretic doctrines, all their opponents become tongue-tied and acknowledge themselves undone. Then the sound of their fame makes the five mountains (of India) vibrate, and their renown flows, as it were, over the four borders. They receive grants of land, and are advanced to a high rank; their famous names are, as a reward, written in white on their lofty gates. After this they can follow whatever occupation they like.

VI.TheCrn*
Next, there is a commentary on the Vrtti-stra entitled CrnT, containing 24,000 slokas. It is a work of the learned Patajal i. This, again, cites the former Stras (Pnini), explaining the obscure points (lit. ' piercing the skin ') and analysing the principles contained in it, and it illustrates the latter commentary (Vrtti), clearing up many difficulties (lit. ' removing and breaking the hair and beard of corn '). Advanced scholars learn this in three years. The labour or merit is similar to that of learning the Ch'un-ch'iu and the Yi-king in China.

VII. The Bhartrhari-sstra**

Next, there is the Bhartrhari-sstra. This is the commentary on the foregoing Curn, and is the work of a great scholar Bhartrhari. It contains 25,000 slokas, and fully treats of the principles of human life, as well as of grammatical science, and also relates the reasons of the rise and decline of many families. The author was intimately acquainted with the doctrine of'sole knowledge' (Vidymtra), and has skilfully discussed about the Hetu and Udharana (the cause and example of logic). This scholar was very famous throughout the five parts of India, and his excellences were known everywhere (lit. ' t o the eight quarters '). He believed deeply in the Three Jewels (i.e. Ratnatraya), and diligently meditated on the 'twofold nothingness' (Snya).4 Having desired to embrace the
[* As Max Mller (1883) and Kielhorn (1883) have pointed out, Crni is a name used for Patajali's Mahbhsya.] [** The date and religion of the celebrated philosopher-grammarian Bhartrhari have been the subject of ample discussion. Despite I Tsing's testimony, Bhartrhari is nowadays generally believed to have been a Hindu (see, for example, Biardeau 1964b) and is sometimes believed to have lived in the fifth century (because the Buddhist logician Dinnga appears to quote him : Frauwallner 1959; cf. Ruegg1959, 57-60). Portions of what is extant of the commentary on the Mahbhsya mentioned here have been published only recently.] 4 The 'twofold nothingness,1 ' both tman and Dharmaare but an empty show.'

16 I Tsing

excellent Law he became a homeless priest, but overcome by worldly desires he returned again to the laity. In the same manner he became seven times a priest, and seven times returned to the laity. Unless one believes well in the truth of cause and effect, one cannot act strenuously like him. He wrote the following verses, full of self-reproach : Through the enticement of the world I returned to the laity. Being free from secular pleasures again I wear the priestly cloak. How do these two impulses Play with me as if a child? He was a contemporary of Dharmapla. Once when a priest in the monastery, being harassed by worldly desires, he was disposed to return to the laity. He remained, however, firm, and asked a student to get a carriage outside the monastery. On being asked the cause, he replied : ' It is the place where one performs meritorious actions, and it is designed for the dwelling of those who keep the moral precepts (Si I a). Now passion already predominates within me, and I am incapable of adhering to the excellent Law. Such a man as myself should not intrude into an assembly of priests come here from every quarter.' Then he returned to the position of a lay devotee (Upsaka), and wearing a white garment continued to exalt and promote the true religion, being still in the monastery. It is forty years since his death (A.D. 651-652).

VIII. The Vkya-discourse

In addition there is the Vkya-discourse (VkyapadTya). This contains 700 slokas, and its commentary portion has 7000 (siokas). This is also Bhartrhari's work, atreatise on the Inference supported by the authority of the sacred teaching, and on Inductive arguments.

IX.ThePei-na*
Next there is Pei-na (probably Sanskrit ' Beda' or ' Veda'). It contains 3000 slokas, and its commentary portion is in 14,000 (slokas).
[* Bhartrhari is not known to have written anything apart from the Mahbhsya commentary and the Vkyapadiya. Takakusu's reference to a Bedvrtti, mentioned by Bhler and occurring in Bhandarkar's Catalogue of MSS in the Deccan College (Takakusu 1896, 225) is not particularly illuminating. Liebich (1930, 266-267) believes " Pei-na" to mean Praklrna, the third part of the Vkyapadiya, which is very long and sometimes regarded as a separate work. This was in fact the viewpoint of Kielhorn, who had already drawn attention to the fact that South Indian manuscripts call this third part (which is linguistically the most interesting part) Praklrnaka, and that Helarja called his commentary on this part Prakirna-praksa (Kielhorn 1883, 227). Also Abhinavagupta (about A.D. 1000) wrote a work called Prakirnakavivaranam, now lost, which was probably a commentary on the third part of the Vokyapadiya. Recently, Aklujkar (1969), has shown that the title Vkyapadiya was in fact originally given to the first two books only. Another possibility is that the name of Punya (raja), a commentator on Bhartrhari's Vkyapadiya, is referred to by the Chinese " Pei-na." There are difficulties with either interprtation as long as it is not certain who the Dharmapla that I Tsing referred to was (cf. Aklujkar 1969, 549, n. 8).]

17 The Method of Learning in the West

The sloka portion was composed by Bhartrhari, while the commentar/ portion is attributed to Dharmapala, teacher of the Sastra. This book fathoms the deep secrets of heaven and earth, and treats of the philosophy of man (lit. 'the essential beauty of the human principles'). A person who has studied so far as this (book), is said to have mastered grammatical science, and may be compared to one who has learnt the Nine Classics and all the other authors of China. All the above-mentioned books are studied by both priests and laymen ; if not they cannot gain the fame of the well-informed (lit. ' much heard/ Bahusruta, or ' Knowing much of the sruti ').

3
Fa Tsang (643-71 2)

Though Hsan Tsang and I Tsing are the most famous of the

Chinese pilgrims who provided information on India, there were other Chinese scholars who referred to Sanskrit and, in such references, to the Sanskrit grammarians. Fa Tsang (643-712) was a Buddhist monk of Sogdian descent and a prolific author who | vec j n China. He was the main authority on the Avatamsaka stra, a Buddhist Mahyna work that taught the sameness of everything and became very Influential in China and Japan. In Fa Tsang's

Account of Exploring the Mysteries of the Avatamsaka Stra, a passage

occurs which deals with Sanskrit grammar. Van Gulik, who stresses that the Chinese knowledge of Sanskrit was generally confined to the script, quotes this passage in order to prove that " it would have been quite possible to draw up a complete Chinese version of Pn'in'i's rules accompanied by a commentary." The quotation proves no such thing, and it does not refer directly to the Sanskrit grammarians. In fact, its explanation of the cases is rather different from the one we are given by Pnini. But some of Fa Tsang's examples may be traced back to the grammatical tradition. The Ksik, for example, uses parasun chinatti ' he cuts with an a x e ' t o illustrate the Instrumental (commenting on Pnini 1.4.42, 2.3.18). Fa Tsang's passage has been published and translated by van Gulik (1956,19-20). It is remarkable that both van Gulik and Fa Tsang talk about eightfold declension, and then proceed to discuss seven cases. As we have seen (on page 13), their practice accords with the Indian tradition.

Account of Exploring the Mysteries of the Avatamsaka Sutra (Seventh Century) Fa Tsang

The cases refer to the (grammatical) rules of the Western countries. If one wants to examine and read the sacred and secular books one has to know the rules for the eightfold declension. If one does not understand these, one cannot know the meaning and arrangement of the text. 1. purusah, the case of direct indication ; for instance, in the sentence 'The man cuts down the tree', this case points directly to that man [nominative]. 2. purusam, the case indicating that to which something happens. As in the sentence ' The tree that is cut ' [accusative]. 3. purusena, the case indicating the instrument with which something is done, as in the sentence 'to cut a tree with an axe' [instrumental]. 4. purusya, the case indicating for whom something is done, as in 1 to cut a tree for a man ' [dative]. 5. purust, the case indicating a causal relation, as in 'to build a house on behalf of a man ' [ablative]. 6. purusasya, the case indicating possession, as in 'the slave belongs to the master' [genitive]. 7. puruse, the case indicating staying with, as in 'the guest stays with his host1 [locative].

4
_ _ m _ A b u Raihan al-Biruni (973-1048)

The expansion of Islam and scholarly curiosity seem to be the two

main factors that explain the large number of Arabic and Persian publications dealing with India. Foremost among these is the Arabic treatise of the Persian scholar Abu Raihn Muhammad Ibn Ahmad al-Brun (973-1048), referred toas "Indica" by its editor and translator, Edward Sachau. This detailed and scholar| y work, the real title of which is An Accurate Description of Ail Categories of Hindu Thought, Those Which Are Admissible as well as Those Which Must Be Rejected, was written in 1030. Al-Brun, who was not only a Sanskrit scholar but also an astronomer, wrote a large number of works on Indian astronomy and mathematics. While writing his " Indica" he was also engaged, as he tells us, in translating Euclid's Elements, Ptolemy's Almagest, and writing a treatise of his own on the construction of the astrolabe into Sanskrit slokas. The " Indica" contains detailed descriptions of Hindu religion, philosophy, literature, and especially astronomy, al-Brun's specialty. In the beginning of his book al-Brun discusses the reasons why "the Hindus entirely differ from us in every respect." The foremost reason he cites is their language, which is " o f an enormous range, both in words and inflections, something like the Arabic, calling one and the same thing by various names, both original and derived, and using one and the same word for a variety of subjects, which, in order to be properly understood, must be distinguished from each other by various qualifying epithets. For nobody could distinguish between the various meanings of a word unless he understands the context in which it occurs, and its relation to both the following and the preceding part of the sentence. The Hindus, like other people, boast of this enormous range of their language, whilst in reality it is a defect" (Sachau 1910, 17-18). This last topic is a theme to which al-Brun occasionally returns, for example, when discussing the variety of geographical names: " O f course, in all of this the Hindus are actuated by the desire to have as many names as possible, and to practice on them the rules and arts of their etymology, and they glory in the enormous copiousness of their language which they obtain by such means" (Sachau 1910, 299). Al-Brun's disapproval is related to the ensuing difficulty of learning "the whole of the language" (pages 228-229): The Hindus and their like boast of this copiousness, whilst in reality it is one of the greatest faults of the language. For it is the task of language to give a name to everything in creation and to its effects, a name based on general consent, so that everybody, when hearing this name pronounced by another man, understands what he means. If therefore one and the same name or word means a variety of things, it betrays a defect of the language and compels the hearer to ask the speaker what he means by the word. And thus the word in question must be droppedin order to be replaced either by asimilar one of a sufficiently clear meaning, or by an epithet describing what is really meant. If one and the same thing is called by many names, and this is not occasioned by the fact that every tribe or class of people uses a separate one of them, and if, in fact, one single name would be sufficient, all the other names save this one are to be classified as mere nonsense, as a means of keeping people in the dark, and throwing an air of mystery about the subject. And in any case this copiousness offers painful difficulties to those who want to learn the whole of the language, for it is entirely useless, and only results in a sheer waste of time.

21 Abu Raihn al-Brn

Another reason for the differences between Muslims and Hindus is that " t h e [Hindus'] language is divided into a neglected vernacular one, only in use among the common people, and a classical one, only in use among the upper and educated classes, which is much cultivated, and subject to the rules of grammatical inflection and etymology, and to all the niceties of grammar and rhetoric" (page 18). Furthermore, the language is very difficult to pronounce ("we have sometimes written down a word from the mouth of Hindus, taking the greatest pains to fix its pronunciation, and afterwards when we repeated it to them, they had great difficulty in recognising i t " (page 18); with characteristic selfconfidence, al-BrQn attributes this to his informants' carelessness). Lastly, the Hindus like to compose their works in meters in order to facilitate their being learned by heart. " Now it is well known that in all metrical compositions there is much misty and constrained phraseology merely intended to fill up the metre and serving as a kind of patchwork, and this necessitates a certain amount of verbosity. This is also one of the reasons why a word has sometimes one meaning and sometimes another" (page 19). The thirteenth chapter of the "Indica" is devoted to "their [the Hindu] grammatical and metrical literature." Though not apparently an admirer of poetry, al-Brn was interested in Sanskrit meters and in the way the Hindus used arithmetic in their metrical system. Most of the chapter is devoted to this topic, but in the beginning he also provides some information on grammatical works and schools (Sachau 1887, 65; 1910,1,135-136; II, 300-301). A further description of the books listed in the text of this article is as follows: 1. An "Aindra" school of grammar has often been construed as a pre-Paninian school of Sanskrit grammarians (last by A. C. Burnell 1875); but a grammarian Indra(gomin), probably much later than Pnini, is known only by name. The confusion may be partly due to traditions which connect the god Indra with the origin of grammar (cf. page 136 of this volume and Patajali's Introduction to the Mahbhsya, quoted in Staal 1969, 501-502). The earliest of these traditions is of Vedic origin : TaiWriyasa m hita 6 A.7 "Speech indeed spoke formerly without manifestation (avykrta). The gods said to Indra: 'do manifest this speech for us'... Indra approaching it from the middle made it manifest (avyakarot). Therefore speech is manifest (vykrta)" (hence vykarana ' making manifest, analysis, grammar'). 2. This is the most popular Buddhist grammar; Candra(gomin)'s date is uncertain (estimates range from the fifth to the seventh century A.D.). 3. The author of the grammar, not his kabia 'tribe', is Skatyana; he was a Jain who probably lived in the ninth century. 4. The reading of the manuscript is pnriti, the result of an effort to render the retroflex feature of Sanskrit " n " (Chatterji 1951, 90; cf. Sachau 1910,11,300-301). 5. This is the oldest non-Paninian grammar. 6. This work is not known to me. 7. Not known to me; this could not be a misspelling of Durghatavrtti, since this work was written in the twelfth century. 8. Deciphered by Kielhorn (Sachau 1910, II, 301); cf. Belvalkar (1915, 91); Renou (1940, Introduction 59).

An Accurate Description of All Categories of Hindu Thought (1030)

Abu Raihn al-Brn

The two sciences of grammar and metrics are auxil ary to the other sciences. Of the two, the former, grammar, holds the first place in their [the Hindus'] estimate; [it is] called vykarana, i.e. the law of the correctness of their speech and etymological rules, by means of which they acquire an eloquent and classical style both in writing and reading. We Muslims cannot learn anything of it, since it is a branch coming from a root which is not within our graspI mean the language itself. That which I have been told as to titles of books on this science is the following: 1. Aindra, attributed to Indra, the head of the angels. 2. Cndra, composed by Candra, one of the red-robe-wearing sect, the followers of Buddha. 3. Skata, so called by the name of its author. His tribe, too, is called by a name derived from the same word, viz. Skatyana. 4. Pnini, so called from it^author. 5. Ktantra, composed by Sarvavarman. 6. Sasidevavrtti, composed by Sasideva. 7. Durgavivrtti. S.Sisyahitvrtti, composed by Ugrabhti. I have been told that the last-mentioned author was the teacher and instructor of Shh Anandapla, the son of Jayapla, who ruled in our time. After having composed the book he sent it to Kashmir, but the people there did not adopt it, being in such things haughtily conservative. Now he complained of this to the Shh, and the Shah, in accordance with the duty of a pupil toward his master, promised him to make him attain his wish. So he gave orders to send 200,000 dirham and presents of a similar value to Kashmir, to be distributed among those who studied the book of his master. The consequence was that they all rushed upon the book, and would not copy any other grammar but this one, showing themselves in the baseness of their avarice. The book became the fashion and highly prized. Of the origin of grammar they give the following account:* One of their kings, called Samalvhana, i.e., in the classical language Stavhana, was one day in a pond playing with his wives, when he said to one of them M udakam dehi, i.e. do not sprinkle the water on me. The woman, however, understood it as if he had said modakam dehi, i.e. bring sweetmeats. So she went away and brought him sweetmeats. And when the king disapproved of her doing so, she gave him an angry reply, and used coarse language towards him. Now he was deeply offended, and in consequence, as is their custom, he abstained from all food, and concealed himself in some corner until he was called upon by a sage, who consoled him, promising him that he would teach people grammar and the inflexions of the language. Thereupon the sage went off to Mahdeva, praying and fasting devoutly. Mahdeva appeared to him, and communicated to him some few rules, the like of which Abul'aswad Addu'alT has given for the Arabic language. The god also promised to assist him in the further development of this science. Then the sage returned to the king and taught it to him. This was the beginning of the science of grammar.
[* This account of the origin of grammar s a patriarchal version of the traditional legend of the origin of the Ktantra school, where it is the ignorant king Stavhana who, when requested by his queen modakom dehi, offered her sweets (Belvalkar 1915, 82-83). Stavhana is the name of a dynasty of South Indian kings (approximately 50 B.C.-A.D. 250).]

5
Trantha (Born I 573)

The interests of the Tibetan Buddhist scholars were similar to

those of the Chinese Buddhists, but they were closer to India, had easier access to a large number of sources, and had been familiar with Sanskrit for a longer period. They also seemed to have had a predilection for describing fantastic events. Among the earlier historians, Bu-ston (1290-1364) provides the most information, though much of it is in mythical form. According to Bu-ston, a great sastra dealing with grammar was first composed in heaven by the god Sarvajana and was subsequently lost. Thereupon the god Indra composed the Indravykarana, which was studied and propagated by the divine perceptor, Brhaspati. It became known in India, but gradually lost popularity. Thereupon Pnini appeared (Schiefner 1869, 294). The work of the Tibetan scholar Trantha (born 1573) on the history of Buddhism in India, which was completed in 1608, also incorporates numerous legends, but its aim appears to be more strictly historical. This book begins with a description of the age of the Buddha and the reign of the king of Magadha, Ajtasatru, who killed and succeeded his father Bimbisra about 490 B.C. About 413 B.C. king Nanda overthrew this earlier dynasty and established a new one, known as the Nanda dynasty. When relating events during the reign of king Nanda, Trantha speaks about Pnini and provides some information about grammars (from the German translation in Schiefner 1869, 53-54). Trantha does not talk about these grammatical works in the manner of a person who is at all familiar with their contents. Kalpa is another name for Ktantra (kalpa, literally * that which holds single parts together,' hence ' bundle,' in particular 'peacock's tail.'

A. Geschichte des Buddhismus in Indien (I6O8):I Trantha

Ein Genosse des Knigs Nanda war der Brahmane Panini, welcher im Westen in Bhlrukavanageboren war. Als er einen Handliniendeuter gefragt hatte, ob er die Wortlehre innehaben werde oder nicht, und dieser es verneint hatte, machte er sich mit einem scharfen Scheermesser Handlinien, wandte sich an alle Meister der Sprachlehre auf der Erde, trieb dieselbe eifrigst und da er noch immer nicht befriedigt war, bannte er durch seine Ausdauer seinen Schutz-Gott herbei. Als dieser sein Antlitz zeigte und a, i, u aussprach, erlangte er alle Stcke der in der Dreiwelt befindlichen Laute. Die Heterodoxen [the Hindus] behaupten, dass es Isvara gewesen, haben jedoch dafr keine eigenen Quellen, die Orthodoxen [the Buddhists] aber sagen, dass es Avalokitesvara gewesen und haben als Quelle die Vorhersagung aus dem Majusnmlatantra; " Der Brahmanensohn Pnini wird sicherlich mit der vollendeten Einsicht eines Srvaka [a Buddhist monk], meiner Vorhersagung gemss, die Majestt des Herrn der Welt durch seinen Zauberspruch herbeibannen ". Dieser Pnini verfasste das Pninivykarana genannte grammatische Stra, welches 2000 Sloka's umfasst und zwar 1000 Sloka's der Wortbildung and 1000 Sloka's der Erluterung. Dies ist gleichsam die Wurzel aller Grammatiken. Vor ihm gab es keine schriftlich abgefassten Sstra's der Wortbildung und da kein System, welches die Sache unter Gesichtspunkte brachte, bestand, so wurden die einzelnen Sprachkundigen, wenn jene von zwei bis zwei Verbindungen anfangend einzelnes zusammenbrachten, fr besonders gelehrt gehalten. Obwohl es in Tibet heisst, dass das Indravykarana lter sei, so wird unten gesagt werden, dass es, obwohl es vielleicht in der Gtterregion frher dawar, in Aryadesa nicht frher erschien. Wenn auch die Pandita's behaupten, dass das ins Tibetische bersetzte Candravykarana mit Pnini, das Kalpavykarana mit dem Indravykarana bereinstimme, so sagt man doch allgemein, dass besonders das Pninivykarana durch die ausgedehnte Ausfhrlichkeit der Bedeutungen und die systematische vollstndige Einsicht etwas sehr Seltenes sei.

B. Geschichte des Buddhismus in Indien (1608): II Trantha

In the fifteenth chapter of his " History," Trantha describes the events that took place at the time of the Buddhist philosopher (crya 'teacher') Ngrjuna, a person around whom numerous legends were woven and who is said to have converted king Udayana to Buddhism. The purohita or domestic priest of king Udayana was friend of Ngrjuna and a grammarian. His name, Vararuci, is in fact another name of Ktyyana, the author of the vrttiks or rules of interpretation (see page 104 of this volume) on Pnini's grammar. Trantha then gives some information about Vararuci and the relationship between the Pninian and the Buddhist systems of grammar (Schiefner 1869, 73-76). The form mamodaksinca as used in this selection seems to be a corruption combining several words as they occur in the sentences describing this episode; for example, in the Kathsaritsgara (6.114-116), the queen first said, modakair deva paritdaya mam, 'don't hit me with water, my lord1 and then explained it later by dissolving the sandhi between mand udakair, as follows: udakaih sica mtvam mam, 'don't (you) splash me with water.' But modaka is also a kind of sweetmeat. The linguistic part of the legend is reported more accurately by al-Brn (see page 22). The legends reported here probably reflect ancient rivalries between grammatical schools. Sanmukhakumra 'the six-faced

25

Geschichte des Buddhismus in Indien : II

youth ' is a name of the god Krttikeya, whose vehicle is a peacock (hence kalapa 'peacock'stail'). Varnasamampya, 'enumeration of sounds,' refers to the pratyhrastra or Sivastra. Patajali, whom tradition identifies with the author of the Yogastra ofthat name, is represented in iconography as a sage who is a serpent below the waist (cf. frontispiece, Plate I).

Der Freund desselben Acarya, der Acarya und Brahmane Vararuci lebte als Purohita des Knigs Udayana. Zu der Zeit kannte eine jngere Gattin des Knigs ein wenig die Grammatik, der Knig kannte sie aber nicht. Zur Zeit als sie im Lusthain im Wasser spielten und der Knig sie mit Wasser bespritzte, sagte sie zu ihm : mamodaksinca d.h. in tibetischer Sprache " Bespritze mich nicht mit Wasser." Der Knig aber verstand in Uebereinstimmung mit der Sprache des Sdens einen in Sesaml gekochten Erbsenkuchen und gab ihr einen solchen. Da dachte die Knigin, dass es besser sei zu sterben als mit einem solchen ochsengleichen Knige zu leben und schickte sich an sich zu tdten, wurde aber vom Knige ergriffen, welcher sich ans Lernen der Sprache machte und von dem Brahmanen Vararuci fleissig lernte, allein daer_nicht im Geringsten vorwrts kam, nahm er Unterricht vom Acarya Saptavarman. Die Geschichte des Acrya Vararuci ist folgende. Der der Buddha-Lehre eifrig ergebene Brahmane, welcher der sechs Werke beflissen war und zu der Zeit als der ehrwrdige Ngrjuna Pandita in Nlanda war, mit ihm bekannt wurde, stammte aus dem stlich von Magadha belegenen Lande Chagala. Als er 12 Jahre hindurch das ehrwrdige Avalokitesvaramantra hergesagt hatte und ihm endlich ein Brandopfer mit Zurstung von 400,000 in Gold gebracht hatte, erschien Avalokitesvara offenbar und fragte, was er wolle. " Ich wnsche durch die acht grossen Siddhi's [magical attainments] das Wohl aller belebten Wesen zu bewirken, und dass du Mahklazu meinem Diener machest." Als ihm dies gewhrt war, konnte er fortan jeden Zauber nach Wunsch vollziehen und durch die acht Siddhi's die Kugelchen u.s.w. erwies er tausendfach den belebten Wesen Nutzen ; die 8000 Siddhi-Besitzer erkannten ihn als ihren Lehrer an, und alle Wissenschaften hatte er von selbst inne ohne sie gelernt zu haben. Als er sich darauf nach dem Sden begeben hatte, wohnte er im Lande des mit grossem Reichthum versehenen Knigs Sntivhana und nachdem er dort durch die Mantra's und Tantra's den lebenden Wesen Nutzen schaffend gelebt hatte, erwies er, als er nach VrnasT kam, wo zu der Zeit der Knig Bhlmasukla herrschte, den lebenden Wesen noch grsseren Nutzen . . . Als er darauf nach dem Sden gekommen war und dort der Knig Udayana die Sprache lernen wollte, aber keinen Lehrer finden konnte, welcher Pnini's Buch vollstndig kannte und als er erfahren, dass der Ngarja [king of snakes] Sesaden Pnini vollstndig kenne, so rief der Brahmane Vararuci diesen durch die Kraft der Mantra's herbei und vermochte ihn eine ausfhrliche Erklrung des ganzen Sinnes von Pnini in 100,000 Sloka's zu geben, welche der Acarya niederschrieb; beide waren aber durch einen Vorhang getrennt. Als 25,000 Sloka's vorber waren, wnschte der Acarya zu sehen, welcher Art sein Krper wre, lftete den Vorhang und erblickte eine grosse sich hinstreckende Schlange, welche voll Schande davon lief. Darauf schrieb der crya selbst die Erklrung weiter, aber es sind dort nicht mehr als 12,000 Sloka's. Diese beiden Werke zusammen sind bekannt unter dem Namen der von dem Ngagelehrten Gram-

26 Trantha

matik. Es w i r d d o r t von der Sprache und anderen Wissenschaften vielfach gelehrt. Endlich soll ihn Mahklaauf seine Schultern genommen und ihn auf den Gipfel des Sumeru nach Prijtaka getragen haben. Da der Knig Udayana der von dem Acrya Vararuci gemachten Erklrung nicht t r a u t e , befahl er dem Brahmanen Saptavarman den Sanmukhakumrazu bannen. Als dieser herbeigebannt war, fragte er, was er w o l l e . " Gieb mir das Indravykarana." Als der G o t t nur die W o r t e Siddhovarnasammnya ausgesprochen hatte, erfasste er die Bedeutung aller der Laute. Frher w u r d e in den in Tibet bekannten Geschichten erzhlt, dass Sanmukhakumra vom Kalpa [see page 25] die vier ersten Capitel d i c t i r t habe und Kalpa als Zusammenfgung der Theile zu fassen sei, so wie in den Pfauenschweiffedern die verschiedenfarbigen Theile zusammengefgt w e r d e n . Allein es verhlt sich nicht so; Kalpa ist von Saptavarman selbst verfasst und die Bedeutung Zusammenfassung der Theile ist die, dass alle nthigen Theile zusammengefasst sind. Ebenso w i r d der Name dieses cryafalsch als Isvaravarman e r k l r t , wie sich die Form Sarvavarman flschlich eingeschlichen hat; Saptavarman bedeutet aber: Siebenpanzer.

The Foundations of Western Scholarship

Plate II

A bronze mage of Patajali, which shows a beard and an enlarged rounded cranium that are the marks of sages in late South Indian iconography. This bronze of the fourteenth or fifteenth century is in a private collection.

6
Jean Franois Pons (1698-1752)

Among the first Western scholars to hear about linguistics in India was Filippo Sassetti (1540-1588). Sassetti was greatly impressed by the Indian discovery that different sounds are produced by the various movements of the mouth and the tongue. He himself extended this idea even further by attributing the large number of sounds in Indian languages to the widespread native custom of chewing betel leaves and areca nuts (Thieme 1957a, 267 note). Several missionaries who worked in India came in closer contact with Sanskrit and wrote grammars of Sanskrit in Latin, for which they most probably made use of the Indian tradition. Around 1660, Henrich Roth, S.J., composed a grammar of Sanskrit, the manuscript of which was recently discovered in the National Central Library in Rome by A. Camps, O.F.M. In 1790 a grammar was brought out in Rome by Paulinus of St. Bartholomew, O.C.D. Little is known about these grammars, and they do not seem to have had any direct bearing on the origins of Indology. Some specific information about the Indian grammarians was provided by another Jesuit who worked in India, Father Jean Franois Pons, S.J. (1698-1752). Pons, who was born in Rodez, France, left for India in 1726. He wrote a letter from Karikal, in Southeast India, on November 23,1740 to Father du Halde, another Jesuit priest. This letter was soon published in Lettres difiantes et curieuses, crites des Missions trangres, par quelques Missionaires de la Compagnie de JESUS, XXVI. Recueil (Paris 1743). Pons refers to an abridgment of a grammar which he had earlier made and sent to Rome, but about which nothing was known until Filliozat discovered the manuscript in Paris in the Bibliothque du Roi, where it had been studied by A. L. de Chzy, who occupied the first chair of Sanskrit (founded for him in 1814) at the Collge de France (Filliozat 1937; for Chzy see below page 50). Pons's manuscript was also studied by Anquetil-Duperron, the first translator of the Upanisads (from a seventeenth-century Persian version), which were in turn made famous by Schopenhauer. The second part of Pons's letter (printed on pages 222-227 of the Lettres difiantes) is reproduced here. Of particular interest is the stress laid on the "small number of primitive elements," themselves not used (i.e., themselves abstract) from which the Sanskrit grammarians are said to derive "the infinite variety of actual forms in use*'; and also the implication that the rules of grammar are described explicitly, so that someone who " knows nothing but grammar" can apply them. The title here chosen from the table of contents of the Lettres difiantes (page 453), i.e., " Richesse et nergie . . . " contains a prfiguration of Humboldt's energeia or Thtigkeit. Anubhtisvarpcrya is the traditional founder of the Srasvata school of grammar, a non-Pninian school called after the goddess of speech, Sarasvat. The Srasvata grammar is indeed greatly abridged, but is probably not older than the thirteenth century A.D. The passage about king Jamour refers to the Jaumara school of grammar, another non-Pninian school, founded in the thirteenth century by Kramadsvara. The school derives its name from its most famous grammarian, Jumaranandin, who is referred to in the manuscripts as mahrjdhirja 'sovereign king of great kings', and who was accordingly ridiculed by opponents as a member of the low weaver caste (Belval kar 1915, 91 -96,108-109).

Richesse et nergie de la langue Samskret, et comment et par qui elle a t rduite en Grammaire (1740)
Jean Franois Pons

La Grammaire des Brahmanes peut tre mise au rang des plus belles sciences; jamais l'Analyse & la Synthse ne furent plus heureusement employes, que dans leurs ouvrages grammaticaux de la langue Samskret ou Samskroutan. Il me parot que cette langue si admirable par son harmonie, son abondance, & son nergie, toit autrefois la langue vivante dans les pays habits par les premiers Brahmanes. Aprs bien des sicles elle s'est insensiblement corrompue dans l'usage commun, de sorte que le langage des Anciens Richi ou Pnitens dans les Vedam ou livres sacrs, est assez souvent inintelligible aux plus habiles, qui ne savent que le Samskret fix par les grammaires. Plusieurs sicles aprs l'ge de Richi, de grands Philosophes s'tudirent en conserver la connoissance, telle qu'on l'avoit de leur tems, qui toit, ce qu'il me semble, l'ge de l'ancienne posie. Anoubhout fut le premier qui forma un corps de grammaire, c'est le Sarasvat, ouvrage digne de Sarasvadi, qui est, selon les Indiens, la Desse de la parole, & la parole mme. Quoique ce soit la plus abrge des grammaires, le mrite de son antiquit l'a mise en grande vogue dans les coles de l'Indoustan. Pania aid du Sarasvat composa un ouvrage immense des rgies du Samskret. Le Roi Jamour le fit abrger par Kramadisvar\ & c'est cette Grammaire, dont j'ai fait l'abrg, que j'envoyai, il y a deux ans, & qui vous aura sans doute t communique; Kalap en composa une plus propre aux sciences. Il y en a encore trois autres de diffrens Auteurs, la gloire de l'invention est principalement due h Anoubhout. Il est tonnant que l'esprit humain ait pu atteindre la perfection de l'art, qui clatte dans ces Grammaires: les Auteurs y ont rduit par l'Analyse la plus riche langue du monde, un petit nombre d'lmens primitifs, qu'on peut regarder comme le caput mortuum de la langue. Ces lmens ne sont par eux-mmes d'aucun usage, ils ne signifient proprement rien, ils ont seulement rapport une ide, par exemple Kru l'ide d'action. Les lmens secondaires qui affectent le primitif, sont les terminaisons qui le fixent tre nom ou verbe, celles selon lesquelles il doit se dcliner ou conjuguer un certain nombre de syllabes placer entre l'lment primitif & les terminaisons, quelques propositions, &c. A l'approche des lmens secondaires le primitif change souvent de figure; Kru, par exemple, devient, selon ce qui lui est ajout, Kar, Kr, Kri, Kir, Kr &c. La Synthse runit & combine tous ces lmens & en forme une varit infinie de termes d'usage. Ce sont les rgies de cette union & de cette combinaison des lmens que lagrammaire enseigne, de sorte qu'un simple colier, qui ne sauroit rien que lagrammaire, peut en oprant, selon les rgies, sur une racine ou lment primitif, en tirer plusieurs milliers de mots vraiment Samskrets. C'est cet art qui a donn le nom la langue, car Samskret signifie synthtique ou compos. Mais comme l'usage fait varier l'infini la signification des termes, quoiqu'ils conservent toujours une certaine analogie l'ide attache la racine, il a t ncessaire de dterminer le sens par des Dictionnaires. Ils en ont dix-huit, faits sur diffrentes mthodes. Celui qui est le plus en usage, compos par Amarasimha, est rang peu prs selon la mthode qu'a suivi l'Auteur de l'/ndiculus Universalis. Le Dictionnaire intitul Visvbhidhnam, est rang par ordre alphabtique, selon les lettres finales des mots. Outre ces Dictionnaires gnraux, chaque science a son introduction, o l'on apprend les termes propres qu'on chercheroit

32 Jean Franois Pons

en vain par t o u t ailleurs. Cela a t ncessaire pour conserver aux Sciences un air de m/stre, telement affect aux Brahmanes, que non contens d'avoir des termes inconnus au vulgaire, ils ont envelopp sous des termes mystrieux les choses les plus communes.

7
Henry Thomas Colebrooke (1765-1837)

It is uncertain how much Sanskrit any of the early missionaries actually knew. The distinction of being the first European who knew Sanskrit well goes to Sir Charles Wilkins (1749M836). Sir William Jones (1746-1794), who was on all accounts the founder of Sanskrit studies in the West, was rather pessimistic about the future of these studies. He wrote to Wilkins on April 24,1784: " Happy should I be to follow you in the same track; but life is too short and my necessary business too long for me to think at my age of acquiring a new language" (Mukherjee 1968, 94). And on October 6,1787: "You are the first European that ever understood Sanskrit, and will, possibly, be the last" (Windisch 1917,1, 23 note). But earlier in the same year, on July 22,1787, Jones had written to George John Spencer (his former pupil and lifelong friend): " I have the delight of knowing that my studies go hand in hand with my duty, since I now read both Sanscrit and Arabick with so much ease that the native lawyers can never impose upon court in which I sit. I converse fluently in Arabick with Maulavis and in Sanscrit with Pundits and in Persian with nobles of the country" (Mukherjee 1968,129). Others judged differently, as we learn from an anecdote related by G. H. Cannon (1958) from William Dick's letter to Sir Walter Scott of August 23,1819 : "Shortly after arriving in Calcutta, Jones found himself using his Persian. He was sitting beside a Persian scholar when several learned Indians came to pay their respects. He addressed them in his ' Persian,' which was so incomprehensible that they thought it was English" (in Sebeok 1966,1,55). Wilkins appears to have been the first to write a grammar of Sanskrit in English, for he informs us that the first printed sheet of his work was destroyed by fire in 1795 (Mller 18702, vi). The earliest grammars written by Englishmen were the result of studies with the help of Indian pandits and were based directly or indirectly on works in the Indian grammatical traditionin particular, on two works in the Pninian tradition, i.e., the Ksik by Jayditya and Vmana (A.D. seventh century) and the Siddhntakaumudof Bhattojidlksita(seventeenth century); and on one in the non-Pninian tradition, i.e., the Mugdhabodha of Vopadeva (thirteenth century ?). The first published and the best (facile princeps, as Max Mller was to call it) among these early grammars was written by Henry Thomas Colebrooke (1765-1837). It was published in 1805 but was never completed, because the printing of Sanskrit characters, manufactured in Calcutta, was so crude that the examples Colebrooke needed to illustrate his statements would have required an excessively large volume (A. W . von Schlegel, page 58 of this volume). Other early grammars were written by H. P. Foster (who delivered his manuscript to the Council of Fort William at Calcutta in 1804) and by W. Carey (published 1806). These four authors must have all been familiar to some extent with Indian methods of grammatical analysis, as their grammars, especially the one by Colebrooke, demonstrate. The question of whether any of the early Sanskritists knew Pnini (Pons had referred to him as Pania) has been raised particularly with reference to Jones, who is known to have studied the Siddhantakaumud; Jones informs us in a letter of August 18,1792 that he finished "the attentive reading of this grammar" (Emeneau 1955,148). It has been doubted whether Jones knew Pnini (Master 1956,186-187). But such discussions result from confusion. Both the Ksik ana the Siddhantakaumud are commentaries on Pnini's Astadhyyand quote, explain, and illustrate the stras or rules of

34 Henry Thomas Colebrooke

this grammar. One difference between the t w o commentaries is that the Ksik adheres t o the o r d e r of rules as given by Pnini, whereas the Siddhantakaumud rearranges the rules in a different order. W h o e v e r studies either of these commentaries is, therefore, faced w i t h a study of Panins's rules, and even a less attentive student than Jones could not fail t o observe that t h e r e is a strakra ' maker of rules,' whose teachings the commentary seeks t o explain. O f course, Sir W i l l i a m might not have k n o w n that Pnini had lived some t w o millenia before Bhattojdksita. But t o say that he knew the Siddhantakaumud w i t h o u t knowing Pnini makes no sense. This conclusion is f u r t h e r c o r r o b o r a t e d by the fact that a statement of Colebrooke's implies that Jones knew Pnini. According t o Colebrooke, Jones called the stras of Pnini, when studied w i t h o u t a commentary, " d a r k as the darkest oracle." C o l e b r o o k e , one of the first and best W e s t e r n Sanskritists, had scientific rather than literary interests. Though he laid the foundations for a variety of Sanskrit studies, he c o n t r i b u t e d especially t o the study of Hindu mathematics, astronomy, law, and epigraphy. Similar t o al-Brn in these respects, he also had no taste for poetry. His son and biographer w r o t e that " t h e false taste of O r i e n t a l poetry especially repelled h i m " ( " B u t t h e n , t h e r e is no evidence that he enjoyed English poetry e i t h e r " : Ingalls 1960,193). He d i d , however, like al-Brn, w r i t e on Sanskrit meter and poetics. Ingalls is of the opinion that in all probability " he read and understood more Sanskrit texts than all the Europeans of his t i m e put t o g e t h e r " (loc. cit. 194). Colebrooke also t u r n e d t o the study of the Indian grammarians, f r o m which " he derived something of the same pleasure he derived f r o m mathem a t i c s " (Ingalls loc. cit., 193). He was a practical scholar and saw t o it that first things w e r e done first. He was largely responsible for the publication of Pnini's grammar in 1809 in Calcutta (begun by Dharandhara and completed by Ksintha). This led in t u r n t o Boehtlingk's editions of the AstdhyyJ. Colebrooke also provided the first reliable and detailed information on the Indian grammarians, in an article " O n the Sanskrit and Prcrit Languages," Asiatic Researches 7 (1803,199-231), f r o m which the first t w e n t y pages, dealing w i t h Sanskrit, are here included ( o m i t t i n g some of the footnotes). There are remarkably few inaccuracies in this account, notwithstanding its wealth of detail. Mention may here be made only of the confusion around Bhartrhari. The philosopher-grammarian by that name w r o t e a commentary on the Mahbhsya (see page 392 of this volume), but did not w r i t e the kriks which occur in the t e x t . He was also not the same as the poet, Bhartrhari, author of the three satakas (each theoretically of a hundred stanzas). A m o n g the interesting remarks Colebrooke made, attention may be drawn t o his characterization of the discovery of the linguistic zero. The characterization of the t e x t , " t h e addition o r the subs t i t u t i o n of one o r more elements," fits context-free rules perfectly well. The chart reproduced at the beginning of this volume gives a survey of the Sanskrit grammarians, t h e i r schools (in capitals), and t h e i r approximate dates. A Vedic column has been added, since the earliest activity of linguistic analysis was setting up the Padaptha o r ' w o r d - f o r - w o r d analysis' of each of the oldest Vedic texts. Pnini mentioned ten predecessors by name (cf. pages 104 I06 of this volume).

On the Sanscrt and Prcrit Languages (1803) Henry Thomas Colebrooke

In a treatise on rhetorick, compiled for the use of Manicya Chandra, Rj ofTirabhucti or Tirht, abrief enumeration of languages, used by Hindu poets, is quoted from two writers on the art of poetry. The following s a literal translation of both passages. "Snscrita, Prcrito, Paischand Mgad'h, are in short the four paths of poetry. The Gods, &c. speak Snscrito; benevolent genii, Prcrito; wicked demons, Paisch; and men of low tribes and the rest, Mgod'h. But sages deem Snscrita the chief of these four languages. It is used three ways ; in prose, in verse, and in a mixture of both." " Language, again, the virtuous have declared to be fourfold, Snscrita [or the polished dialect,] Prcrito [or the vulgar dialect], Apobhronso [or jargon], and Misra [or mixed]. Snscrito is the speech of the celestials, framed in grammatical institutes; Prcrito is similar to it, but manifold as a provincial dialect, and otherwise; and those languages which are ungrammatical, are spoken in their respective districts." The Paisch seems to be gibberish, which dramatick poets make the demons speak, when they bring these fantastic beings on the stage. The mixture of languages, noticed in the second quotation, is that which is employed in dramas, as is expressly said by the same author in a subsequent verse. It is not then a compound language, but a mixt dialogue in which different persons of the drama employ different idioms. Both the passages above quoted are therefore easily reconciled. They in fact notice only three tongues. 1. Sanscrit, a polished dialect, the inflections of which, with all its numerous anomalies, are taught in grammatical institutes. This the dramatic poets put into the mouths of Gods and of Holy personages. 2. Prcrit, consisting of provincial dialects, which are less refined, and have a more imperfect grammar. In dramas it is spoken by women, benevolent genii, &c. 3. Mgod'h, or Apabhransa, a jargon destitute of regular grammar. It is used by the vulgar, and varies in different districts: the poets accordingly introduce into the dialogue of plays a provincial jargon spoken by the lowest persons of the drama1. The languages of India are all comprehended in these three classes. The first contains Sanscrit, a most polished tongue, which was gradually refined until it became fixed in the classic writings of many elegant poets, most of whom are supposed to have flourished in the century preceding the Christian aera. It is cultivated by learned Hindus throughout India, as the language of science and of literature, and as the repository of their law civil and religious. It evidently draws its origin (and some steps of its progress may even now be traced) from a primeval tongue which was gradually refined in various climates, and became Sanscrit in India;
Snscrita is the passive participle of a compound verb formed by prefixing the preposition sam to the crude verb cri, and by interposing the letter s when this compound is used in the sense of embellishment. Its literal meaning then is "adorned ; " and when applied to a language, it signifies II polished." Prcrito is a similar derivative from the same crude verb, with pra prefixed : the most common
I

acceptation of this word is "outcast, or man of the lowest class; " as applied to a language, it signifies "vulgar." Apobhronsa is derived from bhras to fall down : it signifies a word, or dialect, which falls off fro m correct etymology. Grammarians use the Snscrita as signifying "duly formed or regularly inflected ; " and Apabhransa for false grammar.

36 Henry Thomas Colebrooke

Pahlav in Persia, and Greek on the shores of the Mediterranean. Like other very ancient languages, Sanscrit abounds in inflections, which are, however, more anomalous in this, than in the other languages here alluded t o ; and which are even more so in the obsolete dialect of the Vedas, than in the polished speech of the classick poets. It has nearly shared the fate of all antient tongues, and is now become almost a dead language; but there seems no good reason for doubting that it was once universally spoken in India. Its name, and the reputed difficulty of its grammar, have led many persons t o imagine that it has been refined by the concerted efforts of a few priests, who set themselves about inventing a new language; not like all other tongues, by the gradually improved practice of good w r i t e r s and polite speakers. The exquisitely refined system by which the grammar of Sanscrit is taught, has been mistaken for the refinement of the language itself. The rules have been supposed t o be anterior t o the practice, but this supposition is gratuitous. In Sanscrit, as in every other known tongue, grammarians have not invented etymology, but have only contrived rules t o teach what was already established by approved practice. There is one peculiarity of Sanscrit compositions which may also have suggested the opinion that it could never be a spoken language. I allude t o what might be termed the euphonical o r t h o graphy of Sanscrit. It consists in extending t o syntax the rules for the permutation of letters in etymology. Similar rules for avoiding incompatible sounds in compound terms exist in all languages; this is sometimes effected by a deviation from orthography in the pronunciation of words, sometimes by altering one or more letters t o make the spelling correspond w i t h the pronunciation. These rules have been more profoundly investigated by Hindu grammarians than by those of any other nation, and they have completed a system of orthography which may be justly termed euphonical. They require all compound terms t o be reduced t o this standard, and Sanscrit authors, it may be observed, delight in compounds of inordinate length ; the whole sentence too, or even whole periods, may, at the pleasure of the author, be combined like the elements of a single w o r d , and good writers generally do so. In common speech this could never have been practised. None but well known compounds would be used by any speaker who wished t o be understood, and each w o r d would be distinctly articulated independently of the terms which precede and follow it. Such indeed is the present practice of those who still speak the Sanscrit language; and they deliver themselves w i t h such fluency as is sufficient t o prove that Sanscrit may have been spoken in former times w i t h as much facility as the contemporary dialects of the Greek language, or the more modern dialects of the Arabic tongue. I shall take occasion again t o allude t o this topick after explaining at large what are and by w h o m were composed, those grammatical institutes in which the Sanscrit language is framed, according t o the author above quoted ; or by which (for the meaning is ill conveyed by a literal translation) words are correctly formed and inflected. Pini, the father of Sanscrit grammar, lived in so remote an age, that he ranks among those ancient sages whose fabulous hist o r y occupies a conspicuous place in the Puras, or Indian thogonies. The name is a patronymick, indicating his descent from Pain ; but according t o the Paurica legends, he was grandson of Dvala, an inspired legislator. W h a t e v e r may be the t r u e history of Pini, t o him the Stras, or succinct aphorisms of grammar,

37 Sanscrit and Prcrt

are a t t r i b u t e d by universal consent. His system is g r o u n d e d on a p r o f o u n d investigation of t h e analogies in b o t h t h e regular and t h e anomalous inflections of t h e Sanscrit language. He has c o m b i n e d those analogies in a v e r y artificial manner ; and has thus compressed a most copious e t y m o l o g y i n t o a very n a r r o w compass. His precepts are indeed numerous 2 , but t h e y have been framed w i t h t h e u t m o s t conciseness ; and this great b r e v i t y is t h e result of v e r y ingenious methods w h i c h have been c o n t r i v e d f o r this end, and f o r t h e purpose of assisting t h e student's m e m o r y . In Pini's system t h e mutual relation of all t h e parts marks t h a t it must have been c o m p l e t e d by its a u t h o r ; it certainly bears internal evidence of its having been accomplished by a single effort, and even t h e c o r r e c t i o n s , w h i c h are needed, cannot be i n t e r w o v e n w i t h t h e t e x t . It must not be hence i n f e r r e d , t h a t Pini was unaided by t h e labours of earlier g r a m m a r i a n s ; in many of his precepts he cites t h e a u t h o r i t y of his predecessors 3 , sometimes f o r a deviation f r o m a general rule, often f o r a grammatical canon w h i c h has universal cogency. He has even e m p l o y e d some technical t e r m s w i t h o u t defining t h e m , because, as his c o m m e n t a t o r s r e m a r k , those t e r m s w e r e already i n t r o d u c e d by earlier grammarians. 4 N o n e of t h e m o r e ancient w o r k s , h o w e v e r , seem t o be n o w e x t a n t ; being superseded by his, t h e y have probably been disused f o r ages, and are n o w perhaps t o t a l l y lost. 5 A performance such as t h e Pinya g r a m m a r must inevitably contain many e r r o r s . T h e task of c o r r e c t i n g its inaccuracies has been executed by Ctyyana, 6 an inspired saint and law-giver, whose h i s t o r y , like t h a t of all t h e Indian sages, is involved in t h e i m p e n e t r a b l e darkness of m y t h o l o g y . His annotations, e n t i t l e d Vorticas, r e s t r i c t those among t h e Pin ya rules w h i c h are t o o vague, enlarge others w h i c h are t o o l i m i t e d , and m a r k numerous exceptions w h i c h had escaped t h e notice of Pini himself. T h e amended rules of g r a m m a r have been f o r m e d i n t o memorial verses by B h a r t r - h a r i , whose metrical aphorisms, e n t i t l e d Carica, have almost equal a u t h o r i t y w i t h t h e precepts of Pini, and emendations of Ctyyana. If t h e popular t r a d i t i o n s concerning Bhartr-hari be w e l l f o u n d e d , he lived in t h e c e n t u r y preceding t h e C h r i s t i a n / E r a 7 ; f o r he is supposed t o be t h e same w i t h t h e b r o t h e r of Vicramaditya, and t h e p e r i o d w h e n this prince reigned at U j j a y i n i is d e t e r m i n e d by t h e date of t h e Samvat /Era. The studied b r e v i t y of t h e Pinya Stras renders t h e m in t h e highest degree obscure. Even w i t h t h e k n o w l e d g e of t h e key

to their interpretation, the student finds them ambiguous. In the application of them when understood, he discovers many seeming
Not fewer than 3996. Scalya, Grgya, Csyapa, Glava, Scat'yana, and others. 4 In afew instances he quotes former grammars to refute them. 5 Definitions of some technical terms, together with grammatical axioms, are also cited from those ancient works in the commentaries on Pini. They are inserted in a compilation entitled Paribhsh, which will be subsequently noticed. The various ancient grammars of the Sanscrit tongue, as enumerated in a
3 2

memorial verse, are eight in number, and ascribed to the following authors ; viz. Indra, Chandra, Casa, Critsn, Psl, Scatyana, Pini, and Amera Jinndra. 6 This name likewise is a patronymick. 7 A beautiful poem has been composed in his name, containing moral reflections, which the poet supposes him to make on the discovery of his wife's infidelity. It consists of either three or four Satacas, or centuries of couplets.

38 Henry Thomas Colebrooke

contradictions; and, w i t h every e x e r t i o n of practised m e m o r y , he must experience t h e utmost difficulty in combining rules dispersed in apparent confusion t h r o u g h different portions of Pini's eight lectures. A commentary was t h e r e f o r e indispensably requisite. Many w e r e composed by ancient grammarians t o elucidate t h e t e x t of Pini. A most copious one on t h e emendations of his rules was compiled in very ancient times by an uncertain author. This voluminous w o r k , k n o w n by t h e t i t l e of Mahbhshya, o r t h e great commentary, is ascribed t o Patanjali, a fabulous personage, t o w h o m m y t h o l o g y has assigned t h e shape of a serpent. In this commentary every rule is examined at great length. A l l possible i n t e r p r e t a t i o n s are proposed : and t h e t r u e sense and i m p o r t of t h e rule are deduced t h r o u g h atedious t r a i n of argument, in which all foreseen objections are considered and refuted ; and t h e w r o n g i n t e r p r e t a t i o n s of t h e t e x t , w i t h all t h e arguments which can be invented t o s u p p o r t t h e m , are obviated or exploded. Voluminous as it is, t h e Mahbhshya has not exhausted t h e subject on which it treats. Its deficiencies have been supplied by t h e annotations of modern grammarians. The most celebrated among these scholiasts of the Bhshya is Caiyt'a, a learned Cashmirian. His annotations are almost equally copious w i t h t h e commentary itself. Yet t h e y t o o are loaded by numerous glosses ; among which t h e old and new Vivarans are most esteemed. The difficulty of combining t h e dispersed rules of grammar, t o inflect any one verb or noun t h r o u g h all its variations, renders f u r t h e r aid necessary. This seems t o have been anciently afforded in vocabularies, one of which e x h i b i t e d t h e verbs classed in t h e o r d e r implied by t h e system of Pini, t h e o t h e r contained nouns arranged on a similar plan. Both probably cited t h e precepts which must be remembered in conjugating and declining each verb and noun. A catalogue of verbs, classed in regular o r d e r , but w i t h few references t o t h e rules of etymology, is extant, and is k n o w n by the t i t l e of D'htupt'a. It may be considered as an appendix t o t h e grammar of Pini ; and so may his o w n treatise on t h e p r o nunciation of vocal sounds, and t h e treatise of Ysca on obsolete w o r d s and acceptations peculiar t o t h e Vda. A numerous class of derivative nouns, t o which he has only alluded, have been reduced t o rule under t h e head of Uad i, or t h e t e r m i n a t i o n u, &c. ; and t h e precepts, respecting t h e gender of nouns, have been in like manner arranged in Stras, which are f o r m e d on t h e same principles w i t h Pini's rules, and which are considered as almost equally ancient. A n o t h e r supplement t o his grammar is e n t i t l e d Gaapt'a, and contains lists of w o r d s comprehended in various grammatical rules under t h e designation of some single w o r d w i t h t h e t e r m " & c . " annexed t o it. These supplements are due t o various authors. The subject of gender alone has been treated by m o r e than one w r i t e r reputed t o be inspired, namely by Ctyyana, Gbhila, and others. These subsidiary parts of the Pinya grammar do not require a laboured c o m m e n t a r y ; excepting only t h e catalogue of verbs, which does need annotation ; and which is in t r u t h a proper g r o u n d w o r k for a complete review of all t h e rules of etymology, that are applicable t o each verb. 8 The Vrttinyasa, a very celebrated
a

The number of verbal roots amounts to 1750 nearly ; exclusive of many obsolete words omitted in the D'htupta, but noticed in the Stras as

the roots of certain derivatives. The crude verbs, however, are more numerous, because many roots, containing the same radical letters, are

39 Sanscrit and Prcrt

w o r k , is, I believe, a c o m m e n t a r y of this sort. 9 It is m e n t i o n e d by Maitrya Racshita, t h e a u t h o r of t h e D'htu pradpa, as t h e w o r k chiefly consulted by him in compiling his brief annotations on t h e D'htupt'a. A very voluminous c o m m e n t a r y on t h e catalogue of verbs was compiled under t h e patronage of Sayaha, minister of a chieftain named Sangama, and is e n t i t l e d Md'havy vrtti. It t h o r o u g h l y explains t h e signification and inflection of each v e r b ; but at t h e same t i m e enters largely i n t o scholastick refinements on general grammar. Such vast w o r k s as t h e Mhbhshya and its scholia, w i t h t h e voluminous annotations on t h e catalogue of verbs, are not adapted for general i n s t r u c t i o n . A conciser c o m m e n t a r y must have been always requisite. The best that is n o w e x t a n t is e n t i t l e d t h e Csica vrtti, or c o m m e n t a r y composed at Varasi. The anonymous a u t h o r of it, in a s h o r t preface, explains his design ; ' t o gather t h e essence of a science dispersed in t h e early commentaries, in t h e Bhshya, in copious dictionaries of verbs and of nouns, and in o t h e r w o r k s . ' He has well fulfilled t h e task w h i c h he u n d e r t o o k . His gloss explains in perspicuous language t h e meaning and application of each r u l e : he adds examples, and quotes, in t h e i r p r o p e r places, t h e necessary emendations f r o m t h e Vrticos and Bhshya. T h o u g h he never deviates i n t o frivolous disquisitions, n o r i n t o tedious reasoning, but expounds t h e t e x t as succinctly as could consist w i t h perspicuity, his w o r k is nevertheless v o l u m i n o u s ; and yet, copious as it is, t h e commentaries on it, and t h e annotations on its commentaries, are still m o r e voluminous. A m o n g s t t h e most celebrated is t h e Padamanjar of Haradatta Misra; a grammarian whose a u t h o r i t y is respected almost equally w i t h t h a t of t h e a u t h o r , on whose t e x t he comments. The annotators on this again are n u m e r o u s ; but it w o u l d be useless t o insert a long list of t h e i r names, o r of t h e titles of t h e i r w o r k s . Excellent as t h e Csica vrtti u n d o u b t e d l y is, it partakes of t h e defects w h i c h have been i m p u t e d t o Pan in i's t e x t . Following t h e same o r d e r , in w h i c h t h e original rules are arranged, it is well adapted t o assist t h e student in acquiring a critical knowledge of t h e Sanscrit tongue. But f o r one w h o studies t h e rudiments of t h e language, a different arrangement is requisite, f o r t h e sake of bringing into one v i e w t h e rules w h i c h must be r e m e m b e r e d in t h e inflections of one w o r d , and those w h i c h must be combined even f o r a single variation of a single t e r m . Such a g r a m m a r has been compiled w i t h i n a few centuries past by Rmachandra, an e m i n e n t grammarian. It is e n t i t l e d Pracryacaumud. The rules are Pan in i's, and t h e explanation of t h e m is abridged f r o m t h e ancient commentaries ; but t h e arrangement is w h o l l y different. It proceeds f r o m t h e elements of w r i t i n g t o definitions ; thence t o o r t h o g r a p h y : it afterwards exhibits t h e inflections of nouns according t o case, number, and gender; notices t h e indeclinables; and variously conjugated in different senses : the whole number of crude verbs separately noticed in the catalogue exceeds three thousand. From each of these are deduced many compound verbs by prefixing one or more prepositions to the verbal root. Such compounds often deviate very widely in their signification, and some even in their inflections, from the radical verb. The derivative verbs again are numerous ; such as causals, frequentatives, &c. Hence it may be readily perceived how copious this branch of grammar must be. 9 1 have not yet had an opportunity of inspecting either this or its gloss. It has been described to me as a commentary on the Csica vrtti.

40 Henry Thomas Colebrooke

proceeds t o the uses of the cases: i t s u b j o i n s t h e rules of opposition, by which compound terms are formed ; the etymology of patronymicks and other derivatives f r o m nouns; and the reduplication of particles, &c. In the second part, it treats of the conjugation of verbs arranged in ten classes : t o these primitives succeed derivative verbs, formed f r o m verbal roots, or from nouns. The rules concerning different voices f o l l o w : they are succeeded by precepts regarding the use of the tenses ; and the w o r k concludes w i t h the etymology of verbal nouns, gerunds, supines, and participles. A supplement to it contains the anomalies of the dialect, in which the Vda is composed. The outline of Pini's arrangement is simple; but numerous exceptions and frequent disgressions have involved it in much seeming confusion. The t w o first lectures (the first section especially, which is in a manner the key of the whole grammar) contain definitions; in the three next are collected the affixes, by which verbs and nouns are inflected. Those which appertain t o verbs, occupy the t h i r d lecture: the f o u r t h and fifth contain such as are affixed t o nouns. The remaining three lectures treat of the changes which roots and affixes undergo in special cases, or by general rules of orthography, and which are all effected by the addition or by the substitution of one or more elements. 10 The apparent simplicity of the design vanishes in the perplexity of the structure. The endless pursuit of exceptions and of limitations so disjoins the general precepts, that the reader cannot keep in view t h e i r intended connexion and mutual relation. He wanders in an intricate maze; and the clew of the labyrinth is continually slipping f r o m his hands. The order in which Kmachandra has delivered the rules of grammar is certainly preferable; but the stras of Pan in i thus detached f r o m t h e i r context are wholly unintelligible. W i t h o u t the commentator's exposition, they are indeed what Sir W i l l i a m Jones has somewhere t e r m e d t h e m , dark as the darkest oracle. Even w i t h the aid of a comment, they cannot be fully understood until they are perused w i t h the proper context. N o t w i t h s t a n d i n g this defect, Bht't'j Dcshita, 11 w h o revised the Caumudi, has for very substantial reasons adhered t o the Pinya stras. That able grammarian has made some useful changes in the arrangement of thePracriya: he has amended the explanation of the rules, which was in many places incorrect or imperfect: he has remedied many omissions; has enlarged the examples; and has noticed the most important instances where the elder grammarians disagree, or where classical poets have deviated f r o m the strict rules of grammar. This excellent w o r k is entitled Sidd'hnta Caumudi. The author has very properly followed the example of Rmachandra, in excluding all rules that are peculiar t o the obsolete dialect of the Veda, or which relate t o accentuation ; for this also belongs t o the Veda alone. He has collected them in an appendix t o the Sidd'hnta Caumudi; and has subjoined in a second appendix rules concerning the gender of nouns. The other supplements of Pini's grammar are interwoven by this author w i t h the body of his w o r k . The Hindus delight in scholastick disputation. Their grammarians indulge this propensity as much as t h e i r lawyers and t h e i r Even the expunging of a letter is considered as the substitution of a blank. 11 Descendants of Bh't't'oj in the
10

fifth or sixth degree are, I am told, now living at Benares. He must have flourished then between one and two centuries ago.

41 Sanscrit and Prcrt

sophists.12 Bht't'j Dcshita has provided an ample store of controvers/ in an argumentative commentary on his own grammar. This work is entitled Prant'a manram. He also composed a very voluminous commentary on the eight lectures of Pini, and gave it the title of Sabda Caustubha. The only portion of it I have yet seen reaches no farther than to the end of the first section of Pini's first lecture. But this is so diffusive, that, if the whole have been executed on a similar plan, it must triple the ponderous volume of the Mahbhshya itself. I have reason, however, for doubting that it was ever completed. The commentaries on the Sidd'hnto Caumud and Manram are very numerous. The most celebrated shall be here briefly noticed. 1. The Tatwa bod'hini expounds the Sidd'hnta : it is the work of Inynndra Sa ras wat i, an ascetick, and the pupil of Vamanndra Swm. 2. The Sabdendu sec'hara is another commentary on Bht't'j's grammar. It was composed by a successor, if not a descendant, ofthat grammarian. An abridgment of it, which is very generally studied, is the work of Ngsa, son of Siva Bhat't'a, and pupil of Hari Di'cshita. He was patronised, as appears from his preface, by the proprietor ofSrngavra pura.13 Though called an abridgment, this Laghu Sabdndu s a voluminous performance. 3. The Laghu Sbdaratna is a commentary on the Manram of Bht't'j Dcshita, by the author's grandson, Hari Dcshita. This work is not improperly termed an abridgment, since it is short in comparison with most other commentaries on grammar. A larger performance on the same topicks, and with the same title of Sbda ratna, was composed by a professor of this school. 4. Bata sarma Pgodiya, who is either fourth or fifth in succession from Bht't'j, as professor of grammar at Benares, has written commentaries on the Caustubha, Sbda ratna, and Sabdndu sc'hara. His father, Baidyarat'ha Bhat't'a, largely annotated the Paribhshndu see har of Ngj Bhat't'a, which is an argumentative commentary on a collection of grammatical axioms and definitions cited by the glossarists of Pini. This compilation, entitled Paribhsh, has also furnished the text for other controversial performances bearing similar titles. While so many commentaries have been written on the Sidd'hnta Caumud, the Pracrya Caumud has not been neglected. The scholiasts of this too are numerous. The most known is Crsha Pand'ita; and his work has been abridged by his pupil Jayanta, who has given the title of Tatwa chandra to a very excellent compendium.14 On the other hand, Crsha Pand'ita has had the fate common to all noted grammarians; since his work has employed a host of commentators, who have largely commented on it. The Caumudis, independently even of their numerous commentaries, have been found too vast and intricate for young students. Abridgments of the Sidd'hnta Caumud have been therefore attempted by several authors with unequal degrees of success. Of three such abridgments, one only seems to deserve present
Many separate treatises on different branches of general grammar are very properly considered as appertaining to the science of logick. 13 A town on the Ganjes, marked Singhore in Rennel's maps. It is situated above Illahabad.
12 14

Finished by him, as appears from a postscript to the book, in the year 1687 of the Samvat era. Though he studied at Benares, he appears to have been born on the banks of the Tapati, a river marked Taptee in Rennel's map.

42 Henry Thomas Colebrooke

notice. It is the Mad'hya Caumudi, and is accompanied by a similar compendium of annotations, entitled Mad'hya Mnoram. The name indicates, that it holds a middle place between the diffuse original, and the jejune abstracts called Laghu Caumudi, &c. It contains such of Pan in i 's rules as are most universal, and adds t o each a short but perspicuous exposition. It omits only the least common exceptions and limitations. W h e n Sanscrit was the language of Indian courts, and was cultivated not only by persons w h o devoted themselves t o religion and literature, but also by princes, lawyers, soldiers, physicians, and scribes; in short, by the first three tribes, and by many classes included in the f o u r t h ; an easy and popular grammar must have been needed by persons w h o could not waste the best years of t h e i r lives in the study of words. Such grammars must always have been in use; those, however, which are now studied are not, I believe, of very ancient date. The most esteemed is the Sraswata, together w i t h its commentary named Chandric. It seems t o have been formed on one of the Caumudis, by translating Pan in i's rules into language that is intelligible, independently of the gloss, and w i t h o u t the necessity of adverting t o a different context. A n o t h e r popular grammar, which is in high repute in Bengal, is entitled Mugd'habd'ha, and is accompanied by a commentary. It is the w o r k of Vpadva, and proceeds upon a plan grounded on that of the Caumudis ; but the author has not been content t o translate the rules of Pini, and t o adopt his technical terms. He has on the contrary invented new terms, and contrived new abbreviations. The same author likewise composed a metrical catalogue of verbs alphabetically arranged. It is named Cavicalpadruma, and is intended as a substitute for the D'htupt'a. The chief inconvenience attending Vpadva's innovation is, that commentaries and scholia, w r i t t e n t o elucidate poems and works of science, must be often unintelligible t o those w h o have studied only his grammar, and that writings of his scholars must be equally incomprehensible (wherever agrammatical subject is noticed) t o the students of the Piniya. Accordingly the Pandits of Bengal are cut off in a manner f r o m communication on grammatical topics w i t h the learned of other provinces in India. Even etymological dictionaries, such as the commentaries on the metrical vocabularies, which I shall next proceed t o mention, must be unintelligible t o t h e m . It appears f r o m the prefaces of many different grammatical treatises, that w o r k s , entitled Dhtu and Nma pryaa, were f o r m e r l y studied. They must have comprehended, as t h e i r t i t l e implies, " t h e whole of the verbs and n o u n s " appertaining t o the language; and, since they are mentioned as very voluminous, they must probably have contained references t o all the rules applicable t o every single verb and noun. Haradatta's explanation of the t i t l e confirms this notion. But it does not appear that any w o r k is now extant under this t i t l e . The D'htupt'a, w i t h its commentaries, supplies the place of the D'htupryaa. A collection of dictionaries and vocabularies in like manner supplies the want of the Nma pryaa. These then may be noticed in this place as a branch of grammar. The best and most esteemed vocabulary is the Amera csha. Even the bigotry of Sanear chrya spared this, when he proscribed the other works of A m e r a Sinha. Like most other Sanscrit dictionaries, it is arranged in verse t o aid the memory. Synony-

43 Sanscrit and Prcrt

mous w o r d s are collected into one o r m o r e verses and placed in fifteen different chapters, which t r e a t of as m a n / different subjects. The sixteenth contains a f e w homonymous t e r m s , arranged alphabetically in t h e Indian manner by t h e final consonants. T h e sevent e e n t h chapter is a p r e t t y full catalogue of indeclinables, which European philologists w o u l d call adverbs, prepositions, conjunctions, and interjections ; b u t which Sanscrit grammarians consider as indeclinable nouns. T h e last chapter of t h e Ameracsh is a treatise on t h e gender o f nouns. A n o t h e r vocabulary by t h e same a u t h o r is often cited by his commentators under t h e t i t l e o f Ameramd/d. N u m e r o u s commentaries have been w r i t t e n on t h e Ameracsh. The chief object of t h e m is t o explain t h e derivations of t h e nouns, and t o supply t h e principal deficiencies o f t h e t e x t . Sanscrit e t y mologists scarcely acknowledge asingle p r i m i t i v e amongst t h e nouns. W h e n unable t o trace an etymology which may be consistent with the acceptation of the word, they are content to derive it according to grammatical rules from some root to which the word has no affinity in sense. At other times they adopt fanciful etymologies from Puras or from Tantras. But in general the derivations are accurate and instructive. The best known among these commentaries of the Amera csha is the Padra chandrica, compiled from sixteen older commentaries by Vrhaspati surnamed Mucut'a, or at full length Raya Mucut'a Mai. It appears from the incidental mention of the years then expired of astronomical eras, that Mucut'a made this compilation in the 4532d year of the Caliyug, which corresponds with A. D. 1430. Achyuta Jallaci has abridged Mucut'a's commentary, but without acknowledgment, and has given the title of Vyc'hypradpa to his compendium. On the other hand, Bhnuj-Dcshita has revised the same compilation, and has corrected the numerous errors of Mucut'a: who often derives words from roots that are unknown to the language; or according to rules which have no place in its grammar. Bhnuj has greatly improved the plan of the work, by inserting from other authorities the various acceptations of words exhibited by Amera in one or two senses only. This excellent compilation is entitled Vych'ya sud'ha. The Amera csha, as has been already hinted, gives a very incomplete list of words that have various acceptations. This defect is well supplied by the Mdini, a dictionary so named from its author Mdinicar. It contains words that bear many senses, arranged in alphabetical order by the final consonants; and a list of homonymous indeclinables is subjoined to it. A similar dictionary, compiled by Mahswara, and entitled Wiswa pracsa, is much consulted, though it be very defective, as has been justly remarked by Mdinicar. It contains, however, a very useful appendix on words spelt more than one way; and another on letters which are liable to be confounded, such as v and b; and another again on the gender of nouns. These subjects are not separately treated by Mdinicar; but he has on the other hand specified the genders with great care in the body of the work. The exact age of the Mdini is not certainly known ; but it is older than Mucut'a's compilation, since it is quoted by this author. Amera's dictionary does not contain more than ten thousand different words. Yet the Sanscrit language is very copious. The insertion of derivatives, that do not at all deviate from their regular and obvious import, has been very properly deemed superfluous. Compound epithets, and other compound terms, in

44 Henry Thomas Colebrooke

which the Sanscrit language is peculiarly rich, are likewise omitted; excepting such as are especially appropriated, by a limited acceptation, either as titles of Deities, or as names of plants, animals, &c. In fact compound terms are formed at pleasure, according to the rules of grammar; and must generally be interpreted in strict conformity with those rules. Technical terms too are mostly excluded from general dictionaries, and consigned to separate nomenclatures. The Amerocsh then is less defective than might be inferred from the small number of words explained in it. Still, however, it needs a supplement. The Hrval may be used as such. It is a vocabulary of uncommon words, compiled by Purushttama, the author of an etymological work, and also of a little collection of monograms, entitled E'ccshara. His Hrval was compiled by him under the patronage of D'hrta Sinha. It is noticed by Mdinicar, and seems to be likewise anterior to the Viswa. The remaining deficiencies of the Ameracsh are supplied by consulting other dictionaries and vocabularies; such as Helynd'ha's, Vchespati's, the Dharaicsha, or some other. Sanscrit dictionaries are indeed very numerous. Purushttama and Mdinicar name the Utpalin,Sabdrnva and Sansrvrta, as works consulted by them. Purushttama adds the names of Vchespati, Vyd'i and Vicramditya; but it is not quite clear whether he mentions them as the authors and patrons of these, or of other dictionaries. Mdinicar adds afourth vocabulary called Nmaml, and with similar obscurity subjoins the celebrated names of Bhguri, Vararuchi, Sswata, Bplitaand Rantidva. He then proceeds to enumerate the dictionaries of Amera, Subhnga, Helynd'ha, Gverd'hana, Rabhasa Pala, and the Ratnacsha; with the vocabularies of Rudra, Dhananjaya, and Gangd'hara; as also the Dharaicsha, Hrval, Vrhadamara, Tricd'assha and Ratnaml. Many of these are cited by the commentators on Amera, and by the scholiasts on different poems. The following are also frequently cited ; some as etymologists, the rest as lexicographers : Swm, Durga, Sarvadhara Vmana, Chandra, and the authors of the Vaijaynt Nmanid'hna, Haima, Vrhat-nighanti, &c. To this list might be added the Ancrt'ha, dwani manjari Nnrt'ha, and other vocabularies of homonymous terms ; the Dwiructi, Bhuriprayga csha, and other lists of words spelt in more than oneway; and the various Nighantis or nomenclatures, such as the Dhanwantarinighanta and Rjanighanta, which contain lists of the materia medica; and the Nighanti of the Vda, which explains obsolete words and unusual acceptations.15 Before I proceed to mention other languages of India, it may be proper to mention, that the school of Benares now uses the Siddhnta caumud, and other works of Bhattoji, as the same school formerly did the Csic vrtti. The Pracrya caumud, with its commentaries, maintains its ground among the learned of M it'h i la or Tirht. In both places, however, and indeed throughout India, the Mahbhshya continues to be the standard of Sanscrit grammar. It is therefore studied by all who are ambitious of acquiring
15 The Niructi, as explained in Sir Wi 1 1 am Jones's treatise on the I iterature of the Hindus, belongs to the same class with the Nighanti of the Veda: and a small vocabulary under both these titles is commonly annexed

to the Rgvda to complete the set of Upavdas. There s however a much larger work entitled Niructi', and the commentators of it are often cited upon topics of general grammar,

45 Sanscrit and Prcrit

a critical knowledge of the language. The Haricric, w i t h its commentaries by Hlrja and Punjarja, was probably in use w i t h a school that once flourished at Ujjayini: but it does not seem t o be n o w generally studied in any part of India.

The Romantic Period

8
August Wilhelm von Schlegel (1767-1845)

After the foundations for the study of the Sanskrit grammarians had been laid by the early English Sanskritists, the Germans took to its study with the enthusiasm worthy of the Romantic period. The British and especially Colebrooke had had firsthand experience of India. Their scholarship was sound and often related to practical problems. Some of their interests (e.g., in Hindu lawbooks) were due to the requirements of the East India Company. The first English Sanskritists were in fact not altogether dissimilar to some of the Muslim scholars who had visited India earlier. Islam had also in other respects paved the way for the British Raj. The first German Sanskritists were free from such mundane preoccupations, but they were also much less well informed. During the Romantic period, many scholars were especially interested in Indian philosophies and religions. Though not quite so superstitious, they were like the Chinese and Tibetan Buddhists in this respect. The Germans at any rate blamed the British for neglecting the ancient culture of India. However, Colebrooke himself had done the groundwork in this area also in his Essay on the Philosophy of the Hindus (1823). He reacted with characteristic self-confidence to the charges from the continent in a letter to H. H. Wilson of December 24,1827: "Careless and indifferent as our countrymen are, I think, nevertheless, that you and I may derive some complacent feelings from the reflection that, following the footsteps of Sir W. Jones, we have with so little aid of our collaborators, and so little encouragement, opened nearly every avenue, and left it to foreigners, who are taking up the clue we have furnished, to complete the outline of what we have sketched " (Windisch 1917, I, 36). The manifesto of the Romantic tradition in Indian studies was Friedrich von Schlegel's ber die Sprache und Weisheit der Inder (1808). Though this book does not refer to the Indian grammarians, the author says in the Preface that he has "fr die indische Sprache" made use of a manuscript written in Bengali characters. "Es enthlt" (among other things) "eine kurze Grammatik des Sanskrit nach dem Mugdhobdho des Vpodevo" and "den Omoroksha." Friedrich von Schlegel (1772-1829) had tried to study Sanskrit in Paris. He learned his grammar from Alexander Hamilton (17621824), an officer of the British navy and one of the early members of the Asiatic Society of Bengal, who was on parole as a prisoner of war during the Napoleonic wars in Paris, where he had come to collect Sanskrit manuscripts for an edition of the Hitopadesa (Rocher 1968b; Schwab 1950, 74-75). There are conflicting early accounts of the activities of this remarkable Scotsman, now superseded by Rocher's monograph (1968b). In the present context, it is worth mentioning that Hamilton wrote a booklet entitled Terms of Sanskrit Grammar (1814). On May 15,1803, Friedrich von Schlegel wrote from Paris to his brother August Wilhelm that he had been doing very well: Denn vieles, vieles hab' ich erlernt. Nicht nur im Persischen Fortschritte gemacht, sondern endlich ist auch das grosse Ziel erreicht, dass ich des Samskrit gewiss bin. Ich werde binnen vier Monaten die Sakontala in der Urschrift lesen knnen, wenn, ich gleich alsdann die Uebersetzung wohl auch noch brauchen werde. Ungeheure Anstrengung hat es erfordert, da eine grosse Complication und eine eigne Methode des Divinirens und der Mhe; da ich die Elemente ohne Elementar-Bcher erlernen musste. Zuletzt ist mir noch sehr zu Statten gekommen, dass ein Englnder Hamilton, der einzige in Europa ausser Wilkins der es weiss, und

50 August Wilhelm von Schlegel

zwar sehr grndlich weiss, m i r m i t Rath wenigstens zu Hlfe kam. (Kocher 1968b, 45; Windisch 1917, I, 57-58). It seems likely that Friedrich von Schlegel never managed t o learn Sanskrit w e l l . Friedrich's older b r o t h e r , August W i l h e l m von Schlegel (1767-1845), came t o Paris t o study Sanskrit in 1815. A t that t i m e A. L. de Chzy had begun t o give courses on Sanskrit at the Collge de France. Chzy, w h o had studied Sanskrit f r o m t h e manuscript of Pons's grammar which was preserved in the Bibliothque du Roi, does not appear t o have been an exciting scholar o r teacher ( W i n disch 1917,1, 73-74). A . W . von Schlegel was not satisfied w i t h his courses and t r i e d instead t o learn Sanskrit f r o m Franz Bopp, w h o himself had come t o Paris t o learn t h e language. But t o learn Sans k r i t in Paris during those days was an enterprise fraught w i t h difficulties. Bopp had come in 1812, when no Indian languages were studied at all. Chzy's appointment did not do much t o improve the situation. Bopp w r o t e in 1814 that neither he nor Chzy h i m self could understand a t e x t that had n o t already been translated (Windisch 1917,1, 69). This was partly due t o the lack of a good dictionary. The dictionary of H. H. W i l s o n did not appear until 1819 in Calcutta; it was difficult t o get and e x t r e m e l y expensive (Windisch 1917,1, 69). Much o f t h e situation is reflected in an essay " Ueber den gegenwrtigen Zustand der Indischen Philologie," which August W i l h e l m von Schlegel w r o t e in 1819 and which was first published in the Jahrbuch der preussischen Rhein-Universitt, then in French translation in the Bibliothque universelle, and again in the Revue encyclopdique (how much in demand such information must have been !), and finally in Schlegel's own periodical, Indische Bibliothek (1,1820,1-27). It contains an interesting evaluation of the Indian grammarians and a brief description of dictionaries, here quoted from the indische Bibliothek (pages 10-13). The observation that Panini's rules and techniques may be compared to the formulas and methods of algebra, here made by Schlegel, is often found in later works (for example, Whitney, page I40 of this volume). Colebrooke had been even more specific, as we have seen, by drawing attention to techniques of addition and substitution (see page 40). It took more than a century before the formalization that these observations suggested was beginning to be attempted (cf., for example, Staal 1965a, 1965b).

A. lieber den gegenwrtigen Zustand der indischen Philologie (1819)

August Wilhelm von Schlegel

Es sind nun auch bereits drey Original-Werke ber die Indische Grammatik gedruckt: die Sprche des Pn in i nebst einer Auswahl von den Anmerkungen der Scholiasten, die Siddhnta-Kaumudi, und die kurze Sprachlehre des Vopadeva, unter dem Titel Mugdha-Bodha.1 Diese gehren aber nicht zu den Hlfsmitteln fr Anfnger, sondern zu dem ans Licht gefrderten Vorrath von Erzeugnissen der Indischen Gelehrsamkeit. Denn so schwere Bcher knnen nur erfahrne Kenner des Sanskrit lesen, und auch solchen wird Inhalt und Einkleidung noch Schwierigkeit genug machen, besonders da sie mit keiner Uebersetzung oder Erklrung in einer bekannten Sprache ausgestattet sind. Indessen ist es sehr wichtig, siezu haben: diegrammatischen Arbeitender Europischen Philologen knnen in der Folge mit diesen authentischen Quellen verglichen, darnach geprft, daraus berichtigt oder besttigt werden. Die Methode der alten Indischen Sprachlehrer st strenge wissenschaftlich, und sie legen es keineswegs darauf an, die Anfangsgrnde zu erleichtern. Sie sprechen die allgemeinen Gesetze in Formeln aus, welche den algebraischen an Krze gleichen, und mit ihnen den Vortheil gemein haben, dass, wenn man sie einmal begriffen hat, alle darunter befassten Flle mit Sicherheit aufgelst werden knnen. Was sich nicht unter eine Regel bringen lsst, wollen sie dem Gedchtnisse durch allerley mnemonische Kunstgriffe eingeprgt wissen. Sie haben auf Kpfe gerechnet, denen das spitzfindigste bald gelufig wird, und nach dem grossen und dauerhaften Ruhm ihrer Schriften zu urtheilen, haben sie sich nicht betrogen. Das nchste Bedrfniss nach der Sprachlehre sind die Wrterbcher, und hierin sind wir noch lngst nicht so gut berathen als in jenem Fache. W i r haben nichts als den Amara-Kosa,2 freylich in der vortrefflichen bearbeitung von Colebrooke. Allein der Amara-Kosa ist kein alphabetisches, sondern ein metrisch abgefasstes Real-Wrterbuch, dessen Hauptzweck ist, das Geschlecht der Nenn- und Eigenschaftswrter zu bestimmen : diese werden nicht als zwey verschiedene Classen betrachtet, sondern die letzten nur als dreygeschlechtige bezeichnet. In den ersten beyden Bchern sind die Benennungen nach der Folge der Gegenstnde geordnet; dann kommen im dritten Bande vermischte; vieldeutige Wrter, auch die der Biegung nicht empfnglichen. Die Zeitwrter sind ausgeschlossen. Der Herausgeber hat ein alphabetisches Register beygefgt, und die Stelle oder die Stellen, wo jedes W o r t vorkommt, nach der Seiten- und Verszahl angegeben. Es ist also immer ein doppeltes, oft ein mehrfaches Nachschlagen nthig; nicht selten sucht man vergeblich, denn bey vielen andern Verdiensten hat das Buch keineswegs das der Vollstndigkeit. Indessen wird es immer zu Rathe gezogen werden mssen, wenn wir auch knftig ein alphabetisches Wrterbuch besitzen, weil es merkwrdige Aufschlsse ber die Eigenthmlichkeit und den Zusammenhang der Indischen Begriffe von der Geisterwelt, der Natur und dem menschlichen Leben giebt. Da die Ausgabe
The Grammatical Sutras or aphorisms of Pnini with selections from various Commentators. Nagar Character. 2 Vol. 8. Calcutta, 1809. The Siddhnta-KaumudT, a Grammar conformable to the system of Pnini by Bhattoji Diksita. NagorJ Character, I Vol. 4to. Calcutta, 1812. The
1

MugdhiG Bodha, a Grammar by Vopa-

deva. Bengali Character, I Vol. 12, Serampore, 1807. 2 Kosa or Dictionary of the Sanscrit language, by Amara Sinha. With an English Interpretation and Annotations. By H. T. Colebrooke, Esq.

Serampoor, 1808.

52 August Wilhelm von Schlegel

von Colebrooke schon sehr selten geworden ist, (ich besitze sie nur durch die Gte meines verehrten Freundes Sir James Mackintosh) so wre sehr zu wnschen, dass in England ein neuer A b d r u c k veranstaltet werden mchte, der sich bey den Vortheilen der Europischen Typographie leicht um vieles bequemer einrichten liesse. A u f den in Calcutta erschienenen A b d r u c k des Textes von Amara-Kosa nebst drey andern hnlichen W r t e r bchern 3 ohne alle Erluterung, ist dasselbe anzuwenden, was ich oben von den Original-Sprachlehren sagte.

B. Rflexions sur l'tude des langues asiatiques (1832)

August Wilhelm von Schlegel

Though Franz Bopp (1791-1867) acted as August Wilhelm von Schlegel's Sanskrit teacher in Paris, he had himself never had proper training. Whatever the extent of his knowledge, there was no one in Germany at the time who knew more Sanskrit than he did. Yet, out of the translations from the Sanskrit that accompanied his book ber das Conjugationssystem der Sanskritsprache in Vergleichung mit jenem der griechischen, lateinischen, persischen und germanischen Sprache of 1816, only a fragment from the Mahbhrata had not been translated before, and this contained many elementary mistakes (Windisch 1917,1, 69). How his frustrated liaison with Sanskrit relates to his linguistic interest is not known. Bopp himself later considered all his work in Sanskrit subordinate to his general occupation with language. He wrote for example in 1829: " Mir ist von Allem, was Indien anbelangt, die Sprache das wichtigste, und nur in Zergliederung ihres Organismus, in Untersuchungen ber ihr Verhltniss zu den verwandten Dialekten und ihre Bedeutung in der allgemeinen Sprachenwelt trete ich mit wahrer Lust und innigem Vertrauen als Schriftsteller auf" (Windisch 1917,1, 68). Bopp's knowledge of the Sanskrit grammarians was very limited. Had it been otherwise, modern linguistics might look different. This accords at any rate with SchlegeTs interesting opinion : " Herr Bopp hat allerdings grammatischen Sinn: wenn er nur die Indischen Grammatiker fleissiger studirt htte, wenn er nicht immer Originalitt anbringen wollte, wo sie nicht hingehrt, so htte er etwas recht gutes leisten mgen " (writing to Humboldt, June 23,1829: Leitzmann 1908, 244). Bopp's limited knowledge in this area may account for his low evaluation of the "National-Grammatiken."Thus Delbrck: "Bopp hatte sie [viz., the Indian grammarians] wenig studiert, konnte auch in dem damaligen Studium der wissenschaftlichen Entwicklung wenig Nutzen aus ihnen ziehen, htte dann aber freilich auch seine gelegentliche Polemik unterlassen sollen" (Leitzmann 1908, Einleitung, XVI). Boehtlingk had judged Bopp in harsher terms (which he later, in conversations with Delbrck, regretted) in the Preface to his edition of Vopadeva's Mugdhabodha of 1847: " Der zweite Grund (fr den Entschluss, dies Buch herauszugeben) war der, dass Carey und Forster bei ihren Grammatiken Vopadevas Werk zugrunde gelegt haben, und Bopp, der weder bei seinen grammatikalischen noch bei seinen lexikalischen Werken andere als sekundre Quellen benutzt, teilweise dem letzteren von den beiden eben genannten englischen Grammatikern folgt" (Delbrck quoted in Sebeok 1966,1, 264).
3

The Amara Kosa, Medini Kosa, Tricnda Sesa, and HrvalL four

original Vocabularies, I Vol. 8, Nagan character. Calcutta, 1807.

53 L'tude des langues asiatiques

T h e avowed aim of Bopp's Ausfhrliches Lehrgebude der Sanskr ita-Sp rche of 1827, written in Berlin and dedicated to Hum. boldt, is stated in the introduction : " Ich habe die Bearbeitung einer Grammatik der Sanskrit-Sprache in der berzeugung unternommen, dass, nach dem was besonders von Wilkins und Forster in diesem Gebiete Verdienstliches geleistet worden, eine weitere Frderung des Gegenstandes nicht etwa von einer ausgedehnteren Benutzung der eingebornen Grammatiker ausgehen knne, sondern nur von einer unabhngigen Kritik der Sprache selbst*'; with a footnote: "Wilkins und Forster hatten ihre Grammatiken nicht einmal auf Pnini, sondern auf jngere einheimische Systeme basirt" (page v; quoted in Liebich 1891, 42). The materials for this grammar were in fact largely taken from the grammars of Forster and Wilkins, which Bopp held in high regard. He had to rely very heavily on these two, for his was, in fact, the first grammar not written by someone in close contact with native speakers. His judgment about Colebrooke and the Indian grammarians, on the other hand, is typical. He considered Colebrooke's grammar "ebenso lehrreich als Einleitung in das Studium der NationalGrammatiken, als ungengend und hchst dunkel als Lehrbuch der Sprache" (Windisch 1917,1, 54). Bopp's grammar was the topic of a long critical review by Christian Lassen in the Indische Bibliothek (3,1830; cf. also Verbrg in Sebeok1966, I, 229-231). Lassen commented as follows upon the words from Bopp's introduction quoted above: " Der Anfang dieser Stelle wird, ich zweifle kaum daran, bei vielen Anstoss erregen, und ich gestehe, dass, allgemein betrachtet, der darin enthaltene Grundsatz, sich mit abgeleiteten Quellen zu behelfen, und die ursprnglichen zu vernachlssigen, auf keinem Gebiete historischer Forschung gutgeheissen werden kann." This is a curious indictment of Western linguistics by Western philology on behalf of Indian linguistics. Lassen's own views on the usefulness of the Sanskrit grammarians are expressed as follows: W i r mssen also die Texte der Grammatiker kritisch prfen, und, wo sie verdorben sind, herzustellen suchen ; dann zum vollen Besitz ihres Verstndnisses uns hindurch arbeiten ; wir mssen dann die von ihnen vorgetragenen Lehren mit unbefangenem Blicke der Kritik unterwerfen, und bei diesem Geschft mit desto grsserer Vorsicht zu Werke gehen, je mehr zu befrchten st, dass das Gewicht dieser Lehren, welches uns in der That gross sein muss, uns nicht mit sich fortreisse, so dass wir einen eigenen freien und festen Standpunkt nicht behaupten ; und nach allen diesen Studien haben wir noch immer dibselbe Verpflichtung, die Sprache selbst, wie sie im wirklichen Geerauch sich zeigt, zu Rathe zu ziehen, die Sprache des freien Lebens mit der der strengen Schule zu vergleichen [both quotations from Liebich 1891, 42-43]. Bopp was also criticized because he defended the view that the Indian grammarians had created many forms to serve as roots which were not known from Sanskrit texts, and which were therefore not " real " roots. This empiricist note was to have a great future. Bopp himself says in the Preface to his Glossarium comparativum linguae sanscritae (quoted from the third edition of 1867): " Multae sunt a grammaticis Indicis inter radices receptae sunt formae non verae sunt radices, sed verborum denominativorum themata, ut e.c. kumr, kartr, mantr, stom" (' Many forms which are included by the Indian grammarians amongst the roots are not real roots, but are the stems of verbs derived from nouns, as for

54 August Wilhelm von Schlegel

example kumr' etc.). Westergaard defended the opposite point of view and went even f u r t h e r in his Radices linguae sanscritae of 1841 : " Q u u m autem multas radices neque ex nominibus inde derivatis cognoscamus, neque earum usum locis e libris classicis sumptis probare possimus, sunt qui contendant, tales radices o m n i n o non in lingua exstitisse, sed a grammaticis nescio cur mere esse fictas. Mira tarnen assertio, quum tarn paululum literae Indicae n o t a e s i n t " (quoted n Liebich 1891,43: T h e r e are many roots which we do not k n o w f r o m nouns derived f r o m t h e m , and of which we cannot establish the use f r o m classical passages and books. For this reason there are people w h o claim that such roots have never existed in the language at all, but have merely been i n v e n t e d I w o u l d n ' t know w h y b y the grammarians. But such an assertion is astonishing, given the fact that so l i t t l e is k n o w n of Indian litera-

ture'). Though A. W. von Schlegel had started to learn Sanskrit from Bopp, he was soon his equal (as for example his corrections of the latter's translations show: Windisch 1917,1, 76). Schlegel was on the whole inclined to agree with Lassen, and both scholars criticized Bopp for neglecting the Sanskrit grammarians, who in their view constituted the final court of appeal in matters pertaining to Sanskrit grammar. Bopp defended himself against Schlegel, saying that he did not regard further studies of the Indian grammarians as useless and that he had in fact recommended such work to some of his best students. But he excluded himself from such research, at least for the time being, while he continued to arrive at results in linguistics: "Ich selbst mag diese Arbeit nicht unternehmen, so lange wenigstens nicht, als mich ein selbstndiges Forschen und das Streben die Sprache durch sich selbst zu begreifen und die Gesetze zu erkennen, nach denen sie sich entfaltet, zu neuen Resultaten f h r t " (letter to Schlegel of May 26,1829: Windisch 1917, I, 77). Schlegel did not continue with this correspondence, for he noted (as he wrote to Bopp) that "vertrauliche Mittheilungen Bopp unwilkommen seien, sobald eine Divergenz der Meynungen hervortrete" (Windisch loc. cit.). A few years later, A. W . von Schlegel published his Rflexions sur l'tude des langues asiatiques (1832). In the course of this book he discusses several Sanskrit grammars, lastly Bopp's, after which he holds forth on the Indian grammarians (with an excursus on the Vedas and their accents). The passage included here is quoted from the Bonn-Paris edition of 1832, pages 31-37. The program outlined by Schlegel at the end of this passage still awaits completion, though much work toward its partial fulfillment has been done, for example, by Kielhorn, Thieme, and Renou. The monograph on the Sanskrit grammarians which is forthcoming (Joshi, Kiparsky, and Staal, in preparation) will be a further step toward its realization. The beginnings of a "catalogue" (as mentioned under item (2)) had \n fact been provided by the Terms of Sanskrit Grammar (1814) of Alexander Hamilton, the teacher of August Wilhelm's brother Friedrich. Later Renou published his Terminologie grammaticale du Sanskrit (1942,19572). The formalization suggested under (3) has recently begun to receive some attention. Item (4) shows that Schlegel did not know the ganaptha. It may be noted that SchlegeTs statement that the Vedas are in modern times recited in monotone is incorrect, both then and now. M. Bopp s'est occup avec prdilection de l'analyse comparative des langues: il adonn sur ce sujet plusieurs traits, dont l'un

55^ L'tude des langues asiatiques

crit en anglais, 1 remplis d'aperus fins et ingnieux. Sa grammaire est exacte et mthodique; on ne saurait le blmer d'avoir essay de m o n t r e r comment les formes varies du sanscrit dcoulent de certains principes fondamentaux. Mais dans les recherches sur l'unit p r i m i t i v e des langues d'une mme famille, lorsque nous essayons de nous faire une ide de leur formation graduelle, et de remonter une poque de l'antiquit dont il n'existe point de monumens crits, nous sommes sur un autre terrain que quand il s'agit des rgles positives d'une langue fixe par l'usage depuis un temps immmorial. mon avis, M. Bopp a un peu t r o p confondu les deux genres : il accorde t r o p de place ses ides favorites et mme ses hypothses. Les nombreuses innovations qu'il a introduites, ne seront probablement pas approuves par ceux qui pensent que dans un langue anciennement cultive et fixe, il faut respecter l'usage et les autorits classiques. La partie la moins satisfaisante dans toutes ces grammaires c'est la syntaxe : M M . Colebrooke et Bopp ne sont pas arrivs jusque-l; et chez les autres le petit nombre de pages rserves ce chapitre, peut peine mriter ce nom. Cependant la syntaxe est d'une importance majeure dans l'interprtation et la critique des textes. La grammaire, dans l'opinion des Brahmanes, occupe un rang trs-lev parmi les sciences humaines. Ils la mettent dans un rapport immdiat avec la thologie, cause de son utilit pour bien comprendre leurs saintes critures, et pour prserver de t o u t e c o r r u p t i o n ce dpt sacr. Ils considrent le sanscrit mme comme une rvlation. Sans q u i t t e r le point de vue profane d'une origine naturelle, on doit leur accorder qu'un organe aussi parfait de la pense et de toutes les jouissances intellectuelles, est une noble prrogative. L'histoire nous fait voir, que chez plusieurs nations possdant des langues pareilles, l'heureux instinct qui avait prsid leur f o r m a t i o n , s'est perdu ensuite, et que les langues o n t dgnr. Les anciens sages de l'Inde o n t pens qu'il ne fallait pas abandonner le sanscrit aux caprices variables et la ngligence du vulgaire. Ils l'ont enseign de bonne heure par les modles et les prceptes, et ils o n t russi le fixer sans en gner le dveloppement rgulier moyennant la drivation et la composition des mots. Ils o n t approfondi la thorie de ces deux moyens d'enrichir leur langue, tandis que les grammairiens grecs ne se sont pas seulement douts que celaft possible. Lagrammaire a t si anciennement cultive dans l'Inde que les fondateurs de cette science, Pnini, Ktyyana et Patajali, sont devenus des personnages mythologiques. Pnini passe pour le plus ancien ; en y regardant de prs, on voit pourtant qu'il a eu des prdcesseurs, puisqu'il cite dans ses aphorismes huit autres grammairiens. M. Colebrooke pense que les crits de quelques-uns d'entr'eux, peut-tre de tous, existent encore. Mais ceux que j'ai nomms sont les oracles du langage classique. " S i leurs opinions sur quelques points diffrent, d i t M. Colebrooke, on peut opter, mais s'ils sont d'accord, il faut se soumettre leur a u t o r i t . " Cependant on irait peut-tre t r o p loin en disant que, dans aucun cas, il n'est permis de les contredire. Ils sont des tmoins irrcusables sur des questions de fait; mais une partie de leur doctrine est spculative : par exemple l'tymologie, ds qu'elle dpasse le cercle des
Annals of oriental littrature, London 1824, P. I., Analytical comparison of the Sanscrit, Greek, Latin, and Teutonic
1

Languages, shewing the original identity of their grammatical structure. By F. Bopp.

56 August Wilhelm von Schlegel

analogies grammaticales. Ils savaient parfaitement leur langue maternelle, mais ils n'en connaissaient pas d'autre; nous savons moins bien le sanscrit, mais nous pouvons le compareravec d'autres langues, et rectifier par l nos vues gnrales. Nanmoins M. Bopp, dans la prface de l'dition allemande de sa grammaire, congdie formellement les gram mar i ens nationaux du sanscrit; il soutient qu'aprs ce qui en a t extrait dj, ils ne peuvent plus rien nous apprendre. C'est une grande erreur, je n'hsite pas le dire. Je pense au contraire que, pour marcher d'un pas assur dans la critique des textes, il faut tre suffisamment initi dans le systme des principaux grammairiens indiens, pour savoir les consulter au besoin. M. Colebrooke l'a pens de mme: c'est d'aprs ses ordres que deux ouvrages importans, les Aphorismes de Pnini, avec un e x t r a i t des commentaires, et la Siddhnta-KaumudJ, ont t imprims Calcutta. Sans doute, la mthode de ces grammairiens diffre totalement de celle laquelle nous sommes habitus ; elle est f o r t abstruse. Mais en revanche ils se distinguent par une brivet et une prcision admirables, par l'esprit scientifique dans la recherche des principes, et par l'exactitude scrupuleuse qu'ils mettent constater le fait de l'usage. O u t r e l a t e r m i n o l o g i e ordinaire puise dans la langue mme, et approprie seulement un emploi spcial, Pan i n i et ses successeurs ont imagin un autre systme de termes techniques. Ce sont des mots fictifs, des signes abrgs, qu'on peut comparer ceux de l'algbre. Il faut en avoir la cl, sans quoi les Aphorismes de Pnini ressemblent des nigmes plus obscures que les oracles de Bacis; de mme qu'un colier qui ne sait que les lmens de l'arithmtique, ne comprendra rien aux formules algbriques. Quel que soit notre jugement sur cette mthode, nous ne pouvons pas vouloir l'ignorer. Les commentateurs indignes sont ncessaires pour l'intelligence des livres difficiles, et les commentateurs, dans t o u t ce qui a rapport la grammaire, se servent de ces termes techniques. O n voudra bien arriver finalement la lecture des Vedas, de ce monument mmorable de l'antiquit, source premire de la doctrine brahmanique. O r les Vedas sont crits dans un langage surann, dont les licences qui se t r o u v e n t par-ci par-l chez les plus anciens potes piques, ne sont qu'un dernier reste. Pnini marque dans un grand dtail la diffrence du style sacr et de l'usage profane. O u t r e que la connaissance en est ncessaire pour la critique et l'explication des Vedas, ces formes viellies sont intressantes pour la thorie gnrale et (a comparaison des langues. Quelques-unes sont de vraies dviations, des irrgularits que l'instinct grammatical a rejetes plus tard ; d'autres fois la forme des mots et les inflexions anciennes sont plus rapproches de celles qu'on t r o u v e dans des langues affilies, et concourent prouver leur unit p r i m i t i v e . Les rgles de l'accentuation ont t laisses de ct par tous les grammairiens europens du sanscrit. A u j o u r d ' h u i dans les coles des Brahmanes on pronounce les vers des anciens textes uniquement d'aprs la quantit des syllabes, avec une rcitation monotone. Mais jadis le sanscrit a t accentu comme toutes les langues vivantes, et cette accentuation ncessairement a d avoir de l'influence sur laformation de la langue. Pnini en donne les rgles qui ne sont pas faciles comprendre, parce que les accents ne se t r o u v e n t crits nulle part. Ces raisons, auxquelles je pourrais ajouter plusieurs autres, suffiront pour m o n t r e r combien l'tude des anciens grammairiens indignes est importante. Mais elle est d'un abord trs-difficile.

57^ L'tude des langues asiatiques

Les deux ouvrages que je viens de nommer, ne sont accompagns d'aucun mot anglais depuis le titre jusqu' Ter rata; leur extrieur mme est rbutant: c'est un labyrinthe o l'on perd son temps chercher des claircissemens sur telle ou telle matire. Ainsi que tous les livres sanscrits imprims Calcutta d'aprs les ordres de M. Coiebrooke, et excuts parle braire Bbou-Rma, ce ne sont pas des ditions comme nous l'entendons, ce sont des manuscrits multiplis par l'impression. Une introduction gnrale l'tude des grammairiens originaux serait donc un ouvrage fort utile entreprendre. Pour remplir son but elle devra contenir: (1) une anal/se de leur mthode, claircie par des exemples ; (2) un catalogue de tous les termes techniques avec leur dfinition, arrang par ordre alphabtique; (3) la terminologie par signes abrgs et formules, explique dans le plus grand dtail ; (4) un rpertoire de toutes les sries de mots sujets une rgle particulire, lesquelles sont dsignes par le premier mot, plac arbitrairement la tte en y ajoutant et caetera. Horace Hayman Wilson (1784-1860), who was the author of the Sanskrit-English dictionary (1819) and who worked for many years in India with the assistance of pandits, wrote in 1830, in Calcutta, a Memorandum in which he attempted to show that he was the only deserving candidate for the newly established Boden chair of Sanskrit at Oxford. The Bishop of Calcutta forwarded this Memorandum (which stressed, among other things, the usefulness of the knowledge of Sanskrit for the Indian chaplaincy and the mission) to the Principal of Magdalen College, and Wilson was appointed in due course. The Memorandum formulated requirements for the incumbent of the new chair such as were satisfied by Wilson but not, for example, by continental scholars: I should think it an indispensable requisite in the first Professor of the Sanscrit Language, that he had acquired his knowledge in India. It is true that considerable proficiency has been attained by some learned men on the Continent, but it is evident from their publications that their reading has been very limited, and that they are far from possessing any degree of conversancy with the great body of Sanscrit Literature. Their knowledge is, in fact, of the most elementary kind, and restricted to the grammar of the language. The publications of Bopp are chiefly of this description ; and Schlegel has not ventured in translation beyond those works which have been previously translated by English Scholars. With the different departments of Hindu classical literature, with any one of them in a variety of details, and even with its grammar as studied in India, they are unfamiliar; and they must be very incompetent therefore to prepare a Student for this country, or even where general information only is sought, to convey comprehensive and correct notions of the classical writings of the Hindus, of their poetry, mythology, philosophy, and science. . . . If, however, instruction in the language alone be the object to be kept in view, the Lecturer will have to confine himself to the elements of the Sanscrit tongue, its grammar, syntax, and prosody: to explain these satisfactorily, however, he should be familiar not only with thegrammars compiled by European scholars, as Wilkins, etc. but with the original grammars read by the Pundits, the Stras of Panini, the Siddhnta KaumudT, and Mugdhabodha; for although it is by no means advisable to teach the grammar after those systems, they alone furnish a clue to the intricacies of Sanscrit grammar,

C. Controversy with H. H. Wilson

58 August Wilhelm von Schlegel

w i t h o u t which it is not easy t o make a profitable use even of modern European compilations (quoted in Schlegel 1832, Appendix F, 201-202). Though this Memorandum was first treated as a confidential document, the Bishop of Calcutta wrote to the Principal of Magdalen on May 31,1831 : "I could wish this valuable paper, which was drawn up at my request, to be circulated widely " (quoted in Schlegel 1832, 200). When this somehow reached Schlegel, who was in the habit of sending his offprints to Wilson, he went into a fit of rage and wrote Wilson a long letter, which he published as an appendix in his Rflexions in 1832. In this letter Schlegel points out numerous mistakes and inaccuracies in Wilson's own work, suggests that he had left much of the really hard work to the pandits working for him in Calcutta, complains about the high prices of Sanskrit books published in England and the curious restrictions on the sale of inexpensive Indian books outside England, etc. The letter ends with a warning: when Wilson returns to Europe, his activities will be reviewed critically. (Schlegel 1832, 167-168). Toutefois, ne craignez rien de la part de nous autres vtrans. Nous devons imiter l'impassibilit de ces Brahmanes dont nous admirons les sages maximes. Nous nous rappelons que Visvmitra perdit tout le fruit de ses pnitences pour s'tre laiss entraner un mouvement de colre, quoique la provocation ft assez forte. . . . Mais je ne vous rponds pas de nos jeunes Indianistes : ils sont aussi fongeux que zls pour leur tude, et pourraient tre tents de venger leurs ancient matres. Si vous publiez quelque ouvrage, on sera l'afft de vos mprises: et qui n'en commet pas? Si, au contraire, vous ne publiez rien, on dira qu'en partant de Calcutta, vous avez oubli d'embarquer votre savoir [an obvious reference to Wilson's pandit informants]. Croyez-moi, faites votre paix le plutt et le mieux que vous pourrez; je vous offre mes bons offices comme mdiateur. (It is interesting to read, in this connection, what Weber wrote 20 years later: " It is certainly very discouraging to see that Professor Wilson during all the time since he got his one professorship in Oxford, has not succeeded in bringing up even one Sanskrit scholar who might claim to be regarded as one who has done at least some little service to our Sanskrit philology": from a letter of Weber to Salisbury of December 28,1852 (quoted in Sebeok 1966,1,405). These exchanges show, if not atinge of imperialism, at least how knowledge regarding Sanskrit that could almost exclusively be obtained in India, such as knowledge of the Sanskrit grammarians, was beginning to become controversial. Such feelings remained in existence for a long time. In 1899, for example, no less distinguished a scholar than Hermann Oldenberg called Sir William Jones "ein unermdlich coquettirender Schnreder*' (quoted in Sebeok 1966,1,11). The Germans, in turn, were criticized not only by the English but also by the French. The skeptic Jacquemort wrote: " L'absurde de Bnars et l'absurde de l'Allemagne n'ontils pas un air de famille?" (Schwab 1950, 238 note).

9
Wilhelm von Humboldt
(1767-1835)

From 1818 to 1831 A. W . von Schlegel had an extensive correspondence with Wilhelm von Humboldt (1767-1835), in which Sanskrit studies figured prominently (edited by Leitzmann 1908; cf. Windisch 1917,1, 205-207). Humboldt defended Bopp against some of SchlegeTs criticisms and asked Schlegel at the same time numerous questions about the Indian grammarians. Schlegel had unlimited admiration for them and encouraged Humboldt to study them : " Ich hege fr sie [the Indian grammarians] eine grnzenlose Bewunderung, und wollte nur, ich verstnde sie erst vollkommen. Ich zweifle nicht, es wird Ihnen einen grossen Genuss gewhren, in der Folge an die Originalwerke dieser wissenschaftlichen Kpfe selbst zu gehen, da Sie gewohnt sind, schwere Rthsel aufzulsen " (writing to Humboldt in 1821, Leitzmann 1908, 19). As this passage shows, Schlegel knew that his own information on this topic was rather scanty. A letter of 1822, too, makes it clear that he was not sure whether the Sanskrit grammarians had ever dealt with syntax a question in which Humboldt had expressed interest: " Ew. Excellenz legen zuerst Hand an eine schwere wissenschaftliche Aufgabe: die Syntaxis des Sanskrit; denn die paar Bltter bei Carey und Wilkins, welche diese berschrift fhren, sind ja kaum der Rede werth. Sind die Indischen Grammatiker eben so karglaut? Haben sie ihren Scharfsinn so ganz an der Wortbildung erschpft, dass ihnen fr die Wortfgung gar nichts brig blieb? Ich weiss es nicht..." (Leitzmann 1908, 86). Humboldt was close to both Bopp and Schlegel and could not fail to get involved in their battles, and hence in an evaluation of the Indian grammarians. He preferred the Greek grammarians to the " Formelwesen " of the Indian grammatical tradition, as he stated in a letter to Schlegel of June 16,1829, in which he passes judgment on Bopp after stressing that his own position is Fully unprejudiced: Ich stehe ganz unparthesch zwischen Ihnen beiden. Ich kann mich, soviel Belehrung ich auch Ihnen beiden schuldig bin, einen Schler von keinem von Ihnen nennen, Sie haben beide selbst meine Irrthmer und Unkenntnisse mit fast ber hflicher Schonung behandelt, meine langjhrige Freundschaft mit Ew. Hochwohlgeboren, meine lebhafte Zuneigung und Theilnahme an Bopp stellt mich Ihnen beiden nahe, mir kann keine Eifersucht beiwohnen, da ich Sie beide in diesen Studien weit ber mir erkenne. Mein Zeugniss kann mithin von keiner Seite der Partheilichkeit verdchtig se/n. . . . Ich gebe vollkommen zu, dass Bopp die Indischen Grammatiker nicht selbst studirt hat, und ob ich gleich, wie ich augenblicklich sagen werde, dies in ihm gerechtfertigt finde, so billige ich nicht, dass er oft zu sehr dies Studium berhaupt fr nicht nothwendig erklrt. Ich thte das schon nicht, weil gerade ein solches Urtheil mit vielem Studium verbunden seyn msste, es ist aber auch gegen meine Ueberzeugung, wiewohl ich, was hier zu lang zu rechtfertigen wre, keine so vortheilhafte Meinung von den eingeborenen Grammatikern und ihrem Formelwesen habe, als Ew. Hochwohlgeboren, und die guten Griechischen fr weit vorzglicher halte (Leitzmann 1908, 232-233). Humboldt had great admiration for the Sanskrit language, which he seemed mainly to have studied from Wilkins* grammar ("ich habe mein Sanskrit ganz eigentlich zuerst ans ihm gelernt11: writing to Schlegel, June 16,1829), from Bopp's grammar, and from Bopp's and SchlegeTs editions of texts. We do not know how much he knew (he called both Bopp and Schlegel "weit ber mir" in the letter just quoted), but Bopp praised his rapid progress in

60 Wilhelm von Humboldt

a letter to the French Sanskritist Jean-Louis Burnouf (Windisch 1917,1, 83). Humboldt himself published several studies on Sanskrit, for example, on the so-called gerunds ending in -tv and on the hagavad-GJt. To Humboldt, Sanskrit was the zenith of the development of inflected languages. Only because of Sanskrit was a serious and fruitful study of language possible (Dove in Sebeok 1966,1, 98; Windisch 1917,1,82-86). Humboldt was the first to recognize the extent of Sanskrit influence in Southeast Asia (in his ber die Kawi-Sprache auf der Insel Java of 1832). The Indian grammarians are mentioned several times in Humboldt's most well known work, the Introduction to the KawiSprache, published separately under the title ber die Verschiedenheit des menschlichen Sprachbaues und ihren Einfluss auf die geistige Entwicklung des Menschengeschlechts (1836). Recently Chomsky has drawn attention to Humboldt's emphasis in this work on the creative aspect of language use (Chomsky 1964, 17-21 ; 1966, especially 19-28). Humboldt in fact relates some of the views he expresses on this topic to the Sanskrit grammarians. His concept of language as Erzeugnis finds expression in the "Regeln der Redefgung" and in the rules of Wortbildung (Chomsky 1966,19). With respect to the former, Humboldt could not refer to the Indian grammarians, for Schlegel had not enlightened him on this matter. With respect to the Wortbildung he says : " Man kann den Wortvorrath einer Sprache auf keine Weise als eine fertig daliegende Masse ansehen. Er ist, auch ohne ausschliesslich der bestndigen Bildung neuer Wrter und Wortformen zu gedenken, so lange die Sprache im Munde des Volks lebt, ein fortgehendes Erzeugniss und Wiedererzeugniss des wortbildenden Vermgens, zuerst in dem Stamme, dem die Sprache ihre Form verdankt, dann in der kindischen Erlernung des Sprechens, und endlich im tglichen Gebrauche der Rede" (Humboldt 1836,109-110). He then explains why this cannot be merely due to memory. Next he considers how it becomes possible for the vocabulary of a language to form a unity, on account of the Erzeugung of words (Humboldt 1836,111-112).

A. ber die Verschiedenheit des menschlichen Sprachbaues (1836): I

Wilhelm von Humboldt

Die Indischen Grammatiker bauten ihr, gewiss zu knstliches, aber in seinem Ganzen von bewundrungswrdigem Scharfsinn zeugendes System auf die Voraussetzung, dass sich der ihnen vorliegende Wortschatz ihrer Sprache ganz durch sich selbst erklren lasse. Sie sahen dieselbe daher als eine ursprngliche an, und schlssen auch alle Mglichkeit im Verlaufe der Zeit aufgenommener fremder Wrter aus. Beides war unstreitig falsch. Denn aller historischen, oder aus der Sprache selbst aufzufindenden Grnde nicht zu gedenken, ist es auf keine Weise wahrscheinlich, dass sich irgend eine wahrhaft ursprngliche Sprache in ihrer Urform bis auf uns erhalten habe. Vielleicht hatten die Indischen Grammatiker bei ihrem Verfahren auch nur mehr den Zweck im Auge, die Sprache zur Bequemlichkeit der Erlernung in systematische Verbindung zu bringen, ohne sich gerade um die historische Richtigkeit dieser Verbindung zu kmmern. Es mochte aber auch den Indiern in diesem Punkte wie den meisten Nationen bei dem Aufblhen ihrer Geistesbildung ergehen. Der Mensch sucht immer die Verknpfung, auch der usseren Erscheinungen, zuerst im Gebiete der Gedanken auf; die historische Kunst ist immer die spteste, und die reine Beobachtung, noch weit mehr aber der Versuch, folgen erst in weiter Entfernung idealischen oder phantastischen Systemen nach. Zuerst versucht der Mensch die Natur von der Idee aus zu beherrschen. Dies zugestanden, zeugt aber

61 Menschlichen Sprachbaues: II

jene Voraussetzung der Erklrlichkeit des Sanskrits durch sich allein von einem richtigen und tiefen Blick in die Natur der Sprache berhaupt. Denn eine wahrhaft ursprngliche und von fremder Einmischung rein geschiedene msste wirklich einen solchen thatschlich nachzuweisenden Zusammenhang ihres gesammten Wortvorraths in sich bewahren. Es war berdies ein schon durch seine Khnheit Achtung verdienendes Unternehmen, sich gerade mit dieser Beharrlichkeit in die Wortbildung, als den tiefsten und geheimnissvollsten Theil aller Sprachen, zu versenken.

B. ber die Verschiedenheit des menschlichen Sprachbaues (1836): II

Wilhelm von Humboldt

Thus, according to von Humboldt, the Indian grammarians unrealistically regarded the Sanskrit vocabulary as closed and not subject to change. But despite these limitations, they recognized the principles of word formation by which the entire vocabulary is knit together, and which constitute " t h e deepest and most mysterious portion of all languages." In the following pages Humboldt specifies this thesis in greater detail. After referring to pronouns, prepositions, and interjections, he comes to roots. Thereupon, referring to Bopp, to Sanskrit, and to the Indian grammarians, he develops a theory which resembles Bopp's idea that there is a one-to-one correspondence between the smallest morphological and semantic elements of language (cf. Kiparsky 1970). He introduces the notion of postulated or underlying roots, which Bopp, as we have already seen, had repudiated as far as the Sanskrit grammarians were concerned. But according to Humboldt, such roots might be the object of linguistic analysis without being given in actual performance: " . . . die Wurzeln brauchen nicht der wahrhaften Sprache anzugehren Denn die wahre Sprache st nur die in der Rede sich offenbarende, und die Spracherfindung lsst sich nicht auf demselben Wege abwrts schreitend denken, den die Analyse aufwrts verfolgt " (page 115). These roots may be invisible in certain languages (in particular, in inflected languages) and they may have to be postulated. But Humboldt believed that they were quite real in the sense that they had originally existed and might still be found in other languages. This is most clearly expressed in a letter to Schlegel of 1829, in which he contrasts his views with those of Bopp and says that the Indian grammarians, when postulating roots, also isolated them by a kind of purification:

In etwas, das bei Bopp hier noch zum Grunde liegt, bin ich mit ihm verschiedener Meinung. Er betrachtet, wozu freilich auch Colebrooke Anlass giebt, die Wurzeln als grammatische Abstracta. Ich hingegen halte sie fr uralte Grundwrter, die aber in der ganz flectirten Sprache, als solche, verschwinden. Hchstens mchte ich zugeben, dass die Grammatiker, welche sie (wenige Flle ausgenommen) allerdings bloss als wissenschaftliche Hlfsmittel brauchten, sie von gewissen Nebenlauten befreiten oder sonst Lautvernderungen mit ihnen vornahmen, um sie zu durchaus lauteren Quellen aller aus ihnen entspringenden Formen zu machen. Ich suche daher immer nach dem Erscheinen dieser Wurzeln in nicht-Sanskritischen Asiatischen Sprachen, und habe wohl Einiges, aber nicht Vieles bisher gefunden (Leitzmann 1908, 238).
It is in this context that it becomes especially clear how Humboldt's ideas about the creative use of language as manifested in word formation, foreshadowed in fact by Pons' "lmens primitifs" (page 31 of this volume), were influenced by the Indian gram-

62 Wilhelm von Humboldt

Humboldt believed that they were quite real in the sense that they had originally existed and might still be found in other languages. This is most clearly expressed in a letter to Schlegel of 1829, in which he contrasts his views with those of Bopp and says that the Indian grammarians, when postulating roots, also isolated them by a kind of purification : In etwas, das bei Bopp hier noch zum Grunde liegt, bin ich mit ihm verschiedener Meinung. Er betrachtet, wozu freilich auch Colebrooke Anlass giebt, die W u r z e l n als grammatische Abstracta. Ich hingegen halte sie fr uralte G r u n d w r t e r , die aber in der ganz flectirten Sprache, als solche, verschwinden. Hchstens mchte ich zugeben, dass die Grammatiker, welche sie (wenige Flle ausgenommen) allerdings bloss als wissenschaftliche Hlfsmittel brauchten, sie von gewissen Nebenlauten befreiten oder sonst Lautvernderungen m i t ihnen vornahmen, um sie zu durchaus lauteren Quellen aller aus ihnen entspringenden Formen zu machen. Ich suche daher immer nach dem Erscheinen dieser W u r z e l n in nicht-Sanskritischen Asiatischen Sprachen, und habe wohl Einiges, aber nicht Vieles bisher gefunden (Leitzmann 1908, 238). It is in this context that it becomes especially clear how Humboldt's ideas about the creative use of language as manifested in word formation, foreshadowed in fact by Pons' "lmens primitifs " (page 31 of this volume), were influenced by the Indian grammarians and by the long discussions about their "invented " roots between Bopp, Lassen, Schlegel, and others. In his ber die Verschiedenheit, Humboldt then raises the question whether or not the roots thus arrived at are really the smallest elements and cannot be subject to further analysis. Claiming that this is not mere speculation, he arrives at a kind of language atoms, the Wurzellaute, and argues that these are always expressed by the verb, since it is there that they are combined with the greatest variety of additional forms. This leads him again to the Indian grammarians, and he treats at some length their distinction between two kinds of word-forming suffixes, namely krt suffixes and undi suffixes. The facts as we now know them are as follows. Whereas krt suffixes are attached to roots to form derivatives in accordance with Pnini's rules, which specify their syntactic function and often their meaning, undi suffixes are postulated to account for a series of words which are formed from roots in irregular ways. Such words are therefore not subject to the regular analysis provided by Pnini ; they are characterized as ovyutponna 'unanalyzable*. Schlegel had said about these suffixes in a letter to Humboldt (1821) that they" sind allerdings der Theorie zu Liebe ersonnen" (Leitzmann 1908, 16), which is, therefore, not quite correct. Humboldt refers to them in connection with his theory of the Wurzellaute on pages 115-119 of the 1836 edition of ber die Verschiedenheit des menschlichen Sprachbaues, of which the beginning sentences have already been quoted. The work of Pott to which Humboldt refers is Etymologische Forschungen (1833-1836). Pnini refers twice to undi suffixes, so called because a list of them begins with uN 'the suffix uy (where N is a metalinguistic marker: Pnini 3.3.1 and 3.4.75). It is uncertain whether Pnini was in fact the author of such a list; the lists that are presently known are of much later date. One list was published in 1844 by O. Boehtlingk; another, by Th. Aufrecht, appeared in 1859 together with a vnti 'gloss' of Ujjvaladatta. For a recent study see Renou (1956b).

63 Menschlichen Sprachbaues: II

Chomsky's emphasis on t h e value of Humboldt's ideas has been timely but should not obscure the fact that some of Humboldt's formulations can lead rather easily, and have in fact led, to mystification. Contemporary Neo-Humboldtianism (with its admixture of Hegelianism) has a large share of this, and even the Sanskrit grammarians have not remained unaffected. The following is a recent example of a singularly ill-informed and muddled piece of interpretation in this tradition: "Und um diesem Zauber-Worte [brahman] seine ursprngliche Zauber-Kraft zu erhalten, haben die alten Inder schon vor 3000 Jahren eine raffinierte Grammatik erfunden, die in Europa erst in den letzten 100 Jahren erreicht und berholt worden ist. In Griechenland aber ist das magischbeschwrende Wort der Sprache zu dem wissenschaftlichen Begriff"geworden, in dem diese ursprngliche magisch-beschwrende Kraft, um mit Hegel zu sprechen, erst ihre Wirklichkeit und Wahrheit erreicht hat" (Lohmann 1960,183-184). Mit dem Namen der Wurzeln knnen nur solche Grundlaute belegt werden, welche sich unmittelbar, ohne Dazwischenkunft anderer, schon fr sich bedeutsamer Laute, dem zu bezeichnenden Begriffe anschliessen. In diesem strengen Verstande des Worts, brauchen die Wurzeln nicht der wahrhaften Sprache anzugehren ; und in Sprachen, deren Form die Umkleidung der Wurzeln mit Nebenlauten mit sich fhrt, kann dies sogar berhaupt kaum, oder doch nur unter bestimmten Bedingungen der Fall sein. Denn die wahre Sprache ist nur die in der Rede sich offenbarende, und die Spracherfindung lsst sich nicht auf demselben Wege abwrts schreitend denken, den die Analyse aufwrts verfolgt. Wenn in einer solchen Sprache eine Wurzel als W o r t erscheint, wie im Sanskrit, yudh, Kampf, oder als Theil einer Zusammensetzung, wie in dharmavid, gerechtigkeitskundig, so sind dies Ausnahmen, die ganz und gar noch nicht zu der Voraussetzung eines Zustandes berechtigen, wo auch, gleichsam wie im Chinesischen, die unbekleideten Wurzeln sich mit der Rede verbanden. Es ist sogar viel wahrscheinlicher, dass, je mehr die Stammfaute dem Ohre und dem ewusstsein der Sprechenden gelufig wirden, solche einzelnen Flle ihrer nackten Anwendung dadurch eintraten. Indem aber durch die Zergliederung auf die Stammlaute zurckgegangen wird, fragt es sich, ob man berall bis zu dem wirklich einfachen gelangt ist? Im Sanskrit ist schon mit glcklichem Scharfsinn von Bopp, und in einer, schon oben erwhnten, wichtigen Arbeit, die gewiss zur Grundlage weiterer Forschungen dienen wird, von Pott gezeigt worden, dass mehrere angebliche Wurzeln zusammengesetzt oder durch Reduplication abgeleitet sind. Aber auch auf solche, die wirklich einfach scheinen, kann der Zweifel ausgedehnt werden. Ich meine hier besonders die, welche sich von dem Bau der einfachen oder doch den Vocal nur mit solchen Consonantenlauten, die sich bis zu schwieriger Trennung mit ihm verschmelzen, umkleidenden Sylben abweichen. Auch in ihnen knnen unkenntlich gewordene und phonetisch durch Zusammenziehung, Abwerfung von Vocalen, oder sonst vernderte Zusammensetzungen versteckt sein. Ich sage dies nicht, um leere Muthmassungen an die Stelle von Thatsachen zu setzen, wohl aber, um der historischen Forschung nicht wilIkhrlich das weitere Vordringen in noch nicht gehrig durchschaute Sprachzustnde zu verschliessen, und weil die uns hier beschftigende Frage des Zusammenhanges der Spra-

64 Wilhelm von Humboldt

selben von dieser bestimmten A r t ist. Die Undi-Sufnxe begreifen, gerade im Gegentheil, nur Benennungen concreter Gegenstnde, und in den durch sie gebildeten W r t e r n ist der dunkelste Theil gerade das Suffix selbst, welches den allgemeineren, den W u r z e l laut modificirenden Begriff enthalten sollte. Es ist nicht zu lugnen, dass ein grosser Theil dieser Bildungen erzwungen und offenbar ungeschichtlich ist. Man erkennt zu deutlich ihre absichtliche Entstehung aus dem Princip, alle W r t e r der Sprache, ohne Ausnahme, auf die einmal angenommenen W u r z e l n zurckzubringen. U n t e r diesen Benennungen concreter Gegenstnde knnen einestheils fremde in die Sprache aufgenommene, andrentheils aber unkenntlich gewordene Zusammensetzungen liegen, wie es von den letzteren in der That erkennbare bereits unter den UndiW r t e r n giebt. Es ist dies natrlich der dunkelste Thei! aller Sprachen, und man hat daher mit Recht neuerlich vorgezogen, aus einem grossen Theile der U n d i - W r t e r eine eigne Classe dunkler und Ungewisser Herleitung zu bilden. Das Wesen des Lautzusammenhanges beruht auf der kenntlichkeit der Stammsylbe, die von den Sprachen berhaupt nach dem Grade der Richtigkeit ihres Organismus mit mehr oder minder sorgfltiger Schonung behandelt w i r d . In denen eines sehr vollkommenen Baues schliessen sich aber an dem Stammlaut, als den den Begriff individualisirenden, Nebenlaute, als allgemeine, modificirende, an. W i e nun in der Aussprache der W r t e r in der Regel jedes nur Einen Hauptaccent hat, und die unbetonten Sylben gegen die betonte sinken, . . . so nehmen auch, in den einfachen, abgeleiteten W r t e r n , die Nebenlaute in richtig organisirten Sprachen einen kleineren, obgleich sehr bedeutsamen Raum ein. Sie sind gleichsam die scharfen und kurzen Merkzeichen fr den Verstand, w o h i n er den Begriff der mehr und deutlicher sinnlich ausgefhrten Stammsylbe zu setzen hat. Dies Gesetz sinnlicher U n t e r o r d n u n g , das auch mit den rhythmischen Baue der W r t e r in Zusammenhange steht, scheint durch sehr rein organisirte Sprachen auch f o r m e l l , ohne dass dazu die Veranlassung von den W r t e r n selbst ausgeht, allgemein zu herrschen ; und das Bestreben der Indischen Grammatiker, alle W r t e r ihrer Sprache danach zu behandeln, zeugt wenigstens von richtiger Einsicht in den Geist ihrer Sprache. Da sich die Undi-Suffixe bei den frheren Grammatikern nicht gefunden haben sollen, so scheint man aber hierauf erst spter gekommen zu sein. In der That zeigt sich in den meisten Sanskritwrtern fr concrete Gegenstnde dieser Bau einer kurz abfallenden Endung neben einer vorherrschenden Stammsylbe, und dies lsst sich sehr fglich mit dem oben ber die Mglichkeit unkenntlich gewordener Zusammensetzung Gesagten vereinen. Der gleiche Trieb hat, wie auf die Ableitung, so auch auf die Zusammensetzung g e w i r k t , und gegen den individueller oder sonst bestimmt bezeichnenden Theil den anderen im Begriff und im Laute nach und nach fallen lassen. Denn wenn w i r in den Sprachen, ganz dicht neben einander, beinahe unglaublich scheinende Verwischungen und Entstellungen der Laute durch die Zeit, und wieder ein, Jahrhunderte hindurch zu verfolgendes, beharrliches Halten an ganz einzelnen und einfachen antreffen, so liegt dies w o h l meistentheils an dem durch irgend einen G r u n d m o t i v i r ten Streben oder Aufgeben des inneren Sprachsinnes. Die Z e i t verlscht nicht an sich, sondern nur in dem Maasse, als er vorher einen Laut absichtlich oder gleichgltig fallen lsst.

The Golden Days

Plate

A page from a manuscript in the India Office, London, of Panini's Astdhyyi, reproduced by courtesy of Her Majesty's Secretary of State for Foreign and Commonwealth Affairs. The manuscript is modern (see No. 2451, Catalogue Eggeling 1887). The stras, of which groups of twenty are numbered, constitute 3.2.88-144. The text is the following:
bahulamchamdasi//sukarmappamarptrapunyesu krah//somesuah// agnau ceh//karmany agnykhyyrn// karmannir vikriyah//drseh kvanip// rjani yudhikrah //she ca//saptamyrn janer dah / / pancamym ajtau upasargecasarnjnym//anau karrnani//5//anyesvapi drsyate//nisth// suyajor nvanip//jryater atrn//charpdasi lit//litah knaj v//kvasus ca// bhsyrp sadavasasruvah//upeyivnansvnancnas ca// luh//anadyatane la//abhijavacane lrt//nayadi //vibhsskrpkse//parokse l i t / / hasasvator//la ca prasne csannakle //lat sme//aparokseca//nanau p.rstaprativacane//6//nanvair vibhsa// puri lu csme//vartamne lat//latah satrsnacvaprathamsamndhikarane//sambodhaneca//laksanahetvoh kriyyh tau sat//pyajoh sanan// tcchlyavayovacanasaktsu cnas / / indhryoh satr ak#rcchrini//dviso mitre//suo yajasamyoge//arhah prasarpsym// kves tacchlataddharmatatsdhukrisu//trn//alarpkrnnirkrnprajanotpacotpatonmadarucyapatrapavrtuvrdhusahacarah isnuc// nes charndasi//bhuvas ca//gljisths ca ksnuh//trasigrdhidhrsiksipeh knuh //7//samityastbhyo ghinun//sarnprcnurudhnyamnyasaparisrsamsrjaparidevisarpjvarapariksipaparirataparivadaparidahaparimuhadusad visad ru had u hay ujkrdavi vicatyajarajabhajticarpacarmusbhyhanas ca//vau kasalasakatthasrambhah// ape ca lasah//pre . . .

The following is a translation of these rules. Most references through onuvrtti to preceding stras have been supplied between parentheses. There is anuvrtti of 3.2.84 bhte "when the past is expressed " through 3.2.122, but this has been generally omitted in the following translations. Capital letters denote metalinguistic indicatory elements.

(3.2.88) In the Veda sometimes (the suffix KVIP is attached to han ' kill '). (3.2.89) (The suffix KVIP is attached) to krN ' do ' when compounded with su 'well, ' karman 'work, ' ppa 'evil,1 mantra 'sacred formula,' and punya 1 merit.' (3.2.90) (The suffix KVIP is attached) to su ' press ' when compounded with soma 'the soma (-stalk).' (3.2.91 ) (The suffix KVIP is attached) to ci ' construct ' when compounded with agni 'fire altar.' (3.2.92) (The suffix KVIP is attached to ci) when compounded with a term designating a fire altar when an object is referred to [by the entire expression]. (3.2.93) The suffix inl is attached to vikri 'sell ' when an object is referred to. (3.2.94) The suffix KvanIP is attached to drs 'see' (when an object is referred to). (3.2.95) (The suffix KvanIP is attached) toyudh 'fight' and kr 'do' when combined with rj ' king.' (3.2.96) (The suffix KvanIP is attached to yudh ' fight ' and kr ' do ') also with saha ' with.' (3.2.97) The suffix Da is attached to jan ' be born ' when a locative is expressed. (3.2.98) (The suffix Da is attached to jan) when an ablative not denoting a class is expressed. (3.2.99) (The suffix Da is attached to jan) also with a preposition when a name is expressed. (3.2.100) (The suffix Da is attached to jan) with anu when combined with an object. //5// (3.2.101 ) (The suffix Da is attached to jan) also in other cases. (3.2.102) NISJH [that is, the suffixes ta and tavat] is used (when the past is expressed). (3.2.103) The suffix vanlP is used with su ' press ' and yaj ' sacrifice.' (3.2.104) The suffix atRN is used with jr 'grow old.' (3.2.105) The perfect is used in the Veda. (3.2.106) [However,] the suffix KnaC is used optionally in the place of the perfect endings (in the Veda). (3.2.107) The suffix Kvasil is also used (optionally in the place of the perfect endings in the Veda). (3.2.108) In the vernacular (the suffix Kvasil is also used optionally in the place of the perfect) after sad 'sit,' vas ' live,' and sru ' hear.' (3.2.109) upeyivn ' who has come near,' ansvan ' who has not eaten ' and ancnah ' who has studied ' are also used (optionally in the place of the perfect). (3.2.110) The aorist is used (when the

Plate

past is expressed). (3.2.111) The imperfect is used (when the past is expressed) excluding the present day. (3.2.112) The future is used (when the past is expressed, excluding the present day) when a memory is voiced. (3.2.113) But not in combination with yod'that.' (3.2.114) In construction with another sentence there is an option. (3.2.115) The perfect is used when the speaker was not present. (3.2.116) It is also used with the particles ha ' indeed ' and sasvat 'frequently.' (3.2.117) The imperfect is also used in a question about the recent past [the last two rules are divided differently in the printed editions]. (3.2.118) The present is used (when the past is expressed and the speaker was not present) with the particle sma. (3.2.119) Also when the speaker was present. (3.2.120) With the particle nonu in answer to a question. //6// (3.2.121) There is option with the particles na and nu. (3.2.122) The aorist may also be used with the particle pur when the particle sma is not present. (3.2.123) The present tense is used when what is presently happening is expressed. (3.2.124) The suffixes SatR and SnaC are used in the place of the present when coreferential withanoun not in the Nominative. (3.2.125) But (the suffixes SatR and SnaC are used in the place of the present when coreferential with a noun) [even when] (in the Nominative) in modes of address. (3.2.126) (The suffixes SatR and SnaC are used in the place of the present) when they characterize an action or express its cause. (3.2.127) These (suffixes are called) SAT. (3.2.128) The suffix SnaN is attached to p ' purify' and yaj ' sacrifice.' (3.2.129) The suffix CnaS is attached to express a habit, one's age, or a capacity. (3.2.130) The suffix SatR is attached to ' go ' and the causative of dhr ' hold ' when no difficulty is involved. (3.2.131) (The suffix SatR is attached) to dvis ' hate ' in the meaning of'enemy.' (3.2.132) (The suffix SatR is attached) to su ' press ' when there is a connection with a sacrifice. (3.2.133) (The suffix

SatR is attached) to arh ' deserve ' when praise is expressed. (3.2.134) Up to 3.2.177, suffixes are attached to express habit, obligation, and ability. (3.2.135) The suffix trN is attached (to express these meanings). (3.2.136) The suffix isnuC is attached to kr ' do,' preceded by alam and nir, to janA ' be born,' by pra, to pacA ' ripen,' by ud, topatA 'fly,' by ud, to madA 'be mad,' by ud, to rucl 'shine', trap 'be ashamed, ' by apa, to vrtU 'turn,' vrdhU 'grow,' sahA ' prevail,' cor ' move' (3.2.137) In the Veda (the suffix isnuC is attached) to causatives. (3.2.138) (The suffix isnuC is attached) also to bh ' be. ' (3.2.139) The suffix Ksnu is attached to gl ' be exhausted,' i ' conquer,' sth 'stand,' as well (as to bh). (3.2.140) The suffix Knu is attached to trasl 'tremble,' grdhl ' be greedy,' dhrsl ' dare,' ksipl 'throw.' 11711 (3.2.141) The suffix GHinUN is attached to the eight roots beginning with sam ' becalm.' (3.2.142) (The suffix GHinUN is attached) also to prc ' mix ' preceded by sam, to rudh ' obstruct,' by anu, to yam ' hold,' by ON, to yas ' exert,' by ON, to sr 'run,' by pari, to srjA ' emit,' by sam, to devl ' lament,' by pari, to jvarA ' burn,' by sam, to ksipA 'throw, 1 by pari, to ratA' howl,' by part, to vadA 'talk,' by pari, to dahA ' burn,' by pari, to muhA ' err ' by pari, to dusA 'make a mistake,' dvisA ' hate,' druhA ' hurt,' duhA 'milk,' yuj 'join,' kridA 'play,' preceded by ", to vicA 'separate,' by w, to tyajA 'abandon,' raj A ' desire,' bhajA 'dispense,' carA 'move,' by ati and apa, to musA ' rob,' by a, to han ' kill,' by abhy. (3.2.143) (The suffix GHinUN is attached) to kas ' hurt,' lasA ' sport,' katthA ' praise,' srambh ' believe,' preceded by vi. (3.2.144) (The suffix GHinUN is attached) to las ' desire,' preceded by apa as well (as by vi). Stra 3.2.123 has been made use of for dating Patajali (cf. pages 79-81). Many stras in this fragment are discussed by Louis Renou in his article " La thorie des temps du verbe d'aprs les grammairiens sanskrits" (in the last section of this volume, pp. 500-525).

10
Ramkrishna Gopal Bhandarkar (1837-1925)

During the first half of the nineteenth century the foundations for the serious study of the Indian grammarians had been laid primarily by British Sanskritists, and some of the real importance of the subject had been recognized, mainly by German linguists. In the second half of the century this study began to flourish and yield results of lasting value. The most important contributions were due to Franz Kielhorn, a German scholar who spent part of his life in India. His contributions include the edition of Patajali's Mahbhsya, the translation of Ngojibhatta's Paribhsendusekhara, a work of the eighteenth century, and many smaller studies. But there were also many other contributors, among whom O. Boehtlingk, T. Goldstcker, A. Weber, and, a little later, B. Liebich may be mentioned. A special place was occupied by Sir Ramkrishna Gopal Bhandarkar (1837-1925) who without being a pandit himself combined traditional scholarship and Western training (he regarded himself as an Indian scholar " o f the new stamp"). His grammatical training was his main characteristic. 44 Bhandarkar's Physiologie ist die der indischen Grammatik, " wrote Windisch (1921,13). Bhandarkar became the best-known Indian Sanskritist of the nineteenth century. Kielhorn and Bhandarkar were colleagues at the Deccan College at Poona before Kielhorn was appointed at Gttingen. Many important papers by these two scholars, dealing with the Sanskrit grammarians, appeared in the Indian Antiquary, a periodical founded in 1872, published from Bombay, and first edited by James Burgess. Before Bhandarkar and Kielhorn began to write, several important publications appeared, some of which should be mentioned. Boehtlingk published an edition of Pnini's sutras in 1839, which depended largely on the editio princeps of Calcutta of 1809. It was not until 1887 that he brought out a much better edition, this time accompanied by a translation and extensive indexes (reprinted 1964). Boehtlingk also published the text of Vopadeva's Mugdhabodha (1847). In the meantime the Prtiskhya literature, which deals mainly with Vedic phonology and phonetics (to be more exact, which provides for each Vedic text the rules to derive the samhitd from the padaptha), was beginning to become known. The first editor of the Rkprtiskhya was the French Vedic philologist Adolphe Rgnier; it appeared between 1856 and 1858 in the Journal Asiatique. Max Mller had begun an edition and translation of this text in 1856, but it was not published until 1869. Other Prtiskhya texts were edited and translated by A. Weber (1858) and W. D. Whitney (1862, 1871). Another important early work, the Nirukta, dealing with Vedic etymology, was also beginning to be studied ; the text was published in 1852 by Rudolph Roth. In 1861 Theodor Goldstcker (1821-1872) published Pnini: His Place in Sanskrit Literature, a very remarkable monograph which evinces his quite extensive familiarity with the AstdhyyJ (recently reprinted twice: Varanasi 1965, Osnabrck 1966). Goldstcker had studied with A. W . von Schlegel and Lassen in Bonn, and with Eugne Burnouf (the famous son of Jean-Louis Burnouf, mentioned previously) in Paris. Wilson invited him to come to England to prepare a new edition of the Sanskrit dictionary. He was subsequently appointed as an honorary professor at the University College, London, in 1852. Goldstcker planned to bring out an edition of the hAahbhsya, and under his supervision two photolithographic reproductions of manuscripts of that text, one accompanied by Kaiyata's commentary, were pub-

71 Ramkrishna Gopal Bhandarkar

lished (1874a and 1874b; see Plate IV; and see this volume, page 151, note 7). Goldstcker's Pnini, an extraordinarily aggressive book, consists very largely of long refutations of the German Sanskritists, Roth, Boehtlingk (whom he always called the " e d i t o r " of Pnini's grammar), Weber, Max Mller, and others. This work did not in fact establish much. But it raised important problems and devoted much attention to the historical relationships among Pan i ni, Ktyyana, Patajali, the Pratiskhyas, and the Nirukta. Goldstcker defended (as did Thieme more than half a century later) the priority of Pnini with regard to the Pratiskhyas against the prevalent view of Roth, Mller, and Weber. He also attempted to settle Panini's date and tried, with the help of an ingenious argument, to determine that Patajali had written a certain portion of the Mahbhsya between 140 and 120 B.C. Goldstcker's work attracted much criticism, and much of it was well deserved. Albrecht Weber devoted 176 pages of his periodical Indische Studien (5,1862) toa critical review. He tried to show that almost all of Goldstcker's conclusions were untenable, or at least open to doubt. But Weber's own work was not free from defects, and the real shortcomings of Goldstcker became apparent only later. Despite these activities, it was only with Kielhorn that a really new era of scholarship began. This was largely due to the fact that continental scholars such as Bhler and Kielhorn had now begun to deepen their knowledge of Sanskrit in India. Earlier, says Wackernagel, " eine Fahrt nach Indien oder gar lngerer Aufenthalt daselbst kam gar nicht in Frage. Nun aber begann eine ganz neue Periode. Nun wurden europisches und indisches Knnen in Verbindung gebracht, nun der Forschung aller aus dem Boden Indiens eine Flle neuer Schtze zugefhrt" (Wackernagel 1908, 5). Bhandarkar had written a review of Goldstcker's Pnini, which first appeared in a local newspaper, Native Opinion, on August 21 and 28,1864. At Kielhorn's suggestion, it was reprinted in the Indian Antiquary (6,1877,108-113). It is a good specimen of Bhandarkar's style and scholarly qualities and deserves to be republished if only because of the recent reprints of Goldstcker's work. In addition, it is of particular interest because it deals with Pnini's treatment of names and his use of technical terms, both called samjnor sanjn in Sanskrit.

A. Review of Goldstiicker's Panini (1864, 1877) Ramkrishna Gopal Bhandarkar

Dr. Goldstcker is undoubtedly one of the most learned, laborious, and accurate European Sanskrit scholars we have known, and the wide and, in many cases, precise knowledge he has shown of Indian grammatical literature is particularly striking to a Hindu, especially when we call to mind that he has not had the advantage of oral instruction, which is available only in India. Of course a minute knowledge of the complicated and subtle speculations of Indian grammarians can only be acquired after a hard study of at least five years, and from a Pandit-teacher. But much of what they have written is barren and useless, and no European Sanskritist, or Indian scholar of the new stamp, would consider it worth his while to study it. The doctor's critical acumen, the skill with which he has brought together stray facts to illustrate and prove his points, and the success with which he has combated the opinions of several scholars, command our admiration, though we are rather inclined to think he has handled some of his German friends somewhat roughly. His book is, however, not without its weak points, and there are three or four places where it appears to us to be particularly so. It is not our intention at present to write an elaborate review of it, but we will notice one point which bears materially on his theory about Pnini, the Indian grammatical legislator. At page 166, Dr. Goldstcker lays down the four following propositions: 1. That his (Pnini's) Grammar does not treat of those sanjns or conventional names which are known and settled otherwise. 2. That this term sanjn must be understood in our rule to concern only such conventional names as have an etymology. 3. That it applies also to grammatical terms which admit of an etymology, but not to those which are merely grammatical symbols. 4. That such terms as ti, ghu, and bha were known and settled before Pnini's Grammar, but that nevertheless they are defined by Pnini, because they are not etymological terms. These four statements contain, according to Dr. Goldstcker, the principles which guided Pnini in the composition of his work, and are deduced as conclusions from one of his stras, Patajali's hsya on it, and Kaiyata's gloss on the latter. Leaving these points for fuller examination at the end, let us in the first place consider if these principles are worthy of being made the basis of a stupendous grammatical superstructure, and bear an air of truthfulness about them, or if there is any external evidence to support them. According to the first two statements, Pnini does not propose to teach sanjns, and such sanjns only as have an etymology. Does he, then, propose to teach sanjns which are without etymology? The " o n l y " would show that he does propose this. What, then, is meant by sanjns without etymology? Are such sanjns as Panclh, Varanh, Ahgh, which are given by the commentators as instances of this stra and the previous one to which it refers, and which, therefore are the sanjns Pnini, according to them, does not propose to teach,are these sanjns, we ask, without etymology? If they are, according to Dr. Goldstcker, Pnini should teach them. If they are not, no instance can be given of a word existing in the language which is a sanjn without etymology. If we bear in mind that two schools of etymology existed in India, viz. vyutpatti paksa, according to which all words have an etymology, and avyutpatti paksa, according to which some have, and some have not, and that Pnini belonged to the latter, as is asserted and believed by all sstris, such words as panclh and angh are sanjns

73 Goldstcker's Pnini

without etymology. And if this be joined with Dr. Goldstcker's statement it will follow that Pnini should teach them. But as a fact he does not, if we believe the commentators. Now with regard to the vyutpatti paksa, we see that the rule in question contradicts its doctrine, for according to that paksa all words, sanjas included, have etymology, while the rule makes a distinction between words with and words without etymology. If we suppose, then, that Pnini belonged to this paksa, and at the same time that he observed the rule given by Dr. Goldstcker, we must either suppose him to have possessed an extremely illogical mind, or not to have proposed such a rule for his guidance. Upon either view of etymology, therefore, we maintain that the rule laid down in statements Nos. 1 and 2 could not have been followed by Pnini. We perfectly agree with statement No. 1 if it be separated from No. 2, and not interpreted according to the sense of the word sanja given in the latter. In the next two statements, this rule is applied to grammatical sanjas. Such as are settled are not to be defined, but an exception is to be made in favour of such as have no etymology, e.g. ti, ghu, bha, &c. We see no reason why Pnini should select for definition, out of settled sanjas, such as have no etymology. For, both those with and those without etymology are settled, i.e. have a fixed meaning. The mere circumstance of some sanjas having etymology, which may be considered as the reason why they are not to be defined, is immaterial, as the presence of etymology in the one case is nearly the same thing as its absence in the other. The etymology of a technical term is not sufficient to explain its sense, and in some cases it affords no clue to it whatever. How can the etymology of the terms bahuvrlhi, pratyaya, &c. enable one to understand their grammatical signification ? In so far, then, as words with etymology are used in philosophical treatises in a sense different from the etymological, or from that they have in common language, they are in the same predicament as unmeaning words, such as ti, ghu, &c. We see, therefore, no reason why Pnini should have selected the latter for definition, and not the former. Having laid down this theory about Pnini's technical terms, Dr. Goldstcker proceeds to test its accuracy with reference to several sanjas which he knows were settled before Pnini's time, such as pratyaya, pratham, dvltly, tatpurusa, &c, and finds that he has not defined them, as they have an etymology. He then mentions other sanjas, such as karmadhraya, sanyoga, anunsika, hrasva, dJrgha, udtta, anudtta, &c, and since they are defined and possess etymology, he concludes that they must have been first used by Pnini himself. W e cannot help thinking that there is here an instance of the fallacy of reasoning in a circle, or of the anyonysraya of Hindu logicians. In order that Dr. Goldstcker's theory may be true, it is necessary that these defined sanjas possessing etymology should be inventions of Pnini, and they are inventions of Pnini because the theory is true. Or, in plainer terms, the theory is true because these defined sanjas are Pnini's inventions, and they are Pnini's inventions because the theory is true. These defined sanjas may have been settled before Pnini's time, in which case the doctor's theory would be false. And in fact we have reason to believe that such sanjas as hrasva, dJrgha, piuta, udtta, anudtta, &c. were invented before Pnini. We are sorry we have not got any treatise on Siks to refer to just now, but considering that the names for accents and for long and short vowels must have been very early invented by grammarians, as

74 RamkrishnaGopal Bhandarkar

they are t h e most elementary distinctions, and likely t o strike a lingual philosopher before many others, and bearing also in mind that if different terms for these had existed before Pnini, they w o u l d not have been altogether lost, and we should have known t h e m , we are inclined t o believe that the names in question w e r e settled before his t i m e . Dr. Goldstcker himself mentions one such w o r d (dvandva), and is not inclined t o disbelieve that t h e r e may be many more. But the supposition he makes, t o save his theory, that Pnini used t h e m in a sense somewhat different f r o m that in which they w e r e before used, has, in our opinion, no basis. W e have all along used the w o r d definition in Dr. Goldstcker's sense. He seems t o understand by the t e r m definition such a definit i o n as is commonly given in European books, viz. that which unfolds the connotation or comprehension of a t e r m . But the p r i n cipal object of a definition is t o point out or distinguish certain things (defmitum) f r o m the rest, and this may be done in o t h e r ways than by unfolding the connotation. Unfolding the extension or denotation is often an easier process, and may in several cases be resorted t o . Even European logicians call this latter a definition, no less than the f o r m e r . Sanskrit w r i t e r s do not confine themselves t o the f o r m e r , but frequently use the latter and several o t h e r kinds. For instance, in Visvantha Pancnana's Muktvali (p. 71 Asiatic Society's edition) the fallacy anaikntika s defined as that

which is any one of Sdhrana, &c, i.e. anaikntika is either Sdhrana, Asdhrana, or Anupasanhrin, The fallacy is thus defined by enumerating its several kinds. We need not stop here to quote other instances. Any one who takes the trouble will find many in any Sanskrit philosophical treatise. What we maintain, then, is that, so far as this view of definition is concerned (and we are convinced that that is the Hindu view), Pnini has defined the terms pratyaya, tatpurusa, bahuvrihi, &c, which Dr. Goldstcker says he has not; but he has defined them by enumerating the several kinds or individuals contained under them. To Hindu writers such a definition is as good as the other, especially when the latter is difficult to give. We think Pnini in defining terms by enumeration was not guided by any such rule as the learned doctor lays down, but he simply consulted his own convenience. When he found it difficult to give a connotative definition, he gave a denotative one. How difficult would it have been to give a connotative definition of bahuvnhi, for instance, containing as it does such compounds as uttaraprv [Northeast], saputra [accompanied by his son], danddandi [stick against stick], so different from such a one as kamalanayana [lotus-eyed]. We now proceed to examine the principal evidence upon which Dr. Goldstcker's theory is based. As we said before, he quotes astra of Pnini, the bhsya on it, and Kaiyata's gloss on the latter, and deduces his theory from these. When we read this portion of the book for the first time, we were surprised to find that the doctor had not understood one of the passages correctly. The stra referred to is tad asisyam samjnpramnatvt. Dr. Goldstcker's translation:"Such matter will not be taught by me, for it falls under the category of conventional terms which are settled (and therefore do not require any rule of mine; " literally, "for it has the authority of a sanjn or conventional term)." This translation is generally correct. We would, however, translate it more closely, thus:"About that no rule ought to be made, or, that should not be taught, for (the knowledge derived from) the meaning of conventional terms in common usage is an

75 Goldstcker's P i i

a u t h o r i t y in itself." T h e w o r d samja is explained by Patajali as samjanam, which again Kaiyata interprets by sampratyayah, avagamah, i.e. knowledge obtained ( f r o m usage). \n a note o n that p o r t i o n of t h e Siddhnta Ka u mu di (Cow e\\'s edition), w h e r e this stra is explained, w e find t h e following: samjanam iokavyavahrnm evtra pramnatvam, "sanjnsthat is, usagesare here an authority or evidence." The bhsya on this stra is as fol lows :kirn y eth krtrims tighubhdisamjas tatprmnyd asisyam \ netyha \ samjanam samjn\ . Dr. Goldstcker's translation :"When Pnini speaks of conventional terms which he will not teach, because they are settled, does he mean, by this expression, such technical terms as ti, ghu, bha, and the like? No; for sanja is here the same as sanjana, understanding (i.e. a name which has a real meaning, that may be traced etymologically)." We do not see whence he gets the first portion up to "settled." If by implication, we do not think it necessary to understand anything. There is nothing even in the stra which has the sense of the words "which he will not teach, because they are settled." For what Pnini says he will not teach is that something which he has alluded to in the last stra but one, and which we shall explain hereafter. We do not deny that this sense may be inferred from what Pnini actually says. We have, however, a particular objection to the expression "are settled" if it is to be made applicable to the terms ti, ghu, bha &c, and understood to mean "settled before Pnini's time." There is nothing in the original corresponding to the words enclosed in brackets in the above extract, nor is the sense deducible from any word occurring in the Sanskrit passage. There is, no doubt, the word sanjnnam, but we do not know upon what authority Dr. Goldstcker renders it by " a name which has a real meaning that may be traced etymologically." Kaiyata explains it by avagama, sampratyaya, as noticed above, which means ' knowing, comprehending/ as is evident from his use of the word avagati (differing from avagama only in the form and not in the sense of the termination) in the sentence which follows. It is this :tatra yathpo drh sikat vars ity ukte iingasamkhyvisesvagatir utpadyamn pramnam evam pancl varanh ity dv api : "As when one pronounces the words pah, drh, sikath, varsh, the avagati (knowledge or comprehension) of a particular number and gender which is produced is authority, so is it in the case of panclh, varanh/' &c. Our translation of the passage in question is as follows:" Is it on account of the authority of (or evidence afforded by) such artificial sanjns as ti, ghu, bha, &c. that that (the thing mentioned in a previous stra alluded to before) should not be taught," " N o , " says he (GonardlyaPatajali). "Sanja is knowing, comprehending/' Upon the whole, Dr. Goldstcker's translation of these two passages is not very objectionable, but they do not afford any basis for his theory, except for that portion of it which is comprehended in the first statement. But the quotation from Kaiyata is altogether misunderstood. It runs thus : kirn y et iti pratysattinyysrayena prasnah nety heti\ pratysatteh smarthyam balavat \ nahi tighubhdisamjnnm pramnatvam yuktavadbhvasstrasysisyatve hetur upapadyate \ sambandhbhvt I samjanam iti \ avagamah sampratyaya ity arthah \ And Dr. Goldstcker's translation of this is as follows : "The question of Patajali is suggested by the rule of analogy. His answer is in the negative because the context itself has greater weight than (mere) analogy. Now, though such terms as ti, ghu, bha,

76 RamkrishnaGopal Bhandarkar

and t h e like, are settled terms, this circumstance w o u l d not have been a sufficient reason in an etymological w o r k (like that of Pan i ni) f o r leaving them untaught, f o r they have no etymology. ' Understanding' (as Patajali paraphrases sanja), means mentally entering into, understanding t h e component parts of a w o r d (or it means the words which admit of this mental process)." In t h e first sentence of this, t h e w o r d analogy is not, w e t h i n k , a correct translation of pratysatti, though it will do. " P r o x i m i t y " is t h e w o r d that is equivalent t o it, and it ought t o have been used here, f o r a reason which w e shall presently explain. But it is t h e t h i r d sentence that is t h e most objectionable of all. W e have no hesitation in saying that t h e translation here is totally w r o n g , and it is upon this misapprehension of t h e sense of t h e original that t h e doctor's peculiar t h e o r y is based. W e hope o u r readers will excuse us f o r t h e assurance w i t h which w e speak; f o r w e feel that no native scholar acquainted w i t h grammatical phraseology w o u l d ever t h i n k of translating o r i n t e r p r e t i n g t h e passage thus. As Dr. G o l d stcker translates it, he appears t o connect t h e nouns pramnatvam and asisyatve w i t h t h e genitive tighubhdisamjnnm and renders the f o r m e r by " being settled." But asisyatve ought really t o be taken w i t h t h e genitive yuktavadbhvasstrasya; and then t h e translation w o u l d be " f o r leaving yuktavadbhvasstra untaught," instead of " f o r leaving t h e m (i.e. ti, ghu, bha &c.) untaught," as t h e Doctor translates it. Yuktavadbhvasstra is rendered " a n etymological w o r k , " which, if one remembers what t h e stra is about, he will at once see is altogether w r o n g . The w o r d can by no stretch of sense mean that. Sastra means here 'a rule,' as it frequently and primarily does, and not 'a work.' Various instances may be quoted in support of this, t h e last pda of t h e verse about Undi, etac chstram undisu, being one. Sambandhbhvt is rendered as " having no etymology," for which, however, there is not the slightest authority. Sambandha never means etymology: it means connection. Besides, from the context it is clear that the sentence cannot have the sense Dr. Goldstcker attaches to it. For, the whole subject here discussed by these several writers is this:The last but one stra of Pnini is lupi yuktavadvyaktivacane, which is thus explained in the Siddhnta KaumudJ: lupi sati prakrtivalligavacane stah \ panclnm nivsojanapadah panclh \ kuravah | angh | &c, meaning that when an elision called lup takes place, the gender and number (of the noun) are like those of the base; panclh &c. are instances. This requires some explanation. In virtue of the stra tasya nivsah 4,2,69, the termination an should be added to the noun panclh for instance, when we have to form a derivative signifying ' the place of residence or the country of the Pancls,' a race of Kshatriyas (hence the above example from the KaumudJ is worded panclnm nivsojanapadah). Now, this termination is elided in virtue of the stra janapade lup 4, 2, 81. If the termination were not dropped, the word expressing 'the country of the Pancls' would be pnclah. Then the question is, when it is dropped, what should be the gender and number of the noun signifying the countryl Should it be masculine and singular, as the word janapada country is? If so, the derived word would bepanclah. But " No," says Pnini (in the stra lupi yuktavat &c); "the gender and number should be like those of the original base," which is panclh, and, consequently, masculine and plural. Hence the noun signifying the residence or country of the Pancls is panclh. " Now," says Pnini (in the stra tad asisyam samjnpramnatvt), "what is the use of teaching by a rule the number and gender of these?" though

77 Goldstcker's Pnini

he himself, in conformity w i t h t h e practice of f o r m e r w r i t e r s , has done so. " T h e y are t o be learnt f r o m usage, which has Itself an authority, just as t h e gender and number of pah and drh are, and t h e authority of a grammarian is not required. For panclh, angh, &c. in t h e plural are actually t h e names of certain countries, and, as such, ought t o be used in t h e plural, in deference t o t h e existing usage, and there is no necessity of a grammarian's teaching i t . " Upon this Patajali raises t h e question, " Pnini speaks of t h e a u t h o r i t y (of usage in matters) of names. A r e they such names as ti, ghu, bha & c , which have an a u t h o r i t y " (as used by Pnini, not necessarily by any other w r i t e r ) ? " N o , " says he. Kaiyata explains w h y Patajali put t o himself such a question. " He was l e d , " he says, " t o it by t h e p r o x i m i t y of these artificial grammatical sanjas, or that he wanted t o determine which were t h e sanjas meant by Pnini ; because if he did not do so, a reader might, on reading t h e stra in question, be led t o t h i n k first of them (the grammatical names) rather than of any other, on account of t h e i r p r o x i m i t y t o or connection w i t h t h e science he is studying. \n order, therefore, t o avoid all such confusion, he proposes t h e question, and answers it by saying ' N o . ' " W h y not? "(hi) Because," says Kaiyata, " ( t / ghubhdisamjnnm pramnatvam) t h e authority of the grammatical sanjas ti, ghu, bha &c. (na hetur upapadyate) is no reason (as t h e authority of sanjas in common language such as Panclh, Angh, &c. is) why yuktavadbhvasstram (a stra or rule expounding that when a termination is elided by the use of the term lup, the gender and number are like those of the base) (asisyatve) should not be taught." And why is it no reason? " (sambandhbhvt) Because there is no connection " (i.e. no connection between such sanjas as ti, ghu, &c. and yuktavadbhva). This is the whole sense of the three quotations. Yuktavad, i.e. like the base, is the word used by Pnini in the last but one stra (lupi yuktavat), &c. ; and Kaiyata first adds the word bhva to it, when the whole means "the being like the base," and then the word sstram, a ruie, and thus the expression yuktavadbhvasstram signifies literally "the rule about the being like the base," and not an etymological work, as Dr. Goldstcker understands. It will thus be apparent that Dr. Goldstcker's theory is based upon a misapprehension of a passage in Kaiyata; and, now that we have explained its true sense, and have also shown that the theory is not supported by any external evidence, it must, we think, be given up. The first of the doctor's four propositions if separated from the second we agree with, as we have already intimated. Dr. Goldstcker's opinion, that the sanjas ti, ghu, and bha were known before Pnini's time, may be true, for aught we know, but it does not at all follow from anything in the passages commented on. He was, no doubt, led to it by the expression tighubhdisamjnnmpramnatvam, which he renders by "such terms as ti, ghu, bha, are settled terms." We would translate it as the authority of
such sanjas or terms as ti, ghu, bha, & c , and this a u t h o r i t y they

derive from their having been used and defined by Pnini. The whole grammatical literature based on his work does not admit the authority of any other person except him, his continuator and critic Ktyyana, and his bhsyakra. And even if we take Dr. Goldstcker's translation, the expression "are settled terms" does not necessarily mean "settled " before Pnini's time, or by any other person than Pnini himself. Dr. Goldstcker has also misunderstood the sense of the stra pradhnapratyayrthavacanam arthasynyapramnatvt,

78 Ramkrishna Gopal Bhandarkar

which is thus explained in t h e KaumudJ: pratyayarthah pradhanam ity evam rpam vacanom apy asisyam | kutah \ arthosya lokata eva siddheh | i.e. "the saying that the sense of a termination is the principal sense of a word (and that that of the base is attributively joined to it) should not be taught. Why? Because the sense (of a word) is to be gathered from, or is established by, usage.'1 We do not know whence Dr. Goldstcker brings in the idea of a compound and its "principal part" in his translation. We do not think it necessary to enter at greater length into the explanation of the stra in this place. We must here close our remarks; our space does not admit of a more lengthened notice, at least for the present. W e hope our observations will be calmly and patiently attended to by European Sanskritists In several cases, though not in all, native students of Sanskrit have a greater right to be listened to than Europeans. We are also desirous that these few remarks should not give pain to Dr. Goldstcker, who, especially by his articles on our religious difficulties published in the Westminster Review, has shown himself to be our decided friend, who sympathizes with ourfallen condition, and is ready to help us by his friendly advice in our race towards a brighter future. In " O n the date of Patajali and the King in Whose Reign He Lived'* (1872), Bhandarkar turned to Patajali's Mahbhsya and found further evidence to corroborate the date Goldstcker had proposed. The argument is of special interest because it illustrates how the Sanskrit grammarians' carefully recorded observations on the use of tenses could be used for the determination of historical data. This line of research was later developed by others (for example, Liebich 1930, Einleitung zu Anhang III, 265-289). The paper also refers to Pnini's stra 1.1.68, which introduces the distinction between " u s e " and " m e n t i o n " and which was for that reason studied by Brough three quarters of a century afterward (see pages 402-414 of this volume). Bhandarkar's article was published in the Indian Antiquary (1,1872, 299-302). Patajali's date has been dealt with again in a more recent study (Frauwallner 1960), in which a later date than the traditional second century B.C. is suggested ; see also Staal (1967, 48-49). The pivotal stra 3.2.123 occurs on the manuscript page illustrated in Plate III. In Patajali's Mahabhasya or great commentary on Panini, a rule (vrtika) laid down by Ktyyana, is given, teaching that the Imperfect should be used to signify an action not witnessed by the speaker but capable of being witnessed by him and known to people in general. Of this rule Patajali gives two instances: "The Yavana besieged (arunat) Sketa" and "The Yavana besieged (arunat) the Mdhyamikas." The siege of Sketa, therefore, must be considered to have been an event capable of being witnessed by the speaker, i.e. by Patajali himself, in other words, some Yavana king must have besieged Sketa in Patajali's time. Sketa is the usual name for Ayodhy. Reasoning in this way, the late Prof. Goldstcker arrived at the conclusion that the Yavana here spoken of must have been Menandros, King of Baktria, who is said to have pushed his conquests in India to the river Yamun Menandros, according to Prof. Lassen, became king about 144.

B. On the Date of Patajali and the King in Whose Reign He Lived (1872) Ramkrishna Gopal Bhandarkar

79 The Date of Patajal

B.C. Patajali t h e r e f o r e must have lived about that t i m e . But t h e r e is another passage in Patajali n o t noticed by Prof. Goldstcker, in which the name of the king of Ptaliputra, during whose reign he flourished, is given, and which enables us to arrive at the date of the author of the Mahbhsya in another way and from other data. In his remarks on Pan. III. 2-123, Patajali quotes a vartika of Ktyyana, the meaning of which is " A rule should be made teaching the use of the present tense (I at) to denote an action or undertaking which has been begun but not finished." The examples given by Patajali are:" Here we study;" " Here we dwell ;" " Here we perform (as priests) the sacrifices (instituted) by Puspamitra." Then Patajali asks " How is it that Pnini's rule III. 2-123, (vartamne I at), which teaches that the present tense should be used to denote present time, does not extend to these cases?" The answer is, " t h e time here involved is not present time." How not? This question is answered by Kaiyata, whose gloss upon this runs as follows :"The phrase ' here we study' means that study has begun but not ended. When the students being engaged in dining and doing such other things do not study they cannot then properly say 'we study' (according to Pan. III. 2-123, i.e., they cannot use the present tense, for it is not study that is then going on, and consequently the time is not present;) hence the rule by Ktyyana."1 The sense of the whole is, that when an action, such as that of studying or performing the great sacrifices, spreads over many days, the present tense should be used to denote it, if the action has begun but not ended, even though at the time of speaking the speaker may not be actually performing the action. " Here we sacrifice for Puspamitra," is Patajali's example. Now this cannot be an imaginary instance, for such a one would not bring out the distinctive sense that Patajali wishes to convey, namely, that the action has begun but not ended. This example then expresses a fact; i.e., that at the time Patajali wrote, there lived a person named Puspamitra and a great sacrifice was being performed for him and under his orders. If he employed priests to perform the great sacrifices for him he must have been a king; for in the olden days it was Indian kings that propitiated the gods and patronized the Brahmans in this way. The sacrifices were always expensive, and were treated rather as extraordinary festivals than ordinary religious performances. But in another part of the Mahbhsya we are actually told who this Puspamitra was. Pnini (in I. 1, 68) tells us that any grammatical change or operation that he may have in his work prescribed in the case of a certain word ought to be made applicable to that word alone and not to what it signifies, or to its synonyms. This, however, does not hold in the case of his own technical terms. Thus, for instance, to form derivatives in a certain sense from the word agni (fire) the termination eya should, he says, be applied to agni. The meaning of this rule should not be stretched
Pnini vartamne /at III. 2,123, Ktyyana pravrttasyvirme sisy bhavanty avartamnatvt | Patajal i pravrttasyvirme ssitavy bhavanti | hdhimahe \ ha vasmah | ha puspamitram yjaymah | kirn punahkranam na sidhyati | avartamnatvt | Kaiyata pravrttasyet'i \ hddhimaha ity adhyayanam pravrttam prrabdham na ca tad
1

viratam \ y add ca bhojandikm kr yarn kurvanto ndhiyate tad 'dhlmaha iti prayogo na prpnotiti vacanam | Patajal i then proceeds to say that the sense is conveyed by Pan. IlI-2-123 and no new rule is required, but this has no bearing on the present question.

80 Ramkrishna Gopal Bhandarkar

so as t o make it applicable not only t o agn'i, but t o other words also, having the sense ofogni. Vahni for instance also means fire but does not take that t e r m i n a t i o n . But in the case of the technical terms of grammar, the change or operation should be effected in the case of the things (which of course are words) signified by that t e r m . Thus, for instance, when he tells us t o apply a certain t e r m i n a t i o n t o g h u , it is t o be applied, not t o g h u itself, but t o the roots t o which the name gnu is given by him. N o w Patajali, after a long discussion of this rule, in the course of which he shows that it is not wanted, though out of respect for the great crya he does not distinctly say so, tells us that there are some stras in which the rules given are applicable,1, sometimes t o the synonyms of the words,2, sometimes t o the individuals comprised under the species denoted by the words,3, sometimes t o the words alone, and, sometimes t o any t w o of these three. In these cases some indicatory letters ought, he says, t o be attached t o the words t o show t o which, or t o which t w o , of the t h r e e categories the rule is t o be applied. Then in such rules as II. 4, 23, which teaches that a T a t p u r u s a c o m p o u n d ending in the words sabh (court or assembly) preceded by rjan (king) becomes neuter he tells us that j should be attached t o rjan and others, t o show that the rule is applicable only t o the synonyms of rjan and others, and not t o rjan or others themselves, or t o the individuals comprised under the species denoted by rjan and others. A n d the instances he gives t o show that it is not applicable t o individual rjs or kings are Puspamitra-sabh (the assembly o r court of Puspamitra) and Candragupta-sabh2 (the assembly or court of Candragupta) in which we see that the compound is not neuter but feminine. W e thus come t o the conclusion that Puspamitra was the name of a king. N o w we know that the most powerful kingdom during a few centuries before Christ, the sovereigns of which extended t h e i r sway over a large p o r t i o n of India, was that of Magadha, the capital of which was Ptaliputra. A n d Patajali so often speaks of this city in his w o r k 3 that we must infer that he had a great deal t o do w i t h Ptaliputra, and perhaps lived t h e r e f o r some t i m e , and that on that account the city and things concerning it were uppermost in his thoughts. The Puspamitra then that he speaks of in the t w o cases here pointed out, must have been king of Ptaliputra in his t i m e . A n d the fact of his being mentioned along w i t h Candragupta in one of the t w o cases strengthens this inference. For Candragupta the Mauryawas king of Magadha, and t h e r e was no other Candrag u p t a t i l l several centuries afterwards when the Gupta dynasty came into power. N o w looking into the Purnas we find that t h e r e was only one king of Magadha of the name of Puspamitra, the founder of t h e Sunga dynasty, which succeeded the Mauryas. He was the Commander-in-Chief of Brhadratha, the last Maurya king, and usurped the t h r o n e after having killed his master. The ten Mauryas are said t o have ruled the kingdom for 137 years. The
Patajali jitparyyavaconosyoivo rjdyartham\jinnirdesah kartavyah | tato vaktavyam paryyavaconasyaivo grahanam bhavati | kimprayojanam\ rjdyartham sabh rjmanusyaprvm | nasabham isvarasabham | tasyaiva na bhavati | rjasabh tadvise2

snm co no bhavati | puspamitrasabh \ candraguptasabh | 3 See amongst others his comments on -3-2, 11-1-16, II-3-28, 1(1-3-134, and 136 and V-3-57. In the second of these, one of the examples given is anusonam ptaliputram.

81 crya

accession of Candragupta, the first of these ten, has been fixed about 315 B.C. Puspamitra, therefore, must have raised himself to the throne about 178 B.C. The Mtsya Purna assigns him a reign of 36 years, i.e. from 178 B.C. to 142 B.C. It follows then that Patajali wrote his comments on Pan. III. 2,123 some time between these limits. The limits assigned by Dr. Goldstcker, reasoning from the one example he considers, are 140 and 120 B.C. But there is apparently no reason why he should not take into account the earlier years of Menandros's reign. For, according to Prof. Lassen, Menandros must have become king about 144 B.C. The passage in the Mahbhsya, on which I base my conclusion, is not far from the one noticed by Dr. Goldstcker. The latter occurs in the comments on III. 2,111, while the former in those on 111.2,123. We thus see that when this portion of the Bhsya was written, a Yavana king (who must have been Menandros) had laid siege to Sketa or Ayodhy, and Puspamitra was reigning at Ptaliputra; and if we adhere to Lassen's chronology these two things could have happened only between 144 B.C. and 142 B.C.; for there is, J think, no reason to distrust the chronology of the Purnas here, since the date arrived at from the statements contained in them coincide in a remarkable degree with that determined from the evidence of coins. And even supposing that Prof. Lassen's date is not quite accurate, it must be admitted that it cannot be very far wrong. We thus see that Patajali lived in the reign of Puspamitra, and that he probably wrote the third chapter of his Bhsya between 144 B.C. and 142 B.C. And this agrees with the conclusion drawn by Prof. Goldstcker from a statement in another part of the work that the author of the Mahbhsya flourished after the Maurya dynasty was extinct. Since all the passages then, and the different historical events they point to, lead us to about the same period, the date of Patajali so derived must be regarded as trustworthy, and in the History of Sanskrit Literature it is of great importance. The preceding paper of Bhandarkar's, along with another one dealing with the age of the Mahbhrata, led to a long discussion with Weber. These discussions of Bhandarkar with European scholars are interesting in more than one way. Quite apart from their scholarly content, they show quite clearly that Indian scholars were in a better position to contribute to Sanskrit scholarship in such fields as the Indian grammarians (and Indian scholarly (sastra) literature in general) than European scholars had first thought they would be. Discussions of this sort led to interesting reactions on both sides. Bhandarkar expressed his own feelings about European scholarship only much later, when describing to the Bombay branch of the Royal Asiatic Society, his visit to the Vienna Oriental Congress of 1886 (printed in the Bombay Journal (17,1889, 72-95) and reprinted in the Collected Works (1,1933, 332-360)). His is an entertaining observation which throws much light on the general situation of Sanskrit studies in Europe at the time. After describing the main achievements of European scholars in the field of Sanskrit studies, Bhandarkar addressed his Bombay audience as follows (pages 348-351 of the reprinted version): In this work of study and research the Germans, of all the nations of Europe, have been the foremost. Most of the great achievements I have briefly indicated above are due to their patient industry and critical acumen. We have had one great French scholar,

C. Acarya, the Friend of the Student, and the Relations between the Three Acaryas (1876)
Ramkrishna Gopal Bhandarkar

82 RamkrishnaGopal Bhandarkar

and t h e r e are now t w o o r three. Englishmen first of all discovered Sanskrit, as was of course t o be expected f r o m the fact of India's having fallen into t h e i r hands, and we have had first-class English scholars, such as Colebrooke and W i l s o n . But somehow Sanskrit and philological studies have n o t f o u n d acongenial soil in the British isles. W h i l e t h e r e are at present twenty-five German scholars at least w h o have been w o r k i n g in the different branches of Sanskrit literature and have published something, we have not more than five among Englishmen. England employs Germans in connection w i t h her philological w o r k . The best Sanskrit scholar i n t h e country is a German, and the Professor of Sanskrit at Edinburgh and the Librarian of the India Office are Germans. There is a German in charge of Manuscripts in the British Museum and the Assistant Librarian at the Bodleian is a Hungarian. The Germans are the Brahmans of Europe, the French the Ksatriyas, and the English the Vaisyas; t h o u g h , as was the case in India, the Brahmans of Europe have now taken t o a military occupation. The great excellence of German scholarship consists in the spirit of criticism and comparison that is brought t o bear on the facts that come under observation, and in the endeavour made t o trace the gradual development of t h o u g h t and language and t o determine the chronological relations of events. So much for the bright side of the picture. But it has also a dark side, t o shut o u r eyes t o which w i l l do no good t o the cause or t o anybody. The proper and f r u i t f u l exercise of the critical and comparative, or what might be called the historical spirit, depends upon innate ability and a naturally sound judgment. These are not t o be found everywhere, and often we meet w i t h instances in which very comprehensive conclusions are based upon the most slender evidence. Though it is t r u e that a native does not easily look at the language, t h o u g h t , and institutions of his country f r o m the critical standpoint, w h i l e t h e first impulse of an intelligent foreigner is t o do so, still t h e r e are some disadvantages under which the foreigner must labour. He has no full and familiar knowledge of what he subjects t o a critical examination. In the case of European Sanskrit scholars t h e r e is besides always a very strong disinclination t o admit the high antiquity of any book, t h o u g h t , or i n s t i t u t i o n , and atendency t o trace Greek influence everywhere in our l i t e r a t u r e ; while not seldom the major premise in the reasoning s that the Indians cannot have any good in t h e m , since several times in the course of t h e i r history, they allowed t h e m selves t o be conquered by foreigners. Oftentimes the belief that the Brahmans are a crafty race prevents a ful I perception of the t r u t h . O f course, scholars of ability and sound judgment shake off such tendencies and prejudices ; and among these I may mention, since I do not wish t o make invidious comparisons between living scholars, Dr. Muir of Edinburgh and Prof. Goldstcker. But independently of such defects in the exercise of the critical faculty, t h e r e are very i m p o r t a n t branches of Sanskrit literature which are not understood in Germany and Europe. I had a conversation w i t h Dr. Kielhorn on this subject the day after I reached Vienna. I said it appeared t o me that works in the narrative of Purnic style and the dramatic plays were alone properly understood in Europe, while those w r i t t e n in the style of discourse o r w o r k s on philosophy and exegesis w e r e not. He replied that even several of the dramatic plays and w o r k s on Poetics were not

83 crya

understood. Mistakes are constantly made w h e n a scholar end e a v o u r s t o i n t e r p r e t and criticise a w o r k o r passages in a w o r k belonging t o any o f t h e Sastras, as w e call t h e m ; and often t h e sense of passages containing idiomatic expressions in o t h e r w o r k s also is n o t perceived. A scholar reads such a w o r k o r i n t e r p r e t s such expressions and passages w i t h t h e aid of a grammar and a d i c t i o n a r y ; b u t a clear understanding of t h e m requires an amount of previous knowledge which cannot be derived f r o m either. As t o positive command over Sanskrit, I had an illustration in t h e shape of a card which was given t o me by a Professor at t h e C o n gress on which t w o verses in t h e easiest of Sanskrit metres, t h e A n u s t u b h , composed by him, are p r i n t e d . In t h r e e o f t h e f o u r lines t h e m e t r e is violated, and t h e r e is a bad compound in t h e second verse. If t h e study o f G r e e k was n o t successfully carried on in W e s t e r n Europe before t h e fall o f Constantinople d r o v e many learned Greeks into that part o f t h e continent, it is of course n o t reasonable t o expect that Sanskrit l i t e r a t u r e should be p r o p e r l y understood in Europe without instruction from the old Pandits of India. This defect was first of all clearly perceived by those German scholars who spent a good many years in India; and now it has been acknowledged by others also, though there are still some whose reliance on agrammar and a dictionary continues unbounded. And the Germans have already begun to remedy the defect. Dr. Garbe was sent more than a year ago to this country at the expense of the Prussian Government to study Indian philosophy. He lived at Benares for a year and read one or two works with some of the Pandits there, and has recently returned to his country. Dr. Kielhorn has undertaken to publish an edition of the Ksik, an old commentary on Pnini's Stras, containing copious notes and explanations of a nature to enable the European scholar to understand the intricacies of the style of grammatical exegesis.... A little later in his address, after briefly referring to the European wars, Bhandarkar related an incident that led him to another attitude regarding the Germans (page 354 of the reprint): A German Sanskrit Professor once said to me that he liked social equality being given to the natives of India, but not political equality, and that he considered the llbert Bill to be mischievous. I told him that in Ceylon and the presidency towns the native magistrates did actually exercise the power of trying European offenders. He did not know that, he said, but still proceeded to defend his position, and, bringing his oriental learning to his aid, observed, " O h , Buddhism has softened the Ceylonese, so that they might exercise that power; but the case is different in India." I listened quietly, thanking my country's stars that she had not fallen into the hands of Germans. The controversy with Weber about the dates of Patajali and of the Mahbhrata took place about one and a half decades before the Vienna congress which prompted the above excursus. Weber (Indian Antiquary!, 1873, 57-59) complained that Bhandarkar did not know his (Weber's) review of Goldstcker's Pon/n/of 1862 in the Indische Studien, and maintained that the beginning of the first century A.D. was a more likely date for Patajali. Bhandarkar, replying in Indian Antiquary (2,1873, 59-61), maintained that his own conclusions were in no way affected by what Weber had written. It is true that Weber in his review had already referred to the passage containing the expression Pusyamitra-sabh

84 Ramkrishna Gopal Bhandarkar

(Bhandarkar insisted on saying Puspamitra, claiming that the other spelling was due to a scribal error; but here he was wrong). But Bhandarkar shows that his argument depended on the expression iha puspamitram yjaymah ' here we perform the sacrifices (instituted) by Puspamitra1 (Mahbhsya on Panini 3.2.123), which was referred to neither by Goldstcker nor by Weber. Next, Bhandarkar discussed the passage arunad yavano mdhyamikn "the Yavana besieged the Mdhyamika/ from which Goldstcker and Weber had derived different conclusions with regard to Patajali's date. "The truth is," said he in Indian Antiquary (2,1873,60) that the name ' Mdhyamika' has been misunderstood both by Dr. Goldstcker and Professor Weber; and hence, in giving Dr. Goldstucker's argument in my article, I omitted the portion based on that name. The expression arunad Yavano Mdhyamikn makes no sense, if we understand by the last word, the Buddhist school ofthat name. The root rudh means 'to besiege' or ' blockade; ' and the besieging or blockading of a sect is something I cannot understand. Places are besieged or blockaded, but not sects. I am aware that Professor Weber translates this verb by a word which in Engl ish means ' to oppress ; ' but I am not aware that the root is ever used in that sense. By the word "Mdhyamika" is to be understood the people of a certain place, as Dr. Kern has pointed out in his preface to his edition of the Brhat Sanhita, on the authority of the Sanhit itself. We are thus saved the necessity of making astring of very improbable suppositions; and in this way, Professor Weber's argument, based as it is on the hypothesis that the Mdhyamikas alluded to by Patajali were the Buddhist sect ofthat name, falls to the ground. The first of Dr. Goldstucker's passages (the word "Yavana" occurring in both of them), and the passage I have for the first time pointed out, taken together, determine the date of Patajali to be about 144 B.C. And this agrees better with the other passages pointed out by Dr. Goldstcker. For if Patajali lived in the reign of the founder of the Sunga dynasty, one can understand why the Mauryas and their founder should have been uppermost in his thoughts ; but if he lived in 25 A.D., when the Andhra Bhritya dynasty was in power, one may well ask why he should have gone back for illustrating his rules to the Mauryas and Candragupta, and passed over the intermediate dynasties of the Sungas and the Knvas. Weber replied with characteristic promptitude (Indian Antiquary (2,1873, 206-210)) that he was not convinced, since Bhandarkar's arguments (which he apparently did not understand) only showed that Patajali did not live before Puspamitra and that they merely established that, during Patajali's time, "the memory of this king was still cherished by the Brahmans." As to the Mdhyamikas, Bhandarkar's silence in his first article "far from implying that he did not coincide with the interpretation of it given by Goldstcker, would seem to show, on the contrary, that he acquiesced in it, not being yet aware of all the difficulties of the case" (page 207). Though he remained unconvinced, Weber continued, after a few more quibbles, to congratulate Bhandarkar in a friendlier spirit "as a most welcome fellow-labourer in our common studies," and he even condescended to commend his critical spirit and sagacity. In his final reply (Indian Antiquary (2, 1873,238-240)), Bhandarkar complained politely about Weber's lack of good faith, and reiterated his position. If we were, in these

85 crya

days, t o give an instance of a rule such as t h e present tense denotes actions that have begun but not ended, he asked, " s h o u l d w e give such a one as * Johnson edits t h e Rambler', o r * Gibbon is w r i t i n g the History of the Decline and Fall'? Would not, on the contrary, our instances be such as * Drs. Boehtlingk and Roth are compiling a Dictionary of Sanskrit*? I think we should use such as this latter, for in the former the actions of editing and writing have long been over, and consequently they would be of no use to illustrate the rule, which specially requires they they should not be over." At about the same time Weber published a long study about the Mahbhsya in his Indische Studien (13,1873, 293-496), where many of these topics were treated again and equally unsatisfactorily (the discussion about Puspamitra occurs in almost identical terms on pages 309-313). Weber gave translations from the Mahbhsya, in which he distinguished the passages where Patajali quoted Ktyyanafrom passages by Patajali himself. He arrived at the view that the Mahbhsya text in its present form was more the work of his pupils than of Patajali himself. Boehtlingk took this up in two articles in the Zeitschrift der deutschen morgenlndischen Gesellschaft (29,1876,183-190 and 483-490). In the first he gave a full translation of the commentary on Pnini (3.2.123), dealing with the present tense, and of a fragment from the Introduction to the Mahbhsya. He concluded from this that the Mahbhsya must be interpreted as a dialogue between Ktyyana and Patajali, which is quoted and commented on by a third person, who is the author. Therefore, the author must be different from Patajali. "Wollten wir Patajali fr den Berichterstatter halten, dann mssten wir hm eine Selbstverleugnung zumuthen, die uns geradezu in Staunen versetzen wrde" (page 190). In his second article Boehtlingk attempted to corroborate these conclusions further with the help of a translation of the commentary on 6.1.64, a rule which teaches the substitution (for metalinguistic reasons) of retroflex s for the initial dental s of certain monosyllabic verbs. " W i e im ersten Artikel," said Boehtlingk, " erscheinen auch hier Ktyyana und Patajali als einander gegenberstehende Klopffechter. Die Palme der Spitzfindigkeit und des Dnkels knnen wir getrost Patajali zuerkennen " (page 483). Actually it is not impossible that Boehtlingk, in the interpretation of this rule, made use of an excellent and detailed exposition of Pnini's treatment of these verbs, provided by Max Mller in section 103 of his Sanskrit Grammar of 1866. This initial s had puzzled scholars for a long time. It is discussed, for example, in a letter of Schlegel's addressed to Humboldt of 1822 (Leitzmann 1908, 106-107). Western scholars complaining about overingenuity and obscurity in the Indian grammatical texts they have attempted to explain have often been proved wrong. Weber's and Boehtlingk's view regarding both the date and the structure of the Mahbhsya were incorrect. This became increasingly clear from the work of Bhandarkar and especially of Kielhorn. In 1874 Kielhorn published his translation of Ngojibhatta's Paribhsendusekhara (reprinted in 1960), an extremely difficult work and one of the major works of the later Pninian tradition. He was at that time also preparing an edition of the Mahbhsya, which appeared in three volumes in 1880-1885 (second edition 1892-1909; third edition 1:1960, II: 1965). Ballantyne had earlier published a portion of this text in Mirzapore, 1856; and Rjarmasstn and Blasstri had published the text in Baaras, 1872. Goldstcker reproduced two manu-

86 Ramkrishna Gopal Bhandarkar

scripts, as we have seen (1874a and 1874b; cf. Plate IV). Kiel horn (using these manuscripts, among others) not only produced a reliable edition, but specified also where Patajali quoted Ktyyana's vrttiks, thus structuring a text which was unintelligible otherwise (as Weber's and Boehtlingk's hypotheses had amply demonstrated). In addition, Kielhorn wrote aseries of important papers in which he dealt authoritatively with some of the problems brought up by Weber. Many of these papers appeared between 1876 and 1887 in the Indian Antiquary. In the first of these, "On the Mahbhsya" (Indian Antiquary 5, 1876, 241-251), Kielhorn established two important conclusions partly through refutations of the opposite views of Weber: (1) the Mahbhsya was written by Patajali himself, and not by his pupils; (2) the text of the Mahbhsya had not been "several times newly arranged" (as Weber had argued from two passages from the RjatarahginJ and the VkyapadJya), but the text given by the manuscripts was the same as it was about two thousand years earlier. The passages discussed by Weber (and earlier by Goldstcker and A. C. Burnell) established only that the text of the Mahbhsya was recovered, not reestablished or reconstructed, by Candra and others (i.e., about the sixth century A.D.). Weber replied, but his argument remained rather unconvincing (Indian Antiquary 6,1877, 301-307). In 1876 Kielhorn published a booklet that in due course became rather well known, Ktyyana and Patajali (recently reprinted twice: Varanasi 1963, Osnabrck 1965). Here he elucidated the exact relationship (which was later specified with complete precision in his Mahbhsya edition) between the two great Indian grammarians, and their relation to Pnini. In the same year Bhandarkar published an article, "crya, the Friend of the Student, and the Relation between the Three cryas," in the Indian Antiquary (5,1876, 345-350) in which he arrived at a characterization of the three cryas ('teachers') Pnini, Ktyyana, and Patajali, which is in some respects even more accurate than Kielhorn's. Since Kielhorn's booklet is now readily available, Bhandarkar's conclusion at least deserves to be quoted here (page 350).

We thus see (1) that Ktyyana explains and supports the stras, sometimes by raising questions about them and answering them, sometimes without resorting to this procedure; (2) that he amends them, and thus must be understood to criticize them, or find fault with them ; and (3) that he supplements them. Patajali (1) comments on the vrtikos in accordance with his own definition of vykhyna; (2) agrees with Ktyyana; (3) refutes him ; (4) recasts Pnini's stras; (5) affirms that they, or a word or words in them, are not wanted, even in cases when Ktyyana justifies them or defends Pnini ; (6) discusses and explains stras or words in them, notwithstanding that there is no vrtika; and (7) gives supplementary rules called stis, which, however, occur very rarely, very little being left for him to do in this respect, by his predecessors. It will thus appear that in writing the vrtikas, Ktyyana did " mean to justify and to defend the rules of Pnini" also, and that avrtika is often "a commentary which explains; " and that the Mahbhsya contains such varied matter, arguments of such length, so consistent, so well connected, and so subtle, that it by no means deserves the title of " a skilful compilation of the views of Pnini's critics and of their refutation by Patajali," or of a

87 Development of Language and of Sanskrit

" mere refutation o f Katyayana," o r of " a synopsis of arguments f o r and against t h e details of Pnini's system, o r a controversial manual." The only tenable t h e o r y is that Ktyyana's w o r k is an edition o f Pnini w i t h notes, explanatory, critical, and supplem e n t a r y ; and that Patajali's is a commentary on this edition, explaining in detail t h e notes o f Katyayana, b u t discussing at length all points connected w i t h t h e system of Pnini and w i t h grammar generally, w h e t h e r Katyayana notices t h e m o r not, in a manner favorable o r otherwise t o his author. The object of both was t h e same, viz. t o teach grammar by following and explaining t h e system o f Pnini, endeavouring t o perfect it, even though this sometimes required a remodelling of his stras o r t h e i r e n t i r e refutation, and t o complete it by supplying t h e omissions and bringing up t h e knowledge of Sanskrit grammar conveyed t h e r e i n

to their own times. D. Development of Language and of Sanskrit (1883)

First Wilson Lecture Ramkrishna Gopal Bhandarkar In an article in the journal of the Bombay Branch of the Royal Asiatic Soc/et/ (16,1883-1885,181 -189), P. Peterson, a professor at the Bombay Elphinstone College for whom Bhandarkar had served as an assistant, had questioned the date of Patajali for which Goldstcker and Bhandarkar had argued. Bhandarkar replied in the same journal (same volume, pages 199-222 and 343-345). Having been involved in various discussions, the time had come for Bhandarkar to give a more systematic exposition of his ideas. This was done in the Wilson Lectures delivered at the University of Bombay and published in the Bombay Journal (volumes 16 and 17). Of these lectures the most relevant in the present context are I, " Development of Language and of Sanskrit " (Journal of the Bombay Branch of the Royal Asiatic Society 16,1883-1885, 245-274) and VII, " Relation between Sanskrit, Pli, the Prakrits and the Modern Vernaculars" (in the same volume, pages 314-342). Bhandarkar began the first lecture with some general facts about Sanskrit, the origins of linguistic studies in India, phonetics, and different language families, and then he proceeded to analyse a specimen of Vedic Sanskrit. He also gave a sketch of a general theory of the development of language, illustrated with Indian examples, reproducing the communis opinio of the historical linguistics of his day. Next he traced the development of Sanskrit from the Brhmana literature onward, in due course arriving at Patajali, whose style he characterized and illustrated with a translation from the Introduction of the Mahbhsya, beginning with the same passage which Boehtlingk had translated in the Zeitschrift der deutschen morgenlndischen Gesellschaft (29,1876,189-190) (see this volume page 85). He ended the lecture by specifying the types of Sanskrit presupposed by the grammatical analysis of Pnini, Katyayana, and Patajali (pages 266-274). Of special interest is Bhandarkar's translation from Patajali, where the latter discusses the need for rules describing forms that are not in use.

It was in the field of philosophy, dialectics, and exegesis, scriptural or grammatical, that this nominal style was greatly cultivated and developed. The earliest work of the kind we know of is Patajali's Mahbhsya on Ktyyana's Vrtikas or notes on Pnini's Stras. Nearly the whole of the philosophical literature of the Sanskrit is written more or less in the style of disputation. An Indian author does not lead his readers into the processes his own mind

88 Ramkrishna Gopal Bhandarkar

has gone t h r o u g h in arriving at the doctrines he lays d o w n ; in other words, he does not tell us how he has come by the opinions he holds, but lays down these doctrines and sets f o r t h those opinions and conceives objections that may be raised and answers t h e m . O r before actually stating the t r u e doctrine or siddhnta, certain others, more or less opposed are stated, and reasons given in support of t h e m (prvapaksa) which are, of course, refuted. The Mahbhsya is w r i t t e n in this manner, but it differs in a good many respects f r o m later w o r k s of the kind. Unlike the latter, it gives the very words that an opponent, speaking in his o w n person, may be expected t o use. It therefore really consists of a series of dialogues, often smart, between one w h o maintains the prvapaksa, and another w h o lays down the siddhnta. Hence, the language is plain and simple, and the sentences are short, and such as a man may naturally use in ordinary conversation or oral disputation. The nominal style, however, as I have ventured t o call it, in contradistinction t o that of the Brhmanas and also of Yska is observable; but it has of course not yet degenerated into the long compounds and algebraic expressions of modern times, and is perfectly natural. In this respect it keeps pace w i t h the language of the Itihsas and Smrtis. Vtsyyana's Bhsyaon Gotama's Nyya Stra, and Sabarasvmin's on that of Jaimini, as well as the Bhsyas on some of the sacrificial Stras, are w r i t t e n in the same sort of simple and lively style, though however they present a f u r t h e r stage in the d o w n w a r d progress. But gradually this manner of w r i t i n g ceased t o be used, and the philosophical style w e n t on progressing until it has come t o be what it is now. Samkarcrya's Bhsya presents it in a middle stage. The sentences are much longer than those of the earlier w r i t e r s , the construction is more involved, t h e r e is a f r e e r use of a t t r i b u t i v e adjuncts, and the f o r m is that of an essay or a ecture, instead of an oral disputation. But the great Acrya's style is perspicuous t h r o u g h o u t , fluent and charming, and not solidified or petrified, as that of later w r i t e r s is. These latter hardly ever use a verb, and of the cases only a few are t o be met w i t h i n t h e i r w o r k s . The nominative and ablative singular prevail, and long compounds are constantly employed. All our ideas are t h r o w n into the f o r m of nouns, mostly abstract, and even the participles have become rare. This style is the style of formulas rather than of discourse. It has reached its climax in the w o r k s of recent Naiyayikas^ but it has been more or less used by modern w r i t e r s in all the Sastras. The movement which began w i t h a less frequent use of verbal forms and the employment of a t t r i b u t i v e expressions has thus ended in making Sanskrit a language of abstract nouns and compound words. I have used the w o r d style in describing this movement, t o show what i m p o r t a n t changes in the s t r u c t u r e of a language may originate f r o m what is in the beginning but a style or mode of speaking or w r i t i n g . If everybody t h o u g h t and spoke about all matters as the Naiyayika does in his o w n subject our language w o u l d be just like his. It w o u l d almost have no verbs, no participles, and no cases except one or t w o . But it is not so. The movement could not be carried so far in o t h e r subjects. Hence the real classical Sanskrit is the Sanskrit of the Epics, the Purnas, the metrical law-books, the better or earlier specimens of Kvyas or poems and dramatic plays, and of the early philosophical or exegetical w o r k s . A n d if we examine this literature we shall find that the greater use of a t t r i b u t i v e or nominal forms of expression gradually drove out a large p o r t i o n of the Sanskrit verb, and gave

89 Development of Languageandof Sanskrit

a new character t o t h e language, which may be thus described : Very f e w verbal forms are used besides those of such tenses as t h e Present and Future; participles are frequently met w i t h ; t h e verbal forms of some roots, especially of those belonging t o t h e less comprehensive classes, have gone o u t of use, and in t h e i r place w e often have a noun expressive of the special action and a verb expressive of action generally; compound w o r d s are somewhat freely employed ; and a good many of t h e Taddhita forms or nominal derivatives have disappeared, and in t h e i r stead w e have periphrastic expressions. If the treatises of Panini and others had perished, and w e had t o construct a grammar of the Sanskrit f r o m t h e classical literature I have above indicated, o u r V e r b and t h e Taddhita p o r t i o n w o u l d be very meagre. Professor Benfey a t t r i b utes t h e condition which t h e language thus assumed t o t h e i n fluence of t h e Prkrts o r t h e spoken vernaculars. But t h e process appears t o me perfectly natural, and no such influence need be supposed. The change may in some respects be likened t o that which rendered t h e Vedic subjunctive and o t h e r grammatical forms obsolete in t h e later stage of the language. The Prkrts may have given some w o r d s t o t h e Sanskrit, but that they should in this manner have influenced its grammatical s t r u c t u r e is very unlikely. It is more natural t o suppose that it was t h e Sanskrit f r o m which t h e Prkrts evidently sprang which gave t o these latter t h e i r peculiar character. I shall endeavour t o d e t e r m i n e t h e exact relat i o n between t h e Sanskrit and these dialects in t h e concluding lecture. We have thus observed and determined the change that came over Sanskrit after the period that elapsed between the Brhmanas and Yskafrom an examination of the literature itself. But the fact is borne witness to by Ktyyana, who observed it in his time and made it the subject of a few vrtikas. Patajali discusses the points raised in the following manner. An objector or Prvapaksin is introduced, who says: Prv. asty aproyuktah There exist (some) words which are not used ; for instance, usa, tera, cakra, peca. (These are forms of the second person plural of the Perfect.) The Siddhntin, or the principal teacher, who advocates the doctrine that is finally laid down asks: Sid. What if they are not used? Prv. You determine the grammatical correctness of words from their being used. Those then that are not now used are not grammatically correct. Sid. What you say is, in the first place, inconsistent, viz., that words exist which are not used. If they exist they cannot be not used ; if not used, they cannot exist. To say that they exist and are not used is inconsistent. You yourself use them (utter them) and say (in the very breath) there are words which are not used. What other worthy like yourself would you have to use them in order that they might be considered correct? (lit. What other person like yourself is correct or is an authority in the use of words). Prv. This is not inconsistent. I say they exist, since those who know the Sastra teach their formation by (laying down) rules, and I say they are not used, because they are not used by people. Now with regard to [your remark] " W h a t other worthy, &c." [when I say they are not used] I do not mean that they are not used by me. Sid. What then? Prv. Not used by people. Sid. Verily, you also are one amongst the people.

90 Ramkrishna Gopal Bhandarkar

Purv. Yes, I am one, but am not the people. Sid. (Vrt. asty aprayukta iti cen nrthe sabdaprayogt). If you object that the/ are not used, it will not do (the objection is not valid). Prv. Why not? Sid. Because words are used to designate things. The things do exist which these words are used to designate. (Therefore the words must be used by somebody. If the things exist, the words that denote them must exist). Prv. (Vrt. aprayogah prayognyatvt). (It does not follow.) Their non-use is what one can reasonably infer. Sid. Why? Prv. Because they (people) use other words to designate the th i ngs expressed by these words ; for i nstance, kva yyam usith "'where did you live?'] in the sense of *sa; kva yyam tlrnh "'where did you cross?'] in the sense of *tera; kva yyam krtavantah 'where did you act?'] in the sense of *cakra; kva yyam pakvantah "'where did you cook?'] in the sense of peca. (We here see that participles had come to be used for verbs of the Perfect Tense). Sid. (Vrt. aprayukte dlrghasattravat). Even if these words are not used, they should be essentially taught by rules just as long sacrificial sessions are. It is in this way. Long sacrificial sessions are such as last for a hundred years and for a thousand years. In modern times none whatever holds them, but the writers on sacrifices teach them by rules, simply because [to learn] what has been handed down by tradition from the Rsis is religiously meritorious. And moreover (Vrt. sarve desntare) all these words are used in other places. Prv.They are not found used. Sid.An endeavour should be made to find them. Wide indeed is the range over which words are used ; the earth with its seven continents, the three worlds, the four Vedas with their angas or dependent treatises and the mystic portions, in their various recensions, the one hundred branches of the Adhvaryu (Yajur-Veda), the Sma-Veda with its thousand modes, the Bhvrcya with its twenty-one varieties, and the Atharvana Veda with nine, Vkovkya, the Epics, the Purnas, and Medicine. This is the extent over which words are used. Without searching this extent of the use of words, to say that words are not used is simple rashness. In this wide extent of the use of words, certain words appear restricted to certain senses in certain places. Thus, savoti is used in the sense of motion among the Kambojas ; the Aryas use it in the derived form of sava; hammati is used among the Surstras, ramhati among the eastern and central people, but the Aryas use only gam; dti is used in the sense of'cutting' among the easterners, dtra among the northerners. And those words which you think are not used are also seen used. Prv.Where? Sid.In the Veda. Thus, saptsye revatJ revad sa[' on (the god) with seven mouths (i.e., Brhaspati) have you shone riches, o rich ones ? ' : Rgveda 4.51.4] | *yad vo revatJ revatym tarn usa [' you
[* Three sentences quoted from the Mahbhsya have a better reading in Kielhorn's edition (I, 9) : kirn yyam tlrnh ' what did you cross ?,' kirn yyam krtavantah 'what did you do?,1 kirn yyam pakvavantah ' what did you cook?' The second (unidentified) sentence attributed to the Veda also has a reading in Kielhorn's edition (1,10) which is probably more correct:
yad vo revatl revatyam tad usa.]

91 Development of Language and of Sanskrit

have shone on us what is full of riches, o rich ones'] | yan me narah srutyam brahma cakra ['the glorious brahman which you, o heroes, have made me' : Rgveda 1.165.11] | yatr nos cakr ja rasa m tannm [' in which you have determined the age of our bodies' : Rgveda 1.89.9]. We here see that the objector says that certain words or forms are not used by people, and therefore they should not be taught or learnt. The instances that he gives are forms of the perfect of some roots and he observes that the sense of these forms is expressed by using other words which are perfect participles of these roots. These statements are not denied by the Siddhantin, but he does not allow that the forms should not be taught on that account. Though not used, they should be taught and learnt for the sake of the religious merit consequent thereon, just as the ceremonial of long sacrificial sessions, which are never held, is. Then the objector is told that though not used by people, the words may be current in some other country, continent, or world, or they must have been used somewhere in the vast literature of the language. As regards the particular instances, two of them are shown to be used in the Vedas. It thus follows that in the time of Ktyyana and Patajali, such verbal forms had become obsolete, and participles were used in their place. But it must have been far otherwise in the time of Pnini. He gives minute rules for constructing the innumerable forms of the Sanskrit verb. Our grammarians proceeded upon a strictly scientific basis. Nothing is more clear from several observations scattered throughout the work of Patajali, besides those contained in the above passage, than this, that the Indian Grammarians do not give us the inventions of their own brains as they are supposed by some scholars to do. The very perfection of their observation and analysis has rendered them liable to this reproach. But notwithstanding all that, there can be no doubt whatever that they scrupulously adhered to usage. If so, the verbal forms taught by Pnini must have been current in the language at some time. We do meet with them in the Brhmanas, but our grammarian does not include these forms among the peculiarities he has given of the Vedic or Chandas and Brhmana dialect, and thus does not restrict them to those works. They must therefore be understood as having been in use in the Bhs or current language, the grammar of which he teaches in his Stras. And the Bhs that he means must be that which was current in his time. In Pnini's time, therefore, the fluent or verbal style of speech was in use, as I have observed before. But it may be argued that though he refers these forms to the Bhs, the Bhs he means may not be that which prevailed when he lived, but another current before his time and preserved in its literature, on which Pnini based his rules. It matters very little even if we make this supposition. The only effect is that the period when the non-Vedic Sanskrit was rich in verbal forms is placed before Pnini ; but the fact itself that there was such a period is undeniable. The supposition, however, is unreasonable. For it is not at all likely that he should neglect the language prevalent in his time and teach that which was current before him, and speak of it as the Bhs, which word literally signifies the "spoken language" or vernacular. And the occurrence in the Stras of words that became obsolete in later Sanskrit confirms this view. The following are such words:anvavasarga 'allowing one his own way,' niravasita 'excommunicated,' pratyavasna 'eating,' abhividhi ' including,' svakarana 'marrying,' utsajana 'throwing up,' abhresa

92 Ramkrishna Gopal Bhandarkar

'equitableness. 1 You w i l l have seen that the V a r t i k a o f Katyayana which starts the discussion I have translated, and Patajali's comment on it speak of words generally as having ceased t o be used though the examples given consist of certain verbal forms only. It is, therefore t o be understood that the observation is applicable t o other forms and expressions also taught by Pnini which we do not meet w i t h in the later literature. A m o n g such may be noticed upjekr and anvjekr ' t o strengthen,' nivacanekr ' t o be silent,' kanehon, and manohan ' t o fulfil one's longing,'' &c. kanehan occurs in Vskaalso. Most of the verbal derivatives ending in the technical t e r m i n a t i o n namui must also be so considered, such as brhmanavedam bhojayati 'feeds every Brhmana that he finds,' celaknopam vrstah ' rained t i l l the clothes were w e t ' ; svaposam pusnti 'supports by his o w n means,' rdhvasosam susyoti ' w i t h e r s standing/ &c. A good many Taddhita forms taught in Pnini's Stras must also, I t h i n k , be put in the same category. A n d there is another circumstance which shows that Pnini's Sanskrit was more ancient than Ktyyana's. Ktyyana's Vrtikas on Pnini, which I have already spoken t o you of, touch on various points concerning grammar and the system of Pnini. The purpose of a great many of them is the proper interpretation of the Stras, and there are some which supply the links that are wanting in the system, also a good many which teach forms not taught by Pnini, or give the correct forms, when by the strict application of Pnini's rules we arrive at such as are incorrect. N o w , this strict application of Pnini's rules is often in the manner of a quibbling logician, and consequently it was probably never meant by Pnini himself. Again, it may also be allowed that some forms existing in the language may have escaped Pnini's notice. But even after making allowance for al I these considerations, a good many forms taught by Ktyyana are left which it is impossible t o believe Pnini did not observe or know, if they existed in the language in his time. Though not infallible, Pnini was not an indifferent grammarian. He justly deserves the reputation he has all along enjoyed of being the preeminent teacher of grammar. He has noticed even stray facts about the language. If so, the only reasonable supposition is that these forms did not exist in the language at the period when he lived. For instance, according t o Pnini's rules the vocative singular of neuter nouns ending in an such as brahmn and nman is brahman and nman, but Ktyyana in a Vrtika on VIII. 2, 8 tells us it is brahman or brahma and nman or nmo. Pn\n\ teaches that the forms of the dative, ablative, genitive, and locative singular of the feminine of dvitlya and trtlya are optionally like those of the corresponding pronouns, i.e. we have either dvitiyyai or dvitlyasyai, dvitlyyh or dvitJyasyh & c , but Ktyyana in a Vrtika on 1.1, 36 extends this option t o the masculine also, and according t o him we have dviyyah or dvitlyasmai, dviyt or dvitiyasmt & c , while Pnini gives us only the first. Pnini's rule IV. 1, 49 allows of mdtu/n/"only as the feminine of mtula, but Ktyyana gives mtulJ also; updhyynl is not noticed by Pnini, while Ktyyana lays down that as well as updhyyi in the sense of ' wife of the updhyya.' So also ry and ksatriyare according t o Pnini, 'afemale r y a ' and 'afemale Ksatriya,' but Ktyyana gives rynJand ksatriyni as well as ry Professor Goldstcker has used the argument based on the occurrence of obsolete words in the Stras and that set forth in the next paragraph,
1

to prove the archaic character of the language as it existed in Pnini's time, and some of the instances quoted in the text are the same as his.

93 Development of Language and of Sanskrit

and ksatriy. A good many m o r e instances o f a similar nature m a / be given f r o m o t h e r parts o f Sanskrit G r a m m a r . A r e w e t h e n t o suppose t h a t t h e f o r m s brahma, nma, karma &c. o f t h e vocative singular, dvitlyasmai of the dative of dvitlya &c, mtuiJ, updhyym, rynl, ksatriynl and many more such, escaped the observation of such a grammarian as Pnini, or that he did not know them ? Is the supposition that they did not exist in the language in his time not more reasonable? It therefore appears clear to me that the language in Pnini's time was in a different condition from that in which it was in Ktyyana's. The chief differences, to point out which has been the object of this discussion, may be thus stated :In Pnini's time a good many words and expressions were current which afterwards became obsolete; verbal forms were commonly used which ceased to be used in Ktyyana's time, and some grammatical forms were developed in the time of the latter which did not exist in Pnini's. Pnini's Sanskrit must, therefore, be identified with that which preceded the Epics, and he must be referred to the I iterary period between the Brhmanas and Yska. Hence it is that the Brhmanas, as observed before, are the best existing representatives of the language of which Pnini writes the grammar. Ktyyana, on the other hand, wrote when the language arrived at that stage which we have called classical. Thusthen, we have been able to trace three distinct periods in the development of Sanskrit. First, we have the Vedic period, to which the Rgveda Samhit, the Mantra portion of the Yajurveda, and the more antiquated part of the AtharvaSamhit are to be referred. Then commences another period, at the threshold of which we find the Brhmanas, which, so to say, look backwards to the preceding, that is, present the Vedic language in the last stage of its progress towards Pnini's Bhs; and, later on, we have Yska and Pnini. This may be called the period of Middle Sanskrit. And last of all, there is the classical period to which belong the Epics, the earliest specimens of Kvyas and dramatic plays, the metrical Smrtis, and the grammatical work of Ktyyana. Pnini's work contains the grammar of Middle Sanskrit, while Ktyyana's that of classical Sanskrit, though he gives his sanction to the archaic forms of the former on the principle, as he himself has stated, on which the authors of the sacrificial Stras teach the ritual of long sacrificial sessions, though they had ceased to be held in their time. Patajali gives but few forms which differ from Ktyyana's, and in no way do they indicate a different stage in the growth of the language; hence his work is to be referred to the same period. The form which the language assumed at this time became the standard for later writers to fol low, and Ktyyana and Patajali are now the generally acknowledged authorities on vail points concerning the correctness of Sanskrit speech.

E. Relations between Sanskrit, Pali, the Prkrts and the Modern Vernaculars (1883)
Seventh Wilson Lecture Ramkrishna Gopal Bhandarkar

In his seventh and last Wilson Lecture, Bhandarkar addressed himself to the thesis of several European scholars, that the Prkrts were not spoken languages but artificial adaptations. This led in turn to a discussion of the question whether Sanskrit itself was a spoken language. Bhandarkar dealt with this topic lucidly and authoritatively, with the aid of numerous references to the grammarians and quoting passages from the Mahbhsya, including a famous passage concerned with the native speakers of Sanskrit (pages 326-336 of the original). There is positive evidence that Sanskrit was a spoken language. Yska in the Niruktafrequently refers to the Vedic dialect and to another called Bhs, the peculiarities of which mentioned by him are observable in classical Sanskrit. Panini in his Grammar gives a good many rules which are exclusively applicable to the dialect of the Vedas, to which he refers by using the words Chandas, Nigama, Mantra, and Brhmana, and others which are applicable to the Bhs alone, but by far the largest number of his stras have reference to both. Now since Bhs, or the ordinary Sanskrit, is thus distinguished from the dialect of the Vedas, it must be the language in use at the time when these writers lived. " Bhs," as used by them, sa proper name, but in later Sanskrit it acquired ageneric signification and meant language generally. The root from which the word is derived signifies " t o speak," wherefore the original sense of the word as a proper noun must have been the "speech " or "the spoken language." And because this was its signification it afterwards came to denote "explanation." When we ask for an explanation of something that is obscure and unintelligible, what we mean is that the sense should be expressed to us in the ordinary language of men, a language that we can understand. Thus such a sentence as sthitaprajasya k bhs means "what is the Vernacular of sth itapraj a ?" an expression similar to "what is the English of it?" Panini refers certain points expressly to popular usage. He says that the names of countries are conventional, and no grammatical analysis should be given of them, because it is fictitious. These should be used as we find them used. Similarly he says grammarians should not make rules to teach such things as these:That the two words of a compound express the thing denoted by the principal word as qualified by the sense of the subordinate word ; as for instance, rjapurusa, a com pound of rjan "a king" and purusa "a man" does not denote " a king," but " a man," and not " m a n " alone butas connected with a king, i.e. a king's man or officer; and that the base and the termination express the sense of the termination as qualified by that of the base; as aupagava signifies not upagu but a child, and not a child alone but a child as connected with upagu i.e. Upagu's child. For the significations of words are to be learnt from usage. In the introduction to the Mahbhsya Patajali tells us that some persons in his or Ktyyana's time considered the study of grammar to be unnecessary. For said they, "Vedic words or forms we know from the Veda, and those current in popular usage from that usage; grammar is useless." Now the grammar which is thus declared useless is the grammar both of the Vedic and classical Sanskrit ; and the depreciators of the science profess to derive a knowledge of the first dialect from the Vedic books, and of the second not from other books but from popular usage. Hence Sanskrit must have been in the times of those two grammarians a spoken language.

95 Sanskrit, Pali, Prakrts, Modern Vernaculars

Similarly in the passage f r o m the same w o r k which I placed before you in t h e first lecture, you w i l l remember that the objector (or prvapaksin) argues that since usage is the authority upon which the grammarians go, certain verbal forms which are no longer used by people ought not t o be taught by the grammarians, and says that instead of those verbal forms participles are employed. The principal teacher (Siddhntin) does not deny the facts but refers the objector t o the vast literature of the language, where he may find them used, though obsolete at the time. It is evident f r o m t h e whole passage, that Sanskrit was then a spoken language, though some of its verbal forms had fallen into desuetude. I have also shown that the language was considerably changed between the times of Pnini and Ktyyana, and called the Sanskrit that prevailed when Pnini and Yska flourished Middle Sanskrit, and that which was current in the t i m e of Ktyyana classical Sanskrit. N o w these changes from the one f o r m t o the other could not have taken place if the language had been dead or petrified into a merely literary language. I am at a loss t o see why some scholars should find it so difficult t o believe that Sanskrit was a vernacular. If its declensions and conjugations are considered t o o complicated for the language of everyday life, it must not be forgotten that such a fact did not prevent the ancient languages of Europe f r o m becoming spoken languages. And this objection would do equally well against the Vedic dialect, which, or others like which, are regarded as the vernaculars of their times, and which are richer in inflexions than the later Sanskrit. Then it is held that the artificial regularity of Sanskrit makes it improbable that it should have been a vernacular. W h e r e is this artificial regularity? On the contrary, it is the absence of regularity that renders its grammar so difficult and complicated. There is a freedom in the choice of words, expressions, and forms. In every department of its grammar there are innumerable optional f o r m s ; nouns and verbal roots are often declined and conjugated in several ways. One same r o o t in a good many cases forms its special tenses in more ways than one, and in the nominal derivatives, the verbal derivatives, the formation of the feminine, and the uses of cases and tenses there is a freedom which some may consider a license. The only difference is that Sanskrit has had the most perfect grammarians in the w o r l d , w h o observed all the facts of t h e i r language and laid them down as unchangeable facts, and it is this which gives that language a stiffened appearance. Then the Samdhi or euphonic combinations of letters which are necessary in Sanskrit, are regarded as inconsistent w i t h the character of a spoken language. It is however not denied that such combinations are observable in all languages, and particularly so in Latin and Greek; but it is urged that in Sanskrit there is a regularity or universality about them which is not found anywhere else. It should not however be forgotten that Samdhi in the same w o r d and the same compound, and of a preposition w i t h a root is alone necessary. Between different words it is optional, which means that it was on occasions neglected. N o w Samdhi in the same w o r d is necessary by a law of nature. The Sanskrit does not allow a hiatus; and this is a characteristic of most of the modern idioms also. Some languages, such as our Prkrts and the old languages of Europe tolerate it. But the euphonic combination of consonants in the same w o r d is necessary even in Latin; as in r e x = reg-s, scriptus = scrib-tus, cinctum = cing-tum, lectum l e g - t u r n , t r a c t u m f r o m t r a h o , though

96 RamkrshnaGopa Bhandarkar

t h e h like t h e Sanskrit ha stands f o r an original gh\ d and t combine t o f o r m ans, asdefendo, defensum ; sedeo, sessum ; claudo, clausum ; &c. Prepositions are really parts of words, and hence by t h e same law they also must f o r m one harmonious sound w i t h t h e initial letter of t h e w o r d t o which they are attached ; and modern vernaculars have got corruptions of t h e combined words which shows that they must have been used in those forms in t h e colloquial Sanskrit. Sk. paryasta, Pr. pallatta, M. piat: Sk. pratyabhijnn, Pr. paccahina, H. pahicn', Sk. atyartha, Pr. accatta, M. act, &c. N o w as t o compounds, this peculiarity of t h e Sanskrit has been carried t o an extravagant extent by later w r i t e r s , but Pnini allows only certain formations of this nature. These g r e w up as independent words in t h e language, and hence in t h e matter of Samdhi w e r e treated (ike other words. In t h e spoken language t h e euphonic combinations w e have been considering w e r e not consciously made, but t h e words themselves acquired those forms by habitual use in t h e same manner as in o t h e r tongues. The grammarians however discovered and laid d o w n these rules; and t h e practice of using t h e m in books even in combining different words gained g r o u n d , though however many instances in which t h e r e is no such Samdhi are found in t h e Itihsas and t h e Purnas. But if in colloquial speech such a combination was not possible, t h e grammarians do not enjoin that it should be made; and very probably it was not made. A n d traces of many expressions which only a colloquial use of language can generate have been preserved, not so much in t h e literature as by t h e grammarians. Such is one expressive of an intensive or excessive action, composed of t h e Imperative second person f o r m of a r o o t repeated, followed by a verbal f o r m of t h e same in any tense of t h e Indicative and in any person o r number; as khda khdeti khdoti, lit. 'eat, eat, he eats,' i.e. eats much, kuru kurv iti karoti ' d o , do, he does,' i.e. does much. This expression exists in Marthi and is considered so colloquial that no Marthi grammarian has noticed it, as kh kh khto, kor kor karato in which as in Sanskrit kh and kora are t h e forms of t h e Imperative second person singular. A similar expression is used when several actions are a t t r i b u t e d t o t h e same agent ; as odanam bhuksva saktn piba dhnh khdety evyam abhyavaharati 'eat rice, d r i n k barley water, devour fried grain, in this way he fills his stomach' ; which in Marthi is bhtjev pej pi ihy kh asem pot bharato. In this case t h e Indicative should signify a general action of which the Imperatives denote the species, and we may have here karoti ' does ' instead of abhyavaharati. Similarly those innumerable expressions consisting of a form made up by adding am, technically namul, to a root, preceded by a noun, and of a verbal form of the same root must be colloquial; as hastagrham grhnti 'holds by the hand,' jlvansam nasyati 'he perishes,' udarapram bhukte 'eats a stomachful,' yathkram aham bhoksye tathkram bhoksye kirn tavnena ' I, will eat as I eat (as I like), what have you to do with it?' &c. Etymologically hastagrham, jlvansam &c. are accusatives, and they may in these cases be called cognate accusatives, and the expressions somewhat resemble such ones in English, as " run a race," "walk a walk," "die a death," &c. The compounds danddandi, keskesi &c. meaning 'a scuffle in which there is a brandishing of sticks and seizing each other by the hair; ' and a great many others made up of Imperative forms of verbs, or of a verb and its object which are used as nouns, are of a nature suited for the purposes of a light

97 Sanskrit, Pali, Prkrts, Modern Vernaculars

conversation ; as atra khadotamodata varate ' ' ' eat and enjoy ' is t h e rule here," atrsnltapibat varate " 'eat and d r i n k ' is t h e rule here," uddharotsrj tasya dnasaundlr asya grhe " 'take out and give' is what takes place in the house oa bountiful man," jahistamboyam " he is one who constantly says, 'strike the sheafs of corn ' " ; ehisvgat vartate " 'come, welcome to thee' is the practice" &c. Sanskrit was not the only language spoken in the times of Ktyyanaand Patajali. In the Mahbhsya there are several passages which contain allusions to a dialect arising from a corruption of Sanskrit. Thus in the comment on the Vrtika Siddhe sabdrthasambandhe, we are told with reference to the question whether words are created or exist of themselves, that Pnini's rules suppose that they are not created but exist of themselves ; and so is the relation between them and the things they denote, i.e. their power of expression uncreated and self-existent; and according to another interpretation of the Vrtika, the things also which words denote are so. How is it known that all these are self-existent? Because in the affairs of the world or in ordinary life men think of certain things and use words to express them ; they do not set about producing words before doing so. But this is not the case with those things that are created and not self-existent. If a man wishes to do something with an earthen pot, he goes to the house of a potter and says, " Make a pot for me, I have to do something by its means." Now if he has to use words, he does not in the same way go to the house of a grammarian and say, " Make words, I want to use them " ; but immediately that he thinks of this thing and that he uses words (for expressing them). Well then, if words are to be taken from ordinary life (and are not madefy grammarians) what is it that the Sastra (grammar) does: "The Sastra lays down a restriction by observing which a man may attain religious merit. It does what other Sastras in ordinary life do. Thus while it is possible to satisfy hunger by eating anything whatever, it is enjoined that one shall not eat domesticated fowl or pig; and the object is that he may by regulating his conduct thus attain religious merit. In the same way this Sastra (grammar) tells us that while it is possible to express one's meaning by using correct words or incorrect words, correct ones alone which it teaches should be used to secure the religious merit arising therefrom." After this follows the discussion translated in the first lecture; and then we have another of which the following is a portion. Pdrv. Does religious merit arise from a knowledge of correct words or from their use. Sid. What is the difference? Prv. If you say religious merit arises from their knowledge, religious demerit also must arise. For he who knows what words are correct, also knows what words are incorrect. If merit results from the knowledge of those that are correct, demerit must result from the knowledge of those that are incorrect; or greater demerit must arise (from their knowledge), as the number of incorrect words is larger, and that of correct words smaller. For the corruptions of one correct word are manifold ; as, for instance, the corruptions of the correct word gauh : are gvl, gonJ, goto, gopotaiik &c. And the Rsi also indicates (in a passage which is quoted) that the restriction as to correct words concerns their use (and not knowledge). Sid. Well, then, let it be that religious merit arises from their use (and not from knowledge).

98 RamkrishnaGopal Bhandarkar

Purv. If f r o m t h e i r use, the whole w o r l d would obtain heavenly felicity. Sid. And now why should you be jealous if the whole w o r l d obtain heavenly felicity? Prv. N o , certainly, I am not jealous. But what I mean is that it thus becomes useless t o make any effort; such effort only as is f r u i t f u l should be made. If you get the f r u i t w i t h o u t any effort, the effort is useless. (The effort meant is that involved in the study of grammar, i.e. of correct words. People use some correct words at least w i t h o u t studying grammar, and if eternal felicity results f r o m the use of such words they get it w i t h o u t making the effort of studying the subject). Sid. W h y , verily those w h o make the effort w i l l largely use correct words, and w i l l obtain a large amount of heavenly felicity. Prv. That the f r u i t does not f o l l o w the effort is also an observed fact. For t h e r e are persons w h o though they have made the effort are seen t o be incompetent (in the use of correct words), w h i l e others w h o have not, are c o m p e t e n t ; wherefore it is possible the f r u i t , i.e. heavenly felicity, may not follow. Sid. W e l l then, religious merit arises not f r o m knowledge alone, neither f r o m use alone. Prv. From what then ? Sid. Heavenly felicity arises f r o m the use of correct words when it is accompanied by the knowledge that they are correct, derived f r o m a study of the Sastra. A n d thus it goes on. N o w it is clear f r o m all this that correct words, i.e., Sanskrit, was spoken in those days by all, but that incorrect words had got mixed up w i t h it, and the object of grammar was t o teach how t o avoid incorrect words or corruptions, though t h e r e w e r e men w h o could speak correctly w i t h o u t knowing grammar. A n d this is the state in which more or less all languages are at all times; and even at the present day the purpose of grammar is considered t o be t o teach how t o speak correctly. By the way, it w i l l be seen how Sanskrit grammarians distinctly declare that they teach nothing that does not exist, they do not create words, but separate the correct ones f r o m such as are incorrect. But what did Patajali consider t o be the standard of correct Sanskrit, w h o was it that spoke the language correctly, and in whose speech were corruptions observable? This is clear f r o m another passage at the beginning of the t h i r d pda of the first chapter of the Mahbhsya. Patajali interprets the stra bhvdoyo dhtavah: (I, 3,1.) in a manner t o yield such a connotative definition of a dhtu or root as this, that a dhtu is that which denotes action or being. Then a question is raised, if this is the way t o distinguish a r o o t why should a list of roots be given; in o t h e r words, if we have got a connotative definition, a definition by enumeration is not wanted. In this case t h e r e is a difference of opinion between Ktyyana and Patajali. Prv. If you have given a connotative definition now, enumerat i o n should not be made, i.e. a list of roots should not be given. Sid. It should be made. Prv. W h a t for? Sid. The enumeration of (the verbal roots) bh and others should be made for the purpose of excluding nouns (prtipadika) and the verbs that begin w i t h onapoyati. (i.e. If the roots are not actually enumerated, nouns also which express action or being may come under the connotative definition. In the same way it

99 Sanskrit, Pali, Prakrts, Modern Vernaculars

will extend t o roots used in ordinary life, which are not Sanskrit, such as napayati and others.) Prv. W h a t verbs are those which begin w i t h napayati 1 Sid. napayati, vattati and vaddhati. A n d enumeration should also be made in o r d e r that the anubandhas or indicatory letters and accents of roots may be made k n o w n ; i.e., that one may k n o w what t h e accents and indicatory letters of roots are. it is not possible t o k n o w the accents or indicatory letters unless the roots are enumerated. (Anubandhas are certain letters attached t o roots t o denote some conjugational or o t h e r peculiarity belonging t o them.) Prv. N o w those roots whose accents are capable of being inferred, i.e. are udtta, and w h i c h have no anubandhas, but still are inserted in the list, may be o m i t t e d f r o m it. (i.e. W h e n those roots which have the anudtta accent are enumerated, it may be inferred that the rest have the o t h e r o r udtta accent.) Sid. Even these should be enumerated in o r d e r that napayati and others (i.e., c o r r u p t Sanskrit roots) may be excluded. Hereupon Patajali disagrees w i t h Ktyyana and says: Pat. N o . napayati and others will be excluded, because the usage of the educated or Sistas is t o be f o l l o w e d . This usage of the Sistas must be necessarily referred t o in o t h e r cases even ; for instance, in o r d e r t o exclude t h e altered forms of those roots t h a t are enumerated. For in ordinary life, they use kasi for krsi and disi for drsL (The sense is, that since in grammar we f o l l o w the usage of the Sistas or educated, these verbs napayati, vattati and vaddhati, and also kasi for krsi and disi for drsi which are not used by t h e m will necessarily be excluded f r o m o u r connotative definition ; hence for the purpose of excluding t h e m enumeration is not wanted.) N o w napayati is a c o r r u p t i o n of the Sanskrit jnpayati, vattati of vartate, and vaddhati of vardhate, te A t m . being replaced by ti Parasm., and kas ofkrs, and dis of drs. These and such o t h e r c o r r u p t i o n s w e r e in use at t h e t i m e , but Patajali clearly lays d o w n that they w e r e not used by the Sistas or educated people, and t h e r e f o r e they belonged t o the language of t h e vulgar. N o w all these roots are found in the Pali in these very forms, napayati being, however, napayati ; but the reading in the MSS. and the Benares lithographed edition which I have consulted is faulty. So also of the corruptions of the w o r d go given by o u r author, we find gvl in Professor Childers' Pali Dictionary, and gona t h e masculine of go m. Thus our grammarians recognise one language only, the Sanskrit, and these words and forms which are found in the Pali they assign t o t h e speech of t h e vulgar. In another passage still we are t o l d more definitely w h o the Sistas w e r e that spoke the Sanskrit language correctly w i t h o u t studying Pnini's grammar, and whose usage was the standard of correctness. Pnini's Stra V I . 3,109, lays d o w n that such w o r d s as prsodara [ ' w i t h a spotted belly'] should be regarded as correct in t h e f o r m in which they are upadista, i.e., used or uttered ; the changes in t h e m such as the elision, or augmentation of letters or the alteration of t h e i r forms do not obey any general rules laid d o w n by him, but still the words exhibiting those changes should be taken as correct, just as they are used. N o w the Prvapaksin raises the question. Prv. W h a t is meant by upadista 1 Sid. U t t e r e d (used).

100 RamknshnaGopal Bhandarkar

Prv. H o w does it c o m e t o mean that? Sid. The r o o t ' d i s ' signifies ' u t t e r i n g . ' One utters (uses), letters, and says that they are thus upadista. Prv. By w h o m upadista (uttered or used)? Sid. By theSistas. Prv. W h o are the Sistas ? Sid. The grammarians. Prv. H o w so? Sid. The character of a Sista is conferred upon one by the knowledge of the science (Sastra) ; and the grammarians know the science (of grammar). Prv. If a man is made Sista by the science, and the science depends upon a man's being Sista, this is reasoning in a circle. An argument in a circle does not hold. (The circle is, one is Sista, and consequently an a u t h o r i t y in matters of language if he has studied grammar; and grammar itself depends on the usage of the Sistas). Sid. W e l l , then, the character of a Sista is conferred upon one by the country in which he lives and by his con_duct. That sort of conduct must be associated w i t h residence in Aryvarta alone, (lit. that sort of cond_uct must be in Aryvarta alone). Prv. W h i c h is Aryvarta? Sid. The country which is t o the east of the Adarsa, west of Klakavana, south of the Himalaya, and_north of the Priytra. Those. Brahmans in this country of the Aryas w h o do not store up riches (lit. w h o keep only so much grain as is contained in ajar), w h o are not greedy, w h o do good disinterestedly, and w h o w i t h o u t any effort are conversant w i t h a certain branch of knowledge are the worshipful Sistas. Prv. If, then, the Sistas are an a u t h o r i t y as regards language, what function does the AstdhyyT (Pnini's grammar) perform ? Sid. The purpose of the AstdhyyT is t o enable us t o find out theSistas. Prv. H o w is it possible t o find out the Sistas by means of the AstdhyyT? Sid. A student of the Astdhyyl finds a man w h o has not studied the book using words just as they are taught in the Astdhyyl. He then thinks, " V e r i l y , this man possesses some good luck or innate nature by means of w h i c h , though he has not studied the Astdhyyl, he uses words just as they are taught in that book. Verily he knows other words also" (not taught in the Astdhyyl, such as prsodara). Thus, the purpose of the AstdhyyT s t o enable one t o find out w h o is a Sista (in order that he may refer t o him and learn such words as do not obey the rules laid down by Pnini, but still are correct). Here then we have the clearest possible evidence that Sanskrit was the vernacular of holy or respectable Brahmans of Aryvarta or N o r t h e r n India, w h o could speak the language correctly w i t h out the study of grammar. The c o r r u p t language mentioned by Patajali which was composed of correct and incorrect words, that is, a dialect like the Pali must, therefore, have been the vernacular of the other classes. And this is what you may say even w i t h regard t o the modern vernaculars. W h o is it that speaks good or correct MarthT? O f course, Brahmans of c u l t u r e . T h e language of the other classes is not correct MarthT. The w o r d Sista may be translated by " a man of education o r c u l t u r e ; " and this education or culture has, since remote times, been almost confined t o Brahmans. Thus the dialects of the inscriptions of Asoka and the Pali were the vernaculars of the non-Brahmanic classes; but a greater importance must

101 Sanskrit, Pali, Prakrts, Modern Vernaculars

evidently have been attached to them in the times of Asokathan is now assigned to the Marth of the non-Brahmanic classes since they are used by him in the inscriptions. They are however not recognized as independent languages by our grammarians who treated them as we treat the MarthT of the lower classes ; but they were in use and bore the same relation to Sanskrit that low Marth does to high Marth, the English of the lower classes in England to the speech of the higher. And the English of the lower classes contains, as we have seen, a great many such grammatical forms as " I knowed," and "you says," along with others that are correct.

Franz Kielhorn (1840-1908) studied Sanskrit mainly with A. F.

I I

Franz Kielhorn (1840-1908)

Stenzler, who was influenced by Bopp. His first publication was an edition of Sntanava's Phitstra, a treatise dealing with the accent. He then went to Oxford to work with Monier Monier-Williams (to whom Weber had recommended him) on the Sanskrit dictionary. Subsequently Kielhorn proceeded to Poona, this time with recommendations from Max Mller, to take up a professorship of Orient a | Languages (that is, Sanskrit) at the Deccan College. He stayed in India for more than fifteen years and kept in close contact not only with Georg Bhler at the Elphinstone College at Bombay but also with numerous Indian scholars, including Bhandarkar. Bhler and Kielhorn initiated a new era in Western Sanskrit scholarship. Wackernagel says about Kielhorn's work in India: " Kielhorn war nun im Besitz einer Kenntnis der grammatischen Litteratur der Inder, wie sie vor ihm bloss Colebrooke in hnlicher Tiefe besessen hat, nach ihm vielleicht niemand wieder besitzen w i r d " (WackernageM908, 6). One of the first articles written by Kielhorn on the subject of the Sanskrit grammarians, " Der Grammatiker Pnini " (1885), was a reaction to a paper by Pischel, who had defended the view that Pnini had lived in the sixth century A.D. at the earliest. The second part of Kielhorn's article deals with the text of Pnini's grammar as known to the authors of the Kasika. The material of this second half was published later in English in Kielhorn's "The text of Pnini's Sutras, as given in the Ksik-Vrtti, compared with the text known to Ktyyana and Patajali " of 1887. As this second article is also included in this volume, only the first part of " D e r Grammatiker Pnini" will be reproduced here (Nachrichten von der kniglichen Gesellschaft der Wissenschaften und der GeorgAugusts-Universitt zu Gttingen 5, 1885,185-190). Needless to say, Kielhorn's conclusions with regard to the chronology of the early grammarians are now universally accepted.

A. Der Grammatiker Pnini (1885)

Franz Kielhorn

Prof. Pischel hat in der Zeitschr. d. D. Morg. Ges. XXXIX S. 95 98 zwei Verse eines Dichters Pnini aus Nami's Rudratakvylamkratippanaka mitgetheilt und wahrscheinlich zu machen gesucht da der Dichter und der Grammatiker Pnini eine Person seien. Dieser Pnini wrde in das 6te, frhestens 5te Jahrhundert nach Chr. zu setzen sein, und es wrde dann keine so groe Lcke zwischen der AstdhyyJ und der Ksik bestehen ' wie man jetzt annehmen mu, und wie es ganz unwahrscheinlich ist'. Ich will versuchen mit wenigen Worten darzulegen was uns ber das gegenseitige Verhltni der erhaltenen grammatischen Werke der Inder von Pnini bis etwa zur Mitte des 7ten Jahrhunderts nach Chr. bekannt ist. Vorher jedoch bemerke ich da Nami's Worte ebensowenig wie eine anonyme Strophe ein Beweis fr die Identitt des Dichters und des Grammatikers Pnini sein knnen und da die von Nami citirten Stellen eher gegen als fr solche Identitt sprechen. 'Obgleich', so bersetze ich abweichend von Pischel, 'der Gebrauch falscher Formen schon indirect dadurch verboten ist da vom Dichter umfassende gelehrte Bildung (und deshalb natrlich auch eine Kenntni der Grammatik) verlangt worden ist, so tadelt der Verfasser doch noch ausdrcklich den Gebrauch solcher Formen um zu zeigen da man besondere Aufmerksamkeit auf ihre Vermeidung verwenden msse, weil er findet da sogar groBe (und mit der Grammatik vertraute) Dichter (aus Unachtsamkeit) falsche Formen gebraucht haben '. Als Beweise fr den Schlu des Satzes citirt Nami zwei Stellen Pnini's in denen grhya flschlich fr grhhv und pasyaflschlich fr pasyangebraucht worden sind. Ist es aber wahrscheinlich, da der Grammatiker Pnini seine eigenen Regeln VII, 1, 37 samse * naprve ktvo lyap und VII, 1, 81 sapsyonor nityam vergessen haben sollte? Wichtiger ist die Frage nach dem Zeitalter d ieses Pnini. Wenn ich hier zunchst meine eigene Ansicht aussprechen darf, so ist es die da Pnini der vedischen Litteratur weit nher steht als der sogenannten classischen ; da er einer Zeit angehrt in der das Sanskrit mehr war als eine Sprache der Gelehrten. Ich gebe zu, da das Bestreben die Zeit eines Grammatikers aus den von ihm gegebenen Beispielen bestimmen zu wollen, zu einem sichern Resultate deshalb nicht immer fhren kann, weil wir nicht wissen ob solche Beispiele von ihm selbst gebildet oder seinen Vorgngern entnommen sind. Dieselben Beispiele finden sich in den Werken der jngeren wie in denen der lteren Grammatiker. Wre das Mahbhsya verloren gegangen, so wrden wir aus den Beispielen der Ksik fr den Gebrauch des Imperfectums (Ks. Ill, 2,111) und aus der Erwhnung des Pusyamitrafr das Alter dieses Werkes vielleicht dieselben Schlsse ziehn die jetzt fr das Mahbhsya gezogen werden trotz der Thatsache da Patajali ltere Werke ebenso benutzt hat wie es die Verfasser der Ksikgethan haben. Wenn aber Pnini z. B. in Uebereinstimmung mit dem vorherrschenden Gebrauche der altern Sprache das Periphr. Perfectum nur mit dem Verbum kr bildet, whrend die Schriftsteller des 6ten Jahrhunderts k r , bh und as ohne Unterschied gebrauchen ; wenn seine Lehre ber die Verwendung des Aorists durch die Praxis der Brhmanas besttigt wird, whrend die sptere Zeit das Verstndni fr die Unterschiede der Tempora der Vergangenheit verloren hat; wenn seine Regeln fr den Gebrauch der Casus in einem Brhmana bis in die geringfgigsten Details beobachtet werden, Abweichungen von denselben dagegen bei classischen Schriftstellern nicht selten sind, so fhlen wir da

104 Franz Kielhorn

w i r auf festerem Boden stehn, und werden Pnini eher Jahrhunderte vor Chr. als zusammen mit Klidsa, hravi u. a. etwa in das 5te oder 6te Jahrhundert nach Chr. setzen. W i e aber verhlt es sich dann mit der Lcke welche in diesem Falle zwischen Pnini und der Ksik bestehen w r d e und deren Vorhandensein Prof. Pischel als unwahrscheinlich bezeichnet hat? A p r i o r i ist es vollstndig gleichgltig, ob w i r die Ksik200 oder 1200 Jahre nach Pnini verfat sein lassen, denn w i r haben nicht den geringsten G r u n d anzunehmen da sie einer der frhesten oder berhaupt einer der alten Commentare zu Pnini's Grammatik gewesen sei. Die uns bekannten Thatsachen erhrten vielmehr das Gegentheil. Von mehr als fnfzig Regeln knnen w i r b e w e i s e n da sie in der Ksik anders lauten als sie Pnini gegeben hat. Nicht nur werden in jedem Kapitel die Lesarten und Erklrungen A n d e r e r m i t K e c i t , K a s c i t , A n y e , A p a r e , A p a r a h , Eke, P t h n t a r a , S m r t y a n t a r a , S s t r n t a r a , G r a n t h n t a r a u.s. w. eingefhrt, sondern einer der Verfasser berichtet im Eingange seines W e r k e s ausdrcklich, erfasse kurz das Beste von dem zusammen was sich in den C o m m e n t a r e n (zu Pnini) und andern W e r k e n zerstreut finde. Einige Namen der Verfasser solcher Commentare sind uns erhalten w o r d e n . Auer lteren Commentaren haben die Compilatoren der Ksik auch das Mahbhsya benutzt. Das Mahbhsya ist von Bhartrhari c o m m e n t i r t w o r d e n , und da w i r durch M. Mer's fr die Geschichte der Grammatik epochemachende Entdeckung wissen da dieser Bhartrhari vor den Verfassern der Ksik gelebt hat, so verdienen sein Verhltni zum Mahbhsya und seine M i t t heilungen ber die Geschichte dieses W e r k e s unsere besondere Aufmerksamkeit. In der V o r r e d e zum 2ten Bande meiner Ausgabe habe ich gezeigt, da Bhartrhari oft verschiedene Lesarten des Textes des Mahbhsya mitgetheilt und da er noch hufiger die abweichenden Erklrungen anderer Commentatoren erwhnt hat, deren W e r k e verloren gegangen sind. W r d e n w i r schon hieraus schlieen da Patajali und Bhartrhari durch einen vielleicht nach Jahrhunderten zu bemessenden Z e i t r a u m von einander getrennt sind, so w i r d diese Ansicht in vollem Mae besttigt durch die bekannten Verse am Schlsse des 2ten Buches des Vkyapadlya. D o r t nennt Bhartrhari das Mahbhsya das Buch des Rsi; er berichtet da es eine Z e i t gab in der das W e r k nicht verstanden w u r d e , und fgt hinzu da w i r seine W i e d e r e i n f h r u n g in den Kreis der grammatischen Studien dem Acrya Candra und andern Gelehrten verdanken. Absichtlich bergehe ich die Nachricht nach der Candrcrya unter Abhimonyu von Ko s mir gelebt haben soll, und ebensowenig w i l l ich den Stil und den e i g e n t m l i c h e n Sprachgebrauch des Mahbhsya als A r g u m e n t e fr das A l t e r dieses W e r k e s benutzen. Soviel denke ich erwiesen zu haben da es im hchsten Grade gewagt sein w r d e auch nur Patajali in eine so spte Z e i t wie das 5te Jahrhundert nach Chr. setzen zu wollen. Und welcher Z e i t r a u m liegt zwischen Patajali und Pnini ! Das Mahbhsya ist zunchst ein Commentar zu Ktyyana's Vrttikas, aber es ist ebensowenig der lteste Commentar zu jenem W e r k e wie Bhartrhari's der lteste Commentar zum Mahbhsya oder die Ksik der lteste C o m m e n t a r zur Astdhyyi gewesen sind. Denn w i r erfahren von Patajali, da A n d r e nicht nur eine ganze Anzahl von Vrttikas anders als er selbst e r k l r t (vgl. z. B. meine Ausgabe Vol. I, pag. 10, 64, 237, 247, 357, 366, 450, 465, 473 u. s. w.), sondern da sie auch den Text gewisser Vrttikas anders abgetheilt (z. B. I, pag. 193, 422), oder berhaupt anders gelesen

105 Der Grammatiker Pnini

hatten (z. B. I, pag. 179, 314). Auch hatte es in der Z e i t zwischen Kt//ana und Patajali Gelehrte gegeben welche die Lehren des Vrttikakrazu vereinfachen, zu e r w e i t e r n , oder ihnen durch die Hinzufgung nherer Bestimmungen eine exactere Fassung zu geben sich bemht hatten (vgl. z. B. I, pag. 367; 230, 281, 443; II, pag. 103, 273, 304, 397; l, pag. 468, 489; II, pag. 136 u. s. w.). In dieselbe Periode fallen, um von vereinzelt genannten Persnlichkeiten nicht zu reden, zwei Schulen von G r a m m a t i k e r n , welche beide auf der von Kt//ana eingeschlagenen Bahn weiter gegangen waren, und von denen die einen, die Bhradvjiys, in engerem Anschlsse an Kt//ana hauptschlich dessen W e r k zu verbessern oder zu vervollstndigen bestrebt gewesen waren (vgl. I, pag. 73, 136,201; II, pag. 46, 70, 233; III, pag. 199, 230), whrend die andern, die Soungs, in mehr selbstndiger Weise ihre kritischen Bemerkungen gegen Pnini selbst gerichtet hatten (vgl. I, pag. 416; II, pag. 105, 228, 238, 325; III, pag. 76,159; vgl. auch Ksikzu P. VII, 2,17). Was w i r d aus den Generationen von Gelehrten von denen w i r aus dem Mahbhs/a allein Kunde haben, wenn Pnini selbst frhestens im 5ten Jahrhundert nach Chr. gelebt haben soll ? Und dies ist nicht Alles. Die Tradition berichtet da Patajali sein eigenes W e r k verfate als ein lteres grammatisches W e r k , dessen Titel er uns selbst berliefert hat, unverstndlich geworden war und da seine Lehren auf die in jenem W e r k e enthaltenen Lehren basirt sind. W i e dem auch sei, soviel lt sich aus dem Mahbhs/a selbst ersehen da Patajali ein oder mehrere in Versen geschriebene W e r k e vor sich hatte, die n a c h der Z e i t des Kt//ana verfat waren und die Patajali in so ausgedehntem Mae benutzt hat da manche Stellen seines W e r k e s kaum anders denn als Prosaversionen metrischer Originale bezeichnet werden knnen. W e n n nun die Verfasser jener von Patajali benutzten Schriften Kt//ana (II, pag. 121) den Vrttikakra (II, pag. 176) den hagavn Ktyah (II, pag. 97) 'den heiligen Kt/a 1 nennen, gerade wie Bhartrhari den Verfasser des Mahbhs/a den Rsi Patajali gennant hat, so reden sie offenbar nicht von einem Z e i t genossen, sondern von einem Gelehrten der schon fr sieund wieviel mehr fr Patajaliein Weiser der Vorzeit war. Und was fr das Verhltni des Patajali zu Kt//ana gilt, gilt in gleichem Mae fr das Verhltni des Kt//ana zu Pnini. Auch fr ihn ist Pnini schon der Bhagavn Pninih (vergl. Ill, pag. 467), auch er hatte seine Vorgnger in der kritischen Behandlung d e r A s t d h / / I ( v g l . z . B. I, pag. 211, 365; II, pag. 19,133, 216; III, pag. 265, 377), auch er kannte ebenso w i r Patajali Commentare zu Pnini, denn seine Bemerkungen beziehn sich in vielen Fllen nicht sowohl auf den Text der Stras als auf die Erklrungen, von denen sie begleitet gewesen sind. Da das Sanskrt in der Z e i t zwischen Pnini und Kt//ana in grammatischer wie lexicalischer Hinsicht manchen Vernderungen u n t e r w o r f e n gewesen war, da in derselben Z e i t eine neue L i t t e r a t u r entstanden war, da whrend derselben Periode die Sanskrtsprechenden Hindus weite Landstrecken Indiens in Besitz genommen hatten, st von andern Gelehrten erwiesen w o r d e n , und ich glaube nicht da die hierfr wie berhaupt fr das A l t e r der Astdh//T vorgebrachten Beweise einfach durch die Identification eines Dichters und des Grammatikers Pnini zunichte gemacht werden knnen.

B. The Authorities on Grammar Quoted in the Mahbhsya (1887)

Franz Kielhorn

Kielhorn contributed a number of studies on the Indian grammarians to the 1886 and 1887 volumes of the Indian Antiquary. As many of them are full of textual references, only some will be reproduced here; others will be briefly described. Kielhorn's interest was not confined to the grammatical works of the Pninian tradition. In "Indragomin and other grammarians" (Indian Antiquary 15,1886, 181-183), for example, he showed that there was no school of Aindra grammarians prior to Pnini as had been argued by A. C. Burnell (in his Essay on "The Aindra School of Sanskrit Grammarians," 1875), by Peterson and others (including al-Brun, as we saw on pages 21-22). Kielhorn showed, however, that a grammar composed by Indra, whose full name was Indragomin, must have existed as it had been used by Hemacandra (in the twelfth century). After discussing some grammarians referred to by Hemacandra, Kielhorn arrived at the conclusion that Indragomin's grammar was probably related to the grammar of Candragomin (sixth century ? cf. page 158). In "The Cndra-Vykarana and the Ksik-Vrtti" (Indian Antiquary I5,1886,183-185), Kielhorn turned to Candragomin's grammar and found that the authors of the Ksik commentary "diligently used that grammar, although they never actually mention it." In this paper he also contrasted some of Pnini's technical terms with those of Candra. Kielhorn published seven " Notes on the Mahbhsya" in the Indian Antiquary, as follows: 1. cryadesJya (1886, 80-81) 2. Gonikputra and Gonardya (1886, 81-84) 3. On some Doubtful Vrttikas (1886, 203-211) 4. Some Suggestions Regarding the Verses (Kriks) in the Mahbhsya (1886, 228-233) 5. The Authorities on Grammar Quoted in the Mahbhsya (1887, 101-106) 6. The Text of Pnini's Sutras, as Given in the Ksik-V rtti, Compared with the Text Known to Ktyyana and Patajali (1887, 178-184) 7. Some Devices of Indian Grammarians (1887, 244-252) In the first, "cryadesJya," Kielhorn showed with the help of a selection of quotations from the grammatical commentaries that the Mahbhsya is written in the form of a dialogue, in which the principal interlocutors are theSisya, the cryadesJya, and the crya: "The Sisya brings forward his doubts and asks questions; the cryadesJya is ready to solve these doubts and reply to the questions raised, b u t . . . his answers are given hastily and without a full knowledge of all the difficulties of the matter under discussion ; so that finally the crya must step in to overrule him, and to explain the true state of the case " (80). In the second, "Gonikputraand Gonardya," Kielhorn argued that these two names are the names of grammarians quoted by Patajali, and do not denote Patajali himself. The latter view had been defended by other scholars and had led to the assumption that Patajali was the son of Gonik and a native of Gonarda. In the third, " O n some Doubtful Vrttikas" Kielhorn, considering the principles for the reconstruction of the vrttikas of Ktyyana from the Mahbhsya in which they are quoted, discusses a few cases where it is doubtful whether a passage occurring in the Mahbhsya should be regarded as a vrttika. In the fourth, "Some Suggestions Regarding the Verses (Kriks) in the Mahbhsya" Kielhorn showed that most of these

107

verses are not the w o r k of Ktyyana or Patajali, but are bor-

Authorities on
Grammar

rowed from older versified grammatical works, composed after the

Vrtt/kS.

In thefifth article, "The Authorities on Grammar quoted in the Mahbhsya," Kielhorn collected passages or expressions in which older authorities, other than Pnini or Ktyyana, were referred to by Patajali. Though this article, like most of the others, is written mainly for students specifically of the Mahbhsya, it contains information of general interest and is reproduced here (Indian Antiquary 16,1887,101-106). In the preceding note I have tried to show, that the Kriks, which we meet with in the Mahbhsya, are taken from grammatical works composed after the Vrttikas, and that Patajali has probably used the same works, even where he does not actually quote from them. In the present note I intend to collect those passages or expressions, in which Ktyyana and Patajali, or the authors of the verses preserved in the Mahbhsya, are distinctly quoting or referring to authorities on grammar, other than Pnini or Ktyyana.

Prvastra
Grammars older than Pnini are referred to by the term Prvastra, 1 which is used by both Ktyyana and Patajali, as well as in the Kriks, and which occurs six times in the Mahbhsya. According to Ktyyana (Vol. II. p. 205), Pnini may have employed the word upasorjana in the rule IV. 1,14, in the sense of apradhna, in accordance with the usage of former grammars. According to Patajali (Vol. I. p. 248), Pnini has similarly used vrdha for gotra in I. 2, 65. In a Krik in Vol. I. p. 36, the term oksora is said to have been employed in former grammars in the sense of varna, 'a letter.' In Vol. III. p. 104, Patajali refutes a suggestion of Ktyyana's by intimating, that the term citah in P. VI. 1,163, need not necessarily be the Genitive of at, but may be taken to be the Nominative of cito, 'containing a suffix that has the Anubandhac,' the Norn, having been employed by Pnini in accordance with the practice of former grammars, in which that which undergoes an operation was put in the Nom., not in the Gen. case.2 According to Patajali (Vol. III. p. 455), the word ahnah in P. VIII. 4, 7, may, by the same reasoning, be taken to be the Norn, of ohna, not the Gen. of ahan. Lastly, in Vol. III. p. 247 it is suggested that Pnini may have taken the term aun which he uses in VII. 1,18, from an older grammar, a suggestion intended to show, why the operation, which in Pnini's work usually takes place before a termination with the Anubandha does not take place in the case of the terminations under discussion. This last passage has occasioned Patajali's general remark, which has been made much of by the late Prof. Goldstcker,3 that Anubandhas used in former grammars have no effect in the grammar of Pnini.

Kaiyataon P. IV. I, 14:prvastraiabdena prvcryakrtavykaranam ucyote. 2 Kaiyataon P. VI. 1,163:prvovykam ne prathamay kryi nirdisyate ; and on P. VIII. 4, 7:fiurvcryh

kryabhajah sasthy na niradiksann ity arthah. 3 See Goldstcker's Pnini, p. 181 ; Burnell's On the Aindra School of Sanskrt grammarians, p. 40.

108 Franz Kielhorn

From all this we learn little about t h e w o r k s of Panini's predecessors. That some of t h e i r technical terms differed f r o m those used by Pnini, is probable enough, but Ktyyana's and Patajali's remarks regarding t h e particular terms mentioned are hardly of more value than t h e similar statements concerning Prvcryasamjah o r Prcm samjnh of later w r i t e r s . It may also be t r u e that some ancient grammarians, like some modern ones, 4 did use the Norn, in t h e way stated, and that they did employ o t h e r A n u bandhas. I am quite aware t o o of t h e fact, that Panini occasionally does use t h e Norn, case, where we should have expected t h e Gen. ; and ahnah in P. VIII. 4, 7, undoubtedly is t h e Norn, of ahna, just as vanam in VIII. 4, 4 is t h e N o m . of vano, and vhanam in VIII. 4, 8 t h e Norn, of vhana. A t t h e same t i m e Patajali's explanations look t o o much as if they had been invented f o r t h e occasion. A t any rate, t o take t h e w o r d citoh as a N o m i n a t i v e is impossible; and as regards t h e t e r m aun, I cannot help t h i n k i n g that Patajali w o u l d have given his explanation in a more direct and positive manner, had he in this particular instance really been possessed of any authentic knowledge regarding t h e more ancient w o r k s , f r o m which he supposes Pnini t o have b o r r o w e d .

Apisali and Sakatayana

Two only of the grammarians, who are mentioned by Panini_himself, are quoted in the Mahbhsya by name, Sakatayana and Apisali.5 But regarding the former all we are told (Vol. II. p. I 38) is, that in his opinion all nouns are derived from verbs L a statement which has been copied from the Nirukta.6 And of Apisali, only the single rule is referred to (Vol. II. p. 281), that dhenu takes the suffix ka, provided it be not compounded with the negative a. Thus much would appear to be certain, even from this solrtary quotation, that both the author of the Krik, who alludes to Apisali's rule, and Patajali, who more fully explains his remark, were really acquainted with the text ofthat grammarian's Sutra.7 e.g. the author of the Ktantra. stances pniniyam pisalam ksaCompare also in the Kriks such krtsnam, and in Vol. III. p. 125 pisalaconstructions as vlavyo vidram pniniyavydJyagautamiyh 6 (seil, padyate), Vol. II. p. 3I3.The The passage of the Nirukta referred use of the cases in the technical to in the above (Roth's edition, p. 35) structure of Pnini's rules requires a has not yet been satisfactorily exseparate and full investigation. In this plained ; here I would only state that respect, Pnini is most uncertain. He the term samvijnna is used in the undoubtedly employs the Ablative sense of ovyutponna prtipadika in the and Genitive cases in a promiscuous Mahbhsya, Vol. III. p. 436, 1.11, a manner, and he often has the Nominameaning which is not given in our tive, where we should least have dictionaries. The grammar of the old expected it. Nor are the commenSakatayana must have been lost in tators wrong, when they speak of very early times, for, so far as I know, avibhoktiko nirdesa. And from their there is no reference to it in any point of view, I am quite ready to grammatical work later than Pnini. 7 endorse the maxim chandovat strni Regarding Apisali see the preface of bhavanti. [On this point, see Birw Vol. II. of my edition of the Mah1966.] bhsya, p. 20, note. On P. II. 3, 17, 5 Sakatayana is mentioned also in Vol. Kaiyata reports, that Apisali's reading II. p. 120.In Vrt. 3. on P. IV. 1,14, ofthat rule was many akarmany Ktyyana has pisalam adhte. In andara upamne vibhsprnisu. On Vol. I. p. 12, Patajali gives the inP. V . 1 , 21, he states, that for the word
4

109

Authorities on Grammar

Anye Vaiyakaranah ; Anya Acaryah " O t h e r Grammarians" are mentioned by Patajali twice; "other Acryas " only once. In one of these passages (Vol. I. p. 87) the expression "other grammarians" may possibly denote Ktyyana, whose Vrt. 1. on P. VI. 1,144, teaches exactly what the others are stated to have taught. The two other passages are of some interest, because one of them (Vol. I. p. 48) contains the technical term Samkrama, ' a termination having the Anubandha, or k which ordinarily prevent the substitution of Guna and Vrddhi,' a term which has not yet been met with anywhere else; while the other (Vol. III. p. 177), instead of anta ranga, has the word pratyaga which in this technical sense is found in a Krik on P. VI. 4,110, and in the quotation from Gonardyaon P. 1.1, 29. I may add here that Patajali undoubtedly is quoting a rule of other grammarians, although he does not actually say so, in the words laksanam hi bhavati yvor vrddhiprasaga iyuvau bhavata ti in Vol. I. p. 310, and that very probably one or two more rules of others are alluded to elsewhere in the Mahbhsya.8

Eke; in Vrttikas
In the Vrttikas, as they are printed in my edition, Ktyyana seven times introduces other scholars by the word eke, ' Some,' which is always placed at the end of a Vrttika, and for which Patajali generally supplies the verb icchanti 'they maintain.' According9 to the Vrt. on P. I. 2, 38 (Vol. 1. p. 211) only some agree with Pnini, the practical result of which is, that Pnini's rule may be regarded as optional. In Vrt. 1 on P. II. 1,1, Ktyyana explains smarthyam by prthag arthnm ekrthlbhvah; in Vrt. 4 (Vol. I. p. 365) he states, that some take it to be parasparavyapeks. In Vrt. 1 on P. III. 1, 8, he teaches, that the suffix kyac must not be added to a base ending in m nor to an indeclinable; in Vrt. 2 (Vol. II. p. 19) he adds, that, according to some, kyac is added to go, to a base ending in a simple vowel, and to bases ending in n. In Vrt. 2 on
asate of Pnini's rule pisala and Ksakrtsna read ogranthe. Ksakrtsna is besides mentioned by Kaiyata on P. I). 1, 51, where the rule pratyayottorapadayoh is ascribed to him. A rule of the pisalh is given in the Ksik on P. VII. 3, 95. And of the pisalh and Ksakrtsnh it is reported by Helrja, in his commentary on the Prakirnaka, that they had not given the rule tadarham (P. V, 1,117. Bhartrhari merely says tadarham iti nrabdham stram vykaranntare). 8 On the Vrt. 23 on P I. 4, 2 (Vol. I. p. 310) Kaiyata has the remark prasagena vykaranntare laksanam vicroyitum ha yandesd. On the Vrt. 3 on P. II. I, 36 (Vol. I. p. 388) Kaiyata calls the statement vikrtih prakrty, with which the Vrttika begins, a Prvcrya-stra. From the remark in Vol. I. p. 100, 1.18, it appears, that the grammarians whose views are given there, instead of asya cvau (P. VII. 4, 32), had read the rule osya cvv anavyayasya. (The modern Skatyana has the rule cvau csynavyayasyeh ; see also Ganaratnamahodadhi, p. 28). And from Patajali's remark prtipadikasyntah in Vol. II. p. 7 and Vol. III. p. 87, it would seem, that he knew some such rule as is given in the first Phitstra. The Phitstra IV. 6 nyasvarau svaritau would seem actually to occur in Vol. I. p. 262, 1.12, but I have strong reasons to suspect, that in that passage the text given by the MSS. has been interpolated.I purposely have omitted here all references to the Prtiskhyas, or Chandahsstrni.as Patajali calls them. 9 In the following I am merely giving the general import of the Vrttikas referred to, not an accurate and full translation of them.

110 Franz Kielhorn

P. III. 2,146, Ktyyana sa/s, that Pnini has taught the addition of vu to nind, etc., in order to show that the suffixes taught in P. III. 2,134,177 necessarily supersede the suffix pvul of P. III. 1,133; in Vrt. 3 (Vol. II. p. 133) he adds, that, according to some, what is shown by P. III. 2,146, is, that the suffixes mentioned necessarily supersede, not merely pvul but all suffixes taught in general rules. In Vrt. 1 on P. IV. 1, 39, Ktyyana states that, contrary to Pnini's rule, asita and po//taform as/t and palit; in Vrt. 2 (Vol. II. p. 216) he adds, that, according to some, they form asiknl and paliknl in the Veda. In Vrt. 4 on P. VII. 1, 72 he teaches that, contrary to Pnini's rule, the Norn. Plur. Neut. ofbahrj is bahrji; in Vrt. 5 (Vol. III. p. 265) he adds, that according to some, it is bahrji. Finally, in Vrt. 3 on P. VIII. 1, 51 (Vol. III. p. 377) he states, that some object to the interpretation of Pnini's rule by which the words no cet krokam sarvnyat are taken to mean na cet kort sarvnyah. To the above we must add three similar statements, which undoubtedly are Vrttikas, but which have disappeared from the MSS. because Patajali's explanations of them happened to commence with identically the same words.10 The first is dyudrubhym nityrthameke, which should be inserted before 1, 4 of Vol. II. p. 396; the second, upasamastrtham eke which has to be added before the last line in Vol. III. p. 104; and the third, samo v lopam eke, which has disappeared before 1. 8 of Vol. III. p. 425. By the Vrt. 3 on P. V. 2, 97, the repetition of anyatarasym in P. V. 2,109, merely indicates, that P. V. 2, 96 prescribes only the two suffixes lac and matup, in other words, the repetition of anyatarasym is regarded as ajnpaka', in the Vrt. dyudrubhym nityrtham eke Ktyyana adds, that, according to some, the repetition of anyatarasym is not ajnpaka, but is necessary in order that the rule P. V. 2,108 may not betaken to be an optional rule. In Vrt. 1 on P. VI. 1,166, Ktyyana shows that the term jasah of Pnini's rule is superfluous; in the Vrt. upasamastrtham eke11 he adds, that, according to some, jasah is necessary because, without it, Pnini's rule would be applicable also in forms like atitisrau. Lastly, in the Vrt. samo v lopam eke on P. Vlll. 3, 5, Ktyyana states that, according to some, the final of sa m may be elided before skr, which is contrary to Pnini's and to Ktyyana's own teaching. From this, I fear, somewhat tedious exposition it is evident, that Ktyyana was acquainted with the works of other scholars who, before him, had tried both to explain and to amend Pnini's grammar, and who had subjected the wording of the Stras to that critical examination, which is so striking a feature of Ktyyana's own Vrttikas. Those who are familiar with the history of Indian grammar will probably be inclined to suspect, that Ktyyana may have borrowed from his predecessors, even where he does not distinctly refer to them ; certain it is, that he was not the first Vrttikakra.

1 None of the MSS. compared by me give these statements as separate Vrttikas, but the stops put after two of them in some MSS. and the absence of Samdhi between meke and icchonti suggest, that the text of the Vrttikas has disappeared. The Vrttika 'sarvamukhasthnam avarnam eke' may have

disappeared before the words sarvamukhasthnam avarnam eka cchanti in Vol. I. p. 61, I. 21 ; at any rate, Patajali nowhere else uses the phrase eka cchanti, except when he is explaining a Vrttika. 11 Ngojbhatta considers this to be a remark of Patajal i 's.

111
Authorities on Grammar

Vajapyayana, Vyadi, and Pauskarasadi

Compared with this, Ktyyana's references to individual scholars are of slight importance; and it may even be doubted if the three scholars named by him were really all grammarians. According to Vrt. 35 on P. I. 2, 64 (Vol. I. p. 242), Vajapyayana maintained, that words mean a genus, while according to Vrt. 45 (Vol. I. p. 244) Vydi held the opposite doctrine, that words mean individual things.12 In Vrt. 3 on P. VIII. 4, 48 (Vol. III. p. 465) the rule, that atenuis before a sibilant is changed to the corresponding aspirate (vatsah to vathsah), is ascribed to Pauskarasadi.13

Apara ha; Kcid huh : Apara ha; or Apara huh

Patajali most usually introduces the opinions of other grammarians by the phrase opara ha 'another says,' which occurs no less than 83 times in the Mahbhsya. From an examination of the statements so introduced it appears, not only that Patajali knew of grammarians whose views in individual cases differed from those of Ktyyana, or who had tried to add to, to simplify, or to render more exact, and generally to improve on, the Vrttikas of that scholar, but also, and to this I would draw particular attention, that there had been those who, before Patajali, had explained the Vrttikas. Patajali's quotations certainly prove, that others had interpreted or even read certain Vrttikas differently; and more than once he even places before us two different explanations, by others, of one and the same Vrttika. Besides, he introduces, by apara ha, opinions that are at variance with his own, also where he is not explaining Ktyyana; and he employs the same phrase before a number of Kriks. I shall not weary the reader by fully discussing here every one of the many passages which have occasioned these remarks. A few simple examples will, I trust, sufficiently illustrate what I have said above. According to Ktyyana (Vol. III. p. 321) the word samvatsara in P. VII. 3,15 indicates merely, that P. VII. 3,17 is not applicable, e.g., in theformation of dvaisamika, according to another, the same word indicates generally, that words denoting time are nowhere in Pnini's grammar included in the term parimna, and that accordingly we must, e.g., by P. IV. 1, 22 form dvivars, notwithstanding the fact that that rule contains an exception regarding words denoting a measure. On P. I. 3, 25, Ktyyana has the remark (Vol. I. p. 281), upoddevapjsamgatakaranayoh ; another has upoddevapjsamgatakaranamitrakaranapathisu. To the rule P. IV. 2, 7, Ktyyana appends the note (Vol. II. p. 273) kalerdhak', another gives the general rule sarvatrgnikalibhym dhak. In Vol. I. p. 367, Ktyyana defines a sentence to be khytam svyayakrakavisesanam) ['the verb together with indeclinables, kraka-re I at ions and their attributes']; another, simplifying that definition, merely says khytam savisesanam ['the verb together with its attributes']. In Vol. I. p. 468 another permits the two constructions sobhan khalu pnineh or pninin strasya krtih, which is contrary to the teaching of both Pnini and Ktyyana.
12 For Patajali's reference to the Samgraha, which is reported to have been composed by Vydi, see below. 13 Compare the Atharva-prtiskhya II. 6; Toittinyaprt. XIV. 12; Vjasoneyi-prt. IV. 119; and Rk-prt.

CDXXX. It may be noted that, contrary to his usual custom, Ktyyana in his Vrttika puts the name pauskarasdi in the Gen., not in the Norn. case.

112 Franz Kielhorn

In Vol. I. p. 179, Patajali's reading of two Vrttikos is savarne ngrahanam aparibhsyam krtigrahand ananyatvc ca ; another, we learn, reads dananyatvam. In Vol. I. p. 192, Patajali explains the reading jagro'gunavidhih. In Vol. I. p. 314 he shows, that another reads the Wrt. 6 without the particle ca; in Vol. I. p. 422, that another reads dravyasya instead of adravasya. In Vol. I. pp. 10, 20, 64,237,247, 357 and elsewhere, Patajali gives us his own explanations of Vrttikas and also those of another. Regarding the Wrt. 2 on P. VI, 1, 3, he informs us (Vol. III. p. 8), that some supply for trtJyasya the word ekcah, while another supplies vyajanasya; and regarding the Vrt. 2 on P. VI. 4,106 (Vol. III. p. 215), that some supply chandograhanam kartavyam, and others vvacanam kartavyam. In a similar manner he records different explanations in Vol. I. p. 424 and Vol. II. pp. 92 and 171. Again, in Vol. I. p. 390 Patajali himself proposes to substitute bhayabhltabhltibhJbhih for the one word bhayena of P. II. 1, 37, and he tells us, that another would substitute bhayanirgatajugupsubhih. In Vol. III. p. 30 he explains the word apasprdhethm in P. VI. 1, 36, to be a reduplicated form of the word spardh, and he adds, that another derives the same word from apa-spardh. In Vol. III. p. 244, he tells us, that either the rule P. VII. 1, 8 or the rule P. VII. 1,10, is superfluous, and he adds, that according to another the word bahularn of P. VII. 1, 8 and the rule P. VII. 1,10, may be dispensed with. In the same manner he mentions opinions of others, that differfrom his own, in Vol. I. pp. 172, 210, 290 and elsewhere; and he records the views of different scholars in Vol. I. p. 427, and Vol. II. pp. 19,120,151 and 254. That Patajali introduces some of the Kriks by apara aha, I have already mentioned in a previous note.141 will only add here, that the authors of the Kriks themselves allude to the opinions of other scholars, who are referred to by the words Eke or Kecid, in Vol. III. pp. 217and414.

The Bhradvjyh, Saunlgh, and Krstryh

Of individual grammarians or schools of grammarians those most frequently referred to by Patajali are the BhradvjTyas and the Saungas. The former are actually quoted ten times (Vol. I. pp. 73, 136, 201,291; Vol. II. pp. 46, 55, 70, 233; and Vol. III. pp. 199 and 230), and the latter seven times (Vol. I. p. 416; Vol. II. pp. 105, 228, 238, 325; and Vol. III. pp. 76 and 159), but it does not seem at all improbable that some of the statements, which are introduced by the phrase apara ha, or which would appear to contain suggestions of Patajali himself, may likewise really belong to either of those schools.15 Both may be described as authors of Vrttikas, and both flourished after Ktyyana. But, while to amend the Vrttikas of Ktyyana appears to have been the main object of the Bhradvjlyas, the Saungas, so far as we can judge, would seem to have criticized the text of Pnini's grammar more independently. This is indicated also by the manner in which both are quoted in the Mahbhsya. Whereas Patajali usually places the dicta of the BhradvjTyas by the side of those of Ktyyana, as it were, to point out the differences between the two, and to show how the
14 15

ante, Vol. XV. p. 231, note 17. In Vol. II. p. 209, I. 8, it seems as if Patajali himself were attempting to improve on a Vrttika of Ktyyana's ;

from Vol. II. p. 105, I. 7and p. 238, I. 10 we see that he is merely repeating a statement of the Saungas.

113 Authorities on Grammar

former have tried to improve on the latter,16 he generally cites the Saungas in support of his own statements, and without any such distinct reference to Ktyyana's Vrttikas." Thus it happens, too, that in six out of seven cases the remarks of the Saungas are introduced by the phrase evam hi saungh pathanti, preceded by istam evaitat samgrhltam or some similar expression, while the dicta of the Bhradvjyas are always introduced simply by bhradvjlyh pathanti. That of the Vrttikas of the BhradvjTyas which are cited in the Mahbhsya, one (in Vol. III. p. 199) is in verse, I have already had occasion to state in my remarks on the Kriks. A third school of grammarians, the KrostrTyas, is mentioned in the Mahbhsya only once (Vol. I. p. 46). All we learn about them is, that they considered the two rules, P. 1.1, 3 and 52 to be quite independent of each other, and were of opinion, that in any case where both rules might happen to be simultaneously applicable, the former ought to supersede the latter, an opinion which is not shared by Patajali.

Gonikputra, Gonardya, Kunaravdava, Sauryabhagavat, andVdava19

The passages in which Patajali quotes Gonikputra (Vol. I. p. 336) and Gonardya (Vol. I. pp. 78 and 91 ; Vol. fl. p. 76; and Vol. III. p. 309), I have already discussed in my second note (ante, Vol. XV. p. 81); and I have there tried to show, that Gonardya was a writer of grammatical Kriks, who in all probability lived after Ktyyana. About Gonikputra it is difficult to say anything. Later than Ktyyana is also Kunaravdava, for the two statements ascribed to him by Patajali'(Vol. II*. p. 100 and Vol. III. p. 317) are distinctly directed against Ktyyana, whose Vrttikas they show to be superfluous. Whether this Kunaravdava is really the same as Vdava, who together with the Sauryabhagavat is mentioned by Patajali in the difficult passage in Vol. III. p. 421, I have no means of deciding; nor do I know what scholar is meant to be denoted by the term Sauryabhagavat, 'the crya of the town Saurya,' as Kaiyata explains it.20 Ngojbhatta takes Vdava to be the author of the Vrt. 3 on P. VIII. 2,106, a statement, the correctness of which I doubt.

The Samgraha of Daksayana

This work, on which Patajali is reported to have based his own work, is cited in the Mahbhsya only once, in connection with the first Vrttika (Vol. I. p. 6). From that passage we learn, that
16 On P. 1,1. 20, Ktyyana has ghunot been taken from the Mahbhsya, samjnym prakrtigrahanam sidartham, is given in the Ksik on P. VII. 2,17. the hradvjlyas read ghusamjnym In commenting on that passage, Haraprakrtigrahanam sidvikrtrtham ; on datta explains saungh by sungasyP. III. 1, 38, Ktyyana has vidermkit cryasyasisyh. 18 the Bhradvjiyas add niptand vante, Vol. XV. p. 230. 19 gunatvam etc. 1 purposely have omitted in the 17 But the Saungas also more fully above Vrsyyani ; the passage, in explain the meaning of a Vrttika of which his name occurs (Vol. I. p. 258), Ktyyana in Vol. I. p. 416, and they has been copied by Patajali from the improve on another Vrttika in the Nirukta (Roth's edition, p. 31). statement alluded to in note 15 above. 20 A town Saurya is mentioned in Vol. A Vrttika of the Saungas, which has I. p. 474.

114 Franz Kielhorn

the question, as to whether words are nitya or korya, had been fully discussed in the Samgraha, and that the science of grammar had been shown to be necessary, whichever view might be taken regarding the nature of words. Elsewhere we are told that the Samgraha was composed by Vydi ; Patajali himself incidentally calls the author of it Dksyana, in Vol. I. p. 468. Considering the great bulk of the Mahbhsya, it is disappointing that we do not learn from it more regarding the history of Indian grammar, and particularly, that, what we are told in it of the predecessors of Pnini, is wellnigh valueless. But I trust, that my survey of the grammatical authorities referred to by Ktyyana and Patanjali will at least make this much clear, that Ktyyana cannot have been the first author of Vrttikas, and that between him and Patanjali there intervene a large number of writers, writers in prose and in verse, individual scholars and schools of grammarians, who all have tried to explain and to amend the works of both Pnini and Ktyyana. To what extent Ktyyana and Patanjali were indebted to those that went before them, we shall never know; judging from the analogy of the later grammatical literature of India we may, in my opinion, certainly assume, that, like Pnini himself, both have based their own works on, and have preserved in them all that was valuable in, the writings of their predecessors. In conclusion, I would here draw attention to the fact, that instead of the regular terms of the Pnimya and also in addition to them, occasionally, both in the Vrttikas and in the Kriks, we meet with a number of other grammatical termini technici. Most of those terms had doubtless been in use already before Pnini, and they were generally adopted in several of the later grammars, in preference to the more artificial terms of Pnini. But a few are themselves highly artificial symbols, which may have been invented by grammarians later than Pnini, and which remind one of the terms used in XheJainendra, where indeed one of them actually occurs. Thus, Ktyyana occasionally employs the terms svara for Pnini's ac (Vol. I. pp. 59, 123, etc.), vyajana for ha! (Vol. I. pp. 26, 42, etc.), samnksara for ak (Vol. I. p. 24 and Vol. II. p. 19), samdhyaksara for ec (Vol. I. pp. 22, 24, etc.) ; sparsa and aghosa (Vol. I. p. 355); prathama, dvitya, trtlya, and caturtha for the first, second, third, and fourth consonants of the five Vargas (Vol. III. pp. 218, 465, and Vol. I. p. 154); ayogavha, jihvmllya, and upadhmnlya (Vol. I. p. 28 and Vol. III. p. 431). For tat, lut, In, and
lun he has bhavantl, svastanl, bhavisyantJand adyatani(Vol. I. p. 443;

Vol. II. pp. 114,123,160; Vol. II. p. 143; Vol. I. p. 474; Vol. II. p. 114; Vol. III. p. 217). For the phrase samjnchandasoh he uses the artificial term tan (Vol. I. p.488; Vol. II. pp. 99 and 221); and, strange to say, for Pnini's sas, which he himself has, e.g. in Vol. II. p. 199 and Vol. III. p. 107, he employs du in Vol. I. p. 304. In addition to some of these terms we find in the Kriks, paroks for lit (Vol. I. p. 199), krita to denote the Causal (Vol. II. p. 415), and cekrlyita and carkarlta to denote the two forms of the Intensive (Vol. II. p. 232 and Vol. III. p. 359). In the Kriks we also meet with la for lopa (Vol. II. pp. 284, 378, and 425), and with ghu2' (or perhaps dyu) for Pnini's uttarapada (Vol. III. pp. 229, 247, and
318).

21

See ante, Vol. X V . p. 231, note 14.

C. The Text of Panini's Sutras, as Given in the Ksik-Vrtti, Compared with the Text Known to Ktyyana and Patajali (1887)
Franz Kielhorn

In the sixth article of his series of notes on the Mahbhsya, Kielhorn turned to Panini's text itself. This grammar was commented upon in the Mahbhsya, and again some eight centuries later in the Ksik (beginning of the seventh century; see p. 33 of this volume). It is therefore important to see whether the texts presupposed in these two commentaries are the same, and whether they are the same as the text that has been handed down to us. Kielhorn discusses this problem in a fundamental article, "The Text of Panini's Sutras, as given in the Ksik-Vrtti, Compared with the Text Known to Ktyyana and Patajali/' which is reproduced here (Indian Antiquary 16,1887,178-184). As was mentioned before, much of the material presented here had already appeared in the second half of the article " Der Grammatiker Pnini" of 1885, the first half of which has been included earlier in this volume. Considering the almost unrivalled position which Panini's Astadhyyi holds in Indian literature, it may be interesting to inquire, what alterations, if any, the text ofthat work has undergone, and to collect those rules which can be shown to be additions to the original text, or the wording of which has in any way been altered, since the rules were first enunciated by Pnini. To contribute towards the solution of this question, I intend in the present note to show, so far as this may be possible, to what extent the text of the Stras which is given in the Ksik-Vrtti, the oldest extant commentary, differs from the text that was known to Ktyyana and Patajali. In attempting to do this, I shall be mainly guided by the remarks that have been appended to certain Stras by Kaiyata, Nagojbhatta, and Haradatta, and I shall have only few occasions to go beyond, o r t o differ from, what has been already stated by those commentators. But before entering upon the question with which I am more immediately concerned here, I cannot help drawing attention to the fact that the text of Panini's rules has neither in the editions of the AstdhyyJ nor in that of the Ksik-Vrttihowever valuable those editions may be otherwisereceived that critical care and attention, which it undoubtedly deserves. For years I have been content to regard the printed text of the Stras, allowing for some misprints, as trustworthy beyond doubt. It is only lately that I have become somewhat suspicious, and having compared such MSS., as were within reach, I have come to the conclusion, that in the case of a considerable number of rules the printed text differs, more or less, from the text which is furnished by the best MSS., and that wrong readings have in succession crept from one edition into another. A few examples may show this : P. III. 1,109 all the printed texts have etistussvr. Here three old and valuable MSS. of the Ksik, and an old MS. of the AstdhyyJ which I owe to the kindness of Dr. Bhandarkar, have etistussuvr0; the MSS. of the Mahbhsya GAaK, which here as elsewhere give only the beginning of the rule, have etistussu0, and in the Mahbhsya Vol. II. p. 2, where the rule is quoted, the MSS. aK have ssuvr. From this there can be no doubt, that the right reading is ssuvr (not, as in P. VII. 4, 2, ssu-r). P. III. 2, 21 all the printed texts have divvibhnis0, and all accordingly have niskarah in the commentaries. In this case the MSS. of the Mahbhsya are of no value, because they only give the commencement of the rule divvibh0; nor is this rule quoted anywhere in the Mahbhsya. But all the three MSS. of the Ksik omit

116 Franz Kielhorn

niso from the rule and niskaroh from the commentary, and the MS. of the Astdhyay has nis added secund manu in the margin. Accordingly there can in my opinion be no doubt, that Pnini has not taught the formation of the word niskara. Judging from the quotations in Bhtlingk and Roth's Dictionary, divkara, which is taught by Pnini, is an old word, occurring twice in the Atharvaveda, but niskara is not. P. IV. 1, 62 all the printed texts have sakhyasisvti0. This rule is neither treated of nor quoted in the Mahbhsya. The three MSS. of the Ksik and the MS. of the Astdhyay have sakhyasisvi, without /t/\ and so reads Candra. P. IV. 3,119 all the printed texts have pdapdan. This rule is neither treated of nor quoted in the Mahbhsya. The three MSS. of the Ksik, both in the rule and in the commentary, and the MS. of the Astdhyay have ptapdan, and Candra has the rule nmni vtapdan. P. V. 4, 68 all the printed texts have samsnth. The MSS. of the Mahbhsya GaK and originally A, as well as the three MSS. of the Ksik. both in the rule and in the commentary, and the MS. of the Astdhyay read samsnth which singular form is supported by Mahbhsya, Vol. II. p. 438, lines 23 and 25, and p. 443, 1.15, and is no doubt correct. Not taking into account rules such as these, to which I might add a fairly large number of other rules for which the MSS. furnish a better text than the one printed, the Stras of the Ksik-Vrtti, which can be shown to diner from the Stras as known to either Ktyyanaor Patajali, may be treated of under four heads. 1. Excepting as regards the observation of the rules of Samdhi, the wording of the text has remained unchanged, but several consecutive words, which originally were one rule, have been separated so as to form two or even three rules. The technical name for this proceeding is Yoga-vibhga, ' the spl itting-up of a rule (into two or more rules).' 2. One or more words have been added to the original text of a rule. 3. The wording of rules has been altered otherwise than by the addition of one or more words. 4. Whole rules have been added to the original text of the AstdhyyJ. The particulars under each of these four heads are as follows: 1. Yoga-vibhga P. 1.1,17 uah2 and 18 urn originally formed the one rule uo m, and the splitting up ofthat rule into two was first suggested by
1 I may perhaps draw attention here as toa real gem of ingenious Interpretationto the manner, in which this word ti of the above rule has been explained by the author of the Siddhantakaumud (New Bombay Ed. No. 517). It is well known (although nothing is said about it in our Dietionaries), that at any rate in works of the Indian middle ages ti sometimes conveys the sense of prakra or tydi ' w o r d s like this,' ' t h i s and similar words.' When Hemacandra in his Sabdnusosano says pnigrhJtlti, that term means, and is by Hemacandra himself explained to mean,

pnigrhltiprakrh 'words like pnigrhiti,' 'ponigrhlti and similar words'; and when in his Lingnussona he says apotyam ti, he himself tells us that he means apatydayah. This use being well known to Bhattojdksita, that scholar connects the word ti of the above rule of Pnini's with, and in construing the rule, wishes us to place it fter, bhsym, and he then explains bhsym ti to mean bhsdou, i.e., bhsym chandasi ca. Cornment appears superfluous, 2 This word, taken by itself, does not fit into Panini's text, because it is in the Genitive case.

117 Texts o Pan i ni 's Sutras

Ktyyana(Vol. I. p. 72). Panini'sone rule would allow only urfi iti or (according to P. 1.1,14) u iti, while from Ktyyana's t w o rules we also obtain viti.
P. I. 4, 58 prdoyah and 59 upasargh kriyyoge originally

formed the one rule prdaya upasorgh kriyyoge, which has been split in two by Ktyyana (Vol. I. p. 341). To quote an example given by Kaiyata, Pnini's one rule would not allow us to account (by P. VI. 2, 2) for the accent of pr'crya, because here pro would not be termed Nipto. Kaiyata, who knew the Ksik, has the remarkyad prg evo prdoyo iti yogo vibhojyate tod prayojanakothonyo vrttikom\yad tu prdaya upasargh kriyyoga ity eko yogah pothyate tod yogovibhgah kortovyotvena codyote. P. 11.1,11 vibhs and 12 opoparibahiracavah pancamy originally formed the one rule vibhspoporibahirancovah pancamy. The division ofthat rule into two has been suggested by Patajali (Vol. I. p. 380), to make it quite clear that vibhs, as an Adhikro, is valid also in thefollowing rules P. II. 1,13, etc.; for, as Kaiyata observes, anyathehaivsyopayoga sankyeta yogavibhge tv odhikro gamyate. P. IV. 3,117 samjnym and 118 kulldibhyo vu originally formed the one rule samjnym kulldibhyo vu. The division of that rule has been suggested by Ktyyana (Vol. II. p. 317), to enable us to account by the rule samjnym for the words mksika, sragha, etc. (p. 316). Kaiyata appends the notesamjnym kulldibhyo vu iti stram vibhaktam ity orthah. P. V. 1, 57 tad osyo parimnam and 58 samkhyy somjnsamghastrdhyoyonesu originally formed the one rule tad asya parimnam samkhyy samjnasamghastrdhyayanesu, so quoted in Vol. II. p. 343,1.13. The division ofthat rule has not been actually proposed by either Ktyyana or Patajali, but it may justly be argued that Ktyyana's Vrt. 6 in Vol. II. p. 353 would have been superfluous, if to him the words tad asya parimnam had been a separate rule; and Kaiyata and Nagojbhatta are therefore in my opinion quite right, when they say, the former tad asya parimnam iti yogav'ibhgah kartovya ity uktarn bhovati, and the latter (in the Laghusabdendusekhara) uttorena yogavibhgo'tra bhsye dhvanitah. I need hardly add, that the very general rule tad asyo parimnam allows us to account for a number of words, which otherwise could not have been explained by Pnini's rules. P. VI. 1, 32 hvah samprasranam and 33 abhyostrasya ca originally formed the one rule hvah samprasranam abhyastrasya ca. That rule has been split in t w o by Ktyyana (Vol. III. p. 29), in order to account by hvah samprasranam (nou ca samscanoh) for the forms juhvoyisati and ajhovat. P. VI. 1,164 toddhitasya and 165 kitah originally formed the one rule taddhitasya kitah, so quoted in Vol. III. p. 116,1,13. The division ofthat rule has been suggested by Patajali (Vol. II. p. 253, 1.22), who by taddhitasya (citah) wishes to account for the accent of words like Kaunjyon h. But for this new rule such words, being formed with the suffix cpha (P. IV, 1, 98), would be dyudtta by P. VI. 1,197, a rule which here would supersede the rule P. VI.1,163. P. VII. 3,117 idudbhym, 118 out, and 119 acco gheh originally formed the one rule idudbhym oud oc ca gheh, so quoted in Vol. I. p. 116,1. 9, and Vol. II. p. 404,1.15. Ktyyana, after having in Vol. III. p. 342,1.10 divided that rule into the two rules idudbhym and audacca gheh, in 1.14 proposes to divide the latter rule again into the two rules out and oc ca gheh, but in 1. 22 he himself shows

118 Franz Kielhorn

this second division to be unnecessary. (Compare also the KasikaVrtti on P. VII. 3,119). Pnini's one rule would only permit the locative cases krtau, dhenau etc., not krtym, dhenvm, etc. The above are, in my opinion, all rules, in the case of which Yoga-vibhga can with certainty be proved to have taken place. It is true, that according to NgojTbhatta the two rules P. VI. 2,107 udarsvesusu and 108 ksepe also, originally were one rule, apparently because udarsvesusu ksepe has been so quoted in Vol. III. p. 121,1. H a n d p. 133,1.17; but I should not now venture to conclude from the fact that two or more rules are quoted together, that they must necessarily have been regarded as only one rule, unless indeed such conclusion could be supported by other arguments. No less than eleven times we find in the Mahbhsya the quotation ato drgho yai supi ca (P. VII. 3,101 and 102), and yet P. VII. 3,101 and 102 undoubtedly are two separate rules, of which the former has been quoted by itself four times, and the latter twice. Similarly P. VI. 2,143 and 144 have been quoted together eight times, although they are separate rules; and the same might be said of other rules. Besides, the verse in Vol. III. p. 121,1.13 appears to me to prove that P. VI. 2,108 ksepe, even before the time of Patajali, was regarded as a separate rule.

2. One or more words added to the original text of a rule

P. I. 3, 29 samo gamyrcchipracchisvaratyartisruvidibhyah originally was only samo gamyrcchibhym. The verbs vidipracchisvarati and

artisru have been added from Ktyyana's Vrttikas (Vol. I. p. 282). Drs, which also is given by Ktyyana, is mentioned in the ksik only in the commentary, whereas Candra has made it part of the rule. Kaiyata appends to the Vrttikas on P. I. 3, 29 the remark samo gamyrcchibhym ity etvat stram iti vrttikam rabdham. P. III. 1, 95 krtyh prn vulah originally was only krtyh. The addition of the words pra vulah has been suggested by Ktyyana in his Vrt. 1 on Pnini's rule, but shown to be in reality superfluous in Vrt. 2 (Vol. II. p. 81). On the Vrt. 1 Kaiyata has the note
krty ity etvat stram ity ha krtyasamjnym iti. P. III. 1,11.8 pratyapibhym grahes chandasi originally was only

pratyapibhym graheh. The word chandasi has been added by Ktyyana (Vol. II. p. 87). P. ill, 1,126 suyuvapirapilapitrapicamas ca. Here api has in my opinion been inserted from Ktyyana's Vrt. 3 on P. III. 1,124 (Vol. U. p. 88). Dabhi, which is mentioned in the same Vrttika, is in the Ksik given in the commentary on P. III. 1,126. P. III. 3, 122 adhyyanyyodyvasamhrdhrvys ca originally did not contain the words dhra and vya, which have been inserted from Ktyyana's Vrttika on the preceding rule (Vol. II. p. 155). The word avahara, which is mentioned in the same Vrttika, is in the Ksik given in the commentary on P. III. 3,122. In the Mahbhsya, Vol. II. p. 146, I. 20, where the rule has been quoted, the MSS. give it as read in the Ksik, excepting that the MS. K omits from it dhra. Kaiyata on P. III. 3,121 has the remark adhyyastra dhrvyasabdau vrttike darsand abhiyuktaih pra~ ksiptau. P. IV. 1,15, which in the Ksik ends kvarapkhyunm, originally was ending kvarapah. The term khyun has been added from Ktyyana's Vrt. 6 (Vol. II. p. 209), and it occurs also in a Vrttika of the Saungas (Vol. IL p. 105, 1.8; p. 209, 1.8; and p. 238, 1.11; quoted without khyun in the Ksik towards the end of the com-

119 Texts of Pnini's Stras

mentar/ on P. IV. 1,15) as well as in the corresponding rule of Candra's grammar. The original ending of the rule may be seen from Patajali's words kakvarapo yaas ceti on P. IV. 1,16; and Kai/ata has the notestre khyunah ptho'nrsah. P. IV. 2, 2 lksrocansakaiakardamttak originally did not contain the words sakaia and kardama, which have been inserted from Ktyyana's first Vrttika (Vol. II. p. 271) on the rule. Here again Kaiyata has the notesakalakardamayoh stre ptho'nrsah. (Incidentally I may add here that the statement sakalakardambhym anaplsyate which occurs in the Ksik on P. IV. 2, 2, is based on Candra's rule sakalakardamd v). P. IV. 2, 21 ssmin paurnamsiti samjnym. The word samjnym has been added in accordance with Ktyyana's Vrttikas on the rule, but has been declared superfluous by Patajali (Vol. II. p. 275). Kaiyata appends the notesamjngrahanam stre'nrsam iti vrttikam rabdham. P. IV. 2, 43 grmajanabandhusahyebhyas tal originally did not contain the word sabaya, which has been taken from Patajali's note on the rule (Vol. II. p. 279). Gaja, which also has been mentioned by Patajali and which Candra has in the rule, is given in the Ksik on the commentary. P. IV. 4,17 vibhs vivadhavlvadht originally was only vibhs vivadht. VJvadha has been added from Patajali's note on the ruie (Vol. II. p. 329), and is also given by Candra. Haradatta has the notevivadhasabdo vrttike darsant stre praksiptah. P. V. 2,101 prajnsraddhrcvrttibhyo nah originally did not contain the word vrtti, which has been added from Patajali's note 3 on the rule (Vol. II. p. 396) and has also been given by Candra. Here again Haradatta has the notestre vrttisabdo vrttike darsant praksiptah. P. V. 4, 50 abhtatadbhve krbhvastiyoge sampadyakartari cvih originally did not contain the word abhtatadbhve, which has been added in accordance with Ktyyana's first Vrttika on the rule (Vol. II. p. 436). Kaiyata has the noteabhtatadbhvagrahanam vrttike drstvnyaih stre praksiptam. P. VI. 3, 6 tmanas ca prane originally was only tmanas ca, and the addition of prane is Ktyyana's (Vol. III. p. 143). Such evidently is the opinion of Bhattjdksita (Siddhnta-KaumudJ, new Bombay Ed. No. 963), which I now accept as correct. Haradatta, misled by the fact that the Vrttika tmanas ca prane in the MSS. of the Mahbhsya has been put under P. VI. 3, 5, and by Patajali's explanation tmanas ca prana upasamkhynam kartavyam, instead of tmanas ca prana iti vaktavyam, takes the whole tmanas ca prane to be an addition to Pnini's original text (vrttikam evedam strarpena pathitam) ; but the words tmanas ca are necessary for the following rule P. VI. 3, 1, and NgojTbhatta (in the Uddyota and Laghusabdendusekhara) has not, in my opinion, been successful in proving that we can do without those words (atrtmanas ca prana iti visistam vrttikam ity atratyabhsyasvarasdyti vaiykarankhyym ity atra parasya ceti cena parasabdapratidvandvitaytmasabdasyaiva grahanam tad ubhayam caikastram ity huh). P. VI. 3, 40 svngc ceto'mnini originally was only svngc cetah, and amnini has been added from Ktyyana's Vrttika svngc ceto'mnini (Vol. III. p. 156; compare also Vol. II. p. 193, I. 2 and Vol. III. p. 157, 1.11). Kaiyata has the notesvngc ceta ity etvat stram iti matv vrttikrambhah. P. VI. 3, 83 prakrtysisyagovatsahalesu originally was only prakrtysisi, to which agovatsahalesu has been added in accordance

120 Franz Kielhorn

w i t h t h e suggestions of Katya/ana and Patajali. For Kat/ayana has amended Pnini's original rule by adding t o it agavdisu, and Patajali in explaining t h e V r t t i k a has given t h e example sagave savatsya soholya (Vol. III. p. 171). Kaiyata has t h e note agovatsahalesv iti bhsyavrttikadarsant stre kenacit praksiptam. P. V I . 4, Oghasibhasorhali ca. Ktyyana's V r t t i k a on the rule (Vol. III. p. 213) shows that t h e rule originally did not contain the particle ca, which has been added f o r t h e very purpose of making t h e rule, in accordance w i t h Ktyyana's suggestion, more widely applicable. In V o l . III. p. 213, 1.19 some MSS. o f t h e M o h d bhsya read t h e rule w i t h o u t , others w i t h ca. Kaiyata appends the note anyatrpJti vacand vrttikakras cakram na paptheti laksyate. P. VIII. 1, 67 pjant putam anudttam ksthdibhyah originally did not contain t h e w o r d ksthdibhyah, which has been added in accordance w i t h t h e suggestion of Ktyyana (Vol. III. p. 379). Kaiyata has t h e note ksthdibhya ti stre vrttikadarsant kaiscit praksiptam. P. VIII. 3,118 sadisvajyoh (such is the reading of the MSS. of

t h e Ksik) parasya i iti o r i g i n a l l y was o n l y sadeh parasya i iti. Svaji

which is given also by Candra, has been added f r o m Ktyyana's V r t t i k a on t h e rule (Vol. III. p. 451). P. VIII. 1. 73 nmantrite samndhikarane smnyavacanam and 74 vibhsitam visesavacane bahuvacanam originally w e r e 73 nmantrite samndhikarane and 74 smnyavacanam vibhsitam visesavacane. The new division of t h e t w o rules and t h e addition o bahuvacanam are suggested by Patajali (Vol. III. pages 383 and 384), but at t h e same t i m e Patajali himself adds that the w o r d smnyavacanam or, according t o others, visesavacanam may be o m i t t e d f r o m t h e rules. Kaiyata on 73 remarks nmantrite samndhikarana iti stram pathitam tatah smnyavacanam vibhsitam visesavacana iti dvitJyam, and on 74 bahuvacanagrahanam apninlyam iti. Finally, it may appear d o u b t f u l , if the rule P. VIII. 2 , 1 2 f r o m the beginning did contain t h e w o r d kakslvat, because t h e f o r m a t i o n o f t h a t w o r d has been specially taught in Vrt. 7 on P. VI. 1 , 37 (Vol. III. p. 33). The opinions of native scholars are divided on this point, for, while Kaiyata (on P. V I . 1, 37) rejects the V r t t i k a as superfluous (sandlvad asthivad ity atra kakslvacchabdasya niptand vrttikam nrabdhavyam), Nagojbhatta reports that others consider t h e w o r d kakslvat t o be spurious in P. VIII. 2 , 1 2 (etad bhsyaprmnyttatra, i.e. in P. VIII. 2 , 1 2 , kaksivacchabdaptho'nrsa ity anye).

3. The wording of rules altered otherwise than by the addition of one or more words
P. V. 3, 5 etado's. Patajali's remarks on this rule (Vol. II. p. 403) show that t h e reading known t o him was etado'n. Patajali considered the n superfluous, and by doing so suggests the reading etado's. Kaiyata has t h e note ha kec'id asam pathanti kecid anam. P. V I . 1,115 prakrtyntah pdam avyapare. Ktyyana's reading

o f this r u l e was nntah pdam avyapare

(Vol. III. p. 86). B u t f r o m

Vol. III. p. 89, lines 7 and 18, p. 9 1 , I. 8, and other passages in t h e Mahbhsya it appears, that the reading prakrty, instead of no, was known already t o Patajali. In t h e Ksik we have t h e n o t e kecid dam stram nntah pdam avyapara iti pathanti.

121 TextsofPnini's Stras

P. VI. 1,124 ndre ca nityam, and 125 plutapragrhy aci. Patajali's reading of these two rules was 124/ndre co, and 125 plutapragrhy aci nityam (Vbl. III. p. 87, I. 24); but on p. 89, 1.18 he declares the word nityam to be altogether superfluous, and in Vol. 1. p. 66 and Vol. III. p. 53 he cites the rule 125 without nityam. Kaiyata (on Vol. III. p. 87, I. 24) has the noteindre ceti ye stram pathanti plutapragrhy aci nityam iti tu dvityam tanmatenaisa prasnah. P. VI. 1,137 samparyupebhyah karotau bhsane, and 138 samavye ca. In the place of these two rules Patajali has had only the one rule samparibhym bhsanasamavyayoh karotau, which is so quoted in Vol. III. p. 216, I . I , and the first word samparibhym of which has been explained by Patajali in Vol. III. p. 93, 1.13. (Compare the similar explanation of sudurbhym in P. VII. 1, 68, in Vol. III. p. 262, 1.21.) P. VI. 1,150 viskirah sakunirvikiro v originally was viskirah sakunau v, a wording of the rule which was not approved of by Ktyyana (Vol. III. p. 95). In his opinion, viskirah sakunau v would mean, that 'after W, kr takes the augment sut,optionally, when one wishes to denote a bird,' whereas the real meaning of the rule is assumed to be that 'after vi, kr may take sut in case one wishes to denote a bird; ' in other words, the bird may be called vikira or viskira, while in the case of any other meaning the only right form would be vikira. (Differently Goldstcker, Pnini, p. 125). Of the commentators, Kaiyata has the noteviskirah sakunau veti straptham sritya vrttikrambhah] Haradattayath tu bhsyam tath viskirah sakunau vety etvat stram and NgojTbhattaviskirah sakunir vikiro vety anrsah ptha iti bhvah. P. VI. 4, 56 lyapi laghuprvt originally was iyapi laghuprvasya. The Substitution of the Ablative for the Genitive case has been suggested by Ktyyana (Vol. Ill p. 204). In the Mahbhsya the rule has been quoted in its original form in Vol. III. p. 288, lines 4 and 11, and in its altered form p. 191, 1.12, and p. 212, lines 1 and 14. Kaiyata has the remarkkecid cryena lyapi aghuprvasyeti sasthy antamadhypit anye tu laghuprvd iti pacamy antam. P. VII. 1, 25 adddatardibhyah pacabhyah originally was ad datardibhyah pacabhyah. The addition to ad of the Anubandha d has been suggested by Ktyyana (Vol. III. p. 250). Compare also Vol. I. p. 87, lines 17 and 18, and Vol. III. p. 48. P. VII. 3, 75 sthivuklamvcamm (such is the reading of the MSS. of the Ksik) siti originally was sthivuklamucamm siti. The alteration of the wording of the rule has been suggested by Ktyyana (Vol. III. p. 334). In Vol. III. p. 333, 1.15, where the rule is quoted, the best MSS. of the Mahbhsya give the altered form of it, which has been adopted also by Candra. Kaiyata has the notesthivuklamucamam sitlti strapthd ha . . . athav stre tantram an na tu yathopeyivn ity atropasabdo 'tantram iti pradarsanya vrttikrambhah. P. VII. 3, 77 isugamiyamm chah originally was isagamiyamm chah, as may be inferred from Ktyyama's Vrttikaon the rule (Vol. III. p. 334)isu is the reading also of Candra. The Ksik has the remark ya isim uditam ndhlyate te'city anuvartayanti, and Kaiyata appends the noteisugamlti ptho'nrsa ity heses chatvam ahallti. P. VIII. 4, 28 upasargd bahulam originally was upasargd anotparah, as may be seen from Patajali's remarks on the rule. The reading upasargd bahulam has been suggested by Patajali (Vol. III. p. 460).

122 Franz Kielhorn

4. W h o l e rules added t o t h e o r i g i n a l t e x t of t h e Astadhyay p. IV. 1,166 vrddhasya ca pjym is really a V r t t i k a o f Ktyyana's on IV. 1,163 and P. IV. 1,167 ynas ca kutsym is based on the V r t t i k a jJvamdvasyam ca kutsitam on IV. 1,162 (Vol. II. p. 265). As regards, however, the explanation of the t w o rules in the Mahbhsya and in the Ksik, t h e r e is the difference, that in the f o r m e r they are considered obligatory, while in the latter, by supplying v d f r o m IV. 1,165, they are made optional. O w i n g t o the employment of the Genitive cases vrddhasya and ynah, neither rule fits into the t e x t of Pnini's AstdhyyJ. On Ktyyana's Vrttikas Kaiyata has the remarks stresu tu kaiseid vrddhasya ca pjym iti vrttikadarsant praksiptam, and ynas ca kutsym ti stram anrsam iti vacanam. P. IV. 2, 8 kaler dhak is really part of Ktyyana's V r t t i k a on the preceding rule IV. 2, 7 (Vol. II. p. 273). P. IV. 3,132 kaup'mjalahstipadd an and 133 tharvanikasyekalopas ca are really t w o Vrttikas of Ktyyana, which in the Mahbhsya are placed under P. IV. 3, 131 (Vol. II. p. 320). O n 132 Kaiyata has the note apninlyah stresu pthah. Regarding 133 the opinions of the commentators differ; according t o Kaiyata the rule is an original Stra, but Haradatta rightly remarks prvam ca stram dam ca vrttike darsant stresu praksiptam, and on the margin of the MS. a of the Mahbhsya we have the note dam api vrttike drstv stresu praksiptam. P. V. 1, 36 dvitriprvd an ca is really a V r t t i k a of Ktyyana's on the preceding rule. (Vol. II. p. 350). By Candra the w o r d i n g of that V r t t i k a has been altered t o dvitryder an ca. Kaiyata has the note dvitriprvd an cety stresv anrsah ptha iti vrttikrambhah. P. VI. 1, 62 aci sJrsah is really a V r t t i k a o f Ktyyana's on t h e preceding rule (Vol. III. p. 41). Here t o o , Kaiyata has the n o t e vrttikam drstv kaiscit stresu praksiptam. P. VI. 1,100 nityam mnedite dci is really a V r t t i k a o f Ktyyana's on P. VI. 1, 99 (Vol. III. p. 77). Kaiyata again has the n o t e vrttikadarsant stre kaseit praksiptam. P. VI. 1,136 adabhysavyavye'pi teaches the same as, and is clearly based on, Ktyyana's Vrttikas 5 and 6, advyavya upasamkhynam and abhysavyavye ca, on P. V I . 1,135 (Vol. III. p. 92). Kaiyata has the noteadabhysavyavye'pJti strasypthe vrttikapravrttih, and Nagojbhatta adds anrsah stre pthah. Finally, P. VI. 1,156 kraskaro vrksah has been taken f r o m Patajali's notes on P. V I . 1,157 (VoL ill."p. 96). \^ere the Ksik itself has the remark kecid dam stram ndhlyate praskaraprabhrtisv eva kraskaro vrksa iti pathanti. The result of this inquiry then is as follows :The t e x t of the AstdhyyJ, which is given in the Ksik-Vrtti, differs in the case of 58 rules (excluding here the somewhat doubtful case of P. VIII. 2, 12) f r o m the t e x t which was known t o Ktyyana or Patajali. 10 ofthose 58 rules are altogether fresh additions t o the original t e x t (by which I mean here the t e x t known t o Ktyyana or Patajali). 17 rules were f r o m the beginning part of the t e x t , but in the o r i g i nal t e x t those 17 rules did not f o r m 17, but w e r e only 8 separate rules. 19 rules, which also belong t o the original t e x t , have each had one or more words added t o t h e m . The w o r d i n g of 10 original rules has been changed otherwise than by the addition of one or more words, and one rule has been altered in addition t o being split up into t w o rules (P. VI. 1,137 and 138). A l t o g e t h e r the t e x t given in the Ksik-Vrtti (and that of the AstdhyyJ in the editions) contains 20 more Sutras than the original t e x t .

123 Devices of Indian Grammarians

The origin of t h e changes, which t h e t e x t has undergone, can in most cases be traced in t h e Mahbhsya. O u t of 8 cases of Yogavibhga, 5 have been suggested by Ktyyana and 2 by Patajali. In the case of 19 rules, which have received additions, t h e words added have in 13 rules been taken f r o m t h e Vrttikas, in 4 rules from Patajali's notes, and in one rule jointly f r o m Ktyyana's and Patajali's remarks; in t h e case of one rule t h e w o r d added has not been actually taken from a Vrttika, but t h e addition has been made t o comply w i t h a suggestion of Ktyyana's. In t h e case of 12 rules, which have been otherwise changed, t h e changes can in 5 rules be traced t o Ktyyana's and in one rule t o Patajali's suggestions. O f t h e 10 rules, which have been added t o t h e original t e x t , 7 are Vrttikas of Ktyyana, 2 are based on Vrttikas, and one is a note of Patajali's. Have t h e rules of t h e AstdhyyJ s i nce t h e t i m e of t h e composition of t h e Mahbhsya undergone any changes besides those which have been indicated in t h e preceding, and in particular, is there any reason t o suppose that other new rules have been added t o t h e original text? After t h e careful study which I have given t o t h e Mahbhsya and t h e literature connected w i t h it, I feel no hesitation in answering this question in t h e negative. Besides the 1,713 rules, which are actually treated of by Ktyyana and Patajali, nearly 600 rules are fully and about 350 other rules partly quoted in t h e Mahbhsya. A n d as a large number of other rules is absolutely necessary for t h e proper understanding of those rules for which w e have t h e direct testimony of Patajali, and f o r t h e formation of words used by that scholar in t h e course of his argumentsI refer t o t h e numerous quotations at t h e foot of t h e pages in my e d i t i o n w e may rest satisfied that o u r t e x t of t h e AstdhyyJ, o r rather t h e t e x t of t h e best MSS., does not in any material point differ f r o m t h e t e x t which was known t o Patajali.

D. Some Devices of Indian Grammarians (1887)

Franz Kielhorn

The seventh and last article of the series of " notes " on the Mahbhsya deals with "Some Devices of Indian Grammarians." The study of this topic, though fundamental to the study of the Sanskrit grammarians, was resumed only half a century later (for example, in Boudon1938, below pages 358-39l,and Renou1940, Introduction, ch. Ill " Les procds d'interprtation chez les grammairiens sanskrits"). In Pnini's grammar there are certain metarules or paribhss, which state how the sdtras of the grammar have to be used and interpreted. Later commentators added other similar rules, but also introduced numerous other devices. These are the subject of Kielhorn's article (Indian Antiquary 16,1887, 244-252).

In the present note I intend to enumerate, and to illustrate by a few simple examples, some of the devicesother than regular Paribhsswhich the commentators on Pnini are in the habit of resorting to in the course of their discussions. The general aim of these devices is, to secure the right interpretation and proper application of Pnini's rules; to refute objections that might be raised to them ; to extend the sphere of the rules of the AstdhyyJ, so as to make them apply where at first sight they would seem to be inapplicable, and to render additional rules unnecessary; and sometimes also to shorten or simplify those rules. In the Vrttikas of Ktyyana their number is comparatively small ; it is greatly

124 Franz Kielhorn

increasing already in t h e w o r k of Patajali; and, t o a certain extent, one may perhaps venture t o say that, the later an author, the greater is t h e number and t h e more artificial the nature of the devices w i t h which he operates. In accordance w i t h t h e plan of these notes, I shall confine my remarks generally t o t h e w o r k s of Ktyyana and Patajali ; but I shall t r y t o indicate by one or t w o examples, how t h e practice of those older w r i t e r s has been more fully developed in the Ks'ik-

Vrtti. 1. Jnpaka1 ; cryapravrttir japayati; jnpayaty cryah.In addition to what he teaches directly, Pnini teaches many things indirectly. Regarding the interpretation of P. I. 1, 45 ig yanah samprasranam, there arises the question,Is Samprasrana a name for the vowels /', u, etc., substituted for y, v, etc., or is it equivalent to the phrase ' /, u, etc. take the place of/, v, etc' ? Pnini indirectly teaches (japayati), that Samprasra na has both meanings; for, when in VI. 3,139 he says that Samprasrana is lengthened, he shows that Samprasrana denotes the vowels /, u, etc., substituted for y, v, etc., because only vowels can be lengthened ; and, when in VI. 1,13 he rules that Samprasrana shall be substituted for a certain suffix, he shows that the word samprasranam must also be equivalent to the phrase ig yanah ' /, u, etc., take the place of y, v, etc.1 (Vol. I. p. 111). On P. III. 2,16, it may be doubtful whether we should supply only adhikarane from the preceding rule, or also karmani from P. III. 2, 1 ; in other words, whether a word like kurucara, which is formed by the rule, means only kurusu carati, or means also kurms carati. Our doubt is solved by Pnini himself; by giving in III. 2,17 a special rule for bhiksm carati, Pnini clearly intimates that in III. 2,16 we are not to supply karmani. Accordingly, kurucara can only mean kurusu carati, not kurms carati (Vol. II. p. 101). The idea, which underlies the notion o jnpaka, is a perfectly sound one. We must, in the first instance, allow Pnini himself to explain his own work. But as Pnini does not speak out openly, there is the danger lest we should make him suggest more than he really meant to say. And this danger is greatly increased when the AstdhyyJ is regarded as an absolutely perfect work, 2 in which every seemingly irregular or unaccountable proceeding must have been intentionally resorted to for the purpose of indirectly instructing the student. In IV. 2, 42, Pnini teaches that the suffix yan is added to certain bases enumerated in the rule. Patajali here raises the question : 'Why this new suffix? Why is not the suffix yan, which in every respect would serve the same purpose as yan, valid from IV. 2, 40?' And his answer is, that Pnini has purposely employed a new suffix, in order to suggest that this new suffix shall be added to other bases besides those actually enumerated. Here it will be sufficient to point out that Patajali, in the application of the principle of indirect teaching, is going far beyond Ktyyana. Ktyyana, instead of regarding the superfluous yan in P. IV. 2, 42
1 Ktyyana resorts to this device 44 times, Patajali far more frequently. And since Patajali has two cryas to deal with, Pnini and Ktyyana, he is enabled to refer us also to the indirect teaching of Ktyyana, and he has actually done so six times. As regards Pnini, it may be added that

there arejnpakas in his Gonoptha and Dhtuptho, as well as in his AstddhyyJ. 2 no hi kimeid asm'm pasymi sastre yod anorthakam syt [' I don't see anything in this work which is meaningless '] Vol. III. p. 54.

125 Devices of Indian Grammarians

as ajapaka, makes a separate new rule for the w o r d which in his opinion should have been distinctly mentioned by Pnini (Vol. II. p. 279). That japakas are often and rightly made use of in t h e Mahbhsya t o establish t h e validity for Pan in i's grammar of certain general maxims of Paribhss, I have had frequent occasions t o show in my edition of the Paribhsendusekhara. Here, t o o , maxims may be deduced from some peculiar wording of the Stras, which possibly have never presented themselves t o Pnini himself; and occasionally (as in Vol. I. p. 486, Vol. II. p. 64, and elsewhere) the commentators themselves differ both as regards the validity of a particular maxim and the jnpaka by means of which such maxim is sought t o be established. It is strange, that Pnini should have employed t h e same Anubandha n in t h e first and in the sixth of the so-called Sivastras, because by doing so he has made it difficult for us t o decide whether the Pratyhras an and in are formed w i t h the first o r w i t h t h e second n. But did Pnini really adopt this stratagem in order t o suggest, that in every doubtful case of this kind we should have recourse t o t h e (traditional) interpretation of his rules (Vol. I. p. 35),a maxim t o which Patajali draws our attention no less than adozen times in the course of his discussions? On P. V. 1,115 Ktyyana gives t h e special rule, placed in t h e Mahbhsya under V. 1,118, that the suffix vat is added also t o strJ and purns, t o form strlvat and pumvat. Ktyyana considers such a rule necessary, because w i t h o u t it the suffixes na and sna taught for strJ and pums in P. IV. 1, 87, would supersede the suffix vat of P. V. 1,115. According t o Patajali, on t h e other hand, Pnini himself shows that the suffixes taught in IV. 1, 87 do not supersede the suffix vat, inasmuch as he uses the w o r d pumvat in his rule V I . 3, 34. A n d when we object, that even so there would be no reason why we should form also strlvat, we are told that our objection is futile, because (yogpeksam jnpakam) 'the jnpaka (pumvat in VI. 3, 34) has reference t o t h e whole rule (IV. 1 , 87),' i.e. Pnini, by employing pumvat, indirectly teaches that t h e whole rule IV. 1, 87 is superseded by V. 1,115. This example of a yogpeksa j npaka3 naturally leads me on t o 2. Niptana. 4 By incidentally employing a w o r d or any form whatever, Pnini shows that that w o r d or that form is correct; 5 and if such a w o r d or form should happen t o be contrary t o any rule of his, that rule must, in this particular instance, be understood t o be superseded. The incidental employment of a w o r d o r form is thus like a special rule superseding ageneral rule. In sarvanman the initial (dental) n o no man should by P. VIII. 4, 3 be changed t o the (lingual) ; but that change does not take place, because Pnini in 1.1, 27 puts down sarvanmni w i t h a (dental) n (Vol. I. p. 86). On P. III. 3, 90 Ktyyana demands a special rule, t o teach that the root prach before the suffix nah, is not by P. VI. 1,16 changed t o
Patajali in Vol. I. p. 83 ; Vol. II. pp. 81, 238, 347, 365. Compare also avisesena jnpakam in Vol. U. p. 110. The expression yogpeksam jnpakam has been curiously misunderstood in the late Prof. Goldstcker's Pnini, p. 116. 4 Ktyyana in Vol. I. pp. 52, 86, 478 ; Vol. II. p. 406; Vol. III. pp. 103,105, 123,194, 255 ; (compare also 455).
3

Patajali twice as often. In Vol. III. p. 224 Patajali refers us by the phrase niptand etat siddham to a Vrttika of Ktyyana's. 5 Compare Vol. II. p. 413 desyh stranibandhh kriyante i.e. destavyh sdhutvena pratipdy ete stre nibadhyante prasagena sdhutvapratipdanrtham.

126

p.rch; in other words, that prach + nan = prosno, not prsna. Accord-

Franz Kielhorn

ing to Patajali, no such special rule is needed, because Pnini employs the word prasna in III. 2,117. (Vol. II. p. 151). On P. 1.1, 47 Ktyyana makes a special rule to account for bharj and marci. Patajali considers such a rule superfluous, because Pnini has the two words in his Ganas (Vol. I. p. 115). 3. Anabhidhana.6Grammar is not to invent new words or new meanings, but has to concern itself with existing words only, to show which are right words, and to explain their formation and usage. The grammarian need not take into account any possible wrong words which nobody would think of employing; he does his duty if he gives his rules in such a manner as to account for the right words, and to exclude wrong words which people actually do use. Reasoning like this would appear to have led to the device of anabhidhana, which has been frequently resorted to by Ktyyana and Patajali, especially in those chapters of Panini's grammar which treat of the addition of suffixes, sometimes to show that Pnini has said more than he need have said, and sometimes to defend him from the charge of having said too little. If nobody thinks of using a particular word, or of using a word in a particular sense, it may be said that such a word would mean nothing, or would not convey the requisite meaning, and it may therefore seem unnecessary to forbid its use or its employment in that particular sense. According to P. IV. 2 , 1 , a certain suffix is added to a word denoting a colour to signify ' coloured by (or with) that.' The suffix is said to be added ' to a word denoting a colour,' apparently to prevent its addition, e.g., to devadatta, in the expression devadattena raktarn vastram 'cloth coloured by Devadatta.' According to Ktyyana, Pnini might have omitted the words 'to a word denoting a colour,' for the suffix taught by Pnini is added to denote the meaning 'coloured by (or with),' and that meaning would not be denoted by daivadatta in daivadattam vastram. Everybody will understand this phrase to mean ' cloth belonging to Devadatta,' and nobody would employ daivadattam vastram in the sense of'cloth coloured by Devadatta'7 (Vol. II. p. 271). In 111.2,1, Pnini is not obliged to tell us that the suffix, which in accordance with his rule is added in kumbhakra, must not be added to drs, to express the sense of ddityam pasyati, because the word dityadarsa would not convey the requisite meaning, or in other words, because nobody would think of using the word dityadarsa.8 (Vol. II. p. 94). On P. V. 2, 65
Ktyyana, in Vol. I. p. 424; Vol. II. pp. 12, 13, 94,146, 234, 271, 325 ; Vol. III. p. 365; and in other passages, where anabhidhana is referred to by the w o r d uktam or the phrase uktam v. Patajali in Vol. I. p. 177; Vol. II. pp. 25, 250, 274, 307, 308, 309, 319, 334, 341, 351, 358, 382 (twice), 387, 395, 398, 399. I may perhaps draw attention t o the fact, that most of these references are t o Vol. II. of the Mahbhsya, which, generally speaking, treats of Krt and Taddhitasuffixes.
7 The device of anabhidhana may appear so strange, that it is perhaps 6

Kaiyata on P. IV. 2,1 : raktadJnam sabdnm yo'rthah sa eva yadi laukike prayoge pratyayenbhidhlyate tad pratyayo bhavati nnyath prayuktnm sabdnm sdhyasdhuvivekya sstrrambht | devadattena raktam vastram iti vkyyo rtho'vagamyate nsau daivadattam vastram ity ato'vagamyate svasvmisambandhasyaivtah sampratyayt\ 8 See Kaiyata on P. III. 2 , 1 anabhidhnd iti \ nitynm sabdnm dam anvkhynamtram \ na cdityam pasyatitydy arthapratipdanyd ity adarsdayah sabd loke prayujyanta iti sstrenpi sdhutvena nnusisyanta ity

advisable to quote the following from

arthah |

127 Devices of Indian Grammarians

Ktyayana would wish t o alter t h e w o r d i n g of Pnini's rule, so as t o make it quite clear that t h e words dhanaka and hiranyaka, w h i c h , are formed by t h e rule, mean ' a desire for wealth ' and ' a desire for gold,' and do not mean ' desirous of wealth ' and ' desirous of gold.' According t o Patajali, Panini's rule need not be altered ; the suffix taught by Pnini cannot be added in t h e sense of ' desirous of,' (anabhidhnt), because dhanaka and hiranyaka would not convey that meaning (Vol. II. p. 387). These instances will sufficiently prove, that t h e device of onabhidhno, beyond acquainting us w i t h t h e views of t h e commentators w h o happen t o make use of it, is really of no value whatever. W e know that a w o r d cannot be used, o r does not convey a particular meaning, and therefore we want no rule forbidding its use, o r its employment in a particular sense. W h y , we may well ask, do we study grammar at all, if we know beforehand what words cannot be used ? It is right t o add, that at any rate Ktyyana, in general, has had recourse t o anabhidhna only as t o an alternative proceeding, and that t h e weak point of t h e device has been clearly perceived by t h e Indian grammarians themselves. Haradatta, when commenting on P. III. 2 , 1 , says that anabhidhna must be resorted t o , only where t h e authorities tell us t o do so, and that elsewhere we must simply follow t h e rules of grammar. 9 4. Vivaks. 10 Similar t o anabhidhna, and liable t o t h e same objection, is t h e device of vivaks, which is a few times made use of in t h e Mahbhsya, and is more frequently employed in t h e KsikVrtti. Vivaks means ' t h e wish t o say a t h i n g ' ; and vivaks o r iaukikl vivaks, as understood here, is t h e desire of those w h o speak a language t o convey certain meanings by certain words, t h e manner in which people employ t h e words of their language, t h e prevailing and generally understood usage of words (pryasya, i.e. lokasya, sampratyayah). According t o Patajali, Pnini refers us t o this common usage by t h e w o r d iti, which he occasionally employs in a rule. By that iti Pnini, according t o Patajali, indicates that such a rule of his must not be observed generally and under all circumstances, but has its application limited by general usage. The rule must be observed only so far as it may lead t o t h e formation of such words as are used by people, or of words which are really used in t h e particular sense indicated by t h e rule. W e may by P. V. 1,16 say prsdlyam dru [' palatial t i m b e r ' ] , t o convey t h e meaning prsdo'sya drunah syt [ ' t i m b e r sufficient t o build a palace'], but w e cannot by t h e same rule f o r m prsdlyo devadattah [' palatial Devadatta'] : in t h e sense of prsdo devadattasya syt [' Devadatta sufficiently able t o build a palace'], because people would not understand this meaning f r o m prsdlyo devadattah (Vol. II. p. 343). 5. Yogavibhga is t h e splitting up of a rule into t w o or more separate rules. This proceeding has been suggested about 25 times by Ktyyana, and rather more than 70 times by Patajali. Its general p u r p o r t is, w i t h o u t altering t h e w o r d i n g of t h e t e x t of t h e AstdhyyJ, simply by a different division o f t h a t t e x t , t o obviate objections that might be brought against Pnini's rules, and t o
9 10

Haradatta on P. III. 2, I,tac cnabhidhnam yatrptair uktam tatraiva\ anyatra tu yathlaksanam bhavaty eva\ tath ca pathati yathlaksanam bhavaty eva\ tath pathati yathlaksanam aprayukta iti |

Ktyyana in Vol. II. p. 282. Patajali in Vol. II. pp. 275, 342, 393. See also Ksik-Vrtti on P. II. 2, 27; IV. 2, 21, 55, 57, 58, 67 ; IV. 4,125 ; V. 1,16 ; V. 2, 45, 77, 94, 95,1 07, 112,115 ; V. 4, 10.

128 Franz Kielhorn

make those rules teach more than they w o u l d teach otherwise, or that Pan i n i has taught himself. Examples, both f r o m the Vrttikas and f r o m the Mahbhsya, have been already given in my last note. I may add here that the commentators, w i t h o u t resorting t o actual yogavibhga and thus increasing the number of the Sutras, occasionally meet objections by joining the first w o r d or words of one rule on t o a preceding rule, or by dividing the words of the t e x t differently f r o m what they themselves show t o be the ordinary or generally accepted division of it. In Vol. I. p. 272 Ktyyana suggests the possibility of dividing the t e x t of the rule P. I. 2, 11

svaritendhikrah, usually divided into svaritena and adhikrah, into the three words svarite, no, and adhikrah. In Vol. I. p. 271 Patajali proposes to join the word svaritena of the same rule on to the preceding rule, which would then read yathsamkhyam anudesah samnnn svaritena, and to make P. I. 3,11 consist of only the one word adhikrah. In Vol. II. p. 228 Patajali proposes to transfer the word sarvatra from the beginning of P. IV. 1,18 to the end of the preceding rule IV. 1,17. In Vol. II. p. 11 he meets an objection by dividing the words dirghas cbhysasya of P. III. 1, 6, usually divided into dirghas ca-\- abhysasya, into dirghas ca-\- bhysasya. In Vol. III. p. 11 he divides jaksitydayah ^jaksiti^r dayah) of P. VI.1, 6 \ntojaks-\- ity dayah; in Vol. I. p. 152 vareyalopa of P. 1. 1, 58 i nto vare + 'yalopa0, etc. 6. Ekayogah karisyate.11As a single rule may be split in two, so two rules may be joined together, so as to form one rule. The advantage sought to be derived from this device, which has been five times resorted to by Patajali, may be seen from the following example. It may be argued that gunavrddhl in P. 1.1, 3 (ko gunavrddhl) is superfluous, because vrddhih and gunah will be valid from P. 1.1,1 vrddhir ad aie and P. I. 1, 2 ad en gunah. The objection to this would be, that vrddhih of P. 1.1,1 would be valid also in P. 1.1, 2, and that accordingly a, e, o would by P. 1.1, 2, be termed both Guna and Vrddhi. But that objection is met by the suggestion, that P. 1.1,1 and 2 should be made one rule, Vrddhir ad aij ad en gunah. In this single rule the term vrddhih, with which the rule opens, would not be valid in the concluding portion of the same rule, and from this rule both vrddhih and gunah could then be supplied in the following rule (Vol. I. p. 44). 7. Sambandham anuvartisyate; sambandhnuvrtti; sambandhavrttiS2But the objection raised to the validity of the term vrddhih of P. 1.1,1, in P. 1.1, 3, may be met also by another device, which Patajali has resorted to more frequently. In P. 1.1, 2 the whole rule vrddhir ad aie may be regarded as valid, and subsequently gunah and vrddhih may be regarded as valid in P. 1.1, 3. The case in fact would stand thus, P. 1.1,1 vrddhir daie. P. 1.1,2 od en gunah;vrddhir d aie valid from the preceding. P. 1.1, 3 ikah;gunah and vrddhih valid from the preceding. 8. Mandkagatayo 'dhikrh;13 mandkapluti.And there is even a third way of meeting the same objection. There is no reason at all why vrddhih of P. 1.1,1, should be valid in P. 1.1, 2. Like a frog, it may leap across P. 1.1, 2 and alight on P. 1.1, 3. This device has been resorted to by Patajali seven times.
Patajali in Vol. I. pp. 44, 482; Vol. III. pp. 25,162, 315. 12 Patajali in Vol. I. pp. 44, 190, 457, 482; Vol. II. pp. 127,151, 267, 290, 372;
11

Vol. III. pp. 25, 52, 8,148, 238, 271, 410, 425, 431, 433. 13 Patajali in Vol. I. pp. 44, 457, 482; Vol. II. p. 372; Vol. III. pp. 25,161, 314.

129 Devices of Indian Grammarians

9. IstavcT parasabdah. 14 In I. 4, 2 Panini prescribes that of t w o conflicting rules t h e subsequent (para) rule, in t h e order of t h e AstdhyyJ, shall take effect in preference t o t h e preceding rule. N o w Ktyyana on various occasions shows that Pnini's rule is not universally t r u e , and he points o u t a number of purva-vipratisedhas, i.e. instances in which t h e preceding rule must take effect in preference t o t h e subsequent rule. According t o Patajali, on the other hand, t h e special rules given by Ktyyana are unnecessary, and t h e objections o f t h a t grammarian only show that he has not fully understood t h e w o r d para in P. I. 4, 2. Para, amongst o t h e r things, also means ' desired ' (ista), and what Pnini really teaches is, that of t w o conflicting rules it is t h e desired rule that should take effect, i.e. that rule, whatever be its position in t h e AstdhyyJ, the application of which will lead t o correct words. Here again, then, w e ought t o possess a perfect knowledge of t h e language, if we would rightly apply t h e rules of Pnini's grammar. 10. Pratyhragrahana. 15 A t e r m ending w i t h an Anubandha, which at first sight w o u l d appear t o denote a single suffix, r o o t , etc., is occasionally by Patajali explained t o be a collective t e r m denoting a series of suffixes, roots, etc. Thus mtrac in P. IV. 1,15 is not, as might be supposed, t h e suffix mtrac taught in P. V. 2, 37, but is taken t o be a Pratyhra or collective t e r m , formed of mtra in P. V. 2, 37 and t h e Anubandha c of ayac in P. V. 2, 43, and denoting, accordingly, all suffixes from mtrac in t h e f o r m e r up t o and including ayac in t h e latter rule (Vol. I. pp. 106 and 138). The most interesting example occurs in V o l . I. p. 289, and again in V o l . II. p. 47. In t h e older w o r k s of Sanskrit literature t h e Periphrastic Perfect is formed only w i t h kr\ and this is exactly what Pnini teaches in III. 1 , 40, where he says, that km, i.e. kr which has t h e Anubandha f?, is employed in t h e Periphrastic Perfect. Patajali, however, desirous of accounting by Pnini's rules for Perfects such as ham sa, Jhm babhva, explains km in P. III. 1, 40 t o be a Pratyhra, formed of kr in P. V. 4, 50 and t h e Anubandha n of km in P. V. 4, 58, and including therefore bh and as, which in P. V. 4, 5 0 f o l l o w immediately upon kr. 11. Praslistanirdesa. 16 A long or even a short vowel often results f r o m t h e coalition of t w o or more vowels. H o w this simple fact may be t u r n e d t o account in grammatical discussions, may be seen f r o m t h e following examples. In V o l . I. p. 501, Ktyyana states that t h e single vowel (da), which by P. II. 4, 85 is substituted in t h e Periphrastic Future for t h e ordinary personal terminations ti and ta, takes t h e place of t h e whole original terminations (and not merely, according t o P. 1.1, 52, of their final letters), because may be regarded as a combination of t h e t w o vowels c/ + ; and that for this reason Pnini is justified in not attaching t h e Anubandha s t o t h e substitute do (compare P. 1.1, 55). According t o Patajali ; Pnini might similarly have o m i t t e d t h e Anubandha s of t h e t e r m as in P. II. 4, 32 (Vol. I. p. 481), and of t h e same t e r m as in P. VII. 1, 27 (Vol. III. p. 251), because even (short) a may be regarded as a combination of a + a. According t o Patajali, again, oka in P. II. 3, 69 may be regarded as t h e result of t h e combination o f / a + u +
Patajali in Vol. I. pp. 46,194, 306, 404; Vol. II. pp. 237, 279, 337; Vol. III. pp. 18, 99,134, 201, 238, 276. 15 Patajali in Vol. I. pp. 106,138,141, 289, 470, 488; Vol. II. pp. 47,130, 203. 16 Ktyyana in Vol. I. p. 501. Pata14

jali in Vol. I. pp. 47, 88,139,140, 469, 481 ; Vol. II. pp. 46, 52,184, 403; Vol. III. pp. 151, 212, 251, 273, 312. Compare vikranirdesa, Ktyyana in Vol. I. p. 202.

130 Franz Kielhorn

uka, and no additional rule is required t o teach that words like cikJrsu, which are formed w i t h u, are not construed w i t h the Genitive case (Vol. I. p. 469). 12. Ekasesanirdesa. 17 Pnini, according t o the commentators, also employs other terms once only, instead of repeating t h e m . An instance is afforded by the same rule P. II. 4, 85, which has been mentioned under the last heading. In that rule Pnini teaches that daraurasah, i.e. as one w o u l d say, da, rau, and ras, are substituted for the t h i r d personal terminations of the Periphrastic Future. Here it may be objected that, as t h e r e are six such terminations, t h r e e in the Parasmaipada and t h r e e in the tmanepada, and only t h r e e substitutes, the rule P. I. 3,10, which determines the o r d e r in which substitutions take place, w o u l d not be applicable, and that Pnini therefore ought t o have shown in some other way, how the substitution should take place. Such objection is met by the statement that daraurasah is an ekasesa-nirdesa, for daraurasah + daraurasah, i.e. da rau ras-\- da rau ras; and the number of sub-

stitutes having thus been shown to be six, the order of substitution is after all regulated by P. I. 3, 10 (Vol. I. p. 500). To P. I. 1, 27 sarvdJni sarvanmni Ktyyana wishes to append the note that sarva etc., when used as proper names, are not termed Sarvanmni. Such a note, however, is by Patajali regarded as superfluous, because sarvdJni sarvanmni may be taken to stand for sarvdJni sarvdJni sarvanmni sarvanmni. 'sarva etc., are (termed) Sarvanmni; (and the) sarva etc., (here spoken of) are nouns denoting anybody' (not proper names). (Vol. I. p. 88). 13. Avibhaktiko nirdesah.18Pnini on rare occasions does put down in his Stras certain terms without the case-terminations, which we should have expected him to attach to them.19 But this will hardly be considered to justify the commentators in assuming an avibhaktika nirdesa, 'the employment of a term without termination,' in instances like the following. The wording of P. VII. 3, 82 and 83 is mider gunah jusi ca, or, when the two rules are joined according to the rules of euphony, mider guno jusi ca. To meet certain objections, Patajali in Vol. I. p. 47 takes the first rule to consist of the three words mid eh gunah, where the base mid would stand for the Genitive case midah Guna is substituted for the / of mid'); and in Vol. III. p. 335 he further divides gunojusi ca into guna + ujusi ca, where the base guna would stand for the Nominative case gunah ('Guna is also substituted before jus, when jus commences with u'). 14. Luptanirdista.20Occasionally a consonant (usually y or v) is supposed to have been elided in the text of the Stras. Such consonant would of course have to be replaced, when explaining the text. On P. I. 3, 7, Ktyyana demands a special rule, to teach that, contrary to Pnini's rule, the initial c of the suffixes cucup and canap is notan nubandha. According to Patajali, no such rule is required because the two suffixes really begin with the letter y, which has been elided in the text (Vol. I. p. 263).

Ktyyana in Vol. I. pp. 261, 350, 369; Vol. III. pp. 167, 467. Patajali in Vol. I. pp. 88,156, 212, 500. 18 Patajali in Vol. I. pp. 21, 25, 47; Vol. II. p. 46; Vol. III. pp. 242, 335, 414. 19 Compare e.g. P. III. 3,17 sr; III. 3,

17

30 kr; III. 3, 48 vr; VI. 1,184 and VI. 4, 6 nr\ VI. 3, 62eka; VI. 4,142 t/. 20 Patajali in Vol. I. pp. 49, 263 (twice); Vol. II. p. 52; Vol. III. pp. 43, 245, 257. The letters supposed to have been dropped arey, v, and once n.

131 Devices of Indian Grammarians

15. Dvi-krako nirdesah.21More often Panini is supposed to have employed a double consonant, where the actual text of the Stras has only a single consonant. On P. VIII. 3, 5, samah suti Ktyyana suggests that, before the augment sut (in sam-\- skort), s (not ru) should be substituted for the final of sam (= samskart). According to Patajali, Ktyyana's remark is superfluous, because the substitution os has been taught in Pnini's own rule, the wording of which really is samah ssuti (or samasssuti), i.e. 's is substituted for (the final of) sam before sut' (Vol. III. p. 425). Theoretically it is perfectly true that, by the rules of euphony, original samah ssuti might have been changed to samah suti, but how little importance can be attached to Patajali's remark, may be seen from the fact that quite a different meaning has been assigned by him to the same rule samah ssuti in Vol. III. p. 94. 16. Anvarthasamja.22Technical terms are employed for the sake of economy or brevity. When, then, Pnini uses other than short technical terms, he does so to show that the application of those terms accords with their etymological meaning. On the rule 1,1, 23, in which Pnini teaches that bahu, gana, etc. are termed Samkhy, Ktyyana would wish it to be distinctly stated that the ordinary numerals eka, dvi, etc. also are termed Samkhy, because otherwise these numerals would not in Pnini's grammar be denoted by the term Samkhy. But Ktyyana's suggestion is rejected by Patajali, on the ground that Samkhy is an anvarthasamja. The term Samkhy denotes the ordinary numerals eka, dvi, etc., because samkhy is derived from sam-\-khy 'to count together,' which is exactly what the ordinary numerals do (Vol. I. p. 81). 17. Tadvadatidesa.23Or it may be said, that in the rule mentioned Pnini has used the word samkhy in the sense of samkhyavat. Pnini does not say at all that bahu, gana, etc., are termed Samkhy ; what he teaches is, that bahu, gana, etc., are treated as if they were numerals24 (Vol. I. p. 81). 18. Prakarsagati.25As in ordinary life, so here, sometimes the mere fact that Pnini employs a word is sufficient to show that particular stress is laid on such word ; that the word has reference to things which under all circumstances, or in a high degree, or more than other things (sdhlyah), are what the word means. When in 1.1, 56 Pnini uses the term al-vidhi 'an operation depending on a letter,' he means such operations as depend on letters as such, not operations which depend on suffixes that happen to be letters (Vol. I. p. 136). When in III. 1, 94 he speaks of asarpa (i.e. not uniform) suffixes, he must mean suffixes which are asarpa
21

Patajali in Vol. I. p. 139; Vol. III. pp. 94, 254, 425; (dvisakrako nirdesah);Vol. I. p. 155; Vol. II. p. 68; Vol. III. pp. 37,188, 208; (dvila0); Vol. I. p. 170; (dvisa0);Vol. II. p. 20; (dWco0) ;Vol. III. pp. 48, 250; (dvida0);Vol. III. pp. 108, 410; (dwno); Vol. III. p. 257; (dvima) ; Vol. III. p. 351 (divita, trita). This is perhaps the right place for the remark that the wrong reading kkiti ca of the rule P. 1.1, 5 (kiti ca) owes its origin to Patajali's suggestion in Vol. I. p. 269 and Vol. II. p. 132, kakre gakras cartvabhto nirdisyate. Com-

pare a similar expression in Vol. III. p. 110. 22 Patajali in Vol. I. pp. 81, 89, 96, 215, 324, 346, 378; Vol. II. pp. 3, 76. Compare anvarthagrahana in Vol. I. pp. 88, 209, 227, 229, 237, 472; Vol. II. pp. 303, 416; Vol. III. pp. 98,120, 415. 23 Patajali in Vol. I. pp. 81,191, 469. Vatinirdesa in Vol. III. p. 272. 24 Accordingly, Hemacandra in his rule 1.1, 39 has datyatu samkhyavat. 25 Patajali in Vol. I. pp. 136, 261, 283 (twice), 329, 355, 370, 380; Vol. II. pp. 80, 334; Vol. III. pp. 164, 369, 430.

132 Franz Kielhorn

in the ordinary language as well as in the language of grammar, because in the latter all suffixes are asarpa (Vol. II. p. 80). A n d when in I. 3, 3 he speaks of a last consonant, he must be understood t o mean the final consonant of complete roots, suffixes, etc., and Kt//ana need not have attempted t o improve on Pnini's rule (Vol. I. p. 261). 19. A k r o matvarthyah; matublopah. 2 6 To meet objections of various kinds, a noun ending in the vowel a is occasionally regarded as a derivative noun formed w i t h the possessive suffix a; or it is stated that the possessive suffix matup, which should have been added, has been dropped. A simple instance is furnished by the w o r d anudtta in P. VI. 1,158. As the meaning o f t h a t rule (anudttam padam ekavarjam) is, that ' w i t h the exception of one v o w e l , a w o r d contains only anudtta vowels,' one might have expected Pnini t o say (not anudttam padam, but) anudtth pade or anudtth padasya. The actual w o r d i n g of the rule is nevertheless correct, because anudttam does not mean anudtta, but means ' containing anudtta vowels ;' either the possessive suffix has been dropped, or the final a of anudtta is the possessive suffix a. (Vol.

III. p. 97.) 20. Tadarthyat tacchabdyam ;27 sahacaryat tacchabdyam ;28 ttsthyt tacchabdyam.29Things subservient to something else, or things which are in company with something else, or things which are in a particular place, may be called by the names of the things to which they are subservient, or in company with which they appear, or of the place where they are. It is thus that the term samsa in P. I. 2, 43, denotes the rules which teach the formation of compounds (samsrtham sstram; Ktyyana, in Vol. I. p. 214); and that tatpurusah in P. I. 2, 42 means the words forming aTatpurusa-compound (Patajali, against Ktyyana in Vol. I. p. 214). Thus, too, the rule P. 1.1, 30 trtJysamse is supposed to account for the fact that, e.g., in msena prvya dehi ['give earlier by a month '] the word prva does not follow the pronominal declension. (Vol. I. p. 92.) And, by asimilar reasoning, suggestions of Ktyyana have been rejected by Patajali in Vol. II. p. 388 and
p. 437.

As might have been expected, most of these devices which I have collected from the Mahbhsya, have been made use of also in the Ksik-Vrtti. In addition to them, the compilers ofthat commentary have employed other devices among which the two following, with which I conclude this list, are perhaps those most commonly resorted to. 21. Vyavasthitavibhs.30When Pnini teaches that an operation takes place optionally, we naturally understand him to mean, that such operation may or may not take effect in every individual instance which may fall under the rule. This, too, is clearly the opinion of the earlier commentators, as may be seen from various critical remarks which they have appended to some of Pnini's
26

Patajali in Vol. I. pp. 156, 208; Vol. II. pp. 104,179, 376, 377; Vol. III. pp. 30,97,104,171. 27 Ktyyana in Vol. I. p. 214; Vol. II. p. 312. Patajali in Vol. I. pp. 91, 92, 214, 225, 332, 452, (twice); Vol. II. pp. 283, 331, 338, 359; Vol. III. p. 32. 28 Patajali in Vol. I. pp. 180, 202; Vol. ll.pp.284, 309,360, 388.

29

Patajali in Vol. II. p. 437. Compare also Vol. II. p. 218,1.15. Ksik-Vrtti onP. 111.1,144. 30 See e.g. Ksik-Vrtti on P. I. 2, 21 ; I. 3, 70; I. 4, 47; II. 3,17; 60; III. 1,11 ; 90; 143; III. 2,124; III. 3,14; III. 4, 85, 86; VI. 1, 27, 28, 51 .Compare also Mahbhsya, Vol. II. p. 165; III. p. 350.

133 Devices of Indian Grammarians

o p t i o n a l rules. T o obviate such remarks, and generally w i t h a v i e w t o account by Pnini's o w n rules f o r t h e actual facts of t h e language as k n o w n t o t h e m , later grammarians have invented t h e device of vyavastbita-vibhs. A n o p t i o n a l rule need n o t be o p t i o n a l in every case, b u t may be taken t o teach, e i t h e r , t h a t an o p e r a t i o n in part i c u l a r instances necessarily must take place, w h i l e in o t h e r s it is not allowed t o take place at all ; o r , t h a t t h e o p e r a t i o n is really o p t i o n a l in a l i m i t e d n u m b e r of instances only, w h i l e in o t h e r i n stances, as t h e case may be, it must necessarily t a k e place o r may not take place. In I. 2, 21 Pnini teaches t h a t in certain f o r m s roots w i t h p e n u l t i m a t e u o p t i o n a l l y take Guna; dyut thus f o r m s dyutita o r dyotita. A c c o r d i n g t o Ktyyana, Pnini's rule is t o o w i d e ; it should have been l i m i t e d t o roots of t h e first class only. A c c o r d i n g t o t h e Ksik-Vrtti, t h e rule affords an instance of vyavasthitavibbs; t h e o p e r a t i o n t a u g h t by it o p t i o n a l l y takes place in roots of t h e first class; in t h e case of o t h e r roots it does n o t take place at all. By P. III. 1,143 grab f o r m s e i t h e r graha o r graba; but, t h e rule being a vyavasthita-vibhs, groh forms only grabo in t h e sense o f ' a planet,' and o n l y grba in t h e sense o f a 'shark. 1 A c c o r d i n g t o P. I. 4, 47 t h e v e r b abbinivis governs t h e Accusative case; Jayditya, in o r d e r t o account by Pnini's rule f o r t h e Locative case w h i c h also is f o u n d occasionally w i t h t h e same v e r b , makes t h e rule o p t i o n a l by supplying f o r it (by mandkapluti) anyatarasym f r o m P. 1.4, 44, and he t h e n declares P. 1.4, 47 t o be a vyavasthita-vibhs. Abbinivis

in some phrases governs only the Accusative, in others only the Locative. The interpretation of Pnini's rules is again dependent on and regulated by the actual usage of the language. 22. Anuktasamuccayrthas cakrah.31On P. I. 3, 93 luti ca klpah, Patajali shows that the ca ofthat rule is superfluous, because even without it the term syasanoh of the preceding rule would be valid in the rule under discussion ; and he adds the general remark that in like manner all the particles ca of the Astdhyyl may be dispensed with (Vol. I. p. 295). In other places meanings have been assigned by him to ca, which that particle does not convey ordinarily. Thus ca is taken in the sense of eva (avadhranrtha) in Vol. I. pp.381 (P. II. 1,17); 392 (P. I!. 1, 48), 406 (P. 11.1,72); in the sense of iti in Vol. I. p. 415 (P. II. 2,14). More strange even Patajali's proceeding must appear to us, when we see him refute certain criticisms of Ktyyana by means of the particle ca in P. II. 4, 9 and P. VI. 1, 90,a proceeding which, I may add, has been imitated by Jayditya on P. II. 3,16 and P. III. 1, 2. But there is one meaning which Patajali has never assigned to ca, and which in grammar has to my knowledge been assigned to it first by the authors of the Ksik-Vrtti,the meaning indicated in the above heading anuktasamuccayrthas cakrah. The word ca serves the purpose of adding to the words actually enumerated in a rule others that have not been mentioned ; or, in other words, Pnini indicates by the particle ca, that his rule applies to other words besides those actually mentioned by him. On P. V. 1, 7 khalayavamsatilavrsabrahmanas ca, Ktyyana has the note that ratha takes
the same suffix as the words enumerated (Vol. II. p. 339). According t o the Ksik-Vrtti, the particle ca at the end of the rule is by Pnini meant t o show that the suffix taught by the rule is added t o o t h e r words besides those actually enumerated, and the w o r d which
31

See, e.g. Ksik-Vrtti on P. II. 4,18; III. 1,126; III. 2, 30,138,188; III. 3, 119,122; IV. 1, 74, 96,123; IV. 2, 82;

IV. 4, 29; V. 1, 7; V. 4, 25,145; VIL 1, 48; VII. 2,16.

134 Franz Kielhorn

Pai ni had in view in thus employing ca is said t o have been ratha. O n P. III. 2, 30 ndlmustyos ca, Patajali adds t o the w o r d given by Pnini ghat!and others (Vol. II. p. 102); in the Ksik we are t o l d that these very words ghat etc., are suggested by the particle ca of Pnini's o w n rule, n P. IV. 4, 29 Pnini has parimukham co\ the corresponding rule of Candra's grammar is parer mukhaprsvt; according t o the Ksik, the particle ca of Pnini's rule shows that the suffix taught by the rule is added t o o t h e r words besides parimukha, and the w o r d which Pnini had in his mind, when thus employing ca, is the very pariprsva which is actually given by Candra. According t o the Ksik, t h e c o of P. V. 4,145 shows that one also says ahidat or ahidanta, msikadat or msikadanta etc. ;

ahi, msika etc., are the very words which the corresponding rule of Candra actually enumerates, in addition to the words agr, anta etc., which are enumerated by Pnini. Unluckily we do not possess yet a complete copy of Candra's grammar; but judging from what we do possess of it, I feel little hesitation in saying that, wherever the device of which I am speaking is resorted to in the Ksik-Vrtti, the words which Pnini is supposed to suggest by the employment of the particle ca, have invariably been taken by the compilers of the Ksik from the Vrttikas, or from the Mahbhsya, or from thegrammar of Candra. The compilers have invented nothing; not caring for, or having no notion of, the history of grammar, they have tried to show, how Pnini's own rules can be made to account for a number of words, for which Pnini's more immediate successors had given additional rules.32
32

See a paper on caand iti by Dr. Bhler in Wiener Zeitschrift fr die

Kunde des Morgenlandes, Vol. I. p. 13.

The Sceptics and Their Critics

Plate IV

A page from the text of Patajali's commentary on Panini's grammar, Mahbhsya, with Kaiyata's subcommentary, BhsyapradJpa, as it was reproduced in 1 874 by photolithography under the supervision of Goldstcker from an undated manuscript (Goldstcker 1874b). The ten lines in somewhat larger type in the middle of the page constitute the following discussion from the Introduction of the Mahbhsya a discussion that starts with the question of whether in the teaching of grammar, grammatical forms, or ungrammatical forms, or both should be taught:
kirn punaratra jyyah laghutvc chabdopadesah /laghyan chabdopadesah garyn apasabdopadesah ekaikasya sabdasya bahavo'pabhramsh/tad yath/gaur ity asya sabdasya gvgonigotgopotal i kdayo' pabh rarnsh / istnvkhynarp khalvapi bhavati/ athaitasmi chabdopadese sati kirn sabdnm pratipattau pratipadapthah kartavyah gaur asva pu ruso hasti sakunir mrgo brhmana ity evam dayah sabdh pathitavyh nety ha/ anabhyupya esa sabdnrn pratipattau pratipadapthah/evam hisryate/ brhaspatir indrya divyam varsasahasrarp pratipadoktnrp sabdnrn sabdapryanarp provca nrptam jagma/brhaspatis ca pravakt irpdras cdhyet divyam varsasahasram adhyayanaklah nacmtam jagmakirp punar adyatve yah sarvath cirarp jvati sa varsasatarp jvati caturbhis ca prakrair vidyopayukt bhavati//gamaklena svdhyyaklena pravacanaklena vyavahrakleneti/tatracsygamaklenaivyuh krtsnarp paryupayuktarp syttasmd anabhyupyah sabdnrn pratipattau pratipadapthah/katham tarhme sabdh pratipattavyh/ kirpcitsmnyavisesyaval laksanarp pravartyarp yenlpena yatnena mahato mahatah sabdaugn pratipadyeran/ kirn punas tad utsargpavdau kascid utsargah kartavyah kascid apavdah katharp jatyakah punar utsargah kartavyah katharp jtiyako'pavdah smnyenotsargah kartavyah tad yath karmany an/tasya visesenpavdah tad yath tonupasarge kah

" Now which of these is better? The teaching of grammatical forms, because of simpl city. The teaching of

grammatical forms is simpler; the teaching of ungrammatical forms is more complicated. (The reason is that) corresponding to each grammatical form there are many corrupt forms. For example, to the grammatical form gauh ' cow ' correspond the corrupt forms gvl, goni, got, gopotalik, and so forth. Moreover, (in the teaching of grammatical forms) the enumeration of the desired forms arises also. Now if grammatical forms are taught, must this be done by the recitation of each particular word for the understanding of grammatical forms? Must, for example, the grammatical forms 'cow,' ' horse,1 ' man.' 'elephant,' ' kite,' 'deer,' ' Brahman,' be recited? No, says the author [that is, Patajali], the recitation of each particular word is not a means for the understanding of grammatical forms, for there is the following tradition. Brhaspati addressed Indra during a thousand divine years, going over the grammatical forms by speaking each particular word, and still he did not attain the end. With Brhaspati as the instructor, Indra as the student, and a thousand divine years as the period of study, the end could not be attained, so what of the present day when he who lives a whole lifetime lives at most a hundred years? Knowledge is used in four ways: at the time of acquisition, at the time of study and rehearsal, at the time of teaching, and at the time of practice and application. Butsince in the present case the period of acquisition occupies the entire life-span, the recitation of each particular word is not a means for the understanding of grammatical forms. But then how are grammatical forms acquired? Some work containing the general and the particular has to be composed, so that men will acquire vast collections of forms with minimal effort. Of what would this consist? General rules and exceptions. Some general rule and some exception has to be formulated. What kind of general rule and what kind of exception? A general rule has to be formulated in all generality, for example, karmany an [Panini 3.2.1 : " aN is attached to the root when there is an object"]. An exception to

that should be formulated with full particulars, for example, ato'nupasarge koh [Panini 3.2.3: "after a root ending in not preceded by a preverb the affix is Ka" ]." The two lines at the top of the

page continuing with the four lines at the bottom constitute Kaiyata's subcommentary on this passage of the Mahbhsya. See also page 97 of this volume and Staal 1969, 501-502.

si

12
William Dwight Whitney (1827-1894)

From the very beginning of the study of the Indian grammarians there have been scholars who doubted that there ever was a language that conformed to the extremely numerous and complicated rules of Pnini and the other Sanskrit grammarians. This skepticism is not surprising for at least four reasons: (1) Sanskrit was regarded as a language which was no longer spoken; (2) literature in Sanskrit was only beginning to be knowntext editions were relatively few, manuscripts were only beginning to be discovered, and access to them was rare; (3) peopleespecially philologists, but also linguiststend to confound language with a body of literature; and (4) European scholars were not used to such painstaking detail in grammatical description as was displayed by grammatical rules in the Indian tradition. Such specificity and complexity was generally associated only with priestly cunning (there is, in fact, a close relationship between Indian ritualistic studies and the Sanskrit grammarians; see below pages 435-469). Even at present, there are people who are distrustful of a grammar when it looks complicated or contains very many rules. Colebrooke had already referred to such skepticism and brushed it aside (see page 36 of this volume). Bopp during the Romantic period, however, was of the opinion that the Indian grammarians had set up forms to serve as roots, which were in fact merely postulated "underlying" forms, and not " r e a l " (page 53). He was criticized, especially by Lassen (page 53), but also by philologists like Westergaard (who emphasized that too little was known of Sanskrit literature to warrant such a judgment), and, at least implicitly, by linguists like Humboldt (who regarded such postulated roots as real even if they did not come to the surface in a given language; see page 63). We have also seen that Bhandarkar in his lectures not only showed that Sanskrit was a spoken language, but also determined historically which stage of the language was in conformity with the grammatical rules given by specific grammariansin particular, by Pnini, Ktyyana, and Patajali page 91). Kielhorn and Liebich contributed further to this kind of research. Though the Sanskrit grammarians had not become widely known and their study remained a specialty even among Sanskritists, Western scholars had become familiar with at least some of their techniques. Apart from guna and vrddhi, the distinction between roots and suffixes, and similar distinctions of a very fundamental nature, most Sanskrit scholars were acquainted with such specimens of the Sanskrit grammarians as were quoted by Max Mller in his excellent Sanskrit Grammar for Beginners of 1866 (which is much indebted to and collated throughout with Pnini's grammar). In the introduction to this grammar Mller gave the following example of the complexity of Pnini's grammar (quoted from the second edition of 1870, ix-x): By this process [i.e., by collating the whole of the grammar with Pnini] which I have adopted, I believe that on many points a more settled and authoritative character has been imparted to the grammar of Sanskrit than it possessed before; but I do by no means pretend to have arrived on all points at a clear and definite view of the meaning of Pnini and his successors. The grammatical system of Hindu grammarians is so peculiar, that rules which we should group together, are scattered about in different parts of their manuals. We may have the general rule in the last, and the exceptions in the first book, and even then we are by no means certain that exceptions to these exceptions may not occur some-

139 William Dwight Whitney

where else. I shall give but one instance. There is a root jagr, which forms its Aorist by adding sam, Jh, It. Here t h e simplest rule would be that final r before sam becomes r (Pan. VI. 1, 77). This, however, is prevented by another rule which requires that final r should take Guna before isam (Pan. VII. 3, 84). This would give us ajgarisam. But now comes another general rule (Pan. VII, 2,1) which prescribes Vrddhi of final vowels before sam, i.e. ajgrisam. Against this change, however, a new rule is cited (Pan. VIL, 3, 85), and this secures for jgr a special exception from V r d d h i , and leaves its base again as jgar. As soon as the base has been changed t o jgar, it falls under a new rule (Pan. VII. 2, 3), and is forced t o take V r d d h i , until this rule is again nullified by Pan. VII. 2, 4, which does not allow Vrddhi in an Aorist that takes intermediate /', like ajgrisam. There is an exception, however, t o this rule also, for bases w i t h short a, beginning and ending w i t h a consonant, may optionally take Vrddhi (Pan. VII. 2, 7). This option is afterwards restricted, and roots w i t h short a, beginning w i t h a consonant and ending in r, like jgar, have no option left, but are restricted afresh t o Vrddhi (Pan. VII. 2, 2). However, even this is not yet the final result. O u r base jgar is after all not t o take V r d d h i , and hence a new special rule (Pan. VII. 2, 5) settles t h e point by granting to jgr a special exception from V r d d h i , and thereby establishing its Guna. N o wonder that these manifold changes and chances in the formation of t h e First Aorist of jgr should have inspired a grammarian, w h o celebrates them in t h e following couplet: guno vrddhir guno vrddhih pratisedho vikalpanam/ punar vrddhir nisedho'to y an prvh prptayo nava I /

'Guna, Vrddhi, Guna, Vrddhi, prohibition, option, again Vrddhi and then exception, these with the change of r into a semivowel in the first instance, are the nine results.'
Though much new information was becoming accessible, this was not always accompanied by an adequate understanding of the nature of the work of the Sanskrit grammarians. Actually, not only had the skeptical view not died, its extreme form was only beginning to be explicitly formulatedand passionately defended. In 1874, after being dormant for perhaps half a century, when Bopp had defended a much weaker and much more balanced version, the view was put forward by Theodor Benfrey in the Introduction to his Vedic grammar "dass uns von den Indern, diesen grssten Grammatikern der Welt, auf der einen Seite (in den Veden) die wunderbarste Sprache ohne eine sich auf sie sttzende Grammatik hinterlassen worden sei, auf der andern dagegen die wunderbarste Grammatik ohne die Sprache, auf welche sie gesttzt ist. Wir haben daher kein usseres Hilfsmittel, wodurch wir die Richtigkeit dieser letzteren zu prfen, ja auch nur zu controliren vermchten " (quoted in Liebich 1891,44). However, the main proponent of the view that there was a "grammarians' Sanskrit/1 which was "a thing of grammatical rule merely, having never had any real existence as a language" was William Dwight Whitney, the first important American Sanskritist. William Dwight Whitney (1827-1894) studied Sanskrit first at Yale with Edward Elbridge Salisbury, who himself was interested mainly in Arabic but who had studied some Sanskrit with Bopp in Berlin and later with Lassen in Bonn. At an early date Whitney also made use of the second edition of Bopp's Sanskrit grammar, which his elder brother had brought with him from Europe in

140 William Dwight Whitney

1847. Later W h i t n e y studied in Germany w i t h W e b e r , Bopp, and Roth (cf. T. D. Seymour and C. R. Lanman in Sebeok 1966, I, 399-439). He became widely k n o w n also as a general linguist. As a Sanskritist his fame was mainly due t o his w o r k on the Atharvaveda and t o his Sanskrit Grammar, which was first published in Leipzig in 1879 together w i t h a German translation. This grammar was based upon very extensive textual materials, and there is nothing in it that cannot be substantiated by examples f r o m the extant and available corpus of Sanskrit texts. It was in fact mainly praised because of this new, so-called " s t a t i s t i c a l " approach (see page 204) of this volume). As we shall see, W h i t n e y held strong views w i t h regard t o the " r e a l i t y " of roots given by the Sanskrit grammarians. He w e n t even f u r t h e r in his grammar, in that he never set up an underlying r o o t unless it was somewhere (phonetically) manifested in the language (for a fuller discussion of this and o t h e r characteristics of W h i t n e y ' s grammar see McCawley 1967). Needless t o say, W h i t n e y ' s grammar did not rely on Pnini's grammar, which the Preface attacks for its " highly artful and difficult f o r m of about four thousand algebraic-formula-like rules in the statement and arrangement of which brevity alone is had in view, at the cost of distinctness and unambiguousness" (xi). A p a r t f r o m the fact that W h i t n e y appears t o have had a low opinion of Indian scholarship in general (for example, see page 151, note 7), his misinterpretations of the Sanskrit grammarians may be traced back not only t o the general skeptical t r e n d , but also t o his o w n linguistic convictions and philological activities. First, he was convinced that a linguistic description is nothing but an inventory of elements. Chomsky has quoted one of W h i t n e y ' s statements as expressing most clearly perhaps this view of linguistics: " language in the concrete sense . . . is . . . the sum of words and phrases by which any man expresses his t h o u g h t " ( W h i t n e y 1874, 372 in Chomsky 1964, 22). His Sanskrit grammar is a perfect illustration of this approach. Second, W h i t n e y does not appear t o have done any w o r k himself on the Sanskrit grammarians, though he had w o r k e d widely in the Prtiskhya literature and published and translated the Atharvaveda-prtiskhya (in 1862) and the Taittiriya-prtiskhya (in 1871). This does not, however, conflict w i t h his general linguistic o u t l o o k ; on the contrary, it confirms it, for there is a characteristic difference between the Prtiskhya literature and the t r a d i t i o n of the Sanskrit grammarians. Each prtiskhya specifies for a given branch of the Veda how the samhit (or continuous t e x t , in which the sandhi rules, etc. are applied) must be derived f r o m the padaptha (or w o r d - f o r - w o r d analysis, in which the sandhi is dissolved). Each prtiskhya is, therefore, by its very nature confined t o the analysis of a specific corpus of u t t e r ances. The prtiskhyas do not, in general, give linguistic rules, though it often happens that the rules which they provide for the derivation of the samhit f r o m the padaptha coincide w i t h linguistic rules (cf. Staal 1967, ch. II; 1970a). The approach of the Prtiskhya literature was congenial t o W h i t n e y because it accorded w i t h his general o u t l o o k in linguistics. He checked the completeness of the statements of the prtiskhyas w i t h the corpus which they p u r p o r t e d t o describe, and often found t o his satisfaction complete agreement. For example, commenting on the passages of the Taittiriya-prtiskhya which describes the conversion from dental n in the padaptha t o r e t r o flex n in the samhit, he n o t e d : " ! have not discovered in the Samhit any case of a lingual nasal arising in the conversion of padatext into

141 William Dwight Whitney

samhita which is not duly provided for in this chapter" (Whitney 1871,180). And similarly: "The Prtiskhya's enumeration ofthe cases of occurrence ofthe lingual nasal \st so far as I have been able to determine, complete" (Whitney 1871, 281). But this does not mean that Whitney understood the linguistic position ofthe Sanskrit grammarians. Hence the inadequacy of such evaluations as the following, voiced by T. D. Seymour (in Sebeok, 1966, I, 416): "That he [i.e., Whitney] did not discredit and slight the old Hindu grammarians because of any lack of acquaintance with them is shown by his own work and publications in that field. He published not only the Atharva-Veda-Prtiskhya(text, translation, and notes, in 1862), but also a similar edition ofthe Taittirya-Prtiskhya, with its commentary, the Tribhsyaratna, in 1871." Whitney's influential views on the Sanskrit grammarians were expressed in an article in the American Journal of'Philology (5,1884, 279-297) which is here reproduced.

A.The Study of Hindu Grammar and the Study of Sanskrit (1884)

William Dwight Whitney

To the beginning study of Sanskrit it was an immense advantage that there existed a Hindu science of grammar, and of so high a character. To realize how great the advantage, one has only to compare the case of languages destitute of itas for instance the Zend. It is a science of ancient date, and has even exercised a shaping influence on the language in which all or nearly all the classical literature has been produced. It was an outcome of the same general spirit which is seen in the so careful textual preservation and tradition of the ancient sacred literature of India; and there is doubtless a historical connection between the one and the other; though of just what nature is as yet unclear. The character of the Hindu grammatical science was, as is usual in such cases, determined by the character of the language which was its subject. The Sanskrit is above all things an analyzable language, one admitting of the easy and distinct separation of ending from stem, and of derivative suffix from primitive word, back to the ultimate attainable elements, the so-called roots. Accordingly, in its perfected form (for all the preparatory stages are unknown to us), the Hindu grammar offers us an established body of roots, with rules for their conversion into stems and for the inflection of the latter, and also for the accompanying phonetic changesthis last involving and resting upon a phonetic science of extraordinary merit, which has called forth the highest admiration of modern scholars; nothing at all approaching it has been produced by any ancient people; it has served as the foundation in no small degree of our own phonetics: even as our science of grammar and of language has borrowed much from India. The treatment of syntax is markedly inferiorthough, after all, hardly more than in a measure to correspond with the inferiority of the Sanskrit sentence in point of structure, as compared with the Latin and the Greek. Into any more detailed description it is not necessary to our present purpose to enter; and the matter is one pretty well understood by the students of Indo-European language. It is generally well known also that the Hindu science, after a however long history of elaboration, became fixed for all future time in the system of a single grammarian, named Pnini (believed, though on grounds far from convincing, to have lived two or three centuries before the Christian era). Pnini's work has been commented without end, corrected in minor points, condensed, re-cast in arrangement, but never rebelled against or superseded ; and it is still the authoritative standard of good Sanskrit. Its form of presentation is of the strangest : a miracle of ingenuity, but of perverse and wasted ingenuity. The only object aimed at in it is brevity, at the sacrifice of everything elseof order, of clearness, of even intelligibility except by the aid of keys and commentaries and lists of words, which then are furnished in profusion. To determine a grammatical point out of it is something like constructing a passage of text out of an index verborum : if you are sure that you have gathered up every word that belongs in the passage, and have put them all in the right order, you have got the right reading; but only then. If you have mastered Pnini sufficiently to bring to bear upon the given point every rule that relates to it, and in due succession, you have settled the case; but that is no easy task. For example, it takes nine mutually limitative rules, from all parts of the text-book, to determine whether a certain aorist shall be ajagarisam or ajgrisam (the case is reported in the preface to Mller's grammar): there is lacking only a tenth rule, to tell us that the whole word is a false and never-used formation. Since

143 Hindu Grammar and Sanskrit

t h e r e is nothing t o show how far t h e application of a rule reaches, t h e r e are provided treatises of laws of i n t e r p r e t a t i o n t o be applied t o t h e m ; but t h e r e is a residual rule underlying and d e t e r m i n i n g t h e w h o l e : that both the grammar and t h e laws of i n t e r p r e t a t i o n must be so construed as t o yield good and acceptable forms, and not otherwiseand this implies (if that w e r e needed) a condemnat i o n of the w h o l e mode of presentation of the system as a failure. Theoretically, all that is prescribed and allowed by Pnini and his accepted commentators is Sanskrit, and nothing else is e n t i t l e d t o t h e name. The young pandit, t h e n , is expected t o master t h e system and t o govern his Sanskrit speech and w r i t i n g by it. This he does, w i t h immense pains and labor, then naturally valuing the acquisition in part according t o what it has cost him. The same course was f o l l o w e d by those European scholars w h o had t o make themselves the pupils of Hindu teachers, in acquiring Sanskrit for the benefit of Europe; and (as was said above) they did s o t o t h e i r very great advantage. Equally as a matter of course, t h e same must still be done by any one w h o studies in India, w h o has t o deal w i t h the native scholars, w i n t h e i r confidence and respect and gain t h e i r aid ; they must be met upon t h e i r o w n g r o u n d . But it is a question, and one of no slight practical importance, how far W e s t e r n scholars in general are t o be held t o this method : w h e t h e r Pnini is for us also the law of Sanskrit usage; w h e t h e r we are t o study t h e native Hindu grammar in o r d e r t o learn Sanskrit. There w o u l d be less reason for asking this question, if the native grammar w e r e really the i n s t r u m e n t a l i t y by which t h e conserving t r a d i t i o n of the old language had been carried o n . But that is a t h i n g both in itself impossible and proved by the facts of t h e case t o be u n t r u e . N o one ever mastered a list of roots w i t h rules for t h e i r extension and inflection, and then w e n t t o w o r k t o construct texts upon that basis. Rather, t h e transmission of Sans k r i t has been like the transmission of any highly cultivated language, only w i t h difference of degree. The learner has his models which he imitates; he makes his speech after the example o f t h a t of his teacher, only under the constant g o v e r n m e n t of grammatical rule, enforced by the r e q u i r e m e n t t o justify o u t of t h e grammar any w o r d o r f o r m as t o which a question is raised. Thus t h e language has moved on by its o w n inertia, only falling, w i t h f u r t h e r removal f r o m its natural vernacular basis, more and more passively and mechanically into the hands of t h e grammarians. A l l this is like the propagation of literary English or German ; only t h a t here t h e r e is much more of a vernacular usage t h a t shows itself able t o o v e r r i d e and modify t h e rules of grammar. It is yet more closely like t h e propagation of L a t i n ; only that here t h e i m i t a t i o n of previous usage is frankly acknowledged as t h e guide, t h e r e being no iron system of grammar t o assume t o take its place. That such has really been t h e history of t h e later o r classical Sanskrit is sufficiently shown by t h e facts. There is no absolute coincidence between it and t h e language w h i c h Pnini teaches. The f o r m e r , indeed, includes l i t t l e t h a t t h e grammarians f o r b i d ; but, on t h e o t h e r hand, it lacks a great deal that t h e y allow or prescribe. The d ifference between t h e t w o is so great t h a t Benfey, a scholar deeply versed in t h e Hindu science, calls it a grammar w i t h o u t a corresponding language, as he calls t h e pre-classical dialects a language w i t h o u t a grammar. 1 If such a statement can be made
1 Einleitung in die Grammatik der vedischen Sprache, 1874, pp. 3, 4.

144 William Dwight Whitney

w i t h any reason, it w o u l d appear that t h e r e is t o be assumed, as the subject of Hindu grammatical science, a peculiar dialect of Sanskrit, which we may call the grammarians' Sanskrit, different both f r o m the pre-classical dialects and f r o m the classical, and standing either between t h e m or beside them in the general hist o r y of Indian language. A n d it becomes a matter of importance t o us t o ascertain what this grammarians' Sanskrit is, how it stands related t o the o t h e r varieties of Sanskrit, and w h e t h e r it is entitled t o be the leading object of our Sanskrit study. Such questions must be settled by a comparison of the dialect referred t o w i t h the o t h e r dialects, and of them w i t h one another. A n d it w i l l be found, upon such comparison, that the earlier and later forms of the Vedic dialect, the dialects of the Brhmanas and Stras, and the classical Sanskrit, stand in a filial relation, each t o its predecessor, are nearly o r quite successive forms of the same language; w h i l e the grammarians' Sanskrit, as distinguished f r o m t h e m , is a thing of grammatical rule merely, having never had any real existence as a language, and being on the whole u n k n o w n in practice t o even the most modern pandits. The main thing which makes of the grammarians' Sanskrit a special and peculiar language is its list of roots. O f these there are reported t o us about t w o thousand, w i t h no intimation of any difference in character among t h e m , or warning that a part of them may and that another part may not be drawn upon for forms t o be actually used ; all stand upon the same plane. But more than halfactually more than halfof them never have been met w i t h , and never w i l l be met w i t h , in the Sanskrit l i t e r a t u r e of any age. W h e n this fact began t o come t o I ight, it was long fondly hoped, or believed, that the missing elements w o u l d yet t u r n up in some corner of the literature not h i t h e r t o ransacked; but all expectation o f t h a t has now been abandoned. One or another does appear f r o m t i m e t o t i m e ; but what are they among so many?The last notable case was that of the r o o t stigh, discovered in the MitryaniSamhit, a t e x t of the Brhmana period ; but the new roots found in such texts are apt t o t u r n out wanting in the lists of the grammarians. Beyond all question, a certain number of cases are t o be allowed for, of real roots, proved such by the occurrence of t h e i r evident cognates in o t h e r related languages, and chancing not t o appear in the k n o w n l i t e r a t u r e ; but they can go only a very small way indeed t o w a r d accounting for the eleven hundred unauthenticated roots. Others may have been assumed as underlying certain derivatives or bodies of derivativeswithin due limits, a perfectly legitimate proceeding; but the cases thus explainable do not prove t o be numerous. There remain then the great mass, whose presence in the lists no ingenuity has yet proved sufficient t o account for. A n d in no small part, they bear t h e i r falsity and artificiality on the surface, in t h e i r phonetic form and in the meanings ascribed t o them ; we can confidently say that the Sanskrit language, known t o us t h r o u g h a long period of development, neither had nor could have any such roots. H o w the grammarians came t o concoct t h e i r list, rejected in practice by themselves and t h e i r o w n pupils, is h i t h e r t o an unexplained mystery. N o special student of the native grammar, t o my knowledge, has attempted t o cast any light upon i t ; and it was left for Dr. Edgren, no partisan of the grammarians, t o group and set f o r t h the facts for the first t i m e , in the Journal of the American Oriental Society (Vol. X I , 1882but the article printed in 1879 pp. 1-55), adding a list of the real roots, w i t h brief particulars as

145 Hindu Grammar and Sanskrit

t o t h e i r occurrence. 2 It is quite clear, w i t h reference t o this fundamental and most i m p o r t a n t item, of what character the grarmmarians' Sanskrit is. The real Sanskrit of the latest period is, as concerns its roots, a t r u e successor t o that of the earliest period, and t h r o u g h the known intermediates ; it has lost some of the roots of its predecessors, as each of these some belonging t o its o w n predecessors or predecessor; it has, also like these, w o n a certain number not earlier found : both in such measure as was t o be expected. As for the rest of the asserted roots of the grammar, t o account for them is not a matter that concerns at all the Sanskrit language and its history; it only concerns the history of the Hindu science of grammar. That, t o o , has come t o be p r e t t y generally acknowledged. 3 Every one w h o knows anything of the history of Indo-European etymology knows how much mischief the grammarians' list of roots w r o u g h t i n t h e hands of the earlier more incautious and credulous students of Sanskrit : how many false and worthless derivations were founded upon t h e m . That sort of w o r k , indeed, is not yet entirely a thing of the past; still, it has come t o be well understood by most scholars that no alleged Sanskrit r o o t can be accepted as real unless it is supported by such a use in the literary records of the language as authenticates i t for t h e r e are such things in the later language as artificial occurrences, forms made for once or t w i c e f r o m roots taken out of t h e grammarians' list, by a natural license, which one is only surprised not t o see oftener availed of (there are hardly more than a dozen or t w o of such cases quotable) : that they appear so seldom is the best evidence of the fact already pointed out above, that the grammar had, after all, only a superficial and negative influence upon the real t r a d i t i o n of the language. It thus appears that a Hindu grammarian's statement as t o the fundamental elements of his language is w i t h o u t a u t h o r i t y until tested by the actual facts of the language, as represented by the Sanskrit literature. But the principle w o n here is likely t o prove of universal application ; for we have no reason t o expect t o find the grammarians absolutely t r u s t w o r t h y in other departments of t h e i r w o r k , when they have failed so signally in one; t h e r e can be nothing in t h e i r system that w i l l not require t o be tested by the recorded facts of the language, in o r d e r t o determine its t r u e value. H o w this is, we w i l l proceed t o ascertain by examining a few examples. In the older language, but not in the oldest (for it is wanting in the Veda), t h e r e is formed a periphrastic f u t u r e tense active by compounding a nomen agentis w i t h an auxiliary, the present tense of the verb as ' be' : thus, data 'smi (literally dotor sum) ' I w i l l give,' etc. It is quite infrequent as compared w i t h the o t h e r f u t u r e , yet common enough t o require t o be regarded as a part of the general Sanskrit verb-system. To this active tense the grammarians give a corresponding middle, although the auxiliary in its independent use has no middle inflection; it is made w i t h endings modified so
2 I have myself now in press a much fuller account of the quotable roots of the language, with all their quotable tense-stems and primary derivatives everything accompanied by a definition of the period of its known occurrence in the history of the language. 3

Not, indeed, universally; one may find among the selected verbs that are conjugated in full at the end of F. M. Mller's Sanskrit Grammar, no very small number of those that are utterly unknown to Sanskrit usage, ancient or modern,

146 William Dwight Whitney

as t o stand in the usual relation of middle endings t o active, and f u r t h e r w i t h conversion in 1st sing, of the radical s t o ha very anomalous substitution, of which t h e r e is not, I believe, another example in the language. N o w what support has this middle tense in actual use? O n l y t h i s : t h a t i n t h e Brh manas occur four sporadic instances of attempts t o make by analogy middle forms for this tense (they are all reported in my Sanskrit Grammar, 947; furt h e r search has brought t o light no additional examples): t w o of them are 1st sing., one having the f o r m se for the auxiliary, the other he, as taught in the grammar; and in the whole later literat u r e , epic and classical, I find record of the occurrence of only one f u r t h e r case, darsayitahe (in Nais. V 71 .)! 4 Here also, the classical dialect is the t r u e continuator of the pre-classical ; it is only in the grammarians' Sanskrit that every verb conjugated in the middle voice has also a middle periphrastic f u t u r e . There is another and much more i m p o r t a n t part of verbal inflectionnamely, the whole aorist-system, in all its varietyas t o which the statements of the grammarians are t o be received w i t h especial distrust, for the reason that in the classical language the aorist is a decadent f o r m a t i o n . In the older dialects, down t o the last Stra, and t h r o u g h the entire list of early and genuine Upanisads, the aorist has its own special office, that of designating the immediate past, and is always t o be found where such designat i o n is called f o r ; later, even in the epos, it is only another p r e t e r i t , equivalent in use t o imperfect and perfect, and hence of no value, and subsisting only in occasional use, mainly as a survival f r o m an earlier condition of the language. Thus, for example, of the first kind of aorist, the root-aorist, forms are made in pre-classical Sanskrit f r o m about 120 r o o t s ; of these, 15 make forms in the later language also, mostly sporadically (only go, da, dh, pa, sth, bh less infrequently); and 8 more in the later language only, all in an occurrence or t w o (all but one, in active precative forms, as t o which see below). Again, of the fifth aorist-form, the /s-aorist (rather the most frequent of all), forms are made in the older language f r o m 140 roots, and later f r o m only 18 of these (and sporadically, except in the case of grah, vad, vadh, vid), w i t h a dozen more in the later language exclusively, all sporadic except sank (which is not a Vedic root). Once more, as regards the t h i r d or reduplicated aorist, the p r o p o r t i o n is slightly different, because of the association o f t h a t aorist w i t h the causative conjugation, and t h e frequency of the latter in use; here, against about 110 roots quotable f r o m the earlier language, 16 of them also in the later, t h e r e are about 30found in the later alone (nearly all of them only sporadically, and none w i t h any frequency). And the case is not otherwise w i t h the remaining forms. The facts being such, it is easily seen that general statements made by the grammarians as t o the range of occurrence of each f o r m , and as t o the occurrence of one f o r m in the active and a certain other one in the middle f r o m a given root, must be of very doubtful a u t h o r i t y ; in fact, as re4 Here, as elsewhere below, my authority for the later literature is chiefly the Petersburg Lexicon (the whole older literature I have examinedfor myself), and my statements are, of course, always open to modification by the results of further researches. But all the best and most

genuine part of the literature has been carefully and thoroughly excerpted for the Lexicon ; and for the Mahbhrata we have now the explicit statements of Holtzmann, in his Grammatisches aus dem Mahbhrata, Leipzig, 1884.

147 Hindu Grammar and Sanskrit

gards the latter point, they are the more suspicious as lacking any tolerable measure of support f r o m the facts of the older language, But t h e r e are much greater weaknesses than these in the grammarians' t r e a t m e n t of the aorist. Let us first t u r n our attention t o the aorist optative, the socalled precative (or benedictive). This formation is by the native grammarians not recognized as belonging t o the aorist at allnot even so far as t o be put next the aorist in t h e i r general scheme of conjugation ; they suffer the future-systems t o intervene between the t w o . This is in t h e m fairly excusable as concerns the precative active, since it is the optative of the root-aorist, and so has an aspect as if it might come independently f r o m the root d i r e c t l y ; nor, indeed, can we much blame them for overlooking the relation of t h e i r precative m i d d l e t o the sibilant or sigmatic aorist, considering that they ignore tense-systems and modes ; but that t h e i r European imitators, d o w n t o the very latest, should commit the same oversight is a different matter. The contrast, now, between the grammarians' dialect and the real Sanskrit is most marked as regards the middle forms. According t o the grammar, the precative middle is t o be made f r o m every root, and even for its secondary conjugations, the causative etc. It has t w o alternative modes of f o r m a t i o n , which we see t o correspond t o t w o of the forms of the sibilant aorist: the s-aorist, namely, and the /s-aorist. O f course, a complete inflection is allowed it. To justify all this, now, I am able t o point t o only asingle occurrence of a middle precative in the whole later literature, including the epics : that is rlrisJsta, in the Bhgavata-Purna(lll 9, 24), a t e x t n o t a b l e f o r its artificial imitation of ancient forms (the same w o r d occurs also in the Rg-Veda); it is made, as will be noticed, f r o m a reduplicated aorist stem, and so is unauthorized by grammatical rule. A single example in a whole literature, and that afalse one! In the preclassical literature also, middle precative forms are made hardly more than sporadically, o r f r o m less than 40 roots in all (so far as I have found) ; those belonging t o the s and is aorists are, indeed, among the most numerous (14 each), but those of the root-aorist do not fall short of them (also 14 roots), and t h e r e are examples f r o m t h r e e of the o t h e r four aorists. Except asingle 3d pi. (in roto, instead of Iran), only the t h r e e singular persons and the 1st pi. are quotable, and forms occur w i t h o u t as well as w i t h the adscititious s between mode-sign and personal ending which is the special characteristic of a precative as distinguished f r o m asimply optative f o r m . Here, again, we have a f o r m a t i o n sporadic in the early language and really extinct in the later, but erected by the grammarians into a regular part of every verb-system. W i t h the precative active the case is somewhat different. This also, indeed, is rare even t o sporadicalness, being, so far as I know, made f r o m only about 60 roots in the whole languageand of these, only half can show forms containing the t r u e precative s. But it is not quite limited t o the pre-classical dialects; it is made also later f r o m 15 roots, 9 of which are additional to those which make a precative in the older language. Being in origin an optative of the root-aorist, it comes, as we may suppose, t o seem t o be a formation f r o m the root directly, and so t o be extended beyond the limits of the aorist; f r o m a clear majority (about three-fifths) of all the roots that make it, it has no other aorist forms by its side. A n d this begins even in the earliest period ( w i t h half-a-dozen roots in t h e Veda, and t o w a r d a score besides in the Brhmana and Stra); although t h e r e the precative more usually makes a

148 William Dwight Whitney

part of a general aorist-formation : for instance, and especially, f r o m the r o o t bh, whose precative forms are oftener met w i t h than those of all o t h e r roots together, and which is the only r o o t f r o m which more than t w o real precative persons are quotable. H o w rare it is even in the epos is shown by the fact that Hoftzmann 5 is able t o quote only six forms (and one of these d o u b t f u l , and another a false formation) f r o m the whole Mahbhrata, one of them occurring t w i c e ; while the first book of the Rmyana (about 4500 lines) has the single bhyt. Since it is not quite ext i n c t in the classical period, the Hindu grammarians could not, perhaps, well help teaching its f o r m a t i o n ; and, considering the the general absence of perspective f r o m t h e i r w o r k , we should hardly expect t h e m t o explain that it was the rare survival of an anciently little-used formation ; but we have here another striking example of the great discordance between the real Sanskrit and the grammarians' dialect, and of the insufficiency of the information respecting the f o r m e r obtainable f r o m the rules for the latter. Again, the reduplicated or t h i r d f o r m of aorist, though it has become attached t o the causative secondary conjugation (by a process in the Veda not yet complete), as the regular aorist o f t h a t conjugation, is not made f r o m the derivative causative stem, but comes f r o m the r o o t itself, not less directly than do the o t h e r aorist-formationsexcept in the few cases w h e r e the causative stem contains a padded t o d : thus, atisth pat f r o m stem sthpaya, r o o t sth. Perhaps misled by this exception, however, the grammarians teach the f o r m a t i o n of the reduplicated aorist f r o m the causative stem, t h r o u g h the intermediate process of converting the stem back t o the root, by s t r i k i n g off its conjugation-sign and reducing its strengthened vowel t o the simpler r o o t - f o r m . That is t o say, we are t o make, for example, obbhuvat f r o m the stem bhvaya, by cutting off aya and reducing the remainder bhv or bhu t o bh, instead of making it f r o m bh d i r e c t l y ! That is a curious etymological process; quite aside-piece t o deriving varlyas and varisthafrom uru, and the like, as the Hindu grammarians and t h e i r European copyists w o u l d likewise have us do. There is one point where the matter is brought t o a crucial t e s t : namely, in roots that end in u or ; where, if t h e v o w e l on which the reduplication is formed is an u-vowel, the reduplication-vowel should be of the same character; but, in any o t h e r case, an /-vowel. Thus, in the example already taken, bhvaya ought t o make ablbhavat, just as it makes bibhvayisati in the case of a real derivation f r o m the causative stem ; and such forms as ablbhavat are, in fact, in a great number of cases either prescribed or allowed by the grammarians ; but I am not aware of t h e i r having been ever met w i t h in use, earlier or later, w i t h t h e single exception of apiplavam, occurring once in the Satapatha-Brhmana (VI i, 1, 8). Again, the grammarians give a peculiar and problematic rule for an alternative f o r m a t i o n of certain passive tenses (aorist and futures) f r o m the special 3d sing. aor. pass. ; they allow it in the case of all roots ending in vowels, and of grah, drs, han. Thus, for example, f r o m the r o o t da are allowed adyisi, dyisyate, dyit, beside adisi, dsyate, data. W h a t all this means is quite obscure, since t h e r e is no usage, either early or (ate, t o cast light upon it. The Rg-Veda has once (1147, 5) dhyis, f r o m r o o t dh; but this, being active, is rather a hindrance than a help. The Jaim. Brhmana has once (I 321) khyyisyante; but this appears t o be a f o r m anal5

In his work already cited, at p. 32.

149 Hindu Grammar and Sanskrit

ogous w i t h hvayisyate etc., and so proves nothing. The Bhag. Purna has once (VII113, 36) tyit, which the Petersburg Lexicon refers t o r o o t tan ; but if t h e r e is such a t h i n g as the secondary r o o t toy, as claimed by the grammarians, it perhaps belongs rather there. A n d t h e r e remains, so far as I can discover, only asthayisi (Dasak. [ W i l s o n ] , p. 117, 1.6) and anyisata (Ind. Sprche, 6187, f r o m the Kuvalaynanda); and these are with.great probability t o be regarded as artificial forms, made because the grammar declares t h e m correct. It seems not unlikely that some misapprehension o r blunder lies at the foundation of these rules of the grammar; at any rate, the f o r m a t i o n is only grammarians' Sanskrit, and not even pandits' ; and it should never be o b t r u d e d upon the attention of beginners in the language. Again, the secondary ending dhvam of 2d pi. mid. sometimes has t o take the f o r m dhvam. In accordance w i t h the general euphonic usages of the language, this should be whenever in the present condition of Sanskrit there has been lost before the ending a lingual sibilant; t h u s : we have anedhvam f r o m anes -f- dhvam, and apavidhvam f r o m apavis + dhvam ; we should f u r t h e r have in the precative bhavisldhvam f r o m bhavisl-s-dhvam, if the f o r m ever occurred, as, unfortunately, it does not. A n d , so far as I know, t h e r e is not t o be found, either in the earlier language or the later (and as t o the f o r m e r I can speak w i t h authority), a single instance of dhvam in any o t h e r situationthe test-cases, however, being far f r o m numerous. But the Hindu grammarians, if they are reported rightly by t h e i r European pupils (which in this instance is hard t o believe), give rules as t o the change of the ending upon this basis only for the s-aorist; for the /s-aorist and its optative (the precative), they make the choice between dhvam and dhvam t o depend upon w h e t h e r the / is or is not " preceded by a semi-

vowel or h:" that is, apavis + dhvam gives apavidhvam, but ajanis + dhvam gives ajanidhvam, and so likewise we should have janisldhvam. It would be curious to know what ground the grammarians imagined themselves to have for laying down such a rule as this, wherein there is a total absence of discoverable connection between cause and effect; and it happens that all the quotable examplesajanidhvam, artidhvam, aindhidhvam, vepidhvamare opposed to their rule, but accordant with reason. What is yet worse, however, is that the grammar extends the same conversion of dh to dh, under the same restrictions, to the primary ending dhve of the perfect likewise, with which it has nothing whatever to doteaching us that, for instance, cakr and tustu + dhve make necessarily cakrdhve and tustudhve, and that dadhr-i + dhve makes either dadhridh've or dadhri'dhve, while tutud-i + dhve makes only tutudidhvel This appears to me the most striking case of downright unintelligent blundering on the part of the native grammarians that has come to notice; if there is any way of relieving them of the reproach of it, their partisans ought to cast about at once to find it. A single further matter of prime importance may be here referred to, in illustration of the character of the Hindu grammarians as classifiers and presenters of the facts of their language. By reason of the extreme freedom and wonderful regularity of word-composition in Sanskrit, the grammarians were led to make a classification of compounds in a manner that brought true enlightenment to European scholars; and the classification has been largely adopted as a part of modern philological science, along even with its bizarre terminology. Nothing could be more accurate and happier than the distinction of dependent, descriptive,

150 William Dwight Whitney

possessive, and copulative compounds ; only t h e i r t i t l e s ' his man ' (tatpurusa), 'act-sustaining' (? karmadhraya), ' much-rice' (bahuvrlhi), and ' c o u p l e ' (dvandva), respectivelycan hardly claim t o be w o r t h preserving. But it is t h e characteristic of Hindu science generally not t o be able t o stop when it has done enough ; and so the grammarians have given us, on the same plane of division w i t h these four capital classes, t w o more, which they call dvigu ( ' t w o c o w ' ) and avyayibhva (' indeclinable-becoming'); and these have no raison d'tre, but are collections of special cases belonging t o some of the o t h e r classes, and so heterogeneous that t h e i r limits are hardly capable of definition : the dWgu-class are secondary adjective compounds, but sometimes, like o t h e r adjectives, used as nouns ; and an avyoylbhva is always the adverbially-used accusative neuter of an adjective compound. It w o u l d be a real service on the part of some scholar, versed in the Hindu science, t o draw out a full account of the so-called dwgu-class and its boundaries, and t o show if possible how the grammarians w e r e misled into establishing it. But it w i l l probably be long before these t w o false classes cease t o haunt the concluding chapters of Sanskrit grammars, o r w r i t e r s on language t o talk of the six kinds of compounds in Sanskrit. 6 Points in abundance, of major or minor consequence, it w o u l d be easy t o bring up in addition, for criticism or for question. Thus, t o take a t r i f l e or t w o : according t o the general analogies of the language, we ought t o speak of the r o o t grh, instead of grah; p r o b ably the Hindu science adopts the latter f o r m because of some mechanical advantage on the side of brevity resulting f r o m it, in the rules prescribing forms and derivatives : the instances are not few in which that can be shown t o have been the preponderating consideration, leading t o the sacrifice of things more important. One may conjecture that similar causes led t o the setting up of a r o o t div instead of div, ' play, gamble' : that it may have been found easier t o prescribe the prolongation of the /"than its i r r e g ular gunation in devana etc. This has unfortunately misled the authors of the Petersburg Lexicons into t h e i r strange and indefensible identification of the asserted r o o t div ' play' w i t h the so-called root div ' s h i n e ' : the combination of meanings is forced and unnatural ; and then especially the phonetic f o r m of the t w o roots is absolutely distinct, the one showing only short / and u (as in divam, dyubhis), the o t h e r always and only long /"and (as in divyati -divan, and -dy, dyta); the one r o o t is really diu, and the o t h e r did (it may be added that the Petersburg Lexicon, on similar evidence, inconsistently but correctly w r i t e s the roots sJv and srlv, instead of s/V and sriv). It w o u l d be easy t o continue the w o r k of illustration much f u r t h e r ; but this must be enough t o show how and how far we have t o use and t o t r u s t the teachings of the Hindu gram marians. O r , if one prefers t o employ the Benfeyan phrase, we see something of what this language is which has agrammar but not an existence, and in what relation it stands t o the real Sanskrit language, begun Spiegel, for example (Altiranische Grammatik, p. 229), thinks it necessary to specify that cV/gu-compounds do, to be sure, occur also in the Old Persian dialects, but that they in no respect form a special class ; and a very recent Sanskrit grammar in
6

Italian (Pulle, Turin, 1883) gives as the four primary classes of compounds the dvandva, tatpurusa, bahuvrlhi, and avyayibhvaas if one were to say that the kingdoms in Nature are four: animal, vegetable, mineral, and cactuses.

151 Hindu Grammar and Sanskrit

in the Veda, and continued without a break down to our own times, all the rules of the grammar having been able only slightly to stiffen and unnaturalize it. Surely, what we desire to have to do with is the Sanskrit, and not the imaginary dialect that fits the definitions ofPnlni. There is no escaping the conclusion that, f we would understand Sanskrit, we may not take the grammarians as authorities, but only as witnesses; not asingle rule given or fact stated by them is to be accepted on their word, without being tested by the facts of the language as laid down in the less subjective and more trustworthy record of the literature. Of course, most of what the native grammar teaches is true and right; but, until after critical examination, no one can tell which part. Of course, also, there is more or less of genuine supplementary material in the grammarians' treatisesmaterial especially lexical, but doubtless in some measure also grammaticalwhich needs to be worked in so as to complete our view of the language; but what this genuine material is, as distinguished from the artificial and false, is only to be determined by a thorough and cautious comparison of the entire sytem of the grammar with the whole recorded language. Such a comparison has not yet been made, and is hardly even making: in part, to be sure, because the time for it has been long in coming; but mainly because those who should be making it are busy at something else. The skilled students of the native grammar, as it seems to me, have been looking at their task from the wrong point of view, and laboring in the wrong direction. They have been trying to put the non-existent grammarians' dialect in the place of the genuine Sanskrit. They have thought it their duty to learn out of Pnini and his successors, and to set forth for the benefit of the world, what the Sanskrit really is, instead of studying and setting forth and explaining (and, where necessary, accounting for and excusing) Pnini's system itself. They have failed to realize that, instead of adivine revelation, they have in their hands a human worka very able one, indeed, but also imperfect, like other human works, full of the prescription in place of description that characterizes all Hindu productions, and most perversely constructed ; and that in studying it they are only studying a certain branch of Hindu science: one that is, indeed, of the highest interest, and has an important bearing on the history of the language, especially since the dicta of the grammarians have had a marked influence in shaping the latest form of Sanskritnot always to its advantage. Hence the insignificant amount of real progress that the study of Hindu grammar has made in the hands of European scholars. Its career was well inaugurated, now nearly forty-five years ago (1839-40), by Bhtlingk's edition of Pnini's text, with extracts from the native commentaries, followed by an extremely stingy commentary by the editor; but it has not been succeeded by anything of importance,7 until now that a critical edition of the Mahbhsya, by Kielhorn, is passing through the press, and is likely soon to be completed : a highly meritorious work, worthy of European learning, and
For the photographic reproduction, in 1874, of a single manuscript of Patanjali's/Vio/idbhsyo or 'Great Comment1 (on Panini), with the glosses upon it, was but a costly piece of child's play; and the English government, p if tn make thp enternrke a.
7

complete fiasco, sent all the copies thus prepared to India, to be buried there in native keeping, instead of placing them in European libraries, within reach of Western scholars, [cf. Plate IV].

152 William Dwight Whitney

likely, if followed up in the right spirit, t o begin a new era in its special branch of study. Considering the extreme difficulty of the system, and the amount of labor that is required before t h e student can w i n any available mastery of it, it is incumbent upon the representatives of the study t o produce an edition of Pnini accompanied w i t h a version, a digest of the leading comments on each rule, and an index that shall make it possible t o find what the native authorities teach upon each given point : that is t o say, t o open the grammatical science t o knowledge virtually at first hand w i t h o u t the lamentable waste of t i m e thus far unavoidablea waste, because both needless and not sufficiently rewarded by its results. A curious kind of superstition appears t o prevail among certain Sanskrit scholars ; they cannot feel that they have the right t o accept a fact of the languages unless they find it set d o w n in Pnini's rules. It may well be asked, on the contrary, of what consequence it is, except for its bearing on the grammatical science itself, w h e t h e r a given fact is o r is not so set d o w n . A fact in the pre-classical language is confessedly quite independent of Pnini ; he may take account of it and he may n o t ; and no one knows as yet what the ground is of the selection he makes for inclusion in his system. As for a fact in the classical language, it is altogether likely t o fall w i t h i n the reach of one of the great grammarian's rulesat least, as these have been extended and restricted and amended by his numerous successors: and this is a t h i n g much t o the credit of the grammar; but what bearing it has upon the language w o u l d be hard t o say. If, however, we should seem t o meet w i t h a fact ignored by the grammar, or contravening its rules, we should have t o look t o see w h e t h e r supporting facts in the language did not show its genuineness in spite of the grammar. O n the other hand, t h e r e are facts in the language, especially in its latest records, which have a false show of existence, being the artificial product of the grammar's prescription or permission ; and t h e r e was nothing but the healthy conservatism of the t r u e t r a d i t i o n of the language t o keep t h e m f r o m becoming vastly more numerous. A n d then, finally, t h e r e are the infinite number of facts which, so far as the grammar is concerned, should be or might be in the language, only that they do not happen ever t o occur t h e r e ; for here lies t h e principal discordance between the grammar and the language. The statement of the grammar that such a thing is so and so is of quite uncertain value, until tested by the facts of the language; and in this testing, it is the grammar that is on t r i a l , that is t o be condemned for artificiality o r commended for faithfulness; not the language, which is quite beyond our jurisdiction. It cannot be t o o strongly urged that the Sanskrit, even that of the most modern authors, even that of the pandits of the present day, is the successor, by natural processes of t r a d i t i o n , of the older dialects ; and that the grammar is a more or less successful attempt at its description, the measure of the success being left for us t o determine, by comparison of the one w i t h the other. To maintain this is not t o disparage the Hindu grammatical science; it is only t o put it in its t r u e place. The grammar remains nearly if not altogether the most admirable product of the scientific spirit in India, ranking w i t h the best products o f t h a t spirit that the w o r l d has seen ; we w i l l scant no praise t o it, if we only are not called on t o bow down t o it as authoritative. So we regard the Greek science of astronomy as one of the greatest and most

153 Hindu Grammar and Sanskrit

creditable achievements of the human intellect since men first began t o observe and deduce; but w e do not plant ourselves upon its point of view in setting f o r t h the movements of the heavenly bodiesthough the men of t h e Middle Ages did so, t o t h e i r advantage, and the system of epicycles maintained itself in existence, by d i n t of pure conservatism, long after its artificiality had been demonstrated. That the early European Sanskrit grammars assumed the basis and w o r k e d in the methods of the Hindu science was natural and praiseworthy. Bopp was the first w h o had k n o w l edge and independence enough t o begin effectively the w o r k of subordinating Hindu t o W e s t e r n science, using the materials and deductions of t h e former so far as they accorded w i t h the superior methods of the latter, and t u r n i n g his attention t o the records of the language itself, as fast as they became accessible t o him. Since his t i m e , t h e r e has been in some respects a r e t r o gression rather than an advance; European scholars have seemed t o take satisfaction in submitting themselves slavishly t o Hindu teachers, and t h e grammarians' dialect has again been t h r u s t forward into the place which the Sanskrit language ought t o occupy. To refer t o but a s t r i k i n g example or t w o ; in Mller's grammar the native science is made the supreme rule after a fashion that is sometimes amusing in its navet, and the genuine and the fictitious are mingled inextricably, in his rules, his illustrations, and his paradigms, f r o m one end of t h e volume t o t h e other. A n d a scholar of the highest rank, long resident in India but now of Vienna, Professor Bhler, has only last year put f o r t h a useful practical i n t r o d u c t i o n t o the language, w i t h abundant exercises for w r i t i n g and speaking, 8 in which t h e same spirit of subservience t o Hindu methods is shown in an e x t r e m e degree, and both forms and material are not infrequently met w i t h which are not Sanskrit, but belong only t o the non-existent grammarians' dialect. Its standpoint is clearly characterized by its very first clause, which teaches that "Sanskrit verbs have ten tenses and m o d e s " t h a t is t o say, because the native grammar failed t o make t h e distinction between tense and mode, or t o group these formations together into systems, coming f r o m a common tense-stem, W e s t e r n pupils are t o be taught t o do the same. This seems about as much an anachronism as if the author had begun, likewise after a Hindu example, w i t h the statement that "Sanskrit parts of speech are f o u r : name, predicate, preposition, and p a r t i c l e . " Further on, in the same paragraph, he allows (since the Hindus also do so) that " t h e first four (tenses and modes) are derived f r o m a special present stem " ; but he leaves it t o be implied, both here and later, that the remaining six come directly f r o m the r o o t . From this we should have t o infer, for example, that dodoti comes f r o m a stem, but dadtha f r o m the r o o t ; that we are t o divide nasya-ti but do-syati, a-visa-t but a-sic-at, and so on ; and (though this is a mere oversight) that ayt contains a stem, but adot a pure root. N o real grammarian can talk of present stems w i t h o u t talking of aorist stems also ; nor is the variety of the latter so much inferior t o that of the f o r m e r ; it is only the vastly greater frequency of occurrence of present forms that makes the differences of t h e i r stems the more i m p o r t a n t ground of classification. These are but specimens of t h e method of the book, w h i c h , in spite of its merits, is not in its This work, somewhat recast grammatically, is about to be reproduced
8

n English by Professor Perry, of Columbia College, New York.

154 William Dwight Whitney

present form a good one t o put in the hands of beginners, because it teaches them so much that they will have t o unlearn later, if they are t o understand the Sanskrit language. One more point, of minor consequence, may be noted, in which the habit of W e s t e r n philology shows itself t o o subservient t o the whims of the Sanskrit native grammarians : the order of the varieties of present stems, and the designation of the conjugation classes as founded on it. W e accept the Hindu order of the cases in noun-inflection, not seeking t o change it, though unfamiliar, because we see that it has a reason, and a good one; but no one has ever been ingenious enough even t o conjecture a reason for the Hindu o r d e r of the classes. Chance itself, if they had been t h r o w n together into a hat, and set d o w n in t h e i r order as drawn out, could not more successfully have sundered what belongs together, and juxtaposed the discordant. That being the case, there is no reason for our paying any heed t o the arrangement. In fact, the heed that we do pay is a perversion ; the Hindus do not speak of first class, second class, etc., but call each class by the name of its leading verb, as bhd-werbs, ad-verbs, and so on ; and it was a decided merit of Mller, in his grammar, t o t r y t o substitute for the mock Hindu method this t r u e one, which does not make such a dead pull upon the mechanical memory of the learner. As a matter of course, the most defensible and acceptable method is that of calling each class by its characteristic feature as, the reduplicating class, the ya-class, and so on. But one still meets, in treatises and papers on general philology, references t o verbs " o f the f o u r t h class," " o f the seventh class," and so on. So far as this is not mere mechanical habit, it is pedantryas if one meant t o say: " I am so familiar w i t h the Sanskrit language and its native grammar that I can tell the o r d e r in which the bodies of similarly-conjugated roots follow one another in the dhtupthos, though no one knows any reason for it, and the Hindu grammarians themselves lay no stress upon i t . " It is much t o be hoped that this affectation will die out, and soon. These and such as these are sufficient reasons why an exposition like that here given is timely and pertinent. It needs t o be impressed on the minds of scholars that the study of the Sans k r i t language is one t h i n g , and the study of the Hindu science of grammar another and a very different t h i n g ; that while t h e r e has been a t i m e when the latter was the way t o the f o r m e r , that t i m e Is now long past, and the relation of the t w o reversed ; that the present task of the students of the grammar is t o make t h e i r science accessible, account if possible for its anomalies, and determine how much and what can be extracted f r o m it t o fill out that knowledge of the language which we derive f r o m the l i t e r a t u r e ; and that the peculiar Hindu ways of grouping and viewing and naming facts familiar t o us f r o m the other related languages are an obstacle in the way of a real and f r u i t f u l comprehension of those facts as they show themselves in Sanskrit, and should be avoided. An interesting sentimental glamour, doubtless, is t h r o w n over the language and its study by the retention of an odd classification and t e r m i n o l o g y ; but that attraction is dearly purchased at the cost of a t i t t l e of clearness and objective t r u t h .

B. Book Notice of Whitney's Roots, Verbforms and Primary Derivatives of the Sanskrit Language (1886) Franz Kielhorn

Whitney's article naturally evoked much criticism, and most of it was well deserved. The most authoritative of his critics, Kielhorn, refers to it indirectly and rather sarcastically in a short review (in 1886) on Whitney's Roots, Verb-forms, and Primary Derivatives of the Sanskrit Language (which had been published in 1885 as a supplement to the Sanskrit Grammar and which was referred to in a footnote in Whitney's 1884 article). As is apparent from his review, Kielhorn did not count Whitney among those "who more particularly have turned their attention upon the works of the native grammarians." Kielhorn's review, which is here reproduced, appeared in the Indian Antiquary (15,1886, 86-87). Kielhorn's comparison of grammarians with potters is an implicit reference to Patajali's Mahbhsya (ed. Kielhorn, I, 7.288.1): "A man who wants to use a pot goes to a potter and says ' Make a pot, I want to use it.' But a person who wants to use words does not go to a grammarian and say ' Make words, I want to use them.'" (Patajali says this in the context of a discussion where he maintains that the grammarians' rules must account for the actual usage (prayoga) of the people (oka): quoted by Bhandarkar on page 97 of this volume). Also, elsewhere in the Introduction to his commentary, Patajali produced what would be good arguments against Whitney's view that all of the uttered forms of language at least must be accessible (cf. Westergaard, quoted on page 54). In another passage quoted earlier by Bhandarkar (page 89), Patajali discusses at length the view that there are words which are not used (asty aprayuktah: part of a vrttika), ending the discussion by saying (ed. Kielhorn, I, 9.23-24): "To say that there are words which are not used without considering the enormous extent of the use of words is sheer recklessness." Patajali was convinced that the forms of a language are infinite and cannot therefore be studied by exhaustive enumeration (as was done for the Vedic corpus in the Prtiskhya-Iiterature), but only with the help of a grammar consisting of rules and exceptions (the relevant passage is illustrated in Plate IV (cf. Staal 1969, 501-502).

Highly as we value the works of the native grammarians, and convinced as we are that to them is mainly due that rapid progress which the study of Sanskrit has been making during the last century, we may yet, without fear of being misunderstood, venture to say that the time has arrived when their teaching should be subjected, by a comparison with the actual usage of the language, to athorough and searching examination. No one who has given any serious thought to the subject, would suggest that those ancient scholars of India, whose labours have been preserved by Pnini, Ktyyana, and Patajali, would willingly have misled us, or would have invented rules which they did not believe to be warranted by the language as known to them. At the same time it is possible that, in their attempt to analyse and explain the facts of the language, they may have arrived at conclusions which cannot be upheld ; that their desire to generalise may have led them to lay down rules which, true in individual cases, would, if generally observed, give rise to forms or expressions that have never been in actual use; and that the commentators may have given a meaning to their dicta which was not intended by those who originated them. Moreover, it cannot be denied, that the ancient idiom of the Vedas has not received from the grammarians that careful attention and minute description which it deserves, and that their

156 William Dwight Whitney

labours here at any rate must be supplemented by modern research. But a comparison of the teachings of the grammarians w i t h the usage of the language during its consecutive periods is likely t o furnish valuable results in another d i r e c t i o n . If it be t r u e , that t h e grammarians w e r e not like potters w h o fabricate t h e i r wares for those w h o demand t h e m , if what they aimed at was by means of rules and exceptions t o explain the existing words of the language, a careful examination and comparison of the extant literature should reveal what w o r k s must have been known t o the grammarians, and assist us in fixing more accurately the place which they hold in Sanskrit literature. It is f r o m such considerations as these, that we welcome the appearance of the valuable w o r k , the result of years of patient labour, by which Prof. W h i t n e y has laid under deep obligations all w h o are interested in the study of Sanskrit grammar. The book is intended, t o use the author's o w n words, especially as a Supplement t o his Sanskrit grammar, giving w i t h afulness of detail that was not then practicable, nor admissible as part of the grammar itself, all the quotable roots of the language, w i t h the tense and conjugation-systems made f r o m t h e m , and w i t h the noun and adjective (infinitival and participial) formations that attach t h e m selves most closely t o the verb ; and f u r t h e r , w i t h the other derivative noun and adjective-stems usually classed as p r i m a r y : since these also are needed, if one w o u l d have a comprehensive view of the value of a given r o o t in the language. So far as the informat i o n at present available allows, everything given is dated, w h e t h e r found in the language t h r o u g h o u t its whole history, or limited t o a certain period. Veda, Brhmana, Stra, Upanisads, epic poetry, or so-called classical Sanskrit. O f t h e forms taught by the grammarians, which have not yet been met w i t h in actual use, a liberal presentation is made under the different r o o t s : such material being always distinguished f r o m the rest by being put in square brackets. In addition, t o this regular Dhtuptho, the author has given indexes of tense and conjugation-stems, f r o m which it becomes at once apparent, w h e t h e r a particular stem is found only in the earlier or only in the later language, o r occurs in both ; and an index of roots, arranged in reversed alphabetical order, under which we find the interesting note, that " o f the more than 800 roots here recorded as making forms of conjugation, nearly 200 occur only in the earlier language, nearly 500 in both earlier and later, and less than 150 only in the later language." For the later periods of the language Prof. W h i t n e y has drawn his information mainly f r o m t h e St. Petersburg Dictionary; but in the older language he has done much more independent w o r k . He has, namely, himself " g o n e over all the texts of the earlier period accessible t o him, including the as yet unpublished KausTtakiBrhmanaand Kthaka, and the immense Jaiminlya o r TalavakraBrhmana, which has as yet hardly been accessible t o any one else; and f r o m them he has excerpted all the n o t e w o r t h y verbal forms and (less completely )the primary derivatives ; thus verifying and occasionally correcting the material of the Lexicon, supplying chance omissions, and especially filling in not a f e w details which it had not lain in t h e design o f t h a t w o r k t o present in t h e i r ent i r e t y . " The forms taught by the native grammarians have been given chiefly on the a u t h o r i t y of Westergaard's Radices, and for this reason a f e w w r o n g forms, which had found t h e i r way into that very accurate w o r k , have here also been repeated.

157 Kielhorn: Book Notice

Without entering into details, for which this is not the place, we venture to maintain that Prof. Whitney's book will prove of the greatest service not only to the student of Sanskrit generally, but also to those who more particularly have turned their attention upon the works of the native grammarians. To the latter the accurate and full collections contained in the work will probably suggest additional reasons for the belief that the so-called grammarian's dialect accords in a most curious manner with the language of what Prof. Whitney calls the language of the Brhmana period.

13
Bruno Liebich (1862-1939)

Kielhorn's best pupil was Bruno Liebich, who published aseries of works on the Sanskrit grammarians, ranging from Pnini to the non-Pninian schools, in particular, the school founded by the Buddhist grammarian Candragomin (about sixth century A.D.; see pp. 21 and 106 of this volume). In his monograph Pnini. Ein Beitrag zur Kenntnis der indischen Literatur und Grammatik (1891), Liebich continued a line of research initiated by Bhandarkar (pages 91-93) and established that the stage of the language described in Panini's grammar corresponds most closely to the language of two of the older Grhya Sutras, i.e., svalyana and Paraskara (see pages 91-93)434. According to Liebich, the Aitareya-brhmana and the Brhadranyakopanisad are older than Pnini, but the Bhagavad Gt is younger. One chapter of Liebich's monograph is devoted to a discussion of Whitney's views, which the author wishes to test " mit derjenigen Sorgfalt. . . welche einem Gelehrten seines Ranges gebhrt" (Liebich 1891, 45). This discussion, which constitutes Chapter V of the monograph (pages 51-61) and is here reproduced, touches on a number of interesting problems, e.g., the justification for Pnini's classification of nominal compounds. It also points out that a grammarian, when studying his mother tongue, need not confine himself to forms encountered in texts, since he is in a position to reproduce all possible linguistic forms, whenever he needs them, "durch Reflexion"just as a scientist may at any time obtain through experimentation what it would take him a long time to find in nature.

A. Prfung der Argllmente Prof. Whitney's (1891) Bruno Liebich

Eine Prfung der Grnde, welche Herrn Prof. Whitney zur Annahme eines besonderen Grammatikersanskrit gefhrt haben, wird uns zugleich Gelegenheit geben, unsere eigne Ansicht in einzelnen Punkten nher zu entwickeln. Die Beweisfhrung Whitney's zerfllt in zwei Teile. Im ersten (Am. Journ. of Phil. V, p. 282-284) prft er Pnini's Wurzelverzeichnis, unter Hinweis auf den Essay von Hjaimar Edgren : On the verbal roots of the Sanskrit language and of the Sanskrit grammarians.1 Im zweiten (p. 284-291) behandelt er verschiedene Punkte der Formenlehre, welche aus der uns bekannten Literatur gar keine oder eine ungengende Besttigung erfahren. Es war vielleicht nicht glcklich, dass Wh. grade vom Dhtuptha ausging. Wenn wir z. B. ber einen Knstler ein Urteil gewinnen wollen, so werden wir mit einer Prfung seiner besterhaltenen Werke beginnen, nicht mit solchen,welche vielleicht spter berarbeitungen erlitten haben. Das Sabdnussana zerfllt, von diesem Gesichtspunkt aus betrachtet, in drei an Wert ungleiche Teile. Am besten erhalten und ausserdem durch das Mahbhsya gesichert ist der Straptha, der daher, wie er den Hauptteil des Werkes bildet, so auch fr die Kritik der wichtigste st. Der Gana- und Dhtuptha bestehen aus Listen von lauter einzelnen Worten. Wie leicht in solchen Einschiebungen stattfinden konnten, liegt auf der Hand. Aber jener hat vor diesem noch voraus, dass er in der Ksik mit behandelt ist, und so wird sich inbezug auf ihn durch eine kritische Ausgabe derselben und mit dem Beistand, den Vardhamna's Ganaratnamahodadhi gewhrt, hinreichende Sicherheit schaffen lassen. Ob dies fr den Dhtuptha je mglich sein wird, wage ich zur Zeit nicht zu entscheiden. Falls nicht neue Hlfsmittel oder neue Gesichtspunkte fr die Kritik gefunden werden, so wrde ich vorschlagen, bei einer neuen Ausgabe des Dhtuptha diejenigen Wurzeln zu notiren, welche durch den Straptha oder das Mahbhsya gesichert sind, und ausserdem bei jeder einzelnen zu vermerken, ob und in welchem Umfange sie in der Literatur nachgewiesen werden kann. Wh. wird mir vielleicht einwenden, dass er stets von den einheimischen Grammatikern im allgemeinen rede, dass es alsoschliesslich gleichgltig sei, ob diese oder jene Wurzel Pnini selbst oder einem Spteren ihren Platz im Dhtuptha verdanke. Ich mchte aber grade auf den Unterschied zwischen alter und neuer Wissenschaft hinweisen, den wir ja in Indien auch auf andern Gebieten, z. B. der Astronomie treffen. Ich kmpfe hier nur fr die Trias Pnini, Ktyyana, Patajali. Fr die spteren Grammatiker mag alles, was Wh. tadelt, mehr oder minder zutreffend sein ; fr jene alten meines Erachtens nicht. Aber der zweite Teil von Wh.'s Essay beschftigt sich ja ausschliesslich mit Pnini's Straptha. Wenden wir uns daher jetzt zu diesem^ Den bergang bildet der Satz: 'The principle won here is likely to prove of universal application; for we have no reason to expect to find the grammarians absolutely trustworthy in other departments of their work, when they have failed so signally in one'. Nach dem oben Gesagten ein gefhrlicher Schluss, der die Befrchtung erweckt, dass W h . an die folgenden Punkte nicht mehr mit voller Unbefangenheit herantrat.
1

JAOSXI,1-55.

160 Bruno Liebich

W h . geht nunmehr daran, 'an Beispielen den wahren W e r t des einheimischen grammatischen Systems zu ermitteln' und beginnt mit der Conjugation. Sein erstes Beispiel ist das Fut. periphrast. Med. 1 In der lteren Sprache', sagt er (p. 284), ' aber nicht in der ltesten (denn es fehlt im Veda), wird ein periphrastisches Futurum Activi gebildet durch Zusammensetzung eines Nomen agentis mit einem Hifsverb, dem Prsens des Verbums as 'sein ' : so dato 'smi (wrtlich dotor sum) ' ich werde geben ' etc. Es ist ganz ungewhnlich verglichen mit dem andern Futurum, jedoch hufig genug, um zu erfordern, als ein Teil des allgemeinen sanskritischen Verbalsystems betrachtet zu werden. Zu diesem Activum geben die Grammatiker ein entsprechendes Medium, obgleich das Hlfsverbum in seinem unabhngigen Gebrauch keine mediale Flexion hat; es wird gebildet mit Endungen, die so modiftcirt sind, dass sie in dem gewhnlichen Verhltnis von medialen Endungen zu activen stehen, und weiter mit Umwandlung des zur Wurzel gehrigen s in der 1. Sing, zu h eine sehr ungewhnliche Substitution, von welcher es, wie ich glaube, kein andres Beispiel in der Sprache giebt. Nun, welche Sttze hat dieses Medium im wirklichen Gebrauch ? Nur diese: dass in den Brhmana's vier sporadische Flle von Versuchen begegnen, durch Analogie mediale Formen fr dieses Tempus zu bilden (sie sind alle aufgefhrt in meiner Sanskritgrammatik, 947; weitere Nachforschung hat keine neuen Beispiele ans Licht gebracht) : zwei von ihnen sind 1. Sing., von denen eine die Form se fr das Hlfsverbum hat, le andre he, wie in der Grammatik gelehrt wird ; und in der ganzen spteren Literatur, der epischen und klassischen, finde ich das Vorkommen eines einzigen weiteren Falles bemerkt, darsayithe (in Nais. V, 71)! Hier ist auch der klassische Dialekt der getreue Nachfolger des vorklassischen ; nur im Grammatikersanskrit hat jedes Verb, das im Medium conjugirt wird, auch ein mediales periphrastisches Futurum '. Zunchst mchte ich einen kleinen thatschlichen Irrtum Wh.'s richtig stellen. Mit der 1. Sing., welche die Form se fr das Hlfsverbum hat, meint er, wie aus dem bezeichneten seiner Grammatik ersichtlich, prayoktse TS. \\, 6, 2, 3. Nach der einheimischen Grammatik wre diese Form nicht 1., sondern 2. Sing., und das ist sie auch an dieser Stelle, wie der Zusammenhang ergiebt; Mdhava umschreibt sie mit prayoksyase. W i r haben also in der alten Literatur nicht einen regelmssigen und einen unregelmssigen Beleg fr die 1. Person, sondern je einen fr die 1. und 2., beide regulr gebildet. Mein Haupteinwand ist aber allgemeiner Natur und betrifft die meisten folgenden Punkte mit. W h . macht berall die stillschweigende Voraussetzung, dass den einheimischen Grammatikern fr ihr W e r k kein andrer Stoff zu Gebote gestanden habe als uns, oder, was auf dasselbe hinauskommt, er legt zu wenig Gewicht darauf, welche Wahrscheinlichkeit eine Form a priori habe, in der Literatur gebraucht zu werden. Das Erscheinen einer solchen ist doch nicht willkrlich, sondern richtet sich nach dem fr sie vorliegenden Bedrfnis. Nun denn, das Fut. II. ist seltener als das Fut. I., da es eine beschrnktere Bedeutung hat; dieses bezeichnet die Zukunft im allgemeinen, jenes nur die entferntere. Das Med. ist seltener als das Act, die 1. Person seltener als die 3. Welche Wahrscheinlichkeit haben wir also von vornherein, einer 1. Med. Fut. II. in der lteren Literatur zu begegnen?

161 Prfung der Argumente Prof. Whitney's

Der moderne Gelehrte, der die Grammatik einer toten Sprache schreibt und ein Pnini, der die Gesetze des von hm selbst gesprochenen Idioms zu ergrnden sucht, arbeiten auf sehr verschiedene Weise. Jener kennt keinen Zoll breit mehr von der Sprache, als was die Gunst oder Laune des Schicksals in Denkmlern irgend welcher Art ihm aufbewahrt hat. Seine Thtigkeit, soweit sie die Gewinnung des Materials betrifft, besteht darin, die im wirklichen Gebrauch bunt durcheinander gewrfelten Formen zu sondern, zu sammeln und nach gewissen Gesichtspunkten neu zu gruppiren. Er darf streng genommen keine Form ohne Sternchen in seinen Paradigmen auffhren, die er nicht in einem Texte sicher bezeugt gefunden hat. Pnini kannte wahrscheinlich weniger vedische Texte als wir; ob ihm von den Upanishaden, die wir jetzt mit Bequemlichkeit excerpiren, eine einzige zugnglich war, steht dahin. Aber wie der Naturforscher durch das Experiment sich jederzeit Bildungen vor Augen stellen kann, nach denen er in der Natur lange suchen msste, so konnte sich auch jener alle damals mglichen Sprachtypen, so oft er ihrer bedurfte, durch Reflexion herstellen und ihre Merkmale abstrahiren. hnliches thut noch heut jeder gelegentlich mit seiner Muttersprache. Wenn nun Pnini2 uns ausdrcklich sagt, dass die 1. Sing, in diesem Tempus und Genus die Endung he besass, so wrde es mir danach allein schon sehr wahrscheinlich sein, dass man solche Formen vorkommenden Falls wirklich bildete, auch wenn sich in der unabhngigen Literatur kein Beleg dafr fnde; denn wie in aller Welt sollte Pnini darauf verfallen, eine Form mit so ungewhnlichem Laut bergang zu erfinden ? Nun finden wir aber die Form thatschlich belegt, an einer Stelle, wo sie wirklich gebraucht wird : Taitt. Ar. 1,11 lesen wir yasthe ' ich werde opfern ', in Gegensatz gestellt zum Prsens yoje und zur Vergangenheit ayoksi. Ich knnte daher schliesslich die Form mit gleichem Recht fr meine Ansicht in Anspruch nehmen, dass die Brhmana-Sprache und Pnini's Grammatik, beide unabhngig von einander entstanden, doch nur Abbilder ein und desselben Originals darstellen. Die Form dorsayithe im Naisadhya ist anders zu beurteilen. Da ein Unterschied zwischen I. und II. Fut., zwischen Act. und Med. zu Harsa's Zeit nicht mehr empfunden wurde, so verdankt sie ihre Existenz mehr dem Wunsch des Autors, seine Gelehrsamkeit zu zeigen, als einem wirklichen Bedrfnis. Wh. geht jetzt zum Aoristsystem ber und bespricht zunchst dieses im allgemeinen, alsdann speciell den Opt. Aor. oder Precativ im Med., Opt. Aor. Act., zuletzt den reduplicirten Aorist. Auf p. 286 schreibt er irrtmlich die Stellung des Precativs im System hinter den beiden Futuren und dem Conditional, vom Aorist getrennt, auf Rechnung der native grammarians; Pnini wenigstens und seine Schule behandeln den Precativ zusammen mit dem Potential (li) oder vielmehr als eine Unterabteilung des letzteren (slrlin), also unmittelbar neben dem Aorist. Im brigen wre hier dasselbe zu sagen wie beim vorigen Punkte: In den Brhmana's und Stra's hat jedes Tempus der Vergangenheit, nicht minder der Potential und der Precativ oder Benedictiv seine bestimmte Funktion, wie diese auch von Pnini kurz, aberzutreffend charakterisirt wird. In den Epen ist dieses Sprachgefhl verloren gegangen, in der Kunstpoesie bei guten Schriftstellern nach der Grammatik knstlich wiederhergestellt. Wenn daher in der spteren Literatur nicht nur der Aorist, sondern alle Tempora der Vergangenheit immer
2

VII, 4, 52.

162 Bruno Liebich

seltener werden und die Participiaauf ta und tavat an ihre Stelle t r e t e n , so hat das seinen Grund nicht in einem organisch vor sich gehenden Sprach-process, sondern einfach darin, dass diese Formen leichter zu bilden sind. Inwieweit man d o r t noch von einen lebenden Sprache zu reden berechtigt ist, haben w i r an dieser Stelle nicht w e i t e r zu untersuchen. Es folgt in W h . ' s Essay das Passivum des Aorist, Futurum I. und II., Conditional und Precativ, welches nach Angabe der Grammatiker bei gewissen Verben eine verlngerte Nebenform haben

kann, z. B. dsyate oder dyisyate, hanisyate oder ghnisyate. Zunchst will ich feststellen, dass von einem Missverstndnis, wie Wh. gleichsam selbst entschuldigend annimmt, hier nicht die Rede sein kann. Die Bildung dieser Formen geht auf Pn i ni VI, 4, 62 zurck. Ktyyana, der Verfasser der Krik's und Patajali beschftigen sich ausfhrlich mit ihr und geben eine grosse Reihe von Beispielen ; kurz, diese ganze Bildung auf ein Missverstndnis zurckzufhren geht nicht wohl an. Es ist aber auch gar nicht ntig. Wh. sagt: 'Was alles dieses bedeutet, ist ganz dunkel, da es keine Anwendung davon giebt, weder frh noch spt, um Licht darauf zu werfen.'3 Aber die Seltenheit der Nebenformen wrde doch nur dann seine Bedenken rechtfertigen, wenn zugleich gezeigt wrde, dass im Gegensatz dazu die regulren, krzeren Formen relativ hufig wren. Hierber sagt W h . nichts, auch in seiner Grammatik fhrt er an der betreffenden Stelle4 keine Betspiele an, und wenn wir bercksichtigen, dass die einzige unter den oben aufgezhlten Passivformen, deren Vorkommen wir a priori hufiger erwarten drfen, die 3. Sing. Aor. fr unsern Fall nicht in Betracht kommt, da sie eine Bildung fr sich und gar keine Nebenform hat (adyi, aghni), so ist es wohl mglich, dass er berhaupt keine hat auftreiben knnen. Um so unbilliger aber ist es, zu verlangen, dass die besagten N e b e n f o r m e n hufiger sein sollen, und ihr Nichterscheinen den Grammatikern gewissermassen zum Vorwurf zu machen. Nun fhrt W h . selbst aus dem Jaiminlya Brhmana eine Form an, welche genau jener Regel Pnini's entspricht, nmlich khyyisyante', 'aber diese', fgt er hinzu, 'scheint eine mit hvayisyate etc. analoge Form zu sein und beweist folglich nichts'. Ich bezweifle dass diese bequeme Art, sich unbequeme Formen vom Halse zu schaffen, allgemeinen Beifall finden wird, hvayisyate hat hvayisyati und hvayitum neben sich, whrend die entsprechenden Formen von khy nie anders als khysyati und khytum lauten. W i r haben daher kein Recht hier eine Analogiebildung anzunehmen. Auf sol ehe Weise findet man die Wahrheit nicht, sondern verdeckt sie. Andre derartige Formen mgen durch die Ungeschicklichkeit der Abschreiber (vielleicht auch der Herausgeber) verwischt worden sein, da sie sich von den regulren oft nur durch das lange unterscheiden. Kurz, ich kann von diesem ganzen Passus nur den Schlusssatz unterschreiben, dass diese Formation den Anfngern im Sanskrit nicht eingeprgt zu werden brauchte. Im folgenden Punkte (Cerebralisirung der Endungen dhvam und dhve) wird die einheimische Lehre von Wh. ungenau oder wenigstens undeutlich wiedergegeben. Die betreffenden Stra's von Pnini lauten mit Bhtlingks bersetzung: W\\\, 3, 78 nah sJdhvamlunlitm dho 'ngt ' Nach einem vocalisch (jedoch nicht auf
3

Patajali gebraucht diese Formen mehrfach, so gleich anfangs im Mahbhsya (1,11,10) nirghnisyate.

998. Statt adsi muss es dort heissen adisi.

163 Prfung der Argumente Prof. Whitney's

o) auslautenden Stamme t r i t t an die Stelle des dh in sldhvam und in den Personalendungen des Aorists und Perfects der entsprechende crbrale Laut'. 79 vibhsetoh ' Nach dem A u g m e n t / ist in diesem Falle die Substitution nicht n o t w e n d i g ' . Die G r a m m a t i k e n von Mller und Kielhorn fhren ausser den Vocalen noch h und die Halbvocale an, nach denen die Cerebralisirung ebenfalls stattfinden solle, und sie haben ein Recht dazu, da der Pratyhra in auch diese Laute umfasst; zu Gunsten der Bhtlingk'schen b e r setzung lsst sich anfhren, dass Pnini gewiss hauptschlich an vocalisch auslautende W u r z e l n gedacht, und dass auch die Ksik nur aus solchen ihre Beispiele gewhlt hat. Die Formen ojanidhvam etc. sind nach diesen Regeln allerdings falsch, wenigstens nach dieser Auslegung der Regeln ; sie w r d e n r i c h t i g , sobald man, von der einheimischen T r a d i t i o n abweichend, das W o r t nah aus 78 in der folgenden Regel n i c h t f r t g e l t e n liesse. Indessen sind dies doch nicht die einzigen hier in Betracht kommenden Bildungen, w i e es nach W h . ' s W o r t e n (it happens that all t h e quotable examples are opposed t o t h e i r rule) scheinen k n n t e ; W h . f h r t selbst in seiner G r a m m a t i k astodhvam, avrdhvam, cyodhvam auf, und hier

erfolgt die Cerebralisirung ganz in bereinstimmung mit der grammatischen Lehre. Es handelt sich also um einen Sprachprocess, den Pnini bemerkt und dessen Grenzen er festzustellen gesucht hat. Dass dieser Process wirklich vorhanden war, wird durch die in der Literatur vorkommenden Flle gengend bezeugt. Die Grenzen richtig anzugeben, mag hm nicht ganz gelungen sein ; aber vergessen wir auch nicht, dass er sich hierbei nur nach dem Gehr richten konnte, und dass schon ein feines Ohr dazu gehrt, um den Unterschied zwischen einer Endung dhvam mit dentalem oder cerebralem dh berhaupt wahrzunehmen. W i r sehen auch aus der Ksik, dass die Grammatiker in manchen Fllen selbst zweifelhaft waren, was das Richtige sei. Jedenfalls scheint es mir unbillig, aus einem solchen Versehen Pnini, dem wir grade auf dem Gebiet der Cerebralisirung eine Flle feiner und treffender Beobachtungen verdanken, einen schweren Vorwurf zu machen. Wie wir oben sahen, geben die Grammatiker diese Cerebralisirung auch fr die Endung dhve des Perfectums an. W h . findet in der Literatur dafr kein Beispiel. Er kennt zwar auch kein Gegenbeispiel, weil auch diese Form wieder ihrer Bedeutung nach so selten ist, dass man nur durch einen glcklichen Zufall hoffen darf, ihr in der Literatur berhaupt zu begegnen ; aber dies hindert ihn nicht, deswegen in der strksten Weise auf die Grammatiker loszuschlagen.5 Warum ? Weil hier fr die Umwandlung des dh zu dh kein Grund erkennbar sei. Als ob wir in irgend einer Sprache berall die Verkettung von Ursache und Wirkung nachzuweisen vermchten ! Wenn solche Formen in der Literatur wirklich vorkmen, wrde nur ein Dilettant es wagen, sie nach dem Gesetz von Ursache und Wirkung zu emendiren. Man wrde sie, vorausgesetzt, dass Wh.'s Auffassung ber die Ursache der Cerebralisirung im allgemeinen richtig ist, als falsche Analogiebildungen ansehen, wie wir sie berall da beobachten, wo der G r u n d eines Sprachprocesses nicht mehr deutlich empfunden wird. Wenn also dieser Fall darauf deutet, dass Pnini sich auch in Nebensachen davor gehtet habe, durch vorgefasste Meinungen seinen Blick trben zu lassen, wenn ihn seine Ehrfurcht vor der 'gttlichen SarasvatT' vor diesem schlimmsten aller Beobachtungsfehler be5

Er spricht von downright unintelligent blundering, ' handgreif-

lichem unverstndigem Gestmper.1

164 Bruno Liebich

wahrte, so, denke ich, knnen w i r dem gtigen Geschick dankbar sein, welches uns das W e r k dieses Augenzeugen einer fernen Vergangenheit so trefflich erhalten hat. Bis jetzt hat W h . immer nur solche Formen zur Vergleichung herangezogen, die in der Literatur naturgemss so selten sind, dass sieeben kaum verglichen werden knnen.Bei dem grossen Material, das diesem Gelehrten durch seine Sammlungen zur Verfgung steht, scheint es mir im Interesse der Sache bedauerlich, dass er nicht glcklichere O b j e k t e ausgewhlt hat, zumal er (p. 291) versichert, dass es ihm ' leicht sein w r d e , das W e r k der Illustrat i o n viel w e i t e r fortzusetzen '. Denn w i r stehn am Ende seiner Beweisfhrung. Was nun noch folgt, bezieht sich nicht auf den Inhalt der Lehre Pnini's, sondern nur auf die Form seiner Darstellung, aufsein System. Zunchst bemngelt W h . Pnini's Einteilung der Composita. ' Nichts konnte genauer und glcklicher sein als die Unterscheidung von abhngigen, beschreibenden, besitzanzeigenden und copulativen Zusammensetzungen ; nur ihre Titel 'sein Mann ' (tatpurusa), 'die Handlung sttzend' (? karmadhroya), 'Vielreis 1 (bahuvrihi) und ' Paar' (dvandva) beziehungsweiseknnen kaum beanspruchen, der Beibehaltung w e r t zu sein. A b e r es ist das Charakteristikum der H indu Wissenschaft im allgemeinen, nicht fhig zu sein innezuhalten, wenn sie genug gethan hat; und so haben uns die Grammatiker, auf demselben Einteilungsniveau mit diesen vier Hauptklassen, noch zwei andre gegeben, welche sie dvigu (' zweikhig') und avyayibhva ('das Indeclinabelwerden ') nennen; und diese haben kein Existenzrecht, sondern sind Sammlungen von Specialfllen, die zu einer der andern Klassen gehren ' etc. Es ist mir nicht gelungen, festzustellen, welcher Quelle Prof. W h . fr diese Darstellung gefolgt st. Die mir bekannten abendlndischen Grammatiken fhren smtlich nach Pnini's V o r b i l d nicht sechs, sondern vier Hauptklassen von Zusammensetzungen auf; nmlich Tatpurusa, Bahuvrihi, Dvandva, Avyayibhva. Die Karmadhraya sind eine Unterabteilung der Tat pu rusa, die Dvigu nur ein specieller Fall der Karmadhraya. Dass Pnini, wie jeder Systematiker, das Recht hat, Unterklassen aufzustellen und zu benennen, soweit es ihm zweckdienlich erscheint, bedarf keines Beweises. Da brigens W h . zu wissen wnscht, wie die Inder zu ihrer Klassifikation kamen, 6 so lasse ich ihr EnteiIungsprincip folgen, welches, nebenbei bemerkt, so einfach ist, dass es Pnini berflssig erschien, dasselbe im Straausdrcklich auszusprechen (cf. Ks. zu I, 2, 57). N u r beim Bahuvrihi findet sich ein Hinweis darauf. W e n n man nmlich vom zweigliedrigen Compositum ausgeht, so kann man fragen, w o der Schwerpunkt (pradhna) desselben ruht. Hier sind folgende Flle mglich : 1. beide Glieder haben gleiches Gewicht (ubhayapadrthapradhna); 2. der Schwerpunkt ruht auf dem ersten Teil (prvapadrthapradhna); 3. der Schwerpunkt ruht auf dem zweiten Teil (utta rapadarthapradhna) ; 4. der Schwerpunkt liegt ausserhalb des Compositums (anyapadrthapradhna).

Alle vier Flle kommen in der Sprache wirklich vor, ihnen entsprechen die vier Hauptklassen der indischen Grammatik, und zwar der Reihe nach die Dvandva, Avyayibhva, Tatpurusa und
6 Er gebraucht fr dieselbe den nicht sehr geschmackvollen Vergleich :

Tiere, Pflanzen, Mineralien und Kaktusse.

165 Whitney: On Recent Studies

Bahuvrhi. Was die mehrgliedrigen Compositaangeht, so haben in den Dvandva auch hier, w i r vorauszusehen, alle Teile gleiches Gewicht; alle brigen lassen sich auf zweigliedrige zurckfhren. Schiiesslich werden, als Reprsentanten von ' points in abundance, of major o r minor consequence', die noch zu tadeln seien, von W h . die beiden W u r z e l n groh und div aufgefhrt, die richtiger mit grh und dlv oder dl htten angesetzt werden sollen. ber diese Dinge will ich mit W h . nicht rechten. Es ist nicht meine Absicht, das S y s t e m Pnini's in allen Punkten verteidigen zu wollen. Ich denke mir, dass es damals, als es entstand, den Verhltnissen und Bedrfnissen am besten Rechnung getragen hat, da es alle andern aus jener Zeit stammenden verdrngte. Aber es wre nicht gut, wenn es der heutigen Wissenschaft nicht gelingen sollte, an seine Stelle ein vollkommeneres zu setzen. Den sachlichen Inhalt seines Werkes drfen w i r , das hoffe ich gezeigt zu haben, im allgemeinen als wohlverbrgtes, zuverlssiges Material ansehen ; sein System, die A r t , wie er die Formen erklrt und analysirt, mgen w i r beibehalten oder verwerfen, soweit w i r es fr gut finden. Die grossen Verdienste des Herrn Prof. W h i t n e y auf vielen Gebieten der indischen Philologie sind zu bekannt, als dass ich sie hier hervorzuheben brauchte; aber eben wegen der grossen A u t o r itt, die sein Urteil dadurch besitzt, schien mir eine eingehende Widerlegung notwendig. Seine Ansicht ber die indischen Grammatiker ist ihm sozusagen in Fleisch und Blut bergegangen; in seiner Sanskritgrammatik, die doch sonst vom wissenschaftlichen Standpunkt aus die erste Stelle einnimmt, kann man nicht drei Seiten weit lesen, ohne sie reflectirt zu finden ; hier findet der Leser gelegentlich zu seiner berraschung Bildungen, die Pnini ausdrcklich lehrt, getrost als Barbarismen bezeichnet. 7 Auch in Schriften jngerer Sanskritisten, namentlich der amerikanischen Schule, entsinne ich mich hnliche absprechende Urteile ber die einheimische Grammatik angetroffen zu haben. Dieser Auffassung musste nachdrcklich entgegengetreten werden. Z u m Schluss giebt Prof W h i t n e y denen, die sich mit der einheimischen Grammatik nher beschftigt haben, einige Ratschlge, wofr ihm diese dankbar sein werden. W e n n ich mir erlauben darf, als Gegengabe gleichfalls einen Wunsch auszusprechen, so wre es der, aus seiner Grammatik knftig alle Seitenblicke und Anspielungen auf die native grammarians zu entfernen. Er will doch in derselben nur das Sanskrit, wie es sich aus den Literaturdenkmlern abstrahiren lsst, historisch darstellen. Durch einen Verzicht auf diese hchst oberflchlichen und oft ungenauen Bemerkungen wrde sein sonst so vortreffliches W e r k entschieden gewinnen. Whitney answered promptly in an article " O n Recent Studies in Hindu Grammar" of 1893, in which he discusses other works of Liebich's as well as the work of Otto Franke. It appeared in the American Journal of Philology (14,1893,171-197). The problem of the meaning of bhasa in Pnini's grammar, brought up by Franke and brushed aside by Whitney, and the related problem of the meaning of chandas have received much attention in later studies but remain unsolved. Renou (1969, 456-457) queries still: " Is this 'spoken language* a limited and distinct dialect or did Pnini (outside the two zones chandasi and bhsym) hold a view on language which comprised, in an undifferentiated
7

B. On Recent Studies in Hindu Grammar (1893) William Dwight Whitney

Cf. 473, c und Pan. V, 3,56.

166 Bruno Liebich

manner, the chandaswith the exception of archaisms and the bhsV For further references see Wackernagel-Renou 1957, 88 note 317. Nine years ago (in October, 1884) I published in this Journal a paper entitled "The study of Hindu grammar and the study of Sanskrit." It was intended to emphasize the difference between Sanskrit on the one side and Pnjnj and his successors on the other, and to point out the true place of the native grammar as an important division of Sanskrit science, requiring to be studied as such, and not as the foundation of our knowledge of the Sanskrit language. Since that time there have appeared a number of contributions to our knowledge of the Hindu grammar, from the pens of two younger scholars of decided ability, then unknown; and these contributions I propose to examine briefly, especially in order to see how they stand related to the question above set forth. The first of them appeared in 1885, and was entitled "The casesystem of the Hindu grammarians, compared with the use of the cases in the Aitareya-Brhmana" ; it was a doctorate-dissertation by Bruno Liebich ; the author is at present a privat-docent in the Breslau University. Its first part, printed in vol. X of Bezzenberger's Beitrge zur Kunde der indogermanischen Sprachen, was a digest of the system of rules laid down by Pnini for the use of the cases, and was very welcome, as must be every contribution to an easier understanding of the peculiarities and difficulties of the Hindu science. A few words as to the system may not be out of place here. Pnini does not take up the cases as forms of nouns, setting forth the various uses of each, after our manner; he adopts the vastly more difficult and dangerous method of establishing a theoretical list of modes of verb-modification by case, or of ideal case-relations (he calls them kraka, 'factor 1 or ' adjunct1), to which he then distributes the cases. Almost as a matter of course, however, his case-relations or kraka are not an independent product of his logical faculty, but simply a reflection of the case-forms; they are of the same number as the latter, and each corresponds to the general sphere of a case: they are kartar ('actor' = nominative), karman (' act ' = accusative), sampradna (' del i very ' = dative), karana (' instrument' = instrumental), adhikarana ('sphere' = locative), and apdna (' removal ' = ablative). The genitive has no defined character, but is provided for by stating, when all the other case-uses have been rehearsed, that the remainder are those of the genitive. As for the definitions of the case-relations, it may suffice to say that the karman is described as belonging, first, to that which the actor in his action especially desires to obtain or attain (as in " he makes a mat,1' " he goes to the village"): or, second, to that which, though itself undesired or indifferent, is connected with the action in a similar manner. Anything more crude or unphilosophical than this could not well be imagined. There is not an identity between the use of a given case and the presence of its generally corresponding case-relation, because, for example, in a passive sentence, as "the mat is made by him," mat is still called karman or 'act,' though nominative, and him still kartar or 'actor,' though instrumental. Thus there is no recognition of the grammatical category of subject of a verb; and this leads, as could not be helped, to numerous obscurities and difficulties. Then, in the second part of the paper (ibid., vol. XI, 1887), the author proceeds

167 Whitney: On Recent Studies

'y

t o classify under this scheme, in all its headings and sub-headings, its general rules and its exceptions, the facts of case-use in the Aitareya-Brhmana: a careful and creditable piece of w o r k . The results of the comparison are precisely what we should expect t o find t h e m , knowing w e l l , as we do, the relation of the language of the Brhmanas t o Pnini's Sanskrit: t h e r e is agood degree of general agreementas t h e r e w o u l d have been found t o exist even if the Rg-Veda instead of a Brhmana had been compared; since changes of syntactical construction, perhaps even more than changes of f o r m , are of slow progress in every language, leaving the main body of older usages long untouched. Alongside of this agreement are met w i t h just the differences that could not fail t o appear: constructions in the Brhmana that are unnoticed in Pnini, as they are wanting in classical Sanskrit; and especially a host of details in Pnini of which the Brhmana exhibits no examples. There is absolutely nothing t o show, or even t o give reason t o suspect, that any special relation exists between Pnini and this Brhmana any more than any o t h e r of the same class of w o r k s , specimens alike of the Brhmana stage of development of ancient Indian language. The conclusion is that, whatever its defects of t h e o r y , Pnini's case-syntax proves t o be a fairly good practical scheme; and the demonstration of the fact is t o be received w i t h thanks; it is a valuable c o n t r i b u t i o n t o o u r appreciation of the great grammarian. W h e t h e r , however, the author views it in just this light is a little questionable; for he adds as second t i t l e t o his essay " a c o n t r i b u t i o n t o the syntax of the Sanskrit language" which it plainly is not. Is it, forsooth, the Brhmana that he has been examining, t o see w h e t h e r its case-constructions are such as they ought t o be? o r is this part of its grammar now better understood than h i t h e r t o , o r arranged in a manner which w e shall be disposed t o accept as preferable t o , for example, Deibruck's? N o t h i n g of all t h i s ; it is simply that Pnini has been tested by a bit of real language, and the test has t u r n e d out not t o his disadvantage. The misapprehension that something done for Pnini is done for the Sanskrit language is precisely what my f o r m e r paper was especially intended t o discourage. Dr. Liebich adds at the end of his own estimate of the results of his w o r k : " I. The Aitareya-Brhmana is older than Pnini." This w e r e better stated the o t h e r way: namely, that Pnini is later than the Brhmana; since it is really the grammarian, and not this member of the literature, that is under examination. As for the relation itself, it is not only t r u e , but a t r u i s m ; no one having any knowledge of the subject has or could have any question about it ; our author's paper is not a demonstration, but merely an illustration, out of one department of grammar, of a fact already i n c o n t r o v e r t i b l y established on many and sufficient grounds. The author adds as follows : " It (the Brhmana) belongs t o the Vedic period, but t o t h e close of the latter, and stands fairly near t o Pnini (undoubtedly much nearer than t o the Rg-Veda in the o t h e r d i r e c t i o n ) . " Here again we have t r u t h s , but, since t h e r e has been no comparison made between Brhmana and Veda in the paper, they are incorrectly put f o r w a r d as its " results." Further, " 2 . The doctrine of Pnini reposes upon a careful and acute observation of the actual language." Here it is a l i t t l e doubtful w h e r e the stress of the assert i o n lies, and what counter-proposition is intended t o be gainsaid. N o one, certainly, w o u l d t h i n k of denying that Pnini observed and described w i t h remarkable acuteness and t o the best of his ability. N o r , again, I should t h i n k , that he described an actual

168 Bruno Liebich

language" an " rather than " t h e , " for just what language he was dealing w i t h is one of the disputed points. The author's added remarks indicate that he thinks it a book-language; if anything in the rules is not capable of being instanced, it is, he suggests, because so much of the literature has been lost. This seems an untenable view, and has doubtless been since abandoned by him. The question will come up again f u r t h e r on. Four years later (1890), in the same Journal (XV11-2), a kindred topic is taken up by another scholar, Dr. R. O t t o Franke, now aprivat-docent in the Berlin University. The t i t l e of his paper is " T h e case-system of Pnini compared w i t h the use of the cases in Pali and in the Asoka inscriptions." He builds upon the foundation laid by Liebich, adopting the latter's scheme of Pninean case-uses, and looking for correspondences t o them in the dialects confessedly later than Pnini, as the Brhmana was confessedly earlier. Here also he finds all the agreement that could reasonably have been expected ; and, as the ground has been comparatively little w o r k e d over, his w o r k is much more t r u l y a c o n t r i b u t i o n t o the syntax of the dialects of India than is that of Liebich. He brings t o light one very curious thing : that for a problematic rule of Pnini's, declaring the f u t u r e tense t o be usable in describing something recently past, examples are quotable f r o m the Pali, though they have never been discovered in Sanskrit. But his general views as t o Pnini and the Sanskrit seem rather strange. He calls Liebich's little w o r k " a beginning t o w a r d the accomplishment of the very pressing task of determining by internal evidence Pnini's position in the literature, and so, indirectly, that of the Sanskrit "as if nobody, before the appearance of this doctorate-dissertation, had done anything w o r t h y of mention in that direction ; or as if the position of Pnini's Sanskrit in the history of development of Indian language had not long been clear enough. And he points out that, in spite of the partial agreement between t h e case-uses in the Aitareya-Brhmana and Pnini's rules, we ought not t o conclude that the Brhmana was the exclusive, or even the principal, foundation of the rulesas if it could ever enter into the mind of any reasonable person t o draw such a conclusion. He then gives us the same warning in regard t o the Pali, which is even, if possible, more superfluous. He f u r t h e r admits it as possible, though on the whole less probable, that Pnini may have "collected the phenomena of very diverse dialects, and fused them together into an integral w h o l e " t h a n which nothing could well find less t o be said in its favor. But t o the question as t o what the Sanskrit of Pnini really is the same author returns in a special paper entitled " W h a t is Sanskrit?", dated in November, 1889 (though first published in vol. X V I I , 1891, of Bezzenberger's Beitrge). Kather more than half the paper is occupied w i t h the more specific inquiry as t o what Pnini means by bhs, a w o r d that he uses only seven times, or t o o seldom t o set f o r t h its significance w i t h the desirable clearness. ' Popular speech ' is its natural sense ; but the usages quoted from it by Pnini as opposed t o his o w n approved language show that it was no Prakritic dialect (that is the chief result of the author's investigation); and it is as evidently not one of the older Vedic dialects ; there seems t o remain, then, only one possibility: it is essentially Sanskrit, only not what Pnini accepts as good Sanskrit; it includes those words and phrases which, though more or less current, he does not regard as w o r t h y t o be perpetuated.

169 Whitney: On Recent Studies

This conclusion appears t o be a reasonable and safe one. 1 The second half of the paper then deals more directly w i t h the inquiry as t o what Pnini's Sanskrit really is; and the author's opinion is expressed in these terms (pp. 75-76): "Pnini's Sanskrit is accordingly in the main bhs. A n d yet, on the o t h e r hand, it is neither bhs nor a living language." This is not particularly clear; nor is it made very much clearer by the reasonings, and the quotations of the views of others, that follow. It is t o me so strange as fairly t o be called unaccountable that these authors take no notice whatever of the evidence of the dramas upon the subject. In t h e latter we see a condition of society in which educated people talk Sanskrit, w h i l e the uneducated talk Prakrit, in dialects more or less different f r o m one another. So far as I can perceive, t h e r e is not any reason t o question that this state of things was real at the t i m e when those dramas w e r e produced which then set the rule for all f u t u r e t i m e . The speakers all understand one another; the difference between Sanskrit and Prakrit is not yet sufficient t o prevent t h a t ; the Prakrit-speakers can even, in an emergency, put in a phrase of Sanskrit; and, on the o t h e r hand, when King Purravas goes mad, he casts off the restraints of education, and talks in part Prakrit, like a woman. That, now, is just the present character of Sanskrit: an educated or learned dialect, kept in existence, nearly unchanged, by instruction, by learned and literary use, among languages now become so diverse f r o m it that its knowledge is confined t o a very small circle; such, t o o , has been its character for at least t w o thousand years, while the t r u e vernaculars have been g r o w i n g f u r t h e r and f u r t h e r away f r o m i t ; and such must unquestionably have been its character at the outset, when t h e i r divergence, and its separate life, first began. That it was itself originally a vernacular seems t o me a matter of course; nor do I see that any one has the right t o say that Pnini's speech was not a living one, unless he then enters into afull explanation of what he means by a living language as distinguished f r o m it. Sanskrit was the natural successor of the dialects of Veda, Brhmana, and Sutra, and as much " l i v i n g " as any of these had been, when the literary and learned class t o o k it in hand, and, w i t h t h e aid of grammatical science, fortified it against the f u r t h e r effect of the changes that w e r e bringing out of it the various Prakritic dialects (taking that w o r d in its widest sense). There is no absolute line t o be drawn between living and dead languages. If the Sanskrit has never failed of being kept up by a constant t r a d i t i o n f r o m teacher t o pupil, though in a limited class, t h e r e is a real sense in which it has never died, but is still a living tongue. In another and equally correct sense, no language is alive that is not an out-andout vernacular, spoken by a w h o l e community, and having no inferior dialect below it in the same c o m m u n i t y : in this sense, t o be sure, the Sanskrit of the series of grammarians of w h o m Pnini was the chief and virtually the last was not a fully living t o n g u e ; it had Prakritic dialects under it. Moreover, as soon as it t o o k on the character of a learned dialect, it began as a matter of course t o be stiffened into something a l i t t l e unnatural ; no dialect ever fell into the hands of grammarians w i t h o u t suffering f r o m t h e i r pedantry. But I can find no reason whatever for supposing that it was not t h e i r o w n language, the language which they themselves
1

It is, however, rejected by Liebich, in his Ksik1 (p. xxv), to be described

further on. Liebich suggests no substitute.

170 Bruno Liebich

spoke and which they t h o u g h t alone w o r t h y t o be spoken by others, that they set themselves t o describe. W h a t e v e r Pnini's special original part in the w o r k may have been, we know that he left it still abounding in errors, both of omission and of commission ; the i m p o r t a n t additions and corrections of Ktyyana and Patajali, t o say nothing of t h e i r numerous but more insignificant successors, amply prove t h i s ; and it is frankly conceded in many points by these latest students of the system, unlike the scholars of a generation or t w o ago. The task Panini attempted was beyond the power of mortal man t o accomplish, especially in the f o r m adopted by himwhich is one that no sensible man should ever have chosen, yet on account of which, it is very likely, his contemporaries and successors especially admired him, and made him t h e i r supreme a u t h o r i t y . Something like this, in my opinion, is what we have a right t o say that we know about Pnini ; and the investigations of Liebich and Franke, while they bring nothing t o light that contradicts it, merely illustrate here and t h e r e a point in it, and do not add notably t o its amount, because they ignore it all, and assume that the most fundamental facts involved have still t o be established. W h a t we really need f u r t h e r is added precision on a host of points as t o which we have as yet only general knowledge, and particularly a comprehension of how the grammatical system, in all its details, stands related t o the language of the Sanskrit classical literature, which professes t o be governed by it, and yet has evidently had a traditional life of its o w n , simply regulated by the grammar, and has by no means been produced under the latter's dictation. To ask and answer, in all seriousness, such questions as w h e t h e r a certain Brhmana, or w h e t h e r the Pali, is Pnini's Sanskrit, or w h e t h e r that language was a living one, appears t o me the w r o n g way t o arrive at any valuable result. In his conspectus of the views of various scholars as t o the character of Sanskrit, given in the second part of his paper, Dr. Franke quotes w i t h approval and acceptance an odd expression of opinion by W e b e r , made at the very outset of his career, t o the effect that " the development of Sanskrit and of the Prakrit dialects out of t h e i r common source, the Indo-Aryan mothertongue, w e n t on w i t h absolute contemporaneousness (vollstndig gleichzeitig)." But I do not see why this is not an unscientific and untenable proposition. For example, pakkhitta and otto or app are not contemporaneous w i t h praksipta and tm n the historical development of language, any more than Ital. rotto and rotti w i t h Lat. ruptus, -urn in t h e i r various case-forms; and so hodu is preceded in point of t i m e by bhavatu, being a later " c o r r u p t i o n " of the latter, coming t o take its place, as Fr. tait of stabat, or ftes of fuistis. A n d this is t r u e of the great mass of Prakrit words, forms, and constructions; they are developed later than, and c o m e t o be substituted for, the corresponding Sanskrit words, forms, and constructions. If t h e r e w e r e anything t o be found on Indian ground that is earlier than praksipta, and f r o m which it and pakkhitta should have equally descended by a parallel process, then we might have a right t o speak of t h e i r contemporaneity; but that is plainly not the case; it is the Sanskrit forms themselves, and not something older and more p r i m i t i v e than Sanskrit, that the Prakrit words presuppose; they have passed t h r o u g h the stage which the Sanskrit represents. That here and t h e r e exceptions are met w i t h , altered items for which the original is not found in Sanskrit, or is found in Vedic Sanskrit, is w i t h o u t any force whatever as against the

171 Whitney: On Recent Studies

great mass of material of a contrary character; such exceptions to the descent in toto of one dialect from another are the rule in all dialectic history, and might with equal justice be relied on to prove that Italian and French are in their development "absolutely contemporaneous" with Latin. As to the other half or side of the view already quoted, Dr. Franke adds: "That the Sanskrit had become extinct when the Prakrit dialects first began todevelop themselves is false." What this means is quite unintelligible; it seems to go out of the way to deny a doctrine which no well-informed student of language could by any possibility think of maintaining, and it accordingly has no claim to be criticized, but must be simply set aside as valueless. If, for example, tm had ever become extinct, whence should ott or app have originated ? Who would say that the egg had been extinct when the chicken first began to develop itself? But, somehow or other, those whose ancestors had said tm began to say ott i nstead, the one pronunciation passing into the other, with no extinction intervening. It was, however, only a part of the community who did thus; a part, doubtless much the smaller one, continued to say tm; and the two forms went on in currency side by side, as educated and as popular speech, in the same way as in many cases elsewhere in the world ; and tm was Sanskrit, and, with some help and some mishandling on the part of grammarians, has maintained itself in being to this day, in the literature which we call Sanskrit, and which, rather than the grammarians' treatment of it, is the true and proper object of the study of the Sanskrit scholar. Next was produced by Dr. Liebich, in 1891, a valuable collection of studies entitled " Pnini : a contribution to the knowledge of Indian literature and grammar" ; it makes a small octavo volume of 164 pages. The first study, or chapter, deals with Pnini's period ; the author reviews briefly the opinions that have been held by different scholars respecting the matter, and, without attempting to bring any new evidence to bear upon it, comes to the moderate and sensible conclusion that only a certain degree of probability can be arrived at: "after Buddha and before Christ" represents to him the measure of this probability. The second chapter treats of the principal later grammarians who have continued and modified Pnini's work; in regard to the earliest and most important of them the same chronological uncertainty prevails. The third is entitled " Pnini and the remaining literature," and is an attempt to determine where in the succession of the ancient literature of India, from the earliest Veda down, Pnini comes in. It takes as starting-point the wild views of Goldstcker, with their refutation by Weber; it points out further the insufficiency of the evidences relied upon for the prevailing opinion that Yska is earlier than Pnini ; and it then proceeds to its principal task, of applying to the general question a new, a numerical-statistical , method of solution. The author counts off, namely, a thousand personal verb-forms occurring in succession in each of four different monuments of the literaturethe Aitareya-Brhmana, the Brhad-rnyaka, two Grhya-Stras (Asvalyana and Praskara), and the Bhagavad-Glt; representatives respectively of the Brhmana, Upanisad or later Brhmana, Stra, and epic stages of development of Indian speechand then applies to them the rules of the grammar, to see how many and what forms unauthorized by Pnini appear in the several texts. The examination is creditable to the industry and learning of its author, and its results are interesting; we can hardly go further than that and

172 Bruno Liebich

pronounce them important. For they are essentially illustrative o n l y ; they put in a numerical f o r m peculiarities which were already familiarly known t o characterize the different classes of w o r k s instanced. N o t a new item, so far as I can see, is brought t o light; nor is any made more certain than before. Thus, six of the seven classes of Brhmana divergencies drawn out on pages 23-4 have long been recognized as such ; and how many examples of each class may chance t o occur in a given amount of t e x t is a matter of indifference. As for the seventh, represented by a single case, the lengthened final of the imperative krdhJ, it is an e r r o r ; such a protraction does not belong t o the Brhmana language, as indeed, it has no right of occurrence anywhere except in verse; where it appears here (ii 2. 21), it is simply copied f r o m the RgVeda verse (i 36.14) on which the Brhmana is engaged in commenting, and of which it repeats a whole pada (including krdhJ) w i t h merely the substitution of the more regular caranya for carathya in i t ; and the retention of the /" is not improbably even a misreading, such as this Brhmana has in no very small number (it may be added that the author, doubtless misled by Pnini, describes krdhi falsely as a present instead of an aorist imperative). And so also in each of the remaining cases. That is t o say, the matter is not one t o which the numerical method of investigation is well suited ; this would be much better applied between, for example, different texts of the same class, as different Brhmanas, t o see w h e t h e r it would yield any evidence as t o t h e i r respective periods ; and perhaps the part of the whole investigation which is of most value is the comparison which it makes possible between Aitareya-Brhmana and Brhad-Aranyaka, the latter being part of a Brhmana also, but plainly later, as was a matter of course for an Upanisad. Instead, again, of the Bhagavad-Glt, which no one doubts t o be a comparatively recent addition t o the Mahbhrata, it were much t o be wished that the author had selected something out of those parts of the epic which are most probably t o be regarded as its original nucleus, in o r d e r t o cast more light upon the really difficult and h i t h e r t o doubtful question how and how much the epic differs f r o m the classical or Pninean Sanskrit, and why. That Brhmana and Upanisad and Stra antedate Pnini we knew just as certainly before this investigation was made as we know it n o w ; the posteriority of the Bhagavad-Glt, again, could hardly have been questioned, however the case may stand w i t h the earliest epic. The criteria applied t o the divergences of the Gta from grammatical strictness are of a less satisfactory and decisive character. The decided majority (21 against 16) of the irregularities concern the voice of the v e r b ; but, though the looseness of at least the later epic in this regard is certainly excessive, it is likely that Pan mi's rules limiting the employment of the voices are exceptionally artificial and discordant w i t h genuine usage; our author himself so judges examples of them (e.g., p. 28) in connection w i t h the Brhmana. As for the causative perfects w i t h asa (3 in number), Pnini's failure t o authorize them must be either an oversight or a piece of pedantry. Any sucas, since this aorist occurs in Veda and Brhmana, might be deemed a sign rather of antiquity than of modern date. The harvest of results from the chapter, then, must be confessed a rather scanty one. In the sixth and seventh chapters the author returns t o the Aitareya-Brhmana and the Brhad-ranyaka, in order t o see whether any difference of period can be established among t h e i r constituent parts. Here again is, as in the particular noted above,

173 Whitney: On Recent Studies

agood and suitable application of the statistical method, and it leads t o t r u s t w o r t h y and interesting conclusions. In the ranyaka are discovered no notable indications of diversity of age; but in the Brhmanathe author finds good reason t o believe, as had been inferred by others before him, that the concluding chapters are more modern than the rest. Between the parts of the volume devoted t o the first and t o the second examination of these t w o works intervene a couple of chapters, of which the former, the f o u r t h , is headed " Pnini's relation t o the language of India"that is to say, the relation of Pnini's Sanskrit to the other dialects. The chapter is chiefly composed of a succinct statement of the views of other scholars, to which the author then appends his own view; and this is simply a summary of what he has illustrated in the preceding chapters as the relation between Pnini's dialect and the Brhmanaand Stra on the one side and the epic on the other. Then (p. 50) he appends as final result a wholly new and original classification of the entire body of dialects of India. They are divided into three categories: pre-classical, classical, and post-classical. To the classical division are referred, besides " t h e doctrine of Pnini," the Brhmanas and Stras also, which the author has himself previously recognized as pre-Pninean ! This leaves as pre-classical only " t h e samhits of the four Vedas." But the t h i r d division, the post-classical, is still more wonderfully constructed : besides the " independent" epic, it contains the whole literature which we have been accustomed t o know as Sanskrit, namely " Klidsa, etc., originated under the influence of the g r a m m a r " ! W h a t is left t o constitute the classical subdivision " b. Doctrine of Pnini " is very obscure; it can be only Pnini's grammar itself (so that such sentences as idamo rhil, gokuta-dibhyo 'ninitare classical, as contrasted w i t h Klidsa's compositions), and in addition all the works that might, could, would, or should have been w r i t t e n in strict accordance w i t h it, and not merely " under its influence," if there only were any such. N o w I had myself, in my former paper, laid stress on the difference between the purely hypothetical "grammarians' Sanskrit" and the Sanskrit of the literature; but I never went so far as t o maintain, w i t h Dr. Liebich, that the t w o even belonged to different prime divisions of the whole history of language in India (thus, II. b. grammarians' Sanskrit; III. b. Sanskrit of the literature). Just half our author's volume (pp. 82-161) is occupied by t w o studies which are reckoned as Appendix I and Appendix II. The one is a digest of the teachings of the native grammar (Pnini, the Mahbhsya, and the Ksik) respecting the voice-inflection of the verbal roots, as active or middle or both ; the other is asimilar digest for the formation of feminine declension-stems from the corresponding masculines. These t w o appendixes constitute, in my opinion, the substantially valuable part of the volume; they exemplify what needs t o be done for all the various subjects included in Pnini's treatise. The next step, now, should be t o compare in detail the statements thus drawn out w i t h the actual facts of the language as exhibited in the whole series of monuments of the literature, from Vedic down t o classical and epic, in order t o determine what is the relation between the t w o , and then what the former, the prescriptions of the grammar, are w o r t h ; until that is done, no contribution has yet been made t o our knowledge of the language, but only to our knowledge of Pnini. It casts a shade of unreality over the whole subject of voice-conjugation that the voices of the thousand or twelve hundred false roots are

174 Bruno Liebich

not less carefully defined by the dhtu-ptha than those of the eight or nine hundred genuine ones. There is left for our consideration only the fifth chapter, in which the author takes up and attempts t o answer my o w n objections, given in my paper of nine years ago, t o the confusing of the study of Pnini w i t h that of Sanskrit, and the t h r u s t i n g of the grammarians' dialect into the place in our attention which the real" language of the recorded literature ought t o occupy. I propose t o examine here this reply, and see how effective it is. Dr. Liebich's first point is, as was my o w n , the dhtuptha, or list of roots, which is given as part of the material of the grammar, and really even its foundation, since it is upon them that the rules of the grammar profess t o go on and build up the structure of the languageand that not only grammatically but lexically, for the grammar includes the system of derivation, w i t h definition of the modifications w r o u g h t in each root-sense and stem-sense by the added suffixes. O n this point the author offers a criticism which he is obliged himself t o w i t h d r a w in the next paragraph : he first accuses me of treating Pnini rather unfairly, since the dhtuptha was the part of his w o r k most likely t o be deformed by later c o r r u p t i o n s ; but then allows that I was perhaps (as is indeed plainly the case) criticizing the whole system of the grammarians as it lies before us, of which the list of roots objected t o forms undeniably an inseparable part. Bhtlingk gives it in length and breadth in his recent second edition of Pnini, finding nothing else t o put in its place; and it must have gone hard w i t h him, w h o knows what in Sanskrit is real and what is sham better than almost any other living scholar, and w h o has in the Petersburg lexicons done more than any one else t o make plain t h e i r distinction, t o introduce into his w o r k such a mass of worthless rubbish ; I hardly comprehend how he should have prevailed on himself t o do this w i t h o u t exercising his critical acumen upon it, and separating in some way the false f r o m the t r u e . O u r author talks of probable interpolations, and intimates that he deems t h e m posterior t o the great t r i o of Pnini, Ktyyana, and Patajal, acknowledging that my criticisms may be " more or less" applicable t o t h e i r successors. W e l l , I should t h i n k so; and more rather than less. This free and easy way of disposing of the subject is quite characteristic of the whole guild of partizans of the native grammar. It appears impossible t o bring any one of t h e m t o stand up and face fairly the question of the dhtuptha. There are not far f r o m nine hundred real authenticable roots in Sanskrit. W e could believe that the uncritical interpolations of later grammarians might add t o this number a dozen, o r a score, or fifty, or (to take the extreme) even a hundred o r t w o ; but it is the wildest of nonsense (only strong expressions suit the case) t o hold that they could swell the number t o over t w o thousand ! Such increase is thus far wholly unexplained, perhaps forever unexplainable, and certainly most unpardonable; and until it is in some way accounted for the admirers of the Hindu science of grammar ought t o talk in very humble tones. If these roots are not the ones recognized by the wondrous three, when and under what circumstances and by whose influence w e r e the additional t w e l v e hundred foisted in, t o the abandonment and loss of the old genuine ! st ? The difficulty of explaining this seems not less great than that of supposing the whole t w o thousand as old as Pnini himself; both are hard enough ; and, in either event, t h e taint of falsity attaches t o the Hindu system as we know it and are expected t o use it.

175 Whitney: On Recent Studies

As concerns the three points of the middle periphrastic perfeet, the middle precative, and the secondary passive forms, nothing that the author says tends t o change at all the aspect of the case as stated by me; namely, that these are formations which, though taught by Pnini, are wanting in the traditional literary languageas much so as verb-forms from the thousand and more false roots ; they belong t o the grammarians' Sanskrit alone. Just how much or how little excuse Pnini may have had for setting them up, that is a different and a minor question, t o be decided finally by the general result of our examination of Pan in i 's way of w o r k i n g , of selecting what he will adopt and what he will reject. To me they seem artificial and pedantic structures, reared on an obsolete and insufficient or misapprehended basis. The author's well-intended correction of my estimate of prayoktse in TS. ii 6. 23 as 1st sing. I do not find myself able t o accept. The sentence is not, perhaps, absolutely clear; but the presence in it of ate 'for thee' is t o me a tolerably certain indication that the verb is n o t 2 d sing. ('I will employ t o - m o r r o w for thee at the sacrifice,' or 'at thy sacrifice'); no such possessive would be called for (or admissible, I think) if the person were second. And -tse is obviously the t r u e middle analogue t o active -tsmi, as sse t o ssmi and the like; while -the, as given by the grammarians, is absolutely anomalous, being unsupported, so far as I know, by a single other phonetic fact of the language. That it occurs once (but only once)_in the literature, in that very late Vedic document the Taitt. Aranyaka, whose t e x t is in many parts extremely faulty, is beyond question; but I would put forward the suggestion, as by no means an impossible one, that the form is corrupt, and that the 1 st sing, -the of the grammarians is founded solely on it. That the native commentary, it may be added, explains prayoktse in TS. as 2d sing, is not of the smallest particle of importance; an expositor schooled in Pnini would of course do that, and is capable of doing it against the most incontrovertible evidence t o the contrary. Another matter which the author undertakes t o defend against my objections is Pnini's determination of the cases where dh and where dh is t o be used in t h e 2 d plur. endings dhvam and dhve. He is so far successful that he is able t o show the grammarians' rules t o admit in part a different interpretation from that put upon them by the later Hindu authorities, and reported by the European grammars which follow these rather than the language itself. I was careful t o allow for this possibility in so flagrant a case, putting in the caveat " if the Hindu grammarians are reported rightly by their European pupils (which in this instance is hard t o believe)" ; it now appears that a part of the reproach is capable of being shifted from the shoulders of Pnini to those of his later interpreters. But only a part. Pnini uses in the first of his t w o rules one of his customary algebra-like signs, in, which is ambiguous, signifying either simply the /- and u-vowels, or these together w i t h the r- and /-vowels, the diphthongs, the semivowels, and h. But such an ambiguity is itself a palpable blot upon a system that claims t o beso precise, and Pnini's successors are little t o blame, comparatively, if they have chosen the w r o n g meaning. Then, further, it is and must be equally a matter of uncertainty whether this same in is or is not t o be carried over by implication from the first t o the second rule; and this, again, is a characteristic and a pervading difficulty, running through Pnini's entire w o r k , and, as I said in my former paper, involving " a condemnation of the

176 Bruno Liebich

whole mode of presentation of the system as afailure." W h a t are the boasted terseness and exactness of the rules really w o r t h , when in innumerable cases you cannot tell what they mean without first knowing whatthey o u g h t t o mean ?that is t o say, when an acquaintance with the facts of the traditional language is necessary in order to the right interpretation of the grammar's dictum respecting them ? The present is, at the best, a case where the interpreters have been too careless of the facts and the reasons of the facts. But, whatever improved explanation we may apply t o them, there is plenty left t o object t o in Pnini's rules. The2d pi. precative middle is plainly declared t o end in sldhvam or in sldhvam according t o what letters precede the s (which might also be s); and this is senseless. If the ending is -sldhvam, it is so because the form is originally -sl-s-dhvam, with the special precative sibilant between mode-sign and personal ending, as in 2d and 3d sing., -sl-s-ths and -si-s-ta; if it is, on the other hand, -sldhvam, this is because, as in 1st persons and 3d plur., no such sibilant is present, and the ending is originally -sl-dhvam; and no one can speak with certainty upon the point, because, as I have pointed out, not a single example of the form has been brought t o light out of the literature, earlier or later (the probabilities are altogether in favor of s1-s-dhvam, and so -sldhvam) ; but it is perfectly obvious that what precedes the -si- has nothing t o do with determining the matter, any more than w i t h determining the presence or absence of the precative sibilant in the 2d and 3d singular. It is equally plain that in the indicative of the /s-aorist we must always have dhvam (which the known texts also always give), because -idhvam necessarily results from the combination -is-dhvam, w i t h out any reference whatever t o what may precede the -is- ; and the interpreters must regulate themselves accordingly, if they wish to save Pnini's credit. The author thinks he catches me in an error in saying, as concerns this point, that "all the quotable examples... are opposed t o their rule," and brings up against me astodhvam etc. out of my grammar. But this only shows how carelessly or how unintelIigently he has read my paper; for it is distinctly allowed there that the rule as given applies correctly to the s-aorist, and there is quoted the example anedhvam (from anes-dhvam; by the way, this example and its like seem t o show that in in the rule requires t o be taken in its wider sense) : one of the striking things about the matter was that a prescription suiting well the one aorist had been wantonly extended t o include the other, with which it had nothing t o do, its application giving in every instance a different form from the theoretically correct one found occurring in the literature. But Pnini undeniably takes the perfect also into his rule, making its 2d plur. ending t o be dhve ordhve under the same conditions as those laid down for the aorists. The impropriety of the combination and identical treatment of the t w o tenses is clear. The aorist has always at the end of the stem a lingual sibilant anes-, apavis to exercise its euphonic influence upon the dh of the ending, while in the perfect there is none such. That is to say, none unless the endings dhve and dhvam are really by origin sdhve and sdhvam] and this is adoctrine which has found, and perhaps still possesses, some adherents. But it has no foundation whatever in the actual phenomena of Sanskrit, but solely in these blundering rules of the native grammar. Examples of the 2d plur. perfect, indeed, are of exceeding rarity; I am able at present to point to only a single one (dadhidhve, occurring twice in RV.) in

177 Whitney: On Recent Studies

t h e older language. But, if we are t o recognize sdhve in t h e perfect, we plainly ought t o recognize sdhve and sdhvom also in t h e present ( i n d i e , impv., and opt.) and imperfect; and then we should not meet w i t h forms like studhvam, jnJdhvam, bhavedhvam, akrnud-

hvam, but with studhvam and so on. It appears, then, that the only way to save Pnini's reputation in the matter is to strike the syllable lit (meaning ' perfect') out of his rule, as ungenuine; and I would suggest that it was perhaps intruded by the same cunning hand that thrust into the dhtuptha more than a thousand false roots without being detected or deterred ; this latter trick was evidently by far the harder to execute. But Dr. Liebich finds two other defenses to make (both on p. 58). For one thing, we are not justified in asking for a reason why dhvom should in certain cases be converted into dhvam. "As if," he exclaims, "we were able in any language whatever to trace everywhere the connection of cause and effect! " Begging his pardon, I assert that, on the contrary, in the combinations of stem and ending in Sanskrit euphony, we do not meet with any effect of which we may not look for a cause with good expectation of finding it. If we came anywhere upon adhvam without a discoverable reason, we should question its correctness, and hold it probable that some one had blundered, that the text-tradition was corrupt, or the like. On the other hand, if, as is actually the case, we have no dhvam for which we cannot show a perfectly good reason (few as, unfortunately, the instances are), and no dhve at all, and can put against this only the assertion of Pnini and his successors and interpreters that such forms ought to occur without any reason, I submit that the sole acceptable conclusion must be that these grammarians, like grammarians everywhere else, have blundered, and need to be corrected. Our author's remaining plea is one that, it must be confessed, gives a tinge of the comic to the whole discussion. The difference, he points out, between dh and dh is very slight, and it might be unfair to expect Pnini in every case to distinguish the one correctly from the other! That is to say, if Pnini prescribes adh where there is no ground for one, it may be simply the fault of his ear, which caught the sound wrong. Now I have been accused, by the author and others, of insinuating depreciatory things about Pnini, but 1 certainly never went so far as this. If the great grammarian had too dull an ear to distinguish a lingual mute accurately from a dental (like the typical, or mythical, German, who cannot tell t and d apart), what are all his teachings worth that involve phonetic distinctions?The staff is broken over Pnini, and by one of his own partizans. To conclude (after passing without notice the other points made by me; the most important was the grammarians' derivation of the reduplicated aorist from the causative stem instead of from the root directly), Dr. Liebich takes up my criticism of the Pninean classification of compounds, defending and extolling this classification ; and he returns to the same subject, elaborating his view still further, in the introduction to another later publication. "Two chapters of the Ksik."2 According to him, the true scientific principle of arrangement of compounds, which must be regarded as underlying Pnini's scheme, is furnished by syntactical subordination, after the following fashion : 1. In the copulative
2

Zwei Kapitel der Ksik, Breslau, 1892, 8vo, pp. xl, 80.

178 Bruno Liebich

compounds, as devamanusyas 'gods-and-men,' neither element is subordinated t o the other, but both are coordinate; 2. in the determinatives, the f o r m e r element is subordinated t o the latter, either as a case dependent on it o r as an adjective (or its equivalent) qualifying i t : examples are housetop, redbird; 3. in the possessives, both are subordinated together t o a w o r d outside the compound, which they j o i n t l y qualify in the manner of an adjective: for example, redhead, i.e. redheaded, or possessing a red head ; then, 4. t h e r e remains only one other possibility, namely that the second element should be subordinated t o the first, as in atimatram 1 beyond measure' : we might give as English parallel aboveboard o r overboard (also, for the other Hindu variety, consisting of a participle governing a f o l l o w i n g noun, the English spendthrift or hategood; of this variety our author makes no account, because it is Vedic, and unnoticed by Pnini). If, then, we are t o l d , the subordinated element be represented by a minus-sign, and the o t h e r by ap/us, we get thus the four combinations + + . h . H ; and these evidently exhaust all the possibilities of the case. N o w this is in the real Pninean style, and proves Dr. Liebich t o possess a double p o r t i o n of Pnini's spirit, if he be not the great grammarian himself in the latter's nth metempsychosis. Pnini w o u l d have been proud t o adopt it into one of his chapters, t o gether w i t h its algebraic notation, so akin w i t h his o w n . But our author has t o confess that it is not Pnini's o w n scheme; it is only brought out fully and distinctly by a much later successor. Moreover, that Pnini's f o u r t h class, the so-called avyaylbhva compounds, is by no means limited t o examples of t h e f o r m u l a p / u s minus, but includes a number of quite heterogeneous formations. Dr. Liebich is nevertheless confident that he recognized the unique value of the scheme, and had it plainly in mind ; only he sacrificed it, " perhaps w i t h a heavy h e a r t " (Ksik, p. ix), on the altar o f b r e v i t y ! This brings t o our notice, and in a strikingly illustrative manner, another of Pnini's leading characteristics and at the same t i m e greatest weaknesses. The prime object aimed at by him (as in no small measure in the sutro-style everywhere) is brevity, brevity at the cost of every o t h e r desirable t h i n g o f t h e o r e t i c t r u t h , of connection, and, most of all, of intelligibility.Thequality may be one that recommended his w o r k t o those w h o had t o learn it by rote (though in its degree we have the right t o question even that), but it is very much the opposite of a recommendation t o us, and cannot but detract very seriously f r o m our approval and admiration. A n d this especially when we see how capriciously the principle is appliedhow many rules are squandered on details of the most t r i f l i n g consequence, far below others that are o m i t t e d ; on the quotation of o t h e r grammarians (the best way t o confute w h o m was t o leave t h e m unnoticed) ; on the excerption (in more than 200 rules) of scattered particulars out of the Vedic language, which are valueless because they are merely specimens, making no pretense t o completeness, while the motive of t h e i r selection is in many cases beyond the reach even of conjectureand so on. If the grammar were sharply examined w i t h reference only t o this its leading motive, it w o u l d unquestionably be found t o teem w i t h matter for unfavorable criticism. But t h e r e is another and more fundamental difficulty lying behind Pnini's oversight, o r possible sacrifice, in not recognizing the f o u r t h , the plus-minus, class of compounds in its t r u e character, and thus rounding out a perfect scheme of classification, namely t h i s : t h e r e is no such class; Dr. Liebich and his authorities, the

179 Whitney. On Recent Studies

later Hindu grammarians, are deceiving themselves w i t h afalse determination and notation ; t h e avyayJbhva class, however cornposed, is not plus-minus, but minus-minus. By this is not meant that t h e component parts of such compounds do not stand in a plus-minus relation t o one another; but so also do those of t h e ordinary possessives stand in a minus-plus relation ; and if t h e possessive is nevertheless really a minus-minus compound, so is, for the same reason, t h e avyaybhava. The copulative compound, composed of t w o (or more) nouns o r adjectives, is itself noun or adjective accordingly, and is properly reckoned as plus-pius; the determinative is a noun or adjective w i t h preceding limiting w o r d , and it also is noun o r adjective accordingly, and rightly minus-plus. It is different w i t h t h e possessive, because, though this is not less a noun w i t h a preceding limiting w o r d , it has passed through a transformation making of it an adjective, which is t o qualify something outside: mahbhu when it means 'agreat arm ' is determinative o r minus-plus] but when it means 'having a great a r m ' it is changed t o minus-minus. If w e represent t h e adjectivizing influence by a, we shall get the equation (minus-plus)a = minusminus, which is good linguistic mathematics; at any rate, it is only in such a way that t h e possessive comes t o be a minus-minus compound. But precisely t h e same is t r u e of t h e avyayJbhva. Taking, for example, the participial compound bharad-vasu ' bringing wealth,' w e find it made up of a governing w o r d and its objectnoun ; but it is not therefore a noun ; it has been transformed t o an adjective; its accus. sing, and nom. plur. are not bharantamvasu and bharanto-vasu, but bharad-vasum and bharad-vasavas', it has undergone a similar transformation to that of mahbhu, and it is minus-minus] for its formula is again (plus-minus)a = minusminus. But the proper avyayJbhva is not an adjective, but an adverb ; the ph rase ati mtrm ' beyond measu re ' becomes as a compound atimtram ' excessively.' Here is plainly involved a similar fusion and transfer to that already described; and, if we represent the adverb-making force by 6, the proper formula for atimtram is (plus-minus)b minus-minus. But in real truth atimtram is still further from being aplus-minus compound; for to any one who considers the class historically it must be obvious that any such adverb is simply the neuter accusative of an adjective used adverbially, as neuter accusatives, among simple words and compounds of every kind, are wont to be used. For example, the first step from ati mtrm is the common adjective atimtra 'excessive,' of which the formula is (plus-minus)a] then from this comes by another transfer the adverb, with the formula ((plusminus)a)b, or, more briefly, [plus-minus)ab] and, as the adjective was minus-minus, the adverb is doubly so. Whether this double transfer be accepted or not (of course the acceptance does not imply that some of the adverbs have not been made directly, by analogy with the others of more regular development), the asserted plus-minus class is irretrievably lost, and with it the mathematically exhaustive and regular classification of Sanskrit compounds. It has, indeed, never been found that the facts of language could be reasoned on mathematically; and, whenever the attempt so to treat them is made, we have the right to expect to detect a misapprehension, as in the present case. We may now decline to be touched by the spectacle of Panini's "heavy heart," and hold, on the contrary, that Dr. Liebich has probably done him for asecond time signal injustice, in believing him capable of being deceived by an alluring though false theory. The adjective com-

180 Bruno Liebich

pounds w i t h governing p r i o r member, w h e t h e r this be preposition or participle, are sub-classes, w i t h the possessives, of the great class of secondary adjective compounds, as I have located and described them in my grammar; and the avyaylbhvas are no class of compounds at all, but only a g r o u p in the long list of adjective neuter accusatives used adverbially. It may be f u r t h e r mentioned, as a curiously characteristic point, that our author objects (Ksik, p. x i , note 2) t o the name "possessive" as applied by Bopp and his successors t o the " m u c h r i c e " (bahuvrlhi) compounds, because some of them admit of being fairly rendered otherwise than by ' having' or ' possessing,' and because the Sanskrit has no verb ' have,' and therefore Pnini w o u l d not have cast the sense into this f o r m . Then also, it may be inferred, we are w r o n g t o speak of the " possessive" suffixes in and vont, and t o render balin and bolavant by ' having strength, 1 or t o call madiya ' m y ' a " possessive" pronominal adjective or tasya ' his ' a " possessive" genitive. It may be pleaded in reply that, since we name them in our o w n language and not in Sanskrit, we have every right t o cast t h e i r real and undeniable sense into the form of nomenclature that best suits our expression : and that the Hindus themselves put the idea of possession as well as they can into the definitions of these compounds by t h e i r familiar fo r mu la yasy a . . . sa tathoktah: they say, for example, " w h o s e arms are g r e a t " in place of our " having great a r m s " : and it really seems t o amount t o the same thing. A t the close of his chapter, Dr. Liebich, conceiving himself t o have broken the force of all my objections t o setting Pnini above the Sanskrit literature, and his grammatical science above ours, regrets that I have not brought f o r w a r d a happier selection of t h e m . I, on the other hand, t h i n k myself justified in maintaining that, as they all still stand in full vigor, they are a sufficient illustrat i o n and support of my contrary estimate of the native grammar. But I am willing t o add another point, which he indeed almost forces upon my a t t e n t i o n . A t the very end, namely (p. 61), he lifts up hands of h o r r o r at me (as did Speijer, in his Sanskrit Syntax, p. 189, note) for daring t o stigmatize as a barbarism something which Pnini expressly teaches (his alarm makes him see it as double, or worse than double, and he puts it in the plural, as a t h i n g happening "occasionally"). He ought fairly t o have quoted the case, instead of merely referring t o the rule about it. It is t h i s : Pnini teaches that a comparative and superlative adverbial ending may be added t o a personal verb : thus, dadti ' he gives,' dadtitarm ' he gives more,' dadtitamm ' he gives most.' This is precisely as if one were directed t o say in Greek SiSwmrepov (in this case, even the suffix is identical) and hihojairarov. N o w I maintain, and w i t h o u t any fear of successful contradiction, that such formations, no matter w h o authorizes t h e m , are h o r r i b l e barbarisms, offenses against the proprieties of universal Indo-European speech. The total absence of anything like t h e m , or of anything suggesting even remotely the possibility of forming t h e m , in the pre-Pninean language (one might just as successfully seek for suggestions ofSiSajairepov in Homer or Plato), and t h e i r r a r i t y later (no example of -to ma m is ever met w i t h ) , among w r i t e r s t o w h o m a rule of Pnini is as the oracle of a god, is enough t o show that they never formed any proper part of the language. Probably they were jocose or slangy modes of expression (essentially bhs, but far below the level of decent brioso), which some strange freak, perhaps of amusement at their oddity (and Pnini was entitled t o some compensation for

181 Whitney: On Recent Studies

the " heavy heart" which his subserviency to brevity often cost him), led him to sanctionif indeed the rule permitting them be not another interpolation by that mischief-maker who spoiled the list of roots. Dr. Liebich complains of the (presumably disrespectful) references to "the native grammarians" which he finds too frequent in my Sanskrit grammar, and kindly advises me to cast them all out. But this is in the highest degree unreasonable. Considering the place which those grammarians have long occupied in the study of the language, and the influence allowed them by their European successors, and that their ways of viewing and presenting things have determined in large measure the form of universal Sanskrit grammar, it is simply impossible to leave them out of account and unmentioned. I am sure I have been as respectful to them as I possibly could, and probably in the majority of cases quite successfully at least hypothetical I/ respectful, stating their teaching for what it may be worth, and leaving to the future the final determination of its value. It was hardly respectful for him, on his part, to pronounce (in his closing sentence) all my references to them "extremely superficial and often inaccurate," without quoting a single instance to show that they really bear that character. Perhaps, if he had done so, he would have made as signal a failure of it as he has of the attempt to refute the views and reasonings of my former paper. An extended review of Liebich's Pan in i, by Dr. Franke, is found in the Gott. Gelehrte Anzeigen bearing date of Dec. 1, 1891 (pp. 951-83). It is, however, less a detailed examination and criticism of theformer's views than an independent discussion of some of the points involved, carried on with much learning and acuteness. Many pages are expended upon Pnini's classification of the compounds; and here Dr. Franke is far from supporting Liebich's answer to my criticisms ; on the contrary, he takes my side, setting forth the remarkable superficialities and incongruities of Pnini's work in this department, especially as regards the asserted class of avyayibhvas; he makes many points of detail which I have passed without notice in the above discussion of the theoreticgroundwork of the classification. Though dated in the following year, Liebich's Ksik, and its introduction were doubtless written before the appearance of this review; he would hardly have ventured to repeat his views, or would have cast them into a very different form, if he had had before his eyes their condemnation by afellowpartizan of Pnini. In other points, Franke's notice of Liebich's work is mainly laudatory. Thus, he "thoroughly approves," as "very successful " (p. 962), the latter's futile pleadings as to the ending dhvam (including, I suppose, the suggestion of Pnini's dullness of ear), adding, as his own contribution to the controversy, that adh not seldom takes the place of dh in Prakrit, and that Prakritic changes have been known to work their way into Sanskrit. But what has that to do with Pnini's definite prescription of dh in certain conditions which demonstrably have nothing to do with the matter? So in Prakrit, in obedience to the same general lingualizing tendency, n in the majority of cases becomes n; but that would be far from supporting a Hindu grammarian who should teach that a r altered the next following n to n only provided it were itself preceded by the sounds included in the designation in. As for the great question of the 1200 false roots, Dr. Franke slips smoothly over it, merely echoing the other's remark, that it was an " unfortunate proceeding" on my part to commence from that

182 Bruno Liebich

quarter my attack upon the native grammar. Unfortunate, indeed ; but evidently unfortunate only for the grammar: w h o could help starting f r o m that most flagrant, w a n t o n , and inexcusable of all its many weak sides? It is hardly w o r t h while t o say much more than has been already said w i t h regard t o Liebich's Ksik. It is a laborious and useful c o n t r i b u t i o n t o the study of Pnini himself and of one of the most noted comments upon his w o r k , smoothing a little the way t o t h e i r comprehension for those w h o shall approach it hereafter. The author's method is a narrowly restricted o n e ; the rule in Pnini is given, not translated, and then follows a bald recording of the Ksik's exposition, w i t h here and t h e r e brief notes added on one and another point in the latter; f r o m any attempt at an independent explanation, and yet more f r o m any criticism, the author carefully refrains. Thus, of the rule which introduces the whole subject, samarthah padavidhih, the Ksik gives t w o entirely discordant interpretations, illustrating, however, only the latter of t h e m w h i c h is a very strong indication that the commentators were themselves uncertain as t o what meaning really lay hidden in its obscurity; and the translator passes the matter w i t h o u t a w o r d of remark, nor does it occur t o him t o state w h e t h e r in his opinion we ought t o understand ' a w o r d - r u l e is competent,' o r t o force into the t e x t w i t h e x t r e m e violence the sense ' a w o r d in the following rules is t o be taken in connection w i t h its sense' : it is only an illustration of the ordinary principle that you must first find out what a rule of Pnini ought t o signify, and must then, at whatever cost, i n t e r p r e t that signification into it. A n d the continuation is of apiece w i t h t h e beginning. N o one can well avoid being moved t o repugnance by the fantastic obscurity w i t h which the subject is presented ; and we k n o w already that the underlying t h e o r y , the scheme of distinctions and of classification, is a very defective one. To claim, t h e n , that it must be all labored t h r o u g h by the general body of students of Sanskrit, in o r d e r that they may duly understand the subject of Sanskrit compounds, is obviously unreasonable, not t o say absurd. Pnini and his chief commentators must be w o r k e d oyer by a small class of specialists, and not simply translatedthat is a mere beginning of the taskbut brought into such a f o r m as t o be readily understood and assimilated by the mass of scholars. The study is excessively difficult, and on many of the points involved in it certainly seems unattainable. Dr. Liebich confesses (p. i) that he found the rendering of these t w o l i t t l e chapters so hard that he could scarcely keep his courage up t o complete t h e task. Speijer has been a faithful student of the native grammar; but of the discussions and criticisms of points in it on which he occasionally ventures in his Sanskrit Syntax, Bhtl i n g k ( i n a review of the w o r k in Z . D . M . G . X L M 7 9 f f . ) claims t o refute nearly every o n e ; and now Liebich (Ksik, p. i v ) declares Bhtlingk, in spite of his life-long familiarity w i t h t h e subject and his immense e r u d i t i o n , t o have translated Pnini sometimes incorrectly. Rather discouraging that for a student w h o is ambitious t o get his knowledge of Sanskrit directly f r o m native sources! I would be far f r o m saying anything t o discourage the study of Pnini ; it is highly i m p o r t a n t and e x t r e m e l y interesting, and might fairly absorb much more of the labor of the present generation than has been given t o it. But I w o u l d have it followed in a different spirit and for a different purpose and in a different method. It should be t h o r o u g h l y dissociated f r o m the study of Sanskrit,

183 Whitney: On Recent Studies

though never w i t h o u t recognition of what it may finally contribute t o our knowledge of Sanskrit in addition t o what we derive from the literature. As t o what the literature contains, we need no help from the native grammar; it is the residue of peculiar material that we shall value, and that we should strive t o separate f r o m the mass. A n d the study should be made a t r u l y progressive one, part after part of the native system being w o r k e d out t o the last possible degree and the results recorded, so that each generation be not compelled t o begin anew the tedious and unrewarding task. A t the beginning of the introduction t o his Ksik, it is t r u e , Liebich makes the claim that all Sanskrit students need t o master Pnini, if for no o t h e r reason, because the native commentaries cannot be otherwise completely understood, it being k n o w n that they abound more or less in references t o the grammar and demonstrations founded upon it. There would be more in this consideration if the grammatical discussions were not precisely the most worthless part of the comments, which can be in all cases neglected w i t h least fear of loss. W h a t the words mean, what allusions they contain, what is t o be supplied t o complete the sense, which of possible constructions is the right onethese are matters in regard t o which the aid of the commentator is more or less (in p r o p o r t i o n , namely, t o the artificiality of the composition) welcome, sometimes even indispensable; but for the grammatical forms, the derivations, and everything else that Pnini can be quoted for, the case is different. As f o r Syana and his kind, even those w h o make the strongest claims in his favor will hardly vent u r e t o deny that the whole grammatical part of his exposition might be expunged from his t e x t w i t h o u t loss of a j o t or t i t t l e of its value. It may be added that Dr. Franke also, in the first paragraphs of his review of Liebich briefly examined above, shows the same disposition t o exaggerate and misrepresent the claims of Pnini t o attention. He quotes once more, as Liebich had done before him. Lassen's u n w o r t h y insinuation that Bopp's g r o w i n g independence of Pnini was owing t o his ignorance of him ! As if Bopp did not know Pnini, both at first hand and in his European representatives, sufficiently t o judge w i t h full competence what his system was w o r t h , and how far it required t o be fol l o w e d ! There is quite t o o much of Pnini left still in Bopp's grammar; yet t o Bopp belongs the high credit of making the recorded facts of the language for the first t i m e the basis of t h e i r orderly presentation, and of bringing the principles of European grammatical science, and those of a new and developing comparative grammar, t o bear upon Sanskrit. It is owing t o this that he became the real Sanskrit teacher t o Europe, in a manner and degree far beyond the reach of Lassen. Dr. Franke then goes on t o vindicate for Pnini various things t o which he has not the shadow of a just t i t l e : as, 1. that not only for Sanskrit, but also for other Aryan dialects and writings, Pnini is of indispensable importancewhich apparently means nothing more than that some of the phenomena of dialects later than Sanskrit are t o be found noted in his grammar; 2. that the study of his rules has a formally educating influencewhich is, I t h i n k , just the opposite of the t r u t h , since their method is purely mechanical, sacrificing everything else t o brevity, ignoring connection and p r o p o r t i o n , lacking all recognition of the historical element, and therefore necessarily destitute of philosophy (we have seen above that t o o much Pnini has led Dr. Liebich t o doubt the relation of cause and effect in Sanskrit euphony); 3. that it is

184 Bruno Liebich

Panini w h o has taught us t o regard e v e r / w o r d , every ending, even every letter as i m p o r t a n t w h i c h is an accusation laid w i t h o u t any reason whatever against western grammatical science; and 4. that Pnini is going t o aid literary chronology in a way that is h i t h e r t o for the most part only a matter of conjecture and of f u t u r e hope and which therefore, we may answer, it is as yet t o o early t o say anything about; but, if t h e r e are such treasures hid in Pnini, why do not his partizans devote themselves t o bringing them f o r t h , instead of dwelling upon subjects which are far better understood out of the literature itself? Just f o r t y years ago, a German student of more than ordinary ability, in company w i t h w h o m I had w o r k e d for a season under a professor of the highest eminence in Germany, t o o k the degree of doctor of philosophy creditably w i t h a dissertation on one of Klidsa's plays, and went t o England for f u r t h e r study and for employment. He was f o r t i f i e d , among o t h e r things, w i t h a letter of i n t r o d u c t i o n t o a Sanskrit scholar of German b i r t h , then long resident in London. This scholar, on being consulted in regard t o plans and pursuits, t o l d him that all his h i t h e r t o acquired k n o w l edge had no real foundation, and was essentially worthless; that, if he wished t o accomplish anything, he must d r o p all besides and devote himself for t w o or t h r e e years exclusively t o the study of Pnini ; when that had been done it w o u l d be t i m e t o talk of something else. Just how much this rebuff had t o do w i t h t u r n i n g my friend's attention away t o o t h e r studies I do not k n o w ; but, at any rate, until his death some years after he was not heard of f u r t h e r in Sanskrit. Such was, doubtless in its most intense f o r m , the spirit of the devotees of the native Hindu grammar a generation ago. A n d , though it has been in some measure subdued since, it is by no means extinct, when a man of real learning and ability like Dr. Franke can still maintain (in his Casuslehre, etc., noticed above, p. 68, o r p. 6 of the r e p r i n t ) that o u r profounder knowledge of Sanskrit is t o be especially p r o p o r t i o n e d t o our deeper penetrat i o n into Pnini's teachingsagainst which is t o be set, as antidote, the same author's exposure of Pnini's failure in the article of compounds. It is, of course, much t o the credit of Pnini that he exercises such a bewildering fascination over the minds of those w h o involve themselves in the labyrinth of his rulesthough the influence admits, I believe, of a natural explanation. I am fully persuaded that any one w h o should master the Hindu grammatical science w i t h o u t losing his head, w h o should become t h o r o u g h l y familiar w i t h Pnini and escape being Pnini-bitten, w o u l d be able t o make exposures of the weaknesses and shortcomings and needless obscurities of the grammar on a scale h i t h e r t o u n k n o w n .

Otto ^ Boehtlingk ( 1 8 1 51904)

Otto Boehtlingk, the editor and translator of the Astdhyy] and of Vopadeva's Mugdhabodha, and also the first Western linguist who applied Pninian techniques in the linguistic description of another language (in his ber dieSprache der Jakuten of 1851), contributed to this discussion too. His article "Whitney's letzte Angriffe auf Pnini " deals with Pnini's kraka-theory, the roots of the Dhtuptha, nominal composition, and other topics. This paper is reproduced from the Berichte ber die Verhandlungen der kniglich schsischen Gesellschaft der Wissenschaften zu Leipzig, PhilologischHistorische Ciasse 45, 1893, 247-257. The Dhtuptha was later the subject of special studies by Liebich (for example, Liebich 1919-1920,1921,1930). Nominal composition, which figures so prominently in these discussions, has of course been widely studied; for recent investigations cf. Harweg (1964), and Staal (1966a). The kraka theory has only recently again received much attention (Kocher, 1964a, 1964b; Cardona; 1967; Kiparsky-Staal, 1969).

Whitney's letzte Angriffe auf Pnini(l893)

Otto Boehtlingk

Vor Kurzem erschien im American Journal of Philology, Vol. XIV, No. 2, S. 171-197 ein Artikel von Whitney, betitelt On recent studies in Hindu Grammar, der die grammatischen Schriften Bruno Liebich's und R. Otto Franke's kritisirt und bei dieser Gelegenheit auf die Schwchen von Pnini zu sprechen kommt. Als alter Freund und Bewunderer Pnini's halte ich mich fr berechtigt und einigermaassen verpflichtet, zu seiner Vertheidigung die Feder zu ergreifen. Ob in diesem Versuche meine Vorliebe fr Pnini oder Whitney's Groll auf ihn mehr zu Tage t r i t t , mag der geneigte Leser entscheiden. S. 171 fg. heisst es: Pnini does not take up the cases as forms of nouns, setting forth the various uses of each, after our manner; he adopts the vastly more difficult and dangerous method of establishing a theoretical list of modes of verb-modification by case, or of ideal case-relations (he cal Is them kraka, ' factor ' or 'adjunct'), to which he then distributes the cases. Almost as a matter of course, however, his case-relations or kraka are not an independent product of his logical faculty, but simply a reflection of the case-forms; they are of the same number as the latter, and each corresponds to the general sphere of a case: they are kartar ('actor' = nominative), karman ('act' = accusative), sampradna ('delivery' = dative), karana (' instrument' = instrumental), adhikarana ('sphere' = locative), and apdna (' removal ' = ablative). Wenn man nicht wsste, was kraka bei P. bedeutet, wrde man aus Whitney's beinahe mystischer Definition desselben sich schwerlich eine Vorstellung davon machen knnen. Kraka ist nicht jede case-relation, sondern nur die Beziehung eines Nomens zu einem Verbum, und nicht jedem Casus entspricht, wie Whitney behauptet, eine ihm besonders zukommende Beziehung zu einem Verbum : der Genetiv geht hierbei leer aus, was auch W h . nachtrglich bemerkt. Auch ist W h . im Irrthum, wenn er den kartar dem nominativ gleichsetzt. Das grammatische Subject kennt ja P. bekanntlich nicht, sondern nur das logische und dieses lsst er nur im Instrumental und Genetiv (subjectiven Genetiv) auftreten. Im Nominativ kann, wie auch Wh. erwhnt, sowohl der kartar (wenn die Personalendung diesen ausdrckt, d. i. beim Activum), als auch das karman (wenn das Verbum finitum dieses ausdrckt, d. i. beim Passivum) erscheinen. Mit dieser kraka-Theorie, die schwierig und gefhrlich genannt wird, ist Wh. nicht einverstanden. Die Schwierigkeit und Gefhrlichkeit derselben hat aber den khnen, originellen und genialen P. nicht abgeschreckt. Er hat den Versuch gemacht, und dass dieser ihm gelungen ist, werde ich in der Folge an einem Casusbegriff, am Object, klar zu machen versuchen. Dass P. auf diese Kategorien durch die Casus hingeleitet wurde, versteht sich allerdings von selbst; die Aufstellung logischer Kategorien, denen in der Sprache kein Casus entsprche, wre ein ganz unntzes Bemhen gewesen. Es freut mich zu sehen, dass auch Delbrck, dem doch Niemand eine Voreingenommenheit fr Pnini nachsagen oder Schrfe des Verstandes absprechen wird, in seiner so eben erschienenen Vergleichenden Syntax der indogermanischen Sprachen, Th. 1, S. 172 fgg., Pnini's Definitionen der Grundbegriffe der Casus seinen Lesern mitzutheilen fr gut findet und ihnen seine Anerkennung nicht vorenthlt. So heisst es S. 172: An dieser Stelle glaube ich die Sache am besten zu frdern, wenn ich von den Aufstellungen der indischen Grammatik ausgehe, welche sich durch die Schrfe der Fassung vorteilhaft von demjenigen unterscheiden, was in unserer europischen Tradition Gestalt gewonnen hat, und S. 175: Indem Pnini so den Begriff,

187 Whitney's letzte Angriffe auf Pnini

nicht die Kasus zum Eintheilungsgrund macht, erreicht er den Z w e c k seiner Darstellung in hchst vollkommener Weise. A u f S. 172 hlt W h i t n e y ein strenges Gericht ber die Definit i o n des karman, des Objects. As for t h e definitions of t h e caserelations, it may suffice t o say that t h e karman is described as belonging, first, t o that which t h e actor in his action especially desires t o obtain or attain (as in ' he makes a mat', ' he goes t o t h e village'); or, second, t o that which, though itself undesired o r indifferent, is connected w i t h t h e action in a similar manner. A n y t h i n g more crude o r unphilosophical than this could not well be imagined. There is not an identity between t h e use of a given case and t h e presence of its generally corresponding case-relation, because, f o r example, in a passive sentence, as ' t h e mat is made by h i m ' , mat is still called karman or ' a c t ' , though nominative, and him still kartar or ' a c t o r ' though instrumental. Thus t h e r e is no recognition of t h e grammatical category of subject of a v e r b ; and this leads, as could not be helped, t o numerous obscurities and difficulties. Ich gestehe, dass dieses ungerechte und in so schroffer Form ausgesprochene U r t h e i l mich geradezu verblfft hat. Was hat nun Pnini in W i r k l i c h k e i t verbrochen ? Er ist nicht wie unsere Grammatiker von den Casus ausgegangen, sondern von den Casusbegriffen, und hat das Object auf eine originelle und scharfsinnige Weise zu defmiren versucht. Nach dieser Definition musste er in einem Satze w i e tena katah kriyate den Nominativ katah Object und den Instrumental tena Agens benennen, und jeder Unbefangene w i r d zugeben mssen, dass katah in W i r k l i c h keit ein logisches Object und tena ein logisches Subject ist. Dafr, dass das logische Object nicht in den Accusativ, und das logische Subject nicht in den Nominativ gesetzt w u r d e , sorgte Stra 2, 3 , 1 , und dafr, dass das logische Object im Nominativ erschien, 2, 3, 40. Zu diesem letzten Stra vgl. Z D M G . 4 1 , S. 179 fg. Ich sehe also keine numerous obscurities and difficulties. N u n w i l l ich versuchen darzuthun, dass der Gedanke, vom Object und nicht vom Accusativ auszugehen, ein beraus berecht i g t e r und geradezu nothwendiger war. In W h i t n e y ' s Gr. 2 269 lesen w i r : The accusative is especially t h e case of t h e direct object of a transitive verb. Dieses verstehen w i r Alle, weil w i r schon als Schulknaben aus einer lateinischen oder einer anderen G r a m m a t i k gelernt haben, was ein Object und was ein transitives Verbum ist. W e n n aber Pnini etwa gesagt htte: dvitly nedisthann (oder Jpsitatamam) karma sakarmakasya dhtor caste, so wrde ihn Niemand verstanden haben, da es keine Grammatik fr Schulknaben gab, aus der man die Bedeutungen von karman und sakarmako dhtuh htte ersehen knnen. Pnini war also genthigt, zunchst den Grundbegriffeines zu einem Verbum in Beziehung stehenden Accusativs, das karman, zu defmiren. Der Accusativ spielt unter den obliquen Casus die Hauptrolle; daher wurde sein Grundbegriff zuerst bestimmt. Nun mussten consequenter Weise auch die brigen Beziehungen eines Nomens zu einem Verbum errtert verden. Dass Pnini den Nominativ pratham benannte, berechtigt uns wohl zu dem Schlsse, dass er auch ein Verstndniss fr die Wichtigkeit des grammatischen Subjects hatte. Wenn Whitney in der oben aus seiner Grammatik citirten Stelle nach verb noch hinzufgt: and of any word qualifying that object, as attribute or apposition or objective predicate, so ist dieses ein recht mssiger Zusatz, da die Congruenz eines Attributes u.s.w. auch bei anderen Accusativen und den brigen Casus stattfindet. Solche Schwchen wird man bei Pnini nicht finden.

188 Otto Boehtlingk

Sehr schlecht zu sprechen st W h . auf den D h t u p t h a u n d gewiss m i t einigem Recht, da hier Manches v e r d c h t i g ist. W h . b e r t r e i b t aber d i e Sache ein w e n i g . S. 183 sagt er, dass ber 2000 W u r z e l n aufgezhlt w r d e n , v o n denen nach S. 182800 bis 900, nach S. 183 aber not far f r o m nine h u n d r e d echt seien. D i e Zahl d e r untergeschobenen Wurzeln soll nach S.183zwlfhundert betragen. Nach meiner Zhlung enthalten die zehn Klassen 1959 (also beinahe 2000) Wurzeln ; wenn man aber die vollkommen gleichlautenden, in verschiedene Klassen und gruppen vertheilten Wurzeln einfach zhlt, so erhalten wir ungefhr 1770. Von diesen wird aber noch eine bedeutende Anzahl auszuscheiden sein, da die offenbar nur orthographisch von einander abweichenden Wurzeln (wie lad und lal) und die nur durch Hinzufgung eines Nasals sich unterscheidenden (wie ac und ac) bei meiner Zhlung als verschiedene Wurzeln figuriren. Hierzu kommt noch eine Anzahl von Denominativen (wie kumr), die Wh. bei seiner Zhlung der echten Wurzeln wahrscheinlich ausgeschieden haben wird. Die Hunderte von verdchtigen Wurzeln werden theils Abschreiber, theils Erklrer schlechter Texte zu verantworten haben. Hat doch auch der gewissenhafte und kritische Wh. in seinem Wurzelverzeichnisse neue, hchst verdchtige Wurzel- und Verbalformen verzeichnet; vgl. ZDMG. 39, 532 fgg. Auch wrde man, wenn man sich einige Mhe geben wollte, eine nicht geringe Anzahl sogenannter Wurzeln (wie z. B. kusm) auf ihren Ursprung zurckzufhren vermgen und ihnen vom indischen Standpunkte eine berechtigte Existenz zuschreiben mssen. An, so zu sagen, aus den Fingern gesogene Wurzeln glaube ich nicht. Wozu sollte das geschehen sein ? Etwa um mit dem Reichthum der Sprache gross zu thun ? Liebich hatte die Ansicht Whitney's, dass prayoktse TS. 2, 6, 2, 3 die erste Person sei, bekmpft und daraufhingewiesen, dass die von Pnini gelehrte Form auf -the im Taitt. Ar. vorkomme, was brigens auch W h . nicht entgangen war. Hierauf antwortet Wh. S. 184: The author's well-intended correction of my estimate of prayoktse in TS. 2, 6, 2, 3 as 1st sing. I do not find myself able to accept. The sentence is not, perhaps, absolutely clear; but the presence in it of a te 'forthee' is to me a tolerably certain indication that the verb is not 2d sing. (' I will employ tomorrow for thee at the sacrifice', or 'at thy sacrifice') ; no such possessive would be called for (or admissible, I think) if the person were second. And -tse is obviously the true middle analogue to active-tsm/, as ease to ssmi and the like; while -tohe, as given by the grammarians, is absolutely anomalous, being unsupported, so far as I know, by a single other phonetic fact of the language. That it occurs once (but only once) in the literature, in that very late Vedic document the Taitt. Aranyaka, whose text is in many parts extremely faulty, is beyond question ; but I would put forward the suggestion, as by no means an impossible one, that the form is corrupt, and that the 1st sing, -the of the grammarians is founded solely on it. That the native commentary, it may be added, explains prayoktse in TS. as 2d sing, is not of the smallest particle of importance; an expositor schooled in Pnini would of course do that, and is capable of doing it against the most incontrovertible evidence to the contrary. Whitney hat hier und auch schon frher auf eine scharfsinnige und berzeugende Weise dargethan, dass prayoktse eine 1. Sg. sein knne. Aber vom sein Knnen bis zum Sein ist noch ein weiter Sprung. In der Stelle, welche Wh. fr seine Ansicht geltend macht, knnte, wenn nicht die Grammatik dagegen Einsprache erhbe, prayoktse eine 1. Sg. sein, braucht es aber nicht zu sein und ist es

189 Whitney's letzte Angriffe auf Pnini

auch nicht. Dass der Commentar die Form als 2. Sg. deutet, hat auch in meinen Augen kein Gewicht, wohl aber der Umstand, dass er bei seiner Auffassung an te, das er durch tvadlye umschreibt, keinen Anstoss n i m m t , da dieses nicht an das grammatische, sondern an das reale Gebiet streift. Der Opferer und der A n w e n d e r einer bestimmten Formel knnen, brauchen aber wohl nicht verschiedene Personen zu sein. W e n n demnach te fehlte, w r d e man nicht wissen, von wem das Opfer veranstaltet w r d e ; mit anderen W o r t e n : te ist nicht pleonastisch. W e n n aber W h . meinen sollte, dass in einem solchen Falle nicht te, sondern svasmin stehen msste, so htte ich weder fr noch gegen seine Behauptung etwas Entschiedendes vorzubringen. Nach meinem Gefhl sind beide Ausdrucksweisen zulssig. In Delbrck's Altindischer S/ntax finde ich S. 208 Folgendes: In P. w i r d , so viel ich sehe, sv ebenso gebraucht wie in V. Es bezieht sich natrlich in der Mehrzahl der Flle auf d r i t t e Personen, fr eine Beziehung auf eine zweite Person habe ich (wohl zufllig) keinen Beleg n o t i r t . Da prayoktse an und fr sich nach den indischen G r a m m a t i k e r n und nach W h . auch 2. Person ist, und da W h . selbst, wie aus seinen oben angefhrten W o r t e n zu ersehen ist, nicht mit absoluter Gewissheit behauptet und behaupten kann, dass die Form in TS. die erste Person sein msse; so ergbe sich, wenn man seiner, ihm sehr wahrscheinlich erscheinenden Erklrung beistimmte, dass der in Frage stehende Satz sowohl ich w e r d e bei deinem Opfer anwenden, als auch du w i r s t bei deinem Opferanwenden bedeuten knnte. Ist dieses wohl wahrscheinlich? Nun komme ich auf die von Pnini gelehrte, von W h . aber mit aller Entschiedenheit zurckgewiesene Form -the fr die 1. Person zu sprechen. Diese komme, wie er angibt, nur einmal in dem schlecht berlieferten Taitt. Ar. vor und sei falsch, und auf dieser falschen Form im Taitt. A r . beruhe Pnini's -the. Damit ist implicite ausgesprochen, dass Pnini und auch seine Vorgnger nie in den Fall gekommen wren, in ihren tglichen Gesprchen und Disputationen mit Ihresgleichen d i e l . Sg. Med. des periphrastischen Futurum anzuwenden. Credat Judaeus Apella! Htten sie aber eine von der berlieferten verschiedene Form f r die 1. Person gehabt, so w r d e Pnini, da er sie doch nicht durch einen Machtspruch aus der W e l t htte schaffen knnen, sie uns mitget h e i l t und die im T. A. entdeckte neue Form als vedische erwhnt haben. Nach meinem Dafrhalten mssen w i r der Form schon ihres absonderlichen, unserer Deutung sich entziehenden, aber keineswegs deshalb verdchtigen h wegen ein hohes A l t e r zuschreiben. Vielleicht, aber auch nur vielleicht, steht he zu svahe des Duals und svahe des Plurals in nherer Beziehung. Die 1. und 2. Person durften nicht zusammenfallen. S. 184 fgg. erhebt W h . eine neue Anklage gegen P. und zwar wegen des Stra 8, 3, 78 fg., w o von der Cerebralisirung des dh in den Personalendungen dhvam und dhve die Rede ist. Ich gebe gern zu, dass P. sich hier undeutlich ausdrckt, und dass fr die Cerebralisirung im Perfect sich kein Grund nachweisen lsst, dass sie wissenschaftlich nicht haltbar ist. Muss aber dafr gerade P. verantw o r t l i c h gemacht werden ? Ist es nicht denkbar, dass P. einen in der Sprache eingerissenen Fehler, den er nicht erkannte, wohl auch nicht so leicht wie W h . erkennen konnte, einfach verzeichnet htte? Es giebt doch in allen Sprachen falsche Schreibarten, die sich leicht ausmerzen Hessen und doch nicht ausgemerzt w e r d e n . Schreibt doch auch W h . author, obgleich er ganz gewiss weiss, dass das h hier keine Berechtigung hat; ob er auch posthumous schreibt,

190 Otto Boehtlingk

wage ich nicht zu behaupten, wohl aber, dass angesehene Lexicoraphen dieses t h u n . Ueber alle Maassen ungehalten w i r d W h . S. 192 darber, dass P. lehrt, die adverbialen Suffixe tarm und tamm w r d e n auch an ein Verbum fin itum gefgt. This is precisely as if one were d i rected t o say in Greek StScuairepov (in this case, even the suffix is identical) and hihmraTov. Ist dieses etwa ein A r g u m e n t gegen P. ? Ist nicht jedes Glied einer Sprachfamilie erst dadurch zu einer besonderen Sprache geworden, dass es sich selbstndig und eigent h m l i c h vernderte und entwickelte? W h . fhrt f o r t : N o w I maintain, and w i t h o u t any fear of successful contradiction (schreckt mich nicht ab), that such formations, no matter w h o authorizes t h e m , are horrible barbarisms, offenses against the proprieties of universal Indo-European speech. Ist nicht jede Analogiebildung von Hause aus ein Barbarism us, der aber als solcher nicht empfunden w i r d , sonst w r d e er nicht W u r z e l fassen knnen ? Den Barbarismus entdecken nachtrglich die Sprache kritisirende Gelehrte, die aber zum Glck bei der Bildung und V e r v o l l k o m m n u n g einer Sprache nicht mitzusprechen haben. Kann man sich wohl rgere Barbarismen denken als njrrepo, mecTepo u.s.w., w o ep als Theil des Suffixes a u f t r i t t , whrend es in W i r k l i c h k e i t aus dem vorangehenden neTBepo herbergenommen ist, w o es zum Stamme des Zahlworts gehrt? A n diesen Bildungen hat jedoch kein Russe bis zum heutigen Tage Anstoss genommen. W e i t e r heisst es bei W h . : The total absence of anything like t h e m , or of anything suggesting even remotely the possibility of forming t h e m , in the pre-Pninean language (one might just as successfully seek for suggestions of BiSojairepov in H o m e r or Plato), and t h e i r rarity later (no example of-tamm is ever met w i t h ) , among w r i t e r s t o w h o m a rule of Pnini is as the oracle of a god, is enough t o show that they never formed any proper part of the language. Probably they were jocose or slangy modes of expression (essentially bhs, but far below the level of decent bhs), which some strange freak, perhaps of amusement at t h e i r oddity (and Pnini was e n t i t l e d t o some compensation for the heavy heavy heart 1 ) which his subserviency t o brevity often cost him), led him t o sanctionif indeed the rule p e r m i t t i n g them be not another interpolation by that mischief-maker w h o spoiled the list of roots. Ich spreche W h i t n e y jede Berechtigung ab, sich ber gut beglaubigte Erscheinungen, die er sich recht zu erklren vermag, mit solcher Geringschtzung auszusprechen und ihre Existenz sogar in Frage zu stellen, oder sie hchstens in der niedrigen Sprache gelten zu lassen. Weiss W h . berhaupt von der Bhs mehr, als er durch Pnini erfhrt, und mit welchem Recht hat er vor dieser so wenig Achtung? W e n n sogar Schriftsteller nach Pnini, die doch eine von ihm gegebene Regel, wie W h . sich ausdrckt, als das Orakel eines Gottes betrachteten, sich der in Rede stehenden Steigerung so selten bedienen, so kann daraus zunchst nur geschlossen werden, dass sich in ihren Schriften keine Gelegenheit dazu bot, nicht aber, dass dieselbe niemals any proper part of the language gebildet htte. Vielleicht (aber auch nur vielleicht) waren auch P. keine Formen auf tarm und tamm in der Literatur bekannt, w i r d er aber nicht, wie ich schon oben bei einer anderen Gelegenheit bemerkte, in Gesprchen und Disputationen Gelegenheit gehabt haben, sich in der von W h . gergten Weise auszudrcken ? W o h l zum grossen Entsetzen W h i t n e y ' s wage ich sogar zu behaupten,
1

Ein von Liebich gebrauchter Ausdruck.

191 Whitney's letzte Angriffe auf Pnini

dassdie so heftig angefochtene Steigerung eines Verbi finiti gar kein arger Barbarismus ist. Die vollwichtigen, leicht anfgbaren und leicht ablsbaren Adverbialsuffixe tarm und tamm empfand der Inder beinahe als selbstndige Wrter, die er an fertige Indeclinabilia anzufgen gewohnt war; vgl. atitarm, abhitarm, jyoktamm, natarm, notamm, nitarm, pratarm, pratamm, vitarm, samtarm, sutarm, uccaistarm, uccaistomm, nlcaistarm. Ist der Sprung von diesen Adverbien zu pacotitarm u.s.w. etwa gar zu khn ? Zu Gunsten meiner Ansicht kann ich auch hier den besonnenen und nie sich berhebenden Delbrck anfhren; vgl. Vergleichende S/ntax der indogermanischen Sprachen, Th. 1, S. 624, N. Was wrde Wh. erst zu den russischen Ha-Te da habt i h r ! und Hy-Te macht end lieh f o r t ! sagen? Hier ist eine Personalendung mit einer Interjection verbunden worden! Ich mchte die Bezeichnung Barbarismus auf solche Formen beschrnken, die den allgemeinen Gesetzen der Grammatik widersprechen, nur in schlecht berlieferten Texten vorkommen und von keinem alten Grammatiker erwhnt werden. S. 187-191 finden w i r eine recht unerquickliche Polemik gegen Liebich's und der indischen Grammatiker Theorie der Composita, auf die ich nicht nher eingehe. Pnini's Dvigu und Avyaybhva, auf die W h . bekanntlich nicht gut zu sprechen ist, werden hier nicht wieder berhrt; es sei mir aber gestattet, dieselben bei dieser Gelegenheit zur Sprache zu bringen und die Auffassung derselben bei Pnini und Whitney zu vergleichen. Sehen wir uns 13111314 der Whitney'schen Gr. an, so finden wir die, zu den zwei oben genannten Compositis gehrigen Beispiele zusammengestellt und zwar in zwei Abschnitten, von denen der eine Adjective Compounds as Nouns and as Adverbs, der andere Anomalous Compounds berschrieben ist. Also auch wie bei Pnini von den brigen Compositis geschieden, aber theilweise auch auf Composita ausgedehnt, die Pnini nicht dazu rechnet. Whrend bei P. der Dvigu auf das Genaueste defmirt und umgrenzt wird, erblicken wir bei W h . in 1312 eine sehr mangelhafte Definition desselben, und doch soll es der Dvigu Pnini's sein. P. lehrt, dass der Dvigu ein Karmadhraya sei und als Substantiv und Adjectiv erscheine, als Substantiv n. oder f. (/") Sg. in der Bedeutung eines Collectivs. als Adjectiv nicht in der Bedeutung eines Bahuvrhi (also trimrdhan dreikpfig und hnliche Composita keine Dvigu), sondern in den verschiedensten Bedeutungen, die sonst nur durch Anfgung von taddhita's erzielt werden. Dieses st zwar nicht wissenschaftlich ausgedrckt, besagt aber doch das, was wir wissen sollen. Hren wir nun, was W h . in seiner Gr. vortrgt. 1311, der beide Arten von Compositis charakterisirt, lautet: Compound adjectives, like simple ones, are freely used substantively as abstracts and collectives, especially in the neuter, less often in the feminine; and they are also much used adverbially especially in the accusative neuter. Ich kann mich auf kein Adjectiv besinnen, das als Neutrum zu einem Nom. abstr. geworden wre. Die neutralen Participiaauf ta, wie hasitam, werden wohl anders zu beurtheilen sein : diese stehen zu dem impersonal gebrauchten Verbum fin. in nherer Beziehung. Wenn W h . aus substantivirten neutralen Adjectiven Collectiva entstehen lsst, so identificirt er Gattungsbegriffe mit Collectiven. Das neutrale Adjectiv kann als Substantiv, ebenso wie jedes andere Substantiv als Gattungsbegriff verwendet werden, ist aber darum kein eigentliches Collectiv. Ein grammatisches Collectiv entsteht, wenn zwei oder mehr Dinge begrifflich und sprachlich als Ein-

192 Otto Boehtlingk

heit zusammengefasst werden, und ein solches Coliectiv w i r d im folgenden behandelt. Dieser lautet: The substantively used compounds having a numeral as p r i o r member, along w i t h , in part (was soll sich der Leser dabei denken ?), the adjective compounds themselves, are treated by the Hindus as a separate class, called dvgu.tryug, triyojan, tridiv, trilok, trimllsollen also

substantivirte Adjectiva sein ! Ich mchte gern wissen, welcher Sanskritist sich mit dieser Auffassung einverstanden erklren mchte. Whitney hat den Pnini verbessern wollen, hat ihn aber in Wirklichkeit verballhornt. In 131 3 wird der AvyayTbhva behandelt. Auch hier ist die Zurckfhrung des Adverbs auf ein Adjectiv ein wenig gewagt, da ein solches bei den von P. gelehrten Avyayibhvagar nicht nachzuweisen ist. Whitney hat sich bei der Bewltigung des schwierigen Kapitels der Composita grosse Mhe gegeben und Manches, aber nicht Alles, richtiger untergebracht; es leidet aber seine Darstellung an unntzen Wiederholungen und an Erwhnungen und Definitionen eines Compositums an einem ganz falschen Orte. So ist in Chapter V, das von der Declination der Nomina und Adjectiva handelt, die in 323 fg. gegebene Definition zweier Arten von adjectivischen Compositis gar nicht am Platze. Diese brauchten berhaupt nicht erwhnt zu werden, da es bei der Declination gar nicht darauf ankommt, ob ein Adjectiv einfach oder zusammengesetzt ist. In 1294, b werden die adjectivischen Dvigu unter den Bahuvrhi eingereiht; wozu brauchten sie noch einmal 1312, b erwhnt zu werden ? Die Collectiva htte Wh. unter seine Descriptive Compounds2 unterbringen knnen, Hier htte er auch erwhnen knnen, was Pnini lehrt, aber sich nicht von selbst versteht, dass Composita wie triyojanni und trilokh nicht gestattet seien, sondern nur triyojanam und trilokJ, oder nicht com pon i rt trlni yojanni und trayo okh. Das fehlerhafte dasavarsasahasrni trifft man bisweilen an. In 1313, d sind yathkrln und yathcrin an einen unrechten Ort gekommen, da sie mit yathkma und yatkkratu nur das mit einander gemein haben, dass sie auch mit yath beginnen. Man kritisire Pnini, aber auf eine gerechte und urbane Weise. Man verliere indessen nie aus dem Auge, dass dieser geniale Mann bei der Abfassung seiner Grammatik gar kein anderes Ziel verfolgte, als seinen Standesgenossen das Verstndniss der hm bekannten Literatur zu erleichtern und sie zu lehren, wie man in gebildeter Gesellschaft zu reden (darauf htte Wh. mehr achten sollen) und mustergltige Werke abzufassen habe. Diesen Zweck konnte er bei der in Indien herrschenden Methode des Unterrichts nicht anders erreichen, als dass er ihnen ein zum Auswendiglernen geeignetes mglichst kurzes Lehrbuch bearbeitete. Aber er hat auch mehr erreicht: ohne ihn wren unsere Grammatiker3 und Sprachvergleicher gewiss noch nicht da angelangt, wo wir sie heut zu Tage finden. Also Ehre seinem Andenken !
Der indische terminus technicus karmadhraya ist, so viel ich weiss, bisher noch nicht erklrt worden. Ich vermuthe, dass das Wortein Object schuldend d. i. kein Object aufzuweisen habend bedeutet, und dass Pnini damit habe sagen wollen, Composita wie krtakarman, die hier und da wohl als Karmadhraya vor2

kommen, seien nach seiner Meinung nur als Bahuvrhi zu verwenden, 3 Htten, um nur einen Fall zu erwhnen, unsere Grammatiker den Compositis wohl eine solche Aufmerksamkeit geschenkt, wenn ihnen nicht Pnini den Weg dazu geweisen htte ?

Georg Blihler (18371898)

Georg Bhler, a pupil of Benfey's, was no expert on the Sanskrit grammarians, but he contributed to many branches of Indology and, in particular, to the study of Pali and Prakrit. He refuted Whitney with arguments of a different kind, showing that many of the roots of the Dhtuptha, though not found in extant Sanskrit literature, had to be postulated anyway in order to account for Pali and Prakrit roots derived from them. Since many of Bhler's brilliant demonstrations would take us far beyond the scope of the present volume, only some fragments will be included here from "The Roots of the Dhtuptha not found in Literature, " Wiener Zeitschrift fr die Kunde des Morgenlandes (8,1894,17-42), also published in Indian Antiquary (23,1894,141-154 and 250-255).

The Roots of the Dhatupatha Not Found in Literature (1894)

Georg Bhler

In his " Review of Recent Studies in Hindu Grammar" which fills pp. 171-197 of the fourteenth volume of the American Journal of Philology, the continuation of an article on Hindu and European Grammar in the fifth volume of the same periodical, the late Professor Whitney reopens the discussion of a question, which used to sorely vex the soul of the Sanskritists of the last generation, but has since been dropped in Europe, because the progress of indo-Aryan research has shewn very clearly what the solution of the problem is. Professor Whitney, engrossed with his Vedic studies, does not seem to have noticed the labours of the Prakritists. He informs us on p. 182 that there are in the Dhtupthaa "thousand or twelve hundred false roots," and declares that the fact of their "voices being not less carefully defined by the Dhatupatha than those of the eight or nine hundred genuine ones casts a shade of unreality over the whole subject of voice-conjugation." On the next page he condoles with Geheimrath von Bhtlingk, who, in his second edition of Pnini, has given "the whole Dhatupatha in length and breadth, finding nothing else to put into its pjace," though he ought to have known better. Next he severely blames Dr. Liebich, who "talks of probable interpolations and intimates that he deems them posterior to the great trio of Pnini, Ktyyanaand Patajali, acknowledging that his (i.e., Professor Whitney's) criticisms may be more or less applicable to their successors." Turning finally to the Sanskritists of the modern school in general, he throws down the gauntlet to them and winds up with the following peroration : "This free and easy way of disposing of the subject is quite characteristic of the whole guild of partizans of native grammar. It appears impossible to bring any one of them to stand up and face fairly the question of the Dhatupatha. There are not far from nine hundred real and authenticate roots in Sanskrit. We could believe that the uncritical interpolations of later grammarians might add to this number a dozen, or a score, or fifty, or (to take the extreme) even a hundred or two; but it is the wildest of nonsense (only strong expressions suit the case) to hold that they could swell the number to over two thousand. Such increase is thus far wholly unexplained, perhaps for ever unexplainable, and certainly most unpardonable; and until it is in some way accounted for, the admirers of the Hindu science of grammar ought to talk in very humble tones. If these roots are not the ones recognized by the wondrous three, when and under what circumstances and by whose influence were the additional twelve hundred foisted in, to the abandonment and loss of the old genuine I ist ? The difficulty of explaining this seems not less great than that of supposing the whole two thousand as old as Pnini himself; both are hard enough ; and in either event the taint of falsity attaches to the Hindu system as we know it and are expected to use it." Professor Whitney's grievances are therefore: (1) against "the guild of the admirers of Hindu grammar" that they will not to use with Professor Whitney the language of the prize-ring come up to the scratch and fully discuss his objections to the Dhatupatha, though they do answer his strictures on other and less important points : (2) against the Hindu grammarians that their Dhatupatha contains a very large number of verbs, which are not traceable in the accessible Sanskrit literature and which therefore must be "sham" and "false," i.e., if I understand Professor Whitney rightly, inventions either of Pnini or of his successors. If I venture to offer some remarks on the points, raised by the

195 Roots of the Dhtuptha

illustrious Praeceptor Columbiae, my object is t o suggest a definite line of enquiry, which, I think, may lead t o tangible results, valuable alike for Sanskrit and comparative philology, and t o add some practical proposals. In doing so, I must premise that I do not belong t o any guild of partizans of the Vykarana (if such a one exists). Eighteen years of personal intercourse w i t h t h e Hindus have taught me at least something about their many excellent qualities and their weaknesses, which are all clearly discernible in their system of grammar. It shews their great acumen and their pedantry, their laboriousness and their practical sense as well as their feebleness in t h e struggle after an ideal, which is much t o o high for their strength, i am even ready t o bel ieve w i t h the great Mmsrhaka Bhatta, that the Hindu grammarians occasionally resemble " horsemen w h o forget the existence of their steeds." 1 But, strong language on the part of a European or American authority, however great, is insufficient t o persuade me that the Hindu grammarians have invented forms or roots. Such an assertion I could believe only on the evidence of stronger proof than the fact that one, o r a dozen, o r even a score, of scholars cannot find the forms taught. Until that has been furnished, I prefer t o adhere t o my own opinions, which in t h e main coincide w i t h those of Professors Westergaard and Benfey. I must also express my doubts regarding t h e desirability of the use of strong language, in this case and in all other scientific discussions, both f o r personal reasons and out of regard f o r o u r special branch of learning. Professor Whitney's first complaint seems t o me well-founded. I likewise regret that the specialists in Hindu grammar and particularly the able pupils, whom Professor Kielhorn has trained, hitherto have not turned t o the Dhtuptha, and have not availed themselves of the plentiful materials which are ready at hand in order t o carry on and t o supplement the w o r k , begun in so masterly a manner by Professor Westergaard. Since the times of the great Dane the critical treatment of Pnini's Straptha has been begun, and perfectly t r u s t w o r t h y critical editions of the Vrttikas and of their great Commentary, as well as of the Ktantra, have been published. The Paribhss, which are the key t o t h e whole system of Hindu grammar, have been so excellently translated and so carefully illustrated by Professor Kielhorn, that even a beginner may understand their application. The Ksik together w i t h its huge V r t t i , t h e Padamajar of Haradattamisra, Kaiyata's Pradpa, a number of Ngoj's and Bhattoj's grammatical treatises, Bhartrhari's VkyapadTya, Syana-Mdhava's D h t u v r t t i , Skatyana's grammar and t h e Srasvata have at least been printed, be it in their entirety o r in part. And for those, w h o desire t o critically examine these works, there are good old MSS. in t h e public libraries of India, which t h e liberality of the Indian Governments makes accessible t o all Sanskrit students. Finally, t h e Grammars of Candra, Jinendra-Pjyapda, Buddhisgara, Malayagiri and Hemacandra have been recovered in MSS., mostly together w i t h their Angas, as well as Jinendrabuddhi's Ksikvrttinysapanjik, and an apograph of Syana's Dhtuvrtti is lying in t h e library of Elphinstone

Tantravrttika, p. 201, 11. 3-4 (Benares edition):stravrttikabhsyesu drsyate cpasabdanam/osvrdhh katham csvnvismareyuh sacetanhj The sermon, in which Kumrila ex-

pands this text, is highly edifying, and the best Vaiykaranas living have admitted to me that the charges made there are not unfair.

196 Georg Bhler

College, Bombay, which has been transcribed f r o m a MS. (at Nargund), dated w i t h i n a hundred years of the author's time. 2 W i t h these materials, which mostly were not accessible t o Professor Westergaard, or only so in indifferent modern MSS., it is possible t o settle the following points : 1. W h i c h portions of our Dhtuptha were certainly known t o Panini and the o t h e r t w o Munis. 2. W h e t h e r any additions have been made by the later authorities of Pnini's school, Vmana, Jayditya, Jinendrabuddhi and so f o r t h , and w h a t has been added by each. 3. W h a t our Dhtuptha, or the list of verbs in the D h t u v r t t i , owes t o the homonymous treatises of Sarvavarman, Candragomin and the other authors of independent Sabdnussanas. Much of the work that Bhler here suggested be done for the Dhtuptha was actually done later by Liebich in various publications (e.g., 1919-1920,1921, and 1930). A little later Bhler continues as follows:

Turning to Professor Whitney's grievance against the Hindu grammarians, his assertion that they have inserted "false," "sham," or " fictitious" forms in the list of verbs, which, as is acknowledged at all hands, has an intimate connection with their Sabdnussana, is supported in his present paper by the sole argument that he cannot find the verbs, their inflexions and meanings in the literature accessible to him. In his earlier article (Am. Journ. Phil. Vol. V.) he refers to Professor Edgren's paper on the Verbal Roots of the Sanskrit Language (Journ. Am. Or. Soc. Vol. XI. p. 1-55). He greatly approves of his pupil's results and appears to wish them to be taken together with his own argument. Professor Edgren's views coincide with those of sundry authorities in comparative philology, while they disagree from those of the most competent Sanskritists of the last generation. Briefly stated, Professor Edgren's line of argument is as follows: 1. The Dhtuptha contains a great many more roots that cannot be found, than such as are traceable in Sanskrit literature, and the same remark holds good with respect to the inflexions and meanings of the roots. And in spite of a "vast" progress in the exploration of Vedic and Sanskrit works, the proportion of the former had remained in 1882 virtually the same as in 1841, when Professor Westergaard expressed the conviction that every form in the Dhtuptha is genuine and would be found some time or other in inaccessible or unexplored works. Professor Edgren's second proposition is certainly not in accordance with the facts, as will be shewn below. 2. The roots preserved in the grammars and their Angas alone, are barren and mostly have no offspring,are not connected with derivative nouns, such as the genuine roots have produced in great numbers. Only 150 among them seem to have "a possible connection in sense with surrounding or similar nominal forms." This proposition, too, requires considerable modification. 3. Most of the roots, not found in Sanskrit literature, are not represented in the cognate languages. Professor Fick's Wrterbuch shews only 80 roots, solely known through the Dhtuptha, to have belonged to the common stock of the Indo-European speech, and it would seem that in some cases the evidence adduced is too
2

See my Rough List, No. 120. This MS., or its original, will be used for

the continuation of the edition of the Dhtuvrtti in the Benares Pandit.

197 Roots of the Dhtuptha

weak. On the other hand, among the verified roots, 450 have representatives in Greek, or in the Iranian, the Italic, the Teutonic, the Sclavonic and the Celtic languages. 4. On a closer examination the unverified roots shew various peculiarities, which point t o an artificial or fictitious origin. First, t h e majority of t h e m naturally arranges itself into smaller o r larger groups of forms of similar sound and identical in meaning, " t h e analogy of f o r m being such as t o exclude the principle of g r o w t h and decay." The first instance given is the group kev, khev, gev, glev, pev, plev, mev, mlev, sev, meb, peb, mep, lep w i t h the meaning ' t o

honour, to serve,' and with absolutely identical inflexion. To Professor Edgren (p. 15) " it seems, as if, in coining these counterfeits, the guiding principle has been at first to model them in form and sense on some genuine radical, rightly or wrongly interpreted," and he suggests that the above group ' ' leans on the real root sev as its point d'appui." To me it would seem that, in the case quoted, Professor Edgren has made his list unnecessarily long. Sev and sev differ only in pronunciation, and so do pev and peb, as well as mev and meb. To a Hindu the syllables si and si, se and se are absolutely the same thing, and our Dictionaries are full of words, which shew sometimes the one and sometimes the other. Again ba and va likewise are often exchanged. In Northern India (excepting Kasmlr), and in the East, va has been lost completely and, as the inscriptions prove, since ancient times. The ten remaining forms, it would seem to me, are clearly variants of two originals, *skiev and plev, and are due to the same principles of change, which are regularly operative in the Prakrits and not rarely active in Sanskrit, as well as in other Indo-European languages. The pedigree3 stands thus:
*sklev plev

kev khev
f

plev

pev

mlev

lev

gev

i
mev

i
mep lep The form gev has been preserved, I think, in the noun gevaya 'the low ones' (Asoka, Pillar Edict, III.), which is best explained as equivalent to gevakh 'servitors, slaves,'4 The same remarks apply to most of Professor Edgren's other groups, which usually consist of one or two old forms, with numerous dialectic varieties or such varieties as might be expected in the same dialect, according to the laws of Indo-Aryan phonetics. Some shew, too, an intimate con3 Examples of the assumed changes are to be found in Professor E. Miiller's Simplified Poli Grammar, and Professor Pischel's edition of Hemacandra's Prakrit Grammar, as well as in Sanskrit, where, e.g., the same words sometimes shew k and g, like karta or garta, kuipha or gulpha, kirika or girika, samkara or samgara, kuha[ra] and guha[ra], tatka and tadga, lakuta and laguta and where roots are

found ending in k, or equivalents thereof, while the corresponding ones in the cognate languages shew the media. 4 1 withdraw my former proposal to derive gevay from glep dainye, because the Pa|i usually preserves a la preceded by gutturals, and because I find in PN many cases, where aka is represented by aya.

198 Georg Bhler

nection w i t h words of common occurrence in Sanskrit o r in the Prakrit languages. Thus, in the second gana, champ gatau is evidently the parent of the modern Gujarti jhpavum and so f o r t h , and of the Sanskrit j ha mp,j hampa, jhmpana. Again, in his fifth gana, gaj sabdrthe bears the same relation t o the common Sanskrit verb garj as kas t o kars, jap t o jaip and so f o r t h . A n d gajati is probably the parent of gaj a ' elephant,' literally ' t h e roarer, the t r u m peter.' The important fact that a very large p r o p o r t i o n of the roots of the Dhtuptha is Prakritic in f o r m , has apparently not been fully realised by Professor Edgren, though Professors W e b e r , Benfey and many other Sanskritists have repeatedly called attent i o n t o it, both years ago and quite lately. The second point, w h i c h , according t o Professor Edgren, makes the introuvable roots appear artificial, is the fact that so many of t h e m are stated t o have the same meaning. To take only the w o r s t case, t h e r e are, according t o Professor Edgren, 336 verbs, t o which the explanation gatau is appended, and only 65 can be verified in literary w o r k s . The fact, no doubt, looks curious. But it becomes easily intelligible, if one consults the Hindu Sastras as t o the meaning of gat/ or gamana. The Naiyyikas and Vaisesikas say,5 karma pacavidham utksepanam avaksepanam kuncanam prasranarn gamanam, and give us the definition of gamana m / uksepandicatustayabhinnatve sati karmatvavat. They f u r t h e r add, gamanam bahuvidham I bhramanam recanam spandanam rdhvajvalanam tirya-

ggamanam ti // It is evident that the author, or authors, of the Dhtuptha hold the same opinion, and that they mean to say that the roots, marked gatau, denote some kind of motion. It is a matter of course that definitions like bhsane, bhsane, sabdrthe and himsym are likewise intended merely as general indications of the category to which the verbs belong, not as accurate statements of their meanings. The third point, which rouses Professor Edgren's suspicions, is that the same verbs are used according to the Dhtuptha dare/
andarej gatau himsym Ibhsane bhsane or vyaktym vci and

avyaktym vci. Nevertheless, the Sanskrit dictionaries shew that many verbs actually are used with widely divergent meanings, and he might have found without difficulty in English and in other languages a good many instances, exactly analogous to those which have appeared to him so extraordinary in Sanskrit. The problems which the Dhtuptha offers, ought to be approached in a very different spirit and can be solved only by a different method. Taking as correct Professor Whitney's statement (Am. hum. Phil. Vol. V. p. 5 of the reprint) that in all eleven hundred roots are awaiting verification, and likewise Professor Edgren's assertion that 150 among them are connected with nouns occurring in Sanskrit literature, and that 80 have representatives in the cognate languages, the genuineness of 870 forms has still to be proved, and the number of unverified inflexions and meanings is in all probability at least equally great. The first question to be put is, of course, if all that can be done has been done in order to account for them, or if there are still materials unused and unexplored. The next consideration is, whether the author or authors of the Dhtuptha may be supposed to have drawn on other materials than those accessible in the present day and if there are circumstances which could explain the
5

1 take the following definitions from Mahmahopdhyya Bh. Jhalklkar's

excellent Nyayakosa (second edition, 1893, Bo. Sansk. Ser. No. XLJX.)^

199 Roots of the Dhtuptha

apparent barrenness of so many roots as well as the absence of representatives in the cognate languages. Professor Edgren is certainly right in maintaining that a great many Sanskrit w o r k s , and particularly the more ancient ones, have been explored lexicographically since Professor W e s t e r gaard's times. But he is as certainly in e r r o r , when he says the number of verified roots, meanings and inflexions has remained virtually the same. A comparison of the articles on roots in the Petersburg Dictionaries and in Professor W h i t n e y ' s Supplement

with the Radices proves that incontestably. Without counting those roots, which occur in Sanskrit literature, but are not found in the Dhtuptha, Professor Whitney has 120 verified roots, for which Professor Westergaard was only able to quote Pan i ni, the Bhattikvya and the Nirukta, and the smaller Petersburg Dictionary has about a score more. Each Samhit of the Vedas, the Kthaka, the MaitryanTya, the Taittirya and that of the Saunaka Atharvaveds has furnished its contribution. The same remark applies to the Brhmanas, the Upanisads and the Vedngas, among the Stras especially to the huge Kalpa of the pastamblyas. And it must be noted that, with the exception of the Rk and Atharva Sambutas, which have been indexed, the exploration of the printed works is not complete, and that the interpretation even of these two Samhits, is not yet settled. The Reas and the Mantras of the Atharvgi rasas are sti 1 1 a field yatra yuddham kaekaei between the strict philological school and the linguists, and will probably remain so for some time. It is not doubtful that, with an alteration of the method of interpretation, the views regarding the meanings of a certain number of roots and words, and regarding the derivation of the latter will considerably change. After this Bhler gives some illustrations to show that there are many Sanskrit works which have been either partially explored, or not explored at all, and which can furnish facts confirming the statements of the Hindu grammarians. He then continues: But, even after the whole existing Sanskrit literature has been fully explored, only half the task of the root-hunter has been accomplished. He has then to extend his researches to the ancient and modern Prakrits, many of which possess an extensive literature, as well as to the Mixed Language of the first centuries before and after the beginning of our era. The compositions in the oldest types of the Prakrits, which are found in Asoka's Edicts, in the Vinayapitaka, the Paca Nikyas (e.g., in the verses of Jtakas), and other canonical Buddhist books, certainly existed in the third century B. C. This much is evident from the Maurya inscriptions on the rocks and pillars and from those on the Stpas of Sanchi and Bharahut. Their language has preserved numerous forms older than those of the classical Sanskrit of Pnini, and some older than those of the Vedic dialects. Their frequent nominatives plural from masculine a-stems in se and from neuter o-stems in (Professor Oldenberg's discoveries) are Vedic. The not uncommon occurrence of the subjunctive (Professor Pischel's discovery) is another remnant of the language of the Rsis, and such are the imperatives like vajtu (vrojatu), the plural instrumentais of the a-stems in ebhi (Oldenberg), the very common first persons plural in mase, the infinitives in tave, tye, tuye, ase and other forms, which may be gathered from the Pah grammars, or from detached articles and notes of Professors Fausbll, Jacobi, Kern, Kuhn, Leumann, Oldenberg, Rhys Davids, Trenckner, Weber and Zachariae, as well as of Dr. Morris and M. Senart in Kuhn's

200

Zeitschrift,

Bezzenberger's Beitrge, the London Academy, the

Georg Bhler

Journal of the Pali Text Society, the editions of the Asoka Inscriptions, the Mahvastu and in other works. Among the forms, which are older even than the Vedic language, I will only mention the present participles of the tmanepada in mana, mina, mina which the Asoka Inscriptions offer, and which agree with the Greek, Latin and Bactrian endings, and the Aorist addasa ' I saw,1 which goes back, not to Sanskrit adarsam, but to *adrsam, thus corresponding exactly with eSpaKov, and which without a doubt is the older form. A careful investigation of the oldest documents reveals the existence of very many similar cases. Now it might be expected, that such a language should have preserved verbal roots, which were dropped by the classical writers. And Professor Kern has shewn long ago in his Bijdrage tot de Verklaring van eenige Worden in Pali-Geschriften Voorkomende, as well as recently in his Review of Jtakas, Vol. V. (Museum of 1893, p. 10Off.) that this is the case. . . . In lately going over the Jtakas for a different purpose I have noted representatives of some more verbs, for which the explored Sanskrit literature offers no passages, and even of some which Professor Whitney in his Supplement expressly stigmatises as "without a doubt artificial." The verbs which Bhler discusses and illustrates with excerpts from the Jtakas, but also with references to other Prakrits and to modern Indo-Aryan languages, are the following: (1) sighati 'smell,' (2) mundati 'crush,' (3) satati 'pierce,' (4) bhandate 'deride,' (5) kutati 'curl/ (6) amati 'go,' (7) irati 'move,' (8) kujati 'whistle,' (9)gandhayati 'injure,' (10)gra(n)thate 'be hurt,' (11) ndhati 'fade,' (12) maghate 'start,' etc. He then continues: The fundamental maxim, which gives their importance to these researches, is that every root or verb of the Dhtuptha, which has a representative in one of the Prakrits,Pal i, Ma'hrstrl, Mgadh, Saurasen, the Apabhramsas, or in one of the modern Indian Vernaculars must be considered as genuine and as an integral part of the Indo-Aryan speech. Those, who consider such verbs to be "sham," "fictitious" or "artificial " have to prove their contention and to shew, that, and how, the author or authors of the Dhtuptha coined them. This rule, of course, holds good not only for the Indian languages, but mutatis mutandis for all linguistic research. If the grammatical tradition regarding the existence of a certain word is confirmed by the actualities in any dialect of a language, the presumption is that the tradition is genuine. As I do not claim to possess prophetic gifts, I do not care to predict how many hundreds of roots will exactly be verified, when the search has been completed. But it is not doubtful that the majority of those verbs, which Prof. Whitney considers suspicious or fictitious, will turn up, and in addition aconsiderable number of such as have not been noted by the Hindu grammarians. On the other hand, it would be wonderful, if the whole contents of the Dhtuptha could ever be " belegt." For, it has been pointed out repeatedly and must be apparent to the merest tiro in Indian palaeography that a certain proportion of the roots is the result of misreadings. This is, of course, highly probable in all cases where the Dhtuptha gives pairs like yuch and puch or has and us. The characters for jha and are almost exactly alike in the Ngarl alphabet of the ninth, tenth, eleventh and twelfth centuries, just as those for ya and pa in the later MSS. More important is another point, which likewise has been frequently noticed, viz., the fact

201 Roots of the Dhtuptha

that only a small portion of the Vedic literature, known to Pan ini and his predecessors, has been preserved, and that of the ancient laukika Sastra, the Kvya, Purna, Itihsa and the technical treatises only very small remnants have come down to our times. The assertion that the old literature has suffered terrible losses, is admitted by all Sanskritists. It is only a pity that their extent has not been ascertained, at least approximately, by the preparation of a list of works and authors mentioned in the Sabdnussana, the Brhmanas, the Upanisads and the Vedngas. Such a list, especially if supplemented by an enumeration of the numerous references to the spoken language, which Pnini's Sabdnussana contains, would probably bar for the future the inference that a root or form must be fictitious, because it is not found in the accessible literature. This inference is based on aconclusio a minori ad majus, which with a list, shewing what existed formerly and what we have now, would at once become apparent. The lost Skhs of the Vedas and the lost works of the laukika Sastra amount to hundreds. If on an average a third or a fourth of them contained each, as is perhaps not improbable according to the results of the exploration of recently recovered Sambutas and Stras, one or two of the as yet untraceable roots, that would be sufficient to account for all the lost stems. Three other considerations, it seems to me, help to explain some of the most remarkable peculiarities, observable in the materials incorporated in the Dhtuptha, viz., the fact that a certain proportion of the roots really is and will remain isolated, neither derivatives nor cognate forms being traceable in the IndoAryan or in the Indo-European languages, and the indisputable fact that many roots may readily be arranged in groups, similar in sound and identical in meaning and inflexion. Both these peculiarities, as stated above, have been used by Prof. Edgren in order to prove that the verbs, shewing them, must be fictitious. And it has been pointed out, that the number of the isolated and barren verbs is not so great as Prof. Edgren supposes, the inflected forms or representatives of a certain proportion being found in the Prakrits and in the unexplored Sanskrit literature. Nevertheless, a certain number of instances will remain, which requires accounting for. With respect to the second fact, it has been pointed out that many of the curious variants are clearly dialectic and derived from lost or preserved parent-stems in accordance with phonetic laws valid in the Prakrits and in Sanskrit.6 The chief considerations, which in my opinion do account for these peculiarities are (1) the great length of the period, during which the materials of the Dhtuptha were collected, (2) the enormous extent of the territory from which the Hindu grammarians drew their linguistic facts, and (3) the great diversity of the several sections of the Indo-Aryans inhabiting this territory. It is admitted at all hands that Pnini's Sabdnussana is the last link in a long chain of grammatical treatises, which were gradually enlarged and made more and more intricate, until the Hindu system of grammar became a science, which can be mastered only by a diligent study continued for years. According to the unanimous tradition of the Hindus, the Vykarana is a Vednga,
6

A perusal of Prof. Per Person's Wurzelerweiterung und Wurzelvariation would perhaps convince Prof. Edgren that many Indo-European

roots may be arranged in ganos, similar to those in which he has arranged so many verbs of the Dhtuptha.

202 Georg Bhler

i.e., a science subservient t o t h e study of t h e Veda, and it is highly probable that the older Hindu grammars exclusively o r chiefly explained t h e Vedic forms, just like t h e oldest Kosas, t h e Nighantus, include very little that is not derived f r o m Vedic texts. In Pnini's grammar t h e Vedic language is of minor importance. Its chief aim is t o teach t h e correct forms of t h e laukikJ bhs f o r the use of students of Sanskrit. The road, that leads f r o m t h e Vedanga t o the independent Sabdnussana, is a long one, and has not been traversed in one o r a few decades. Centuries were required in order t o effect t h e change. For in India processes of development are particularly slow, except when extraneous impulses come into play. To t h e conclusion that t h e prehistoric period of t h e Vykarana was a long one, point also Pnini's appeals t o t h e a u t h o r i t y of numerous predecessors. He not only mentions ten individual earlier teachers, but also t h e schools of t h e N o r t h and t h e East, and his grammar shews indeed very clear traces that it has been compiled f r o m various sources. N o w , if Pnini's Stras are t h e final redaction of a number of older grammatical w o r k s , the same must be t h e case w i t h his Dhtuptha. For t h e arrangement of all Indian Sabdnussanas presupposes t h e existence of a Dhtuptha, and t h e r e is no reason t o assume that t h e older grammars were deficient in this respect. It may be even suggested that t h e occasional discrepancies between t h e teaching of t h e Dhtuptha and rules of t h e Sabdnussana, t h e existence of which has been alleged, as well as t h e inequality in t h e explanatory notes, appended t o t h e roots, are due t o an incomplete unification of t h e various materials which Pnini used. Similar instances of what looks like, o r really is, carelessness in redaction 7 are not wanting in other Stras. In t h e Introduction t o my Translation of Apastamba's Dharmastra 8 1 have pointed o u t that, though Apastamba condemns in that w o r k t h e raising of Ksetraja sons and t h e practice of adoption, he yet describes in t h e Srauta Stra the manner in which a " s o n of t w o f a t h e r s " shall offer t h e funeral cakes, and that Hiranyakesin has not t h o u g h t it necessary t o make the language of t h e several parts of his Kalpa agree exactly. But, if Pnini's Dhtuptha must be considered as a compilat i o n f r o m various w o r k s , dating f r o m different centuries and composed in various parts of India, it is only t o be expected that it should contain many verbs which had already in his t i m e become obsolete and isolated, many variants o r dialectic forms. This supposition becomes particularly credible, if t h e extent of t h e t e r r i t o r y is taken into consideration, f r o m which t h e ancient grammarians d r e w t h e i r linguistic facts. It extends f r o m t h e Khyber Pass and t h e f r o n t i e r of Sindh in t h e W e s t , about 71 E. L , t o beyond Patn in t h e East, in 86 E. L , and f r o m t h e Himalaya t o t h e Vindhya range, w h e r e t h e Narmad, t h e mekhoi bhuvah, divides t h e U t t a r p a t h a f r o m t h e Daksinpatha, or roughly reckoning f r o m t h e twenty-second t o t h e t h i r t y - f i r s t degree N. L. The Aryan population of this large tract was divided into a very great number of tribes, clans, castes and sects, as well as of schools
1 say advisedly ' looks like or really is carelessness,' because it is always possible that the Strakras intentionally left contradictory rules unaltered in order to indicate an option. Very clear cases of carelessness in the working up of different
7

materials, do, however, actually occur, e.g., in the grammatical and lexicographical works of Hemacandra. 8 Sacred Books of the East, Vol. II. p. xxiii p. 130, note 7.

203 Roots of the Dhtuptha

of Vaidiks, Pandits and poets, and owed allegiance t o the rulers of perhaps a dozen or more different kingdoms. In historic India t r i b a l , sectarian, political and other divisions have always strongly influenced the development of the languages, and have caused and perpetuated dialectic differences. It seems difficult t o assume that matters stood differently in prehistoric times, when there was not, as later, one single w o r k which was generally considered as the standard authority of speech by all educated Aryans. The diversity of the words and forms in literary works and in the speech of the educated classes probably was very great and the task of the earlier grammarians, who had t o make their selection from them very difficult. This difficulty was, it might be expected, not lessened by their method of w o r k i n g . Even in the present day Indian Pandits rarely use any of the scientific apparatus, of which European scholars avail themselves. Indexes, dictionaries and "Collectanea," such as are at the service of the Europeans, are unknown t o t h e m . They chiefly trust t o memory, and w o r k in a happy-go-lucky sort of way. Even when w r i t i n g commentaries, they frequently leave their quotations unverified or entrust the verification t o incompetent pupils. The enormous quantity of the materials and the deficiencies in the system of w o r k i n g them up, explain why none of the Vedas or other old books have been excerpted completely, while the diversity of the materials and the length of the period, during which the collections were made, fully account for the occurrence of dialectic, and of isolated or obsolete, forms in the list of roots. In my opinion it is only wonderful that they are not more numerous. I now come t o the real object of my paper, the practical suggestions for the continuation of the search for roots and forms and for an organisation of this search. On the one hand it is necessary that all the unpublished Dhtupthas together w i t h their commentaries should be edited critically w i t h good indexes and that the same should be done w i t h the Sanskrit Kosas, which furnish the tradition regarding the derivatives. On the other hand, all accessible Sanskrit, Pali and Prakrit books and MSS., as well as the Vernacular classics ought t o be read and excerpted by competent scholars, w i t h a view t o the preparation of a Dictionary of Indo-Aryan Roots. This Dictionary ought t o contain, not only the roots, included in the Dhtuptha, together w i t h their meanings and inflections, verified and unverified, as well w i t h the corresponding forms of the Prakrits and Vernaculars, but also those verbs, which the grammarians have o m i t t e d , whether they are found in Vedic, Sanskrit, Prakrit or Vernacular literature or speech. If the materials are arranged methodically and intelligibly, and if a good index is added, such a book would be of very considerable use t o all linguists, who study any of the Indo-European languages. And if the excerpts are made w i t h the necessary care, a portion of them can be made useful for the Sanskrit, Pali and Prakrit dictionaries of the future. The magnitude of the undertaking would preclude the possibility of its being carried out by one or even by a small number of students. The co-operation of a great many would be required, not only of Europeans and Americans, but also of the Hindus of the modern school, who alone can furnish the materials for the very important Vernaculars. Moreover, a careful consideration of the general plan would be necessary, as well as the settlement of definite rules and instructions for the collaborators. Perhaps

204 Georg Bhler

one of the next International Oriental Congresses will be a suitable occasion for the discussion of such a scheme, and of its details as well as of the great question of ways and means. I believe, that if the idea finds the necessary support, the appointment of a permanent international C o m m i t t e e w i l l be advisable, which should supervise the preparation of the w o r k and the indispensable preliminary labours. A small beginning has already been made w i t h the latter by the Imperial Austrian Academy's Series of Sources of Sanskrit Lexicography, of which the first volume has appeared and the second, containing Hemacandra's Undiganastra w i t h the author's commentary, is ready for the press, w h i l e the t h i r d , the Mankhakosa w i t h its commentary, has been undertaken by Prof. Zachariae. It is a matter of congratulation that the Council of the Socit Asiatique has expressed its willingness t o cooperate and has commissioned M. Finot t o edit the Ajayakosa on the same principles, which Prof. Zachariae has followed in preparing the Anekrthasamgraha. I have hopes that the Austrian Academy w i l l sanction the issue of some more volumes, including also some Dhtupthas, e.g., those belonging t o Hemacandra's grammar and t o the Ktantra. Prof. Lanman, the German O r i e n t a l Society, the Asiatic Society of Bengal and other corporations or individuals publishing editions of Sanskrit texts would each agree t o undertake a f e w volumes, the necessary auxiliary editions might be prepared w i t h o u t t o o great a delay and w i t h o u t t o o heavy a strain on the resources of one single body. A t the same t i m e it w o u l d be quite feasible t o begin w i t h the excerpts f r o m the literary w o r k s , the results of which could be published preliminary in the Journals of the various Oriental Societies and in the Transactions of the Academies. The f o r m of publication ought t o be such that they could easily be used by the editor or editors of the Dictionary, and the original excerpts, done according t o uniform principles, might be deposited for f u t u r e reference in the libraries of the learned bodies, publishing the results. W i t h a well considered plan, which might follow partly the lines o f t h a t , adopted for the new Thesaurus Totius Latinitatis, the " Dictionary of Indo-Aryan R o o t s " might be completed w i t h i n the lifetime of those among us w h o at present are the madhyamavrddha Sanskritists. If the idea is ever realised and a standard book is produced, agreat part of the credit w i l l belong t o Prof. W h i t n e y . In his Supplement, which, in spite of my different views regarding the character of the linguistic facts handed down by the ancient Hindus and regarding various details, I value very highly and in his justly popular Sanskrit Grammar, the statistical method has been first applied t o Sanskrit, and these t w o w o r k s mark a decided advance in the study of the ancient Brahmanical language.

The Transition

16
Bernhard Geiger (1881-1964)

The long controversies on the nature and the value of the linguistic work of the Sanskrit grammarians did not do the subject much good. Actually, Whitney's forceful opinions may have acted as a deterrent to many young Sanskritists. During a long period of transition, roughly coinciding with the first quarter of this century and marked by its paucity of contributions, only Bruno Liebich kept the torch burning by a series of discerning and valuable publications following his monograph on Pnini and including Zwei

Kapitel der Ksik: bersetzt und mit Einleitung versehen (1892), Das Cndra-vykarana (1895), Cndra-vykarana: Stra, Undi, Dhtuptha (1902), Candra-vrtti: Der Original-Kommentar Candragomin'szu seinem grammatischen Stra (1918), Zur Einfhrung in die indische einheimische Sprachwissenschaft IIV (1919-1920), Materialien zum Dhtuptha (1921), Konkordanz Pnini-Candra (1928), and KsJrataraginJy KsJrasvmin's Kommentar zu Pnini s Dhtuptha
(1930). Liebich did for the non-Pninian school of Candragomin what had earlier been done for the Pninian school. The Cndra school, especially popular among Buddhist scholars, did not survive in India but was very influential (generally in translation) throughout the Buddhist world from Tibet to Ceylon and Java. More directly in the Kielhorn tradition, during the period of transition, one study appeared by a pupil of Kielhorn, Bernhard Geiger. Geiger had first studied at Vienna and had written a dissertation on Arabic literature. He continued his studies in Prague, Bonn, and Gttingen, where he studied grammar with Kielhorn and Iranian with F. K. Andreas. In 1908 at the instigation of Kielhorn he published the study, reprinted here, which qualified him for an appointment at Vienna as " Privatdozent fr altindische und altiranische Philologie und Altertumskunde" (Frauwallner 1961, 89-90). He also published studies on Indian music (jointly with E. Felber) and on Iranian. After the German occupation of Austria, Geiger emigrated to the United States, where he worked at the Asia Institute from 1938 to 1951, and taught at Columbia from 1951 until his retirement in 1956. He died in 1964. The article of 1908 discusses some rules in Pnini's grammar as interpreted and commented upon by Patajali in his Mahbhsya and by Kaiyata in his subcommentary PradJpa (of the eleventh century). In this study Geiger translates the relevant passages from the Mahbhsya and deals with, among other things, Pnini's concept of asiddha, ' regarded as not having taken effect,' which corresponds to the notion of rule ordering in contemporary linguistics. An example of Pnini's technique is the following: Rule 6.1.71 implies that the infix t called tUK (capitals stand for metalinguistic elements) is inserted after a short vowel before the suffix ya called LyaP, i.e., in the context: short vowel LyaP. (1)

For example: prakr + LyaP - prakrtya. Rule 6.1.101 states that two homorganic vowels combine into one corresponding long vowel (2). For example, danda -f agram

- dandgram.
Similarly, (2) would yield: adhi + / + LyaP -> adhiya. But the correct form in Sanskrit is adhltya. This can be obtained if (1) has also applied. But (1) is applicable only if the left context is a short vowel, and in adhi ya it is long. Pnini succeeds in deriving

the correct result by postulating a metarule (6.1.86), which applies

208 Bernhard Geiger

t o o t h e r rules as well and w h i c h states that (2) is asiddha ' regarded as not having taken effect' w i t h respect t o (1). In modern linguistics this result w o u l d be obtained by o r d e r i n g (1) before (2). A n o t h e r example is the f o l l o w i n g : Rule 6.4.101 states that the imperative ending hi is replaced by dhi after the r o o t hu- and after roots ending in a class of sounds denoted as jhAL, i.e.:
hi dhi

/!ihAL_

(3)

(e.g., bhind + hi - bhinddhi). Rule 6.4.119 states that the vowel element of the roots dand dho- (together called GHU) and the root as 'be* are replaced by e before the ending hi, i.e.: f*GHlh

I as I - / -

hl

,.

,A,

(where *GHU denotes the vowel element of GHU; e.g., da + hi -> dehi). But the imperative of as- is edhi 'be/' Now s, but not e, is included in the class jhAL Hence edhi cannot be derived from the underlying form as-hi unless (3) is declared asiddha with respect to (4), or (4) with respect to (3). This requirement is indeed fulfilled as a direct consequence of metarule 6.4.22. Geiger deals with these and with similar topics in " Mahbhsya zu P.VI,4,22 und 132 nebst Kaiyata's Kommentar: bersetzt, erlutert und mit einem Anhang," Sitzungsberichte der philosophisch-historischen Klasse der kaiserlichen Akademie der Wissenschaften zu Wien (160,1909, VIII, 1-76).

Mahabhasya ZU PVI, 4 , 2 2 und 132, nebst Kaiyata's Kommentar (1909) Bernhard Geiger

Vorwort Whrend eines glcklichen Studienaufenthaltes in Gttingen erhielt ich bei der Lektre des Mahbhsya von meinem hochverehrten Lehrer, Herrn Geheimrat Professor F. Kielhorn, die Anregung, die Abschnitte VI, 4, 22 und 132 des Mahbhsya zu bearbeiten. Sie empfahlen sich hiezu besonders deshalb, weil sie in sich abgeschlossen sind, die Anwendung und Bedeutung einer ganzen Reihe wichtiger grammatischer Kunstgriffe und Erklrungsmethoden kennen lehren und typische Beispiele der Vorzge und Schwchen des Mahbhsya liefern. Angesichts der groen Schwierigkeiten, die das Verstndnis des Mahbhsya bietet, schien mir eine solche Arbeit nur dann berechtigt zu sein und Nutzen zu versprechen, wenn sie imstande wre, auch den der Methode des Mahbhsya Unkundigen mit ihr vertraut zu machen. Dieses Ziel schwebte mir bei der bersetzung und den Erluterungen vor und bestimmte mich dazu, auch den Kommentar Kaiyata's vollstndig zu bersetzen, trotz der Schwierigkeiten, der die Darstellung hier begegnete. Die bersetzung des Kommentars schien mir unerllich zu sein, weil Kaiyata's Erklrungen oft viel komplizierter und schwerer verstndlich sind als das Bhsya selbst und die Kenntnis anderer Stellen des Mahbhsya und die des Paribhsendusekhara voraussetzen. Leider erwies sich mein ursprnglicher Plan, den vollstndigen Text von Kaiyata's Kommentar beizugeben, als undurchfhrbar. Ich habe in den meisten Fllen nur die ersten Worte einer Anmerkung von Kaiyata zitiert und hievon nur bei besonders schwierigen Stellen eine Ausnahme gemacht. Auch wichtigere Parallelstellen aus anderen Teilen des Mahbhsya sowie manche schwierigere Bemerkung der Ksik habe ich bersetzt, den Paribhsendusekhara, dessen bersetzung von Kielhorn mir Vorbild war und fr das Verstndnis von Kaiyata's Kommentar die grten Dienste leistete, habe ich mglichst oft herangezogen, und im Anhang gebe ich einige, wie ich glaube, notwendige Ergnzungen und Erluterungen, die in den Anmerkungen nicht untergebracht werden konnten. Die Vrttikas Ktyyana's habe ich zumeist unbersetzt gelassen, da Patajali sie genau pharaphrasiert. Fr die bersetzung des Bhsya habe ich Kielhorns Ausgabe [Mbh.] bentzt. Kaiyata's Kommentar war mir nur in der gedruckten Ausgabe Benares 1887 zugnglich, die von sinnstrenden Fehlern nicht frei ist. Den Dhtuptha habe ich nach der Ausgabe von Westergaard zitiert. An Abkrzungen kommen zur Verwendung: Kty. = Ktyyana. Kaiy. = Kaiyata. Pat. = Patajali. Ks. = Ksik. Siddh. K. = Siddhnta-Kaumudl. Paribh. mit folgender Seitenzahl bezeichnet Ngojlbhatta's Paribhsendusekhara, Text und Translation von F. Kielhorn. Bombay Sanskrit Series. Bombay 1868-74. Paribh. Nr. = Paribhs Nr. [in dem eben genannten Werke]. Ms. K. bezeichnet ein Herrn Prof. Kielhorn gehriges Manuskript von Kaiyata's Kommentar, mit dem er auf meine Bitte einige Stellen verglichen hat, an denen ich von meinem Text abweichende Lesungen vermutete. Es empfiehlt sich, bei der Bentzung dieser Arbeit berall dort, wo der Paribhsendusekhara zitiert st, auch noch Kielhorns bersetzung dieses Werkes zu Rate zu ziehen. Die Kenntnis von

210 Bernhard Geiger

Paribh. N r . 38 setze ich voraus. ber die in ihr enthaltenen Termini vgl. Paribh. Transi, p. 185, n. 1. Diese A r b e i t war schon abgeschlossen und ich war eben dabei, einige Verbesserungen und Bemerkungen, die Prof. Kielhorn mir wenige Tage vorher gesandt hatte, zu v e r w e r t e n , als ich die erschtternde Nachricht von dem pltzlichen Tode meines lieben, unvergelichen Lehrers erhielt. Nun erreicht ihn mein Dank fr seine so freundliche, teilnahmsvolle Frderung dieser A r b e i t nicht m e h r ! Es war sein W u n s c h , da durch die bersetzung und Erluterung eines Abschnittes des Mahbhsya das Verstndnis der indischen Grammatiker, deren grter Kenner er war und fr deren Kenntnis er das meiste getan hatte, gefrdert werde. Mge dies der vorliegenden A r b e i t gelungen sein !

Einleitung
Obwohl Kielhorn schon im Jahre 1876 in der grundlegenden Abhandlung ,Ktyyana and Patajoli : their relation to each other, and to Pnini' [Bomba/ 1876] das wahre Verhltnis der drei groen Grammatiker endgi'ltig aufgezeigt hat, begegnet man noch immer einer unrichtigen Auffassung ihres gegenseitigen Verhltnisses, besonders aber einer ungerechten Beurteilung Pat.'s. So urteilt selbst Bhtlingk noch in der Einleitung [p. XVII] zu seiner Pninibersetzung [Leipzig 1887] folgendermaen: ,Wenn Kty. solche Versehen1 [sc. Pn.'s] rgt, sucht Pat. sie auf irgendeine spitzfindige Art zu bemnteln. Wer an einer solchen Kritik des groen Grammatikers Gefallen findet, mge sich in das uns jetzt in musterhafter Ausgabe vorliegende Mahbhsya vertiefen.' Mit dieser Behauptung bekennt sich Bhtlingk, wenigstens soweit Pat. in Betracht kommt, zu dem alten Standpunkt Webers, nach welchem Kty. ein Gegner Pn.'s und Pat. dessen Verteidiger gegen die Angriffe Kty.'s sein soil. Es mag also nicht berflssig sein, wenn ich an einem der hier bersetzten Abschnitte, Mbh. zu VI, 4, 22, der sich hiezu vortrefflich eignet, zeige, wie sehr die Webersche Auffassung gegen den tatschlichen Sachverhalt verstt. Wie verhlt es sich nun in diesem Abschnitt zunchst mit der Gegnerschaft Kty.'s gegen Pn. ? Die zwei ersten Vrttikas enthalten Erklrungen zweier Ausdrcke des Stra. Die folgenden sechs Vrtt. (3-8)
Bhtlingk fhrt folgendes Beispiel .grerer Unachtsamkeit' Pn.'s an: P. Ill, 3, 90 lehrt die Anfgung des Suffixes na (na) an die Wurzel prach. Da dieses na ein /-Suffix ist, wrde sich nach VI, 1,16 die Substitution
des Samprasrana r fr r, also prsna,
1

ergeben. Kty. bemerkt deshalb im Vrtt. 1, das Stra msse durch ein Verbot des Samprasrana ergnzt werden. Darauf entgegnet Pat., die Form prasna komme niptant zustande, d. h. dadurch, da Pan, in III. 2,117 das Wort prasna erwhnt und so ein Verbot gegen die Anwendung von VI, 1,16 statuiert. Pat.'s Ansicht haben auch die kritische und feinfhlige Ks. und Siddh. K. zu III, 3, 90 akzeptiert. Da Pn. sich eines so

leicht auffallenden Versehens htte schuldig machen knnen, halte ich fr vllig ausgeschlossen. Und die Anwendung eines niptana erscheint mir um nichts knstlicher als irgendeine andere Andeutung oder als so manche Paribhs, deren Kenntnis und Anwendung wie ich mit Goldstcker, Pnini, p. 114, glaube Pn. vorausgesetzt haben mu. Aber davon abgesehen, wre der Vorwurf der Spitzfindigkeit wegen der Annahme eines niptana doch auch Kty. gegenber gerechtfertigt, der in dem ganz analogen Falle sarvanman [1,1, 27 Vrtt. 1] und auch sonst fters [vgl. Ind. Ant. vol. XVI, p. 245, n. 4] zur Rechtfertigung Pn.'s ein niptana annimmt.

211 Mahbhsya

fhren Zwecke des Sutra an, dienen also zur Verteidigung Pan.'s geger[ die eventuelle Behauptung, da das Stra berflssig sei. Dagegen konstatieren die Vrtt. 9 (welches in 10 begrndet w i r d ) , 11,13 und 14 Ausnahmen von dem Stra. Vrtt. 12 erklrt, da die in 9 und 11 konstatierten Ausnahmen nicht bestehen, wenn in dem Stra das W o r t samnsraya hinzugefgt w i r d . Die V r t t . 15 und 16 nennen Fehler, die sich aus jeder der zwei Alternativen ,prgbht' (d. i. Geltung des Stra bis VI, 4,129) und , bht' ( = saha tena, d. i. Geltung bis VI, 4,175) ergeben. Naturgem kann nur eines dieser zwei Vrttikas gegen Pn. gerichtet sein. Aus Vrtt. 7 ist ersichtlich, da Kty. die zweite Alternative annimmt. Der grere Teil der Vrttikas dieses Abschnittes dient also zu Pn.'s Rechtfertigung. Pat., der angebliche Verteidiger Pn.'s, bettigt hier seine Parteinahme fr Pn. zunchst in der Weise, da er die zur Rechtfertigung Pn.'s bestimmten Vrtt. 3-8 zum Teil auf spitzfindige A r t widerlegt und hiebei einigen Regeln Gewalt antut. Er verteidigt ferner die gegen das Stra gerichtete Schlubemerkung des Slokavrttikakra. Das Vrtt. 14 wohl das einzig berechtigte unter den kritikbenden Vrttikas dieses Abschnittes lehnt Pat. nur zur Hlfte ab, whrend er die zweite Hlfte akzeptiert. Andrerseits weist Pat. die Behauptung Kt/.'s [ V r t t . 12] zurck, da die in Vrtt. 9 und 11 erwhnten Ausnahmen durch Hinzufgung von samnsraya vermieden werden, und erklrt, da der verlangte Zusatz schon in dem W o r t e atra des Stra zum Ausdruck komme. Mit demselben A r g u m e n t t r i t t Pat. auch dem Vrtt. 13 entgegen. Sub Vrtt. 12 widerlegt er den im Vrtt. 10 angefhrten Grund bahirangalaksanatvt. Und schlielich weist er auch die Vrtt. 15 und 16 zurck. Erzeigt, da weder die im V r t t . 15 genannten Ergnzungen durch die Annahme der A l t e r n a t i v e ,prg bht' notwendig werden, noch auch die im V r t t . 16 verlangte Konstatierung von Ausnahmen durch die W a h l der A l t e r n a tive bht (= saha tena). Schon diese Inhaltsangabe lehrt, wie unzutreffend die W e bersche Ansicht ist. Der eben e r r t e r t e Abschnitt bietet aber auch Gelegenheit, auf eine Eigentmlichkeit des Mahbhsya hinzuweisen, die bei der Beurteilung desselben nicht auer acht gelassen werden darf. O b w o h l nmlich Pat. bei den V r t t . 12 bis 16 von dem zweiten Teil des Vrtt. 14 abgesehen fr Pn. einzutreten scheint, fllt es doch auf, da er hier eigentlich berhaupt nicht fr Pn. Partei ergreift. Denn er weist alle Vrttikas, welche Zwecke des Stra angeben, zurck, ohne andere Zwecke namhaft zu machen. Man gewinnt infolgedessen den Eindruck, als ob Pat, das Stra VI, 4, 22 als gnzlich berflssig betrachtete. In der Tat leitet er von der W i d e r l e g u n g des Vrtt. 8 zu den Slokavrttikas mittels des Satzes ber: yadi tarhy ayamyogo nrabhyate (.wenn d e m n a c h diese Regel nicht aufgestellt w i r d ' ) . Und zwischen der letzten Bemerkung des Slokavrttikakra, die gegen die A n w e n dung des Stra gerichtet ist, und den. Ausnahmen von dem Stra, die Kty. aufzhlt, w i r d die Verbindung durch den Satz hergestellt: rabhyamne 'py etasmin yoge (,auch w e n n diese Regel aufgestellt w i r d ' ) . Pat. begibt sich damit pltzlich auf den Standpunkt Kty.'s, da das Stra notwendig sei, und k r i t i s i e r t von diesem Standpunkt aus, ohne ihn jedoch als seinen eigenen zu bezeichnen, die f o l genden Vrttikas. Aus den Entgegnungen auf die V r t t . 15 und 16 ist nicht ersichtlich, welche Stellung Pat. dem Stra gegenber einnimmt. Er sucht nur zu beweisen, da die in diesen Vrttikas gennanten Fehler sich aus keiner der beiden A l t e r n a t i v e n ergeben.

212 Bernhard Geiger

O b etwa die Annahme einer von ihnen nach Pat.'s Ansicht andere Fehler zur Folge hat, erfahren w i r nicht. Und doch wre das scheinbare Resultat der W i d e r l e g u n g von V r t t . 15 und 16, d. i. der Standpunkt, da keine der beiden Alternativen Zustze oder Verbote notwendig mache, da also beide k o r r e k t sind und gleichzeitig zu Recht bestehen, widersinnig und ganz gewi nicht der Standpunkt Pat.'s. Pat. ist hier also ebensowenig prinzipieller, unentwegter Parteignger Pan.'s als Kty. prinzipieller Gegner Pn.'s ist. Es ist berhaupt v e r k e h r t , zu glauben, da Gegnerschaft oder V o r e i n genommenheit fr Pn. die Grundstze sind, von denen Kty. und Pat. sich leiten lassen. Kty. macht nicht nur Zustze und Verbesserungen ; er t r i t t auch f r Pn. ein und in zahlreichen Fllen n i m m t er eine gegen diesen gerichtete Behauptung durch ,na v' (,oder auch nicht'), ,siddham tu' (,doch ist es in O r d n u n g ' ) u. a. m. zurck und lehrt, auf Grund welcher Auffassung ^vijnnt1), Bedeutung (,vivoksitatvt'), oder mit Hilfe welcher Paribhs die in vorhergehenden Vrttikas genannten Schwierigkeiten vermieden werden. Viel deutlicher zeigt sich der Mangel jeglicher Voreingenommenheit bei Pat. Er macht von dem Spielraum, den das so feindurchdachte System Pn.'s seiner Findigkeit noch offen lt, den ausgiebigsten Gebrauch und erhebt bei jeder sich darbietenden Gelegenheit Zweifel und Einwendungen . g l e i c h g i l t i g , o b es s i c h u m e i n e R e g e l Pn.'s o d e r e i n e B e m e r k u n g K t y . ' s h a n d e l t , u n d o h n e R c k s i c h t d a r a u f , o b es e i n e B e h a u p t u n g K t y . ' s f r o d e r g e g e n P n . i s t . Pat. prft und wgt nach allen Seiten hin ab und sucht immer noch w e i t e r e Grnde, neue Auswege, a n d e r e M g l i c h k e i t e n der Erklrung ausfindig zu machen. So k o m m t es, da er fters wie z. B. fr le Form sdhi zu Vrtt. 3 ir einen einzigen Fall gleichzeitig mehrere Mglichkeiten prsentiert. Und dies berechtigt uns zu dem Schlu, da nicht in jeder uerung Pat.'s auch seine persnliche berzeugung zum Ausdruck k o m m t , und da nicht jeder Abschlu einer Diskussion auch eine prinzipielle Entscheidung bedeutet. 2 Ein t y p i sches Beispiel mge dies veranschaulichen. Im Kapitel VI, 4, 22 [Mbh. vol. Ill, p. 190; s. unsere bersetzung] behauptet der Slokavrttikakra, da bei ahritarm [aus ahri-ta + tarm] die Elision der Endung -ta nicht als asiddho betrachtet zu werden brauche, damit die Elision von tarm v e r h i n d e r t werde. Es gelte nmlich aus einer frheren Regel das W o r t ,knitil f o r t . Pat. verteidigt diese Ansicht. Anders verhlt sich Pat. in derselben Frage im Kapitel VI, Dies st vielfach unverkennbar und brigens eine natrliche Folge der Methode des Mahbhsya, d. i. seines Diskussionsstiles. Wenn Pat. irgendeine Einwendung erhebt, so kommt darin in vielen Fllen gewi nicht sein eigener Standpunkt zum Ausdruck, sondern eine Auffassung, die vielleicht mglich wre, eine Meinung, die ein Teilnehmer an der fingierten Diskussion uern knnte. Und zwar sind dies oft recht weit hergeholte EinWendungen. In manchen Fllen wie z. B. bei upadidiye zu VI, 4, 22 Vrtt. 14 werden sie sofort mit einer treffenden Entgegnung abgetan. In
2

anderen Fllen dagegen wie bei babhvatuh zu demselben Vrtt. sind noch weitere Erwiderungen und Verteidigungen denkbar und die Diskussion kann bei einigem guten Willen lnger fortgefhrt werden, bis sie schlielich in eine Sackgasse gert, So werden scheinbare Inkonsequenzen Pat.'s leicht verstndlich. In dieser Weise ist wohl auch der von Kaiy. [zu Mhb. Ill, 193, 1.1.] gergte Fehler zu beurteilen, der in der Annahme einer Form bahusun (statt bahusvan) besteht. In dieser Hinsicht ist die Bemerkung Kaiy.'s zu Mbh. Ill, 190, 1.10 beachtenswert.

213 Mahbhsya

4 , 1 0 4 [ M b h . v o l . Ill, p. 214]. Kty. verteidigt d o r t in den drei Varttikas 3 die Regel gegen die eventuelle Behauptung, es msse in der Regel ta hinzugefgt w e r d e n , damit nicht nach der Elision des ta auch die von tarm erfolge. Das erste V r t t i k a lehrt, da ta und tarm nicht gleichzeitig abfallen knnen, weil [nach 1,1, 61] luk nur fr ein Suffix und nicht fr eine Verbindung von Suffixen e i n t r i t t . V r t t . 2 stellt fest, da auch nachher, d. i. nach erfolgtem Abfall desto, das folgende Suffix nicht abfallen knne, w e i I d i e El is i o n des ta [nach V I , 4, 22] als asiddha z u b e t r a c h t e n s e i . Nach V r r t . 3 wre auch noch ein anderer G r u n d mglich : das Prinzip, da eine Handlung (Operation) bereits vollzogen sei [und nicht ein zweites Mal vollzogen werden soll]. W h r e n d Pat. das V r t t . 3 ablehnt, n i m m t er mit den W o r t e n ,tasmtprvoktv eva parihrau1 die zwei ersten Varttikas ausdrcklich an, betrachtet also im Gegensatz zu der vorher erwhnten Stelle den Abfall von ta als asiddha. A b e r unmittelbar nach dieser Feststellung fhrt Pat. f o r t : , O d e r a b e r [die Elision von ta ist nicht asiddha, sondern] ,,/cri/t/" gilt [aus VI, 4, 98] f o r t ' . Und er verteidigt nun diese Ansicht in derselben Weise wie Mbh. vol. Ill, p. 190. Man t u t aber Pat. auch unrecht, wenn man glaubt, da seine Rolle sich in migen dialektischen Spielereien und spitzfindigen Haarspaltereien erschpfe. Mag man auch bei so mancher seiner Diskussionen das Gefhl haben, da der Scharfsinn in ihnen geradezu mibraucht w i r d , so verdienen sein Scharfblick und seine Schlagfertigkeit doch an vielen anderen Stellen alle A n e r k e n n u n g , seine Ansichten und A r g u m e n t e den Vorzug v o r denen Kty.'s. Und zwar nicht nur d o r t , w o Kty. fr Pn, e i n t r i t t , sondern im W i d e r s p r u c h mit Bhtlingks eingangs z i t i e r t e r Behauptung vielfach gerade in denjenigen Fllen, in welchen Kty. Versehen Pn.'s .rgt 1 . Da der verallgemeinernde V o r w u r f , Pat. bemntele in spitzfindiger Weise die von Kty. gergten Versehen Pn.'s, der Bedeutung Pat.'s w o h l nicht gerecht w i r d , mgen einige Beispiele aus dem hier bersetzten T e x t e zeigen, die sich leicht durch solche aus beliebigen anderen Teilen des Mahbhsya vermehren lieen. Was zunchst Kty.'s und Pat.'s verschiedene Deutungen von otro [VI, 4, 22 V r t t . 2 und Pat. zu V r t t . 12] betrifft, so scheint zwar die Analogie von prvatrsiddham [VIII, 2 , 1 ] fr die Auffassung Kty.'s im V r t t . 2 zu sprechen, wonach atra besagt, da eine Regel von VI, 4, 23 an , bis bho1 nur ebendaselbst, d. h. nur in bezug auf eine ebendemselben Abschnitt angehrige Regel als asiddha zu betrachten ist. W e n n man aber erwgt, da Pn. sich sonst in adhikras der grten Knappheit des Ausdruckes befleiigt, die Bestimmung der Geltungsgrenze eines adhikra der Erklrung berlt und in unserem Straden Geltungsbereich schon durch den Zusatz , bhV abgrenzt, so w i r d man die Auffassung Pat.'s doch fr mglich halten drfen, nach welcher das W o r t atra ausd r c k t , da eine als asiddha betrachtete Regel dieses Abschnittes ebendarauf beruhen mu, worauf die Anwendung der andern Regel dieses Abschnittes beruht. Dann kme der Zusatz samnsraya, Diese Varttikas, die einzigen zu VI, 4,104, beweisen schlagend die Unrichtigkeit der Behauptung Goldstckers [Pnini, p. 120]: ,ln consequence, his [sc. Kty.'s] remarks are attached to those Stras alone which are open to the censure of abstruse3

ness or ambiguity, and the contents of which were liable to being completed or modified: he is silent on those which do not admit of criticism or rebuke.' Goldstcker meint natrlich nur die Kritik von seiten Kty.'s.

214 Bernhard Geiger

den Kty. verlangt, schon in atra zum Ausdruck. Diese Auffassung Pat. ist vielleicht nicht zwingend, aber doch annehmbar und keineswegs spitzfindig. Ihr schlieen sich die Ks. und Siddh. K. an. Dem Nachweis Pat.'s [zu V r t t . 6], da bei gatah, gatavn die Elision des Nasals [VI, 4, 37] nicht als osiddha betrachtet werden mu, damit die Elision des a [VI, 4, 48] v e r h i n d e r t werde, seiner W i d e r l e g u n g der V r t t . 7,13,16, seiner Zurckweisung des von Kty. im V r t t . 10 angefhrten Grundes bahirahgalaksanatvt sub V r t t . 12, sowie der W i d e r l e g u n g der Vrttikas zu VI, 4,132 w i r d man ohne Vorbehalt zustimmen. Aus all dem geht hervor, da man das Studium des Mahbhsya wohl nicht blo demjenigen empfehlen darf, der an unfruchtbarer spitzfindiger K r i t i k Gefallen findet und damit vorlieb nimmt. Die Bedeutung des Mahbhsya besteht zunchst darin, da es die Vrttikas des Kty. in vortrefflicher Weise erlutert, vor allem aber darin, da es das Verstndnis von Pn.'s Grammatik in hohem Mae f r d e r t und einen Einblick in den wahren Sinn und Zweck einer Regel vielfach erst ermglicht. Fr ein tiefer eindringendes Verstndnis des Pnineischen Systems ist das Studium des Mahbhsya zweifellos unentbehrlich. Darf ich schlielich noch einen praktischen Grund anfhren, so verweise ich darauf, da die Ks. die zahlreichen zutreffenden Deutungen und Entscheidungen Pat.'s aufgenommen hat, und da infolgedessen viele Stellen der Ks. ohne Kenntnis des Mahbhsya schwer verstndlich oder berhaupt unverstndlich sind. Der erste der hier bersetzten zwei Abschnitte des Mahbhsya behandelt das Stra VI, 4, 22. Mag nun Pat.'s Erklrung des W o r t e s atra [zu V r t t . 12] richtig sein oder nicht, so besagt dieses Stra doch folgendes: Eine in dem Abschnitt VI, 4, 22 bis bha gelehrte, in W i r k l i c h k e i t bereits in Kraft getretene Operation ist in bezug auf eine andere Operation desselben Abschnittes, die nach der Ausfhrung der ersten Operation eintreten soll oder mte, so zu betrachten, als o b sie asiddha, d . h . n i c h t i n K r a f t g e t r e t e n w r e . Der Umstand, da Pn. in dieser Regel asiddhavat sagt, whrend es VI, 1, 86 asiddhah und Will, 2,1 asiddham heit, hat die Erklrer zu Errterungen ber die Bedeutung des vat [in asiddhavat] veranlat. Die verschiedenen Ansichten hierber finden w i r am Beginn von Kaiy.'s Kommentar zu VI, 4, 22. Es heit d o r t : ha kvacid upamnopameyayor abhedam vivaksitv smndhikaranyena nirdesah kr yate: Dyam Brahmadatta iti. Sstre3 pi : satvatukor asiddhah; lit kit; goto nid iti ca. Tatra smarthyd atidesapratipattih. Kvacit tu pratipattilghavya bhedopakrame vatin nirdesah kriyate: Brahmadattavad ayam iti. Ihpi : asiddhavad atrbhd iti. Anyetv huh: svsrayam api yath syd[vgl. M b h . v o l . II, p. 66, I. 2] ityevamartham vatkaranam; tena debhatur ity atra svsrayaikahalmadhyagatsrayv ettvbhysalopau bhavata iti. Etad apare na mrsyanti. Saty asati v vatv atidesesv tidesikviruddhasvsrayakrynivrttih; siddhatvsiddhatvayor virodht katham vatin siddhatvasya prpanam 1 katham v siddhatvsiddhatvayor visayavibhgo labhyate 1 sthnivad itydau tu vatim antarena samjn syd iti vatkaranam atidesam gqmayat svsrayaprptyartham [so Ms. K. ; Text : artho] vijnyate. Snasor allopa iti taparakarant kvacit siddhatvam sakyam anumtum; anyath stm, sann itydv to Dsiddhatvl

lopprasahgt kirn taparatvenal ,Wenn man sagen will, da zwischen dem, womit verglichen wird, und dem Verglichenen selbst kein Unterschied besteht, so drckt man dies im Leben (iha) bisweilen durch die Gleichstellung aus, [indem man z. B. sagt]: ,, Dieser st [ein zweiter] Brahmadatta." [Ebenso] auch im grammatischen

215 Mahbhsya

Lehrbuch [z. B. VI, 1, 86] : M In bezug auf den E i n t r i t t von s und in bezug auf das Augment t ist [ein ekdesa als] asiddho [zu betrachten] " ; [I, 2, 5] : ,, Eine Personalendung des Perfektums ist [wie] ein k/t-Suffix [zu b e h a n d e l n ] " ; [VII, 1, 90]: ,, Nach go ist [die Endung eines starken Kasus wie] ein n/t-Suffix [zu behandeln]." In diesen Fllen erkennt man schon aus der Sachlage, da es sich um eine bertragung 4 [und nicht Identifizierung] handelt. Manchmal aber bedient man sich, um die W a h r n e h m u n g [der bertragung] zu erleichtern, der Bezeichnung durch vat, wenn man auf [die Betonung] des Unterschiedes abzielt, [indem man z. B. sagt]: ,, Dieser ist w i e Brahmadatta." [So] auch hier [in VI. 4. 22]: ,, Bis bha ist [eine bereits in Kraft getretene Operation in bezug auf eine andere Operation] ebendaselbst [so zu betrachten], wie w e n n sie nicht in Kraft getreten w r e . " Andere dagegen sagen : ,, Die Setzung von vat hat den Zweck, da auch die aus ihm [d. i. dem verglichenen siddha] selbst sich ergebende [Operation] eintreten mge; 5 auf diese Weise erfolgen bei debhatuh [aus dambh] die Subs t i t u t i o n des e und die Elision der Reduplikation, die au dem in der M i t t e zwischen zwei ei n f a c h en Konsonanten stehenden [Vokal a] beruhen, welcher eben auf ihm [sc. dem Siddha-seln] selbst b e r u h t . " 6 Dies wollen wieder andere nicht zugeben, [welche einwenden] : ,,Ob nun vat dasteht oder nicht, so w i r d doch bei bertragungen [ n u r ] diejenige aus ihm [d. i. dem Verglichenen] selbst sich ergebende Operation nicht aufgehoben, welche der bertragenen [Operation] nicht widerspricht. W i e soll man es [hingegen bei debhatuh], da doch Siddha-se\n und Asiddha-seln einander ausschlieen, mittels vat erreichen, da [eine Operation] siddha sei ? Oder wie soll man [wenn das aus
4

Nmlich des fr den einen Begriff Geltenden auf den andern. 5 Hier wird der gewaltsame Versuch gemacht, dem vat eine hnliche Bedeutung beizulegen, wie sie dem vat in sthnivat [1,1, 56] eigen ist. Wenn z. B. I, 3, 28 lehrt, da bei han nach der Prposition in nicht transitiver Bedeutung das tmanepadam eintritt, so gilt nach 1,1, 56 dasselbe auch fr das Substitut vadh [II, 4, 44]. D. h. das fr den sthnin [d. i. hon] Geltende wird auf das Substitut bertragen; dieses hrt nicht auf zu funktionieren, sondern bildet die Basis fr die bertragene Operation. Man erhlt also auer hata auch vadhista. Das tmanepadam ist demnach nicht nur sthnysrayam, sondern auch svsrayom, d. i. dessrayam. In unserem Falle stehen die Begriffe asiddha und siddha einander gegenber. Das vat in asiddhavat soll auch hier angeblich ausdrcken, da das aus asiddha sich Ergebende auf siddha bertragen werde, ohne da dieses zu funktionieren aufhrt. Hier kann natrlich nicht wei bei sthnivat eine einzige Operation in Betracht kom-

men, sondern auer der bertragenen, aus der Eigenschaft asiddha sich ergebenden Operation (asiddhatvsrayam) soll auch eine andere, auf der Eigenschaft siddha beruhende
(svsrayam, d . i . siddhatvsrayam)

eintreten drfen. 6 Wenn in da-dambh-atuh nach VI, 4, 24 der Nasal elidiert worden st, so kommt das a der Wurzel zwischen zwei e i nfache Konsonanten zu stehen und dann erfolgen gem VI, 4,120 Substitution von effro] und Elision der Reduplikation. So erhlt man debhatuh. Wenn jedoch wei es VI, 4, 22 verlangt die Elision des Nasals von dambh [VI, 4, 24] als asiddha betrachtet wird, so kann VI, 4r 120 nicht eintreten. Deshalb fordert Vrtt. 5 zu VI, 4,120 einen Zusatz zu dieser Kegel. Nach der von Kaiy. erwhnten Auffassung von vat dagegen kommt debhatur dadurch zustande, da auch svsrayam, d. i. siddhatvsrayam kryam eintreten darf, d.h. die aus der in Kraft getretenen (siddha) Operation VI, 4, 24 sich ergebende Operation VI, 4,120.

216 Bernhard Geiger

siddho sich Ergebende a u c h eintreten darf], zu einer Unterscheidung der Wirkungsbereiche des S/dd/ia-seins und As/ddho-seins gelangen ?7 Bei sthnivot [1,1, 56] dagenen wre [das W o r t sthn/"] ohne vat eine Bezeichnung [des Substitutes], 8 und daraus erkennt man, da die Setzung des vat die bertragung [des fr den sthnin Giltigen auf das Substitut] andeutet und den Zweck hat, da [die fr den sthnin geltende Operation] auch in bezug auf jenes [sc. das Substitut] selbst sich ergebe. 9 Daraus [jedoch], da in [VI, 4,111]: ,,snasor allopah" [dem o in at] ein t nachgesetzt ist, kann man erschlieen, da [eine Operation dieses Abschnittes t r o t z VI, 4, 22] bisweilen auch siddha sein kann. Denn welchen Zweck htte anderenfalls das Nachfolgen des t, da doch bei stm, asan [im Augenblick der Elision des a] das Augment noch nicht in Kraft getreten wre und infolgedessen die Elision [des a] sich gar nicht darbieten w r d e ? " '10 Hinsichtlich der Bedeutung des W o r t e s asiddha in unserem Stra verweist Kty. im Vrtt. 1 auf VI, 1, 86 Vrtt. 1-5. Ich lasse hier die bersetzung des Bhsyazu diesen Vrttikas folgen, da sie fr das Verstndnis von Kty.'s Auffassung unseres Stra von Wichtigkeitsind: P. VI, 1, 86: Satva-tukorasiddhah. ,ln bezug auf den E i n t r i t t von s [ f r s ] und [die Anfgung] des substituieren.' Dazu Pat.: Anityo 'yam D. h. wie ist es dann bei dem vidhir iti. ,D. h. diese Regel [VI, 4, 22] Gegensatz zwischen siddhatva und ist nicht immer giltig.' Und Kaiy. asiddhatva berhaupt mglich zu unerklrt: Snasor attva iti \ Asiddhatvaterscheiden, in welchem Falle siddhasynityatvajnpanya takrah krtah. tvsrayam kryom, und in welchem Falle die bertragene, aus asiddha sich Nitye tv asiddhatve sann itydv to 'siddhatvl lopo na bhavisyatiti kirn ergebende Operation eintreten soll ? 8 tannivrttyarthena takrenal TensiD. h. man wrde interpretieren: ddhatvbhvd [Text: tena s/'] dambha unter dem sthnin ist der desa ettvam siddhyati. ,Das t [in at] ist gemeint. Dann entstnde aber der gesetzt, um anzudeuten, da das Fehler, da das in I, 3, 28 gelehrte Asiddha-se\n [einer Operation] nicht tmanepadam nur bei dem Substitut durchwegs gilt. Angenommen aber, vadh, nicht aber bei dem sthnin han, das Asiddha-seln gelte durchwegs, so eintreten wrde Vgl. p. 11, Anm. 1 wird bei [der Bildung von] san usw., und Mbh. vol. I, p. 133 (Anfang). 9 Vgl. Ks. zu 1,1, 56: Vatkaranam kirn? da ja [zur Zeit der Anwendung von Sthny desasya sarpjn m vijnyitiVI, 4,111] das Augment noch gar svsrayam api yath syt. Ao yama- nicht vorhanden wre [vgl. p. 24, Anm. 1], dessen Elision nicht eintreten hanah [I, 3, 28]; hata, vadhistety knnen ; wozu wre unter d iesen Umtmanepadam ubhayatrpi bhavati. stnden das t ntig, welches [nach I, Hinsichtlich der bertragung vgl. 1, 70] ausdrcken soll, da diese Mbh. zu 1,1, 56 Vrtt. 1 : . . . Guruvad [Elision des Augmentes o] unterasmin guruputre vartitavyam iti gurau bleiben mge? [san usw. wrden aber y at kryam, tad guruputre 'tidisyate. ohne Schwierigkeit gem VI, 4, 22 Evam ihpi sthnikryam dese gebildet werden, wenn Pn. in VI, 4, 'tisdisyate. 10 111 a-lopah statt al-lopah gesagt Diese Bemerkung will das Zustandehtte. Das t mu also noch einen ankommen von debhatuh erklren und deren Zweck haben, u. zw. anzudeubezieht sich auf das Slokavrtt. Mhb. ten, da VI, 4, 22 bei debhatur nicht vol. Ill, p. 219: Snasor attve takrena jnpyate tv ettvassanam. rDa aber bei gelten mge.] Auf diese Weise kommt also dadurch, da das Asiddha,,snasor" [VI, 4,111] at vorliegt, so sein [bei VI, 4, 24] nicht stattfindet, wird durch den Buchstaben t die die Substitution von e [fr o] zuVorschrift angedeutet, [bei dambh stande.' nach VI, 4,120 gegen VI, 4, 22] e zu
7

217 Mahbhsya

Augmentes t st [ein ekdesa als] asiddha [zu betrachten].' Z u welchem Zweck w i r d dies gelehrt ? Satvatukor asiddhavacanam desalaksanapratisedhrtham utsargalaksanabhvrtham ca|| V r t t . 1. Da in bezug auf den E i n t r i t t von s [fr s] und [die Anfgung] des Augmentes t [ein ekdesa] asiddha sei, w i r d gelehrt, damit 1. die durch das Substitut bedingte [Operation] verboten werde, und 2. die durch das Ursprngliche 1 1 bedingte [Operation] stattfinde. Zunchst 1 . , damit die durch das Substitut bedingte [Operat i o n ] verboten w e r d e : [z. B.] ko^sicat, yoDsicat. W e n n [hier nach VI, 1,109] die Substitution des einen Vokales [ o f r o + a] vollzogen ist, w r d e sich gem [Vll\t 3, 59]: , Nach in' E i n t r i t t von s [ f r s ] ergeben. W e i l [aber VI, 1,109] als asiddha betrachtet w i r d , findet er nicht statt; 1 2 2. damit die durch das Ursprngliche bedingte [ O p e r a t i o n ] stattfinde: [z. B.] adhltya, pretya. W e n n [hiebei nach VI, 1,101 und 87] die Substitution des einen Vokales [J, bezw. e, fr /' + /', bezw. a + /'] vollzogen ist, so w r d e sich das Augment t, [welches nach VI, 1, 71 nur] ,an eine Krze' [angefgt w i r d ] , nicht ergeben. E s t r i t t [aber] ein, weil [VI, 1,101 und 87] als asiddha betrachtet werden. 1 3 Ist dies der Z w e c k [des W o r t e s asiddha] ? Was ist denn da gegen einzuwenden ? Tatrotsargalaksanprasiddhir utsargbhvt|| V r t t . 2. D o r t , bei adhltya, pretya, kann doch die durch das Ursprngliche bedingte Operation nicht zustande kommen. Weshalb? W e i l das Ursprngliche nicht mehr vorhanden ist. [In VI, 1, 71] w i r d g e l e h r t : ,An eine Krze [ w i r d t angefgt']; aber hier [in adh-ya, pre-ya] sehen w i r keine Krze mehr. E i n w u r f A b e r es ist doch vermge des W o r t e s asiddha vollstndig korrekt. 1 4 E r w i d e r u n g Asiddhavacant siddham iti cen nnyasysiddhavacand anyasya bhvah 1 1 V r t t . 3.15 W e n n jemand sagt, es sei vermge des W o r t e s asiddha v o l l stndig k o r r e k t , so t r i f f t dies nicht zu . W a r u m ? Daraus, da
11 12 Utsarga bezeichnet sonst eine allAus kos asicat wird nach VIII, 2, gemeine Regel im Gegensatz zu 66 kar asicat, nach VI, 1, 113 ko + apavda, der Spezial- oder Ausu asicat. nach VI, 1, 87 ko asicat. nahmsregel. [Vgl. Mbh. vol. I, p. 6 und nach VI, 1,109 (ekdesa): ko 'sicot. Paribh., Transi, p. 321, n. 1 ] . An Da diesses o nach VI, 1, 85 auch als unserer Steile dagegen dient utsarga An(aut des folgenden betrachtet zur Bezeichnung des sthnin, des werden kann [k=os'mcat], mte ursprnglichen Elementes, das durch nun nach VIII, 3, 59 s fr s eintreten, das Substitut aufgehoben wird. ZwiDiese durch das Substitut o bedingte sehen sthnin und desa besteht ja ein Operation wird jedoch durch VI, 1, hnliches Verhltnis wie zwischen 86 verhindert. 13 utsarga und apavda. Vgl. Kaiy. zu In diesem Falle hat VI, 1, 86 nicht unserer Stelle: Utsargah sthnl smden Zweck, eine durch das Substitut nyenotsrstatvt (,U. st der sth., weil [/fr/ + /] bedingte Operation zu dieser als etwas Allgemeines [durch verbieten, sondern die durch den den desa] aufgehoben wird'); Kaiy. utsarga [adhi + i-] bedingte Operation zu VI, 4, 22 Vrtt. 1: Utsargasabdena [VI, 1, 71] zu ermglichen. 14 smnyavisayatvasdharmyt sthny D. h.: weil das Substituais asiddha abhidhiyate (,. . . weil ihm [sc. dem bezeichnet wird, ist der utsarga als sthnin] eine generelle Funktion vorhanden zu denken, zukommt'); Ks., ed. Kalkutta, p. 183; 15 Vrtt. 3-5 sind gleich 1,1,57 Utsrjyate, desena hdhyata ity utsargah Vrtt. 4-6. sthnl.

218 Bernhard Geiger

das eine [se. der ekdesa] als asiddha bezeichnet w i r d , folgt nicht das Vorhandensein des andern [sc. des utsarga]. Denn dadurch, da das eine als asiddho bezeichnet w i r d , k o m m t nicht das andere wieder zum Vorschein. 16 W e n n nmlich auch der Mrder des Devadatta gettet worden st, so kehrt doch [dadurch] Devadatta nicht in die Existenz zurck. Tasmt sthnivadvacanam asiddhatvam ca|| V r t t . 4. Deshalb mte [in dem Stra] gesagt werden : es [sc. das Subs t i t u t ] verhlt sich w i e der sthnin und es ist asiddha. [ U n d zwar] verhlt es sich bei adhltya, pretya w i e der sthnin, whrend es bei ko ^sicat, yo ^sicat asiddha ist.17 Sthnivadvacannarthakyam sstrsiddhatvt | Vrtt. 5. [Doch] ist es unntig zu sagen ,wie der sthnin'. Warum ? Weil die Regel asiddha ist. Durch dieses [ W o r t asiddha] wird nicht bewirkt, da die Operation asiddha ist, sondern es bewirkt, da die Regel asiddha ist. Die den ekdesa betreffende Regel gilt als asiddha in bezug auf die Regel ber das Augment t.18 Da Kty. in dem Vrtt. 1 zu VI, 4, 22 auf diese Vrttikas verweist, nimmt er offenbar auch fr das Stra VI, 4, 22 an, da nicht eine Operation, sondern eine Regel als asiddha zu betrachten sei. Denn gegen die Auffassung, da eine Operation als asiddha anzusehen sei, knnte Kty. in den Fllen edhi, sdhi [Vrtt. 3], in denen das Stra den Zweck hat, da die durch den utsarga bedingte Operation eintreten mge, geltend machen, da nach der Ausfhrung der Substitution ein utsarga nicht mehr vorhanden ist. Demgem bemerkt Kaiy. zu Pat.'s Frage asiddhavacanam kimartham am Beginn dieses Abschnittes : ha sstrasya kryrthatvt kryasya prdhnyd asiddhatvena bhvyam. Tadasiddhv api sthnino nivrttatvt tallaksanam kryam na prpnotity avyptim matv prasnah. taro vypakatvc chstrsiddhatvam pradesntara eva vyavasthpitam [so Ms. K. ; Text: evam sthpitam] manyamna ha: asiddhava16 Kaiy.: Krysiddhatvsrayenedam ucyote, desena sthnino nivartitatvt saty opy desasysiddhatve sthninah pratypottyabhvt. .Dies wird behauptet von dem Standpunkt aus, da [die bereits vollzogene] Operat ion asiddha sei; denn da der sthnin durch das Substitut aufgehoben worden ist, kann der sthnin, auch wenn das Substitut asiddha st, doch nicht wieder eintreten.' 17 Kaiy.: Satve 'siddhatvam, sthnivadbhve tu svsrayasynivartant satvaprasagah. ,ln bezug auf den Eintritt von s [fr s] st [der ekdesa] asiddha [und nicht ,wie der sthnin1 zu behandeln]. Wenn er sich aber wie der sthnin verhielte, so wre [noch immer] der Eintritt des s mglich, weil das aus hm [sc. dem ekdesa] selbst sich Ergebende nicht aufgehoben wre.' Vgl. die Errterungen ber vat p. 2.15 ff. 18 Kaiy.: Tatas ca prvam tuksstram pravartate, pascad ekdesasstram

ity uktam bhavoti. .Damit st gesagt, da zuerst die tu/c-Regel [VI, 1, 71] eintritt und nachher die ekdesaRegel [VI, 1,101, bezw. 87].' Diese mte nmlich als paro- und n/tya-Regel gem Paribh. Nr. 38 frher eintreten als die tuk-Regel. Nach Kty.'s Ansicht dagegen wird folgender Vorgang beobachtet: bei od/i/ + / + ya ergeben sich gleichzeitig VI, 1,101 [ekdesa] und 71 [U/k]; da nun in bezug auf diese Regel jene als asiddha [ = nicht vorhanden] zu betrachten st, tritt trotz Paribh. Nr. 38 die Regel VI, 1, 71, also die durch den utsarga i -f / bedingte Operation, zuerst ein und dann erst nach 101 das Substitut /". In bezug auf die Verwandlung des sins ist es natrlich gleichgiltig, ob man die Operation oder die Regel asiddha sein lt. Denn hier ergibt sich VIII, 3, 59 erst nach der Ausfhrung von VI, 1,109.

219 Mahbhsya

cana uktam iti. .Weil eine Regel eine Operation zum Zwecke hat, und infolgedessen die Operation die Hauptsache ist, mu [die O p e r a t i o n ] asiddha sein. W e i l aber, selbst wenn diese [ O p e r a t i o n ] asiddha ist, der sthnin [durch sie] aufgehoben w o r d e n ist, so ergibt sich nicht mehr eine durch diesen [sthnin oder utsarga] bedingte O p e r a t i o n . In der Meinung, da aus diesem Grunde die Definition

[von asiddha, d . i . desalaksanapratisedhrtham utsargaiaksanabhvrtham ca] zu eng sei,19 wird die Frage [nach dem Zweck von ,asiddha'] gestellt. Der andere dagegen denkt daran, da schon an einer anderen Stelle festgestellt worden ist, die Regel [und nicht die Operation] sei asiddha, weil dies vollstndig [d. i. fr alle Teile der Definition von asiddha] zutrifft.' Nach Kty.'s Auffassung wrde also bei edhi, sdhi [Vrtt. 3] die durch den utsarga bedingte Operation VI, 4,101 frher eintreten als die sich gleichzeitig darbietenden Regeln VI, 4,119 und 35, weil diese in bezug auf VI, 4, 101 als asiddha [ = nicht vorhanden] zu betrachten sind. In den Fllen dagegen, in denen der Eintritt einer durch das Substitut bedingten Regel verhindert werden soll, t r i t t die Regel ber die Substitution ein, und diese Regel wird nun in bezug auf die zu verhindernde Regel als asiddha betrachtet. In diesen Fllen ist also der Vorgang genau so, wie wenn die Operation selbst als asiddha angesehen wrde. W i r werden natrlich Kty.'s Besorgnis wegen des utsarga nicht teilen. Denn wenn die Substitution, d. i. die bereits vollzogene Operation, als asiddha betrachtet wird, ist eben trotz Kty. er'utsarga als noch vorhanden zu denken. Was den Geltungsbereich unseres Stra betrifft, so habe ich schon oben bemerkt, da Kty. annimmt, es gelte einschielich des Abschnittes ber bha [VI, 4,129-175], also bis VI, 4,175. Pat. begngt sich damit, die Vrtt. 15 und 16 zurckzuweisen und zu zeigen, da weder aus der Annahme der Alternative ,prg bht' [d. i. bis VI, 4,129], noch auch aus der Wahl der Alternative , bht' [= saha tena] die von Kty. genannten Fehler sich ergeben. Eine bestimmte Antwort gibt Pat. nicht. In der Diskussion zu VI, 4,149 Vrtt. 3 wird es als offene Frage behandelt, ob ,prag bht' oder ,saha tena' gilt. Die Ks. zu VI, 4, 22 erklrt gleich Kty., da das Stra bis zum Schlu des Adhyya, also bis VI, 4,175 anzuwenden sei. Ebenso urteilt die Siddh. K. zu VI, 4, 22: i ta rdhvam pdaparisampter abhJyam .[eine Regel] von hier an weiter bis zum Abschlu des Pda heit eine bis bha gelehrte [Regel].' Zu dem eben besprochenen Abschnitt ist schlielich noch Candra V, 3, 21 [ed. Liebich, Abhandl. f. d. K. d. Morgen!, vol. 11, Nr. 4, p. 101]: ,Prg yuvor avugyug asiddham samnsraye' zu vergleichen. Der zweite der im folgenden bersetzten Abschnitte ist Mbh. zu VI, 4,132. Kty. erklrt das W o r t th im Stra fr berflssig, weil Formen wie prasthauhah auch durch Substitution des gewhnlichen Samprasrana u fr v zustande kommen. Demgegenber erklrt Pat., th habe den Zweck, die Par i bhs ,asiddham bahiragaaksanam antaragaiaksane' anzudeuten, welche verhindert, da prasthauhah usw. auf die von Kty. angegebene Weise gebildet werden. Diese Ansicht teilt auch die Ksik.
Avypti bedeutet, da ein Merkmal in einem Teile des zu Definierenden nicht vorhanden ist. Vgl. die Dipik zu Tarkasamgraha 2: Laksyaikadesvrttitvam avyptih, yath goh kapilatvam. In der Definition von
19

asiddha trifft, wenn die O p e r a t i o n als asiddha betrachtet wird, dieses Merkmal kryam (Operation) zwar fr die erste Hlfte der Definition zu, nicht aber fr die zweite.

220 Bernhard Geiger

bersetzung p, v i , 4, 22: Asiddhavad20 atr bht. Welchen Zweck hat das W o r t asiddha ?21 Asiddhavacana uktam|| Vrtt. 1. [,Es ist schon dort gesagt worden, wo von asiddha die Rede war']. Was st gesagt worden ? Dort [VI, 1, 86 Vrtt. 1 ] ist schon gesagt worden: ,Die Bezeichnung [des ekdesa] als asiddha in bezug auf den Eintritt von s und in bezug auf [die Anfgung] des Augmentes t hat den Zweck, da die aus dem Substitut sich ergebende [Operation] verboten werde und die durch das Ursprngliche bedingte [Operation] stattfinde.'22 Auch hier hat das W o r t asiddha den Zweck, da 1. die durch das Substitut bedingte [Operation] verboten werde, und 2. die durch das Ursprngliche bedingte [Operation] stattfinde. Zunchst 1., da die durch das Substitut bedingte [Operation] verboten werde, [z. B. in] gahi,jahi; gatah, gatavn : wenn [nach VI, 4, 37] die Elision des Nasals [von gam]23 und [nach VI, 4, 36] der Eintritt von ja [fr hon] vollzogen ist, so ergeben sich [VI, 4, 48] ,Elision des a [von ga- vor den rdhadhtuka-Suffixen ta und tavat]24 und [VI, 4,105] .Abfall des Suffixes hi hinter o' [von ga- und ja-]. Weil [aber VI, 4, 37 und 36] als asiddha betrachtet werden, findet dies nicht statt. 2. Da die durch das Ursprngliche bedingte [Operation] stattfinde, [z. B. in] edhi, sdhi: wenn bei den Wurzeln as und ss [vor hi gem VI, 4,119] der Eintritt von e [fr das s von as]25 und [nach VI, 4, 35] die Einsetzung von s [fr ss] vollzogen ist, so ergibt sich nicht die [nach VI, 4,101] durch einen K o n s o n a n t e n (mit Ausnahme der Nasale und Halbvokale) bedingte Substitution von dhi [fr hi]. Weil [aber VI, 4,119 und 35] als asiddha betrachtet werden, t r i t t [das durch das Ursprngliche, d. i. as und ss, bedingte dhi] ein.26
Kaiy.'s Errterungen ber die Bedeutung von vat s. Einleitung, p. 215 ff. 21 Kaiy.'s Bemerkung hiezu: i ha sstrasya etc. s. in der Einleitung, p. 218. 22 V g l . d i e Ks. zu V I , 4 , 2 2 . 23 Das Imperativsuflfix hi ist nach III, 4, 87 nicht pit, also gem I, 2, 4 /t. Kaiy. bemerkt zu gahi, da der Prsenscharakter a [sop: 111,1, 68] von gam nach II, 4, 73 abgefallen ist. 24 Kaiy.: Avayavalopinm ato lopo nstity etatparibhsrtham upadesagrahannuvrtty bhsyakrah sampdayisyati ,Dies wird der Verfasser des Bhsya [zu Vrtt. 6; p. 226] durch [die Annahme] richtigstellen, da das W o r t upadesa [aus VI, 4, 37] fortgilt, um die Paribhsanzudeuten: ,Bei [Stmmen], welche [wie gam das m] einen ihrer Teile durch Elision verlieren, findet die Elision von a nicht statt.' Vgl. p. 227, Anm ; 50.
25 20

Das o fllt gem VI, 4,111 aus, da hi Srvadhtuka und apit [III, 4, 87],

also /t [I, 2,4] ist. 26 Kaiy.: Edhti : paratvn nityatvc ca prvam ettvam; sdhity atrpi nityatvtprvam sbhvo; 'l-vidhitvc ca dhitve nsti sthnivadbhvah. .Zuerst [d. h. vor VI, 4,101] t r i t t e [VI, 4,119] ein, weil [VI, 4, 119] eine poro-und n/tyo-Regel st [d. h. vor oder nach dem EintrittvonVI, 4,101 eintreten kann; und zwar auch nach dem Eintritt des dhi deshalb, weil nach 1,1, 56 fr das Substitut dhi dasselbe gilt, was fr den sthnin hi gegolten hat]. Und auch bei sdhi t r i t t so [VI, 4, 35], weil [diese Regel] nitya ist, zuerst ein [d, h. vorderpora-Regel VI, 4,101. Vgl. Paribh. Nr. 38], Und es gilt nicht etwa hinsichtlich des Eintrittes von dhi [VI, 4,101, die Regel 1,1, 56], da [die Substitute e und s sich] wie die sthnin's [as und ss verhalten], da es sich [bei 101 entgegen dem Verbot in 1,1, 56] um eine Vorschrift handelt, welche auf L a u t e n [d.i. hier den Auslauten der sthnins as und ss] beruht.' Da also die Substitution von dhi nicht nach 1,1, 56 zustande

221 Mahbhsya

W e l c h e n Z w e c k hat denn aber das W o r t atra ?27 A t r a g r a h a n a m visayrtham 2 8 | | V r t t . 2. [Durch atro] wird der Geltungsbereich zum Ausdruck gebracht. Es soll ein [in dem Abschnitt] bis bha gelehrte Regel h i e r [d. h.] in bezug auf eine ebenfal Is [in dem Abschnitt] bis bha gelehrte Regel asiddha sein. [Hingegen] soll [das Stra] in den folgenden Fllen nicht gelten : abhji, rgah,29 upavarhanam.30 Welches sind nun aber die Zwecke dieser Regel ? Prayojanam saittvam dhitve || Vrtt. 3. Der Eintritt von so und von e ist ein Zweck hinsichtlich des Eintrittes von dhi, [z. B. bei] edhi, sdhi : wenn bei den Wurzeln as und ss der Eintritt von e [VI, 4,119] und der von s [35] vollzogen sind, so ergibt sich nicht der [nach VI, 4,101] durch einen K o n s o n a n t e n (mit Ausnahme der Nasale und Halbvokale)
kommen kann, bedrfen wir zur Bildung von edhi und sdhi der Regel VI, 4,22. 27 Kaiy.: Yathngasyetydayo 'dhikr etc. ,Wei sonst adhikros nach A r t von ,agosya' [VI, 4,1] u. a. m. auch ohne die besondere Hervorhebung der [Geltungs]grenze aus ihrer Kommentierung als eine bestimmte Grenze besitzend verstanden werden, so wird auch der adhikra ,asiddhavat' [in unserem Stra als] bis bha [geltend] verstanden. Weil [also] dort der Ausdruck , bht' den Zweck hat, den Geltungsbereich abzugrenzen, brauchte das W o r t atra nicht gesetzt zu werden. Dies ist der Sinn [der Frage].' 28 Kaiy.: visistoyo visayah etc. ,Um den [schon durch , bht'] bestimmten Geltungsbereich als charakterisiert durch [die Abhngigkeit beider Operationen von] einem gemeinsamen [Element] zu bezeichnen. Infolgedessen ist ein auf einem gemeinsamen [Element] beruhende [Operation] asiddha, hingegen eine nicht [auf einem gemeinsamen Element] beruhende [Operation] siddha: dies ist der vom Verfasser der Vrtti kas angenommene Zweck [des Wortes atra] [vgl. Vrtt. 12]. Auch der Verfasser des Bhsya [Pat.] wird spterhin [zu Vrtt. 12] eben diesen Zweck des Wortes atra feststellen. Jetzt dagegen nimmt er [vorlufig] an, der Ausdruck , bht' habe den Zweck, im Interesse leichterer Erkennbarkeit die Grenze des adhikra anzugeben, und k o m m e n t i e r t [nur], durch das W o r t atra werde ausgedrckt, da der Bereich, in welchem [die eine Operation des Abschnittes bis bha] als asiddha zu betrachten ist, sich nur auf eine [ebenfalls in dem Abschnitt] bis bha [vorkommende] Regel erstrecke.' Was Kaiy. hier von Kty. behauptet, ist unrichtig. Denn das W o r t visaya, das Kty. gebraucht, bedeutet,Geltungsbereich' und nichts weiter. [Vgl. 111,1, 92 Vrtt. 1 : tatragrahanam visayrtham.] Und dies st ja eben der Grund, weshalb Kty. im Vrtt. 12 den Zu sat zsamnsraya verlangt. Vgl. p. 235, Anm. 78. 29 In a-bhaj-i kann nach VI, 4, 33 der Nasal ausfallen und w i r erhalten a-bhaj-i. Da nun Pnultima (upadh) des Stammes ist, t r i t t nach VII, 2,116 Vrddhi ein: abhji. Dies wre nicht mglich, wenn VI, 4, 33 asiddha wre; denn dann wre a nicht Pnultima. In bezug auf die auerhalb des Abschnittes VI, 4, 22 bis bha stehende Regel VI1, 2,116 ist also eine im Abschnitt bis bha gelehrte Regel nicht asiddha. hnlich bei rgah: VI, 4, 27 und VII, 2, 116. Kaiy.: ,Die nach den Regeln . . . [VI, 4, 33 und 27] erfolgende Elision des n gilt nicht als asiddha in bezug auf die [Substitition von] Vrddhi, welche nach der Regel [VII, 2, 116]: ,,fro, welches Pnultima ist", vollzogen werden soll.' 30 Kaiy.: Vrhi vrddhv ity asya vrmher ocy anititi nalopo gune siddho bhavati. ,Die Elision des n erfolgt gem [VI, 4, 24 Vrtt. 2]: ,,Von der Wurzel vrmh d. i. derjenigen, [von welcher es im Dhtup. 17, 85 heit]: ,vrhi in der Bedeutung strken1 vor einem Vokal, wenn er nicht das Augment / ist" ;

222 Bernhard Geiger

bedingte E i n t r i t t von dhi [fr hi]. W e i l [aber V I , 4,119 und 35] als asiddha betrachtet w e r d e n , t r i t t [dhi] ein. 31 E i n w u r f Was zunchst den E i n t r i t t von so betrifft, so macht er [das Stra] nicht n o t w e n d i g . Ich w e r d e folgendermaen sagen : so hau [ V I , 4, 35] ist gleich ss hau,32 Der Laut s ist [nach (VIII, 2, 66 und) VIII, 3,17] zu y33 g e w o r d e n . D o r t [sc. bei dem Substitut ss] erfolgt [also] h i n t e r dem s [nach VI, 4, 101] der E i n t r i t t von dhi; das s aber w i r d gem [VIII, 2, 25] : ,auch v o r einem [ m i t ] dh [anlautenden Suffix]' elidiert. 3 4 O d e r aber ich w e r d e [das Stra ndern und] sagen : hau. Auch auf diese W e i s e ergibt sich f r den Lauts [ E i n t r i t t von dhi nach VI, 4 , 1 0 1 ] . Es gilt [nmlich in VI, 4, 34 und 35] ,fr die Pnultima' [aus 24] f o r t : wenn fr die Pnultima [von ss] eingesetzt w o r d e n ist,35 so erfolgt h i n t e r dem s [nach VI, 4,101] E i n t r i t t des dhi und gem [VIII, 2, 25] : ,auch v o r einem [ m i t ] dh [anlautenden Suffix]' Elision des s. O d e r aber ich w e r d e sagen [das Stra sollte lauten] : na hau. Nachdem d o r t [in V I , 4, 35: na hau] der E i n t r i t t von / [fr das von ss: V I , 4, 34] v e r b o t e n w o r d e n ist, erfolgt h i n t e r dem s der E i n t r i t t des dhi und gem [VIII, 2, 25] : ,auch v o r einem [ m i t ] dh [anlautenden Suffix]' Elision des s. Was ferner den E i n t r i t t des e [in der W u r z e l as: V I , 4 , 1 1 9 ] betrifft, so w i r d man ihn als Ausnahme von der [in 111 g e l e h r t e n ] Elision [des a von as] betrachten, und es ergibt sich [aus 119] nicht Elision dess. 36 [diese Elision] ist siddha in bezug auf [die Substitution von] Guna [VII, 3, 86].' Denn sonst knnte VI!, 3, 86 nicht eintreten, weil keine ku rze P n u l t i m a vorhanden wre. 31 So auch Ks. zu VI, 4, 22 und Siddh. K. zu VI, 4,119 und 35. 32 Kaiy. : Ihetvobdhanyo ssir etc. ,Der Sinn ist: hier [in VI, 4, 35] wird, um den Eintritt des [in 34 gelehrten] / zu beseitigen, ss allein a Is Substitut fr sos vorgeschrieben. Da also [das Substitut] auf einen Konsonanten endigt, ist der Eintritt des dhi [ohne Anwendung von VI, 4, 22] vollstndig korrekt.' Nach dieser Interpretation verhalten sich VI, 4, 34 und 35 folgendermaen: VI, 4, 34: Bei ss t r i t t unter den angegebenen Bedingungen /[frd] ein; VI, 4, 35: vor hi dagegen t r i t t ss [ohne bergang des in /'] ein. 33 Welches nach VIII, 3,19 abfllt. 34 Kaiy.: Dhi sakresicoetc. ,Es wird [hier] nicht [Slokavrtt. zu VIII, 2, 25] angenommen, da bei dem s vor einem dh Elision [nur] fr das s des Aoristes (,s/C) [und nicht auch fr das auslautende s einer Wurzel] zu lehren sei". Es wird [nmlich von Pat. zu VIII, 2, 25 im Gegensatz zum Slokavrttikakra] ausgefhrt werden, da nur cokdhi [mit Elision des Wurzelauslautes s, und nicht cakddhi] gebildet wird.' 35 D. h. VI, 4, 34 und 35 stehen in folgendem Verhltnis: VI, 4, 34: fr [die Pnultima] von ss t r i t t i ein; VI, 4, 35: vor hi dagegen t r i t t [fr die Pnultima von ss] dein. 36 Kaiy. : Snosor ollopa ity asya lopasypavdo etc. ,Als Ausnahme von der in der Regel ,snasorallopah' [VI, 4, 111] gelehrten Elision [des a von as] wird [in 119] Eintritt von ean Stelle des a vorgeschrieben. Weil diese Regel ,snasor allopah1, welche den [allgemeinen] Bereich [der Elision des ] ausdrcken soll, bei [119] fortgilt, so erfolgt nun durch diese [sc. 119] innerhalb des Bereiches der Elision [von a] Eintritt von e [fr o]. Dieser Sinn ergibt sich nach seiner Behauptung.' D. h. VI, 4,111 lehrt a1 1 g e m e i n Ausfal I des a von as vor einem Srvadhtuka, welches k/t oder /tist; VI, 4,119 aber lehrt die Ausnahme, da vor dem Suffix hi das ebenfalls Srvadhtuka und it ist a von S nicht elidiert wird, sondern da dafr e eintritt. VI, 4,119 lehrt also nicht die Substitution von efr das s von as. Das s selbst fllt dann nach VIII, 2, 25 ab.

223 Mahbhsya

Hilopauttve || Vrtt. 4. Die Elision von hi bildet eine Veranlassung [zur Anwendung von VI, 4, 22] in bezug auf den Eintritt von u: wenn [nmlich] bei kuru [nach VI, 4,106] die Elision des [Srvadhtuka] hi erfolgt ist, so ergibt sich nicht gem [VI, 4,110] : ,wenn auf [den Prsenscharakter] u ein Srvadhtuka folgt 1 Eintritt von u (fr das a von kar-u]. W e i l [aber VI, 4,106] als asiddha betrachtet w i r d , t r i t t [110] ein. 37 Einwurf Auch dies ist keine Veranlasung. Er w i r d [nmlich] dort [zu VI, 4,110 Vrtt. 1, Schlu] sagen, das W o r t srvadhtuka [in 110] habe den Zweck, da das u [fr a] eintreten mge, auch wenn ein Srvadhtuka nur eben v o r h e r d a g e w e s e n ist.38 Tstilopenyandes addvidhau || Vrtt. 5. 1. Die Elision [der Endung] ta [in der 3. sing. aor. pass.], 2. die Elision [des a] in der Wurzel as, sowie 3. die Substitution des Halbvokales fr [das / der] Wurzel / (.gehen1) bieten Veranlassung [zur Anwendung von VI, 4, 22] in bezug auf die Vorschrift ber die [Anfgung der] Augmenten und : 1. [Die Elision des ta, z. B.] akri, aihi: wenn [nach VI, 4,104] die Elision des ta [Substitutes fr u] vollzogen ist,39 so ergeben sich nicht mehr die [nach VI, 4, 71 und 72 nur], vor /i/n 1 eintretenden Augmente a [bei kr-i]40 und [bei Jh-i]. W e i l [aber VI, 4,104] als asiddha betrachtet w i r d , treten sie ein. 2. Die Elision [des a] in der Wurzel as und die Substitution des Halbvokales fr [das / der] Wurzel / (.gehen') geben Veranlassung [zur Anwendung von VI, 4, 22], bei san, yan : wenn
37

Kaiy. : Kurv iti e t c . , W e n n [nach

Anwendung von III, 1, 79 und III, 4, 87] kr^-u + hi vorliegt, so [bieten sich gleichzeitig zwei Regeln dar: 1. VI, 4, 106: Abfall des hi und 2. VII, 3, 84: Guna des Wurzelvokals wegen des folgenden Ardhadhtuka (III, 4,114) u;] weil es spter gelehrt wird [poro] als der Abfal I des hi, tritt zuerst Guna ein [und wir erhalten kar-\-u-{-hi. Nun bieten sich gleichzeitig zwei Regeln dar: 1. VI, 4,106: Abfall des hi und 2. VI, 4,110: Substitution von i/fr o vor einem Srvodhtuko. VI, 4,110 ist zwar poro, doch] erfolgt noch vor dem Eintritt des u der Abfall des hi, weil dieser nitya ist [d. h. vor oder nach der Substitution des u eintreten kann]. Dann aber [sc. nach dem Abfall des Srvadhtuko hi] ergibt sich der Eintritt des u [fr a] nicht mehr. Doch erfolgt er dadurch, da der Abfall des hi als asiddha betrachtet wird.' Dies ist auch die Ansicht der Siddh. K. zu VI, 4,110. 38 In der Diskussion zu VI, 4,110 Vrtt. 1 wird nmlich bewiesen, da das Wort ,srvadhtuke' in dieser Regel berflssig wre, da aus VI, 4, 106 ,utoh' zu ergnzen sei. Daraus schliet Pat., da srvadhtuke im

Sinne von bhutoprve 'pi srvadhtuke zu fassen sei. Man knnte nun sagen, srvadhtuke sei gnzlich berflssig, weil nach dem Abfall des hi die Substitution von u fr a gleichwohl gem 1,1, 62 erfolgen knne. Dagegen wendet dort Kaiy. ein: Atro na lumatngasyeti pratyayolaksanapratisedhd uttvam na syd iti bhtaprvagoty srvadhtukagrahand bhovati. ,Weil hier die aus dem [durch luk abgefallenen] Suffi r[/i/]sich ergebende [Operation VI, 4,110] durch [die Regel 1,1, 63]: ,,nicht an einem Stamme [bei Ausfall des Suffixes] durch luk, lup oder slu (lumat)" verboten wird, wrde u [fre] nicht eintreten knnen; also tritt dieses kraft des Wortes srvadhtuka ein nach dem Prinzip ,,wenn etwas frher dagewesen st".' Die Ks. zu VI, 4,110 stimmt Pat. zu. 39 Diese Regel tritt vor den Augmentregeln ein, ,weil sie nitya ist1 [Kai/.]. 40 Der Bildungsproze bis zur Anwendung der Augmentregeln ist: kr + c/Z + ^un); nach III, 1, 66 und VII, 2, ta 115: kr+(cin) + ta, und nach VI, 4, / 104: kr/.

224 Bernhard Geiger

in den W u r z e l n / und os [die Substitution des] Halbvokales [VI, 4, 81], bezw. die Elision [des a: 111] vollzogen sind, so ergibt sich, da [diese W u r z e l n dann] nicht mehr m i t einem Vokal beginnen, nicht das A u g m e n t [VI, 4, 72]. W e i l [aber VI, 4, 81 und 111] als asiddho betrachtet w e r d e n , t r i t t es ein. E i n w u r f Was zunchst die Elision [des a] in der W u r z e l as betrifft, so bietet sie keine Veranlassung [zur A n w e n d u n g von VI, 4, 22]. Das Vorgehen des Lehrers [Pn.] lt erkennen (jnpayati), da das A u g m e n t a grere Kraft besitzt [d. h. f r h e r e i n t r i t t ] als die Elision [des a], da er ja in [der Regel VI, 4 , 1 1 1 ] : ,snasorallopah' [dem d] ein t nachfolgen lt.41 Auch die Substitition des Halbvokales fr [das /' der] W u r z e l / (.gehen 1 ) gibt nicht Veranlassung [zur A n w e n d u n g von VI, 4, 22]. Hinsichtlich der Substitution des Halbvokales w i r d man eine T e i l u n g d e r R e g e l [ V I , 4, 82] v o r n e h m e n : [Zunchst l e h r t VI, 4, 81 allgemein] : ,Fr [das / der] W u r z e l / (.gehen 1 ) t r i t t [vor Vokalen] der Halbvokal ein.' Hierauf [VI, 4, 8 2 A ] : ,Fr das [auslautende] i [des Stammes], wenn er mehr als einen Vokal enthlt' ; [ d . h.] auch fr [das auslautende] / [des Stammes], welcher mehr als einen Vokal enthlt, t r i t t bei / (,gehen') der Halbvokal ein. Hierauf [ V I , 4, 82B] : ,Fr [ein / ] , dem nicht eine Konsonantenverbindung vorangeht, [erg. aus 81 :] t r i t t der Halbvokal e i n ' , und zwar nur 42 fr ein [auslautendes] /', wenn [der Stamm] mehr als einen Vokal [ = eine Silbe] enthlt. 4 3 D. h. Pan. htte a-lopoh sagen knnen. Wenn er ausdrcklich at setzt, so betont er [vgl. 1,1, 70], da bei as das k u r z e a , n i c h t aber ein langes a, elidiert wird. Er deutet dadurch an, da die Elision des o der Wurzel ausgefhrt wird, nachd e m das Augment a eingetreteten st. Kaiy.: Taparakaranasya prayojanam stm, sann tydv ti krte ,vrnd ngam ballya' ity ekdesam bdhitv m bhd krasya lopa iti ; yadl prg opah syt tato 'najditvd to 'bhvt tallopanivrttaye taparatvam na kartavyam syd ity arthah. ,Der Sinn ist: die Nachsetzung des t bezweckt, ba bei [der Bildung von] stm, san usw. nach erfolgtem Eintritt des Augmentes [also in -\-as-\-tm usw.] wobei der [Eintritt des] ekdes [fr + o: VI, 1, 90] durch [die Paribh. Nr. 55]: Eine den Stamm betreffende [Operation, d. i. VI, 4, 111] hat grere Kraft als eine [die Verbindung von] Lauten betreffende [Operation: VI, 1, 90]" verhindert wird nicht das [sondern das a] elidiert werde. Wenn dagegen v o r h e r [d.h. vor Eintritt des Augmentes ] die Elision [des a von as] erfolgte, dann brauchte, da [die Wurzel] nicht mehr mit einem Vokal beginnen wrde, und infolgedessen
41

das A u g m e n t a g a r n i c h t e i n t r e t e n k n n t e , nicht noch [dem a] ein t nachgesetzt zu werden, [um auszudrcken], da n i c h t die Elision dieses [Augmentes o] stattfinde.' san wird also ohne Anwendung von VI, 4,22 folgendermaen gebildet; as + an; -\-as + an\ +s -\-an=san.
42 D. h.: whrend bei der ersten Teil regel [VI, 4, 82A]: ,eranekcah' aus 81 nich ,inah' zu ergnzen st, gilt bei der zweiten Teilregel [VI, 4, 82B]: ,asamyogaprvasya' nicht mehr ,inah' aus 81,sondern nur noch ,eranekcah' aus 82A. fort.

Die zweite Teilregel, die sich nicht mehr auf die Wurzel / (.gehen') bezieht, gibt also den Sinn der ungeteilten Regel 82 wieder. Fr die Bildung von yan kommt die erste Teil regel [82A] in Betracht, die besagen soll, da der Halbvokal bei der Wurzel / ferner eintritt, wenn der Stamm mehr als e i n e n Vokal enthlt also schon mit dem Augment a versehen ist. Kaiy. : tam antarennekj in na bhavatlti prvam d bhavatlty anumlyate; Jyatur itydau yogavibhgam antarena yanah siddhatvd dvidhyartham eva yogavibhgo vijnyate; tena nityo 'pi yandeso 'krta ti na bhavati. ,Ohne das Augment enthlt [der Stamm

43

225 Mahbhsya

Fr alle [im Vartt. 5 genannten Flle] aber lassen sich [Schwierigkeiten durch die Annahme] vermeiden, [aus VI, 4, 62] gelte f o r t ,upadese' [ d . i. ,bei der im grammatischen System gebrauchten G r u n d f o r m ' ] . Unter dieser Voraussetzung treten die Augmente a und a [nur] ein, solange [die W u r z e l n ] sich im Zustand der im grammatischen System vorliegenden Grundform befinden. 44 Oder aber [aus VI, 4, 46] gilt f o r t ,rdhadhtuke, [,vor einem rdhadhtuka].*5 Oder aber es handelt sich bei [der Regel V I , 4, 71]: Ju-la-lrksv at' um eine Bezeichnungsweise, die [in lu, Ion, Ir je] zwei / wiedergibt; 4 6 [die Regel besagt demnach : die Augmente a und treten] bei folgendem lu usw. [ein], sofern diese [noch] mit dem Laut / beginnen. 47 E i n w u r f Bei jeder [dieser drei Annahmen] kommen aijyata und aupyata [3. sing. impf, pass.] nicht zustande. 48
von] / (.gehen1) nicht mehr als einen Vokal, also t r i t t [gem 82A] z u e r s t [d. . vor der Substitution des Halbvokales] das Augment dein: so wird gefolgert. Da nun bei lyatur usw., [wo der Stamm ja auch mehr als einen Vokal enthlt], der Halbvokal [durch die allgemeine Regel 81] ohne eine Teilung der Regel [82] zustande kommt, so ergibt sich die Auffassung, da die Teilung der Regel einzig und allein im Interesse der Regel ber das Augment [72] vorgenommen wird. Demnach findet die Substitution des Halbvokales, obwohl sie [in bezug auf 72] nitya ist, nicht statt, solange das Augment nicht vorgesetzt worden ist.' yon wird also gebildet: i-\-an; -\-i+an; -f-/+ofi yan. ber Yogavibhgo vgl. Kielhorn, Ind.Ant. XVI, 247. [Cf. this volume, page 127] 44 Kaiy.: Antaragan api vidhln bdhitv lundyupadesa evdtau bhavata ityarthah. .Der Sinn ist: die Augmente o und o treten [nur] vor der im grammatischen System vorliegenden Grundform von lu usw. ein, indem sie sogar Antaraga- Rege In verdrngen.' Wenn z. B. san gebildet werden soll, so st der upadesa des Ion: as + la. Hier bieten sich nun, durch la veranlat, g l e i c h z e i t i g zwei Operationen dar: 1. die Substitution der Personalendung fr la, 2. die Vorsetzung des Augmentes. Die erste Operation st antaraga, weil die Veranlassung zu ihrer Ausfhrung nher liegt [s. unten zu Mbh. Ill, p. 223, I. 6] als die Veranlassung zum Vollzug der zweiten Operation, die deshalb bahiroga st. Die AntaragaOperation sollte nun gem Paribh. Nr. 50 frher eintreten. Infolge der Ergnzung von ,upadesel aber erfolgt zuerst die Vorsetzung des Augmentes; denn solange noch der upedesa des lu usw. vorliegt, also die Substitution der Personalendungen fr lu usw. noch nicht erfolgt ist, sollen die Augmente eintreten. 45 Kaiy.: Ardhadhtukagrahannuvrttismarthyd akrtesu lundesesu lvasthym labdhrdhadhtukasamjnym adttau bhavata ty arthah. Vermge des U m s t a n d e s , da das W o r t ,rdhadhtuka' fortgilt, treten die Augmente a und ein, solange die Substitutionen fr lu [usw.] noch nicht vollzogen sind, whrend noch der durch la [gekennzeichnete] Zustand [vgl. Ill, 4, 77] vorhanden st, dem die technische Bezeichnung rdhadhtuka eigen st [wogegen nach III, 4,113 die substituierten Personalendungen srvadhtuka heien].' 46 D.h. jedes / = / / . 47 Erg.: und nicht die entsprechenden Personalendungen fr sie substituiert worden sind. Dieselben drei Annahmen sowie derfolgende Einwurf auch Pat. zu VI, 4, 74. 48 Aijyata z. B. w i r d [nach d e r Ks. zu VI, 4, 72] gebildet: .Wenn la eingetreten ist [: yaj+lan], so [ergeben sich] in [diesem] durch la gekennzeichneten Zustand [gleichzeitig 1. Substitution von ta fr la und 2. Vorsetzung des Augmentes o; aber] vor dem Eintritt des Augmentes a wird die Substitution [von ta] fr la vollzogen, weil sie antaraga st [vgl. Anm.44. W i r erhalten yaj +ta]. Dann [bieten sich gleichzeitig 1. das Augment a und 2. nach 111,1, 67 das Zwischensuffix (vikarana) yak dar;

226 Bernhard Gelger

E r w i d e r u n g Er [Slokavrtt. zu VI, 4, 74] w i r d folgendes sagen: ,Bei vokalisch anlautenden [Stmmen] k o m m t [aijyatausw.] m i t Hilfe des Augmentes o [und nicht ] zustande. 49 Anunsikalopo hilopllopa/or jabhvas ca || V r t t . 6. Die Elision des Nasals gibt Veranlassung [zur Anwendung von VI, 4, 22] hinsichtlich der Elision von hi und a, sowie ferner der E i n t r i t t von j a [fr hon], bei gahi; jah'r, gatah, gatavn: wenn [nach VI, 4, 37] die Elision des Nasals und [nach 36] der E i n t r i t t des ja vollzogen sind, so ergibt sich gem [den Regeln] .Nach a [Elision] von hi' [VI, 4,105] und .[auslautendes] a w i r d [vor einem rdhadhtuka] elidiert' [48], Elision [des hi in ga-hi und ja-hi und des a in ga-tah und ga-tavn]. Sie findet [aber] nicht statt, weil [VI, 4, 37 und 36] als asiddha betrachet werden. doch] verdrngt der vikarana, da er nityo ist, das Augment a [und wir erhalten yaj+ya + ta. Man knnte dagegen einwenden, da auch die Anfgung des Augmentes a nitya sei, weil sie sowohl vor als auch nach dem Eintritt von yak erfolgen knnte; aber] das Augment a ist [nach Paribh. Nr. 43] onitya, weil sich [nach dem Eintritt von yok] eine andere Wortform ergeben wrde. Denn nach erfolgtem [Eintritt von yak] htte jenes [Augment a] bei dem auf das Zwischensuffix [yak] ausgehenden Stamm einzutreten; wenn [yak] hingegen nicht eingesetzt worden ist, [htte das Augment a] bei der bloen Wurzel [einzutreten]. Eine Regel ist aber anitya, wenn sie sich [nach dem Eintritt einer zweiten gleichzeitig sich darbietenden Regel] bei einer [hiedurch] genderten Wortform ergbe. [Einwurf]: Aber gem [Paribh. Nr. 44]:,Hinter einer genderten Wortform' wre doch [auch die Anfgung des] vikarana [yak] anitya, [da die Wortform, hinter welcher er eintreten wrde, wenn das Augment frher eingetreten wre, anders wre, als vor dem Eintritt des Augmentes]? [Die Ks. gibt keine Antwort. Man hilft sich, indem man sagt: der vikarana ist nach Paribh. No. 46 nitya]. Nach erfolgtem [Eintritt des] vikarana findet vor [dem Eintritt des] Augmentes a [die sich gleichzeitig darbietende Substitution des] Samprasrana [i fr y nach VI, 1,15] statt, da diese eben nitya st [und wir erhalten ij-\-ya-{~ta]. Da nun ein vokalisch anlautender Stamm entstanden ist, tritt nach VI, 4, 72 das Augment ein [und wir erhalten gem VI, 1, 90: aijyata].' Diese Bildung von aijyata usw. wre nach der Behauptung des Oppenenten bei keiner der drei Annahmen mglich. Vg. Kaiy.: Trisv api parihresu usw. .Bei allen drei Annahmen zur Vermeidung [von Schwierigkeiten] wrden die Wurzeln yaj usw., da in dem durch la gekennzeichneten Zustand [also vor dem Eintritt des k/t-Suflfixes yok] Samprasrana [VI, 1,15] nicht eintritt, nicht mit Vokalen beginnen: infolgedessen knnte das Augment nicht eintreten folglich wrde sich das Augment a ergeben.' Wir bekmen dann: a-\-yaj + an; a-\-yaj-\-ta; a-\-yaj-\-ya-\-ta und schlielich a + ij+ya + ta, was jedoch nach VI, 1, 90 nicht aijyata ergeben kann.
Kaiy. : tas cet i yat strarn tad atas ceti kriyate, acity adhikrc ca hali vrddhyabhavah. Katham yann, sann iti, yvatenastyor yanlopayoh krtayor vrddhir nstil Antaragatvad vrddhi r bhavisyatity adosah. Nnsrayatvc ca vrnd agam baliya iti nsti. Krtdym vrddhv sann iti taparakarand kralopbhvah. no 'pi vrddher aikrasya yan na bhavisyatlti ; no yan er iti yogavibhgd krntasyeno yanvidhnt tadabhvc cydese [Text: ca ydese (sic!)] krte yann iti bhavisyati. ,Das Stra ,tas ca' [VI, 1,
49

90] wird gendert in ,atasca' [d. h. auch fr das Augment o + Vokal tritt Vrddhi ein], und aus dem [hier fortgeltenden] Adhikra, ,vor einem Vokal' [VI, 1, 77] ergibt sich, da vor einem Konsonanten Vrddhi nicht eintritt. [Einwurf]: Wie kommen aber [unter dieser Voraussetzung] yan, san zustande, da sich doch bei den Wurzeln /' und as, wenn [in a-\-i -\-an und a + as + an] die Substitution des Halbvokales [fr /] und die Elision

227 Mahbhsya

E i n w u r f Was zunchst die Elision des Nasals betrifft, so gibt sie keine Veranlassung [zur A n w e n d u n g von VI, 4, 22]. Bei der Elision des o [VI, 4, 48] st [nmlich] ,upadesel [aus 37] zu ergnzen. 50 G e g e n e i n w u r f W e n n ,upodesel zu ergnzen st, so e r g i b t sich d o r t [III, 1, 80] nicht dhinutah, krnutah." E r w i d e r u n g Dies bedeutet keinen Fehler. Mit dem W o r t e upadeso w i r d nicht auf die ursprngliche Form Bezug genommen, sondern auf des rdhadhtuka w i r d Bezug genommen. [ G e m e i n t ist ein Stamm], welcher beim A n t r i t t eines rdhadhtuka an die im grammatischen Lehrsystem gebrauchte Form [der W u r z e l ] auf a auslautet. 52 [des o von as] vollzogen sind, Vrddhi nicht mehr ergibt? [Erwiderung]: Es liegt kein Fehler vor, denn [die Substitution von] Vrddhi wird [noch v o r diesen Operationen] eintreten, weil sie [als nherliegend: vgl. p. 225, Anm. 44] eine Antara/igo-Operation ist. Und weil ferner [die beiden Operationen] auf v e r schiede nen [Elementen, die Vrddhi auf dem Augment o-f-folgendem Vokal, der Halbvokal, bezw. die Elision, auf den Suffixen] beruhen, so gilt hier nicht [die Paribh. Nr. 55]: ,Eine den Stamm betreffende [Operation, hier: Substitution des Halbvokals, bezw. Elision des o] hat grere Kraft [ t r i t t frher ein] als eine [die Verbindung von] Lauten betreffende [Operation, hier: die Vrddhierung].' [Vgl. dagegen Paribh. Text, p. 60, I. 9f. (Transi., p. 303): Yot tu samnanimittikatvarpasamnsrayatva evaiseti tan na]. ,Wenn nun Vrddhi [zuerst] eingetreten ist, so findet bei san infolge der Nachsetzung des t [hinter dem a von allopah in VI, 4,111 nach 1,1,70] nicht noch Elision des statt. Bei der Wurzel / hinwiederum wird fr die Vrddhi ai [aus a + i + an] nicht noch der Halbvokal eintreten. Da [nmlich] durch die Teilung der Regel in ,inoyan er' [VI, 4, 82 A; vgl. p, 224] die Substitution des Halbvokales nur bei der Wurzel /', sofern sie auf auslautet, vorgeschrieben wird, und weil dieses [/' nach der Vrddhierung in ai-\-an] nicht mehr vorhanden st, wird nach vollzogener Substitution von y [fr ai gem VI, 1, 78] yan zustande kommen.' 50 D. h. bei einem Stamme, dessen ursprngliche Form d. i. die im grammatischen System gebruchliche Form der Wurzel auf a auslautet, wird dieses a vor einem rdhadhtuka elidiert. Ga- hingegen ist das Ergebnis einer an der upodesa-Form gam vorgenommenen Operation. Hier kann also VI, 4, 48 berhaupt nicht angewendet werden. Vgl. p. 220, Anm. 24. 51 Dhinutah wird aus der upodes'a-Form dhivi [Dhtup. 15, 84] = dhinv [VII, 1, 58] gebildet, indem nach 111,1, 80 an dhinv das Suffix u angefgt und [ g l e i c h z e i t i g ] fr das auslautende v ein a substituiert wird. W i r erhalten dhino + u + toh. Das a wird nun wegen des folgenden rdhadhtuka-SufTixes u nach VI, 4, 48 elidiert, so da sich dhinutah ergibt. Wenn jedoch bei VI, 4, 48 upadese zu ergnzen wre, knnte die Elision des a nicht erfolgen weil nicht dhina- upadesa-Form st, sondern dhinv. 52 Eigentlich: in [dem Stadium] der upodeso-Form, [wie sie] beim Antritt eines rdhadhtuka [erscheint]. Kaiy.: rdhadhtukopadesakle yad akdrntam agam, tasyrdhadhdtuke parato opa iti strrthah. Tatra dhinvikrnvyor a cet/ usw Weil in [der Kegel III, 1, 80]: ,,Bei dhinv und krnv t r i t t [u und] auch a [fr v] ein" zugleich mit dem [rdhadhtuka-]SufT\x u [die Substitution des] a vorgeschrieben wird, so lauten dhinv und krnv in der upodeso-Form, [wie sie] beim Antritt eines rdhadhtuka [erscheint], auf o aus; also kommt unter jener Voraussetzung [tatra; d. i. gem der Annahme ,rdhadhdtukofxjdese']die Elision des a [VI, 4, 48] zustande. Und weil dieser [lopa, welcher ein Substitut (desa) des ursprnglichen a ist], durch ein folgendes [Element, sc. das u] bedingt ist, sich also [gem 1,1, 57 in bezug auf die Regel VII, 3, 86, welche den dem sthnin vorangehenden Laut / betrifft], wie der sthnin [se. a] verhlt, so erfolgt bei dhinutah usw. nicht [die Substitution

228 Bernhard Geiger

Auch der E i n t r i t t von ja [fr han] g i b t keine Veranlassung [zur A n w e n d u n g von VI, 4, 22]. Hinsichtlich der Elision von hi [VI, 4,105] w i r d [nmlich] eine Teilung der Regel [106] v o r g e n o m men w e r d e n . [Zunchst 105]: .Nach einem a [erfolgt Elision] von hi.' Hierauf [106A.] : ,Auch nach ul ; [d. h.] auch nach einem u findet Abfall des hi statt. Hierauf [106B.]: . W e n n es ein Suffix ist' ; .wenn es ein Suffix ist', ist in beiden Fllen [sc. in 105 und 106A.] zu ergnzen. 53 E i n w u r f A b e r w a r u m heit es denn [im V r t t . 6] ,die Elision des Nasals in bezug auf die Elision von hi und a, sowie der E i n t r i t t von ja', w a r u m heit es nicht ,die Elision des Nasals und der E i n t r i t t von ja in bezug auf die Elision des a und hiV E r w i d e r u n g Damit nicht die [in I, 3 , 1 0 g e l e h r t e ] Entsprechung [der in gleicher Zahl einander gegenberstehenden Glieder des Satzes] der Zahl nach zur A n w e n d u n g komme. [ D e n n ] die Elision des Nasals gibt Veranlassung [zur A n w e n d u n g von VI, 4, 22 auch] in bezug auf die Elision von hi, in : ,Mandki tbhir gahi1 [VS. 17, 6 ] ; ,rohidasva ihgahi1 [VS. 1 1 , 72]; ,marudbhir agna gahi' [RV.1,19,1].54 Samprasranam avarnalope || V r t t . 7. [Die Substitution von] Samprasrana gibt Veranlassung [zur A n w e n d u n g von VI, 4, 22] hinsichtlich der Elision des Lautes a, von] Guna [fr das / von dh'm-].' D. h. trotz der erfolgten Elision ist der Stamm in bezug auf VII, 3, 86 noch als auf o [dhin{a)-\-uJrtah'\, also auf z w e i Laute ausgehend zu denken. Das / ist dann nicht Pnultima, und VII, 3, 86 kann nicht angewendet werden. Vgl. Siddh. K. zu VI, 3, 48: rdhadhtukopadesakle yad akrntam tasykrosya lopah syd rdhadhtuke pare. Ebenso urteilt offenbar auch die Ks., die zu VI, 4, 22 unter den Fllen, in denen diese Regel zur Anwendung kommt, zwar gahi und jahi, nicht abergotoh und gotavn erwhnt. 53 Kaiy.: Samudypeksant ; ,asamyogoprvd' ity atra tu bhedenpeksand nantaryd uta sambadhyate, na tv ata iti ,[Pratyayt ist in beiden Fllen zu ergnzen], weil es [zum Vorhergehenden] im Verhltnis der Anreihung steht; dagegen ist hier bei ,asamyogaprvt1, weil dessen Beziehung [zu dem Vorbergehenden] in einer Unterscheidung [d. i. Einschrnkung] besteht, [nur] ,utas' [106A] gemeint, da [asamyogaprvt] unmitelbar [auf ,utas'] folgt, nicht aber ,atas' [105].' D. h. pratyayt st auch in 105 zu ergnzen, asamyogaprvt dagegen nur in 106A. Die Regel 105 besagt demnach: nach a, jedoch nur wenn es Suffix ist, erfolgt Abfall von hi. Ja-hi, bei dem dies nicht zutrifft, also nicht die Anwendung von VI, 4, 22. Wenn das Varttika lautete, I II 1 ,anunsikalopa-jabhvv; allopa2 hilopayor', so wrde sich nach I, 3,10, ebenso wie die Glieder 1 und I einander entsprechen, auch hilopa nur auf jabhva, nicht aber auch auf anunsikalopa beziehen. Doch beweist [nach Kty.'s Ansicht] die Form gahi [aus gam-hi], da anunsikalopa auch in bezug auf hilopa als asiddha zu betrachten ist. Kaiy.: Yadyapi prayojankhynaparatvd usw. ,Fr diesen Satz wrde zwar die Entsprechnng der Zahl nach nicht gelten, weil er [nicht ein Stra st, sondern] die ausschlieliche Bestimmung hat, Zwecke [eines Stra] anzugeben. Aber weil man auf Grund der Beobachtung, da sonst eine Reihenfolge den Grund fr ein proportionales Verhltnis [ihrer Glieder: vyavasth] bildet, dasselbe auch in unserem Falle vermuten knnte, so ist, um dies gnzlich auszuschlieen, [im Varttika] diese A r t des Ausdrucks gewhlt worden.' Zu gahi bemerkt Kaiy.: Atrpi pratyayd ity asyobhayoh sesatvd dher lugabhvah sidhyati. .Daraus, da ,pratyayt' in beiden [Regeln: 105 und 106A.] zu ergnzen st [vgl. oben nebst Anm. 53], ergibt sich, da das hi auch hier [bie ga-hi] nicht abfllt.'
54

229 Mahbhsya

[z. B.] maghonah (ace. pi.), maghon, maghone. W e n n [nmlich in magha + van + ah gem VI, 4,133 die Substitution des] Samprasrana [ u f r v] vollzogen [und nach VI, 1,108 u fr u + a eingetreten] ist, so ergibt sich gem [der Regel VI, 4,148]: ,Fr / und o [vor der Femininendung /" und vor einem Taddhita]1 Elision [des a in magha + un - f oh v o r dem Taddhita un (aus van)]. 55 W e i l aber [VI, 4,133] als asiddha betrachtet w i r d , t r i t t sie nicht ein. 56 E i n w u r f Dies ist keine Veranlassung [zur Anwendung von VI, 4, 22]. Er w i r d [nmlich im Bhsya zu IV, 1, 7] folgendes sagen : Maghavan ist ein [etymologisch] nicht abgeleiteter Nominalstamm. 57 Rebhvallope || Vrtt. 8. Der Eintritt von re [fr ir] gibt Veranlassung [zur Anwendung von VI, 4, 22] hinsichtlich der Elision von , in : ,Kim svid garbham prathamam dadhra pah' [vgl. RV. 10, 82, 5]. W e n n [nmlich in da-dh + ire gem VI, 4, 76] der Eintritt von re [fr ir] erfolgt ist,58 so ergibt sich [in da-dh-re] nicht gem [der Regel 64] : ,A w i r d auch v o r dem [einem Ardhadhtuka vorgesetzten] Augment / elidiert' Elision des . W e i l [aber VI, 4, 76] als asiddha betrachtet w i r d , t r i t t sie ein. 59
55

Kaiy.: Maghona iti : mogham osystiti cchandaslvanipau ceti vanip ,[Maghavn] bedeutet: ihm ist eine Gabe eigen [vgl. V, 2, 94]. Es est [nmlich mit dem Taddhita-Sufx] von [gebildet], nach [dem Vrtt. zu V, 2,109]: ,lm Veda [werden] auch noch [die Suffixe] / und van [im Sinne von mat (matup: V, 2 94) angefgt].' 56 Asiddham bahiragam antaraga iti paribhsaysamprasranasysiddhatvarn nstiy esa hi paribhs vaha d ity atra jnpitatvd bhcchstriy tasym pravartamnym akralopasamprasranayor asiddhatvd antaragabahiragayor yugapad anupasthnn, njnantarya iti pratisedhd v. ,[Die Substitution des] Samprasrana ist [nur gem VI, 4, 22 und] nicht auf Grund der Paribhs [Nr. 50]: Eine ahiranga-Regel st asiddha in bezug auf eine Antaraga- Regel" als asiddha zu betrachten. Denn wenn dieses Paribhs welche nmlich eine [im Abschnitt] bis bha [VI, 4, 22175] gelehrte [Regel] ist, weil sie in [der Regel VI, 4,132]: ,vhat'[s. Pat. hiezu] angedeutet wird zur Anwendung gelangt, so sind sowohl die Elision des a [d. i. die AntaragaOperation], als auch [die Substitution des] Samprasrana [d. i. die ahirangaOperation gem VI, 4, 22] als asiddha zu betrachten, und infolgedessen sind die Antaraga-und die ahirangaOperation nicht g l e i c h z e i t i g zur

Stelle [vgl. nheres pp. 236 ff., nebst Anm.]; oder die Paribhs [Nr. 51]: ,,[Eineo/i/rorigo-Operation ist] nicht [als bahiraga, also auch nicht als asiddha zu betrachten, wenn die Anorogo-Operation] auf der unmittelbaren Folge eines Vokales [beruht]" verbietet [hier die Anwendung der Paribh. Nr. 50].' 57 Van ist also nicht als Suffix (pratyaya), und speziell Taddhita zu betrachten. Jene Operationen, welche sich im Falle einer etymologischen Herleitung des Wortes maghavan ergeben wrden hier die Elision des a [VI, 4,148] sind also unmglich, und VI, 4, 22 hat deshalb hier keine Gelegenheit zur Anwendung. Vgl. Paribh. Nr. 22: Undayo 'vyutpannni prtipadikni. Kaiy.: Tatas ca taddhite usw. ,Und deshalb t r i t t die vor einem Taddhita vorgeschriebene Elision des a nicht ein. Zumal da [VI, 4,133] auch deshalb nicht [in bezug auf 148] asiddha sein knnte, weil [die zwei Operationen] nicht [auf einem gemeinsamen Element] beruhen [vgl. zu Vrtt. 12], mu man die Alternative, da [maghavan] etymologisch nicht ableitbar st, annehmen.' Vgl. Undistras [ed. Aufrecht] 1,158. 58 Kaiy.: Nityatvt. .Weil [diese Operation in bezug auf VI, 4, 64] nitya st [und nach Paribh. 38 zuerst eintritt].' 59 Soauch Ks. undSiddh. K.zu V l . 4 , 7 6

230 Bernhard Geiger

E i n w u r f Auch dies g i b t nicht Veranlassung [zur A n w e n d u n g von V I , 4, 22]. Der E i n t r i t t von re [fr ir] ist [nmlich] vedisch. Und die Personalendungen des Perfektums 6 0 sind [nach III, 4,117] im Veda auch Srvadhtuka. U n t e r dieser Voraussetzung (tatra) e r g i b t sich aus [der Regel I, 2, 4] : ,Ein Srvadhtuka, welches nicht pit ist, gilt als /7/t-Suffix1, da [re] ein it-Sufix ist, und demgem erfolgt die Elision des [in da-dh + re] nach [der Regel V I , 4, 112]: ,Fr das d von (s)n [dem Prsenscharakter der IX. KL] und das der reduplizierten Stmme [ v o r einem Srvadhtuka, welches k / t o d e r n/t ist]. 61 W e n n demnach diese Regel [ V I , 4, 22] nicht aufgestellt w i r d :

Zwecke des Slokavrttikaklra 62 Ut tu krah katham or vinivrttau ; ,Wie soll denn aber in kr u [fr o] eintreten, nachdem [der Prsenscharakter] u verschwunden ist V
60

Die sonst nach III, 4,115 rdhadhtuka sind. 61 Kaiy. : Srvodhtuke sapslau dvirvacane 'ghor iti itvanisedhd [Text: itva-] kralopah ,Wenn [der Reihe nach] das Srvadhtuka [re] eingetreten, die Abwerfung des Prsenscharakters a (sap) [111,1, 68 wegen des Srvadhtuka ; II, 4, 75] und die Reduplikation [VI, 1, 10 und 8] erfolgt ist, wird das [des Stammes gem VI, 4,112] elidiert, da ja durch [VI, 4,113]:,,[/ t r i t t an die Stelle von vor einem konsonantisch anlautenden Srvadhtuka, jedoch] nicht bei da und dh (ghu)" der Eintritt von 1 [fr ] verboten wird.' 62 Kai/.: .Nachdem die vom Verfasser der Vrttikas [Kty.] genannten Zwecke [von VI, 4, 22] widerlegt worden sind, werden die vom Slokavrttikakra behaupteten Zwecke vorgefhrt.' 63 In VI, 4,110 gilt nmlich aus 106 ,utah' fort. 64 Die genannten Formen werden folgendermaen gebildet: kr-\- vah; nach 111,1, 79: kr - f u + vah] nach VII, 3, 84, die als para-Regel frher als VI, 4,108 und 109 eintritt : kar + u -f vah ; nach VI, 4,108 und 109: kar +

f vah
{ mah

und kar -f yt. Die eben erfolgte Elision des u mu als asiddha betrachtet werden, wenn nun nach 110 kurvah usw. zustande kommen sollen. Kaiy.: Ut tu kra iti, srvadhtuke para ti. Atha usw. ,Wenn ein Srvadhtuka [auf das Suffix u] folgt': ,[E n wu rf:] Aber warum erfolgt denn

der Eintritt des u [fr a] nicht [nach dem Wortlaut von VI, 4,110], wenn nur ein Srvadhtuka [ohne vorangehendes Suffix u] folgt? [ E r w i d e r ung:] Dies ist nicht mglich. [Denn sonst] wrde der Eintritt des u [fr o] nur dort erfolgen, wo die Elision [des Suffixes] u vollzogen worden st [wie in kurvah aus kar + vah], nicht aber bei kuruta usw., weif hier [das u] als Innensuffix [kar- und Srvadhtuka trennend] dazwischentritt. Wenn man dagegen geltend machen wollte, da ja auch bei kurvah usw. [das Suffix u] noch trennend dazwischenstehe, weil [dessen Substitut lopa~\ sich [nach 1,1, 57] wie ersthnin [u] verhalte, so [antworten wir]: nein ! Ein derartiges [fortdauerndes] Dazwischenstehen [eines elidierten Vokales gem 1,1, 57] wird [sonst nur] auf Grund eines [in einem Stra darauf hinweisenden] Wortes angenommen. [Nach der Interpretation srvadhtuka eva paratah" aber enthlt VI, 4,110 keinen Hinweis auf das Suffix u]. [ E i n w u r f : ] Wenn man nun aber annimmt, [da 110 infolge Ergnzung von Utah" aus 106 zu interpretieren sei:] ,,Wenn auf [das Suffix] u ein Srvadhtuka folgt," so wird ja doch sogar dann, wenn die Elision des [Suffixes] u als n/t/a-Operation [schon vo r der Substitution von Guna: VII, 3, 84] vollzogen worden ist, der Eintritt des u [fr o] dadurch erfolgen, da der /opa fr jenes [Suffix u] sich [nach 1,1, 57] wie der sthnin [u] verhlt [also:

231 Mahbhsya

[Pat.] W e n n hier, bei kurvah, kurmoh, kuryt, [gem V I , 4 , 1 0 8 und 109] die Elision des [Prsenscharakters] u vollzogen ist, so e r g i b t sich [in kor + vah usw.] nicht gem [ V I , 4 , 1 1 0 ] : , W e n n auf [den Prsenscharakter] u63 ein Srvadhtuka f o l g t ' , der E i n t r i t t v o n u [ f r das a v o n kar-]*4

ner api cet i katharn vinivrttih | ,Und wie sollte denn auch ni vor it verschwinden V [Pat.] Und auch hier, bei krisyate aus krayati, ergibt sich nicht [gem der Regel VI, 4, 51]: ,Fr [das Suffix] ni [vor einem
Suffixes u] erfolgen. Daraus ergibt sich, da [die Substitution von] Guna nicht nitya ist, und zwar gem [der Paribh. Nr. 45:] Eine Regel [.welche nitya sein sollte,] ist nicht nitya [tritt also nicht zuerst ein], wenn sie sich [nach dem Eintritt der zweiten sich gleichzeitig darbietenden Regel nur noch] mit Hilfe irgendeiner anderen Regel ergeben wrde." [So wrde sich hier Guna nach vollzogener Elision des u nur noch mit Hilfe von 1,1, 62 ergeben]. Dann aber [d. h. wenn also Guna erst nach der Elision eintritt] ist das a [in kar-vah] ein Element, welches erst dem durch das Substitut [se. opa] ersetzten Vokal vorangeht; wenn also an einem) solchen [o] die Vorschrift [sc. VI, 4, 110] ausgefhrt werden soll, so kann [das Substitut opa] nicht [nach 1,1, 57] wie der sthnin [d. i. das Suffix u] behandelt werden. [Vgl. im Anhang s. sthnivadbhva. Demnach wird VI, 4, 22 angewendet werden, indem VI, 4,108 und 109 als asiddha betrachtet werden]. [Einwurf:] Srvadhtuka, also in kurvah, aus kar[u] Aber es ist doch nicht mglich, da -\- vah, auf u-vah; vgl. im Anhang s. pratyayolaksanam]. [Erwiderung:] die Elision des [Suffixes] u asiddha sei, da sie nicht [auf einem gemeinsamen Also folgendermaen: der Verfasser Element] beruht; denn vor [den des Bhsya hat nicht angenommen, da Buchstaben] m und v [vgl. VI, 4,107] [die Elision des Suffixes u] sich wie der wird dei Elision des u vorgeschriesthnin [u] verhalte, da er ein anderes ben, vor dem [auf u] folgenden SrvaVerfahren zur Vermeidung [der dhtuka [in seiner Gnze, d. i. vor vah, Schwierigkeit, u. zw. VI, 4, 22] mah] wird der Eintritt des u [fr a] angeben wollte. Oder aber [wir vorgesch reiben? [Erwiderung:] mssen, da die Anwendung von 1,1, Dies trifft nicht zu; denn [nicht vor 57 sonst berechtigt wre, annehmen:] m und v, sondern] vor einem mit m zuerst erfolgt die Elision des [Suffixes] oder v beginnenden Suffix wird die u, weil sie nitya ist, und nachher [die Elision des [Suffixes] u vorgeschreiben. Substitution von] Guna [fr r]. Wenn Es ist also tatschlich die Abhngignmlich die Elision des u vollzogen ist, keit [der Elision des u] vor einem so mu gem [der Regel 1,1, 62], da gemeinsamen [Element] vorhanden.' das durch das [elidierte] Suffix Be[Vgl. zu Vrtt. 12.] Die Regel VI, 4, dingte eintritt, [die Substitution von] Guna [trotz der vollzogenen Elision des 22 wird hier also angewendet. u + vah; VI, 4,108: kr + vah; VII, 3, 84: kar + vah und VI, 4,110 mit Hilfe von 1,1, 57 aus kar[+u] + vah: kurvah]; geschweige denn hier, wo zuerst [nach VII, 3, 84die Substitution von] Guna [fr r] erfolgt, weil [Vit, 3, 84] para ist, und erst nachher [also unmittelbar vor der Anwendung von VI, 4,110] die Elision des [Suffixes] u [,die nicht nitya ist]; und wenn dann die Vorschrift [VI, 4, 110] fr [das Element o ausgefhrt werden soll,] welches dem Vokal [d. i. dem Suffix u] voranging, als er noch nicht durch das Substitut [se. opa'] ersetzt war [vgl. im Anhang s. sthnivadbhva], so verhlt sich [gem 1,1,5/ der opa fr u] tatschlich wie der sthnin [u]. Hingegen soll [VI, 4,110] nicht [auf Grund von 1,1, 62] als eine durch das [elidierte] Suffix [u] bedingte [Operation] eintreten; denn [diese Operation] beruht [nicht ausschlielich auf u als einem Suffix, sondern] auf dem Buchstaben [u in Verbindung mit dem folgenden

232 Bernhard Geiger

Ardhadhatuka, jedoch] ohne das A u g m e n t /" die Elision des [Kausativ-Suffixes] ni (= /). 65 Abruvatas tava yogam imam syl luk ca cino nu katham na tarasya || ,Wenn du diese Regel nicht gelten lassen w i l l s t , w i e sollte da nach ein nicht auch der Abfall von tara erfolgen V [Pat.] Und hier, bei akritarm, ahritarm, w i e sollte da nicht [gem V I , 4,104] der Abfall des auf ein [Suffix i der 3. sing, aor. pass.] folgenden tara erfolgen ?66 Erwiderung des Slokavrttikakra Cam bhagavn krtavmstu tadartham tena bhaved iti ner vinivrttih | , ,,Ca" [,und auch'] hat der Meister [Pn.] doch nur zu dem Zwecke gesetzt, da dadurch vor it das Verschwinden des ni erfolge.' [Pat.] Hier in [der Kegel VI, 4, 62]: ,Vorsya (fut. und condit.), sie [s-aor.], slyut (precat.) und tsi (periphr. fut.) im Impersonale und Passiv kann bei [Verbalstmmen], die in ihrer im grammatischen System vorliegenden [einfachsten] Form vokalisch auslauten, ferner bei han, grab und d.rs auch [so verfahren werden], als ob bei ihnen ein [die Endung / der 3. sing. aor. pass.] vorlge; [dann] t r i t t auch noch das Augment / ein.' Was denn sonst noch [ca]? U nd au ch die Elision des [Kausativ-Suffixes] ni.67 Mvor api ye ca tathpy anuvrttau ,Sowohl ,mvor' wie auch ,ye ea gelten noch fort.' [Pat.] Auch hier, bei [der Bildung von] kurvah, kurmah, kurydt [d. i. in VI, 4,110] wird ,vor m und v' [107], sowie ,und vor y' [108] noch fortgelten. 68
65 Das Fut. pass, des Kausativums von kr ist entweder 1. gleich der tmanepada-Form krayisyate, welche folgendermaen gebildet wird: kr + ni + te; nach VII, 2, 115: kri + te; Vil, 3, 84: kre-te. Dann nach III, 33, VII, 2, 70 und VIII, 3, 59: kre + i-sya + te und schlielich nach VI, 1, 78: kroyisyate. Oder aber 2. der mit dem Kausativ-Suffix ni (/') gebildete Stamm kr -{-ni- wird so behandelt, als ob er auf ein [d. i. das Suffix /' der 3. sing. aor. pass.] endigte, und in diesem Falle wird an krigem VI, 4, 62 im Futurum das Suffix sya mittels des Augmentes / angefgt. W i r erhalten kri -f- i-sya + te. Diese Vorsetzung des Augmentes / mu nun als osiddho betrachtet werden, wenn das auslautende / (ni) des Stammes kri gem VI, 4, 51 elidiert werden soll. Vgl. Mbh. vol. Ill, p. 206 das Slokavrtt.: itcsiddhas, tena me lupyate nir . . . . Kaiy. : Prakrtipratyaysrayatvd usw. ,[Die Vorsetzung des] Augmentes / (it) und die Elision des ni sind von einem gemeinsamen [Element] abhngig, da sie beide auf dem ursprnglichen [d.

h. nicht mit dem Augment / versehen] Suffix [sya] beruhen.' [Vgl. zu Vrtt. 12]. 66 Akdri st aus akarita entstanden, indem das auf ein (/) folgende Suffix ta gem VI, 4,104 abgefallen ist. Nach der Anfgung von tarm mte derselben Regel zufolge auch das Suffix tora abfallen, wenn es nicht als asiddha betrachtet wrde. Vgl. im Anhang s. visayavisayibhva. 67 Das ,ca' soll besagen, da auer der Elision des ni, die aus VI, 4, 51 fortgilt, auch noch /teintritt. Da sich also die Elision des ni schon aus VI, 4, 62ergibt, bedarf es zu ihrem Eintritt nicht erst der Anwendung von VI, 4, 22. 68 Danach wre VI, 4,110 zu interpretieren: Fr o t r i t t uein vor einem Srvadhtuka, wenn dieses mit m, v oder y beginnt. Damit ist gesagt, da diese Substitution vor den mit m, v und y beginnenden Endungen ohne Rcksicht auf die Elision des Suffixes u [vgl. Kaiy. p. 230, Anm. 64, Anfang] erfolgt. Kaiy. : Ata ut srvadhtuke ityatra usw. .Weil in [der Regel VI, 4,110]: ,,Fra t r i t t uein, wenn ein Srvadhtuka folgt" ,mvor' und ,yeca'

233 Mahbhsya

cinluki ca kita eva hi luk syat || ,Und bei dem Abfall nach ein soll nur der Abfall eines k/toder n/t-Suffixes gelten.' [Pat.] Auch bei dem Abfall [des Suffixes] nach ein [VI, 4,104] gilt das vorher erwhnte W o r t Mnit' fort. 6 9 W o ist es erwhnt w o r d e n ? In [der Regel VI, 4, 98]: [Die Paenultima] von gam, han, jan, khan und ghas w i r d elidiert v o r einem [vokalisch anlautenden] kit- oder n/t-Suffixe, ausgenommen vor an [Aoristcharakter a].' E i n w u r f Aber dieses [kit in 98] st doch durch den Lokativ [kiti] ausgedrckt, und hier [in 104] bentigt man ein durch den Genetiv ausgedrcktes [d. i. kita h].70 E r w i d e r u n g Dieser Ablativ cinah [.nach ein'] w i r d an die Stelle des Lokativs ,kiti' den Genetiv setzen, 71 indem [die Regel 1,1, 67] g i l t : ,Wenn [ i m Lehrbuch] etwas durch den Ablativ ausgedrckt ist, so [erfolgt die Operation] an dem darauffolgenden [Element]. 7 2 Selbst wenn diese Regel aufgestellt wird, 7 3 Siddham vasusamprasranam ajvidhau || Vrtt. 9. mu man [als Ausnahme] konstatieren, da [die Substitution von] Samprasrana [als] in Kraft getreten [zu betrachten ist] hinsichtlich einer Regel, welche einen Vokal betrifft. Z u welchem Zwecke denn ? Damit bei [der Bildung der] Akkusative plu ral is
fortgelten, so wird nach der vollzogenen Elision des [Suffixes] u der Eintritt von ufr a erfolgen: dies ist der Sinn. Aber auch' wendet Kaiy. ein ,mit Hilfe der [ergnzten] Buchstaben m usw. [d. i. v und y] erfolgt bei kurmah usw. nicht der Eintritt des u, weil durch sie das [ W o r t ] srvadhtuka [in 110] in einer Weise nher bestimmt wird, da sich ein Widerspruch in der Aussage [vkyabhedo, der Regel 110] ergibt.' D. h. durch die Ergnzung von, ,mvor' und ,ye ca' hat die Regel eine ihrem Zweck widersprechende Einschrnkung erfahren; denn die Bildung von kurutah, kurvanti, deren Endungen nicht mit m, v oder y beginnen, wre nun nicht mehr mglich. 69 Bei dieser Annahme kann nur die Endung ta abfallen, da sie ein ri/'tSuffix st, whrend tara(p) und tama(p) als p/t-Suffixe nicht abfallen. 70 Luk solI ja a n d i e S t e 11 e des k/toder ri/'t-Suffixes treten. Dasjenige aber, an dessen Stelle ein anderes Element tritt, wird in der Grammatik nach 1,1, 49 durch den Genetiv ausgedrckt; wenn etwas hingegen im Lokativ angegeben st, so wird nach 1,1, 66 die Operation an dem vorangehenden Element vorgenommen. Bei der Ergnzung des Lokativs kiti knnte also ein kit- oder nitSuflfix nicht Gegenstand der in VI, 4, 104 vorgeschriebenen Operation sein. 71 D. h. er wird bewirken, da der Lokativ im Sinne eines Genetivs aufzufassen ist. 72 Vgl. die Paribh. Nr. 70: Ubhayanirdese pacamlnirdeso bally an ,Wenn [in einer Regel zwei Elemente] durch beide [Kasus, das eine durch den Ablativ, das andere durch den Lokativ] ausgedrckt ist, so besitzt das durch den Ablativ ausgedrckte grere Kraft.' D. h. in einem solchen Falle gilt nicht die Regel 1,1,66, wonach der Ablativ als Genetiv aufzufassen wre, sondern 1,1, 67. Demnach besagt VI, 4,104: Nach ein t r i t t , wenn unmittelbar darauf ein kit- oder ri/t-Suffix folgt, fr d eses Suffix/ukein.[Vgl.Einleit. p. 213 f.] Die Regel VI, 4, 22 wre also auch in diesem Falle unntig. Im Text werden hier die Slokavrttikas im Zusammenhang noch einmal vorgefhrt. 73 Kaiy.: Anekaparihrsrayena usw. ,wenn sie zu dem Zwecke angewendet wird, damit sich ein Verstndnis nicht erst auf umstndliche Weise dadurch ergebe, da man mehr als eine Annahme zur Vermeidung [von Schwierigkeiten] macht.'

234

papusah, tasthusah; ninyusah, cicyusah', luluvusoh, pupuvusah, wenn

Bernhard Geiger

[nach VI, 4,131 die Substitution des] Sa mp rasa rana [u fr das v] von vas vollzogen worden ist, gem [VI, 4, 64] : ,Vor einem Vokal' [hier vor us] die Elision des [von p und sth] und die brigen [Operationen, d. i. 82 und 77] stattfinden mgen.74 Aber aus welchem Grunde sollten sie denn nicht zustande kommen? Bahirangalaksanatvd asiddhatvc ca || Vrtt. 10. [Die Substitution des] Samprasra na [VI, 4.131] ist ja doch eine Bah i ranga-Regel,75 und sie ist auch [gem VI, 4, 22 in bezug auf VI, 4,_64, 82 und 77] als asiddha zu betrachten. Attvam yaiopllopayoh pasuso na vjn ckhyit ckhyitum || Vrtt. 11. Man mu [ferner als Ausnahme] konstatieren, da der Eintritt von [als] siddha [zu betrachten] ist in bezug auf die Elision eines y und die eines . Weshalb denn? [Wegen],pasuso na vjn1 [RV.V, 41,1]. Wenn [bei der Bildung von] pasusah der Eintritt
74 Deracc. pi. part. perf. act. von p wird gebi Idet: popo -f vas + os. Da os eine vokalisch anlautende Endung eines schwachen Kasus ist, ist papvosein /ia-Stamm, es wird also nach VI, 4,131 Samprasrona substituiert: papo -\- uos -f- as; daraus wird nach VI, 1, 108 und VIII, 3, 59 [wo Bhtlingk unrichtig bersetzt ,am A n f a n g eines Suffixes', vgl. die Ks.]: papa + us -\-as. Die Substitution des Samprasrana darf nun nicht als asiddha betrachtet werden, weil das nach VI, 4, 64 nur vor einem v o k a l seh anlautenden rdhadhtuko elidiert wird. hnlich verhlt es sich mit den zwei anderen Gruppen von Beispielen, bei denen die Regeln 82 und 77 in Betracht kommen. Man kann nun einwenden, da dem rdhadhtuka vas nach VII, 2, 35 das Augment /' vorgesetzt werden sollte. Die Ausfhrungen Kaiy.'s hierber: ,nanu cntorangatvdd it bhvyam' usw. bis ,prvam in na provartate' gebe ich im Anhang s. ,nimittpye noimittikasypy apyah1 wieder. 75 Denn ihre Ursache liegt auerhalb der Ursache der Antarango-Regel VI, 4, 64, wie folgende Darstellung zeigt: papa -f vas -\- os papa + us -f as
y

I VI, 4, 64: Elision des

76 Kaiy. erklrt zunchst die Bildung von pasusah [ = pasum sanoti,Vieh verschaffen']: Nach III, 2, 67 t r i t t an die Wurzel san das krt-Suffix vit (v) [welches nach I, 2, 46 die Wurzel zum Nominalstamm macht. Nach VI, 1, 67 verschwindet es zwar wieder, doch treten auch nachher gem 1,1, 62 die durch dieses Suffix bedingten Operationen ein]: pasu-san + (v) + S ; Substitution von fr n vor vit nach VI, 4, 41 (und von s fr s nach VIII, 3, 108): pasu-s-\-as. Die Substitution von o darf nun nicht als asiddha betrachtet werden, wenn gem VI, 4,140 die Elision des erfolgen soll. Kaiy. bemerkt ferner: Loksanapratipadoktaparibhs tv llope nsrlyate, ovyptiprosongt ,Bei der Elision des [VI, 4,140] wird jedoch die Paribhs[Nr. 105] ber Ausdrcke, welche sowohl etwas erst aus einer Regel sich Ergebendes bezeichenn, als auch das durch ihren Wortlaut Gegebene, nicht angenommen. Denn sonst wrde sich ergeben, da [die Regel 140] nicht alle Flle umfat [ovypti, vgl. p. 219, Anm. 19].' D. h. wenn die Paribhs hier glte, wrde t in VI, 4,140 nur das der auf auslautenden Wurzeln bezeichnen knnen, nicht aber ein sekundres , welches nach 41 fr n substituiert worden st. Auf ein soches wrde VI, 4,131 .Sam- sich 140 dann nicht erstrecken. Aber

a bedingt durch
u: antaroga

prasrana ufr
v bedingt durch as: bohl ranga

die Bildung von pasusah lehrt, da 140

auch auf ein sekundres o angewendet wird, da somit in diesem Falle die Paribh. 105 nicht gilt. Vgl. Paribh. Transi., p. 486, n. 2.

Vgl. Paribh. Nr.50.

235 Mahbhsya

von a [fr das n von san: V I , 4, 41] als asiddha betrachtet w i r d , so ergibt sich infolgedessen nicht gem [der Regel 140] : ,Fr des [auslautende] einer W u r z e l ' die Elision des .76 W e n n bei [der Bildung von] ckhyit, ckhyitum der Eint r i t t von d [ f r n: V I , 4, 43] als asiddha betrachtet w i r d , so ergibt sich gem [der Regel 49] : ,Fr ein auf einen Konsonanten f o l gendes y [ v o r einem rdhadhtuka]1 die Elision des y.77 Samnsrayavacant siddham || V r t t . 12. .Dadurch, da man [in der Regel] hinzufgt,, welche von dem gleichen [Elemente] abhngig st", ist [das Stra hinsichtlich der erwhnten Ausnahmen] vollstndig k o r r e k t ' . [ N u r eine O p e r a t i o n , ] welche auf dem gleichen [Elemente] beruht [ w i e die zweite O p e r a t i o n ] , ist asiddha. Diese [in den V r t t . 9 und 11 genannten Regeln] aber beruhen nicht auf dem gleichen [Elemente]. Zunchst hier, in den Akkusativen pl. papusah, tasthusah; ninyusah, cicyusah; uluvusah, pupuvusah, beruhen die Elision des a und die brigen [Operationen] auf -vas [us], das Samprasra na [dagegen] auf der Endung [as] des auf -vas ausgehenden [Stammes]. Bei pasusah [beruht] der Eintritt des [fr das n von san] auf dem [Suffix] vit, die Elision des dagegen auf der Endung des auf vit ausgehenden [Stammes]. Bei ckhyit, ckhyitum [beruht] der Eintritt des auf dem [Intensivcharakter] ya [lies : yanyttvam], die Elision [des y hingegen] auf dem rdhadhtuka [i-tr] des auf ya ausgehenden [Stammes]. Sollte dieses [ W o r t samnsraya in dem Stra] ausdrcklich hinzugefgt werden ? Durchaus nicht! W i e soll man es aber [aus der Regel] verstehen, wenn es nicht ausdrcklich erwhnt wird ? Vermge des Wortes atra. Einwurf Aber es ist doch etwas anderes als Zweck des Wortes atra behauptet worden? Was ist behauptet worden ? [VI, 4, 22 Vrtt. 2:] , Das W o r t atra hat den Zweck, den Geltungsbereich anzugeben. Erwiderung Dies [sc. der Geltungsbereich] ergibt sich schon vollstndig aus dem Adhikra [,asiddhavat'].78

78 Kaiy. : Khanater ya\ dvirvacant Nach Katy.'s Ansicht [Vartt. 2] gibt paratvdye vibhsenyttvedvirvacane ca , bht' den Bereich derjenigen trj irpam .Wenn an die Wurzel khan Regeln an, welche asiddha sind, und [nach III, 1,22] der Intensivcharakter ,atra' den Bereich derjenigen Regeln, ya angefgt worden ist [: khan-ya-], in bezug auf welche jene asiddha [die Wurzel] auf die Reduplikation sein sollen. Nach Pat. dagegen ergibt [VI, 1. 9] folgt usw. [d. h. nach VII, 4, sich schon aus dem Adhikra [vermi85 das Augment nuk (n) an das a der ttelst der Erklrung] der Abschnitt Reduplikationssilbe getreten ist: VI, 4, 22 bis bha als Bereich der als ca-khan-ya-], die optioneile Substitu- asiddha zu betrachtenden Regeln, tion von [fr n zunchst an der whrend , bht' den Bereich der Wurzel] und [dann] auch an der Regeln angibt, in bezug aufweiche Reduplikation [VII, 4, 83] erfolgt ist jene asiddha sind. ,Atra' wre also [: c-kh-ya-], so tritt das mit [dem berflssig, wenn es nicht wie Pat. Augment] / versehene [rdhadhtuka] annimmt den Zweck htte, austrc (tr) an.' Wir erhalten [VI, 4, 48] zudrcken, da beide Regeln von dem ckhyit. Wenn aber die Substitution gleichen Element abhngig sein des als asiddha betrachtet wrde, mssen. Vgl. die Ks. zu VI, 4, 22: Asialso noch ca-khan-y- als vorhanden ddhavad ity ayam adhikro yad ta zu denken wre, mte nach VI, 4, 49 rdhvam anukramisyma adhyyapadas auf n folgende y abfallen, und wir risamptes tad asiddhavad veditavyam; wrden flschlich ckhit erhalten. bhd iti visayanirdesah: bhasam-

77

236 Bernhard Geiger

E i n w u r f Hier bei papusah, cicyusah, luluvusah w u r d e n [ V a r t t . 10] zwei Grnde [dafr] namhaft gemacht [,da die Elision des usw. nicht zustande kommen kann, nmlich] : der Umstand, da [VI, 4,131] eine Bahiranga-Regel und [auerdem gem VI, 4, 22] asiddha sei. [Die Annahme,] da sie [gem VI, 4, 22] asiddha sei, mag als an jener Stelle [Vrtt. 12] beantwortet gelten. 79 A u f [den G r u n d , da VI, 4,131] eine Bahiranga-Rege\ sei, st aber noch nicht e r w i d e r t worden. 8 0 Erwiderung 8 1 Dies bedeutet keinen Fehler. [Denn] eine Bahiraga- und eine Antoronga-Operation sind zwei Begriffe, die
sabdanad yod ucyate totra kartavye [.wenn eine (zweite Operation) dort (d. h. nach einer Regel) vollzogen werden soll, welche biszur Erwhnung von bha (d. i. bis VI, 4,129 oder 175) gelehrt wird']. Atreti samnsrayatvoprotipattyartham : tac ced atra yotro [bhavoti st zu streichen !] tod bhc chstnyam vidhyate tadsrayam eva bhavati, vysrayam tu nsiddhavad bhavatity arthah ,. . . [Eine in dem Abschnitt bis bha gelehrte Operation st osiddha,] wenn diese dort, wobei eine andere in dem Abschnitt bis bha gelehrte Operation vorgeschreiben wird, und nur eben davon abhngig ist . . .' In dem in diesem Kapitel angefhrten Beispiele handelt es sich um ein gemeinsames Element, an w e l c h e m die Operation der zweiten Regel vollzogen wird; es ist zwar nicht, wie bei der ersten Operation, Ursache, aber doch ein Element, von dem die zweite Operation abhngig ist. Es kann aber auch Ursache der zweiten Operation sein, und dann haben beide Operationen die gleiche Ursache. Vgl. im Anhang s. samnsraya. Vgl. noch die Ks. zu unserem Stra: Atragrahanam kiml Papusah pasya usw.
79 Kaiy.: ,Durch die Annahme, da es auf die Abhngigkeit von dem gleichen [Element] ankomme. 80 Kaiy.: Yadyapi njnantaryo ity ayam usw. .Obwohl hier das [in der Paribh. Nr. 51 enthaltene] Verbot zur Anwehdung kommen knnte: ,,[Eine bereits ausgefhrte ahirangaOperation ist] nicht [als asiddha zu betrachten, wenn die AntaragaOperation] auf dem unmittelbaren Folgen eines Vokales [beruht;" hier die Elision des auf dem folgenden u von us], so will dieser Zusatz [des Bhsyakra] doch zu verstehen geben,

da [hier] auch die Mglichkeit der Anwendung der Paribhs [Nr. 50] nicht besteht, wei I [wie im folgenden gezeigt wird] eine Ursache [fr ihre Anwendung] gar nicht vorhanden ist.' 81 Das Folgende findet sich in hnlichem Zusammenhang bei Pat. zu VIII, 3 , 1 5 V r t t . 2. 82 Kaiy. zu VIII, 3,15 Vrtt. 2 erklrt pratidvandvibhvinou : dvandvam yugmam ucyate; pratiklam parasparaviruddham dvandvam pratidvandvam, tatro bhovatah. Z u unserer Stelle bemerkt Kaiy.: Na hi vastvantarnapeksam antaragam usw. . . . bis padny osminn arthe yojyante ,Es gibt nicht irgendeine Antaraga- oder BahiragaOperation, die, ohne sich auf den anderen Gegenstand [d. i. auf eine zu ihr gehrige bahiraga-, bezw. AntaragaOperation] zu beziehen, nur fr sich selbst zu Recht bestnde; denn sie liegen nur dann vor, wenn sie beide sich gleichzeitig in der Weise darbieten, da eine von der andern abhngt, so wie bei [der Bildung von] syona [aus si -f- + na (vgl. Paribh. Transi, p. 222, n. 2) die Substitution des] Halbvokales [fr /] und die von Guna fr / sich gleichzeitig darbieten. Bei papusah dagegen bieten sich nicht [beide Operationen] gleichzeitig dar: in dem Zeitpunkt, in welchem das Samprasrana sich darbietet [d. i. so lange -vas noch vorliegt], besteht keine Mglichkeit fr die Elision des und die brigen [Operationen, vgl. p. 234], weil eine Ursache [fr ihren Eintritt, d. i. ein Vokal] nicht vorhanden ist; und in dem Zeitpunkt, in welchem die Elision des [vor einem Vokal] sich darbietet, st [die Substitution des] Samprasrana schon ausgefhrt. Und deshalb ist das Samprasrana nur erst die Ursache der Elision des usw.; folglich gelangt

237 Mahbhsya

(einander widerstreitend) zu zweien auftreten. 8 2 In welcher Weise ? W e n n eine Antaranga-Operation vorhanden ist, so ist [gleichzeitig auch] die oh/rongo-Operation da; und wenn eine Bahiroga-Operdt\on vorhanden ist, so ist [gleichzeitig auch] die Antaranga-Operation da.83 Hier aber [bei papusah usw.] sind die Antaraga- und die oh/rongo-Operationen nicht gleichzeitig zur Stelle. 84 W e n n [nmlich] eine o/i/rongo-Operation nicht in die Erscheinung getreten ist,85 so bietet sich [auch] eine AntaragaOperation nicht dar. Unter diesen Umstnden (tatro) istdiefangeb-

hier die Paribhas [ N r . 50] nicht zur Anwendung. Dies st der Sinn. [Ei n w u r f : ] W e n n es sich so verhlt, so [kann man dagegen einwenden:] auch bei pacvedam [aus pacva i dorn] usw. bietet sich, so lange [die ahirango-Operation, d. ,] die Substitution von Guna nach a [ f r a - f /: VI, 1, 87] nicht vollzogen st, [die AntoragaOperation, d. i.] die Substitution von ai gem . o / f r e 1 [III, 4, 93] nicht dar; weil somit [die o/i/ro/igo-Operation] die bedingende Ursache und [die Antoroga-Operation] das hierdurch Bed ingte ist, wre [auch hier] die Mglichkeit der Anwendung der Paribhas nicht vorhanden. [In W i r k lichkeit gilt sie jedoch bei pacvedom. Vgl. Paribh. Transi, p. 271, n. 4.] [ E r w i d e r u n g : ] N u n , dann w i r d [pratidvandvibhvinau] folgendermaen anders e r k l r t : die Paribhs [ N r . 50] .Eine o/i/rongo-Operation ist asiddho in bezug auf eine Antaraga-Opera.tion 1 st in [der Regel V I , 4,132]: ,ln vh t r i t t [als Samprasrana fr v] ein' angedeutet [vgl. Paribh. Text p. 44, I. 3 ff.; Transi, p. 23537 samt Noten], sie ist also eine in dem Abschnitt bis bha gelehrte [Regel]; oder aber: obwohl [diese Paribhs im Vrtt. 8] zu [der Regel I, 4, 2:] .Bei einem Konflikt [zweier Regeln] ist die spter erwhnte zu vollziehen' ergnzend hinzugefgt worden st, so w i r d sie [doch auch] zu einer im Abschnitt bis bha gelehrten [Regel], weil [nach Paribh. N r . 3] Termini technici und Paribhss sich mit jeder Operation [bei der diese Termini vorkommen oder diese Paribhss eine Rolle spielen] zu einem Satze vereinigen, somit auch die Substitution des Samprasrana [VI, 4,131] in vas [der Paribh. N r . 50] eine Sttte gewhrt. W e n n dann die im Abschnitt bis bha

gelehrten Operationen, nmlich die Elision des und die brigen, vollzogen werden sollen, so ist in bezug auf diese die [in demselben Abschnitt gelehrte] Paribhs [gem VI, 4, 22] asiddha, t r i t t also nicht ein. Und wenn sie nicht e i n t r i t t [d. h. das Samprasrana nicht als bahiraga betrachtet w i r d , also auch nicht asiddha ist], so finden [verursacht durch das Samprasrana] die Elision des und die brigen Operationen statt. Dies st der Sinn der Auseinandersetzung [des Bhsya]. Die einzelnen W o r t e [derselben] werden in diesem Sinne [d. i. in dem der z w e i t e n Interpretation] gebraucht.' Vgl. im Anhang s. antaraga und p. 229, Anm.56. 83 Kaiy. erklrt ,sati' durch buddhypeksite .sobald man mittels der W a h r nehmung auf sie Bezug genommen hat', d. h. sobald man sie im Bildungsproze der Form zum ersten Male wahrgenommen hat. Vgl. Kaiy. zu VIII, 3,15 Vrtt. 2: saty antaraga ti: antarage buddhypeksite tadapeksay bahiragam bhavati, bahirage cpeksite tadapeksayntarangam bhavaty antaragavyapadesam labhata ity arthah. 84 Nmlich im Augenblick der A n wendung des Paribhs. Kaiy. zu VIII, 3,15 V r t t . 2 erklrt yugapatsamavasthnam durch: ekasym buddhv apeksitapa raspa ram avasthnam ,ein Eintreten in der Weise, da man schon bei einer einmaligen Wahrnehmung jedes auf das andere bezieht'. 85 Nmlich deshalb, weil sie im Augenblick der Anwendung der ParibhsgemVI, 4,22in bezug auf diese als asiddha betrachtet werden mu. Vgl. im Anhang s. antaraga.

238 Bernhard Geiger

liehe] Bahiraga-Operatlon nichts w e i t e r als die Ursache der [angeblichen] Antaranga-Operation. 8 6 Hrasvayalopllops cydese lyapi || V r t t . 13. Man mu [als Ausnahmen] konstatieren, da 1 . die [Subs t i t u t i o n d e r ] Krze, 2. die Elision des y und 3. die Elision des a hinsichtlich der Substitution von ay [fr das Kausativ- Suffix /] v o r [dem Suffix des Gerundiums] yap (ya) als siddha zu betrachten sind, [in den Beispielen :] 1. prasamayya gatah, pratamayya gatah ; 2. prabebhidayya gatah, pracecchidayya gatah ; 3. prastanayya gatah,

pragadayya gatah. Wenn [nmlich] 1. die [Substitution] der Krze [fr die Lnge vor ni: VI, 4, 92], 2. die Elision des y [VI, 4, 49] und 3. die Elision des [Stammauslautes] a [VI, 4, 48] als asiddha betrachtet werden, so ergibt sich nicht gem [der Regel VI, 4, 56]: ,Vor lyap (ya) nach einem auf eine Krze folgenden [Wurzel konsonanten] ' die Substitution von ay [fr /]. 87 Erwiderung Auch hier vermeidet man [Schwierigkeiten] in der Weise [,da man erklrt]: ,Es ist dadurch vollstndig korrekt,
Kaiy.: Paribhsy asiddhatvd apravrttv ti bhvah usw. .Damit ist gemeint: unter der Voraussetzung, da die Paribhs nicht eintritt da sie asiddha ist. Und zwar ist die Paribhs asiddha [gem VI, 4, 22] infolge der Abhngigkeit von dem gleichen [Element], indem das Samprasrana allein es ist, wovon sowohl die [Anwendung der] Paribhs, als auch die Elision des und die brigen [Operationen] abhngen.' Vgl. Paribh. Transi, p. 265, n.2. In allen drei Fllen handelt es sich um die Bildung des kausativen Gerundiums. Und zwar werden mit Prpositionen zusammengesetzte Verba gewhlt, weil nach VII, 1, 37 in der Komposition lyap (ya) fr ktv (tv) substituiert wird. 1. Der Kausativstamm st prasam-fn/-, daraus nach VII, 2,116: pra-sm + /-. Fr die lange Pnultima wird nun nach VI, 4,92 die Krze substituiert [Bhtlingk ungenau: .bewahrt die Krze']: pra-sam + /-.Wenn nun [nach VII, 1, 37 statt tv] das Suffix ya angefgt worden ist, so wird ay fr das / substituiert, und wir erhalten prasamayya. ,Wenn jedoch die [Substitution der] Krze [o] als asiddha betrachtet wird, so geht dem [auslautenden Wurzel konsonanten] m nicht eine Krze voran, und infolgedessen bietet sich die Substitution von ay [fr /] nicht dar' (Kaiy.). Im Dhtup. 19, 70 erscheint sam unter den Wurzeln mit stummem m (mtah) als ,mit', sofern es nicht ,darsana' bedeutet.
87 86

2. .Hinter dem auf das [Intensivsuffix] ya (ya) ausgehenden [Stamm] von bhid t r i t t das Kausativ-Suffix nie (i) ein' (Kaiy.). W i r erhalten prabebhid -f ya -f /-; nach VI, 4, 48: pra-bebhid + y-\- /-; nach VI, 4, 49 [Elision des / v o r
d e m rdhadhtuka / ] : pra-bebhid -\~ /-;

nach VI, 4, 56: prabebhidayya. ,Wenn hier die Elision des y als asiddha betrachtet wird, so folgt [in prabebhidy + '-] Q\c (') nicht auf einen [auslautenden Wurzelkonsonanten] mit vorangehender Krze, da ja noch das y dazwischen steht, und die Substitution von ay [fr /] bietet sich infolgedessen nicht dar' (Kaiy.) 3. .Unter der berschrift,Die auf a endigenden Wurzeln" wird [im Dhtup. 35, 78] erwhnt: ,,stana und gada, zur Bezeichnung himmlischer Gerusche" (Kaiy.). Nach der Anfgung des Kausativ-Suffixes /', erhlt man gem VI, 4, 48: prastan-iund nach VI, 4, 56: prastanayya. ,Wenn hier die Elision des a [VI, 4, 48] als asiddha betrachtet wird, ergibt sich nicht die Substitution von ay [fr /: 56], da [in prastana-i-] ein o dazwischensteht [, also das / nicht auf einen Wurzelkonsonanten mit vorangehender Krze folgt].' So auch die Ks. zu VI, 4, 56. Kaiy. : abhuvatur ti : bh atus vuk usw. .[Zunchst] bh -f- atus, [dann wird] vuk (v) [angefgt]; wenn dieses als asiddha betrachtet wird, bietet sich [die Substitution von] uva (uv) dar. [E i n wu rf:] Aber [die Anfgung von] vuk [VI, 4, 88], die [in bezug auf einen speziellen Fall]
89 88

239 Mahbhsya

da man sagt, [eine Operation sei asiddha,] wenn sie von dem gleichen [Element] abhngt. [Vrtt. 1 2 . ] Wieso? A u f dem [Suffix] ni beruhen diese Regeln [VI, 4, 92; 49; 48], auf dem [Suffix] yap (ya) hinter ni [dagegen] die Substitution von o/. 88 Vugyutv uvayanoh || V r t t . 14. Man mu [als Ausnahmen] konstatieren, da die [Anfgung der] Augmente vuk (v) und yut (y) hinsichtlich [der Substitution von] uva (uv) und yon (des Halbvokales) als siddho zu betrachten sind. 1. Babhvatuh, babhvuh : wenn [hier die Anfgung von] vuk (v) [in bobh + v + atuh: VI, 4, 88] als asiddho betrachtet w i r d , so bietet sich infolgedessen [nach 77 die Substitution von] uva (uv) [fr das ] dar; 89 2. upadidiye, upad idly ate \ wenn [hier die A n fgung won] yut (y) [in upadidl + y + e: VI, 4, 63] als asiddha

betrachtet wird, so bietet sich infolgedessen [nach 82] die Substitution des Halbvokales [y fr 1] dar.90
vorgenommen wird, whrend [fr diesen Spezialfall nach der allgemeinen Regel VI, 4, 77 auch die Substitution von] uva sich notwendigerweise [gleichzeitig]ergeben wrde[nprpte], stellt doch eine Ausnahme [opavda] von dieser [Substitution des uva] dar [und hebt diesse auf]; es gilt ja der Grundsatz [Pat. zu 1,1, 47, Vrtt. 1]: ,Wenn auch [eine allgemeine und eine Spezialregel] gleichzeitig eintreten knnten, so wird doch [jene durch diese] aufgehoben.' [Nach dem Eintritt der Spezialregel VI, 4, 88 kann also die allgemeine Regel 77, d. i. die Substitution von uva, nicht mehr eintreten.] [Erwiderung:] Dies trifft [hier] nichtzu. [Denn nur] diejenige [allgemeine Regel], welche in dem Augenblick vorhanden ist, in welchem die sie verdrngende [Spezialregel] zur Anwendung gelangt, wird von der verdrngenden [Spezialregel] vllig aufgehoben und gelangt nicht mehr zur Anwendung; uva [nach der allgemeinen Regel] jedoch tritt nicht [gleichzeitig] in die Erscheinung, da es [gem VI, 4, 22] in dem Augenblick als asiddha [ = nicht vorhanden] zu betrachten ist, in welchem vuk [nach der Spezialregel] eintritt, [und zwar nur deshalb eintritt,] weil es [sonst, d. i. nach dem Eintritt der allgemeinen Regel] keine Gelegenheit [zum Eintritt] htte; [uva] kann also [durch die Spezialregel] nicht [ganz] verdrngt werden ; und so tritt es denn, wenn [die Anfgung von] vuk vollzogen st, aus dem Grunde ein, weil diese [gem VI, 4, 22 in bezug auf die Substitution von uva] als asiddha [ = nicht in Kraft getreten] betrachtet wird.' Es werden nmlich zwei Arten von Ausnahmsregeln (apavda) unterschieden: 1. Spezialregeln, die gleichzeitig mit oder nach dem Eintritt der allgemeinen Regeln zur Anwendung kommen knnten und diese gem der Maxime ,yena nprpte yo vidhir rabhyate, sa tasya bdhako

bhavati' [Paribh. Text p. 65, I. 8f.; Transi, p. 321 f.] vol I stand ig auf heben. Von solchen Spezialregeln sagt man, da sie die allgemeinen Regeln apavdatvt aufheben; 2. Spezialregeln, die nicht gleichzeitig mit oder nach dem Eintritt der allgemeinen Regeln zur Anwendung kommen knnten, und die nur deshalb, weil sie sonst berhaupt nicht Gelegenheit zur Anwendung htten [anavaksatvt], die allgemeinen Regeln aufheben. Wenn sich nach der Ausfhrung einer solchen Spezialregel die allgemeine Regel noch darbietet, so gelangt sie nachher noch zur Anwendung. Vgl. Paribh. Text p. 67, I. 5: kvacittusarvathnavaksatvd eva bdhakatvam; I. 8: tatra bdhake pravrtte yady utsargaprptir bhavati tad bhavaty eva, und

speziell Paribh. Nr. 58. Unsere Regel VI, 4, 88 ist aus den von Kaiy. genannten Grnden eine Spezialregel der zweiten Art. Nach ihrer Ausfhrung bietet VI, 4, 22 der allgemeinen Regel VI, 4, 77 die Mglichkeit der Anwendung. Vgl. Paribh. Transi, p. 329, n.4. 90 So auch Ks. und Siddh. K. zu VI, 4, 22 und 63. Vgl. Candra V, 3, 21, wo yukfr unser yut steht.

240 Bernhard Geiger

E r w i d e r u n g Was zunchst vuk betrifft, so st fr dieses [eine Ausnahme] nicht zu konstatieren. Ich werde [nmlich in VI, 4, 88] vuk berhaupt nicht erwhnen. Ich werde folgendermaen sagen : Bei bh w i r d vor [den vokalisch anlautenden Personalendungen im] A o r i s t u n d Perfektum fr die Paenultima substituiert." 9 1 Hiebei w i r d der E i n t r i t t von an Stelle derjenigen Paenultima erfolgen, welche vorliegt, wenn die Substitution von uva (uv) [fr das von bh gem VI, 4, 77] bereits vollzogen ist.92 E i n w u r f A b e r wenn [man] auch in dieser Weise [ m i t V), 4, 88 verfhrt], woraus geht denn dies hervor, da nmlich der Eint r i t t von fr diejenige Paenultima erfolgen w i r d , welche vorliegt, wenn die Substitution von uva (uv) bereits vollzogen ist, da es hingegen nicht fr diejenige Paenultima eintreten mge, welche gegenwrtig vorliegt, [d. i.] fr das bh [von babh-atuh]l E r w i d e r u n g Dies bedeutet keinen Fehler. [Denn] ,fr u ()' [aus VI, 4, 83] 93 gilt [hier] noch f o r t . A u f diese Weise w i r d [der E i n t r i t t des ] an Stelle des u [in babhuv-atuh] erfolgen. 94 E i n w u r f Zugegeben, da babhvatuh, babhvuh [auf diese Weise] zustande gekommen sind, 95 so k o m m t doch das Folgende nicht zustande, [nmlich] babhva, babhvitha. Aus welchem G r u n d e ? Denn, wenn [in babh + tha nach VII, 3, 84] Guna und [in babh + a nach VII, 2,115] Vrddhi [fr das ] s u b s t i t u i e r t ' w o r d e n sind, ist der Laut u [fr den eintreten knnte] nicht vorhanden. 9 6 E r w i d e r u n g Hier bieten sich [die Substitutionen von] Guna und Vrddhi gar nicht dar. A u s welchem Grunde? W e i l [die Regel 1,1,5:] ,Auch vor einem [Suffix] m i t stummen k oder n es verbietet. W i e s o sind [a und tha] kit-[Suffixe]? Gem [der Regel I, 2, 6 ] : ,Auch nach indh und bh'. E i n w u r f Dann weisen w i r eben diese [Vorschrift], da [bei bh die Personalendungen des Perfektums als] kit [zu betrachten] sind, durch [die Anfgung von] vuk (v) zurck. 9 7

D. h. auer der Streichung von vuk in 88 wird auch noch eineTeilung der Regel (yogovibhgo) 89 vorgenommen. 92 D. i. fr die Paenultima u in babhuvatuh, so da wir babhvotuh erhalten. 93 Oh st gen. sing, von u, welches nach 1,1, 69 zur Bezeichnung von u und dient. 94 Kaiy.-.Uvarnasyopadhy d bhavatity evam srayand usw. .Weil ja angenommen wird, fr den Laut u, welcher Paenultima st, tretet/ein; und nur wenn [die Substitution von] uva (uv) vollzogen st, wird der Laut u Paenultima. [Der Gegner aber], welcher der Ansicht st, die Geltung [der Personalendungen des Perf. von bh] als kit gem [I, 2, 6]: ,nach indh und bh' werde [durch die Vorschrift ber vuk] zurckgewiesen [vgl. im folgenden], sagt ,,bhavet siddham".' 95 Denn die Personalendungen des Dualsund Plurals sind nicht pit, also nach I, 2, 5 k/t-Suffixe, die gem I,

91

1, 5 weder Guna noch Vrddhi bewirken. 96 Hier wird die im Folgenden erwhnte Regel I, 2, 6 ignoriert, nach welcher alle Personalendungen des Perfektums von bh k/t-Suffixe sind, also nach 1,1,5 Guna und Vrddhi verbieten. Die Endung der2. Sing., thai (tha), wird also, da sie nach III, 4, 82 fr s/p substituiert worden ist, als pit betrachtet, es t r i t t demnach gem VII, 3, 84 Guna ein. Die Endung der 3. (und 1.) Sing, ist nach III, 4, 82 nal; nach VII, 2,115 wird also fr das auslautende des Stammes Vrddhi substituiert. Wenn wir so babhav-itha und babhv-a erhalten haben, ist ein u, fr welches eintreten knnte, nicht vorhanden. 97 D. h. die Regel, I, 2, 6 ist berflssig, denn die Substitutionen von Guna und Vrddhi werden verhindert, wenn wir vuk in der unvernderten Regel VI, 4, 88 belassen. Wenn wir nmlich vuk

241 Mahbhsya

Erwiderung Aber hier [in I, 2, 6] wird doch durch [die Vorschrift], da [die Personalendungen des Perfektums von bh als] kit [zu betrachten] seien, [die Vorschrift ber] vuk zurckgewiesen. 98 Was ist nun aber hier magebend ?Das W o r t vuk ist maegbend. Denn selbst wenn [die Personalendungen nach I, 2, 6] kit sind, mten hier gleichwohl Guna und Vrddhi eintreten. Warum? [Denn] dies [sc. I, 1, 5] ist ein Verbot [der Substitution] von Guna und Vrddhi, sofern sie [den pratyhra] ik betreffen. Und diese [in VII, 2, 115 gelehrte Vrddhi] ist nicht eine Vrddhi, welche [den pratyhra] ik betrifft."
angefgt und bobhuvitho, babhv-a erhalten haben, sind die Regeln VII, 3, 84 und VII, 2,115 nicht anwendbar, weil der Stamm nicht mehr vokalisch auslautet [vgl. Ks.: ig-, bezw. ajantasyo]. Und zwar erhlt das Augment vuk den Vorzug vor den sich gleichzeitig darbietenden Substitutionen von Guna und Vrddhi:bhuvo vuko nityatvd iti nyyt ,nach dem Grundsatz: weil das an bh [angefgte] vuk (v) nitya ist' [Kaiy.]. Vgl. Mbh. zu I, 2, 6 Vrtt. 1, wo erklrt wird, da die Regel I, 2, 6 berflssig sei, bhuvo vuko nityatvt: bhavater api nityo vuk; krte 'pi [sc. gune] prpnoty akrte 'pi. Dazu Kaiy. : Oh supity ata or iti nnuvartata iti bhvah. Ekadesavikrtasynanyatvc ca sabdntaraprptyabhvah ; satym api v sabdntaraprptau krtkrtaprasangitvd vuko nityatvam, vuki krte gunasya prptir eva nstiti naitayos tulyabalatvam .Dies besagt, da nicht aus [der Regel]: ,oh supi' [VI, 4, 83] ,oh' [,fr ul\ vgl. p. 240] fortgilt. Und da ja eine Sache, welche an einer Stelle gendert worden st, dadurch nicht zu einer andern Sache wird, so wrde sich [auch infolge der Substitution von Guna] nicht eine andere Wortform ergeben. [Wenn die Wortform mit Guna als von der ursprnglichen Wortform vershieden zu betrachten wre, wrde vuk nach Paribh. Nr. 44 n i c h t nitya sein]. Oder aber, auch wenn sich [infolge der Substitution von Guna] eine andere Wortform ergbe, wre vuk [nach Paribh. Nr. 46] auch schon deshalb nitya, weil es sich sowohl nach als auch vor dem Eintritt [von Guna] darbieten wrde. Wenn [also] vuk [zuerst] eingetreten st, ergibt sich [die Substitution von] Guna berhaupt nicht mehr. Diese beiden [Regeln VI, 4, 88 und VII, 3, 84] haben also nicht gleiche Kraft.' [Vgl. Paribn. Nr. 42.] Dasselbe gilt natrlich auch fr Vrddhi Vgl. Siddh. K. zu VI, 4,
88 : nityatvd
98

vug gunavrddhl ceti

bdhate.

K a i y . : Kiti

gunavrddhinisedhd

usw. ,Weil [die Regel 1,1, 5:] ,,Auch vor einem kit oder it" Guna und Vrddhi verbietet, und weil, wenn [nach VI, 4, 77] uva (uv) eingetreten st, der Eintritt von fr den Laut u [in VI, 4, 88 (ohne vuk) + 89 A] vorgeschrieben ist.1 Vuk ist also berflssig. 99 Kaiy. : Ac nitity atreka ity anupasthnd iglaksanatvbhvah ,Da in [der Regel VII, 2,115]: Fr [einen Laut des pratyhra] ac vor [einem Suffix] mit dem anubandha oder n" nicht ,,ikah" vorliegt, st eine Beziehung [der dort gelehrten Vrddhi] auf ik nicht vorhanden.' Acah lt vielmehr erkennen, da fr die in dieser Regel vorgeschriebene Vrddhi das Verbot in 1,1, 5, welches sich nur auf ik bezieht, nicht gilt. Dagegen bezieht sich die Guna-Regel VII, 3, 84, welche bei babh-itha in Betracht kommt, auf ik [vgl. die Ks.]; das in I, 1, 5 enthaltene Verbot mu sich also tatschlich auf VII, 3, 84 erstrecken. Deshalb st das W o r t ,guna' in dem Satze des Bhsya ,sytm evtra gunavrddhl' nicht am Platze. Hierber bemerkt Kaiy.: Gunagrahanam prasangoccritam ; sigunasyegiaksanatvt sidhyati hi pratisedhah ,Das W o r t ,,guna" ist nur wegen der Gelegenheit [d. h. nebenbei] erwhnt; denn da [die Substitution von] Guna [gem VII, 3, 84] vor [der Endung der 2. sing.] si (sip) [also nach, 1,1, 56 auch vor dessen Substitut tha] sich auf ik bezieht, so kommt das Verbot [von Guna durch 1,1, 5] tatschlich zustande.' So erklrt Kaiy. zu I, 2, 6 auch die Frage des Bhsya nach dem Zweck dieser Regel folgendermaen: Vinpi strenestam sidhyati saty api cestam

242 Bernhard Geiger

E r w i d e r u n g U n t e r diesen Umstnden 1 0 0 [sage ich] folgendes: man bedarf w e d e r des vuk noch auch [der Annahme], da [die Personalendungen nach I, 2, 6] kit seien. Guna und Vrddhi mgen hier [in bobh-itha, babh-a] e i n t r e t e n ; f r diejenige Pnultima, welche v o r l i e g t , wenn [die Substitution von] Guna und Vrddhi vollzogen, und [nach V I , 1 , 78] av und v eingesetzt sind, w i r d der E i n t r i t t von erfolgen. 101 W i e s o ? In , oh' [ V I , 4, 83] ist auch der Laut a (und a) zum Ausdruck gebracht. 102 E i n w u r f Dann bietet sich aber [ V I , 4, 83] auch hier bei [der Bildung d e r ] Akkusative plu ral is kJllapah, subhamyah dar.103 E r w i d e r u n g Die Elision [des von -p und -y nach VI, 4, 140] w i r d in diesen Falle [die Regel V I , 4, 83] vollstndig aufheben. 104 no sidhyati, babhvety ajloksonotvod vrddher igloksanotvbhovt protisedhprosongt. Gunomtronisedho evo tu bobhvitho, ohom bobhveti co nittvbhvopokse syd ti prosnah. .Auch ohne das Stra [I, 2, 6] kommt die erwartete [Form] zustande, und selbst wenn es gilt, kommt die erwartete [Form] nicht zustande; denn da bei [der Bilding von] bobhvo [in VII, 2, 115] die Beziehung der Vrddhi auf ,,oc" vorliegt, folglich eine Beziehung derselben auf ,,/'k" nicht vorhanden ist, hat [das Stra I, 2, 6 in Verbindung mit 1,1,5] nicht die Mglichkeit, [Vrddhi] zu verbieten. Vielmehr wre [das Stra] nur ein Verbot von Guna allein bei [der Bildung von] bobhvitho und der 1. sing, bobhvo in dem Falle, da [die Endung der 1. sing, nal nach VII, 1, 91] nicht n/t ist [also nicht, wie die 3. sing., Vrddhi, sondern nach VII, 3, 84 Guno bewirkt]. Dies ist der Sinn der Frage.' 100 D. h. da also erwiesen ist, da die Regel, I, 2, 6 berflssig ist, weil zwar bobhvitho, nichtaberdie3. sing. babhva zustande kme, wenn ihre Endungen als kit betrachtet wrden. 101 Der Opponent beharrt also bei der am Beginn der Diskussion gemachten Annahme eines Yogovibhgo mit gleichzeitiger Streichung von vuk. 102 Kaiy.: Akrokroyor d gunekrte asi-asos ceti prvoikdesenn nirdest ,Denn die Laute a und u [die nach I, 1, 69 zugleich ihre Lngen reprsentieren] sind [in oh] in der Weise zum Ausdruck gebracht, da, nachdem [fr den Vokal u] hinter dem a [und dieses selbst] Guno [o allein] substituiert worden ist, gem [VI, 1,110]: Vor [der Endung] os des Ablativs und Genetivs singularis" das vorangehende [o] allein substituiert wird.' D. h. o + u + os nach VI, 1, 87: o + oh, und dies nach VI, 1,110: oh. In Wirklichkeit ist oh natrlich gen. sing, von u. 103 Kaiy.: Oh suplty atrvarnosypi nirdesd yanprasagah. ,Da [nach d e r vorausgehenden Interpretation] in [der Regel VI, 4, 83]: Fr o (a) und fr u () [tritt] vor einer [vokalisch anlautenden] Kasusendung [der Halbvokal ein]" auch der Laut zum Ausdruck kommt, wrde sich [die Substitution des] Halbvokales [fr das von -p und -y] darbieten.' Dies wre jedoch ein Nonsens. 104 Die Regel VI. 4,140 lehrt die Elision des auslautenden a einer [durch Anfgung von vie (III, 2, 74; VI, 1, 67) zum Nominalstamm gewordenen] Wurzel, jedoch nur bei einem bha-Stamm, d. h. vor den vokalisch anlautenden Endungen derschwachen Kasus. Nach dieser Regel wird also im Acc. pl. aus killo-p + (vic) -\-as: killapah. Die Deutung von oh = a ()+u () + oh hat aber zur Folge, da sich gleichzeitig mit VI, 4,140 auch VI, 4, 83 darbietet. Es fragt sich also, welche dieser Regeln zuerst eintritt. Und gelangt VI, 4,140 zuerst zur Anwendung, so wrde sich gem VI, 4, 22 nachher doch noch 83 darbieten. Die Antwort auf die zweite Frage ist im Bhsya in dem Worte bdhoko gegeben. Vgl. Kaiy.: Lopo 'treti: porotvd ti bhvoh. to dhtor iti loposyvoksoh somyogoprvdkrnto dhtuh ; okroproslesosyottorotrvoksoh: bobhvo, bobhvitheti. Kiilopo ity otrobhoyoprosoge porotvd krolopoh, viprotisiddhe csiddhotvom no bhovotlti japoyisyote. .Gemeint ist: weil [VI, 4,140] eine spter gelehrte [Regel] st. Der Elision nach [140]: Fr das einer Wurzel" bietet Gelegenheit [zur

243 Mahbhasya

E i n w u r f Dann aber bietet sich [VI, 4, 83] hier dar, [nmlich bei] klllapau, klllaph [nom. pi.] 105 Erwiderung Unter diesen Umstnden [sage ich] folgendes : [Aus VI, 4, 77] gilt ,vyor' [d. i. ,fr u() und fr /(f)1]106 fort. Dadurch werden wir [in oh = a + u + as] den Laut u speziell [d. i. als allein wirksam] kennzeichnen [und VI, 4, 83 interpretieren]: ,0h' [soweit es] ,vyoh' [ist].107 Hier [in VI, 4, 88] gilt jetzt [nur noch] ,oh' fort, whrend ,vyoh' [fortzugelten] aufgehrt hat.108 Auch fr die [Anfgung von] yut (/) [in upadidi-y-e nach VI, 4, 63] soll nicht [eine Ausnahme] konstatiert werden. Mit Hilfe des Ausdruckes yut wird nicht [upadidy] gebildet werden. Der AusAnwendung] die Wurzel [pa], die auf endigt, welches der erste Laut der Verbindung [o () -f u ()] ist; die Verschmelzung des Lautes a () [mit u () zu o, also oh, d. i. VI, 4, 83] hat in einem zweiten Falle Gelegenheit [zur Anwendung, nmlich bei]: babhva, babhvitha. Da sich nun bei [der Bildung von] klllaph beide [sc. lopo und oh, gleichzeitig] darbieten, t r i t t [nach I, 4, 2] die Elision des [VI, 4, 140 zuerst] ein, weil sie spter gelehrt wird. Und da dort, wo [zwei sonst der Bildung v e r s c h i e d e n e r Formen dienende Regeln bei der Bildung einer ei nzigen Form] einander im Wege stehen, [die spter gelehrte, also zuerst eingetretene Kegel] nicht etwa [gem VI, 4, 22] als osiddha [in bezug auf die andere] betrachtet werden darf, wird [durch die Paribh. Nr. 40] angedeutet werden.' Vgl. im Anhang s. vipratisedha. Kaiy.: Asarvanmasthna ity onuvartand usw. ,Denn da [in I, 4,18 aus 17],,Nicht vor den Endungen der starken Kasus' zu ergnzen ist, also der Terminus bha [bei den starken Kasus klllapau, klllaph] nicht vorliegt, bietet sich [hier] die Elision desd[von-pd; VI, 4,140] nicht dar.' Es wrde sich also nur nach VI, 4, 83 Substitution des Halbvokales fr ergeben. Kaiy.: Yodeva yvor iti prakrtam usw. ,Was [in VI, 4, 77] als ,,yvor" vorgekommen ist, dasselbe st [hier] mit einer Umkehrung der Reihenfolge des / als ,,vyor" bezeichnet.' Yvor st gen. pl. von /' + u. Die Umstellung in vyor hat wohl darin ihren Grund, da das / bei ,oh' berhaupt nicht in Betracht kommt. 107 w e sonst oc/' oer anac u. a. m.
106 105

aus frheren Regel in einschrnkendem Sinne zu ergnzen sind, so soll hier ,oh' in VI, 4, 83 durch das aus 77 zu ergnzende, ,vyoh' eingeschrnkt werden. Oh bedeutet ,fr a und fr u'; vyor bedeutet,fr u und fr /'. ,Or vyoh' besagt also: ,fr a und fr u, [und zwar nur insoweit es] fr u und /' [geschieht].' Durch die Ergnzung von ,vyoh' wird also das in diesem nicht enthaltene o von ,oh' aus seiner W i r k samkeit ausgeschaltet, und solange ,vyor' fortgilt, bedeutet ,oh' demgem trotz seiner Zusammensetzung aus a -f- u + as nur ,fr u'. W o dagegen ,vyor' nicht mehr zu ergnzen ist, bedeutet ,ohl seiner Zusammensetzung gem ,fr o und fr u'. Kaiy.: Tatrkrokrasamudyanirdese 'pi usw. .Obwohl dort [in ,,oh"] die Verbindung der Laute a und u zum Ausdruck kommt, so t r i t t doch, weil [oh] durch ,,vyor" nher bestimmt wird, nur fr den Laut u der Halbvokal ein, nicht aber fr a1. Damit ist der Einwurf betreffs klllapau, klllaph zurckgewiesen. 108 Kaiy.: Iheti: Bhuvo lulitor d upadhy tyatra. .Hier, in [der Regel VI, 4, 88(ohnevuk) + 89A]: Beibh t r i t t vor den Endungen des Aoristes und Perfektums fr die Pnultima ein ein".' Die Ausgabe von Benares 1887 hat flschlich: bhuvovug lulitor usw. Vuk st zu streichen, denn der Opponent hat den Standpunkt ,nrtho vuk npi kittvena' nicht verlassen. Da ,vyor' bei VI, 4, 88 nicht mehr fortgilt, bedeutet das hier allein zu ergnzende ,oh' ,fr a und fr u', und zwar kommt fr babhva, babhvitha nur das a in Betracht. Das Resultat dieser Diskussion ist demnach: Vuk wird aus VI, 4, 88 eliminiert, und diese erhlt infolge eines Yogavibhga und

244 Bernhard Geiger

d r u c k y ut hat einen andern Z w e c k . W e l c h e n ? Da zwei / hrbar sein mgen. 109 E i n w u r f Es gibt keinen Unterschied in Bezug auf das H r e n eines oder m e h r e r e r y h i n t e r einem Konsonanten. 110 Ist denn nun aber [eine Regel] nur vor ,0ha1 [d. i. bis incl. VI, 4,128] als asiddhazu betrachten, oder einschlielich dieses [Abschnittes mit dem A d h i k r a ,bha', d. i. bis VI, 4,175]? W o h e r denn aber dieser Zweifel ? Es w i r d dies nmlich [in VI, 4, 22] durch [die Prposition] (an) ausgedrckt, und an erzeugt doch einen Z w e i f e l . So z. B. [entsteht in dem Satze] : ,Es hat bis Ptaliputra geregnet' der Z w e i f e l , ob [ n u r ] v o r Ptaliputra oder e i n s c h l i e l i c h desselben. Ebenso e r g i b t sich auch hier der Zweifel : vor ,bha' oder ei n.sch I i e l i c h [der G e l t u n g ] desselben ? Und w o r i n besteht denn da der Unterschied ? Prg bhd iti cecsunmaghonbhgunespasamkhynam || V r t t . 15. W e n n man a n n i m m t , da [VI, 4, 22] v o r ,6ha' gelte, so mte hinzugefgt w e r d e n , da [das Stra] bei 1. sun, 2. maghon und 3. bei [der Substitution von] Guna in bh [dem Substitut fr bahu] auch noch hinzugerechnet w e r d e n mge. 111 1. sunah (acc. pl.), sun, sune: wenn [in svan-as nach VI, 4, 133 die Substitution des] Samprasrana erfolgt ist, so bietet sich [bei suan-as] .Elision des a von an' [VI, 4,134] dar.112 W e n n dagegen Konsonanten wird ein Halbvokal oder Nasal vor einem Halbvokal oder Nasal elidiert" Elision vorgeschrieben wird, also zwei /tatschlich fakultativ (pakse) eintreten, so bedarf man [hiezu] nicht noch des Augmentes yut (y); also wird selbst fr den Fall, da ein Unterschied im Hren [eines 109 oder mehrerer y hinter einem Dies ist die Erwiderung auf den zweiten Teil des Vrtt. 14. Der Sinn Konsonanten angenommen wird], ist: wir substituieren in upadidi-e nicht etwa infolge der Vorschrift zunchst nach VI, 4, 82fr /"den Halbdieses [yut in VI, 4, 63] der Halbvokal vokal und erhalten upadidye; VI, 4, 63 [fr das in upadidJ-e] substituiert.' 111 lehrt nur, da hier zwei / hrbar sein Denn in diesen drei Fllen handelt sollen, da also upadidyye gebiIdet es sich um Regeln, die in dem Abwerde. Da also VI, 4, 82 zuerst einschnitt mit dem adhikra ,bha' [VI, 4, t r i t t , bedrfen wir der Regel VI, 4, 129175] stehen. 112 22 auch in diesem Falle nicht. Kaiy.: Suna iti : svan sas iti sthite 110 Kaiy.: Vyajanoporosyeti : vyanjont samprasranam; vrnd ngasya baiiparasyeti sup supeti samsah (,Vy. ist yastvt prvaikdesam bdhitvllopah ein Kompositum gem [II, 1, 4]: prpnoti. ,Wenn svan + as (acc. pl.) Eine Kasusform mit einer Kasusvorliegt, so erfolgt [die Substitution] form," nmlich aus vyanjantporasyo'). des Samprasrana [fr v]\ da nun Srut\bhedapakse 'pi yono maya iti [nach Paribh. Nr. 55] eine den Stamm dvirvacanavidhnd dhalo yamm yami betreffende [Operation: VI, 4VII, 4 opa iti pakse lopavidhnt pakse yak- incl.] grere Kraft besitzt [also radvayam bhavaty eveti nrtho yuteti frher eintritt] als eine [die Vertadvidhnasmarthyd yan na bhavati. bindung von] Lauten betreffende, so .Weil [einerseits] die Verdoppelung ergibt sich [in suan-as] die Elision des in [VIII, 4, 47 Vrtt. 1 :],.Eines Halba [VI, 4,134], mit Verdrngung der vokales nach einer Muta oder einem Substitution des einen vorangehenden Nasal (uern)" vorgeschrieben wird, [Lautes ufr u + a: VI, 1,108].' Denn auf der anderen Seite aber in [der VI, 4,134 ist rigam, VI, 1,108 aber Kegel VIII, 4, 64]:,.Nach einem der Ergnzung von ,oh' folgende Gestalt: [Or] bhuvo lulitor d upadhyh. Babhva, babhvitho werden gebildet, indem trotz I, 2, 6 Guna und Vrddhi fr das von bh substituiert werden, und fr die Pnultimaa, bezw. , welche dann vorliegt, eintritt.

245 Mahbhsya

f r diese [Substitution des Samprasarana g i l t ] , da [eine Regel] e i n s c h l i e l i c h dieses [Abschnittes ber bha] als asiddha betracht e t w i r d , w i r d dadurch, da diese [Substitution als asiddha betrachtet w i r d , das V e r b o t [ V I , 4 , 1 3 7 ] : ,Nicht aber, wenn [das a von an] auf eine Konsonantenverbindung folgt, welche auf v o d e r m endigt' [gegen 134] zur Geltung kommen. 1 1 3 E r w i d e r u n g Auch wenn f r diese [Substitution des Samprasarana: V I , 4,133 angenommen w i r d , da nur eine Regel] v o r bha als asiddha betrachtet w i r d [.Samprasarana also siddha ist], so bedeutet dies f r diese [Substitution des Samprasarana] keinen Fehler. W i e s o ? Es gibt hier keinen Unterschied, ob nun das Verschwinden [des a von an] durch die Elision des o [134] erfolgt, oder dadurch, da [nach V I , 1,108] der vorangehende [Vokal u fr u + a] eintritt. Einwurf 114 Der Unterschied besteht in folgendem : wenn das Verschwinden [desa] durch die Elision des o [VI, 4,134] erfolgt, so wrde sich [nach VI, 1,161] notwendigerweise der durch das Verschwinden des Udtta bedingte Akzent [fr die darauffolgende unbetonte Endung] ergeben.115 Erwiderung Hier bietet sich der durch das Verschwinden des Udtta bedingte Akzent nicht dar. Aus welchem Grunde? Weil [die Regel VI, 1,182:],Nicht hinter go, svan und einem im Nominativ singularis [auf] a [ausgehenden Stamme]' dies verbietet. Einwurf Dies ist nicht ein Verbot gegen den durch das Verschwinden eines Udtta bedingten Akzent.Wogegen denn? [Nur] gegen die Betonung der Endung des Instrumentals und der folgenden [Kasus, d. i. gegen VI, 1,168].116 Und wo es sich nun Wenn die Substitution des Samalso aus suan- mit falschem Akzent prasarana als asiddha betrachtet suna erhalten]. Wenn dagegen [in wird, also noch svan-as als vorhanden suan- nach VI, 1,108] die Substitution zu denken st, folgt das a auf eine Kon- des einen [Vokales u fr u -f-o] erfolgt, sonantenverbindung, welche auf v so ist gem [der Regel VIII, 2, 5]: endigt. Die Elision nach 134 wird also ,,Ein einziger [Vokal] als Substitut [fr durch das in 137 enthaltene Verbot einen unbetonten] samt einem verhindert. udttierten [Vokal] ist udtta" das 114 Die .Einwrfe' enthalten den [fertige] Wort in der ersten [Silbe] Standpunkt desjenigen, welcher udtta [d. i. sun mit richtigem Akbehauptet hat, da VI, 4, 22 auszent]; denn das a der Wortform svan nahmsweise auch noch in den ist infolge der Betonung des Suffixes genannten drei Fllen hinzugerechnet [an: III, 1, 3] End-uddtta.' [Bei Bhtl. werden mte, wenn dieses Stra zu VIII, 2, 5 ist also das Wort .voransonst nur vor ,bha' glte, also den gehenden' zu streichen.] Standpunkt des Upasamkhynavdin, 116 D. i. also der Kasus mit Ausnahme wie Kaiy. ihn nennt, im Gegensatz zu des Akkusativs und des Nominativs. dem des Pratykhynavdin, des Kaiy. : Upasamkhynavdy ha : naiseti ; Opponenten. So bemerkt Kaiy. hier: trtiydisvarasyeti : sv ekca iti prpta,Der Upasamkhynavdin sagt, um den sya; tatra pratisiddhe 'pi trtiydisvare Unterschied aufzuzeigen.' ope saty udttanivrttisvaraprasanga iti 115 Kaiy.: Anudttanimittatvd udttaviseso 'sti, na gosvann ity asya tu iopasya. Ekdese tu saty ekdesa udttenisedhasya phalam svabhym svabhir nodtta ity dyudttam padam bhavati, iti haldau vibhaktv asti. ,Der U. svasabdkrasyapratyayasvarensagt: ,,Naisa"\ [dann] ,,[Ein Verbot] ntodttatvt. .Denn die Elision des gegen die Betonung der Endungen udttierten [Vokales a] hat ihren des Instrumentals usw,": [d. .] Grund [vgl. ,yatra' in VI, 1,161] in der gegen diejenige, welche sich aus [VI, unbetonten Endung. [Wir wrden 1,168]:,,Eines im Lokativ pluralis
113

246 Bernhard Geiger

nicht um die Endungen des Instrumentals und der folgenden [Kasus] handelt, also bei sunah (acc. pl.)? 117 E r w i d e r u n g Dann [sagen w i r ] folgendes: w i r lehren nicht, da [VI, 4,182] eine [bestimmte] Regel [sc. VI, 1,168] verbiete, sondern es ist dies ein Verbot gegen die aus irgendeiner Regel sich ergebende Betonung einer Endung [berhaupt]. 1 1 8 ergibt [also auch den nach VI, 1,161, eintretenden udttanivrttisvara'].1 Nach Kaiy. wre also der Gedankengang: VI, 1,182 verbietet die Betonung der Endungen des Instr. usw. nur fr svabhym, svabhir, nicht aber fr diejenigen Endungen des Instr. usw., bei denen nach 161 der udttanivrttisvara eintreten mte [d.i. bei sun, sune]', und (tarhi) dort, wo nicht ein Instr. usw. vorliegt, also beim acc. pl. sunah, wie sollte 182 da den udttanivrttisvara verbieten? Kaiy. scheint miraber in der Unterscheidung zwischen svabhym und sun zu weit zu gehen. Ich mchte interpretieren: VI, 1,182 verbietet nicht wie der Pratykhynav. offenbar meint den udttanivrttisvara bei svan berhaupt, das Verbot in 182 erstreckt sich vielmehr nu rauf die Endungen des 117 Sa evho: yatra tarhlti: crthe tarhi- Instr. und der folgenden Kasus [also sabdah. Na kevalam sun, sune ity atra auf sun, sune, svabhym usw.]; wie sollte 182 unter diesen Umtrtiydisvare pratisiddhe 'sty udttanistnden (tarhi) den udttanivrttisvara vrtti[svara]p rasan go, 'pi tu yatra trtlydyabhvah sunah pasyeti tatrpy fr die Endung des Akkus, pl. verbieten? Kaiy. erwhnt am Schlu udttanivrttisvaraprasanga ity arthah. Na hi atra tad api sakyate vaktum : yena dieser Diskussion des Bhsya, da pthah usw.] vor kenacid api laksanena prptasya trtly- bisweilen [kvacittu 1 ,yatra tarhi der Satz ,evam tarhi yena disvarasya pratisedha iti. .Ebenderselbe kenacit prptasya trtiydisvarasya ' [sc. Upasamkhynav.] sagt [ferner] (,dann [sage ich] folgendes: [182 ist ein ,,yatra tarhi". Das Wort tarhi steht Verbot] gegen die Betonung [der im Sinne von ca: nicht nur bei sun, Endungen] des Instr. usw., die sich sune bietet sich der durch das Veraus einer beliebigen Regel ergibt') schwinden des Udtta bedingte Akzent eingeschoben erscheint. [In diesem [fr die Endung] dar, selbst wenn die Falle wre Kaiy.'s Unterscheidung Betonung der Endungen des Instruzwischen svabhym und sun berechmentals und der folgenden [Kasus tigt]. Bei der uns vorliegenden durch VI, 1,182] verboten wird [vgl. Lesung aber msse tarhi im Sinne von die vorhergehende Anm.], sondern ca erklrt werden, weil tarhi sonst auch dort, wo kein Instrumental oder ein verschiedenes Subjekt zu haben ein folgender [Kasus] vorliegt, im acc. pflege. pl. sunah [aus sun-ah] bietet sich ein durch das Verschwinden des Udtta bedingter Akzent [fr die Endung] dar. Dies ist der Sinn. Denn hier kann man nicht einmal einwenden, da [VI, 1,182 nicht speziel1168, sondern] diejenige Betonung der Endungen des Instrumentals usw. verbiete, die sich aus irgendeiner beliebigen Regel Kaiy.: Nasvekca ityasyaiva usw. ,Dies [sc. VI, 1,182] ist nicht nur ein Verbot gegen die Regel [168]:,,Eines im Lok. plur. einsilbigen [Stammes]," sondern gegen jeden Udtta berhaupt, der sich fr eine Endung ergibt [also auch fr die des Akk. pl. sunah]. Dies ist der Sinn. Aus dem Worte
118

einsilbigen [Stammes]" ergibt; obwohl unter dieser Voraussetzung (tatra) die Betonung der Endungen des Instrumentals usw. [durch 182] verboten wird, so liegt doch ein davon verschiedener Fall vor, wenn sich [in suan-, sun-e] nach dem Eintritt der Elision [des o] der durch das Verschwinden des Udtta bedingte Akzent [fr die Endung gem 161] darbietet. Das Verbot M nicht nach go, svan"[182] aber uert seine Wirkung [nur] bei svabhym, svbhir vor einer mit einem Konsonanten anlautenden Kasusendung [, vor welcher Samprasrana(W\, 4,133), also auch Elision des d (134) und udttanivrttisvara (VI, 1,161) sich nicht ergeben. VI, 1,182 erstreckt sich also nur auf Flle, in denen der udttanivrttisvara sich nicht darbietet].'

247 Mahbhsya

E i n w u r f W o nun aber die Betonung einer E n d u n g nicht in Betracht k o m m t , [ w i e bei] bahusnJ?"9 E r w i d e r u n g [ W i e wre es], wenn man diese [Regel VI, 1,182] vielleicht doch als V e r b o t auch gegen [jedweden] durch das Verschwinden des Udtta bedingten A k z e n t betrachtete? 120 E i n w u r f Solches ist nicht mglich. [ D e n n ] es w r d e sich auch hier darbieten, [nmlich bei] ki/mr/'.121 E r w i d e r u n g U n t e r diesen Umstnden [sagen w i r ] folgendes : das Vorgehen des Lehrers [ d . i. Pn.s] deutet an, da sich bei svan berhaupt nicht ein durch das Verschwinden des Udtta bedingter A k z e n t einstellt, da er ja [Ganaptha 8 1 , 64] das W o r t svan unter gaura usw. anfhrt [und IV, 1 , 41 die Bildung der Feminina dieses W o r t mittels ms, d. i. des betonten f, l e h r t ] . Er macht [also] wegen des En-udtta [von sunf] eine besondere Anstrengung. Denn [dieser] k o m m t schon durch nlp (/") zustande. 122 ,,gegen die Betonung e i n e r E n d u n g " [geht hervor, da dieser [sc. der Pratykhynav.] annimmt, das W o r t ,,vibhakti" gelte [in 182 aus 168 fort].' 119 Kaiy.: ,Hier spricht der Upasamkhynavdin ,,yatra tarhi1'.' Bahavah svnoyasym iti usw. ,[Bahusunl] ist ein Bahuvrihi mit der Bedeutung eine [Fahrstrae, rathy; vgl. Mbh. vol. II, p. 204, I. 3 v. u.], in der sich viele Hunde befinden". Daselbst ist nach Ansicht desjenigen, welcher die Elision des a behauptet, [an bahusvan] nach [der Regel IV, 1, 28]: ,,An einen auf an [auslautenden Bahuvrihi], dessen Pnultima [a] elidiert wird " nlp [d. i. das unbetonte Femininsuffix 1, angetreten]. Weil nun nach [der Regel V!, 2,1 75]: ,,Nach bahu [als erstem Glied eines Bahuvrihi], wenn dadurch die Vielheit des im zweiten Gliede Ausgedrckten [bezeichnet wird], wie bei der Negation [als erstem Glied: VI, 2,172]" [bahusvn] den End-udtto besitzt, so bietet sich [in bahusvan-1 infolge der Elision des d] der durch das Verschwinden des Udtta bedingte Akzent dar [und wir erhalten flschlich bahusunf].' Da in bahusn nach VI, 1, 68 eine Endung nicht in Betracht kommt, kann sich VI, 1, 82aufdiesen Fall nicht erstrecken, wenn es als Verbot gegen die Betonung jeder Endung von go, svan usw. aufgefat wird. Der udttanivrttisvara wird also durch diese Regel nicht verhindert. Kaiy.: ,Der Pratykhynav. sagt ,,yadi punar". Damit st gemeint: das W o r t ,,vibhakti" (Kasusendung") gilt nicht [aus VI, 1,168 in 182] fort.' D. h. das Verbot in 182 soll sich nicht
120

nur auf Kasusendungen, sondern auch auf jeden udttanivrttisvara bei svan usw. erstrecken. 121 D. h. wenn VI, 1,182 den udttanivrttisvara verbte, so wrde dies auch fr die in der Regel erwhnten Stmme gelten, welche im Nom. sing, auf o ausgehen, also auch fr kumra, dessen Femininum nach IV, 1, 20 mittels ip gebildet wird und nach VI, 1,161 den udttanivrttisvara erhlt: kumra + = kumrf. Bei Bhtl. zu VI, 1, 182 ist das ganz unbegrndete W o r t .einsilbig' zu streichen.
122 Sunf wrde auch durch Anfgung von nlp (unbetontem /") Zustandekommen: sudn+z nach VI, 1,161 = sunf. Wenn Pn. trotzdem die Anfgung von s (f) lehrt, so deutet er offenbar an, da bei svan niemals der udttanivrttisvara eintritt. Kaiy. : ,Der Pratykhynavdin sagt ,,evam tarhi". Mag auch der durch das Verschwinden des Udtta bedingte Akzent nicht durch [die Regel VI, 1, 182]: Nicht nach go, svan" verboten sein, so stel It er sich doch bei svan infolge einesJnpa/co nicht ein. Dies ist der Sinn. So hat der Opponent (pratykhynavdin) bewiesen, da kein Unterschied besteht [ob nun das Verschwinden des a von svn nach VI, 4,134 oder VI, 1,108 erfolgt]. 1 D. h. aus der Elision des a nach VI, 4,134 ergibt sich kein Fehler; das Stra VI, 4, 22 braucht also bei der Bildung von sun usw. nicht hinzugerechnet zu werden, wenn man annimmt, da es nur v o r ,bha' gelte. Bahusnl wird mittels nlp und Elision des gebildet: bahusun -(- 1 [wobei man aus dem Jnpaka ersieht, da nicht nach VI, 1,

248 Bernhard Geiger

2. Maghonah (akk. pl.) f maghon, maghone: wenn [in maghavanas nach VI, 4,133 die Substitution des] Samprasra na [fr v] vollzogen ist, so ergibt sich nach [der Regel VI, 4,148] : ,Fr / und fra' Elision [des o vor-un]. Wenn hingegen fr diese [Substitution des Samprasrana gilt], da [eine Kegel] ei nsch I iel ich dieses [Abschnittes ber ,bha'] als asiddha zu betrachten sei, so wird dadurch, da sie als asiddha betrachtet wird, [VI, 4,148] nicht eintreten. Erwiderung Aber auch wenn fr diese [Substitution des Samprasrana gilt], da [eine Regel] v o r ,bho' als asiddha zu betrachten sei, bedeutet dies fr diese [Substitution] keinen Fehler. Wieso? Er [Pat. zu IV, 1, 7] wird [nmlich] sagen: Die Wortform maghavan ist ein etymologisch nicht abgeleiteter Nominalstamm.123 3. [die Substitution von] Guna in bh, [bei der Bildung von] bhyn : wenn [nach VI, 4,158] der Eintritt von bh [fr bahu] erfolgt ist, so ergibt sich [nach 146 die Substitution von] Guna fr das . Wenn dagegen fr diesen [Eintritt von bh gilt], da [eine Regel] ei nsch I iel ich dieses [Abschnittes ber ,bha'] als asiddha zu betrachten ist, wird dadurch, da er [sc. der Eintritt von bh] als asiddha betrachtet wird, [die Substitution von Guna] nicht erfolgen.124 161 Oxytonese eintritt, da also das u den Ton erhlt] = bahusuni. Kaiy. bemerkt noch: Vidyote tu visesah; allope usw. .Doch ist [in Wirklichkeit] ein Unterschied vorhanden. Wenn nmlich die Elision desa [VI, 4,134] gilt, so mu [nach IV, 1, 28] hip eintreten, weil [bahusvn in diesem Falle ein Bahuvrihi auf an ist], dessen Pnultima elidiert wird: [wir erhalten also] bahusuni. Wenn dagegen diese [Elision] nicht stattfindet, so wird [nach IV, 1,12ohne ein Femininsuffix] bahusv gebildet. [Denn] auch das fr goura usw. [darunter svan] vorgeschriebene Suffix ms [IV, 1, 41 ] bietet sich wegen des Adhikra ,,Nicht hinter dem untergeordneten Gliede [eines Kompositums": IV, 1,14] nicht dar. [In dem Bahuvrihi bahusvan ist nmlich svan nach I, 2, 43 uposarjana]. Am Schlu [unseres Kapitels] wird ja auch festgestellt werden, da [eine Regel] einschlielich dieses [Abschnittes ber bha] asiddha ist. In diesem Falle mu [weil VI, 4,134 dann durch 137 verboten wird] eben nur bahusv gebildet werden, gleichwie suparv [nach IV, 1,12 ohne /"]. Dadurch aber, da die Abfassung [dieser Diskussion] sich nur von dem einen Ziele leiten lt, einen Unterschied hinsichtlich des Akzentes zu beseitigen, ist dies in Bhsya nicht richtig dargestellt worden. [D. h. es blieb unbeachtet, da bahusuni berhaupt nicht gebildet werden kann, wenn die Elision nach VI, 4,134 nicht erfolgt.] Aber auch bei [der Kegel IV, 1,13]: ,,Dp [das unbetonte Femininsuffix , tritt beliebig] in beiden Fllen [d. i. nach -man und nach einem auf an ausgehenden Bahuvrihi ein]" ist [von Pat. am Schlu] festgestellt worden, da nur bahusv [und nicht bahusk] gebildet werden darf.' Zu der folgenden textkritischen Bemerkung Kaiy.'s vgl. p. 246, Anm. 117, Ende. 123 Vgl.zuVrtt. 7.
124

Kaiy.: Nanu ca bhbhvasysiddha-

tvd usw. ,Aber wenn der Eintritt des [Substitutes] bh als asiddha betrachtet wird, so ist infolgedessen dieses [Substitut] doch [so anzusehen, als ob] noch [der sthnin, d. i.] das Wort bahu [dastnde]; es wrde sich also [gem ,,utsargalaksanabhvrtham ca" (vgl. p. 217)] tatschlich Guna [bei bh] ergeben? Dieses bedeutet keinen Fehler. Weil nmlich in bezug auf die [an bh] zu vollziehende [Substitution von] Guna der Eintritt von bh als asiddha betrachtet wird, so erfolgt der Eintritt von bh, nachdem [die Substitution von] Guna schon vorher [an bahu] vollzogen worden ist. Und fr Guna und den Eintritt von bh ergibt sich nicht etwa der Fehler, da man in einen Circulus (cakraka) gert;

249 Mahbhsya

E r w i d e r u n g Auch wenn f r diesen [ E i n t r i t t von i da [eine Regel] v o r ,bhal als asiddha zu betrachten ist, so bedeutet dies f r ihn [ d . i. den E i n t r i t t von bh] keinen Fehler. W i e s o ? W e g e n der Aussprache m i t langem [] w i r d [ d i e Substitution von Guna] nicht erfolgen. 125 E i n w u r f D i e Aussprache m i t langem [] hat einen andern Z w e c k . W e l c h e n ? [ D i e Bildung v o n ] bhman.*26 E r w i d e r u n g Dieses k o m m t schon infolge seiner ausdrcklichen Erwhnung 127 zustande. Welches ist die ausdrckliche Erwhnung? [In der Regel V I , 2 , 1 7 5 : ] ,Bahor navad uttarapadabhmni.*28 Oder aber es gelte [die Annahme], da [eine Regel] e i n sch I iel ich dieses [Abschnittes mit dem Adhikra] ,bhal alsasiddha zu betrachten sei: A bhd iti ced vasusamprasranayalopaprasthdinm pratisedhah|| Vrtt. 16. Wenn [man annimmt], da [eine Regel] b i s [ e i n s c h l i e l ich] 129 zum [Abschnitt ber], ,bha' [als asiddha zu betrachten ist], so m u i . fr [die Substitution von]Samprasrana in vas. 2. fr die Elision von y und 3. fr [die Substitution von] pra, stha usw. ein Verbot konstatiert werden : 1. [Bei der Bildung der Akkusative pl.] papusah, tasthusah ; ninyusah, cicyusah] luvuvusah, pupuvusah:1130 nachdem [die Substitution von] Samprasrana [nach VI, 4,131] vollzogen worden ist, kommen, wenn diese als asiddha betrachtet wird, infolgedessen die Elision des [64] und die brigen [Operationen, sc. 82 und 77, welche sich nur] ,vor einem Vokal' [ergeben] nicht zustande. Erwiderung Dies bedeutet keinen Fehler. [Im Vrtt. 12] ist [nmlich] folgendes gelehrt worden : Es ist vollstndig in Ordnung,

denn bei einem Clrculus trifft man je nach der erwnschten [Form] eine [ihr entsprechende] bestimmte Entscheidung.' D. h. wenn die Operation a die Operation b, und diese wieder a veranlat, so sieht man zu, welche Form bei den Autoritten erwnscht ist; wenn diese durch b erreicht ist, t r i t t a nicht wieder ein. In unserem Falle liegt ein cakroka nicht vor, weil der Eintritt von bh nicht direkt Guna veranlat, sondern in bezug auf dieses asiddha ist, also nicht erst die .erwnschte Form' fr das Eintritt oder Nichteintreten von Guna magebend ist. 125 D. h. Pnini sagt ausdrcklich bh mit langem , um anzudeuten, da Guna sich nicht mehr ergibt. 126 Kaiy.: Bhatvbhvd atra gunbhvah. .Weil [der Stamm vor dem Suffix man] nicht ein 5ho-Stamm st, so kann hier nicht Guna eintreten.' Das lange knnte also nicht den Zweck haben, den Eintritt von Guna in bhman zu verhindern. Das soll vielmehr andeuten, da vor dem Suffix -man fr

bahu zwar bhu mit kurzem u substituiert wird, da aber trotzdem bhman gebildet wird. 127 ber den Terminus niptana vgl. Kielhorn, Ind. Ant, vol. XVI, p. 245. [this volume, page 125]
128

Kaiy. : Hrasvnte

'py dese

kriya-

mne usw. .Obwohl das auf eine Krze ausgehende Substitut [bhu] eingesetzt wird, so wird doch bei bhman infolge der ausdrcklichen Erwhnung [dieses Wortes] die Lnge eintreten. Dies st der Sinn.' 129 K a i y . : Prg bhd ty asya paksasya usw. ,Da [ bht] als Gegenteil der einen Alternative prgbht" gebraucht wird [lies: pdiyamnatvd statt mnd~\, so st [die Prposition] im Sinne des Einschlusses [abhividhi, vgl. II, 1,13] zu verstehen.' 130 Kaiy.: Papsa itydinm parihrtnm api usw. .Obwohl papsah usw. [schon im Vrtt. 12 als nicht in Betracht kommend] zurckgewiesen worden sind, werden sie doch wieder vorgebracht, um zu zeigen, da der vorhin [sc. im Vrtt. 9] aufgezeigte

250 Bernhard Geiger

wenn man sagt, [eine Operation sei asiddha,] wenn sie von dem gleichen [Element] abhngig ist. Wieso? A u f [dem v v o n ] vas [beruhen] die Elision des und die brigen [Operationen], auf der Endung des auf vas endigenden [Stammes die Substitution des] Samprasrana. 2. Die Elision von / [in] saun balk [,ein mit der Sonne in gleicher Richtung (Hhe) befindlicher Kranich'] : wenn dasjenige a, welches [nach VI, 4,148] vor [dem Taddhito] an (a) elidiert wird, als asiddha betrachtet wird, so bietet sich infolgedessen nicht nach [der Kegel 149]: ,Vor V die Elision des/dar. 131 Erwiderung Auch hier vermeidet man [Schwierigkeiten] in der Weise [,da man erklrt]: Es ist dadurch vollstndig korrekt, da man sagt, [eine Operation sei asiddha,] wenn sie von dem gleichen [Element] abhngig ist. Wieso ? Auf [dem Taddhita] an (a) beruht die Elision des Lautes a, auf dem /"des auf an (a) endigenden [Stammes] die Elision des y.132 Fehlersich gerade bei dieser Alternative [sc. t,saha tena"] ergibt.' Vgl. zu Vrtt.12. 131 Kaiy.: Souriti: sryenaikadig ityan usw. ,Saun: [um auszudrcken] in gleicher Richtung (Hhe) mit der Sonne befindlich" wird [nach IV, 3, 112 an srya das Taddhito an (a) gefgt [welches nach VII, 2,117 Vrddhi bewirkt]; gem [der Regel VI, 4, 148]: Fr/ und fr a" erfolgt die Elision des dem an vorangehenden a]; hierauf tritt [gem IV, 1,15 das Femininsuffix] np () ein; nun wieder gem [der Regel VI, 4,148]: Fr / und fra" Elision des an (o) [vor/"]. Wenn nun hiebei beide Elisionen von o [d. h. nicht allein die von an vor /"] als asiddha betrachtet werden, so ist infolgedessen das y nicht Pnultima [des Stammes vor /, sondern drittletzter Buchstabe], und so bietet sich die Elision des y [149] nicht dar. [Einwurf:] Aber auch wenn [man annimmt, da eine Regel nur] vor bha" als asiddha zu betrachten ist, so st, weil dann beide Elisionen [d. h. auch die des a vor an] siddha wren, das y nicht Pnultima [des Stammes vor i, sondern letzter Buchstabe], folglich bietet sich die Elision [des y (lies: yalopa0) hier] ebenfalls nicht dar. [Denn] auch [die Annahme, da der fr on substituierte opa] sich wie der sthnin [an] verhalte [vgl. Pat. zu VI, 4,149 Vrtt. 12], ist [nach 1,1, 58] verboten, weil es sich um eine Vorschrift ber die Elision eines y handelt. [Erwiderung:] Es liegt [bei der Annahme ,,prgbht"] kein Fehler vor, wenn man upadh [in VI, 4,149] nach der Methode ,wenn es nur vorher dagewesen st" auffat [also: fr y, wenn es auch nur vorher Pnultimagewesen st"]. Oderauf Grund des Wortlautes [d. . ,,fry, solange es Pnultima ist"] wird die [umgekehrte] Reihenfolge gewhlt werden, nmlich zuerst die Elision des y, nachher Elision des a [d. i. an], obwohl die Elision des a (an) nitya st [, also gem Paribh. Nr. 38 frher eintreten sollte]' D. h. man bildet: srya + a(n) = saury -f- a(n) ; saury +a(n) + l. Weil nach VI, 4,149 das y nur elidiert wird, wenn es Pnultima eines Stammes ist, und weil es nicht mehr Pnultima wre, wenn a(n) vorher abfiele, wird zuerst y als Pnultima des auf a(n) ausgehenden Stammes wegen des folgenden / elidiert: saur -f a(n) -f- /", und jetzt erst erfolgt die Elision des a(n). So wrde also nach Kaiy. derjenige, welcher prg bht" annimmt, beweisen, da diese Annahme keinen Fehler zur Folge hat. 132 Dieser Auffassung stimmt auch die Ks. zu VI, 4,149 zu, wo es heit: Ani yo yasyeti opas usw. .Diejenige Elision, welche [fr das a] wegen des folgenden a(n) gem [der Regel 148]: Fr / und fra" erfolgt, ist nicht als asiddha zu betrachten, weil es nicht [von dem gleichen Element] abhngt [wei die Elision des a(n)]. Dagegen wird die [Elision, welche fr das a(n)] wegen des folgenden J eintritt, als asiddha betrachtet, [weil sie von demselben Element abhngt wie die Elision des y], und infolgedessen wird das y als Pnultima des auf a(n) ausgehenden

251 Mahbhsya

3. Bei [den S u b s t i t u t i o n e n v o n ] pra, stha usw. [ i n ] preyan, stheyn: w e n n [ d i e S u b s t i t u t i o n e n v o n ] pra, usw. [ V I , 4 , 1 5 7 ] als asiha betrachtet werden, so ergibt sich nicht gem [der Regel 163] : ,Ein einsilbiger [Stamm verbleibt] in seiner ursprnglichen Gestalt' das Verbleiben der ursprnglichen [d. i. unvernderten] Formen [pra usw.].133 Erwiderung Dies bedeutet keinen Fehler. Wie sich zwar dadurch, da [die Substitutionen von] pra, stha usw. als asiddha betrachtet werden, das Verbleiben der ursprnglichen Formen [pra, stha usw.] nicht ergibt, ebenso wird doch auch die Elision des letzten Vokales samt dem etwa darauffolgenden Konsonanten [VI, 4,155] nicht eintreten.134 P.VI,4,132:Vhat. Warum wird t nicht vorne angefgt? Aus [der Regel 1,1, 46, welche besagt] :,Vorne [wird angefgt], was ein t zum Anubandha hat' ergibt sich, [da t] vorne [antritt] ?135
ha-Stammes von suryo [in saury(a)-i] wegen des folgenden /'elidiert.' Anders Siddh. K. zu VI, 4,149: agasyopadhy yasyo lopah syt, sa ced yah srydyavayavah. ,Fr ein y, welches Pnultima eines [noch unvernderten] Stammes ist, soll [vor einem Taddhita oder ] Elision eintreten, wenn dieses y einen Bestandteil von srya usw. bildet [, nicht aber fr das y der von srya usw. abgeleiteten Stmme].' Diese Interpretation der Regel schliet sich zum Teil Pat.'s Erklrung zu VI, 4,149 Vrtt. 3 an und in der Ausdruckweise dem Kommentar Kaiy.'s zu dieser Stelle, unterscheidet sich aber von Pat.'s, bezw. Kty.'s, Ansicht dadurch, da dort das W o r t upadhyh der Regel als berflssig erklrt wird. Kaiy. erklrt dort: hi taddhite ca yad agam ansritarpavisesarn tasya yakrasya lopah, sa ced yakrah srydyavayavo bhavatiti strrthah. Es mte also nach VI, 4,155 der letzte Vokal von pra usw. elidiert werden. 134 Zum Verstndnis der folgenden Ausfhrungen Kaiy.'s sei daran erinnert, da nach VI, 4, 22 Vrtt. 1 das W o r t asiddha einen doppelten Zweck hat: 1. da die durch das Substitut bedingte Operation verboten werde, 2. da die durch das Ursprngliche (utsarga = sthnin) bedingte Operation stattfinde. Kaiy. : Nanv desalaksanah prakrtibhvo usw. ,[E i n wu rf:] Aber man kann doch einwenden: das durch das Substitut [pra usw.] bedingte Verbleiben der unvernderten Form [pra
133

usw.] besteht nicht zu Recht, wenn [die Substitution] als asiddha betrachtet wird. Dann aber bietet sich wiederum die Elision des auslautenden Vokales samt dem etwa darauffolgenden Konsonanten [von pra usw.] dar, da sie sich aus dem [noch als vorhanden zu denkenden] sthnin [priya usw.]ergibt?[Erwiderung:] Dies bedeutet keinen Fehler. [Denn] wie sollte sich Elision des auslautenden Vokales fr den sthnin [priya usw.] ergeben, der gar nicht vorhanden ist, da er durch ein Substitut [pra usw.] aufgehoben wrde, welches [noch vor der allgemeinen Regel VI, 4,155] eingetreten ist, weil es sonst keine Gelegenheit [zum Eintreten] gehabt htte [vgl. Paribh. Nr. 57-58]? Selbst wenn hiebei, noch b e v o r [die Substitute pra usw.] als asiddha betrachtet werden, die Elision des ti [von priya usw.] vollzogen wird, so bietet sich trotzdem nach dem Eintritt der Substitute, da diese [in bezug auf VI, 4,155] asiddha sind, nicht eine durch diese [Substitute] bedingte Elision von ti dar; noch auch eine durch das Ursprngliche [priya usw.] bedingte: denn bei den Ursprnglichen wre sie schon vorher eingetreten. Und weil [infolge des Ass/d/io-seins von pra usw.] keine Gelegenheit [fr den Eint r i t t von tilopa] vorhanden ist, so ergibt sich fr pro usw. auch nicht der Fehler eines Circulus (cakraka).' Vgl. den hnlichen Fall p. 248, Anm. 124. 135 Kaiy. : Yadyapi cchvoh sd ity atrsya usw. .Obwohl [von Pat.] bei [der Regel VI,4,19]: cchvoh sd [anunsike

252 Bernhard Geiger

E r w i d e r u n g Durch [das W o r t ] tsamprasranam' [VI, 4,131] w i r d die Stelle des Halbvokales [v] in Beschlag genommen. 136 E i n w u r f W e n n es sich so verhlt, Vha dvacannarthakyam samprasranena krtatvt || V r t t . 1. so ist fr vh das W o r t th unntz. W a r u m ? ,Weil es durch Samprasra na vollbracht w i r d ' , [d. h.] schon durch [die Substitut i o n des] Samprasarana [u f r v] k o m m t [die zu bildende Form] zustande. W i e k o m m t denn [auf diese Weise] die Form prasthauhah (acc. pl.) zustande ?137
ca]" dargelegt worden ist, da dieses [t] mit dem Anubandha th [und nicht t] versehen ist, fat er es [hier] trotzdem nach dem bloen Gehr [als tit] und stellt demgem die Frage.' Vgl. die Ks. zu VI, 4,19: Othas thitkaranam etyedhatythsviti visesanrtham ; vha d ity oyam api thid eva ,Die Setzung von th mit dem Anubandha th hat den Zweck, [es als das th der Regel VI, 1, 89] zu bezeichnen: ,,Wenn [auf a () das e der Wurzeln] /', edh, oder ein (th) folgt, [so t r i t t als alleiniges Substitut Vrddhi ein]". Auch dieses [t der Regel] ,,Vha t" hat eben diesen Anubandha th.' Die Bemerkung Bhtl.'s [in der .Erklrung dergramm. Eiern.' s. th] ,Der Ausgang st bedeutungslos' st demnach unrichtig. 136 Kaiy.: Vkyasya bhvivarnasya v usw. ,Ob nun der Terminus Samprasrana als Bezeichnung des Satzes [,,igyanah": 1,1, 45] oder des Lautes, der eintreten soll, verstanden wird [vgl. Mbh. zu 1,1, 45; Ind. Ant. vol. XVI, p. 244] [this volume, page 124], t r i t t th an die Stelle des Halbvokales. Aber warum wird th n i c h t u nm i t t e l bar v o r den Halbvokal gesetzt? Weil man auch anderen Orten, an denen Samprasrana erwhnt wird, annimmt, da ,,yanahlt ein Genitiv [zur Bezeichnung] der Stelle [1,1, 49] ist, so ist auch hier nur der Genitiv, der die Stelle bezeichnet [, an der etwas eintritt], richtig angewendet. Wie es denn auch heit: Siebzehn Substitute machen es notwendig, da [der Genitiv zur Bezeichnung] der Stelle [an der etwas eint r i t t ] , gebraucht wird. Die Ansetzung [von th] mit dem Anubandha th aber [lies: thitkaranam tu] bezweckt [Substitution von] Vrddhi [nach VI, 1,89].' 137 Die Vrttikas 2 und 3 geben an, wie die Formen prasthauhah, visvauhah usw. auch ohne uth zustande kommen wrden. Kaiy.: Prasthavh as ti sthite usw. .Wenn prasthavh [ + nvi] -h as vorliegt, so t r i t t , nachdem [die Substitution von] Samprasrana [u] fr das v [VI, 4,132] und die alleinige Substitution des [dem ] vorangehenden [u gem VI, 1,108] erfolgt ist, Guna o ein. welches auf dem Suffix nv/[III, 2, 64; VI, 1,67; 1,1, 62; VII, 3, 86] beruht. Wenn sodann nach [der Regel VI, 1,88]:,. Vrddhi [ist alleiniges Substitut], wenn [auf a ()] ein Diphthong folgt" Vrddhi eingesetzt worden ist, so kommt prasthauhah usw. zustande.' Man knnte nun einwenden, th sei unentbehrlich bei der Bildung von slyhah aus slivh und prauhah aus pravh, weil sich nach der Substitution des Samprasrana u fr v und von Guna slyohah und prohah [aus pra -\-oh-\-ah gem VI, 1, 94] ergeben wrden. Gegen eine derartige Einwendung richten sich die folgenden Worte Kaiy.'s: Anakrnte copapade usw. .Wenn ferner das Vorderglied [des Kompositums] nicht auf o ausgeht [sli-], so erscheint im Veda nicht [wie sonst nach III, 2, 64] nvi [hinter der Wurzel vah]\ folglich kommt auch fr das Zustandekommen von slyhah usw. die Vorschrift von th nicht in Betracht. Oder wenn man [dort nvi] anwendet, so wird doch slyhah von der in der Bedeutung .fhren' auftretenden Wurzel h vor folgendem kvip[\\\, 2, 61] gebildet werden. Und auch wenn eine [auf o auslautende] Prposition das Vorderglied bildet [wie dies bei pra-vh -f as der Fall wre], t r i t t im Veda nach vah nicht das Suffix nvi ein, aus dem sich [nach VII, 3, 86 Guna und dann nach VI, 1, 94 die Substitution eines einzigen Vokales] in Gestalt des [auf die Prposition] folgenden, Vrddhi verdrngenden [o] ergeben wrde, in

253 Mahbhsya

Gunah pratyayalaksanatvt || V r t t . 2. Mit Hilfe d e r das Suffix [nvi] betreffenden Regel [ V I I , 3, 86] wird Guna eintreten. Ejgrahand v.rddhih || Vrtt. 3. Vermge des Wortes ,ej" [.Diphthong' : VI, 1, 88] wird [sodann fr o + o] Vrddhi eintreten. Erwiderung Unter diesen Umstnden [sagen wir] folgendes: Wenn der Meister [Pn.], obwohl [prasthauhah usw. auch ohne th] tatschlich zustande kommt, th [als Samprasra na] fr [das v von] vh lehrt, so deutet er dies speziell an, da die folgende Paribhs gilt : Eine Bahiranga-Regel ist als nicht in Kraft getreten [bezw. als nicht vorhanden] zu betrachten in bezug auf eine [zu vollziehende] An ta ranga-Regel.138 Welcher Zweck ist in der Andeutung dieser [Paribhs] enthalten ? [Die Bildung von pacvedam, pacmedam: weil nmlich lichen Leben ergibt. [Vgi. Paribh. Transi., Preface, p. IVf.] Denn die Menschen beschftigen sich [zunchst] mit dem, was ihre eigene 138 Person betrifft [vgl. pratyagavart Kaiy.: Samprasranam yajdipraloko laksyate: Mbh. vol. I, p. 145, II. tyayanimittabhasamjnsrayatvd usw. ,[Die Substitution des] Samprasrana, 23 ff.; Paribh. Text, p. 49, II. 1Off.]? die bahiraga ist .weil sie von dem Ter- [Erwiderung:] Dies trifft nicht zu. [Denn nur] dort, wo eine Anzaragaminus bha [VI, 4,129, also von einem und eine Bar/rarga[-Operaton] sich vokalisch oder mit Halbvokal gleichzeitig darbieten, soll die anlautenden Suffix] abhngt, der An tarar ga [-Ope rat on zuerst] einselbst durch die Suffixe der Wurzeln treten, weil man die dem gewhnyaj usw. [VI, 1,15] bedingt ist, ist als osiddha zu betrachten in bezug auf die lichen Leben angehrige Maxime annimmt. Hier dagegen st die zu vollziehende [Substitution von] Antararga[-Operaton] durch die Guna, welche antaraga ist, da sie ah/ranga[-Operation] bedingt, die nicht [auch vom Terminus bha, dem gewhnlichen Leben angehrige sondern von dem innerhalb gelegenen Maxime kommt also nicht in Betracht.' nvi] abhngt. Guna wrde also nicht D. h. wie der Mensch am Morgen der eintreten, und wenn dieses nicht Reihe nach zuerst seine eigenen vorhanden ist, wrde [aus prastha + uh + oh] nicht die Form [prasthauhah] Angelegenheiten besorgt und dann die seiner Freunde usw., so geht man zustande kommen. So deutet denn auch dort, wo eine Antaraga- und die Vorschrift von th die Paribhs eine ah/rariga-Operation sich ,,asiddham" [usw.] an.' Mit yajdipratyayanimittabhasamjnsrayatvt meint gleichzeitig darbieten, der Reihe nach vor und vollzieht zuerst die nherKaiy.: Wenn in VI, 4,132 das Sampraliegende Antararga-Operation. In srana von Suffixen [u. zw. mit diesem Falle st die Paribhs nyyasiVokalen oder Halbvokalen anddh. Wo aber die Antararga-Operalautenden] abhngig gemacht wird, so tion sich erst darbietet, nachdem die hat dies darin seinen Grund, da fr ah/ranga-Operation vollzogen yaj usw. [darunter vah: Dhtup. 23, worden st, da st die Paribhsjnpa35] in VI, 1, "\ 5 Samprasrana vor kasiddh. Im ersten Falle bedeute asigewissen Suffixen vorgeschrieben ddham ,als gar nicht vorhanden zu bewird. Vgl. die graphische Dartrachten', im zweiten Falle, .[obwohl stellung Paribh. Transi, p. 236. schon eingetreten, doch] als nicht Kaiy. schliet mit folgender Bemereingetreten zu denken*. Vgl. auch kung: Nanu naitaj jnpakasdhyam im Anhang s. antaraga. W i r bilden usw. ,[E i n w u rf:] Aber zu dieser also : prastha-vh + ah ; nach VI, 4,132 [Paribhs] sollte man doch nicht erst auf Grund eines Jnpaka gelangen, und VI, 1,108: prastha-h + ah und nach VI, 1, 89: prasthauhah. da sie sich schon aus dem gewhnwelchem Falle die Vorschrift von th notwendig wre [, um Vrddhi zu bewirken].' Vgl. dazu Paribh. Transi, p. 235, n. 3 und p. 238 f.

254 Bernhard Geiger

die Bahiraga-Regel [ V I , 1, 87, d. i. die Substitution von] Cuna nach a [fr dieses + /] als osiddha betrachtet w i r d , t r i t t die AntaragaRegel [III, 4, 93, d. i.] die Substitution von ai [fr e] nicht ein. 139

Anhang !. Antaraga und bahiranga[Zu p. 236, Anm. 82.] Kaiyata trgt zwei verschiedene Deutungen des Satzes vor, da eine Antaragaund ein o/i/rongo-Operation sich gleichzeitig darbieten mssen. Die erste Deutung gibt den Worten des Bhsya den Sinn : zwei Operationen sind antaraga und bahiraga, wenn ihre Vollziehung sich noch v o r der A n w e n d u n g der Paribhs gleichzeitig darbietet, nicht aber in dem Falle, wenn die Antaranga-Operation sich erst nach und infolge der Ausfhrung der ohiranga-Operatlon ergibt. Wenn diese Auffassung richtig wre, drfte die Paribhs nicht nur bei der Bildung von pacvedam aus pacva dam [vgl. Paribh. Transi, p. 271, n. 4], sondern auch bei der Bildung von visvauhah usw. [vgl. ibid. p. 236, n. 1], fr welche die Paribhs doch zunchst angedeutet sein soll, nicht zur Anwendung gelangen. Denn in diesen Fllen ergeben sich die Antaraga-Operationen erst nach und infolge der Ausfhrung der oh/ronga-Operationen. Nach der zweiten Deutung sind zwei Operationen antaraga und bahiraga, wenn im A u g e n b l i c k der A n w e n d u n g der Paribhs beide gleichzeitig zur Stelle sind. Dies ist nun bei pap-us + as nicht der Fall. Aus den von Kaiyata genannten Grnden ist hier die Paribhs nach VI, 4, 22 in bezug auf die oh/ronga-Operation als asiddha zu betrachten. Und in dem Augenblick, in welchem die Paribhs angewendet wrde, wre die ohirongo-Operation [d. i. die Substitution des Samprasra na] als asiddha zu betrachten, und infolgedessen wrde sich auch ihr Korrelat, die AntaragaOperation, nicht darbieten. Da die zweite Interpretation auch der Ansicht Patajalis entspricht, geht aus dem Bhsya zu VIII, 3,15 Vrtt. 2 hervor, wo ein hnlicher Fall errtert wird. Dort behauptet Patajal i von einer Antaraga-Operation [VIII, 3,15] und einer Bahiranga-Operztlon [VII, 2,117], da sie sich nicht gleichzeitig darbieten, und nennt als Grund : asiddhatvt. Nach Kaiyatas Erklrung bedeutet dies : weil die Antaranga-Regel VIII, 3,15 gem VIII, 2,1 als asiddha zu betrachten ist in bezug auf die in VI, 4,132 angedeutete Paribhs; weil also die AntarangaRegel nicht vorhanden ist, so ist auch VII, 2,117 nicht eine zu ihr in Beziehung stehende ah/ranga-Regel ; die Bahiraga-Paribhs gelangt also nicht zur Anwendung, und infolgedessen ergibt sich VII, 2,117. Auch die Ks. zu VI, 4, 22 akzeptiert die zweite Deutung und bemerkt : Es hi paribhs bhc chstrly ; tasym pravartamnym vasusamprasrandlnam bhc chstrlynm evsiddhatvd antaragabahiragayor yugapatsamupasthnam nsti paribhs na pravartate. ,Yugapatsamupasthnam' bedeutet also nicht, da beide Operationen sich vor der Anwendung der Paribhs noch unausgefhrt darbieten mssen, wie bei syona. Die o/i/ranga-Operation darf vielmehr bereits vollzogen sein, wie bei pacvedam. Dies stellt NgojTbhatta, Paribh. Text, p. 43, I. 15 ausdrcklich fest in dem Satze: Antarage kartavyejtam tatklaprptikam ca bahiragam asiddham ity arthah. ,ln bezug auf eine zu vollziehende Antaranga-Operation ist eine Bahiranga-Operstlon,
In pacvedom gilt nmlich der ekdesae[fra + i] nach VI, 1,85
139

auch als Auslaut des Vorangehenden. Vgl. Paribh. Transi, p. 271, n.4.

255 Mahbhsya

sowohl wenn sie bereits zustande gekommen ist, als auch wenn sie sich gleichzeitig m i t jener darbietet, als asiddha zu betrachten.' Kaiy. : Popusa iti : nanu cantara ngatvd it bhvyam usw. :140 Aber [vor vos] mte doch [nach VII, 2, 35 und Paribh. Nr. 50] das Augment / (it) eintreten, da es antaraga ist [,das Samprasranafr v in vas aber bahiraga] ? Und das it verschwindet auch dann nicht mehr, wenn [seine Ursache, da v von vas verschwunden ist, d. i. die Substitution des] Samprasrana [fr v] vollzogen worden ist. Denn der Verfasser des Bhsya [Pat.] hat die Paribhs [Nr. 56] nicht angenommen [.welche lehrt]: ,Wenn [beim Eintritt einer Bahiranga-Regel] die Ursache [einer Antarago-Operation] verschwindet, so verschwindet auch das durch jene Ursache Bewirkte'. 141 Und ebenso ist auch [im Bhsya] zur [Regel VI, 3, 138] : ,Vor ac (cu) [wird der Endvokal des Vordergliedes verlngert]' gelehrt worden : ,Hier stellen andere [Lehrer] ein Verbot gegen [den Eintritt des] pratyaga [d. i. der Anta ranga -Kegel VI, 1, 77] vor ac (cu) auf; dies mte doch auch hier [d. i. in der Regel selbst, irgendwie] bewerkstelligt werden [.wenn VI, 3,138 berhaupt eintreten soll].' Dies ist deshalb gesagt worden, weil die Substitution des Halbvokales [fr das /' in prati-ac + as nach VI, 1, 77] usw. [d. i. die Substitution von fr a + o in pra-ac + as nach VI, 1,101, welche Regeln antaraga sind,] nicht aufgehoben werden, auch wenn [ihre Ursache, d. i.] das a [von ac infolge des Eintrittes der Bahiranga-Regel VI, 4,138] verschwunden ist.142 Und ebenso ist auch [im Bhsya zum Vrtt. 2] bei [der Regel VI, 4,19]: ,Fr cch [ = t + ch] und v [wird auch vor einem Nasal s, bezw. substituiert]' bemerkt worden : ,Hier mu notwendigerweise eine besondere Bemhung gemacht werden, damit nicht [in prasna aus prach + na nach VI, 1, 73] das Augment t (tuk) vorhanden sei ; denn [die Anfgung von tuk: VI, 1, 73] bietet sich
Den korrekten Text findet man Paribh. Transi, p. 313, n. 1. 141 Vgl. Paribh. Text p. 65, I. 3: Kim cais bhsye na drsyate. Tad uktam asiddhavatstre [VI, 4, 22] Kaiyatena: nimittpye naimittikasypy apya iti paribhsy bhsyakrtnsrayand iti. Durch nimittpye usw. gibt Kaiy. die Paribh. Nr. 56 wieder, welche lautet: Akrtavyhh Pniniyh ,die Anhnger des Pnini stellen keine besonderen Erwgungen [ber das Eintreten einer Regel] an [d. h. kmmern sich nicht um ihr Eintreten, wenn eine Ursache derselben verschwindet].' Nach Paribh. Text p. 61, I. 9 ff. st der Sinn dieser Paribhs, da eine Antaranga-Operatlon berhaupt nicht stattfindet, wenn nachher, infolge des Eintretens der BahiragaRegel, die Ursache der AntaragaOperation verschwnde. Nach Kaiy.'s Auffassung [nimittpye usw.] dagegen besagt die Paribhs, da das b e r e i t s ei n g e t r e t e n e Resultat
140

II. Nimittpye naimittikasypy apyah [Zu p.234, Anm.74.]

einer Antarongo-Operation wieder verschwindet, wenn seine Ursache verschwindet. Weil aber diese Paribhs so folgert Kaiy. weder von Pat. erwhnt, noch auch in den folgenden Beispielen [ac und tuk] angewendet wird, kommt sie auch in unserem Falle, d. i. fr it, nicht in Betracht, ,/t mte also als antaraga eintreten, und durch dieses [nicht aber durch das u in pap + i-us + os] wrde dann [nach VI, 4, 64] die Elision des bewirkt werden. Unter diesen Umstnden wre jene [Bemerkung Pat.'s ,samprasrane krte1 zu VI, 4, 22 Vrtt. 9] ganz ungereimt.' [Paribh. Text p. 64, I. 9 f.] Wenn aber it eintrte, wrden wir papyusah erhalten. 142 D. h. weil die Paribh. Nr. 56 nicht anerkannt wird. Anstatt von dieser Gebrauch zu machen, antwortet Pat., Pnini habe in VI, 3,138 durch die Vorschrift der Lnge fr den dem ac vorangehenden Vokal angedeutet, da

256 Bernhard Geiger

deshalb [zuerst, d. i. v o r V I , 4 , 1 9 ] dar, weil sie ontoraga ist.143 Auch dies ist auf G r u n d der Ansicht behauptet w o r d e n , da das [als ontaraga] bereits eingetretene t (tuk) nicht verschwindet, auch wenn [dessen Ursache ch verschwunden, d. h. gem der Bahiranga-Regel V I , 4 , 1 9 ] s [ f r ch] s u b s t i t u i e r t w o r d e n ist.144 E r w i d e r u n g U n t e r diesen Umstnden 145 [sagen w i r ] f o l gendes : wenn man, indem man das fertige W o r t 1 * 6 [papusah] v o r Augen hlt, die Auflsung [desselben in seine Bestandteile] v o r n i m m t und die einzelnen Teile [nebeneinander] hinstellt, so liegt nicht eine Antarago- und eine Bahi ra ga[-Rege\] v o r . W e n n nmlich [die Teile] pap + vas + as dastehen, so t r i t t [zunchst] das Samprasra na [fr v von vas] ein, indem dieses, weil es nitya ist,147 das it verdrngt. Und da es sich [, wenn nachher gem Vli, 2, 35 it eintreten soll] um eine auf einen Laut [d. i. v des sthnin vas beruhende] Regel handeln wrde, gilt nicht [die Regel 1,1, 56], da [das Substitut u] sich wie der sthnin [v(as)] verhlt, und infolgedessen bietet sich [die Anfgung von] it nicht dar. Oder aber148 [die Substitution des] Samprasarana [VI, 4,131] und dasjenige, was darauf beruht [d. i. die Elision des ] besitzt als [Pratipadavidhi] grere Kraft [als VII, 2, 35] und in diesem Bereich [sc. einer Pr atipada-Regel] t r i t t it nicht frher ein.149 die Antarongo-Regel VI, 1, 77 [Substitution des Halbvokales fr diesen Vokal] nicht eintritt. Vgl. Paribh. Text p. 64, 1.10 ff. und Transi, pp. 317 f. und318, n.1. 143 Die besondere .Bemhung' besteht darin, da Pn. in VI, 4,19 die Substitution von s fr cch, und nicht fr ch, lehrt. Vgl. Paribh. Transi, p. 306,n.1.
144 D. h. Pat. hat dies behauptet, weil er die Paribhs nicht annimmt. Denn wenn er sie angenommen htte, wrde t(uk) berhaupt nicht eintreten, oder es wrde [nach Kaiy.'s Deutung der Paribh.] das schon eingetretene t(uk) wieder verschwinden. Dann aber htte die Bemerkung Pat.'s keinen Sinn, da [durch die Schreibung cchvoh statt chvoh] eine besondere Bemhung gemacht werden msse, damit man nicht pratsna statt prasna erhalte. Vgl. Paribh. Text p. 64, 1.12 und Transi, p. 306, n. 1 ; p. 318 f. und 318, n. 2.

D. h. dadie Paribh. 56 nicht gilt und infolgedessen /t eintreten mte. 146 Nityah sabdah ,das [beim Sprechen fr einen Begriff einzusetzende] fertige Worte', im Gegensatz zu kryah sabdah, ,dem [mit Hilfe grammatischer Regeln erst] zu bildenden W o r t ' , Vgl. Mbh. vol. I, p. 3, 1.18; p. 6, 1.12 und p. 7 (unten) f. 147 Denn das Samprasrana kann sowohl vor als auch nach der An-

145

fgung von it eintreten; it dagegen ist anitya, weil es gem ,valdeh' in VII, 2, 35 nur v o r der Substitution des Samprasrana angefgt werden kann. 148 Auch hier zerlegt man zunchst das fertige W o r t in die Bestandteile pap + vos -f- as, aus denen es hervorgegangen ist. 149 Da hier protipadavidhnt oder pratipadavidhitvt zu ergnzen ist, ergibt sich aus Kaiy. zu IV, 1, 82 [vgl. Paribh. Transi, p. 311, n. 2]: . . . Tatra samprasranam baliyah pratipadavidhnd iti tatra krte valditvbhvd in nstlti siddham papusa iti. ,lm diesem Falle besitzt [die Substitution des] Samprasrana grere Kraft, weil sie fr den Ausdruck [vos: VI, 4,131] in seiner durch den Wortlaut gegebenen Bedeutung vorgeschrieben wird. Wenn demnach [das Samprasrana] substituiert worden ist, so t r i t t it nicht mehr ein, weil [das rdhadhtuka] nunmehr nicht mit [einem Laut des pratyhra] ,,val" [hier v] beginnt. So kommt denn papusah tatschlich zustande.' Vgl. zu dem analogen Beispiel sedusah Paribh. Text p. 63, I. 10 ff. : . . . pratipadavidhitvt prvam samprasrane valditvbhvd i ta h prptireva neti. . . Dazu Transi, p. 313 f. In VI, 4,131 erscheint ,vas' in eben dieser, durch den Wortlaut gegebenen Gestalt, st also pratipadoktam, whrend es sich bei der Anwendung von VII, 2, 35 aus dem Terminus

257 Mahbhsya

IM. P r a t y a y a l a k s a n a m [ Z u p. 230, A n m . 84] Zu Kaiy.'s Bemerkung ,Pratyayalaksanam tu varnsrayatvn m bht1 vgl. Mbh. vol. I, p. 161, 1.12: Atha dvitlyam pratyayagrahanam kirn arthaml Pratyayalaksanam yath syd, varnalaksanam m b'hd iti. ,Aber welchen

Zweck hat denn [in 1,1, 62] das zweite W o r t pratyayal Da [nur] die aus dem Suffix selbst [in seiner Eigenschaft als Suffix] sich ergebende [Operation] eintreten mge, [hingegen] soll nicht [eine Operation] stattfinden, die sich aus [ihm als] einen [bloen] Buchstaben [und Bestandteil des folgenden] ergibt.' Dazu Kai/. : Pratyayalope tallaksanam ity ucyamne pratyayasya yatra krye nimittabhvah pratyayarpsrayena varnarpatsrayena v tat sarvam syt sarvanmno vastumtraparmarsitvt. Pratyayagrahane tu sati pratyayanimittam eva kryam pratyayalope bhavati, na varnarpatnimittam. Raikulam iti: avayavadvrentra pratyayasypy desam prati nimittatvam asty eva, pratyaysrayas tv ydeso na bhavatJti na pravartate. .Wenn [in der Regel] gelehrt wrde : Nach der Elision eines Suffixes [trittgleichwohl] das durch dieses (tad) Bedingte [ein]', so wrde berall dort, wo ein Suffix bei einer Operation eine Ursache bildet, jede derartige [Operation] stattfinden, ob sie nun auf [jenem in seiner Eigenschaft] als Suffix beruht, oder ob sie auf ihm nur zum Teil und] insofern beruht, als es in Form eines Lautes [und Bestandteiles der Gesamtursache] erscheint; denn das Pronomen [tad] vergegenwrtigt ausschlielich den Gegenstand [d. i. pratyaya im allgemeinen, ohne eine einschrnkende Bestimmung]. Wenn dagegen das W o r t pratyaya vorhanden ist, so [besagt dies, da] nur eine durch das Suffix [selbst] bedingte Operation nach der Elision des Suffixes eintritt, nicht aber [eine Operation], die [nur] insofern [durch das Suffix bedingt ist], als es in Gestalt eines Lautes erscheint. Raikulam [aus ryah kulam] : [Wenn in ryah kulam nach II, 4, 71 das KasussufFix as abgefallen ist, sollte in rai-kulam gem 1,1, 62 die durch das abgefallene Suffix bedingte Operation VI, 1, 78, d. i. die Substitution von y fr ai eintreten]. Das Suffix [as] ist zwar tatschlich mittels eines Teiles [d. i. des Vokales a] Ursache fr die Substitution [von y fr ai], doch beruht die Substitution von y nicht auf dem Suffix [selbst] und t r i t t deshalb nicht ein.' Vgl. Paribh. Nr. 21. IV. Vipratisedha [Zu p. 242, Anm. 104.] Die Definition von vipratisedha st in dem Vrtt. 1 zu I, 4, 2 enthalten, welches Pat. folgendermaen wiedergibt : Dvau prasagau yadnyrthau bhavata ekasmims ca yugapatprpnutah sa vipratisedhah. .Wenn zwei [Regeln] sich [sonst] zur [Bildung] verschiedener [Formen] darbieten und bei [der Bildung] einer einzigen [Form] sich gleichzeitig ergeben, so ist dies ein vipratisedha.' In einem solchen Verhltnis stehen, wie Kay. zeigt, die Regeln VI, 4,140 und 83, die verschiedenen Zwecken dienen und sich bei der Bildung von kJllapahgleichzeitig darbieten. Nach
,val', also erst aus einer Regel, erWort apavda der Paribh. Nr. 38 gibt. Die Substitution des Samprapratipadavidhi einsetzt:-Poron/tyntosrona [VI, 4,131] ist also eine rongapratipadavidhayovirodhisamnipte pratipadavidhi. Und da sich ein pratipatesm mithahprasage parabaiiyastvam. doktam frher darbietet [,sighropasthi- D. h. je zwei dieser Arten von Regeln tikatvt'], der abgeleitete Ausdruck stehen einander im Wege, wenn sie aber spter [.vilambopasthitikah' : zusammentreffen. Wenn sie sich vgl. Paribh. Text. p. 104, 1.11 f.], so gleichzeitig darbieten, besitzt jene tritt naturgem auch e\nepratipadaArt grere Kraft, welche [in obiger vidhi frher ein. Dies drckt Kaiy. zu Aufzhlung] spter genannt ist. Vgl. VII, 2, 98 dadurch aus, daerfrdas Paribh. Transi, p. 314, n. 2.

258 Bernhard Geiger

I, 4, 2 t r i t t also die para-Regel VI, 4,140 zuerst ein. Indem nun die Paribh. N r . 40: ,Sakfdgatau vipratisedhe yad bdhitam lad bdhitam eva1 bestimmt, da bei einem vipratisedha die durch die para-Rege\ aufgehobene prva-Regel als v o l l s t n d i g a u f g e h o b e n zu betrachten ist und nicht mehr angewendet werden kann, deutet sie gleichzeitig an, da dieprva-Regel hier V I , 4, 83 auch dann nicht noch e i n t r i t t , wenn in bezug auf sie die para-Regel nach VI, 4, 22 als asiddha betrachtet werden mte. Im Bhsya weist Pat. durch das W o r t bdhaka auf die Paribhs[,. . . tad bdhitam eva'] hin. Mit dem W o r t jnpayisyate aber nimmt Kaiy. nicht auf ein in einer Regel Pn.s enthaltenes Jnpaka bezug. Denn auch Pat. meint zu I, 4, 2 Vrtt. 7 nicht Pnini, sondern einen andern crya, wenn er bemerkt : Pathisyati hy crya h : sakrd gatau vipratisedhe yad bdhitam tad bdhitam eveti. [Vgl. Kiel horn, Ktyyana and Patajali p. 24f.] Diese Paribhs, welche Paribh. Transi. p.189ff. erschpfend erklrt ist, erwhnt Pat. zu VI, 3, 42; 139; VI, 4, 62; VII, 1, 26; 54. V . V i s a y a v i s a y i b h v a [ Z u p. 232, A n m . 66.] Pat. zu VI, 4,104: ,ln [der Regel ber] den Abfall [des Suffixes] hinter ein mte das W o r t ta gesetzt werden. Zu welchem Z w e c k e ? Damit [der Abfall] nicht auch hier, in akritarm, ahritarm erfolge. Cino luki tagrahannarthakyam samghtasypratyayatvt|| Vrtt. 1. In [der Regel ber] den Abfall des [Suffixes] hinter ein ist das W o r t ta unntig.Weshalb? ,Samghtasypratyayatvt l : [d. h.] warum erfolgt nicht der Abfall eines Komplexes [von Suffixen, d. i. -ta + tara + am]? W e i l [dies] nicht ,ein Suffix' ist. Es w i r d ja [1,1, 61] gelehrt: ,[Das Verschwinden] eines Suffixes heit Ink, slu oder up1, und ein Komplex [von Suffixen] ist nicht ,ein Suffix'. E i n w u r f W e n n dem so st, so bietet sich doch, nachdem die Elision des [ersten Suffixes] ta vollzogen ist, die das folgenden [Suffixes taro] dar? ErwiderungTalopasyacsiddhatvt|| Vrtt. 2. Die Elision des ta w i r d als asiddha betrachtet, und weil sie asiddha ist, w i r d [auch die des folgenden Suffixes tara] nicht erfolgen. Hiezu bemerkt Kaiy. : Tagrahanam iti : idam asminn asiddham iti bhedanibandhanatvd visayavisayibhvasya cino lug ity asya laksanasya bhedbhvd akritarm ity atra opasysiddhatvbhvt tarapo 'pi luk prpnoti ; evam sati pratyayatraytmakasya samudyasya lukprasagah. Itaras tu yugapat pratyayatrayasya lukprasago 'nenokta iti matvha cino lukti. . . Pratilaksyam laksanabhedd asti visayavisayibhvah. ,Weil ein Objekt [einer Operation in einer und derselben Regel] selbst auch ein Objekt [in bezug auf welches es asiddha ist] nur unter der Bedingung hat, da die Teilung vorgenommen wird M [und] dieses gilt als asiddha in bezug auf jenes", und weil eine [derartige] Teilung der Regel ,cino luk1 nicht vorhanden ist, so ergibt sich bei akritarm, dadurch da die Elision [des ta] nicht asiddha [in bezug auf den Abfall von tara] ist, auch der Abfall [des Suffixes] tora. [Ta wre nmlich sonst visaya der ersten Operation (luk) und gleichzeitig, als asiddha, visayin in bezug auf den Abfall von tara]. Unter diesen Umstnden wrde sich der Abfall des aus drei Suffixen bestehenden Komplexes [ta + tara + am] darbieten. Der Opponent aber glaubt, jener habe behauptet, da der Abfall der drei Suffixe sich gleichzeitig darbiete, und sagt deshalb [im Vrtt. 2] ,cino luki1 [u.s.w.]. . .; [zu Vrtt. 2:] Man mu dagegen [d. i. gegenber der Behauptung, da ta in der Regel notwendig sei] daraufhinweisen, da infolge [der tatsch-

259 Mahbhsya

lichen Annahme jener] Teilung dieser Regel das Objekt [ta] tatschlich [innerhalb dieser Regel] selbst ein Objekt [d. i. den Abfall von tara] hat [, in bezug aufweichen sein eigener Abfall als asiddha betrachtet w i r d ] . ' Noch deutlicher setzt dies Pat. zu 1,1, 57 Vrtt. 6 auseinander. D o r t w i r d nmlich zunchst behauptet, da zur Erklrung von vyvoh usw. wegen der Regel V I , 1 , 66 die Regel V I , 1, 11 i n t e r p r e t i e r t werden msse: Fr /', u, r, I und ihre Lngen werden vor einem Vokal die entsprechenden Halbvokale substituiert, [und diese Regel ist als asiddha zu betrachten, wenn sich die Elision der Halbvokale (y oder v) darbietet]. Dagegen w i r d eingewendet, da nur eine Regel in bezug auf eine a n d e r e Regel asiddha sein knne (anyad anyasmin), whrend hier die Regel VI, 1 , 77 in bezug auf sich selbst asiddha wre. Darauf w i r d e r w i d e r t : Tad eva cpi tasminn asiddham bhavati. Vaksyati hy cryah: cino luki tagrahannarthakyam samghtasypratyayatvt talopasya csiddhatvd iti.Cino luk cino luky evsiddho bhavati. ,Es ist doch auch eine bestimmte [Regel] in bezug auf sie selbst [ohne da dies in der Regel ausdrcklich gesagt wird] asiddha. Denn der Lehrer [Kty.] wird [VI, 4,104 Vrtt. 1] sagen ,Cinoluki' usw. [d. h.]: Der A b f a l l [des Suffixes] nach ein ist in bezug a u f d e n A b f a l l [eines anderen Suffixes] nach ein asiddha. Kaiy. zu 1,1, 57 nennt die wrtliche Auffassung der Regel ,sakrtptha' [.einmalige Lesung']. Visayabhedt tu bhedsrayand asiddhatvam snyate. ,Infolge der Teilung des Objektes [der Regel in ein Objekt und Subjekt] aber gelangt man auf Grund der Annahme der Teilung dazu, da [das Objekt] asiddha ist.' Vgl. Ks. zu VI, 4,104: Akritarml ahritamm ity atra talopasysiddhatvt taraptamapor na lug bhavati, cino lug ity etadvisayabhedd bhidyate.

VI. Samanasraya [Zu p-.231, Anm. 64 (Schlu).]

1. samnsraya : 2. vysraya: vah mah

kur[+u] +

1. Die Elision des u [VI, 4,108 und 109] beruht auf dem [ganzen] mit v, bezw. m, anlautenden Srvadhtuka-SufT\x vah, bezw. mah. Auf dem gleichen Element beruht die zweite Operation [VI, 4,110]. Die Elision des u ist demnach samnsraya. 2. Nach dieser Auffassung beruht die Elision des u nicht auf dem ganzen Suffix vah, bezw. mah, sondern nur auf ihren Anfangsbuchstaben v, bezw. m, welche bei der zweiten Operation [VI, 4, 110] nicht in Betracht kommen ; denn diese beruht auf dem ganzen Suffix vah, bezw. mah. Die Elision des u ist in diesem Falle vysraya, also nicht asiddha. V I I . Sthnivadbhlva [Zu p. 231, Anm. 64 (Mitte und gegen Schlu).] Zu den Bemerkungen Kaiy.'s: ,Tatas cndistdaeahprvasya. . .' und nachher: ,tatas cdistdacahprvo 'kra iti tadvidhau nsti sthnivadbhvah' vgl. Pat. zu 1,1, 57 Vrtt. 1 : Yo 'ndistd aeah prvas tasya vidhim prati sthnivadbhva distc caiso 'cah prvah. ,ln bezug auf die Operation an einem [Element], welches dem Vokal vorangeht, solange er noch n i c h t d u r c h das S u b s t i t u t ersetzt ist, verhlt sich [das Substitut] wie der ursprngliche [Vokal] ; dieses aber geht [erst] dem durch das Substitut ersetzten Vokal voran.'

The Modern Period

Plate V

A page from a manuscript in the India Office, London, of Patajali's Mahbhsya with Kaiyata's subommentary hasyaprodipa, reproduced by cou rtesy of the Secretary of State for Foreign and Commonwealth Affairs. This manuscript was written in A.D. 1787-1788 (see No. 4981, Catalogue Keith 1935). It is MS. B of Kielhorn's edition. The four lines in the middle of the page constitute the following passage from the beginning of the Introduction of the Mahbhsya :
sabdah yenoccritena ssnlrpg lakakudakhuravisninrn sarnpratyayo bhavati sa sabdah athav prattapadrthako loke dhvanih sabda ity ucyate tadyath sabdarn kuru masabdarp krsh sabdakryayarp mnavaka ti dhvanim kurvann evam ucyate tasmt dhvanih sabdah kni punar asyasabdnussanasya prayojanni . raksohgamalaghvasarndehh prayojanarp raksrtharn vednm adhyeyarp vykaranarn lopgamavarnavikrajno hi samyag vedn pari playisyatti hah

" (What then is) the word ' cow ' ? That by means of which, when uttered, there

arises the notion of creatures with dewlap, tail, hump, hooves, and horns. Or else the sound, which in everyday life conveys a particular meaning, is called a word. One refers to sounds when saying " produce a word," "do not produce a word," "the boy produces words." Therefore a word is sound. Now what are the fruits of this teaching of words? Preservation, modification, tradition, abbreviation, and removal of doubt constitute its fruit. We should study grammar for the preservation of the Vedas. For he who knows the linguistic zero, the verbal augment, and the ways in which sound is modified will be in a position to properly protect the Vedas. Modification (. . .)." The beginning of this passage with its inquiry into sabda has given rise to a great many interpretations and controversies (see, for example, Brough's interpretation, page405. The five lines at the top of the page continuing with the five lines at the bottom constitute Kaiyata's subcommentary on this passage of the
Mahabhsya.

17
Leonard Bloomfield
(1887-1949)

Under the heading "The Modern Period," a final selection of papers will now be presented which together illustrate a remarkable diversity of approaches. The " modern period " includes roughly the years 1925-1960. At the beginning of this period there are two publications which can hardly be regarded as very important contributions themselves, but which contrast in significant and instructive ways when considered together. Leonard Bloomfield (1887-1949), the well-known linguist who is generally regarded as the founder of American structural linguistics, was the nephew of the Sanskritist, Maurice Bloomfield. Leonard studied Indo-European and especially Germanic at Harvard, Wisconsin, Leipzig, and Gttingen. He studied Sanskrit with Herman Oldenberg (mainly in Indo-European perspective) and Karl Brugmann. That Leonard Bloomfield held the Sanskrit grammarians in high regard is clear from many passages in his works. In Language (1933), for example, he called Pnini's grammar "one of the greatest monuments of human intelligence" and substantiated this as follows: " It describes, with the minutest detail, every inflection, derivation, and composition, and every syntactic usage of its author's speech. No other language, to this day, has been so perfectly described" (page 11). In Linguistic Aspects of Science (1939), he characterized the contribution of the Sanskrit grammarians to linguistics in the following sentences: "Around the beginning of the nineteenth century the Sanskrit grammar of the ancient Hindus became known to European scholars. Hindu grammar described the Sanskrit language completely and in scientific terms, without prepossessions or philosophical intrusions. It was from this model that Western scholars learned, in the course of a few decades, to describe a language in terms of its own structure" (page 2). It is interesting to note that Bloomfield was attracted to the Indian grammarians in the first place because of their alleged positivism. Moreover, the Linguistic Aspects of Science, from which these lines are quoted, forms part of the International Encyclopedia of Unified Science, the neopositivist series which derived great inspiration from the ViennaCircle, and in which such scholars as Rudolf Carnap played an important part. In turn, construing praise by Bloomfield as an indisputable symptom of taxonomic leanings, later linguists have erroneously regarded Panini as the embodiment of the taxonomic approach (e.g., Lees 19632, xix). Bloomfield made a more technical contribution to the study of Pnini's grammar in a paper published in 1927, when he was Professor of German and Linguistics at Ohio State University. The discussion mainly deals with an apparently "unwonted repetition " in the stra and in the accompanying gana list. Bloomfield's "whimsical humor" (Bernard Bloch in Sebeok1966,11,517) is apparent in a excursus on undergarments. What is generally striking in this paper is his purely technical and atheoretical approach. What is perhaps more specifically characteristic of Bloomfield is his empiricism, which becomes apparent at the end of section IV of the paper. There Bloomfield assumes, "as would any modern linguist," that it isonly natural that Pnini "did not concern himself with such far-fetched theoretical possibilities as someone's taking it into his head to say svalid dhaukate" (the expression means 'to come near licking like a dog'). But the Sanskrit grammarians were always looking out for possible examples (udharana) and counterexamples (pratyudharana), for precisely such cases.

265 Leonard Bloomfield

Unlike Pnini, Bloomfield does not seem t o have t h o u g h t of these predictive implications of t h e notion of grammatical rule. Bloomfield's paper " O n some rules of Pnini " was published in the Journal of the American Oriental Society (47', 1927, 61-70).

On Some Rules of Panini(l927) Leonard Bloomfield

In the first chapter of the first book of his Grammar, Panini sets up the term pronoun (sarvanman) as a designation of certain words. He does this so that in other passages he may, without repeating the list of these words, describe their inflectional and other peculiarities (e. g.f dative singular masculine and neuter, pronominal sarvasmai, as opposed to nominal pade, dvoya, 7,1,14). These passages are of course listed in the Indices of Boehtlingk's second edition (Leipzig, 1887)1. The rules defining the term pronoun read as follows (1,1): 27. sarva etc. are pronouns, 27. sarvadini sarvanamni, 28. option in direction-com28. vibhs diksamse bahuvrpound exocentric, hau, 29. na bahuvrhau, 29. not in exocentric, 30. trtly sa mse, 30. instrumental-compound, 31 . dvamdve ca, 31. and copulative, 32. option before nom. pi. -as, 32. vibhs jasi, 33. and prathama, carama, -taya, 33. prathamacaramataylpralpa, ardha, katipaya, nema. dhakatipayanems ca, 34. prva, para, avara, daksina, 34. prvaparvaradaksinottauttara, apara, adhara in rpardharni vyavasthym spatial relation not name, asamjnym, 35. sva not in kinsman or prop35. svam ajntidhankhyym, erty appellative, 36. antararh bahiryogopasarh36. antara in conjunction with vynayoh. outside and undergarment. The list (gana) Sarva Etc. that goes with Rule 27 reads with Boehtlingk's numbering, as follows: 1. sarva; 2. visva; 3. ubha; 4. ubhaya. 5. d-atara; 6. d-atama; 7. tara; 8. anya; 9. anyatara. 10. tv; 11. tva; 12. nema; 13. sama; 14. sima; 15. prva-par-'varadaksino-1 ttar- par- dharni vyavasthym asamjnym; 16. svam ajntidhankhyym ; 17. antararh bahiryogopasamvynayoh. 18. tyad; 19. tad; 20. yad; 21. etad; 22. adas; 23. dam; 24..elka. 25. dvi; 26. yusmad; 27. asmad; 28. bhavat-u; 29. kirn. Bhattojidksita, SiddhntakaumudJ(ed. Vasu, Allahabad, n. d.) places 7 after 9, 22 after 23, and reads tvatfor 10. This last variant is mentioned by Jayditya, Ksik (my copy is a reprint of the Benares 1876-8 edition), which otherwise agrees with Boehtlingk's reading. The traditional interpretation of these rules, as given, e.g., in the two books just named, and, except for one detail, accepted by Boehtlingk, is as follows: (a) Whatever word is designated as a pronoun will nevertheless not be a pronoun [i.e., lack pronominal characteristics) in an exocentric compound, in an instrumental determinative compound, and in a copulative compound ; except that it may (optionally) have pronominal characteristics in an exocentric compound of points of the compass (e. g. northeast) and in the nominative plural masculine (in the other genders the nominal and pronominal inflections here coincide) of a copulative compound.
1 Apart from Pnini himself and his commentators, my obligation is to Boehtlingk's indices and to the writings of B. Liebich : Pnini (Leipzig, 1891 ) ; Zwei Kapitel der Ksik (Breslau,

1892); Candra Vrtti (AKM 14, Leipzig, 1918) ; Zur Einfhrung in die indische Sprachwissenschaft. I-IV (Heidelberg
Sr 1919 ff.).

267 On Some Rules of Pnini

(p) Designated as pronouns are the words in the List Sarva Etc. ; numbers 15,16,17 w i t h the limitations there stated. To this constituency is t o be added the fact that the first six words in Rule 33 (which are not otherwise treated as pronouns) optionally have pronoun character in the nominative plural masculine, e. g. prathame beside pratlnamh, And f r o m this constituency is t o be subtracted the fact that (12) nema and the words under 15,16,17 (in the senses there stated ; in other senses they are not pronouns) optionally lack pronoun character in the nominative plural masculine, e. g. nemh beside neme, prvh beside pn/e. In the case of (17) antara the restriction means that this w o r d is a pronoun when it is a synonym of outside and in the sense of

undergarment. II What first strikes one in this passage is the repetitious and clumsy treatment of the words in 15,16,17 of the List. They are there cited not in stem form, but as inflected neuters, and restrictions of meaning are added, contrary to the usual form of the Lists. The option for nom. pi. mase, is stated by repeating in Rules 34, 35, 36 not only the entire section of the List, but also the restrictions of meaning,in contrast with Pnini's usual elegant brevity. No wonder that Ktyyana's one comment on Rule 34 is, "And needlessness of repetitive citation in the Text of Rules of avara etc., owing to reading in the Text of Lists." (I cite from S. D. Kudla's edition of Patajali's Mahbhsya, vol. 1, Nirnaya Sagar Press, Bombay 1917, which contains also Kaiyata's PradJpa, Ngesa's Uddyota, and selections from a Chaya, commentary). Patajali tries to discover what Pnini meant to indicate by this unwonted repetition. Did he mean to set up the restrictions under which these words have pronoun character? No, for that has been attained by the List. Did he mean to show that these words are not to be included when he tells us (5, 3, 2) that (25) dvi etc. are in certain cases excepted from pronominal treatment? (We interrupt to ask why anyone could be tempted to include 15,16,17 in a statement about "25 and the fol lowing"; of this more, immediately). No, says Patajali, for Pnini permits himself exceptions in matters of sequence. Pnini's real intent, he concludes, was to show us, by a repetition which otherwise would be purposeless, that the words in 15,16,17 and Rules 34, 35, 36 belong with Rules 32 and 33, /'. e., have option of nominal form in the nom. pi. mase,e. g.f prvh beside prve. The second of the suppositions which Patajali rejects would be intelligible only if in the List 15,16,17 came after (25) dvi, etc. Only if that were the order could anyone suppose, even for a moment, that they were included in the " dvi etc." of Rule 5, 3, 2, and only then could one suppose, be it mistakenly, that Pnini's repetition was meant to indicate that the words in 15,16,17 constitute a special group and are not to be involved when an exception is stated for (25) dvi etc. And, indeed, Kaiyata tells us that by some scholars (13) tyad and the following (including therefore (25) dvi and its successors) are read before (15) prva and the following. This shows us that since Patajali's time the order of the List has been changed. Patajali's reason for rejecting this second possible motive is of interest. The reason is that when Pnini at 5, 3, 2 makes a statement about " dvi etc." this need not include the subsequent prva

268 Leonard Bloomfield

and its followers, for Panini permits himself exceptions in matters of sequence. The Master, sa/s Patajali, has given us a formal indication of this by making an obvious and otherwise inexplicable except i o n t o the sequence announced in 8, 2 , 1 . This rule says that from there t o the end of the Grammar each rule is t o be taken as uneffected w i t h regard t o preceding rules. Yet, at 8,3,13 he gives a rule which can apply only if the subsequent rule 8, 4, 41 be taken as already effected, and apply w i t h i n just limits only if 8, 4, 53 be taken as already effected. He points out that Ktyyana recognizes this in his comment on 8, 3,13. The motive which Patajali finally attributes t o Panini seems inadequate. If Pnini meant t o tell us that the words in 15,16,17 allowed of nominal nom. pi. mase, he could have added the w o r d "prva-etc." t o Rule 33. He would not even have had t o specify the extent of "prva-etc", since these words stood at the end of the List. Ktyyana seems t o have the better of the argument : his objection is valid. It is n o t e w o r t h y that Ktyyana cites our words not as "prva etc.", but as "avara etc." Kaiyata says that he (or rather, Patajali in paraphrasing him) does this prakrrtham " f o r the sake of the special meaning" (/. e., t o symbolize that Rules 34, 35, 36 are avara " p o s t e r i o r " t o the reading of the same words in the List, which is prva " preceding"). Ngesa, however, says that in reality Ktyyana says "avara etc." because avara ought properly t o precede, since its syllables are short (Ktyyana on Pnini 2, 2, 34) and because it begins in a vowel and ends in short a (Pnini 2, 2, 23 : the edition reads ajyantatvt, for ajdyadantatvt). Patajali touches upon the peculiar form of Ktyyana's citat i o n . He opens his discussion of Ktyyana's critique by asking: H o w do we know that the T e x t of Lists comes first and that therefore the reading of 34 in the Text of Rules is a repetition ? By the symbolism that in the f o r m e r we have "prva etc.", that is, " preceding etc." and in the latter "avara etc.", that is, " f o l l o w i n g etc." ? N o , for in the latter, t o o , we have "prva etc." Yet it is t r u e that the Lists come first, the Rules second. The Master indicates this t o us by his later reference (7,1,16) t o " t h e nine beginning w i t h prva."That is, the fact that Pnini says " n i n e " decides for Patajali that the List is meant, since only there does prva head a group of nine words, and since the mention o f " n i n e " could have no other motive than t o refer us explicitly t o the List: what w i t h the final position of prva etc. in the List, the w o r d " n i n e " w o u l d otherwise be superfluous. The fact that Pnini's reference is t o the List and not t o the Rules apparently decides for Patajali that the List is t o be read first. This shows that Patajali, at any rate, read prva and not avara as the first w o r d in Rule 34. W e shall find reason t o believe that this was Pnini's own reading. Accordingly, Ktyyana's "avara etc." remains unintelligible except as a laconic indication of a flaw, namely that Pnini's w o r d i n g disagrees w i t h the rules of order in copulative compounds (as above cited from Ngesa). From Pnini's general practice it appears that he did not begin his rule w i t h avara (or w i t h one of the other words w i t h initial vowel) because t o do so would have added a syllable (though not a mora) t o the length of his rule; he arranges his words so as t o merge or elide as many vowels as possible. Beyond this, I cannot account for the order in which Pnini cites these stems.

269 On Some Rules of Pnini

III Beside the repetition, the strange form of numbers 15,16,17 n the List, their historic change of place, and the later reference with " nine," our rules contain several peculiarities which the commentators do not mention. Why does Rule 33 cite its words in the masculine, when the normal form of citation is neuter (Rules 34, 35, 36)? Why does the word ca " a n d " stand in Rule 33?The commentators ignore these points, probably because their discussion ofPninean methods had worked out the principles that the gender in which Pnini cites words is not necessarily significant, and that the placing of ca " a n d " does not necessarily indicate the end of an enumeration. At least, these decisions are given by NagojT (Ngesa) in his Paribhsendusekhara, Numbers 73 and 78 (Kielhorn's translation, Bombay Sanskrit Series, No. 2,1868-74). Finally,a point which lay outside the view of the ancients, in the light of historical linguistics the doctrine of the rules as they are traditionally interpreted is most surprising. In the first six words of Rule 33 we see the lapse, widespread in Indo-European languages, of nominal words into pronominal inflection, especially in the nom. pi. mase. In Latin and Greek this has, of course, involved all the o-stems (and gone on even to the j-fem nines); in Germanic it has involved the whole strong adjective declension. The pronominal form of the nom. pi. mase, is favored also in the copulative compounds, Rule 32. But in the case of nema in Rule 33 and of the words in Rules 34, 35, 36 we are taught the opposite: nema has pronominal forms throughout, but in the nom. pi. mase, may have nominal form, nemh beside neme; and the other words have pronominal forms throughout (except for an option in the ablative and locative singular, masculine and neuter, 7,1,16), but in the nom. pi. mase, also nominal form, e. g., prvh besideprve.

IV Pnini and Ktyyana recorded the facts of a standard colloquial language spoken by them as their mother-tongue and as the medium of everyday life; if, like ours, it was archaic in comparison with the normal dialects of the time, it was, like ours, native to many speakers, who with a little training were able to decide what could and what could not be spoken. The later grammarians, including, I venture to say, Patajali, were in no position to criticize the facts thus given by their predecessors ; for them Sanskrit was a second language, spoken to be sure, but preserved and by them acquired through a literary tradition, much as was classical Latin for a learned Italian of, say, the fifth century A.D. They could judge only of the form of Pnini's presentation. What is more, the text of Pnini was canonical. If Patajali found a repetition in our passage, it was his task to divine what the Master could have meant by this unusual proceeding. If he found a discrepancy in the order of the rules in the Eighth Book, he could at best take it as an intentional and formal indication on the part of the Master, to the effect that exceptions of order occur in the Text. He could not like Boehtlingk, see in the order of 8, 3,13 and 8, 4, 41 one of the slips inevitable in a human construction of such size and complexity. He could not, as would any modern linguist, conclude from the order of 8, 3,13 and 8, 4, 53 that Pnini did not concern himself with such far-fetched theoretical possibilities as someone's taking it

270 Leonard Bloomfield

into his head t o say svaliddhaukate. For later students, Panini, Ktyyana, and Patajali were the Three Seers, and especially the ultimate interpretation of Patajali was canonically binding. If Ngoj found a single passage where, under Patajali's interpretat i o n , the gender in which a w o r d is cited is indifferently chosen, or a single passage where, under Patajal i's i n t e r p r e t a t i o n , the w o r d " a n d " has no particular bearing, then he was obliged t o conclude that Panini could not have intended these features t o be generally significant.

V Probably the peculiarities of our passage are t o be explained as follows. The citation of the words in Rule 33 is in masculine f o r m because it is only the masculine nominative plural that is involved in the rule. This helps t o set off these words f r o m the following, and incidentally saves a short vowel. The w o r d ca " a n d " in Rule 33 has the same value as in Rule 31 and in many other passages of the Grammar: it is used in the last of a series of ( t w o or more) rules that are additively coordinated. This habit, taken f r o m ordinary speech, serves in the Grammar t o show where such an enumerative series ends. Thus, in Rule 31 " a n d in a copulative," the w o r d " a n d " shows that " in a copulat i v e " is the last of the t h r e e places where sarva etc. are not pronouns. Technically, we may phrase this by saying that the w o r d " a n d " shows the continued validity (anuvrtti) up t o this point, and the cessation of validity (nivrtti) at this point of the w o r d na " n o t " in Rule 29. Similarly in Rule 33: the w o r d " a n d " concludes the series of rules ( t w o in number) which tell where o p t i o n in the nom. pi. mase, is given. In technical language: the w o r d " a n d " in Rule 33 marks the continued validity in this rule of the words " o p t i o n before nom.-pi. - a s " of Rule 32 and the cessation of this validity at the end of Rule 33. This indication is reinforced by quoting the words in masculine f o r m . Thus the subsequent rules, 34, 35, 36, have nothing t o do w i t h option in the nom. pi. The words in these subsequent rules w e r e not intended by Pnini t o stand in the List. These t h r e e rules simply state that these words, w i t h these restrictions of sense, are pronouns. These words are treated in the Text and not in the List because they require restriction : they are pronouns only in the meanings here given. A f t e r Pnini's t i m e and before Ktyyana's, some t i n k e r e r added them bodily, w i t h o u t reducing them t o stem f o r m , and w i t h the statements of restriction, t o the end of the List Sarva Etc. There Ktyyana and Patajali, and, according t o Kaiyata, some later students found t h e m . This c o r r u p t i o n of the List involved not only the repetition which Ktytyana criticized and Patajali t r i e d t o explain, but also another discrepancy: Rule 5, 3, 2 excludes " d v i etc." f r o m a characteristic of pronouns which almost any page of Sanskrit shows t o hold good for prva and its group (adverbial forms like prvatra). Patajali explains this, as we have seen, as an exception t o order. A later arranger of the List solved this difficulty by moving these final numbers of the List back t o a place before 18 tyad; here they no longer interfered w i t h "dvi e t c . " of Rule 5, 3, 2 or w i t h rules about "tyad etc." Pnini's reference at 7 , 1 , 1 6 t o " t h e nine beginning w i t h prva" would have lacked the mention o f " n i n e " and w o u l d have read simply "prva etc.", had he intended these words t o stand, as

271 On Some Rules of Pnini

Patajali found them, at the end of the List Sarva Etc. We have seen how Patajali tries to explain the mention of the number "nine." In reality, 7,1,16 refers to Pnini's Text at 1,1, 34, 35, 36, and the number " nine" is necessary to secure inclusion of all the words. This reference shows, incidentally, that for Pnini prvo, and not avara, actually stood first in Rule 34. The word nema in Rule 33 repeats number 12 of the List. The traditional interpretation, then, has it that the first six words in Rule 33, not being in the List, are normally nouns, but are in the nom. pi. mase, capable of pronominal inflection ; but that nema, being in the List, is normally a pronoun, but is in the nom. pi. mase, capable of nominal inflection. Not to repeat the other arguments, one may surmise that if Pnini had meant this, he would have put nema into a separate rule. Probably nema did not stand in Pnini's List; he treated it in Rule 33 on a par with the six other words there cited : in Pnini's language nema had nominal character except for the option of a pronominally formed nom. pi. mase. In Sanskrit literature the word nema is little used, certainly not enough to carry on a (literary) tradition of its Pninean inflection. Later grammarians were dependent for information on this point upon their interpretation of Pnini's statements and upon whatever examples they could find in old books. Probably some scholar who knew pronominal forms of the word in Vedic texts (perhaps nemasmin RV10, 48,10) inserted it in the List. Boehtlingk in his edition of Pnini and more fully in the Petersburg Dictionary, s. v. antara, suggests the correct interpretation of bahiryoga- in Rule 36: " in conjunction with 'outside' " means " in contrast with ' outside.' " The word yoga is used in the same way at 4 , 1 , 48; everywhere else in Pnini it means " in (actual) context with " so that the only possible alternative meaning of bahiryoga- would be " in connection with the word bahih, 'outside.' " There is no parallel for Kaiyata's interpretation, "as a synonym of'outside.' " which was taken over by Bhattoji and Jayditya. This faulty interpretation does not go back to Ktyyana and Patajali. The former says : "The mention o f undergarment' is purposeless, because this effect is attained by bahiryoga.No, for, rather, it has the purpose of applying to a pair of petticoats and the like." Patajali expands these remarks, rejecting the latter. The Chy points out that this discussion was necessitated by the absence of the word khy "appellative" in the rule, which otherwise could have been interpreted like Rule 35, giving " in an appellative for undergarment." Now, Ktyyana's second sentence is obscure, and Patajali's interpretation of it seems far-fetched,2 but so much is clear, that these scholars did not know antara in a sense of "outer," but interpreted bahiryoga as " inner."
2

Ktyyana's remarks may perhaps be interpreted thus: In his and Panini's speech one could say "an inner" for "an undergarment." K. questions whether this need be specially stated, since this usage would seem to be included under the definition "in contrast with outside" (Vartikal), but decides that, after all, it needs to be stated, since "an inner" refers to either of two petticoats or the like, worn one over the other, while "in

contrast with outside " might be taken as appropriate only to the inside one (Vrtika2). This, however, is not Patajali's interpretation. He paraphrases the two Vrtika's and explains and refutes the second by saying: Where this is not known: which is the inside one (ontariyom) and which the outside one (uttanyam), here, too, that person who has first made examination, for him it is decided: this is the innerand

272 Leonard Bloomfield

VI
In our view, then, the List Sarva Etc. should be read without (12) nema, (15) prva etc., (16) sva, (17) antara; and the Rules would be interpreted thus: (a) As interpreted by the ancients. (b) Designated as pronouns are the words in the List. Also the words prva, para, avara, daksina, uttara, apara, adhara in the sense of a spatial or temporal relation when not specialized appellatives; the word sva when not an appellative for " kinsman " or " property" ; the word antara when meaning " inner" or " undergarment." (Later, at 7,1,16, we learn that these nine words have optionally nominal inflection in certain forms; the nom. pi. mase, however, in the pronominal senses, is always prve, etc., never prvh). The words prathama, carama, -taya (/. e., v/ords formed with this suffix), alpa, ardha, katipaya, nema optionally have the pronominal form in the nom. pi. mase, but otherwise have only nominal forms.
this the outer. (Kaiyata's gloss: In the case of a pair of petticoats of equal size and not donned, it is not known whether a given one is the outer or inner." Here too "; as soon as the identity of the undergarment is determined by forethought, then, too, it s a case of " conjunction with outside"). Thus it appears that Patajali's interpretation of Ktyyana's second Vrtika is: An undergarment is still an undergarment even when not recognized as belonging inside. His refutation is that, as soon as one recognizes it as an undergarment, one also recognizes that it belongs inside. He has underestimated Ktyyana's point, probably because the expression "an inner" for an undergarment was unfamiliar to him. A different interpretation is quoted by Kaiyata: But others say: The word "and so forth" is included in the Vrtika 2: this means that in the case of a set of three or four petticoats, since the third and fourth are not in "conjunction with outside," the word " undergarment" deserves separate mention.

8
Barend Faddegon (1874-195 S)

Barend Faddegon, Professor of Sanskrit at the University of Amsterdam, was the author of a few papers in linguistics and of numerous studies in the field of Sanskrit (mainly concerning Indian philosophy, the Smaveday and Pnini). Though hardly known as a linguist, he had sound judgment in linguistic matters. Faddegon was original both as a scholar and as a man and wrote In a very personal style. The more temperate of his contemporaries reacted to his publications with mixed feelings. Perhaps Renou's judgment on Faddegon's Studies on Pnini's Grammar (1936) may be taken as representative and just: "ouvrage un peu fantasque, mais sduisant" (1940,10, note 1). Earlier Renou had commented on Faddegon's Studies in the following terms: " i l a du got pour les formules algbriques et verse lui-mme aisment dans ces enchanements par association qu'il a relevs chez l'auteur qu'il dcrit" (Orientalistische Literaturzeitung 40, no. 5,1937, 319). A good specimen of Faddegon's style and outlook is the motto of his Studies on Pnini's Grammar: Goldstcker has admirably attacked Bhtlingk, but for Bhtlingk we forget Goldstcker; and Whitney had admirably attacked Pnini, but for Panini we forget Whitney. I adore Bhtlingk because he reveals to us the spirit of Panini, I adore Pnini because he reveals to us the spirit of India, I adore India because it reveals to us the Spirit, the Spirit. A more specific and very insightful comment on Pnini's syntactic-semantic theory of kraka is also characteristic: Evidently Pnini tries in this analysis to separate the ideational aspect from the linguistic expression, an attempt which the Occidental linguists of the latter half of the nineteenth century have condemned, misled as they were by the hope of being able to understand language through the exclusive study of its phonal and morphological aspect, i.e. its articulative utterance and the association-system underlying declension and conjugation, as if the application and imitation of physics and a mechanistic psychology were the last word of moral science. And so besides the injustice done to a pioneer of grammar who lived about twenty-five centuries ago by associating with him supposed results of modern grammar, it is even questionable whether Pnini has not something still to say to us [from page 18 of the same book]. Faddegon had one outstanding pupil in the field of Pninian studies: Herman Eldert Buiskool (1884-1963), at first a schoolteacher and later a civil servant who was mostly engaged in committee work relating to the frequent proposals for spelling reform in Dutch. Buiskool wrote one book, his dissertation for Faddegon, entitled Prvatrsiddham. Analytisch onderzoek aangaande het systeem der Tripdvan Pnini's Astdhyyl(1934), republished in 1939 as The TripdJ, being an English Recast of Prvatrsiddham (An Analytical-Synthetical Inquiry into the System of the Last Three Chapters of Pnini's Astdhyyl). In this work Buiskool gave a penetrating analysis of some of the fundamental techniques of Pnini's grammar, largely relating to the relative strength of rules and to the concept of asiddhatva (see page 207). Probably the earliest of Faddegon's contributions to the study of the Sanskrit grammarians was a paper read at the First International Congress of Linguists on April 13,1928: " T h e Mnemotechnics of Pnini's Grammar I: The Siva-Stra," Acta Orientalia 7, 1929, 48-65. It is a remarkable, although perhaps unintentional, indication of each author's evaluation of the Sanskrit grammarians,

274 Barend Faddegon

that the Sanskritist Faddegon's paper was read at a linguistic congress, whereas the linguist Bloomfield's paper was published in an orientalist j o u r n a l . There are inaccuracies in this paper. For example, it is argued in 13 that v may be dispensed w i t h , but that this is not so follows f r o m a rule mentioned earlier: W i t h o u t v it is impossible t o state 8.3.7, which amounts t o n - > nTsAR/ # chAV~aM.

Hence we have bhavms cinoti, bhavms tarati, but bhavn karoti. The Sivastra has been the subject of a number of studies by modern authors. Apartfrom Skld (1926), mentioned by Faddegon, there are studies by Breloer (1929), K. C. Chatterji (1934a), Thieme (1935a: beginning of Chapter 4), C. K. Raja (1957), Staal (1962a), Misra (1966, section 2.1), Cardona (1969; cf. Staal 1970b). Skld had defended the view that the Sivastra was not the work of Pnini, but it has become increasingly clear that this list of sounds is very closely interrelated with many of Pnini's grammatical rules. The traditional view that the Sivastra was revealed to Pnini by Siva is related to some of the legends connected with Patajali (see this volume page 3).

The Mnemotechnics of Pnini's Grammar I: The SivaStra (1929)

Barend Faddegon

The subject I have chosen for my paper does not seem to possess any actuality.1 Even Prof. Liebich, who has given so much time to the interpretation of Pnini, acknowledges the study of this grammar to be exclusively the work of the specialist.2 It is many years ago since the study of Pnini was used in Europe as a grammatical guide for Sanskrit. At the present time one need only understand the technical expressions borrowed from Pnini by the commentators of Indian literature; Pnini's grammar itself has lost its general interest. And yet Bhtlingk praised Pnini as a genius and his grammar as a masterpiece, thus claiming for it a lasting value. But before giving up the modest attitude of Liebich, we must consider whether Bhtlingk is right, for in his laudation he is more assertative than argumentative. To prove this I shall now read out to you a quotation from Bhtlingk's preface bearing on this matter. "The order of the stras may here and there seem strange to us, but it is planned and carried out in a consistent and admirable manner. A perfect conciseness and complete avoidance of all repetitions has been the aim of the author and without any doubt he has attained this. The more thoroughly one studies Pnini's grammar, the more one is struck by the acuteness and the successful mastery of the vast matter, shown in it. It is in its kind a masterpiece of the first rank. The more recent grammars in which the unchangeable order of Pnini's stras has been tampered with in order to string together everything bearing on the same matter, cannot be understood without the aid of extensive commentaries, which of necessity continually refer to passages either long past by or not yet arrived at; indeed these grammars must be considered as failures." So runs the quotation from Bhtlingk. The reader, however, who reads Pnini for the first time, will be painfully struck by the fact that the text of the first chapters totally contradicts the assertion of Bhtlingk. Numerous references to later passages are also needed here and are indeed given by Bhtlingk himself in his translation. And even these references are not sufficient in number. Already at the second chapter of the first book the reader is overwhelmed by its grammatical intricacies; and when he has finished the eight books, he only possesses a chaotic impression of details. The help afforded by Bhtlingk is insufficient in order to determine the merits of Pnini exactly; we want more insight into the total plan, the execution of details, and the idea underlying the pregnant form of expression. Here I shall restrict myself t o the last-mentioned point, Pnini's mnemotechnical system, and again specially to the mnemotechnics of the introductory Siva-Stra. By the Siva-Stra is understood a collection of 14formulae of phonetical contents, which precedes the eight books, the Astdhyyl. Each formula, also called astra, contains a group of speechsounds, to which stra is added at the end a mute letter, a so-called onubondha, which possesses a mnemotechnicfunction. With reference to this Siva-Stra31 have two propositions which I wish to defend : r Proposition I. The Siva-Stra has a double purport. The chief
2

See Supplement 16. Cf. B. Liebich, Zwei Kapitel der Kslk, Breslau 1892, p. I sq. 3 In order to simplify my argumentation I purposely treat the author of

the Siva-Stra and the author (or more correctly the authors) of the AstdhyyT as one and the same person. Cf. Supplement 2.

276 Barend Faddegon

p u r p o r t is a phonetical classification. The sutra presupposes the traditional alphabet and together w i t h this alphabet affords the means for a concise phonetical terminology. In this respect the Siva-Stra deserves praise. Proposition U. The subordinate p u r p o r t of the Siva-Stra Is t o afford the means of formulating concisely euphonic and morphological rules. Although on the whole very interesting and ingenuous, this grammatical use of the Siva-Stra in many cases degenerates into subtlety. In order t o prove my first proposition I have put together three tables : table A, containing the Siva-Stra; table B, showing the phonetical basis of this stra; and table C which explains the phonetical basis of the traditional alphabet. 4 Alphabet and Siva-Stra complement each other. For we see that Pnini forms f r o m both his pratyhras, i.e. technical denotations of phonetical classes. He uses ku for denoting the group of the five velars k, kh, g, gh, as found in the alphabet; similarly cu, tu, tu, pu respectively for the four other groups. From the Siva-Stra Pnini forms several pratyhras all according t o the same method ; for denoting a group he takes the first consonant of this group and then adds t o it the mute letter, the anubandha which closes the g r o u p ; for instance the pratyhra ak denotes a, i, u, r, /, the five original monophthongs of Sanskrit; the pratyhra ec denotes e, o, ai, au of which the t w o first were originally diphthongs and the t w o last are still diphthongs; in the same way (see stra 5 and 6) the pratyhra yan is a denotation for the semivowels as defined by the Hindus, i.e. our semivowels and liquids. An additional rule bearing on the indication of the quantity of vowels may be o m i t t e d here (see Supplement 6). The alphabet which Pnini used as complement t o the SivaStra was not his o w n invention. If it had been so, this fact w o u l d have been carried down t o us by t r a d i t i o n . But, since we are neither sure that Pnini was the author of the stra, in this way we cannot settle t h e i r chronological order. However, when we compare the alphabet w i t h the stra, we see that the f o r m e r is the older of the t w o and that the latter is derived f r o m it. In the case of the contact-consonants, namely, the alphabet shows a more logical order than the stra, while at the same t i m e we can prove that the illogical order of the stra is due t o the desire of obtaining pratyhras which may be useful for the formulation of the phonetical and morphological rules. Thus the alphabet enunciates the contactconsonants in regular order f r o m the velar series up t o the labial series ; on the other hand the Siva-Stra places for instance the jh and bh before the gh, and the ph before the ch. Putting aside this irregularity of o r d e r 5 l shall show w i t h a f e w examples the phonetical importance of the Siva-Stra. For this p u r p o r t I have composed the table B. This table differs f r o m table A, i.e. the stra itself, in so far as all the mute letters of minor importance are left out. The partial stras 1 and 2 are arranged in one line, which contains the sounds of the pratyhra ak; the stras 3 and 4 a r e combined into ec w i t h omission of the anubandha ; the h of stra 5 is put apart; the rest of stra 5 is put together w i t h stra 6 as yan; stra 7 remains the same, as am; stra 8 and 9 are conjoined asjhas; s t r a l O remains the same as jas; stra 11 requires a special hypothesis, namely, that in the first mental plan4

See Supplement 1.

See below the discussion on P. 7, 3, 101-102, and Supplement 5.

277 Panini's Grammar I: The Siva-Stra

ning of the stra the onubandha v followed th and not t; as a result of this surmise the five first consonants of stra 11 form the pratyhra khav and its three remaining consonants together w i t h stra 12 the pratyhra cay, finally the stras 13 and 14 remain the same. The proof for the hypothesis concerning the original place of the anubandho v is given in the Supplement 12 and 13; for the present I want t o draw the following conclusions from table B. 1st conclusion. In the alphabet voiced and voiceless speechsounds alternate w i t h o u t o r d e r ; but the author of the stra, by reversing the order of the consonants and by placing the h twice (once in stra 5 and again in stra 14), enables us t o denote all the voiced speech-sounds by the pratyhra as, and all the voiceless speech-sounds by the pratyhra khar, whilst at the same time we can combine the four continuants into the one pratyhra sal. 2nd conclusion. In the new arrangement, as met w i t h in the stra, the nasals can be combined w i t h the semivowels into the pratyhra yam, which is an advantage when considering the great acoustic affinity between semivowels and nasals. 3rd conclusion. Table B shows a remarkable regularity in the choice of the anubandhas as technical consonants. The stops k and c are used for the t w o divisions of the vocalic sounds; the nasals n and m for the acoustically connected semivowels and nasals; the continuants s and s for the t w o divisions of the voiced stops; the semivowels v and y for the t w o divisions of the voiceless stops ; the liquids r and / for the t w o classes of the continuants. 4th and final conclusion. The Siva-Stra is an- old and meritorious classification of the speech-sounds as met w i t h in Sanskrit. W i t h this last conclusion the proof for the first proposition is completed. I shall now repeat my second proposition : The Siva-Stra affords in its subordinate p u r p o r t the means for formulating very concisely euphonic and morphological rules. Although this grammatical use of the stra is on the whole ingenious and interesting, in many cases it degenerates into subtlety and artificiality. The merits of the Siva-Stra in its grammatical aspect are very evident. For instance, the pratyhra en in the t h i r d partial stra and the pratyhra aie in the fourth stra enable the author t o formulate concisely the rules for the guna and vrddhi of fand . Similarly the pratyhra yan, contained in the 5th and 6th stra, and the pratyhra k, contained in the 2nd and 3rd stra, by their correspondence in order facilitate the formulation of the rules bearing on the changes of semivowels into their correspondent vowels and vice versa. Many other similar examples in the entire stra prove its grammatical usefulness. The weak side of the stra, its subtlety, requires a fuller explanation. As an example I have chosen the pratyhra ya, contained in the stra 5 up t o 8. This pratyhra is only once made use of, namely in P. 7, 3,101-102, which runs as follows : ato dlrgho yai and supi ca. It would take up too much time t o explain in full the algebraic expression of these stras ; in ordinary language they express the rule that a long is substituted for the short of the thematic conjugation and the thematic declension in those cases where the personal ending or the case-ending begins w i t h one of the consonants contained in the pratyhra yan. W h e n in order t o apply this rule we compare all the beginnings of the personal endings and case endings w i t h the consonants of the pratyhra ya, then we find that these t w o series have only four consonants in common, namely, m, v, y and bh. Thus when we

278 Barend Faddegon

f o l l o w the general rules of these sutras 101 and 102 and at the same t i m e apply the exceptions t o those rules, as given by the stra 103 and following, we arrive at the fact that the Sanskrit language possesses such forms as pacmi, pacvah, pacmah; vrksya and vrksbhym. Evidently, Pnini was obliged here t o change in the SivaStra the alphabetical order of the contact-consonants and t o insert the onubandha after the 8th stra; for, if he had used the pratyhra yas, then he had t o formulate a separate exception for the second person plural of the tmanepada, such as pacadhve and apacadhvam.

The method of formulating the rules too broadly and then leaving it to the reader to find out from the context the proper limitations to these rules has, no doubt, always roused enthusiastic admiration among Hindu grammarians ; on the contrary the modern European, who does not cultivate oral memorizing to such an extent as the Hindu does, and even when memorizing does not appreciate the saving of a syllable, will consider the process artificial and misleading. And now I have proved my two propositions, the first one maintaining that the Siva-Stra as a phonetical classification deserves the highest praise, and the second that the Stra in grammatical respects, although often ingenious, shows a tendency towards artificiality. Liebich considers Pnini's grammar a work of interest only for the specialist; this modest attitude, we may already conclude, is too modest. For, although Pnini is no longer used for grammatical instruction, he still claims the attention of every linguist who feels interested in the history of his own study; and, in doing so, at the same time he rewards the reader with a sudden insight Into two typical features of the Hindu mind, its acute originality and its queer subtlety. Bhtlingk has on his own authority praised Panini's grammar, in its composition, as a masterpiece of its kind. Whether he is right or wrong jn his judgment, cannot be decided by a mere discussion on the Siva-Stra. Still I am inclined to think that a more complete analysis of only this introductory stra might show that even in the subtlety of Pnini there lies genius.

Supplement to Chapter I
1. T a b l e s s h o w i n g t h ep h o n e t i c a l basis o f t h e S t r a a n d t h et r a d i t i o n a l a l p h a b e t . A. The Siva-Stra.
r

Siva-

Speech-sounds
1 a u

Anu-

. ,. bandhas

Speech-sounds
gh dh dh jbgdd kh ph ch th th

_ i

_i

, ,, bandhas
s s

Anu-

2 3 4 5 6 7 8

C
eo
Q\ OU

n k
c
t

9 10 11 12 13 14

c t t kp
sss

hyvr

y
r

I
m nn jhbh

n m

279 Pnini's Grammar I: The Siva-Stra

B. The Phonetical Basis of the Siva-Sutra. Place of articulation Character of sound S.-S.
DO r +->

labio-velar

pala

pala

"O

dem tal

es
lab. vel. 15
_r

C C -O

_o
DO

o E

I vowels

1-2 3-4 5 5-6 7

8-9 10 11 11 12 13 14

e
ai

o au r m bh b gh g kh n dh d

III semivowels II nasals II voiced stops II voiceless stops

n
dh d th t

m s s

ph P

ch c

th t

[W] y
r I

IV continuants

C. The Phonetical Basis of the Traditional Alphabet. Place of articulation

gloti

Character of sound I vowels

pharyngal a

velar

palatal

labiovelar u oau

mrdental labial dhanya rf //

il eai

II stops III semivowels IV continuants h

kkhg gh

cch j jh y s

tthd dhn r

tthd dhn

pphb bh m
V

1 s

2.TheauthorshipoftheSiva-Sutra. Pnini begins the first stra of his grammar with the word vrddhi as a kind of precative formula or mongola and the last stra of his work lla a" is a clear reference to this first stra. For in free paraphrase this stra means " in the beginning of my work I made the fictitious supposition that a and are homogeneous sounds and thus I called the vrddhi of the a, but now I take back this supposition and give to the a its real phonetical value." This beginning and this end stamp the eight books as a complete composition, from which the Siva-Stra is exempted. Should Pnini have intended to claim it as its own, he would have inserted it in his work. The name Siva-Stra, i.e. Stra taught by the god Siva, moreover shows that Indian tradition does not ascribe these introductory formulae to Pnini himself. This legendary name, which is of more recent origin than the names aksora-stra or pratyhrastro (according to Bhtlingk, edition, p. 1), is also interesting,

Anu

pha

las

280 Barend Faddegon

because it reminds us of the narrative which proclaims Siva t o be Panini's teacher in grammar (Bhtlingk, edition, Einleitung, p. VIII and Kathsaritsgara, 4, 20 sqq.). Most likely the Siva-Stra is of earlier date than the AstdhyyT. However, f r o m this conclusion we must draw another conclusion, namely, that Pnini must have b o r r o w e d many of his grammatical rules f r o m his immediate predecessors, for the SivaStra is closely interwoven w i t h the grammar and many of its anubandhas are inserted for the p u r p o r t of facilitating t h e formulat i o n of grammatical rules. W h e n I mention Pnini in my article, I really mean the grammarians t o w h o m we are indebted for Siva-Stra and AstdhyyT. For the prehistory of Pnini's grammar cf. B. Liebich, Zur Einfhrung in die indische einheimische Sprachwissenschaft, II,

Sitzungsber. d. Heidelberger Akad. d. Wissensch., phil.-hist. KL, 1919,15. Abh., especially p. 43-45, and Th. Goldstcker, Pnini, London 1866, p. 181. Further see below 16. 3. The o r d e r of t h e c o n t a c t - c o n s o n a n t s in t h e SivaStra. When we compare the useful alphabetical arrangement of the contact-consonants in their rows and columns with the Siva-Stra, then we notice that the columns of the alphabet in reverse order (nasals, voiced aspirates, &c.) become the rows of the stra.r 4. The p h o n e t i c a l p r i n c i p l e s of alphabet and SivaStra. A glance at the tables of 1 shows that each of the four phonetical classes of the alphabet moves forward from the pharynx to the lips, the only exception to this being the u-group; that the Siva-Stra follows two systems of arrangement, the order palatal, labial, velar, &c. for the voiced consonants, and the order velar, labial, palatal, &c. for the voiceless stops. 5.The reason f o r t h e i r r e g u l a r ity o f o r d e r in t h e Siva-Stra. The author of the Siva-Stra has here and there deviated from the logical order of the alphabet in order to arrive at a mnemotechnical system. Thus e-o, ai-au, y-v are grouped into pairs like i-u in order to facilitate the rules for guna, vrddhi and samprasrana. and m are placed in front of the nasals to obtain the pratyahra am, the three nasals which at the end of a word are doubled according to the same rule. For the place of jh and bh see 9; for kh, ph, k and p 11. 6. The i nd cat on of t h e v o w e l - q u a n t i t y . At, it, ut, rt, It indicate the short monophthongs, at, &c. the long monophthongs. A represents both a and , and according to the commentators likewise the nasalized a and , and the a lengthened by pluta. The same rule holds good for the other monophthongs. 7. The i n s e r t i o n o f t h e anubandhas o f s u b o r d i n a t e i mportance. The principal anubandhas afford the means for a phonetical classification ; the anubandhas ofsubordinate importance have only a grammatical significance, see 8-11. 8.Theanubandhanattheend ofthe firststra. We may distinguish the anubandhas n at the end ofthe first and sixth stra as n, and n2; and accordingly the pratyhra an
as an and an2.

281 Pnini's Grammar I : The Siva-Stra

T h e pratyhara an, is used in P. 6, 3,111 : a, , u are lengthened t h r o u g h t h e loss o f a f o l l o w ing dh or repha (punrakta, &c.) ; P. 7, 4,13 sq. : , J, are shortened before the taddh/ta-suffix ka(jnak, kumrik, &c); P. 1,1, 51 : a stra bearing on the alternations r : r, r : ir, f : ur ; P. 8, 4, 57: a rule for the nasalization of final monophthongs. 9. The anu band ha n. Cf. 5 on the group e-o. 10. The anubandha t. This anubandha is only used in the pratyhara at, which contains the vowels, the h and the semivowels with the exception of/. See: P. 8, 3, 3 and 9: a Vedicsandhi-ru\e; P. 8, 4, 2: a rule for the transition of n into n; here the reader is obliged by the context to eliminate the repha. This artifice of formulating a rule too broadly and leaving the reader to find the restrictions himself in the context, may be termed " implied restriction." It is a device, used by Pnini with virtuosity and care, and which leads to conciseness and avoids ambiguity; P. 8, 4, 63: a facultative sandh/-rule for s, followed by vowel, y, v or r; s + h, of course, does not occur. 11. T h e anubandha . The anubandha is only met with in the pratyhara ya, P. 7, 3,101-102. The first of these two stras is interesting as implying two " implied restrictions." On the one hand the pratyhara ya is limited toy, v, m, bh\ and on the other hand the technical term srvadhtuke (taken from 7, 3, 95) is restricted in meaning by its context. In the stra ato drgho yai [i.e. srvadhtuke] the genitive means " is substituted for," the t of at indicates the short a as such; the locative means "when follows." The translation accordingly runs: the [correspondent] long vowel, [namely ] must be substituted for a before a srvadhtuka-suif\x beginning with y, v, m and bh. Now there are three classes of srvadhtuka-suff\xes : 1. the suffixes of the present tense (a accented and unaccented, nu, n, &c, with the exception of u); 2. the personal endings with the exception of those of the perfect; 3. a small class of suffixes which form nouns from the stem of tenses, e.g. the participle-suffixes -na and -ant; the Vedic infinitive-suffix -adhyai; all the suffixes of this third class begin with a vowel.6 Of these three classes of sn/adhtuka-s uffixes the third is excluded because of its beginning with a vowel ; the first is excluded, because the rule applies only to such suffixes which follow an a, namely the a of the present system. Thus the term srvadhtuka is limited by the context to the meaning of personal ending. N o t e . In abhavam and bhavanti the last a is short, because Pnini analyses here abhav-am and bhav-ant'r, according to 6,1, 97 the a of the stem is elided before the a of the ending. 12. The anubandha v, its o r i g i n a l place and t h e advantages of its r e m o v a l . As the Siva-Stra is in principle a phonetical classification, one would expect to find the anubandha v after th. What has been the reason for placing it after t? In other words, what advantage
See Bhtlingk's edition of Pnini, p. 183* s.v. sit!.
6

282 Barend Faddegon

aprasn, i.e. (padasya)1 no (ru)2 chavy (ampare)3, aprasn. padasya* = padasynte, P. 8,1,16; ru2 = the indication of an (unoriginal) final r or one of its substitutes such as visarga, s, &c, P. 8, 3,1 ; ampare3, P. 8, 3, 6. The entire stra is closely connected with 8, 3, 2 and 4. Translation : ru (i.e. s before t, th, s before t, th, s before c, ch, see 8, 2, 66; 8, 3,15 and 34; 8, 4, 40 sq.) is substituted for n at the end of a word, when the next word begins with one of the consonants of the pratyhra chav fol lowed by one of the speechsounds of the pratyhra am. At the same time the preceding vowel is nasalized or is followed by an additional anusvra. The rule does not apply to the word prasn. For instance in bhavn + chdayati, the of bhavn becomes either am or art), and the n is changed into s. N o t e I. The technical substitutions of Pnini's are evidently no " laws of sound change" in our sense of the word. N o t e II. The rule affords again an example o f " implied restriction." For am in ampara contains as apratyhra the vowels, the h, the semivowels and nasals. But practically speaking7 the voiceless contact-consonants are only followed by vowels and semi-vowels, the only exception to this being tman. On the other hand, if we examine the consonants not included by am and which form the pratyhra jhal, then we see that there is only one case in which a chc/v-consonant is followed by a jha/-consonant, namely the word tsarati. Thus the rule given by the addition ampare amounts to this, that in bhavn tsarati the and n remain unchanged. For the rest, since ampare is only a supplement taken from the precedent stra, we are not quite sure whether Pnini himself intended to state this exception. 13. No disadvantages c o n n e c t e d w i t h t h e removal of v. No difficulties arise from the removal of v. For instance, if we want to express that under certain circumstances an aspirate loses its aspiration, we simply say that jhas becomes jas and khav (or even khay) becomes cay, for in the latter case the cav- (or cay-) consonants are eliminated from the khav- (or khay-) group by " implied restriction." See P. 8, 4, 53 sqq. and 8, 2, 39 for the form in which this principle is applied. 14. The ingenious pregnancy in Pnini's use o f t h e Siva-Stra. First e x a m p l e . The imperative 2nd pers. sing, ofthe verbs Hh and dvis is lldhi and dviddhi. How can we construct these forms with the aid of Pnini's grammar? N o t e . A modern linguist would explain these forms with the aid of "phonetical laws" and psychical "associative influences,"8
7

has t h e r e been in removing it, and are there no disadvantages connected w i t h doing this ? In searching for an answer t o these questions, we notice that the strange position of v and the irregular o r d e r of the voiceless contact-consonants are due t o the same cause, the wish of Pnini t o f o r m the pratyhra chav, which includes the t w o series ch, th, th, and c, t, t. A peculiarity which strikes us in these t w o series is t h e i r parallelism t o the series s, s, s. A n d f u r t h e r we notice that the pratyhra chav is only once met w i t h , P. 8, 3, 1, nas chavy

For the following see the dictionaries.

See the excursus,

283 Panini's Grammar I: The Siva-Stra

but f r o m Panini we only expect grammatical description in pseudomathematical f o r m . *lih-dhi > *lidh-dhi; P. 8, 2, 31 : h > dh, if h is samyogdi, i.e.

the first component of a consonant-group (8, 2, 29), whilst the suffix begins with a jha/-consonant (an/ consonant with the exception of nasal or semivowel), 8, 2, 26; *lidh-dhi > lidh-dhi; P. 8, 4, 41 : mrdhan/izing9 owing to assimilation ; *lidh-dhi > lidhi; P. 8, 3,13: dh dropped before dh; 6, 3,111 : / > /"through the loss of a fol lowing dh. On the other hand: *dvis-dhi > *dvid-dhi\ P. 8, 4, 53: jha/-consonant changed into the nearest related jos-consonant before jhas-consonant; dvid-dhi > dviddhi) P. 8, 4, 41 : mrdhanyizing through assimilation. (Thed is preserved before dh; cf. P. 8, 3,13.) 15. Second example o f p r e g n a n c y in Pnini's use of t h e Siva-Stra. How to build up with the aid of Pnini's grammar the flexion of the stem and root //h; especially of-lit, nom. sing.; -litsu, loc. pi. ; alet, 2nd pers. sing, imperf. ; leksi, 2nd pers. sing, present. The nominative -lit and the 2nd person a et follow the same explanation. We start from the theoretical forms *lih-s and *aleh-s *lih-s > *lih, and *aleh-s > *oleh ; P. 6,1, 68 : the flexional s is dropped after a consonant ; *lih > *lidh, and *aleh > *aledh; P. 8, 2, 31 : h > dh "ante" = at the end of a word ;10 (8, 2, 32-35 state special rules which do not apply to our case) ; *lidh > lid, and *aledh > aled; P. 8, 2, 39: jhalm jaso 'nte, ajha/-consonant becomes the nearest related jas-consonant at the end of a word ;10 lid remains //dor > lit, and aled remains aled or > aleV, P. 8, 4, 56: free choice between jas-consonant (a voiced non-aspirate) and car-consonant (car really means both voiceless non-aspirate and voiceless continuant, but here only voiceless non-aspirate owing to " implied restriction "). We now turn to the construction of the verbal form leksi and start for that purport from the theoretical form *leh-si. *leh-si > *ledh-si; P. 8, 2, 31 : h > dh (cf. samyogdi 8, 2, 29 and jhali 26); *ledh-si > leksi ; P. 8, 2, 41 : dh > k before s, and P. 8, 3, 57 and 59: s > s after k. The construction of the form litsu is more difficult. One would expect *liksu according to the same argumentation as given for the verbal form leksi. Here the rule of stra 8, 2, 41 is annulled by other rules and principles. P. 1, 4,14: sup-ti-antam padam, apada is that which ends in a declinational or conjugational ending; 17: sv-disv asarvanmasthne, the term pada is also applied to the nominal stem when it precedes a flexional ending provided no "ending of a strong case" follows; 18: yaci bham, the term bha is applied to the nominal stem
9 I leave the term mrdhanya untranslated, although Prof. Lanman has suggested the very near translation "domal" for it; "concave-linguals" would be the best phonetical term;

and the conversion of dental into concave-lingual might be called concavation. 10 A fuller explanation of these stras is given below.

284 Barend Faddegon

before a suffix which begins w i t h / or oc ( = a vowel). The last t w o of these three stras must be explained in agreement w i t h the general principle laid down in P. 1, 4,1 for the chapters 1, 4 sqq. According t o this principle stra18 contains a restriction t o stra 17. Consequently the stras 1, 4 , 1 4 ; 17 and 18 give an exhaustive classification for the nominal stem-forms into a strong forms or stem-forms before sarvanmasthna ( 1 , 1 , 42-43), podo or weak forms the ending of which begins w i t h bh or s, and y " bha" or weak forms the ending of which begins w i t h a vowel or yo (this last condition refers t o the rules for derivation). The fact that the same t e r m podo is applied t o a " c o m p l e t e w o r d , " i.e. a w o r d w i t h a flexional ending, as well as t o the weak nominal stem before a consonantal ending means in Pnini's pregnant language that the pada-stem follows the same euphonic rules as the " complete w o r d " in the phonetical coherence of the sentence. N o w P. 8 , 1 , 1 6 runs podosya, which means that up t o 8, 3, 54 we must supplement the expression podosyo t o all the stras which by t h e i r context demand this. Consequently P. 8, 2, 31 hodhah (... ante ca) and 8, 2, 39hol'am jaso 'nte must be understood as padosyonte, which expression includes the end of a nominal stem before a consonantal ending: *lih-su > *lidh-su > *lidsu > litsu (cp. P. 8, 4, 55 and 42). N o t e . Since Pnini does not mention the stem dvis explicitly, one might t h i n k that he has overlooked the declension of the stem dvis, in which case I have beaten Pnini in subtlety. But considering the great care which he has bestowed on the description of the consonantal declension, this is not likely. 16. T h e p u b I c a t i o n o n P n i n i by H a n n e s S k I d . W h e n preparing my paper for the Congress of Linguists I had forgotten that Skld had already published his penetrating studies on Pnini (Popers on Pnini ond ndion grommor in general, Lund 1926). In the chapter Facts and conjectures about the Siva-Sutras Skld has treated the same matter as myself, but according t o a different method and w i t h a different aim.

Excursus

T h e f o r m s dviddhi

and

lldhi.

A strictly phonetical development of the imperatives of dvis and lih w o u l d have made them similar. For modern comparative linguistics teach us the following series : IE. *dvis-dhi -** *dvizdhi>* *dvizdhi-+ *dvizdhi-+ Skr. *dvJdhi. IE. **ligh-dhi-> *ligdhi-> *iizd'hi - > *izdhi'^> Skr. lldhi. In the transition zdh-^zdh (the 3rd transition of the 2nd series), or in general in the transition f r o m the combination palatal continuant + dental contact-consonant into the combination mrdhanya continuant + mrdhanya contact-consonant we have t o do w i t h a reciprocal assimilation. In the palatal continuant the tongue is only raised in the middle whilst the point is lowered towards the back of the b o t t o m teeth ; in the mrdhanya continuant both the middle and the point of the tongue are raised whilst the surface between is concave. Under the influence of the following dental stop the palatal continuant is changed into the mrdhanya continuant, and owing t o this mrdhanya continuant the dental itself is changed into a mrdhanya stop. N o t e . Both the palatal continuant and the mrdhanya continuant occur in the individual pronunciation of English sh. O t h e r

285 Pnini's Grammar I: The Siva-Sutra

examples of the same assimilation : IE. *oktou > *osto- > Skr. asta-; IE. *vikto- > Skr. vista-; IE. *ugdhi (imperative of V wek) > *uzdhi > *uzdhi ( > Skr. uddhi). The z originated from z before d/i was transferred by " phonetical analogy" to forms in which z was followed by bh\ thus *lizbhis > *lizbhis. A second normal sound change converted z before the voiced stops g, >h, j into d (whilst dj > jj), and z before the voiced stop bh intod; thus *dvizbhis > dvidbhis and *iizbhis > lidbhis. Dialectically z was converted intod also before the homorganicd/i; thus *dvizdhi > dviddhi, *uzdhi > udd/i/ (cf. J. Wackernagel, Altindische Grammatik I, Gttingen 1896, p. 251, 271). By a third normal and general sound change z was dropped before dand dh, and z before d and dh with simultaneous lengthening of the precedent vowel ; thus *lizdhi > lldhi. To summarize, lldhi is the regular phonetical form of the universal language, and dviddhi and uddhi are preserved formations of an old dialect (other explanations are given by Wackernagel, l,p.175sq.,149candp.177, 150 in fine).

19
Kshitish Chandra Chatterji (1896-1961)

A prominent feature of the work of the modern period is the large number of contributions made by Indian scholars. This is only natural; an Indian Sanskritist is in a much better position to study the Indian grammatical tradition than is a Western scholar. The reason that this feature is nevertheless inadequately represented in this volume is that most Indian contributions have been of a rather technical nature. On the one hand, Indian papers often deal with the interpretation of specific rules and are thus beyond the more general scope adopted here. On the other hand, they often presuppose such a high degree of fluency in Sanskrit that they could not be incorporated here without being largely rewritten. Among early monographs written by Indian scholars, special mention may be made of I. S. Pawate's The Structure of the Astdhyy(1935?), a rare work in which the author argues that many rules of Pnini's grammar are not by Pnini. This line of research was developed much later by Robert Birw, especially in Studien zu
Adhyya III der AstdhyyJ Pninis (1966).

Kshitish Chandra Chatterji (1896-1961), who studied in Calcutta and taught comparative philology and Sanskrit at Calcutta University for thirty-five years, wrote several papers on the Indian grammarians as well as a monograph, Technical Terms and Technique of Sanskrit Grammar (1948), and a lively annotated translation into English of the Paspas or " Introduction " to Patajali's Mahbhsya (second edition 1957). These works are full of valuable information and relate interesting and often amusing traditions among the grammarians. The article reproduced here deals with various criticisms directed at the Sanskrit grammarians by scholars of other schools, in particularthe ritualist philosophers of the Mimms (who themselves had contributed in a very original way to the study of Sanskrit syntax and semantics; see Edgerton 1928,1929; Staal 1962b). These criticisms had never been the subject of a separate study, though they were not unknown (for example, Kumrila's stravrttikabhsyesu driyate cpasabdanam ' ungrammatical expressions occur in the rules, the vrttiks and the commentary* was quoted by Bhler, see this volume, p. 195, note 1). Chatterji quotes from several MJmms sources: the MJmms Stra attributed to Jaimini (abbreviated J.S.), of uncertain date; Sabara's commentary on this stra, probably written in the fifth century A.D.; Kumrila's Tantrayrttika (T.V.), of the eighth century; and Prthasrathi Misra's SstradJpik, of the fourteenth century. He also quotes a logical work, the NyyamanjarJ (N.M.), probably written in the tenth century. The controversies among the grammarians and the logicians have only recently begun to be studied (see Matilal 1966,1968). Chatterji quotes rather extensively from his sources and provides free translations, which for that reason are not placed within quotes. In fact, his article consists very largely of such free translations and summaries of the texts. I have omitted most texts quoted in the footnotes, retaining the references, but I have added a number of translations, always in square brackets. "The Critics of Sanskrit Grammar" is reproduced from the Journal of the Department of Letters, University of Calcutta 24,1934, 3.1-21. The beginning of this article refers to Patajali's discussion on s/sta, where this circularity is also mentioned (see above page 100). t h e references to incorrect words also echo Patajali (see above page 97). The final mention of the fact that there are always people who are cleverer than others is a well-known theme in philosophical texts (for example in Sakara, and Bhartrhari).

The Critics of Sanskrit Grammar ( 1934) Kshitish Chandra Chatterji

"'

Parthasarathi Misra in his Sastradpika sums up the charges against grammar in the following Krik: nirmlatvd vigltotvn naisphalyd vedabdhant/ prvaparavirodhc ca nsya prmnyasambhavah// No authority can possibly attach to grammar, for it has ' nothing authoritative as its basis, because its adherents themselves differ widely in their opinions and cast aspersions on each other, ; because it serves no useful purpose, because its teachings are in conflict with the Vedas and lastly because it is self-contradictory. 1. Now as regards the first point. We derive all our knowledge of grammar from the writings of the three sages Pnini, Ktyyana and Patajali whence grammar is described as trimunivykaranam ['the grammar of the three sages']. But.Pnini and others are human beings and as such cannot be regarded as free from the four defects bhrama (mistake), pramda (carelessness), viprolipso (desire to deceive) and karanptava (defect of the senses). Thus grammar being a product of the human brain must, in order that its authority may be established, be shown to be based on the Vedas, which it is not possible to do. No doubt we find mention of many technicalities of grammar, of sabda ['correct expression '] and apasabda [' incorrect expression '] in the Vedas themselves. We find, for instance, such texts as tasmd brhmanena na mlecchitavai ['therefore brahmans should not speak barbarously'], mleccho ha v esa yad apasabda ['a barbarism, i.e., an incorrect expression ' ] , ekah sabdah samyak prayuktah svarge loke
kmadhug bhavati ['a single expression well pronounced fulfills

one's wishes in heaven '], hitgnir apasabdam prayujya prayasciWym srasvatJm istim nirvapet [' when the hitgni priest uses an incorrect expression he should, for expiation, perform an oblation to Sarasvati '] tasmd vykrt vg udyate [' therefore speech is manifest'], etc. But these texts refer to the Prtiskhyas which deal exclusively with Vedic words and accents and have nothing to do with grammar in the sense in which we are considering it here. Should you say that the science of grammar like the science of medicine is based on the usage of the Sistas or experts, we must ask you to define clearly what you mean by Sista. Do you mean by Sista those who use correct words like go, etc., or those who use corrupt words like gavl, etc., or those who use both ? In the first case the vicious circle is inevitable, for grammar is based on the usage of the Sistas and Sistas are those who are versed in grammar. In the second case, it has to be admitted that coachmen and others who use such corrupt words as gdv, etc., are Sista and that grammar based on the usage of these coachmen1 who habitually use corrupt words lays down rules for the formation of correct words a statement which is self-contradictory.2 In the third case it is difficult to conceive what useful purpose would be served by grammar which prescribes a heterogeneous mass of correct and corrupt words.
By the same brilliant process of reasoning a philosopher arrived at the following conclusion as to the origin of language: "The first men as yet speechless came together in order to invent speech, and to discuss the most appropriate names that should be given to the perceptions of the
1

senses and the abstractions of the mind." 2 In the Mahbhsya, however, we find the tables turned on the grammarian who discovered an apasabda in the language of his charioteer (ed. Kielhorn.Vol.il, p. 557).

288 Kshitish Chandra Chatterji

Sistas such as the authors of the Kalpa-sutras, Smrti-texts, MTmms and Grhya-stras have been found t o use any number of " i n c o r r e c t " words. The nom. neut. sing, of itaro [ ' o t h e r ' ] is not taram but \torot a f o r m found in such sentences as samnam itarac chyenena ['the other similar t o a falcon ' ] , yet Masaka in various places uses such incorrect expressions as samnam taram jyot'istomena [ ' t h e other similar t o the jyotistoma sacrifice'] and the author (of the Chndogya-stras) himself has said : samnam taran) gavaikhikena [ ' t h e other similar t o the Gavaika descending node']. According t o Panini's rule svar'itaitah kartrabh'iprye kriyphale (1.3.72) a root w i t h an indicatory svarita vowel or a r o o t w i t h an indicatory takes the Atmanepada when the benefit of the action accrues t o the agent, yet in the t e x t bah'ispavamna'ih sadasi stuvlran ['they chant w i t h Bahispavamna praises in the sacrificial hut 1 ] where the nom. t o stuvJran is three priests and where, consequently, the benefit of the action does not accrue t o the agent (the merit arising from the performance of the sacrifice going t o the yajamna [' patron of the sacrifice'] and thus where we should expect Parasmaipada as in the sentence yajanti yjakh [ ' t h e priests sacrifice'] Atmanepada has been used regardless of grammar. Similarly in the strapratyasitv pryascittam na juhuyah

[' having sat down in the opposite direction they should not perform the expiation ceremony '] Asvalyana has not used yap [' the suffix -ya'] in pratyasitv though it is a compound in direct violation
of Pnini's r u l e : samase 'naprve ktvo yap (VII.1.37). In jyen-

ksino jya [' having anointed the eyes with clarified butter'] on the other hand, yap has been used in jya even though it is not a compound. Similarly in his work on Siks, Nrada writes 3 pratyuse brahma cintayet[{ at dawn (pratyusas'i) he should meditate on brahman '] where the expression pratyuse is on a par with words like gvl, etc. Similarly in the line jntrah santi mety uktv [' having said : they are my witnesses'] (VIII. 56) where we should expect jntrah santi ma ty uktv Manu has joined me ity uktv into Sandhi regardless of the doctrines of grammar. Similarly in the MTmms-stra gavyasya ca taddisu ['and incases like that of gavya'] (1.1.18) the word gavya which can be correctly used only in the sense of gor v'ikrah [' modification of cow'] or gor avayavah [' part of cow'] has been used in quite a different sense, viz., gavm ayana ['going of cows']. Similarly in the stra dyvos tatheti cet ['and ifyou say it is similar of the two heavens'] (J.S., IX. 3.18) instead of saying dyvprthivyoh ['of heaven and earth '], dyvoh has been used against all canons of grammar. Similarly, the author of the Grhya-stras has made use of the expression mrdhany abhij'ighrnam ['smelling at another's forehead ' (as a token of affection)] where he should have said mrdhani abhighrnam, the substitute jighra (for ghr) being out of place, for Pnini in his rule pghrdhm ... (VII.3.78) distinctly lays down that j'ighra is to be substituted for ghr only before srvadhtuka affixes and lyut ['the suffix -ana'], as every schoolboy knows, is an rdhadhtuka affix. Even in the Nirukta which is the complement of grammar there are many uses which do not conform to the rules of grammar, such as brhmano bruvant (Brhmana is so called because he speaks). While stating the derivative meaning of the word Brhmana as applied to the frogs in the stanza samvatsaram sasaynh [' after lying low for a year . . . ' ] , Yska with a view to show that
3

In the printed text the reading is

usasi brahma c'mtayet (11.8.1).

289 Critics of Sanskrit Grammar

this particular application of the w o r d Brhmana is due t o t h e i r habit of speaking, uses the w o r d bravana instead of substituting vac for br according t o Pnini's rule bruvo vacih (11.4.53) and saying vacano. (In the printed texts of the N i r u k t a , however, we find brhmana vratacarino' bruvnh.) There is no end of incorrect words in tinosas and purnas. W e also find words like ubhbhyam [' by, for, f r o m both ' (for, ubhbhym)], etc., used by the author of w o r k s like the Hastisiks. Vlmki, for example, uses such an expression as tad anantaram tubhyam ca rghavasya ca [' immediately after that for you and of Rghava'] and Dvaipyanasuch a sentence asjanmejanmeyad abhyastam [ ' w h a t is accumulated in b i r t h in b i r t h ' ] . 4 In sentences like yugapad ubhbhyam dantbhym yah prahrah sa ubhbhyah

['striking with both tusks at the same time is with both (plural)'], etc., words have been used by Plakpya, Rjaputraand others regardless of grammar.5 Nor can Grammar be held authoritative because it is generally recogn ised as a vednga [auxi I iary science to the Veda] for, according to us, it is not siks [phonetics], kalpa [ritual], etc., that constitute the six Vedngas but the group of six, viz., sruti [verbal authority], inga [evidence], etc. Or if this does not satisfy you, if you insist on the generally accepted enumeration of the six Vedngas in which grammar is included, even then we have no objection, but grammar must be held to mean not grammar in general but the grammatical speculations scattered in Vedic literature or the Prtiskhyas. The text tasmd brhmanena niskranam sadago vedo'dhyeyo jeyas ca ['therefore the Veda and its six auxiliaries must be disinterestedly studied and known by a brahman '] may refer to either of these two. Thus we find the science of grammar is not based on solid granite but on heaps of sand. 2. Now we come to the next point.6 It is well-known that Sanskrit grammar is based on the stras of Pnini, the vrttikas7 of Ktyyana which explain, criticise and supplement the rules of Pnini, and the Bhsya of Patajali which explains the rules of Pnini and Ktyyana and often severely criticises the latter. Of these three Patajali is regarded as more authoritative than the two other sages, as he, coming much later than they, had an opportunity of observing a much greater number of actual forms.8 Kaiyata says (on 11.4. 26) munidvayc ca bhsyakrah pramnataram adhikalaksyadarsitvt ['the author of the commentary (i.e. Patajali) has greater authority than the other two sages because he has observed more linguistic usage']. ... Now these three sages often differ from one another. Ktyyana is openly hostile to Pnini. He subjects the rules of Pnini to a rigorous test, finds many of them wanting, rejects some of them and suggests supplementary rules to remedy the defects of others. His rules often end with iti vcyam or iti vaktavyam ['thus it should be said '] meaning that Pnini ought to have said this but has not said.
4 5

N.M., p. 414. T.V., p. 259. 6 For the sake of convenience, the second and fifth points have been taken up together. 7 uktanuktaduruktarthovyaktikdri tu

varttikam ['the vartikkas explain the meaning of what is said, not said or imperfectly said ' (in astro)] 8 Haradatta's Padamajar, Benares ed., p. 7.

290 Kshitish Chandra Chatterji

The Bhasyakra professes the highest regard for Panini, thinks his rules perfection itself, says at the beginning of his treatise : pramnabhta cryo darbhapavitrapnih sucv avakse prnmukha upavisya mahat yatnena stram pranayati sma tatrsakyam varnenpy anarthakena bhavitum kirn punar iyatstreno [ ' t h e competent teacher, having sat down facing East w i t h purifying darbha grass in his hand in a pure place, composed the rules w i t h great care hence it is impossible that a single sound should be w i t h o u t meaning, much less an entire r u l e ' ] (ed. Kielhorn, Vol. I, p. 39) and f u r t h e r on (under Pnini, VI. 1, 77) nitye yah skalabhk samase tad artham etad bhagavms cakra smarthyayogn na hi kicid asmin pasymi sastre yad anarthakam syt [' If the compound is obligatory according t o t h e f o l l o w e r s of Skalya, the preceptor has fixed the appropriate meaning by way of the meaning of the context, for I don't see anything in this treatise which is w i t h o u t meaning']. Yet he takes all sorts of liberties w i t h the rules of Pnini, sometimes rejects them wholesale (as the seven rules in the apdna section), sometimes goes out of his way t o change them (as in upasargd anotpara, VIII, 4, 28, which he changes t o upasargd bahulam), splits up single rules of Pnini into t w o one part prescriptive and the other merely illustrative (as in the case of samnasya cchandasy amrdhaprabhrtyudarkesu VI. 3. 84, etc.), sometimes charges Pnini w i t h carelessness as when he speaks of the repetition of tad in tad adhhe tad veda (IV. 2. 59) as pramdakrtam cryasya ['a mistake has been made by the precept o r 1 ] and so on, and turns and twists the rules of Pnini in all conceivable ways. Sometimes he raises objections which he cannot solve; like the Frankenstein monster they prove t o o strong for him. Let us take some more concrete examples. It has been pointed out that in the very first rule of Pnini the canons of both grammar and rhetoric have been violated. The rule is vrddhir d aie meaning , ai and au are t o be known as vrddhi vowels. N o w it is a commonplace of grammar and rhetoric that of the subject and the predicate, since the subject is more or less known t o us and the predicate supplies some new information about the subject, it is the subject that is t o be placed first, for we always find it easy t o pass f r o m the known t o the u n k n o w n . The violation of this rule constitutes the defect known as vidheyvimarsa.9 In the rule vrddhir d aie the predicate vrddhi ought t o have been placed after d aie, just as Pnini himself says ad en gunah. Patajali points out the defect and apologises for Pnini dam ekam cryasya mangairtham mrsyatm mngalika cryo mahatah sstraughasya mangairtham vrddhisabdam ditah prayukte / mangaldJni hi sstrani prathante vJrapurusakni bhavanty yusmat purusakni ca, adhyetras ca vrddhiyukt yath syuh [' let the preceptor be forgiven for this once on account of auspiciousness. The preceptor, intent on auspiciousness, uses the w o r d vrddhi at the beginning of his great compilation of science for the sake of auspiciousness. For only such works as have auspiciousness at the beginning t h r i v e well and make mankind strong, long-lived, and prosperous '] (ed. Kielhorn, Vol. I, p. 40). This apology is lame and completely breaks down in the case of the rule aprkta eklpratyayah (I.2.41) where commentators are The commentator of the Vyaktiviveka (Trivandrum ed., pp. 15-16) tries to show that there are indica9

tions in the writings of Pnini, Ktyyana and Patajali about many of the rhetorical defects.

29) Critics of Sanskrit Grammar

forced t o admit, t h e inversion of the order of t h e subject and predicate merely serves t o show that the grammarians do not recognise the defect known as vidheyvimarsa. The non-mutation of the final palatal into the corresponding guttural is the other defect in this rule. This also Patajali explains away w i t h the help of a rule applicable only t o t h e Vedas for according t o him chandovatstrni bhavanti ['stras are like the Veda'] (ed. Kielhorn, V o l . I, p. 37,1.4). Frequently we find sharp differences of opinion among t h e grammarians themselves on many points. 10 The w o r d patita for instance is t o be derived w i t h the suffix i ta according t o Pnini and ta according t o Patajali. From t h e s t r a / a t tad etebhyah parimne vatupV.239) it appears that in tvn the affix is vatup, but from the vrttika daVatoV arthavaisesyt the suffix appears t o bedvatu. 11 According t o Pnini the compound of na w i t h a substantive is optional, whereas according t o Ktyyana it is obligatory. 1 2 Further, according t o Patajali those words are t o be regarded as correct t h e precise meanings of which may be easily ascertained f r o m the practice of experts and which have been in use in these particular senses f r o m t i m e immemorial. A n d what do we find in Pnini ?13 From t h e beginning t o the end his w o r k abounds w i t h words and expressions like ti, ghu, bha, etc., which were never in use before his t i m e and which must consequently be regarded as 'apasabdas.' A strange sight thisthe w o r k of an exponent of correct words bristling w i t h incorrect w o r d s ! W h a t is stranger still is that Panini himself has violated his own rules at almost every step. Take for instance such a simple

rule as janikartuh prakrtih (\ A30). In this rule the word janikartuh may well be regarded as an instance of the maximum of error in the minimum of space. According to the dictum ikstipau dhatunirdese (III.3.198.2) ik has been tagged on to jan to indicate that the writer is speaking of the root Jan. The stra does not certainly prescribe ablative ending for the producer of the root. The word jani has here evidently been used in the sense of " t h e act of being born " a meaning which it can never have according to the rules of grammar. Further the rule trjakbhym kartar (II.2.15) forbids the compounding of an objective genitive with a word ending in the agent suffix trc or aka.u Pnini has gone out of his way to compound jani with kartuh and thereby violated one of his own rules. He has violated another rule of his, viz., gamahanajanakhanaghasm opah kity anai (VI.4.98) by not dropping the vowel of the root jan and changing n into . Then again Pnini uses the expression samjnapramnatvt (1.2.35) where the samsa in samjnpramnatva is absolutely indefensible for he himself has forbidden such compounds in his rulepranagi/na ... (11.2.11). In this instance his noble example has been followed even by his commentator and critic Ktyyana who constantly uses such expressions as dambherhalgrahanasya jativcakatvt siddham (1.2.10.1), nyabhvyam tu klasabdavyavyt

(1.1.2.11), etc. Then again he frames the rules aksanahetvoh kriyyoh (III.2.126) and samudrbhrd ghah (IV.4.118) in direct violation of
Tantravrttika, nandsramaed., p. 246 11 Sabdakaustubha, p. 6. 12 Sabara on J. S.r X . 8. 4.
10 13 14

SyanaonTaittirlyaSamhit, 1.1.1. Another provision of this rule had been violated in tatprayojako hetus co Pnini, I.4.55.

292 Kshitish Chandra Chatterji

his o w n teaching alpctaram (11.2.3), I.e., in a Dvandva compound, a w o r d w i t h a smaller number of vowels is t o precede. Then again he has his rules samm samm vijyate (V.2.12) and bandhuni bahuvrihou (VI.1.14) in direct violation of all known canons of number and gender, for bandhu substantive is never neuter and sarna is always plural. In the rule gnvbhyo'n (IV.3.57) it is difficult t o understand why he uses the plural number. Many of his rules are ambiguous and ambiguity is the one fault a stra should not labour under. In t h e case of his use of t h e pratyhras an and in we do not know whether he intends the first or the second n. The writings of Ktyyana are also full of such ungrammatical words as ssvata, etc. And Patajali does not lag behind his t w o masters in his use of incorrect words and expressions. In aviravikanyyena [ ' t h e saying about ovi and avika' both avi and avika mean 'goat, 1 but an expression meaning 'goat's flesh ' can be derived f r o m the latter only (i.e., vikam)] grammar requires that the first component of the compound should drop its vibhakti [' ending, 1 i.e., resulting in avyavikanyyena]. In anyath krtv coditam anyath krtv parihrah [' having done it one way, an injunction, having done it another way an exclusion '] the rule anyathaivamkathamitthamsu siddhprayogas cet (III.4.27) requires the addition of namul ['the suffix -am1, i.e., anyathkram krtv . . .]. Then again such expressions as sivabhgavata, sakya cnena svammsdibhir api ksut pratihantum [' it is possible with this with dog-flesh etc. even to tramplestrike'], etc., clearly violate the fundamental rules of grammar. Sabdnm sabdapryanam ['going over the grammatical expression of the grammatical expressions '] is manifestly tautologous.15 Let us now examine in some detail the first few lines of the great commentary of this highest authority on grammar. He begins his book with the sentence atha sabdnussanam [' now (begins) the investigation into grammatical expressions1]. Critics have taken exception to the compound sabdnussanam on the ground that a compound with the objective genitive in such cases is barred by the rule karmani ca (11.2.14). He then asks kesm sabdnm ['whose grammatical expressions']? This sabdnm cannot refer to the word sabda in the compound sabdnussanam for the simple reason that it is a subordinate member of a compound.16 Then again Patajali's division of words into two classes ' laukika' and ' vaidika' is not correct as most of the words and forms current in the language are common to both Vedic and classical Sanskrit, the only distinction between them consisting in the abandonment by the latter of many superfluous Vedic expressions. Further instead of adducing gauh [' cow 1 ], asvah [' horse'], purusah [' man ' ] , host/" ['elephant 1 ], etc., as instances o laukika words, he should have mentioned some words or grammatical forms which are peculiar to classical Sanskrit and which have been described by ancient grammarians as occurring in the bhs only. As instances of Vedic words Patajali mentions san no devir abhistaye [' peace be the goddess to us, to our aid ' (beginning of the Atharvaveda)], etc. words which are also of frequent occurrence in classical Sanskrit whereas he should have mentioned grbhnmi [' I grasp ' ] , dattvya [' having given ' ] , etc., which are purely Vedic. It cannot be urged, T.V., p. 260 See also N.M., p. 143, 11.18ff.
15 16

T.V., p. 261.

293 Critics of Sanskrit Grammar

in his defence, t h a t these are sentences, not w o r d s , and as such differ in t h e i r very nature f r o m classical Sanskrit w h e r e t h e o r d e r of w o r d s is free, whereas the o r d e r of w o r d s is fixed and immutable in t h e Vedas, for, t h e o r d e r of words in popular speech also is fixed in many cases as indrgni [' Indra and Agni ' ] , pitputrau ['father and son ' ] , etc., and secondly grammar lays d o w n rules for words only and has nothing t o do w i t h sentences. Then again strictly speaking it is gvl, gon, etc., that are aukika sobda's and not go and it is these w o r d s that ought t o be explained in grammar, for can you point out a single person w h o even though versed in the Vedas, even t h o u g h conversant w i t h the t e x t ' one should use correct words only,' even though devoted t o the practice of virtuous conduct, even though of refined intellect, even though averse t o t h e practices p r o h i b i t e d in t h e Sastras, even though a Brhmana learned in the Vedas, uses correct w o r d s only? 17 A n d t h e i r examples have been followed by later w r i t e r s on grammar. Bhatti w r o t e an epic poem t o illustrate the rules of grammar and t h e very first w o r d that he composed is grammatically indefensible.

In the line obhn nrpo vibudhasakhah parantapah ['there was a king, a friend of the gods, a destroyer of his enemies'], abht should have been babhva for the king lived long long ago and was beyond the range of vision of the poet (rajas cirtatvat, kaveh paroksatvc cd). At the beginning of his Prodipa on the Mahbhsya, Kaiyata writes bhsybdhih kv atigambhJrah ['the ocean of the commentary is exceedingly deep'] where the epithet atigambhJrah vitiates the compound bhsybdhih for according to the rule upamitam vyghrdibhih smnyprayoge (11.1.56) an object of comparison is compounded with words like vyghra, etc., provided the common quality is not stated in words. Krttikeya Siddhnta begins his commentary18 on the Mugdhabodha thus : yady aham snkrttikeyo mandabuddhis tathpi ca prvakovidapanthnam avalambycikJrsisam dhJr vah prnjaiir yce varam ekam padntatah yadartham yasya sacintya tasya kry pariskriy [' Now I, Sri Krttikeya, slow-witted as I am, wished to follow the path of earlier scholars, I beg you with folded hands, o wise ones, to grant me finally one wish, that the task which I carefully considered for its own sake may be fully accomplished ']. Here prvakovidapanthnam ought to be patham, for pathin at the end of a compound is changed into patha, and vah in the third line ought to be yusmn for these enclitics cannot be used at the beginning of a sentence nor immediately after a vocative which begins the sentence. Similarly Bhattoji while explaining the rule halo'nantarh samyogoh (1.1.7) writes ajbhir avyavahith ['adjoining (to) aj\ i.e., adjoining vowels]. Here ajbhih is a grammatical blunder, it should have been agbhih. No doubt ga might lead to confusion with the instrumental plural of ak. Bhattoji should therefore have written svaraih [' (to) vowels ']. As Ngesa says : atra
kutvam nyyam/ata eva hayavastre 'aco'ksu' iti bhsyaprayoge kutvam drsyate/svarair avyavahit iti tcit vrttih. The same remark applies t o ac-sandhi and anusvrasypy ac-tvt.

As a matter of fact some curse seems to be connected with the study of grammar. Just as in our Catuspthls when students are lectured on the suddhitattva (which prescribes the period of mourning in the case of death of one's kin) some bereavement to
17

N.M., p. 406.

18

MS. in the collection of the Asiatic Society of Bengal.

294 Kshitish Chandra Chatterji

either teacher or pupil is inevitable, just as when lessons are given on the vyadhikarana ['difference of locus'] section of Nyya dealing with abhva [' negation ' ] , some calamity is sure to befall the scholars or their teacher; similarly lessons on grammar are sure t o impair the mental efficiency of the teacher and the taught. It has been said : vrttih stram tila msh kapotrl kodravodanah ojadya pradtavyam jadJkaranam uttamam ['glosses, rules, sesame seeds, beans, an evil bird, ...(?) when this is to be imparted it causes utter stupidity even to a clever person ' ] . 3. As for the utility of the science of Grammar. Grammar serves no useful purpose for the very simple reason that Pnini who may be regarded as the father of this science has not stated any. In the case of Mlmrps and allied sciences the authors themselves have stated in no ambiguous terms the utility of the study of these particular sciences. It cannot be said that the purpose of the study of grammar is too well-known to be stated, for how can it be well-known when we with all our efforts cannot discover it and when no two men agree on what it is.19 Of the four ends of human existencevirtue, wealth, love and salvationnot one can be said to be promoted by the study of grammar. Religious merit accrues from the performance of sacrifices, making of gifts, speaking the truth, offering oblations to the f\re and so on and we learn about these not from grammar but from Vedic injunctions, from theSmrtis of Manu, Yjnavalkya and others, from the practice of the pious and from the epics and Purnas. It is well-known that economics and politics teach one of the best means of acquiring wealth and not grammar, for we find scholars well-versed in grammar suffering the pangs of poverty throughout their lives. Success in love may be secured by a study of Vtsyyana's Kmasstra. It is absurd to suppose that a repetition of tiddhnan, etc., and a knowledge of the distinction between dental ' n ' and cerebral ' n ' will raise a man's worth in the eyes of his beloved. And lastly selfknowledge has been held to be the means of attaining salvation and not grammatical rules. As Sankarcrya says prpte snnihite'tra marae na hi raksati dukm karane. When death which is always near us comes, grammatical aphorisms like dukm karane cannot save us.20 Ktyyana after racking his brains to find out the purposes served by grammar mentions the following : (/') the preservation of the Vedas, (//) ha, (Hi) gama, (iv) simplicity and (v) absence of doubt. /. The preservation of the Vedas. It is not knowledge of grammar but connexion with the preceptor that helps in the preservation of the Vedas. The preceptor gives lessons on the Vedas and as soon as the pupil while repeating the lessons makes the slightest deviation in pronunciation or accentuation, his fellow pupils at once come down upon him and by rebukes, ridicule and repeated instructions make him properly recite the Vedas.21 //. ha. In the books dealing with sacrifices all the sacrifices are not treated in detail ; one sacrifice of each class is dealt with in detail (this is known as prakrti) and only the variations in the case of the others ofthat type are mentioned. These are known as vikrts. In certain cases, in the vikrtiyaga's the vibhakti ['ending'],
19 20

TV., p. 256. N. M., p. 411, N.M., p. 411.

21

T,V., p. 262,

295 Critics of Sanskrit Grammar

etc. have t o be changed. Katyyana and Patajali are of the opinion that one cannot make the necessary changes in the case of the vikrti sacrifices unless he is conversant w i t h grammar. This also is not t r u e . Uhas may relate t o mantras, samans and samskra (different modes of purification). Grammar cannot possibly have anything t o do w i t h the last variety, and as for the first t w o , the practice of the yjnika's or those versed in sacrificial lore will be enough t o instruct us in the changes t o be introduced. So grammar is superfluous. ///. Vedic Texts. There is the Vedic t e x t brhmanena niskrano dharmah sadago vedo dhyeyo jeyas ca which means a Brhmana

should study dharma, i.e., the Veda with the six angas or subsidiary studies and understand them and of the six angas grammar is the most important. We have al ready dealt with the text and shown how sadaga here means the group of sixsruti, inga, etc., and how even supposing it to refer to siks [' phonetics '], etc., vykarana as included in the group of six refers either to the grammatical speculations in Vedic literature or to the Prtiskhyas. It may be pointed out in this connection that it is absurd to say that the text brhmanena, etc., shows one of the purposes of the study of the Vedas. It would be more correct to say the text tells of the reason why we should study the Vedas. v. Simplicity. It is absurd to speak of grammar as the simplest means of acquiring a knowledge of words when it is common experience that even after years of patient labour one fails to master the science. One feels inclined to suspect that what the Vrttikakra meant was gaurava [' heaviness'] or cumbrous means and has euphemistically used the word ghava ['simplicity'] (by what is known as vipantalaksana [' metaphorical usage of opposites ']).22 v. Removal of doubts. It has been said that when a doubt arises as to the meaning of certain words and expressions, a knowledge of grammar helps us to take into account the accents, etc., and to determine the correct meaning. This is hardly correct, for it is well-known the various doubts that every sentence gives rise to are dispelled not by grammar but by MTmms.23 Grammar deals with words only, whereas Mlmms deals with words as well as sentences. When we come across the Vedic text akth sarkar upadadhti, i.e., anointed pebbles are to be placed and are in doubt whether the pebbles are to be anointed with ghee or oil, it is Mlmms that comes to our help and tells us that the subsequent mention of ghee in the arthavda makes it quite clear that in the present case it is ghee that has to be used as the ointment and not oil. And what do the grammarians themselves do when they are in doubt as to the precise meaning of an expression in their precious rules? Do they not rely on the authoritative explanations on the point? What harm is there, therefore, in relying on traditional explanations for determining the meaning of such compounds as sthlaprsatJ [' coarsely spotted (cow) ' ; when accentuated differently, 'coarse, spotted cow'], etc.? Where then does grammar come in? Thus to say that raks [' preservation ' ] , etc., are the purposes served by a study of grammar is an insult to the intelligence. We may summarily dismiss the further reasons adduced in support of the study of grammar, for since these principal purposes have
22

N.M.,p.412.Tantravrttika, pp. 265-66. VI. 1-2.

23

N.M., p. 412.

296 Kshitish Chandra Chatterji

failed t o establish the utility of grammatical studies, it would be like leaning on a broken reed t o expect anything f r o m the subsidiary reasons.24 It has been said by Bhartrhari that the t r u e nature of sounds (words) can only be learnt w i t h the aid of grammar (tattvvabodhah sabdnm nsti vykaranttate). This is entirely w r o n g . W h a t he ought t o have said is tattvvabodhah sabdnm nsti srotrendriy-

ttate ['the true nature of words can only be learnt with the aid of the sense of hearing'] (T.V., p. 260). 5. Lastly the rules of grammar are, in many cases, in conflict with the Vedas. To explain the formation of the word kleya [name of a chant in the Smaveda] Pnini has laid down the rule kalerdhak (IV.2.8) which means that the affix dhak is to be added to the word kali in the sense of drstam sarna [' intuited chant1]. Similarly vmadevya is formed according to the next rule vmadevddyaddyau ca (IV.2.9), in the sense vmadevena drstam sarna ['chant intuited by Vmadeva1]. In the Vedas, however, we find yad akiayat tat keyasya kleyatvam [' that which drives forward is the essence of kleya'] from which it is clear that kleya is not derived from the base kali but from the causal verb ko//-. Similarly we find po rtviyam rchams tsm vyuh prste vyavartata/tato vmam vasu samabhavat/ tan mitrvarunv apasyatm tv abrtm vmam v dam devebhyo1jani tasmd vmadevyam ['the waters came in proper time and the wind set on their back. Then Vasu joined the lovely one (yma). Mitra and Varuna saw him and they said, verily, this Varna is born from the gods. Hence (the name) vmadevya ']. Thus we see that the waters in their monthly courses had connection with the air and gave birth to treasure which accrued to the gods. Hence the name vmadevya (vide Maykhamlik on Sstradipik, 1.3.9). Similarly for the etymology of the word ypa. In the Undistra's the word yupa is derived from the root yu with the suffix -p (kuyubhyn ca 111.27). In the Vedas however it is derived from the root yupyad ayopayat tad ypnm ypatvam [' what conceals that is the essence of yupas ']. And so on. Now since in all cases of conflict between asrut/ ['revelation'] and usmrti ['tradition '] it is the smrti that must go to the wall, we must reject grammar. Thus it has been said25Only those who are possessed by the devil, or afraid of the royal rod or cursed by their parents need take pains in the study of grammar. The Nyayamajar concludes the Mlmmsaka's diatribe on grammar with the following words :
sarvath durvyavasthitam sabdnussanam / y as ca vykhytrnm uktanuktaduruktanirksanaprayatno yas ca vcakamtrvarndhikyamisapurah saralaksanaparicodanaprakro yac cedam vykhytrvacanam 'ha na bhavaty anabhidhnt'iti, yac ca vyptisiddhau saralam upyam apasyatm krtiganavarnanam yac ca pade pade bahulavacanam tat sutarm aparisuddhim anussanasya darsayatti I anye tu sobheti cJrnam iti na yti pratibhettum iti mtur anuharatti phalinabarhinau dhsJti kndislka iti bhrjisnur iti ganeya iti varenda iti laksyasamgrahabahiskrtasmrtisandehaviparyayapratipdakatvaiaksanaskhalitam vip\atam pninitantram manyamns tatra mahntam ksepamatnisuh sa tu sthlodaraprya itlha granthagauravabhayn na likhyate / nanu yadi laksanasya pranet pninir na samyag darsayaty atra vivaranakrs ca ntinipunattasah, kmam anyah sdkrtabuddhir bhavisyati vrttikrs ca praudhatarattastayo bhavisyanti tebhyah sabdalaksanam aviplutam avabhotsymaha iti /
24

N.M., p. 412.

25

N.M., p. 418.

297 Critics of Sanskrit Grammar

naitad asti, tesm apy abhiyuktatarh kecit preksanta eva dosam tesm apare tesm apy apare / tad evam anavasthprasangn nsti nirmalam anussanam iti klesyaiva vykarandhyayanamahvratagrahanam / tath ca brhaspatih pratipadam asakyatvl laksanasypy avyavasthnt tatrpi skhalitadarsand anavasthaprasangc ca marannto vydhir vykaranam iti // The science of grammar is not final by any means. The fact that commentators try to find out what has been said, what has been omitted and what has been wrongly asserted, the fact that faults are found with the rules under the pretext that a word contains more mtrs or more letters than are absolutely necessary, the words of the commentator 'the particular thing prescribed does not take place here because such a word is not in use,' the mention of krtigana's (i.e., lists which are not exhaustive but which have to be supplemented from usage) as no easy way is found for including all words, the use of the word bahulam ['variously'] at every stepall this shows clearly and distinctly the utter inadequacy of the science of grammar. Others have found fault with the science of Pnini thinking it to be worthless as it labours under the defects of ambiguity and falsity inasmuch as it does not include within its scope such expressions as sobh ['splendid '], drnam ['conduct'], noytipratibhettum, mtur anuhrati [' he resembles his mother'], phalinabarhinau dhasi, kndiiika [' running away'], bhrojisnu ['shining'], ganeya ['calculable'], varenda, etc. That is patent on the surface and so is not discussed here for fear of increasing the bulk of the volume. Well should you say, ' If Pnini, the composer of the rules does not expound properly and if the commentators are also far from being very clear-sighted in these matters, it must be admitted that there will be others whose intellect is as clear as the needle, and the writers of the vrttis will also be men of clearer insight, from them we shall learn the rules for theformation of words correctly; ' that is not possible. People more learned than they find fault with what they say, and these in their turn are subjected to criticism by people more learned and they by others. So.since there is thus no finality there can be no instruction of words free from defects and consequently the taking of the great vow of the study of grammar merely brings on distress. As Brhaspati says : As it is not possible to lay down rules for every word, as the rules again are not perfect, and as even in them slips and errors are found and consequently there is no finality in the science, grammar is afatal disease, which baffles the skill of the physician at every step, the symptoms of which again are not permanent, and in the diagnosis of which the physician is apt to make mistakes.

Paul T h i e m e ( b o r n 1905)

Paul Thieme, who studied at Gttingen with the German Sanskritist E. Sieg, considers himself connected through upadesaparampar (uninterrupted tradition of teaching) with Kielhorn (Thieme 1957b, 47). Thieme spent the years from 1933 to 1935 in India, where he studied grammar and was taught in particular by Pandit Kamalknta Misra, who was then at Deraganj near Allahabad. Thieme taught Sanskrit at Gttingen, Breslau, Halle, Frankfurt, and Yale (from 1954); at present he is back in Tbingen. In 1935 he published Pnini and the Veda: Studies in the Early History of Linguistic Science in India, a rare book printed and published by the Globe Press at Allahabad. With this publication Thieme joined the ranks of the great Pniniya scholars of the past; he also re-embodied the forcefulness and argumentativeness of his predecessors. In his later work, he continued to concentrate on the two areas mentioned in the title of his book, the Sanskrit grammarians and the Veda. In the same year that Pnini and the Veda was published in India, there appeared a long study in the Nachrichten von der Gesellschaft der Wissenschaften zu Gttingen, Philologisch-historische Klasse (Neue Folge, Bd. 1, Nr. 5,1935,171-216), entitled "Bhsyazu vrttika 5 zu Pnini 1.1.9 und seine einheimische Erklrer. Ein Beitrag zur Geschichte und Wrdigung der indischen grammatischen Scholastik.*' This study, included in this volume, shows how a grammatical discussion on the homorganic status of r and / was developed by the commentators through the centuries. The primary issues are whether the rule which is under discussion is adequate to cover the relevant facts of the language and whether it is required, or whether other rules which have already been adopted suffice to account for these facts. Here descriptive adequacy is not considered enough; the results have to be obtained in the correct manner. Thieme notes how the grammarians' technical use of certain terms captures this difference (prpnoti 'obtains (but) incorrectly,' sidhyati 'obtains correctly'). The discussion is carried out in two rounds and made explicit in Thieme's translations. In the first round it takes place between an opponens, a defensor and the crya who "settles the doubt"; in the second, between the crya and the ekadesin, who voices new criticisms but who is only partly familiar with the facts. The examples and counter-examples are on a par with such forms as svalid dhaukate ' t o come near licking like a dog,' to which Bloomfield had taken such exception (see page 270). Thieme notes that even a pandit "wallowing in Sanskrit" would rarely have occasion to say hotrkrah 'the vowel /ofthe Hotr priest' or upalkrlyati 'he wants in addition an /.' But he rightly enlarges upon the relevance of the discussion for grammatical theory. The discussion starts with Patajali's Mahbhsya and culminates in Ngojibhatta's Uddyota and Laghusabdendusekhara (LSS), works which constitute the foundation of the so-called new school of grammar in the eighteenth century. Thieme translates the relevant passages of the difficult LSS in full. The intervening stages, representing the old school, are the Ksik (seventh century), which Thieme describes as a practical but not always consistent manual; Kaiyata's Pradpa (not earlier than the eleventh century), "pedantically meticulous"; Haradatta's Padamajaand Bhattojdksita's Sabdakaustubha (seventeenth century), the latter full of artificial anxieties; and Bhattojidksita's Siddhantakaumud with its "subtle ingenuity."

A. Bhasya zu v r t t i ka 5 zu Pnini I. 1.9 und seine einheimischen Erklrer (1935) Paul Thieme

Nachdem Katya/ana Panini's Definition : tulyosyoprayatnam savarnam (1.1. 9) so gendert, beziehungsweise interpretiert hat, da sich der Sinn ergibt: Was die gleiche Artikulationsstelle 1 ) und die gleiche Artikulationsweise im Munde2) wie etwas anderes hat, heit mit Beziehung auf dieses 'gleichlautig'", fgt er ein Postulat hinzu : rkralkrayoh savarnavidhih (vrtt. 5 zu Pn. 1.1. 9)3). An dieses vrtt. schliet Patajali eine Errterung, die ich nebst ihren Reflexen in der einheimischen Literatur erlutern mchte, da sie ein typisches Beispiel der Darstellungs- und Argumentationsmethode der indischen Grammatiker liefert, als wie gering immer sich ihr sachliches Interesse herausstellen mag. Sie zerfllt in drei Teile: I. Paraphrase (vkydhyhra) des vrtt. und Beispiel (udharana). II. Diskussion, ob fr das vrtt, eine 'notwendigmachende Veranlassung' (proyojano) besteht. Die Teilnehmer an der Diskussion nenne ich in der folgenden bersetzung 'Defensor [des vrtt.] 1 und 'Opponens [des vrtt_.]\ den Verknder der [Zweifel-JBeschwichtigung (samdhna), 'crya'. Der Acrya gibt mit dem samdhna eine vollstndigere Erklrung des vrtt., als sie in I angedeutet st. III. Diskussion, ob sich bei der Annahme des samdhna, d. h. des konsequent interpretierten vrtt., Fehler (dosa) ergeben. Ich nenne den Verteidiger des samdhna 'Acrya', seinen Opponenten, 'Ekadesin'. Dem Ekadesin ist nur ein Teil der Wahrheit bekannt, und er befrchtet deshalb, da Fehler resultieren, wenn man das vrtt. anwendet. Er wird schlielich mit einer richtigen Antwort widerlegt, die er selbst auf eine Frage des Acrya erteilt 4 .
1 Die Artikulationsstellen sind: Hals (kantha), Schlund (tlu), Kopf[mitte] (mrdhan), Zhne (danta) usw. 2 Die Artikulationsweisen im Munde: verschlossen (sprsta), halbverschlossen (satsprsta), offen (v'tvrta), bedeckt (samvrta) usw. Die Verschlulaute sind 'sprsta', die Halbvokale 'isatsprsta', die einfachen Vokale 'vivrta', kurzes a 'samvrta' usw. Vgl. Pat. zu vrtt. 3 zu 1.1.10 und zu vrtt. 10 zu SS. 3, 4. Von diesen 'Artikulationsweisen im Munde' sind zu unterscheiden die auerhalb des Mundes gelegenen, wie Tonhaftig-undTonlosigkeit.Aspiriertund Unaspiriertheit usw. Vgl, Pat. zu vrtt. 2 zu 1.1. 9. 3 [(Wenn die in den vorhergehenden vrtt. geforderte Formulierung von Pn. 1.1.9 angenommen wird, dann) mu ausdrcklich gelehrt werden, da] fr 'gleichlufige' Laute gltige Regeln [auch] fr r und / [gltig sind]". Zur Ausdrucksweise vgl. z. B.

vartt. 2 zu SS. 3, 4: plutydisv ajvidhih

[(Wenn die in vrtt. 1 geforderte Hinzufgung von t gettigt wird) dann mu ausdrcklich gelehrt werden, da] fr 'ac' gltige Regeln [auch] fr 'plutierte' [Diphthonge] [gltig sind]". vrtt. 13 zu SS. 3,4
tulyarpe samyoge dvivyajanavidhih

[(Wenn Teile eines Lautes durch Nennung des ganzen Lautes nicht mitgenannt werden) dann mu ausdrcklich gelehrt werden, da] fr zwei aufeinanderfolgende Konsonanten gltige Regeln [auch] fr eine gleichgestaltige Konsonantengruppe (z. B. tt, kk) [gltig sind]". 4 Die Abtrennung von Rede und Gegenrede ergibt sich unzweideutig aus dem Text und ist von der einheimischen Tradition bereits richtig durchgefhrt. Da meine Benennung derTeilnehmerwillkrlich ist, habe ich schon angedeutet: Sie dient lediglich praktischen Zwecken.

300 Paul Thieme

bersetzung des Bhasya zu vartt. 5 zu 1.1. 9. I. (Paraphrase und Beispiel) ,,Der Name ' gleich lautig' mufr r und / [ausdrcklich]5 gelehrt werden. hotr + Ikrah > hotrkrah.
II. (Erste Diskussion) Opponens: Was ist die notwendigmachende Veranlassung [fr diese spezielle Angabe]" ? Defensor: [Dies mu gelehrt werden,] damit Substitution des langen Vokals [in hotrkrah aus hotr - r Ikrah] nach akah savarne dirghah (Pn. 6.1.101 )6 statthabe". (Erstes Argument.) Opponens : ,,Dies (i. e. die Substitution des langen Vokals in hotrkrah aus hotr-\- Ikrah) [bildet] nicht eine notwendigmachende Veranlassung. [Denn Ktyyana (selbst)] wird lehren : 'savarnadirghatve rti *r7 vvacanam, Iti */ 7 vvacanam8 (vrtt. 1 und 2 zu 6.1.101)" 9 . (Erstes Argument.)
5 Kaiyata: rkaralkarayor iti sthanabhedn na prpnotity rambhah Dies

vrtt. mu speziell unternommen werden, da [fr rund/ der Name 'gleichlautig'] sich nicht ergeben wrde, insofern sie verschiedene Artikulationsstellen haben". r und / haben die gleiche Artikulationsweise im Munde: vivrta. Die Artikulationsstellen jedoch sind beziehentlich 'Kopf[mitte]' (mrdhan) und 'Zhne', 6 ,,Nach einem ak (o, /', u, r, I nebst ihren Lngen) vor einem gleichlufigen Laut wird der lange Vokal allein fr den vorhergehenden und den folgenden Laut substituiert". 7 Ich gebe die hier zitierten Laute durch *r und */ wieder, da sie weder mit gewhnlichem kurzen, noch mit gewhnlichem langen r und / identisch sind, wie sich aus der Diskussion des Bhsya selbst und auch aus der Behandlung der einheimischen Tradition deutlich ergeben wird. Trapp, Die ersten fnf hnikas des Mahbhsyam ins Deutsche bersetzt und erklrt (Leipzig 1933) S. 264 (nebst Anm. 180,181 ) hat das nicht bemerkt und folglich die ganze Diskussion nebst Kaiyata miverstanden. Auch hat er bersehen, da Kielhorn in den zitierten vrtt. nicht die gewhnlichen Zeichen fr r und J, sondern eigens geschaffene aksara verwendet. 8 Zu Pn. 6.1.101 mu gelehrt werden, da vor kurzen r entweder *r, und vor kurzem / entweder */ substituiert wird [oder der 'lange' Vokal]".

9 Da / niemals ans Ende eines Wortes oder Namens zu stehen kommt, mu sich das zweite vrtt. auf einen Fall wie hotr -|- kroh beziehen. Es lehrt demnach die Formen hot*lkrah und hotrkrah fr hotr-\- Ikrah und wir brauchen nicht vrtt. 5 zu 1.1. 9 um hotrkrah nach 6.1.101 zu bilden, wie Defensor (1. Arg.) behauptet hat. Kaiy.: . . . tatra Ivvacanam ty atra dirgha ty anuvartate. tatra Iti *lsabde vikalpite 'prpta eva pakse dlrgho bhavisyati,, In dieser [Hinzufgung Kty.'s zu 6.1.101] gilt in dem Satz 'es mu gelehrt werden, da entweder */ substituiert wird' fort: 'der lange Vokal' (aus 6.1.101 akah savarne dirghah) [soda zu ergnzen st: 'oder der lange Vokal']. Indem also vor kurzem / [die Substitution von] */ als A l t e r n a t i v e gegeben st, wird der 'lange' Vokal beliebig substituiert, obgleich ersieh [ohne dies] nicht ergeben wrde". Zur Entscheidung der Frage, welcher 'lange' Vokal substituiert wird, haben wir uns an Pn. 1.1. 50 sthne 'ntaratamah ,,im Falle einer Substitution tritt der nchstbenachbarte Laut an Stelle [des Originals]" zu halten. 'Nchstbenachbart' dem Original r + / (in hotr-{-Ikrah) wrde entweder ein langes f oder ein langes [sein. Der letztere Laut fehlt der Sprache. Es kann also nur r substituiert werden. Kaiy. : sa ca bhavann Ivarnasya dirghbhvd rvarnasyntaratamyd dirgha rkro bhavisyati ,,Wenn nun der 'lange' Vokal substituiert wird, so wird, da ein zu / gehriger 'langer'

301 Bhsya und Erklrer

Defensor: [Gut. Du magst recht haben. Die Substitution des 'langen' Vokals (f) in hotr-\-kr ah > hotrkrah ergibt sich aus dem vrtt. zu 6.1.101. Sie kann also nicht die notwendig machende Veranlassung fr unser vrtt. sein. Aber es ergibt sich eine andere notwendig machende Veranlassung:] Da diese [alternativische Substitution des 'langen' Vokals (r) und des */ vor /, die durch das vrtt. zu 6.1.101 erreicht ist] vor (nach unserm vrtt.) 'gleichlautigem' Laut (i. e. wor / nach r) statthabe, nicht aber in den folgenden Fllen: dadhy krah, madhv Ikrah (aus dadhi + Ikrah, modhu -f Ikrah, wo / dem vorausgehenden Laut unter keinen Umstnden 'gleichlautig' ist)10. (Zweites Argument.) Opponens: Was [Kty. als] 'savarnadlrghatve rt [gelehrt hat], das werde ich als 'rtah' (nach kurzem r") lehren11 (. e. : ich werde vrtt. 1 zu 6.1.101 formulieren : rta *rvvacanam). Und dann [werde ich lehren] Iti (*lvvacanam). 'Auch vor / wird entweder */ substituiert [oder der lange Vokal]'. (Das gilt jedoch) nur [unter der Voraussetzung] 'rtah' (nach r) 1 2 ". (Zweites Argument.) Defensor: Diese [alternativische Substitution des 'langen' Vokals und des *r bezw. */ wor r und /, die in den vrtt. zu 6.1.101 gelehrt wird] mu nicht [ausdrcklich] gelehrt werden [wenn Vokal nicht vorhanden, das 'lange' r substituiert werden, da es dem r [in hotr -f- Ikrah] nchstbenachbart [unter den 'langen'-Vokalen] ist". 10 Kaiy.: . . . yady avidhya savarnasamjnm tad ucyate 'gmtrasya Iti tatkryam syt: dadhy Ikra ti. tasmd vidhey savarnasamjn ,,Wenn dieses [vrtt. zu 6.1.101] gelehrt wird, ohne da man den Namen 'gleichlautig1 [fr r und I] gelehrt hat, dann wrde die durch dieses vrtt. veranlate Operation schlechtweg fr 'ak' vor kurzem / gelten, z. B. [fr / -f / in] dadhy Ikrah etc. Deshalb mu der Name 'gleichlautig' [fr r und /] gelehrt werden". Trapp meint, der Defensor wolle einem andern Zweck des vrtt. zu 6. 1.101 angeben. Abgesehen davon, da die Diskussion sich darum dreht, einen Zweck (eine 'notwendigmachende Veranlassung') fr vrtt. 5 zu 1.1. 9, und nicht fr irgend ein anderes vrtt., zu finden, vermag man nicht einzusehen, wie die Veranlassung fr das vrtt. zu 6.1.101 sein knnte, die von ihm gelehrte Operation auf bestimmte Flle einzuschrnken, ohne da diese Einschrnkung in der Formulierung enthalten wre. 11 Nicht 'werde ich beifgen' (Trapp), und nicht rtah gen.: 'fr r' (Trapp), sondern abl.: 'nach r'. Vgl. Sabdakaustubha zu Pn. 1.1. 9 . . . rty ity apaniya rta iti pancamyantapthenaiva sarvasamajasyam Alles ist schon in Ordnung, wenn man 7t/" [aus dem vrtt,] e n t f e r n t und den A b l a t i v Vtah1 einsetzt ('liest')". 12 Kaiy. kommentiert die Form, in der der Opponens die vrtt. zu 6.1.101 akah savarne dirghah lehrt. 1. rta *rvvacanam es mu gelehrt werden, da nach r entweder *r substituiert wird (oder der 'lange' Vokal)". Kaiy.: rkrasya rkra eva savarna ity rty eva bhavisyati,, [Die Substitution] wird statthaben nur vor r (obgleich das nicht ausdrcklich gelehrt wird), da eben nur r einem r 'gleichlautig' st" (i. e. er lt 'savarne1 aus dem stra fortgelten). 2. ti*lvvacanam ,,es mu gelehrt werden, da vor / entweder */ substituiert wird (oder der 'lange' Vokal)" nur nach r ('rtah' gilt fort aus dem ersten vrtt.)". Kaiy.: dam csavarnrtham. tena Iti rpadvayarn siddham Und dies [vrtt.] wird fr ungleichlautige Laute [gelehrt]" (i. e. er lt savarne nicht mehr fortgelten). Folglich sind die zwei Formen vor / (nach r) korrekt". . . . rkre . . . hotr krah, hot* rkra iti rpadvayam. kre . . . hotrkrah, hot*lkra iti rpadvayam Vor r (i. e. in hotr-f- rkrah) ergeben sich zwei Formen: hotrkrah und hot*rkrah, und vor / (i.e. in hotr + Ikrah) ergeben sich zwei Formen: hotrkrah und hot*lkrah".

302 Paul Thieme

unser vrtt. gelehrt w i r d , da die betreffenden Operationen sich in diesem Falle von selbst ergeben] 13 . (Drittes A r g u m e n t ) . Opponens : Diese [alternativische Substitution des langen Vokals (f) und des * r und */] mu unter allen Umstnden (i. e. ob man v r t t . 5 zu 1 . 1 . 9 ablehnt, w i e der Opponens, oder annimmt, wie der Defensor) gelehrt werden. [Die Ansicht, da w i r bei A n nahme von v r t t . 5 zu 1. 1. 9 die v r t t . zu 6. 1. 101 streichen knnen, ist unrichtig.] [ N a c h 6.1.101 w i r d der 'nchstbenachbarte' ' l a n g e ' Vokal substituiert. N u n ] w i r d gelehrt, da ein 'ac1 genannter Laut je nachdem er die Lnge von u, oder l hat, 'kurz', 'lang' oder ' p l u t i e r t ' heit ( 1 . 2. 27). Und weder * r noch */ heit 'ac'u [kann demnach auch nicht 'lang' heien, und folglich nicht nach 6.1.101 substituiert werden, obgleich es dem Original 'nchstbenachbart' i s t ] " . (Drittes Argument.) Defensor: Ich werde [ausdrcklich] lehren, 15 da *r und */ 'ac heien 16 . [Dann kann ich ihre Substitution nach 6.1.101 auch ohne v r t t . 1 , 2 erreichen.] Und zwar mu das unter allen Umstnden (/. e. ob man die zwei vrtt. zu 6.1.101 ablehnt, wie der Defensor, oder annimmt, wie der Opponens) gelehrt w e r d e n . [ D i e Ansicht, da w i r * r und */ nicht 'ac1 zu nennen brauchen, wenn w i r die vrtt. zu 6.1.101 lehren, ist unrichtig.] [Es mu gelehrt werden,] damit die Substit u t i o n des plutierten Vokals statt hat. ( Z . B. :) hotr-\-rkrah> hot*rkrah, (nach Pn. 8. 2. 86 w i r d gebildet:) 'hotrlkraV ; hotr + lkrah>hot*!krah, (nach Pn. 8. 2. 86 w i r d gebildet:) 'hotllkral'*7". (Viertes Argument.)
13 Trapp miversteht den Defensor wiederum. Er meint, es handle sich darum, die vom Opponenten vorgeschlagene nderung des vrtt. zu 6.1.101 zurckzuweisen, wobei dann das folgende Argument vllig seinen Sinn verliert. Die Frage st vielmehr, die Formen r, *r, */fr r -\-r und r-\-\ mit Hilfe des vrtt. 5 zu 1.1. 9 auch ohne vrtt. 1, 2 zu 6.1.101 zu erhalten. Kaiy. : dvv apy etau dirghau *r */ iti. tatra rkre kadcid r iti dlrghah kodcid *r iti. kre 'pi r */ ity etou bhavisyatah satyrp savarnasamjnym iti bhvah Die Meinung ist: Diese [in vrtt. 1, 2 zu 6.1.101 genannten Laute] *r und */ sind alle beide 'lang'. Insofern sich das so verhlt (tatra), wird vor r manchmal der 'lange' Vokal r und manchmal [der 'lange' Vokal] *r (als 'nchstbenachbarter' 'langer' Vokal nach 6.1.101) substituiert. Und auch vor / werden r und */ (nach 6.1.101) substituiert, unter Voraussetzung, da der Name 'gleichlautig' [fr r und /] statt hat (und somit 6.1.101 auf r + / Anwendung finden kann)". 14 'ac' heit ein Laut entweder, wenn

er in SS. 1-4 aufgefhrt ist, oder, wenn er einem der dort aufgefhrten Laute 'gleichlautig' ist (nach 1.1. 69). Es ist demnach klar, da die Laute, von denen hier die Rede st, weder mit gewhnlichem r und /, die SS. 2 tatschlich aufgefhrt sind, identisch, noch ihnen 'gleichlautig' sein knnen. Das heit, sie haben entweder andere Artikulationsstellen, oder eine andere 'Artikulationsweise im Munde.' Die Entscheidung dieser Frage behalte ich fr spter vor, wie auch die Behandlung der Bemerkungen Kaiy.'s die Trapp (Anm. 185 zu S. 265) grausam entstellt hat. 15 Nicht 'ich behaupte' (Trapp). 16 Kaiy.: saty actvedrghasamjanayor bhavisyati ,,Wenn die beiden [Laute *r und */] 'ac' heien, dann wird der Name 'lang' fr sie statt haben." 17 Da nach Pn. 1. 2. 27 ebenfalls nur ein 'ac' den Namen 'plutiert' empfngt, ist die Substitution von berlangem *r und */ nach 8. 2. 86 nur unter der Voraussetzung mglich, da diese berlangen Laute 'ac' heien. Kaiy.: tvaypi *rvvacanam *lvvacanam iti bruvat 'ctvam anayor vaktavyam. anyath vidhnamtram

303 Bhsya und Erklrer

(Hiermit st die Diskussion abgeschlossen.) Der Opponens will vrtt. 5 zu 1.1. 9 ablehnen, aber die zwei vrtt. zu 6.1.101 beibehalten.18 Der Defensor will vrtt. 5 zu 1.1. 9 biebehalten, aber die zwei vrtt. zu 6.1.101 ablehnen. Beide, Defensor und Opponens, sind gezwungen ausdrcklich zu lehren, da *r und */ 'ac' heien. Ihre Vorschlge scheinen sich ungefhr die Wage zu halten19, da vrtt. 5 zu 1.1. 9(21 Moren) und die vrtt. zu 6.1.101 (24 Moren) beinahe gleich 'schwer' (guru) sind.) [Zweifelsfrage (ksepa) :] ,,Was aber ist unter diesen Alternativen {otro) besser"? [Zwejfel-Beschwichtigung (samdhna) :] Acr/a: ,, Est ist besser, lediglich den Namen gleichartig [fr r und /]zu lehren [als lediglich die vrtt. zu 6.1.101]. ,, Denn dann ist nicht nur die Substitution des langen Vokals20 [r und */fr r + / ] schon (ohne weiteres vrtt.) in korrekter Weise erreicht (nach 6.'1.101 )21. ,.sondern es wird (dann) auerdem durch Nennung von r [in der Grammatik] Nennung von / impliziert22 (nach Pn. 1.1. 69). [Zum Beispiel :] Nach rty'akah (Pn. 6.1.128)23 [erhlt man] khatva rsyah (fr khatva -{-rsyah), mla rsyah (fr ml -{-rsyah). [Bei Annahme des vrtt. 5 zu 1.1. 9] ergibt sich auch das folgende in korrekter Weise : khatva Ikrah, mla Ikrah.
anayoh syat nackaryam plutah. sati tv octve tbhym trimtrayor api svarnyd grahane sati plutasarpjn bhavisyatiti plutasiddhih Auch du, der du die vrtt. zu 6.1.101 lehrst, mut lehren da *r und */ 'ac' heien. Wenn das nicht geschhe, dann wrde fr sie lediglich statt haben, was gelehrt ist (da sie nmlich fakultativ fr r -\-r und r -f- / substituiert werden), aber nicht die fr ein 'ac' gltige Operation, nmlich [die Substitution] des 'plutierten' Vokals. Wenn sie jedoch 'ac heien, dann werden sie auch die dreimorigen Laute (*rl und */?) einbegreifen, dasie 'gleichlautig' sind (nach 1.1. 69). [Diese werden also ebenfalls 'ac' heien und] folglich wird der Name 'plutiert' [fr sie] statt haben, und da der Name 'plutiert' statt hat, ergibt sich in korrekter Weise die Substitution des 'plutierten 1 Vokals (nach 8. 2. 86)". In der Fassung: rta *rvvacanam, Iti *tvvacanam. Gegen die nderung des Wortlauts st nichts einzuwenden, da sie die Zahl der Moren nicht vermehrt. 19 Kaiy.: ekaikasmin samhitym . . . rpadvayam sdhyam. taccobhayath sidhyati . . . iti smyam ubhayoh paksayoh Vor r sowohl als vor / mssen zwei (Substitutions)formen erreicht
18

werden, wenn Sandhi statt hat. Und diese ergeben sich korrekt auf beiderlei Weise. Deshalb sind die beiden Alternativen im Gleichgewicht". 20 Nag.: isatsprstavivrtarpodlrgho varna ity arthah. Der Sinn ist: 'Substitution des langen Vokals in der Form 'isatsprsta' (*/) und 'vivrta' (*r)". ber die Ausdrcke Jsatsprsta und vivrta s. u. 21 Beachte, da der crya mit dem Ausdruck dlrghatvam caiva hi siddham bhavati die Behauptung des Defensors (1. Arg.) akah savarne dirgha iti dirghatvam yath syt, die bereits im 'Beispiel' (I) angedeutet war, aufnimmt, jedoch in einem weiteren Sinne, indem nunmehr nicht nur hotrkrah (fr hotr + Ikrah) sondern auch hot*ikrah (fr hotr + Ikrah) 'notwendigmachende Veranlassung' ist.
22

Hierdurch ergibt sich eine weitere Serie von Veranlassungen, wodurch erst die Schale wirklich zu Gunsten des Defensors sinkt. 23 , , V o r k u r z e m r nach einem 'ak' wird ein einzelner Laut nicht substituiert, aber der 'kurze' Vokal [fr etwaiges 'langes' 'ak'], nach der Lehre des Skalya".

304 Paul Thieme

Nach v supy Apisaleh (Pn. 6 . 1 . 92)24 [erhlt man] uparkrlyati und uprkrlyati (fr upo+ rkriyati). [Bei Annahme des vrtt. 5 zu 1.1. 9] ergibt sich auch das folgende in korrekter Weise: upalknyati und uplkriyati (fr upa + IkrJyati)". MI. (Zweite Diskussion) Ekadesin : ,,Wenn denn durch Nennung von r Nennung von / impliziert ist (und wir also berall, wo Pn. V nennt, zu ergnzen haben ' und / '), ergibt sich nach ur an raparah (Pn. 1.1. 51)25 [flschlich]26, da auch ein fr / [substituiertes a, /, u] von r gefolgt ist." (Erste Befrchtung.) crya: ,,lch werde [ausdrcklich]27 lehren, da ein fr / [substituiertes a, i, u] von / gefolgt ist. Und zwar mu das unter allen Umstnden (i. e. ob das vrtt. 5 zu 1.1. 9 besteht oder nicht) gelehrt werden. [Die Befrchtung, da die Annahme des vrtt. ein falsches Resultat verschulde, ist nichtig.] Wenn r und / n icht 'gleich laut ig' heien, wird [meine Regel] dazu dienen eine positive Vorschrift aufzustellen28 (in diesem Falle wrde ein Substitut fr / eben zunchst nur a, i oder u lauten, und es mte gelehrt werden, da ein /zu folgen hat : al, il, ul), und wenn sie 'gleichlautig' heien, wird ebendieselbe [Regel] dazu dienen, rzu verhindern29 [in diesem Falle wrde ein Substitut fr / zunchst ar, ir, ur lauten, und es mte gelehrt werden, da es nicht so, sondern al, il, ul zu_ lauten hat)". Ekadesin : [Gut. Da der Acrya ausdrcklich lehren wird, da a, i, u als Substitute fr / von / gefolgt werden, ergibt sich bei Annahme unseres vrtt. nicht flschlich nach 1.1. 51, da sie von r gefolgt sind. Und da der Acrya gezeigt hat, da er es, auch wenn unser vrtt. nicht gilt, lehren mu, kann ich auch nicht einwenden, da vrtt. 5 zu 1.1. 9 eine 'Erschwerung' des Wortlauts der grammatischen Regeln bedingt.] Zur folgenden Rege! denn: rasbhym no nah samnapade (Pn. 8. 4.1)30 ist Nennung von /-gefordert31 (vrtt. 1 zu 8. 4.1 rasbhym natva rkragrahanam32), um [Substitution von n fr n in] mtrnm, pitrnm zu erreichen. Diese [Substitution] ergibt sich (bei Annahme unseres vrtt., auf Grund dessen durch Nennung von rauch / impliziert ist) [flschlich] auch
24

Nach einem Prverb vor einer mit kurzem ranlautenden denominativen Verbform wird entweder vrhi oder guna allein substituiert, nach der Lehre des pisali". 25 ,,Wenn a, i oder u fr r substituiert werden, sind sie [je] von r gefolgt". 26 prpnoti 'erbigt sich flschlich' st im Bhsya und der grammatischen Literatur stets das Gegenteil von sidhyati 'ergibt sich in korrekter Weise'. na prpnoti 'ergibt sich flschlich nicht', na sidhyati 'ergibt sich nicht in korrekter Weise'. 27 Zu Kaiy.'s Deutung von vaksymi s. u. 28 'vidhyartham'. Unrichtig Trapp: 'Zur Regelung der Vorschrift'. 29 i. e. eine spezielle Ausnahme

(apavda) zu einer generellen Vorschrift (utsargo) aufzustellen. Ein 'vidhi' lehrt eine Operation, die sich flschlich nicht ergibt; in Sanskrit mag man 'vidhi' erklren mit 'aprptavidhnam'. Ein 'apavda' verhindert eine Operation, die sich flschlich ergibt; in Sanskrit mag man 'apavda' erklren mit 'prptabdhanam'. 30 ,,Nach r und s wird fr n, wenn es sich im gleichen W o r t befindet, n substituiert" [ob es nun unmittelbar folgt, oder durch einen Vokal, Halbvokal. Guttural, Labial usw. getrennt ist: 8. 4. 2]. 31 'coditam' meint stets 'in einem v r t t . gefordert'. 32 In Pn. 8. 4.1 mu r [ausdrcklich] genannt werden."

305 Bhsya und Erklrer

m fojgenden Fall : 'klpyamnam33 posya' " . ( Z w e i t e Befrchtung.) A c r / a : N u n denn, auch wenn r und / n i c h t 'gleichlautig' heien, warum hat [die Substitution von n fr n] nicht statt im folgenden Fall : prakjpyamnam34 pasya" ?

Ekadesin : ,, Ich werde [ausdrcklich] lehren : 'Wenn [n von r oder s] durch einen Palatal, einen Lingual, einen Dental, ein /, ein sor (i. e. ein s, ein s oder ein s) getrennt ist, [wird] n nicht [substituiert] (i. e. ich werde Kty.'s vrtt. 1 zu 8. 4. 2 annehmen, dessen Notwendigkeit trotz vrtt. 2 zu 8. 4. 2 durch vrtt. 3 zu 8. 4. 2 etabliert ist.) Ein anderer sagt: Ich werde [ausdrcklich] lehren : [n wird] nicht [fr n substituiert,] wenn es [von r oder s] durch die drei mittleren Gruppen (Palatale, Linguale, Dentale) oder durch /, s, s getrennt ist'35. (Halbvers). Und auerdem [werde ich lehren] (ca\), da durch Nennung von Lauten Teile von Lauten [ebenfalls] gennant werden36. (Ich werde die von Kty. zu SS. 3, 4 behandelte Frage, ob diese Paribhs anzunehmen ist (vrtt. 6-10) oder nicht (vrtt. 11-13), im erstem Sinne entscheiden.) [Dann kann ich erklren :] das /, das in / (phonetisch = 9/9) enthalten ist, wird den Grund dafr abgeben, da [in praklpyamnam] das Verbot gilt [da bei Trennung durch / nicht n fr n substituiert wird]". Acrya: Wenn das so ist37 [da nmlich da in / enthaltene / die Substitution von n verhindert], dann ist die [im vrtt. 1 zu 8. 4. 1 geforderte] Nennung von r in 8. 4.1 zwecklos. [Da du sagst:] 'Auerdem [werde ich lehren], da durch Nennung von Lauten Teile von Lauten [ebenfalls] genannt werden', wird das r, das in r (phonetisch = 9r9) enthalten ist, den Grund dafr abgeben, da [in mtrnm etc.] n [fr n] substituiert wird] [nach der Regel 8. 4 . 1 ,
33

In diesem Fall st n nur durch einen Labial, Halbvokal und einen weiteren Labial von / getrennt. 8. 4.1 in der von Kty. geforderten Form wrde demnach nach 8. 4. 2. statthaben mssen. 34 Kaiy.: praklpyamnam iti krtyoco iti natvoprosagah In 'prakjpyomnam' ergibt sich eine zu weite Anwendung der Substitution von n nach Pn. 8. 4. 29." Pn. 8. 4. 29 krty ocah 'Fr ein n in einem krt-Suffix, welches auf einen Vokal folgt (in diesem Fall fr das n von -mono-) wird n substituiert nach einem r [nicht nur wenn es sich im gleichen W o r t befindet, sondern auch] nach einem Prverb [mit r] (in diesem Falle pra)". 35 Ka,\y.'.prvasmdayamvisesah: prvotra sarontorbhto 'pi sakrah sksnnatvanimittatvd vyavadhyakatvena nsrlyate. ha tu tasynupdnam eva M Der Unterschied [dieser zweiten Formulierung] zur ersten ist der folgende: In der ersten Formulierung wird s nicht als eine Trennung kon-

stituierend aufgefat, obgleich es in dem Ausdruck sor enthalten ist, da es [auf Grund von Pn. 8. 4.1] direkte Ursache der Substitution von n ist. In der zweiten Formulierung st es gar nicht erst erwhnt." In andern Worten: Zwischen den zwei Formulierungen besteht kein sachlicher Unterschied. 36 Zur nheren Erluterung s. u. 37 Es steht nicht fest, ob dieangefhrte Paribhs wirklich gltig ist. Zu 8. 4.1 werden deshalb zwei weitere Mglichkeiten erwhnt, mit deren Hilfe vrtt. 1 zu 8. 4.1 abgelehnt werden kann (vgl. Pat. Ill p. 452 Z. 15 ff. und Z. 19 f.). Patajali nimmt brigens auch Kty.'s Vorschlag, Pn. 8. 4. 2 negativ zu formulieren (vrtt. 1 u.8.4. 2), nicht an, wie der 'Ekadesin'. Auch das macht keinen Unterschied, da er durch Interpretationsmanver dasselbe erreicht, was Kty. und der 'Ekadesin'durch nderung erreichten.

306 Paul Thieme

da nach r ein n fr n substituiert w i r d ] . (Da du also vartt. 1 zu 8. 4.1 nicht zu lehren brauchst, kannst du dein A r g u m e n t nicht aufrecht erhalten, da sich bei Annahme von v r t t . 5 zu 1 . 1 . 9 etwas Falsches aus v r t t . 1 zu 8. 4.1 e r g i b t ) " . Aus den bersetzten Diskussionen haben w i r gelernt, da bei Annahme des v r t t . 5 zu 1. 1. 9 rkralkrayoh savornovidhih, 'er-

wnschtes (hotr + Ikrah > hotrkrah, hot*lkrah; rnla + Ikrah > mia ikrah (neben mlaikrah); upa + Iknyoti > upalkrlyati, uplkrlyati) sich in korrekter Weise ergibt' ('istasiddhih'), und Befrchtungen, bei seiner Annahme ergben sich Fehler dosaprptih'), nichtig sind. Vom praktischen Standpunkt aus gesehen ist das 'erwnschte' Resultat gewi vllig belanglos. Auch ein im Sanskrit schwelgender Pandit hat wohl selten Gelegenheit z. B. von dem 'vokalischen /des Hauptpriesters' zu sprechen oder 'sich auerdem ein / zu wnschen', und sicherlich wird er nicht das Bedrfnis empfinden, fr solche Ausdrcke gleich zwei, alternativ korrekte, Formen von der Grammatik geliefert zu bekommen. Um so interessanter ist die Errterung selbst. Sie zeigt die ganze Kunst patanjaleischer Darstellung. In abstraktester Form werden die Gesichtspunkte fr und wider gegeneinander ausgespielt, werden uns die verschiedenen berlegungen in ihren Resultaten vorgefhrt. Die Ausdrucksweise ist knapp genug, Miverstndnisse des Interpreten entschuldbar zu machen, und doch wieder so genau, da die falsche Auffassung eines Arguments sich unweigerlich beim nchsten verrt und rcht. Uns ldt sie schlielich noch deshalb zu nherer Betrachtung ein, weil die sprachlichen Tatsachen, um die es sich handelt, leicht zu berblicken sind, und wir uns gnzlich auf die logischen Schwierigkeiten konzentrieren knnen, die uns auf einer kurzen Wanderung durch die panineische Literatur begleiten werden. Zunchst haben wir uns klar zu machen, welcher Art die Laute sind, die nach vrtt. 1, 2 zu 6.1.101 fr r + r und r + /substituiert werden. Ich habe sie durchgngig mit *r und */ umschrieben ; in den Ausgaben des Bhsya und der Kommentare finden wir sie teils durch spezielle aksara bezeichnet, teils durch gewhnliches langes r und J, und, am ftesten, durch gewhnliches kurzes r und /, die nur gelegentlich durch diakritische Zeichen markiert sind. W i r werden sehen, da jedenfalls die Autoren der Prakriyund der Siddhntakaumudi sicherlich selbst kurzes r und / verwendeten, als sie ihre Werke fixierten. Was Kt/yana, der Autor der vrtt., sich vorgestellt hat, mag vorsichtigerweise dahingestellt bleiben. Seine Ausdrucksweise ist zu kurz und nicht eindeutig genug, um sichere Schlsse zuzulassen. Patajali's Errterungen (oben unter II) dagegen zeigen deutlich, wie er die beiden Laute auffate, obgleich er es nicht ausd r c k l i c h feststellt. Der 'Defensor' und der 'Opponens' sind sich nmlich darber einig, 1. da die beiden Laute nicht lac heien, wenn sie nicht ausdrcklich durch Hinzufgung einer Angabe so gennant werden. (Opp. 3undDef. 4.) Sie sind also weder in den SS. aufgefhrt, noch den dort aufgefhrten Lauten r und / 'gleichlautig' (ob. S. 301 Anm. 2). 2. da sie nur deshalb nicht 'lang' heien, weil sie nicht 'ac heien. (Opp. 3 und Def. 4.)

307 Bhsya und Erklrer

Sie sind also nicht 'kurz', sondern 'zweimorig' und heien 'lang', sobald ihnen der Name 'ac zugeteilt w i r d . 3. da sie neben r nach 6.1.101 substituiert werden k n nen, sobald sie den Namen 'lang' erhalten, vorausgesetzt, da v r t t . 5 zu 1 . 1 . 9 gelehrt w i r d . (Def. 3 und O p p . 4.) * r g i l t also fr ebenso 'nchstbenachbart' zu r + r wie r, und */ gilt fr ebenso 'nchstbenachbart 1 zu r + / wie r. W e n n nun * r einem r nicht 'gleichlautig' ist, und gleichzeitig einem r-\-r fr ebenso 'nchstbenachbart 1 gilt w i e r, kann es sich von r nur durch die 'Artikulationsweise im Munde' unterscheiden, nicht aber durch die 'Artikulationsstelle'. Im letzteren Fall knnte es nmlich einem r-\-r nicht fr ebenso 'nchstbenachbart' gelten wie r, auf Grund der Paribhs 'yatrnekavidham ntaryam tatro sthnata ntaryam ballyah38. hnliches gilt f r * / : Es kann sich von / nur durch die ' A r tikulationsweise im Munde' unterscheiden, da es sonst einem r + / nicht fr ebenso 'nchstbenachbart 1 gelten knnte wie f. Die 'Artikulationsweise im Munde' von r und / ist 'offen' (vivrta), wie die aller einfachen Vokale auer o. Diejenige von * r und */ mu, w i e w i r gesehen haben, verschieden sein. Das nchstliegende ist, sie als 'halbverschlossen' (Jsotsprsta) zu betrachten, wie die Halbvokale (y, v, r, I), und anzunehmen, da Pat. sich etwa das folgende gedacht hat : In r and / ist ein halbvokalischer 39 und ein vokalischer Bestandteil 4 0 hrbar. W e n n nun einmoriges r oder / gelngt w i r d , so mag man entweder durch Lngung des vokalischon Bestandteils eine ' o f f e n e ' (vivrta) Lnge erhalten (r), die im Fall des / in der Sprache nicht existiert, oder durch Lngung des halbvokalischen Bestandteils eine ' h a l b v e r s c h l o s s e n e ' (Jsotsprsta) Lnge: * r bzw. */. Diese so gebildeten Laute entsprechen den oben genannten Erfordernissen. 1. Als 'halbverschlossen' sind sie 'offenem' r und / nach 1 . 1 . 9 nicht 'gleichlautig' ; 2. als zweimorig erhalten sie nach 1 , 2. 27 den Namen 'lang', sobald sie 'ac' genannt werden ; 3. als von gleicher Artikulationsstelle w i e r bezw. / sind sie ebenso 'nchstbenachbart' zu r + r bezw. r + / wie r: 1 T + r T. * f sind z w e i m o r i g und lingual, 2. r + / ist zweimorig und halb lingual, halb dental, *r ist zweimorig und lingual, */ ist zweimorig und dental.

Die einheimischen Erklrer

Bevor wir uns dazu wenden, einen kurzen Blick auf die Reflexe der behandelten Bhsyastelle in einigen Hauptwerken der panineischen Literatur zu werfen, mchte ich mich ber zwei prinzipielle Punkte aussprechen, die im einzelnen an der zur Errterung stehenden Diskussion zu demonstrieren die eigentliche Absicht dieses Aufsatzes ist. Es handelt sich dabei um nichts Neues, aber um etwas das, wie mich bednkt, in der Praxis oft vergessen oder vernachlssigt wird, und das auch schon deshalb von Wichtigkeit ist, weil
38

Vgl. Paribhsendusekhara 13. B e z e i c h n e t als 'repha' u n d 'la', siehe oben S. 305.

39

40

Genannt 'bhakt, z. B. Pat. zu vrtt. 2 zu 8. 4.1 (III 452 Z. 9).

308 Paul Thieme

es an eine Frage r h r t , die uns in hnlicher Weise bei der Erklrung jedes alten indischen Textesentgegentritt : die Frage einer k r i t i schen und gerechten W r d i g u n g der einheimischen ' T r a d i t i o n . ' Erstlich : Was w i r die grammatische T r a d i t i o n ' nennen, ist offenbar nicht eine Tradition in d e m Sinne, da in ihr einfach alte, auf Patajali selbst zurckgehende Erklrungen zugnglich w e r d e n , die durch sisyaparampar sich mndlich f o r t g e e r b t htten. Eine s t r i k t e T r a d i t i o n ' besitzen w i r in der berlieferung des Vedatexts seit der Z e i t der Redaktoren : Man kann Lieder und sogar Bcher auswendig lernen, und der heilige Text konnte sich vllig unversehrt erhalten ohne andere Hilfe als den mndlichen U n t e r r i c h t und die bliche chorusmige Rezitation, die als korrigierender Faktor nicht unterschtzt werden darf. Gedanken und Inhalte jedoch kann man nicht mechanisch berliefern, wie die traurige Verfassung der einheimischen Veda i n t e r p r e t a t i o n zeigt 41 . T r a d i t i o n e l l ' in der einheimischen G r a m m a t i k e r - L i t e r a t u r ist lediglich die Methode: Das beraus ernste Bemhen, die Lehren des grammatischen Dreigestirns in der richtigen Weise zu interpretieren, das heit, durch Anwendung der im Mahbhsya selbst entwickelten Prinzipien die k o r r e k t e n Sprachformen durch k o r r e k t f o r m u l i e r t e Regeln zu erhalten. Die auerordentlichen Erfolge der einheimischen Auslegung sind zu danken einmal dem Scharfsinn und der, Intelligenz der 'Pandits', zum andern der Bescheidenheit und Geduld, mit der man innerhalb dieser hochentwickelten Gelehrtenkaste sich um eine w i r k l i c h grndliche Kenntnis der klassischen T e x t e vor allem andern bemht. Unsere 'kritisch-historische' Erklrungsweise verfgt ber reichere Fragestellung und vielfltigere Arbeitshypothesen, es pflegt ihr jedoch, von glnzenden Ausnahmen abgesehen, die solide Grundlage zu fehlen, die dem Pandit selbstverstndliche Voraussetzung ist, und die leider auch ein Kursus bei einem modernen SstrT uns nicht verschaffen kann. Zweitens : Einem grammatischen Kommentator k o m m t es meistens nicht auf die einfache Erklrung des W o r t s i n n s an. Er hat seine eigenen Lehrmeinungen, die er entweder denen seiner Vorgnger gegenberstellt, oder ihnen unterschiebt. Die kritischapologetischen Bemhungen Ktyyana's und Patajali's finden ihre Fortsetzung in folgenden Jahrhunderten. Ktyyana's stndige Fragen, ob eine bestimmte Formulierung einer Lehrmeinung unk o r r e k t , ungengend oder berflssig sei, sind nicht v e r s t u m m t bis auf den heutigen Tag. Er erffnet den Reigen der nachpanineischen 'Aufsteller gltiger Lehrmeinungen' (siddhntasthpakh), als dessen wichtigstes und praktisch letztes Glied w i r NgojTbhatta betrachten drfen, der heutzutage als der Begrnder der 'neuen Schule' (navyh) im Gegensatz zur lteren (prdnh : Kaiyataund Bhattoji Dlksita) gilt. Die neue Schule nun unterscheidet sich von der lteren in einem wesentlichen Punkt: in der Stellung zu Patajali. Auch die V e r t r e t e r der lteren Schule glauben an seine A u t o r i t t . Sie handhaben jedoch seine Methode verhltnismig unabhngig, indem sie versuchen, durch Einfhrung neuer sozusagen nur im
41 Da der durchschnittliche Rgvedin, der den gesamten Text in makelloser Reinheit zu rezitieren versteht, nicht im Stande ist, ber ein Lied etwas zu sagen, das nicht in der AnukramanT steht, kann ich aus persnlicher

Erfahrung bezeugen. Man kann von ihm nicht einmal erwarten, da er mit Syana vertraut st, oder da er sich in Sanskrit auszudrcken versteht,

309 Bhsya und Erklrer

Geist patanjaleischer Kunstgriffe, das Lehrgebude Pnini's auf eigene Faust zu e r w e i t e r n und vernderten Umstnden anzupassen. Ngojbhatta geht von der Voraussetzung aus, da alles zur 'Interpretation ' Pnini's Notwendige im Bhsya vorgebracht st, da w i r also keine A u t o r i t t besitzen, neue Gesichtspunkte einzufhren, oder Sprachformen zu rechtfertigen, die im Bhsya nicht erwhnt oder gebraucht sind. Es k o m m t infolgedessen nur darauf an, die Lehrmeinungen Patajali's zu erfassen und herauszuarbeiten. A u f diese Aufgabe k o n z e n t r i e r t er denn seine ganze Aufmerksamkeit und seinen ganzen Scharfsinn. Es gelingt ihm in der Tat, nicht nur die gesamte Bhsya-Interpretation, die bis dahin von Kaiyata beherrscht war, von Grund auf zu reformieren, sondern auch das Lehrgebude der 'PniniyfV von zahlreichen 'Ausklgeleien ' (phakkik) und manch einer 'gequlten I n t e r p r e t a t i o n ' (klistam vykhynom) panineischer Regeln zu subern. In seinen, in der blichen Weise als Kommentare oder Subkommentare eingekleideten Darstellungen pflegt er nach Auseinandersetzung der Ansicht Kaiyata's oder Bhattoji Dlksita's seine eigene, fast stets schlagend originelle und richtige Auffassung mit der bescheidenen W e n d u n g einzuleiten : pare tu ... 'Andere jedoch [sagen]', die von seinen Erklrern treffend paraphrasiert w i r d : bhsyatattvavidas t u ... 'Diejenigen jedoch, die die wahre Meinung des Bhsya kennen...'

1. Die Ksik42 (Erste Hlfte des 7. Jahrhunderts n. Chr.)

Ks. zu Pn. 1.1.9... rkralkrayoh savarnasamjn voktavy. hotr + krah, hotrkrah. Ubhayor rvarnasya ca varnasya cntaratamah savarno dlrgho nstity rkra43 eva dirgho bhavati. Der erste Satz ist identisch mit Patajali's 'Paraphrase' (I). Da die Ksik 'hot*!krah' hier nicht erwhnt, drfen wir schlieen, da des Defensors Lehrmeinung (Arg. 3) nicht angenommen ist, das heit, da der Verfasser es nicht fr richtig hlt, diese Form bereits aus vrrt. 5 zu 1.1. 9 abzuleiten und folglich die vrrt. zu 6.1.101 zu streichen. Andrerseits befindet er sich im Widerspruch mit dem Opponenten, mit dem er nur darin bereinstimmt, da die vrtt. zu 6.1.101 gelehrt werden mssen. Denn der Opponens erhlt auch die Form 'hotrkrah1 aus eben diesen vrtt. und lehnt vrtt. 5 zu 1.1.9ab(Arg.1 und 2). Die Ksik fhrt fort: Dafr beide, nmlich das rund das / [in hotr + krah] ein 'nchstbenachbarter' 'gleichmtiger' 'langer' [Vokal] nicht vorhanden ist [da ein langes / der Sprache fehlt], wird lediglich fais 'langer' [Vokal] (nach 6.1.101 akah savarne dJrghah) substituiert". Diese Erklrung der Substitution von f in hotr + krah > hotrkrah ist im Bhsya als offenbar selbstverstndlich nicht gegeben. Kaiy. formuliert hnlich44; wir werden sehen, da man spter noch genauer ist. Ks. zu Pn. 6.1.101 lehrt, wie erwartet, die beiden vrtt. und erhlt durch sie: hot*rkrah neben hotrkrah fr hotr + rkrah und hot* I kroh neben hotrkrah fr hotr + kroh. Dann fhrt sie fort : rkralkrayoh savarnavfdhir ukto; dirghaZitiert nach der Ausgabe von Pandit Bla Sstr, Benares21898. Sie st leider, wie auch die Ausgaben der folgenden von mir zitierten Werke, durchaus nicht frei von Druckfehlern, in denen sich in einer
42

fr uns recht lstigen Weise des Pandits souverne Verachtung des geschriebenen Wortes ausdrckt, 43 Ausgabe: rkra. 44 S. oben S. 300 Anm. 9.

310 Paul Thieme

pakse tu samudayantaratamasya kriyate.

Ivornasya dlrghasyabhavad rkara evo

Es st gelehrt worden (zu 1.1. 9), da r und / 'gleichlautig' heien. (Daraus ergibt sich, da rti und Iti in den vrtt. richtig paraphrasiert werden als 'vor einem gleichmtigen f und \or einem gleichmtigen /' : Def. Arg. 2.) Senn die Substitution des 'langen' (Vokals) gewhlt wird, dann wird lediglich /"gettigt, da ein der Summe (r + /) [ebenfalls] nchstbenachbartes langes' / nicht vorhanden ist". Aus der Ausdrucksweise geht unzweideutig hervor, da *r und */ nicht fr 'lang' gelten. Damit ist jedoch nichts ber ihre tatschliche Dauer ausgesagt. Es bedeutet lediglich, da der Vorschlag des Def. (Arg. 4), fr sie den Namen 'ac' zu lehren, abgelehnt ist, und zwar offenbar deshalb, weil die Substitution des 'plutierten' Vokals in dem gewi absurden Satz: 'He, du / des Hauptpriesters', ein verzichtbares Resultat erscheint. Wenn man aber das 4. Argument des Def. nicht annimmt, dann mu man dem Opp. folgen, welcher behauptet, da die zwei vrtt. zu 6.1.101 unter allen Umstnden zu lehren sind, ob man nun vrtt. 5 zu 1.1.9 annimmt oder nicht (Arg. 3). Die Ksik stimmt also den Argumenten 2 des Defensors und 3 des Opponenten zu, whrend sie die Argumente 3 und 4 des Defensors ablehnt, und demnach auch den Siddhnta des Acrya. Die Begrndung knnen wir ihren knappen Stzen nicht unmittelbar entnehmen, es ist jedoch offensichtlich, da praktische Rcksichten der Deutlichkeit und Einfachheit fr eine Lsung verantwortlich zu machen sind, die in keiner Weise Patajali's wirkliche Ansicht wiederspiegelt. 2. Kaiyata45 (frhstens 11. Jahrhundert n. Ch.) W i r haben Grund anzunehmen, da in Kaiyata's 'Bhsyapradlpa' nicht viel originelle Arbeit steckt. Der Autor selbst versichert, da er 'traditionsgem' erklren will und gibt Bhartrhari als seine Quelle an46. Seine Knappheit und bersichtlichkeit, und wahrscheinlich auch seine kluge Beschrnkung auf einfache und naheliegende Interpretationen haben dem Pradlpa jedoch den Vorrang vor anderen Bhsyakommentaren erobert. Auch wir tun gut, Kaiyata als erste Hilfe zum Verstndnis vorzugsweise zu behandeln. Er gibt uns freilich nicht eine einfache Erklrung des W o r t sinns, er paraphrasiert nur gelegentlich. Das Bhsya ist ja in leichtem Sanskrit abgefat, und es kommt verhltnismig selten vor, da wir Wendungen oder Ausdrcke nicht verstehen. Eine einfache bersetzung der Worte bereitet kaum eine Schwierigkeit. Diese beginnt erst spter: Nachdem wir den Wortsinn (artha) verstanden haben, mssen wir uns die Meinung (bhva) klar machen, die Motivierungen der Argumente und Behauptungen, die im Bhsya kurz und nackt hingesetzt sind. Bei diesem Versuch, das Skelett des Textes mit Leben zu erfllen, hilft uns Kaiyata mit seinen Deutungen. Ich habe bereits in den Funoten zu meiner bersetzung alle die Bemerkungen Kaiyata's aufgefhrt, die ich fr richtig halte, und auch meine eigenen Hinzufgungen durchweg auf ihn gegrndet. Der Beweis der Richtigkeit wird durch die Tatsache
45

Zitiert nach der Ausgabe von Pandit Sivadatta D. Kuddla: Patajali's Vykarana Mahbhsya with Kaiyata's PradTpa, Bombay 1917.

46

Einleitung Vers 5 und 7. Zum Verhltnis Kaiyata's und Bhartrhari's vgl. Kielhorn, Mahbhsya II2 Preface S.19, Anm.1.

311 Bhsya und Erkirer

geliefert, da nur unter der Voraussetzung dieser Motivierungen das Bhsya einen konsistenten Sinn hat : Nur wenn r und / verschiedene Artikulationsstellen haben, braucht das.vrtt. 5 zu 1.1.9 gelehrt zu werden ; nur wenn aus 6.1.101 der Ausdruck drghah in den vrtt. fortgilt, und nur wenn r allein der 'lange' 'nchstbenachbarte' Vokal zu r + / ist, hat das Argument 1 des Opponens, nur wenn sich ohne vrtt. 5 zu 1.1. 9 als Substitut fr dadhi + Ikrah etwas Falsches nach vrtt. 2 zu 6.1.101 ergibt, hat das Argument 2 des Defensors einen Sinn usw. Aber auch Kai/ata hat nicht darauf verzichtet, Deutungen in seinen Kommentar aufzunehmen, die wir teils ablehnen, teils als nicht notwendigerweise richtig dahingestellt bleiben lassen mssen 1. Der Acr/a (1. Antwort) sagt, da er ausdrcklich lehren will, da ein o, /' oder u als Substitut fr / von / gefolgt ist. W i r erhalten z. B. fr upa + kriyati zunchst: up + guna + kriyati (Pn. 6 . 1 . 92); guna meint a, e oder o(Pn. 1.1.2); als zu whlender Vokal ergibt sich o als dem a + / 'nchstbenachbart' (Pn. 1.1. 50): up -\~ a + kriyati; und schlielich soll resultieren : upalkrlyati. Patajali lst jedoch das Versprechen des crya nicht ein, wie er auch nicht den Namen ' ac' fr *r und */ tatschlich lehrt. Sein Bhsya hat eben nur die Aufgabe zu diskutieren, was im Text der Grammatik hinzugefgt oder gendert werden sollte und was nicht: die resultierenden Lehrstze sind nicht formuliert. Es bleibt Sptem berlassen, die Folgerungen zu ziehen. Bezglich *r und */ einigt man sich denn, wie wir noch sehen werden, sie in den SS. aufzufhren, als das krzeste und beste Mittel zu erreichen, da sie ac heien. Bezglich der Substitute fr / hilft map sich seit der Ksik in der folgenden Weise: Man fat das a in SS. 6 (an) als nasaliert, d. h. nach 1.3.2. ais anubandha, auf47 und lt in der Regel Pn. 1.1. 51 ur an rapara h das ra mit diesem anubandha gebildet sein, sodaes also die in SS. 5 und 6 genannten Laute rund / einbegreift. Dem betreffenden Substitut wird dann nach 1.1. 50 beziehentlich r oder / folgen, je nachdem es fr r oder / steht. Anstatt also ' ausdrcklich zu lehren', interpretiert man mit einem Kunstgriff. Kaiyata projiziert nun diese knstliche ' Interpretation ' ins Bhsya. Er erklrt den Ausdruck des crya (1. Antwort) vaksymi ' ich werde [ausdrcklich] lehren ' mit vykhysymlty arthah D e r Sinn ist: Ich werde interpretieren", und fhrt fort: rapara ity atra ra iti an iti sakrkrena pratyhra srlyate. tatrntaratamyd rkrasyn raparah, Ikrasya laparah In dem Ausdruck raparah (in Pn. 1.1. 51) wird ' ra' [nicht als der Name des Konsonanten r, sondern] als ein Sammelausdruck (pratyhra) aufgefat, (der nach 1.1. 71 gebildet ist) mit dem ebenfalls (als anubandha) wirkenden a in SS. G (an). Bei dieser Auffassung wird (nach Pn. 1.1. 51) fr r ein von rgefolgtes, fr / ein von / gefolgtes a, oder u substituiert." Aber er gert in die Brchej bei dieser Interpretation von vaksymi wird der folgende Satz des crya logisch unkorrekt : asatym savarnasamjnym vidhyartham ... bhavisyati, denn die Interpretation von ra in 1.1. 51 als r und / hilft uns nichts, wenn r und / nicht 'gleich laut ig' heien, da dann in dem uf) in 1.1. 51 nur r und nicht auch /genannt ist. Kaiyata ist also gezwungen gegen das Bhsya, das dasselbe (tad eva) Mittel fr beide Flle ob nun vrtt. 5 zu 1.1. 9 aufrecht steht oder nicht anwenden will, zu behaupten:
47

Das ist mglich, insofern die Nasalierung eines Vokals in der AstdhyyT niemals schriftlich bezeichnet, sondern

aus der Interpretation zu folgern st: protijnnunsikyh PninJyh.

312

asatyam savarnasamjayam

ur an rapara ity atra

Ikaragrahanam

Paul Thieme

kartavyam bhavati Wenn [fr r und /] der Name 'gleichlautig' nicht [gelehrt] ist, dann mu in Pn. 1.1. 51 der Ausdruck / hinzugefgt werden_48". 2. Die vom Acryazitierte Regel 6.1.12849 wird nicht nur so interpretiert, da man nach ihr beliebig mla rsyah etc. statt mlarsyah (nach 6.1. 87) fr ml + rsyah, sondern auch so, da man hotrrsyah statt hotrsyoh (und hot*rsyah) (nach 6.1.101) erhlt (vgl. vrtt. 1 und Bhsyazu 6.1.128). Wenn nun hotrrkrah nach 6.1.128 korrekt ist, dann, mu auch hotr krah korrekt sein : Sofern / und r 'gleichlautig' heien, nach 6.1.128; im andern Fall mag man Pn. 6.1.127 in Ansprach nehmen. Kaiyata zu Opponens 2 operiert demnach mit je drei Sandhiformen : hotrrkrah, hotrkrah, hotrrkrah und hotrlkrah, hotrkrah, hot*ikrah> und je einer Auersandhiform : hotr rkrah und hotr krah.50 Ngojbhatta gelingt es im Uddyotazu zeigen, da Kaiyata diesem komplizierten Spiel nicht ganz gewachsen ist und sich in seinen selbstgeschaffenen Schwierigkeiten verfngt (unten S. 320f.). Die Kernfrage der ersten Diskussion (II) bildet, wie wir sahen, die Auffassung der Laute *r und */. Kaiyata bemerkt richtig, da das dritte Argument des Defensors nur einen Sinn hat unter der Voraussetzung, da er beide Laute fr lang 1 und somit ihre Substitution nach 6.1.101 fr gegeben hlt. Der Opponens bekmpft diese Ansicht mit dem Hinweis, da beide Laute nicht 'ac' heien, was wiederum nur einen Sinn hat unter der Voraussetzung, da sie weder mit gewhnlichem r und / identisch, noch ihnen 'gleichlautig' sind (o. S. 301 Anm. 2). Deshalb sagt Kaiyata zu Opp. 3 : anye tv Jsatsprstakaranatvd anayor rkralkrayosca vivrtatvt tbhym tayor agrahand anactvam huh ,,Andere jedoch [deren Ansicht ich teile] sagen, da jene beiden (i.e. *r und */) nicht 'ac heien, weil sie von [den in SS. 2 als 'ac' aufgefhrten] r und / nicht (nach 1.1. 69) einbegriffen werden, da ihre Artikulation 'halbverschlossen' ist, r und / jedoch 'offen ' sind." 51 Wie wir uns weiter bei Ablehnung der vrtt. zu 6.1.101 und bei Annahme von vrtt. 5 zu 1.1. 9 die Substitution von r, bzw. *r und */, nach 6.1.101 vorzustellen haben, errtert er spter (zu m ' ksepa ') i n folgender Weise : dvayor rkrayo rephadvayayuktatvd vivrtatvc ca kadcid rephadvayayukta *r52 bhavati kadcid vivrta rkrah53 kre 'pi kadcid rkrntaratama rkrah kadcid krntaratama *lkrah.54 ,,Da die zwei r-Vokale [in hotr + rkrah] zwei konsonantische r enthalten, und da sie [zu gleicher Zeit] ,offen' sind, wird manchmal das zwei konsonantische r enthaltende *r[als 'nchstbenachbarter' 'langer' Vokal], und manchmal 'offenes' r [als 'nchstbenachbarter' 'langer' Vokal] (nach 6 . 1 . 101) substituiert. Und auch wenn / folgt [in hotr + krah], wird manchmal das dem r [von hotr] 'nchstbenachbarte' r, manchmal
Die korrekte Interpretation der Bhsyastelle erst bei Nagojbhatta, siehe unten S. 320. 49 Oben S. 303. 50 Die Tendenz, mit fakultativer Regeln mglischst viele Sandhiformen zu gewinnen, wchst seit Patajali. Den Rekord hlt wohl Bhattoji DTksita, der sich in der Siddhntakaumud zu 8. 3. 34 gleich 108 verschiedene Aussprachemgli48

chkeiten fr das W o r t somskorta ausrechnet! 51 Trapp bersetzt (Anm. 185): 'weil r und ! zugleich leicht berhrt und offen gesprochen werden und wegen dieser beiden sich widersprechenden Eigenschaften nicht als Vokale erfat werden knnen', (sie!) 52 Ausgabe: r. 53 Ausgabe: rkrah. 54 Ausgabe: Ikroh.

3(3 Bhsya und Erklrer

das dem / [von karah] 'nchstbenachbarte' */ [als 'langer1 Vokal] (nach 6.1.101) [substituiert]." Kai/ata gibt also dieselbe Deutung, zu der wir oben55 unsere Zuflucht nahmen : Nur wenn r als 'offen,' *r und */ jedoch als 'halbverschlossen1 gilt, hat die Argumentation des Bhsya einen Sinn. Kai/, ist nur etwas genauer in der Bestimmung der Laute *r und */, als wir es zu sein wagten : W i r begngten uns mit der Annahme, die Lngung von r und / zu *r und */ sei erreicht durch Lngung des in r und / enthaltenen konsonantischen Bestandteils. Kaiyata rechnet aus : In r ist ein konsonantisches r = { Mora, und in / ein konsonantisches / = { Mora enthalten. Da ein 'kurzer' Vokal (nach 1. 2. 27) eine Mora betrgt, bleibt je \ Mora fr den vokalischen Bestandteil. In *r und */ dagegen sind, nach seiner Meinung, bezeihentlich zwei konsonatische r = 1 Mora, und zwei konsonantische / = 1 Mora enthalten, soda eine Mora fr den vokalischen Bestandteil bleibt. Wenn sich demnach Kaiy.'s Erklrung im wesentlichen mit der unsern deckt, so fragt sich, ob er oder seine Quelle sie auf demselben Wege erreicht, das heit, ob er sie ebenfalls durch Interpretation erschlossen hat, oder ob sie sich auf mndliche Tradition grndet, in dem Sinne, da die Aussprache, die Patajali selbst fr *r und */ im Unterricht verwendete, durch si sy apar amparo noch bekannt gewesen wre. Auch abgesehen davon, da es sich offensichtlich mehr um eine theoretische als um eine praktische Eigentmlichkeit handelt, knnen wir unumwunden die erste Alternative bejahen. Kaiyata kennt nmlich auch eine andere Ansicht. Ihre Vertreter erklren, da *r und */ deshalb nicht 'ac' heien, weil sie zwei und eine halbe Mora betragen.56 Sie haben also die Auffassung von *r und */ als 'halbverschlossen' entweder nicht gekannt oder nicht geteilt: ardhatrtJyamtratvd iti kecid huh ,,Manche sagen [da *r und */ nicht 'ac heien], weil sie [je] 2 Moren betragen". Nach Pn. 1. 2. 27 hat nun ein 'ac' entweder ein (u), oder zwei (), oder drei (l) Moren. Die zitierten Gelehrten mssen also die Lngen des r etwa in der folgenden Weise analysiert haben :

r + r ( = 19 + r + 1i + r) = f(3 + 1 r) = 2 Moren r + r (=r. 2 + i r + 2 + i r) = *r(4 + 1 r) = 2 Moren.

Kaiy. zeigt nun, da diese Deutung des Satzes : na ca rkra Ikro vj asti, falsch ist, da sie sich spterhin in der Interpretation nicht bewhrt. Er sagt nichts ber eine tatschliche, feststehende Aussprache, der sie zuwiderliefe: saty actve dlrghasamjnnayor bhavisyati. yesm tu matam ardhatrtJyamtrv etv iti tesm mate saty apy actve dvimtratvbhvd etayor drghasamja na prpnoti. tasmd dvimtrv etv abhyupagantavyau ,,[Er (i.e. der Defensor) sagt: 'Ich werde lehren, da *r und */ 'ac heien', indem er voraussetzt:] Wenn *r und */ 'ac heien, dann werden sie den Namen 'lang' (nach 1. 2. 27) empfanS. 3 0 7 . Nach Trapp (Anm. 185) werden gewhnliches r und / von einigen Grammatikern als 'zweieinhalbzeitig' angesehen ! In derselben Anmerkung bemerkt er, da eine Vorschrift, die nur von einigen Grammatikern vertreten wird, immer wahlfrei sei.
56 55

*r und */ seien zweieinahlbzeitig, von Kaiy. unzweideutig abgelehnt wird, Es handelt sich eben gar nicht um eine 'Vorschrift', sondern um eine 'Ansicht'. Falsche Ansichten jedoch werden auch in Indien bekmpft und abgewiesen: sie gelten nicht als fakultativ ebenfalls richtig.

W i r werden sehen, da die Ansicht,

314 Paul Thieme

gen [und demnach nach 6.1.101 substituiert werden]. Diejenigen jedoch, die da annehmen, da *r und */ je zwei und eine halbe Mora enthalten, mssen annehmen, da, auch w e n n * r und */ 'ac' heien, der Name 'lang' sich fr sie nicht ergibt, da sie nicht zwei Moren [sondern zwei und eine halbe] enthalten (und demnach 1. 2. 27 nicht Anwendung finden kann). Deshalb mssen * r und */ als zweimorig angesehen w e r d e n . " W i e ich an diesem Beispiel vor allem zeigen w o l l t e , ist also Kaiyata's 'Lampe' zur Aufhellung der Meinung des Bhsya auch fr uns von unschtzbarem W e r t . N i c h t jedoch, weil seine Deutungen auf alte Tradition zurckgehen, sondern vielmehr, weil sie auf wohlerwogenen Grnden der Interpretation beruhen, deren Gewicht w i r nachzuprfen durchaus imstande sind, wenn w i r in der Grammatik und im Bhsya selbst gengend Bescheid wissen. W i r drfen natrlich seine Deutungen verwerfen, wie zum Beispiel seine Erluterung zu vaksymi (oben S. 311), falls w i r nmlich nachweisen knnen, da sie keine W u r z e l im Bhsya haben. W i r drfen uns jedoch nicht bermtig ber sie hinwegsetzen, oder gar unterlassen, sie berhaupt erst zu verstehen. Sonst laufen w i r Gefahr, aufs neue mizudeuten, was indische Gelehrte bereits vor Jahrhunderten richtig herausgefunden haben, und bersetzungen zu liefern, deren w i r uns vor dem durchschnittlichsten VykaranaSstrlzu schmen htten.

3. Haradatta's Padamajar57 (spter als Kai/ata) und Bhattoji Dksita's Sabdakaustubha58 (17. Jahrhundert) In der
Padamajar zu Ks. zu Pn. 1.1. 9 (p. 55 ff.) am Ende (p. 58 f.) gibt Haradatta zunchst eine Erklrung des Siddhntader Ksik. Er ist dabei, wie es brigens schon Jinendrabuddhi war, genauer in der Motivierung der Substitution des langen r, als Kaiyata und die Ksik: ,,Der Sinn des Satzes der Ksik 'ubhayoh' usw. ist: Dafr beide (r + /) [zusammen] ein 'nchstbenachbarter' 'langer' Vokal nicht existiert,59 und da auch fr / ein ' nchstbenachbarter' 'langer' Vokal nicht existiert, wird fr beide lediglich r (nach 6.1.101) substituiert". Dann geht er dazu ber, eine Paraphrase des Bhsya zu vrtt. 5 zu geben. Den Wert solcher Paraphrasen in der Padamajar hat schon Kielhorn betont.60 Sie ergnzen Kaiyata in vorzglicher Weise. Whrend Kaiy. in seinem Kommentar gewissermaen nur Antworten auf unausgesprochene Fragen gibt, die man zum Text des Bhsya stellen kann, gibt Haradatta eine zusammenhngende, knappe aber bersichtliche Darstellung der Diskussionen. Er benutzt Kaiyata, mit dem er teilweise wrtlich bereinstimmt, sucht ihn jedoch in logischer Schrfe zu bertreffen. So fgt er zu Kaiy.'s Deutung, da *r und */ 'halbverschlossene1 'zweimorige' Laute seien, welche zwei konsonantische r bezw. / enthalten, eine noch genauere Angabe ber ihre tatschliche Konstitution hinzu : modhye dvau rephau (bezw. : lakrau), tayor ek mtr. abhito 'jbhakter apar ,,ln der Mitte von *r (bezw. */) befinden sich zwei konsonantische r (bezw. /), welche eine Mora
57

Zitiert nach der Ausgabe von Bhradvja Damodara SstrT (Reprint from Pandit), Benares 1898. 58 Zitiert nach der Ausgabe von Rma Krsna SstrT, Benares 1898. 59 Vgl. vrtt. 18 zu 1.1. 50 und Bhsya:

ubhoyor yo 'ntaratamas tena bhavitavyam. 60 Mahbhsya II2 p. 11 f. Trapp hat zu seinem Schaden auf diese Hilfe gnzlich verzichtet.

315 Bhsya und Erklrer

ausmachen ( Mora + i Mora = 1 Mora). Die vokalische Partikel auf beiden Seiten [des rr bezw. //] macht eine Mora aus." 61 W i r knnen also umschreiben : * r = rr, */ = //. W i e Kaiyata das A r g u m e n t 1 des Opponens versteht, w i r d nach dessen Ansicht 'dlrghah' aus 6.1.101 in vrtt. 2 zu 6.1.101 fortgelten, und deshalb beliebig auch der 'lange' Laut r in hotr - f Ikrah substituiert. Haradatta macht sich Sorge, wieso sich das notwendigerweise ergibt, und beschwichtigt sie mit der Annahme, da in Iti *jvvacanam der Ausdruck 'v' im Sinne der Hinzufgung 62 stehe, d. h. ' u n d ' bedeute. 63 Rhmt sich Haradatta im Eingang seines Kommentars 'von dem Mangobaum 64 der Grammatik die Bltendolde der (richtigen) W o r t e gepflckt zu haben' (V. 4), so hat Bhattoji Dlksita im Sabdakaustubha vor ' das Kaustubhajuwel des (richtigen) W o r t e s aus dem Ozean des von der Schlange65 verkndeten Bhsya herauszuholen [ w i e die G t t e r den Kaustubha aus dem gequirlten Ozean]' (V. 3). Z u v o r erweist er seine Ehrerbietung dem grammatischen Dreigestirn : Pnini, Ktyyana und Patajali, und den Auf- oder Feststellern der gltigen Lehrmeinungen (siddhntasthpakh) Bhartrhari usw. (V. 2). 66 Den Sinn, den man nur mhevoll und durch Bewltigung vieler W e r k e erhalten kann, sollen sich die 1 Guten ' restlos und ohne Mhe aus seinem Buch aneignen (V. 4). Der Dlksita hat also nicht die Absicht, die Interpretation eines alten W e r k e s zu erleichtern, sondern vielmehr die, es zu ersetzen, und zwar durch Darstellung und Untersuchung der Lehrmeinungen nicht nur des ursprnglichen Verfassers, sondern auch anderer, folgender 'siddhntasthpakh.' Der 'Sinn,' den man sich aneignen soll, st nicht einfach das Verstndnis des Bhsya sondern der endgltige Sinn aller bisher geleisteten grammatischen A r b e i t : Die Bildung der k o r r e k t e n W o r t f o r m e n auf Grund korr e k t f o r m u l i e r t e r Regeln. Der bekannte auf Haradatta's W e r k gemnzte Nyya 67 gilt auch fr. den Sabdakaustubha. Die Erklrung des vrtt. 5 zu 1 . 1 . 9 im Sabdakaustubha (S. 151 ff.) ist also zu verstehen als eine Widergabe nicht nur der E r r t e r u n g Patajali's, sondern auch der Feststellungen Kaiyata's und Haradatta's, mit denen sie oft w r t l i c h bereinstimmt, und die
Da man auch diese Ansicht im Bhsya begrnden kann, zeigt NgojTbhatta, siehe unten. 62 Auf Grund von N i r u k t a i . 4 vet/' vicranrthe . . . athpi samuccoyrthe bhavoti, st eine entsprechende Angabe in die Kosa aufgenommen. 63 tatra iti *vovacanam ity otro vsobdo dirghosya samuccoyrthas. tenprpta eva dirghobhavisyati,, In diesen vrtt. steht nun das W o r t 'v' in vrtt. 2 um die Hinzufgung von 'dirghah' zu erreichen. Deshalb wird der 'lange' Vokal (r) [fr r + / ] substituiert werden, obgleich er sich [flschlich] nicht (nach 6.1.101) ergibt (insofern r und / nach meiner Meinung nicht 'gleichlautig' heien)". Die Ausgabe liest statt Iti *lvvacanam : ni *rvvacanam, was nicht
61

richtig sein kann, da der 'lange' Vokal (r) sich als Substitut fr r -f r korrekt nach 6.1.101 ergibt, ob man nun r und /als 'gleichlautig' betrachtet oder nicht. 64 sahakrapdapt: Der Sahakra ist eine besonders s duftende Abart des gewhnlichen mra. 65 Patanjali wird gern als Schlange (sarpo, phonin, ngo, sesa) vorgestellt, 66 Vgl. diehnliche Gegenberstellung von munitroya und vaiykaranasiddhnta im Einleitungsvers der SiddhntakaumudT. 67 Siehe G. A. Jacob, Laukikanyynjali 1112 (Bombay 1911), S. 7: Anodhite Mahbhsye vyarth syt Podomojar, adhite 'pi Mahbhsye vyarth s Padamoja.

316 Paul Thieme

sie vllig ausbeutet, jedoch ihrerseits in Genauigkeit und logischer Schrfe bertrifft. Kaiyata's Bemerkung zu O p p , 2, da nach dessen Ansicht im v r t t . 1 zu 6.1.101 'savarne' aus 6.1.101 fortgelte, whrend vrtt. 2 'asavarnrtham' gegeben sei (o. S. 301 A n m . 12), w i r d mit einem negativen Beleg (pratyudharana) gerechtfertigt : dhtr + amsah ergibt nur dhtramsah nach 6 . 1 . 77. Seit Def. 3 w i r d gefordert (auer von der Ksik, s. o. S. 310), da fr * r und */ der Name *ac gelehrt werden mu, damit sie (nach 1. 2. 27) 'lang' heien knnen, und damit Substitution des plutierten Vokals (nach 8. 2. 86) fr sie statthabe, insofern der Name ' p l u t i e r t ' (nach 1. 2. 27) nur einem 'ac gegeben werden darf ( K a i / . , Haradatta). Dksita przisiert: insofern ein 'plutierter 1 Vokal nach 1 . 2. 28 (acas ca) nur fr ein 'ac' substituiert werden

kann, tadvidhne hy acas ceti paribhsay aca ity upatisthati ,,Denn wenn Substitution des plutierten Vokals gelehrt wird, so t r i t t auf Grund der Interpretationsregel 1. 2. 28 'fr ein ac' hinzu68 (soda wir also 8. 2. 86 zu interpretieren haben : Fr ein ac, welches 'lang' usw. ist... wird der 'plutierte' Vokal substituiert"). W i r erreichen jedoch nicht die Substitution des r i c h t i g e n 'plutierten' Vokals. Wenn nmlich lediglich gelehrt wird, da *r und */ 'ac' heien, so ist damit nicht gesagt, da sie auch 'an' sind. Dann knnen sie aber berlanges *rl und * / ' nicht (nach 1.1. 69) einbegreifen, diese heien also nicht (nach 1. 2. 27) 'plutiert.' Folglich wird man flschlich als 'nchstbenachbarten' 'plutierten' Vokal rl substituieren : ata eva acsamjnmtrenpi na nistarah Deshalb ist die Schwierigkeit auch nicht damit beseitigt, da man fr *r und ^lediglich den Namen 'ac' lehrt." Sie mssen vielmehr in den Sivastra unter den 'ac' a u f g e f h r t werden (varnasammnye tau pathanJyou) ; dann erst gelten sie auch als 'an' und begreifen (nach 1.1. 69) auch die berlangen Vokale *rl und *ll ein, die dann als 'nchstbenachbarte' 'plutierte' Vokale richtig nach 8. 2. 86 substituiert werden. 4. Die Siddhntakaumud 69 In der ueren Anlage seiner Werke zeigt der Dksita wenig Erfindungsgabe : Der Sabdakaustubha ist im Grunde nichts als ein 'Haradatta redivivus.' Und auch die SiddhntakaumudI ist nach einem bekannten Muster gearbeitet, sie folgt in Plan und Darstellung engstens der Prakriykaumudl70 Rmacandra's (15. Jahrh.). Wie Rmacandra lehrt Bhattoji DTksita alle drei vrtt. (vrtt. 5 zu 1.1. 9 und vrtt. 1, 2 zu 6.1.101) in der Siddh. Kaum., setzt also die im Sabdakaustubhafestgestellte Lehrmeinung, die die theoretisch beste Mglichkeit darstellt, nicht in die Praxis um. Sonst mte er *r und */ in den SS. lehren und die beiden vrtt. zu 6.1.101 streichen. Vrtt. 5 zu 1.1. 9 erscheint (Siddh. Kaum. p. 8) in der Formulierung: rlvarnayor mithah svarnyam vcyam. die der Prakr. Kaum. (p. 26) entlehnt st. Zu diesem Wortlaut'vcyam' ist nicht mitzurechnen, ist zunchst zu bemerken, da mithah 'gegenseitig' hinzugefgt ist.
Auf Grund der Paribhs: kryakam samjnparibhsom (Paribh. Indus. 3). 69 Zitiert nach der Ausgabe von Pandit SivadattaSstrT, Bombay 1926. 70 Ed. K. P. Trivedi, Bombay 1925.
68

Ich schulde es dem gelehrten und sorgfltigen Herausgeber zu betonen, da diese Ausgabe in dem oben (S. 309 Anm. 42) geuerten Urteil nicht inbegriffen ist. Sie wird allen wissenschaftlichen Ansprchen gerecht.

317 Bhsya und Erklrer

Kaiy.'s, im Sabdakaust. gebilligte, Ansicht 71 , da mithah sich aus der Interpretation ergebe, ist also nicht angenommen. Die Hinzufgung schafft natrlich unmittelbarere Klarheit. Da sie jedoch das vrtt. 'schwerer' macht, ist in anderer Beziehung abgekrzt w o r den. Statt savarnavidhih (Si Moren) heit es : svarnyam (7 Moren), und statt rkralkrayoh :rvarnayoh, wobei durch Auslassung des ersten -kra 4 Moren, und durch Ersetzung von krayoh durch varnayoh eine halbe Mora, gespart sind. In der Praudhamanoram 72 , die er selbst eine 'vykhy' der Siddh. Kaum, nennt (Einleitungsvers), die jedoch in W a h r h e i t lediglich eine Rechtfertigung der Formulierungen und Lehrmeinungen der Kaumud ist73, e r k l r t Bhattoji (p. 15), wie das K o m positum rjvarnayoh aufzulsen ist. Nmlich entweder als Genetiv eines Kompositums : 'die beiden Laute r ('T = N o m . von r-) und / ('d' = N o m . von / - ) ' , oder als Kompositum des Genetivs : ' des r ('uh') und des / ('u/'), der beiden Laute'. H i e r m i t vermeidet er den Vorwurf, da das Kompositum gegen Pn. 2. 2.11 verstoe. Die v r t t . zu 6.1.101 rti *rvvacam, Iti *!vvacanam erscheinen in der Prakr. Kaum. (p. 67f.) als rti savarne rv, Iti Iv7\ und in der Siddh. Kaum. (p. 27) als rti savarne r v, Iti savarne I v7*. Die Formulierung des Dksita ist 'schwerer', hat jedoch den Vorteil, seine Lehrmeinung klar erkennen zu lassen : Er nimmt das Arg. 2 des Def. an, da die vrtt, nur unter der Voraussetzung, da r und / 'gleichlautig' heien, einen korrekten Sinn haben, aber auch das Arg. 3 des Opponens, da die vrtt, gelehrt werden mssen, da *r und */ nicht 'ac' sind. Die Wiederholung von savarne im zweiten vrtt. macht es ganz deutlich, da er dlrghah n i c h t fortgelten lt. Dem Einwand, da fr *r und */ der Name 'ac1 gelehrt werden mu, oder, wie er selbst im Sabdakaust. sagt, da *r und */ in den SS aufgefhrt werden mssen, um die Substitution des plutierten Vokals (nach 8. 2. 86) zu erreichen, mag er mit der Antwort begegnet haben, da der plutierte Vokal nur in Vokativen, und auch hier in nicht-letzter Silbe nur fakultativ, substituiert wird, nieKaiy. zu Pat. rkralkrayoh savarnasamjn vidhey: atra canayoreva srutatvn mithah savarnasamjn vijnyate, na tv etayor anyena saheti bodhyam Da in diesem vrtt. eben nur r und / gehrt werden, wird erkannt, da der Name 'gleichlautig' gegenseitig gilt; nicht jedoch hat man zu verstehen: [der Name 'gleichlautig1] fr diese zwei und einen andern [Laut]." 72 Zitiert nach der Ausgabe von Pandit Rma SastrT Mnavalli, Benares 1885. 73 Wie brigens im Titel auch angedeutet: Praudhamanoram 'Die ppige Geliebte' und = ' [Der Kommentar (v/dkhy)] der das Denkorgan (manah) der Fortgeschrittenen (praudha) beschwichtigt (ram) '. Belvalkar vermutet (Systems of Sanskrit Grammar p. 47), da der Kommentar 'Praudhamanoram'
71

genannt sei, um ihn von der 'Blamanoram' zu unterscheiden, die er eine verkrtze Fassung (agridgement) der Praudham. nennt, und zweifelnd ('perhaps') demselben Autor zuschreibt (vgl. Generalindex unter Blamanoram). Das mag jedoch auf sich beruhen bleiben, da der Autor der Blam. sich selbst Vsudeva nennt, von Bhattoji als einer von ihm verschiedenen Persnlichkeit spricht (vgl. Blam. ed. Trichinopoly1910 p.2: tac ca Praudhamanoramym svayam eva mlakrt prapacitam eva), u n d schlielich sogar NgojTbhatta zitiert (z. B. p. 13: sstrsiddhatvakrysiddhatvayoh phalabhedas tu Sabdendusekharevyaktah, vgl. LSS. p. 40f.), noch auch die Blamanoram ein 'abridgment' der Praudhamanoram genannt werden darf. 74 Ich verzichte absichtlich auf diakritische Zeichen !

318 Paul Thieme

mand jedoch praktische Gelegenheit hat, z. B. den Vokal '/' anzureden oder anzurufen. Ich kann nicht sagen, ob Bhattoji diesen eigentlich recht vernnftigen Einwand 75 irgendwo ausdrcklich vorgebracht hat. Da er bestand, ergibt sich aus Ng.'s Bemerkung zu vrtt. 5 zu 1 . 1 . 9 (Bhsya I): krasabdo devatvclty eke. ,,Einige [sagen], das W o r t 'Ikra' bezeichne eine G o t t h e i t " . Das kann nur als eine A n t w o r t auf einen derartigen Einwand verstanden werden. Da der Dksita * r und */ nicht 'ac' nennt, kann er sie auch nicht 'lang' heien. So sagt e r : Z u beiden v r t t . mu [ausdrcklich] gelehrt w e r d e n , da es sich [bei diesen Lauten] um je eine Lautdoppelheit handelt, die z w e i Moren betrgt (aber nicht 'lang' heit, da die Lautdoppelheit nicht den Namen 'ac' trgt). Die erste Lautdoppelheit (*r) hat zwei konsonantische r in der Mitte. Diese machen e i n e Mora aus. A u f jeder Seite [dieses rr] ist eine Vokalpartikel. Diese macht eine weitere Mora aus ( + rr + d). Die zweite Lautdoppelheit (*/) hat zwei konsonantische / in d e r Mitte. Alles weitere wie im ersten Fall." Der aufmerksame Leser w i r d noch eine w e i t e r e Konsequenz beobachten. Da der Dksita * r und */ nicht als 'ac' ansieht, lt er keinen Sandhi eintreten und sagt: savarne r, savarne /. Gewi, auch Ktyyana sagt rti *r . . ., Iti */. . . Dies kann man jedoch, auch wenn r und / 'ac' heien, nmlich m i t 6.1.127, rechtfertigen. Schlielich braucht Bhattoji nicht anzugeben, da *r und */ 'halbverschlossen' sind. Diese Angabe ist ja nur notwendig, wenn man zeigen w i l l , da sie zwar gewnlichem r und / nicht 'gleichlautig', jedoch ihnen 'nchstbenachbarte' 'lange' Laute sind. Die Praudhamanoram verteidigt die Beschreibung von * r und */ als 'zweimorige Lautdoppelheiten' : yat tu prc vykhytam dlrghe prpte hrasva rkra kras ca vidhlyata iti tad Bhsya-Kaiyatdivirodhd upeksyam. ,,Was jedoch der 'Frhere' e r k l r t hat: '[In den vrtt. rti. . . rv etc.] w i r d der 'kurze' r- und /-Vokal gelehrt in einem Fall, w o ein 'langer' Vokal sich ergeben w r d e ' , das ist nicht zu beachten, da es im W i d e r s p r u c h zum Bhsya und zu Kaiyatausw. steht." Der 'Frhere', der so flschlich * r und */ als 'kurze' Laute auffat und offenbar vom Schriftbild v e r f h r t sie mit gewhnlichem r und / identifiziert, ist Vitthala, der Kommentator der PrakriykaumudT (siehe Prakr. Kaum. p. 67). Er steht m i t seinem I r r t u m allein in der gesamten panineischen Literatur, die zwar berwiegend *r und */ in der Seh r i f t mit r und / wiedergibt, sich jedoch ber die besondere N a t u r dieser Laute mindestens seit Patajali vllig einig ist.

5. Ngojbhatta (18. Jahrhundert) Seinen Ruhm als Grammatiker verdankt Ngojbhatta vor allem drei Werken : dem Uddyota76, einem Kommentar zum Pradpa und zum Mahbhsya, dem Laghusabdendusekhara77 einem Kommentar zur Siddhntakaumud und dem Paribhsendusekhara78, einer Monographie ber d i e grammatischen Interpretationsregeln, die von Pnini nicht ausdrcklich gelehrt sind. Dies letztere Werk ist durch
75 Den wir ja auch fr die Ksik voraussetzten, oben S. 310. 76 Zitiert nach der ausgbe von P. Sivadatta D. Kuddla (Mahbhsya, Bombay1917).

Zitiert nach der Ausgabe von Narahari Sstri Pendse, Benares 1927 ff. (Abgekrzt = LSS.) 78 Ed. Kielhorn, Bombay 1868. (Abgekrzt: PS.)

77

319 Bhsya und Erklrer

Kielhorn's meisterhafte bersetzung79 und Erklrung allgemein zugnglich, und deshalb hervorragend geeignet als Einfhrung in die Grammatik, wie sie in Indien betrieben wird, zu dienen. Der Ehrenplatz, den Ngojbhatta als der Begrnder der 'neuen' Schule der indischen Grammatik hlt, seine fast autoritative Geltung bei den modernen Grammatikern80, kommt ihm mit Recht zu. Er beherrscht das Bhsya unendlich viel besser, als irgend jemand vor ihm, und berragt seine Vorgnger gleicherweise in Grndlichkeit, Scharfsinn und Originalitt. Kaiyatasagt von sich, da er den Ozean des Bhsya langsam ausschreitend gleich einem Lahmen auf der von Bhartrhari gezimmerten Brcke berquert habe (Einleitung V. 7), womit er nicht nur seine Sorgfalt in der Auswahl der richtigen Interpretation charakterisiert, sondern auch die Vorsicht, mit der er fast gnzlich an der Erklrung der einzelnen Stelle klebt, mit der er darauf verzichtet, eigene neue Interpretationen zu versuchen. W i r haben an einem kleinen Beispiel gesehen, da Haradatta und Bhattoji Dlksita wesentlich auf ihm beruhen, da auch sie keine entscheidenden Neuerungen einfhren : Sie verschrfen, verfeinern, treiben auf die Spitze gelehrte aber inspirationslose Grbler. Ganz anders verhlt es
79 Paribhasendusekhara, Part II: Translation and Notes, Bombay 1873 ff. 80 Natrlich versucht man auch, ber ihn hinauszukommen und seine Gedanken fortzuentwickeln. Die Freude am 'Zerbrechen' (khondano) frherer Lehrmeinungen ein Kunst, die Nag. erfolgreichst in seiner Kaiyata-Kritik entwickelt und am Beseitigen von Unstimmigkeiten durch nterpretatorische Spitzfindigkeiten (phakkik) ist, namentlich in Benares, auch heute noch lebendig. Wer sich von der modernsten Richtung einen Begriff machen will, mag sich in die Vijay vertiefen, den beliebtesten Kommentar zum PS.

Ich gebe je ein Beispiel fr ein 'Zerbrechen' und einen Interpretationskunstgriff. a) Einige Grammatiker meinen, der Grundsatz, da ein Augment (gama) ein Teil des Elementes wird, zu dem es hinzugefgt wird, und deshalb bei Nennung des augmentlosen Elements auch das Element m i t Augment genannt st (Paribh. 11), komme nicht in Anwendung, wenn es sich um ein Augment zu einem Laut handelt. Nag. zu Paribh. 11 verwirft diese Ansicht, da sie drei Bhsyastellen widerspreche (vgl. Translation p. 57 ff.), und begngt sich lediglich mit der Feststellung, da Paribh. 11 keine allgemeine Gltigkeit hat. Die Vijay sucht nun zu zeigen, da die von Ng.

amgezogenen Bhsyastellen anders zu interpretieren seien, und jene Ansicht doch zu Recht bestehe. b) Die Bemerkung Ng.'s zu Paribh. 6: ddivisaye tu sorvdesatvam vinnubandhatvasyaivbhvennuprvyt siddham ,,Insoweit da usw. in Frage kommen, ergibt sich aus der Reihenfolge [der grammatischen Operationen] (da sie fr das ganze Original und nicht fr seinen letzten Laut substitutiert werden), insofern der Name 'onubondha' berhaupt nicht vorhanden st, solange da usw. nicht fr das Ganze substitutiert sind", steht im Widerspruch zu seiner eigenen Ansicht (vgl. Translation p. 34 Anm. 1 Abs. 2), die im LSS (zu Pn. 7.1.17, p. 281) und bei Kielhorn I.e. auseinandergesetzt ist. Man interpretiert den zitierten Satz deshalb, indem man vina statt mit sarvdesatvam mit abhvena konstruiert, in der folgenden knstlichen Weise: ,,Auch ohne das NichtVorhandensein des Namens 'onubandha' (i. e. auch wenn d in da usw. schon von vornherein 'anubondha' heien, wie es nach LSS in der Ordnung ist) ergibt sich korrekt, da da usw. fr das Ganze substituiert werden auf Grund der Aufeinanderfolge (von da und , die in da verborgen st, und dem Suffix den Charakter eines aus mehreren Lauten bestehenden Substituts: d-f gibt)".

320 Paul Thieme

sich mit Nagojbhatta. Er hat den Unterschied selbst gefhlt und in seiner bescheidenen Weise zum Ausdruck gebracht: Dem Versprechen Kaiyata's, das Bhsya 'traditionsgem' (yathgamam) erklren zu wollen (Einleitung V. 5), stellt er sein eigenes Prinzip im Uddyota gegenber (Einleitung V. 4): . . . Bhsya-Pradipavykhynam kurve 'harn tu yathmati.

Ich jedoch interpretiere das Bhsya und den Pradlpa (nicht traditionsgem, sondern) wie ich es fr richtig halte". Er braucht keine Brcken und Krcken. Wie ein groer Virtuose handhabt er mit scheinbarer Leichtigkeit und mit absoluter Sicherheit Patajali's schwieriges Werk, spielt er mit den im Bhsya vorgebrachten Gesichtspunkten, Mglichkeiten und Lehrmeinungen ein kongenialer Denker, der sich niemandem beugt als dem 'Herrn der Schlangen' (LSS. V. 3 : natv PhanJsam . . .), das heit dem Gott Siva, dem Offenbarer der Grammatik, und dem Patajali81. Siegreich behauptet denn Ng. das Feld gegen Kaiy. auch bei der Erklrung unserer Bhsyastelle im Uddyota. Es mu fallen zunchst Kaiy.'s Deutung des vaksymi des crya(oben S. 311):
vacanasya kvpy adarsand ha vykhysymiti. anye tu lanstrasthkrasynunsikatve 'oto Irntasya' ity atra bhagavn Pninir akrarp noccrayet pratyhrenaiva nirvht. tasmd aprvam vacanam kryam ity eva Bhsysaya ucita ity huh Kaiy. sagt: ' [ D e r Sinn von

'vaksymi' ist:] Ich werde interpretieren', da sich der Lehrsatz (da ein an als Substitut fr /von /gefolgt ist) nirgends [tatschlich ausgesprochen] findet. Andere jedoch [deren Meinung ich teile] sagen : Wenn das in SS. 6 (an) stehende a nasaliert wre [und demnach (nach 1.3.2) den Namen it empfangen wrde], dann wrde der Erhabene Pnini in der Regel 'ato rntasya' (7. 2. 2) kein / aussprechen, da er [in diesem Fall] lediglich durch einen (nach 1.1. 71 gebildeten) pratyhra ra') [in welchem sowohl r als / enthalten wren] seine Absicht erreichen knnte. Deshalb ist allein als Ansicht des Bhsya angemessen, da ein neuer Lehrsatz aufgestellt werden mu82. Unrichtig sind auch einige der Bemerkungen Kaiy.'s zu Opp. 2 (oben S. 311 f.), wo es heit: asamhitym rkre hotr rkr a iti rpam . . . kre 'samhitym hotr kra iti Wenn nicht engster Zusammenschlu (vgl. Pn. 1. 4.109) vorliegt, so ergibt sich, falls r folgt, die Form hotr rkrah, falls /folgt, hotr kr ah (insofern die auf Pn. 6.1. 72 folgenden Regeln nur unter der Voraussetzung 'samhitym' gltig sind)". Ng. bemerkt, da in einem Kompositum stets 'engster Zusammenschlu' [der Glieder] statt haben mu (samase samhit nity), und da eine solche Form (wie hotr-^ krah) auer in einem Kompositum nicht vorkommt (asamse tv Jdrsam rpam durlabham). Das im Bhsya stehende hotr krah (p. 62 Z. 28, 63 Z. 7) ist [nicht eine in der Sprache verwendete Form83, sondern] das
Zu Patajali als Schlange vgl. oben S. 315 Anm. 65. Kontrastiere die Anrufung des grammatischen Dreigestirns und der siddhntasthpakh bei Bhattoji Dksita oben S. 315. 82 Ausfhrlicher LSS. (p. 9 ff.) zu Siddh. Kaum, zu den SS. (lanstre 'krasco r)auch das a in SS. 6 heit 'it'"). Hier heit es am Ende der von pare tu . . . ity huh eingeklammerten Errterung: ,,Deshalb(i. e. aus den
81

hier dargelegten Grnden) ist die Annahme, das o in SS. 6 sei nasaliert, ohne Autoritt. In der Regel ur an raparah ( 1 . 1 . 51) mu / [ausdrcklich] genannt werden, wie auch die 'Gleichmtigkeit' von r und / [ausdrcklich gelehrt werden mu] ". 83 Vgl. Ng. zu Pn. 6.1. 72 in LSS (p. 185): Und zwar ist'somhitd' die Aussprache mit einer natrlichen Trennung [der Worte] durch die Zeit

321 Bhsya und Erklrer

Aussprechen des Zustandes, in w e l c h e m d i e S u b s t i t u t i o n des langen Vokals zu erfolgen hat (bhsye tu hotr Ikra iti dJrghapravrttiyogyadasoccranam). Das heit, es handelt sich hier um eine grammatische Abstraktion. Ich habe diese Tatsache in meiner bersetzung angedeutet, indem ich in diesen Fllen hotr + Ikra h etc. umschrieben habe. Wichtiger ist das Folgende: Kaiy. meint, da bei Annahme der von Opp. 2 vorgetragenen Ansicht sich nicht nur hotrlkrah, sondern auch beliebig hotrlkrah als Sand h form fr hotr-\-Ikrah korrekt ergbe: Skale hotrlkrah "Wenn man der Ansicht des Skalya(Pn. 6.1.127f.) folgt, ergibt sich hotrlkrah". Hierzu Nag.: evam Skale hotr Ikra ity api cintyam, savarnatvavidhyakavacanbhve etadvicrasattvena tatra rty aka ity asyprpteh. iko 'savarna ity api na samsa iti nisiddham. nityagrahanam tu tatra bhsye pratykhytam ity huh Man sagt [mit Fug], da ebenso auch Kaiy.'s [Satz] 'Skale hotrlkrah1 bedenklich sei, da Pn. 6.1.128 ([Skalyasya] rty akah) sich hier nicht ergibt, insofern der Lehrsatz, der den Namen 'gleichlautig' [fr r und /] lehren wird, [noch] nicht existiert, insofern seine Existenz erst noch berlegt wird. (Wir knnen also rti in 6.1.128 noch nicht als 'vor r u nd / interpretieren.) [Auch kann man nicht sagen, da die Sandhiform hotrlkrah sich bei Ablehnung von vrtt. 5 zu 1.1. 9 aus Pn. 6.1.127 (iko 'savarneSkalyasya hrasvasca) ergbe. Denn] auch das Eintreten von 6.1.127 ist durch [die zustzliche Bemerkung] 'nicht in einem Kompositum' verboten. [Man knnte freilich sagen, da vrtt. 1 zu 6.1.127, welches diese zustzliche Bemerkung lehrt, nur von 'ewigen84 Komposita spricht]. Der Ausdruck 'ewig' [des vrtt.] ist jedoch im Bhsyazu 6.1.127 zurckgewiesen." Die Sandhiform hotrlkrahwohl zu unterscheiden von der grammatischen Abstraktion hotr-\- Ikrahergibt sich folglich erst bei Annahme des vrtt. 5 zu 1.1. 9 durch den Acrya nach 6.1.128. Ng. sagt deshalb zum 'samdhna' : khatva Ikra iti hotrikra ity asyopalaksanam '[Durch Nennung des Falles] khatva Ikrah ist [der Fall] hotrlkrah elliptisch (ebenfalls) genannt85".
einer halben Mora. Diese Regel (Pn. 6.1. 71 sa m h ityam 'in engstem Zusammenschlu') dient zur Verhinderung der in diesem Kapitel gelehrten Operationen im Falle man [mit einer Trennung der Worte] durch die Zeit einer weiteren halben Mora ausspricht. Die alten [Erklrer] sagen jedoch, da bei einer Trennung vin ber einer halben Mora weder Korrektheit (der gesprochenen) noch Verstndnis (der gehorten Worte) [von Pn.] angenommen werde. Fr diese hat denn Pn. 6.1. 72 kein (spezielles) Anwendungsgebiet. (Wenn man nmlich Worte im Gebrauch stets im engsten ZusammenSchlu aufeinanderfolgen lassen mte, wie etwa die Glieder eines Kompositurns oder Stamm und Suffix, mte die Kegel so interpretiert werden, da sie in der ganzen Grammatik gltig ist.) Dies ist der Sinn (der Feststellung Bhattoji's 'iti adhikrtya' 'nachdem 6.1. 72 als K a p i t e l berschrift gegeben worden ist')''. 84 i. e. durch Anfgung eines taddhita unauflslich gewordenen, 85 Wenn wir der Ausgabe der Padamanjar! trauen drften, wrden wir annehmen mssen, da diese Interpretation auf Haradatta zurckgeht. Hier lesen wir nmlich (S. 59): dam api siddham bhavati 'rty akah' khatva ikrah mtr Ikrah "[Wenn vrtt. 5 2u 1.1. 9 gelehrt wird] ergibt sich auch das folgende korrekt nach 6.1.128: khatva Ikrah, mtrlkrah." Leider verdankt das Beispiel mtrlkrah sein Dasein ganz offenbar lediglich einer Liederlichkeit des Herausgebers, und steht fr das dem Bhsya entnommene mala Ikrah. Man beachte die hnlichkeit der aksara fr tr und la\

322 Paul Thieme

Nag. bleibt jedoch nicht bei der K r i t i k Kai/.'s stehen. Seine treffende Bemerkung zum Ausdruck 'dlrghatvam* im ' samdhna habe ich m i r (oben S. 303 A n m . 20, 21) angeeignet, da sie erst die w i r k l i c h e Pointe herausbringt. Anderes ist weniger naheliegend, und hat fr uns den W e r t einer K r i t i k des Bhsya, das doch nicht alle Seiten des komplizierten Problems in Betracht gezogen hat. Nag. hat freilich ein Mittel zur Hand, diese Unterlassungssnden zu rechtfertigen, ein M i t t e l , das natrlich in s t r i k t e r Weise 'patanjaleisch' ist. Z u O p p . 2 heit es im Uddyota: tena ItJti. siddhnte f p/ vyavasthitavibhsaynabhidhnena v hotr-\- kre prvavrttikam garni + kre cottaravrttikam na pravartate. evam kr iti dJrghasya rkre pare prvasya tasyaiva Ikra uttarasya cpravrttis tata eva iti bhvah

,,Kaiy. sagt: '[Wenn man vrtt. 5 zu 1.1. 9 streicht und die vrtt. zu 6.1.101 in der Form lehrt, die Opp. 2 vorschlgt] dann [ergibt sich richtig] vor / usw.' [Er schweigt ber die Frage, ob sich dasselbe ergibt, wenn man Kty. folgt.] Die (diesem Schweigen) [zu Grunde liegende] Meinung86 ist: ,,Auch wenn man die vrtt. zu 6.1.101 in der Form, in der sie von Kty. gegeben sind, und auerdem vrtt. 5 zu 1.1. 9 lehrt ('siddhnte 'pi') (und folglich vrtt. 1 zu 6.1.101 rti *rv . . . interpretieren mu 'vor 'gleichlautigem' kurzen r o d e r / wird beliebig *r substituiert', und weiter annehmen mu, da beide vrtt. auch dann gelten, wenn / oder r vorausgehen, da r und / ja auch diesen 'gleichlautig' sind), hat das erste vrtt. (zu 6.1.101) nicht statt in hotr^r krah, und das zweite vrtt. nicht in garni + krah ('der /-Vokal der Wurzel garni'), da es sich (in vrtt. 1 und 2 zu 6.1.101) um eine vyavasthitavibhs handelt, oder aber weil die Meinung nicht ausgedrckt ist. So hat, aus eben demselben Grunde, das erste vrtt. nicht statt, wenn der lange Vokal der Wurzel kr von r gefolgt ist, noch das zweite vrtt., wenn derselbe Vokal von / gefolgt ist." Da also fr r + / und f-f-r nicht *r, und fr / + / und f + / nicht */ substituiert wird Mglichkeiten, die das Bhsya vllig auer Acht gelassen hat erreicht Ng. zunchst durch die Annahme einer 'vyavasthitavibhs'. Wenn eine vibhs 'vyavasthit' ist, hat die durch diese vibhs als beliebig statthabend gelehrte Operation in gewissen Fllen nicht, oder aber in gewissen Fllen alleinig statt.87 Inwieweit der Geltungsbereich solcher vibhs beschrnkt ist, hat man aus der Richtigkeit der sich ergebenden Formen festzustellen
(laksynusrd vyavasth bodhy, PS. p. 101). In unse r m Falle htte

man also festzustellen, da die in vrtt. 1, 2 zu 6.1.101 als beliebig statthabend gelehrte Substitution von *r und */ in den genannten Fllen nicht statthat, da das Resultat *rfr r + / usw. und */fr / + / usw. falsch wre. Ng. hat jedoch selbst im LSS. (zu Pn. 6.1.123, p. 151) nach Aufzhlung der Flle, in denen das Bhsya eine vyavasthitavibhs
86 Ng. ist ein ritterlicher Gegner. Solange Kaiy. sich nicht durch eine unzweideutige uerung blogestellt hat, nimmt Ng. an, da er etwas Richtiges meint. In dieser Hinsicht ist er typisch 'scholastisch'. Er unterscheidet sich von andern wesentlich nur dadurch, da er niemals dem Wortsinn Gewalt antut. 87 Vgl. PS., Paribh. 99 und Translation

p. 471 ff. Z. B.: nach Pn. 8. 2. 56 drfen wir bilden trono benen troto, in 'devatrtoh' hat jedoch die als beliebig gelehrte Operation n i c h t statt. Nach 6.1.123 drfen wir bilden gavgrom neben go'grom und goagrom (6.1.122) usw., i ngavksam hat jedoch die als beliebig gelehrte Operation a l l e i n ig statt.

323 Bhsya und Erklrer

annimmt, festgestellt, da w i r keine A u t o r i t t haben, mit diesem Interpretationsmittel auf eigene Faust zu operieren 8 8 . Deshalb versucht er eine andere Interpretation und sagt: anobhidhnena v ,,oder aber [und das st die richtige Auffassung] weil [der Sinn] nicht ausgedrckt ist". Den Schlssel zum Verstndnis dieser W o r t e bietet uns der PS. Zu Paribh. 13 heit es

dort (p. 13 Z. 13) : sthnyarthbhidhnasamarthasyaivdesateti siddhntt nach dem Prinzip, da nur dasjenige ein Substitut werden kann, was fhig ist, den Sinn des Originals auszudrcken . . .". Die an sich grammatisch korrekten Formen hot*karah usw. werden demnach nicht fr hotr-\- rkrah usw. substituiert, da sie nur fhig sind, den Sinn von hotr^ krah usw. auszudrcken. Es handelt sich hier nicht um eine grammatische Interpretationsmaxime, sondern um einen sprachphilosophischen Grundsatz, der natrlich seine Wurzel ebenfalls im Bhsya hat. Wenn Ng. meint, da ein Substitut fhig sein mu, den Sinn des Originals auszudrcken, spricht er nicht von einzelnen Lauten, die nach der Ausdrucksweise der Grammatik fr einzelne Laute substituiert werden. Denn nicht einzelne Laute, sondern nur Lautgruppen haben einen Sinn (vgl. z. B. Pat. zu vrtt. 15 zu SS. 5, bes. p. 32 Z. 7 ff.). Er fat also hot*lkrah als Substitut fr hotr+ Ikrah auf, und nicht */ als Substitut fr r + / . Da dies der sprachphilosophisch richtige Standpunkt ist, geht z. B. aus dem Bhsya zu vrtt. 5 zu 1.1. 20 (p. 75 Z. 10 ff.) hervor, wo sich der berhmte Vers findet: sarve sarvapaddes DksJputrasya Pnineh ekadesavikre hi nityotvam nopapadyote. Alle [Substitute] des Sohns der DksT, Pnini, sind Substitute fr ganze Worte. Denn wenn nur ein Te i I [eines Wortes] gendert wird, stimmt die [Annahme der] Ewigkeit [des Wortes] nicht". Ausfhrlicher und vielseitiger als im Uddyota sind die Lehrmeinungen ,die man aus unserer Bhsya-Diskussion abstrahieren kann, im LSS. untersucht und dargestellt. Da es sich um vielleicht besonders schwierige Stellen handelt, gebe ich eine vollstndige bersetzung. 1. LSS. (p. 45) zu Siddh. Kaum. (p. 8) : rjvarnayor mithah svarnyom vcyam. [Das Kompositum] 'rlvornayoh' ist zu zerlegen : 7 und der /Laut89. Unterbleiben der Substitution [in r/-] ist korrekt nach Pn. 6.1.12890.
88

Und man darf nicht sagen, da Pn. 6.1.124 (/ndreca ,,vor Indra wird immer ava fr das o von go substituiert") zwecklos sei, insofern die stndige Substitution von ava vor Indra sich durch [Annahme einer]
vyavasthitavibhs korrekt (nach 6 . 1 .

89

Richtet sich gegen die Ansicht der Praudhamanoram(ob. S. 317). 90 Identisch mit Praudhamanoram ad I.e. und Uddyota zu rkralkrayoh (Bhsya I), an welch letzterer Stelle auch eine andere Ansicht erwhnt ist:
vrttike 'py evam eva ptha ti prmni-

123) ergebe (wie die stndige Substitution von ava in gavksa, s. oben Anm. 1). [Und zwar deshalb nicht] da diese Regel (6.1.124) gegeben ist um anzudeuten, da fr [die Annahme] von] vyavasthitavibhss, die im Bhsya (zu 7. 4. 41 usw.) nicht ausdrcklich gelehrt sind, keine Autoritt besteht".

kh Diejenigen, fr die [der tatschliehe Gebrauch in einem Lehrsatz] Autoritt st, sagen: '[Pat. sagt rkrakrayoh] da es ebenso im vrtt. steht'". Das wird wohl deshalb zugefgt, weil es wunderlich erscheint, da Pat. diesen Sandhi nicht von vornherein als 'prayojana' vorgebracht hat, und ihn auch spter nicht

324 Paul Thieme

,,Allein um dieses yartt. willen w i r d / in den SS. namentlich aufgefhrt [nmlich in SS. 2: r k] (nicht aber um der Substitution des 'plutierten ' Vokals in kllptosikha ! (nach 8. 2. 86) willen, wie im Bhsyazu SS. 2 behauptet wird 9 1 . Gilt nmlich dieses vrtt., wrde / auch dann 'ac' heien, wenn es in den SS. nicht aufgefhrt wre, indem es als 'gleichlautig' nach 1 . 1 . 69 in der Nennung von reinbegriffen wre. Es wrde demnach auch dreimoriges II nach 1. 2. 27 'plutiert' heien und fr das 'ac' / nach 1. 2. 28 substituiert werden knnen). [Denn] wenn / in der SS. nicht namentlich aufgefhrt wre, knnte dieses vrtt. [durch welches dann der Grund, den Pat. fr die Auffhrung von / angibt, hinfllig w i r d ] berhaupt nicht gelehrt werden, da man befrchten mte, da / nicht existiert. ,,(Nun knnte man einwenden, da diese Auffassung gegen Pat. gehe, da ja dieses 'prayojona' fr die Auffhrung von / im Bhsya zu SS. 2 nicht bercksichtigt st. Das ist jedoch nicht der Fall.) [Da sich dies so verhlt] das ist am Ende des Bhsyazu SS. 3 auch angedeutet (wenn auch nicht ausdrcklich ausgesprochen). (Und zwar in der folgenden Weise:) ,, Nachdem man die Befrchtung vorgebracht hat, da sich (wenn man das Prinzip ablehnt, da durch Nennung von Lauten Teile von Lauten ebenfalls genannt werden) eine durch das Vorhandensein zweier Konsonanten bedingte Operation 9 2 flschlich erwhnt. Wie wir sahen, ist er ja nach Ng.'s Ansicht im samdhna elliptisch genannt (oben S. 321). 91 Im Bhsya zu SS. 2 wird untersucht, warum / in den SS. namentlich aufgefhrt wird. Ich gebe eine kurze Paraphrase: / kommt nur in Ableitungen von der Wurzel krp wirklich vor, z. B. in klpti-; klpta-. Das / dieser Formen st nun nach 8. 2.18 fr r substituiert, ist also wie alle nach 8. 2.1 gelehrten Substitutionen in Rcksicht auf eine vorhergehende Regel als nicht vorhanden zu betrachten. Alle vorhergehenden Operationen werden folglich an dem r von \/krp gettigt, fr welches dann (nach 8. 2.18) der jeweils 'nchstbenachbarte' /-Laut eintritt. In 8. 2. 86 dagegen ist das Substitut / bereits vorhanden, da 8. 2. 18 vorhergeht, und es ist deshalb notwendig, auch / 'ac' zu nennen. Kty. erklrt aus diesem Grunde im vrtt. 1 zu SS. 2, da / aufgefhrt sei, um die Substitution des plutierten Vokals zu erreichen. Er entrkftet das in vrtt. 4 mit dem in Paribh. 37 formulierten Prinzip, da etwas, das eine nderung in Bezug auf einen seiner Teile erfahren hat in diesem FalI ist ri (r) zu a/i (/) geworden nicht als etwas Anderes betrachtet wird, als was es vorher war. Insofern / in klpta- demnach immer noch als r betrachtet wird, heit es 'ac'. Pat. lehnt es jedoch ab, die Substitution des plutierten Vokals in kptasikhal auf diese Wiese zu erklren, da wir so zwar erreichen wrden, da / 'ac' heit, andrerseits aber das fr r gegebene Verbot der Substitution des plutierten Vokals (8. 2. 86 . . . anrto . . .) auch fr / gelten mte. / ist also nach seiner Meinung in den SS. notwendig. Kty. will freilich die genannte Schwierigkeit durch nderung der Formulierung von 8. 2. 86 beseitigen (vrtt. 5 zu S S 2.) Pat. bemerkt dazu: Die Auffhrung des /-Lautes in SS. 2, die den Zweck hat die Substitution des plutierten Vokals (nach 8. 2. 86) mglich zu machen, ist von Kty. unter nderung einer Regel (nmlich 8. 2. 86) zurckgewiesen worden. Das ist, als ob man eine latvC!) von einem hohen Bambusstamm herunterholt (i. e. eine unverhltnismige Anstrengung). Die brigen Grnde, die Kty. im vrtt. 1 fr die Notwendigkeit, / in den SS. aufzufhren, angibt, sind in den vrtt. 2 und 3 nebst Bhsya widerlegt. 92 Wie die in 8. 2. 86 gelehrte Substitution des plutierten Vokals, die fr einen 'positionslangen' Vokal statt-

325 Bhsya und Erklrer

nicht ergbe in safyanth usw. (i.e. savvatsarah, yallokam)93 (soda man also nicht salyyantahl usw. bilden knnte), da (zwar kk \nkukkutah\ nicht 'hol' heit, sondern 'samyoga', insofern das in den SS. genannte k nicht das 'gleichlautige' kk' mitnennt, da k nicht 'an' heit, yy usw. jedoch den Namen 'hal' erhalten mten, da) durch Pn. 1.1. 69 (an udit savarnasya cpratyayah) bei Nennung von y usw. (i. e. v, I) [die im hal von Pn. 1.1.7 halo 'nantarh samyogah genannt sind] auch das Doppel-/ usw. (i. e. fy, vv, II), das eine ganze Mora betrgt, mitgenannt ist94, insofern Pn. 1.1. 69 sich auch auf die Konstituenten eines pratyhra (in diesem Fall 'hol' in 1.1. 7) erstreckt95 [soweit sie 'an' heien, in diesem Fall also y, v, /], wird dort nmlich gesagt: Da ein eine [ganze] Mora betragender Konsonant nicht gelehrt ist, und da etwas, das nicht gelehrt ist, nicht existiert, und da etwas, das nicht existiert, nicht erhalten werden kann, ist Doppel-y usw. wenn 1.1. 69 statthat, nicht mitgenannt96 (kann also auch in dem hal von Pn. 1.1.7 nicht mitgenannt sein, und ist deshalb als 'Doppelkonsonanz' zu erkennen). Auch [kann man] nicht [gegen dieses vrtt. einwenden], da sich die Substitution des ' plutierten ' Vokals (nach 8. 2. 86) in kllptasikhal flschlich nicht ergebe, da [wenn rund /gleich lautig heien] das Verbot ' nicht fr ein kurzes f (8. 2. 86) [auch fr /] gelten msse97. Denn es ist kein Fehler, wenn man vermutet, da r und / sich manchmal gegenseitig nicht erfassen (obgleich sie 'gleichlautig' heien), da sie verursachen, da [die Wurzeln mit
haben darf. 'Positionslang' (guru) ist ein Vokal vor einer ' Konsonantenverbindung' (samyoga). Den Namen 'samyoga' erhalten nach 1.1.7 aufeinanderfolgende 'hal' genannte Laute. 93 Aus sam -f- yant usw. nach 8. 4. 58. 94 Pat. zu vrtt. 13 zu SS. 3 (p. 26 Z. 24 ff.): i ho tu katharp 'sayyant, savvatsa roh, ya/okam, tafiokam' iti yatraitad osty an savarnn grhntiti. Uddyota: yatraitad astiti Bhsye. haio 'nantar ity atra yakrdibhir asya grahane haldvaybhva iti bhvah. Im Bhsya (i. e. n i c h t im Pradpa) [heit es]: yatraitadasti usw. Die zugrundeliegende Meinung st: Da in Pn. 1.1. 7 auch yy usw. durch y usw. genannt sind, [in yy usw.] nicht zwei [sondern nur ein 'hal'] vorhanden". 95 Bezglich der in Pn. 1.1.10 (njjhalau ,,ein 'ac' und ein 'hal' sind nicht gleichlautig") genannten pratyhra hat zwar 1.1. 69 nicht statt, da erst nachdem die Definition der 'Gleichlautigkeit' vollstndig gegeben ist, der Ausdruck savarnasya in 1.1. 69 verstanden werden kann. An andern Stellen der Grammatik, wo ein pratyhra genannt wird, hat jedoch 1.1. 69 statt, und nennen die in den pratyhra enthaltenen Laute auch ihre 'gleichmtigen' Partner, z. B. das in akah 6.1.101 enthaltene /' auch langes /Vgl. Bhsyazu vrtt. 4 zu 1.1.10 (p. 64 Z. 11-18). 96 Bhsya(zu vrtt. 13 zu SS. 3 (p. 26 Z. 26 f.) anupadistam sat katham sakyarp vijntum asac ca katham sakyam pratipattum W i e kann etwas, das nicht gelehrt ('gezeigt') worden ist, als existierend erkannt werden? Und wie kann etwas, das nicht existiert, erhalten werden"? Uddyota: anena Ikrpthe tasysattvasankay tayoh svarnyavidhir apy asakya iti scitam Hierdurch st angedeutet, da, wenn / [in den SS.] n icht aufgefhrt wre, auch der Lehrsatz, da [r und /] 'gleichlautig' heien, nicht gelehrt werden knnte, da man zu befrchten hatte, da / nicht existiert". Derselbe Einwand wird im Bhsya gegen die Anschauung erhoben, da / in den SS. nicht gelehrt zu werden braucht, insofern es als von r nicht verschieden aufgefat werden kann. (Siehe oben S. 324 Anm. 91).
97

326 Paul Thieme

symbolischem r und /] verschiedene anubandha haben " 98 . 2. LSS. (p. 148) zu Siddh. Kaum. (S. 27): (1) ni savarne rv: hotrkrah, hotrkrah. Iti savarne I v: hotlkrah, pakse rkrah svarnyt: hotrkrah. rti rv, Iti Ivety ubhayotrpi vidheyam varnadvayam (2) dvimtram. dyasya madhye (3) dvou rephau, tayor ek mtr. (4) obhito 'jbhakter apar. dviyasya tu madhye (5) dvau lakrau. sesam prgvat....

(1) [Bhatt. sagt:] 'vor einem gleichmtigen r\ da er will, da das im Stra(6.1.101 akah savarne dlrghah) stehende W o r t 'gleichlautig1 fortgilt. Er erreicht damit, da die vrtt. in Fllen wie dadhy rkrah (fr dad'hi + rkrah) usw. nicht statthaben". Die auf Grund dieser vrtt. (als Substitute) herzustellenden Laute heien 'ac', da sie [in den SS.] unmittelbar nach a i un (SS. 1) aufgefhrt werden (mssen)100. Daraus ergibt sich, da fr sie der
Kaiy. sagt (zum Schlu des Bhasyazu scheut er sich spezielle Vermutungen als berechtigt anzuerkennen, die nicht vrtt. 5 zu 1.1. 9) : rditm Iditm ca bhedennubandhaniresd bhedena cop-im Bhsya ausdrcklich geuert sind. Er entnimmt als der Tatsache, dnd anubandhakryesu parasparada r und / verschiedene anubandha grahanbhvd vyatikarbhvah. Die Wurzeln mit dem anubandha r und die sind, die allgemeine Vermutung, da Wurzeln mit dem anubandha / werden sie sich manchmal nicht erfassen. Nach der Erklrung eines Kommentanicht gegenseitig durcheinandergetors (LSS p. 45 Anm.) beruht seine bracht, weil, da die anubandha [r und Ansicht auf dem logischen Prinzip: /] als verschieden [im Dhtuptha] asati bdhake pramnnm smnye aufgefhrt sind und als verschieden paksaptah ,,Wenn nicht eine spezielle [in der Grammatik] zitiert werden, Einschrnkung vorhanden ist, haben r und / sich gegenseitig nicht erfassen, 'Erkenntnismittel' ['Schlufolgerung' sofern von anubandha bedingte (anumna) usw.] eine Tendez zum Operationen in Betracht kommen". Allgemeinen". In diesem Fall Hiegegen hat schon DTksita im handelt es sich um ein 'kalpana' Sabdakaustubha(zu SS, 2, p. 51) ein(Vermutung), eine Abart des anugewendet, da wir daraus, da die mna. Wie man z. B. aus dem VorVerschiedenheit der anubandha r und handensein von Rauch auf Feuer im / einen Sinn haben mu, zwar schlieallgemeinen, nicht aber auf ein en drfen, da ein fr eine Wurzel bestimmtes groes oder kleines mit dem anubandha I gelehrte OperaFeuer schlieen kann, ist auch in tion nicht auch fr eine Wurzel mit diesem Fall nur eine allgemeine dem anubandha r gilt. Nachdem wir dies geschlossen haben, hat jedoch die Vermutung mglich. Verschiedenheit einen Sinn, und wir drfen nicht noch weiter schlieen, da auch eine fr eine Wurzel mit dem anubandha r gelehrte Operation nicht fr eine Wurzel mit dem anubandha I gi It. Deshalb und wegen der Form kptas\kha\ nimmt er an, da die Auffhrung von / in den SS. den Zweck hat, anzuzeigen, da die 'Gleichmtigkeit' von r und / nicht allgemein gltig ist (I. c. . . . rkralkrayoh svarnyasynityatrp jnpayitum kartavya eva Ikropadesah). Nag. hat oben eine andere 'Veranlassung' fr die Auffhrung von /gegeben, die er mit dem Bhs/a begrnden zu knnen glaubt. Auch Vgl. oben S. 301 (Def. Arg. 2). Das ist zwar in der Siddh. Kaum, nicht geschehen, im Sabdakaust. wird jedoch Auffhrung unter den 'ac' verlangt (siehe oben S. 316). Ng. definiert die Stelle, an der *r und */ zu denken sind. Man erinnere sich, da nach vrtt. 6 ff. nebst Bhsya zu SS. 5 auch die 'unangeschirrt ziehenden' Laute (visarjaniya, jihvmiya, upadhmnlya, anusvra usw.) in den SS. aufzufhren sind, was ebenfalls niemals in die Praxis umgesetzt wird. Im Uddyota bemerkt Ng. zu Kaiy. zu Def. 4: [Kaiy. sagt]: 'Wenn *r und */ 'ac' heien'. Die zugrunde liegende Meinung ist: 'dasie unter
100 99 98

327 Bhsya und Erklrer

p l u t i e r t e Vokal k o r r e k t substituiert wird 1 0 1 , und da sie (wenn sie fr r + r bezw. r + / substituiert werden) nicht von r gefolgt sind (was nach 1 . 1 . 51 der Fall sein mte, wenn man sie vor dem ersten anubhandha der SS. (n) lehren wrde). (2) [ B h a t t . sagt: 'in beiden Fllen ist ein] zweimoriger [ D o p pellaut herzustellen]'. ,,Eben weil sich das so verhlt, sind diese zwei v r t t . im Bhsya zu Pn. 1 . 1 . 9 zurckgewiesen, indem man sich daraufsttzt, da sich das erwnschte Resultat, welches darin besteht, da [die Substitution von * r bezw. * / f r r + r bezw. r + / ] beliebig statt hat, sich [auch ohne diese v r t t . ] allein durch 6.1.101 k o r r e k t ergibt, ,,insofern sich [als 'nchstbenachbarte' 'lange' Substitutionsvokale] k o r r e k t ergeben ,,A. fr die zwei r (in hotr-\- rkrah) 1. zuweilen der zwei konsonantische r in sich tragende Laut (*r), insofern er 'nchstbenachbart' ist auf Grund der Eigenschaft102 welche darin besteht, da er (wie das Original) zwei konsonantische r enthlt [r (ara) + r (did) benachbart zu *r (drrd)], 2. zuweilen der 'offene' Laut (r), insofern er nchstbenachbart ist auf Grund dessen (i.e. der Eigenschaft), da er (wie das O r i ginal) offen ist [r (did) - f r (prd) benachbart zu r (ara)] ; ,,B. wenn / f o l g t (in hotr-{- Ikroh) 1. zuweilen der zwei konsonantische / enthaltende Laut (*/), insofern er 'nchstbenachbart' st auf Grund dessen (i. e. der Eigenschaft), da er (wie das Original) ein konsonantisches / enthlt [f (ara) + / (did) benachbart zu */ (dlId)], 2. zuweilen der offene Laut f, insofern [er 'nchstbenachbart' st, insofern] er [dem Original] hnlich 103 ist auf Grund dessen (i. e. der Eigenschaft), da er offen st, und insofern er [dem Original]
den 'ayogavho' (visorjonlyo usw.) aufzufhren sind1." Das heit: da sie als aufgefhrt zu d e n k e n sind. 101 Vgl. Sabdakaust., oben S. 316. 102 Nachbarschaft kann beruhen auf Artikulationsstelle (sthno), Sinn (artha), Eigenschaft {guno) und Zeitdauer (promano). Vgl. PS. zu Paribh. 13. Eine Eigenschaft ist auch die 'Artikulationsweise1. Ng. zhlt im folgenden nicht alle 'nchste Nachbarschaft' begrndenden Faktoren auf, sondern nur diejenigen, die uns veranlassen unterschiedliche und doch gleichwertige 'nchste Nachbarschaff anzunehmen, r und *r sind dem Original gleich nchst benachbart, soweit Artikulationsstelle und Zeitdauer in Betracht kommen, die in beiden Fllen die gleichen sind. Auerdem haben sie je eine Eigenschaft, die dem andern fehlt. Von diesen Eigenschaften ist die Rede. 103 'Nahe benachbart' und 'hnlich' sind synonym. Vgl. z. B. die Gleichung ontorotomoh = sodrsotomoh, Siddh. Kaum, zu 1.1. 50 (p. 13). Da ntaryam als Abstraktum zu ontoratamah 'nchst benachbart1 vergeben ist, verwendet Nag. smyam 'hnlichkeit' als Abstraktum zu antarah 'nahe benachbart1. Whrend */ schon insofern es I enthlt, 'nachstbenachbart' zu r-f / st, da es keinen andern 'langen' Ihaltigen Laut gibt, st f insofern es r enthlt nur 'nahe benachbart' zu r und /, da es diese Eigenschaft mit *r teilt. Da es 'nchstbenachbart' ist und deshalb mit Ausschlu von *r substituiert wird, verdankt es der weiteren, unterscheidenden Eigenschaft, da es 'offen' ist. Man vergleiche das Schulbeispiel fr 'nchste Nachbarschaft' auf Grund einer Eigenschaft: vgghorih nach 8. 4. 62 fr vg + horih. Hier st das dem g 'gleichlautige' gh substituiert, insofern es als'tnend' und 'aspiriert' dem h 'nchstbenachbart' ist, whrend das dem g ebenfalls 'gleichlautige' kh nicht substituiert wird,

328 Paul Thieme

hnlich st auf Grund dessen (i. e. der Eigenschaft), da er ein konsonantiches r enthlt [r (ara) + / (9/9) benachbart zu r (ara)]. (3) [Bhatt. sagt: 'In der Mitte des ersteren Lautes (*r) befinden sich] zwei konsonantische r. [Diese machen eine Mora aus]'. Die Autoritt auch fr diese Angabe ist die nmliche Bhsyastelle. Denn nur wenn * r z w e i konsonantische r enthlt, ergibt sich in korrekter Weise die Einschrnkung104, da dieser Laut (*r) substituiert wird, wenn das Original zwei r enthlt ( r - f r ) , der andere Laut (*/), wenn das Original ein / enthlt ( r - f /). (4),,[Bhatt. sagt:] 'Um herum [befindet sich je eine Vokalpartikel usw.]'. Es ist zu ergnzen 'die zwei konsonantischen r-Laute'. Und zwar steht dieser [aus dem vorhergehenden zu ergnzende Ausdruck] im Akkusativ, da er mit dem W o r t 'um herum' konstruiert ist. ,,Die [der Angabe, da sich in * r eine Vokalpartikel (9) auf jeder Seite der zwei konsonantischen r-Laute [rr] befindet, zu Grunde liegende] Meinung ist: ,,Da, insofern eine Vokalpartikel von je einer halben Mora (pi) auf jeder Seite (des konsonantischen Bestandteils) in den in SS. 2 gelehrten Lauten [r und /] angenommen w i r d , es billig ist, da es sich ebenso verhlt 105 [soweit * r in Frage kommt]. ,,[Da sich eine Vokalpartikel auf beiden Seiten des konsonantischen Bestandteils von gewhnlichem r und / befindet, i. e. 1) da r und / nicht mit einem r bezw. / beg i n nen , sondern mit einem 9, und 2) da r und / nicht mit einem r bezw. / seh I i e e n , sondern ebenfalls mit 9, wird angenommen aus folgenden Grnden :] 1. Bei Errterung der Alternative, da durch Nennung eines Lautes ein [diesem Laut identischer] Teil eines [andern] Lautes [mit] genannt ist ist106, welche Alternative auf jeden Fall anzunehmen ist, damit z. B. die Substitution von n in einer Form wie mtrnm sich korrekt (nach 8. 4.1) ergibt (insofern dann V in 8. 4.1 auch den dem r identischen Teil des r (ara) und r (aara) mitnennt, und w i r somit zu interpretieren haben 'n nach r, mag dieses r nun allein stehen oder in r bezw. /"enthalten sein'), w i r d im Bhsyazu SS. 3 zwar [erstlich] die Befrchtung ausgesprochen, da in [einer Form wie] pralya [Hinzufgung von] tuk (nach Pn. 6 . 1 . 71) [sich ergeben wrde] (insofern nach dem genannten Prinzip auch der zweite Teil des ( = u + u) in dem Ausdruck 'kurzer Vokal' einbegriffen wre) 107 , in [einer Form wie] khatvbhih [Substitution von] ois [fr
insofern es als lediglich 'aspiriert' dem h nur hnlich ('nahe benachbart') st. Siehe LSS. zu 1.1. 50 (p. 84): Warum st [in 1.1 . 50] das Superlativsuffix (ontoro-tomoh) verwendet? Damit in vg gharih usw. nicht das dem vorhergehenden Laute 'gleichlautige' kh (nach. 8. 4. 62) substituiert wird, welches gettigt werden wrde, da es [dem Original] hnlich ist auf Grund dessen, da es aspiriert ist." 104 'yyavasth' : veiste visaye 'vasth 'Statthaben in speziellen Fllen', 'Beschrnktsein'. 'Beschrnkung' im transitiven Sinn ist 'niyamah'. Whrend fr irgend eine andere Annahme keine 'billige' Begrndung gegeben werden knnte, 106 vrtt. 6 zu SS. 3 vornoikodes vornogrohanena cet Wenn durch Nennung von Lauten Teile von Lauten [ebenfalls genannt werden] . . .". Das heit, z. B. durch Nennung von konsonantischem r auch der in r (sra) steckende konsonantische Teil, durch Nennung von a auch die Teile des langen 0 = 0 + 0. 107 Pat. zu vrtt. 7 (p. 24 Z. 6f.): . . . lyo, prolyo. hrosvosyo piti krti tug bhovotlti tuk prpnoti.
105

329 Bhsya und Erklrer

bhis] (nach Pan. 7.1. 9) [sich ergeben wrde] (insofern nach dem genannten Prinzip auch der zweite Teil des ( = a - f ) in dem Ausdruck 'kurzes Q einbegriffen wre)108, und in [einer Fgung wie] vc tarati [Eintritt] des durch [das Vorhandensein von] zwei loc genannten Lauten bedingten [Suffixes than] (nach Pn. 4. 4. 7) [sich ergeben wrde] (insofern nach dem genannten Prinzip auch die beiden Teile des ( = a + a) in dem Ausdruck 'ac' einbegriffen wren, und somit vc- zwei 'ac' enthielte)109, und [zweitens] eine Beschwichtigung gegeben, welche darin besteht, da [man annimmt, da] die Vermutung notwendig gemacht ist, da [obgleich der genannte Grundsatz besteht] eine grammatische Operation, die bedingt ist durch einen Teilen [eines andern] Lautes hnlichen, selbstndigen Laut, nicht statt hat, wenn diese Teile des [andern] Lautes nicht als abgetrennt wahrgenommen werden110), da es einen Sinn haben mu, da tuk ausdrcklich als [in gewissen Fllen] nach einem langen Laut (6.1. 75) [anzufgend] gelehrt ist111,
Pat.zu vrtt. 8(p. 24Z.11 f.):khotvbhih, mlbhih. oto bhisa ois (7.1. 9) ty oisbhvah prpnoti. 109 Pat. zu vrtt. 8 (p. 24 Z. 26 f.): ekovarnavoc co drgho bhovatti voktavyom. kirn proyojonoml vc tarotlti dvyajloksonos thon m bhd ti. 110 Pat. zu vrtt. 9 (p. 25 Z. 4): nvyopovrktosyvoyovosyvoyovsrayovidhir bhovoti. Z. B. wird u in nicht als von dem ersten Bestandteil des getrennt wahrgenommen, es wird nur abstrahiert; r in r (ara) dagegen wird als von dem vokalischen Bestandteil getrennt wahrgenommen, man hrt in r tatschlich zwei verschiedene Laute. Soweit u in in Frage kommt, gilt also eine durch einen 'kurzen' Vokal bedingte Regel nicht, whrend fr den konsonantischen Bestandteil des r eine durch r bedingte Regel Anwendung findet. Dieser Grundsatz ist im Bhsya durch einen Beleg (drstnta) aus dem tglichen Leben gerechtfertigt, ist also 'lokonyyosiddho' (vgl. zu diesem Terminus Kielhorn, PS. Translation p. IX f.). Da es jedoch mglich und erlaubt ist wenn auch nicht notwendig auch fr einen solchen Grundsatz ein 'jnpoka' zu finden, bemerkt bereits Kaiy., da die verschiedenen jnpoko, die von Patajali zu vrtt. 7 und 8 in einem speziellen Sinn genommen werden (siehe im folgenden !). den allgemeinen Satz vermuten lassen, der im vrtt. 9 durch einen lokanyyo begrndet ist,
108

und stellt Nag. an unserer Stelle das Prinzip als 'jnpokosiddho' dar. 111 Pat. zu vrtt. 7 (p. 24 Z. 7 ff.): ,,Das von Pnini eingeschlagene Verfahren macht die Vermutung notwendig (jnpayati), da (obgleich das in vrtt. 6formulierte Prinzip gilt) eine durch einen 'kurzen' Vokal bedingte Regel nicht statt hat, wenn ein 'langer' Vokal vorliegt, insofern er [ausdrcklich] (in 6 . 1 . 75) verordnet, [da das Augment] tuk vor ch nach einem 'langen' Vokal [angefgt wird] (wofr keine notwendig machende 'Veranlassung' bestnde, wenn dies sowieso nach 6 . 1 . 73 geschhe. Die Regel 6 . 1 . 75 wrde also ohne das erschlossene Prinzip sinnlos sein, und erhlt einen Sinn, wenn esgilt.) (Opponens:) ,,Das ist nicht 'etwas, das eine Vermutung notwendig macht' ('jnpoko', Trapp: 'beweiskrftig' ! Kielhorn Paribh. S. Translation p. V: ,,Jnpoko . . . is applicable to any term employed by Pnini, or to any rule given by him, or in short to any proceeding of his, which would be meaningless or superfluous (vyortho), or for which it would be absolutely impossible to assign a reason, if a particular Paribhs did not exist, but which appears necessary and serves a purpose (i. e. is choritrtho), as soon as and only when that Paribhs has been adopted, and which on that account indicates the existence ofthat Paribhs etc."). Denn diese Regel hat eine andere

330 Paul Thieme

d a es einen Sinn haben mu, da [an das in 7 . 1 . 9 genannte a] ein t angefgt ist112, und d a es einen Sinn haben mu, da das in 4. 4. 7 stehende ( W o r t ) 'nau-' ausdrcklich genannt ist 113 ; jedoch w i r d w e d e r die Befrchtung ausgesprochen, da z . B. in prtar rtam sich [nach] Pn. 8. 3.14 (Schwund des r vor r) [ergeben mte] (insofern nach dem genannten Prinzip der konsonantische Bestandteil des r durch den Ausdruck 'vor r' mitgenannt wre), und ,,z. B. in tad Ikrah sich [nach] Pn. 8. 4. 60 [/ fr den Dental (d)] [ergeben mte] (insofern nach dem genannten Prinzip der konsonantische Bestandteil des / durch den Ausdruck 'vor /' mitgenannt wre), noch auch eine Beschwichtigung (solcher Befrchtung) gegeben 114 . (Hieraus geht denn klr!ich hervor, da der konsonantische Bestandteil des r und / sich nicht am Anfang dieser Laute befindet, somit r in prtar rtam berhaupt nicht vor r, sondern vor einen Vokal (9), und d in tad krah berhaupt nicht vor /, sondern vor einen Vokal (9) zu stehen kommt.) 2. A n d r e r s e i t s heit es im Bhsyazu SS. 5115: 'die (Laut) part i k e l , welche auf r (in r) folgt'. 'notwendig machende Veranlassung'. Was ist diese Veranlassung? (Opponens:) [Die Veranlassung ist, da Pnini im Sinn hat:) 'Ich werde in 6.1. 76 lehren, da [nach einem 'langen' Vokal an Ende eines Wortes Antritt von tuk] beliebig ist. (6.1. 75 mu demnach gelehrt werden, um 6.1. 76 mglich zu machen.) ,, [Wenn sich das so verhlt,] dann [macht das von Pn. eingeschlagene Verfahren die Vermutung des obigen Prinzips notwendig] insofern er zwei Regeln lehrt (wo eine einzige gengen wrde). Denn wenn das Prinzip nicht bestnde (torath), wrde er nur (in einer einzigen Regel) formulieren dlrght podntd v [statt: dirght. padntd v]". (Da es schwerer ist, zwei Regeln statt einer zu lehren geht aus Paribh. 121 hervor, wenn auch das hieraufgestellte Prinzipals zu weitgehend abzulehnen ist.) 112 Pat. zu vrtt. 7 (p. 24 Z. 12): taporokoronasmorthyn no bhovisyoti Da es einen Sinn haben mu, da [an 0 in oto bhiso ois (7.1. 9)] ein t angefgt ist, wird [Substitution von ois fr bhis in khotvbhih] nicht statthaben. (Wenn nmlich der Ausdruck 'nach ot' ('kurzem' 0) auch den zweiten Teil des bezeichnen sollte, wrde es gengen, einfach zu sagen 'nach 0'".
113 Pat. zu vrtt. 8 (p. 25 Z. 1 f.): . . . nougrahonom jnpokom dirghd dvyojloksono vidhir no bhovotiti Die Nennungen von nou- (in 4. 4. 7) ist ein jnpoka, da eine Operation die durch [das Vorhandensein von] zwei 'ac' bedingt ist, nicht nach einem 'langen' Vokal statt hat [obgleich es bei Annahme des genannten Grundsatzes als zwei 'oc' enthaltend gelten mu], (Wenn nmlich solche Operation statt htte, wrde die spezielle Nennung von nou- in 4. 4. 7 sinnlos sein.)" 114 Man knnte sagen, da die Flle 'prolyo' usw. die Flle 'prtor rtom' usw. elliptisch mitbezeichnen. Dagegen sprichtdiesorgfltige Behandlung, die Pat. jeder einzelnen der aufgefhrten Befrchtungen zu teil werden lt, und entscheidend die Tatsache, da die fr prolya vorgebrachte Beschwichtigung fr angebliches prtor + rdtan nicht in Frage kme. Es htte eine andere Beschwichtigung vorgebracht werden mssen, die uns auch darber aufklrt, warum in mtrnm usw. der in vrtt. 6 gelehrte Grundszta gilt, in prtor rtom (angeblich = protor -\rdtom) aber nicht. 115 Dies ist ein merkwrdiger Irrtum: der angefhrte Satz findet sich im Bhsya zu SS. 3 (p. 26 Z. 4 f.). Vgl.

331 Bhsya und Erklrer

(Hieraus geht klrIich hervor, da der konsonantische Bestandteil der in SS. 2 gelehrten Laute sich auch nicht am Ende dieser Laute befindet.) (5) [Bhattsagt: 'In der Mitte des zweiten Lautes (*/)] sind zwei konsonantische /'. Die [zu Grunde liegende] Meinung ist: da */ ein Genosse des erstgenannten Lautes (*r) ist, welcher zwei Konsonanten enthlt116 (insofern *r und */ in vrtt. 1, 2 zu 6.1.101 in gleichem Atem gelehrt werden). ,,Wenn117 dieses (vrtt. 1, 2 zu 6.1.101) gelehrt wird (atra), so ist die Substitution des ersten Lautes (*r) korrekt nur dann, wenn es sich um ein Einzelsubstitut handelt, dessen Original zwei kurze r sind (r + r); und die des zweiten Lautes (*/) nur dann, wenn es sich um ein Einzelsubstitut handelt, dessen Original ein kurzes r und ein / ist118. ,,Eben weil sich das so verhlt (ist durch 7t/' im ersten vrtt. nicht auch / genannt, und durch 7t/" im zweiten vrtt. nicht auch r119, und) wird, wenn / folgt nicht *r, und wenn r folgt, nicht */ substituiert. ,,Eben weil sich das so verhlt, ist im ersten vrtt. ausdrcklich gelehrt vor 'ku rzem' r, und hat der Ausdruck 'vor kurzem /' im zweiten vrtt. einen Sinn. Und [eben weil es sich so verhlt] ist es logisch einwandfrei, wenn im Bhsya zu 1.1.9 gesagt wird : 'Fr 7t/" werde ich 'rtah' lehren', und : 'indem man sich auf diese vrtt. sttzt braucht man nicht zu lehren, da r und / 'gleichlautig' heien [um hotrkrah fr hotr+ Ikrah zu erhalten]120. Jene beiden Laute (*r und */) sind 'halbverschlossen' [zu nennen], da wir keine Autoritt dafr besitzen, die (gegenber r drd und / dh] hinzugetretene Artikulationsweise, welche die eines konsonantischen r ist [*r=drrd, */ = 5//], zu ignorieren. Wenn sie nmlich 'offen' wren, dann wrden auch sie [wie f] durch offenes r und / mitgenannt, und bereits das vrtt. [zu 6 . 1 . 101] wrde sinnlos sein121, insofern [das Resultat] sich bereits aus 6.1.101 korrekt ergeben wrde".

auch Bhsya zu 8. 4.1 vartt. 2 (p. 452 Z.9). 116 Im PS. zu Paribh. 103 zeigt Nag., da wir aus dem Bhsya zu 2. 3. 8 lernen, da n u r h n l i c h e Dinge als Genossen behandelt werden (p. 103 Z. 8 f.: tena [karmopravacamyayukte dvitiy (2. 3. 8) iti stre bhsyena] hi sadrsnm eva prayoge sohyobhvo bodhitah). Nachdem Ng. die Bemerkungen Bhatt.'s erlutert hat, geht er dazu ber, zwei Feststellungen zu treffen, die sich nicht bei Bhatt. finden. us Warum sich das so verhlt, ist im Uddyota errtert (siebe oben S. 322). 119 Da wir zu solcher Annahme berechtigt sind, hat Ng. zu vrtt. 5 zu 1.1. 9 (siehe oben S. 326) errtert. 120 Whrend alle die in den mit 'eben
117

weil sich das so verhlt' (ata eva) eingeleiteten Stze die Richtigkeit der Behauptung voll besttigen, die sich demnach im Einklang mit dem Bhsya befindet, kann man nicht sagen, da die Formulierung Kty.'s oder auch die des Opponens ein 'jnpaka' enthlt, das uns ein Prinzip vermuten lt. Alles angefhrte autorisiert uns jedoch, die Substitutionen nur in den von Pat. genannten Fllen fr richtig zu halten. Danach haben wir dann die Interpretation einzurichten. 121 Wenn Kty. das vrtt. lehrt, sind *r und */ noch nicht 'ac' genannt. Erst nachdem das geschehen ist, kann man es zurckweisen. Es ist jedoch nicht von Anfang an sinnlos. Wenn *r und */ 'offen' wren, wrde es berhaupt nicht vorgeschlagen worden sein.

332 Paul Thieme

Es ist ein langer und ein dorniger W e g , der vom 'Phamsa' zum 'Ngesa' f h r t , und es ist nicht zu v e r w u n d e r n , da er oft gescheut, und immer wieder der Versuch unternommen w i r d , das Bhsya ohne die Hilfe grndlich verstandener Kommentare zu meistern. Ich habe hier dartun wollen, da es nicht nur notwendig ist, die W e r k e jener 'Spteren' zu lesen : w i r mssen auch ihren individuellen Charakter verstehen, ihre verschiedenartigen Betrachtungsweisen w e r t e n und in Rechnung stellen. Es besteht ein gewaltiger Unterschied zwischen der kurzen, aufs Praktische gerichteten und nicht immer konsistenten Ksik und der messerscharfformulierten, spitzfindigen und geschlossenen Siddhntakaum u d l ; zwischen Kaiy.'s pedantisch-genauem Pradlpa, in dem die Schwierigkeiten des Bhsyatexts mit nchternem Bemhen angegriffen werden^ und dem Uddyota, der sie meisterhaft beherrscht ; zwischen dem Sabdakaustubha mit seinen grblerischen berlegungen und selbsterfundenen phakkiks, und dem Sabdendusekhara mit seiner originellen und sicheren Virtuositt, seinen strikten Prinzipien. Sind auch die Glieder der Kette, die Bhartrhari und Ngojbhatta verbindet, von unterschiedlichem direkten W e r t fr die Interpretation des Bhsya, so knnen w i r doch kaum eines vernachlssigen, wollen w i r nicht das Verstndnis des nchsten in Frage stellen. Und wenn auch Ng. nicht w i r k l i c h das letzte W o r t haben soll, und fr unseren Geschmack immer noch zu viele A n nahmen macht, so scheint es mir doch auer Frage, da der W e g ins Bhsya ber das grndliche Kennenlernen dieses genialen Scholastikers f h r t . Es hat seine tiefere Bedeutung, da Kielhorn den Paribhsendusekhara bersetzte und erluterte, bevor er an die Herausgabe des Bhsyaging. Als ein Ausdruck der brahman sehen Civilisation, jener Lebenseinrichtung, die Indiens hervorstechendste Eigenart bildet, hat 'die Weisheit des Pandits', der diese Seiten gewidmet waren, schlielich noch ein hheres Interesse, als die einer schtzensw e r t e n Hilfe, die sie uns bei unsern philologischen Bemhungen sein kann. Sie zeigt uns harte Z g e : w i r finden in ihr all den Stolz, all die Unnahbarkeit des Brahmanen, der fern von der Masse eiferschtig seinen alleinigen Besitz, seinen einzigen Schatz htet, die Wissenschaft, welche s e i n e n Schutz gesucht hat, und die den Uneingeweihten sich nicht ergeben mag. W i r finden in ihr aber auch die f r o m m e Bescheidenheit des 'Zweigeborenen', dessen hchstes Glck es ist, die heiligen Schriften seiner Vorvter zu studieren und sich mglichst W o r t fr W o r t anzueignen, der die Siege und Niederlagen weltlicher Reiche auf indischem Boden vergessen, aber die Sprache der Urzeit seines Volkes bewahrt hat bis auf den heutigen Tag.

B. On the Identity of the Vrttikakra (1937-1938) Paul Thieme

In Pnini and the Veda, Thieme had dealt, among other things, with the conceptual and historical relationships between Panini's grammar and the Prtiskhya literature, the treatises attached to each of the schools or branches (skh) of the Veda. Among these, the Vjasaneyi Prtiskhya, which belongs to the W h i t e Yajurveda, is attributed to a certain Ktyyana. This raises the problem of the relationship between this Ktyyana and the Ktyyana who was the author of the vrttikas on Pnini's stras. Though Goldstcker had defended the identity of the two Ktyyana's, this view had been generally rejected by Western scholars. Thieme returned to it, and brilliantly defended it in " O n the Identity of the Vrtti-

333 The Identity of the Vrttikakra

kakra," here reproduced from Indian Culture 4,1937-1938,189209. The demonstration culminates in the analysis by Ktyyana, the vrttikakra, of the description of the pronunciation of the Subrahmany* a Vedic chant which plays an important role in the Vedic ritual. Here Ktyyana, the vrttikakra, shows himself to be not only a linguist conversant with the technicalities and theoretical issues raised by Pnini's rules, but equally at home in the technicalities of the more practical realm of the Vedapthaka, the priest who recites the Veda at the sacrifice. In my little book ' Pnini and the Veda' I have raised afresh, amongst other points, the question of the relative age of Pnini's work and the so-called Vjasaneyi Prtiskhya (properly to be styled : Ktyyaniya Prtiskhyastra). Following others I thought a comparison of such rules of the two Stras as agree in their purport, but show a somewhat different wording, to be especially helpful towards a final settlement of this old problem. I arrived at the conclusion that the rules of the Prtiskhya, though often lacking in that utmost brevity observable in the AstdhyyT, at several instances appear to reveal a deliberate endeavour to improve on formulations of Pnini's. Particularly convincing in this respect seemed to me the difference of Pnini 1.1.9 tulysyaprayatnam savornam and V. Pr. 1. 43 samnasthnakaransyaprayatnah savarnah, since this latter definition adds to the former an essential element (sthnakarana), which also the Vrttikakra proposes to supply when expressing in vrtt. 2 on Pnini 1.1.9 the desire to replace Pnini's rule by the more accurate one: sye tulyadesaprayotnom savarnam, his expression desa comprising the terms sthna and karana\ Yet my arguments have not been able to carry conviction. Even in face of an instance as the one given just now, Prof. Keith still maintains that ' it is easy to explain his (Pnini's) deviations from the V. Pr. by the desire to condense the matter of the latter.' 2 Now I am ready to admit for argument's sake that, as matters stand, the decision of the dilemma essentially rests on the appreciation of certain general considerations and, in the end, must be of subjective nature.3 Prof. Keith holds Pnini 1.1. 9 to be a condensation of V. Pr. 1. 43 ; I hold V. Pr. 43 to be an improvement on Pnini, but though having the weighty support of the Vrttikakra,
* [The translation of the Subrahmany runs as follows: "Subrahmanyom! Subrahmanyom! Subrahmanyom! O Indra, come! O possessor of bay horses, come! O ram of Medhatithi, o wife of Vrsanasva! O cow impregnator! O lover of Ahaly! O Brahman of the Kausika family! O usurper of the name of Gautma! Today (respectively: tomorrow, in two days, in three days) come to the Soma pressing, O generous one! Gods, Brahmans, come, come, come! " The inserted piece (asau yajate . . .) means: " N . sacrifices, the son of N. sacrifices, the grandson of N. sacrifices, the descendant of N. sacrifices, the father of N. sacrifices, the grandfather of N. sacrifices, the greatgrandfather of N. sacrifices, the father, the grandfather, the greatgrandfather of those that will be born sacrifices." Specimens of the Subrahmany as they are still chanted can be heard on an LP record album The Four Vedas (Asch Mankind Series, New York 1969).] 1 o.e., p. 92f. 2 Indian Culture, Vol. II, p. 741. 3 Thus O. Strauss, Deutsche Literaturzeitung 1936, p. 880, who is ndined however to accept my view as the more probable one.

334 Paul Thieme

I can see no short way t o convince my opponent that my appreciat i o n of the facts is correct. For the briefer w o r d i n g is Pnini's, and for somebody w h o takes brevity of expression t o be probably a sign of young age, this may be a sufficient argument.

If, then, the correctness of my contention be still open t o doubt, this much my lengthy discussions must have shown, t h a t it is at least possible t o accept Pnini's p r i o r i t y , and t o consider it on par w i t h the contrary assumption. This being, at least for argument's sake, admitted, we are faced w i t h the f u r t h e r question, w h e t h e r the author of the V r t t i k a and the author of the V. Pr., w h o is referred t o by his commentator Uvataas ' Ktyyoncrya,' may not be identical, after all. For it is hard t o agree w i t h Prof. Keith's view that the identity of a name like Ktyyana creates 'no probability at all ' w i t h reference t o the identity of the authors. 4 The probability created may be afaint one, it may even be altogether, deceptive, but anyone t h i n k i n g t h e V. Pr. t o be later than Pnini, must feel a keen suspicion, which he wants t o be either removed or confirmed by good reasons. Prof. Keith is persuaded that the V r t t i k a and the V. Pr. are by quite different hands : 'It seems t o have been f o r g o t t e n by Dr. Thieme that W e b e r [ n o t e : Ind. Stud., V, 103ff. ; XIII, 444; Keith, TS., I, p. C L X X I ] long ago adduced points in which the t w o w o r k s differed in t e r m i n o l o g y , and that unless and until the facts in question are explained away, they f o r m a very powerful argument against the identity of the t w o authors.' 5 W h e n A l b r e c h t W e b e r w r o t e the 5th volume of the ' Indische Studien '6 he was not yet in possession of either a manuscript or an edition of the Mahbhasya. W h a t e v e r he knew of the w o r k of the Vrttikakra was culled f r o m the old Calcutta edition of Pnini, where a number of ' v r t t i k a s ' are quoted, and f r o m Goldstcker's ' Pnini '. He w r o t e the 13th volume 7 after just having gone t h r o u g h the whole Bhsyafor the first t i m e , and many are the misunderstandings that occurred t o him in the difficult, really temerous, task of mastering the great and intricate w o r k in one sitting. As a matter of fact, he added here nothing new t o what he said in the 5th volume w i t h respect t o o u r problem : which merely proves that arguments do not lie on the surface. It must be remembered that all notions about the Vrttikakra had t o be of the haziest description until Kiel horn, after having shown a way t o distinguish the part of the Bhsya belonging t o Patajali f r o m that belonging t o Ktyyana, 8 presented our science w i t h his monumental edition. 9 W e b e r ' s argumentslike those of Goldstckerare nevertheless still of interest. But this interest is almost altogether a historical one. There is something t r u l y pathetical about the passion w i t h which these great pioneers t r i e d t o w r i n g evidence f r o m witnesses they could not really understand half; t h e r e is something t r u l y admirable in the acumen by which now Goldstcker and then W e b e r somehow managed t o find a way-mark in the all-overspreading darkness. Prof. Keith has summed up those points of W e b e r that in his
Keith, I.e., p. 742. 'I.e. 6 Berlin, 1862. 7 Berlin, 1873.
4 8

Ktyyana and Patajali, Bombay 1876. 9 BSS. 1880-1885.

335 The Identity of the Vrttikakra

opinion still hold good in his Translation of the TS., I p. C L X X I , t o which he refers me. T h e / run as follows : ' 1 . The t e r m jit occurs in both [the V r t t i k a and the V. Pr.] w i t h a different sense; 2. the terms used in the Prtiskhya (sim, mud, dhi, bhvin) are not found in the Vrttikas; 3. the V r t t i k a (on II. 4. 54) makes khy have the original form ofks; the Prtiskhya (IV. 164) repudiates this v i e w ; 4. the Prtiskhya (IV. 120) provides for the regular change of a mute before a nasal into a nasal ; the V r t t i k a (on VIII. 4. 45) leaves it optional except in the Bhs before a nasal affix.' It is not difficult t o show that W e b e r ' s points even in this sifted f o r m do not at all deserve the value Prof. Keith attaches t o t h e m . Let us start w i t h points 3 and 4. For point 3 Prof. Keith relies on W e b e r , Ind. Stud. V, p. 119, w h o quotes asiddhe sasya yavacanam vibhs as t h e first vrtt. on

Panini 2. 4. 54. But from Kielhorn's edition it becomes clear that this is only the third vrtt. while the first runs thus: caksiah ksnkhynau, which means : '[If a suffix called rdhadhtuka follows,] both ksn or khyn are substituted for caksi [not only khya as Pnini teaches].' The following vrtt.s say that instead of ksn and khyn, there might also be taught one single substitute starting with kh and s (i.e. khsn), from which are to be derived the forms with khy and ks- by means of an optional rule to be inserted in Pnini somewhere after 8. 2 . 1 . Further vrttikas discuss the benefit that might be derived from our doing so.
V. Pr. 4.164 khyteh khayau kasaui0G rgyah sakhyokhyamuk-

hyovarjam. 'Grgya changes the kh and y of root khy to k and s respectivelyexcept in sokhya, ukhya, mukhya.' This cannot be said, as Prof. Keith appears to believe, to repudiate the view that khy is derived from ks; it is, on the contrary, in perfect agreement with the first vrtt. on 2. 4. 54, which recognizes khy and ks side by side, while Pnini knows only khy. Point 4. Having as its chief and immediate object to give rules for the conversion of the VS. Padaptha into the Samhitptha, the V. Pr. naturally has to teach only such sandhi phenomena as are observable in the latter.11 Hence it says in 4.120 [sparso' pacamoh 4.117] pacame pacamam 'a mute is changed into a nasal if a nasal follows1, having explicitly stated before that this rule does not apply inside a word (4.116 nntahpade svorapancamntohsthsu). Having as its object to define the sandhi phenomena of a spoken language and several Samhits, the Astdhyay naturally has to teach optional validity of its rules rather often12. So in 8. 4. 45. It is clearly in no way surprising that Ktyyana, even if he be the author of the V. Pr., does not take exception to it. The fact of this rule being strictly observed in his own Samhit could give him no reason to forbid its optional validity in other Samhits or in the Bhs. And when adding to Pnini 8. 4. 45 that it is always
Thus correctly Weber's ms. A. cf. AV. Pr. 1, 2 evom iheti co vibhsprptam smnye with the commentator's exposition as given by Whitney. 12 cf. Patajali I, p. 4001. 9ff. ovasyam kholv osmbhir dam vaktovyam 'bahulam, anyatarosym, ubhayath, v, ekesm' iti. sarvavedaprisadam hJdam sstram. totra naikah ponthh
11 10

sakya sthtum. 'We can indeed not -avoid teaching [with general expressionslike:] 'often1, 'on one alternative', 'both ways', 'optionally', 'according to the teaching of some'. For this Sastra [of Pnini] is a grammar of the whole Veda. This being so, he cannot take one way only [and imply that other usages are incorrect].'

336 Paul Thieme

applied in the Bhas if the consonant is followed by a suffix (yaro 'nunsike protyaye bhsym nityavacanam), he distinctly suggests that it is not always applied in a Samhit (cf. e.g. vgmin): if he had failed t o add bhsym t o his own teaching, then and only then we should have a right t o speak of a contradiction, viz. t o V. Pr. 4.116. N o w w e may t u r n t o point 2. The expression sim (= Panini's k), mud ( = Panini's sar), dhi (= half vowels and nasals), jit (= Pnini's khar) and bhvin (= all vowels except a, a) are all used in the V. Pr., that is t r u e enough. But they all are explicitly defined. W h o would have been able t o understand them otherwise? W h e n commenting on the Astdhyyi, Ktyyana could not but prefer the short expressions formed by Pnini's rules. W h e n w r i t i n g the V. Pr., he could not employ Pnini's pratyhras w i t h o u t having given also the Siva Stras and some f u r t h e r rules teaching how t o put them t o t h e i r proper use. The rules which serve this purpose in Pnini ( 1 . 3. 3 , 1 . 1 . 7 1 , 1 . 3. 9) are at the same t i m e useful also in other respects (for the formation of sup, tin, etc.) : in the V. Pr. they would have been given just for the sake of a few soundpratyhras. Occasionally there occur in the V r t t i k a expressions like

svara (instead of ac), vyajana (instead of hoi), sandhyaksara (instead of ec), sparse?, prathama, dvitlya, trtJya, caturtha, jihvmDllya, and upadhmnJya, all of which can be found also in the V. Pr. Such terms, of course, were common property, they were understood by everybody. The Vrttikakra occasionally even employs samnksara (ec), when the V. Pr. says sim, aghosa (khay), when the V. Pr. says jit.'3 As the compiler of the V. Pr., Ktyyana did not want, I assume, to presuppose acquaintance with the Astdhyyi in those Veda students likely to study his book. Most probably they knew as little of it as modern Vedapthakas would. As the commentator of the Astdhyyi, Ktyyana did not want, I assume, to presuppose acquaintance with his little treatise on the V.S. in his readers. Why should he have meant the Vrttika only for Vjasaneyins ? I cannot see any inherent impossibility in such assumptions. The only really decisive argument against the identity of the two Ktyyanas in Weber's opinion is of course the fact that jit has a different sense in a vrttika (vrtt. 7 on 1.1. 68) and in the V. Pr. (here = Pnini's khar), though Pnini does not employ the term jit himself. Prof. Keith also accords this point precedence before the others. Now it is somewhat misleading to say that jit is a term with the Vrttikakra. This would suggest that he uses it the way he uses du (vol. I p. 304 = Pnini's sas) or tan (vol. I p. 488, vol. II pp. 99 and 221 = samjnchandasJ)u, which two terms he nowhere explains and may have taken from some other source.15 In reality the case of jit in vrtt. 7 on 1.1. 68 is of a different character. In his Vrttika on Pnini 1.1. 68, Ktyyana discourses on the difficulty that a word given by Pnini in his rules sometimes denotes only words for the representatives of the different kinds of the conception named by the word, as for example vrksa in 2. 4.12, where we have to understand that the rule is to be applied only to the different names of the trees like plaksa, nyagrodha etc.,
13 All the above quoted terms from the Vrttika have been collected by Kielhorn, Ind. Ant., XVI, p. 106. 14 Kielhorn, I.e. 15

Do they not resemble curiously the sim, mud, dhi, jit of the V. Pr. with their apparent arbitrariness.

337 The Identity of the Vrttikakra

but not t o the w o r d vrksa itself; sometimes its synonyms, like sva in 3. 4. 40, where we have t o understand the w o r d sva or any other w o r d for ' p r o p e r t y 1 ; sometimes its synonyms only, like rjan in 2. 4. 23, where we have t o understand any w o r d for ' king ' but the w o r d rjan itself; and sometimes words for the representatives of the different kinds of the conception named and the w o r d given itself, like matsya in 4. 4. 35, where we have t o understand the different names offish and the w o r d matsya itself. To remove this difficulty, he proposes t o teach the following rules in addition t o Pan in i 1 . 1 . 68 : (1 ) A w o r d t o which an s is attached as anubandha belongs t o the first class (vrtt. 5) ; (2) a w o r d t o which a p is attached, t o the second (vrtt. 6); (3) a w o r d t o which a j is attached, t o the t h i r d (vrtt. 7); (4) a w o r d t o which aj/i is attached, t o the last (vrtt. 8). Consequently we must add an s t o vrksa in 2. 4 . 1 2 ; a p t o sva in 3. 4. 40; a j t o rjan in 2. 4. 23; and a j / i t o matsya in 4. 4. 35. The terms sit, pit,jit,jhit are, then, not terms Ktyyana takes so t o speak f r o m his o w n private vocabulary, but terms he would use if Pan i n i had taught the rules proposed. I cannot see, why Ktyyana in considering this possibility of an addition t o Pan ini's grammar should feel bound t o choose another sound but j for his t h i r d rule, just because he has employed jit in the V. Pr. in a different sense. I should find it much more astonishing that he does not mind proposing as anubandha p, though Pan in i has already employed p as anubandha for a different purpose. But quite apart from all this, i do t h i n k arguments like W e b e r ' s 1 decisive jit' are all but worthless. Even if Ktyyana were not consistent, we should be quite w r o n g t o press the point. Does not Pan in i himself use homonymous terms over and over again ? W h a t valid inference could be drawn f r o m jit being used in the V r t t i k a in the sense of having j as anubandha1, and in the V. Pr. in t h e sense of ' surd mutes and sibilants' when Pnini in his w o r k uses an as a pratyhra for a, i, u, and as a ter m for a krt (3. 2.1 etc.) and a taddhita ( 4 . 1 . 83 etc.); or p as a t e r m for the feminine endings in o ( 4 . 1 . 1 . etc.), and as apratyhra for all case terminations f r o m the of the instrumental singular up t o the sup of the locative plural (7. 2.112); or ak as a pratyhra for a, /, u, r, I, and in the sense of ' having no k' (6.1.132 etc.), etc. etc. : examples lie on the way of anybody w h o is ready t o take the t r o u b l e t o look. I cannot regret having ' f o r g o t t e n ' W e b e r ' s points as referred t o by Prof. Keith. The instances supposed t o show that the author of the V. Pr. and the V r t t i k a ' in several important respects completely differ in opinion on phonetic points '16 were taken f r o m misinterpreted passages. The deviation in the terminology recognizable as far as sim, mud, dhi, jitare concerned t h r o w s no light. N o b o d y can deny an author the right t o express himself the way he chooses and thinks appropriate t o the particular occasion. If I should see anything significant in terminological usages, it w o u l d be the circumstance that the Vrttikakra does use Prtiskhya terms occasionally: svara ' v o w e l ', vyajana 'consonant', sandhyaksara ' d i p h t h o n g ' , sparsa ' mutes and nasals', prathama ' s u r d m u t e ' , dvitlya 'surd aspirate m u t e ' , trtya 'sounding m u t e ' , caturtha 'sounding aspirate mute,' jihvmulJya, and upadhmnlya',117 that Keith, TS., I, p. CLXXL To these above mentioned terms I only want to add one more: vikra in vrtt. 16on Pnini 1.1. 66, 67
17 16

stands for the Pninean term desa (1.1. 48, 56, 8. 3. 59), as vikra in V. Pr. 1.1. 33, etc.

338 Paul Thieme

beside at, et, ot, etc. he also has the Pratisakhya way : okara, ekara, okra etc., and that the V. Pr. sporadically employs Pninean terms like tin 1. 27; luk 3.12; /up 1.114; et, ot 1.114; 4. 58. Is'it not so that at least this last point has t o be 'explained away' by rather farfetched assumptions if we are t o believe the V. Pr. t o be older than Pnini ?

II dvirbaddham subaddham bhavati (Patamali III, p. 119, 1.121)

But I shall not rely on terminological usages. Again I am ready to admit for argument's sake that they cannot prove either theory. They just open out the possibility ofthat identity I suspect. To get a solid philological base from which we may approach the hard task of proving it conclusively, we have to confront single rules of Pnini and the V. Pr. Since my former endeavours have proved unconvincing, I shall again take a number of rules and ask again and again the decisive questions: Is it likely that Pnini has condensed the formulation of the V. Pr., or is it likely that the V. Pr. has wanted to improve on Pnini ? And : Is it possible to recognize any connection between the deviations of the V. Pr. and vrttikas of Ktyyana's ? I need not apologize for the following discussions in part being rather intricate. He who wants to understand old Indian grammarians, must follow them into their subtleties, and he who wants to settle their relative age with better arguments than were at the command of Weber and Goldstcker, must understand them. 1. Pnini says in 1.1. 66, 67 tasminn iti nirdiste prvasya, tasmd ity uttarasya 'when something is given in the locative case, the rule applies to what is preceding it; when in the ablative case, to what is following it.' Hence we have to understand when we read for example the loc. aci in 6.1. 77, that this rule applies to what is preceding ac\ when the abl. dvyantarupasargebhyah in 6. 3. 97, that this rule is preceding oc; when the abl. dvyantarupasargebhyah in 6. 3. 97, that this rule applies to what is following dvi, or antar, or a preposition. The V. Pr says in 1.134,135 tasminn iti nirdiste prvasya, tasmd ity uttarasydeh, which is longer by the word deh. But should we apply here the theory of condensation, we would be badly advised. For Pnini, in order to complete his second rule, has to teach besides in 1.1. 54deh parasya 'when [a substitute is taught to step] in the place of something that is following [something given in the ablative case], [the substitute steps] in the place of its first sound.' Only now we can construct correctly Pnini 6. 3. 97 dvyantarupasargebhyo 'pa It ' long 1 is substituted for the first sound of ap that is following dvi, or antar, or a preposition.' The V. Pr. need not give this rule. One might ask, why it stopped at changing the second rule. Why did it not say correspondingly in 1.134: prvntasya, especially as it does not teach a rule like Pnini 1.1. 52 alo 'ntyasya 'when [a substitute is taught to step in the place of something; it steps] in the place of the last sound ' ? An answer to this question I should not be able to give, if I had not the first vrttikaon Pnini 1.1. 66, 67: nirdistagrahanam nantaryrtham, which means that the expression nirdiste in 1.1. 66 can have only the purpose of making it clear that a rule containing a locative applies only to that which is preceding immediately the word put in the locative,

339 The Identity of the Vrttikakra

that is the last sound. Else Panini could have simply said 'tasminn iti prvasya.' Yet this may not be obvious enough. Looking ahead, we find a rule that complements V. Pr. 1.134, 135. In V. Pr. 1.145 we read: purvottaroyor uttarasya 'when [a rule w o u l d apply simultaneously] t o what is preceding and t o what is following, it [has t o be understood t o refer only] t o what is following.' 18 This rule has, strangetosay, no application in the Prtiskhya. N o r does Pnini give it. Did he o m i t it as useless?This can hardly be. For in the Astdhyyi t h e r e are sundry rules containing both a locative and an ablative case. Hence Ktyyana formulates the

vrtt. 3 on 1.1. 66, 67: ubhayor nirdese vipratisedht pacamnirdesah ' when both (a locative and an ablative case) are given, the giving of the ablative [will be the stronger one according to Pnini 1. 4. 2 vipratisedheparam kryam], since there is a conflict.' Those on the lookout for contradictions between the Vrttika and the V. Pr. might, however, point out that this vrtt. just shows that Pnini need not have taught a special rule like V. Pr. 1.145, since the dilemma can be solved already by 1. 4. 2. Further they might point out that the V. Pr. in 1.159 teaches a rule of identical purport with Pnini 1. 4. 219, hence the man who wrote the vrttika could not have thought 1.145 necessary in the V. Pr. either. I. must however disappoint them : Ktyyana rejects in vrtt. 13 his first view that in a dilemma the giving of the ablative must be stronger according to Pnini 1. 4. 2, by showing that technically there would be no 'conflict' 20 , and gives as his final view (siddhnta) in vrtt. 17 that both the operations, the one concerning what is preceding and the one concerning what is following, would have to apply if Pnini 1.1. 66, 67 is left as it standseven if one should have recourse to some special assumption. I need not enter into the technical details of this assumption, which anyhow would not be applicable in the V. Pr. What I wanted to draw attention to, is the fact that there exists a connection between V. Pr. 1.145 and vrtt. 3ff. on Pnini 1.1. 66, 67. They are children of the same thought.
2. Pnini says in 1.1.11 Jdded dv vaca na m pragrhyam 'an J or ,

or e expressing duality is called pragrhya.' The V. Pr. says in 1. 92, 93 pragrhyam, ekrekrokr dvivacannth. Again, at a first look, the theory of condensation appears to work. There are however other points worth consideration. I need not be long on the V. Pr. saying ekra etc. instead of et etc. Also the AV Pr., which is later than Pnini even in Prof. Keith's opinion,21 uses the former expressions, quite apart from the fact that the Vrttikakra himself does not avoid them. On the assumption that Pnini reformulated V. Pr. 1. 92, 93, we should have to account for Pnini not leaving the order of the vowels as it was given in the V. Pr. by saying edfdd ... We could easily do so by suggesting that he preferred to follow the order of the alphabet. On the assumption that the V. Pr. modelled its rules on
18 Of identical construction is V. Pr. 1.144 samnikrstaviprakrstayoh samnikrstasy and e.g. Patajali I, p. 71,1. 16 laksanapratipadoktayoh protipadoktasyalva. 19

See below. As defined in vrtt. 1, 2 on 1. 4. 2. 21 Indian Culture, Vol. II, p. 741.

20

340 Paul Thieme

Pnini, we should have t o account for its changing the natural order as reflected in a supposable Jkrokraikrh. W e could easily do so by pointing o u t that t h e r e exist dual forms in e (type : male, pcete) as well as ai ( t y p e : pacvahai). A formulation Jkrokrai-

krh would leave it doubtful, whether the former or the latter type was meant. Is Pnini's dvivocanam a condensation of dviVacandntam? Pnini's rule may be translated as above. In this case Jddet is the subject of the proposition, dvivacanam, the attribute of the subject. This seems obvious, but creates a difficulty: the rule does not apply to the ending e of aform like pcete, since here we have not an e expressing duality, but an ete. Neither has Pnini called ete 'pragrhya1, nor has he stated that an e that happens to stand in the end of an element expressing duality, is also called ' pragrhya'. This is formulated by the first vrtt. on Pnini 1.1.11 : Jddayo dvivacanam pragrhya iti ced antyasya vidhih ' if [Pnini means to say that] etc.., when expressing duality, are called lpragrhya', a special rule must be given for [an e] that forms the end [of an element expressing duality]. 1 Pnini's rule may also be constructed by taking dvivacanam as subject, and Jddet as its attribute. If we do so, Jddet can, according to Pnini 1.1. 72, be understood to denote something ending in f, G, e. This possibility is considered in vrtt. 2 and accepted as unobjectionable in vrtt. 3. A third possibility is to understand 'word ' as subject, and both Jddet and dvivacanam as its attribute. Now we can translate according to Pnini 1.1. 72: 'a word that ends in J, , e and in an element expressing duality.' This possibility is considered in vrtt. 4 and accepted as unobjectionable in vrtt. 5. Now everybody will agree that Pnini must have meant what I have translated first. For both alternative constructions create a serious difficulty with respect to the next rule (1.1.12 adaso mat), where only Jddet can be the subject, and is taken as such by the Pninyas, who yet, following Patajali's siddhnta (I. p. 68, I. 6f), make dvivacanam the subject in 1.1.11. It is obvious that Ktyyana's vrttikas 2 ff. are nothing but scholastic devices in defence of Pnini's wording. On the assumption that Pnini condensed the dvivacanntam of the V. Pr. into dvivacanam, we have to believe that he did not notice that he was putting the worse for the better. This is the more unlikely as the V. Pr. wording removes with one stroke all difficulty. It may be noted that it is the only formulation that cannot be projected by any interpretative device into Pnini's rule. For dvivacannth 'as endings of [a word] expressing duality' is a tatpurusa, and for Pnini 1.1.11 can be got from Pnini 1.1. 72 only the bahuvnhi: dvivacannta 'whose end is [something] expressing duality', 3. Pnini says in 6.1.158 anudttam padam ekavarjam. This admits of two interpretations. Either: 'a word has no udtta but one1 (so Patajali III, p. 98, I. 25 f.). Or: 'a word is anudtta except for one [vowel].' It is only this latter interpretation that Ktyyana considers in the Vrtt i ka. He is not satisfied with the rule. When forming a word like malakJja, we ought to make its first vowel udtta according to 6. 2. 82, and the last but one vowel of the element preceding jo according to 6. 2. 83 also. It is desirable that only the latter rule applies : malakJjah. In order to obtain this result, it would be

341 The Identity of the Vrttikakra

necessary t o recognize a ' conflict ' b e t w e e n 6. 2. 82 and 6. 2. 83, w h i c h w o u l d a l l o w us t o apply Panini 1 . 4. 2. vipratisedheparam kryam. Yet according to the definition of'conflict' (yipratisedha) as given by Ktyyana in vrtt. 1 and 2 on Pnini 1. 4. 2, there is no ' conflict' between 6. 2. 82 and 83, since it is quite possible for both rules to apply simultaneouslythe one concerning the first and the other the third vowel, and the eka in 6.1.158 being understandable only in the sense of'the one [for which an udtta (or primary svarita) is explicitly taught] 1 . This22 is expressed by the first vrtt. : anudtte vipratisedhnupapattir ekasmin yugapatsambhavt ' if [we define with Pnini a word to be] anudtta, we do not obtain a 'conflict' [in cases where several udttas are taught], since [several udttas] might be substituted at the same time'. The second vrtt. proposes to remove the difficulty by changing Pnini's dfinition : siddham tv eknanudttatvt ' but it is alright if we teach [instead of anudttam padam ekavarjam]: eknanudttam padam ('a word has only one vowel that is not anudtta').' Now there is a 'conflict' between 6. 2. 82 and 83 (and between similar rules), for now they cannot apply simultaneously, since from 6. 2. 82 we have to learn that the first vowel alone is udtta, and also from 6. 2. 83 that the last but one before ja alone is udtta. Ktyyana is careful not to propose ekodttam padam, for this would neglect the cases where a primary svarita is taught (6.1.185, etc.). The reflection of Pnini 6.1.158 is found in V. Pr. 2.1 svaritavarjam ekodttam padam. The decision whether this has been condensed by Pnini, or whether it is meant as an improvement on Pnini might have been doubtful if we had not the Vrttika. Having the Vrttikawe cannot but recognize that V. Pr. 2.1 wants to avoid what could be objected to Pnini. Its formulation equals the one proposed by the Vrttikakra, only that in the Vrttika there has been found out a more concise form of a truly cunning simplicity. We may compare the three definitions in the following way: Pnini 6.1.158 anudttam padam ekavarjam : concise, but objectionable (from Ktyyana's point of view). V. Pr. 2 . 1 . svaritavarjam ekodttam padam : not concise, but unobjectionable. Vrttikakraeknanudttam padam: both concise and unobjectionable. Here I may be forgiven if I quote what I wrote in my ' Pnini,' p. 93, after discussing Pnini 1.1.9 with vrtt. 1 and 2, and V. Pr. 1.43: ' Pnini 1.1.9: tuiysyaprayatnam savarnam: concise, but not precise. V. Pr. 1. 43 : samnasthnakaransyaprayatnah savarnah : not concise but precise. Vrttikra: sye tulyadesaprayatnam savarnam : both concise and precise.' 4. A vocative is not accented except in the beginning of a sentence or of a line of a verse. This is expressed by Pnini thus: 8 , 1 , Gpadasya Mpadt 18 anudttam sarvam apddau 19 mantritasya ca. By the V. Pr., thus: 2. 2. anudttam 17'padaprvam mantritam annrthe ' pddau.
22

I hope to have fat homed Ktyyana's meaning correctly. His objection is, no doubt, very subtle, and Patajali

has easy play in showing that it may be dropped (III, p. 98, I. 20 ff.).

342 Paul Thieme

There are the following differences: (a) The V. Pr. has left out a w o r d corresponding t o padasya ; (b) mantrita is put in the genitive case by Pnini, in the nominative in the V. Pr. ; (c) the V. Pr. has padaprvam instead of padt; (d) it has left out the expression sarvam; (e) it has added annrthe. a. It is not quite correct t o say that the V. Pr. has left out a w o r d corresponding t o padasya. For a corresponding w o r d is taught, only it need not be repeated. It is valid f r o m 2.1 svaritavarjam ekodttam padam, which equals Pnini 6.1.158 anudttam padam ekavarjam. Here lies no argument. W e could say, of course, that the V. Pr. has deliberately t r i e d t o arrange matters so as t o be enabled t o save one 'padam'. But it could well be maintained that Pnini, for reasons of his o w n , has deliberately chosen a different disposition of the accenting rules ; and that in any case he could not imitate the procedure of the V. Pr. since he wanted t o put pada the first time in the nominative case and the second t i m e in the genitive. b. W e have, then, in reality t w o nominatives in the V. Pr. and t w o genitives in the AstdhyyT. This difference of construction cannot prove much. It is well known that when Pnini says: ' f o r x (gen.) [is substituted] y (nom.)', the V. Pr. say: x (nom.) [is changed] t o y (ace.).' W e are quite used t o that. It must, however, be pointed out that the Vrttikakra is at pains of setting right the significance of the genitive padasya. If it is understood as ' in place of apodo' (according t o Pnini 1 . 1 . 49), we have t o construct the rule 8. 2. 4 udttasvaritayor yanah svarito 'nudttasya t h u s : ' a svarita vowel is substituted for the last sound (according t o 1 . 1 . 52) of a w o r d that ends in an anudtta vowel (according t o 1 . 1 . 72) that is following a yan which has been substituted for an udtta or svarita vowel (according t o 6 . 1 . 77).' This means that we can obtain the correct accent of forms like kumryau and kisoryu, but not of forms like kumryh and kisoryh (cf. v r t t . 4 on 8 . 1 . 1 6 , 1 7 w i t h Patajali). Consequently, padasya must be taken as an a t t r i b u t i v e g e n i t i v e (vrtt. 5)and added as such t h r o u g h out the chapter. Padasya . . . sarvam in 8 . 1 . 1 8 hereby receives the meaning: ' t h e whole of apada', padasya. . . anudttasya in 8. 2. 4 : ' in the place of an anudtta vowel of apoda' etc. It must be admitted, of course, that t h e r e lies no obvious argument here e i t h e r . c. The V. Pr. has padaprvam instead of padt. This is, no doubt, due t o its having taught in 1.135 that if something is given in the ablative case, the rule applies t o the first [sound] of what is following. It is for the same reason that the V. Pr. says, for example, in 6.11 khytaprvam, when Pnini could have used the ablative (cf. atiah in 8 . 1 . 28). Pnini is free t o use the ablative because he has employed the w o r d sarvam in 8 . 1 . 1 8 . d. By leaving out sarvam the V. Pr. becomes, t h e n , really more cumbrous. 23 As if foreseeing this objection against the f o r m u l a t i o n It should be noted, however, that it may use the ablative whenever a misunderstanding cannot arise. In V. Pr. 2.17 and 6.11 one might understand that only such vocatives and verb forms are meant as start with an udatta vowel before the rule is taught, type: gne, pacat. In 2. 9. ytha grbhobhuvo' gnibhyah, for example,
23

the ablative is unobjectionable, since the rule could not possibly be concerned with the first sound of ytha. Somewhat diffrent s the case of the ablative in 4.134 udattc cnudttam svaritom, since udtta means here a syllable 'containing an udatta vowel ', and svarita a syllable 'containing a svarita vowel' according to V. Pr.4.1.

343 The Identity of the Vrttikakra

of the V. Pr. Ktyyana tries to prove in the Vrttikathat Pnini need not have employed the word sorvam in 8.1.18 : sorvavocanam ander anudttrtham ti eel luti pratisedht siddhom ' if [one should maintain that] 'the word sarvam has been employed [in 8.1.18] for the purpose [of obtaining substitution] of an anudtta vowel that is not the first [of what is following, in spite of 1.1. 54]', [the answer would be that the substitution of an anudtta for such vowel] is already in order because of the prohibition [given in 8.1. 29] with respect to a periphrastic future [which prohibition would be without any purpose if an anudtta had to be substituted only for the first vowel according to 8 . 1 . 29, since the first vowel in a periphrastic future is necessarily always anudtta]'. Against this one might object that Pnini 1.1. 54, according to Ktyyana's own words (vrtt. 1 on 1.1. 54), is a special exception (apavda) to the general injunction (utsarga) pronounced in 1.1. 52: alo 'ntyasya 'a substitute steps in the place of the last sound', and if Pnini by his prohibition in 8 . 1 . 29 had indicated that in this chapter he did not want 1.1. 54 to apply, he might yet have meant 1.1. 52 to apply, and that consequently the word sarvam in 8.1.18 is necessary to remove this wrong impression. This objection is voiced in vrtt. 2
on 8 . 1 . 1 8 alo ' ntyavidhiprasagas tu ' [ i f sarvam were not employed

in 8.1.18] there yet would wrongly apply Pnini 1.1. 52.' Now Ktyyana sets out to show in vrtt. 3 and 4 that Pnini has given an indication that 1.1. 52 should not apply either in this chapter. For if it applied, he need not have taught the rule 8 . 1 . 51, since the last vowel of an ordinary future form is always anudtta (vrtt. 3), nor need he have employed the expression anta in 8. 2. 7, but could simply have said nalopah prtipadikasya (instead ofprtipadikntasya), since it would anyway be clear from 1.1. 52 that opa could be substituted only for the end of aprtipadika (vrtt. 4 on 8.1.18 and vrtt. 6 on 8.1.16,17). In fact it may be suspected that vrtt. 6 on 8.1.16,17 has only been given as an alternative solution, beside the one mentioned in vrtt. 5 (above b), of the difficulty pointed out in vrtt. 4, because Ktyyana could not take padasya as an attributive genitive in 8.1.18 if he cancelled the expression
sarvam.2*

But we need not insist on the latter point. We only ask : Why ever does the Vrttikakra try so hard to prove that sarvam in Pnini 8.1.18 is superfluous? Can it be anything else but the special reason suggested above? Even this point may not be obvious enough. Let us, then, turn to the last. e. Vrtt. 5 on Pnini 8.1.18: samnavkye nightayusmadasmaddesh ( = vrtt. 11 on Pnini 2.1.1) says that the loss of accent (according to 8.1.18,19 etc.) and the substitution of vm and nau etc. (according to 8.1. 20ff) ought to have been taught [not only for the case of a vocative and the respective forms of yusmad and asmad following a word, but also for the case of their standing] in the same sentence [as the word they follow].
130,131 sovritavantsvaritoh; udtta6.1, 223), which again is not quite vn udttah.On the other hand, in sufficient in Patajali's opinion (III, order to prevent the ablative in p. 119, I. 21ff.). Without defect Pnini 8. 4. 66 (and in other cases) is T. Pr. 14. 29 udtttparo 'nudttah from causing a wrong application, svaritam, 30 vyanjanntarhito'pi. 24 Ktyyana has to give the rule of Patajali explains vrtt. 6 in a difinterpretation: haisvaraprptau ferent way, without accounting for vyajanam avidyamnavat (vrtt. 2 on Ktyyana's va.

344 Paul Thieme

N o w we have an obvious argument. It suffers no doubt that the one additional expression in V. Pr. 1.17: annrthe and this v r t t i k a a r e children of the same thought, though the formulation of the v r t t i k a : samnavkhye seems by far happier. But then, Ktyyana, in order t o make it clear had t o give a definition of the concept vkya in v r t t . 9 and 10 on Pnini 2 . 1 . 1 : khytam svyayakrakavisesanam vkyam', ekati. That the V. Pr.'s annrthe may betaken strictly in the sense of ' in one sentence', appears f r o m Ml. S. 2 . 1 . 46 arthaikatvd ekam vkyam . . . 5. The Vrttikakra is not satisfied w i t h Panini 1. 2. 39 [ekasruti 33] svaritt samhitym anudttnm. If we take the plural anudttnm at its face value ' f o r several anudtta vowels', the rule applies only when more than t w o anudtta vowels follow a svarita (vrtt. 1 on 1. 2. 39) ; if we take it t o mean ' for anudtta vowels ' = ' f o r any anudtta vowel ', we can apply it only t o the next that follows the svarita, since the plural would be void of any special force (vrtt. 2). A solution of the dilemma is given in vrtt. 3: anekamapJti tuvacantsiddham " i t is in order if we teach: 'also several anudttas'." Everybody w i l l admit the difficulty of construing this additional anekam api in Pnini's rule. In reality we could only say e i t h e r : svaritt sa m h ityam anudttnm aneknm api, o r : . . . anudttasynekasypi. V. Pr. 4.138 reflects Pnini 1. 2. 39 t h u s : svaritt param anudttam udttamayam. Here Ktyyana's addition w o u l d fit perfectly Rather, it does! For 4.139 actually runs: anekam api. O n l y very powerful arguments indeed could make me believe that this is a coincidence created by chance.25 The other differences between Pnini's w o r d i n g and that of the V. Pr. (samhitym: param, ekasruti : udttamayam) yield no obvious arguments. 6. Pnini says in 1. 4. 2: vipratisedheparam kryam ' w h e n t h e r e is a conflict [between t w o rules], the one that comes later must be applied.' The V. Pr. in 1.159 : vipratisedha uttaram balavad alope. There are t h r e e deviations : (a) Instead of kryam, the V. Pr. reads balavat; (5) the V. Pr. has added the expression alope; (c) Pnini says param, the V. Pr. uttaram. a. It is certainly not obvious w h e t h e r kryam or balavat should be preferable. But it is interesting, though it proves nothing by itself, that when referring t o what is expressed in Pnini 1. 4. 2 by kryam, the Vrttikakra does not use this w o r d , but balyas ' s t r o n g e r ' : v r t t . 9 on 7 . 1 . 1 : vipratisedht tu tpo balJyastvam, and on 1. 4. 2 itself tells us that Pnini has given his rule because whenever t h e r e arises a conflict of t w o rules, neither w o u l d apply since both would have the same strength : v r t t . 5 on 1. 4. 2 : apratipattir vobhayos tulyabalatvt. In v r t t . 8 on 1. 4. 2 he adds that Pnini ought t o have taught beside param ' the one that comes later ', also antaragam ' t h e one the cause of which presents itself first.' He refers t o this addition in vrtt. 9 on 6.1.108 by the expression antaragabalyastva. Even when he does not employ the expression ballyas itself, his construction shows that in his mind he had a f o r m u l a t i o n not like: vipratisedhe param kryam, antaragam ca, but l i k e : vipra25

Also the fact of V. Pr. 4.138 being taught after 4.134 (corresponding to Pnini 8. 4. 66) is in accordance with

the remarks of the Vrttikakra on 1. 2. 32.

345 The Identity of the Vrttikakra

tisedhe param ballyah, antaragam ca. Thus in v a r t t . 9 ff. on 1. 4. 2 we have a number of constructions of the type svoro lopt (vrtt. 18): '[a rule on] accent [is stronger] than [a rule on lopa]', and t h r o u g h o u t the Vrttika, whenever Ktyyanafeels called upon t o state that an x is effected (according t o 1. 4. 2) and not a y, because there is a ' c o n f l i c t ' between the t w o , he always says: 'x (nom.) [is stronger] than y [abl.] since t h e r e is conflict [between the t w o and x is the one that comes later] ' (cf. vrtt. 1 on 1. 2. 5 ; v r t t . 4 on 2 . 1 . 69 ; v r t t . 1 on 6. 2 . 1 2 1 , etc.). b. The addition of olope in V. Pr. 1.159 is instructive. It certainly cannot be accounted for by the argument that the more archaic author has not yet found out means of arranging his rules in such a way as t o make our rule universally valid. On the contrary, the V. Pr. avoids a fault of Pan ini's. The Vrttikakra not only has t o add antaragam ca (vrtt. 8) t o the latter's rule ( w i t h many applications set f o r t h in v r t t . 10 ff. on 1. 4. 2), but also t o teach in v r t t . 25 that luk is stronger than lopa, substitution of yan etc., and, on a number of occasions, t o name cases where the rule has t o be i n v e r t e d : v r t t . 4 on 3. 4. 77; v r t t . 1 on 5 . 1 . 2; v r t t . 9 and 10 on 6 . 1 . 1 2 ; v r t t . 1 on 6. 4. 48; vrtt. 10 and 11 on 7 . 1 . 96. It is well known that Patajali evades the embarrassment created by Pan ini's rule and procedure contradicting each other so frequently, by taking para in the sense of sta 'desirable' and understanding Pan i n i 1. 4. 2 t o mean : ' if t h e r e is a conflict of t w o operations the one that is desirable must be applied ' (I, p. 306, I. 9 f. and often). c. If Pan ini says param and the V. Pr. uttaram, the t h e o r y of condensation w o u l d of course maintain that Pan i n i has deliberately chosen the shorter w o r d param for the longer w o r d of his predecessor. But I w o u l d be at a loss t o explain why Pan in left unchanged uttarasya of V. Pr. 1.135 in 1 . 1 . 67, where parasya w o u l d have been the more f i t t i n g as parasya is employed just a few rules before (in 1 . 1 . 54) synonymously. N o r can I easily account for the V. Pr. having replaced param by uttaram, for it also uses para (e.g. in 3. 3) synonymously w i t h

uttara. Both authors, this is the only possible inference, did not mind whether they said para or uttaranotwithstanding the latter being the longer expression. It is necessary to emphasize this point. It appears as if scholars, when talking of the ' brevity' of grammatical rules, do not always take an altogether correct view of the character of this brevity. I think, because there is always in their mind that last Paribhs of Ngojlbhatta's collection : 'Grammarians rejoice over the saving of [even] the length of half a short vowel as over the birth of a son.' Enjoying the sublime irony of this witticism, which seems to voice what we feel when faced with Pninean rules like ko yan aci (6.1. 77) or a a (8. 4. 68), we are apt to forget that, like any witticism, it ought to be taken with a pinch of salt. It would be wrong, of course, to rely only on the comparative recency of our maxim. For it can be shown that already Patajali held somewhat similar views. So, when he maintains that Pan ni having produced his work with great careholding a bushel of darbha grass in his hand, sitting on clean ground, his face to the east, it would be impossible that even one sound be without purpose (I, p. 39 1.10 ff.); or when he calculates that the expression yvoh has the length of 3 short vowels, while the synonymous expression inah would count only 3, and asserts that Pan in i must

346 Paul Thieme

have had a special reason t o choose the f o r m e r lengthier one, nstead of the latter (I, p. 35 l.12ff.). On the other hand, however, Ngojbhatta is, no doubt, right when remarking that ' the question raised [in the Bhsya] is generally only, w h e t h e r in a rule which is made up of several words a w o r d can be saved, but not w h e t h e r a mtr (or half a mtr) can be economized.' 2 6 It is not difficult, in point of fact, t o recognize that Pnini, though striving after brevity w i t h great eagerness, often does not mind employing long words when he easily could have avoided t h e m . He is ingenious in finding out ways of being brief, but he is not pedantic about it. Be this however as it may. Essential for us now is the question, w h e t h e r the Vrttikakra can be supposed t o share the view of 1 b r e v i t y ' implied in our Paribhs. The answer can only be a decided ' N o '. W h e n e v e r the Vrttikakra is about t o shorten some rule of Pnini's, he proposes t o cancel a whole expression that appears superfluous. Never does he want t o replace some w o r d by a s h o r t e r synonym. He is a logician of no small acumen, and sometimes his pruning-knife appears t o cut rather t o o sharp. But he raises no point that w o u l d be, materially or logically, altogether irrelevant. Looking at things in a natural way, it does seem irrelevant w h e t h e r one should say e.g. porasya or uttarasyo in 1 . 1 . 67. He w h o thinks that t o Ktyyana it was not, must prove it. It is simply w r o n g t o ascribe t o him a view for which we only have later authorities, and which, even w i t h t h e m , is only occasionally and for distinct purposes brought t o bear upon the i n t e r p r e t a t i o n of Pnini. If Ktyyana really t h o u g h t porosyo preferable t o the uttarasyo of Panini 1 . 1 . 67, why did he not say so in his Vrttika? These specimens may suffice. They are not meant t o exhaust the arguments, and on request I could easily increase t h e m . They were meant as a series of experiments executed t o test several theoretical possibilities. For only such possibility can be accepted as likely as holds good if applied t o single facts. Several tests led, as might have been foreseen, t o ambiguous results. W e cannot decide, for example, w h e t h e r the expression param in Pnini 1. 4. 2 is meant as an improvement on uttarom in V. Pr. 1.159, or vice versa (6 c). W e did not detect a single point, where Pnini w o u l d have been shorter and better at the same t i m e . A few tests favoured the t h e o r y of V. Pr. being younger: dvivaconantam in V. Pr. 1. 9327 is better than dvivacanom in Pnini 1.1.11 (2); the addition alope in V. Pr. 1.159 is b e t t e r t h a n Pnini's silence about the exceptions t o his rule 1. 4. 2 (6 b). The majority of facts, however, only revealed t h e i r significance after the deviations of the V. Pr. f r o m Pnini had been looked at in the light of vrttikas. It was only a v r t t i k a that made plausible the additional r u l e V . Pr. 1.145 (1); the formulation of V. Pr. 2.1 (3); the expression annrthe in V. Pr. 1.17 (4e). On the other hand, only the formulation of V. Pr. 4.138,139 made understandable the w o r d i n g o f v r t t . 3 on Pnini 1. 2. 39 (5). Leaving aside all o t h e r points and questions, we have t o admit that the Vrttikakra must have well known the rules of the V. Pr., and that he must have t h o u g h t them t o compare favourably w i t h
26

cf. Kielhorn's edition of the Paribhsendusekhara, p. 115, 1.12ff.; Translation, p. 526.

27

cf. also AV. Pr. 1. 75, 76 [Ikrokrau 74]dvivacanntau, ekros ca.

347 The Identity of the Vrttikakra

Pnini's. So much so that he puts himself out t o prove that Panini 6.1.158 contains an objectionable definition (3), which is by no means obvious; that the expression sarvam in Pnini 8 . 1 . 1 8 is superfluous (4d), which nobody will admit easily. Pondering over all this I cannot help feeling that all those w h o t h i n k identity of such a name as Ktyyana t o create any, however slight, probability at all, w i l l be forcibly inclined t o believe that the V. Pr. and the V r t t i k a are by the same hand. Those who do n o t w e l l , they may assume that the Vrttikakra's father, or grandfather, o r great-grandfather, or cousin, or uncle, or any other male relation of his in the ascending line, has composed the V. Pr., and t h a t Ktyyana has devoted careful study t o it. All this is, as yet, by no means ruled out. But, I t h i n k , everybody will have t o a d m i t : (1) that the V r t t i kakra knew the V. Pr. w e l l ; (2) that the probability of the V. Pr. being younger than Pnini is stronger than the contrary, if detailed comparisons can yield any result at all.

NI te khalvapi vidhayah suparigrhit bhavanti yesu laksanam p rapan cas ca (Patajai I, p. 400,1.8)
I know that I have been unfair in taking Prof. Keith by his word and pretending to believe that it is Weber's arguments that prevent him from accepting the identity of Ktyyana, the author of the V, Pr., and Ktyyana, the Vrttikakra. As a matter of fact, he cannot lay much store by them, since that he did not think it worthwhile to examine them with the help of Kielhorn's Mahbhsya text neither in 1914, nor in 1936. He will hardly feel sorry now they are shown to be of no value; he will be convinced that there are other, better arguments available to close the breach. I am even afraid lest he should heed all the interpretative details I have given here above, as little as those I gave before. He may still think that I am trying to prove what cannot be, and that my diving into technical subtleties only tends to obscure a clear issue. What is it that makes Prof. Keith accept any argument put forward in favour of the priority of the V. Pr., bad as it may be, and that makes him take easy any argument to the contrary? It can be nothing else but the general impression he has derived from the study of Pnini's work, the Vrttika, and the V. Pr. As to the first, Prof. Keith's impression can only be that, may Pnini be agenius or a more or less skilful compilator, the Astdhyyl evinces a very considerable degree of knowledge and acumen, of insight into the structure of the language it describes, of technical routine in arranging and representing factsin brief, that it testifies to a high stage reached by the science of grammar in his time. As to the V. Pr., his impression must be that it reveals in part an endeavour to define certain grammatical facts by general rules, similar to those given in Pnini's grammar; and in part a deplorable incapacity of grasping the significance of others, which seem very simple. To give only two examples : In 6.1 the V. Pr. teaches that a verbform is anudtta if it follows another word belonging to the same sentence; in 2.14 it teaches that srutam is anudtta when preceded by ha, aiming at V. S. 7. 9 mmd ih srutam hvam, where srutam must be a verbform. In 4.164 the V. Pr. teaches that Grgya changes the kh and y of root khy to kand s respectively; in the same rule it puts down the absurd addition: 'except in sakhya, ukhya, mukhya.'

348 Paul Thieme

Led by this impression, w h i c h , no doubt, w i l l be shared by many, he thinks it quite unlikely that the V. Pr. should be younger than Pnini. The facts seem t o allow of one interpretation only. The science of grammar as represented in the V. Pr., is on its march towards the perfection attained at the t i m e of Pnini. The V. Pr. is Pan ini's precursor, as the dawn is the precursor of the day. As t o the Vrttikakra, Prof. Keith's impression must be that its author being later than Pnini, it must be more perfect than the AstdhyyT. Hence it is impossible t o ascribe it t o the same author as the V. Pr. These impressions appear very plausible, and the conclusions formed on them seem t o simply compel acceptance. W h e n in some places Pnini's formulation obviously is less happy than the one of the V. Pr., it becomes very easy t o account for it by such assumptions as: Pnini has b o r r o w e d unintelligently, or: in Pnini's grammar the consideration of brevity is allowed t o override even intelligibility and logical correctness. It is difficult, if not impossible, t o prove such like assumptions t o be w r o n g in every single case. Yet I t h i n k we can look at things in a somewhat different way. Let us t r y t o see, w h e t h e r it leads us t o absurd consequences. If I should state the general impressions I have derived f r o m the study of Pnini and the V. Pr. in a few short sentences, I should say: Admirable as Pnini's w o r k is as a whole, I cannot deny that he has overdone his ingenuity and partly fallen sacrifice t o it. He is so brief as t o be often obscure and not seldom even illogical : he is so subtle as t o be ambiguous, and not seldom even incomprehensible. In order t o understand rules of his that are not exceptionally simple, it is necessary first t o know what they are supposed t o teach : to-day, when his language does not any longer live, but has t o be learned in school, a scholar w h o wants t o freely handle and master his injunctions, must possess a stupendous memory and a tremendous amount of learning in the vast literature discussing the implicit suggestions, silent assumptions and principles underlying his formulations or supposed t o underlie t h e m . There is hardly anything admirable in the V. Pr. But I cannot but acknowledge that if it is not quite free, it is yet more free than Pnini, of the defects of indistinctness, ambiguity, and obscurity. It may contain some hard passages, but nowhere is it necessary t o make its author implicitly suggest, or silently assume, anything, or establish by some artificial device any principle that w o u l d adjust his f o r m u l a t i o n . In o r d e r t o apply its teaching it is necessary t o know the Padaptha of the V.S., t o have a certain idea of the meanings of the words occurring t h e r e i n , and t o observe carefully what the author says. Pnini addresses subtle intellects, scholars w i t h scientific rather than practical interests; the V. Pr., just any ordinary Vjasaneyin. Led by these impressions, I t h i n k it quite natural t o assume that the V. Pr. is w r i t t e n by someone w h o knew Pnini, but did not want t o follow t o o closely his risky ways of teaching. Doubly so, if I bear in mind that as far as the Bhs is concerned an incorrect f o r m , even when used a t a public occasion, is no serious matter, while even one incorrect sound uttered in the recitation of a vedic verse, is bound t o bring down bad misfortune. Is not Pnini himself inclined t o be careful when referring t o vedic forms (e.g. in 3 . 1 . 1 2 3 ; 3 . 1 . 4 2 ; 7.1.43)?

349 The Identity of the Vrttikakra

I can find no difficulty in accounting for the deviations of the V. Pr. f r o m Pnini. They all may be based on reasonable motives, as I have taken some t r o u b l e t o show. There remain the cases where the V. Pr. appears t o contradict itself, where it appears t o follow t w o distinctly different methods, of which the one looks archaic (V. Pr. 2 . 1 4 ; 4,164 second half), the other modern (V. Pr. 6.1 ; 4.164 first half). The most natural course t o explain contradictions of this kind in a w o r k , is t o trace them t o the individuality of its author. Consequently I should say: There are t w o souls living in the breast of Ktyyana, the author of the V. Pr. The one aspires high : it strives t o vie w i t h Pnini in the abstract sphere of scientific t h i n k i n g and logical acumen. It makes him give the definition (laksana) 6.1 anudttam khytam mantritavot ' a verb is unaccented after the way of the vocative ' ; and 4.164 khyteh khayau kasau . . . . The o t h e r one is bound down by the practical necessities of life: it keeps in the low region where the Vedapthaka breathes, w h o knows the t e x t of his Samhit and nothing else, w h o does not care about the sense of what he recites, relying on a dogma expressed by MT. S. 1. 2. 39 (mantrnarthakyam) or by Kautsaas quoted in N i r . 1.15 (. . . anarthak hi mantrh). It is this soul that makes him give an amplification (prapaca) t o his definition 6.1 in the seemingly superfluous rule 2 . 1 4 ; that makes him add t o his definition 4.164 the absurd exception sakhyokhyamukhyavarjam, which guards against any w r o n g application of his teaching on the part of the unintelligent. If I were asked t o give my general impression of the V r t t i k a kra in one short sentence, I should say: These t w o souls live also in Ktyyana, the Vrttikakra. This wants some proof. I will give it in detail, for it forms my strongest argument. W i t h o u t it, I should probably have left matters as they stand, considering Prof. Keith's view as possible, though not likely. T h r o u g h o u t his w o r k the Vrttikakra does not deem it necessary t o add a wealth of vedic details t o Panini's rules. He is satisfied, for example, w i t h adding t o Pnini 6. . 94 in v r t t . 6 the general r e m a r k : emandisu cchandasi, which sums up V. Pr. 4. 53 sumudrasyemams, tvemams, tvodmann iti ca; or w i t h adding t o Pnini 2 . 1 . 2 in v r t t . 6:param api cchandasi, which reflects the circumstantial rules V. Pr. 2.18 and 19. Occasionally he explicitly states that it is impossible so t o complement by one short rule an injunction given by Pnini as t o make it comprise the vedic detail : v r t t . 1-4 on 6 . 1 . IP In v r t t . 1 on 1.1.6 he wants t o cancel the expressions dldhJand vevl because these roots occur only in the Veda, and because for the Veda [no general rules can be given but only] rules that describe afterwards what has been observed t o occur (drstnuvidhitvc ca cchandasi) . . . .29 For a student of the Astdhyyl it is sufficient t o know that such and such phenomena do occur in the Veda, he need not t r o u b l e about t h e i r exact where and when. The attempt t o reformulate Pnini's vedic rules w i t h a view of making them a safe guide for Vedapthakas of any descript i o n , w o u l d indeed be u t t e r l y impossible. It is different when Pnini expounds on some vedicdetail that is easy of verification, because it only can occur in one special context.
28 29

Quoted by Wackernagel, Altindische Grammatik, I, p. LXV n. 1.

Compare e.g. Patajali on vrtt. 1 on 6.4.141 (III, p. 225, I. 3), quoted by Wackernagel o.e., p. LXV n. 5.

350 Paul Thieme

W i t h such a detail are concerned Pnini's rules on the accent of the Subrahmanyformula, 1. 2. 37, 38. The Subrahmany runs t h u s : subrahmanyo3msubrahmanyo3msubrahmanyo3m indrgaccha hariva gaccha medtither mesa vrsanasvasya mengaur vaskandinn ahalyyai jrakausika brhmana gautama bruvna ityahe (i.e. : adya, svah, dvyahe, or tryahe) sutym gaccha maghavan dev brahmna gacchatgacchatgacchata.30 D u r i n g the Agnistoma, on t h e day preceding t h e pressing of t the Soma_, there is on a certain occasion (Lty. S. 1. 3.18; Drhy. S. 1. 3.18; p . S. 1 1 . 20. 3; Man. S. 2. 2. 5. 9; Kty. S. 8. 9.12) inserted in the Subrahmany, before the sentence svah sutym gaccha. . . (Lty. S. 1. 3. 20;'Drhy. S. 1. 3. 22), the following piece : asau (i.e. the Yajamna ' N. N.1) yajateamusya (i.e. ' N . N . ' s ' ) putro yajateamusya pautro yajateamusya napt yajateamusya pit yajate(amusyh pit yajate) amusya pitmaho yajate (amusyh o o ) amusya prapitmaho yajate(amusyh o o)janisyamnnm pit pitmahah prapitmaho yajate.3^ Pnini's rules run t h u s : 1. 2. 37 [ekasruti durt sambuddhau ( 1 . 2. 33)] na subrahmanyym svaritasya tdttah 38 devabrahmanor anudttah ' [Everything is] of equal pitch when it is a case of calling f r o m afar, [ b u t ] not in the Subrahmany formula. Here, however, an udtta vowel is substituted for asvarita; an anudtta is substituted for a svarita in the words deva and brahman'. They are supposed t o teach the following accentuation : subrahmanym ndrgccha, hrva gccha, mdtither mesa, vfsnasvasya men, gar vaskandinn, ahalyyai jra, kaska brhmana, gatma bruvna, svh (dvyah etc.) sutym gccha maghavan dev brahmna gcchata. The Vrttikakra has t w o possibilities of viewing and examining these rules. This reveals what I have called his t w o souls. His first possibility is t o examine the logical indemnity of Pnini's teaching itself. He indulges in it in his V r t t i k a on Pnini 1.2. 32. He says: ' Pnini 1. 2. 32-40 ought t o be taught after Pnini 8. 4. 66 in o r d e r t o make apply correctly Pnini 1. 2. 39' ( v r t t . i ) . T h i s means: The substitution according t o 8. 4. 66 of a (dependent) svarita for an anudtta that follows an udtta, has t o be considered as not having taken effect in any preceding rule according t o 8. 2 . 1 . Hence Pnini 1 . 2. 39 w o u l d apply only t o such anudtta vowels as follow an (independent) svarita that has been substituted according t o a rule like 6.1.185.
30

SB. 1.1.10,11 omits gacchotgacchotgocchata.SB. 3. 3. 4. 7ff. quotes subrahmanyo3m only twice; omits gaccha maghavan; quotes gacchata only once.T. 1.12. 3ff. gives the text only till bruvna. Lty. S. 1. 3. Iff. has etvadahe instead of ityahe (1. 3.1); leaves optional the sentence dev brahmna gacchatgacchatgacchata (1. 3. 3); leaves it optional to say gaccha maghavan (1. 3. 5), or simply gaccha (1. 3. 4), or to leave out both words.I have marked by hyphens the places where one

must inserta pause (according to Lty. S. 1.3.6, 7). Drhy. S. 1. 3. 2ff. substantially agrees.See also CalandHenry, L'Agnistoma, p. 65. 31 A more elaborate form is given by Caland-Henry, o.e., p. 119. I have tried to keep closely to the wording suggested by Lty. S. 1. 3.18-20; Drhy. S. 1. 3.18-20. The hyphens are inserted according to Lty.S. 1. 3. 9; Drhy. S. 1. 3. 9. The other srautastras (quoted above) allow no inference as to the exact wording,

351 The Identity of the Vrttikakra

' A n d also f o r t h e sake o f t h e second part o f Pnini 1 . 2. 37 (svaritasya tdttah)' ( v r t t . 2). This means: T h e s u b s t i t u t i o n o f an udtta f o r a svarita in t h e Subrahmany w o u l d apply again o n l y t o an independent svarita. Consequently we should get only subrahmanym instead of subrahmanyom (with svarita according to 6.1.185), but not gccha etc. for gccha etc. (with svarita according to 8. 4. 66). 'And also in order to prevent substitution of a svarita (according to 8. 4. 66) for [the anudtta] following the udtta that is substituted for a svarita (according to 1. 2. 37)' (vrtt. 3). This means: If 8. 4. 66 follows 1. 2. 37, it has to apply after 1. 2. 37 again. We should consequently get first gccha for gccha, and afterwards gccha for gccha. 'And also for the sake of Pnini 1. 2. 40' (vrtt. 4). This means: Also this rule would apply only to such anudtta vowels as precede an udtta or zsvarita that is taught in rule standing before Pnini 8.2.1. All these objections are perfectly correct. Pnini has not followed his own plan with all desirable care. The V. Pr. is in this respect blameless : the rules corresponding to Pnini 1. 2. 39 (V. Pr. 4.138) and 1. 2. 40 (V. Pr. 4.135) follow the rule corresponding to Pnini 8. 4. 66 (V. Pr. 4.134). But Ktyyana knows how to free Pnini of the net he has thrown over him. In his last vrttika on 1. 2. 32 he gives as his final view this: 'Thefact of Pnini teaching the substitution of an anudtta [for a svarita] in the words deva and brahman (devh for dvh, brahmnah for brhmnah) gives an indication that Pnini 8. 4. 66 has [to be considered to have] taken effect [as far as Pnini 1. 2. 32-40 is concerned, in spite of Pnini 8. 2.1].' This means : Since Pnini 1. 2. 38 would be perfectly void of sense and purpose if it did not concern the secondary svarita, and if Pnini 8. 4. 66 would apply afterwards again, we are entitled to infer that this paragraph (1. 2. 32-40) is meant to follow 8. 4. 66. As often, Ktyyana's subtlety has been a match for Pnini's. He has scrutinized his great predecessor's formulation, pointed out a logical flaw, and hit on a striking solution for the difficulties resulting: the rules in question are put in the wrong place. Then he has topped it all by finding out that Pnini himself has given an indication that he wanted us to interpret him as if he had proceeded correctly. Taken all in all, a neat specimen of scholastic acumen ! His second possibility is to examine the practical usefulness of Pnini's rules from the point of view of a priest who has to recite the subrahmany. He might say, firstly, that it is not clear whether durt sambuddhau is still valid from 1. 2. 33, and, secondly, that it is not clear which accentuation is really the correct one. Having formulated a negative rule by just forbidding to recite with equal pitch, Pnini has, strictly speaking, left open manifold possibilities. Theoretically, an accentuation like indr agccha would answer his definition also. Ktyyana must have thought likewise. In his Vrttika on 1. 2. 37 he actually rewrites the whole rule. He puts a prapaca in the place of Pnini's laksana. It runs thus: subrahmanyym okra udttah (vrtt. 1) kra khyte, paradis ca (vrtt. 2) vkydau ca dve dve (vrtt. 3) maghavanvarjam (vrtt. 4) sutyparnm antah (vrtt. 5) asv ity antah (vrtt. 6) amusyety antah (vrtt. 7) syntasyopottamam ca (vrtt. 8) va nmadheyasya. ' [Pnini ought to have said thus :] In the Subrahmany the vowel o is udtta (1); the vowel when followed by a verbform [is

352 Paul Thieme

udatta]; also the first of the following [syllables] (2); also the t w o first [syllables] in the beginning of each sentence [are udtta] (3); except [those of the w o r d ] maghavan (4) ; the end [of the words] followed by suty [is udtta] (5); the end of [the w o r d designated by] ' N . N . ' [is udtta] (6); the end of [the w o r d designated by] ' N.N.'s ' [is udtta] (7) ; [the end] and the last but one [syllable] of [a w o r d ] ending in sya [is udtta] (8); [the last but one syllable] of a name [is udtta] or [ n o t ] (9).' The last 4 rules must be concerned w i t h the piece inserted on the last day before the pressing : it is open t o doubt, w h e t h e r Pnini wanted t o include it in his rule. Let us t r y , then, t o apply Ktyyana's teaching t o the w h o l e ! V r t t . 1 : The vowel o is actually always udtta in the Subrahmany. N o t only in subrahmany3m, which is given as example by Patajali, but also in the inserted piece: putr [yajate], pautr [yajate], pitamah [yjate], prapitamah \yjate], V r t t . 2a: The vowel a is always udtta when followed by a v e r b f o r m . N o t only in gaccha, which is given as example by Patajali, but also in the inserted piece: napta [yajate], pit [yjate], and in the end : gacchata. V r t t . 2 b : The first of the following syllables is also udtta, says Ktyyana. Obviously we have t o accent not only gaccha, which is given by Patajali, but also: napt yjatepit yjate. W e even must go a step f u r t h e r , and construe paradis ca not only w i t h kra, but also w i t h okra (as may be indicated by the ca). For if we accent napt yjate, it is only logical that we accent likewise: putr yjate, pautr yjate, pitamah yjate, prapitamah yjate. V r t t . 3: The t w o first syllables of each sentence are udtta. This provides us w i t h the accentuation : ndrgccha, hrva [gaccha], mdtither mesa, vfsnasvasya men, gar vaskandin, hiyayai jara, kaska brhmana, gatma bruvna. All the following vrttikas are special exceptions (apavda) t o this last general statement (utsarga), for they teach only the accent of words that happen t o stand in the beginning of sentences of the Subrahmany. V r t t . 4 : maghavan, which of course in reality belongs t o the sentence ityah sutym gaccha, and not t o dev brahmna gacchata, is unaccented. V r t t . 5 : ady, dvyah, tryah, which do start a sentence, are accented on t h e i r last syllable only. V r t t . 6: The end of the w o r d designated by ' N . N . ' (asau) is udtta. For simplicity's sake, let ' N . N . ' be called by a name used by Patajali in his examples: grgya, dksi, or devadatt. First of all, the rule informs us that in sentences like grgyo yajate, dksir yajate, devadatto yajate, v r t t . 3 is superseded. For grgyo yjate, devadatt yjate we do not want any f u r t h e r r u l i n g : t h e i r accents result already from vrtt. 1 and 2. N o r is it possible t o say that vrtt. 6 would supersede t h e m . For neither is it an apavda w i t h respect t o v r t t . 1 and 2, since it applies also in cases that do not fall under the jurisdiction of these rules, nor can it be stronger by v i r t u e of its being taught later, since t h e r e is no ' conflict' 3 2 . dksir yajate is different. The name has normally an udtta in the first syllable. W e get, however, by v r t t . 6 only dksir yajate. A n d while everywhere else the syllable following an udtta becomes udtta itself before an anudtta, we seem t o have a single
32

See above.

353 The Identity of the Vrttikakra

udtta here. I think it is obvious that Katyayana has given his last v r t t i k a f o r an instance like this. As translated above, 33 it gives us permission t o accent the last but one syllable of a name. W h o wants t o use the name daksi can consequently say: dksir yajate, getting the first udtta from vrtt. 9 and t h e second from vrtt. 6. N o w everything is in order. Vrtt. 7: The end of a w o r d designated by ' N . N . V (amusya) is udtta. W e learn from this, again first of all, that in sentences like : grgyasya putro yajate, devadattasya putro yajate, vrtt. 3 is superseded. Secondly, that dkseh, which may have an udtta in the first syllable according to vrtt. 9, is also accented on the last: dksh putr yjate. Vrtt. 8: The end and the last but one syllable of a word ending in sya is udtta : gorgysy putr y jate, devadattsy putr yjate. Vrtt. 9 : See on vrtt. 6 and 7. The accents of the inserted piece are, then, according to Katyayana: gargy (devadatt) yjate (dksr yjate) ; gorgysy (devadattsy, dksh) putr yjate; napt yjate; pit yjate; pitamah yjate; prapitmah yjate.34 It thus becomes obvious that the lengthy vrttikas do not add or change anything with respect to the content of Pnini's rule: In the Subrahmany a svarita is replaced by the udtta. The one point on which the Vrttikakra differs is the accent of devh and brahmnah. According to Pnini 1. 2. 38 the svarita in these words is replaced by the anudtta. According to Ktyyana vrtt. 1 on 1. 2. 38 this is only the opinion of some authorities. Consequently he wants himself to accent: dvh, brhmnah, which restores complete analogy of accent throughout the Subrahmany. For our particular investigation, however, it is not essential what Ktyyana is teaching, but how he is teaching it. We may even leave the question open whether he actually is quite distinct and unambiguous. We saw that without some interpretative efforts,
Patajali understands vrtt. 9 in a different way. He believes it to teach optional correctness of devadattasya pit yjate beside devadattsy pit yjate, which is evidently wrong. 34 It would apparently be wrong to apply Ktyyana's teaching to the last sentence, which would have to be accented thus: janfsyamnnm pit pitamahh prapitmah yjate. It is possible that Katyayana did not know the use of it, seeing the many variations in the Subrahmany as pointed out in Lty.S. (above p.350 n. 30). Nor does he appear to have considered the cases amusyh pita yajate etc. Drhy. S. 1. 3. 23ff. also teaches the accentuation of the inserted piece. It differs from Katyayana in considering also the accent of feminine genitives and of the last sentence. It does, moreover, not provide for any name that is not antodtta. It runs thus: 1. 3. 23 arthanirvacanam uccntam
33

'what expresses [only] the sense [of the nominal stem] (i.e. a nominative) has a high-pitched (i.e. udtta) end ' (the name, putrh, putrh, napt, pit, pitamahh, prapitmahah). 1.3. 24 vaibhakte ca sydau vaibhaktas caiva ' what is preceding a sya etc. (i.e. yh) that belongs to a case termination, and [the vowel] belonging to the case termination [is highpitched] also ' (-sy, -yh pit etc.). 1. 3. 25 uccc ca nice nJcam ' a lowpitched (i.e. anudtta) syllable that is following a high-pitched and preceding a low-pitched syllable (i.e. a svarita) [is high-pitched] also ' (. . . pit pitmahh prpitamah . . .). 1 . 3. 26 janisyamnnm ti madhye dve 'the two [syllables] in the middle of ' janisyamnnm [are high-pitched]' (janisyamannm). 1. 3. 27 y ajes cdih 'the beginning of root yaj [is high-pitched] also' (putr yjate, pit yjate etc.).

354 Paul Thieme

his last v a r t t i k a is misunderstandable, and that if I am right, Patajali did misunderstand it. Moreover, Patajali hardly realized the full bearing of t h e first vrttikas, as interpreted above. For us is only essential Ktyyana's unmistakable intention of being distinct, and his obvious impression that Pnini was not distinct enough. Let us admit that theoretically Pnini's formulation is ambiguous, W e have t o acknowledge, at t h e same t i m e , that it reveals insight into t h e linguistic phenomenon as such. It clearly grasps the essential feature of t h e accenting particularity of t h e Subrahmany. In this respect it is unsurpassable. Let us admit that theoretically Ktyyana's formulation is clearer. W e have t o acknowledge, at t h e same t i m e , that it is much more circumlocutory, and that it does not reveal any insight into the linguistic phenomenon. It is superficial and mechanical t o the last degree, though in its o w n way ingenious enough. Is it not t h e author of t h e Prtiskhya w h o spoke t o us in these Vrttikas? Let us, t o facilitate o u r argument, apply one last test. Let us suppose t h e V r t t i k a and Bhsya on Pnini 1. 2. 37 had been lost, and by some lucky chance t h e r e had been found, in o u r days, some palmleaf containing only Ktyyana's 9 rules w i t h o u t Patajali. W h a t w o u l d philologists do w i t h it? Some w o u l d say: 'These rules must be pre-Pninean, f o r t h e expressions okra, kro, khyta have been replaced by Pnini by t h e shorter ones : ot, at, tin.' I am afraid Prof. Keith w o u l d be amongst t h e m . For in his opinion ' everything points t o t h e condensation of t h e AstdhyyT as t h e carrying t o perfection of an endeavour t o attain brevity for its o w n sake [ n o t e : e.g. ku for kavarga].'35 A n d if he finds it ' q u i t e impossible t o believe that V. Pr. 1. 55 amtrah svaro hrasvah [etc.] is an attempt t o improve on Pnini [ 1 . 2. 27 kio 'jjhrasvadJrghaplutah]',36 he cannot believe that o u r rules are an attempt to improve on Pnini 1. 2. 37 either. Not only is Pnini's rule about five times shorter, to us it appears also to contain the clearer formulation. And how much more clever is its point of view! Prof. Keith thinks Pnini 1. 2. 27 'absurd ', yet he maintains it to be a ' refinement', though a ' not very happy' one. Pnini 1. 2. 37 is not absurd at all. It just looks a very happy refinement of our rules. Most scholars should say: 'These rules must betaken from some of the Prtiskhyas, in which terms like okra, kra, khyta are in permanent usebe they young (as the AV. Pr.) or old (as the V. Pr.). If so, they must belong to either a Smaveda Prtiskhya, since the priest reciting the Subrahmany is a Smavedin, or a Prtiskhya to the White Yajurveda, since according to the Vjasaneyins it is the Adhvaryu himself who recites the Subrahmany. '37 Upon this they would find that the V. Pr. not only indeed uses the terms quoted above, but also that it is well acquainted with expressions of the type ' maghavanvarjam' (cf. e.g. V. Pr. 1. 87; 1.131 smajapanynkhavarjam : Pnini 1. 2. 34 ajapanykhasmasu;V. Pr. 2.1 ; 4. 21 ; 4.164). Further they would ask themselves, which word is referred to by the neuters dve dve (vrtt. 3) and upottamam (vrtt. 8), and they could not but recognize that it must be aksara n. They would now remember that V. Pr. 1. 99 says : svaro 'ksaram ' a vowel is cal led aksara [together with the
35 36

Indian Culture, Vol. II, p. 745. I.e., p. 746.

37 Caland-Henry, o.e., p. 64, No. 49 n.2.

355 The Identity of the Vrttikakra

preceding consonants] ', and uses aksara in this sense in 4.129, and makes us supply aksaram t o sv a rita m in 4.132,134 etc. etc. N o w as regards ' doctrine, which is the sole criterion available', 38 they would certainly b e s t r u c k by the fact that the beginning of our rules does not seem t o presuppose acquaintance w i t h the most simple grammatical facts and categories. But soon they would come across the terms khyta and vkya, which betray quite a decent standard of grammatical knowledge. After just having forgiven the author his unintelligent first rules, they would hit on the f o u r t h , which shows that either the author or those for w h o m he w r o t e were not able t o discern that maghavan belongs t o the sentence : ityah sutym gccha maghavan, though its accent forbids t o construe it w i t h the f o l l o w i n g : dv brhmna gcchata. Again they would remember that the V. Pr., too, though presupposing acquaintance w i t h the concept khyta (finite verb) in 6.1, does not expect in 2. 14 its students t o realize that srutam in V.S. 7. 9 is an khyta ; that the V. Pr., though presupposing acquaintance w i t h the concept mantrita (vocative) and sasthl (genitive) in 2.17,18, does not expect in 2.19 its students t o realize that apm napt in V.S. 8. 24 is not a vocative, but a nominative ; that the V. Pr., though presupposing acquaintance w i t h the grammatical abstraction khyti (root khy) in 4.164, does not expect in the same rule its students t o realize that the khy of sakhya, ukhya, mukhya has t o do nothing whatsoever w i t h the khy of root khy. Those would not be the worst of our hypothetical scholars, who t r y t o find a place in Sanskrit literature for our hypothetical pal m leaf, that w o u l d insist on the circumstance that just this contradictory side by side of a respectable standard of knowledge and mechanical, insipid pedantry, which we find in our rules, gives its individual character t o the V. Pr. ; and that the t w o fighting t e n dencies of stating each single case by itself and of giving general, comprehensive directions eventually spoil each other's game as well in our rules as in the V. Pr. Sceptics might rely on the Subrahmany not occurring in the V.S. But theirs would be a weak argument: the author of the V. Pr. could well have taken special interest in it, since it is recited by the Vjasaneyin Adhvaryu. They might point out, further, that once we allow the author t o take his point of view as he chooses, these rules are quite a chef-d'uvre in t h e i r skilful arrangement and artful disposition, which makes use of the really surprising accident that all the vowels o, and all the vowels before a form of a verb, are udtta in the Subrahmany. In this respect hardly one piece of the V. Pr. can compare w i t h t h e m . Yet these sceptics certainly would have a hard stand. If, however, anyone should suggest that our rules are really taken from the Vrttikahe might refer t o the expressions okra, kra, vkya and the type maghavanvarjam being not foreign t o its style, he would be answered by the same argument, on which Prof. Keith declines 39 t o believe in the identity of the Vrttikakra and the author of the V. Pr.: 'The plain fact is that the Vrttikakra is far advanced in grammatical knowledge beyond the author of these rules.' Fortunately we have not t o deal w i t h an anonymous pal m leaf. Fortunately we are not placed in the awkward position of having t o prove that our 9 rules were w r i t t e n by the Vrttikakra. Fortunately by indulging in his Vedapthaka inclinations, Ktyyana
38

Keith, I.e., p. 746.

39

I.e., p. 742.

356 Paul Thieme

himself has led ad absurdum all the arguments by which Prof. Keith still wants t o settle the relative chronology of the Prtiskhyas. Ktyyana's rules show distinctly that in chronological questions we can rely neither on the brevity of technical terms, nor on the brevity of expression in general, nor on the more or less 'advanced ' doctrine. They show more. They show that the suspicion of Ktyyana, the Vrttikakra, and Ktyyana, the author of the V. Pr., being identical was not vain. It is not only easy t o assume that the scholar who penned the V r t t i k a o n Pan in i 1. 2. 37 s responsible also for the V. Pr., it is almost unavoidable when it is considered that t h e V r t t i k a shares, in its expressions as well as in its method of representing linguistic facts, marked peculiarities w i t h the V. Pr. I defy scholars t o find out a similarly close relationship between any t w o other grammatical works of Indian antiquity. The proof that the Ktyncrya, who w r o t e the V. Pr., is not the Vrttikakra now lies w i t h the sceptics. W i t h W e b e r ' s arguments they cannot defend t h e i r case. They have t o explain away the relationship of the terminology of the V r t t i k a and the V. Pr., which was unknown t o W e b e r , and the surprising coincidences of a number of deviations of the V. Pr. f r o m Pan in i w i t h objections raised by Ktyyana against the very points of Pnini's grammar that are removed by these deviations, coincidences, which are far more numerous and characteristic than those pointed out by G o l d stcker. They have t o explain how, if t h e i r idea of the development of linguistic studies and teaching technique in India is right, Ktyyana could ever have w r i t t e n the V r t t i k a on Pnini 1. 2. 37. W i l l they not rather admit that t h e i r simple and seemingly easy assumptions (a lucid w o r k : early, obscure: y o u n g ; diffuse: early, brief: young, etc.), have not taken in account the complicated nature of historical development, which seldom, if ever, follows a straight line? Is it not obvious that they have neglected t o consider that books are not w r i t t e n by schools, but by single men, by individuals, w h o naturally had t h e i r own taste, t h e i r own inclinations and preferences, and little cared when adopting new styles, new methods, and new points of view, w h e t h e r t o later centuries t h e i r procedure meant progress or not? Unfortunately, in Indian literature it is but seldom possible t o recognize the author behind his w o r k . In the case of Ktyyana, I t h i n k , we can do so; butas soon as we catch a glimpse of his personality, his writings cease t o be measurable by the simple standards generally accepted, and thus prove t h e m t o be a r b i t r a r y and u t t e r l y unreliable.

We have seen that the first information available in Europe on the

I I

Pierre Boudon

Sanskrit grammarians was due to the French Jesuit Jean Franois Pons (see pages 30-32). Pons' Sanskrit grammar, available in manuscript, enabled A. L. de Chzy to occupy the first chair for Sanskrit at the Collge de France in 1814. But despite the excellent traditions of Sanskrit studies which developed in Paris during the nineteenth century, there were hardly any French scholars who contributed directly to the study of the Indian grammarians. Of course, the Sanskrit grammarians were studied for the historical data they provide. The well-known French Sanskritist and Indologist Sylvain Lvi (1863-1935; cf. Renou 1936) for example in a brief communication of 1891, ingeniously demonstrated that Ktyyana must have been a contemporary of one of the few kings (e.g., Piyadasi) who were referred to in inscriptions with the honorific title devnmpriya ' dear to the gods.' What is remarkable about this compound is that its first member retains the Genitive ending. According to Pnini 3.6.21, it must therefore have a pejorative meaning. Ktyyana, however, allows for an honorific use by listing devnmpriya among the exceptions to Pnini's rule. Since Patajali again reverts to Pnini's view and accepts only pejorative meanings for compounds of this kind, and since later usage continues to confirm these meanings only, Ktyyana must have been a contemporary of those few kings which were honored with the title devnmpriya. This fixes his date around the middle of the third century B.C. In another note (Lvi 1906-1908), Sylvain Lvi analyzed Pnini's treatment of preverbs in 1.4.80-82, together with Ktyyana's vrttiks (but omitting Patajali's comments, "qui supposent une extrme familiarit avec les dtails du systme de Panini," page 277). He concluded that Pnini treated a stage of the language intermediate between Vedic and " le terrain encore anonyme du sanscrit qui nat" (page 279). Sylvain Lvi also directed the excellent dissertation of a visiting Indian scholar, Vasudeva Gopala Paranjpe, Le Vrtika de Ktyyana. Une tude du style, du vocabulaire et des postulats philosophiques (1922). But it was his best pupil and eventual successor, Louis Renou (see pages 432-525), who restored the balance almost singlehandedly and placed French Indology in the forefront of Indian grammatical studies. At the time Renou was beginning to contribute to the study of the grammarians, another French Sanskritist, Pierre Boudon, worked in the same field. Boudon contributed only one publication: the following article, entitled "Une Application du raisonnement par l'absurde dans l'interprtation de Pnini (les jnpakasiddhaparibhs)", which appeared in the journal asiatique (230, 1938, 65-121)! Like Buiskool's book, Boudon's article was based in the first place on Kielhorn's edition of the Paribhsendusekhara, the one publication that still provides the best introduction to the study of the later grammarians. Boudon's article deals with metarules (paribhs), some of them artificial, and also with problems connected with the relative strength of rules. The application of Pnini's rules sometimes gives rise to contradictions. These are interpreted as apparent contradictions which in fact serve a purpose; they are devices indicating (jnpaka) that a paribhs is needed. Such a paribhs, which is inferred from a jnpaka, is called a jnpakasiddhaparibhs.

Une Application du raisonnement par l'absurde dans l'interprtation de Pnini (les japakasddhaparibhs) (1938)
Pierre Boudon

I "ardhamatralaghavenaputrotsavam manyante vaiykaranh"

C'est un postulat admis par les commentateurs grammaticaux sanscrits que Pnini n'a jamais employ une demi-mtr qu'elle ne ft indispensable. Or on rencontre de loin en loin dans l'AstdhyyJ une lettre, un mot, parfois tout un stra, ou encore un procd de composition, qui premire vue sont inutiles ou inexplicables. Ce qui fait croire l'inutilit de ces lments, dans la rgle o ils se trouvent, c'est le plus souvent l'existence d'une autre rgle (ou d'une maxime) qui semble exprimer la mme ide. Ailleurs telle nonciation parat superflue en ce que l'objet qu'elle vise aurait pu, semble-t-il, tre atteint sans elle, vu l'apparence d'une double possibilit d'interprter ou d'appliquer une rgle (ou une srie de rgles), l'une avec, l'autre sans cette nonciation. Les commentateurs de Pnini en premier lieu Patajali qui ne pouvaient admettre une drogation leur postulat, se sont efforcs de donner un sens ces nigmes : ils ont, selon le cas, soit ni la valabilit universelle de la rgle ou maxime apparemment concurrente, soit rejet l'une de deux possibilits d'interprtation, pour ne laisser subsister que celle qui comporte l'utilisation de l'lment autrement superflu. Et la maxime ou interprtation gnante est rfute par l'absurde: sa fausset (au moins partielle) rsulte de ce qu'elle aboutirait une consquence inadmissible, la prsence d'un mot ou procd inexplicable dans les stras de Pnini. Le mot ou procd ainsi sauvegard devient le signe rvlateur d'une maxime d'interprtation, d'une paribhs. Une maxime tout fait gnrale (paribh. 51 : eknt anubandhh) enseigne que es anubandha2 font partie de ce quoi ils sont attachs. Si cette maxime tait toujours vraie, le mot sit de 1,1, 55 : "anekl sitsarvasya"3 serait inutile. Car un substitut multilittre
La numrotation suivie est celle du recueil de Ngojbhatta, Paribhsendusekhara (Ed. Kielhorn, Bombay, 1868; Part II. Translation and notes, Bombay, 1874). Cet ouvrage tant frquemment cit dans ce qui suit, les abrviations PS. et Kielh. renvoient, la premire au texte sanscrit, la seconde la traduction, que j'ai gnralement suivie pour la partie technique des sections I et \\ de ce travail. 2 Les anubandha sont des lettres supplmentaires que Pnini attache aux suffixes, substituts, augments, racines, pour en marquer les diffrentes proprits quant l'accentuation, modification du thme, etc. Ainsi les dsinences e, as, as, i des datif, ablatif, gnitif et locatif singulier, figurent dans Pnini sous les formes techniques e, asi, as, ri/ avec l'anubandha commun ri, dont le rle est de rappeler conventionnellement les caractristiques communes ces
1

quatre dsinences et dfinies par les rgles 1,4, 6, 7, 3,111, etc. (L'anubandha / de asi distingue la dsinence de l'ablatif de celle du gnitif). "Anubandha" a pour synonyme dans Pnini le mot " t", lequel est toujours employ comme second terme du bahuvnhi dsignant le mot qui a pour anubandha la lettre constituant le premier terme du compos. Les quatre dsinences ci-dessus sont des " nit" , i.e. des lments qui ont pour it (pour anubandha) fa lettre ri. Le mot "it" sert par suite de dsignation collective dans les rgles relatives ces dsinences. 3 1 . 1 . 55: " Un substitut multilittre (anekl) er un substitut sit prennent la placedetout(l'original)". Cesont l deux exceptions la rgle gnrale de substitution "alo 'ntyasya" (1,1, 52), qui enseigne qu'un substitut prend la place de la dernire des lettres de l'original.

359 Raisonnement par l'absurde dans Pnini

(sous sa forme technique) pourrait ne consister qu'en un substitut rel d'une seule lettre accompagn de son anubandha, et la catgorie des sit rentrerait dans celle des anekl. Mais si le mot sit tait inutile, Pnini aurait viol son principe d'conomie; ceci tant une absurdit, il faut conclure que la paribhs 5 est fausse, au moins en ce qui concerne la rgle 1 , 1 , 55. Telle est la rfutation par l'absurde et sa conclusion limite, partir de laquelle,par une deuxime dmarche qui se prsente sous les dehors d'une induction, mais qui n'est peut-tre qu'une citationest invoque une maxime nouvelle destine corriger la premire. C'est la paribhs 6: " nnubandhakrtam onekltvom", 14 La m u l t i l i t t r i t n'est pas cre par les anubandha". Le mot sit, de superflu qu'il apparaissait d'abord, prend maintenant sa valeur et mme une double valeur, puisqu'en mme temps qu'il indique la paribhs 6, il apparat indispensable aussitt qu'on admet celle-ci. De vyartha ce mot devient caritrtha*. Il est dit lejnpaka i. e. l'indice de la paribhs 5 . La jnpakasiddhaparibhs, avec le raisonnement qui prtend l'tablir, est i n t r o d u i t e dans le Bhsya, soit dans le stra mme qui contient le jnpaka, propos d'une critique faite au t e x t e du stra, t e x t e que la paribhs vient justifier, soit, plus souvent, ailleurs, en rponse diverses objections. C'est ainsi que dans le Bhsya, 1,3,9, est souleve la question : les anubandhafont-ils partie de ce quoi ils sont attachs? L'un des interlocuteurs rpond qu'ils en font partie. Sur quoi un autre fait observer qu'il en rsulte trois difficults . . . l'une relative au sarvdesastra ( 1 . 1 , 55). Avec cette thse en effet tous les substituts seraient anekl] par suite le substitut aut(=au) prescrit par 7 , 1 , 84 6 , par exemple, serait p a r i , 1, 55 substitu la totalit du thme div, alors qu'il doit tre substitu, par 1 , 1 , 52, la finale seulement. L'objection est en fin de discussion carte par Patajali au moyen de la paribhs 6 appuye sur son jpaka: "yad apy uktam sarvdesa ti tatrpy cryapravrttir jnpayati nnubandhakrtam anekltvam bhavatJtiyadayam sitsarvasyetyha". " E n ce

qui concerne l'objection introduite par le mot sarvdese, cela aussi (il y a rponse car) il est une procdure de Pnini qui nous
4 Ces deux conditions sont ncessaires au moins selon Ngoj. pour qu'il y ait vritable jnpaka. Car il n'admet pas qu'une paribhs puisse tre indique d'une manire conventionnelle par un mot ou procd qui, la maxime adopte, ne serait pas caritrtha, (Voir PS. rfutation des jnp. de 93,119, justification du jnp. de 28). D'autre part, si le mot ou procd a une raison d'tre dans l'hypothse de l'inexistence de la paribhs, il n'est pas valable non plus comme jnpaka. (Voir rfutation des jnp. de 52 [premier jnp. allgu], 90,120, justification des jnp. de 17, 28, 53, 55). 5 Le raisonnement qui procde du jnpaka trouve son illustration la plus intressante dans l'tablissement des

jnpakasiddhaparibhs, qui ont en gnral pour but de dfendre, en l'interprtant, le texte des stra. Mais ce mode d'argumentation revient sans cesse dans le dialogue du Bhsya, o il est galement employ, pour tayer la critique des stra, soit, le plussouventtort, par l'interlocuteur qui n'est que partiellement inform (ekadesin), soit, valablement, par celui qui, sachant fond la doctrine (siddhntin), exprime la pense de Patajali. De ces jnpaka de stra on trouvera deux exemples ci-aprs, pages 382-385 et 386-388. 6 7,1, 84: diva out (sau) "ouest substitu (la finale de thme) div devant su (=s, dsinence du nominatif singulier)": dyauh.

360 Pierre Boudon

suggre cette maxime: " La multilittrit n'est pas cre par les anubandha", savoir le fait qu'il a dit sitsarvasya". Cette formule, peu prs invariable, dont Patajali se sert pour i n t r o d u i r e une jnpakasiddhaparibhs, prsuppose le rejet de la thse contradictoire par une rduction l'impossible qui, pour n'tre pas toujours exprime 7 , n'en est pas moins l'opration capitale de l'argumentation qui prtend fonder la paribhs. A u t r e exemple. La paribhs 30 vyapadesivad ekasmin* enseigne: " Une opration subie par un mot complexe raison d'une dsignation spciale (par exemple du type taddi ou tadanta) qui dfinit cette complexit, est subie de mme par le mot simple c o r r e s p o n d a n t " . Ainsi la rgle 4 , 1 , 95, ota i " (dans le sens de descendant), le suffixe i vient aprs un thme a", (c'est--dire, p a r i , 1, 729, adanta) s'applique non seulement dans le cas Daksa-\7 Deux fois (pour ne parler que des jnpaka de paribhs), on rencontre la majeure du syllogisme conditionnel explicitement formule: Bh. ,1,1,1 (paribh.109 jnp.: udttah de 7,1, 75): "yadi bhedak gunh syur udttam evoccrayet ". "Si leurs qualits suffisaient diffrencier les lettres, Pnini aurait (simplement) prononc udtta (le substitut ana, sans lui assigner cette qualit par le mot "udttah")". Bh. 7, 2, 98(paribh. 52 jnp. : anuvrtti ex (7, 2, 91 per) 7, 2, 95, 96 in 7, 2, 98 de " maparyantasya" \yady atrnya ekavacandesh syur maparyantnuvrttir anarthik syt". II s'agit des composs du type tvatputrah ou tvoddhitam de:

yusmad -\-nas -\-putrah > yusmad + putrah (2, 4, 71) > tvatputrah (7, 2, 98); yusmad + e+hitam > yusmad + hitam . . . > tvaddhitam . . . L'lisionpar/uk(2,4, 71)dela dsinence du gnitif (as) ou du datif (ne) est effectue en premier lieu. Aprs quoi, par 7, 2, 98, tva est substitu la partie du pronom qui a pour limite inclusive ma (maparyanta), c'est--dire yusma. La validit, littralement le passage par roulement (anuvrttir) de "maparyantasya" dans 7,2, 98 est indispensable, pour empcher que tva, substitut multilittre (anekl), ne soit substitu, en vertu de 1,1, 55, la totalit du pronom yusmad, Le raisonnement de Patajali nous dit quelle condition (inadmissible) la validit de "maparyanta" serait inutile. "[Si l'antarangaparibhss'appliquait ci, 7, 2, 95 (96), antaraga,

l'emportant sur 2, 4, 71, bahiraga, et par suite] S'IL POUVAIT Y AVOIR D'AUTRES SUBSTITUTS DU SINGULIER [ savoir tubhya, mahya de 7, 2, 95 (tava, marna de 7, 2, 96) que les substituts tva, ma de 7, 2, 98 (d'o conflict), ces derniers l'emporteraient en qualit d'apavda, et en vertu de la maxi me que les substituts d'un apavda correspondent au mme original que les substituts d'un utsarga], L'ANUVRTTI DE "MAPARYANTASYA" SERAIT INUTILE ". L'anuvrtti de mapary0 ne peutqu'tre indispensable. Il s'ensuit que les substituts tva, ma de 7, 2, 98 sont les seuls possibles, et ce parce que, dans le conflit 2, 4, 71 7, 2, 95 (96), c'est la rgle 2, 4, 71 qui s'applique en premier lieu, lidant par luk les dsinences du gnitif et du datif et supprimant du mme coup la possibilit des substituts prescrits par 7, 2, 95 (96) au devant de ces dsinences. Par l se trouve tablie la paribhs 52 " antaragan api vidhn bahirango lug bdhate". " Une rgle de luk bahiraga l'emporte mme sur des rgles antaraga". La paribhs 52 est une restriction l'antarangaparibhs (paribh. 50) explique ci-aprs pages 362-366 les notions d'apavda et d'utsargasontdfiniesci-aprs p.375, n.67. 8 Littralement : "comme le dsign, (ainsi) dans le cas du simple". 9 1 , 1 , 72 : "yena vidhis tadantasya", littralement : " l'occasion de quoi il y a rgle, (rgle vaut) pour ce qui a pourfin cela", c'est--dire " Une opration grammaticale prescrite pour tel lment (lettre, suffixe etc.) s'applique ce qui finit par cet lment".

361 Raisonnement par l'absurde dans Pnini

i = Daksi, mais aussi dans le cas A (Visnu) + in = I ( u n descendant d e A), bien q u e le t h m e A, t a n t fait d ' u n e seule l e t t r e ne soit pas capable par l u i - m m e de r e c e v o i r la dsignation " a d a n t a " . Si c e t t e paribhs tait u n i v e r s e l l e m e n t valable, la sparation des d e u x rgles 5, 2, 86 "prvd / n / V ' 1 0 e t 87 "saprvc c a " 1 1 n ' a u r a i t aucun sens. C a r u n e rgle applicable ce q u i f i n i t par prva (ici 5, 2, 87, o , saprvaprva = prvnta) serait aussi applicable au s i m p l e prva, e t il a u r a i t suffi de d i r e tlprvt saprvd nih1'. La s p a r a t i o n des d e u x rgles ne p e u t q u ' t r e indispensable. C'est d o n c q u e prva n'est pas t r a i t c o m m e prvnta. C e t t e c o n s t a t a t i o n suggre u n e m a x i m e n o u v e l l e , q u i l i m i t e la p r c d e n t e : "vyapadesivadbhvo ' p r t i p a d i k e n a " , " Le t r a i t e m e n t vyapadesivad ( i . e. la p a r i b h . 30) ne s'applique pas un t h m e n o m i n a l " . Ceci est la paribhs 32 12 . Les deux paribhs 6 et 32 ont un caractre limitatif et n'existent qu'en fonction d'autres maximes. C'est l le caractre du plus grand nombre des paribhs qui reposent sur un jnpaka. Il y a conflit entre l'utilit d'un mot ou procd de Pan ni et la valeur, au moins universelle, d'une maxime (paribhs des recueils ou paribhsstra); ce conflit se rsout par une maxime nouvelle qui corrige la premire, formule en termes trop absolus. Les autres jnpakasiddhaparibhs sont des maximes indpendantes, mais on peut faire ici une nouvelle distinction. Certaines de ces maximes (9, 87, 88, 89, 50 [cas 1]) peuvent se ramener au type restrictif en ce qu'elles constituent des drogations ce qu'on serait tent de prendre pour des vrits premires grammaticales. Par exemple, on croirait que renonciation d'un
10 5, 2, 86 : le suffixe ni ( = in) vient aprs le mot prva " (dans le sens de anena krtam, c'est--dire en vue de dsigner la personne qui est la premire accomplir une action quelconque). Example: prvam gatom (bhuktam, pltom, etc.) anena= prvl. 11 5, 2, 87 : "(le suffixe ni) vient aussi aprs (le mot prva) prcd d'un autre mot" (dans le mme sens). Exemple : krtaprvl (katam). 12 La paribhs 32 apparat dans la discussion du vrt. 15 "tasya ca" 1. 1, 72, vrt. ncessaire pour expliquer le driv rauna, de ronl + an (4, 2, 78), la formation duquel s'oppose prcisment la paribhs 32. Celle-ci a sa raison d'tre dans les nonciations strnta (4, 2, 60), dasnta (5, 2, 45), o l'emploi de "anta" indispensable, vu le vrt. 3 1,1, 72 " samsapratyayavidhau pratisedhah" (qui enseigne que dans les rgles de composs et de suffixes un prtipadika ne dsigne que lui-mme et non ce qui finit par ce prtipadika), ne peut pas indiquerqu'il y a drogation la paribhs 30, ni par dispenser de la paribhs 32. La discussion se termine ainsi : " cryapravrttir japayati vyapadesiva-

dbhvo 'prtipadikeneti yad ayarp prvd inih saprvc cety ha. naitad asti npakam/ asti hy any ad etasya vacane prayojanam. kirn saprvt prvd inim vaksymlti yat tarhi yogavibhgam karoti.j itarath hi prvt saprvd'mirity eva bryt.1' "II est un procd de Pnini qui indique : " Le traitement vyapadesivad ne s'applique pas un prtipadika", savoir le fait qu'il a dit "prvd nih" (quand d'autre part il disait) ''saprvc ca" [celui qui parlecomprend : "aprs un mot (quelconque) prcd d'un autre mot"]. Cela n'indique rien, car il y a un motif autre cette nonciation. Lequel? (Le motif est que Pnini a ceci en tte :) C'est aprs lesaprva "prva" (i. e. c'est lorsque le mot lment antcdent est prva), que je veux prescrire le suffixe ni. ("prva" de 5, 2, 86 est donc destin passer dans le stra suivant; n'tant pas vyartha, ce mot ne peut pas tre jnpaka). Alors c'est la sparation des deux rgles (qui est le jnpaka), car autrement (i.e. si la paribh. n'existait pas), Pnini dirait (en une seule rgle) "purvt saprvd inih".

362 Pierre Boudon

suffixe individuel ne dnote que ce suffixe: la paribhsa 87 enseigne que le suffixe an, lorsqu'nonc au locatif il est la cause d'une opration grammaticale, dnote aussi le na tcchJlika. Les autres paribhs du mme type autonome (79, 80, 81, 82, 86, 90, 50 [cas 2]) n'ont pas p r o p r e m e n t parler un caractre restrictif, mais d t e r m i n e n t le choix dans un cas douteux. La racine p, dans Pnini, dnote-t-ellepdt/ ou pibatil Selon la paribhsa 90, eue ne peut dnoter que pibatiu. Dans l'une comme dans l'autre de ces deux catgories de maximes indpendantes, il s'agit de deux possibilits a priori d'int e r p r t a t i o n ou application des rgles (ou suites de rgles). Dans l'une tel lment est indispensable, dans l'autre il serait superflu ; une rduction l'impossible rejette celle des deux hypothses qui ne t i e n t pas compte de ce qui se t r o u v e devenir ainsi le jnpaka de l'autre hypothse. L'antarangaparibhs (paribh. 50), deux aspects et deux jnpaka correspondants, offre un exemple de chacune de ces deux dernires varits de paribhsa. Elle enseigne que la succession naturelle des rgles n'est pas toujours possible (cas 1) et permet de choisir entre deux oprations qui se prsentent simultanment (cas 2). Un premier jnpaka est le mot th de 6, 4,132 15 . th ( = ) est un samprasrana spcial qui a pour but laformation d'un accusatif pluriel comme visvauhah. O r il semble qu'on pourrait y arriver aussi bien par le samprasrana ordinaire, comme le m o n t r e la comparaison que voici : visva -f vh -\-nvi-\- sas visva -f- uh + O + ah 6, 4,132 "vhah"
visva u h O a/6,1,1,10817

visva
visv
13

oh
au h

O
O

ah7,3,8616
oh 6 , 1 , 8 8 1 8

visva + vh -f nvi + sas visva + h +O -f ah 6, 4,132 "vah th" visva h 0 ah 6,1,108 visv au h oh 6,1, 8919 ainsi comprise s'appliquerait indment bhinatti, chinatti, etc. Par suite, Pataj. interprte " puganta laghpadhasya" comme le gnitif d'un samahradvandva de puganta (tatpu rusa) + laghpadh (karmadhraya), gnitif d'un substantif en apposition ko au lieu de gnitif d'un adjectif qui qualifierait agasya. D'o, la traduction : "guna devant srva- et rdhadhtuka de \'ik d'un aga, quand cet k est soit un puganta, soit une laghpadh". Sur puganta tiquet "tatpurusa" voir Ngoj., Uddyota, ad 1,1, 3, vrt. 6. 17 6,1,108: samprasranc ca (ekah prvaparayoh prvah) "Aprs la voyelle d'un samprasrana (et devant voyelle de la racine), la prcdente, i. e. la voyelle du samprasrana est le simplesubstitutde la prcdente et de la suivante". Cela veut dire : Le samprasrana, i. e. la transformation en voyelle de la demi-voyelle qui prcde la voyelle a, s'accompagne de la disparition de cet a de la racine.

87 : "tcchllike ne 'nkrtni bhavanti " "(Les oprations) dtermines par le suffixe an ont lieu aussi devant le na tcchilika" (na de 4, 4, 62). 14 90: " lugvikaranlugvikaranayor alugvikaranasya" " De deux (possibilits d'interprtation:) racine vikarana uk ( = racine de la 2 classe), racine vikarana autre que uk, (c'est la seconde qui prvaut et une rgle ne vaut que) pour la racine vikarana autre que uk". 15 6, 4,132 : vha th (samprasranam bhasya) " est le samprasrana de la demi-voyelle de vh, lorsque vh est un thme bha (i. e. devant suffixe commenant par y ou une voyelle. Pan., 1,4,18). 16 7,3,8G:pugantalaghpadhasya caiko agasyagunahsrvadhtukrdhadhtukayoh). Traduction ordinaire: "guna devant srva- et rdhadhtuka (Pn^, 3, 4,113 et 114) de l'/k (/, Lf, r, /) d'une base (aga) qui a l'augment puk ( = p) comme finale ou une brve comme pnultime". Mais la rgle

363 Raisonnement par l'absurde dans Pnini

tant donn que le suffixe nvi20 (3, 2, 64) n'est ajout la racine vah que si celle-ci est en composition avec un premier terme termin par a, il n'y aurait jamais d'autre substitution oprer, aprs samprasrana et guna, que celle de au pour a + o, et cette substitution serait ralise par 6 , 1 , 88, sans le secours d'un second procd de formation. L'avantage de celui-ci, c'est qu'il permet d'arriver visvauhah sans passer par le guna de 7, 3, 86. Il semblerait naturel qu'une succession de rgles applicables un thme d t s'effectuer sans obstacle dans l'ordre des transformations du thme, chaque tat intermdiaire tombant automatiquement sous la rgle nouvelledont il raliserait les conditions. Si ce principe s'appliquait ici, i. e. si le gunade 7, 3, 86 tait possible aprs le samprasrana de 6. 4,132 (lue "vhah " ) , le mot th serait inutile. Ceci ne pouvant tre, on conclut qu'est seul praticable le procd qui part d'une rgle "vha th". Pourquoi l'opration de 7, 3, 86 n'est-elle pas ralisable aprs celle de 6,4,132, condition ncessaire et suffisante de la premire 21 ? Explication : le samprasrana, bien qu'effectu, est considr par convention comme irrel (asiddha) au moment o il s'agit d'oprer la substitution du guna. Et ce qui dtermine Vasiddhatvam de 6. 4,132 en face de 7, 3, 86 est rvl par l'examen des conditions des deux rgles dans leur application au cas

visva + vah

+ nvi + c/s uh oh 7, 3, 86 Antar1 6, 4,132 "vhah" Bahir

7, 3, 86 dpend du suffixe nvi dont l'effet subsiste en vertu de 1,1, 62 mme aprs substitution du lopa. 6, 4,132 dpend du suffixe os, au moins mdiatement, puisque c'est l'initiale vocalique de ce suffixe qui fait que vah est un thme bha.22 Les deux causes sont
18 6 , 1 , 88 : vrddhir eci (t ekah prvaparayoh) "Aprs a (finale du mot prcdent) et devant diphtongue (initiale du mot suivant), la vrddhi est le simple substitut de la prcdente et de la suivante". 19 6 , 1 , 89 : etyedhatythsu (t vrddhir ekah prvaparayoh) ''Aprs a (finale du mot prcdent) et devant e de eti ou edhati ou devant th ( = de 6, 4,132), la vrddhi est le simple substitut des deux: prcdente et suivante". 20 QV ( = v ) est un suffixe pour lequel la rgle 6,1, 67 prescrit la substitution d'un lopa (ce que je reprsente par un zro), i. e. un suffixe qui n'apparat pas dans la ralit. En vertu de 1, 2, 45, aucune racine ne peut tre thme nominal, moins (cf. 1, 2, 46) qu'un suffixe primaire n'y soit ajout : d'o ncessit de suffixes imaginaires pour oprer, sans aucun changement rel, I e passaged h t u > p r t i pad ika.

Bien que le v qui constitue le suffixe lui-mme disparaisse compltement par 6,1, 67, l'effet du suffixe n'en subsisterait pas moins en vertu de 1, 1, 62 : le guna de l'/'k pnultime prescrit par 7, 3, 86 aurait lieu, n'tait la paribhs 50, comme si le v de l'rdhadhtuka nvi tait rellement prsent.
21

Le samprasrana et la substitution 6,1,108, qui en est insparable, peuvent tre considrs comme une seule opration. Cf. texte de la paribhs 119, o 6,1,108 est dsigne simplement par sa relation avec le samprasrana : " samprasranam
tadsroyarp ca kryam . . . " , " Le

sampr. et l'opration qui en dpend

22

II faut donc admettre ici avec Ng., sur l'autorit de Pataj., qui fait du mot th le japaka de l'antaragaparibhs (voir note 24), la possibilit de la causation mdiate (paramparay nimittatvam). Cf. (Kielh., p. 232-233) le

364 Pierre Boudon

l'une par r a p p o r t l'autre i n t r i e u r e et e x t r i e u r e . 7, 3, 86, d o n t la cause est i n t r i e u r e , est d i t e anta ranga; 6, 4,132, bahiraga.23 D'o la maxime: " Un bahiraga dj effectu est considr comme irrel lorsqu'un antaraga d o i t t r e effectu." 2 4 C'est le p r e m i e r aspect de I antarangaparibhs, caractris par l'antri o r i t du bahiraga. Cet aspect c o m p o r t e deux possibilits: 1 A exige l'application pralable de B. L'effet de celui-ci tant asiddha, c'est--dire v i r t u e l l e m e n t suspendu, A est inapplicable; 2 L'effet de B constitue un obstacle pour A ; B devenant asiddha, A devient applicable. 25 C'est la p r e m i r e de ces possibilits q u ' i l l u s t r e l a f o r m a t i o n de visvauhah. Ainsi s'explique la ncessit dans 6, 4,132 du mot th, qui p e r m e t la v r d d h i de 6 , 1 , 89, celle de 6 , 1 , 88 tant impossible, cause de l'inapplicable guna i n t e r mdiaire de 7, 3, 86. cas vrtrahan -j- kvip + bhym, o cette possibilit, faute de la mme autorit, n'est pas admise : l'lision de \'n final d'un pada (8, 2, 7) n'est pas considre comme bahiraga, bien qu'elle soit cause mdiatement par le suffixe bhym, lequel est cause que vrtrahan est un pada (voir Pan., I, 4,14).
23

PS., p. 4 1 , 1 . 1 2 : " nimittam

atrngasabdena eva grhyate sa Cf.

sabdarpam bdasstre

tasya pradhnatvt".

25 De cette seconde varit d'application du cas 1, on peut citer, dfaut d'exemple rel que je n'ai pas 24 rencontr, une illustration thorique : L'antarangaparibhsest introduite la formation de adhitya de adh -\- i -\par Pataj. dans le commentaire de 6, k t v > adhi + i+tv >adh + tvd(2,2, 4,132 en rponse trois vrt. qui 28 et 6,1,101) > adhl + ya (7,1, 37) > dclarent le mot th inutile, vu que adhtya (6,1, 71). Cette dernire son objet pouvait tre ralis par opration, insertion de l'augment tuk l'application successive de 7, 3, 86 et ( = t) prescrite aprs voyelle brve, ne 6,1, 88. La rplique est donne dans devrait pas avoir lieu dans le cas adh cette fin de dialogue : "evarp tarhi -f ya. Un moyen de rsoudre la diffisiddhe sati y ad vha tham ssti ta] cult serait de faire appel la parijnpayaty cryo bhavaty es paribhssiddham bahiragalaksanam antaraga- bhs 50, laquelle ressortit le conflit des rgles 6,1,101 et 6 , 1 , 71 dans le laksana i ti. Kim etasya japane prayojanam pacvedam pacmedaml cas asiddhatvd bahirangalaksanasyd adhi + / gunasyntarangalaksanam aitvam na A bhavatiti", "Eh bien (puisque l'objet 6,1,101 6,1,71 du mot th serait ralis ainsi) lefaitque Pnini prescrit th (comme sampra6,1,101, qui est B parce qu'elle porte srana spcial) du thme vh indique sur deux mots (v. infra, p. 366, n. 30) l'existence de la maxime "asiddham tant considre comme irrelle, le

Kielh, p. 223 : "The word aga in this (Paribhs) denotes only a formal cause (such as a letter or a combination of letters), because in a work which teaches the formation of words main importance attaches to the wordform."

raga i. e. de lasubstitution de guna prescrite aprs a (6,1, 87), l'antaraga, i. e. la substitution de ai (3, 4, 93) n'a pas lieu." Par 3, 4, 93, ai est substitu e faisant partie des suffixes de la premire personne l'impratif. La diphtongue e qui rsulte du sandhi a -f- /', quivalant en vertu de 6 , 1 , 85 la premire des deux voyelles, c'est-dire l'a final de pacva, pacma, appartient l'impratif et de ce fait tomberait sous 3, 4, 93. Par l'antarangaparibhs cette diphtongue devient asiddha, et par suite 3, 4, 93 inapplicable.

.. . ." Quel est le cas (par exemple) qui peut motiver cette indication? pacvedam (de pacva -f dam), pacmedam (pacma + dam) : tant donn l'irralit (conventionnelle) du bahi-

compos adhi serait cens revenir l'tat adhi -\- i et l'addition du tuk serait normale. En fait Pnini a fait appel une autre solution : la mention du mot tuk dans 6,1, 86 " satva-tukor

365 Raisonnement par l'absurde dans

La paribhas 50 a un second aspect, suggr par un jnpaka particulier, le monosyllabe an de 6 , 1 , 95, mot indispensable pour arriver une forme comme khatvodh. O r il semble t o u t d'abord que Pan ni aurait pu faire l'conomie de ce monosyllabe, vu les deux possibilits thoriques que voici :
khatv + a -j- dh khatv a + dh 6,1,101 26 khatv o dh 6,1,87 27 khatv + + dh khatv-\- o dh 6 , 1 , 87 khatv o dh 6,1,95 2 8

II s'agit ici de deux oprations (6,1,101 et 6 , 1 , 87) qui se prsentent simultanment. Par laquelle commencer? Si l'ordre 6, 1,101 6 , 1 , 87, tait permis, le second procd, c'est--dire le mot an de 6,1, 95 serait superflu. Le premier procd est donc illicite. Et l'explication est celle-ci: en face de 6,1, 87, 6,1,101 est asiddha, mais dans un sens diffrent. Dans le cas 1, l'une des rgles (B) tait dj effectue quand se prsentait l'autre. La suspension artificielle de B faisait que A pouvait ou non s'appliquer, selon que l'effet de B constituait pour A un obstacle ou une condition pralable. Ici les deux rgles se prsentent ensemble; aucune des deux n'exige l'application ou la suspension de l'autre. Le caractre asiddha du bahiranga consiste cder le pas l'antaraga, qui est effectu d'abord. La raison de l'osiddhatvam est donne par l'examen des conditions respectives des deux rgles dans leur application au cas

khatv + + dh 6,1,101 6,1,87

6.1. 87 est antaraga parce que ses causes ( + ) se prsentent d'elles-mmes en premier lieu ; 6,1 ; 101, bahiranga, parce que se causes (khatv + -dh) se prsentent en second lieu (voir Kielh., p. 243, n. 2). L'exemple khatvodh illustre une nouvelle maxime: " Un bahiranga qui est applicable en mme temps qu'un antaraga est asiddha en face de celui-ci". C'est l'aspect 2 de la paribhas 50: A est effectu en premier lieu ; aprs quoi, si les causes de B subsistent, B est effectu, sinon non.

asiddhah (ekdesah)", " Un ekadesa (i.e. un substitut forme simple d'un original constitu de deux lments, cf. 6 , 1 , 84) est asiddhaquand il s'agit de substituer s ou d'ajouter le tuk". Par l, dit Nag. (PS., p. 53,1. 4) le mot tuk de 6 , 1 , 86 indique l'inconstance de l'antarangaparibhs. Si, dans lecasadhtya, l'antaragapar nes'applique pas pour la seule raison qu'elle est anitya, elle peut s'appliquer ailleurs dans un cas analogue; autrement dit cette possibilit o A devient applicable du fait de la suspension de B, rentre dans le mcanisme de la paribhs. 26 6,1,101 : akah savarne drghah {ekah

prvaparayoh) "Aprs une ak (a, i. u, r, /) et devant homophone, la longue est le substitut des deux : prcdente et suivante". 27 6 , 1 , 87 : d gunah {ekah prvaparayoh)" Aprs a (finale du mot prcdent et devant voyelle autre que a, initiale du mot suivant), le guna est le substitut de la prcdente et de la suivante". 28 6 , 1 , 95 : om-nos ca {t pararpam ekah prvaparayoh) "Aprs a et devant om et an (i. e, la diphtongue qui rsulte du sandhi : prfixe -fvoyelle initiale d'une forme verbale), le simple substitut des prcdente et suivante est constitu par la suivante"

366 Pierre Boudon

L'antaraga 6 , 1 , 87 effectu, on a khatva + odha. Ceci, par 6, 1 , 88, donnerait* khatvaudh, d'o ncessit de la mention spciale " a n " dans 6 , 1 , 95, qui prvoit le s u b s t i t u a forme de la l e t t r e suivante (pararupam) pour les cas du type khatv + odh. Le monosyllable on est ainsi le jnpaka de la paribhs 50, cas 2. Les deux aspects de cette paribhs o n t t runis sous une seule f o r m u l e : " asiddham bahiranga m anta range.29"

Il L'Argumentation Base sur le Jnapaka et sa Valeur Logique

La rduction l'absurde n'est pas autre chose qu'un syllogisme conditionnel du type tollendo-tollens: la destruction du consquent dans la mineure entrane comme conclusion la destruction de l'antcdent. Ce raisonnement pour tre rigoureux suppose une condition essentielle: il faut que la mineure qui nonce le postulat pralable, pivot du raisonnement, soit vraie. Que vaut ce postulat selon lequel il n'y a pas de mot vyartha dans l'Astdhyyll "Vyartha," entendu au sens strict qui l'oppose caritrtha, qualifie une nonciation ou un procd de Pnini, qui en l'absence de telle paribhs, ne serait pas indispensable, ou but alors le mot adya est-il employ? La paraphrase de Ngoj. (PS., p. 43, Du mot adya son tour il y a union 1.15) distingue nettement les deux avec l'aggrgat (odha)." cas : "antarage kartavyejtam tatk30 Bien qu'il ne soit pas explicitement laprptikam ca bohirogam asiddham enseign par Patanjali que ce qui est ity arthah". Chaque cas comportant padadvaysraya est bah i ranga (cf. deux possibilits d'application, on a Kielh., p. 269, n. 3), cela ressort de les quatre schemes suivants: passages comme celui qui prcde. cas 1 : B, non A C'est en tout cas admis par Kaiyata B,A dans son commentaire 7, 2, 98, procas 2 : A, B pos du compos gomatpriyah de gomat A,non B + su -\-priya : " gomatpriya iti atrntaLe cas 2 de la paribhs 50 est rang ekapadsrayatvt sulopdayah suggr clairement, mais sans indicasyuh padadvaysrayatvd bahirago lug tion de jnapaka, dans le Bhsya 6,1, 108, propos de la formation adya -f itigomnpriyaitisyt...", "(Si l'antarangaparibhs s'appliquait) ici, auraient + dh : "Kim punar ihntarangam kim bahiragam yvat dve pade s r ity alieu le lopa du suffixe su (6,1, 88) et savarnadirghatvam api bhavaty dguno autres oprations qui sont antaraga 'pi. Dhtpasargayor yat kryam tad parce qu'elles dpendent d'un seul antaragam kuta tat prva upa- mot, le luk (du mme suffixe par 2, 4, 71) tant bahiraga parce qu'il dpend sargasya hi dhtun yogo bhavati ndyade deux mots [au moins indirectesabdena kirn artham tarhy adyasabdah prayujyate athdyasadb asyapi ment, vu que 2, 4, 71 prescrit le luk tu samudyena yogo bhavati". " Mais du suffixe casuel d'un mot qui devient prtipadika, c'est--dire qui entre en quel est donc ici l'antaraga, quel, le composition], et on obtiendrait bah ranga, puisque, l'une et l'autre *gomnpriyah". Si le suffixe disportant sur deux mots (et donc tant paraissait par lopa, toutes les oprathoriquement B30 l'une et l'autre), tions lies au suffixe (7,1, 70, etc.) et la substitution de longue homoauraient lieu, en vertu de 1,1, 62, phone s'applique, et la substitution de mme aprs disparition du suffixe. guna prescrite aprs al L'opration Mais la paribhs 52 enseigne qu'une qui porte sur la racine et son prfixe, rgle de luk l'emporte mme sur un celle-l est l'antaraga. Comment antaraga. Du fait de sa disparition cela? L'opration premire en effet par luk, l'effet du suffixe disparat avec c'est l'union du prfixe avec la racine, ui (1,1,63). non avec le mot adya. Dans quel
29

367 Raisonnement par l'absurde dans Pnini

serait inexplicable, quand sa valeur grammaticale. Et de fait (sinon en thorie) " v y a r t h a " n'quivaut pas toujours " inutile absolument" : on peut citer, d'aprs le tmoignage indirect, sinon explicite, de Patajali ou de Ngoj, eux-mmes trois exemples de mots grammaticalement vyartha lafois et de par ailleurs utiles. C'est d'abord le prtendu jnpaka de la paribhs 56 " akrtavyhh PninJyh,"3* le mot " samarthnm" de la rgle 4 , 1 , 82

" samarthnm prathamd v," dont la traduction ordinaire est celle-ci : " (Les suffixes prescrits par les rgles qui suivent s'ajoutent) aprs le premier des mots en construction (d'une locution de l'un des types 4 , 1 , 92, 4, 2 , 1 , etc., la locution n'tant pas exclue par le driv, vu que le suffixe n'est prescrit que) facultativement." Le stra ainsi compris est un adhikra (rgle gouvernante) dont l'influence s'tend jusqu' la fin de5,2, i.e. sur toutes les rgles des suffixes taddhita. Mais il ne gouverne celles-ci que par l'intermdiaire d'un certain nombre d'adhikra subordonns, dont chacun indique le sens d'une srie de suffixes. Ainsi 4 , 1 , 92 "tasypatyam" gouverne les rgles des suffixes patronymiques (4,1, 83 et suiv.). Les " mots en construction" de l'adhikra gnral sont les nonciations tasypatyam (4,1, 92), tena raktam rgt (4, 2, 1), tatra bhvah (4, 3, 53), etc. des adhikra infrieurs. En consquence, chaque rgle de suffixe taddhita doit tre interprte travers les deux adhikhrasuperposs. Ainsi 4 , 1 , 95 "ata i " signifie: " Le taddhita i s'ajoute facultativement aprs un thme finissant par a, quand ce thme correspond au premier des mots en construction d'une expression du sens de " tasypatyam." De stthita, par exemple, on forme, au sens de " stthitasypatyam,"

un driv stthita + i = sautthitih.

Mais c'est un autre sens que Patajali attribue l'adjectif samartha. Il dsignerait selon lui les mots "capables d'exprimer un sens," c'est--dire les mots sous leur forme dfinitive, tels qu'ils sont aprs sandhi de leurs parties constituantes.32 Ainsi stthita et vksamna sont dits samartha par opposition su-utthita et vi-ksamna, formes asamartha. L'adhikra 4,1, 82 signifierait donc que le taddhita i, par exemple, doit s'ajouter stthita et non su-utthita. C'est sur cette interprtation de samartha comme synonyme de krtasandhi que repose le jnpaka "samarthnm." A supposer en effet que pour arriver sautthitih on parte d'un asamartha su-utthita, auquel cas deux oprations, 6,1,101 33

" Les Paninens ne soutiennent pas (lathse : persistance de l'effet d'une rgle quand les causes de celle-ci ont disparu) " . 32 Bhsya, 4 , 1 , 8 2 : . . . kirn punah samartham arthbhidne yat samartham kirp punas tat krtavarnnuprvikampadam... Kaiyataexplique: knavarnnuprvikam tiltasyoiva oke 'rthapratipdanyaprayogtsamarthatvam iti bhvah. " . . . parce qu'on n'emploie dans le langage ordinaire en vue d'exprimer un sens que le mot de cette espce (le mot qui est krtavarna), ce mot a qualit de samartha."

31

krtavarna = krtasandhi, voir Nagoj. PS., p. 64,I. 2. II faut supposer, si l'on veut que samarthnm se rapporte prathamd, que la seconde acception n'est pas exclusive mais ne fait que se superposer la premire : " Les taddhita s'ajoutent aprs le premier des mots en construction d'une expression analytique . . .tant entendu d'autre part qu'il ne peut s'agir que de mots "capables de sens" i. e. de mots forme dfinitive. 33 6,1,101 : akah savarne drghah V. supra, p. 365, n. 26.

368 Pierre Boudon

antaraga et 7, 2,117 34 bahiraga, sont simultanment applicables, 35 la rgle 6,1,101 est, en v e r t u de la paribhs 50 (cas 2), effectue en p r e m i e r lieu et le taddhita i ne peut que s ' a j o u t e r a stthita. A u t r e m e n t d i t , si l'antarangaparibhs s'applique, samarthnm est superflu. O n devrait conclure que la paribhs ne s'applique pas. C'est ce que f o n t certains grammairiens. Constatant en o u t r e que le conflit des deux rgles en question a ceci de particulier que les causes de l'antaraga 6,1,101 ( savoir u + u) disparatraient par l'application de bahiraga, si celui-ci s'appliquait en p r e m i e r lieu, ces grammairiens o n t voulu faire du mot samarthnm le jnpakad'une paribhs (56) qui prtend q u ' " u n antaraga voit son effet disparatre (ou n'a pas lieu), dans le cas o ses causes disparaissent u l t r i e u r e m e n t (ou mme par la simple considration qu'elles disparatraient) par l'application du bahiraga." 3 6 Mais paribhs et jnpaka sont rejets par Ngoj., qui dclare que l'antarangaparibhs s'applique rellement dans la f o r m a t i o n de sautthitih, et s'exprime ainsi sur le mot samarthnm : " Bien qu'il ne soit que la r p t i t i o n de ce que nous savions dj par l'antarangaparibhs, ce mot nous met en garde c o n t r e l ' e r r e u r qui consisterait c r o i r e qu'il en est pour tous les taddhita comme pour na, qui dans visuna, etc., est ajout une f o r m e (visu + ok) o n'a pas t fait lesandhi." 3 7 Par l Ngoj. reconnat I o que samarthnm est vyartha du point de vue s t r i c t e m e n t grammatical, puisqu'il

7, 2,117 : taddhitesv ocomadeh (niti vrddhih) " La vrddhi est substitue la premire voyelle du thme devant taddhita it ou nit". 35 II y a conflit exactement entre 6 , 1 , 101, d'une part et, d'autre part, 4 , 1 , 95 immdiatement suivie de 7, 2,117. (Cf. paribh. 50 o de mme l'opration B tait en ralit double : 6,4, 132 + 6,1,108.) Commence-t-onpar, 1,101, i s'ajoute stthita, d'o sautthitih; si l'on commenait par le suffixe, celui-ci s'ajouterait suutthita, d'o *svutthitih.
36 Telle est la double porte que donnent la paribh. 56 ses partisans. Voir PS., p. 61, 1.10 " bahirangenntaragasya nimittavinse posct sambhvite 'ntaragam no" et p. 62, I. 8 " etatpravrttau ca nimittavinsasambhvanpi nimittam". 37 PS., p64,l. 1.6 et suiv: "somorthnm iti strosthasamarthograhanom tu visuna ity dv akrtasandheh pratyayadarsanena sarvatra tathbhramavranya nyyasiddhrthnuvdaiva", (Sur visuna, voir vrt. 2 5, 2,100.) Selon Kaiyata, samarthnm est destin empcher la vrddhi qui s'appliquerait en vertu de la paribh. 55, d'aprs laquelle une rgle de l'angdhikra (ici, 7, 2,117) est plus forte qu'une rgle de sandhi (ici 6 , 1 ,

34

101). Ainsi si la Paribh. 55 s'applique, samarthnm empche la vrddhi. Si la paribh. 55, par inconstance, ne s'applique pas, c'est la paribh. 50 qui s'applique normalement, et le mot samarthnm a mme objet que la paribhs. Dans un cas comme dans l'autre ce mot n'est pas jnpaka de 56. Cf. Kielh., p. 316, n.1. La paribh. 56 est mentionne dans Kaiyata mais non dans Patajali. Kaiy. (6. 4, 22) signale mme qu'elle n'est pas admise par Patajali. Certains invoquent la paribh. 56 pour expliquer la forme papusah d'o l'augment it (prfix au suff. vas par 7, 2, 35, antaraga) disparatrait lorsque sa cause, l'initiale v du suff. vas, a disparu par le samprasrana de 6, 4,131, bahiraga. Kaiy. rpond ainsi : "papusa itijnanu cntarangatvd it bhvyam. Ina ca samprasrane krte 'pito nivrttir nimittpye naimittikasypy apya ity asyh poribhsy bhsyakrennsrayant. . ." " L'augment it n'est-il pas possible en qualit d'antaranga? Et il n'est pas vrai que mme aprs le samprasrana (opration B qui dtruit ci lacausede A, l'effetdeA, i.e.) l'augment it disparaisse, car la maxime " disparition de la cause, disparition de l'effet" n'est pas admise par l'auteur du Bhsya . . . "

369 Raisonnement par l'absurde dans Pnini

est une rptition (anuvada); 2 que cette rptition a nanmoins une Utilit. Tel est encore le cas du soi-disant jnpaka de la paribhs 9310, le mot sthne de 7, 3, 46,38 exprim aprs toh, alors qu'il est sousentendu, conformment 1 , 1 , 49,39 aprs atah. atah dsigne l'objet de la rgle: a original du substitut prescrit; toh dsigne ce qui qualifie l'objet de la rgle: original de l'ex-substitut devenu original du substitut prescrit. L'emploi par Pnini du mot sthne aprs atah indiquerait, au dire de certains, que sthne, c'est--dire le paribhsstra 1 , 1 . 49, ne pouvait pas tre suppl ici, et par suite qu'une paribhs se sous-entend avec ce qui est enseign dans une rgle, non avec ce qui nefait que donner un renseignement touchant l'objet de la rgle. 40 La paribhs est rejete par Ngoj., qui dclare q u e l , 1, 49 ne signifiant pas que sthne doive tre suppl avec tous les gnitifs, la mention expresse de sthne dans une rgle n'indique rien. 41 Si le mot est exprim ici, c'est seulement dans un but de clart: "tatra sthnegrahanam tu spastrthameva" (PS., p. 97,1.2). Un troisime exemple est le mot ubhayesm de 6,1,17. 4 2 Ce mot aurait pu tre omis, car il dsigne deux catgories de verbes nonces dans 6,1,15 et 16, dont la validit dans 6 , 1 , 1 7 allait de soi. Le mot tant vyartha, on'a voulu en faire le jnpaka d'une paribh. (119), selon laquelle " le samprasrana et l'opration qui en dpend (== 6,1,108) ont plus de force que les oprations concurrentes," 4 3 paribhs qui expliquerait laformation du parfait des deux groupes de verbes en question. Lorsqu'en effet a t opr le redoublement de vyodh par exemple, deux rgles sont applicables l'tat vya + vyadh -f- a: 6 , 1 , 1 7 (samprasrana) et 7, 4, 60 (haldisesah), qui enseigne que des consonnes de la redouble seule la premire est retenue. En v e r t u de 1 , 4, 2, 7, 4, 60 devrait s'appliquer d'abord puisque postrieure (para), et le y de la syllabe redouble tant ainsi lid, le samprasrana serait alors substitu au v de la mme syllabe, ce qui d o n n e r a i t * uvydha. C'est le contraire qui a lieu et cela s'expliquerait par la paribhs 119. Mais le mot ubhayesm est un jnpaka sans valeur car il ne devient pas caritrtha aprs adoption de la paribhs. Celle-ci est rejete par Ngoj., qui dclare, d'aprs le Bhsya, que le mot
38 1 , 1 , 49 : sasthi sthneyog " La sixime (vibhakti, . e. le gnitif), implique (en grammaire) la relation dnote par le mot "sthne", " la place de". 39 7, 3, 46 : udicm tah sthne yakaprvyh (ata id na) "/ n'est pas substitu la place de a qui est substitu (par 7, 4,13) la place d'un d (du fminin) prcd de y ou k. (Interdiction conforme la doctrine) des Septentrionaux (et par suite facultative"). Exemple : ksatriyakou ksatriyik. 7, 3, 46 est une exception 7, 3,44, d'aprs laquelle/'est le substitut de a qui prcde le k appartenant un suffixe, quand suit la terminaison fminine , except la fin d'un bahuvrhi. Exemple : karaka-\- = krik. 40 Paribh. 9310 : vidhau paribhsopatisthate nnuvde. 41 Si 1,1, 49 n'enseigne pas que sthne doive tre sous-entendu avec tous les gnitifs, sthne peut ne pas tre sousentendu (peut tre exprim) sans que 1,1, 49 cesse de s'appliquer. La mention expresse de sthne aprs tah ne peut donc pas signifier la non-application de 1,1, 49, ni par suite indiquer la paribh. 9310. 42 6,1,17 : lity abhysasyobhayesm samprasranam) "sampras, de la syllabe du redoublement des deux (groupes de verbes : vacydi (6,1,15) et grahdi (6,1,16)), quand les suif, de lit suivent" (i. e. au parfait), 43 Paribh. 119 : samprasranam tadsrayam ca kryam balavat.

370 Pierre Boudon

ubhayesm a le mme objet qu'aurait la paribhasa, vu qu'il a le sens de eva et indique que des deux rgles applicables au redoublement des verbes en question, seul le samprasrana d o i t avoir lieu. 44 A ces mots auxquels est assigne par Patajali ou Ngoj. une valeur conventionnelle, 4 5 on peut ajouter une autre catgorie, certainement aussi f o r t r e s t r e i n t e , de mots techniquement vyartha ceux que Pnini peut avoir employs dans une i n t e n t i o n pdagogique. Si cette i n t e n t i o n concidait en gnral avec la recherche de la concision, elle pouvait exiger de temps en temps qu'il y f t fait drogation. Ainsi peut-on essayer de justifier deux ou t r o i s lments d o n t cause de leur i n u t i l i t p r o p r e m e n t grammaticale on a voulu faire des jnpaka. Soit la paribhs 84 " samsntavidhir anityah."*6 Elle aurait pour jnpaka la prsence de m o t rjan dans le ganaamsVd/, qui i n t e r v i e n t dans 6, 2,193 : " prater amsvdayah tatpuruse (antodtth) " Dans un tatpurusa dont le premier terme est prati, les mots amsu, etc. ont l'aigu sur la finale." Or le mot rjan, quand il figure la fin d'un tatpurusa, prend le suffixe tac par 5, 4, 91, 47 et ce suffixe tant cit, un mot comme pratirja est dj antodtta par 6,1,163.48 On raisonne donc comme suit: si la rgle 5, 4, 91 tait d'une application constante, il n'y aurait pas d'autre compos pratiraja que celui qui est form au moyen du tac, et la mention de rjan dans les amsvdi serait superflue. Il s'ensuit que la rgle du tac est anitya, comme le sont, dit la maxime ainsi suggre, toutes les rgles des suffixes samsnta (de 5, 4, 68 la fin du pda). Il est bien vrai que l'objet direct (prescription de l'aigu sur finale) de la mention du mot rjan dans le gana amsvdi tait obtenu sans cela. Mais ce rappel est-il inutile? Il est probable que Pnini (si Pnini est l'auteur des gana)49 a insr dessein le mot rjan, de manire faire un gana complet de tous les mots accentus sur la finale dans les mmes conditions (i.e. dans un tatpurusa premier terme "prati"), y compris un mot qui, portant dans son suffixe samsnta l'indice de cette accentuation, aurait pu la rigueur tre omis. Si l'on admet cette explication, le mot rjan

C f . B h s y a , 6 , 1 , 1 7 : . . . dam tarhy ubhayesm grahanasya proyojanam / ubhayesm abhysasya samprasranam eva yath syd yad anyat prpnoti tan m bhd ti. kirn cnyat prpnoti haldisesah ... 45 II y a lieu de signaler aussi les quelques lments rejets par Patajali, ainsi que ceux qui, bien que rellement vyartha, ne sont pas discuts dans le Bhsya, et au sujet desquels Ngoj. (PS., p. 94, l.10etsuiv.)s'exprime ainsi : " bhsycritaprayojannm sautrksarnm pryndv adrstamtrrthakatvakalpany evaucityt". Ngoj. vient de dire que les paribh. 931"5 sont rejeter parce qu'elles ne figurent pas dans le Bhsya et que nous ne sommes pas autoriss considrer comme correcte une maxime, mme tablie par un jnpaka, si elle n'est pas donne dans le Bhsya.

44

II ajout : " e t parce que la seule manire convenable (d'expliquer) certaines lettres employes dans les stra et dont le motif n'a pas t discut dans le Bhsya, c'est de supposer qu'elles ont simplement pour but de procurer celui qui les tudie des mrites religieux". 46 " Une rgle de suffixe samsnta ( = suff. que prend un mot en fin de compos) est inconstante." 47 5,4,91 : rja-ahas-sakhibhyas tac (tatpur usasya) " Le suffixe tac s'ajoute aux thmes rjan, ahas, sakhi en fin de compos, quand le compos est un tatpurusa". 48 6,1,163 : citah (antodttah) " Les (mots termins par un suffixe) cit sont accentus sur la finale " . 49 Sur ce point, voir Goldstcker, Pnini, p. 131, n. 154. [see this volume 264],

371 Raisonnement par l'absurde dans Pnini

d u gana amsvdi ne peut pas suggrer la paribhsa 84, qui se rvle d'ailleurs c o m m e une fausse paribhsa. Patajali, qui termine le commentaire du stra, 2,197 par la formule "vibhs samsnto bhavati," o, l'on peut voir une premire forme de la paribhsa "samsntavidhir anityah," ne parle pas du prtendu jnpaka rjan, mais constate que Pnini a employ le compos pddanmrdhasu sans suffixe samsnta. La rgle 6, 2, 197 " dvitribhym pddanmrdhasu bahuvrihau (antodttah)" enseigne qu'aprs dvi et tri les mots pad, dot et mrdhan ont l'aigu sur la finale dans un bahuvrhi. Deux vrttika observent que si Pnini a voulu dsigner le thme mrdha, il faut ajouter mrdhan ; s'il a voulu dire mrdhan, il faut ajouter mrdha, puisque l'aigu se rencontre galement sur la finale de trimrdh et sur celle de trimrdhah. D'accord avec un troisime vrt. Patajali dclare50 que renonciation au moyen du thme mrdhan suffit englober les deux cas, car l'accent prescrit par 6, 2,197 pour la finale d'une forme trimrdh se retrouve, en vertu de 6,1,161, 51 aprs addition au thme mrdhan du samsnta sa (5, 4,115),52 sur la finale de trimrdhah. Suit une discussion sur le conflit des rgles 6, 2,197 et 5, 4,115. Accentue-t-on la finale du thme avant d'ajouter le samsnta, ou ajoute-t-on d'abord le samsnta, qui se trouve ainsi recevoir l'accent ? Les deux rgles tant nitya,53 6, 2,197 l'emporte paratvt (Pn. 1, 4, 2), et l'on a la succession : 6, 2,197 (accent sur la finale de mrdhan), 5, 4,115 (samsnta), 6,4,144 (elision du ti, i.e. de la partie an de mrdhan), 6,1,161 (accent sur le samsnta). La conclusion est donne de la faon suivante:54
"Mais convient-il de discuter cela? [la question de savoir si 6, 2, 197 doit tre effectue avant ou aprs le samsnta, car mme aprs addition du samsnta, le second terme du compos n'en est pas moins mrdhan. O r , 6, 2, 197 concerne le second terme, non le compos.] N'est-il pas vrai que Pnini a employ de faon non quivoque le thme finale n puisqu'il a dit mrdhasu7. S'il avait employ le thme finale o, il aurait dit mrdhesu. C'est le suffixe samsnta qu'a pour objet (en dfinitive) cette discussion; car (en agissant) ainsi (qu'il le fait, c'est--dire en disant n\rnasu)y Pnini nous indique que le samsnta est facultatif."

Patajali parle-t-il du seul samsnta sa, ou du samsnta en gnral ? Il est probable qu'il s'est servi dessein d'uneformule gnrale qui justifiait les drogations dj courantes de son temps aux rgles des samsnta, tel le fminin supath, form sans le kap
Ce qui suit n'est pas la traduction littrale, mais seulement un abrg du Bhsya. 51 6,1,161 : anudttosyo ca yatrodttalopah (udttah) " Reoit aussi l'udtta une voyelle non accentue qui dtermine l'lision de la voyelle accentue prcdente " . 52 5, 4,115 : dvitribhym sa mrdhnah (bahuvhau) " Le suffixe sa (= a) vient, dans un bahuvrhi, aprs le thme mrdhan venant aprs un premier terme dvi ou tri". 53 II s'agit de la notion technique de . nitya, dfinie par la paribh. 42 : krt50

krtaprasagi nityam tadviparitam anityam "Une rgle qui s'appliquerait aussi bien avant (aprs) qu'elle s'appi que aprs (avant) I'effectuation d'une rgle concurrente est nitya; daus le cas contraire, anitya". Cf. Kielh., p. 209, n. 1. 54 Yuktam punar dorn vicrayitum jnanv anensamdigdhena nakrntagr aha nena bhavitavyam yvot mrdhasv ity ucyote yady akrntagrahanam syn mrdhesv iti bryt sais samsntrth vicran \ evam tarhi jnpayaty cryo vibhs samsnto bhavatti.

372 Pierre Boudon

de 5, 4,152, et qu'on rencontre Bhas., 7 , 1 , 1 7 (supath nagar). Quoi qu'il en soit, le compos en mrdhasu employ par Pnini n'implique rigoureusement que la non-application du samsnta sa. Quant la prsence de rjan dans le gana amsvdi, elle n'a pour Patajali aucune valeur d'indice puisqu'il ne la mentionne pas; ce n'est probablement qu'un pseudo-japaka invent par ceux qui dans la suite o n t donn sa forme dfinitive la fausse paribhs. 55 Un autre souci qu'a pu avoir Pnini dans la rdaction de ses sQtra, c'est celui de simplifier l'explication des formes rares ou difficiles. Si dans son systme l'conomie des rgles a plus d ' i m portance que l'conomie des mots (voir paribh. 121 : " padagauravdyogavibhgogarlyn"), il a d juger, du point de vue du lecteur, comme galement entache de gaurava la procdure qui obligerait recourir deux rgles pour expliquer une forme, alors qu'une seule moyennant une lgre addition suffirait. De cette manire peut-on essayer d'expliquer les deux prtendus jnpaka de la paribh. 92. dnonciation ja, au moyen de la longue, dans 7, 3, 79,56 permet d'expliquer les formes jnti,jyate, etc., par cette seule rgle; avec renonciation ja, ces formes exigeaient l'application successive de 7, 3, 79 et 7, 3,101. 5 7 De mme, grce au monosyllabe t, la rgle 6, 4,160 "jydd Jyasah"58 suffit pratiquement (lafusion des deux homophones allant de soi) rendre compte dejyyn, qui ne s'expliquerait autrement que par l'application de 6, 4,160 (lue il jyd Jyasah11 a v e c a n u v r t t i de " / o p a h " ex 6, 4,158) et de 7, 4, 2559 (allongement de a dejya). 6 0
L'inauthenticit de la paribh. 84, dj signale par Goldstcker (Pnini, p. 113, n. 131), se trouve confirme par l'examen des justifications auxquelles lafait servir l'auteur de la Durghatovrtti, livre de casuistique grammaticale, o elle n'est pas invoque moins de sept fois. Voir Durghatovrtti ofSaranadeva, edited with notes by T. Ganopoti SostrJ, Trivandrum, 1909 : 2,2, 30 : apetamaithilm, solcisme contre 2, 2, 30 (uposarjanom prvam), si tatpurusa selon 2 , 1 , 38, est interprt comme un bahuvrhi qui n'a pas pris lesuff. /cap (5, 4,153). 4 , 1 , 44 : ananyogurvys ex ononyagurvJ (m s licite par 4 , 1 , 44) est interprt comme un bahuvrhi sans lekop. 5, 4, 78 : brohmovorcosoh, ablatif au lieu de varcasdt, forme normale avec le samsnta oc de 5, 4, 78. L'explication par l'inconstance est ici d'autant plus inadmissible qu'il s'agit d'une rgle faite expressment pour le mot vareos. 5, 4, 92 : gmivartamnhar est interprt comme un visesanasamsa sans le toc de 5, 4, 92. 5, 4,132 : dhrtadhanusom est
55

interprt comme un bahuvrhi sans le suffixe ona de 5, 4,132. 5, 4,153 : varatonu, et 7, 3,108 : sutanu, solcismes contre 7, 3,108 si thmes en u, deviennent rguliers par 7, 3,107, parce qu'interprts comme des vocatifs de bahuvrhi termins par des thmes nadi (voir 1, 4, 3) qui n'ont pas pris le suffixe kap (5, 4,153). 7, 3, 79: jnjanorj (siti) "ja est substitu kjn et jan devant un suffixe sit". 57 7, 3,101 : oto dirgho yai (srvodhtuke) " o est substitu o devant un srvadhtuka commenant par une lettre du pratyhra yon ( = demivoyelles -f nasales -\-jh et bh). 58 6, 4 , 1 6 0 : jyd ad iyosoh ' ' aprs jyo, est substitu (la premire lettre de) iyas". Cf. 1,1,54 59 7, 4, 25 : akrtsrvodhtukoyor dirghoh (yi kiti) " Une longue est substitue la voyelle finale du thme devant suffixe commenant par y, qui est kit ou it, mais non devant un krt ou un srvadhtuka". 60 Une forme un peu complique est le plus souvent le rsultat de plusieurs rgles, celles-ci tant des rgles d'une certaine gnralit. Mais I convient
56

373 Raisonnement par l'absurde dans Pnini

Ce sont l explications aussi satisfaisantes que celles qui font du monosyllabe ot et de renonciation jo les jnpaka d'une paribhs " agavrtte punarvrttv avidhih," selon laquelle de deux rgles successives de l'angdhikra 61 qui portent sur le mme mot seule la premire peut avoir lieu. Patajali (Bhs., 7 , 1 , 30) qui formule cette paribhs avec l'addition du m o t " nisthitasya," indique par l qu'elle ne rend compte que des formes qui sont correctement tablies sans l'intervention d'une seconde rgle de l'angdhikra, 62 celles o l'on constate l'application de deux rgles successives s'expliquant par l'inconstance de la paribhs: ainsi " dvayoh" obtenu par 7, 2,102 et 7, 3,104 et deux fois employ dans les stra (1, 2, 59 et 5, 3, 92), ce qui semble bien indiquer que la paribhs 92 tait inconnue de Pnini. 63 Il n'y a qu'un petit nombre de cas o il soit possible de donner d'un mot d'apparence inutile une explication extra-grammaticale. C'est dire que les quelques mots rellement superflus, du point de vue technique de lagrammaire, qu'on peut rencontrer dans Pnini ne peuvent pas infirmer pratiquement le postulat de l'conomie paninenne, et le raisonnement qui repose sur ce postulat
de considrer part le cas des formes rares ou d'exception : si Pnini tait amen faire une rgle en vue du seul comparatif jyyn, ou en vue des temps sit des deux racines a et jan, il devait tre tent de lafaire unique, d'y inclure, si possible, tout ce qui concernait la formation de ce comparatif ou de ces temps exceptionnels. L'adhikra "agasya" (6,4,1) gouverne le quatrime pda du livre 6 et les quatre pda du livre 7. 62 En fait le Bhsya ne cite que deux cas o la paribhs 92 pourrait s'appliquer (et dans aucun des deux elle n'est indispensable); elle expliquerait : 1 La non-application de 7, 3, 86 aprs 7, 3, 78 dans pibati, ce qui rendrait inutile le vart. 1 "pibater gunapratisedhah" 7, 3, 78. (Bhs., 6, 4,160 et 7, 3,79). Lorsque de p + sop -{-tip on a obtenu, par 7, 3, 78, pib + a + ti, le guna devrait tre substitu la pnultime / en vertu de 7, 3, 86; celle-ci tant une seconde rgle de l'angdhikra, on peut dire qu'elle est exclue par la paribh. 92. Mais une autre explication est possible : le substitut prescrit par 7, 3, 78 est piba (au lieu de pib), auquel cas 7, 3, 86 est inapplicable et la paribh. 92 inutile. 2 La non-application de 7, 3,103 aprs 7, 2, 90 dans la formation de yusmabhyam (Bhs., 7,1, 30). Cette forme peut tre ainsi obtenue : yusmad + bhyas > yusmad -f bhyam (7, 3, 30 lue " bhyaso bhyam")^> yusma -{-bhyam
61

(7, 2, 90). Ici par 7, 3,103, e devrait tre substitu Va de yusma devant bhyam. 7, 3,103 tant une seconde rgle de l'angdhikra, on peut recourir l'explication de la paribh. 92. Autre explication : 7,1, 30 est lue "bhyaso'bhyam". Laformeenquestion est alors ainsi obtenue : yusmad + bhyas > yusmad -\-abhyam (7,1, 30) > yusma + abhyam (7, 2, 90) > yusmabhyam (6,1, 97), et la paribh. 92 est inutile. 63 D'aprs certains (ke cit), dit Ngoj., il n'y aucune formeenvuede laquelle cette paribhs soit indispensable (. . . anay paribhasayd na kimcil laksyam sddhyate) et les assertions contenues dans le Bhs. 6, 4,160, 7, 3, 79 et 7,1, 30 doivent tre considres comme paroles de celui qui n'est que partiellement instruit de la doctrine (ekadesyuktir). Kielh., p. 440, n. 1, signale : d'aprs Pyagunda, "kecit" indique que Ngoj. n'approuve pas cette thse; d'aprs d'autres, "kecit" signifierait " bhsyatattvavidah". D'o le point d'interrogation dont Kielh. fait suivre le mot " rejected " dans sa liste des jnpakasiddhaparibhs (prface, p. x). "kecit" ne peut pas dsigner Ngoj., vu que l'opinion de celui-ci est toujours introduite par "pare tu". Nanmoins, comme Ngoj. expose la thse du rejet sans lafaire suivre de rfutation on peut supposer qu'il la considre comme vraisemblable.

374 Pierre Boudon

est gnralement valable. Mais ce raisonnement ne permet qu'une conclusion limite. De la maxime gnrale invoque, la constatation faite dans la rgle du jnpaka ne constitue qu'un cas particulier. C o m m e n t s'opre le passage du cas particulier la maxime ? Y a-t-il induction ? Et d'abord que vaudrait cette induction ? Parce que le thme "prva" de 5, 2, 86 ne dnote pas prvnta (supra, p. 361 ), s'ensuit-il que t o u t thme nominal chappe la maxime " vyapadesivadekasmin" ?

L'induction n'est possible, disent les logiciens, que s'il y a enumeration suffisante de donnes singulires, l'induction partir d'un cas unique n'tant valable que si l'attribut du jugement singulier est essentiel au sujet. Or dans les propositions de la grammaire il n'y a pas d'attribut essentiel, ni par suite d'induction valable partir d'un cas singulier.64 La maxime 32, si elle n'avait pas d'autre fondement que son jnpaka n'aurait qu'une valeur de probabilit: on souponnerait qu'il en est pour tous les thmes nominaux comme pour le thme "prva," faute de voir pourquoi Panini aurait agi de telle manire dans tel cas et non ailleurs. La valeur de probabilit des maximes formule universelle ne serait encore acceptable comme telle qu' condition de s'appuyer strictement sur la constatation premire. Condition en fait ralise le plus souvent. C'est la comprhension de la rgle du jnpaka qui dtermine celle de la paribhs. Par exemple, le mot karma, driv de karman par le suffixe na tacchlika (4, 4, 62) est trait par Pan ini dans 6, 4,172 comme un driv par an\ maxime correspondante (paribh. 87) : les oprations dtermines par le suffixe an le sont aussi par le na tcchJlika.65 Mais il est des paribhs qui n'auraient mme pas valeur de vraisemblance; ce sont celles dont laformule ne procde pas exactement des indications du jnpaka, soit qu'elle les dpasse (paribh. 66), soit qu'elle reste en de (paribh. 106). La paribhs 66 a pour jnpaka le mot "akitah" de 7, 4, 83 " drgho 'kitah" : " Une longue est substitue (la voyelle finale du) redoublement l'intensif moyen et actif (littralement: devant yon [ = yat suffixe de I1 intensif] et devant uk de yan), quand ce redoublement ne prend pas d'augment kit." "Akitah" a pour but
Je mets part un cas comme celui de la paribh. 27, qui nonce un principe allant de soi : samjdvidhau pratyayagrahane tadantagrahanam nsti" "Dans une rgle de dfinition, a mention d'un suffixe n'quivaut pas (contrairement 1 , 1 , 72) la mention de ce qui finit par ce suffixe". Un suffixe pour tre dfini comme tel, doit ncessairement tre distingu de ce qui finit par le suffixe. Dans ce cas l'esprit passe sans hsiter, et mme sans faire d'induction, de la constatation singulire faite dans la rglejnpaka (1, 4,14 : suptiantam padam) la maxime gnrale.
65 87 : tacchllike ne 'nkrtni bhavanti. Cette paribhsa pour jnpaka le mot karma de 6,4,172 "krmas tcchJlye" " (Existe le mot) karma dans le cas d'habitude caractristique " (c'est-64

dire pour dsigner celui qui a l'action [karman] pour habitude, un homme d'action). 6, 4,144enseigne : "d'un thme bha(v. Pan. 1, 4,18) le ti (partie commenant avec la dernire voyelle) est lid devant taddhita". 6, 4,167 enseigne : "d'un thme finissant par an, la syllabe an demeure inchange devant le suffixe an". Pnini, par la rgle 6, 4,172, veut fixer la forme karma (de karman -+- na tcchilika); or l'lision du tide karman serait rguWre par 6, 4,144. Si 6, 4,172 a un sens, autrement dit si le driv karma mrite une mention spciale, c'est qu'il est comme rrgulier par rapport 6, 4,167 : par consquent les oprations dtermines par le suffixe an le sont aussi par le na tcchilika.

375 Raisonnement par l'absurde dans Pnini

de sparer du domaine de 7, 4, 8366 le domaine des rgles 7, 4, 84, 85, etc., qui enseignent l'addition d'augments anubandha k (nik, nuk, etc.) au r e d o u b l e m e n t de l'intensif de certains groupes de verbes, toutes rgles qui seraient des apavda,67 par r a p p o r t un utsarga7, 4, 83 f o r m u l " drghah." Soit 7, 4, 85,68 qui prescrit I1 augment nuk (== n) aprs l'a bref d'un r e d o u b l e m e n t l'intensif moyen et actif, quand la racine finit en nasale. Si, dans la f o r m a t i o n de l'intensif de yam, par exemple, la prdominance du suppos apavda (qui serait ici un anavaksa) s'exerait normalement, c'est--dire si 7, 4, 85 s'appliquait d'abord, la f o r m e yamyamyate une fois pourvue du nuk n'aurait plus un redoublement en voyelle finale et ne c o u r r a i t aucun risque de t o m b e r sous l'utsarga "drghah." A u t r e m e n t d i t , " a k i t a h " serait superflu. Conclusion inadmissible. Ce qui oblige constater que dans le cas des deux rgles 7, 4, 83 (lue "drghah") et 7, 4, 85, qui o n t t r a i t l'une et l'autre au r e d o u b l e m e n t intensif, la prdominance normale et t renverse : l'utsarga aurait prcd (et par l rendu impossible) l'apavda. D'o, ncessit dans 7, 4, 83 de "akitah" pour empcher cela.69
66 7, 4, 83 : dlrgho 'kitah (abhyasasya yoniukoh [ = yai yanluki co]). 67 Un apavda ou rgle particulire (par opposition utsarga, rgle gnrale) est dfini par la maxime "yena nprpte yo vidhir robhyate sa tosya bdhakobhovati". "Alors qu'une rgle (gnrale) s'appliquerait ncessairement ( un mot), une rgle (particulire) faite (en vue du mme mot) exclut la rgle (gnrale) " . Il y a deux varits d'apavda :

1 L'anavaksa (cf. paribh. 58), rgle particulire "dpourvue d'occasion ' ' d'application, et donc de raison d'tre, moins qu'elle ne soit applique de prfrence la rgle gnrale. L'exclusivisme d'un anavaksa a son principe dans le conflit avec l'utsarga (PS., p. 55, 1.1 : virodhe bdhakotvom); par suite une telle rgle n'est exclusive qu'autant qu'elle doit s'appliquer elle-mme : l'utsarga s'applique ensuite s'il est encore applicable; 2 L'apavda au sens strict (cf. paribh. 57), rgle particulire qui serait applicable mme aprs la rgle gnrale (les deux tant donc svaksa). Elle exclut l'utsarga absolument (PS., p. 55, I. 2 : vinpi virodham saty api sambhave bdhakatvam). Si le mot apavda peut dsigner l'une ou l'autre des deux varits de rgles particulires, il va de soi que l'instrumental apavdatvena (ou l'ablatif apavdatvt) ne dsigne que l'apavda au sens strict, en s'opposant hanavaksatvena (tvt). Cf. Kielh.,

p. 329, n. 4. 7, 4, 85 : nug ato 'nunsikntosyo (abhysasya yalukoh [ = yai yaluki ca]). 69 Ngoj. (PS. p. 73-74) explique ainsi le japaka : "dlrgho 'kitah (7, 4, 83) ity akidgrahanam asy jnpakam / anyath yamyamyata ity atra nuki krte 'najantatvd dirghprptau tadvaiyarthyam spastam eva " " L'emploi-du mot "akitah" de la rgle "dlrgho 'kitah" est le japaka de cette paribhs. Autrement (i. e. si la paribhs n'existait pas), une fois, dans yamyamyate par exemple, le nuk ajout ( la syllable du redoublement, cette syllable) ne se terminerait plus en voyelle; tant donn par suite l'inapplicabilit de la longue, "akitah" serait manifestement superflu " . (La rduction l'absurde qui dans Patajali aboutit la paribhs se transforme naturellement en une preuve par l'absurde de la paribhs dans le PS., recueil de monographies sur les maximes supposes connues.) Kielh., dans sa paraphrase de Ng. (p. 349, n. 2), commet une erreur en disant que "akitah" a pour but d'empcher que 7, 4, 83 ne soit applique aprs 7, 4, 85. Kielh. suppose donc qu'en vertu de la paribh. 66, l'utsarga "drghah" deviendrait applicable aprs l'apavda. Or cela est inadmissible. Si 7, 4, 85 a pour effet de crer une situation (redouble termine en consonne et non plus en voyelle) qui rend 7, 4, 83 inapplicable, ce n'est pas une pari68

376 Pierre Boudon

Une gnralisation naturelle partir de cette constatation aurait consist dire : " Q u a n d il s'agit de modification du redoublement intensif, un apavda n'exclut pas un utsarga." O r . la pari-

bhs 66 est ainsi conue: "abhysavikresu bdhyabdhakabhvo nsti," "Quand il s'agit de modification du redoublement les rgles ne s'excluent pas l'une l'autre," tous les genres de conflit tant supprims. Cette formule, si elle tait donne comme le rsultat d'une induction ne se justifierait pas, car elle dpasse en comprhension la constatation du jnpaka. Il y a d'autre part des paribhs dont la porte n'est pas universelle mais limite quelques cas. Telle la paribhs 106, qui a pour jnpaka l'anubandhap de daip.70 tant donn que, par 6,1, 45,71 est substitu la diphtongue finale des racines du Dhtuptha, et qu'en vertu de la paribhs 7,72 l'anubandhap n'empche pas daip d'tre considre comme racine en ai, l'anubandha commun ne peut avoir pour but que d'inclure les deux racines dp et daip dans une nonciation commune, celle de l'exception adp de la rgle 1,1, 20 " ddh ghv adp",73 Mais cette explication se heurte la paribhs 105 " laksanapratipadoktayoh pratipadoktasyaiva grahanam" " Lorsqu'une nonciation pourrait dnoter lafois quelque chose d'original et quelque chose rsultant d'une rgle grammaticale, elle ne dnote que ce qui est original." 74 Si cette maxime tait universellement valable, renonciation d (de ddh) ne pourrait pas dnoter la racine daip, qui n'est qu'une
bhasa qui peut faire que 7, 4, 83 devienne applicable aprs 7, 4, 85. Et " akitah" ne peut pas tre destin empcher ce qui est impossible. En ralit le cas yamyamyate se prsente de lafaon que voici. Normalement, . e. en l'absence de la paribh. 66, 7, 4, 85 exclurait 7, 4, 83 ("dirghah") anavaksatvt. En vertu de la paribh. 66, la prsance srit renverse : c'est 7, 4, 83 qui, s'appliquant d'abord, rendrait 7, 4, 85 impossible. Le rle de "akitah" consiste, en sparant le domaine des deux rgles, supprimer le conflit (et la solution qu'en donne la paribh. 66); il n'y a plus ni apavda ni utsarga et le redoublement de yam ne relve que de la rgle d'augment 7, 4, 85. A noter que dans les quatre exemples donns dans le Bhsya 7, 4, 82, l'effet de la paribh. 66 consiste dans le renversement de la prdominance: Normalement: dodhaukyate : 7, 4, 83 para exclurait 7, 4, 59 acikarot : 7, 4, 94 para exclurait 7, 4, 79-93 mimmsate : 3,1,6 antaraga exclurait 7, 4, 79 ajiganat : 7, 4, 97 anavaksa En vertu de 66, on a : 17,4,59 7, 4, 79-93 7, 4, 79 7, 4, 60
70

2 7, 4, 82 7, 4, 94 3,1,6 7,4,97

Le jnpaka, mme quand il est cit, comme ci, indpendamment de tout stra, n'en est pas moins li un stra. Le jnp. de 106, c'est en ralit l'anubandha p de la racine daip, en tant que celle-ci intervient dans 1,1, 20. 71 6,1, 45 : ad eca upadese 'siti (dhtoh) " est substitu lafinale d'une racine qui dans l'enseignement (de Pnini, i. e. dans le Dhtuptha) finit en diphtongue (ec), sauf quand suit un suffixe sit". Exemple de glai, on a glt, gltum, mais glai -f sap + tip = glyati. 72 7 : nnubandhakrtam anejantatvam, littralement : " la non-terminaison en diphtongue (ec) n'est pas cre par lesanubandha". 73 " Les diffrentes racines da et dh s'appelent "ghu", except les racines d pourvues de l'anubandhap" 74 Ce n'est l qu'une des deux significations que comporte cette paribhs. Cf. Kielh., p. 486.

exclurait

7, 4, 60

377 Raisonnement par l'absurde dans Pnini

racine da drive (par application de 6 , 1 , 45), et la racine daip s'exclurait d'elle-mme. Inutile ds lors de l'inclure dans adp, d'o inutilit de l'anubandhap. Force est de conclure que la paribhs 105 ne s'applique pas ici : "da" peut dnoter les racines drives (do, den) aussi bien que les racines originales (dan, dudan). Si partir de cette constatation il s'agissait de faire une induction analogique, on tablirait qu'il en est de mme pour toutes les racines capables de dnoter, ct de la forme originale, une forme drive par application de 6 , 1 , 45, c'est--dire pour toutes les racines en . A u lieu de cela, la paribhs invoque (106: gmdgrahanesv avisesah) n'accorde ce privilge qu' trois racines. Induction qui serait absolument a r b i t r a i r e ; ce n'est qu'a posteriori, aprs avoir lu t o u t Pnini, qu'on pourrait d t e r m i n e r le domaine de cette paribhs. 75 Enfin il est une maxime qui dans Patajali n'existe qu' l'tat de constatation singulire. C'est la paribhs 91 "prakrtigrahane nyadhikasypi grahanam,"76 qui a pour jnpaka le m o t acai de 7, 3, 56 "her acai (ku-has-abhyst)." C e s t r a nous d i t q u ' " U n e gutturale est substitue la lettre h de hi (hinoti) aprs un redoublement, mais non au can (aoriste redoubl du caust if)." Or le stra 3,1, 32 " sandyant dhtavah" enseigne que le thme du dsidratifettous autres thmes drivs sont des "racines" et, en cette qualit, indpendants de la racine primitive et exclus par lasimple nonciation de celle-ci. Si ce paribhsstras'appliquait 7, 3, 56, le can, forme redoublement fait sur le thme du causatif, s'exclurait de lui-mme, et le mot acai serait superflu. Il s'ensuit que le paribhsstra ne s'applique pas. C'est cette constatation que se borne Patajali, la fin de la discussion sur le stra 7, 3, 56 et ses deux vrttika, dont le Bhsya ne fait que reprendre la conclusion77 : "Quand Pnini dit "acai", qu'est-ce qu'il a en vue? (Il pense des expressions comme) "prjhayad dtam ".
La paribh. 106 est nonce dans le Bhsyai, 1, 20, sans indication de jnpaka, la fin de la discussion sur l'interprtation de "ddh". L'cole de Bhradvja enseigne que la dfmition des " ghu" doit comporter le mot "prakrti" : l'appellation "ghu" doit s'appliquer aux nonciations da etdh, en tant qu'elles sont des thmes "originales" (prakrti s'oppose ici anukarana, " imitatif", i.e. nonciation purement matrielle). Celai cause des suffixes sit, qui ne permettent pas la transformation diphtongue > a de 6 , 1 , 45, ce qui n'empche pas les racines en diphtongue qui se trouvent au devant d'un sit de se comporter en racines ghu (ainsi concernant 8, 4,17); et 2 cause des racines vikrta, i. e. transformes par 6,1, 45, qui dans l'application de 8, 4, 17 se comportent galement en racines ghu. L'interprtation de ddh l'aide de la notion de prakrti expi que dans ce second cas la drogation la paribh. 105.
75

Cette thse est rfute par Patajali de lafaon suivante. Inutile de faire intervenir le mot prakrti dansi, 1, 20. En ce qui concerne les suffixes sit, cette intervention est ndispensable, I est vrai, dans 8,4,17, en vue de la racine ma, qui doit tre lue "prakrtim", mais "prakrti" doit tre en mme temps rapport au mot "ghu" qui prcde (ce qui dispense d'employer "prakrti" dans 1,1, 20). En ce qui concerne les vikrta, ladrogation la paribh. 105 trouve sa justification dans la paribh. 106, qui corrige 105 : " . . . vikrtrthena cpi nrthah/ dosa evaitasyh paribhsy laksanapratipadoktayoh pratipadoktasyaiveti gmdgrahanesv avisesa ti". 76 " Une racine employe (en grammaire) dnote non seulement la racine simple, mais aussi tout ce qui rsulte de l'addition cette racine du suffixe ni (suffixe du causatif)". 77 Bh., 7, 3, 56 : "acaiti kimartham prjihayad dtamj Vrt. I. hes cari/ pratisedhnarthakyam angnyatvt.

378 Pierre Boudon

Vrt. I. L'interdiction de la racine hi au can est superflue, vu qu'(au can) il s'agit d'une base diffrente. " L'interdiction de la racine hi au can (dit Ktyyana) est superflue. Pourquoi? Parce qu'(au can) il s'agit d'une base diffrente. La base (sur laquelle est form le can) est termine par ni; elle est diffrente (de la racine primitive). Aprs qu'un lopa a t substitu (par 6, 4, 51) au suffixe ni, ce n'est pas une base diffrente. En vertu de la rgle (ici 1 , 1 , 59) qui veut qu'un substitut se comporte comme l'original, (le lopa tant considr comme quivalent au suffixe ni) il s'agit d'une base diffrente." Vrt. II. Ou plutt (l'interdiction au can) est un jnpaka, qui a pour but de nous faire connatre que la gutturalisation du causatif a lieu ailleurs (qu'au can). 44 Alors en agissant ainsi, Pnini nous indique que (dans le domaine) du causatif, ailleurs (qu'au can), la gutturalisation a lieu. Quel est l'objet pratique de cette indication? C'est que dans prajighyayisati la gutturalisation se trouve correctement tablie." Ainsi donc l'exception 3 , 1 , 32, indique par le jnpaka, ne vaut, selon Ktyyana et Patajali, que pour la racine hi et seulement en tant qu'elle tombe sous la rgle 7, 3, 56.78 Et la paribhs 91, sous laforme qu'elle a dans le PS., n'est qu'une gnralisation arbitraire dont lafausset est confirme par laformule sous laquelle se prsente le vrt. 2 8, 4, 34: ce vrt. enseigne qu'aux racines numres dans la rgle il y a lieu dfaire l'addition des mmes racines "nyantnm", ce qui n'aurait aucun sens si la racine simple dnotait aussi le causatif79.

Le suffixe du causatif lid par 6, Bh. : hes cai pratisedho 'nartha4, 51 est, en vertu de 1,1, 59, consikah kim karanam angnyatvt \ nyantam tat / agom anyad bhavati dr comme prsent en vue du redoublement prescrit par 6,1,11. Ce lope krte nngnyatvam sthnivaredoublement est donc fait sur le dbhvd angnyatvam. Vrt. 11. jnpakam tu anyatra nya- thme du causatif. 78 dhikasyo kutvavijnnrtham. D'aprs Ng., la paribhs 91 a pour Bh. : evam tarhi jnpoyoty cryo domaine les rgles qui enseignent la 'nyatra nyadhikasya kutvarp bhovoti. substitution d'une gutturale, c'est- kim etasya jnpane prayojanam dire le groupe 7, 3, 54-59 (PS., p. 92, I. prajighyayisatity atra kutvam siddham 15 : yam ca kutvavisayaiva). Mais le vrt. de Ktyyana et la phrase de bhavati." Patajali ne peuvent se rapporter, du Le can de hi est ainsi obtenu : fait du mot "anyatra", qu' une rgle hiJrnic + du domaine de laquelle le can est exclu, di + lu> a + hy + / + can -\-1 c'est--dire 7, 3, 56; ce qui laisserait (6,4,71,3,1,48,7,2,116) supposer que les formes gutturalia + hy + O-f a -ft sation qui, se justifiant par la paribhs (6,4,51) 91, sont obtenues par d'autres rgles a + hay + O + a + t que 7, 3, 56 (par exemple le cari de hon, (7,4,1) ajlghanat, obtenu par 7,3, 55) n'taient a ja + hay + 0 -f a + t pas en usage aux temps de Ktyyana (6,1,11,1,1,59) et Patajali. aji +hay + O+ a + t 79 Ng., qui cite d'ordinaire tous les (7,4,93, 7,4,79) textes propres confirmer ou infirmer aj hay a t une paribhs, ne mentionne pas (7, 4, 94).

379 Raisonnement par l'absurde dans Pnini

Intressante est la comparaison des trois dernires paribhas et des trois jnpaka correspondants. Ceux-ci (akitah de 7, 4, 83, l'anubandhap de daip, c'est--dire en ralit le m o t adaip inclus dans addp de 1 , 1 , 20, acai 6e7, 3, 56) o n t mme caractre: il s'agit de trois nansamsa dont l'emploi par Pnini indique que tel mot (ou telle opration, cas de akitah) n'est pas exclusif de tel autre comme on l'attendrait normalement. Les trois jnpaka o n t aussi mme valeur. O r du premier Patajali passe une maxime gnrale, et qui dpasse mme, en comprhension, la constatation du jnpaka, du deuxime, une maxime applicable trois racines, et du troisime, une proposition qui ne vaut que pour la rgle du jnpaka. Il est manifeste que dans chacun des cas la porte de la maxime tait connue par Patajali indpendamment du jnpaka. Et de fait les paribhas, quand le t e x t e ne varietur n'en tait pas connu par la t r a d i t i o n (cas des paribhs authentiques), o n t t tablies a posteriori, d'aprs le plus ou moins grand nombre de faits qu'elles devaient s e r v i r a expliquer. C'est ce que semble bien indiquer le principe de Pyagu nd a "yvat vinnupapattis tova to jpyatvam" " A u t a n t est suggr par un jnpaka qu'il est ncessaire pour viter une incorrection " . Cela revient dire que le jnpaka n'a aucune valeur d'infrence gnrale. Il indique selon le cas une maxime universelle ou particulire; il peut mme, si besoin est, c o m p o r t e r une double conclusion. (Voir Kielh., p. 37, n. 2 et p. 172, deux exemples de jnpaka double fin). De l'examen des exemples qui prcdent il rsulte que le raisonnement bas sur le jnpaka, alors mme qu'il aboutit la constatation rigoureuse d'un fait particulier, ne permet pas d'aller au del. Il ne pourrait lgitimement fonder ni les paribhs porte limite (83,106, etc.), ni les paribhs formule gnrale, qui ne seraient alors que des gnralisations de vraisemblance. Si les jnpaka sont impuissants faire connatre les maximes d'interprtation, il convient d'admettre que Patajali et autres commentateurs ou tenaient ces maximes d'une t r a d i t i o n sre, ou bien les o n t inventes de leur propre chef. Tel est en effet le principe de la distinction entre paribhas authentiques, remontant Pnini, (et mme probablement, comme on essaiera de le d m o n t r e r , antrieures Pnini) et paribhs inauthentiques, inventes aprs Pnini.

NI Paribhasasutraet Paribhas Authentiques non Formules par Pnini. L'Antriorit de Celles-ci prouve par l'Impuissance du Jnpaka les Infrer Pourquoi Pnini a-t-il formul certaines paribhs et non les autres, bien qu'il y en ait parmi celles-ci qui ne soient pas moins importantes que les paribhsstra. Une premire rponse est donne par la paribhs 116: " jnpakasiddham na sarvatra" "Ce qui est fond sur un jnpaka ou un nyya [selon Ng. le mot jnpaka a ci ce double sens] n'est pas d'une application constante." Ce que Ngoj. commente ainsi : " La
celui-l. Le vart. 2 8, 4, 34 est signal dans la Durghatavrtti (1, 2, 2), propos de la forme udvejita, o l'on constate le guna malgr la rgle 1, 2, 2 "vija it (/t) ", qui enseigne qu'aprs la racine vij un suffixe prcd de l'augment it est it et par suite (en vertu de 1,1, 5) exclutgunaet vrddhi. L'auteur justifie "udvejita" en disant qu'il s'agit d'une forme du causatif. Rfutant ensuite l'objection qu'en vertu de la paribhs91 renonciation de la racine vij inclurait aussi le causatif, l'auteur cite le vrt. qui dtruit la paribhs.

380 Pierre Boudon

mention par Pnini de certaines maximes, alors mme qu'elles taient tablies par un jnpaka ou un nyya, a pour but de faire connatre que les autres sont a n i t / a . 8 0 " L'explication de la paribhs 116 n'aurait en soi rien d'inadmissible. Mais elle suppose q u ' i l / a deux catgories absolument distinctes de maximes, avec coexistence des caractres constant et f o r m u l dans l'une, inconstant et non f o r m u l dans l'autre. O r l'une des plus importantes paribhs de Pnini, le vsarpastra ( 3 , 1 , 94)81, est considr par Pnini lui-mme comme anitya. C'est sur cette constatation que Ngoj fonde la paribhs 68, d p o u r v u e de jnpaka direct 8 2 . Il y a l une objection srieuse c o n t r e l'explication de la paribhs 116. D'autre part cette explication prsuppose que toutes les maximes

80 PS., p. 110, I. 9 : nyyajnpakasiddhnm api kesmcit kathanam anyesm onityotvabodhanya ty arthah. Cf. Kielh., p. 509 : "When some (rules such as 1,1, 56) have been actually given (by Pnini) although they are established by jnpakas and nyyas they must be understood (to have been given by him) for the purpose of informing us that other rules (which are similarly established but have not been actually given by him) are not universally valid." Le paribhsstrai, 1, 56 : " sthnivad deso 'nalvidhou" ,si Pnini ne l'avait pas formul, pouvait tre restitucomme suit : La partie positive "sthnivaddesah" (" un substitut se comporte comme l'original ") est tablie par une maxime de sens commun (nyya) : Celui qui prend la place d'un autre en prend aussi les attributs (tatsthnpannetaddharmalabhah). Quanta la restriction "analvidhou" elle serait suggre par un jnpaka, le mot " lyap" de 2, 4, 36 "ado jagdhir lyap ti kiti (rdhadhtuke)", " jagdh est substitut de ad devant lyap et devant un rdhadhtuka qui commence par t (voir paribhs 33) et a k pour anubandha" (exemple ktv). Lyap (-(-/a) est substitut de ktv ( = tv) quand le verbe est pourvu d'un prverbe sauf o (7,1, 37). Si lyap se comportait toujours comme ktv, il et t superflu, dans 2,4, 36, de l'noncer part. Cette nonciation ne peut qu'tre indispensable : lyap ne se comporterait donc pas comme ktv concernant l'opration de2,4,36,qui a lieu devant suffixe COMMENANT PAR T. D'o la restriction " analvidhau", " . . . sauf quand l'opration dpend des lettres (de l'original)".

3 , 1 , 94 : vasarpo 'striyam (dhtoh) "Aprs une racine (i. e. quand il s'agit d'un suffixe krt), un (suffixe prescrit par un apavda) non uniforme (avec un suffixe prescrit par un utsarga) n'est que facultativement exclusif, sauf s'il s'agit du fminin." Exemple : 3, 1,133 (utsarga) prescrit nvul (= aka) et trc ( tr) aprs toutes les racines; 3,1,135 (apavda) prescrit ka (= a) aprs racines pnultime k, etc. Le suffixe ka, vu qu'il est asarpa, n 'exclut pas les suffixes nvul et trc : on a vikseptr et viksepaka ct de viksepa 82 PS., paribhs 68 (ktalyuttumunkhalarthesu vsarpavidhir nsti) dam ca vsarpavidher anityatvt siddham tadanityatve jnpakam crhe krtyatrcas ca (3, 3,169) ti \ tatra h cakrasamuccitalin knyatrcor bdh m bhd iti krtyatrjgrahanam kr'iyate . . . "Cela (paribhs 68) est tabli par l'inconstance du vsarpavidhi. Et ce qui indique l'inconstance de celuici, c'est la rgle 3, 3,169 "arhekrtyatrcas ca (Un)", o la mention des suffixes krtya et trc empche qu'ils ne soient exclus par le suffixe lin, coordonn (ex 3, 3,168) par le mot ca." Si le vsarpavidhi s'appliquait ici, i. e. si lin, suffixe d'apavda, n'excluait que facultativement les suffixes d'utsarga krtya (3,1, 95) et trc (3,1, 133), il serait superflu de mentionner ceux-ci dans 3, 3,169.
83 Goldstcker, Pnini : his place in Sanskrit Literature. London, Berlin, 1861, p. 107-118. 84 Goldstcker(p. 110) signale pour le rfuter un autre argument qu'son avis on pourrait tre tent d'invoquer en faveur de la postriorit des paribhs par rapport Pn. Mais cet argument, tir de la dfinition par Vaidyantha

81

381 Raisonnement par l'absurde dans Panini

formules et non formules ont pour inventeur Panini, thse qui sera rfute dans ce qui suit. Une autre explication est celle de Goldstcker83. Si les recueils de Ngoj. et autres, dit-il en substance, pouvaient tre considrs comme un tout indivisible, les "vieuxgrammairiens" que Ngoj., dans l'introduction du PS., cite comme son autorit ne pourraient pas avoir prcd Pnini, puisqu'il y a une paribhs qui contient le mot " pninyh " . Mais ces recueils, pourdiverses raisons, n'tant pas originaux, on ne peut juger de l'ge des maximes que par leur contenu (p. 110-113)84. Cette constatation faite, Goldstcker argumente comme suit (p. 113-115). Des deux hypothses, antriorit, postriorit des paribhs par rapport Pnini, la premire est la plus probable, vu l'indispensabilit du grand nombre de ces maximes pour une correcte application des stra. Nanmoins l'hypothse de la postriorit des paribhs ne serait pas impossible si Pnini n'en avait formul aucune. Or il en aformul un certain nombre. L'omission des autres n'est donc pas affaire de principe: l'explication n'est donne que par l'hypothse selon laquelle les paribhs omises'sont celles qui existaient dj85. des mots jnpako et nyaya, ne serait possible que si cette dfinition pouvait tre tendue aux paribhs ellesmmes, ainsi que le fait Goldstcker. Vaidyantha, commentant les mots " jnpakanyyasiddhni" employs par Ngoj. dans l'introduction du PS, s'exprime ainsi : " tatraitacchstnyaligam japakam \ etacchstralokatantrntaroprasiddhayuktir nyyah". Goldstcker (p. 108) explique : "The paribhss . . . have been defined by Vaidyantha . . . as "axioms (the existence and authority of) which are established by certain stras of Pnini, and axioms (the existence and authority of) which are established by the method that governs other works, but is applicable to Pnini also . . . In other words, these paribhss are, according to the grammarians quoted, special axioms referring to Pnini exclusively, and general axioms which avail for his grammar as well as for other works. The "certain " stras of Pnini which indicate that such Paribhss are in existence and are required for a proper application of the rules, are called jnpakaand the method of other authors which indicates that those paribhss are applied as well to them asto Pnini, bearthe nameof Nyya. " (Cf. p. 118, o G. rfute la dfinition de Vaidyan. et la remplace par une autre). Sans entrer dans l'examen de la traduction mme de G., je note seulement que ce que vise Vaidyan. dans sa double dfinition, ce n'est pas la panbhasa, mais le procd qui la fonde. Ce qu'a certainement vu G., qui applique nanmoins la paribhsellemme ce qui est dit du procd. Or la distinction est ici d'importance, du moins en ce qui concerne la premire partie de la dfinition. Si les jnpakasiddhaparibhs pouvaient tre qualifies d'"exclusivement pninennes" (telle est l'interprtation contestable qu'en dfinitive donne G. du mot "etacchastryo"), cela signifierait qu'elles on t inventes par Pnini et donc formules aprs lui, conclusion qui pourrait ensuite tre applique avec quelque vraisemblance la totalit des paribhs [ce n'est qu'ainsi que j'arrive comprendre la porte de l'argument en question]. Mais les jnpaka peuvent tre dits " exclusivement pninns ", que les paribhs authentiques aient t formules avant ou aprs Pnini, puisque dans les deux cas Pnini les a appliques, les jnpaka tant les traces d'application. La dfinition de Vaidyan., qui (dans sa premire partie) concerne les seuls jnpaka, ne peut donc, mme entendue la faon de G., fournir aucun argument touchant l'ge des paribhs. 85 II y a lieu de distinguer soigneusement la question de l'authenticit des paribhs, et celle de la premire formulation des paribhs authentiques. On constate ds l'abord qu'il y a des paribhasa qui n'ont pu qu'tre connues de Pnini (par exemple

382 Pierre Boudon

Selon Goldstcker, en effet, Panini ne serait pas l'inventeur de t o u t le systme grammatical qui nous est parvenu sous son nom. Il aurait repris l'uvre et les procds de ses devanciers, se contentant d'y faire additions et perfectionnements. En ce qui concerne les paribhs, il n'aurait formul que celles qui taient indispensables pour l'interprtation de son seul ouvrage, omettant celles qui existaient dans la t r a d i t i o n grammaticale et pouvaient s e r v i r a l'interprtation des uvres de ses prdcesseurs aussi bien que de la sienne p r o p r e . Goldstcker t r o u v e une confirmation de son hypothse dans le sens qu'a constamment le mot jnpaka dans les vieux commentateurs et notamment dans Patajali. Ce mot dsigne un stra qui se rfre un autre stra (stra ordinaire ou paribhsstra), celuici toujours antrieur. La mme relation, conclut Goldstcker, doit exister entre jnpaka et paribhs non formules : celles-ci o n t d exister avant ceux-l, si l'on veut que Patajali soit d'accord avec lui-mme dans la dfinition du mot jnpaka. Sur l'autorit de cette dfinition, qu'il croit t r o u v e r dans la formule yogpeksam jnpakam qui clt le commentaire de Patajali sur le vrt. 10 1 , 1 , 23 86 , Goldstcker dclare: qu'un jnpaka concerne l'application d'une rgle; et par suite ne peut pas prcder mais d o i t suivre la rgle qu'il indique. Q u ' u n jnpaka concerne t o u j o u r s , directement ou indirectement, selon une distinction qui sera faite dans la suite, l'applicat i o n d'une rgle, c'est certain, mais tel n'est pas le sens de l'expression invoque par Goldstcker, dont la mprise se double d'une fausse distinction entre le stra 5 , 1 , 23 d'une part et les stra 5, 2, 51, 52, (53) 87 , d'autre part. Selon Goldstcker ces derniers seulement seraient jnpaka de 1 , 1 , 23, parce que seuls ils concerneraient l'application de cette rgle. O r le stra 5 , 1 , 23, considr dans sa relation a v e d , 1, 23 ne diffre en rien de l'un quelconque des trois autres; qu'il suffise de comparer

paribhs 6), et d'autres qui lui sont non moins certainement trangres (par exemple paribhs 84). Dresser la liste exacte des deux catgories supposerait rsolu le problme de l'authenticit, lequel ne peut tre tudi que dans le cadre de chaque paribhs, Tout fait indpendante est la question de la formulation, qui n'a de sens qu'en ce qui concerne les paribhs authentiques (les fausses paribhs ne pouvant qu'tre entirement fond et forme postrieures Pnini). Le problme est celui-ci : les paribhs authentiques ont-elles t formules avant ou aprs Pnini ? Dans la premire hypothse, Pnini les a connues par une tradition ou des ouvrages grammaticaux antrieurs o elles taient dj formules. Dans la seconde, Pnini, inventeur des paribhs, les aurait appliques pour la premire fois; elles auraient t extraites de son livre au moyen des

jnpaka et formules aprs lui, soit par Patajali, soit par d'autres (par exemple les "vieux grammairiens" de Ng.) chez qui Patajali les aurait trouves. 86 1 , 1 , 23; bahu-gana-vatu-dati sarpkhy " Les mots bahu, gana, les mots en vatup (c. tvat), les mots en dati (kati, etc), sont des samkhy (-f des numraux)". 87 En ralit G. n'oppose 5,1, 23 qu' 5, 2, 51 et 52, parce que le manuscrit de Kaiyata qu'il cite ne mentionne que ces deux stra. Mais 5, 2, 53 ne peut pas tre spar des deux autres; les trois stra ont exactement mme caractre quant la relation aved, 1, 23: 5,2, 51 : sat-kati-katipaya-caturm thuk (dat > dati paratah). 52 : bahu-pga-gana-saghasya tithuk (dat > dati paratah.) 53 : vator thuk (dat > dati paratah).

383 Raisonnement par l'absurde dans Pnini

5 , 1 , 23 : vator id v (ex 22 : kan > kanah)88 et 5, 2, 53 : vator ithuk (ex 48 : dot > dati paratah)89 Dans les deux stra le collectif vatup est trait conformment la dnomination de samkhy qu'il a reue par 1 , 1 , 2.3, les augments prescrits tant l applicable un suffixe (kan) ici dtermin par un suffixe (dat), qui l'un et l'autre n'appartiennent qu'aux samkhy. Si Pnini, dans la dfinition 1 , 1 , 23, avait omis le mot vatup, le lecteur tait mme de le suppler en s'appuyant son gr sur 5 , 1 , 23, ou 5, 2, 53, car l'un et l'autre stra seraient galement inexplicables si les mots en vatup n'taient pas des samkhy. Les deux stra o n t donc en d r o i t mme valeur de jnpaka. Et l'quivalence est la mme entre 5 , 1 , 23 et 5, 2, 51 ou 52. En pratique il existe une diffrence entre 5 , 1 , 23 et le groupe 5, 2, 51-53. Ceux-ci s'opposent celui-l, non comme des jnpaka un non-jnpaka (ainsi que le croit Goldstcker) mais comme des jnpaka particuliers et rigoureux un jnpaka inference gnrale et indtermine. C'est, dans l'esprit de Patajali, un jnpaka conventionnel, relatif l'usage verbal concernant la notion de nombre, c'est--dire destin justifier toutes les expressions o un mot, mme tranger la catgorie des numraux ordinaires, est trait comme un numral. Tel est le sens de l'expression yogpeksam jnpakam, qui clt la discussion sur le vrt. 10 (bahvdlnm agrahanam jnpakt siddham) entre matre et lve, l'un A(dversaire), l'autre D(fenseur du stra, ou t o u t au moins faisant des objections la suppression du stra): A. L'emploi des mots bahu, etc, pouvait tre vit (i.e. le stra 1 . 1 . 23 peut tre supprim). 90 D. Par quel moyen alors dans les rgles d'opration (pradesa)91 applicables aux samkhy, y aura-t-il une correcte interprtation du mot samkhy? A. Cette interprtation est assure par un jnpaka. D. Quel jnpaka?
5 , 1 , 23 : "/testfacultivement l'augment du suffixe kan venant aprs un numral qui finit en vat." Exemple tvatkah ou tvatikah. 89 5, 2, 53 : " ithuk est l'augment d'un thme finissant en vat, quand suit dat" (suffixe des ordinaux, voir 5, 2, 48). Exemple yavatm pranah = yvatithah it se place au devant de kan; ithuk aprs le mot en vatup. Cf. 1,1,46: dyantau takitau. kanah ( 5 , 1 , 23) et vator (5, 2, 53) sont des gnitifs, vator de 5 , 1 , 23 est un ablatif au sens de 1.1. 67. 90 A. bahvdJnm grahanam sakyam akartum. D. kenednlm samkhypradesesu samkhysampratyayo bhavisyati. A. jnpakt siddham tat. D. jnpakam kirn. A. yad ayam vator id veti samkhyy vihitasya kano vatvantd tarn ssti. D. vator eva taj jnpakam syt. D. nety ha / yogpeksam
88

jnpakam. 91 pradesa dsigne la rgle d'opration par opposition soit la rgle de dfinition, soit la rgle (ou maxime) d'interprtation, qui intervient comme subordonne dans la premire. Ce sens apparat nettement dans le passage du PS consacr aux paribh. 3 et 4 (yathoddesa0 et kryaklapaksa), o le mot pradesa revient plusieurs fois, toujours oppos samjn ou paribhs. Cf. notamment PS., p. 3, I. 7 : samjnsstrasya tu kryaklapakse na prthagvkyrthabodhah kirn tu pradesavkyrthena sahaiva, "Une rgle de dfinition, au kryaklapaksa, n'a pas de sens considre isolment, mais seulement en union avec le sens exprim par la rgle d'opration (o elle intervient)", et 1.10 " . . . koryajnnam ca pradesadesa eva . .", " l a connaissance d'une opration (grammaticale) n'a lieu qu' l'endroit du pradesa".

384 Pierre Boudon

A. Le fait que Pnini par la rgle vator id va (S. 1. 23) enseigne qu'aprs un mot finissant en vatup, it est l'augment du suffixe kan prescrit (par 5 , 1 , 22) aprs une samkhy. D. Ce jnpaka ne vaudrait que pour les vatup. A. Non, un jnpaka vaut pour l'usage,92,,

Cette dernire rplique est commente par Kaiyata de la faon suivante: " Parce que ce jnpaka (5,1, 23) serta la rfutation de cette rgle, qu'on ne comprenne pas pour cela qu'il existe en vue de cette (seule) rgle. Au contraire (le sens est:) il concerne les usages (verbaux, dans le domaine de la catgorie nombre), d'o son nom " relatif l'usage". (En ce qui concerne les mots dfinis par la rgle 1,1, 23) le fait que Pnini, par les rgles 5, 2, 52, 53, 51, prescrit (tel ou tel) augment, "quand dot suit", ind que qu'il ya opration de samkhy"93. Cette dernire phrase n'oppose donc pas les trois stra 5, 2, 51-53 au stra 5,1, 23 comme des jnpaka un non-jnpaka. La mprise de G. est d'autant plus surprenante que la phrase de Kai/., telle qu'il la cite, ne mentionne comme jnpaka de 1,1, 23 que les deux rgles 5, 2, 52 et 51. Texte point inadmissible et qui supposerait ceci : c'est parce qu'il considre 5,1, 23 comme jnpaka particulier des samkhy en vatup que Kai/, se dispense de faire appel 5, 2, 53. Autrement dit, 5,1, 23, indpendamment de son rle de jnpaka gnral, aurait la mme valeur prcise relativement aux vatup que 5, 2, 52, 51 relativement bahu-gana et les mots en dati. Et dans ce cas raison de plus de ne pas l'opposer celles-ci comme un non-jnpaka des jnpaka. Cette variante sur la citation des jnpaka par Kaiy. est d'importance minime. L'intrt est dans le mot yogpeksam. Qu'il n'y ait pour Kaiy. qu'une faon de le comprendre, c'est ce que confirme Ngoj. Ngoj., en dsaccord avec Pataj., n'admet pas le rejet du stra 1,1,23, parce que, mme en conversant la FORMULE du stra, c'est-dire mme si n'existe pas, comme donne infrer, la DONNE formule par le stra, les rgles 5, 2, 51-53 ne sont pas vyartha, mais indispensables pour indiquer la possibilit du suffixedat aprs pga, etc., et par suite ne peuvent pas tre jnpaka de 1,1, 2394. Ayant ainsi donn son opinion personnelle sur le Bhsya, Ng. passe l'interprtation de Kaiy. : " D'aprs certains, dit-il 95 , le stra vator id va indique (et justifie) le caractre samkhy des autres oprations, mme caractrises (en elles-mmes) par l'absence de cause consistant en dnombrement, dont on constate dans les formes usuelles qu'elles ont eu lieu comme si dtermines par la qualit de samkhy. Par l est tabli le caractre samkhy des oprations subies par adhika, etc96. Voil pourquoi Patajali dit yogGoldstcker fond les rpliques 4, 5, 6 en une seule. La phrase yad ayam ..., qui affirme la valeur de jnpaka de 5, 1, 23, devient dans sa traduction interrogative et se trouve ainsi nie par la rplique finale. 93 Kaiy. : yogpeksam iti \ asya yogasya pratykhynd etadyogpeksam iti na boddhavyam \ kirn tu yogan apeksata iti yogpeksam / yad ayam bahupgaganasaghasya tithuk (5, 2, 52) vator ithuk (53) satkati (51) iti dati parata gamam sdst taj japayoti bhavati samkhykryam iti.
92

Ng. : y ad aya m iti. . . pgdibhyo dadartham bahuganeti stre krte 'py jnpakvasyakatvam ti bhvah. 95 Ng : ke cit tu vator d vety anena samkhynakaranatvbhvavatm apy anyesm prayogesu drsyamnasamkhykrynm samkhyakryam jnpyate / tendhikdmm api siddham tad ha yogpeksam iti prayogpeksam ity arthah \ etena pgdibhyah sasdiprasango 'pi varita ity huh. 96 Bien qu'aucune rgle ni japakade Pnini ne permette d'attribuer au mot adhika la qualit de samkhy, un corn-

94

385 Raisonnement par l'absurde dans Pnini

peksam; " relatif l'usage" (prayogapeksam), voil ce qu'il veut dire. Grce cette interprtation se t r o u v e carte du mme coup (api) l'application indue des suffixes sas, etc. aprs pga, etc. Ainsi disent ( K a i / , et autres, commentant la pense de Patajali) " . Ngoj. approuve-t-ii tacitement l'interprtation selon K a i / . du jnpaka 5,1,23 et de son qualificatif yogpeksaml La considret-il comme suspecte comme le laisserait croire l a f o r m u l e " k e cit tu" ? Une traduction plus simple de yogpeksam jnpakam, qui vient en rponse l'objection " C e jnpaka ne vaudrait que pour les vatup", serait " N o n , il vaut pour (toute) la r g l e " . Mais c'est l pour Kaiy. un contresens contre lequel il met en garde avec une insistance qui mrite d'tre prise en considration. D'autre part cette traduction ne rsout aucune des difficults que pose le stra 1 , 1 , 23 ; elle en cre de nouvelles. De quel d r o i t un jnpaka qui n'implique la qualit de samkhy que pour les vatup, vaudrait-il pour les autres mots de 1 , 1 , 23. Si son inference peut tre arbit r a i r e m e n t tendue trois mots qui sont hors de son domaine, mieux vaut alors l'appliquer partout ou il est utile de le faire, c'est-dire tous les mots que l'usage t r a i t e comme des samkhy. C'est cette solution qu'a probablement choisie Patajali. Le stra 1 , 1 , 23 soulevait en effet deux difficults, l'une antrieure au rejet du straet l'une des causes du rejet : c'est que ce stra est incomplet, vu qu'il omet adhika et autres mots qui ne diffrent en rien quant l'application des samkhypradesa des quatre mots qu'il mentionne. Comme pour ceux-ci la qualit de samkhy est implique par des jnpaka, Pataj., la suite de Ktyyana, avait deux raisons de condamner le stra. Mais le rejet de 1 , 1 , 23 par les rgles 5, 2, 51-53 soulevait une deuxime difficult: il fallait empcher que l'infrence des trois jnpakastra ne s'etendt, comme le voudrait la logique, tous les mots qu'ils contiennent. A u t r e m e n t pga, sagha, katipaya subiraient invitablement, t o u t comme bahu et gana, etc., les oprations dtermines par la qualit de samkhy, par exemple l'adjonction des suffixes sas (5, 4, 43), dh (S, 4, 20), krtvasuc(S, 4,17), etc. O r il se t r o u v e qu'il y a deux rgles, mme valeur de jnpaka, qui impliquent l'une et l'autre la qualit de samkhy pour les mots en vatup: 5 , 1 , 23, et 5, 2, 53. Une seule suffisant cet office, Patajali fait jouer l'autre le rle de jnpaka conventionnel qui lui permet d'chapper aux difficults susdites. Du moment en effet que Patajali attribue 5 , 1 , 23 une capacit d'infrence indtermine, signifiant qu'il y a opration de samkhy pour tous les mots qui, diffrents des numraux o r d i naires, sont traits par l'usage l'gal de ceux-ci, sont exclus du mme coup tous les mots exclus de ce t r a i t e m e n t par l'usage. Ainsi s'explique que les jnpaka particuliers 5, 2, 51 et 52 impliquent la qualit de samkhy sans restriction pour les mots bahu, gana, kati, t o u t en n'impliquant que la possibilit de prendre le suffixefat pour les mots pga, sagha, katipaya.

En rsum, le stra 1,1, 23 serait tacitement rejet au moyen des trois jnpaka 5, 2, 51-53 ; mais Patajali se contenterait pratiquement dfaire appel au jnpaka conventionnel 5,1, 23, qui englobe les trois autres et en mme temps supple ce qu'ils ont d'insuffisant. Parce que Patajali devait tre tent de recourir cette convention, il est probable que Kaiy. interprte exactement
pos comme adhikasastika n'en est pas moins conforme aux rgles 2,1, 51 et 7,1,15, qui concernent l'une et l'autre les composs dont le premier terme est une samkhya (cf. vart. 9 1,1, 23). Mme cas pour adhyardha et ordha par rapport aux rgles 2,1, 51 et 5,1, 22(Cf. vrt. 5et71,1,23).

386 Pierre Boudon

la pense de Patajali. Laformule yogapeksam jnpakam,au moins selon Kaiyata (or c'est travers Kaiy. que Goldstcker prtend comprendre Patajali), ne signifie donc pas qu'un jnpaka concerne l'application d'une rgle, mais bien qu'un jnpaka concerne " les usages " . Cette faon de comprendre l'interprtation de Kaiyata est encore confirme par le m o t prayoga, par lequel Ngoj.traduisant la pense de Kaiyata 97 glose yoga, car prayoga, qui s'oppose sastra comme la pratique la thorie, ne peut dsigner ici que l'usage verbal qui tend chapper aux rgles, et ne s'y ramne qu'au moyen d'artifices comme celui qu'on vient de voir 98 . Une rgle-jnpaka, ajoute Goldstcker, ne peut pas prcder mais doit suivre la rgle qu'elle indique. Qu'en fait il en soit ainsi, cela rsulte de l'arrangement des stras de Pnini. Mais on ne peut pas dire qu'il doit en tre ainsi. Dans un o r d r e des stra qui serait diffrent, le jnpaka pourrait prcder sans cesser d'tre jnpaka. En ce qui concerne en particulier les jnpita qui sont des samjn- ou des paribhsstra, la position antcdente ne se conoit invitable (et encore du seul point de vue de Pnini, e x t r i e u r ici la notion de jnpaka) que par rapport aux jnpaka que l'on peut appeler directs, ceux que constitue un cas d'application du samjn- ou du paribhsstra. Il en va t o u t autrement pour les jnpaka inference indirecte, tel celui qui est cit par Goldstcker p. 117 n. 136 et qu'on rencontre dans le commentaire du vrt. 1 la rgle 1,1,3 ko gunavrddh. Ce vrt. (iggrahanam tsanhyaksaravyanjananivrttyartham) enseigne que 1,1, 3 a une triple valeur ngative: il exclut de la substitution de guna et vrddhi la lettre a, les diphtongues et les consonnes. Sous son troisime aspect, 1,1,3 signifie donc "na vyajanasya guno bhavati".Mais, dit le contradicteur99, cette vrit n'a pas besoin d'tre formule; elle est implique par un jnpaka: lacaractrisation par l'anubandhad (ditkaranam) du suffixe da prescrit par 3, 2, 97100 aprs la racine jan compose avec un mot reprsentant un locatif. Led signifiant l'lis ion du ti (partie du thme qui commence avec la dernire voyelle, cf. 1,1, 64), on a par
97 G. (p. 116, n.134) cite la glose yogapeksam iti prayogpeksam iti arthah comme si c'tait l'opinion personnelle de Ngoj. En ralit, cette glose, tant comprise entre les mots ke cit . . . huh, est cense celle de Kaiy. dont Ng., dans tout ce passage, ne fait qu'interprter l'interprtation. 98 II faut reconnatre que l'interprtation de Kaiy. peut soulever des objections, l'une constitue par l'emploi de yoga (au moins au pluriel ; cf. Kaiy.: yogn apeksate et Nag. : prayogesu) au sens de prayoga. Qu'on prfre l'interprtation de Kaiy. ou Tinterprtation plus facile (trop facile) selon laquelle "ce jnp. concerne toute la rgle", une chose est certaine, c'est que la traduction de Goldstcker est inadmissible. D'ailleurs "application de rgle" est une expression complexe, dont l'quivalent sanscrit ne peut tre qu'un compos. Ce compos

existe. C'est sstrapravrWh (v. Mahbhsyam, Introd., comm. du vrt. 15), ailleurs laksanapravrttih (texte de la par. 111 : parjonyaval laksanapravrttih) ou encore strapravrttih (PS., p. 3, I. 12), compose premier terme variable selon le mot choisi pour signifier " r g l e " , mais dont le second terme est pravrtti pour signifier "application " . " Bhsya ad 1,1, 3, vrt. 1 : . . . vyanjananivrttyarthenpi nrthah / cryapravrttir jnpayati na vyajanasya guQO bhavatlti yad ayam janer dam ssti katham krtv jnpakam ditkarana tat prayojanam dititi tilopo yath syt / yadi vyajanasya gunab syd ditkara-. nam anarthakam syt. 10 3, 2, 97 saptamym janer dah. " Le suffixe da vient aprs la racine jan, au sens d'un locatif" (i. e. quand la racine jan est compose avec un mot reprsentant un locatif).

387 Raisonnement par l'absurde dans Pnini

exemple au sens de " upasarejatah ", uposara + jan - f da = upasarajah. O r si leguna prescrit devant srva-et rdhadhtuka pouvait t r e substitut d'une consonne, cette " c o t i s a t i o n " duc/oserait superflue, car aprs substitution du guna, les t r o i s voyelles, celle de la racine, celle du guna, celle du suffixe, se ramneraient par 6 , 1 , 97 la d e r n i r e d ' e n t r e elles, d'o la mme f o r m e upasarajah. Mais comme l e d du suffixe da de 3, 2, 97 ne peut q u ' t r e indispensable, ce suffixe dit d o i t reprsenter l'unique procd d e f o r mation praticable. Conclusion : le guna d'une consonne est illicite. Ce jpaka est discut et finalement rejet dans le dialogue que voici 1 0 1 : " C e t t e prescription du suffixe da aprs jan que vous inv o q u e z n'indique rien (japakam na). C'est seulement si (l'objet de la rgle 3, 2, 97, i.e. la formation de upasaraja, etc.) pouvait tre obtenu par application d'un principe "guna est substitut de consonne" que la rgle tablie par Pnini aurait valeur de jnpaka. O r (les formes en question) ne peuvent pas tre obtenues par le guna (de la finale) ejan. D'o peut-on savoir en effet que leguna prescrit est a et non pas plutt e ou o102? En vertu de l'affinit (prosodique), une consonne, valant une demi-mtr, ne peut pas avoir d'autre substitut que la lettre a, d'une mtr. Mme ainsi, c'est la voyelle nasalise qui se substitue (ce qui est incorrect). Par sa fusion avec la voyelle suivante la nasalise reviendra l'tat pur. " yad apy ucyote joner davacanam japakam na vyajanasya guno bhavatlti siddhe vidhir rabhyamno jopakrtho bhavati na ca janer gunena sidhyati / kuto hy etajjaner guna ucyamno 'kro bhavati na punar ekro v syd okro veti. ntaryato 'rdhamtrikasya vyajanasya mtriko 'kro bhavisyati. evam apy anunsikah prpnoti. pararpena suddho bhavisyati. evam tarhi gamer apy ayam do vaktavyah \ gamer ca guna ucyamna ntaryata okrah prpnoti / tasmd iggrahanam kartavyam/l 102 Celui qui faisait appel au jnpaka supposait deux possibilits d'arriver upasaraja, etc. La prescription par Pnini de l'une (3, 2, 97) impliquait l'exclusion de l'autre (guna de la consonne par une rgle 7, 3, 84 conue sans le mot " i k a h " ) . Rponse : c'est seulement si le guna de la finale de jan reprsentait bien une premire possibilit que la rgle 3, 2, 97, moins d'tre vyartha, impliquerait que le guna de la consonne est illicite. Mais en fait le guna ne permet pas d'obtenir upasaraja. En d'autres termes, 3, 2, 97 serait indispensable mme si d'une manire gnrale le guna de la
101

consonne tait licite. Elle ne peut donc pas tre jnpaka de la proposition contraire : " na vyajanasya guno bhavati". Je traduis d'aprs l'dition Kielhorn. Le texte de l'dition Sivadatta D. Kudla (Bombay, 1917) est le suivant " yad apy ucyatejanerdavacanaml na vyajanasya guno bhavatti II siddhe . . . " L'diteur qui con sidre les mots "janerdavacanam" comme une interpolation, pense avec la Chy que Patajali reprend simplement la donne 3 du vrt. : " na vyajanasya guno bhavati". Il est plus probable que c'est le jnpaka qui est d'abord rappel et ni avant d'tre rfut en forme. On peut ainsi conserver les deux mots suspects en coupant ' ' yad apy ucyate janer davacanam na", le fragment "vyajanasya guno bhavati" se construisant alors avec siddhe. Mais de quelque faon qu'on ponctue, il n'y a qu'une interprtation possible de siddhe, qui a pour sujet, comme sidhyate de la ligne suivante, ce qui est l'objet de la rgle 3, 2, 97, c'est--dire la formation de upasaraja, etc. Cf. Bhsya, 7, 2,102, jnp. discut aprs la troisime krik et o le mot siddhe a mme valeur.

388 Pierre Boudon

(S'il en est) ainsi (pour les drivs de jan, i.e. si l'on peut arriver upasaraja, etc., par le guna de la consonne, il reste) alors (une autre objection: rappelez-vous que) le mme suffixe da doit tre prescrit (par 3, 2,101) aprs la racine gam. Or si Ton applique ici le procd du guna, c'est la lettre o qui, en vertu de l'affinit (selon l'organe de prononciation) se substitute indment la consonne. (La d/tisation du da, lequel est valide de 3, 2, 97 dans 3, 2, 101, est donc l'unique possibilit de former les drivs de gam et ne peut pas avoir la valeur de jnpaka que vous lui attribuez). Voil pourquoi le mot ik (en ce qui concerne le sens 3 que lui donne le vrt.) devait tre employ dans 1 , 1 , 3 " . Ainsi la "d/tisation " par Pnini du suffxe/a n'est pas jnpaka de 1 , 1 , 3 (3), car si mme da peut tre regard comme inutile en ce qui concerne les drivs dejan, il est destin passer de 3, 2, 97 dans 3, 2, 101, o il reprsente l'unique possibilit de former certains drivs de la racine gam. Parce qu'il aurait une raison d'tre mme en l'absence de la vrit enseigne par 1,1,3 (3), il n'est pas jnpaka de cette vrit, qui par suite devait tre formule par Pan ini 103 . C'est l un exemple de ces pseudo-jnpaka qu'invoque si souvent l'un des interlocuteurs l'appui d'une objection contre un sdtra ou un vrt. et qui sont finalement rejets par Patajali. La rgle 3, 2,101 " anyesv api drsyate" (traduction ordinaire : "(Le suffixe da) se rencontre (aprs la racine jan) mme quand elle est en composition avec d'autres (mots que ceux qui sont numrs dans les prcdents stra) " est comprise par Patajali de la faon suivante : "(Le suffixe da) se rencontre aussi dans le cas d'autres (racines quejn)", l'une de ces racines tant gam.-Trapp (Die
ersten fnf Annikas des Mahbhsyam ins Deutsche bersetzt u erklrt,
103

Leipzig, 1933, p. 215, n. 76), croyant que Patajali fait allusion la rgle 3, 2, 48, qui prescrit le suffixe da aprs la racine gam prcde de diffrents mots (par exemple sarva, d'o sarvagah), explique ainsi l'objection de la dernire rplique : " Es msste also fr das m, o eintreten und das stumme d wre zwecklos." C'est exactement le contraire que serait la conclusion de Patajali (mme s'il s'agissait de 3, 2, 48) : c'est parce que le guna, en tant que reprsent ci par o, ne permettrait pas d'arriver sarvaga et autres formes semblables, que l'anubandhad de da serait dclar indispensable (comme il l'est en fait) en tant qu'unique moyen de formation. Ce n'est pas l'inutilit du d ce qui en ferait un jnpaka, mais au contaire son indispensabilit ce que l'empche d'tre un jnpaka,

que Pataj. a besoin de dmontrer. Car Pataj. dfend ici laformulation du stra 1,1, 3 (aspect 3) et laformulation du stra n'est dfendable qu'autant que la chose formule n'est pas suggre par un jnpaka. Mais ce n'est pas du da de 3, 2, 48 qu'il s'agit. L'anubandhad de ce do a beau tre indispensable, Pataj. n'en serait pas plus avanc, si le d du da de 3, 2, 97 tait reconnu inutile. C'est l'indispensabilit de celui-ci qui est mise en cause par le contradicteur et qu'il s'agit de dmontrer. Or Pataj. vient dfaire une concession : il a implicitement reconnu que, le guna de la consonne permettant d'obtenir upasaraja, etc., la d/tisation du da de 3, 2, 97 pouvait tre considre comme INUTILE EN CE QUI CONCERNE LES DRIVS DE JAN. Il s'agit nanmoins de lui trouver une raison d'tre pour l'empcher d'tre un jnpaka. Le suffixe da est valide dans les rgles suivantes jusqu' 3, 2,101 inclusivement. Si dans toutes ces rgles il s'agit de la racine jan et par suite d'une drivation uniforme, l'inutilit du caractre dit du suffixe se trouve reconnue unefois pourtoutes. Si au contraire, "onyesu" de 3, 2,101 signifiant "dans le cas d'autres racines" (il n'importe en rien que ce soit ou non l'intention de Pnini; il ne s'agit ici que de l'intention de Pataj.), la racine gam est

389 Raisonnement par l'absurde dans Pnini

C'est en mme temps un exemple de jpaka indirect. Le raisonnement part d'une hypothse affirmative ("si le guna de la consonne tait l i c i t e " ) que le jnpaka exclut parce qu'il en reprsente l'quivalent pratique; d'o, conclusion ngative. Par suite 3, 2, 97 (pour celui qui invoque le jnpaka) est en relation avec le principe contenu dans 1 , 1 , 3 , non comme illustration directe, vu qu'en soi elle n ' a r i e n v o i r avec l'application des substituts guna et v r d d h i , mais simplement parce qu'elle exclut le principe contraire 1 0 4 .Inverse est le mouvement dans le cas des stra 5, 2, 51-53, jnpaka directs de 1 , 1 , 23 : on part d'une hypothse ngative ("si bahu, etc., n'taient pas des samkhy1'), contredite par les japakastra, qui sont des cas d'application de la dfinition des samkhy; d'o, conclusion affirmative. O r si dans ce dernier cas les jnpaka ne peuvent que suivre le japitastra, le fait (dans 5, 2, 5153) de t r a i t e r comme numraux d'autre mots que les numraux ordinaires exigeant de la part de Pan ini (pour qui 5, 2, 51-53 n'avaient pas valeur de jnpaka) un avertissement pralable ses lecteurs, savoir la dfinition 1 , 1 , 23, il n'en est pas de mme pour un jnpaka indirect comme celui de 3, 2, 97. Celui-ci pourrait, sans inconvnient pour lui-mme, prcder le prtendu japitastra, vu qu'il en est indpendaut. Il vient aprs parce qu'en vue d'un certain arrangement des stra qui prsentait des avantages didactiques Pan in i a plac dans le premier pda le plus grand nombre de ses samjn- et paribhsstra. Qu'au lieu de faire ainsi il les et distribus par petits groupes selon les adhyya o, ils devaient i n t e r v e n i r : le stra " ko gunavrddh" ( 1 , 1 , 3) pourrait t r e numrot 7 , 1 , 3 : le d du suffixe situ dans 3, 2, 97 (pour celui qui l'invoque) n'en en serait pas moins le jnpaka.

susceptible de prendre le da, on est en prsence d'un cas diffrent: le guna de la consonne tant inapte former de gam les drivs voulus, le caractre dit du suffixe est ci indispensable. Mais do n'est valide dans 3, 2,101 que parce qu'il est nonc dans 3, 2, 97. Ainsi le d du da de 3, 2, 97 aurait une raison d'tre mme en l'absence de la proposition " na vyajanasya guno bhavati." Il ne peut pas tre le jnpaka de cette proposition, qui par suite devait sous les espces du mot ik tre formule par Pnini. Telle est la seule manire, d'ailleurs conforme aux commentaires de Kaiy. et Ng., de comprendre la rfutation par Pataj. de ce pseudo-jnpaka. A noter que l'interprtation qu'admet ici Pataj. de "anyesu" n'a pas t invente pour la circonstance; elle est conforme au commentaire qui est fait en son lieu du stra 3, 2,101 : " anyebhyo' pi drsyata iti vaktavyam \ ihpi yath syt / akh utkh parikh." La Ksik est encore plus catgorique: "apisabdah sarvopadhivyabhi crrthah . . ." " Le mot api signifie

que (la prescription du suffixe da) est indpendante de toute condition". Formule reprise par Ng. qui, plus logiquement, attribue cette valeur
au mot drsyate " . . . anyesv api drsyata ti stre drsigrahanasya sarvopadhivyabhicrrthatvena . . . " L'interprta-

tion de "anyesu" comme impliquant (entre autres choses) "anyebhyas" "aprs d'autres racines," semble donc appartenir la tradition des commentateurs. 104 Les jnpaka appels ici indirects le sont en ce sens seulement que la conclusion, tant ngative, ne peut pas tre prouve par une trace positive d'application du principe ni, mais seulement par la non-application du principe contraire. Ce que Kielh. en parlant des jnpaka de paribhs appelle jnpaka indirects (3, 3,169, jnp. de 68 (v. supra, p. 380, n. 82); atm de 2, 2.19, jnp. de 75) ce sont des jnpaka qui ne permettent d'arriver la paribhs que par l'intermdiaire d'une premire inference dont la paribhs est une consquence.

390 Pierre Boudon

II y a des jnpaka qui pourraient tre situs aprs la rgle qu'ils impliquent (ou sont censs impliquer). Et mme dans le cas des jnpaka qu'on a appels directs, parce qu'ils consistent en une trace positive d ' i n t e r v e n t i o n d'une autre rgle, lordre suivi par Pnini ne correspond pas une ncessit absolue. On ne voit donc pas comment la position relative dans la grammaire de Pnini du jnpita- et du japakastra pourrait impliquer une relation parallle entre la paribhs et son jnpaka. A u lieu de se placer sur le plan des donnes crites, peut-tre convient-il de se m e t t r e dans l'esprit de celui qui, dans le dialogue de Patajali, fait appel au jnpaka. (\\ n'est ici question que des jnpaka qui se r l r e n t un stra). Cet interlocuteur (cas gnral) a commenc par nier la raison d'tre d'un stra (ou d'une partie de stra); il en suppose l'inexistence. " T e l l e rgle (ou partie de rgle), d i t - i l , aurait pu t r e omise: tel mot (ou procd) d'une autre rgle en implique l ' o b j e t " . La v r i table relation qui existe entre le jnpaka comme tel et le stra ou lment cens inexistant et par l mme japya, c'est la relation logique du connu l'inconnu. La relation de fait note par Goldstcker est quelque chose d'accidentel qui ne peut rien prouver touchant la relation jnpaka-paribhs. Que les paribhs aient prcd Pnini, c'est--dire leurs jnpaka, ou qu'inventes par Pnln'i et formules aprs lui, elles aient suivi leurs jnpaka, l ' o r d r e : rgle implique, rgle indicatrice, peut se concevoir tel qu'on le constate dans Pnini. L'hypothse de l'antriorit des paribhs par r a p p o r t a Pnini,fonde par Goldstcker sur la ncessit d'expliquer leur absence de YAsthyyl et sur le fait gnral que Pnini n'est pas l'inventeur de t o u t son systme grammatical, n'est donc pas confirme par la constatation que, dans le corps des stras, le jnpita prcde le jnpaka. Par contre elle t r o u v e une confirmation dans la conclusion de l'analyse du procd d'argumentation qui a pour point de dpart le jnpaka: les jnpaka tant inaptes faire connatre exactement les paribhs, il serait invraisemblable que Pnini, s'il en tait l'inventeur, les et omises systmatiquement pour laisser la sagacit de ses commentateurs le soin de les ext r a i r e de son t e x t e , puisque c'est chose impossible; si les paribhs ont t omises, c'est parce qu'elles taient dj connues 105 . Ainsi peut se dfendre lathse de Goldstcker mme prive de son principal argument. La discussion de la f o r m u l e " yogpeksam jnpakam" adonn l'occasion de voir des jnpaka (ceux de 5, 2, 51-53) dont l'infrence tait rigoureuse. Il ne peut en tre autrement lorsque le jnpitastra est une dfinition (cas de 1 , 1 , 23), et que chacun des mots dfinis a son jnpaka. Comme il ne s'agit de t i r e r de celui-ci qu'une constatation singulire, l'applicabilit te) mot de telle dsignation, il ne saurait tre question d'induction et le prasangnumna (rduction l'absurde) suffit donner la conclusion voulue. D'autre part on a vu que les paribhs ne peuvent pas, en tant que formules gnrales, t r e fondes sur Jes jnpaka. Faut-il en conclure que le mot jnpaka aurait deux valeurs diffrentes selon le cas, ou qu'il aurait dans les deux cas (incorrectement dans le second cas) la mme valeur prcise de hetu ou sdhana? N i l'une ni l'autre de ces deux suppositions n'est admissible.

Cette conclusion ne vaut rigoureusement que pour les jnpakasiddhaparibhs (les plus importantes et les plus nombreuses). Mais on peut,

105

comme le fait Goldstcker de sa propre conclusion, l'tendre sans trop de risque l'ensemble des paribhs.

391 Raisonnement par l'absurde dans Pnini

En ralite les mots japaka etjapoyati ne se r a p p o r t e n t pas t r o i t e m e n t l ' o p r a t i o n 'inference. Il est p r o b a b l e q u e Patajali a e m p l o y dessein un m o t d'ample acception p o u r dsigner une o p r a t i o n q u i , c o m p o r t a n t t o u j o u r s sa base une inference (celle du prasangnumna), t a n t t n'tait pas a u t r e chose (cas de certains j n p i t a s t r a , par e x e m p l e , 1 . 1 , 23), t a n t t tait plus q u ' u n e i n ference (inference s i n g u l i r e ^ rappel de l a m a x i m e : cas n o t a m m e n t des paribhs) 106 . En ce q u i concerne ce d e r n i e r cas, ce n'est qu'au m o y e n de la f i g u r e pars p r o t o t o q u e l ' a r g u m e n t a t i o n d e Patajali a pu t r e t r a i t e 'anumna et la paribhs d'anumnik, mots q u ' o n r e n c o n t r e dans Ngoj 1 0 7 . C e t t e v o l u t i o n q u i t e n d a i t faire du m o t japaka le s y n o n y m e exact de hetu tait favorise par le fait q u ' o n avait d a t t r i b u e r au japaka une capacit d ' i n f r e n c e t o t a l e p o u r j u s t i f i e r les fausses paribhs: ainsi la paribhs 91 q u i , n'existant dans Patajali q u ' a l'tat d e c o n s t a t a t i o n singulire, est i n d m e n t passe dans la suite au rang de f o r m u l e gnrale. L ' e m p l o i par N g o j , des mots anumna, anumnikl, pas plus q u e l ' e m p l o i des mots japaka, japayati par Patajali, ne p r o u v e rien t o u c h a n t l ' o r i g i n e des paribhs a u t h e n t i q u e s . Il reste tabli qu'elles ne p e u v e n t pas t r e infres par leurs japaka. C a r ces japaka ne sont en ralit q u e des indices rappelant les maximes sous-jacentes, q u e Pnini avait suivies, mais q u e Patajali et autres connaissaient par la t r a d i t i o n g r a m m a t i c a l e avant m m e d'en r e t r o u v e r la t r a c e dans les stra. Ceci ne vaut n a t u r e l l e m e n t q u e p o u r les paribhs a u t h e n t i q u e s , celles q u ' a connues Pnini. En ce q u i concerne les fausses paribhs, le japaka n'est q u e le p r t e x t e q u i leur a d o n n naissance, prtexte constitu par un fait mal interprt, partir duquel on tablissait une maxime d'une extension suffisante pour recouvrir tous les cas de la mme espce qui ne s'accordaient pas autrement avec Pnini. C'est de mme que de nos jours les juristes sont obligs d ' " interprter" les codes pour les adapter des situations que le lgislateur n'avait pas prvues. Tout comme les institutions et les murs les langues voluent; des formes nouvelles s'imposent qui ne tiennent pas compte des grammairiens. Mais il tait admis que les stra du rsi Pnini dtenaient la clef du langage. Il suffisait de savoir les interprter. L'aide des maximes traditionnelles ne le permettait pas toujours : de l les fausses paribhs.
106 Kielh. (prf. p. v, n. 3) remarquant que Patanjali considre exceptionnellement comme japaka un mot ou procd que la paribhs correspondante ne rend pas caritrtha (par exemple l'emploi de l'anubandha n dans deux Sivastra, jap. de I, le mot apratyayah de 1,1, 69, jap. de 19 observe que dans ce cas le mot japayati est employ par Patajali dans un sens moins limit. Il est probable que japayati est toujours employ dans le mme sens et que ce n'est jamais un sens limit. Il semble que ce soit Ngoj. qui ait cherch systmatiser la notion de japaka. Dans son commentaire de la paribhs I, par exemple, il a soin d'employer le mot bodhyate (non jnpyote), parce que cette paribhs ne possde pas un

japaka selon l'ide rigoureuse qu'il s'en est faite (cf. Kielh., p. 2, n. 3). On a vu d'autre part qu'il n'admet pas que les rgles 5, 2, 51-53 soient japaka de 1,1, 23, parce que, tant indispensables comme jap. de l'adjonction du suffixe dat aprs pga, etc., elles ne seraient pas vyartha mme en l'absence de l'infrence que leur attribue Patajali. 107 Notamment PS., paribh. 50 (p. 45,
1.2: kirn ca prvatrety asya pratyaksatvena tennumniky asy bdha

evodtah), o Ng. oppose la paribhs, qui repose sur une inference (anumnikl) et le stra 8, 2 , 1 , objet de perception, celui-ci par suite l'emportant sur celle-l (cf. PS., p. 45,1. 9 etsuiv.).

K A . ^ Subramania Iyer (born 1896)

A unique place in the grammatical tradition is occupied by the grammatical philosopher Bhartrhari. His commentary on Patajali's Mahbhsya and his Vkyapadya were already known to ITsing (pages 15-16), but his work is only now beginning to receive due attention again. The reason is two-fold. On the one hand, only one incomplete manuscript of the commentary on the Mahbhsya is known; it is now preserved in Berlin. Kielhorn published a few fragments of this manuscript, and also established that Bhartrhari " had access to a very extensive commentational literature, superseded by his own work as his own writings have in turn been superseded by those of later authors" (Preface to the second volume of the first Mahbhsya edition: Kielhorn 1880-1885; this Preface has unfortunately been omitted from the third revised edition of 1965). The first volume of an edition of the Berlin manuscript has only recently been published for the first time (Swaminathan 1965). Bhartrhari's Vakyapadya, on the other hand, was not only badly edited and published in parts; it remained largely unintelligible. This was partly due to the fact that this text, unlike most other grammatical works, disappeared from the traditional curriculum at an early date and had rarely been studied. The foremost authority on the VkyapadJya is K. A. Subramania Iyer (born in Kerala in 1896), who studied in Paris with Sylvain Lvi and Foucher, and in London with Barnett, Rhys Davids, and de la Valle Poussin (a refugee in London during the First World War). Subramania Iyer, who for many years has been Professor of Sanskrit at Lucknow University, has occupied many distinguished positions in the Indian academic world. He has edited parts of the text of the Vkyapadya (Kndal, Poona1966; Knda III, P a r t i , Poona1963), translated the first chapter (Poona 1965) and published a monograph on Bhartrhari (Poona 1969). Earlier, when working with the published texts then available to him, containing " many passages where the text was very doubtful and did not make much sense" (1963, vii), he published some articles on the Vkyapadya, namely, "The Conception of Guna among the Vaiyykaranas," New Indian Antiquary (5,1942,121-130; dealing with the philosophical notion of guna 'quality* as refined by the grammarians, not with the linguistic notion of guna as introduced in Panini1.1.2);and "The Point of View of the Vaiyakaranas," Journal of Oriental Research (18,1948, 84-96, Madras University). The latter is reproduced here with the omission of almost all of the textual material quoted in the footnotes and with the addition of some translations within square brackets. "After failing for many years to attract the attention due to such an important work, the Vkyapadya is now at last coming into its own " (Gray 1968, 70); see, apart from Subramania Iyer's works and Brough (1951b; see pages 40l-423ofthis volume), Ruegg (1959), Sastri (1959), Bhattacharya(1962), Pandeya(1963), Kunjunni Raja (1963), Biardeau (1964a and 1964b), Joshi (1967), and Aklujkar (1969). For reviews of Pandeya and Kunjunni Raja see Staal (1966b). For a general survey of the Sanskrit philosophy of language, see Staal (1969).

The Point of View of the Vaiykaranas* (1948)

K. A. Subramania Iyer

To explain the forms of a language is the main purpose of the science of grammar. For this, it is necessary to isolate and analyse the notions which are expressed by the forms ofthat language. As these notions are, to a great extent, the products of the social factors which govern that language, they may or may not be the same as those of other languages. Grammar is not the only discipline which is concerned with notions. Logic and philosophy are equally concerned with them. Hence the influence of logic and philosophy on grammar in the West until the 19th century when the method of observation of facts, already in operation in the physical sciences, was extended to linguistic phenomena. In India, as elsewhere, logic and philosophy share with grammar a partly common vocabulary. Countless are the passages in Sanskrit grammatical literature where the concepts of this or that system of philosophy are brought in for explaining the facts of the Sanskrit language. And yet our grammarians knew that the point of view of grammar was quite distinct from that of the systems of philosophy. This distinction is pointed out by Helrja on many occasions in his commentary on the Vkyapadya. But, before we present his views on the subject, it is necessary to study the few passages in the Mahbhsyaon which the views of Helrja are based. Vykarana is a Vednga, but it is not attached to any particular Veda. It is common to all of them. Patajali makes this point clear in connection with the use of the word bahulam ['variously'] in P. II. 1.57 and P. VI. 3.14. Though the word bahulam found in the stras is traditionally interpreted in four ways1 so as to include all the facts and details which can come under the stras in question, Pnini actually mentions some of these details in the subsequent stras. In addition to the word bahulam he sometimes uses other words like v, ubhayath, anyatarasym, aikasm. Patajali explains the diversity of usage by saying that the science of Vykarana is common to all the Vedas and this diversity of usage is necessary to cover all the facts found in the various branches of the Vedas.2 All that we have to note here is that, according to Patajali, the science of grammar is not attached to any particular Veda or to any branch of it, but is common to all. It is sarvavedapdrisada. We will see, in a little while, whatform this idea assumes in the later grammatical literature. Another statement of Patajali which throws light on the grammarian's point of view is found in the Bhsya on vt. 13, of the Paspashnika. To the objection that if the knowledge of the correct word leads to spiritual merit, a knowledge of corrupt forms, inevitable in a close study of grammar, must necessarily lead to demerit, Patajali answers:sabdapramnak vayam\yacchabda aha tad asmkam pramnam/sabdas ca sabdajnne dharmam aha npasabdajnne dharmam/3 [' We go by the authority of sabda. What sabda says is our authority. And sabda says that merit accrues from the knowledge of correct expressions, not that demerit accrues from the knowledge of incorrect expressions ' ] . Here Patajali means the Vedas by the word sabda and refers to the well known sruti: ekah sabdah samyagjntah sstrnvitah suprayuktah svarge loke kmadhug bhavati/4 [one word properly
* Paper read in the Classical Sanskrit Section of the 15th All-India Oriental Conference, Bombay. 1 Nysaon Pnini III.3.1.
2 3 4

M. Bhsya on 11.1.58(57). M. Bh. onVt13, Paspashnika. M. Bh. on P. VI. 1.84.

394 K. A. Subramania Iyer

known and well pronounced n accordance w i t h the t h e o r y will grant wishes in heaven ' ] . The sruti speaks only of merit resulting f r o m a knowledge of the correct words, and not of demerit due t o a knowledge of incorrect ones. But in another similar context containing the same words, sabda does not mean the Vedas. It means merely the w o r d in general. The point is raised w h e t h e r in the sentence ayam dando hornena [' this (is a) stick, take w i t h it ' ] , the danda is the agent (kart) of the action o f being 1 , the meaning of the w o r d osti which is understood here, or the instrument of the action of taking (hara). Against the view that it is the f o r m e r , the objection is raised that it is, after all, w i t h the danda that the action of taking is done and that, therefore, it should be considered t o be the instrument of that action rather than as the agent of an action which is not mentioned in the sentence. To this objection the answer is given that, for grammarians, it is the w o r d which is pramna, a u t h o r i t y . W h a t ever the w o r d presents, they accept. 5 In the sentence in question, the words as they stand, present the danda as the agent of the action o f " b e i n g " , which, though not mentioned, is understood here, as in all o t h e r cases w h e r e no other action is openly expressed. Thus the danda is, at first, the agent of the action o f being' and then only does it become the instrument of the action of taking. This is at least the case if we go by what the words present, apart f r o m what the reality may be. Thus we see t w o ideas in Patajali : (1) that Vykarana is not confined t o any particular Veda: (2) that the grammarians go by what the words present rather than by how things really are. By sabda, Patajali means sometimes the Vedas and sometimes the w o r d in general. These t w o ideas are made use of frequently by Helrja w h i l e explaining Bhartrhari's Vkyapadya and we need not doubt that they w e r e quite familiar t o Bhartrhari himself. But both these ideas have undergone modifications: The idea that the science of grammar is common t o all the Vedas is changed into the idea that it is common t o all the systems of philosophy. It is found that Bhartrhari in his Vkyapadiya often expounds a grammatical doctrine, not only f r o m his o w n point of view, but also f r o m t h e point of view of some system of philosophy or other. Often an idea belonging t o some system of philosophy is made use of t o explain and justify a particular idea of grammar. The question arises : W h y should he do it ? W h y did he not content himself w i t h explaining it as an Advaitin which he was? Helrja justifies this by saying that Vykarana as a sastra belongs t o all the disciplines. If linguistic facts can be explained f r o m as many points of view as possible, so much the better for the science of grammar. This does not preclude a particular author f r o m having a preference for his own point of view. Bhartrhari, for instance, has a preference for t h e Advaitic point of view, and he has t r i e d t o explain most of the facts and notions of grammar f r o m that point of view. 6 But his w o r k is remarkable for t h e bringing in of o t h e r points of view on
5 sabdapramnaka vayom / yac chabda word says is our authority. And the aha tod asmkom pramnam j sabda word says here that something is ceha sattdm aha - ayam danda / this (is a) stick; " is" is implied. That astlti gamyate \ sa dandah kart bhstick which is agent, when combined tvnyena sabdenbhisambadhyamnah with another word, becomes instrukaranam sampadyate[' we go by the ment']. M.h. on P. II, 1.1. 6 authority of the word. What the Helrja on Vk III. Ja. verse 35.

395 Point of View of the Vaiykaranas

many occasions. A f e w examples w i l l make this point clear: It is the view of the Vaiykaranas that, when words are used, three things are understood by us:(1) the f o r m of the w o r d , consisting of an entity over and above the sequence of sounds heard; (2) the meaning; (3) the intention of the speaker. 7 O f these, the first one is closest t o the w o r d . It is understood by the hearer in any case, even if he does not understand the second and the t h i r d . Between this and the second meaning the relation is vcyavcakabhva. The first is vcoka [' expressing '] and the second is vcya ['expressed ' ] . This is usually understood as referring t o the objects of the w o r l d . Between the first and t h e t h i r d , the intention of the speaker, the relation is said t o be kryakranabhva [' relation between cause and effect ' ] . It is the intention of the speaker which calls up particular words for use. They are, therefore, looked upon as the effects o f t h a t intention. 8 The question now arises : W h y should Bhartrhari speak about kryakranabhva at all, considering that grammar, as a science, is chiefly concerned w i t h the other relation namely, vcyavcakabhva, between the w o r d and the meaning ? The answer given is that Vykarana as a discipline is common t o all the systems of philosophy. Its notions and explanations must be such that they can be acceptable t o the followers of all the systems of philosophy. 9 Some hold that the w o r d does not point t o any external object, but only refers t o the intention of the speaker. It is in order t o respect t h e i r views (tanmatopaskrrtham) that Bhartrhari speaks about kryakranabhva. If meaning is nothing more than the intention of the speaker, the relation between the w o r d and the meaning is naturally kryakranabhva, because it is the intention of the speaker which calls up this w o r d or that in speech. 10 A n o t h e r doctrine which is expounded in the Vkyapadiya is sattdvaita, the view that all words, nay, even parts of a w o r d like roots and suffixes, ultimately have satt or " B e i n g " as t h e i r meaning. This ' Being' is the Supreme Universal which is found in all the objects of the w o r l d and which binds them all together in one reality. The distinctive features of each object are comparatively unreal. In this view, even negative entities are credited w i t h a kind o f Being'. 11 This ' Being' is essentially identical w i t h Brahman. It is clear that Bhartrhari here speaks as an A d v a i t i n . But he f u r t h e r points out that the Snkhya Philosophy is also in keeping w i t h the doctrine o sattdvaita. According t o this system, the first e v o l u t e f r o m Prakrti is Mahat or Buddhitattva, as it is called. It is t o be regarded as " t h e most universal stage which comprehends w i t h i n it all the buddhis of individuals and potentially all the matter of which the gross w o r l d is f o r m e d . Looked at f r o m this point of view, it has the widest and the most universal existence, comprising all creation and is thus called ' mahat', (the great one)". 1 2 A l l the o t h e r evolutes proceed f r o m this principle and are absorbed into it at the t i m e of dissolution. This great principle is essentially ' Being' and all the evolutes proceeding f r o m it share this ' Being'. Thus the Snkhya system also, Helrja points out, favours the doctrine of sattdvait.13 Side by side w i t h the view that the science of grammar is sarvaprsada and, therefore, brings in notions and ideas c u r r e n t in
7

Helrja on Vk. III. Sam. verse 1. Helrja on Vk. III. Sam. verse 1. 9 Helrja on Vk. III. Sam. verse 1. 10 Helrja on Vk. III. Sam. verse 1.
8

Helrja on Vk. III. Ja. verse 34. Das Gupta. A History of Indian Philosophy, Vol. I, p. 249. 13 Hel. on Vk. III. Ja. verse 34.
12

11

396 K. A. Subramania Iyer

other sastras t o explain grammatical notions, t h e r e is the other view that grammar is not bound t o accept an idea simply because it is current in some o t h e r discipline or in the w o r l d . W h e t h e r such an idea should be made use of by Vykarana is a matter of convenience only. It was utilised if it was convenient t o do so. O t h e r w i s e it was not. Thus t h e Vaisesika conception of Guna 14 has been utilised by Vykarana in explaining some f o r m s : T h e w o r d gunavacana occurring in P. IV. 1.44; V. 1.124; in a vr. on V. 2, 94. and in VIII. 1.12. refers t o this Vaisesika conception. 1 5 But this conception is not enough t o explain all the forms which occur in the Sanskrit language. A n o t h e r conception of it is mentioned here and t h e r e in the Bhsya and it is explained in the gunasamuddesa of the 3rd knda of the Vkyapadya. It is a conception peculiar t o Vykarana and it is derived f r o m the forms of the Sanskrit language, and it is meant t o explain them. 1 6 Similarly, the popular conception o f I inga' : stanakesavatl strl syl lomasah purusah smrtah / ubhayor antaram yac ca tadabhve napumsakam / / 1 7

[' women possess (long) hair and breasts, men are known for hairiness, / what is different from both and does not have these (qualities) is neuter'] is found inadequate to explain all the diversity of forms relating to I inga found in the Sanskrit language. Patajali therefore rejects it and says: tasmn na vaiykaranaih sakyam
laukikam Ungarn asthtum / avasyam ca kascit svakrtnta stheyah // 18

['therefore the ordinary concept of inga cannot be used by grammarians; and some self-made notion has to be set up'] that Vaiykaranas must evolve their own conception o f linga' and proceeds to do so. This idea that Vaiykaranas have a right to evolve their own notions is expressed by later writers also. P. I. 2.58 teaches the use of the plural number after a word which primarily expresses jti ['genus ; class ']. But for this stra, only the singular number could be used, because jti is one and so it would be normal to use the singular number. Now one can say, brhmanah pjyah [' a brahman should be honored '] or brhmanah pjyh [' brahmans should be honored '] in the same sense. To this somebody objects that it is wrong to say that ' jti ' is one. Jti has no number at all. It is the dravyaor ' vyakti ' in which the jti resides which has number.19 This is answered by saying that in these matters grammarians do not accept the views of other sastras. They evolve their own notions.20 To them jti is one because the word presents it as such, and it is, therefore, natural to use the singular number after a word expressive of it. This reference in the Bhsya and in the later literature to svakrtnta ['self-made (notion)'], the particular doctrine of the Vaiykaranas suggests that they have their own point of view from which their doctrines and notions are derived. What this point of view is has been indicated by Patajali in that passage
where he says : sabdapramnak vayam / yac chabda ha tad

asmkam pramnam / [' we go by the authority of the word. What the word says is our authority']. The idea contained in this passage has been utilised by Helrja very frequently in his commentary on the Vkyapadiyam. For the grammarian, 'artha' does not mean the external reality but whatever the word brings to the mind.
Kanda-Vaisesika-Stral.1.16. Kanda-Vaisesika-Stra 1.1.16. 16 See the author's paper on " T h e Conception of Guna among the Vaiykaranas"(N. I. A., Vol. V., No. 6,
15 14

Sept.1942). M. Bh. on P. IV. 18 M. Bh. on P. IV. 19 Helrja on Vk. 20 Helrja on Vk.

17

1.3. 1.3. III. Ja. verse 52. III. Ja. verse 52.

397 Point of View of the Vaiykaranas

A r t h a does not mean vastvartha but sabdrtha, not reality, but the meanings of words. Individual words bring something t o the mind and the sentence as a whole also brings something t o the mind. Both these things are included in the expression 'sabdrtha'. Grammar studies both these things in order t o evolve notions which will explain the forms of the language. Grammar is satisfied if these notions conform t o what we understand f r o m words, no matter w h e t h e r they conform t o reality or not. Grammar does not look at reality directly in theface. As Helrja puts i t :

sabdapramnaknm hi sabda eva hi yathrtham abhidhatte tathaiva tasybhidhnam upapannam / na tu vastumukhaproksatay / 21 ['for to those whose authority is the word, the word designates what it corresponds to, and its designation is accordingly appropriate; but it is not for looking reality directly in theface']. Not to look at reality directly in the face is as good as not philosophising and Helrja sometimes makes it quite clear that the grammarian is really not concerned with philosophy proper. Thus while explaining the different conceptions of Time mentioned by Bhartrhari in the Klasamuddesa such as that it is an entity which exists apart from the mind or that it is a mere construction of the human mind, Helrja says that Bhartrhari is not really concerned with what time is philosophically, but that he is anxious to examine and analyse that something which is responsible for our putting the Sanskrit verb in different tenses as in abht ['was'], asti [' is'] and bhavisyati ['will be']. That something may not be able to stand close philosophical scrutiny, but if it serves the purpose of explaining the different tenses, one would have to accept it.22 Similarly in the kriysamuddesa, the question is: What is action? The answer given by Bhartrhari on thebasis of the Bhsya passages is that it is a process, something having parts arranged in a temporal sequence. It is not directly perceptible, but it is to be inferred. Each moment or part may be looked upon as action, in which case, it will also be inferrible only and not directly perceptible. These parts may be further subdivided and the smaller parts will also be actions. There will come a time when the part cannot be further sub-divided. It cannot then be called action at all. Such an atomic point may be directly perceptible but that will not make actions so because such a point cannot be called action at all. Only that can be called action which has parts arranged in a temporal sequence. After having clearly explained all this, Helrja adds that for grammarians the real question is not whether an action has actually parts or not, but whether the verb presents it as such. The answer is that verbs do present action, however momentary, in nature, as something having parts which cannot co-exist, but are arranged in a temporal sequence. And Vaiykaranas go by what the words present to us.23 It is pointed out that a notion arrived at by the Vaiykaranas from their own point of view, may sometimes agree with popular ideas rather than with those accepted by some system of philosophy. Vaisesikas think of the whole, the avayav, as existing in the parts, theavayavas. That is an idea to which they have come by a logical analysis of reality. They also specify the particular relation by which the whole exists in the parts. It is samavya, inherence. The popular conception, however, is that the horn of a cow exists in the cow and not vice versa. The part exists in the whole. Hel21 22

Helrja on Vk. III. Sam. verse 66. Helrja on Vk. III. K. 58.

23

Helrja on Vk. III. Kri.10.

398 K. A. Subramania Iyer

raja points out that the language follows the popular conception in this matter. In the expression gavi srgam [ ' t h e horn is in the c o w ' ] the locative suffix is affixed t o the w o r d go [ ' c o w ' ] which denotes the whole of which srga is a part. If one followed the Vaisesikas in this matter, one would have t o say srge gauh [ ' t h e cow is in the horn '] and skhym vrksah [ ' t h e t r e e is in the branch']. 2 4 It is mainly this point of view which the Vaiykaranas adopt in defining the various grammatical categories such as the different krakas, gender, number, person, aspect (upagroha) etc. dealt w i t h in the 3rd knda of the Vkyapadya. This naturally results in certain distinctive notions. It w i l l not be out of place here t o draw the attention of the reader t o a few of these distinctive notions. Regarding the meaning of individual words, t h e r e are t w o views c u r r e n t among grammarians, associated w i t h the names of t w o ancient grammarians mentioned in the Vrttikas of Ktyyana, namely, Vjapyyanaand Vydi. One view is that all words, nay, even parts of words, denote primarily the Universal and only secondarily the Particular. If we apply the same w o r d , say, ' cow ' or ' t r e e ' t o a large number of objects it is because we see some common characteristic in all of t h e m . This common characteristic is the universal or 'jti ' as it is called. The w o r d 'asva' [' horse'], therefore, primarily denotes 'asvatva' [' horseness'], the w o r d ' g o ' [' cow 1 ] denotes gotva [' cowness '] and so on. If this is t r u e , then the w o r d 'jti ' must also denote a universal present as a common characteristic in all the universals. But such a view goes against the Vaisesika view that there cannot be a universal in a universal. They argue that t o accept a universal in a universal, would lead t o anavasth or ' regressus ad i n f m i t u m ' . W h e r e w o u l d one stop in the process of postulating universals ? W h y not postulate a t h i r d universal in the second one and so on ?The best t h i n g w o u l d be t o stop at the very first universal and not go any f u r t h e r . But the grammarian replies that this kind of reasoning might be all right f r o m the Vaisesika point of view, but not for himself. He has his own point of view. His chief concern is t o find out the nature of meanings conveyed b y w o r d s . W h a t he finds isthat in all universals as conveyed by words, t h e r e is a common characteristic which can be looked upon as another universal. The existence of the first universal was postulated just because a common characteristic was experienced in the individuals or particulars, followed by the use of the same name t o all the individuals. A similar common characteristic is experienced in all the universals as conveyed by words and that justifies the use of the w o r d 'jti ' for all of t h e m . W h e r e t h e r e is identity of cognition and of name, a universal has t o be postulated and in the universals as presented by words t h e r e are both. A n d for grammarians, it is what words convey which matters. 25 If we go by what words present, there can be not only a universal in a universal but many o t h e r things which are o r d i narily looked upon as guna [ ' q u a l i t y ' ] or kriy [ ' a c t i v i t y ' ] may t u r n o u t t o be universals. If a guna is presented by words as something which persists as a common feature in many things, it becomes a universal for the grammarians. 26 That is probably the reason why Pnini has applied the w o r d smnyato the meaning expressed by the w o r d 'syma' " d a r k " in the compound 'sastrl24

Helrja on Vk. III. Ja. 11.

" Helrja on Vk. III. Ja. 11.

399 Point of View of the Vaiykaranas

syama' " d a r k like a k n i f e " formed according t o the sutra upamanni smnyavacanaih.21 Samnya is another name for ' j t i ' . In the compound sastnsym the w o r d syma expresses a guna and, as it expresses a common feature between a sastrl and whatever is compared t o it, the grammarians look upon it as jti. Similarly, action can be presented as jti by words. W e use the expression ' pacati 1 , ' he cooks', in a variety of circumstances. The person w h o cooks, the thing cooked, the fuel and the utensils used for cooking, may all be different and yet the notion of cooking and the expression ' pacati ' persist. This also shows that the w o r d presents action as 'jti'. 2 8 Even though philosophers like Vaisesikas make a distinction between jti, guna, kriy, and dravya, the grammarians believe that it is all a question of how words present them and the words can present the first t h r e e also as dravya [' substance ' ] . They have t h e i r o w n definition of dravya. A n y t h i n g which is presented by words as something t o be characterised, distinguished f r o m o t h e r things (bhedyo) is a dravya. W h a t e v e r can be referred t o by the demonstrative pronouns dam " t h i s " and tad " t h a t " is a dravya:

vastpalaksanam yatra sarvanma prayujyate / dravyam ity ucyate so1 rtho bhedyatvena vivaksitah /'/ Vk. III. Dra. verse 3. "where a pronoun is used to refer to a thing it refers to a substance and its meaning is expressed by differentiation (from other things) ". If words present 'jti ' as something to be differentiated, as a visesya, then it becomes a dravya: sarvanmapratyavamarsayogyatvam, the fitness to be referred to by a pronoun as 'this' is the characteristic of a dravya. This view is traced by Helrja to so ancient a writer as Yska, whose statement: ada iti y at pratlyate tad dravyam "what is inferred from 'that' is a substance" is quoted by him.29 The expression is significant because it makes the whole thing dependent upon the ' vivaks' of the speaker. It makes it quite clear that what is defined in the verse quoted above is not the artha called dravya, but the meaning of the word 'dravya' or rather the nature of the thing presented by the word 'dravya'.30 Anything can be presented by words as something to be differentiated. For instance, movement or action is so presented in "sukham sthlyate" [' it is pleasant to stand '] where the action of standing is presented as athing and it is determined or modified by the word 'sukham'. The meaning of the verb "sthTyate" is, therefore, dravya. In the sentence, "suklataram rpam " ['the color is whiter'] a quality is so presented. Here ' rpa', though a quality, is presented by words as athing to be qualified or determined by white. It is, therefore, a dravya. The action in ' sthTyate ' can be referred to by the pronoun " kim ", another indication that it is a dravya. It is easy to see that this conception of dravya is very different from the Vaisesika conception of it. The essence, then, of the Vaiykarana point of view is that it does not look at reality in the face directly, but only at reality as presented by words. But while they knew this distinctiveness of their point of view, their literature is full of passages which make one wonder whether they did not often forget this and indulge in a direct analysis of reality. Whether they are discussing the nature of'jti ' or 'guna' or ' kriy', their language often makes one think that they are discussing, not reality as presented by words,
26 27 28

Ibid. P. 11.1 55. Helrja on Vk. III. Ja. verseil.

Vk. IM. Kri. 20. 29 Helrja on Vk. IIII. Dra. verse 3. 30 Helrja on Vk. IIII. Dra. verse 3.

400 K. A. Subramania Iyer

but reality itself. It is true that a writer like Helaraja frequently reminds himself and his readers that, for grammarians, artha is 'sabdrtha' and not ' vastvartha'. But this frequent reminder to himself is perhaps the best proof that it is not easy to discuss the nature of sabdrtha, without unconsciously straying into a consideration of the nature of'vastvartha'. Perhaps the fact that the word 'artha' in Sanskrit can and does mean both 'vastvartha' and 'sabdrtha' also made it difficult to separate the two. Some of the problems discussed in Vaiykarana literature and the answers given also show that the grammarians did not always succeed in keeping the two kinds of'artha' absolutely distinct. One set of such questions which they have discussed relates to action. What is action ? Is it perceptible or can it only be inferred ? Is there such a thing as action apart from that which is active? And they are answered as follows. Action is something which has parts arranged in a temporal sequence. It cannot be directly seen but has to be inferred. It is quite distinct from ' dravya'. These questions are more appropriate to philosophy than to grammar. The same thing can be said of the grammarians' treatment of gender. In languages like English, we have two words, sex and gender, to denote the distinction found in the objects of the world and that found in words respectively. In Sanskrit, ' linga' has to denote both, and this fact may have ultimately led to the grammarians coming to the conclusion that what is called ' liga1 is a property of things and not of words. And they have invoked the Snkhya philosophy in determining this property of things. The idea that it is a property of words was also known to them. They discuss it only to reject it. Thus Vykarana oscillates between philosophy and linguistics, while it is conscious all the time that its proper sphere is something distinct from that of philosophy.

John B r o u g h ( b o r n 1917)

During the century and a half after Colebrooke, the Sanskrit grammarians were studied in England by continentals (Kielhorn and Goldstcker; cf. Bhandarkar, this volume, page 82). In such a handbook as The Sanskrit Language (1955) by T. Burrow, the Boden Professor of Sanskrit at Oxford, Pnini is referred to only n passing. This situation began to change when John Brough (born 1917), who had studied at Edinburgh and Cambridge and who became Professor of Sanskrit, first at London and now at Cambridge, directed attention to the Sanskrit grammarians. Brough published three articles within the span of three years, in which he made some use of contemporary linguistics as it was then being developed by his colleague at London, J. R. Firth. Brough writes about the problems discussed by Indian grammarians: " it is to a large extent the rediscovery of these problems by modern linguistics during the last twenty orthirty yearswhich renders possible a better understanding of the Indian theories." Brough's articles deal mainly with Patajali and Bhartrhari. During the same period another English scholar, W . S. Allen, directed attention to the Prtiskhya and Siks literature, which resulted in Phonetics in Ancient India (1953). This work was in fact preceded by S. Varma's Critical Studies in the Phonetic Observations of Indian Grammarians (1929). Brough dealt with the stories and legends connected with the origin and development of the Indian grammarians (such as were related by Trantha, see pages 25-26) in a different context. In his Selections from Classical Sanskrit Literature (1951) he published and translated a relatively little-known section of the Haracaritacintmani (of the twelfth century) dealing with the origin of grammar (pages 2-21). This episode, entertaining and eminently readable as it is, is beyond the scope of this volume. In "Audumbaryana's Theory of Language," Bulletin of the School of Oriental and African Studies (14,1952, 73-77) Brough dealt with Bhartrhari's thesis that language consists of sentences, while words are merely artificial abstractions. He made it seem probable that this thesis was already held by Audumbaryana, an ancient authority referred to in the Nirukta. In "Theories of General Linguistics in the Sanskrit Grammarians," Transactions of the Philological Society (1951, 27-46), which is reproduced here, Brough discusses the notion of sphota, the distinction between use and mention (or antonymy) and other fundamental semantic concepts. For a recent study on sphota, together with the text and translation of a seventeenth-century text on semantics, see Joshi (1967). The passage dealing with the definition of sabda occurs on Plate V. Brough's emendation apratapadrthaka is not supported by the recently published text of Bhartrhari's MahbhsyatJk (Swaminathan 1965, 5).

A. Theories of General Linguistics in the Sanskrit Grammarians (I9SI)

John Brough

It is well known that the discovery of Sanskrit by the West at the end ofthe18th century provided the operative stimulus for the development of the comparative study of the Indo-European languages. It has also been recognized that the Pninean analysis of Sanskrit into a system of roots, stems, and suffixes pointed the way to the method which has prevailed in Indo-European studies to the present day. It is true that roots and suffixes were not entirely new concepts to Europe, but it remains doubtful whether the method would have been applied with such thoroughness if it had not been for Pnini's example. It is customary to add at this point the deprecatory remark that Pnini was, of course, aided in his analysis by the extraordinary clarity of structure of the Sanskrit language; but we are apt to overlook the possibility that this structure might not have seemed so clear and obvious to us if Pnini had not analysed it for us. But while we in the West have acknowledged a debt to Pnini in the matter of formal analysis, we have paid less attention to the theories of general linguistics and linguistic philosophy to which the Indians devoted much thought. 1 I wish in this paper to discuss a few of the most important aspects of the Indian theory, not merely as a matter of antiquarian curiosity, but because in their extraordinary linguistic and philosophic acumen these ancient authors are still, I believe, worthy of our respect.2 As Bhartrhari himself puts it, the Goddess of Learning does not smile on those who neglect the ancients.3 As has been said from time to time, 4 linguistics is faced at the outset with the difficulty that in making statements about language, language is apparently turned back on itself. At first sight the situation is similar to the old philosophical problem of how consciousness can be conscious of itself. The difficulty here, however, is not, I believe, a fundamental philosophical one. It is merely that the habits of everyday language make it troublesome to state the matter in a clear manner. At the same time, everyday language has already provided us with the method of dealing with linguistic facts, by permitting words not only to be used but also to be quoted. The point is so obvious and familiar that it is difficult to realize its cardinal importance, and even to be aware of it in ordinary grammatical discourse. The discussion of this topic in
It is clear from scattered references that many of these problems were discussed as early as Yskaand Pnini; and the Samgraha of Vydi, known only from quotations, and Patajali's Mahbhsya both dealt incidentally with such topics. The earliest extant work specifically devoted to the philosophy of grammar is the Vkyapadlya (VP) of Bhartrhari (c. 7th century A.D.). Later writers elaborated Bhartrhari's exposition, but while continuing the tradition of "the Grammarians " as a definite school of thought opposed to the standard philosophical schools, added little of first-rate significance. 2 I do not, of course, wish to exaggerate in this matter. The majority of
1

the problems discussed by the Indians are in fact familiar in modern linguistics and logic (though still perhaps insufficiently appreciated by many professional teachers of languages); and it is to a large extent the rediscovery of these problems by modern linguistics during the past twenty or thirty years which renders possible a better understanding of the Indian theories. In this connection I am particularly grateful to my colleague Professor J. R. Firth, with whom I have had numerous Informal discussions on these topics. 3 VP .493, anupsita-vrddhnm vidyntiprasdati. 4 For example, J. R. Firth, IPS., 1948, p. 128.

403

General Linguistics in the Sanskrit Grammarians

Sanskrit centres round Panini's statement5 svam rupam sabdasysabdasamjn, "A word (in a grammatical rule) which is not a technical term denotes its own form." The immediately obvious interpretation is, of course, that a statement about a particular word is not to be taken as applying also to its synonyms. Thus, Kenou explains it,6 "(... il s'agit) du mot en tant que forme propre, (non en tant que porteur d'un sens, autrement dit: qu'il n'englobe pas les mots de mmes sens que lui)." This follows closely Ktyyana's preliminary statement (vrttika 1): sabdenrzhagater, arthasysambhavt, tadvcinah samjnpratisedhrtham svamrpagrahanam. "Since by the word is understood the thing-meant,7 and since the thing-meant is impossible (in this context), the expression ' its own form ' is to prevent the word being taken as a name denoting any word which expresses the same thingmeant." As Patajali says, with a touch of humour, when we say " Fetch the cow", or " Eat the curds", it is a thing which is fetched, and a thing which is eaten ; but when we say " Agni (fire) has the suffix-eya",8 we clearly cannot add the suffix to the embers.9 But since the use of the word agni in the stra brings to mind the fire, the inference might then be that "words meaning fire" are intended. This, however, seems to be merely a primafacie view (prvapaksa), since Ktyyana continues in vrttikal: na v, sabdaprvako hy arthe sampratyayas, tasmd arthanivrttih. "Alternatively, this is not the real intention of the stra, for the understanding of the thing-meant is preceded by the word ; hence, in the grammatical context, the thing-meant is ruled out of court." The question of synonyms therefore does not arise, since when the word agni is used in the grammatical statement the signatum is not the fire, but is (apparently) the same as the signans. The difference in logical status is obvious if we compare "A sheep is grazing in the field", with " 'Sheep'10 has an anomalous plural". The svam rupam reference (in the logician's terminology, the autonymous use of a word) is conventionally marked in our writing by quotation-marks. As Bhartrhari puts it, "Justas a technical term like vrddhi, while linked to its own form, is also attached to what is named by it, viz. the speech-entities symbolized by aaic (i.e. , ai, au), so likewise this word agni (in the stra), while linked to the word agni (in everyday use, i.e. the meaning 'fire'), is also attached to the sound agni, which (in this context) has the word agni as the thing-expressed. The word which is uttered (in ordinary usage) is certainly not the one which partakes of the operation (of adding the suffix). But in conveying this other sense, its power (to

.1.68. La Grammaire de Pdnni (Paris, 1948), p.13. 7 The Sanskrit artha is as many-sided as the English "meaning", and "thingmeant" is a convenient device to indicate what I understand to be the sense of the Sanskrit here. The term was first introduced, I believe, by Sir Alan H. Gardiner (Theory of Speech and Language, p. 29). Its use here, however, does not imply an accep6

tance of Gardiner's consequent application of the term " meaning". 8 Pnini, iv.2.33, agnerdhak. 9 sabdenoccritenrtho gamyate: gm naya dadhy asnety artha niyate 'rthas ca bhujyate. arthasysambhavt: ihavykarane 'rthe kryasysambhavah. agner dhag iti na sakyate 'garebhyah paro dhakkartum (Kielhorn, vol. i, pp. 175-6). 10 Note that English does not permit the autonym to take an article.

404 John Brough

convey the normal sense) is not impeded." 11 This is important, since otherwise grammatical discussion would be futile. Although agni and ' agni' symbolize in different ways, they are indissolubly connected, 'agni1 being the name (somjn) while agni is the thingnamed (samjin). Both the name and the thing-named, it seems, must be understood as members of classes.12 Otherwise extraordinary complications of thought arise. Bhartrhari notes two such possible interpretations : "Some consider that the stra 'A word denotes its own form ' means that the particular (which appears when the grammarian utters the rule agnerdhak) is the name, and that it is the class attached to the particular which (is the thing-named and which) undergoes the grammatical operations. Others hold that it is the particular as the thing-named which is the purport of the stra (and that it is the class which is the name; the grammatical operations being thus attached to the particular, since) in any given instance there appears only a particular whose understanding is brought about by the class."13 It is, of course, easier to be clear in such discussions where the grammar of one language is described in another language. Thus, a translation into English of a Sanskrit passage on Sanskrit grammar shows at once which words are autonymous, since these, of course, remain in Sanskrit. This is also obvious in the case of a two-language dictionary: the heading-words, in the first language, are clearly autonymous (svam rpam), while the defining words in the second language express their own meanings. The distinction, though not so immediately apparent, is, of course, equally present in a one-language dictionary, and in logical and mathematical definitions, as well as in everyday usage ( " W h a t does ' x' mean ?" ; " W h a t did you say?""I said V " ) . 1 4 The importance of this for logic is very considerable.15 It is fundamental in the discussion of the possibility of stating the syntax of a logical language in that language itself, and is important for the theory of a hierarchy of languages.16 But these matters must be reserved for a later discussion.
VP i.59-61 : vrddhydayo yath sabdh svarpopanibandhanh daic-pratyyitaih sabdaih sambandham ynti samjibhih, agni-sabdas tathaivyam agni-sabdanibandhanah agni-srutyaiti sambandham agni-sbdbhidheyay. yo ya uccryate sabdo niyatam na sa kryabhk, anya-pratyyane saktir na tasya pratibadhyate. 12 This is not stated directly, but seems plausible from the relation of normal words with things, VP i.15, yathrtha-jtayh sarvh sabdkrtinibandhanh, "Just as all thing-classes depend upon word-classes . . ."; VP iii.1.6, sv jtih prathamam sabdaih sarvair evabhidhyate, tato 'rthajtirpesu tadadhyropakalpan, "All words first of all express their own class, thereafter they are fictionally
11

superimposed on the forms of classes of things." 13 VP .68, 69: svam rpam ti kaiscittu vyaktih samjopadisyate, vyaktau kryni samsrst jtistu pratipadyate. samjninm vyaktim cchanti stragrhym athpare, jtipratyyit vyaktih pradesespatisthate. 14 Cf. MBh. . p. 176, yo 'pi hy asv hyate nmn, nma yadnena nopalabdham bhavati tad prcchati, kirn bhavn heti. VP i.57, ato 'nirjatarpatvt kirn hety abhidhyate. 15 See for example Carnap, The Logical Syntax of Language, p. 156. 16 Bertrand Russell, An Inquiry into Meaning and Truth, p. 62. Some at least of the paradoxes, however, which are used as arguments for the hierarchy, appear to be capable of a different solution.

405 General Linguistics in the Sanskrit Grammarians

What then can be said of the nature of these words which as linguists we quote and discuss?The definition given at the beginning of the Mahbhsya is well known : "What then is this word (sabda) 'cow' ? . . . It isthat by means of which, when uttered, there arises an understanding of creatures with dewlap, tail, hump, hooves, and horns."17 In this definition, however, the word " uttered " (uccrita) must mean something more than the mere production of sounds, since Patajali proceeds to contrast it with a commonsense view held by non-linguists, namely that the word consists of the actual sounds of the instance. The commentators therefore are doubtless right in understanding the sense to be that the word is the symbol which, when brought to light (abhivyokto) by the pronunciation of the sounds, brings about the understanding of the meaning. The nave view, that the word is the sound, is usually read as follows: atha vpratltapadrthako loke dhvanih sabda ity ucyate, tadyatha, sabdam kuru, ma sabdam krsJh, sabdakry ay am mnavaka iti dhvanim kurvonn evam ucyate. This is perhaps possible if we take the author to mean that the popular view looks on the sound itself as the direct conveyor of the meaning. But pratltopadrthoka merely repeats the sense of sampratyaya in the first definition, and it seems to me that the intended contrast is better brought out if we understand it as apratltapadarthoka. The sense would thus be, "Alternatively, in everyday life the actual sound, which does not itself convey the meaning, is called sabda, e.g. ' Make a noise', ' Don't make a noise', 'This boy is noisy'so it is said when one makes (linguistic?) sounds." The tradition of the commentators interprets the first definition here as concerning the sphota, a term which is normally opposed to the speech-sound, dhvani or nda. This, however, can be accepted only with the reservation that the sphota for Patajali was a rather different conception from that found in the later grammarians, however pardonable the commentators' assumption may be that the fully-fledged theory is already present in the Mahbhsya. This term sphota, which is of prime importance for Indian linguistic theory, has unfortunately been subjected by modern writers to a great deal of unnecessary mystification. Thus, for example, Keith19 describes the sphota as " a mysterious entity, a sort of hypostatization of sound, of which action sounds are manifestations". Similarly, S. K. De20 writes, "Some philosophers propounded and the grammarians took it for granted that a word has intrinsically a word-prototype corresponding to it. The sphota is not exactly this word-prototype, but it may be explained as the sound of a word as a whole, and as conveying a meaning apart from its component letters (varnas). The sphota does not contain exactly the sounds of the word in the order peculiar to the letters, but the sounds or something corresponding to them are blended indistinguishably into a uniform whole. When a word is pronounced its individual sounds become reflected in some degree in the order of the sphota in which the particular sounds are comprised ; and as the last sound dies away, the sphota, in which the idea corresponding to all these sounds is comprised, becomes manifest and raises to our consciousness the idea thus associated. The sounds
17 MBh. i. p. 1. atha gaur ity atra kah sabdah?. . . yenoccritenassnlngla-kakuda-khura-visninm sampratyayo bhavati sa sabdah. 18 Kielhorn, in his edition, adds a punctuation mark after this word,

but this has no authority from the commentators. 19 A. Berriedale Keith, A History of Sanskrit Literature, p. 387.
20

Studies in the History

of

Sanskrit

Poetics, vol. ii, p. 180.

406 John Brough

of a w o r d as a whole, therefore, and apart f r o m those of t h e constituent letters, reveal t h e sphota." It is not surprising that Professor De goes on t o describe this as a " s o m e w h a t mystical c o n c e p t i o n " ; but w i t h all respect, I feel that this statement hardly does justice t o t h e grammarians' t h e o r y ; and, indeed, it is hardly t o be wondered at if t h e western reader, in t h e face of numerous comparable accounts, should come t o t h e conclusion that t h e sphota-theory represents a departure f r o m lucidity w h i c h , coming as it does f r o m men whose professional task was t h e clear presentation of linguistic facts, is quite inexplicable. First, it must be made clear that t h e sphota is not a " h y p o statization of s o u n d " . Its fundamental attachment is t o t h e o t h e r side of t h e linguistic situation, namely, t h e meaning. In its nontechnical sense sphota means simply " a bursting, a splitting o p e n " , and it is normally defined in its linguistic sense as " t h a t f r o m which the meaning bursts f o r t h , i.e. shines f o r t h , in o t h e r words t h e word-as-expressing-a-meaning (vcaka)". 21 The sphota then is simply t h e linguistic sign in its aspect of meaning-bearer (Bedeutungstrger). The suggestion of t h e commentators that Patajali's first definition of sa bd a refers t o t h e sphota is t h e r e f o r e t r u e t o t h e extent that it is t h e meaning-bearer which is in question. But t h e essential nature of t h e meaning-bearer was not seen by Patajali in t h e same light as it was by later thinkers. 2 2 This is clear f r o m his discussion of Pnini i.1.70 (taparas tatklasya), where he explicitly distinguishes speech-sound (dhvani) and sphota. The Pninean system uses t h e convention that a statement about a short vowel appl ies also t o t h e corresponding long vowel (and where necessary the prolated, pluta). W h e r e , however, it is necessary t o restrict a statement t o t h e one length, a t is added. Thus, a = a and a; at = a only. The stra, however, states that t h e t restricts t h e vowel t o the same time-length ; and Ktyyana (vrttika 4) raises t h e possible objection that if a rule containing this t is u t t e r e d in t h e rapid style of diction (druta) it w i l l be necessary t o add that t h e medium (madhyama) and slow (vilambita) styles are also included, since they differ in time-length. This he answers in vrttika 5, by stating that t h e vowels themselves are fixed (avasthit varnh) and that the styles of diction (vrtti) depend upon t h e speech-habits of t h e speaker. Patajali's discussion of this is interesting 2 3 : vakt kascid sv abhidhy bhavati, su varnn abhidhatte, kascc d r e n a kascc cratarena, tadyath: tarn evdhvnam kascid su gacchati kascic drenagacchati kascic ci ratarena gacchati. rathika su gacchaty svikas d r e n a padtis ciratarena.visama upanysah. adhikaranam at rdhv vrajati-kriyyh. tatryuktam
Kondabhatta, Vaiykarana-bhsana (Bombay, 1915), p. 236; Ngesabhatta, Sphotavda (Adyar Library, 1946), p. 5: sphutati praksate 'rtho 'smd ti sphoto vcaka ti yvat. Mdhava, Sarvadarsanasamgraha (ed. Abhyankar, p. 300), gives the double explanation that the sphota is revealed by the letters, and itself reveals the meaning: sphutyate vyajyate varnair ti sphoto varnabhivyagyah, sphutati sphutibhavaty asmdartha iti sphoto 'rthapratyyakah.
21 22

There is no evidence to show that Pnini knew the term sphota or any meaning-theory comparable to the later discussions. The term itself occurs first in the Mahbhsya. Ngesa, it is true (Sphotavda, p. 102), ascribed the doctrine to Sphotyana, who is quoted by Pnini (vi. 1.123) on a point of morphology. But this is rather like ascribing a theory of roots to Racine. 23 Kielhorn's edition, i. p. 181.

407

General Linguistics in the Sanskrit Grammarians

yad adhikaranasyavrddhih rsau sytm.evam tarhi, sphotah sabdo, dhvanib sabdagunah. katham? bheryghtavat.24 tadyath bheryghtah. bherlm hatya kascid vimsati padni gacchati kascit trimsat kascic cat v rim sat. sphotas catvn eva bhavati, dhvanikrt vrddhih. "One speaker is rapid in his utterance, pronounces the sounds rapidly, another is slow, and athird still slower. Similarly, one man travels the same road rapidly, a second slowly, a third still more slowly. Thus, a charioteer goes rapidly, a horseman slowly, and a pedestrian still more slowly. (The pupil objects :) 'The analogy is inexact. In this case the road forms the (unchanging) substratum of the action of going. But in the other case this is not applicable, since the length and shortness belong to the substratum itself.' (The teacher answers :) It is the same in the other case also, the sphota (the unchanging substratum) is the word, the sound is merely an attribute of the word. How? Like a drum-beat. When a drum is struck, one drum-beat may travel twenty feet, another thirty, another forty. But the sphota is of precisely such and such a size, the increase in length is caused by the sound."25 From this context it is clear that the word as the sphota, in Patajali's view, consists of a fixed pattern of letters, with long and short vowels ; and this is confirmed by his statement that "the sphota is of precisely such and such a size". In the more developed theory of Bhartrhari, such an attribution of size to the sphota would be unthinkable. I trust that this explanation does not attribute too much significance to tvn eva. Punyarja, it is true, in commenting on VkyapadJya i.49, uses the expression tvn eva
sphoto vicitrm vrttim anuvidhatte; but I suspect that this is a direct

reminiscence of the Mahbhsya passage, and that it is strictly incompatible with the later theory. The principal argument for my interpretation, however, is that Patajali's problem here is the distinction between vowel-length as ordinarily understood in linguistic analysis, and absolute vowel-length of instances such as might be measured instrumentally. This in the later theory is the distinction between theprkrta-dhvani and the vaikrta-dhvani ; while the sphota in Bhartrhari's sense is really irrelevant to the problem. It would thus seem that Patajali's sphota (except in so far as it is for him the meaning-bearer) is really comparable to Bhartrhari's prkrta-dhvani. The commentators, being acquainted with the later theory, naturally point out that the speed of utterance belongs to the vaikrta-dhvani, but fail to observe that the contrast of the latter with Bhartrhari's sphota does not answer Patajali's problem. For Patajali, then, it would seem that a word as asphota is fundamentally a structure consisting of aseries of consonants and long or short vowels, in other words a structure which can be analysed as a succession of phonematic units. In the same way it is possible for him to talk of the sphota of a single letter (yarna). Thus, in discussing Pnini's statement krpo ro Iah26 (" in the root krp-, r is replaced by /", hence for *karpta we have kalpt, etc.), he points
24

This should be printed as a varttika. The analogy here is clearly suggested by the literal meaning of sphota, so that in the case of the drum it might betaken as the "burst" of sound. But whether or not this is the
25

starting-point of the doctrine, the later discussions, as already noted, are more concerned with the "burst" of meaning occasioned by the linguisticsign. 26 viii.2.18; MBh. vol. i, pp. 25-6.

408 John Brough

o u t that strictly this does not make allowance for such forms as kjpta < *krpta, since only the consonant r is mentioned in the rule, and not the vowel r. He ingeniously suggests that the stra may be analysed as krpa (for krpel) uh rah Iah, where uh and r-ah are respectively the genitives of r and r, while la-h is the nominative of/. This, however, is not completely satisfactory, as / still does not appear. The final solution is : athavobhayatah sphota mat ram nirdisyate; rasruter asrutir bhavatlti, " a l t e r n a t i v e l y , in both cases ( r a n d /) it is only the sphota which is taught in t h e stra, i.e. 'an 27 r-sound is replaced by an /-sound ' . " This can be approximately rendered in modern terminology, " In both cases the phoneme is meant, i.e. 'an allophone of the r-phoneme is replaced by an allophone of the / - p h o n e m e ' . " It is of interest t o observe that Patajali realized that f o r the phonology of Sanskrit it is convenient t o regard r a n d r as belonging t o the same phoneme. Bhartrhari, like Patajali, starts f r o m the observation that the w o r d can be considered under t w o aspects, as sound, or as meaning-bearer (VP i.44) : dvv updnasabdesu sabdau sabdavido viduh eko nimittam sabdnm aparo Y t h e prayujyate. " In meaningful language, linguists recognize t w o (entities which can be called) words : one is the underlying cause of words, the o t h e r is attached t o the meaning." Here the " underlying cause of w o r d s " is clearly t o be interpreted as the abstract sound-pattern which underlies instances of t h e utterance of the w o r d , while t h e o t h e r which this utterance reveals is t h e sphota, which in t u r n gives rise t o t h e meaning. Thus, for Bhartrhari, the sound (dhvani) is something more than the instance, and the sphota, so far f r o m being a time-series pattern is, in fact, repeatedly stated t o have neither t i m e nor parts. The t i m e - o r d e r of t h e dhvani is merely a means (upya) f o r revealing the timeless and indivisible sphota (VPi.48): ndasya kramajtatvn na prvo na paras28 ca sah akramah kramarpena bhedavn ivagrhyate. " T h e fact that the sound is produced serially is no argument for considering t h e sphota t o be (capable of the predicates) ' b e f o r e ' o r ' a f t e r ' : having no order itself, it is only apparently perceived under the disguise of o r d e r and as possessing parts." This concept of the sphota appears t o have arisen under the influence of arguments in t h e philosophical schools. The Nyya philosophers, for example, held that the meaning of a w o r d was presented t o the mind by t h e last sound, aided by t h e memoryimpression of the preceding sounds. This, however, is linguistically unsatisfactory. Even granting the hypothesis that the data are received as aseries of atomic perceptions, it is necessary t o postulate in addition that we remember not only the impressions, but also t h e i r o r d e r ; and even then all that w o u l d be present t o t h e mind is a collection of sounds in agiven order, and not a meaning-bearing
V. Trapp, Die ersten fnf Annikas des Mahbhsyam (Leipzig, 1933), p. 89, seems to me to obscure the matter by importing the definite article: "An Stelle der Gehrsempfindung ' r' t r i t t die Gehrsempfindung ' I ' " . 28 So both India Office MSS.; Benares edition, aparas; Punyarja, prvatraparatva-; cf. MBh. i. p. 355, prva27

parayoh, Laghumanjs, p. 162, ayam prvo 'yam para iti. On the other hand, paurvparya is the regular form; and cf. also (dependent on this) prvparbhvt, MBh. i, 356, vrttika. There is probably a reference here to Pnini i. I. 66-7, tasm/nn iti nlristeprvasya; tasmd ity uttarasya.

409 General Linguistics in the Sanskrit Grammarians

word. The sounds by themselves have clearly no capacity to attach themselves directly to a meaning, otherwise the collection of the first three sounds of the word manage would present to the mind the word man ; and this, in fact, does not happen. To deal with the situation adequately, it is necessary to postulate a meaning-bearer which is not identical with the collection of sounds, but is related to this collection in such away as to be capable of being revealed by it. In other words, for linguistic purposes the word must be considered as a meaning-bearing unit, a single symbol whose parts are not relevant to it qua symbol, just as the fact that the written symbol y contains a part similar in shape to v is strictly irrelevant to its symbolic employment. Thus the sphota is simply the word considered as a single meaningful symbol. In this conception of the sphota, it seems to me that there is nothing " mysterious" : it is merely an abstraction to assist us in the handling of our linguistic material. In fact, most linguistic discussions implicitly assume such asp/iota, if only as a point of reference. The fact that the sphota itself cannot be pronounced is a characteristic shared with the phoneme and any other linguistic abstraction. And the fact that the Indians themselves appear to have given "ontological status" to this abstraction, and to have considered it as a sort of quasiKantian " Wort-an-sich ", does not detract from the linguistic appropriateness of their observations. This sphota, the word located in the mind (sabdo buddhisthah, VP i.46) is revealed by the sounds produced in a fixed sequence (VPi.85,86): ndair hitabljym antyena dhvanin saha vrtti29-paripkym buddhau sabdo 'vadhryate. asatas cntarle yn chabdn astti mnyate pratipattur asaktih s grahanopya eva sah. " W i t h the last sound, the word is grasped in the mind (of the hearer) where the seed has been sown by the sounds, and which has been brought to ripeness by the telling over in order (avrtti) of the sounds. And as for the non-existent words which the hearer considers to be present in the interval (before the utterance is completed)30this is merely the incapacity of the hearer: it is simply a means to the comprehension (of the word actually being uttered)." The false attribution of time to the sphota may arise in two ways : either we may measure the actual time of an instance, or we may construct an abstract phonological time-pattern (VP i.75-8): sphotasybhinnaklasyadhvaniklnuptinah grahanopdhibhedena vrttibhedam pracaksate. svabhvabhedn nityatve hrasva-dlrgha-plutdisu prkrtasyadhvaneh klah sabdasyety upacaryate. varnasyagrahane hetuh prkrto dhvanir isyate vrttibhede nimittatvam vaikrtah pratipadyate. sabdasyordhvam abhivyakter vrttibhede tu vaikrth dhvanayah samupohante sphottm tair na bhidyate. "According to the differences in the specific cause of its comprehension (in individual instances), men attribute differences in speed of utterance (vrtti) to the sphota which is not divided in time, and merely reflects the time of the sound. Similarly, in the case of the short, long, and prolated vowelssince, on the view that these are permanent, they are intrinsically distinctit is the time-pattern
29

Benares edition and India Office MSS., vrtta-.

30

In examples such as man: manage.

410 John Brough

of the primary sound which is metaphorically attributed to the word (the sphota) itself. The l primary sound ' (prkrta-dhvani) is defined as the cause of the perception of the letters (phonemes), the 'secondary sound ' (vaikrta-dhvani, literally ' modified ') is the causal factor underlying differences of diction. But it is only after the word has been revealed that the secondary sounds are presented to the mind as differences of diction ; hence (a fortiori) the essential nature of the sphota is not disrupted by these." This last observation is most important. I may, for example, hear two utterances identical in absolute length, but it is only after I have understood the words themselves that I can interpret one as sita and the other as sita. The same consideration can, of course, also be applied to other linguistic happenings. In my own speech, for example, I normally distinguish in pronunciation the three words poor, pore, and paw. Many speakers of Southern English, however, pronounce all three (in certain contexts) as pho:. But it is only when I have understood which word is intended that I can start to consider the nature and extent of the vaikrta-dhvani modifications which such a speaker has imposed on the pattern of my own pronunciation, which by an understandable prejudice I consider to be identical with the prkrta-dhvani. Bhartrhari's analysis therefore envisages three aspects of the language situation : (1) the integral linguistic symbol, the sphota, which we may for convenience distinguish typographically as AGNI. This, of course, is not the " pronunciation " of the sphota, since it cannot be pronounced, but is merely the name of it, just as we say, for example, " t h e t-phoneme." (2) The prkrta-dhvani, agni, the phonological structure, the sound-pattern of the norm ; or, from another point of view, the name of the class of which the various instances are members. (3) The vaikrta-dhvani, agni, the individual instance, noted in purely phonetic terms. This, of course, we do not normally perceive in language-communication, since we receive it as a series of sense-data which the brain is conditioned to elaborate and interpret as a finished Gestalt. Accordingly, in a given instance it is apparently the prkrta-dhvani which is presented to the consciousness of the hearer. Even so, it is not felt by the hearer as something separate from the sphota : and normally, in everyday conversation, all that we are immediately conscious of is the meaning (VPi.82): sphotarpvibhgenadhvaner grahanam isyate kaiscid dhvanir asamvedyah svatantro 'nyaih praksakah. "Some consider that the perception of the dhvani is indissolubly linked with (the perception of) the form of the sphota] others hold that the dhvani itself is not perceived (as such) [i.e. we are not normally aware of the phonemes when we hear a w o r d ] ; and others consider the dhvani to be an independent manifesting agent" [as is clearly seen in the speech of a parrot or agramophone record]. The three views are, of course, not mutually exclusive, and in differing circumstances one or other will commend itself. It will thus be seen that the sphoto-doctrine, so far from being something " mysterious", is in fact of central importance for the theory of language-symbolism. The fact that it has been so generally neglected by western Sanskritists appears to be due for the most part to two reasons. The first is the unfortunate mistranslation of sabda as "sound ". Thus, Cowell 31 translates Mdhava's statement varntirikto varnbhivyangyo 'rthapratyyako nityah sabdah sphota
31

Sarva-darsana-samgraha, translated

E. B. Cowell and A. E. Gough, p. 211.

411
General Linguistics in the Sanskrit Grammarians

iti tadvido vadanti, as "And . . . (say the wise in these matters). . . this sphota is an eternal sound, distinct from the letters and revealed by them, which causes the cognition of the meaning". This \s hardly an incentive to further investigation. A retranslation, however, removes the objection : "The abiding word which is the conveyor of the meaning . . . is called the sphota by the grammarians." The second reason for neglect has been the fact that on the basis of the sphota-theory there was erected a metaphysical superstructure, in which the transcendental Word was seen as the firstprinciple of the universe, in a manner somewhat analogous to the Xyoc, doctrine of St. John's gospel. It has therefore been assumed that the theory was in all respects mystical, and that the translation of sphota as "transcendental word" 3 2 was entirely adequate in all circumstances. Hence, it has been overlooked that the doctrine is in the first place founded upon observation and interpretation of the actual speech situation. In fact, when an opponent questions the existence of the sphota, the grammarian replies, not that it requires a mystical insight to perceive it, but that its justification is direct perception of the facts.33 Similarly, Ngesa-bhatta says34. dam ekam padam ekam vkyam iti pratyayah sphotasattve tadekatve ca pramanam, "The justification for the existence of the sphota and for its unity is the realization 'This is one word, one sentence'." The later development of the theory details eight classified varieties of sphota35 : 1. varna-sphota 2. pada-sphota 3. vkya-sphota - (vyakti-sphotas) 4. akhanda-pada-sphota 5. akhanda-vkya-sphota 6. varna-jati-sphota 7. pad a-j at i -sphota 8. vkya-jti-sphota Here also the fundamental argument is meaningfulness (yacakatvam). Thus, the Ietter-sphota (1) is justified on the grounds that a meaning is understood, for example, from suffixes such as -h, -ti, in rmah, pacati. Similarly the alternations in kar-, kr-, /cur-, cakar-, are clearly functional.36 On the other hand, the impossibility of
E.g. O. Strauss, Altindische Spekulationen ber die Sprache und ihre Probleme, ZDMG, N.F. 6,1927, pp. 99151. B. Liebich, ber den Sphota, ZDMG, N.F. 2,1923, pp. 208-219, contributes nothing to the understanding of this question. A. Foucher, Le Compendium des Topiques Tarkasamgraha d'Annambhatta (Paris, 1949), Introd., p. xix, describes the sphota as " La mystrieuse et fulgurante relation qui clate entre le son et le sens, entre le mot et l'ide". The sphota, however, is not a relation, but the word itself. 33 Sarvadarsanasamgraha (ed. Abhyankar), p. 299, pratyaksam evtra pramanam, gaur ity ekam padam ti nnvarnti ri ktai kapadvagateh sarvajanlnatvt.
32
34

Mahbhasya-pradipoddyota

(Bibl.

Ind.), v o l . i, p. 1 1 .
35

See, f o r e x a m p l e , N g e s a - b h a t t a , ( A d y a r L i b r a r y , 1946); Sabda-kaustubha

Sphotavda

Bhattoji-dlksita, 1933), p p . 7 f f .

( C h o w k h a m b a S k t . Ser., Benares,

The discussion of the question as to whether or not the individual letters in a word have meaning is not far removed from modern discussions on the phoneme. Thus, it may be said that letters are meaningful on the grounds that meaning can be understood from roots, suffixes, and particles which consist of a single letter; and also since the substitution of a different letter can produce a different word, while the non-perception of a single letter may make it

36

412 John Brough

discriminating exactly how much of the word conveys the thingmeant, and how much the case-relationship (e.g. in rmya, ramena, haraye, harau, harln)} makes it necessary to postulate a word-sphota (2). Similarly, sandhi-forms such sdadhdam, "this is curds," point to the desirability of a senten ce-sphota (3). But inasmuch as the word is still thought of as consisting of stem and suffix, and the sentence as consisting of words, these sphotas (2 and 3) do not satisfy the linguist's requirements. As Bhartrhari had already shown, in language as we find it in the world as an object of study, there are no letters in the word, and no words in the sentence.37 The analysis into letters and the distribution of meanings between stem and suffix, or between the words in the sentencethese proceedings, as Ngesasays, are the occupation of grammarians.38 In actual usage, the word conveys its sense as a unit, and hence the undivided-word-sphota (4) takes the place of (2). Even this is not completely satisfactory, since isolated words (padas) do not occur as meaningful utterances in ordinary language (apparent exceptions being one-word sentences). Therefore the grammarian admits the reality only of the undivided-sentence-sp/iota (5). The preceding sphotas are merely fictional (klpanika) constructs of the grammarian. Here there arises a philosophical controversy as to whether the sphota is a particular (yyakti) or a universal (jti). To meet the needs of those who believe in the latter, the class-sp/iotas (6, 7, and 8) are provided. V. Krishnamacharya39 states that the ancients held the jti-sphota view, the moderns (i.e. Ngesa and his contemporaries) the vyakti-sphota. The ascription of the jti-sphota view to Bhartrhari is supported by Bhattoji-dlksita, 40 who quotes in this connection some stanzas from VP. iii. 1.33 : sambandhibhedt sattaiva bhidyamngavdisu jtir ity ucyate tasym sarve sabd vyavasthith. 11 Being divided into cows and so forth through distinctions of those things in which the relationship subsists, it is Existence which is called the Class (par excellence) ; and in this class all words have their being." This, however, is no argument for holding that Bhartrhari accepted the jti-sphota theory, and still less is the other passage quoted by BhattojT (VP i.94) : anekavyaktyabhivyangy jtih sphota ti sm.rt kaiseid vyaktaya evsy dhvanitvena prakalpith. In this stanza, Bhattoj presumably took the first line as a complete statement; but the interpretation of Ngesa-bhatta,41

impossible to understand the meaning. But from another point of view the letes are meaningless in themselves, since the hearer does not perceive a meaning from each letter separately. (MBh. I. 220. arthavanto varn dhtuprtipadika-pratyaya-niptnm ekavarnnm arthadarsand varnavyatyaye crthntaragamand varnnupalabdhau cnarthagateh ; anarthaks tu prativarnam arthnupalabdheh.) In similar fashion some modern writers have considered their phonemes to be the smallest

significant segments of a word.their "significance" lying in their differentiation-value. 37 VP i. 73: pade na varn vidyante varnesv avayav na ca vkyt padnm atyantam praviveko na kascana. 38 Ngesa, Laghumanjs, p. 5, tat tad arthavibhgam sastramtravisayam. 39 Introd. to his edition of Ngesa's Sphotavda, p. 24. 40 Sabda-kaustubha, p. 9. 41 Sphotavda, p. 99.

413
General Linguistics in the Sanskrit Grammarians

which takes the first phrase down to kaiscit, seems preferable42: "Some consider that the sphota is the class revealed by the various individual instances, and they consider that the members of this class are the dhvanis." Thus, on this view, as Ngesa explains,43 there appears in the particular only the sounds and nothing more. The distinction between the two views, therefore, is not quite what it would appear at first sight._B_oh sides, in fact, accepted the concept of the class, but while the jti-school considered the sphota to be merely a class whose members were not themselves sphotas, the other school maintained that the sphota was in fact present in the particular. That Bhartrhari held the latter views is further supported by the fact that in his discussion of the definition of a sentence he appears to give preference to the view that the sentence is eko 'navayavah sabdah, "asingle integral languagesymbol," i.e. vyakti-sphota, rather than ajtih s'amghtavartinJ "a class residing in a collocation (of words)". Thus the class-sphoto theory is closely similar to a modern view put forward by Kaplan and Copilowish,44 who define a sign (including a linguistic sign) as "a class of sign-vehicles all having one and same law of interpretation ". Merely as terminology, this is extraordinarily awkwardRussell, for example, stumbles over this awkwardness and talks of the loudness of a "sign ", forgetting for the moment that a "class of sign-vehicles" cannot have loudness. But the authors themselves have not taken their definition seriously, for they discuss the conditions under which a sentential sign may be true. Clearly a class cannot be true or false. Thus they have propounded a jti-sp/iota theory, but implicitly assume in their discussion a vyakt i -sphota view. It would seem, in fact, that the \dX\-sphota view is philosophically unworkable. In Bhartrhari's view, then, the primary linguistic fact is the undivided sentence-sp/iota. Just as a bare root has no meaning in the world, 45 so also the meanings of individual words are merely hints or stepping stones to the meaning of the sentence. This is a plain linguistic fact, which has none the less been clear to very few philosophers, either in India or elsewhere; and though familiar enough in modern linguistics, is still constantly overlooked in many discussions on meaning. Until it is thoroughly understood, no real progress can be made in the central linguistic problem of semantics. Bhartrhari's discussions of these questions is full of valuable observation, and it is hoped to offer an account of these on a later occasion. The present paper is intended merely as an introduction to some of the aspects of the Indian theory which are still of vital importance to us to-day. In this very difficult field I can hardly hope to have rendered the subtle arguments in so brief a compass in an entirely satisfactory manner. To quote yet again from Bhartrhari (.34) : yatnennumito 'py arthah kusalair anumtrbhih abhiyuktatarair anyair anyathaivopapdyate. "When clever logicians have with great pains deduced a result to their own satisfaction, then others still more able come along, and
42

This is also in accordance with Punyarja's commentary on Bhartrhari, which introduces the stanzaas matntaram ha. 43 bid., tadvyajakavyaktayas ca dhvanayaeva, natadatirikt iti.

44

Mind, Oct. 1939; discussed by Russell, Inquiry into Meaning and Truth, p. 184. 45 VP ii. 212, dhtvdlnm visuddhnm laukiko Ytho na vidyate.

414 John Brough

reach a totally different conclusion." If nothing else, I t r u s t that this preliminary presentation w i l l offer an incentive t o those more able t o prosecute this study f u r t h e r .

Brough reverted t o Indian semantics in the following article, " S o m e Indian Theories of Meaning," Transactions of the Philological Society (1953,161-176). This paper paved t h e way for later w o r k by K. Kunjunni Raja, f o r example, Indian Theories of Meaning (1963).

B Some Indian Theories Of Meaning (1953) John Brough

In a paper to the Society in 1951,1 I gave some account of ancient Indian theories on the relationship between sounds and meaningful words which is summed up in the doctrine ofsphota. The present paper is intended to supplement that discussion and to indicate those aspects of Indian theory on the more general topics of meaning which I feel preserve the greatest interest for modern linguistictheory. One of the earliest pieces of practical linguistics of which we have any record is the composition of the pada text of the Rgveda. This is an analysis of the somhit, or connected text as uttered in recitation, into its constituent words in the form in which they would appear in isolation. This involved the resolution of the rather complicated junction-features which the connected Sanskrit sentence exhibits, and in many places the pada text did in fact amount to an interpretation at a time when the connected text was beginning to suffer from obscurity. Alongside this analysis of words from the sentence there was developed the study of the meanings of the words thus derived, and the results of this etymological study is summed up for us in theNirukta of Yska. It is clear that just as phonetics arose in India chiefly as a means to preserve the mode of utterance of the Vedic hymns, so the study of words and of the meanings of words was undertaken in the first place primarily to meet the needs of Vedic ritual and the text material required by it. It was thus on a basis of words and of word-meanings that the study of Vedic exegesis took shape in what was later known as the Mlmms school of philosophy. And, indeed, throughout the development of Indian linguistic thought, the relationship between word and sentence, between the word-meaning and the sentence-meaning, remains a central problem. The Mmms school developed elaborate canons of interpretation, and this organized body of linguistic doctrine later played an important part in the discussions of lawyers when interpreting legal injunctions, and on the other hand did much to stimulate the development of logic. In passing, one might note that the Mlmms preoccupation with the injunctions of Vedic texts with regard to religious duties was not without its effect on logical theory. The typical sentence with which they are concerned is in the imperative mood, and although later Indian logic deals largely in indicative sentences, the linguistic thought of philosophers in India was not so strictly confined to indicative propositions as that of logicians inthe west. This influence can be traced in the terms vidhi and pratisedha, originally meaning injunction and prohibition, but in later texts occasionally used also to apply simply to positive and negative statements. I do not propose here to give a detailed account of all the types
1 Theories of General Linguistics in the Sanskrit Grammarians, TPS1951,

pp. 27-46. [This volume, pages 4024I4]

415 Some Indian Theories of Meaning

of definitions of asentence which occur in Indian w r i t i n g s ; but mention should be made of the main types. As early as the Ktyyana-srauta-stra the Mlmms type of definition appears, in a purely ritual manual. A sentence, it is said, is that which is nirkksam2: that is t o say, something which has no requirement or y' expectation of words outside itself t o complete its meaning. It is, of course, realized that the expectancy which holds between words in the sentence is a grammatical one, since, for example, a sentence containing a pronoun requires the evidence of a neighbouring sentence t o identify the pronoun. Accordingly a commentator on this passage interprets the rule which follows, mithah sambaddham " it is mutually linked ( w i t h other sentences)" t o mean that it may be necessary t o complete the meaning of the sentence by understanding words f r o m preceding sentences. The fact that the requirement of knks is a grammatical one was not always fully understood and we find even Bhartrhari criticizing the Mlmms definition on the grounds that its knks would imply that a passage of several grammatical " s e n t e n c e s " w o u l d have t o be considered as one sentence. 3 Later the normal statement of the conditions for a sentence is that it must be a collection of words possessing knks, yogyat and satti. In effect, however, it is only the first which is of real linguistic importance. It is the desire or requirement of an individual w o r d or words in the sentence for others t o complete the meaning, the factor which distinguishes a sentence f r o m a string of words " c o w horse man elephant". The second factor, yogyat, really involves a judgment on the t r u t h or falsity of a statement, or the sense o r nonsense of a sentence. The example quoted most frequently as a breach of this condition is " He wets it w i t h f i r e " . Into this category also fall such logical puzzles as " t h e round square". The t h i r d condition, satti or samnidhi, is that the words should be contiguous in t i m e . It is said that words u t t e r e d w i t h the interval of a day intervening between each w o r d cannot produce a sentence. This again is not a linguistic condition. It is of course only t o be expected that the early stages of linguistic t h e o r y in India, as elsewhere, should show a certain navet and it is not surprising t o find generally current in Indian philosophy, outside the writings of the professional grammarians, the idea that an individual w o r d possesses an individual word-meaning or, in the case of nouns, that the w o r d is the name of a t h i n g . This view is fossilized in the regular philosophic t e r m for thing or object, namely padrtha, literally l l meaning of a w o r d , that which a w o r d means". The t w o main schools of the later Mlmms w e r e sharply opposed in t h e i r theories of the sentence. The Bhtta school on the whole seems t o preserve the more p r i m i t i v e attitude. According t o them words have in themselves meanings, and as the words are uttered in a sentence, each w o r d performs its task of expressing its meaning, and the sentence is the summation of these meanings, The Prbhkara school, on the other hand, held the more sophisticated t h e o r y that the individual words did not express any meaning until they w e r e united together into a sentence. This was upheld by an appeal t o the method whereby a child learns its o w n mother tongue. They pointed out that it was by hearing sentences " f e t c h the c o w " , " f e t c h the h o r s e " , and so f o r t h , that the child came
2

KSS i. 3. 2 tesm (sc. mantrnm) vkyam nirknksam. 3 VP ii. 3 ff. The grammatical sen-

tence is here identified on the basis of the Varttika definition, eka-ti " possessing one finite verb".

416 John Brough

gradually t o understand that the animal which he saw on each several occasion was, in fact, either a cow or a horse and that the action performed by his elders was the act of fetching. These t w o views were named respectively obhihitnvoya-vda and anvitabhidhna-vdo, terms which are troublesome t o translate by concise English expressions. Roughly speaking, the first is the t h e o r y that the sentence is "aseries of expressed w o r d - m e a n i n g s " , and the second is that the sentence is " t h e expressed meaning of a series (of w o r d s ) " . A t the beginning of the second book of the VkyapadJya, Bhar t r h a r i gives a list of definitions and quasi-definitions of a sentence. Fiveofthese are grouped by the commentator u n d e r t h e traditional Mmms designations. Thus the view that the sentence is a unified collection (samghta) and the view that it is an ordered series (krama) are aspects of the abhihitnvayo-vda ; while the o t h e r t h r e e belong t o the anvitbhidhna-vdo. These are, that the sentence is defined by a verbal expression (khyta-sabda) or by the first w o r d (pa da m dyam) or by all the words taken separately w i t h the feature of mutual requirement or expectancy superadded (prthak sarvapadam sknksam). All these views, of course, imply the feat u r e of expectancy, and the first and second are t o be explained w i t h reference t o this feature, since the verb or the first w o r d is only what it is in view of its ties w i t h the o t h e r words in its own sentence. All these theories are adversely criticized by Bhartrhari and they need not be considered in detail here. They are none the less of some interest as evidence of very vigorous argument and debate on linguistic topics in ancient India. All these earlier discussions on the nature of the sentence accept w i t h o u t question the fact that there are such things as words (pada), and that it is possible t o a t t r i b u t e t o these words something which can be called t h e i r meaning (artha). The most developed theory, namely that of the anvitbhidhnavda, did t o some extent foreshadow the later development in the grammatical schools, since it denied the words conveyed a meaning except in the cont e x t of a sentence. Like the other views, however, this t h e o r y continued t o regard the words as real and actual constituents of language. They were the units which in fact operated in linguistic communication, and since they were actually present, it merely remained for the grammarian t o detect t h e i r presence by means of a grammatical analysis. The statement of procedure is quite in accordance w i t h many modern statements. Thus a root or suffix is analysed out on the basis of a paradigm, and complete words were recognized on the basis of substitution in sentences. It remained for the professional grammarians, of w h o m Bhartrhari is the leading ancient spokesman, t o draw attention t o the fact that although this process of analysis could give some account of language f r o m the formal aspect, and though it possessed a distinct value for teaching and for the explanation of texts, it was entirely inadequate as a basis for a t h e o r y of language-meaning. To Bhar t r h a r i and his school words w e r e , in fact, artificial constructions of the grammarian, and looked on f r o m the point of view of language functioning in the w o r l d , they were unreal (asatya). This extraordinary relegation of words t o the realm of fictions is not at all easy t o grasp at first sight, and I hope I may be pardoned if I dwell at some length on this topic, since it seems t o me of considerable importance for fundamental linguistic theory, and hence also for philosophy, in so far as the latter may be a " c r i t i q u e of language".

417 Some Indian Theories of Meaning

It is, of course, a commonplace in modern linguistics that the sentence is the primary datum. But such a statement may imply no more than an attitude comparable with the anvitbhihnavda. I suspect that it is the latter type of view which is often implicit in statements described as "analysis at the word level". 4 It demands something of an effort for the beginner brought up on an alphabetic system of writing to appreciate fully that a word is not a succession of letters or phonemes or segments which are then rammed together"realized " has been a popular term in this connection. But on the contrary the word is what it is, and any account in terms of syllables, letters, phonemes, segments, prosodies, is merely an analysis, an attempt at description which may be more or less adequate. A similar effort, though perhaps a still more difficult one, is required to grasp the significance of Bhartrhari's theory of the unreal nature of words. It is important to realize that this theory is not derived from a priori speculation, but is the result of a careful examination of what happens when we speak or listen in ordinary conversation. We do not in fact express ourselves or understand what is spoken in a series of meaning-units. After a sentence has been understood we may look back at it, analyse it into words, and maintain that we discern words in it. But if we do so during the course of the utterance itself, we are apt to lose the meaning of the sentence. The situation is perhaps analogous to the experience which some of us had at the recent International Congress of Linguists (London, 1952) when Professor Delattre played to us records of series of synthetic vowels, each vowel being made up of two musical formants. According to the method of focusing the attention, one could hear the record either as a series of vowels or as two converging musical scales, but not as both simultaneously. The essence of the matter lies in discriminating clearly between language in operation, and language-material considered and described by a grammarian. Bhartrhari's view is simply that words and "word-meanings" belong to the latter sphere. They constitute an apparatus (not necessarily adequate) for the description of language events, but (roughly speaking) do not themselves "exist" in the events described. This theory of the non-reality of words not unnaturally met with strong opposition from other Indian philosophers, and Bhartrhari provides us with a number of the arguments which they brought against it. Typical of these is the argument from our experience of sentences which contain an unknown word. If for example a townsman, who has not previously heard of the bird in question, hears the sentence, " Fetch a cuckoo from the woods," he instinctively assumes that he has not understood the sentence because he does not know the word "cuckoo" ; and as the objector points out, he does know the words " Fetch . . . from the woods". On Bhartrhari's view, however, a better description of the situation
Such an analysis may be justified where the forms of words are the chief concern; but considerable contortions and an embarrassing set of fictions (e.g., the "central core of meaning " of a word) seem to be needed if we attempt to construct a descriptive statement of meaning on the basis of a word analysis. Such
4

a wholesale rejection of word-meanings from scientific discussion does not of course mean the advocacy of the abolition of lexicography, but rather the recognition of the essentially practical and pragmatic nature of lexicographical statements of meaning.

418 John Brough

would be that he cannot understand the w o r d t h a t is, he cannot a t t r i b u t e a word-meaning t o itbecause he has not understood the sentence. This is, at first sight, perverse and paradoxical, but if patiently considered it will be seen t o have much in its favour. It follows as a corollary that the piece " Fetch . . . f r o m the woods " is not the same as similar phrases which occur in other sentences, for example, " Fetch a tree f r o m the w o o d s " ; and this situation Bhartrhari unreservedly accepts. 5 It is of course clear that any meaning which we a t t r i b u t e t o the fragment " Fetch . . . from the woods " is different in the t w o cases, since for example the method of transport w i l l be different. A t the most, therefore, we can say that the apparently identical fragments in the t w o sentences are similar but not the same. On this view, the distinction between " f o r m u l a e " and " f r e e expressions " is not so clear-cut as Jespersen w o u l d have us believe. 6 O f the f o r m e r category, he says, " O n e may indeed analyse such a formula and show that it consists of several words, but it is felt and handled as a unit, which may often mean something quite different f r o m the meaning of the component words taken separately." But this holds also for free expressions, and as Jespersen himself realizes elsewhere, 7 the " meaning of the component words taken separately" is something which cannot be determined apart f r o m a c o n t e x t ; and once this is granted, the distinction between the t w o categories amounts t o little more than t h i s : that a w o r d meaning analysis is more congenial t o agrammarian in the one case than in the other. The apparently objective c r i t e r i o n upon which Jespersen relies t o diagnose a f r e e expression, namely, subs t i t u t i o n in sentential functions, is as we have seen explicitly rejected by Bhartrhari as being in fact illusory. The occurrence of homophones in a language has always provided grammarians w i t h an interesting problem, and almost all w r i t e r s on the t h e o r y of grammar have discussed the factors which enable a language t o tolerate such homophones w i t h o u t giving rise t o ambiguities. Bhartrhari gives a list of such factors, of which the most i m p o r t a n t are vkya, sentence-context, and prakarano, situational context. As a typical modern statement of the same matter I might quote Sir Alan Gardiner: " The polysmie of words . . . does not matter in the least, because the hearer always has the situation t o guide him in choosing that type of meaning which is appropriate t o the c o n t e x t . " 8 This statement conveys the position roughly; but it seems unlikely that the hearer actually chooses
5 The objection is raised in VP ii. 74, word are in themselves meaningless, and answered in ii. 94. The nave perso also are the words (pada) in a senson (mdha) thinks that he perceives tence.] Thus the substitution of the the same meaning in the parallel word pika for vrksa produces an enportions of the two sentences (vonat tirely different sentence; and if there pika an Jyatm; vand vrksa an iyotrn); is doubt as to the meaning of one but this, as the commentator remarks, word, then the whole sentence is not is a misconception due to the serial understood (pikdiyogt sakalam nature (krama-vast) of the linguistic evtyantavilaksanam ; ekapadrthasamsentence-symbol. [Thus for Bhartrhari dehe sakalam evjntam vkyam ty the nave view is completely analoucyate). 6 gous to the suggestion that in y and v O. Jespersen, The Philosophy of due to the linear nature of our Grammar, p. 18f. 7 writingthere s a common part. Cf. Op. cit., p. 66. 8 also VP ii. 416 : just as the letters in a TPS1951, p. 60.

419 Some Indian Theories of Meaning

between the various meanings or types of meaning of the word he certainly does not do so consciously in his native language ; and it might be better t o say that the nearer does in fact understand the sentence, and that this understanding, if afterwards utilized by the lexicographer (who also must " understand " the sentence), will enable the latter t o state that, in such-and-such a situation, in such-and-such a verbal context, etc., the given w o r d can be extracted analytically, and such-and-such a " w o r d - m e a n i n g " a t t r i b u t e d t o it, this word-meaning being different f r o m other meanings of the same " w o r d " in different contexts and different situations. Further, it is necessary t o recognize that when we talk of " t h e w o r d x w i t h meaning A", " t h e w o r d x w i t h meaning ", the identification of the t w o x's as " o n e w o r d " is a mere practical convenience for lexicography and exegesis, something which belongs not t o the material but t o one method of describing the material, and that this method is not necessarily the best approach t o a satisfactory description of language in operation. W e are apt t o say f r o m t i m e t o t i m e , when struggling w i t h a difficult passage in a foreign t e x t , that we know all the words, but that the meaning of the sentence escapes us. This however is a delusion. In such circumstances we are presumably a t t r i b u t i n g t o one or more of the words a " m e a n i n g " which has not been extracted f r o m this particular context, and the obvious comment is that we do not know all the words, since our knowledge does not include the manner of t h e i r occurrence in the context in question. In practice, of course, a more general, if vaguer, aura of meaning extracted f r o m similar contexts frequently gives us a sufficient clue; but this leads us in the first place t o an understanding of the meaning of the sentence as a whole, and only afterwards, by an analysis of this understanding, t o the a t t r i b u t i o n of meanings t o the individual words. These considerations are of the first importance for those of us w h o are concerned w i t h ancient texts and hypothetical forms in Indo-European or other conjectural languages. The pursuit of the meanings of words in ancient texts is a highly skilled art, and the best w o r k which has been done in this field has substantially added t o o u r u n d e r s t a n d i n g of texts. But it is an art which requires a delicate tread. W h e n we inquire into the meaning of a w o r d in an ancient language we are really juggling w i t h possible translations of sentences in which the w o r d occurs, until we finally succeed in finding a mode of translation in which a single w o r d or phrase of the English appears t o correspond more or less t o the ancient w o r d in question. W e then say that this English w o r d or phrase is a meaning of the ancient w o r d . Here we at least have in the texts sentences which by one means or another we understand after a fashion ; and historical and comparative studies frequently enable us t o glean f r o m texts in related languages useful hints towards this understandingthough it is important t o remember that these methods can at best protect us, as a near-scientific c o n t r o l , against specific errors, but can never prove a positive case. W h e n we come t o the prehistory of words, however, we have no sentences at all. The only conclusion that we can reach is that it is therefore impossible t o talk of the meanings of, for example, Indo-European roots, except in a very different sense of the t e r m " meaning". Indeed the vagueness of the meanings a t t r i b u t e d t o Indo-European roots by w r i t e r s on this subject is an indication of t h e vagueness of what is meant by meaning in this context.

420 John Brough

Having characterized the sentence as " a s i n g l e undivided u t t e r a n c e " which conveys asingle undivided meaning, Bhartrhari proceeds t o indicate what he understands t o be the nature of this sentence-meaning. One cannot claim that what he sa/s is a definit i o n , and indeed the t h e o r y itself really implies that definition as ordinarily understood is an impossibility. The important point is that the sentence-meaning is grasped as a unity. The situation is compared t o our apprehension of a picture. This we perceive as in some sense a unity, and although we can analyse the field of vision, and say that this part of the picture is blue, this part w h i t e , and so f o r t h , none the less, we are normally aware of the integrated whole. If on this analogy we proceed t o explain the sentence on the basis of an analysis into words, we are in fact merely giving a commentary on it in what are ultimately other words, not words of the sentence itself. The idea here is closely similar t o that expressed by W i t t g e n s t e i n when he maintains that a proposition can only show what it has in common w i t h the fact, and that this cannot be said in language, since any attempt t o do so can only produce other propositions sharing the same logical form. 9 In the end the utmost that can be said of the meaning of asentence according t o Bhartrhari is that it is grasped by an instantaneous flash of insight (pratibh) The same w o r d is used in later times w i t h reference t o the insight of a great poet, and in such contexts may be reasonably translated as poetic genius. W e are all, in fact, in a greater or lesser degree poets in our composition of sentences and in our understanding of the utterances of our fellows. And when we have understood a sentence, we cannot explain t o another the nature of this understanding. 1 1 A l t h o u g h it is an acquired faculty, understanding a language is in its operation very similar t o the instinctive behaviour of animals. 12 It is unnecessary t o labour the point that the meaning of a sentence is not necessarily grasped f r o m a knowledge of the dictionary meaning of the words. A few examples, however, may be given of the way in which sentences frequently produce an " implied " sense over and above what appears t o be the literal sense. As Bhartrhari's commentator points out, when a mother says, " T h e tiger eats little boys who c r y , " she does not in fact literally mean that t h e r e will follow an actual eating by a tiger. Rather she

Troctatus, 4.12-4.1212. plavanadikriyasu kah VP i. 119,145. jtyanvayaprasiddhsu 11 VP i i. 146: dam tad iti snyesam anprayokt mrgapaksinm. khyey kathamcana : pratytmavrttisi"Who alters the note of the cuckoo ddho s kortrpi na nirpyote: "This in spring? Who teaches the spider to (pratibh) cannot in any way be exweave its web? Who impels the birds plained to others in terms such as ' It and beasts in their eating and mating, is this'; its existence is ratified only in their enmities, or in their flight, in the individual's experience of it, and i nal I the other act ions determined and the experiencer himself cannot by heredity?" The commentator on describe it." the latter verse in fact uses the term 12 Bhartrhari points the analogy in a pratibh, where we should say "anipair of verses (VP i i. 151,152): mal instinct" (pratiprnyhrdikriy svaravrttim vikurute niyatndipratibhvast, "Theactions madhau pumskokilasya kah : of animals, eating, etc., differing from jantvdayah kulydione animal to another, are determined karane kena siksitah. by a beginningless pratibh," i.e., a hraprtyabhidvesapratibh which is not learnt).
10

421 Some Indian Theories of Meaning

means w h a t is m e a n t by t h e s e n t e n c e , " D o n ' t c r y . " 1 3 S i m i l a r l y , if a t r a v e l l e r says t o his c o m p a n i o n , " W e m u s t g o , l o o k at t h e s u n , " t h e m e a n i n g c o n v e y e d is n o t s i m p l y t h a t o f l o o k i n g at t h e s u n , b u t

rather that the companion should realize how late in the day it is.14 Again, in response to the command, " See that the crows do not steal the butter," not even a child is so literal-minded as to interpret it to mean that he can allow the dogs to steal the butter.15 Examples of this sort are a direct invitation to formulate a theory on the hypothesis that a sentence can be said to have a literal meaning. This is something which in our normal linguistic discussions we are very apt to take as axiomatic; but it will be apparent after what has been said that from Bhartrhari's point of view it is more of the nature of a postulate which we ourselves lay down as the condition for constructing specific systematic statements, as a practical convenience in handling the material. For Bhartrhari himself, the examples quoted above were probably simply further indications of the unsatisfactory nature of a theory depending upon word-meanings. But the commentator does in fact interpret them on the basis of metaphorical transfer of meanings (laksan). This was the standard interpretation in later grammatical writings, and we find for example the explanation that "crows" in the sentence quoted stands metaphorically for "crows and other animals which might steal the butter." 16 The theory of literal and metaphorical meaning was further extended in the 9th century by nandavardhana in the Dhvanyloka. This is primarily a treatise on poetics; but as the basis of his aesthetic theory, the author carries out an elaborate analysis of poetic meaning. He had inherited from earlier theorists the distinction between primary and transferred or metaphorical senses of words (abhidh and laksan), and in addition to these he postulated a third potency of language which he called the capacity to imply or reveal a meaning other than the literal meaning (vyanjan). The central term of the theory, namely dhvani, which has frequently been translated as suggestion, is said by Ananda himself to be directly taken from the grammarians, though the relationship between his use of it and the use in grammar has perhaps been insufficiently clarified by modern writers. In brief, just as the sound of utterances (dhvani in the grammarians' sense) reveals the word (sphota), so a poem is said to be dhvani when it reveals a meaning over and above the literal meaning and when the revealed or implied meaning has at the same time aesthetic value. In this theory we thus leave the more abstruse levels of philosophic linguistics, and cometo more practical affairs, namely, the description and classification of meaning types as they occur in literature. Ananda's work in fact seeks to unite the two traditions of the grammarians on the one hand and the formal rhetoricians on the other. Ananda's basic postulate is that utterances possess a literal meaning, and can also convey a further meaning. The scheme of classification which he adopts is fairly detailed and I can give only the outlines of it here. The main subdivision is into two types, first, the type where the literal sense is not intended (avivaksita13 Commentary to VP ii. 322: yath rudantam vyghro bhaksayatiti blosyocyate, na tatra vyghrabhaksanam vostusthity sambhovi, kevalam m kadcit tvam rodlr iti rodananisedha eva tasya kriyate . . . 14 15

VP ii. 312. VP ii. 314: kkebhyo raksyotm sarpr t blo lp coditah: upaghtapare vkye na svdibhyo na raksati. 16 So far example Ngesa Bhatta, Laghumanjs, p. 123 (kkdi).

422 John Brough

vacya); and second, the type where the literal sense is in fact intended, but subserves the implied sense (yivaksitnyapara-vcya). The first of these is again subdivided into t w o : the type where the literal sense is completely set aside (atyantatiroskrta-vcyo), and the type where the literal meaning is shifted (arthontorosamkramita-vcya). The first of these embraces what we should normally call metaphor; but it is, so to speak, motivated metaphor, where the metaphorically used words are employed with the definite intention of conveying their associations, or producing a striking eect. The second sub-variety is an interesting one, and covers cases where a word is used in an enhanced or diminished sense. Edgerton 17 compares this with the "emphasis" of the classical western rhetoricians, quoting Quintilian's definition ; though in fact the point of view here is somewhat different. Typical examples are, " O n l y when favoured by the rays of the sun are lotuses lotuses" ; " Let men continue to give the moon as a simile for her face; none the less, in the final analysis, the poor moon is the moon." Of much greater interest is the second main subdivision, where the literal sense is intended. The chief type here is that where poetic emotion or mood (rosa) is conveyed. It is of great interest to see the term artha " meaning," enlarged to include all that is conveyed by a poem. In accordance with the grammarians' views on the unity of the sentence-meaning, the dhvani-theory to a large extent operates in terms of larger unities and not individual words. A t the same time it is possible from another point of view to indicate that the operative factor in producing the overtones of the implied meaning may on occasion be a single word or phrase. Thus in one example an old hunter says to a tradesman who is seeking merchandise, " How can you expect us to have elephants tusks or tiger skins, so long as my daughter-in-law wanders about the house with dishevelled hair?" Here, says nanda, the dhvani arises not from the sentence as a whole, but from the phrase "dishevelled hair," since this indicates to the hearer that the hunter's son, who ought to be out hunting, is in fact spending his time in dalliance with his newly wedded wife. 19 Similarly, when in a drama the king Udayana is told that the queen has perished in afire, he calls to mind in his anguish her beauty: " Those eyes of hers glancing wildly round in terror . . ." Here the word " t h o s e " heightens the emotion conveyed by the stanza and underlines for the sensitive audience the poignancy of the king's memories of very different circumstances.20 But though it is reasonable for analysis to take account of features of this sort, Ananda fully realizes that in other cases we must take the whole stanza,

17 F. Edgerton, " Indirect Suggestion in emphasis (Institutio, viii, 3, 83 ff.). 18 Poetry: A Hindu Theory of Literary To bring out the idea, we can offer Aesthetics," Proc. American Philosoparaphrases such as " i.e., lotuses in phical Society, Ixxvi, 1936, p. 700. the fullest sense of the word; lotuses Edgerton seems to imply that the with all the qualities of beauty which whole of the avivaksita-vcya category make them worth calling lotuses"; could be compared with emphasis, and in the second example, " i.e., only though in fact only the arthntarasam- the moon and nothing else." 19 kramita type is really analogous to the Dhvanyaloka, iii. I, vrtti. 20 first of Quintilian's two varieties of Ibid. iii. 4.

423 Some Indian Theories of Meaning

o r even t h e whole poem, as instrumental in conveying t h e poetic meaning. 21 The extant Sanskrit writings on linguistic t h e o r y and on rhetoric f o r m a very extensive literature, and the foregoing account is necessarily a mere outline sketch of some of the most interesting aspects of the Indian theories. One important point which I should like to stress is the realization of the Sanskrit rhetoricians of the need for an explicitly formulated theory of language-meaning as a basis f o r a t h e o r y of poetics. Most philosophic discussions of meaning confine themselves t o a relatively small portion of language behaviour, namely, statements which describe o r r e p o r t a state of affairsthe propositions of the natural sciences, or, more generally, such statements as are traditionally handled by logic. This part of language possesses enormous importance and prestige, and is also the least difficult t o deal w i t h in a more o r less clear fashion. But its t r e a t m e n t frequently suffers f r o m a forgetful ness of t h e fact that propositions (or the formulae of symbolic logic) are none the less language; and I w o u l d suggest that a w i d e r linguistic understanding is most desirable, both for philosophy and f o r poetic t h e o r y . O f colloquial language, W i t t g e n s t e i n remarks 22 that it is " a part of the human organism and not less complicated than it. The silent adjustments t o understand colloquial language are enormously complicated." This is sufficient t o dismiss the subject f r o m the consideration of logic, and it is of course quite reasonable that the logician should limit his field in this way. The linguist however must include w i t h i n his survey all types of language behaviour, f r o m logic t o literature. Wittgenstein's implication clearly isthat logic can construct a logical language which can be understood w i t h o u t these " s i l e n t adjustments " ; and it has frequently been claimed in modern times that the aim of logic should be the construction of syntactical rules which will prevent nonsense. I t r u s t that the present paper w i l l show that the f o r m e r hope is certainly a vain illusion, and that the latter is probably so. Logic, mathematics, linguistics, science in general, all convey t h e i r messages in language, and this language, however technical, cannot be understood save in a manner which is fundamentally similar t o the understanding of everyday language. As t h e ancient Indian might say, t h e utterances of the costermonger, the language of t h e great poet, and the formulae of t h e atomic physicist are all in some sense manifestations of the same divine Vk.
21 See the discussion in the early part of book iii (summed up in iii. 2), where the types of dhvani are classified as arising from individual sounds (varna), words (or parts of words, suffixes), sentences, "stylistic struc-

ture " (samghatan, i.e. the style measured by the incidence of cornpound words), or the whole poem or epic. 22 Tractatus, 4.002.

Ylltaka Ojhara ( b o r n 1923)

Buddhism was officially introduced in Japan in the sixth century. Buddhist scholarship developed in its wake, and Sanskrit studies began to flourish within the larger framework of research on Indian Buddhism. While much information on India used to reach Japan through Chinese intermediaries, the eighteenth century brought a revival that broke away from this dependence. Though the Japanese interest has largely remained confined to Buddhist studies (just as the Indian interest itself remains largely confined to Brahmanical studies), Sanskrit scholarship has rapidly increased in depth and width ever since this time. Among non-Buddhist topics treated by Japanese Sanskritists, the Hindu philosophical systems rank first; these had in fact influenced Buddhist philosophy in India, and some had even become part of the Buddhist canon in China (for a general evaluation of Sanskrit and Indological studies in Japan, see Renou 1956a). The leading contemporary Japanese expert on the Sanskrit grammarians is Yutaka Ojihara (born 1923), who studied in Philadelphia and Paris (with Louis Renou), and at the Bhandarkar Institute in Poona. He now teaches at Kyoto University. Ojihara has contributed aseries of very careful specialized articles (including several translations) to Japanese journals, some of them written in Japanese, others in French. Among the latter are the "Causeries Vykaraniques," IIV, in different issues of the Journal of Indian and Buddhist Studies during the period from 1958 to 1967. Together with Louis Renou, Ojihara published two volumes of a translation of the first pda of the Ks'/k; the third volume is by Ojihara alone. The fourth of the "Causeries Vykaraniques," "Jti 'genus' et deux dfinitions pr-patajaliennes," is reproduced here. It deals with a problem which led the grammarians to invoke two different notions of jt/ 'genus, generic term.' In Sanskrit a bahuvrlhi compound may be formed to express " he whose wife (bhry) is a young girl (kumri)" (in analyzed form: kumn bhryyasya). Since the description itself refers to a man, the compound stem cannot have a feminine ending (as in kumri-bhry) but must be provided with a masculine ending. The " masculinization " may be confined to the end of the compound (as in kumri-bhrya), but it may also affect the ending of its first member (as in kumra-bhrya). The two definitions of jot/ are invoked to account for such alternatives. Ojihara's study is reprinted, with the author's corrections, from the Journal of Indian and Buddhist Studies (16,1967, 459-451 [Japanese reverse pagination]).

Jati genus 'et deux definitions pr-patajaliennes (1967)

Yutaka Ojihara

Le mot jti figure vingt trois fois au total dans la Grammaire de Pnini, et cela, une seule exception prs, manifestement au sens de "genre" ou "espce."1 Cet tat de choses donnera en lui-mme prsumer que les sista " hommes de culture" de l'poque de Pnini taient dj familiers avec des lments de logique plus ou moins labore. Quant au rle capital que joue chez Patajali le mme mot dans ladite valeur, il ne serait gure opportun de le dcrire ci dans tous ses dtails. Bornons-nousnoterque, en posant la fameuse quadripartition des mots, Patajali reconnaissait en jti "genus" la premire des quatre "causesde production (des mots)" ' (sabda) pravrttinimitta' bien que ce dernier terme tel quel soit d'origine post-patajalienne.2 Ce n'est videmment point l'auteur de ces lignes, dont le souci habituel se concentre sur l'aspect "opratoire" de la Grammaire, qu'incombe latche d'exposer comment la notion de jti "genus" allait revtir une importance croissante, ceci sur le plan dcidment philosophique, chez les thoriciens ultrieurs reprsentant diverses branches ou tendances de l'rudition indienne. Il sera, par contre, assurment de notre devoir de rappeler expressment que le plus ancien essai connu de dfinir le terme en question remonte une poque pr-patajalienne, tant attest sous la forme de deux krik cites au dbut du Bh. ad 4 . 1 . 63.3 Nous marquerons dsormais par [A] et [B] respectivement, soit les deux krik elles-mmes, soit les dfinitions qu'elles donnent du terme jti tel qu'il est nonc (JATEH l'Ablatif) dans le s. 4 . 1 . 63.4 Voici maintenant le texte de la portion intresse du M_ahbhsya(d. Kielhorn, 11,225,13-21): JATER ity ucyote, kjtir nma / opara aha [A] ' akrtigrahana jatir [B] 'pradurbhavavinasabhyam sattvasya yugapad gunaih / lingnm ca na sarvabhk \ asarvaiingam bahvarthm sakrdkhytanirgrhy torn jti m kavayo viduh / / ' gotram ca caranaih saha //' gotram ca caranni ca //
kah punar etayor jtilaksanayor visesah/ yath prvam jtilaksanam (A), tath kumrbhrya ti bhavitavyamj yathttaram (B). tath kumrabhrya iti bhavitavyam / /

1 Cf. O. Bhtlingk, Panjni's Grammatik (Leipzig, 18872), p. 232*, s. u. jti La seule exception est JTYANTT dans le s. 5. 4. 9: il s'agit ici du mot jtipris en tant que pure forme. 2 Bh. ad praty.-s. 2 vt. 1 (d. Kielhorn, 1,19, 20sq.): catustayi sabdnm pravrttih. jtisabd gunasabdh kriysabd yadrcchsabds caturthh. 3 Le fait n'a que peu retenu, semble-til, l'attention des spcialistes modernes de la philosophie indienne du langage (d'autant moins peut-tre que Bhartrhari, pas davantage que Helarja, ne reprend ces krik dans son "jtisamuddesa", ainsi que s'intitule le chapitre initial du 3e knda, "pada-k" ou "prakJrnaka-k", du Vkyapadya). Parmi les plus rcents

ouvrages, on ne trouve aucune rfrence ces krik chez MM. D. Seyfort Ruegg, B. Bhattacharya, R. C. Pandeyaet Kunjunni Raja; une rfrence chez Mlle M. Biardeau, v. ci-dessous n. 5; une rfrence chez M. H. Scharfe, v. ibid. 4 S. 4 . 1 . 63: JTER ASTRIVISAYD AYOPADHT // (ATAH 4) (PRTIPADIKT1) (STRIYM 3) (NTS 40). La mention mdiane du s. soulve un problme extrmement compliqu: cf. notre Sur /'nonc paninen ASTRJVISAYA ( 4 . 1 . 63) : deux interprtations et leur rapport avec (e Ganaptha (Adyar Library Bulletin, vols. 3 1 - 3 2 = V . Raghavan Felicitation Volume, p. 125 sqq.)

426 Yutaka Ojihara

Nous estimons aussi de notre devoir de prsenter ici un nouvel essai de traduction des krik, d'autant plus que les traducteurs qui nous ont prcd ne semblent pas en avoir saisi la porte implicite compte tenu de l'observation terminale de Patajali, 5 o, doivent entrer en jeu des considrations bel et bien " o p r a t o i r e s . " Dfinition (A) 6 : "Est une jti: (1) ce dont la comprhension surgit en fonction d'une forme corporelle (dtermine) 7 ; (2) ce qui (a) n'est pas susceptible de (valoir travers) la totalit des (trois) genres (grammaticaux) 8 et qui, (b) une fois (qu'on l'a) signal nommment (, titre d'instruction, par rfrence , un individu donn), (nous) est (dsormais) reconnaissable (sur d'autres II n'existe notre connaissance que deux essais de traduction savante des krik : 1 L. Renou, Terminologie grammaticale du sanskrit (Paris, 1942), p. 148: [A] "ce qui se laisse percevoir par sa forme (krti; ou comprendre comme forme N.), ce qui ne participe pas tous les genres ( = n'est pas un adjectif), se laisse sitt nonc reconnatre (en d'autres individus Pr.), et (dsigneen particulier) lafamille (gotra) ainsi que les sectes {corana)1' . . . [B] "ce qui fait apparatre et disparatre l'objet, est en connexion avec les qualits en mme temps (qu'avec l'objet), ne comporte pas tous les genres, a des sens multiples (atteignant tous individus Pr.) ". 2 M. Biardeau, Thorie de la connaissance et philosophie de la parole dans le brahmanisme classique (Paris, La Haye, 1964), p. 48: [A] " Lajti, c'est ce que fait apprhender Ykrti, et elle ne participe pas de tous les genres (grammaticaux). Reconnue aprs avoir t conce une seule fois, elle est aussi le gotra et les diffrentes coles vdiques". . . . [B] "Du fait que chaque tre individuel apparat et disparat en mme temps que ses qualits, les potes savent que la jti comporte une multiplicit d'objets". H. Scharfe, Die Logik im Mahbhsya (Berlin, 1961), p. 141 se contente de reproduire la traduction Renou 'wegen ihrer[ = zwei Zitatverse] Schwierigkeit', dit-il. 6 La krik [A] se retrouve cite et commente non seulement, comme il va de soi, dans les deux sous-comm (entaires) du Mahbhsya, savoir Pr (adpa) de Kaiyata et Ud(dyota) de Ngesa, mais dans plusieurs autres ouvrages: ainsi, pour ne nommer que ceux que nous avons consults, la
5

Ksika sous 4.1. 63 et le N(ysa) ad loc. ; la S(iddhnta) K(aumud) 518 avec ses sous-comm., notamment le Laghus(abdendusekhara) de Ngesa et la B(la)M(anoram); le Sabdakalpadruma (s.u. jti) contenant une trs utile citation de Durgdsa, grammairien du dbut 17e s. (citation puise de toute vidence sa "Subodh", comm. sur le Mugdhabodha de Vopadeva: cf. Aufrecht, Cat. Cat., I, p. 256); le Nyyakosa (s.u., la krik n'tant d'ailleurs explique qu'en partie). Ces textes, toutefois, ne laissent sentir la divergence des opinions que sur le deux points signals ci-dessous nn. 9 et 10, si bien que, dans les notes qui suivent, nous n'indiquons en principe que celle d'entre les gloses indignes qui constitue la source la plus directe du passage concern de notre article. 'krtigrahan' = avayavasamnivesavisesavyangy (Pr.) A ct d'avayavasamnivesa, le mot samsthna est employ non moins souvent pour gloser 'krti'. Il semble que, en ce qui concerne un animal, seul passe pour un avayava ce qui fait partie de son ossature: les deux cornes sont censes tre des avayava d'une vache, ayant donc rapport Ykrti de cette dernire, alors que nos cheveux sont considrs comme choses extrieures n'affectant point notre krti (cf. ci-dessous nn. 17 et 22). Ainsi donc, nous rendons 'krti' dlibrment par "forme corporelle". Les auteurs ultrieurs citent surtout souvent ce premier quart de la krik [A], en le rapprochant non sans raison du Nyya Stra 2. 2. 67: ' krti r jti lingkhy'. 8 ' imgnm na sarvabhk' (seil, y) sarvni lingni na bhajate (N.).
7

427 Jti "genus"

individus) 9 ; ou encore, (3) une appellation patronymique conjointement (chacune des appellations des) diverses coles de science sacre,10 Dfinition (B)11 : (1) "Ce qui (a) (, bien qu'tant d'existence perenne, se manifeste et se cache) mesure que nat et prit une substance (donne, tout en s'unissant cette dernire) en mme temps que diverses qualits (s'y unissent comme on le sait),12 qui (b) n'assume pas tous les (trois) genres (grammaticaux), et qui (c) vise plusieurs (individus de manire en tre concomitant),13 c'est l ce que les sages comprennent sous le nom de jti" (2) (il faut comprendre,) en outre, une appellation patronymique, et (chacune des appellations des) diverses coles de science sacre. La dfinition [A] est tripartite: 14 la clause [1] dtermine la notion de jti proprement dite de telle faon que, pour peu qu'il satisfasse la prsente condition, un mot donn s'intitulera un "nom gnrique" (jtisabda) sans qu'il / a i t plus considrer s'il satisfera ou non les conditions ultrieures (notamment [2a]); la clause [2] a pour effet de rhabiliter au rang de jti, ceci sous la double condition [a + b], les mots qui ont t limins par [1 0 ] 15 ; la clause [3] rhabilite encore, titre d'ailleurs nettement accessoire, certains d'entre les mots qui ont t limins deux
' sakrdkhytanirgrhya' = ekasym vyaktau kathand vyaktyantare kothanam vinpi sugrah (SK.). D'aprs Durgdsa(cf. ci-dessus n. 6), ainsi que le N. du moins en puissance, la prsente teneur se rattache aussi la clause [1 ], d'une manire d'ailleurs toute diffrente de celle dont elle fait partie de la clause [2]. Sabdakal pad ru ma (s. u. jti) : . , . ajtahamsasya hamsarn drstavato 'pi tasya samsthnena hamsatvam vyajitum nasakyate, tihamsasypijtitvam nyatam t'iprvalaksanasya dosah. evam ... iti dvitiyalaksanasya dosah. iti dosadvayam apakarturn dvayor laksanayor visesanam ha sakrd.. .iti. . .tena, Idrso hamsa ity upadese hamsam drstavatas tasya samsthnena hamsatvam vyajitum sakyata evti prvalaksanasya na dosah. 1 ' gotram caranaih saha ' = apatyapratyayntah skhdhyetrv ca sabdah (SK.). Pour ce que la glose adhyetr implique au sujet de 'carana', cf. cidessous n. 26. Le terme 'gotra' est prendre ici dans sa valeur "courante", signifiant donc lapatya(mtra)\ Le seul dissident, ce dernier propos, est Ngesa dans son Laghus., qui veut mme ici la valeur "technique" du terme. Pour l'opposition laukikagotra/ pribhsikagotra, cf. notre KsikVrtti (adhyya I, pada 1) traduite et commente, 3e partie (Paris, 1967),
9

p. 108, n. 13. 11 A part Pr. et Ud., le Laghus. est le seul commenter la krik [B] parmi les oeuvres nommes ci-dessus n. 6. 12 ' prdurbhvavinsbhyrp sattvasya' = sattvasya [=]dravyasya prdurbhvavinsbhym yvirbhvatirobhvau prpnoti. yavaddravyabhvimty arthah (Pr.). virbhvet/d/. nityatvd utpattivinssambhavd evam uktam (Ud.). ' yugapad gunaih ' = gunair yugapad (seih ya) dravyena sambadhyate. yath nirgunasya dravyasyopalambho na bhavati, evam tirahitasya (seil. dravyasya) apty arthah (Pr.). L'essence de la clause rside dans la premire moiti de la teneur, par contraste avec la moiti restante qui ne constitue qu'une constatation explicative. Ud.: gunair iti. dam atra laksane svardpakathanam eva. 13 ' bahvarthm ' = sarvavyaktivypinlm ... arthasabdo %tra visayavdd (Pr.) visayavct/.. . jtysrayatvd vyaktlnm visayatvoktih (Ud.), 14 BM.: . . . iti cet, na [ ! ] . . . bhdsyoktatrividhajter vivaksitatvd ityabhipretya bhsyoktatraividhyam prapaeayatikrtigrahan jtir ti, prathamti sesah . . . 15 Pr.: aprptaprpanrtham cdam vacanam. Ud.: prvena (seil, vacanena) yatrprptam jtitvam, tadartham dam (seil, vacanam) ity arthah.

428 Yutaka Ojihara

reprises par [1] et par [2]. La dfinition [B] est bipartite si l'on y intgre la clause complmentaire [2], due peut-tre Patajali lui-mme (et dfait identique [A 3]): la clause [1], propre la krik et remarquable par son tendue, vise unifier deux clauses distinctes dans l'autre dfinition, savoir [A 1] et [ A 2 ] , en sorte qu'un mot donn passera pour tre un " nom gnrique" s'il satifait lafois la triple condition [a + b + c], dont le terme essentiel, malgr sa teneur trop elliptique pour ne pas tre obscure, est videmment [a], surtout sa premire moiti 16 (tandis que, en substance, [b] ne diffre pas de [A 2 a], ni [c] de [A 2 b]). Illustrons maintenant comment fonctionnent l'une et l'autre dfinition. Dfinition (A): [1] lgitime comme " n o m gnrique" go "vache", tant donn une " f o r m e corporelle" spcifique (caractrise par deux cornes, etc.) 17 ; tato " rivage" pour la mme raison, ceci sans tenir compte de [2 a], cd., que le mot vaille ou non travers les trois genres grammaticaux (cf. ci-dessous [B 1 ]18). [2] lgitime comme " nom gnrique" brhmana " brahmane', mot qui, bien qu'limin par [1] (la " f o r m e corporelle" dont il s'agit n'tant pas autre chose que celle d'un tre humain quelconque),19 satisfait [a] par l'absence de genre nt.,20 ainsi que [b] du fait que l'tat de brahmane, signal sur un individu donn, ne manquera jamais de s'appliquer, sinon ses enfants (qui peuvent tre d'une caste mixte), du moins, ses parents, grands-parents, etc.21 Un nom de qualit ou adjectif (gunasabda), ainsi munda "chauve," est limin, non seulement par [1] (la " f o r m e corporelle" de ce dont il est question n'tant pas change par l'absence ou la pousse des cheveux),22 mais aussi par [2] ceci dfaut de satisfaire [a].23 Ce n'est donc pas un " n o m gnrique." Un nom propre (samjnsabda), ainsi Devodatta, est limin, non seulement par [1 ] (pour la mme raison que celle note ci-dessus pour brhmana), mais aussi par [2] ceci dfaut de satisfaire [b]. 24 Ce n'est donc pas un " n o m gnrique," sauf l o s'applique

[3].
[3] lgitime comme " nom gnrique" Aupagava, nom patronymique 'Upagu, tir avec le suffixe [an] -a- selon 4 . 1 . 83 et 92 25 ; Kotha, nom d'un des corana yajurvedin.2*

Cf. ci-dessus n. 12, in fine. N.: gotvdayo hi visndimatsamsthnavyangyatvd krtigrahanh. 18 BM. : jalasamipapradesa krtivisistas tatah, atas tatatvam krtivyangyatvj jtih. 19 N.: etena laksanena brhmanatvadayojtivises na samgrhlth. na hi te samsthnavyangyh. . . ydrsam hi samsthnam brhmanasya, ksatriydeh samsthnam api tdrsam eva. . . 20 Ceci conformment Amara 3. 5. 37: ' strpumsayor . . . dvicatuhsatpadoragh'. 21 BM.: brhmanatvam tu putrapautr17

16

dau yady api na sugamam brhmant ksatriyym utponnasya brahmanatvbhvt (seil, kirn tu mrdhavasiktatvt), tothdpi pitrdau sugamam eva. 22 BM. : mundatvam nma viluptasarvakesatvam. tat tu nakrtivyagyam, kesadasym api tadkrteh sattvt. 23 BM.: sarvalingatvt. Le mot, en effet, est attest assez souvent au nt., comme dans siro mundam. 24 Ud. : na devadattatvam kvacit pindantare pratiyate. yady api pindntarasya devadatta iti samja, tathpi vcakasdrsyam eva, na tv arthasdrsyam tatra pratiyate.

429 Jti "genus"

Dfinition [B] : [1] lgitime d'un seul coup go et brhmana chacun comme " nom gnrique," [a] tant videmment satisfait de part et d'autre, aussi bien q u e [ b ] (== [ A 2a], v. ci-dessus); le genre nt. fait dfaut galement go) et [c] ( = [ A 2b], v. ci-dessus) ; de mme, tata malgr le t r i p l e genre assign au mot (depuis Amara 1. 9. 7) ceci du fait ou bien que la condition [b] n'est que de nature approximative (pryika), ou bien que ledit t r i p l e genre (notamment, la reconnaissance de tata- nt.) n'est d, qu' une e r r e u r chez des lexicographes. 27 Les noms propres sont dfinitivement limins par [1], dfaut de satisfaire [c] ( = [A 2b], v. ci-dessus) exception faite d'ailleurs de ceux d'entre eux qui font l'objet de [2] ( = [A 3], v. ci-dessus) ; les adjectifs le sont galement, ceci dfaut de satisfaire [b] ( = [ A 2a], v. ci-dessus). T o u t cela veut-il dire que l'une et l'autre dfinition mnent partout un mme rsultat? La rponse doit tre ngative, puisque Patajali nous y contraint par sa remarque terminale (d'ailleurs f o r t nigmatique). Considrons le cas mis en cause, celui du mot kumra " j e u n e h o m m e , " la lumire des deux dfinitions: Voici un " nom g n r i q u e " suivant la dfinition [ A ] , le mot s'accommodant avec la clause capitale [1] du fait que la jeunesse est bel et bien dote d'une " f o r m e c o r p o r e l l e " spcifique, caractrise par la finesse dans l'ensemble des membres. 28 A en juger par la dfinition [B], ce n'est pas l un " nom gn r i q u e " ceci dfaut de satisfaire [1a], en ce sens que la j e u nesse n'est pas du t o u t de nature se manifester aussi longtemps que dure telle ou telle " s u b s t a n c e " sattva' = dravya29). Voil donc au moins un cas de divergence entre les deux dfinitions de jti, et cette divergence n'ira pas sans se rpercuter, le cas chant, sur le plan " o p r a t o i r e " : Au mot kumra-, conu comme " nom g n r i q u e " suivant la

25

Le mot est limin par [2], ceci non un kriysabda et non un]tisabda{N.= pas en raison de [b], qu'il pourra Durgdsa), et ce n'est que grce satisfaire d'une faon analogue celle la mention ' coronoih saha' dans la note plus haut pour brhmono, mais prsente dfinition dejat/, que lui du moins en raison de [a]: un mot sera transfre la qualit de " m o t patronymique s'emploie bel et bien gnrique" (Laghus.).N.: caranosomme au nt., ainsi oupogovorn kulam. bdas tv odhyoyonokriyosombondheno Cf. Amara 3. 5. 45 sq. : ' ondyants . . . provrttotvt kriysobdo evo [no tu] trisu '. jtisobdo ity otos todorthosyo jtitvo26 A en juger par certaines obprotipdonyo caranaih sahty uktom. servations indignes, il n'est plus Ajouter les lments mis entre question ici, semble-t-il, de discuter crochets au texte, d. Chakravarti, I, comment le mot sera limin tant par 862,16. 27 [1 ] que par [2]. Le mot en cause n'est, Laghus. : l'mgnm co no sorvobhk strictement parlant, pas la mme [A 2a] iti. . .toc co pryikom ti kecit. chose que le nom du Sage Katho mais, . . .totam trisu [Amara 1.9,7] tydi il en est un driv lointain, ralis en kosokrtm promdoh, tyonye. 28 passant par 4. 3. (104 et) 107 ainsi que Pr.: kaumdram. . .krtigrahanatvj 4. 2. (59 et) 64, le sens littral tant jotir. . . Ud.: krtigrahanatvd iti. donc "celui qui rcite le texte probolyogotosksmovoyovosornnivesosyo mulgu par (le Sage) Kotho" (Laghus.); sorvotrokojatiyotvad iti bhvoh. 29 se rapportant ainsi l'action de Pr.: uttoretu loksone koumrom jtir rciter sur le plan de sa "cause de no bhovoti, oyovoddrovyobhvitvt. production " , le mot est par dfinition

430 Yutaka Ojihara

dfinition [ A ] , s'appliquera comme de juste le suffixe fm. [ns] -Jselon le su. 4 . 1 . 63 (en raison de l'nonc JATEH), d'o le driv fm. kumrJ- " j e u n e f i l l e " (en liaison avec 6, 4.148): spcifions cette dernire forme comme kumn-[A]. A la mme base, estime toutefois non " g n r i q u e " suivant la dfinition [B], le suffixe fm. applicable ne sera plus le suffixe [ns] 4 . 1 . 63 ( dfaut de satisfaire JATEH), mais le [np] -J-que le s. 4 . 1 . 20 enseigne pour une base notant le " premier g e " (VAYASI PRATHAME), d ' o , kumrJ- au fm. (comme prcdemment): mettons ici kumdn-[B]. Aucune diffrence, 30 protestera-t-on sans doute. Mais qu'on garde prsent l'esprit ce fait d'importance capitale, savoir que kumn-[A] est un " nom g n r i q u e " tandis que kumar-[B] ne l'est point. O r , lorsqu'il s'agira de f o r m e r selon 2.2.24 un compos bahuvrhi correspondante l'expression analytique ' kumrJ bhry yasya' " c e l u i qui a pour pouse une jeune fille (avant la pubert)," il y aura certes lieu d'tudier comment l a f o r m e provisoire kumnbhry- va tre affecte par la " masculinist ion " (pumvadbhva) dont t r a i t e la section 6. 3. 34-42. Le m o t fm. kumrJ-, tant ici appos l'autre fm. bhry-qu\ suit, fera-t-il bien l'objet du su. 6. 3. 34 enjoignant la " masculinisation " prcisment pour un tel lment? 3 1 Mais attention ! cette " masculinisation " sera p r o hibe son t o u r par le su 6. 3. 4 1 , quand le m o t fm. en question est un " nom gnrique. 3 2 C'est dire que t o u t dpend maintenant de la nature du fm. kumrJ- : S'agit-il ici de kumrJ-[A], " nom g n r i q u e " ? Alors, c'est la rgle ngative 6. 3. 41 qui d o i t e n t r e r ici en vigueur de manire entraver la " masculinisation " selon 6. 3. 34: kumri-bhry- passe kumrJ-bhrya- par application du s. 1. 2. 48 t o u t seul. S'agit-il de kumn-[B], qui n'est pas un " nom g n r i q u e " ? La " m a s c u l i n i s a t i o n " selon 6. 3. 34 doit alors y prendre effet, sans que la p r o h i b i t i o n pose par 6. 3.41 intervienne en aucune manire: kumn-bhry- passe kumra-bhry- d'abord, puis kumrabhrya- selon 1. 2. 48. Nous voici parvenu, esprons-nous, claircir toute la section intresse du 4 . 1 . 63 Bh. Il ne restera qu' savoir, titre d'pilogue, si Patajali portait ou non sa prfrence tacite l'une ou l'autre de ces deux dfinitions dejti. Rappelons, tout d'abord, que ce qui a t observ ci-dessus concernant le mot kumra- (fern. kumrJ-) vaut aussi bien pour le mot yuvan- "jeune" (fm. yuvoti-, selon 4 . 1 . 77 en liaison a v e d . 4. 17 et 8. 2. 7), dans la mesure o il est dit d'un tre humain.33 Or, Patajali lui-mme cite, dans son Bh. ad 4.1.1 vt. 19 (d. Kielhorn, II, 195, 7), la forme yuvatitar- ou la base prsuffixale
30 Ceci mme au point de vue accentuel: kumr(-[A], le ton suffixal -- d 3.1. 3 tant seul maintenu selon 6.1.158 (et vt. 9 ad loc); kumri-[B] = kumr0',-l-[kumra-J-, le [np] -/- tant atone selon 3.1.4, tandis que l'undi-s. 418 (numrotation d'aprs SK.) fait valoir l'oxyton kumra- en tant que nom d'agent driv de kam- "briller" (!). 31 S. 6. 3. 34: STRIYHPUMVAD... SAMNDHIKARANE STRIYM

... // (UTTARAPADE 1.). S. 6. 3.41 : JTES CA// (STRIYH PUMVAD 34) (NA 37). 33 Noter toutefois que ces deux mots ne sont pas tout fait synonymes, yuvan appartenant un stade de maturit plus avance que kumro: ' mrdusamsthnavyangy kumratvajtih, kathinasamsthnavyangy yuvatvajtih', ainsi qu'il est dit dans un sous-comm. du Laghus.
32

431 Jti "genus"

yuvati- se t r o u v e m a i n t e n u e l'abri de la " masculinisation " selon 6. 3. 35 34 ceci, f o r c e nous est de l ' e n t e n d r e , par l'application

yuvati- de la rgle prohibitive 6. 3. 41 portant sur un " nom gnrique" fm. : autrement dit, la dfinition [A] de jti est sous-jacente au prsent emploi patajalien. S'il en est ainsi, va-t-on conclure avec Kaiyataque Patajali tait partisan implicite de la dfinition [A]?35 Non, dclare Ngesa: qu'on n'ignore pas un autre emploi patajalien, qui va prcisment l'encontre d'une telle conclusion: savoir la forme yuvajni- qui figure dans le Bh. ad1.1. 57 (d. Kielhorn, 1,142,11 sq.) en tant que compos bahuvrhi reprsentant 1 yuvati r jy y asya, et qui doit prsupposer la dfinition [B] de jti, puisque la " masculinisation " 6. 3. 34 s'est bel et bien effectue sur le membre antrieur yuvati-, au mpris total de la prohibition 6. 3. 41 portant sur un " nom gnrique" fm. La conclusion de Ngesa est que, tant donn ces deux tmoignages contradictoires manant d'un mme matre Patajali, les deux dfinitions de jti, [A] et [B], sont estimer gales en autorit. 37 *
34

S. 6.3.35: TASILDISV KRTVA = SUCAH// (STRIYH PUMVAD . . . 34). Le suffixe comparatif -tora-, [tarap] 5. 3. 57, est enjoint certes dans les limites inclusives 5, 3, 7 (enseignant [tasil] -tas) 5.4.17(einseignant [krtvasuc] -krtvas). 35 Pr.: prvoktam eva Iaksanam bhsyakrasyabhimatam.. .tath ca yuvatitar ti NYP stra udharanam dadau, TASILDISU iti prptasya pumvadbhvasya JTES CA iti nisedht.

Le passage de jy - jni- s'explique par 5. 4.134 et 6.1. 66. 37 Ud. : vastutas tu iaksanadvayapranetrkarsidvayaprmnyj jtikryasya yuvasabddau vikalpah. ata eva yuvajnih (ity atra pumvattvam, Laghus.), yuvatitar (ity atra tadabhvas ca, ibid.) itydau dvividham apy anusthnam bhagavatah. * Nos plus vifs remerciements sont dus M. Jacques May pour avoir donn tous ses soins mettre au point la prsente rdaction franaise.

36

Louis R e n o u (18961966)

This selection of readings is completed with some specimens of the work of Louis Renou (1896-1966), the late Professor of Sanskrit at the Sorbonne, who was not only the greatest Western Sanskritist of the last decades (and "the most complete Sanskritist," as V. Raghavan aptly called him: 1956, 20) but also the leading expert on the Sanskrit grammarians. Renou, who had studied with Sylvain Lvi, contributed to all branches of Sanskrit and Indology (see Filliozat 1967a and 1967b for an evaluation of his work and for a bibliography). In the field of vykarana he published and translated a very technical text of the twelfth century, Saranadeva's Durghatavrtti (in two volumes, each in three fascicles, 1940-1956). He published a dictionary in two parts, one dealing with vykarana and the other with the Vedic linguistic treatises: Terminologie grammaticale du sanskrit (1942; 19572 cf. Thieme 1958). This finally fulfilled Schlegel's wish (see page 57), though it was in fact preceded by Hamilton's booklet of 1814 (see page 54). He brought out a new translation of Pnini (1948-1954; re-edited together with the text of the stras: 1961 ; cf. Thieme 1956), largely based upon the interpretations of the Kos/kdand of Purusottamadeva's/isvrtt/ (twelfth century). He translated, together with Ojihara, the beginning of the Ksik. Apart from these major publications, Renou published numerous important monographs, smaller studies, and articles. Basic information and references are contained in three contributions which provide together the best introduction presently available to the study of the Sanskrit grammarians: (1) the Introduction to the Durghatavrtti (1940); (2) Section VII of the " Introduction gnrale," an annotated re-edition in French of the Einleitung of Wackernagel's Altindische Grammatik (1957); and (3) the article " Pnini " in Current Trends in Linguistics, volume 5 (1969). Renou also completed two series of studies: tudes pninennes (five articles in the Journal asiatique in 1953 and 1956) and tudes vdiques et pninennes (16 volumes, 1955-1967). It is clear from this incomplete sketch that it is difficult to make a representative selection from Renou's work on the Indian grammarians. Four articles will be republished here. Renou devoted a number of studies to the place which the vykarana occupies in the whole of Sanskrit technical and scholarly literature. Since few have been as qualified as Renou to cover authoritatively so wide an area, and since this volume has so far concentrated mainly on the Sanskrit grammarians themselves without effectively showing what an important role they played in Indian culture and civilization, three of these not primarily historical but conceptual and comparative studies, are included. They relate to ritual studies, philosophy, and poetics, respectively. The fourth article deals directly with Pnini's treatment of a grammatical topic. In all his work Renou made use of numerous abbreviations. Those with which only Sanskritists may be expected to be familiar are in the list of abbreviations on pages xxii-xxv. The first and longest of Renou's comparative studies on grammar is devoted to the ritual. We have seen on page 90 that Patajali compared grammatical rules for forms that are not in use with the rules framed by the ritualists for the performance of sacrificial rites that are in fact never performed. Indeed, the existence of the Sanskrit grammatical tradition would be wholly unintelligible without the equally unparalleled background of the Vedic ritual. In the following study, Renou relates the analysis of grammar to the practice of Veda recitation; shows how the concept of

433 Louis Renou

stra ' r u l e ' is used in ritual and grammatical texts and how the concept of paribhs ' m e t a r u l e ' arose; studies the use of a number of technical expressions in ritual and grammatical stras; and finds parallels in the ritua! literature for many of the basic concepts and terms of grammar. Renou's article bristles w i t h names and references t o this ritual literature, which is vast and complex. A brief sketch of the structure of this collection of texts may therefore be in order. The four Vedas themselvesRgveda (RV), Yajurveda (YV), Smaveda (SV), Atharvaveda (AV)are subdivided into various schools or recensions, many of which have developed around a corpus of four collections of texts: the samhit (S) consisting mainly of hymns and ritual formulas (mantras); the brhmana (B) consisting mainly of prose interpretations of the samhit, often from a ritualistic perspective; the ranyaka (A), appendixes t o the former; and the upanisads (U), f u r t h e r appendixes tending toward more general speculation. Each of the schools possesses technical texts in stra form which describe the ritual in precise detail. The most fundamental among these are the srautastras (SS), dealing w i t h public ceremonies. N e x t come the grhyastras (GS), dealing w i t h domestic ritual. N e x t come a variety of texts, for example, the dharmastras (DS), relating t o religious law, or the prtiskhyas (Pr), dealing w i t h recitation, pronunciation, and phonetics of the particular Vedic school t o which they belong. The following table provides a simplified outline of the main names. Many of the stra texts belong t o the period 500-200 B.C. (cf. Liebich's conclusions w i t h regard t o Praskara- and svalyanagrhyastra, page 158). Ritualism developed later into a fullfledged system of philosophy, the Mimms (see page 286). Renou reverted several times t o the topics treated in the following study, last perhaps in "Sur le genre du stra dans la littrature sanskrite, "Journal asiatique (251,1963,165-216). The article reproduced here is "Les Connexions entre le rituel et la grammaire en sanskrit," Journal asiatique (233,1941-1942,105-165).

spas

434 Louis Renou

rd rd C rd t->

rd

+-

_c

>
rd C

>

E rti
+-* <

rti

rti

rd
JC:

ird r~

iid

8. (J -^

2 e
CL -

i<

o
"O CL

<

Z < u

>

rd

>

E irti

A. Les Connexions entre le rituel et la grammaire en sanskrit (1941-1942)

Louis Renou

Si ladatation de lagrammaire de Pnini1 demeure aujourd'hui encore incertaine, malgr les efforts qui ont t faits depuis longtemps pour la fixer,2 en revanche la situation de ce texte relativement la littrature indienne en gnral ne laisse gure de place au doute. On a eu depuis longtemps le sentiment,3 et plus prcisment depuis les travaux de Liebich,4 que la description pninenne s'applique un tat de langue qui, en littrature, est reprsent au plus prs par les traits du rituel domestique. On peut dire: du rituel en gnral, si l'on limine dans les Stradu culte solennel les archasmes nombreux, provenant des mantra qu'ils reclent, dans les dharmastra les vulgarismes ou dialectismes qui ont port croire que certains de ces traits chappaient la norme des grammairiens.5 Comme enfin la littrature des Brhmana n'est pas au fond sur un autre plan que celle des Stra rituels et qu'on a renonc instituer une priode des Stra indpendante de la priode des Brhmana, on peut mettre en fait que les sources littraires de Pnini sont la prose vdique dans sa totalit. 6 Sans doute on rencontre de temps en temps l'affirmation que Pnini dcrirait une bhs, une " langue parle" distincte de la langue religieuse. Mais en ralit les rgles de sa grammaire qu'il donne pour applicables la bhs sont de nombre et d'importance minimes; on n'a jamais su linguistiquement en tirer aucune donne prcise. A l'exception des quelque 250 stra validit vdique, qui sont signals par la mention chandas! ou un terme analogue, l'ensemble de la description englobe indistinctement la langue
Abrviations : les abrviations usuelles des textes sanskrits, et notamment P. Pnini, M. Mahbhsya, RV. Rgveda, AV. Atharvaveda, Br. Brhmana (AB. Aitareya Br.,SB. Satapatha Br., etc.), SS. Srautastra (KSS. KtyyanaSS. etc.), Pr. Prtiskhya (RPr. Rk Pr., APr. Atharva Pr. etc.), M. Mmms(sutra), Ve. Vednta(stra). 2 C'est plutt par l'effet d'une conviction subjective que par suite de preuves directes qu'on maintient P. au IVe s. avant l're, date o avaient conduit le situer les premires recherches, fondes sur des indices incertains. Voir sur ce problme la bibliographie chez Winternitz Ind. Litt. Ill p. 383 Keith Skt Liter, p. 425, auxquels ajouter plus rcemment Pathak Ann. Bhand. XI p. 59 Skld Papers on P. p. 24 Man. Ghosh P.Siks p. LU C. V. Vaidya Skt Liter. 3 p. 146. 3 Bradke ZDMG. XXXVI p. 470 Kielhorn IA. XV p. 87 Bhandarkar J BoRAS. XVI (1885) p. 20 Wackernagel Ai. Gr. I p. XXXIII ; les rserves de Delbrck Ai. Synt. p. VII Franke BB. XVI p. 74 demeurent thoriques. 4 Dans son Pnini (1891) notamment, o, se fondant sur les formes ver1

bales attestes, I tablit que la description de P. repose sur un tat trs voisin de, sinon identique , celui des grhyastra et que AB. et BAU. en particulier sont antrieurs P. Dans un article des BB. XI p. 308 il prcise que toutes les rgles casuelles de P. concordent avec l'usage de l'AB.: rsultats qui ont t leur tour tendus par Wecker (BB. XXX p. 91 ) aux Upanisad anciennes, en sorte qu'on admet couramment depuis lors que BAU, ChU. au moins sont prpninennes, et probablement l'ensemble des Up. dites "anciennes". 5 Sans qu'il faille ncessairement en tirer les conclusions chronologiques que Bhler Sacred Books of the ryas I. p. XLVI ZDMG. XL p. 705 croyait pouvoir formuler. Aussi bien avant qu'aprs P., tout un ct de la littrature sanskrite demeure hors de l'influence des rgles des grammairiens. 6 Les essais qu'on afaits pour dlimiter les textes connus de P. aboutissent des conclusions analogues : cf. depuis Weber Ind. Stud. I p. 141 et Goldstcker Pan. p. 129 jusqu' R. Mookerji IA. Lll p. 21 et surtout Thieme Pan. and the Veda, passim.

436 Louis Renou

religieuse et la langue profane ou bhasa. Bien loin que la bhsa s'oppose la langue religieuse, elle en est si proche qu'une mme description, sauf dans des cas exceptionnels, suffit rendre compte des deux domaines : autrement dit l'Astdhyay se situe en un point de l'volution indienne o la langue profane commenait peine manifester son autonomie en regard des disciplines du culte, de l'exgse et de l'cole brahmaniques. Les monographies qui ont t consacres tudier tel ou tel t r a i t de langue mnent presque invariablement souligner les accointances de l'Astadhyay avec les faits de la prose vdique pris dans leur ensemble, expurgs de leurs rsidus archaques. 7 Il est vrai que la plupart des traits de langue anciens que nous trouvons chez Pnini sont attests nouveau dans la l i t t r a t u r e classique, quelque moment de son dveloppement: mais c'est que cette l i t t r a t u r e s'est emplie peu peu, par purisme ou par pdantisme, de formes qui ont t calques sur les enseignements grammaticaux. La manire mme dont Pnini les prsente, en jonction avec les faits du domaine vdique, par exemple l a " t m s e " du prverbe et du verbe (BSL. X X X I V p. 96), laisse dceler qu'il les puise dans une t r a d i t i o n t o u t e voisine de l'tat ancien, " la f r o n t i r e entre le domaine surann du chandas et le terrain encore anonyme du sanscrit qui n a t " (S. Lvi MSL. X I V p. 279). Quelques-uns de ces traits, d'ailleurs, n'ont pas survcu la prose vdique; leur prsence chez Pnini prend ds lors une porte chronologique eminente: tel est le cas pour les drivs en -uka(avec construction accusative), qui sont limits aux Brhmana; pour l'emploi du gnitif partitif, qui n'a gure dbord la t r a d i t i o n vdique; defaon moins stricte, pour lavaleur " hors l a v u e " (paro'kse) du parfait, la valeur " d ' a u j o u r d ' h u i " (adyatane) de l'aoriste, ainsi que pour nombre de prescriptions particulires visant l'emploi de la voix moyenne. Inversement les stra ignorent les faits de langue que nous voyons attests partir de l'pope: ainsi il limite laformation des noms-racines en han- aux trois termes brahmahan-, bhrnahan-, vrtrahan-, ce qui reflte exactement, mantra mis part (cf. le bahulam III 2 88), l'usage vdique. 8 Il ne mentionne que l'auxiliaire kr- dans laformation du parfait priphrastique: c'est aussi le seul qu'utilise normalement la prose ancienne. 9 Il ignore en fonction de p r o h i b i t i f l'impratif avec ma, et plus f o r t e raison le f u t u r : le second emploi est inconnu aux textes vdiques, le premier n'y figure que trs accidentellement. De mme il ne donne pas la construction accusative pour le verbal en -tavant-, laquelle est postvdique; il ne parle pas (sauf dans un cas d'espce) de l'utilisation Voir, pour le subjonctif ou l'infinitif, nos Monogr. sktes I p. 43 ou II p. 57. Dans quelques cas, il est vrai, la description de P. ne concide pas entirement avec ce que les textes nous livrent, ainsi quant l'absolutif en -am dit namul (cf. MSL. XXIII p. 373) ou quant l'adverbe en -tarm (IHQ. XIV p. 132) : mais le principe de laformation est tout de mme ancien, l'image qui en rsulte est archasante et au total des cas de ce genre ont nettement moins d'importance que les faits inverses de concordance.
7

Quelques formes nouvelles, visiblement empruntes la languefamilire, apparaissent ChU. VI115 2 sq. ; par ailleurs on ne trouve gure qu'une formation instantane, matsyahanSB., ou analogique de formes mantriques, yajahon-. 9 L'auxiliaire as- apparat de faon trs isole dans des textes rcents SSS. GB. et portion finale de AB.; bh- est plus rcent encore. On sait que Patajali a rintroduit artificiellement la mention de sa et de babhva dans la teneur des stra.

437 Le Rituel et la grammaire

comme prdicat du participe en -vas-, qui a disparu aprs les mantra pour ne reparatre qu'avec l'pope; il n'autorise pas les formes du type dantena, pdena qui supplantent en skt classique les formes radicales data, pad, et qu'enseigneront les grammairiens partir de Candra; il ne mentionne pas l'usage, si habituel partir de l'pope, du gnitif en fonction du datif d'attribution. 1 0 Il est vrai qu'en revanche quelques faits propres la prose ancienne sont ignors de Pnini, ainsi l'optatif en -ayta des Brhmana tardifs et des Kalpastra. Mais il est prsumer qu'il s'agit l d'un tat de choses que Pnini entendait ne pas sanctionner. O n n'en saurait dire autant des faits prcdemment cits. Il y aurait lieu d'examiner l'ensemble des mots donns en exemple chez Pnini ou qui drivent expressment d'une rgle 11 : on verrait qu'une p o r t i o n notable d'entre eux se retrouve dans les textes vdiques ou appartient la zone vdique de la phrasologie. 12 O n a fait valoir que plusieurs prescriptions de Pnini se rapp o r t e n t un usage parl (Wackernagel A i . Gr. I p. XLIII) ; combien plus visent des aspects religieux. Le nom d e p a t n i l V I 33 est dfini par la participation de l'pouse au sacrifice, yajasamyoge: dfinit i o n qui surprend Patajali et l'amne expliquer une expression comme vrsolasya patn "pouse d'un hors-caste" par: femme "assimile une patn" ; les grammairiens ultrieurs, ainsi Candra II 3 30 ou Hemac. Il 4 51 t r o u v e n t plus simple d'adopter une dfinit i o n profane. Les allusions aux faits rituels sont nombreuses dans l'Astdhyyi: on en t r o u v e propos du t e r m e samstva- III 3 31 (et cf. prastra- 32), des drivs en -ya- btis sur un nom de yajus IV 4 125, en -ika- sur un nom de sacrifice V 1 95, en -Jya- sur un nom d'officiant 135 (et cf. le suffixe -iva-136), propos des dvandva dsignant des coles ou des rites II 4 3 sq. Le phnomne de la pluti, qui n'est pour ainsi dire pas attest hors de la langue sacre (Wackernagel op. c. p. 298), rsulte de l'observation des rcitations liturgiques, de mme que l'ensemble si minutieusement ordonn des rgles sur l'accentuation, que les grammairiens postrieurs, Usage que Candra notera de manire adventice. Les exemples date ancienne sont trs rares, ainsi ChU. 111 3 :cf. OerteISBBayer. 1937 8 p. 132 KZ. LXIII p. 206. Le mot bhavisnu- n'est donn par P. que comme vdique : il ignore l'emploi qu'en fait l'pope; au reste, dans l'ensemble de laformation en -isnu-, l'adhsion la prose ancienne est sensible. Les adjectifs en -ma-, ceux en -trima- mis part, datent de l'pope et comme il est naturel ne sont enseigns que chez Patajali; de mme les adjectifs du type sudarsana-, duryodhana-; de mme l'emploi de paraspara-, le type tmantrtlya-, la construction avec l'accusatif des prpositions rte et vina. Tout le traitement des prpositions chez. P. est archaque; cf. aussi le fait que les drivs en -as- sont transfrs
10

massivement dans les undi au lieu d'tre traits avec les autres suffixes primaires dans le straptha. 11 Cette recherche n'a gure t reprise depuis que Schroeder ZDMG. XXXIII p. 194XLIX p. 161 asgnale les termes manant de MS. et de KS.; cf. aussi quelques indications de Bradke ibid. XXXVI p. 470 pour les textes Mnava, de Thieme Pan. and the Veda passim pour les mantra en gnral. 12 Bornons nous signaler les noms en -yya-, lment -y- radical, que P. cite III 1 127 sqq. : la plupart sont des termes rituels : smnyyaparicyyakundapyya- dhyy- upacyya- (sur ce dernier, v. Schroeder I.e. p. 163 et HirSS. XXV 4 47). Ktyyana ad P. 11 6 signale comme superflue la mention des racines didhl- vevl-, parce qu'elles sont limites au Veda,

438 Louis Renou

dans la mesure o ils en tiennent compte, rangeront parmi les faits vdiques 13 ; de mme encore l'extension donne la thorie du gotro, les prescriptions relatives aux astrismes, etc. Lorsque Wackernagel op. c. p. LXIV n. conteste que la grammaire soit sortie de l'tude du Veda, la remarque est juste dans la mesure o il appert que Pnini vise une description systmatique qui dpasse les cadres de l'exgse et les besoins d'un srotriya; mais on ne saurait d o u t e r que sagrammaire ne plonge dans un milieu o les proccupations religieuses taient au premier plan et d o n t elle a conserv l'empreinte l mme o elle s'en est mancipe. Et d'abord dans l'enseignement de la phontique: l a t h o r i e de la gemination dans les stra se fonde visiblement sur le varnakrama (le t e r m e lui-mme de kramo " m a r c h e pas pas" est l'un des modes de rcitation du Veda) tel qu'il est expos dans les Prtiskhya. Peu i m p o r t e que ces textes soient antrieurs ou, comme l'affirmait rcemment M. Thieme avec des raisons trs valables, postrieurs Pnini (le peu qu'ils apportent de terminologie grammaticale p r o p r e m e n t dite a t o u t e s les apparences d'tre e m p r u n t aux grammairiens). Il existait un courant puissant d'tudes sur la phontique applique au Veda, et lagrammaire " g n r a l e " , le smnya que vise APr. I 2, en a subi l'effet: comment expliquer autrement l a t h o r i e du pragrhya chez Pnini, dans laquelle plusieurs rgles n'ont de valeur que pour le padaptha et dont le nom mme et son analyse par pra-groh- rappellent ncessairement la technique du padaptha ?
13 L'un de ces stra accentuels est commun P. et au KSS 1819 (cf. aussi VPr. 1130 sq.), comme l'a reconnu Weber ISt. X p. 423 qui conclut une source commune aux trois textes. Il s'agit de la description de la monotonie (ekasruti) et, sauf peut-tre la premire mention (durt sa mbuddhau), tout ce qu'en dit P. n'a d'application que dans la liturgie, comme le prcisera Ktyyana ad P.1 2 37 (v. le dtail chez Thieme Ind. Cu. IV p. 205 [this volume, pages351-356]). En l'empruntant comme il est probable P., le KSS. a restitu au rituel un bien qui lui est propre. L'aikasrutya est dfini svSS. I 2 9 " rapprochement maximum (parah samnikarsah) des tons udatta, onudtta, svarita" : cette dfinition singulire s'inspire peut-tre de celle de P. I 4109 sur la somh/t. 14 C'est ce qui demeure de plus certain aprs la polmique qui a oppos Bat. Ghosh IHQ. X p. 665 Ind. Cu. IV p. 387 NIA. Il p. 59 (qui voyait dans P. 11 16 un emprunt direct RPr.) Thieme IHQ. XIII p. 329 Ind. Cu. V p. 363 Chattopdhyya IHQ. XIII p. 343 Keith Ind. Cu. Il p. 742 Chaturvedi NIA. I p.451. 15 Ces drivs en graha (pragrhya se dit pragraha dans TPr. et HirSS. XXI 2

33) sont d'ailleurs propres aux Pr. o ils dsignent des faits dupada-ou du krarna-ptha : avagraha, que P. emploie incidemment, avec la mme nuance que VPr., pour dsigner le phonme aprs lequel intervient une "sparation " dans un mot complexe du padapatha (cf. avagrah- "tenir spar" PB. VI 7 22 et l'emploi plus technique de l'expression padvagrham AB.SSS. "en sparant les mots" au cours d'une rcitation); vigraha "tat disjoint" d'un mot, i. e. son emploi simple, hors de la composition : terme que la grammaire reprend au sens de "rsolution, analyse" d'un compos; parigraha "entourage" de iti par un mot rpt, terme emprunt au rituel qui dsigne par l l'encerclement de la vedi au moyen de tracs concentriques; upagraha "annexion " d'un-e en finale LSS, Vil 811 DSS. Il 2 4 PuspasG. p. 518, ou d'un visarga SamhUB. p. 17 : terme propre aux "anciens matres" d'aprs la Ks. ad P. VI 2134 comme dsignation du gnitif, ou d'aprs Helar, ad Vkyapad. Ill 121 comme nom de la "voix", de la diathse verbale. Il y a. encore udgrha, nom d'un samdhi chez RPr., etnigraha, nom d'un dosa dans le mme texte.

439 Le Rituel et la grammaire

Ce n'est pas un hasard si le mot somhita qui dsigne la " d i c t i o n c o n t i n u e " chez Pnini dsigne aussi un mode de rcitation du Veda: survivance manifeste en grammaire d'un vaste emploi labor dans les cercles de phontkiens du Veda, et qui asa rsonance mystique particulire dans les Aranyaka (A. Ill 1 5 S. Vil 2 sqq. et cf. aussi TA. VII 3 1, o l'enseignement de lsamhit est qualifi 'upanisad " d o c t r i n e secrte"). 1 6 Le mot onorsa chez Pnini " propre (au t e x t e , i. e. au podaptho) n'manant pas d'un rsi" appartient aux mmes milieux, et il faut observer que la plupart des mentions de docteurs cits dans les stra de Pnini sont faites propos de donnes de phontique: c'a t l visiblement le point de dpart de l'enqute grammaticale. La structure de l'Astadhyay est la mme que celle des textes rituels: c'est la composition en stra ou " a p h o r i s m e s " comme on t r a d u i t conventionnellement 1 7 : phrases en prose domines par le souci de la concision, qui lui-mme rsulte d'exigences mnmoniques. A vrai dire ce nom de stra couvre des textes de type assez divers: par exemple dans la t r a d i t i o n du rituel le Vdhlastra ou le Baudhyana ne sont que des Brhmana affubls du nom de stra; inversement le Smavidhnabrhmana et les Brhmana mineurs du Smaveda laissent transparatre la composition en stra. Ce mode de rdaction a d se constituer une poque prcise; il rpond la pense de verser en manuels la masse des enseignements, oraux condenss dans les coles vdiques; ce qui ne veut pas dire qu'on n'ait pas de t o u t temps, et quelquefois par simple pastiche, continu compiler des stra: ainsi les tardifs s t r a d u Smkhyafaits l'imitation des textes des autres darsana, ou bien les stra des grammairiens htrodoxes, depuis le Ktantra jusqu'au Saupadma, faits l'imitation de Pnini, ou encore les stra de Ruyyaka traitant de la rhtorique, 1 8 les Silpastrade Nrada,
16 II est probable aussi que le terme tout voisin de samdhi, vit dans P. et, relativement encore, dans M., et que le PradpaadSivasu. 3-4 dclare emprunt aux "anciens matres", remonte aussi l'enseignement vdique o nous le voyons, travers les Pr, avec des inflexions beaucoup plus riches que dans les textes grammaticaux. Comme samhit, samdh'i comporte dans les ranyaka un jeu migrammatical, mi-mystique difi sur la notion de "jonction ", A. Ill 1 2, 2 2. 17 Le terme sort apparemment de la technique du tisserand, comme tantra qui dsigne les actes auxiliaires du processus rituel, qui se retrouvent toujours les mmes d'un rite l'autre et constituent la "trame" du sacrifice (le terme apparat aussi en grammaire partir de M. avec une lgre translation de sens: "valeur normative" d'une dsinence, etc., dans un nonc; et cf. svatantra P. dit de l'agent, "qui a sa propre trame", donc

autonome); cf. aussi les multiples mtaphores de la langue vdique fondes sur la racine tan- associe yojo- odhvara- dhl- etc. (et, absolument, SB. XIII 2 5 2) et sur lesquelles repose l'emploi gnralis de cette racine dans le Kvya. Mme acception rituelle pour tontuSB. XI 5 5 13. Une dfinition tardive du stra est celle du Madhvabhsya (d. C. Palle p. 10) "ce qui a aussi peu de mots que possible, ne prte pas amphibologie, contient l'essence, est universel [visvotomukho: ou " multiface", comme propose Ghate Vednta p. 5], n'a pas de (mots superflus comme sont les) stobha, est au-dessus de toute critique". [Depuis Vyup. LIX 142]. 18 Mais les autres rhtoriciens rdigent sous forme de "versets" ou krik; il n'y a pas lieu de prsumer que ces krik remontent des stra perdus, mme pour ceux, tels Vmanaou Mammata, dont les manuscrits ou les commentaires donnent souvent l'uvre comme tant faite de stra.

440 Louis Renou

les Bhaktisutra et les Saivastra qui traduisent les spculations des religions sectaires.19 Les stra de haute poque sont, outre ceux de Pnini (auxquels on peut joindre les Phitstra qui les suivent de trs prs par le style et par la terminologie), les Prtiskhya et leurs annexes (Upalekha, Bhsika, Upanidna, etc.) et plusieurs Anukraman, celles de Ktyyana notamment. Les stra de Pigala sont de type plus archasant qu'archaque. Dans le rituel l'apoge du genre est marque par le srautastra, relativement tardif, de Ktyyana, mais tous les traits du culte solennel ou priv, textes srauta, grhya, dhorma, sulba (etc.) y participent, bien qu'avec de notables diffrences de forme. Aux dharmastra on peut rattacher les aphorismes 'arthasastra attribus Brhaspati. Le style nominal, reprsent de faon rigoureuse par les stra grammaticaux 20 et que reprendront les stra philosophiques, 21 cde la place, dans le rituel, un style verbal caractris par l'indicatif descriptif, l'optatif prescript if, l'absolut if d'enchanement temporel : cet tat de choses dcle l'influence vdique, marque une priode que n'a pas encore atteinte la prvalence de l'expression abstraite. L'ablatif de cause dans les noms abstraits, dont les origines en sanskrit ne semblent pas remonter au-del des Upanisad

En revanche, les natastra mentionns par P. ( ct des bhiksustra, ceux-ci d'interprtation incertaine) semblent, concuremment d'autres indices, autoriser poser un Ntyasstra primitif deforme stra, cf. De Skt Poetics I p. 27. On l'a revendiqu aussi pour le Kmasstra et l'Arthasstra: nos textes de Vtsyyana et de Kautilya prsentent en effet des traces de rfection, et les anciens stra percent et l sous le revtement de bhsya. De mme, dans les parties introductoires du Nirukta, le style stra est encore reconnaissable sous la prose continue. Le procd consistant noyer des stra ou des vrttika dans un bhsya a t frquent dans la littrature technique, depuis le Mahbhsya jusqu' certains commentaires du Nyya par ex. Une autre forme de dgradation des stra est. la mise en sloka: ainsi dans le Rkprtiskhya, o il est assez ais de dceler les stra primitifs, conformes ceux de tous les autres Pr. Dans les textes rituels aussi le stra et le bhsya alternent; plusieurs srautastra sont crits, en totalit ou en partie, dans une prose de type brhmana. Mais comme ce sont les textes les plus anciens de la srie, il est difficile de croire que le style stra y a t aboli au profit du style brhmana. Peut-tre y aurait-il profit

rechercher dans un srautastra comme celui de Baudhyanas'il y a effectivement des traces de forme stra: Hillebrandt GGA. 1903 p. 949 l'affirme et a mme relev des lments de vrttika (il cite Bau. II15), qui montreraient que la dialectique des grammairiens et des logiciens a exist aussi, du moins titre fragmentaire, dans la littrature rituelle. Pnini n'emploie comme forme personnelle du verbe que drsyate (en regard de darsant prfr par les stra rituels ou philosophiques, qui font grand usage du terme et avec des implications plus ou moins charges de sens); toutefois il utilise volontiers le verbe personnel pour dfinir des valeurs de suffixes, tad gacchati, tad adhlte, tad veda, tadarhati, etc. Cf. encore les tardifs yurvedastra. Il existe desSaktistra, ainsi que des stra tantriques, ceux-ci en consciente imitation du kalpa vdique, savoir les Prnanda, Agastya, Bhrgava. Jusque la Bhaktamla du vieux hind qui s'inspire du style stra (Grierson JRAS. 1909 p. 608).
21 20 19

II n'y a pas un seul verbe expressif, par exemple, dans les Yogastra. Les deux MTmms, et surtout la seconde, ont d'assez nombreuses formes verbales, mais de valeur faible, comme darsayati braviti pratisedhati upapadyate ssti smaryate.

441 Le Rituel et la grammaire

( O l d e n b e r g A i . Prosa p. 29 n. 1 e t Br. J e x t e p. 229 note comme exceptionnel l'emploi 'avinsitvt B A U . IV 3 23 sqq.), fait dfaut dans les stra grammaticaux, lesquels, fidles apparemment aux exigences premires du genre, dcrivent e t ne justifient pas22. Il est connu des Prtiskhya, mais d ' u n usage trs limit 2 3 . Dans le r i t u e l l'ablatif de cause est usit, mais de faon ingale: on le t r o u v e s u r t o u t dans les portions eparibhs; a t o u t p r e n d r e il est plus f r q u e n t dans le dharma que dans le srauta, o toutefois Ktyyana, Hiranyakesin e t Apastamba n o t a m m e n t (sur ce d e r n i e r , v. Garbe d. Ill p. xv) le prsentent de manire assez suivie. Il est probable que son dveloppement a une valeur chronologique. L'opposition si tranche qui se marque cet gard e n t r e les stra de la g r a m maire ou de la phontique et ceux du rituel ou de la philosophie rpond celle q u i , dans l a g r a m m a i r e mme, spare les stra des vrttika. Les stra rituels et s u r t o u t les stra philosophiques c o m binent avec l'ancien schma du stra descriptif le schma nouveau du stra dialectique et interprtatoire, dont le modle parfait est fourni par le grammairien Ktyyana. Une des exigences du style stra est la formulation desparibhs ou rgles gnrales d'interprtation : c'est l'une des six sortes de stra reconnues pareillement par lagrammaire et par la Mimms. On sait que Pnini a dissmin dans l'AstdhyyT, et surtout dans le premier pda, des stra qui constituent des axiomes valeur gnrale, " illuminant comme une lampe lagrammaire entire" (PradTpaad M. 11 49 vt. 4) et que latradition appelle des poribhsstra. L'un des plus notables est vipratisedhe param kryam I 42 " lorsqu'il y a prohibition mutuelle (entre deux rgles simultanment applicables, la rgle) effectuer est celle qu i est ultrieure". 24 Les philosophes du rituel ont emprunt cette paribhs; les Mi. Xl\ 4 37 donnent la formule vipratisedhe
22

La formule est deThiemeZII. VIII p. 24. Ce principe suffit nous permettre de juger comme interpols les s. I 2 53 57 de P. (sur lesquels v. en dernier lieu Faddegon Stud, on P. p. 57) qui, outre les ablatifs de cause, contiennent un emploi moderne du mot promana au sens de " norme " (P. n'a que le sens d e " mesure", de mme RPr. " mesure" d'un mtre); le sens de " norme", rare encore dans M., apparat dans un passage de PGS. 18 13 (qui se rfre la sruti) et dans LSS.-DSS. ainsi que dans une kr. Kaus. CXLI 24. Ils contiennent aussi un emploi moderne de l'optatif s/t, qu'on retrouve dans la Mmms et dj dans le rituel, notamment dans KSS. I 217, 6 18 IX 9 19 et plus spcialement dans syd v I 5 3, 814 VIII 219 etc.
23

discussion implicite entre un prvapaksin et un siddhntin la manire des vrttika. La phrasologie par na + ablatif figure VPr. I 3 comme KSS. I 418, 5 8 et sera reprise abondamment par les stra philosophiques, qui font usage aussi de l'idiome iti cen na, emprunt aux vrWka grammaticaux (l'emploi de ced dans P. et dans les Pr. est tout diffrent). Sauf erreur, on n'a dans aucun texte en stra l'instrumental servant justifier la cause fournie l'ablatif, instrumental dont useront tous les bhsya tard ifs. 24 Le terme vipratisedha apparat aussi dans le rituel en concurrence avec virodha (qui le glose chez M.) et, isolment, avec vibdhamna pSS. XXI 1 20 (qui montre un cas, rare date ancienne, de la racine bdh- dite de rgles qui s'entravent: emploi de M. et des grammairiens en gnral, repris aussi M. Ill 3 8), et de l dans les su. philosophiques: toutefois, dans les NyS. le mot a le sens de contreobjection V1 41 sq. ou mme d'un

On l'a au dbut de VPr. (ainsi que l'optatif syt), portion peut-tre surajoute au texte primitif; ainsi que VPr. V 43 sq. APr. IV 6 RPr. X 1 9 et notamment XI 66 sq. o figure une

442 Louis Renou

param, inflchie d'ailleurs vers une valeur diffrente " lorsqu'il y a prohibition mutuelle (entre ce qui est en vue du rite et ce qui est en vue de l'homme), c'est l'autre (i.e. ce qui est en vue du rite) qui est effectuer". 25 Mais les stra philosophiques connaissent aussi la prvalence de l'ultriorit en un sens voisin de celui o l'entendent les grammairiens. Elle est constante au fond de la notion a'uttara-mlmms applique au Vednta en tant que spculation postrieure et suprieure lafois la MTmms premire; elle est latente dans nombre d'argumentations de Sakara. Un des Ml. s., savoir VI 5 54, en fait tat lorsqu'il enseigne paurvparye prvadaurbalyam prokrtivot " lorsqu'il y a connexion entre lment antrieur et lment ultrieur, l'antrieur est le plus faible, comme dans l'archtype"; l'archtype en effet, bien que prescrit antrieurement aux ectypes, est plus faible que ceux-ci, doit autrement dit cder le pas toute rgle particulire: ici encore il s'agit de l'application d'un principe grammatical bien connu, aux termes duquel la rgle gnrale (utsarga, nyya, smnya) cde le pas la rgle particulire (apavda, visesa), cf. la paribh. 57 et le s. I 2101 du Sarasvatkanthbhar. utsargpavdayor apovdo vidhir balavn.2* Pnini, on le sait, n'a pas formul toutes les paribhs auxquelles son uvre pouvait donner cours : certaines, qui taient ncessairement prsentes son esprit, ont t laisses de ct, soit parce qu'videntes, soit parce qu'existant dans des recueils spars. En fait nous avons des recueils de paribhs, mais de date gnralement tardive et qui englobent avec ces authentiques axiomes des maximes nouvelles qui sont de simples gnralisations, d'autres qui sont dduites d'une rgle de grammaire en vue d'obtenir certains rsultats pratiques. Le problme se pose, qu'on ne peut aborder ici,27 de savoir lesquelles de ces paribhs sont d'origine pninenneou pr-pninenne. Rappelons seulement que les paribhs grammaticales sont plus rigoureuses que celles des techniques voisines : elles dsignent des rgles dont l'existence est requise par un indice rvlateur (jnpaka) puis dans la substance mme d'un stra; elles supposent un raisonnement d'infrence, fond sur le postulat implicite du caractre ncessaire de toute teneur pninenne. Hors de la grammaire, les paribhs se prsentent dans des conditions diverses. Dans les Prtiskhya elles sont assez rares et mal circonscrites; leur place, leur manque partiel de ncessit, d'autres indices encore, invitent les considrer comme de simples emprunts aux coles de grammairiens. Dans l'APr. nous n'en trouvons qu'une, savoir ntaryena vrttih I 95 " le traitement (des phonmes) a lieu d'aprs l'affinit" : cet axiome joue le mme
simplepratisedha 111 14. Aarthovipratisedha et sobdo(para) de M. rpondent les mmes expressions dans M. I 2 36 III 3 36 V1 26; cf. aussi rpa HirSS. 11 16c//wmaVSS. 11 1 70 M. III 3 40 ko\pQ SSS. XII114 5. 25 L'argument du parrthatva se retrouve ailleurs, avec des nuances diverses: M. 11 18 I sert justifier la thorie de la prennit des mots. Dans le rituel mme on le trouve appliqu aussi un cas de conflit entre deux possibilits contradictoires: (ptha kramrthakramayor) virodhe 'rthastatparatvt KSS. I 5 5 "lorsqu'il y a conflit entre l'ordre de la teneur et l'ordre du sens, le sens l'emporte parce que la teneur dpend du sens"; de mme I 416 orthadravyavirodhe 'rthasmnyam tatparatvt. 26 Inversement il y a parodaurbalya M. Ill 3 14 dans une squence de facteurs, qui rappelle celle de la paribh. grammaticale 38. 27 Cf. en dernier lieu Boudon J. as. 19381 p. 101. [this volume, pages 379-391].

443

role que vikn yathsannam VPr. 1142 (et cf. aussi yathntaram

Le Rituel et la grammaire

RPr. I 56) et repose comme ce dernier sur P. 11 50 sthne 'ntaratamah. Dans le TPr. les s. I 50 padagrahanesu padam gamyeta et 51 api vikrtam rappellent respectivement le pbh. su. P. 11 68 et la pbh. 37 ekadesavikrtam ananyavat.2* Le Rktantra et le Smatantra n'ont pas de paribhs. Le RPr. compte une maxime gnralisante, savoir nyyair misrn apavdn protJyot I 53 (Mller 54), qui est toute proche des paribh. grammaticales 62et 63: l'emploi du mot opavda "exception, rgle particulire-entravante" (qu'on retrouve plusieurs fois en ce sens RPr. et APr. dans la recension de Vishva Bandhu Sstr) sent son origine grammaticale.29 Une autre maxime sarvasstrrtham pratikantham uktam I 54 donne le mme enseignement que la paribh. 110; la formule padavac ca padyn I 61 (Mller 62) vulgarise P. 11 62, enfin padam padntdivad ekavarnam II 6 (Mller 110) rappelle dyantavad ekasmin de P. 11 21 (et cf. VPr. 1152 sa evadir antas ca)* Dans les textes rituels, les paribhs forment des sries plus ou moins compactes de stra situes gnralement en tte de la portion srauta; elles sont d'autant plus compltes, semble-t-il que le texte est plus rcent. Dans pastamba et Baudhyana elles se groupent dans une section finale. Les grhyastra, qui ne font que prolonger les srauta en guise d'annex, n'avaient pas besoin de paribhs spciales, et de fait n'en possdent pas ; toutefois quelques maximes de caractre secondaire marquent le dbut de plusieurs textes, Apastamba, Gobhila, Khdira, ainsi que Kausikaqui n'a pas de srauta derrire lui ; dans Hiranyakesin les paribhs sont disperses au hasard des cas particuliers auxquelles elles s'appliquent, conformment l'axiome grammatical (pbh. 3) kryaklam
28

Seul le su. I 25 avec son annexe 26 n'a pas de correspondant en grammaire: sannam samdehe. Le s. I 23 (avec 24) est plutt un samjnstra, dfinissant la valeur du nominatif dans un nonc. La pbh. grammaticale yanthntaram est reprise dans un texte de kalpa: Kaus. VII110. 29 De mme nyya, qui a aussi chez les grammairiens tardifs le sens de " rgle gnrale" opposkapavda; Ng. ad M. Ill 4 67 vt. 7/8 Pradpa ad M. Il 3 1 vt. 1 (imprim: nyyya). Le terme plus usuel est utsarga. Le sens 'utsarga est proprement "ce dont l'application est sujette tre leve (par une rgle particulire, par l'opavda)", valeur qui est sensible aux commentateurs (cf. Prad. V11 86 vt. 1 et la pbh. cite M. I p. 463 2) et qui explique aussi l'acception cT'lment original " (donc sujet tre lev par le substitut) que revt quelquefois le terme. Tout ceci drive du sens de 11 leve, suspension " connu par le rituel (v. les ex. chez BR.), d'o "renonciation" M. IV1 3, " r e m placement" d'une ide fausse par une ide vraie Sank ad. VeS. Ill 3 9. Le

sens de " rgle gnrale " n'a gure dbord les commentaires grammaticaux; cf. cependant Sabarasv. Ill 7 41. 30 Le VPr. est assez riche en paribhs, par quoi ce texte souligne son adhsion plus marque aux traditions grammaticales. Certaines de ces maximes sont empruntes littralement P.: savoir celles sur la valeur deseas, sur le locatif 1134 = P. 11 66, sur l'ablatif 1 3 5 = P. M 67, sur le gnitif 136 = P. 11 49, quoi VPr. ajoute l'accusatif 133 et l'instrumental 139. D'autres ont trait la manire de citer les phonmes I 36 sq. (cf. aussi TPr. 119 sqq.), au sens de iti ibid. (TPr. 116 sq.), la manire de dsigner un varga 64, la valeur englobante d'une voyelle brve 63 (RPr. I 55). La pbh. vipratisedha uttaram balavad alope 1159 suit de trs prs P. I 4 2; prvottarayor uttarasya 145 est analogue M. 11 66 vt. 3. Enfin samkhytnm anudeso yathsarpkhyam 143, presque identique P. I 3 10, se retrouve quant au fond dans la pbh. rituelle bahusu bahnm anudesa nantaryayogah avSS. II16.

444 Louis Renou

samjnparibhsam " samja et paribhs (entrent en fonction non au moment o elles sont enseignes, mais) au moment o les oprations (qu'elles concernent sont prescrites)". Le SnkhGS. prsente V 10 3 une paribhs noye dans la description. 31 Il est probable que plusieurs de ces maximes ont prcd le texte o elles figurent: on les retrouve sous des formes similaires dans la plupart des coles. La paribhs enseignant qu'on reconnat l'endroit o prend fin un mantra par le commencement du mantra suivant, paradina prvntah KSS. I 3 9, se laisse attester sous un aspect moins concis dans uttardih prvntalaksanam LSS. 11 3 DSS. I 1 3, uttarasydin prvasyntam_vidyt HirSS. 11 30, u prvasyvasnam BhrSS. I 2 1 , u pyidyt ApSS. XXIV 2 4, dinottarasya prvasyntam vidyt MnSS. 1115. Ces expressions rappellent curieusement les termes pardi, prvnta qui dans le Mahbhsya dfinissent la place des "accrments" entre le thme et le suffixe. Mais le schma dominant est celui qui met en relief une prvalence entre deux notions contradictoires (cf. dj ci-dessus p. 441), soit le type dravyasamskravirodhe dravyam ballyah/ arthadravyavirodhe 'rtho balJyn pSS. XXIV 3 47 sq., mantracodanayor mantrabalam KSS. I 5 7, vipratisedha rtur naksatram ca balJyah HirSS. Ill 2 30; analogue III 8 28 et toute la srie dbutant avec le su. 32, dans laquelle en particulier la juxtaposition angtpradhnam fait rplique exacte ce que serait un *upasarjant pradhnam en grammaire. Ce schma de prvalence est ralis aussi dans la Mlmms: vipratisiddhadharmnm samavye bhyasm syt sadharmatvam XII 2 22, agagunavirodhe 25. Il est trs frquent dans les paribh. grammaticales. C'est ainsi que la maxime prasangd apavdo balyan svSS. II 22 M la rgle particulire (ou : exception) prvaut sur la rgle qui devait se prsenter" (i. e. sur la rgle gnrale, smnyavidheh Nr.) reparat dans la grammaire et indirectement dans les Ml. Su., v. ci-dessus p. 442. De mme la paribh. srutir balyasy numnikd crt HirSS. XXV11 127 " un enseignement exprs prvaut sur un traitement rsultant d'une inference" (analogue vipratisedhe srutilaksanam balJyah XXVI 8 9), concide avec la paribh. pninenne n 104 srutnumitayoh srutasambandho balavn, qui elle-mme retentit dans l'opposition sruti/l'mga Ml. Ill 314. Ces rgles gnrales que nous venons de citer pour la Mmms ne constituent cependant aucun gard un corpus de paribhs analogue celui de lagrammaire ou du rituel. Les stra philosophiques sont de deux types, ou bien explicites (Nyya, Vaisesika, Yoga), et ils n'ont que faire des implications que les paribhs ont pour rle de dvelopper; ou bien elliptiques (les deux Mmmsa) et apparemment ddaigneux de faciliter au lecteur l'intelligence du texte. La concision dans les deux Mimms, qui conduit supprimer des lments essentiels et amoindrit en fait l'intelligibilit (Deussen Syst. d. V. p. 29 G hate Vednta p. 5 Thibaut Ved. S. transi. I p. xiv) est aux antipodes de la concision pninenne, o tout ce qui importe est formul.
Par contre, ce qu'on appelle le paribhss. de Bau(GS.) n'est qu'un fragment de brhmona superpos un complmentau GS., Caland ber d. S. d. Bau. p. 30. Sont aussi de simples maximes introductoires les s. initiaux de plusieurs textes de dharma, Baudh., Vas., etc. Hors de l on trouve encore des paribhs dans l'introduction la SarvnukramanT,
31

dans SVB. I 5 2 dans es sulbastra (Thibaut On theSu. p. 6), dans le dbut de Pig. Chandahs., dans la seconde strophe de l'Aryabhatya (trad. Clark p. X), chez Amara et d'autres lexicographes, dans les Prnandastradu tantrisme, chez Candra Liebich Konkordanz p. 49, chez Kaumra Lders SBB. 1930 p. 526.

445 Le Rituel et la grammaire

La composition en stra, sous sa f o r m e la plus stricte, entrane l'omission des mots qui se laissent dduire tacitement du stra antrieur et valent ainsi dans le stra suivant par l'effet de ce qu'on appelle anuvrtti ou " rcurrence". 3 2 Les commentateurs du rituel font tat quelquefois de V anuvrtti, par e m p r u n t vident aux choses de lagrammaire (ainsi ad KSS.'l 6 16, 8 8 et 46 VI11 33, 2 1 , ad HirSS. X X I V 8 37): mais il est clair qu'elle est loin de s'appliquer dans le rituel avec la rigueur que lui donnent les grammairiens, o les gloses en arrivent justifier, abusivement d'ailleurs, t o u t e rptit i o n de mots. Ici au contraire la rptition 3 3 n'est nullement vite et n'est jamais justifie. Les adhikra ou " rubriques gouverna n t e s " de lagrammaire, avec leur validit soigneusement mesure, n'ont qu'un assez faible cho dans le r i t u e l , o la n o t i o n plus banale de " c h a p i t r e " , " t t e de chapitre" 3 4 l'a e m p o r t sur celle de rubrique porte prcise, se confondant en fait avec une anuvrtti de longue dure. Dans la langue rituelle, adhikra est normalement la "qualification " effectuer un r i t e , le " d r o i t " , d i t le M. Nyyaprak. 225 (et cf. Sank, ad VeS. I 3 25), j o u i r du f r u i t qu'on attend de l'acte r i t u e l . Mais c'est l un aspect de l'emploi plus gnral de " c o n c e r n e m e n t " qui rend compte aussi du sens grammatical. Il est d i t ainsi d'un certain puisage, qui n'est qu'une modification du puisage du dadhi, qu'il r T ' e s t pas c o n c e r n " ou " c o m m a n d " par les ordonnances valant pour le soma, anadhikrtahf somadharmair dadhigrahavikratvt ApSS. XII 7 1 5 (analogue SOS. XVI20 3DSS.XII212). 35 Tous les commentaires de stra raffinent sur la valeur des particules, co v tu api eva hi atha: ils cherchent introduire par le biais de ces mots certaines indications smantiques qu'a manqu exprimer la composition lacunaire du stra. Le cas le plus remarquable est celui de la particule v, laquelle les stra grammaticaux confrent constamment le sens de " titre facultatif", qui repose en somme sur l'omission d'une des deux parties de l'alter-

Le terme est du M. et ne parat pas avoir franchi la zone des commentaires grammaticaux; cf. toutefois Sakaraad VeS. 11 24. Mais le verbe anuvrt- figure dans le rituel avec un sens trs voisin, ainsi svSS. V 3 11 au causatif: "faire valoir nouveau " dans une prire une divinit dj invoque. 33 Punargrahana KSS. I 4 6 sq., punahprayoga pSS. 1X1 7, anuvda KSS. IV 3 18, abhysa passim: mmes termes que chez les grammairiens. Les MT. su. ont aussi abhysa passim, anuvda II 4 25 III 412 (qui est comme en grammaire une " remarque complmentaire" sans valeur prescriptive, un " rsum" d'une chose dj dite); en outre, dvirukta III 6 2 punarukta II 4 8 punarvacana 11415. La nivrtti "cessation" 1 anuvrtti apparat pSS. X X I V 1 41 dans un sens presque identique celui que donneront ce mot les grammairiens.
34

32

vaksymah, que nous retrouvons dans les textes de philosophie. Le terme du rituel est prakarana ("contexte" en grammaire et dans la M.: ekavkyatm panno vkyasamhah dfinit le Jaim. Nyyamlvistara en termes nettement grammaticaux), celui de la philosophie adhikarana (en grammaire: " rfrence, plan " ) . Mais adhikra est connu de TPr. XXII 6 (VPr. IV177 incertain) APr. 11 utilise le terme prtijn et de quelques commentateurs du rituel, ainsi Nr. ad svSS. II 3 24 1111 1 VIII 717Dhanv. ad DSS. XII11 1 ; le sens de " s u j e t " d'un passage persiste dans les VeS. 11312, 21 III 3 3 et Sakara 11 19. 35 Emploi analogue dans la MT.: ainsi III 8 31 o le terme se dit de dtails de l'archtype qui n'ont pas " place" dans l'ectype, ou 22 de directives "concernant" une autre personne, ou encore III 3 1 de mots "qui visent" une espce; analogue III 2 20, 7 35
V 1 1 4 VI11 1 3 .

M a r q u e p a r l ' e x p r e s s i o n atha . . .

446 Louis Renou

native. Avec cet emploi caractristique concide un changement dans lafonction du m o t : va cessant d'tre enclitique peut figurer l'initiale du stra. 36 Cette acception se retrouve dans les Pr., quoique sensiblement moins dveloppe: on relve un exemple dans APr. (IV 27), un dans VPr. (1132), plusieurs dans les Phits. et dans le RPr., et dans ces deux textes, aussi, comme chez Pnini, l'initiale du stra ou du pda, savoir Phits. 112 et 16 RPr. VI 9 et 26. Le mme emploi a t not il y a longtemps (dans le BR.) pour le KSS. o il est assez frquent, et o il apparat galement l'initiale, ce que Bhtlingk omet de d i r e , soit 11 13 XI11 26 X X I I 3 5 X X I V 7 8. Il ne semble pas qu'aucun autre t r a i t de kalpa l'atteste. C'est un cas remarquable de convergence entre un t e x t e rituel et les coles de grammairiens. 3 7 La particle ca a pour effet d ' a t t i r e r dans l'nonc o elle f i gure un lment de l'nonc qui prcde: c'est l'anukarsana de Ktyyanaet de Patajali (leTribhsyaratnase sert du t e r m e anvkarsana), Vanvdesa de TPr. X X I I 5. Les commentaires rituels en font tat, eux aussi, d'ordinaire par l'expression karsana ou par la simple mention casabdt.3* L'emploi le plus notable est Vanuktasamuccayrthatva, c'est--dire ca = " e t c . " : il y a l comme on sait une valeur dont les grammairiens postrieurs Patajali ont fait grand usage (Kielhorn IA. X V I p. 251 [this volume, p. 133]) et qui leur a permis d'enrichir peu de frais l'enseignement des stra. La question s'est pose il y a longtemps si ce sens vaut galement pour les stra rituels et philosophiques, o les commentaires le donnent quelquefois (Vas. X I 2 Vaises. 11 6 NyS. V 2 1 d'aprs les Navya, etc.). 39 Il faut reconnatre qu'aujourd'hui o la littrat u r e rituelle est mieux connue, nulle part l'interprtation d'un ca par " e t c . " ne parat s'imposer. Le rituel et la philosophie ne sont
P. fait usage aussi des mots anyatarosym et vibhs (ce dernier, substantif fig), qui ne semblent gure avoir dpass la grammaire: cependant vibhs est attest isolment dans Puspas. VI 241 APr. I 2, et VPr. V15 a anyataratah. Les commentateurs de stra ainsi que les stra philosophiques (parfois dj ceux du rituel, v. BR. et cf. encore DSS. VIII 3 29, 41 ) ont souvent vikalpa (vikalpena, vikalpot) que les grammairiens utilisent aussi, mais qui est encore rare dans M. On trouve aussi, basse poque, pakse (paksena), ventuellement dvaidha. pSS. XII11 15 emploie krtkrta, glose vaikalpika. Les termes inverses nityam et sarvatra "defaon ncessaire" sont en usage dans le rituel comme en grammaire. 37 Par ailleurs les commentaires donnent va le sens de avadhrane (BR. s. v. v n 4), paksavyovrttau ou prvapoksanirsrthe, vikolpe. Yjnik. ad. KSS. 111 17 Mi 3 25 VIII 5 25f Nr. ad svSS. 112 5 connaissent le sens de vyavasthitavikalpa comme les gram36

mairiens depuis Candra (Kieihorn IA. XVI p. 251 [this volume, p. 132]). Il y a aussi un va crthe Nr. ad svSS. 11 25 Srbhsya II 2 38 III 3 54 samuccayrthah S lokav., codans. 109. Api v et na v, propres aux vrtt. grammaticaux, figurent dans les textes rituels ainsi que dans la M. (Paranjpe p. 58). Ca est aussi samuccayrthe "visant englober" (le terme grammatical de samuccaya se trouve aussi dans le rituel, v. BR.); accessoirement vrthe Yjn. ad KS. 11 10, avadhranrthe I 3 29, tvarthe HirSS. IX 7 9 SrTbhsya II 1 2, ivrthe HirSS. 11 14 vikalprthe II 3 4 etc.
39 38

Cf. ce sujet la vieille controverse entre Bhler W Z K M . l p. 13 affirmant l'authenticit de cette acception, et Knauer Fest. Bhtlingk p. 62 Bhtlingk ZDMG. X X X I X p. 484 XL p. 145 (ad Bhler X X X I X p. 706) XLI p. 516 Sachs. Ber. 1895 p. 152, qui ne voient l qu'une habile invention des commentateurs; cf. encore Franke Genuslehren p. 49 ZDMG. XLVIII p. 84 Aufrecht LU p. 273.

447 Le Rituel et la grammaire

pas comme lagrammaire des domaines o la pression de la ralit et l'volution de la langue tendent constamment faire clater les anciens cadres. 40 Maintenant le vocabulaire: et d'abord les termes qui, sans tre des noms techniques de grammaire, se rencontrent lafois chez les grammairiens et chez les ritualistes. Ainsi vidhi " rgle prescriptive". Le mot en grammaire s'oppose pratisedha " rgle p r o h i b i t i v e " en tant qu'il souligne le caractre positif de la prescription ; il s'oppose niyama " rgle qui restreint (une autre rgle a n t r i e u r e ) " en tant qu'il comporte un enseignement nouveau : aprvovidhih M. I. p. 31216. Ces oppositions se r e t r o u v e n t dans le r i t u e l , te) que l'labore du moins la Mlmms: le seul t r a i t qu'elle ajoute est que dans le vidhi le verbe est l'optatif ou l'impratif et ce fait concorde avec Pnini chez qui l'emploi fondamental de ces deux modes se note par le mot vidhi. Quant au reste, le vidhi de la Mmms est aprva comme celui des grammairiens (est aprva aussi \dLprakrti KSS. IV 3 21), cf. vidhir v sydaprvatvt M I 2 19, mais l'aprva est rig en un principe d'effet surnaturel de l'acte r i tuel (cf. le t e r m e voisin adrsta). Ceci est de la spculation : l'origine Qprva glose en quelque sorte le t e r m e codan " injonction " (cf. Sabarasv. ad II 1 5), qui lui-mme est un quivalent de vidhi" et que les grammairiens emploient sous la forme codita, codyate avec le sens de " requrir au moyen d'un vrttika". D'autre part l'opposit i o n vidhi/ pratisedha est prsente aussi bien dans le rituel (o protisedho s'applique naturellement aussi l'interdiction d'une chose concrte, d'un mets, etc.) qu'en philosophie (cf. l'index du Ml. Nyyaprak. ou I'Arthasamgraha p. v u ; " o b j e c t i o n " NyS.). 42
Mmes conclusions au sujet de iti, pour lequel les commentaires connaissent aussi le sens de "etc.": mmes rfrences, et en outre Keith J RAS. 1910 p. 1317 Gaastra JSS. p. XXVII. Le commentaire de HirSS. Ill 1 1 1 , l'Upaskraad VaisS. X 1 4 connaissent un iti prakre conforme aux grammairiens; iti hetau ad HirSS. XXVII 3 57. L'emploi de iti pour noter la fin d'une section est propre date ancienne au trait de Pingala et au VPr. (Weber ISt. IV p. 92); on le note dans les VaisS. en fin des sections II VI et X, ainsi qu' la fin des YoS. L'expression itikarana de M. et des Prtiskhya (R., V., Upalekha) se retrouve aussi SSS. I 2 25. Yjnik.ad KSS. I 7 8, 8 46 recourt l'artifice de Yavibhaktiko nirdesah des grammairiens (Kielhorn IA. XVI p. 249). L'emploi pregnant des cas qu'enseignent pour les stra grammairiens et phonticiens n'avait gure l'occasion de s'appliquer dans le rituel: cependant un gnitif sthdne est reconnu par ex. par Agnisv. ad LSS. VI1104. Un emploi remarquable chez. P. est relui de a et de prk rfrant un stra ultrieur: on les
40

retrouve dans le rituel, appliqus la fois un rite ultrieur et au passage textuel qui le dcrit: ainsi pour pSS. VII114 20 KSS. I I 2 2 X I I 4 1 , pour prk pSS. X 21 5 X I X 1 18. Le bahulom de P. n'a pas dpass le domaine voisin des P r : APr. (dans les deux versions) VPr. Nir. avec un trs petit nombre d'attestations. 41 Codan est attest dj dans le rituel (codana KSS.) et glos en gnral par vidhi. Le codansabda KSS. I 101 est l'impratif des praisa. Le terme est dfini kriyyh pravartakam vacanam Sabarasv. 11 2, c'est-dire sruti. Vidhi dsigne aussi un texte de brhmana (M. I2 53 II 1 30), alors que dans PGS., ct de vidheya mantra et de tarka = brhmana, il semble se rfrer au kaipa.
42 Le mot nisedha n'apparat pas en grammaire avant Kaiyataet Purusotama (on le trouve aussi dans quelques Siks), bien que la forme verbale nisidh- soit vdique et que nisedha mme s'applique des sman dans les coles du SV. Il est fort rare dans les stra philosophiques : un seul ex. NyS. et cf. nisiddha VaisS. V 2 23.

448 Louis Renou

Quant niyoma, le mot est aussi du rituel : comme en grammaire, niyamasabda KSS. I 4 8 dsigne la particule eva, et vkyaniyama M. I 2 32 vise la " f i x a t i o n (de l'ordre des mots) dans une phrase (de mantra)" ; la dfinition du Mm. Nyya-Prak. pakse 'prptasya yo vidhih 243 " prescription (non de ce qui n'tait absolument pas acquis auparavant, comme dans le cas du vidhi p r o prement dit, mais) de ce qui n'tait optionnellement pas acquis" est conforme l'esprit des grammairiens. Aniyama (VeS. Il 3 37, 51 et dj KSS. I 3 6_ainsi qu'an y ata) quivaut " a r b i t r a i r e " comme y thk m J ApSS. 11115 3 DSS. Il 2 20 (et v. BR.; kmya SSS.) : le Mahbhsya use du t e r m e voisin kmacra qui drive de SBLe rituel ne fait pas grand usage au sens de " r g l e " des termes yoga, Iaksana, sastra qui de faon variable ont accd cette acception chez les commentateurs grammaticaux. Cependant on a pour yogar(d'ordinaire " c o n n e x i o n , " comme d'ailleurs chez P. mme) KSS. 1216 (mais: sambandha Yjnik. ad l o c ) , pour sastra M. I 3 24; Iaksana ( " r g l e " M. et RPr. XIII 31 : 739), qui garde son sens propre de " c a r a c t r i s t i q u e " et que le Slokav. Il 9 glose par nimitta/pramna, a t supplant par l'mga.

On sait le rle essentiel que jouent les notions 'antaraga/ bahraga dans l'argumentation des grammairiens partir de M. Ces termes n'apparaissent gure hors de lagrammaire qu'en un passage des YoS. (cf. aussi Vcaspatim. ad YoS. Il 45 Smkhyaprav. Ill 35 IV 8 V115): cette concidence est mme l'un des arguments sur lesquels on s'est appuy pour tenter d'tablir l'identit des deux Patajali (Liebich SBHei. 1919 4 p. 7,1921 7 p. 57). Mais la valeur de ces termes est sensiblement diffrente dans les YoSQ. de ce qu'elle est en grammaire, savoir " moyens internes (immdiats)" et "externes (indirects)" pour atteindre au yoga.*4 Des termes importants du raisonnement grammatical se fondent sur les drivs de la racine dis-', nirdesa est n'nonc" M. (cf. dj nirdista P.), c'est--dire l'nonc par excellence, les stra en tant que principe d'autorit. Dans le rituel et la philosophie rituelle, T'nonce" est naturellement le texte de lasrut/, extrieur aux stra et qui les commande. De cette translation rsulte qu'a disparu dans le rituel tout le ct tonnamment formaliste de lagrammaire, l'autorit attache la lettre mme de l'enseignement, et les vastes consquences qu'en tirent les coles grammaticales. Le rituel, qui situe au del et hors des stra le principe d'autorit, n'a pas connu dans sa forme et dans sa teneur l'extrme exigence et les multiples implications qui ont t le fait de la doctrine grammaticale.45
43

Niyamo " r g l e " en gnral, NyS. III 211 ; " rgle de composition, nibandhona" M. I 3 12. Vyavasth note de part et d'autre soit une fixation qui esta instaurer sparment pour tel et tel cas d'espce KSS. I 3 4 M. Il 4 26, soit une restriction, quivalent passif de niyama. 44 Skstsvarpopakrokatvtl prarnparyenopakrakatvt Maniprabhad loc: d'autres commentaires mettent en vidence le caractre de ncessit de Vantaroga. 45 Nirdesa est usit aussi de manire

plus lche, en un sens voisin de vyapadesa (un autre driv de dis- commun lagrammaire et au rituel) "dsignation, nonciation ", ainsi VeS. Il 3 36,1 22 M. Il 4 26; "assignation" d'une oblation la divinit M. IV1 29 etc. D'autres noms de I'" nonc " ou "teneur" communs tous les stra sont vaccina et grahana. Le premier sens de grahana o tre la valeur concrte de "fait du puiser, de saisir pour l'oblation une part du liquide", atteste partir deSB.; de

449 Le Rituel et la grammaire

Le mot uddesa, que les NySu. opposent nirdesa comme " d o n n e gnrale" " d o n n e particulire", a pour contrepartie en grammaire, soit upadesa (mme sens, kr. chez M. I V 1 73 ; ou " e x p l i c a t i o n d i r e c t e " par opposition uddesa " e x p l i c a t i o n par description " M. I 3 2 init.), soltonudesa: I'uddesa est alors un lment de teneur qui "dsigne par avance" un lment ultrieur corrlatif appel anudesa. Upadesa est d'ordinaire "enseignem e n t " : toutefois le sens propre de " rfrence" est conserv P. I 4 7 0 ( parrthah prayogah Ks.) et repris NyS. 111 6 5 , 2 39 etc. ; le N i r . 11 III 21 entend par ce t e r m e le pronom soit dictique, soit anaphorique; 4 7 enfin le mot est rang VaisS. IX 2 4 parmi les synonymes de hetu "cause". desa est galement, au point de dpart, " indication, prescription " (Nr. ad AsvSS. 1118 glose par vidhi), notamment dans la f o r m u l e rituelle andese SB. SS. passim 11 lorsqu'il n'y a pas d'indication ( c o n t r a i r e ) " ; le mme sens a pntr dans plusieurs Prtiskhya (et dans des textes annexes de mme poque). Mais d'une part le t e r m e tend fonctionner en corrlation avec anvdesa pour dsigner " une premire affectat i o n " , prathamdesa, ainsi N i r . IV 25 o il sert pour les faits d'anaphore pronominale, et AsvSS. Ill 4 1 0 (cit dans la n. prcd.) o il vise l'emploi d'un mantra dans l'archtype. D'autre part des'a dans la langue des grammairiens a revtu ds l'origine le sens de " s u b s t i t u t " . O n voit bien comment cette acception s'est tablie : le substitut est la chose effectivement " n o n c e " en regard du p r i mitif qui s'efface devant lui : M. 11 56 aprs v r t t . 11 le dfinit pertinemment yo 'bhtv bhavati " c e qui est maintenant, n'ayant pas t auparavant". 48 Le t e r m e inverse est sthnin "lment p r i m i t i f " , p r o p r e m e n t " c e qui (tait) telle place" (et n'y est plus), sens que confirme l'autre valeur du mot, galement pninenne, "lment souse n t e n d u " , opp. kprayujyamna "lment effectivement e m p l o y " . Le rituel ne connat le t e r m e qu'au sens de " q u i est sa place" (AsvSS. IIIJ3 19 d i t d'une devat et glos kasmims cit karmani yastavy; ApSS. X X 1 1 18 d i t d'une donne qui " a s a place" dans

tel rite).
l "fait de saisir, de noter, notation, mention" : al'mgagrahane KSS. XV 2 13 "quand il n'y a mention d'aucun signe particulier", mantravidhis cdigrahanena LSS 11 2 " un mantra not par le dbut est prescrit (comme valant) dans sa totalit"; le terme est glos promena ad HirSS. Ill 1 2. 46 Mme sens 'anudesa VPr. 1143, qui concerne plus prcisment l'anaphore pronominale II 7. La valeur propre "nicdtton" est conserve KSS. XVIII 615. Uddesa "description" Nir. 115 ( = upadesa Skand.); "explication" XII 40(=prot/jndSkand.); "tradition " SSS. 1111. 47 Glos uccrana, abhivyakti Skand. Ailleurs le Nir. se sert au mme sens 'anvadesa, qui est aussi un terme de P., qui dansTPr. I 58 traduit de manire plus gnrale une " rfrence un lment antrieur" et qui dans le rituel, ainsi svSS. Ill 410, concerne l'emploi dans l'ect/pe d'un mantra appartenant en propre l'archtype, c'est--dire relevant de \'desa. Il y a ici comme ailleurs correspondance exacte entre les termes de lagrammaire et ceux du rituel, si loignes les unes des autres que soient les notions qu'ils circonscrivent. Upadesa syn. de vidhi Slokav., autp. s. 11. 48 Hors de la grammaire ce sens de "substitut" n'est pass que dans les tardifs Pr. de l'AV. ainsi que dans le Pratij. 22 et la YjnavSiks152. Il y est fait allusion NyS. Il 2 39. Les M. I 116 emploient varnntara et l'on a plus souvent, dans le rituel et dans la philosophie, pratinidhi, qui chez P. signifie "rplique, contre-partie"; cf. aussi upalingin Nir. X17.

450 Louis Renou

Enfin atidesa, rgle de grammaire corn portant^ un transfert d'application, figure avec une valeur analogue AsvSS. IX 1 2 sq. (et cf. atidis- SB. SS. cits chez BR.); le mot est repris VeS. Ill 3 46 (mais simplement " rfrence " NyS. V 1 6). C'est dans les M. Su. qu'il atteint toute son ampleur: il dsigne les rgles qui, formules pour l'archtype, sont transfres l'ectype sans avoir tre nouveau nonces l'occasion de chaque cas particulier. Un autre terme d'origine grammaticale, qui a pris une grande importance dans la Mmms, est celui de seso. En grammaire il caractrise certaines formations comme le bahuvrhi selon P. Il 2 23, le gnitif selon II 3 50, un groupe de taddhita selon IV 2 92 en tant qu'elles valent pour les cas " restants", dans les emplois qui " restent" couvrir, n'ayant pas t prvus par les stra antrieurs. Une valeur analogue se retrouve dans le rituel, ainsi VSS. 14110 (rare) et dans les M. Il 1 33 III 719. Mais en gnral, dans ce dernier texte, sesa dsigne un lment auxiliaire du rite, ce qui n'entre pas dans le schma du pradhna] c'est un autre nom de l'aga, et qui s'est donn un corrlatif dans ses'm = pradhna Ml. Nyyaprak. 105; il est dfini par la notion de parrthatva Ml. Ill 1 2 "(ce qui existe) en vue d'une autre chose". A vrai dire les ritualistes hsitent sur l'amplitude exacte du terme et Sabarasv. ad III 1 1 voque les deux interprtations de "auxiliaire" et de "cas restants", qui audit passage se trouvent concider. Le Vednta son tour oriente l'antithse sesa/sesin vers des valeurs spculatives.49 Reste le vocabulaire proprement grammatical. Ici le travail de Liebich nous prcde. Dans un mmoire paru en 1919 dans les Sitz. Ber. de l'Acadmie de Heidelberg sous le titre Historische Einfhrung und Dhtuptha, le regrett savant a signal de remarquables concordances entre les termes techniques de la grammaire et ceux des textes religieux. Mais son tude tait ncessairement incomplte puisque d'une part elle se fondait presque exclusivement sur lathorie du verbe, que d'autre part elle ne faisait gure intervenir le rituel et son prolongement mmmsiste. Les faits replacs dans un plus vaste ensemble confirment les rsultats auxquels Liebich tait arriv. L'unit auquel le grammairien a affaire est le mot. Le terme sabda qui signifie " m o t " dans la langue courante n'est pas vritablement un terme grammatical. C'est, dit M. I p. 1819, " un point dans l'espace comportant perception auditive, susceptible d'tre saisi par l'entendement, clair par l'emploi ". Mais ceci est une dfinition labore; le sens_premier est "son " : l'oreille est le lieu des sabda SB. XIV 5411 (BAU. Il 411), 6 2 6 (III 2 6): c'est la base des phonmes TPr. XXI11 (o le Tribhsy. glose dhvani) et, joint aux phonmes, c'est la base du langage XXIII 3 : il est produit par l'air avec mlange (samkara) [d'autres lments] VPr. I 6. Mais bien qu'en vdique (o le terme apparat depuis la VS.) sabda
49 Cf. sur l'emploi en MT. Strauss SBB. 1931 p. 265. Les formes si usuelles prpta et siddha, qui scandent l'argumentation dans M. (d'une part "ce qui est acquis" titre provisoire ou de manire inadquate, d'autre part "ce qui est acquis" titre dfinitif), et que reprendront les bhsya ultrieurs de toute espce, n'ont pas laiss grande trace dans le

rituel (prpti KSS. cit BR.; siddha dans la formule de Man. p. Bhr. passim siddham istih samtisthate "Visti mene terme d'aprs (le paradigme dj acquis"), ni dans les stra philosophiques (prpta M. Ill 6 3 aprpta I 29 asiddha VaisS. Il 2 29; vikraprpta NyS. Il 2 45). Sur utsarga et apavda, v, cidessus p. 442.

451 Le Rituel et la grammaire

dsigne aussi un simple bruit, 5 0 il s'identifie de bonne heure la parole, yah kos co sabdo vg eva s SB. X I V 4 3 10 ( B A U . I 5 3) " t o u t ce q u i est son est p a r o l e " . Chez les grammairiens et les phonticiens, sabda est un lment, mais non pas ncessairement un m o t : un suffixe, une dsinence (VPr. Ill 17), une syllabe (Bhr Siks26, 30, etc.), un phonme (P. IV 3 64 VIII 3 86); en t a n t que m o t , il s'applique toutes les formes (dsinentielles et mme drives) d ' u n t h m e donn. 5 1 C'est sabda, non pada, q u i figure dans les spculations relatives la p r e n n i t des mots (sabdanityotva, M. passim, M. 11 5 VaisSu. Il 2 32); lesabdrtho "sens des mots (en t a n t que groupes p h o n i q u e s ) " se distingue d u padrtha "sens du m o t " oppos au vkyrtha "sens de la p h r a s e " . Sabda apparat dans plusieurs passages d u rituel o il est question de substituer t e l " m o t " t e l autre dans un mantra ou un vidhi: il s'agit du m o t en t a n t que f o r m e nonce, non en t a n t q u ' u n i t grammaticale : l'expression sabdavikrp SSS. V11 3 dfinit ainsi Vha, la f o r m u l e sabde Wipratipattih KSS. I 4 9 signifie qu'en t e l cas " i l n'y a pas lieu de modifier un m o t " du mantra p o u r l'adapter un acte nouveau. 5 2 Le t e r m e grammatical est pada, q u i dsigne une f o r m e acheve en regard d e p r a k r t i ou aga (v. ci-aprs) " b a s e " , une f o r m e une en regard de vkya " e n s e m b l e constituant un nonc c o m plet, phrase". 5 3 C'est ce qui se t e r m i n e par une dsinence, e n seigne P. I 4 14 d f i n i t i o n que r e p r e n d r o n t les NyS. Il 2 56
Encore " s o n " danssabdakarman P., et aussi N i r. IX12. A. 1 1 1 2 5 joue avec le double sens de "son " du luth et "son articul" mis par le corps humain. 51 Du moins dans le RPr. IV 23 : 242. Plus frquemment en ce sens on a pravda, en plusieurs passages de RPr., de l'Upalekha et dans TPr. XIII 9 o le Tribhsy. glose "nonc (valable non seulement pour laforme telle qu'elle est pose, mais pour toute forme qui s'en distingue par) une diffrence de genre et de dsinence, par (adjonction d'un membre de) compos ou d'un suffixe secondaire". Lesens depravada dans le rituel est "mention " faite dans un mantra pSS. XVIII 2210 XIX 18 4MSS.V1 414 HirSS. Il 7 68 (o le comm. glose pada). Ce sens est l'origine de l'emploi des Pr. Le Nir. donne au mot la nuance d ' " expression " (et pratiquement "dsignation qualificative, pithte" 1113). On asabda au sens de "thme nominal" RPr. IV 52: 271 Uvata; "terme technique" de la grammaire TPr. XXII 3, glos sstra Tribhsy. 52 De mme dans les stra philosophiques o le terme, demeur en grande partie voisin de dhvani, s'oppose kartha "sens"; "tmoignage oral", c'est--diresruti, NyS.ll 22.
50

Le sens prcis de " phrase " , i. e. "verbe avec adverbes, dterminants casuels et leurs appartenances ", ventuellement "verbe (seul)" M. Il 1 1 vt. 9 sq. n'est ralis que par la grammaire, o d'ailleurs la valeur prdominante demeure celle d'"expression analytique" ( peu prs ce que RPr. appelle vigraha), oppose vrtti "expression synthtique". Mais les M. Su. n'inflchissent en somme qu'assez lgrement leterme lorsqu'ils emploient vkyabheda "dissociation d'une phrase en deux " (expression frquente des comm. grammaticaux) I 2 25 111 47 1111 19,21 par opposition ekavkya(t) III 1 20, 4 9 (et mme VeS. Ill 4 24), vkyrtha I 2 40, vkyasampti I11 48, gunavkya " phrase secondaire, subordonne" II 3 2, (vkynvaya "connexion de phrases" VeS. I 419) et surtout vkyasesa I 2 22, 29, 3 13, 4 24 (VeS. I 412) qui dsigne un "complment un vidhi", formant avec ce vidhi une seule unit de sens et d'emploi rituel, et qui repose sur la doctrine grammaticale de I'"ellipse" que rend le mme terme dans Nir. XII 22 par ex. ( = P. vkydhyhra) : ainsi les stradu rituel sont dits, par contraste avec le vidhi ( Brhmana), avkyasesa, en ce sens que leur teneur ne comporte

53

452 Louis Renou

sous la forme te vibhaktyantah padam; ou, dit plus prcisment Kt. 11 20, l'ensemble de la base et du suffixe accompagnant la perception d'un sens. L'intervention du "sens" dans la dfinition remonte VPr. Ill 1, qui traduit pada par artha "(ce qui comporte un) sens", tout en maintenant l'autre dfinition, purement phontique, aksarasamudya "agrgat de syllabes" VIII 50 (et: aksaram v " o u bien encore syllabe unique" ou, comme dira Sabarasv. 11 5 p. 48 aksarny evo padam).5* D'o drive cet emploi ? L'tymologie indienne par padyate gamyate rtho 'nena), que donne par exemple Uvata ad VPr. Ill 1 et qu'illustre AA. Il 2 2, nemne videmment rien. Il faut situer le terme dans les conditions vdiques de son emploi. Le sens de " m o t " est relativement tardif. Il est vrai que Geldner, dans son souci de moderniser le Veda, a soutenu que l'acception tait dj aucun lment (prescriptif) suppler. La formule samdigdhesu vkyasest M. I 4 29 est une vritable paribhs "dans les cas douteux, (on dcide) sur la base du complment d'nonc". Autres ex. de vkya M. Il 2 23, 26r 3 7, 41 7 III 3 14 VaisS. V11 1 NyS. II 1 59. La dfinition samhatyrtham abhidadhti padni Sabarasv. Ill 3 14 est emprunte la grammaire, et l'explication du vkya comme arthaikatvd ekam vkyam sknksam ced vibhge syt M. Il 1 46 " la phrase est une unit en raison de l'unicit d'objet, si, lorsqu'on l'analyse, (les parties) requirent (chacune un supplment)". Le vkya est l'un des six promana de l'interprtation verbale M. Ill 3 14. Le terme n'est pas ancien dans l'usage : rare encore chex P., il ne figure, outre Nir., que dans des textes vdiques tardifs comme Brhaddev., APr. (recension Vishva Bandhu), Bhsikas., Sarvnukram. Pour le rituel, BR. ne cite aucun exemple : cependant le vkyabheda est mentionn KSS.XXVI 213, qui en un autre passage note que l'nonc des yajus forme une " phrase" lorsqu'une fois dit, il ne requiert aucun mot de plus (vkyam nirknksam I 3 2 : aspect complmentaire de M. Il 1 46 prcit). On a vkyasesa HirSS. XXV 3 12, sampti XXVI 4 4. Mais simplement "nonc" de mantra pSS. VII113 7 XXIV 3 51. A partir des DhSu., le sens classique de " paroles, ordre verbal, etc." est acquis. Plus prcisment encore, pada chez P. vise le traitement de fin de mot; le terme corrlatif est alors bha, terme fictif, peut-tre abrg de bhakta
54

"faisant partie de", qui dsigne le traitement interne. Dans l'usage que font les grammairiens du terme pada, la notion de finale est prpondrante (cf. padya APr. au sens de padntya " relatif la fin de mot"). Le membre de compos est senti comme un mot parce que sa finale est traite comme fin de mot, d'o les expressions prva= et uttara-pada, que connaissent aussi les Pr. (RPr. forme mme le driv prvapadya " relatif au membre antrieur") et, partiellement, Nir.; la tentative d'exprimer le membre par parvan "articulation " APr. (notamment dans jaratparvan) VPr. Nir. (eka= et aneka-parvan) dans la langue plus image de ces textes, n'a pas eu de succs. D'autres composs, purement grammaticaux, sont tmane= et 'parasmaipada "mot pour soi " (moyen) et " mot pour autrui " (actif), d'o pada "voix" dans padavyavasth, remplaant le terme dsuet upagraha. Des formes analogues mais qui d'aprsPrad.adM.VI3 8vt.1 appartiennent la langue non technique sont tmane= et parasmaibhsa que M. cite titre d'exemples et qu'emploie APr. dans la recension Vishva Bandhu; cf. aussi adhytmam " par rfrence soi ", i. e. " la voix moyenne" Rktan. 196. Prtipadika "thme nominal ", proprement "ce qui est (pareil soi-mme) en chaque mot" est repris APr. ainsi que GB. I 24 et 26. Pancapad " les cinq formes (fortes du nom flchi) " APr. repose sur l'emploi prcit de pada/ bha. Upapada " mot annexe" ou "adjoint", repris VPr., figure en un passage au moins du rituel, HirSS. XXV 4. 2.

453 Le Rituei et la grammaire

tablie dans le RV. I 72 6, la faveur d'un double sens (cf. la note ad loc. ; dans le Glossar il ajoute ce passage RV. VII 87 4 AV. X 8 6). C'est f o r t peu probable: il n'y a pas lieu de dissocier ces quelques phrases des nombreux passages o le t e r m e dsigne en contexte plus ou moins mystique le " s j o u r " de la divinit (Bergaigne Rel. Vd. Il p. 76 et cf. ci-aprs). Le sens de " m o t " est probablement connu d'un Br. tardif, KB. X X V I 5, o pada se juxtapose rc, ardharca, pda, varna. Cette enumeration a subi une rfection : la forme ancienne en est prserve SB. X 2 6 13 o le t e r m e pda fait dfaut et o aksara figure au lieu de varna: il n'est nullement certain, par suite, quepada y signifie autre chose que " v e r s " . Dans le mme Br. XI 5 6 19 il est recommand, si on ne peut lire un t e x t e entier, de lire au moins ekam devapadam, c'est--dire sans doute moins " un mot divin " qu'un " passage de t e x t e sacr" (devatvcy ekam vaidikapadam Say.). Enfin les insertions l i t u r giques sont qualifies PB. XI113 22 d'ekapada et tryaksara: ici encore le sens de " m o t " donn par BR. ("one-footed " Caland) est improbable. D'aprs Liebich un emploi intermdiaire a achemin pada " s j o u r " kpada " m o t " : celui de " q u a r t de s t r o p h e " , soit " v e r s " au sens restreint du t e r m e , " p i e d " selon la conception indienne. Le vers est le " pied " ou le " pas" de Sarasvat comme telle rgion du ciel ou de l a t e r r e est le " pas" d'Agni et de Soma. En cet emploi, pada a t progressivement remplac par son doublet pda qui, signifiant " pied " au propre, repose sur la comparaison de la strophe typique quatre vers avec l'animal quadrupde (la vache), par l'entremise de " q u a r t i e r , quart". 5 5 Au contraire, dans pada, l'acception " q u a r t de s t r o p h e " est ancienne: on la t r o u v e dj dans l'hymne nigmes 1164 du RV., applique au " pied " du mtre jagat(J), au v. 23 ; dans le mme hymne, au v. 45, il est dit que le langage consiste en quatre pada ou " q u a r t i e r s " , dont l'un est le langage des hommes: cette valeur archaque est mconnue par M. qui dans la Paspas entend par ces quatre pada les quatre parties classiques du discours (nom, verbe, prverbe, particule). AV. IX 1019 il est question du pada de la strophe ( " s t e p " , t r a d u i t encore W h i t n e y ) ; d'autres mentions sont relever dans VS. ainsi que dans plusieurs Br. et Stra, mais l'emploi est dans l'ensemble en disparition devant pda. C'est certainement par archasme que la Brhaddev. le donne, ainsi que le N i r u k t a o l'expression pdaprana I 4 sqq. et 9, qui rpond au pdaprana de P., s'oppose vkyaprana.56

En fait, pada comme pda sont les formes thmatiques tires du nom racine pad/ pad qui ds l'origine de la tradition dsignait des "quarts de strophes" lorsqu'il figurait au dernier membre de composs numriques: ainsi astapadpithte de vc RV. VIII 76 12, dvipad et catuspad pithtes de vka RV. 1164 24 (et dans le mme hymne, v. 41 ekapad, dvi etc.) ; par extension apad SB. XIV
Pda au sens de "quart de strophe, vers" n'apparat pas avant KB. prcit, o pada et pda sont en contigut, et AB. IV 4. Ce sens est courant dans les Pr. et P. en porte la trace dans l'expression pdaprana qu'il a en commun (quoique avec une valeur grammaticalement un peu diffrente) avec VPr. RPr. (et cf. Brhaddev. Il 90). Figurment pda = mla AsvSS. 1117.
55

En outre, en fin de composs dsignant une structure strophique, ekapad, dvi, tri et dans padapakti, nom d'un mtre, RPr. VS. SSS., qui est analyser en pacapada paktih RPr. SB. Le driv padya signifie "qui consiste en pac/o= vers" KB. ou "envisage du point de vue des pada" RPr. PB.

56

454 Louis Renou

8 1 5 1 0 = B A U . V 1 4 7 ( ct 'ekapad, etc.) et padanusaga SB. VIII 6 2 3 "appendice d'un vers'1. Le contact vident entre pod et podo (doubl par podo) n'exclut pas-que le sens de " m o t " ait pu se constituer directement sur pada " s j o u r " du Veda. La notion de " s j o u r " dans les montra confine souvent celle d'une dsignation, d'un nom, le nom (nman) inversement est un autre aspect du " s j o u r " ou plutt nom et sjour superposs caractrisent la personnalit divine (Geldner Z D M G . L X X I p. 317) : les expressions padam goh, padam veh, druhas padam, etc. sont des priphrases qui circonscrivent le " nom " mme de la vache, de l'oiseau, du dol ; au v. VII 87 4 les trih sapta nma " les trois fois sept n o m s " de la vache sont t r o i t e ment relis auxpadasya guhy, aux "arcanes du s j o u r " , comme mahat padam A V . X8 6" le grand s j o u r " glose en quelque sorte jaran nma " le nom a n t i q u e " ("substance a n t i q u e " Henry) qui prcde. D'autre part, dans les mantra rgvdiques, pada et aksara sont plusieurs fois rapprochs (III 31 6, 55 1 X 13 3): ce rapprochement est significatif, un stade de la langue o ces deux termes s'acheminent vers des notations linguistiques en partant l'un et l'autre de valeurs abstraites, " s j o u r s " et "lment imprissable". Le mot dhtu " racine verbale" 5 7 est une cration de la grammaire; les deux APr., R P r , et VPr., ainsi que le R k t a n t r a e t le N i r u k t a emploient incidemment le mot, mais il est visible que c'est l un simple rappel de la thorie grammaticale. Dhtu parat d'abord n'avoir aucun tenant dans la l i t t r a t u r e vdique. Les plus anciens emplois du mot sont des composs numriques tels que tridhtu (et trivistidhtu), saptadhtu (et aussi, sans doute par analogie, sudhtu) qui ne sont autres que des adjectifs ou des adverbes de rpartition " t r i p l e ( m e n t ) , septuple(ment), multiple( m e n t ) " . La valeur rpartitive n'est pas moins nette dans le driv adverbial dhtusas du SB. Knva (Caland I p. 54) et de BauSS. X X 11. La situation est donc analogue celle de pada (pda) qui au sens de " v e r s " apparat postrieurement aux composs numriques ekapad, etc. ; on retrouvera des faits parallles propos de guna. Sans doute il existe un dhtu sous l a f o r m e simple RV. V 44 3, exemple unique pour ce t e x t e (et douteux d'ailleurs, cf. Oldenberg N o t e n ad l o c ) ; il ne nous semble pas exclu que laphrasesoc ca dhtu ca ne contienne latmse d'un ancien *saddhtu quivalent de sudhtu. En t o u t cas les autres emplois du mot en simple qu'on t r o u v e dans le rituel sont clairement dduits de l'emploi corn positionne! : les expressions tridhtu, pancadhtu ApSS. Il 92(analogue I 4 1 0 VSS. 121 21 sq. HirSS. I 8 4 sq.) sont reprises par la formule distributive dhtau dhtau (4) qui en drive, et sur laquelle les commentateurs n'ont aucune explication plausible f o u r n i r . La phrase de C h U . VI 5 1 - 3 tredh vidhyate/tasya yah sthavistho dhtuh " . . . se r p a r t i t

en trois : la portion la plus grossire . . . " souligne que le mot relve de la catgorie distributive sous sa forme la plus abstraite: dhtu est pour ainsi dire la nominalisation d'un adverbe en -dh\ de mme tridhtu JUB. IV 23 6 rappelle caturdh qui prcde et entrane son tour un pluriel dhtavah au sens de " parts ",58 II ne faut donc voir dans le terme grammatical de dhtu autre chose
Pratiquement aussi " v e r b e " , notamment dans le Nir. Ce sens est la base des termes grammaticaux rdhodhtuko et srvodhtuka "qui appartient laforme verbale (rduite
57 58

Mme valeur dans prathamam dhturn dadhti KSS. XVI 3 27; le comm. interprte par mrtpraksepo. Mme origine numrique dans ayujodhtn kurvan Kaus. Il 22 ayugdhtni ynni

de) moiti" ou "entire".

KSS. I 314.

455 Le Rituel et la grammaire

que " p a r t " , " l m e n t d'un t o u t scable". Comme dans le cas de pada, un emploi d'origine abstraite, associ des reprsentations numriques. L'acception de " racine" dans GB. 11 26 sq., passage connu de mystique du langage, drive naturellement de la grammaire. Le t e r m e guna dsigne, comme on sait, un aspect a{r) e o du vocalisme, substitu dans des cas dtermins aux formes de base r i u, qui ne portent pas de nom spcial, et qui sont susceptibles aussi de se " r e n f o r c e r " en o{r) ai au, sous le nom de vrddhi.59 Ce sens, qui est propre aux grammairiens, et qui de l a pass f u r t i v e ment dans RPr. X110: 622 et dans N i r . X 17, s'explique par une spcialisation en partant de "caractristique secondaire", valeur bien connue du t e r m e , qui a diversement volu et que prsentent aussi les textes grammaticaux. Dans l'acte rituel on distingue l'lment fondamental ou prodhna, l'lment secondaire ou guna.60 La notion de guna s'applique des choses trs diverses, depuis un processus rituel " a u x i l i a i r e " jusqu' un simple qualificatif: ainsi il est dit KSS. VI 7 23 qu'en guise de svistakrt pourvu de qualificatif on aura en telle circonstance un svistakrt sans qualificatif, sagunasthne 'gunah. Le t e r m e s'oppose aussi dravya "substance sacrif i c i e l l e " (gteau, etc., c o m m . ad KSS. I 513) et il est dfini (comm., passim) upakraka ou sdhanabhta ; il prcise le t e r m e vikrta ApSS. X X I 241 ou W k r o X I V I 1. Le nombre est^un guna KSS. X X I I 8 1 4 . Ces nuances de " q u a l i f i a n t " (Yjnik. ad KSS. I 4 1 7 g l o s e guna par krsnatvdi " f a i t d'tre noir et autres q u a l i t s " ) , cette opposition avec dravya se r e t r o u v e n t en grammaire o guna tend dsigner l'adjectif, dravya le substantif; l'expression gunavacana figure dans les deux domaines. La " q u a l i t " est conue de part et d'autre comme un a t t r i b u t " s e c o n d a i r e " , " a u x i l i a i r e " partir de la substance. Les mmes caractres s'appliquent au mot guna dans les M. S. qui confrontent eux aussi le t e r m e avec dravya, ainsi 1118 o il est dfini y air dravyam cikrsyate; avec pradhna II 1 6; avec

mukhya III 3 9. En outre, paralllement aux paribhs grammaticales, il dsigne l'emploi secondaire, fgur d'un mot I 247 (cf.

Le mot vrddhi s'explique de luimme. Outre sa valeur littrale il a peut-tre une nuance auspicieuse que met en vidence l'emploi du terme au dbut de l'Astdhyyl et aux fins 'adhyaya dans VPr. Hors des grammairiens, on le trouve dans VPr. V 29 et (vrddha) APr. IV 55 Pusp. 1 1 1 1; vrddha au sens de pluta Rktantra 44 et 68 SamhUB. p. 29 LSS. VII 8 5 Pacavidh. I 40. Vrddhi oppos nirhrsa HirSS. XXV 1 10, hrsa avec le sens d"'allongement" (de voyelle) NyS. II 2 55. Augmentation du volume sonore d'un mot prononc par plusieurs personnes lafois M. 1111 et 1 7. Samprasrana, qui dsigne soit une voyelle issue de semi-voyelle, soit le processus de vocalisation, est

59

I1" extension " : c'est le sens du groupe samprasr- en vdique, et notamment dans l'emploi plus technique de l'Anupadas. (cit BR. s. v.). Autre explication Edgerton JAOS. LXI p. 222. 60 Un autre nom du guna est aga, qui s'oppose aussi pradhna et dsigne les portions du rite, comme les prayja et les anuyja, qui se rptent sans changement d'un sacrifice l'autre et constituent le tantra. C'est sans doute cet aspect d'" invariant" qui a dtermin l'emploi du terme en grammaire pour dsigner chez P. la portion du mot "aprs laquelle un suffixe est prescrit, en prsence dudit suffixe", soit la "base" nominale ou verbale.

456 Louis Renou

Sabarasv. yat sambandhini stotavye sambandhyantaram stuyate I 210) ; analogue VeS. 1 6 II 3 3,4 2 N y S . IV 1 56.61 On sait d'autre part que dans la philosophie Smkhya (Yoga) le mot guna note l'un des trois " a s p e c t s " ou " t a t s " de laprakrti. Bien que l'usage associe parfois cette notion avec la valeur commune du mot comme " q u a l i t " (cf. W o o d s Yogasu. p. 148 n.), il semble qu'il faille partir d'une donne plus abstraite, de caractre numrique et rpartitif, comme celle qu'on a pose la base de dhtu et de podo. En effet dans les textes vdiques les expressions dviguna, triguna s _ e prsentent frquemment au sens de " d o u b l e , t r i p l e " (triguna ApSS. VI11 1 2 quivaut trivrt du passage parallle SB. Ill 7 1 20). A ct figure quelquefois guna l'tat simple, au sens de " f i l , lment d'un cordage t r e s s " , en sorte qu'une expression comme triguna aurait pour sens original " ( c o r d e ) trois fils " . Il / a t o u t e apparence cependant pour que ce mot guna soit dduit du compos multiplicatif, comme l'a pens par exemple O l t r a m a r e T h o s . brahm. p. 239 n. 2: le mot a trs peu de vitalit, il figure toujours ct du compos, ainsi uttaram uttaram gunam uttamam karoti BhSS. VII 9 3 avoisine dviguna, triguna 8 19 et 9 2. Mme dans TS. VII 2 4 2 o guna apparat seul, la formule gune gunam donne l'impression de rsoudre un ancien multiplicatif dviguna ou un itratif gunam-gunam. Bref le t e r m e en cet emploi concret n'a pas plus d'authenticit que n'a dhtu dans tridhtu, que

n'aurait vrt ou varttu dans trivrt et trivarttu.

Reste le passage galement numrique AV. X 8 43 o l'on a voulu voir une allusion anticipe aux trois guna du Smkhya (guna pr-smkhya, dit Senart J. as. 1915 2 p. 158); il est parl l du pundarka, figurant sans doute le corps humain, avec ses neuf portes (navadvram), ferm par (les) trois guna (tribhir gunebhir vrtam). Le vers est entendre en liaison avec un autre passage (X 2 29) o il est question de la citadelle (du brahman) ferme par l'amrta et, un peu plus loin (32), du rceptacle (kosa) trois rais, trois tais. Rien de t o u t celane rappelle les guna de la philosophie classique: nous sommes en pleine fantasmagorie numrique de type essentiellement vdique, et l'expression tribhir gunebhih ne fait que rsoudre un plus ancien trigunam " t r i p l e m e n t " . Le t e r m e pradhna, qui est antrieur Pnini (selon qui I 2 56 la notion de pradhna est asisya, n'a pas tre enseigne), dsigne en grammaire, par opposition upasarjana62 la relation " principale" en regard de la relation " s u b o r d o n n e " ; dans un compos en particulier, le " d t e r m i n " en regard du " d t e r m i n a n t " , purusa- en regard de rjan- dans rjapurusa- (ou de rjah dans En tant que caractristique secondaire, guna dsigne l'accent NyS. Il a 55, la proprit d'un phonme (sonorite, aperture, etc.) RPr. XIII 21 : 7 29, ge lieu et le mode articulatoires, etc.) Uvataad RPr. XII119 : 727, (l'tat de somdhi et de poda) APr. 11. Dans un passage du Nir. 112 et 14 II 1 il est question des mots dans lesquels " la formation et l'accent se signalent par un guna propre la base (dont ils sont issus) ", c'est--dire par une modification radicale (la variante de guna auxdits passages est significativement vikra); Strauss ZDMG. VI (1927)
61

p. 115 rend "par une proprit qui rfre au verbe de base". 62 Qui semble tre rest peu prs confin dans les disciplines grammaticales. Le mot dsigne le "versement" de l'oblation KSS. XII 5 9 et cf. l'expression upasarjani (dpah), l'eau qu'on a verse sur la farine, l'eau de ptrin. Il est notable cependant qu'nart. ad SSS. 117 5 connaisse le terme comme synonyme 'upasarga et de guna : on le retrouvera dans les comm. de la ML, ainsi que dans l'padev.

457 Le Rituel et la grammaire

l'expression anal/tique rajah purusah). Dans le N i r u k t a et la Brhaddev., le t e r m e caractrise une strophe en tant qu'elle s'adresse la divinit t i t r e " p r i m a i r e " , c'est--dire de manire " d i r e c t e " : c'est un quivalent de pratyaksa ( N i r . VI11 sqq.) et le t e r m e oppos est ici nipta ou naighantuka (aussi paroksa). Dans le rituel, d'o l'emploi semble maner, le mot dsigne par opposition aga (y. ci-dessus p. 450) les portions essentielles du rite, celles qui lui donnent son individualit et qui varient de sacrifice en sacrifice: sorte de " pr-tabli " rituel (telle est la valeur tymologique) 6 3 . Prokrt/ signifie " b a s e " : le mot est glos mua, yoni (le comm. de HirSS. X X V 1 10 le donne pour un quivalent de nimitta ou de krana) et dsigne les rites qui une fois dcrits ne seront plus rpts quand on t r a i t e r a d'autres crmonies. C'est par rapport la prakrti qu'est instruite \ vikrti ou " v a r i t " , i' " e c t y p e " si l'on prfre par r a p p o r t a I' " a r c h - t y p e " . Le M. Nyyaprak. 107sq. dfinit prakrti " (forme rituelle) d'o la vikrti t i r e ses lments auxiliaires (aga)" et aussi " a c t e dans lequel les aga ne sont pas atteints par une injonction de t r a n s f e r t " , autrement dit, dans lequel ils sont directement prescrits. La mme valeur de " base" existe chez les grammairiens et les phonticiens : donc " radical " oppos kpratyaya, ou bien " t a t primitif, originel " du mot, oppos son " t a t m o d i f i " que note le t e r m e vikrti et plus souvent vikra (le vikra est signal dans la teneur des stra par une dsinence d'accusatif, enseignent VPr. 1133TPr. I 28). C'est sur la base d'une disquisition grammaticale que les M l . 11 10 emploient prakrti au sens de " f o r m e o r i g i n e l l e " = sthnin, et vikrti de " f o r m e s u b s t i t u e " = desa. L'instrumental adverbial prakrty est frquent aussi bien en grammaire qu'en rituel (et cf. Ml. Ill 3 34, 6 2), avec le sens de "dans l'tat p r i m i t i f " , pratiquement "sans modification " : le sens premier est ncessairement " c o n f o r m m e n t l'archtype" : l'usage du t e r m e en grammaire se dcle donc comme un e m p r u n t la langue rituelle. 6 4 O n a vu que pratyaya est le t e r m e oppos prakrti. Lesens de " s u f f i x e " s'est instaur en grammaire de manire stable et constante, 65 mais hors l il n'apparat que dans des commentaires tardifs ou bien dans un passage de GB. (I 24, 26) dont les attaches avec lagrammaire sont videntes. Dans le N i r u k t a 115 il est d i t qu'on ne peut sans le secours de l'analyse tymologique reconnatre le sens (arthapratyaya) d'une strophe vdique; pratyayrthe RPr.
Dans les NySu. IV1 S6pradhna dsigne par opposition guna le sens " propre" du mot, qui en grammaire est not plus souvent par mukhyo. La terminologie du Smkhya, dont nous avons vu ci-dessus l'application peu correcte qu'elle fait du mot guna la dsignation de substances (alors que partout ailleurs guna et drovyo s'opposent rigoureusement), utilise le terme pradhna, de manire peu adquate, comme un quivalent de prakrti "matire (originelle)". 64 Vikarana dsigne aussi l'origine une " modification " : on retrouve ce sens, appliqu un phonme, dans une citation faite chez Uvata ad VPr. Ill 135; appliqu au sens d'un mot,
63

dans Nir. I 3; appliqu l'accentuation du verbe, dans Bhsikas. Il 1. Dans lagrammaire proprement dite, partir de M., le mot se spcialise pour dsigner les affixes du prsent et de l'aoriste, proprement les "modificateurs". 65 II apparat aussi, par emprunt la grammaire, dans l'APr. (les deux recensions) et peut-tre dans VPr. V 13 (vibhaktipratyay) ; dans TPr. V 7 en tout cas l'acception serait plus large qu'en grammaire puisque le mot englobe l'augment verbal. "Suffixes" se dit antakarana ou nmakarana dans le Nir., aussi ibid., upabandha, qui dans le rituel signifie "connexion"; peuttre aga Rktantra 127.

458 Louis Renou

VI 34: 410semble signifier " q u a n t lafonction " . Dans les Sutra rituels o le t e r m e apparat quelquefois, une acception non technique, " f a i t ou moyen de reconnatre, de raliser (une n o t i o n ) " est gnralement dcelable: ainsi dans lokapratyayt (et pratyayt seul X X I I 3 44) " d u fait que (la chose) est reconnue dans l'usage p r o f a n e " KSS. XII11 9; mnyopratyaya " c e qui a la t r a d i t i o n pour (domaine) reconnu " Kaus. I 2; "signe d i s t i n c t i f " BDhS. I 2012. Les commentairesglosent le plus souvent par pratyate, c'est--dire jnyate; celui de SSS. XI113 6glose par proti " c o n c e r n a n t " , ce qui fournirait pour pratyoya un point de dpart abstrait qui irait de pair avec celui de plusieurs autres termes grammaticaux. De mme l e c o m m . de BSS. X X I V 1 : 1 8 5 1 et 12 glose promana " c r i t r e " . 6 6 Est-ce de laque peut s o r t i r l'acception de " s u f f i x e " ? Sans doute les grammairiens exposent (M. Ill 1 1 vt. 3) que le suffixe est pradhna, l'lment essentiel dans le mot, parce qu'il n'a pas fait l'objet d'un "enseignement a n t r i e u r " , et le M. Nyyaprak., dveloppant cette ide, dit que laracineou l e t h m e e s t s u b o r d o n n au pratyaya. Plus gnralement pratyoyo est glos dans la grammaire " c e qui est enseign, ce qu'on a fait comprendre, fait se raliser" (pratyyyamna), ainsi M. et Prad. 11 69 init. : " o n appelle pratyaya, dit le Prad. ad III 1 1, ce dont le sens est ralis par le radical " . Mais t o u t ceci semble avoir t construit sur la valeur non technique du t e r m e , celle d o n t j e rituel a conserv la trace et que r e p r e n d r o n t les Mmmsistes (Sabarasv. 1 1 5 p . 44pratyyakah sabdah " le mot est ce qui fait comprendre, ce qui communique (un sens)" ; dans le mme contexte, pratyaya quivaut "signification " ) . Il semble que, pour rendre compte de l'acception de "suff i x e " , il faille p l u t t partir du sens de " q u i s'agrge , qui s u i t " , sens qui est attest pour le groupe verbal prati-i- dans le RV., et duquel serait sorti son t o u r le sens d' " a t t e i n d r e , de raliser". O r ce sens de " q u i s u i t " est prcisment attest dans plusieurs Prtiskhya (largement dans RPr., ainsi que VPr. et Puspas.) o le locatif pratyaye quivaut apare " q u a n d (tel mot ou tel lment d'un autre mot) s u i t " . Il existe dans les mmes textes un emploi parallle pour un autre driv de la racine /-, savoir udaya, et cet emploi est mme accidentellement pass chez Pnini. Dans l'usage qu'elle fait du mot pratyaya, la l i t t r a t u r e des Prtiskhya nous parat donc plus archaque que Pnini, chez qui cependant le su. N11 Iparas ca marque la fois que la dissociation pratyaya/ para tait effectue, et que persistait le souvenir d'une identit. L'origine est plus apparente dans vibhakti " d s i n e n c e " , dont Liebich a reconnu exactement l'emploi premier. Hors de lagrammaire le mot est attest au sens de " d s i n e n c e " dans les deux recensions d'APr., dans VPr., Rktan., N i r . , Brhadd. (et plus tard NyS. Il 2 38 et 56 M. I 3 29); on retrouve aussi le sens grammatical dans GB. I 24 et 26 sq. et dans NidS. Ill 9 IV13. Vibhakti est connu dans t o u t e la prose vdique, soit au sens gnral de " d i s t r i b u t i o n " ( " r p a r t i t i o n " du gteau entre Agni et Visnu AB. 11 6, " d i v i s i o n " de la victime animale VI11 6, " d i s t i n c t i o n " entre les trois pressurages VI 5 3 ; analogues TS. TB.) ; soit par spcial isation smantique, pour dsigner les modifications, la " f r a g m e n t a t i o n " que subit un mot tel qu'ogn/- dans divers mantra o il figure, identiques par ailleurs, au cours des rcitations deyjy. C'est la fois la division casuelle d'un nom et la discrimination qui grce cette division s'opre entre les mantra qui p o r t e n t ces diverses formes. On dira
66 Autres exemples du rituel : KSS. XXII 3 46 XXV1 3 BSS. XXII 2:118

10XXII112 :169 7 L$S. X 4 4 BDhS. I 11 24Pitrme. Calandp.40 8.

459 Le Rituel et la grammaire

donc " p r o n o n c e r une vibhakti" (yibhaktim uktva) ApSS. V 28 8 TS. 15 2 3, " f a i r e une vibhakti'' TS. ibid. ou encore " placer (dodhti) des vibhakti" BSS. Ill 2: 7 0 4 , 3: 7017, c'est--dire rciter les mantra en agne, agnim etc. Telle istak est appele vibhakti BSS. (passim) du fait qu'elle s'accompagne d'un mantra modifications (vibhaktimantra). Qu'en cet emploi le m o t ne fasse que spcialiser le sens de " v a r i t " ou " v a r i a t i o n " , on le voit par PB. qui signale une vibhakti du t o n dans un sman X 9 1 , une vibhakti dans la modulation de la finale 1 0 1 , dans Vida 11 1, et qui donne paralllement une vibhakti ou nom d'Agni'71 et 5 ou d'Indra 8 1. Ibid. IV 8 7 le dixime jour du rite, qui reprend des motifs antrieurs, est d i t "consister en stoma, en mtres, en vibhakti dj utiliss" (ptastoma, pta-

cchandas, ptavibhaktika) : il s'agit moins des dsinences que des formes en lesquelles un nom se divise: mais il n'est pas douteux que la valeur grammaticale du terme ne soit en quelque manire devance dans cette affectation rituelle. 67 Un autre terme o l'influence religieuse est sensible est upasarga " prverbe", terme connu de la grammaire ainsi que des Prtiskhya et du Nirukta. On le trouve aussi GB. I 24 et 27 et dans la section des GS. relative au nom donner l'enfant (p. XV 9 Hir. Il 410 Bau. Il 1 27). Le sens propre est "chose verse en sus, accessoire" (cf. ci-dessus upasarjana): cette valeur est directement conserve dans le rituel, o une divinit est upasrsta SSS. 117 5 lorsqu'elle est saguna (v.r.), c'est--dire accompagne d'pithtes, par exemple Agni dans agnim tapasvantam: on reconnat ici dj un domaine touch par lagrammaire. Analogue SSS. XVI111 12 AsvSS. VI 3 15 o upasrjet est glos misrayet. Upasargin LSS. IV 8 21 dsigne un jour qu'on "ajoute" aux trente jours du mois. D'autre part, dans la liturgie, upasarga signifie " interpolation": PB. XII 13 22 vise l'insertion de groupes " unpedes" et trisyllabiques comme pra vaha, mitro na (dans SV. I 302-304). Chez les Aitareyin (AB. IV 4 1) ce sont les sakvarqui servent 1 upasarga, c'est--dire des portions de strophes mahnmm.6* C'est aussi vers le sens de " mention incidente" qu'est orient l'origine le mot nipta, qui dans latradition grammaticale dsigne les particules, ainsi appeles d'aprs Nir. I 4en ce qu'elles " t o m bent" en des sens divers, ou plutt en ce qu'elles "tombent" au cours d'une phrase comme des lments tout faits, sans processus formatif. Le sens de " particule" est attest dans plusieurs Pr. ainsi que dans GB. I 24et 26. Mais la valeur plus ancienne d ' " incidence" est la base des expressions prvanipta chez P., paranipta chez M. ainsi que de la racine nipat-, du driv niptana.69 D'autre part
67 Autres rfrences TS. 11 5 6, 5 2 2 V 7 1 1 KS. VIII 5 IX I XXII 8 MS. I6 4, 7 3 T B . 11 5 6 , 8 5 , 3 1 1 et 6 T A . I 9 5 JB. Ill 330 KB. I 4 SB. Il 2 3 26 pSS. V 28 6 et 9, 29 5 MSS. I 6510 V 1 26 svSS. I 3 6, 6 3 II 8 5 LSS. IV 5 1 18 BSS. X 23 sq.; en outre M. dans la Paspas, not. p. 310. Mais vaibhakta DSS. I 3 24 se dit d'une finale comme -sya "appartenante la dsinence" du gnitif sing. Le Bhsyaad NyS. Il 2 38 entend vibhakti = vibhga "distinction " . 68 Cf. encore NidS. Il 12 (cit Caiand ad PB. p. 306). Un vers de neuf syllabes

est upasrsta KB. XVI11 ; analogue SA. IM. 69 Qui dsigne en grammaire une forme que l'auteur d'un s. ou d'un vrtt. a " laiss tomber" toute faite dans la teneur, et qui par suite revt d'aprs les paribhs une autorit particulire. Le terme n'apparat qu'une fois dans la littrature technique non proprement grammaticale, savoir RPr. XII 26 : 708 o il sert de glose tymologisante nipta. Niptana au sens pninen se dit pratikantha dans ce texte,

460 Louis Renou

le N i r . (et sa suite la Brhadd.) dsigne par nipota la " mention i n c i d e n t e " d'une divinit dans un hymne, par opposition kpradhna. Enfin svSS. VI 1 4 1 4 note par niptita des expressions comme meda uddhrtam prsvatah sronitah qui sont " t o m b e s " au milieu d'un montra.10 Les phonmes n'ont en principe pas de nom, sauf les phonmes accessoires (les ayogavha des phonticiens) qui, tant de prononciation malaise l'tat isol, avaient besoin d'une dsignation particulire. Seul le phonme r c o m p o r t e un nom spcial, parce que dans le t r a i t e m e n t phontique il posait certains problmes (du point de vue notamment du padaptha) dans ses relations avec le visarga et avec la sifflante. Le t e r m e de repha est connu de M. et glos ra + ipha; l'analyse rappelle celle d e T P r . 119 et de HVPr. I 39 en r + epha, et suppose que les liens rattachment le mot la racine riph- se sont relchs : en effet, dans les Prtiskhya, le verbal riphita (quelquefois aussi rephin) dsigne le visarga du samhitptho " c h a n g e n r, rhotacis". Cet emploi remonte aux SS. o figure riphita SSS. I 2 9sq. (yiriphita AB. V 4 3 au sens probable de " d muni de - r - " ) , riphyate AsvSS. I 5 10, ainsi que repha SSS. I. c. svSS. l 2 1 8 H i r S S . X X I 2 3 4 S a m h U B . p. 17 rephin svSS. 1511 sq. Le Vaidikbharana ad TPr. 119 connat pour riphyate le sens de " t r e d c h i r " , dit d'une toffe, ce qui confirme l'origine onomat o p i q u e d u t e r m e ( " g r a t t e r " ApSS. XII 22 7? Cf. Caland ad l o c ) . La dsignation du phonme a lieu par le mot kra attach a, i, u, ka, ta, etc. : cet emploi, enseign M. Ill 3 108 vt. 3 (o kra est senti comme suffixe), remonte TPr. 116 VPr. I 37, mais il n'est pas connu de Pan in i qui, par souci de brivet, pose le phonme l'tat nu, ainsi uh "aprs un - r - " etc. L'emploi de kra aprs un nom de phonme apparat dans les Brhmana de la seconde priode AB. (V 32 2) KB. (XI 5 X I V 3) JB. (n 182) et se poursuit dans les Stra rituels, les Aranyaka, les Upanisad et d'autres textes. Il sort de la notation des interjections rituelles hinkra, omkra, svhkra, vasatkra, svadhkra, etc. qui sont attestes ds le YV. et l'AV. et demeurent frquentes dans t o u t e la l i t t r a t u r e vdique; les formes verbales correspondantes svhkr- et hikr- sont mme connues depuis le RV,, qui dsigne ou figure le nom de I' " c l a i r " par une forme haskra, soit peut-tre " c e qui fait (ou : fait faire) has!" : on aurait l une premire bauche de la catgorie en kra. Le mme mot kra a servi tayer des particules, cakra, evakra, ventuellement mme une forme verbale aytkra, SB. 17 312. Nous n'avons en somme, dans le type akra, kakra, qu'une spcialisation incidente. Le nom vritable du phonme est varna. Pnini l'emploie f o r t peu, 'encontre des Prtiskhya o il abonde. Il est bien tabli dans la prose religieuse_ partir d'AB.-KB., dans plusieurs stra de type srauta et dans les ranyaka. Bien qu' l'tat simple il dsigne
70 Un terme grammatical de contenu analogue uposargo, mais plus cornprhensif, est celui de gati : on entend par gati tous les lments, par ex. les formes en - ou un adverbe comme sat, qui fournissent avec le verbe kr- un groupe compos. Il n'est gure douteux qu'il ne s'agisse du mot banal gati "moyen d'accs, voie" (cf. la glose gamyate Padamaj. I 4 60 "ce par quoi sens est obtenu " et cf. pratyaya),

encore qu'en tant que nom grammatical il soit du masculin, anomalie note Padamaj. VI 2 49. C'est par emprunt vident la grammaire que l'APr. dit par Vishva Bandhu Sstri 1111 emploie le terme. En phontique, smavdique, gati est retirement d'une syllable de stobha par addition de /, u Puspas. p. 520 Rktantra(Van der Hoogt Vedic Chant p. 20).

461 Le Rituel et la grammaire

le plus souvent les phonmes en gnral (et aussi dans des composs c o m m e sarvavarna TPr., va r nas m m nya N i r . ) , il ne figure gure, aprs un nom particulier de phonme, que pour des voyelles brves : constatation que confirme l'enseignement de TPr. I 20 aux termes duquel varna dsigne les trois premires voyelles. De fait, P. et M. n'ont que les formes avarna, varna, uvarna (aussi rvarna) ; de mme les Pr. ; de mme encore le Ktantra (Burnell A i n d r a School p. 112 Goldstcker Pn. p. 37); il faut descendre basse poque pour r e n c o n t r e r des composs comme kavarna, savarna. Partout la voyelle brve devant varna implique la longue correspondante, autrement dit avarna = akra + kra ( + 3kra). Cette limitation semble bien indiquer que le sens au dpart n'tait autre que "coloration ".71 Un nom analogue, d'origine manifestement religieuse et plus encore mystique, est le nom de la "syllabe", aksara.12 La valeur tymologique du terme est pleinement ressentie par les textes : na ksarati "ce qui ne s'coule pas" est attest Nir. XII112 Vaidikbhar. ad TPr. l 2 M. (krik cite n. prcd.), ainsi que dans plusieurs ouvrages rituels. Le RV. ici encore prlude ce rapprochement verbal lorsqu'il juxtapose ksaraty aksara m dans l'hymne sotrique I16442.73 Le sens de "syllabe" est fix depuis l'AV. et le YV. ; le mot est affect surtout aux choses de la mtrique; on sait le rle considrable que jouent les mtres et les noms des mtres dans la spculation religieuse, et que les mtres se fondent sur un dcompte rigoureusement syllabique. C'est dans ce domaine, comme on l'a rappel, que s'est constitue sans doute la notion du " m o t " . Dans le RV. l'emploi 'aksara fait difficult. Il n'est gure douteux qu'au premier plan le mot ne dsigne en tant qu' " imprissable" la parole sacre. Mais la question est de savoir si cette affectation mystique prsuppose ou non une valeur grammaticale. Bergaigne le croyait (tudes sur le lexique du RV. s. v.); Oldenberg aussi inclinait l'admettre (ZDMG. LXIII p. 293), tout en pensant que le sens de "syllabe" tait vis de manire assez indirecte pour laisser place d'autres connotations ; Neisser (Wrterb. s. v.) se suffit avec " parole du sacrifice", ct de quoi il maintient la suite d'Oldenberg, pour lefm. aksar, le sens de "vache". Geldner
71 Cf.rakta " c o l o r " , ranga "colorat i o n " , pour dsigner dans plusieurs Pr. et Siks les phonmes nasaux et spcialement les voyelles nasalises. 72 Le terme est parfois employ au sens de "phonme" : d'aprs une krik chez M. I p. 3610 aksara est le nom du " phonme" dans un sdtra antrieur, stra que Kaiyata cite sous la forme varn aksarni. Il dsigne aussi la voyelle, en tant qu'lment essentiel et support de la syllabe (la consonne est svarnga TPr. XX11). Cette double concidence s'exprime dans les dfinitions de \'aksara des Pr.: "voyelle avec consonne ou anusvra, ou bien voyelle seule" RPr. XVIII 32 :1033 (analogue VPr. I 99 sqq.) et simplement "voyelle" APr. I 93 Rktan.46; d'autre part "conglomrat

de phonmes ou phonme unique" VPr. Vill 48 sq. Les composs samnksara "voyelle pure" proprement "de type gal " (terme des anciens matres d'aprs le Pradpa et attest de fait dans les Pr.), sarndhy aksara "diphtongue" M. (Prad., mme remarque; proprement "phonme de jonction") Pr., ainsi que SSS. I 24 (qui oppose de terme prakrtyksara " phonme originel ") svSS. I 5 9 MSS.V1 1 11 GB.I 27, gardent la trace de cette valeur. Pour "diphtongue", RPr. a aussi de faon image dviyoni.
73 Sank, ad VeS. I 3 10 reprend la formule na ksarati ; il y ajoute asnute qui mane aussi de M. I p. 36 9. On trouve encore na ksiyate Nir. et M. I. c, JUB. 14311.

462 Louis Renou

Glossar propose " parole ternelle' ' et aussi "lment primitif de la strophe" (au fm. "discours"), mais dans sa traduction il rend par "syllabe" 1164 39 et 41. Il nous semble que le sens de "syllabe" n'est pas encore form, qu'on s'y achemine en partant de "(base) imprissable (du discours)" et par association avec pad " pied " qui, comme on l'a vu, se dirige en mme temps vers "ligne mtrique". Les deux termes sont voisins en plusieurs passages: il est dit. Ill 31 6 que Sarama, lafemme "aux beaux pieds" (lisez en mme temps: "aux beaux vers"), " a d i r i g les aksara" i. e. les paroles: ce doit faire allusion, comme le pense Geldner, la joute oratoire entre Sarama et les Pani; il est dit ailleurs (116441) que la bufflesse (symbolisant apparemment la parole comme tonnerre) a mugi, devenant ekapaddvipod, c'est--dire se fragmentant en les multiples versets de la posie sacre, devenant enfin sahasrksar dans l'espace suprme. Pareillement " la parole faite de mille (lments)" VI115 9 se dirige vers Agni pour l'hommage. Le terme voisin est mme pada III 551 lorsqu'il s'agit de rappeler qu'au temps o apparurent les premires aurores " le grand aksara s'est propag au sjour de la vache", mahadvi jaje aksarampade goh, ce qui, transcrit en moderne, donnerait " la grande syllabe est ne dans le domaine de la parole". Bien loin de prsupposer un aksara "syllabe", ces passages mettent en vidence un emploi cosmogonique qui, favoris par la jonction avec pad(a) et la technique mtrique, s'engagera dans la dsignation de la "syllabe". Toute cette imprgnation mystique a survcu, s'est dveloppe dans les Upanisad et les Aranyaka, o il s'est tabli une sorte d'quivalence entre aksara et brahman-, on sait qu' l'origine le mot mme de brahman servait noter la " parole sacre". Les deux termes ont achev leur jonction, cf. BAU. Ill 8 8 sqq. ChU. I 1 5 et 9. L'expression jlvksara AA. Il 3 8 quivaut jJvtman. Cf. aussi VeS. 1310. Sur les noms de groupes de phonmes; il y a peu dire: vyajana "consonne", qui n'est usit en grammaire qu' partir de M., mais que connaissent les phonticiens (et qui figure comme terme grammatical dans plusieurs ouvrages rituels SSS. 1119, 212X1111 8 svSS. I 5 9 LSS. V11016 HirSS. X X I 2 34), signifie en propre "chose qui suggre, signe visible". 74 C'est le sens qui esta la base de RV. vyajana_"ornement, parure" ( cot 'abhyajana), et plus directement d'svSS. VII11213 "expression qui suggre" (scaka sabda v.r.) ou Nir. VI113, dit de l'pithte en tant qu'elle "suggre" un nom personnel (ainsi V.rtrahan comme vyajana d'Indra). Par ailleurs, le terme dsigne en vdique les_"signes caractristiques" de lafemme (seins, etc.) ou de l'homme ApSS. VIII 61 BhSS. VIII 6 5, les " marques de pubert'^ApDhS. Il 26 12 HirSS. XXVII 5_20/, les symptmes de maladie HirSS. XXVII 3_36, les condiments ApSS. VIII 4 1 1 , les rgles (cf. Iaksana) X 15 13; SA. superpose dans le mme emploi la valeur de "consonne" celle d' " indices corporels". La parole, dit AA. Il 3 6, "se manifeste par les occlusives et les spirantes" sparsosmabhir vyajyamn (yak): c'est tout le contenu e vyajana qu'implique cette formule. 75
Non pas "ce qui est manifest par la voyelle adjacente " Tribhsy. ad TPr. XX11 Vaidikbhar. ad I 6, ni mme "ce qui manifeste" (le sens) Uvataad RPr. I 5, mais "signe" permettant de caractriser un mot dont la "tonalit" estfournie par la voyelle
74 75

(svara "voyelle" et "ton "). Un autre driv de la mme racine est vyakti qui a abouti "genre" chez P.; c'est d'aprs la Ks. I 2 51 une expression des anciens matres pour linga. Le mot est rare chez les gram-imairiens et l'emploi connu de "man

463 Le Rituel et la grammaire

Les voyelles sont dites svara " s o n , rsonance", mot qui a abouti paralllement " t o n " , ton musical, ou ton accentuel. Pnini n'emploie pas svaro au sens de " v o y e l l e " , vu qu'il utilise un pratyhra, mais ce sens est connu de plusieurs SS._ainsi que du SamhUB. Un passage mystique d ' A A . Ill 2 5, repris SA. VIII 8 sq., m o n t r e un usage ambivalent du t e r m e : les svara sont compars aux " t o n s " du luth divin, dont les sparsa ("occlusives", p r o p r e m e n t " c o n t a c t " ) seraient les touches. SamhUB. p. 23 enseigne que les consonnes sont traverses (vypta) par le svara.76 Le " c o m p o s " se dit sarnosa, t e r m e pass plusieurs Pr. (APr. dans les deux recensions, VPr., R P r , N i r . ) . Le sens propre est " c o m b i n a i s o n " , et l'on peut discerner travers le RPr. (ainsi X 1 6 : 605) un emploi un peu plus large du mot englobant toutes les formes divisibles par avagraha, c'est--dire ce qu'ailleurs les Pr. appellent ingya " a m o v i b l e " . Dans le rituel le sens est " unification de deux choses" (ekkarana glose ad HirSS. X X V 1 26 sq.), ainsi festation individuelle, individu" oppos kjti est assez tardif. Plusieurs Siks emploient vyakti au sens d'"hiatus". Liga est aussi un nom du "signe" devenu nom du "genre" (grammatical), ce dernier sens ayant pass dans la recension nouvelle d'APr., dans VPr. IV170, Brhadd. GB. I 24. On est parti de pumliga et stliga Kaus. LX15PitrmeSG. p.29 6, 452, "signe distinctif" da masculin, du fminin. Le igo est dans toute la littrature rituelle (ainsi que dans Nir. Brhadd. Sarvnukram.) la "caractristique" d'un mantra, le Stichwort consistant d'ordinaire en ladivinit qui est dsigne, mais qui ventuellement peut consister aussi en le " genre " de la personne laquelle le mantra s'adresse: ainsi au passage cit de Kaus. le yajus adress l'homme (au donateur) est dit pumliga, celui adress la femme (l'pouse) est stliga. Ce sens est repris et dvelopp diversement dans les stra philosophiques, notamment dans M. Ill 3 3, 3 33 et 42, 4 31 VII11 27 X 5 5; le M. Nyyaprak. 90 le dfinit par smarthya. De mme vacana " (ce) qui exprime " (cf. gunavacana, etc.) en est venu signifier "nombre (grammatical) " (galement d'aprs Ks. I 2 51 un nom des anciens matres) parce que le mot tait
associ eka, dvi et bahu.

Les noms des genres sont fonds sur la rpartition des tre anims en " mles" et "femelles"; le neutre napurnsaka (A. emploie astrlpumn II 3 8) est ce qui n'est " ni mle[, ni femelle] ". Les tentatives par vrsan/

yos AB. S B., qui amoraient des nuances grammaticales, n'ont pas abouti. Le SB. X51 2 parle des briques de l'autel qui portent des noms masculins, fminins ou neutres comme les membres de l'homme, et il ajoute (3) qu'on dit istak (" brique"), non istakah ou istakam, car (les briques) sont appeles d'aprs la forme de langage (vc) et tout ici-bas est vc, qu'il soit fminin (femelle), masculin (mle) ou neutre (ni l'un ni l'autre). Ce passage montre clairement la concidence du genre naturel et du genre grammatical; il indique le point o commence une notion de grammaire, se dgageant avec peine d'une notion de la langue courante. Cf. sur ce problme OertelSBBay.1943n7. 76 Diverses triades de phonmes s'instaurent dans des contextes analogues: aux passages cits d'A. S. (et aussi ChU. Il 22 3), le groupe svara/ sparsaj sman. Ailleurs (A. Il 24) les sman ou "spirantes" sont opposs aux vyajana (peut-tre compris en un sens restrient : occlusives sourdes) et aux ghosa " phonmes sonores " comme le souffle (prna) est oppos au corps et l'me. L'sman est identifi ou prna aussi A. Ill 2 1 . Ceci confirme la valeur premire du mot "souffle (chaud), vapeur". Prna mme afailli atteindre une valeur grammaticale RPr. lntrod.4(etcf. A. Ill 2 6S. VII111) o il est dit que le phonme -s- (une spirante!) s'appelle prna, comme -n- s'appelle bala (selon ChU. Il 22 5 ce sont les voyelles qui sont prononcer avec ' ' force ").

464 Louis Renou

A B . 11 14 o il est dit que les saisons sont cinq par sarnoso de l'hiver et de lasaison froide, ou bien LSS. X 1 2 1 4 qu'il y a sarnoso (du rle de Vadhvaryu, dans la description ultrieure) avec (ce qui a t dit prcdemment) du hotr. Une nuance voisine est " r s u m " : ainsi KB.IV 5 enseignant que tel sacrifice est le somsa de tel autre, d'o l'adverbe samsena " s o m m a i r e m e n t " . N o t e r que la dfinition du compos comme " unification " se retrouve en grammaire : samsasyoikrthotvo M. I 2 42 vt. 1. Sarnosa au sens de samudya NyS. Il 2 38 d'aprs le Bhsya. Les noms des composs sont t a n t t des exemples de la formation, tatpurusa, bahuvrhi, dvigu,71 t a n t t des formes qui suggrent ce dont il s'agit, ainsi avyaybhava ou karmadhraya Dvandva signifie " paire, c o u p l e " : du dvondva " c o u p l e " nat le mithuna " a c c o u p l e m e n t " AB. Ill 5 0 4 e t passim. Il s'agit dans le rituel d'objet associs par paire, mais non ncessairement composs TS. 16 9 4 SB. M 1 2 2 V 3 3 1 4 ; les huit dvandva alludes N i r . VII 4 IX 35 sont les n 29 36 de Nigh. V 3 et comprennent des formes comme ortn ou dev jostr. N o t e r qu'en passant en grammaire le mot est devenu masculin (masculin aussi au sens rituel NidS. VII 7), d'aprs les autres noms de composs, sauf dans VPr. Il 48, 55 V 28 qui conserve le genre neutre. Les cas grammaticaux ne p o r t e n t pas de dsignations spciales: ils sont nots par des indices numriques, pratham, etc. 79 Ce mode de dsignation doit provenir du rituel o une foule de notions (jours, rites, modes musicaux, etc.) taient voques par des o r d i naux. En revanche les fonctions que notent les cas dans leur relation avec le procs verbal, les kraka ( p r o p r e m e n t " c e qui fait s'effect u e r l'action v e r b a l e " ) , p o r t e n t des noms d'aspect f o r t e m e n t individualis, parmi lesquels prdomine un groupe de drivs de la racine kr-*: karman, karano, kartr, adhikarana. Karman " a c t i o n

(en tant qu'objet direct, objet du verbe transitif)" appartient aux milieux rituels o le mot, depuis le RV., dsignait l'acte par excellence, c'est--dire le rite. De mme kartr qui dans les deux do77 Mme type de dsignation dans nadl, nom des thmes en -/"- (--) P., danssraddh Kt. Il 110, agni 111 8. Cf. ci-aprs krt et krtya. 78 Nom obscur : il s'agit d'un compos form de l'apposition de deux noms, donc peut-tre " (fait de deux noms) qui portent l'acte (le procs verbal) " [de manire gale], qui sont en smnydhikaranya. 79 Cf. aussi pacami " impratif" Kt. 1111 18 et 26 Hem. Ill 3 8, saptami "optatif" Kt. Ill 1 20 et 25 Hem. Ill 3 7.
80

Nir. Il 28 VI 22, galement un nom des anciens matres d'aprs le Dhtuprad. Il 71 ; cekrlyito " intensif" (moyen) Kt. Ill 214 et 43, 3 7 : ce nom est pareillement donn comme des anciens matres Prad. IV1 78, mais le Nir. I. c. qui dsigne le verbe coskyate comme un carkarita montre ainsi que cekrlyita a d tre fabriqu secondairement. Le mme texte V11 a aussi cikrsita "dsidratif". Les formes kurvant krta karisyant de I'AB. IV 31 3,29 3 V1 3 n'ont pas t retenues par l'usage comme dsignations temporelles :
les f o r m e s e n bh- e t vrt l ' o n t

C e t t e r a c i n e , q u e est aussi la base

de kraka, fournit chez P. les noms du suffixe primaire (krt) et du verbal d'obligation (krtya), lesquels sont d'ailleurs aussi des exemples de la formation. En outre chez M. krita "causatif" (nom des anciens matres d'aprs le Dhtuprad. X170 et attest en effet Nir. 113 et APr. IV 91); carkarta "intensif" (actif) M. ainsi que

emport. Un mot tel que taddhita "suffixe secondaire" ne constitue pas un exemple de la formation, mais donne l'une des valeurs qui selon V1 5 commandent l'emploi d'un tel suffixe. Cf. aussi P. tadrja, APr. IV 29 tanmn'm. Nir. VI 28 tatprepsu (APr. IV 29 prepsu).

465 Le Rituel et la grammaire

mairies est I' " a g e n t " . Karana, d'aprs l'usage gnral en vdique, devait tre un quivalent plus banal de karman: " a c t e " dans les mantra, " f a i t d'excuter" dans la prose: l'emploi en grammaire au sens de " ( n o t i o n d ') i n s t r u m e n t " provient du fait que de nombreux noms suffixe -ana- comportaient une valeur instrumentale. Adhikarana est plus difficile: en grammaire c'est la notion locative, la " location " , mais le sens propre, rvl par l'expression samndhikarana, est " rfrence, plan " . C'est en gros la valeur dans les textes religieux: dans TA. VII 3 1 adhikarana dfinit la srie de " rfrences" qu'impliquent les termes adhilokam, adhijyotisam, etc. : c'est en somme la nominalisation du prverbe adhi " e n ce qui concerne". 8 1 Apdna " n o t i o n d'ablatif, a b l a t i o n " fait contre-partie sampradna " notion dative, dation " , comme apdtr kpradtr TB. 17 3 1 . Sampradna est la " t r a d i t i o n " , mais l'attache du mot au verbe samprayam- " d o n n e r " est sensible, ainsi SA. IV15. Hetu " c a u s e " est aussi un kraka: c'est l'incitation agir transcrite grammaticalement par le causatif. Le t e r m e est rare en vdique, bien que le livre X du RV. prsente dj l'expression prpositionnelle banale hetoh gnitif " cause d e " . Il est probable que Pnini sent la forme comme un nom d'agent, " c e l u i qui i n c i t e " . Nisth qui dsigne le verbal en -ta- est u n t e r m e de la langue religieuse, au sens d'achvement: il se dit HirSS. X X V I I 3 58 = ApDhS. 1115 2 4 d ' u n e crmonie qui marque la " l i m i t e " ( partir de laquelle commence la capacit suivre le dharma), c'est--dire qui constitue un " a c h v e m e n t " ; de mme HirSS. X X V I I 6 24 ApDhS. Il 29 11 " c o m p l m e n t " , dit des connaissances propres aux femmes et aux sdra, qui ne sont qu'un complment celles des hommes. C'est aussi " a c h v e m e n t " qui convient C h U . VII 20 1, o le mot est rattach la racine verbale nih-sth-.*2 Avasna " p a u s e " est p r o p r e m e n t le " d t e l a g e " , le relais et le relche. La pause est marque, on le sait, par la relaxation des effets du samdhi. Wackernagel A i . Gr. I p. 301 estime avec raison qu'on est parti de la notion de pda et qu avasna dsignait d'abord la fin du pda (et dans lepadaptha la fin du mot, puisque t o u t m o t dans cette rcitation est suivi d'une pause). C'est cette affectation ancienne qu'il faut v o i r sans doute dans le driv avasnya VS. X V I 33, pithte de Rudraen tant qu' "ayant rapport avec le pda" : le t e r m e contigu est slokya "ayant rapport avec l a s t r o p h e " . Lopa " p e r t e " , amuissement d'un phonme (cf. aussi la racine verbale up-, atteste notamment dans les Pr., reprise par M. et s u r t o u t par le Kt. etc. ; le verbal lupta, les termes up uk lu lumant forgs par P.) est attest plus largement dans le rituel pour dsigner quelque lment du rite, tel j o u r sacrificiel, etc., qu'on omet, mais souvent aussi djtel mot, tel groupe de mots, telle syllabe qu'on " s a u t e " dans une rcitation. 8 3 La valeur ancienne
81 Mais= sthna DSS. 1 42 BSS. XXIV 12:196 8; = sdhanadravya BSS. XXIV 2 :186 2 (cf. chez P. o adhikaroa est glos par drovyo Ks. passim et M. IM 1 vt. 21); "fonction" SSS. M 8BhSS.V1413VIII617;"section" d'ouvrage dans les s. philosophiques. 82 Glos gurususrsdi Sank., d'o "attention on a t u t o r " Mller; "Hervorwachsen" Deussen (qui propose aussi : " Wurzelung"), d'o

"growing f o r t h " Hume; "pratique parfaite" Senart. Mieux "Abschluss" Bhtlingk. 83 TS. II! 2 9 5 KB. XXVI 4 pSS. 110 21 II 19 3 VIII 8 5 XXIV414(oppos vivrddhi) BhSS. 11015 DSS. XV 3 15 HirSS. VIII 8 26 KSS. IV322SSS. 11119 2 LSS. Ill 7 7 V11 14. Avec une valeur plus grammaticale SSS. I 210XII11 8 XIV 4018 svSS. I512VSS. 11 1 67 DSS. IV 3 20 XV 4 7 LSS. 13 21 II 5 22 III

466 Louis Renou

est " r o m p r e " : elle figure sous la forme du causatif ropayati, rrupot en des passages de TS. TB. (cits chez BR.) o il est mis en garde contre telle pratique qui aurait pour effet de " r o m p r e " une partie du sacrifice: Sy. glose mohayati, bhrntimjanayati, mais l'ide est celle d'une continuit rompue, et par suite d'une dperdition. Pluti (piuta) dsigne chez les grammairiens et s u r t o u t chez les phonticiens la prolongation de la dure d'une voyelle jusqu' trois mores. Le t e r m e figure avec cette valeur dans plusieurs textes rituels, qui connaissent mme les formes causatives plvanq MSS. V 1 1 11 pDhS. 1517 plvoyati MSS. I.e. HirSS. X X I 2 34 SSS. 121 svSS. I 5 7 dont lagrammaire n'use pas. Le sens originel est " f l o t t e r , nager", d'o " s e m o u v o i r " ( lafaon de quelque chose qui flotte ou nage): le SB. X I 5 513 parle des jus de soma qui " c o u l e n t " entrelacs des jus de soma, somh somairvyatisakth plavante, des atomes lumineux qui " n a g e n t " (dans l'air) 1 X 4 1 8, de l'homme qui " s e m e u t " avec ses membres, restant immobile avec (le gros de) son corps X I I 2 4 8, etc. Il semble qu'il y ait l un emploi imag, lequel permet d'induire que le mot est pass des Prtiskhya lagrammaire et non l'inverse. 84 Cette srie de concordances entre les habitudes linguistiques du rituel et de la t h o r i e grammaticale atteste qu'on a affaire des disciplines issues des mmes milieux, rpondant des besoins complmentaires. L'une et l'autre ont pour norme la pratique des sista, des spcialistes (GobhGS. Ill 3 29, 5 38 d'une part, M. VI 3 109 de l'autre). En prsence de t e l mot particulier, il n'est pas ais de reconnatre s'il sort des cercles de grammairiens ou des cercles de ritualistes: l'absence de t o u t e chronologie textuelle, le paralllisme gnral des techniques dans l'Inde ancienne, rendent cette recherche alatoire. Dans beaucoup de cas cependant il est visible que le point de dpart est dans les textes religieux, la valeur grammaticale apparaissant comme une spcialisation l'intrieur d'une acception rituelle mieux articule. La masse, l'importance de la l i t t r a t u r e religieuse, l'indniable p r i o r i t des mantra et des

417VI 1014sq. VI111 6 KSS. XIX 7 6 dans le mantra a vajam vjinoagman, SamhUB. p. 16 sq. NidS. Il 10. La le prverbe G tombe et l'on ramne racine cyu- au mme sens GB. I 26 n'a sa place (pratyhrtya) la forme ogmon, rien donn. Les N/S. \\ 2 55 emplode manire obtenir un nouveau ient upamardo (ekarpanivrtti). mantra ogman vjam vjinah. 84 Nom- (samnam-) se dit dans le L'abhinidhna, dfini par RPr. VI rituel d'un mantra qu'on " incline", 17 : 393 comme lasgrgation et le c'est--dire qu'on modifie pour l'adapvoilage de consonne devant consonne, ter un besoin nouveau : c'est une c'est--dire une articulation incomdsignation image de Vha. L'emploi plte de la premire consonne par (relevable pSS. XIV1 9 XIX 26 6 opposition samyoga qui dsigne la XXIV 3 25 HirSS. IX 2 4 XX 4 9 Kaus. "jonction " normale (cf. RPr. VI 24 : LX 20 LXII112 svGS. Ill 8 7) a peut400), est dans le rituel " le fait de tre incit les phonticiens utiliser dposer ct ou dessus " (donc, conlaforme nom- (nati nmin vinam) pour tact sans jonction) KSS. V1 31 XXV 3 dsigner I'" inclinaison" d'une dentale 13 et cf. abhinidh-SB. Abhinihita est en crbrale. le nom d'un samdhi o l'a- initial s'est Le terme technique pratyhra a "rapproch" de-e-o final (pour tre son domaine prpar par l'usage que absorb par ceux-ci): "misen contact" fait le rituel du groupe verbal SB. 13 412. pratyhr- : ainsi selon DSS. XV 4 7,

467 Le Rituel et la grammaire

formes liturgiques qu'ils prsupposent, invitent voir de ce ct l'origine. Hors mme des correspondances de vocabulaire, cette l i t t rature rvle par ses proccupations gnrales comment des notions de grammaire ont pu s o r t i r peu peu de valeurs religieuses. Ds un hymne du Rgveda, l'hymne la Parole (yc) X 7 1 , nous voyons la parole magnifie comme art de l'expression : cet hymne qui fait prvoir pour ainsi dire le dveloppement de la grammaire est en mme temps le plus vieux t e x t e de la potique indienne. L'origine du langage est reporte aux temps primitifs o sous l'gide de Brhaspati " le matre de la parole sacre", il fut " imparti des noms aux choses", et qu'ainsi " s e rvlaavec amour ce qu'il y avait en elles de meilleur, de pur, et qui demeurait cach". Ainsi que les Brhmana le raffirmeront souvent, le langage est conu comme la partie extriorise, explicite (nirukta) de l'ensemble mystrieux (paro'ksa) des choses informulables (anirukta). Les vieux Sages ont impos la parole la pense " c o m m e on clarifie les grains par le c r i b l e " : ce qui est voqu ici, c'est donc la cons t i t u t i o n d'une langue raffine, d'un " s a n s k r i t " grce auquel " la beaut s'imprima sur la parole". Cette langue raffine est au service du r i t e : ce n'est pas un hasard si le t e r m e samskrta et la srie qui s'y apparente dsignent les prparations rituelles, les actes matriels (adhisrayana, paryagnikarana, ptrapratapana, etc.) qui o r n e n t et ajustent un rite, et aussi les crmonies de conscration ou " s a c r e m e n t s " qui marquent les tapes de la vie d'un rya. Mais cette parole, poursuit l'hymne X 7 1 , n'est pas donne tous : les uns ne la voient pas ni ne l'ententent: ils cheminent dans ('"illus i o n " (moyo) et leur langage est "sans fleur et sans f r u i t " ; ils ne connaissent pas le chemin de l'acte, entendez: du rite. D'autres au contraire la ralisent: eux la parole " a ouvert son corps comme son mari une femme aux riches a t o u r s " . Le problme de la rpartition de la parole a maintes fois occup les Brhmana. La plus ancienne mention s'en t r o u v e dans l'hymne nigmes 1164 du RV. ; au v. 45 il est dit que le discours " mesure quatre parties", dont " t r o i s sont tenues secrtes, on ne les met point en b r a n l e " ; seule est connue la quatrime: c'est le langage des hommes. Il ne faut pas voir ici, naturellement, d'allusion aux quatre parties du discours, comme croit ou feint de croire la Paspas (p. 3 I. 26) encore que prcisment cette d i s t r i b u t i o n quadripartite a d tre prsente l'esprit de Yska et des auteurs des Prtiskhya quand ils ont rparti les mots en noms et verbes, particules et prverbes. Il ne faut pas davantage en croire le SB. IV1 3 1 6 , selon qui les quartiers cachs sont le langage des btes, quadrupdes, oiseaux, animalcules. Il s'agit bien plutt, pensons nous avec Geldner (RV. ad loc.) et Strauss ( Z D M G . VI 1927 p. 102), de la partie transcendante du langage, ce qu' date ultrieure on dnommera le brahman, dont il est dit, comme de la vc, que l'homme n'est en tat d'en reconnatre qu'une minime partie. Car t o u t e la spculation sur le brahman se fonde en dernire analyse sur une diffrenciation linguistique o le brahman " parole sacre, mot magique" englobe lafois le mot matriel (ce qu'on appelle significativement sabdabrahman, et dont Ytman est une manifestation) et le mot transcendant, le parabrahman. C'est cette distinction qui dans les cercles de philosophes du langage fournira la base de l'opposition dhvani: sphota. Du point de vue vdique, le nom (nman) est l'essence des choses, et le " m o t " (pada) est le lieu conu comme une entit. La ralit est la parole divine

468 Louis Renou

exprime. Monas et vacforment une unit SB. 1 4 4 1 5 . La pense (mati), c'est la parole, car c'est par la parole que pense t o u t tre ici-bas VII11 2 7. C'est le plus puissant des souffles KS. X I X 1 0 . Lorsqu'on respire, ce n'est pas par la respiration, c'est par la parole qu'on d i t : j'ai respir KB. W 7. Le Soi t o u t entier entre dans la parole et est fait de parole, ibid. D'autre part, la parole joue un rle primordial dans la cration : elle est cratrice, elle est le dmiurge. La cration est l'uvre de la parole, c'est par la parole que t o u t a t fait ici-bas SB. VII11 2 9. Les trois mots sacrs bhh bhuvah svah deviennent les trois mondes X11 6 3. On pourrait multiplier ces citations, dont on t r o u v e r a d'autres spcimens encore chez Deussen Gesch. Philos. 11 p. 205 Hopkins Connecticut Acad. X V p. 26 S. Lvi Doctrine du sacrif. p. 22 Oldenberg W e l t a n schauung p. 78 Scharbau Idee der Schpfung p. 123. Il n'est pas surprenant si dans ces conditions lagrammaire et les thories grammaticales ont joui dans la pense indienne d'un prestige unique. La grammaire est considre comme un membre du Veda, un Vednga, 85 voire le plus important et le p r e m i e r : pradhnam satsv agesu vyakaronam M. I p. 1 19, prathamam chandasm agam Vkyapad. 111 ; ce dernier t e x t e ajoute que lagrammaire est " la science suprme, filtre-purifiant de toutes les sciences" (pavitram sarvovidynm adhividyam), " la voie royale exempte de d t o u r s " (ajihm rjapaddhatih). Le Sarvadarsanasamgraha, qui m e t t r a le Pninimata sur le plan d'un darsana, dit la fin de la section affrente que lagrammaire "vise raliser l'objet suprme de l ' h o m m e " (paramapurusrthasdhanatay). Elle est le Veda des Veda, C h U . VI11 2 et 4, 2 1 . Lorsque laPaspas(p. 1 14) veut dfinir les buts fondamentaux de la grammaire, elle ne t r o u v e faire valoir que des notions d ' o r d r e religieux: la raks ou prservation du t e x t e sacr, \'ha ou modification apporte un mantra pour qu'il s'adapte un besoin nouveau, Va gama ou t r a d i t i o n . D'ailleurs t o u t ce qui dans la Paspas est dit de la grammaire se rfre en fait au Veda. Lorsqu'au vrtt. 5 il est enseign que le domaine d'application des mots est constitu par les quatre Veda, ceci est entendre au sens littral. L'adjonction mme que Patajali propose ce sujet souligne la prminence des choses vdiques : car les domaines annexes qu'il mentionne, la vkovkya, Vitihsa, lepurna, levaidyaka sont certainement des aspects particuliers de la l i t t r a t u r e sacre.86 Ceci m o n t r e que, mme l'poque de Patajali, aprs le grand effort vers la constitution d'une bhs que t r a h i t son uvre, lagrammaire tait encore dfinie par rapport aux besoins de l'exgse, comme une sorte 'ancilla theologiae, appele d'ailleurs jouer aussi le rle d'une matresse. La pense indienne a pour substructure des raisonnements d ' o r d r e grammatical. La Mmms, en tant que prolongement et pour ainsi dire jurisprudence du r i t u e l , implique une masse de donnes philologiques qui remontent en fin de compte lagram-

En fait l'Astadhyay), qui est largement scularise, est probablement le reflet d'un trait antrieur, plus proche du chandas. Les noms des prvcrya cits chez Pnini sont pour la plupart des noms de phon-

85

ticiens du Veda, connus par les Pr,, et en mme temps de ritualistes. 86 Vkovkya, qui pourrait faire doute, dsigne d'aprs Kaiyata les nigmes rituelles,

469 Le Rituel et la grammaire

maire 8 7 : d'abord parses discussions sur la prennit du m o t (et le caractre prissable des phonmes), 88 sur la relation entre le m o t et le sens et le problme du sa m band ha ou " c o n n e x i o n " (Strauss SBB. 1932 p. 486), sur le p r o t o t y p e ternel du m o t ou sphota*9 lesquelles concident avec celles du Mahbhsya (p. 483) et reprsentent le fonds le plus ancien de la Mmms (p. 470). Ensuite par t o u t e son organisation interne, par le lien qu'elle tablit entre l'injonction vdique ou vidhi et son aspect linguistique, savoir l'optatif ou / m , par le rle qu'elle a t t r i b u e laforce de ralisation incluse dans la forme verbale de l'injonction, labhvan, contrepartie du krakatva et de la sakti des coles grammaticales (v. ldessus Edgerton Language IV p. 174), II y a l une vritable scolastique grammaticale sous-jacente l'hermneutique ritualiste. Le Vednta semble premire vue bien loign de ces p r o blmes. Mais il suffit de lire attentivement Sakara ou Rmnuja pour constater que leur pense est constamment taye et n o u r r i e de raisonnements grammaticaux. Rappelons la thorie du sesa, lev la hauteur d'un principe thique (et labor d'ailleurs sur une armature de Mmms), la notion de bheda " d i f f r e n c i a t i o n " , de vypti " e x t e n s i o n (d'un phonme, d'une fonction, transitivat i o n ) " ; la notion de guna " a t t r i b u t " ou " q u a l i f i c a t i o n " commande le problme du brahman. Le postulat du primat de l ' u l t r i o r i t mane des paribhs grammaticales. Le principe du smnydhikaranya ou " c o m m u n a u t de rfrence (de deux noms par rapport l'action v e r b a l e ) " (symbolise en grammaire par le karmadharaya) est la base du monisme, o l'axiome tat tvam asi se rsout en une relation de sujet prdicat. Toute glose indienne, dans son dtail, dans son cheminement littral, est ptrie de grammaire, mais c'est aussi lagrammaire qui relie et domine quelquesunes des dmarches les plus hautes de la pense indienne.
Dans sa discussion sur la question si le mot dsigne l'espce ou l'individu, M. I 2 64 vt. 35-59 (traduit Strauss ZDMG. LXXXI p. 137) est un avantcoureur de la M. depuis Jaimini, UpavarsaetSabarasvmin, du Nyya depuis Gotama et Vtsyyana (Strauss I. c. 151). De mme il y a des contacts troits entre tel passage de M. sur le genre grammatical et le genre naturel (M. IV1 3, traduit Strauss Festg. Garbe p. 84) et l'argumentation des logiciens. Sur les analogies entre M, et les Yogas., v. en particulier l'Introd. de Woods p. 15. 88 Ce problme crucial est abord M. 11 6-23 et repris dans les NyS. et les VaisS., cf. notamment Abegg Festschr. Wackernagel p. 255. Le point de dpart en rside dans une rflexion toute primitive sur la nature et lafonction du Veda, dont on retrouverait sans peine la trace jusque dans les Hymnes, En tout cas la discrimination entre le mot krya "opratoire" et ssvotika "ternel" est djformule RPr. XII114 : 722.
87

C'est une date bien postrieure que se sont levs chez les MTmmsistes des critiques de lagrammaire, sentant en celle-ci une adversaire de la rvlation (quelques textes runis ci Ksh. Ch. Chatterjee J. Dep. Letters XXIV [this volume, pages 287-297]). Noter que les NyS. Il 2 55 usent d'une terminologie grammaticale, mais non pninenne: guna au sens d'"accent", upomarda, hrsa, vrddhi au sens d'"allongement", eso, sieso. 89 Littrature : Abegg Fest. Windisch p. 188 Liebich ZDMG. LXXVII p. 212 Strauss p. 131 Jacobi LVI p. 399 Chakravarti Ling. Specul. p. 42,127 et passim Philos. Skt Gr. p. 84 Varma Critical Studies p. 172 Dasgupta Study of Pat. (Appendicel). [Seethis volume, pages 405-414], La doctrine, laquelle sauf le Yoga aucune cole philosophique n'a vraiment adhr, mane des grammairiens; ladiscussion la plus attentive sur ce sujet est celle de Kumrila, dans le chapitre du sphotavda du S lokavrttika.

B. Grammaire et Vdanta (I9S7) Louis Renou

The second comparative study presented here is "Grammaire et Vdanta," Journal asiatique (245,1957,121-133). Kenou takes one of the most well-known philosophical texts in Sanskrit (part of which he translated in Sakara: Prolgomnes au Vdanta, 1951) and studies concepts and patterns of reasoning that may be traced back to the grammarians. Sakara's commentary on the Brahmastra or Vedntastra was probably written in the eighth century A.D. Prenant pour exemple le commentaire de S[ankara] sur les [Brahma]s[tra] par commodit, nous ajouterons aux rfrences par adhyya, pdaet s. les pages de la traduction Thibaut, essayons de voir l'usage que fait S. de notions ou de formes de raisonnement manant ou susceptibles d'maner des coles grammaticales. vrai dire, les citations expresses de textes grammaticaux sont en nombre trs rduit: l'index de Thibaut(-Winternitz) n'en livre que trois: a) P[nini] I. 4,30 sur l'Abl. de cause matrielle (prokrti), appuyant la thse que le b[rohman] est bien la cause matrielle du monde (I. 4,23 p. 285). b) P. VI. 4,158 sur le sens du mot bhman ( ' ' fa i t d'tre beaucoup"), mot que le s. lui-mme (et en tout cas S. ad loc.) I. 3,8 p. 162 interprte comme un des noms du b. (le terme vient de ChU. VII. 23 et 24). c) Ad I. 4,11 p. 260 enfin, l'expression pancajanh de la BAU. (implique dans le s. suivant n 12) est interprte l'aide de P. II. 1, 50, aux termes duquel de pareils composs ayant un nom de nombre au membre antrieur sont des samjn: pancajanh, en l'occurrence, dsignera une classe d'tres et nommment, en vertu du s. 12, " le souffle et autres (essences) analogues". Beaucoup plus souvent, les rgles pninennes demeurent l'tat latent, comme il est d'ailleurs naturel, dans la trame du commentaire. D'ordinaire, dans la portion initiale du dveloppement consacr tel ou tel s., celle qui contient la paraphrase. Comme dans les cas prcdents, l'allusion aide lucider quelque forme, quelque emploi embarrassant du su. ou, plus souvent encore, de l'Up. vise par ce s.1 C'est l'emploi des cas ( quoi se rfre dj I. 4,23 prcit) qui donne lieu au plus grand nombre de remarques. Au cours des efforts faits par S. pour rduire l'unit les propositions upanisadiques et tendre es identifications avec le b.,2 U s'agissait en effet
1 Mentionnons pour mmoire les allusions des donnes qui sont moins de grammaire que de " philosophie grammaticale", encore que Patajali y ait touch le premier. Ainsi sur le sphoto I. 3,28 p. 204 considr comme le mot (peru globalement, par opposition aux phonmes successifs qui le

la connexion des mots est avec l'espce ou le genre, non avec l'individu, krtibhis ca sobdnm sombandho no vyoktibhih : thme bien connu de discussion, depuis Patajali. 2 Ici joue l'occasion le concept du poryyo, de la synonymie. Ainsi 1.1,22 p. 83, aprs qu'il a eu montr qu'ksa composent), tosmt sphoto evo sabdoh signifiait b., S. ajoute que les synop. 206. Ou bien sur le dhvani, ibid. p. nymes d' signifient galement b. En 208, considr comme la "tonalit", revanche, les "symboles" (protlko) d'o dpendent les distinctions 'udotels que le " nom ' ' ne sont pas aptes tta,anudtto et autres, tannibondhons exprimer le b. IV. 1,4 p. 341 ; la cocodttdayo visesh. Ibid. enfin, dans ce ordination grammaticale (smnymme riche contexte de rflexions dhikoranyo) qui lie ventuellement paralinguistiques(p.2O2), il est dit que ces symboles au mot "b." n'est pas

471 Grammaire et Vdnta

de m o n t r e r que t e l cas grammatical quivaut t e l a u t r e , o u d u moins n'entrave pas l'assimilation. Ainsi le Loc. de sjour (dhra, m o t q u i chez P. I. 4,45 glose adhikarana) " d a n s le ciel " quivaut l ' A b i . de l i m i t e (marydy "au-dessus d u c i e l " 1.1,27 p. 96.^De mme le Loc. et N n s t r . III. 2,7 p. 142: dans " il s'unit avec l ' t r e " , un Loc. ("dans l ' t r e " ) est e n t e n d u , en r a i s o n n e passages c o m plmentaires o figure effectivement " d a n s l ' t r e " : ainsi d i t - o n dans l'usage profane " il va vers la m e r par le G a n g e " o u " d a n s le Gange", indiffremment. Mais, au moins au titre de prvapaksa, une diffrence entre Nom. et Loc. est admise I. 2,5 p. 112. Il ne fallait pas que le krakatva grammatical ft un obstacle dirimant aux quivalences notionnelles. Ailleurs (II. 3,10 p. 21), S. a jug utile de prciser la valeur de l'Abl. afin d'expliquer le texte "de l'air naquit le f e u " vyor agnih de la TU., texte qui semble en contradiction avec d'autres affirmant que le feu naquit du b., tejaso brahmajatve. La difficult est rsolue en invoquant la diffrence entre origine immdiate et origine ultime. D'autres passages montrent la ncessit de respecter la distinction casuelle: dans pituh pita ou prnasya prnah, Gn. et Nom. ont chacun leur rle 1.1,23 p. 87 Ailleurs (III. 2,22 p". 169), S. note que le b., dans tel passage de la BAU. (II. 3,1), n'est mentionn qu'au Gn., en sorte qu'il n'est pas dominant, mais figure seulement en tant que dtermin par ses propres formes, brahma tu rpavisesanatvena sasty nirdistam . . . na svapradhnatvena.4 Enfin (IV. 2,12 p. 373), le prvapaksin discute sur l'quivalence_entre Gn. et Abl. dans na tasya/na tasmt prn utkrmanti BAU. IV. 4,5 : le Gn., qui exprime la relation, la connexion en gnral,5 est ci "dtermin dans le sens d'une relation spcifique par l'Abl.", pancamy sambandhavisese vyavasthpyate. Hormis ces faits d'emploi casuel, qui prolongent et justifient la nette sparation qu'oprait P. entre catgorie casuelle et emploi des cas, seuis sont relever des traits disperss. Le prvapaksin 1.1,13 p. 66 discute sur le sens de -maya- dans nandamaya TU. : est-ce un vikrasabda,6 est-ce un lment notant "abondance" (prcurya)7! Aprs une longue disquisition, S. se dcide conserver le sens usuel "fait de flicit" (s. 19). Le mme s (1.1,19) fournit la matire de quelques autres allusions de grammaire: S. p. 76 autorise qu'un mot masculin en l'occurrence, un pronom se rfre au mot b., lequel est neutre: ainsi seront acquis pour le Vednta une srie de passages dcrivant un tre qui en apparence est htrogne au b. Le mme contexte
une identification, pas plus que la coordination thoriquement exprimable dans la phrase gaurasvah "le buf (est) un cheval " n'aboutit identifier rellement le buf au cheval. 3 L'Abl. est not pzravadhi " l i m i t e " dans la Ks. l. 4,24 et le mot marydd lui-mme figure P. I. 4,89. 4 Le Nom. exprime le svtantrya ou "autonomie", comme chez les grammairiens, 1.1,5 p.49-50,ainsi dans"(le solei I) brle, donne la lumire " . 5 Samhanhasmnyavisay hi sasthh Cf. encore 1.1,2 p. 16. Le terme sambandho figure dans la Ks. II. 3,50. Noter la distinction que fait S. entre Gn. du karman et Gn. du sesa pour expliquer le compos brahmajijns I. 1,1 p. 12 : le karman, "dsir de connatre !eb. " s'accorde mieux l'criture que les'esa "dsir de connatre li au b.". 6 Mot employ par la Ks. IV. 3,143 et qui d'ailleurs reprend une teneur de P. lui-mme IV. 3,134. 7 Mot reprenant le prcuryena prastutarn de Ks. V. 4,21.

472 Louis Renou

cite les noms techn. pratipadika p. 74, anukarsana p. 73 (onanukrsya "sans qu'il / ait lieu d'attirer [dans un texte un lment emprunt un texte antrieur] "), adhikro p. 72, qui, au sens de " influence gouvernante (d'un mot mis en tte d'un dveloppement)", reflte l'emploi grammatical du terme. 8 Ailleurs: upapada 1.1,7 p. 57 au sens de " mot servant d'attribut ou de dtermination accessoire". Sarvanman " pronom " 1.1,24 p. 93. Bahuvrthi 1.1,2 p. 16 (avec l'pithte tadgunasamvijnno, cf. Terminol. s. u. guno). Kormadhoroya I. 3,15 p. 180, passage o S. propose d'expliquer brohmaloko par " le monde (qui est) b.", afin d'viter l'attribution au b. d'une notion concrte distinctive. L'explication du dsidratif pipatisati "elle va tomber" (dit de la berge du fleuve) 1.1,5 p. 52 sert montrer comment les Smkhya s'entendent poser que le prodhno, entit notoirement dnue d'intelligence, est la cause du monde; l'exemple est connu en grammaire, Ks. III. 1,7. Le recours l'tymologie, malgr les facilits qu'il donnerait l'argumentation, est vit. Un exemple est le mot setu " pont" MuU. II. 2,5: S. enseigne que ce terme n'implique pas ncessairement l'ide d'une "autre rive", ide qui aurait pour consquence fcheuse que le sjour du ciel et de la terre serait distinct du b. : conformment l'tymologie par si- " lier", l'ide est celle de "tenir ensemble", "soutenir" I. 3,1 p. 156.9 Les cas qui prcdent sont des emprunts la Grammaire, valables pour des besoins limits, pour alimenter des gloses particulires. Tout autres et plus intressants sont les cas qui suivent, o un raisonnement de type grammatical entre en jeu. Ici se pose une question de principe. Ces mmes procds se retrouvent dans les commentaires sur la Mmms, parfois aussi dans ceux sur le Nyya. On ne peut strictement dmontrer qu'ils aient pris naissance chez les grammairiens, sauf l o le contenu est nettement grammatical. Nous savons que S. tait nourri de Mmms, nous constatons que des tranches entires de son argumentation en drivent. L mme o l'on penseatteindreun lementgrammatical, c'est travers la Mimms qu'il a pu parvenir S. A vrai dire, il existait un mode de raisonnement lmentaire, dont la plus

Ailleurs, odhikara est simplement "subject-matter", sens attest dans les s. mmes II. 3,12 et 21 III. 3,3, et qui avoisine celui eprakarona, plus frquent. Le mot anuvrtti figure notamment 1.1,24 p. 91 ; 3,40 p. 232. Notons en passant qu'il y'n a pas de glose "grammaticale" sur des particules telles que co ou v (v III. 1,7 sert carter une objection, comme plus souvent tu); des gloses sans porte grammaticale affectent atha (au dbut de l'uvre), iti (passim), evo (passim), tath II. 4,1. 9 Des paribhsgrammaticales sont misesen jeu III. 1,25 p. 131 utsorgpavdayos ca vyavasthitovisoyatvom " l a rgle gnrale et l'exception ont des domaines del i mits", axiome d'ail leurs

en contexte mmamsiste (Garge Citations in Sabara-bhsya p. 264) : "ne fais pas violence la crature" et "qu'il offre un animal tel dieu" ont des sphres d'application distinctes, Ou bien I. 1,4 p. 39 o S. traite des deux types de ngation; prasajya et paryudso, galement dans le cadre mmamsiste, Garge p. 265 et note de Thibaut ad loc. D'origine non moins grammaticale est un axiome tel que itaretaravisesanavisesyabhva IV. 3,1 p. 383; 2 p. 385 " relation rciproque entre attribut et terme pourvu d'un a t t r i b u t " ; passim aussi dharmadharminorabhedah, cf. Terminol s. u. dharma. La notion 'abheda " non-diffrence" a un substrat grammatical patent,

473 Grammaire et Vdanta

ancienne expression se t r o u v e dans le Mahabhasya,10 mode qui comportait I usage d'un certain nombre de paribhs ou clefs interprtatoires ; beaucoup de ces paribhs sont passes haute poque dans le Mahbhsya lui-mme. Les autres systmes n'en offrent pas l'quivalent, et elles ont fait loi, mutatis mutandis, dans le domaine philosophique, t o u t au moins dans l'une et l'autre Mlmms qui, plus qu'aucun autre systme, exigeaient un code de conventions. L'argument prliminaire, chez S., se fonde volontiers sur le sens tabli (prosiddha), sur l'usage (proyoga) attest (dorsono), t o u t comme les grammairiens font tat de l'usage (loko), de l'aptitude e x p r i m e r (obhidhno). Ce "sens tabli " f o u r n i t des valeurs directes, immdiatement sensibles. Il repose en principe sur la convention ou rdhi I. 3,40 p. 232. Entre deux valeurs tablies, il advient que l'une " monte (plus) vite la conscience" sghrom huddhim rohoti 1.1,22 p. 82. C'est l'emploi profane ou usuel du mot. La mthode vdntique, sur le plan de la Semasiologie, consiste pour une bonne part carter ce sens naturel, r e t r o u v e r une valeur indirecte, outrepasser en somme l'usage: ainsi le mot onno signifiera, non " n o u r r i t u r e " , mais " t e r r e " (dans C h U . VI. 2, 4) II. 3,12 p. 23. O n reconnat ce forage linguistique si commun lascolastique indienne depuis le N i r u k t a : l'exgse du Rgvedaen a fourni le prtexte et en a accrdit l'habitude. Mais, ici comme ailleurs, le recours la grammaire forme le propre du prvapaksa; le Vedntin accompli le transcende: c'est ce que font, plus ou moins ouvertement, toutes les disciplines classiques de l'Inde. D'autres considrations priment en effet l'usage ou l'apparence. Ainsi le prokorono ou " c o n t e x t e " . Toutefois il y a un risque abandonner le sens direct au profit du contexte I. 3,8 p. 166 tatra prakoronnurodhena srutih parityakt syt: on contrevient au

principe bien connu de la Mmms (Garge op. c. p. 276), allud par S. lui-mme III. 3,49 p. 262, aux termes de quoi la valeur directe (sruti) a plus de force que le signe-caractristique (lingo), le lingo plus que la connexion syntactique (vokyo), le vkya enfin plus que le prokorono. On reconnat l cette scrutation sur la "force" de tel type de rgles par rapport tel autre, domaine essentiel de la thorie des paribhs.
10

Qu'on pense la squence peu prs obligatoire : paraphrase littrale, expos du problme en litige (samsoyo), thse antrieure, thse dfinitive (faisant suite des objections partielles, des rpliques, des aperus ne comportant qu'une part de vrit Thieme GN. 1935 [this Volume page 299]). Cette squence se retrouve dans toute la scolastique indienne; mais nulle part, sauf peut-tre chez Sabarasvmin, elle n'est si voisine du modle patajalien que chez S., en dpit des sicles qui sparent ces deux matres. L'entre en matire si commune chezS..telle que 1.1,22 p.81 " une question se pose : le mot Ether (ChU. I. 9,1) dsigne-t-il le b. suprme

ou bien l'ther-lment? Pourquoi cette question? Parce que l'un et l'autre sont attests " kirn ksasabdena param brahmbhidhiyata uta bhtksam iti/ kutah samsayah/ ubhayotra prayogadarsant rappelle tant d'attaques toutes semblables chez Patajali. L'emploi de prpta pour dsigner un rsultat apparemment, provisoirement, acquis, en fait annulable par un raissonement ultrieur, est aussi frquent chezS. qu'en Grammaire. Cf. encore l'usage du terme proyojana I. 1,4 p. 38 " motivation " (non "final end " Thibaut) et une formulefamilire comme kim tova tena syt (bid.) kim krtarp syt (passim).

474 Louis Renou

La " v a l e u r d i r e c t e " c o m p o r t e elle-mme des amnagements; on peut faire appel d'une i n t e r p r t a t i o n moins informe une autre mieux informe, ainsi IV. 2,1 p. 364 o le pdrvapaksin t i e n t que c'est la parole mme qui se rsorbe dans la pense (au moment de la m o r t ) , cette explication seule faisant justice au "sens d i r e c t " ; mais le siddhntin comprend " la fonction de la parole se rsorbe . . . " : fonction et chose affecte de fonction sont comprises comme non-distinctes, vrttivrtti motor abhedopacrt. En ralit, S. use, suivant les convenances, de l'un ou l'autre de ces procds. Il fait tat, trs souvent, du prokorono, malgr la faiblesse relative de cet argument. Ainsi, dans un passage (1.1, 24 p. 90) o il entend m o n t r e r que la lumire (j/ot/'s, C h U . III. 13, 7) est bien le b. Les "signes caractristiques" sont galement invoqus, ainsi dans le dveloppement 1.1, 23 p. 85 o S. aboutit l'quation prno = b. Si le personnage sis dans le disque solaire est le b. mme (1.1,11 p. 63), c'est en raison du lingo, savoir, " le fait de n'tre associ aucun mal". 1 1 La " c o n n e x i o n s y n t a c t i q u e " joue un rle eminent (Garge op. c. p. 262). Si le mot " l u m i r e " (au passage dj cit, p. 93) signifie b., c'est que la phrase o il figure est lie la prcdente par le pronom yod, qui f o r m e un somorthyo, une " c o m m u n a u t smantique". Ce qui compte est moins laforce d'une donne initiale I. 4,19 p. 274 que la valeur du dveloppement entier, ibid. p. 275, ou bien la pluralit des connexions 1.1, 29 p. 99, o le t e r m e sombondhobhman figure dj dans le su. Ainsi en Grammaire, o Vadhikra initial est souvent perdu de vue mesure qu'on avance dans le chapitre qu'il commande. C'est la connexion de tel lment avec tel autre qui peut rendre un t e x t e intelligible 1.1,26 p. 96. Le vkyo peut provoquer un transfert en propulsant un mot hors de son domaine naturel 1.1, 24 p. 93 (contexte dj cit), svavisayc chakyprocyvoyitum. Le mot otideso " t r a n s f e r t " joue un rle eminent en Mmms (Garge p. 280) comme en Grammaire. On le t r o u v e encore 1.1, 20 p. 78 et 79, o il est dit que le puruso sis dans le disque solaire a des caractres qui se laissent transfrer au purusa sis dans l'il (passage dj cit). 12 Toute interprtation qui prserve l'unit d'une phrase vaut mieux que celle qui entrane une scission : principe connu des grammairiens ( T e r m i n o l . s. u. ekovkyo), chez qui une paribhs invite ne dduire de la "scission de r g l e " (yogovibhgo) que les rsultats "souhaitables". Mais d'aprs la Mmms (Garge p. 256), le vkyobhedo a une signification un peu diffrente: l'axiome entre en vigueur l o une mme proposition semble contenir deux t injonctions distinctes. C'est le risque que commande d'viter S. I. 2, 2 p. 108; 3,14 p. 177; III. 3, 58 p. 279 (cf. aussi 1.1, 31 p. 103) et qui lui permet de rfuter un prvapaksa. En ce dernier passage (III. 3, 58), l'argumentation consistant unir plusieurs donnes en un mme ensemble syntactique est rejete en posant que les phrases successives ne possdent aucun indice qui marquerait P ' a p p t e n c e " (knks) de l'une d'elles vers un lment compltif : chacune d'elles se suffit elle-mme, preuve qu'on a bien

La notion de iga mane du Rituel plutt que de la Grammaire, JA, 19411942, p. 153 [this Volume page 463]. En Grammaire, l'emploi peu prs corrlatif est celui de loksana. 12 Une consquence de la connexion

11

non-observe est l'erreur consistant abandonner un sujeten cours pour en attaquer un nouveau I. 1,15 p. 68; 24 p. 90 et passim, prakrtahnprakrtoprakriye.

475 Grammaire et Vdnta

affaire une " phrase " : la notion d'akaksa est connue aussi en Grammaire. La " c o n n e x i o n " est plus que la simple p r o x i m i t , qui en Grammaire peut servir de base une i n t e r p r t a t i o n , t o u t en passant pour moins " f o r t e " que le smorthya ( T e r m i n o l . s. u. pratysatti). S. use et lde l'argument de p r o x i m i t ; il distingue une p r o x i m i t moins grande et une plus grande (sa m n h tata r a) 1.1,19 p. 75. Mais il enseigne 1.1, 23 p. 86 que la p r o x i m i t ne prvaut pas contre laforce d'une " p r o p o s i t i o n c o m p l m e n t a i r e " (vkyasesa), en v e r t u de quoi, par exemple, prna sera identifiable b. Cette notion du vkyasesa repose, du point de vue grammatical, sur l'ellipse, cf. le Mahbhsya, T e r m i n o l . s. u., comme l'indique clairement l'expression iti vkyasesah 1.1, 2 p. 16. C'est un aspect particulier de la connexion. A ct des " p r o p o s i t i o n s absolues" (dont certaines sont dites "dcisives", nirnayavkya 1.1, 2 p. 19), il y a les vkyasesa'. le mot figure dans les su. mmes, I. 4 , 1 2 , ainsi que dans les MS. C'est souvent le vkyasesa qui donne la clef d'un Texte, ainsi la signification du compos (dj cit) pacajanah I. 4,12 p. 260 est prcise par le v " c e u x qui connaissent le souffle du souffle", d'o suit quepacajana (sur le plan vdntique) gale prna. La complmentarit rsulte d'un mme emploi rituel et d'une unit smantique, cf. le passage mmmsiste 1.1,4p. 23. La sruti ou valeur directe l'emporte, comme de juste, sur la p r o x i m i t III. 3, 55 p. 273, laquelle est faible p_ar nature III. 3, 25 p. 224, passage galement d'inspiration mimmsiste. Ainsi l'emploi simultan de la Grammaire et de la Mmms, ou si l'on prfre de la Mmms substructure grammaticale, p e r m e t t r a de saisir des gradations, partant des prfrences, des applications privilgies pour telle proposition qu'on cherche distinguer par rapport telles autres. Un autre artifice, plus productif encore, est la diffrence qui s'opre entre valeur primaire (mukhya) et valeur secondaire (gauna). Il s'agit l d'une des tendances fondamentales, la plus importante mme sans doute, de t o u t e la smantique indienne, de toutes les thories fondes de prs ou de loin sur la smantique, commencer par l'Alamkrasstra. 13 L'origines'en t r o u v e clairement du ct des coles grammaticales. Une paribhs conserve chez Patajali pose que " s i un mot a une valeur secondaire en mme temps qu'une valeur primaire, l'opration le concernant se rfre sa valeur primaire (seule)". Ceci rappelle d'un peu loin l'axiome prcit de la Mmms, aux termes duquel la sruti ou " v a l e u r d i r e c t e " prvaut sur t o u t e autre i n t e r p r t a t i o n . Chez S. aussi le sens primaire est prfrable au sens impliqu 1.1, 22 p. 82; 26 p. 96 ;14 IV. 2,1 p. 364 (passage dj cit). La dist i n c t i o n entre sens " d r i v " et sens primaire ne s'avre que l o existe une diffrence de fait entre les choses nonces 1.1, 4 p . 42

prasiddhavastubhedasya gaunatvamukhyatvaprasiddheh, II est vain de poser avec les Smkhya un sens figuratif au mot " il considra"
13 Le sens "secondaire" n'entrave pas le principe, connu aussi des grammai riens, de Vekrthatva du vocabulaire (racines verbales exceptes) : principe rappel par S. I. 1, 7 p. 56; 22 p. 82. Les grammairiens ont le juste sentiment que les acceptions secondaires d'un mot, si aberrantes soient-elles, ne contredisent pas son

unit smantique essentielle. II s'agit ici de la connotation du mot b.dans ChU III. 12, 7 : "sens primaire" veut dire que le mot b. ne signifie " rien d'autre que b.". C'est autour et par rapport au mot b. que s'agite tout l'effort smasiologique du Vednta sankarien.
14

476 Louis Renou

de [ U p . pour justifier que le sujet de ce verbe soit le neutre sat " l ' t r e " I. 1, 6 p. 53.15 Ce qui parat devoir dsigner des essences extrieures au b., par exemples les vocables " r i v e " , " mesure", " c o n n e x i o n " , " s p a r a t i o n " III. 2, 31 p. 175 et qui les dsigne selon le pdrvapaksin n'est en ralit que mtaphore; c'est ainsi que le b. est appel " r i v e " parce qu'il ressemble une rive d'une certaine manire. Soit le Texte relatif l'ther II. 3,1 p. 3 et suiv., lequel est t a n t t donn comme sans origine, t a n t t comme n du Soi. C o m ment expliquer cette contradiction ? A-t-on le d r o i t d'unir syntactiquement les deux propositions? On se ralliera plutt la position d'un "sens secondaire" : le Texte admettant une certaine origine de l'ther (s. 3 p. 5) un sens secondaire, autrement d i t figuratif. Tel m o t ainsi sambhtah (s. 5 p. 7) peut tre primaire dans telles conditions, secondaire dans telles autres. Le mot b. luimme est pris secondairement quand il se rfre la n o u r r i t u r e , primai rement quand il vise la flicit (ceci d i t propos de T U . III. 2-6, ibid.). L'quivalence, au mme passage, "topas b." est figure, c'est une bhakti ou " participation " assimilatrice. Les mots " naissance" et " m o r t " , lorsqu'ils s'appliquent au Soi, sont bhkta ou purement mtaphoriques, d'aprs le s. mme II. 3 , 1 6 p . 28: leur emploi primaire entacherait la permanence du Soi. Faut-il reconnatre le sens figur dans le cas de choses pour lesquelles les mots (vdiques) sont notre seul instrument de connaissance (promana) ? Ce point de vue est rejet par S. 1.1, 7 p. 56, qui y voit le risque d'un " manque de confiance" gnralis, sorvatrnsvsaprasangt; il consent que certains mots aient l'une ou l'autre valeur suivant les circonstances. 16 Il discute encore 111. 1 , 1 0 p . 1 2 0 s u r l e sens direct du m o t carona " c o n d u i t e " et le sens connotatif " rsidu des oeuvres " : ici les t. techn. employs sont srauta d'une part, oksonika de l'autre, qui font penser aux paribhs grammaticales bien connues oprant avec sruto/lksonika ou sruta/onumito (Terminol. s. uu.). Un autre aspect de la valeur " s e c o n d a i r e " est celui qu'on observe III. 3, 42 p. 255: on voque l ce principe mlmmsiste qu'un arthavdo ou " proposition nonant un tat d f a i t " ne forme qu'une matire secondaire, alors que les portions concernant le rsultat ou " f r u i t " d'un acte enjoint f o r m e n t une " matire princip a l e " et du point de vue linguistique sont entendre au sens direct, littral, toth hi gunavda padyeta pholopadese tu mukhyavdopapattih. Parfois le sens secondaire est prfrer. On enseigne IV. 4, 4 p. 408 que l'me dlivre est indivise avec le Soi suprme: l o un Texte nonce une sparation, c'est la non-sparation qu'on
15 Audits. estattest le motgauna, que S. glose par aupacriko "rsultant d'uneapproximation(mtaphorique)". 16 Toutefois il ne faudrait pas prendre un seul et mme mot, tantt au sens primaire avec tel attribut, tantt au sens secondaire avec tel autre, en un seul et mme dveloppement; I en rsu Itrait un " manque d'uniformit", vice qu'il faut tout prix viter II. 4, 3 p. 77, na hy ekasmin prakarana ekasmims ca vkya ekah sabdah sakrd

uccarito bahubhih sambadhyamdnah kvacid mukhyah kvacid gauna tyadhyavastum sakyamj vairpyaprasangt. Le sens secondaire est admis titre optionnel (v du s.) 111.1,7 p. 110 : il s'agit du mot anna dans " les Vaisyasont Vanna des rois" viso 'nnam rjnm, et analogues. Aussi III. 2, 4 p. 4 propos de l'individu qui, en dormant, "cre des chariots, etc." selon BAU. III. 3, 9 et 10.

477 Grammaire et

admettra, sur la base du "sens secondaire", bhedanirdesas tv abhede 'py upacaryate.

Vdnta

L'objet propre du Vdnta, tel que l'entend S., est de ramener l'unit les Textes et les notions inscrites, Thibaut p. LXVI. De manire analogue, l'objet de la Smrti savante est de rsorber les conflits virtuels entre les sources crites ou ceux entre droit et coutume. Celui des commentaires pninens est de concilier l'usage avec les rgles, quand ce n'est pas de permettre l'application non contrarie de deux rgles affrentes une mme opration. On ne peut accepter l'ide d'un conflit, mme entre les mantra yuktam/avirod'ht: pas davantage entre les diverses critures.17 De l l'emploi du vkyasesa (ci-dessus dcrit) et du "sens secondaire", qui voque de loin l'ontaragotvo et le bahiragatva des raisonnements grammaticaux. Un axiome banal est la " conformit de vues des textes rvls " srutismnya 1.1, 21 p. 81. Toutefois, il faut distinguer. De mme que certaines rgles de grammaire sont de simples dveloppements (proponed) d'une rgle antrieure, que d'autres sont des dfinitions, d'autres enfin des rappels d'une chose tablie (pnuvdo), non pas des injonctions proprement dites, de mme il y a dans la Mmms des arthavda (S. y fait allusion 1.1, 4 p. 23 et cf. III. 3, 42 cit ci-dessus), lesquels n'ont d'autorit que par l'effet de leur enchanement tel ou tel vidhi. De mme aussi que dans les s. grammaticaux le rptition d'un mot, d'une rgle donne une rgle prochainement conscut ive, serait une faute si ellen'avait quelque motivation (prayojana) particulire, de mme en Sruti la rptition de deux Textes indique qu'il faut les sparer l'un de l'autre, sinon la validit de l'un des deux serait compromise III. 3, 35 p. 242. Pareillement, rappelle encore S., la rptition dans la Mmms montre que deux actes rituels sont distincts. La reprise du mot b. I. 2,15 p. 127, a une incidence sur la valeur du mot prdicat, en l'occurrence ko et kha: le prdicat serait conu comme une simple qualit et ne pourrait servir de support une mditation. La rptition oriente lasmantique: le mot "flicit" tant repris dans TU. au sens de " 5 . " , il s'ensuit queje " Soi fait-de-flicit" n'est autre que leb. 1.1,12 p. 65. Chez S., comme chez les grammairiens, l'attitude initiale, vis--vis des teneurs rptes, est celle d'un " d o u t e " qui se dclanche, d'un problme qui se pose.
18 Le conflit est admis, du moins Le terme figure chez S. I. 1, 4 p. 27; titre prliminaire, I. 4,14 p. 263 et 265, 3, 42 p. 234 ("Qui est le Soi?" Rpour un domaine n'intressant pas la ponse "Celui qui est dans le cur" : doctrine, celui des choses cres. La s'agit-il l d'un anuvdo sur la nature srutivipratipatti de III. 2, 31 p. 175 du Soi migrant, dj connue par demeure au niveau des apparences. d'autres sources; ou bien le passage Noter que certains mantra et brvise-t-il tablir la nature dudit Soi? hmana, au dbut des Up., n'appartienArgumentation qui fait penser nent pas labrahmavidy, bien qu'elles \'aprpta= et laprpta-vibhs des en avoisinent l'enseignement, III. 3, 25 coles grammaticales) II. 3, 47 (les p. 223 : ceci exprime un fait dont nous Textes visent enseigner la nonavons aujourd'hui le sentiment, diffrence; sur la diffrence, il n'y a savoir que le dbut des Up. plonge que des anuvdo, comme on en a sur encore dans la substance des Brahmales choses naturellement tablies, na et n'chappe que progressivement svabhvaprpta). cette gangue. 17

et les brhmana 1.1,15p. 68, mantrabrhmanayos

cajkrthatvam

478 Louis Renou

Nous venons de mentionner l'expression " pour servir de support la mditation " : S. en effet fait intervenir la " mditat i o n " pour r rendre compte de certains attributs qu'au premier examen la Sruti confre au b. : ces fantmes o n t t voqus simplement, nous d i t - o n , " p o u r servir la m d i t a t i o n " , upsanrtham: ainsi 1.1, 24 p. 91 propos des pseudo-localisations du b. En G r a m maire, les mentions de lieux (udJcm, etc.), les noms de matres figurent pjrtham " honoris causa" : elles sont dnues de valeur injonctive par elles-mmes. Il ne serait pas excessif de m o n t r e r que les artifices de la my ou " altrit (phnomnale) " 1.1, 20 p. 80, de la ll ou " j e u (phnomnal)" II. 1 , 33 (s. !), du nmarpa ou " n o m - e t - f o r m e " II. 2, 2 p. 369, considrs comme autant d'obstacles l'identification fondamentale, reposent lointainement sur des postulats grammaticaux: la distinction entre la forme oprationnelle (krya) et la forme ralise (siddha), ou encore cette " non-ralisation " technique que rsume en des perspectives illimites le clbre prvotrasiddham de Pnini. Toutes les dmarches de S. postulent la recherche d'un tat second permettant d'annuler la multiplicit que f o u r n i t l'tat lmentaire. C'est aussi, en un sens, l'effort de la Grammaire.

C, Grammaire et Potique en sanskrit (1961) Louis Renou

The t h i r d comparative study is " Grammaire et potique en sans k r i t , " tudes vdiques etpninennes (8,1961,105-131 ). Here Renou takes a basic treatise on poetics, t h e Kvyapraksa, and studies some of the relationships between poetics and syntax. The Kvyapraksa was written by a Kashmir brahman, Mammata, who lived in the eleventh century A.D. Renou deals with discussions on the grammaticality of poetic forms; poetic analysis of long nominal compounds; expressions for poetic comparison or analogy; ellipsis; cases where a member of a compound is syntactically related to words outside the compound ; and other topics.

1. Prenant pour base le Kvyapraksa (KP.),1 notre objet limit est de rechercher ce que la Potique ajoute notre connaissance de lagrammaire, telle qu'elle rsulte de P(nini) et de son cole.

Texte cit d'aprs l'd. du Bhandarkar Oriental Research Institute (1950); pars(tra)'s et, ventuellement, par pages. Le dcoupage des kr(ik) en su.'s, ajoutons-le, n'a pas de justification interne (cf. en dernier Sivaprasad Bhattacharyya, d. du KP. 1, p. LXIV), pas en tout cas celle que peut possder le Rk-Prtis. l'autre texte o cette double division est atteste , savoir de nous rappeler qu'un sectionnement ancien en su.'s a t refondu en forme de kr.'s. La seule raison d'tre des pseudo-s". du KP. est qu'ils concident avec des dveloppements distinctifs de la v(rtti); c'est la v. qui, par sa disposition mme, commande cette fragmentation

en su.'s ou qui, l'inverse, amne grouper deux ou plus de deux kr. en une seule masse, laquelle est paradoxalement considre aussi comme un su. Sur les relations de la kr. et du s., cf. la dfinition de Bharata6.11 (d. p. 264 du vol. ! 2 etcf. K.M.Varma Seven Words in Bharata p. 69). Le KP. occupe une position centrale en Potique, centrale en ce sens qu'il condense et repense tous les problmes prcdemment abords; mais ce n'est pas un texte "de base", car le destin de l'Alamkrasstraa voulu qu'il n'existt point, au dpart de cette tradition, un corpus de su.'s comme il y en avait l'ore de la plupart des disciplines erudites ou spculatives.

479 Grammaire et potique en sanskrit

II convient de distinguer ici deux points. On trouve des allusions disperses plus nombreuses dans certains chapitres, ainsi dans celui des Dfauts, VII des faits de morphologie, avec renvoi explicite ou implicite, P.,2 allusions qui servent appuyer quelque interprtation, illustrer quelque trait du smantisme " potique". On trouve ensuite des faits de langue donns indpendamment du vykarana, pour en dvelopper les enseignements ou mme en dborder le cadre. C'est ce qu'on pourrait appeler les faits " linguistiques", par contraste avec l'autre srie, qui rassemble les faits "grammaticaux". Ce second aspect doit videmment nous retenir davantage, tant plus intressant et plus novateur. Encore faut-il noter tout de suite qu'il n'est pas toujours ais de voir le moment exact o la Potique se dissocie de la Grammaire. Nous laisserons de ct, comme de peu d'importance, les apports de phontique,3 qui se rsument en de menues disquisitions sur des cas de sandhi et 'asandhi, en des apprciations sur le caractre expressif de certaines consonnes. On peut omettre aussi les axiomes touchant la ncessit des tudes grammaticales pour l'apprentissage de la matire potique: axiomes qui font dfaut, au surplus, dans le KP., trait strictement proccup de technique et qui limite au minimum les exordes qui s'talaient dans la littrature antrieure.4
2

Ni dans le KP., ni dans les autres Traits que nous avons pu parcourir, il ne semble surgir d'enseignement remontant une source autre que P. et son cole (Mahbhsya, accessoirement Ksik-v.). Isol et sans porte est le fait que le Sah. Darp. 10.19 (v). signale que chez les Kalpa on dsigne par yinn-yi-nam ce que P. rendait par kyac-kya-namul. 3 Sur l'accent (vdique) et les trs rares allusions qui y sont faites en Potique, cf. Ruyyakacit H. Jacobi ZDMG. 62, p. 420 note, P. V. Kane History of Sanskrit Poetics2 p. 70 (cas de ndrasatru, qui d'ailleurs est allgu en Potique, non en partant du Veda, mais travers Patanjali). 4 Ainsi d'aprs Bhm, 1.9 la Grammaire vient en tte des savoirs requis pour le pote : sabda la dsigne, juxtapos abhidhna et artha (ou abhidhnrthal), qui vise l'tymologie (et ventuellement la smantique). Vm. 1.3,4 parle aussi de sabda{smrt), suivi o'abhidhna(kosa). Naturellement le kvya lui-mme est dfini le plus souvent par la combinaison de sabda et 'artha, mais o il s'agit de "forme" et de "sens", non pas de "morphologie" et de "smantique" Chez Bharata (chap. 6, dbut), il y a une utilisation imprvue du mot nirukta, au sens probable de "dfinition

reposant sur une analyse grammaticale" (cf. Varma, op. cit., p. 69 qui a tudi en dtail cette notion), dans les prliminaires la thorie du rasa : il s'agit d'un souvenir du "nirukta" vdique, mais dtourn de son sens ancien. Il y a en outre (d., chap. 1 4 = trad. Man. Ghosh chap. 15) une brve description des principes de la Grammaire, comme prliminaire aux rgles de prosodie. Dj donc chez Bharata comme dans la Potique proprement dite, la Grammaire est traite en simple "servante", en dpit des loges dont et l elle fait l'objet : sur ces loges ou l'affirmation de la ncessit de l'tude grammaticale, v. des rfrences chez Kane op. cit., p. 374 ou S. K. De Sanskrit Poetics 1, p. 10. Le texte le plus explicite sans doute, avec insertion d'un passage entier emprunt la Paspas, est la Kavyammmsa. D'autre part, on semble prouver une certaine antinomie entre Potique et Grammaire, depuis l'instant o se dveloppe la thorie du dhvani (cf. S. K. De 2, p, 162) : car, si la Grammaire se fonde sur les valeurs expresses du langage, le dhvani s'vertue dgager les valeurs latentes. Une indication plus prcise est celle que donne le chap, grammatical de Bhm. 6.29, savoir qu'on peut

480 Louis Renou

2. A l'inverse de certains Traits plus anciens il s'agit, en fait, des uvres de Bhm[aha] et de Vm[ana] , qui groupent des faits de langue, grammaire aussi bien que lexique, dans un chapitre t e r m i n a l , le KP. ne comporte aucun dveloppement linguistique autonome. Les citations de P. manquent l'intrieur des kr(ik), voues laformulation la plus abstraite, la plus ellipt i q u e ; elles sont rares dans la v(rtti). D'autant plus utile demeuret-il de relever l'emploi que fait ce t e x t e de valeurs linguistiques. Mais voyons d'abord en quoi consistent ces chapitres t e r m i naux de Bhm. et de Vm. ? Ils ont peu v o i r avec la potique proprement dite, si ce n'est par cet a t t r i b u t gnral qu'on y pourvoit le pote de conseils pratiques concernant des emplois prfrer, d'autres viter. A part quelques gnralits, on n'y t r o u v e en effet qu'une suite de cas concrets, suite succincte chez Bhm., beaucoup plus ample chez Vm. Ces cas sont allgus dans l'ordre mme o ils devraient figurer s'ils taient traits dans l'Astdhyy,5 ce qui confirme bien l'origine exacte de cet enseignement. Ceci dit, Bhm. et Vm. sont assez diffrents l'un de l'autre, malgr quelques concordances de dtail. Bhm, a souci de marquer que laformation dont il t r a i t e est non seulement acceptable, mais recommandable; il reste, cet gard, voisin de la Potique dans son ensemble, laquelle se fonde sur le choix. Ainsi pose-t-il que les suffixes " y u c " , " k u r a c " , " varac", " snuc" sont "singulirem e n t " (yisesena) utilisables en posie (48). Ou bien, dsireux de prciser une option pninenne, il enseigne qu'on doit prfrer mrjanti hmrjonti (31). Il admet l'emploi du participe parfait dans la bhs (42), alors que les Grammairiens laissent souvent la ques-

recourir l'upasamkhyono et l'/st/, phoniquement {kasta) et insolites, comme peuvent l'tre en franais qui sont des "additifs" quasi norcertains imparfaits du subjonctif. maux faits P. dans le Bhsya, mais 5 Ceci a t dj relev pour Vm. par qu'il faut viter d'user du yogovibhga Cappeller Stilregeln p. ix (dont nous ou "scission de rgle", cet artifice suivons ci la numrotation des s.), qui, bien qu'tant connu parfois du mais c'est galement vrai de Bhm. II. Bhsya lui-mme, se dveloppe surtout s'agit donc d'une " durghatavrtti" chez les commentateurs postrieurs et tend l'abus; une paribhs (n114) l'tat embryonnaire; les donnes de Bhm. et de Vm. offrent du reste plus cherche en limiter les effets, yogavid'un point de contact avec le texte de bhgd istasiddhih "d'une (nouvelle Saranadeva; nous en signalerons rgie obtenue par) scission (en deux) plusieurs. Mais le grammairien va plus d'une rgle (ancienne, on ne peut loin que ses devanciers dans l'effort tirer que) les rsultats (qui apparaisde "conciliation ".Ad P. 3.2,162, il citesent) souhaitables (pour tel cas un enseignement de Vm. (40) cond'espce) " (la paribhs n'est pas cernant la nuance du driv en -ura-: reprise, l'inverse de tant d'autres, en fait, Vm. s'tait born instaurer dans les su. du Trait grammatical de sous forme d'une isti une indication Bhoja. Bhm. fait aussi allusion au dj donne dans la Ksik, indication Nysa, ci-dessous 8. Des dtails qui n'avait pas laiss de provoquer aussi prcis font dfaut chez Vm. II quelque remous parmi les comn'est que nature! que des allusions de mentateurs, cf. nos tudes de gr. skte grammaire dans une strophe potique 1, p. 122. Ainsi lagrammaire de Bhoja soient considres comme un dfaut, (1.4,233) pose d'emble la valeur ainsi KP. su. 81, p. 413, moins, ajoute rflchie (karmakartari), mais ajoute l'auteur, que lastrophe incrimine que, selon certains, la valuer active ne soit mise dans la bouche d'un grammairien. Quant aux aoristes de la (-transitive) est seule correcte. str. cite p. 414, i Is sont la fois rudes

481 Grammaire et potique en sanskrit

t i o n controverse. 6 Il arrive aussi qu'il restreigne, sur lafoi de l'usage potique, un emploi q u i chez P. se tenait dans un cadre t r o p large : i m p o r t a n t e est son observation (40) touchant les formations " kvip", qu'il souhaite v o i r se l i m i t e r l'lnstr. sing, et au Loc. p i . Il circonscrit aussi (39) le d r o i t qu'on a t h o r i q u e m e n t de t i r e r des formes nouvelles en partant d ' u n gana pninen. 7 Il pose enfin (32) q u ' o n ne peut, d'une rgle gnrale c o m m e P. 1.2,64 (sur les ekasesa), d d u i r e qu'un n o m b r e d t e r m i n de formes, celles savoir q u i e x p r i m e n t la relation dfinie par P. 67, c o m m e varunau, indrau, bhavau, sarvau, mrdau : il s'agit donc d'endiguer un usage q u i risquait de dpasser les bornes permises. 8 Jamais Bhm. n'carte de f r o n t un enseignement de l'cole pninenne; ses rares noncs ngatifs sont conformes des pratisedhastra de P., ainsi ceux qui visent le sasthtatpurusa d ' u n n o m en -tr-(36) ou en -aka-.9 Les deux kr. ultimes sont orientes un peu diffremment, attirant notre attention sur une formation verbale assez rare, au sujet de quoi l'tudiant pourrait trbucher, l'aspect en -at- du participe des prsents redoubls, et la 3e pers. plur. corrlative en -ati, ou bien encore le -/- de liaison du type roditi, svapiti. 3. Chez Vm., lathorie grammaticale a plus clairement l'allure d'une sabdasuddhi ou d'un sausabdya, une " purification des formes du langage", comme les donnes qu'grnent tout au long des s. pninens les commentateurs tardifs ainsi la Bhsvrtti, et dj, timidement, laCandravrtti et la Ksik, et qui se dploient plein dans la Durghatavrtti. On trouve d'abord chez Vm. des formations que l'auteur tient pour incorrectes, ainsi avaihi 71, apnganetr 72, slistapriya 73 (cit aussi Durgh. ad 6.3,34, qui justifie la forme), nidrodruk 88, nispanda 89, ogulisaga 90; propos de srvara 52, Vm. n'a pas fait l'effort qu'on pouvait attendre pour justifier la drivation, effort auquel se livre sans peine Durgh. ad P. 4.3,11. Ces mots et plusieurs autres sont-ils considrs par Vm. comme absolument incorrects, ou seulement comme sujets caution ? On peut se le demander, voyant les vocables fort divers dont l'auteur se sert pour stigmatiser ces emplois, tantt anvesya, cintya, durlabha, tantt avadya, ou encore a(nisedhya) ou simplement (88) no.10 Dans la majorit des cas, les formes, vicieuses selon l'explication grammaticale obvie, peuvent cependant se concilier avec la Grammaire au prix d'une explication moins apparente. Ainsi en va-t-il de sudat69: ou bien l'on donnera une valuer cumulative au CA de P. 5,4,145 (ce que fait dj la Ks., reprise par la Bhsv.), ou
Les cas de mrjanti/ mrjanti rsulte en partie de P. (7.2,114), en partie (et surtout) d'une sti dans Bhsya 1.1,5 (trad. Ojihara-Renou p. 28). Quant au participe parfait, question fort embarrasse, cf. nos tudes de gr. skte 1, p. 99 et 133 et, en dernier lieu, notre article sous presse du JAs. 1961 sur la Thorie des temps d'aprs les grammairiens sanskrits [see pages 500-525 of this volume]. 7 Sur Vkrtiganatva de P. et les problmes que pose la prsence des gana, cf. l'introduction l'd. critique du Ganaptha, ouvrage (sous presse)
6

de R. Birw. Cf. Durgh. ad P. 67 et le long bhsyo ad P. 64, notamment autour du vt, 46 (liga-vacanasiddhih). 9 Cf. ci-dessous 8. Ibid., galement, la question, un peu obscure, de Yavyaylbhva. 10 Plusieurs de ces mots sont repris Durgh., passim. Le NA pninen ne signifie jamais qu'une forme (en usage) est viter, mais seulement que l o les rgle(s) positive(s) prcdemment nonce(s) cesse(nt) de valoir dans les cas d'espce qui sont noncs ensuite,
8

482 Louis Renou

bien l'on considrera sudatJ comme un " n o m " (samjn) et l'on recourra P. 5.4,143 (passage o justement la Bhsv. cite le mot en exemple); la Durgh. ad 5.4,141 connat encore un autre biais. Vm. 45 fait appel aussi au CA, comme il se retranche ailleurs (79) derrire le BAHULAM, comme il fait tat du niptana 43 (pour sobh, mot tudi aussi Durgh. ad 3.3,102) et 85 (pour vira ma).u La technique des gana llustratifs (krtigana) permet de "justifier" plusieurs formes embarrassantes, ainsi sous 14,16,18, 33, 75. Il advient qu'on ait utiliser une analyse (autre que celle normalement attendue) 37, 38, 84, une tymologie distincte 76, 77, 78 (ainsi dans praplya, qui viendrait, non de la racine p-, mais de la racine pi), un dcoupage du mot incrimin (ainsi 82, o le futur vetsyasi est expliqu par vetsy asi). Certaines formes relvent d'une interprtation par l'analogie (c'est l une tentative plutt rare d'explication de type scientifique): ainsi l'Instr. mukhena dans mukhena trilocanah 26 est donn comme une extension licite de P. 2.3,20, aksn knah\ ha en fonction de prtrit (46) rsulte de l'analogie des autres formes verbales en -a, toutefois Vm. ajoute aussitt que les bons crivains usent de ha au sens de prsent. Enfin il y a des explications qui, sans faire tat d'une rgle de grammaire spciale, prennent prtexte d'une valeur "gnrique" (sarvatra), ce qui permet laforme litigieuse d'chapper l'application d'un su. grammatical : ainsi pour drdhabhakti 73 bis (manque Cappeller) ou sakya 25 (avec appui d'une citation tire du Mahbhsya.12 Plus intressants, de notre point de vue, sont les cas o une forme est dclare correcte, malgr la Grammaire, sur lafoi de l'usage qui l'accrdite. Cette pression de l'usage vivant a t une tendance naturelle, que le temps n'a pu que renforcer: de fait, il est peu d'auteurs de kvya qui, tel ou tel moment, n'outrepassent les limites de la stricte adhsion P. Ainsi Vm. 83 admet kmayna au lieu de kmayamna, pourvu, ajoute-t-il curieusement, qu'un pareil emploi repose sur une tradition, autrement dit, ne fasse pas novation. 13 De mme pour ssvata 53 (cf. Durgh. ad 4.3,23) ; te et me au sens de tvay et maya 11. Mais ces cas sont peu frquents. Vm. est domin par le souci de "concilier", plus encore, par celui de maintenir la puret du langage en attirant l'attention sur les barbarismes, apabhramso raksyah 23. Il se tient ainsi mi-chemin entre les tendances quelque peu mcanistes des grammairiens et le comportement qualitatif des poticiens. L'option qu'il donne entre smagrya et smagr (58) ne fait que codifier un enseignement de vykarana.u Quelques notations de caractre plus global ont lieu au su. 62, o Vm. marque que le suffixe -tarasert aussi pour des valeurs non strictement "comparatives" ; ou bien, dans Yadhyya 1 (du 5e adhikarana), quand il expose le sens de la double ngation (9) ou l'emploi du duel (17, avec recours l'autorit, non de P., mais de la Mmams). Ce sont l les rares
11 Sur viramo, cf. un cas analogue cit l'auteur a eu en vue, non le cas dans Bhsv. ad P. 7.3,34 et dans spcial de kmayno (mot pique), Durgh. ibid., p. 109 (udyama, uparamais le groupe des formations simima). laires. La v. s'appuie sur une (pseudo-) 12 Sur drdhabhakti, cf. Durgh. ad P. paribhs, n93/2, dont on a fait grand 6.3,34; sur sakya suivi d'un infinitif, v. abus par ailleurs. nos Monogr. sktes 2, p. 48, touchant la 14 Cf. les rfrences chez Debrunner doctrine des grammairiens. Suffixe p. 397 bas. 13 Anadi ced. Il est donc probable que

483 Grammaire et potique en sanskrit

donnes qui ne reposent pas sur l'exercice d'un artifice technique.15 4. Dans le KP. comme dans la majorit des Traits de potique, les composs nominaux sont de beaucoup la catgorie grammaticale la plus souvent allgue.16 Non pas que ces textes aient du neuf nous apprendre sur la structure interne du compos ; ce n'est pas leur rle. Mais justement, par contraste avec la Grammaire qui ne met en uvre que des formations lmentaires, des cellules bimembres,17 la Potique, elle, traite des composs longs, ou plutt elle en confirme l'existence, l'importance stylistique, elle en mesure (adensit. Considrant les Qualits en tant qu'elles se rfrent au mot (sabdoguna) car, remarque-t-on au su. 95, les Qualits rsident en fait dans les sentiments et ne sont prsentes que secondairement dans les mots , le KP. (s 98) enseigne qu'elles ont pour facteur "suggrant" (vyajaka), soit les phonmes, soit le compos nominal, soit enfin l'arrangement (racan: entendez, une sorte d'harmonie pouvant rsulter de l'arrangement des mots dans la phrase).18 Or, il est ais de voir que le compos nominal est un indice plus important que les phonmes ou mme que l'arrangement; en fait, si l'on prend l'ensemble des Traits, on s'aperoit qu'il est prsent dans toutes les dfinitions, alors que les autres signes n'y figurent que temporairement. Des trois Qualits que distingue le KP., le mdhuryo ou " suavit", typique du "sentiment" Erotique et plus encore, nous dit-on, du Pathtique, du Frustratif ( l'intrieur de l'Erotique), du Quiescent, comporte soit absence totale de composs (avrttf),19 soit du moins prsence de composs de dimension moyenne (madhyavrtti), s. 99: ceci vade pair avec la prsence de phonmes doux, l'absence de phonmes rudes, l'allure "coulante" de la diction.

15 Nous relverons quelques autres faits emprunts ces mmes sections de grammaire (8). 16 Cf., outre les 4 8, l'emploi des composs dans l'upam 9 et 10 dans la thorie des Dfauts 12. Bharata (15.34 et 35 trad. Man. Ghosh = 14.34 de l'd.) croit devoir dfinir le compos (luptavibhaktir nmnm ekrtham samharat samso'pi) au terme de sa brve exposition de grammaire (cidessus p. 106 n. 2). C'tait pour lui une manire de conclusion. 17 Le principe dans l'AstdhyyT (cf. BSL. 52, p. 103 et n. 1) est de laisser la formation libre, tout au moins pour le bahuvrhi et le dvondva ( l'exception du dvandva singulier, qui est rglement); bien que limit en apparence, le tatpurusa est aussi, en fait, largement ouvert, cf. notamment P. 2.1,57 2.2,6 et 8, comme le montrent, par preuve inverse, les interdictions nonces sous P. 2.2,10 17.

Enfin SAHA SUP 2.1,4 pouvait frayer la voie diverses facilits. Les composs plurimembres ne sont signals que rarement par les ce, ainsi quelques dvandva donns en contreexemples ad P. 2.4,6; cf. aussi (pour les dvandva) Bh. ad 2.2,34 vu. 1-3 et 6; pour les bahuvrlhi, Bh. ad 2.2,24 vt. 1, spculant sur la teneur quelque peu quivoque ANEKAM de P. (ad loc). Le type skaprthiva 2.1,69 vt. 8 atteste indirectement la prsence de t r i membres (utta rapada lopin). 18 Disposition ternaire qu'on retrouve ailleurs 9 et 11. 19 Vrtti au sens de " compos" est un emploi qu'on trouve et l en Grammaire, ainsi Bh. 2.1,1 init. ou Ks. 2.1,34 et 40: c'est un aspect particulier du mot au sens de "expression synthtique", englobant donc la drivation nominale, d'o la formule samsataddhitnm vrttih Bh. ad 2.1,18 vt. 2.

484 Louis Renou

La Qual t de " f o r c e " ou d"'ampleur puissante" (ojos, glos proudhi)20 exige, l'inverse, des composs longs (vrttidairghyo), indice que corrobore l'emploi de certaines consonnes ou combinaisons consonantiques, su. 100; on ajoute parfois aussi : la construction tendue ou heurte. La troisime Qualit, le prosada ou " limpidit", su. 101, ne possde pas de critres formels, ou du moins pas de critres positifs: nous de les dduire par comparaison avec les deux autres Qualits. Le prosoda est utilis pour exprimer tous les rasa ( l'inverse de Vojas qui exprime l'Hroque, plus encore le Rpugnant, plus enfin le Furieux) et convient toutes les dictions. 5. Cette thorie a une assez longue histoire. Les trois Qualits du KP. se sont substitues (comme le note expressment la v.) aux dix catgories de Vm., catgories qui taient dmunies de signes linguistiques directs. Mais ces signes surgissent ailleurs, de temps autre: ainsi dj chez Bharata (16.100 = trad. Man. Ghosh 17.99) la Qualit dite somata " galit" comporte des traits ngatifs: pas trop de mots " broys" (ou : pas de mots " broys" l'extrme), pas d'expressions redondantes ou difficiles comprendre, nticrnapadair yukt, na ca vyarthbhidhyibhih/durbodhanois ca . . ., ce qui semble signifier: pas trop de mots non composs (qui laissent une impression hache au discours)21 : l'absence d'excs, l'unit l'intrieur de la stance est recommande, unit que viole un contre-exemple cit Vm. 3.1,12, o nous voyons un premier hmistiche fait de mots simples, le second ayant un compos trs long. Toutefois la situation du compos nominal n'apparat de plein fouet qu' propos de la description de l'ojas. Ds Bhar. 16.105 ( trad. 17.102), on enseigne que l'ojas abonde en mots composs, somsavadbhir bohubhih . . . padair yutam (les pithtes qui suivent, vicitraih, snurgaih, udraih, n'ont pas d'incidence directe sur le langage). Dandin 1.80 dfinissait Vojas par le mot " samsobhyastva, dominance (en nombre) des composs".22 Vm. 3.2,2 reconnaissait pour l'une des manifestations de l'ojas envisag comme Qualit de sens le sarnosa,23 c'est--dire, non pas direcSur ojos et les implications du mot en Potique, v., outre les Manuels d'ensemble, la monographie de J. Gonda Ancient-lndianojos, notamment p. 37. 21 Le mot crna sert de t. techn. pour dsigner un certain type de composition en prose (Vm. 1.3,24) comportant des composs " point trop longs " (ainsi que des mots empreints de "douceur"), par contraste avec le type dit utkalik(prya) 25. D'aprs l'Agni Pur. 1.13, le crnaka (caractris selon 1.9 par le petit nombre de composs; compos se disant vigraha, mot qui en Grammaire dsigne plutt l'analyse des membres du c) domine le genre littraire appel khyyik, cf. ci-dessous 6 fin., alors que l'utkalik (1.9) a des composs longs. Le vrttagandhi, qui chez d'autres thoriciens est simple20

ment, comme son nom l'indique, un type mixte prose/vers ou une prose versifie, est ici (1.10) caractris par des composs point trop difficiles (utkota). Remontant jusqu' Bharata (d. 14.38 trad. Man. Ghosh 15.35), nous voyons que crna, crnapada, est simplement contrast nibaddha (bandha): la question des composs demeure non explicite, 22 L'expression est reprise, par ex., dans SKbh. 1.71, le mdhurya tant ci signal par l'existence de mots spars 1.68, la susabdat reposant sur la "clart des sup et des tin " 1.72: c'est l'quivalent du prosado qu'on a ailleurs. Dans Sah. Darp. n 608 sqq.r le mddhuryo a peu ou point de composs, l'ojas en a beaucoup; rien n'est dit sur le prosada. 23 Ce samso de Vm. est allud KP. su. 96 (v.).

485 Grammaire et potique en sanskrit

t e m e n t le " c o m p o s " au sens grammatical, mais la facult de resserrer en un mot (ce qu'on pourrait aussi dire en plusieurs) : c'est le samksepa de Bhoja, le slesa de Bharata. Il y a l un essai, normal chez Vm., pour dpasser le niveau linguistique, t o u t en s'appliquant le transcrire autrement. Le Sr. K. Abhar. 1.36, dfinissant l'antonyme de l'ojos, parle d'une structure brise (khandayan rJtim) et ajoute " asamasta, absence de composs". Quant au rle majeur de l'ojas dans la l i t t r a t u r e , il suffit d'voquer ce que dit Dandin, loc. cit. : l'ojas est l'me de la prose; mme en posie, les auteurs (autres que les Mridionaux) y t r o u v e n t leur unique recours, etad (ojos) gadyasya jvitam/padye 'py adksintynm idam ekam parayanam.2* 6. Une autre classification o interviennent les composs nominaux est celle des Styles ou rti (parfois mrga ou " chemin " ) . Ils y sont mme la marque fondamentale, t o u t autant que dans les Qualits; au surplus, il y a u n lien t r o i t entre Qualits et Styles, comme on le constate du premier abord. Ces styles dnominations rgionales (ou censment telles, comme les noms des prkrits) nous sont dcrits, non dans le KP., mais dans les traits antrieurs, ds Dandin 1.42, qui note que les dix Qualits au nombre desquelles figure Vojos sont l'me du vadarbha marga, l'inverse de ces Qualits tant le propre du gauda vortman.25 C'est aprs Dandin que se fait l'association entre ojas et Gauda's (ou gaudy n t / j : cette gaudJy nti selon Vm. 1.2,12 a les Qualits d'ojas (et de dpti ou " c l a t " ) , d'o l'on peut infrer la prsence de composs longs. A cette nt/s'oppose la va da rbh (Vm. 19) qui, du moins sous sa f o r m e " p u r e " , est exempte de composs; enfin la pncll (ibid. 13) qui, ayant les Qualits de suavit et de t e n dresse (saukumrya), doit (est-on amen penser) se c o m p o r t e r , quant aux composs nominaux, comme la Qualit mdhurya. La vaidarbh rpond donc la Qualit prasda, malgr un certain dcalage. Dans les textes ultrieurs, les caractres formels se prcisent et durcissent. Ils sont trs explicites dans le Sr. K. Abhar. 2.28 sqq. qui dfinit la vaidarbhl comme dnue de composs, \dL pncll comme ayant des composs de cinq ou six membres, \a gaudy comme faite de mots composs " t r s v i g o u r e u x " , samasttyudbhatapada; les trois autres " s t y l e s " sont l'vantik, qui ades composs de deux trois membres, laly, qui mlange les mots composs aux mots simples, enfin \a mgadh, avec son " s t y l e b r i s " , khandanti. Analogue dans le Srng. Prak. 3 p. 212 sqq. 2 6 : gaudy, composs trs longs, pncll, composs point t r o p longs, vaidarbhl, pas de composs, ltly, peu de composs. Dans le Sah. Darp. n 625 sqq., la vaidarbhl, peu ou pas de composs, gaudly, beaucoup de composs, pncll, des formations de cinq ou six membres, ltl, intermdiaire entre vaidarbhl et pncll. De mme dans l'Agni Pur. 4.9. Dandin qui, nous l'avons rappel, ignore ces subdivisions, admet pourtant (1.46) que les Gauda prfrent ce qui n'est pas " t r o p conventionnel, ntirdham", ce qui revient
24

II s'agit des Orientaux (cf. 1.83), ce qui rejoint la rpartition entre Gauda's et Vaidarbha's; les Vaidarbha sont assimilables aux Mridionaux d'aprs 1.60. 25 Dandin admet qu'il y ait plusieurs mrga, mais n'en dcrit que deux, vaidarbho et gauda. Il signale le

premier (1.44) comme tant fait de bandhagaurava "compactness of (syllabic) binding " Belvalkar, ce qui ne s'accorde gure aux caractristiques qui seront instruites plus tard. 26 V. Raghavan, Analyse du Srg. Prak., p. 200, qui cite encore d'autres textes.

486 Louis Renou

dire qu'ils s'intressent aux formations analysables: l'exemple cit l'appui atteste en effet un compos long, comportant une "analyse" morphologique et smantique relativement complexe.27 7. Ici intervient dans KP., su. 102, un lment nouveau de quelque intrt, qui n'appartient pas aux Traits surtout descriptifs des poques antrieures, mais puise dans le Dhvanyloka. Les caractres servant dfinir les Qualits sont modifiables selon le sujet parlant (vaktr), la chose dont on parle (vcya), le genre littraire (prabandha): ainsi l'ojaset partant, le compos lourdest admissible quand parle Bhima(sena), personnage incarnant le rasa Furieux, mme si le contenu de lastrophe ou legenre dont elle relve ne s'accordent pas l'ojas. De mme a-t-on l'ojas l o le sujet trait le commande (en l'occurrence, la dcapitation du dmon Kumbhakarna), bien que le sujet parlant soit un simple hraut (yaitlika) et le genre un ntaka, type peu propice aux longs composs.28 En effet, le genre littraire affecte l'aspect du langage, en ce sens, dit la v.t que dans les oeuvres dramatiques les composs lourds sont viter mme s'il s'agit de rendre le rasa Furieux: fait, ajoutet-on, qui n'est cit qu' titre d'exemple. 29 Le Dhvanyloka formule ce point une remarque assez pntrante (3.5): on distingue d'abord la structure de phrase, samghatan, selon qu'elle est dnue de composs, ou " pare" de composs moyens, ou enfin pourvue de composs longs : nous voyons ici la morphologie pousse au premier plan, sans mixture d'autres facteurs, stylistiques non plus que phontiques. Un peu aprs (3.6) est note la diversit du langage en fonction du rasa: c'est nouveau la question des composs qui la signale, composs longs dans l'Erotique, mots simples dans le Furieux (et autres rasa), mme s'il est admis qu'en principe l'ojas convienne au Furieux, la suavit ou la placidit l'Erotique ou certains drivs de l'Erotique. Cette constatation contradictoire amne aussitt une question : tant acquis (par d'autres Traits) que \'ojas est caractris par les composs longs, comment une structure sans composs sera-t-elle apte rendre la " f o r c e " ? Ladpti qui se manifeste l'occasion du rasa Furieux n'est autre que la force ou ojas: y a-t-il donc une " f o r c e " qui peut se manifester sans composs?
27

Selon la KvMT (trad. fr. p. 49), la gaud est riche en composs, la pacal pauvre, la vaidarbh en est dmunie. Le mme ouvrage, p. 101 et suiv., dveloppe une indication du Visnudharmott. 3.4, suivant laquelle les diverses catgories d'tres surnaturels ont chacune leur langage; l encore, les composs (du moins dans KvMT., non dans Visnudharmott.) sont signals au nombre des caractristiques eminentes, aussi bien de la langue des dieux que de celles des Vidydhara, des Gandharva, des Yoglnl, enfin des Nga, sans qu'il y ait d'ailleurs (et pour cause !) des variations trs marques de l'une l'autre. Ce texte prcieux montre bien comment le problme des composs est au centre nerveux de la

thorie "ancienne" de l'Alamkara, au carrefour de la description des guna, des rlti, des rasa, et mme des vrtti, qui sont associes aux rlti dans le chap. 3 de la KvMT. 28 Ceci est rsum dans Hem. s. 103: la structure se conforme parfois au sujet trait sans gard pour la personne qui parle ni le genre littraire, parfois en revanche elle est adapte au genre en ngligeant sujet parlant et sujet trait, kva cid vaktrprabandhnapeksay vcyaucityd eva . . . kva cid vaktrvcynapeksh prabandhocit eva.
29

Sur un plan plus gnral, il y a la diffrence entre posie et prose (diffrence effleure Dandin 1.80 dj cit) et celle entre les divers types de prose, ci-dessus p. 484 n. 21.

487 Grammaire et potique en sanskrit

L-dessus cette remarque: la structure linguistique et les caractres e x p r i m e r sont deux choses distinctes. Il arrive qu'une structure composs longs entrave la perception d'un rasa; s'il s'agit de rendre la tristesse ou la souffrance d'amour, on aurait t o r t de prfrer les composs longs, les moyens y sont bons. Les composs longs peuvent n'tre pas hors de propos pour dpeindre un caractre noble-et-fier, dhroddhata, entendez: mme dans le cont e x t e de l'Erotique o on ne les attend gure. 30 Ce qui importe, c'est la qualit de " l i m p i d i t " ou " p l a c i d i t " : s'il n'y a pas " placid i t " , une structure mme exempte de composs sera inapte voquer la tristesse o la souffrance d'amour; s'il y a cette qualit, des composs moyens n'y feront pas obstacle. Tel est le sens vers lequel le pote doit p o r t e r son effort. Quant aux genres littraires tudis en fonction du rasa, le Dhvany. note que mme dans Vkhyyik (o l'on s'attend des composs longs), 31 ces composs sont viter s'il s'agit d'un rasa ne les comportant pas: le rasa prvaut sur le " g e n r e " . Dans les drames, on s'abstiendra de composs en t o u t cas: ici le " g e n r e " l'emporte. Mais c'est pour rendre le Furieux ou l'Hroque qu'il convient s u r t o u t d'en t e n i r c o m p t e : ainsi, il ne saurait y avoir uniquement des mots simples dans Vkhyyik, ou des composs t r o p longs dans le drame, quelles que soient par ailleurs les rgles prsidant ces deux rasa. Tel est l'enseignement, rudimentaire par certains cts, mais insistant et t o u t de mme fondamental par d'autres, que livrent les thoriciens. La question du compos nominal, bien que prsente de manire moins pressante, est aborde de nouveau propos de la Comparaison (9 et 10), elle revient dans le chapitre sur les Dfauts (12). Mais reprenons d'abord les quelques d o n nes rsultant des chapitres grammaticaux de Bhm. et de Vm., chapitres d o n t nous avons dj brivement parl. 32 8. Ces donnes n'intressent certes pas l'essence mme de la potique; mais elles sont utiles du point de vue linguistique. Il est clair que Bhm. et s u r t o u t Vm. s'appliquent l i m i t e r l'usage de certains composs dont la Grammaire leurs yeux t i e n t l'ventail t r o p ouvert. D'abord chez Bhm. : limitation des avyayJbhva quant l a f o r m e casuelle (c'est un raffinement, d'ailleurs assez obscur, apport par Bhm. 6.34 P. 2.4, 83 et 84, raffinement s'appuyant sans doute sur le BAHULAM du grammairien). Limitation du gana tisthagu Bhm. 35, cf. P. 2.1,17, rgle au sujet de laquelle le vt.
30

KP. su. 81, p. 431 donne le cas d'un compos long qu'on n'attend pas dans une ambiance d'Erotique. Sousentendre: il est tout de mme acceptable, servant intensifier le dplaisir ressenti par l'amant qui souffre. 31 Autre avis sur Vkhyyik dans l'Agni cit p. 112 n. 1. Sur ce terme techn., v. en dernier S.K. De Problems of Skt. Poetics p. 65 qui rappelle (p. 74) que ce genre littraire est en prose de type crna. Pourtant le modle convenu de Vkhyyik n'est-il pas le Harsacarita (cf. KP., fin du chap. VIN), qui consiste presque tout entier en composs immenses? Hem, loc. cit.: dans Vkhyyik, mme

pour l'Erotique, I ne doit pas y avoir trop de mots "suaves"; dans lakath, mme pour le Furieux, il n'y a pas user d'un style "sauvage"; dans le drame, mme pour le Furieux, il n'y a pas de composs longs. 32 Rpartition des composs selon les mtres, cf. Suvrttatilaka2.17 (on note l'absence de composs en vamsastho); 23 (c en premier hmistiche, en mal i ni); 27 (pas de c en prthvl); 37 (c au second hmistiche, en srd/a); tout cela parat bien arbitraire, comme les particularits euphoniques adjointes, et la mention inattendue du participe prsent (11).

488 Louis Renou

1 donnait dj la spcification d'emploi "ka/owsese", celle mme qu'exploite ensuite Bhm. Interdiction (Bhm. 36) du sasthtatpurusa pour un nom en -tr-, interdiction conforme P. 2.2,15, mais avec cette prcision notable que, aux dires du poticien, les personnes " c u l t i v e s " et l'auteur du Nysa admettraient un tel compos: de fait, le Nysa instaure (ad P. 2.2,16) un " kva cid", succdant l'artifice du jopana, et cite en exemple la forme bhayasokahartr (variante: hantr), qui ressemble f o r t l'exemple vrtrahantr que donnera Bhm. dans des conditions similaires, pour le rejeter il est vrai, alors que le Nysa le dclarait acceptable. 33 La description chez Vm. est plus articule. La remarque la plus instructive concerne la distinction entre bahuvrhi et tatpurusa, distinction sur laquelle la Grammaire tait muette, si ce ne'st l'enseignement indirect qu'on pouvait t i r e r du fait que le champ des samsnta y tait ouvert de manire quasi illimite pour les bahuvrhi (SEST P. 5.4,154), alors qu'il tait maintenu dans des bornes assez serres pour les tatpurusa. Il est donc significatif que Vm., ds le dbut de son expos (5.1,7 et 8), dconseille l'emploi des karmadhraya susceptibles d'tre pris pour des bahuvrhi (et l'inverse): vrapurusa ne saurait signifier " q u i a de valeureux servit e u r s " (il y faut vrapurusaka): cf. aussi la discussion sur caturasrasobhin 5.2,59 et la v. affrente. Quant l'ordre des membres, la tendance chez Vm. est, comme nous l'avons vu, concilier: on attend *adharabimba " lvre (semblable au) bimba" par P. 2.1,56, mais bimbdhara, qui est l'expression usuelle (Vm. 5.2,16), s'explique par rfrence au g. skaprthiva, ce refuge t r o p commun pour maintes difficults, l'int r i e u r mme de la t r a d i t i o n vykaranique. Au su. 23, Vm. se rsigne admettre certains manquements la squence normale des lments du dvandva, manquements qui sont d'ailleurs couverts d'avance par le t r o u b l e mme que j e t t e dans la thorie lapidaire de P. 2.2,34 la multiplicit des vtt. subrogs audit su. Seul kharostrau (Vm. 28, au lieu de ustrakharau) est censur comme intolrable. Plusieurs types de mots ne peuvent e n t r e r en composition, ainsi le type madhupipsu (14) ne ressortit aucune catgorie reconnue, mais peut s'expliquer par P. 2.1,24 vtA {gana gami-gmi). Dhnyasastha (19) a l'air incorrect cause de P. 2.2,10, comme aussi pattraptiman et analogues (2) cause de P. 11 : Vm. maintient la seconde p r o h i b i t i o n , lve la premire. Pour arihan (et analogues, 37) brahmavid (et analogues, 38), qui pareillement contreviennent, quoique de faon diverse, P. 3.2,87, le premier groupe est tax d'incorrect, le second peut tre ramnag si l'on appelle au secours P. 2.2,8+3.2,76. Un compos d'apparence malais expliquer, tel que mlalola (18) pourra s'intgrer dans un gana : en l'occurrence, sous la r u brique commode des mayravyamsakdi. En revanche, on ne sait t r o p sur quelle base Vm. 21 i n t e r d i t les bahuvrhi dont les membres sont en relation discordante entre eux (les vyadhikarana bahu), interdiction qui n'est leve que pour les seuls janman et analogues: le vt. 10 ad P. 2.2,24 parat bien rejeter de tels bahuvrhi d'une manire gnrale, mais le rejet est ensuite jug superflu (vt.18), du fait que l'usage d'ores et dj ignore dtelles formations {anabhidhnt). O n voit que Vm. a estim utile de prohiber ce qui,

33 Ceci montre que le Nysa juge des faits en grammairien, incorporant la rigueur des formes de type pique

et puranique; plus puriste, Bham. se tient aux normes du kvya, qui exclut soigneusement ces mmes formes.

489 Grammaire et potique en sanskrit

au niveau de langue du Bhsya, passait encore pour insolite.34 9. Revenant aux enseignements de potique proprement dite, on observera qu'un groupe massif de traits linguistiques est mentionn propos de la Figure qui vient gnralement en tte de l'expos sur les Alamkra, savoir Vupam ou Comparaison.35 Alors que les autres Figures, qui toutes ont un lien de dpendance implicite par rapport Vupam, sont dcrites sans qu'interviennent des critres morphologiques, Vupam au contraire est mise d'emble en connexion avec la structure grammaticale. Ainsi le KP., s. 127, enseigne que, sous sa forme " pleine" (prn), Vupam se prsente, soit en phrase (analytique), soit en compos nominal, soit dans un driv (secondaire). On retrouve ici le classement ternaire qui prvaut en Potique, alors qu'il est absent en Grammaire. On retrouve aussi le compos parmi les facteurs ou indices de Vupam. La Comparaison de phrase (vkyag) est celle qui est introduite par les particules iva, yath, v et analogues.36 La Comparaison par driv (taddhitag) s'exprime au moyen du suffixe adverbial -vat. Ces deux premiers types d'upam sont tantt direct ou perceptible (sraut), tantt indirect ou reposant-sur-le-sens (rth). Subdivision d'apparence purement " potique", voire philosophique, mais qui ne laisse pas, au moins quant k-vat, d'utiliser une donne pninenne. En effet, P. 5.1,116 avait pos le suffixe -vat au sens du Locatif (math urvat "comme Mathur") : ou du Gnitif (devadattavat "comme [celui] de Devadatta"): -vat, observent les poticiens, quivaut ici va et dclanche comme iva une Comparaison directe: le pouvoir-inhrent (smarthya) cet lment engendre l'ide d'une relation, par le seul fait qu'on le " peroit" dans la phrase, tout comme les dsinences du Gnitif engendrent l'ide d'une connexion ou sambandha, de par

D'autres cas sont mentionns, touchant l'expression du fminin au membre antrieur (73), touchant aussi le samsnta (60 dj cit, 66 et suiv.), laforme du membre ultrieur (84), la phontique jonctionnelle (90-92), le genre dans un dvondva (27), enfin la relation du pronom avec un membre subordonn de compos (5.1,11), question qu'on peut retenir de prfrence aux autres, ne serait-ce que parce qu'elle n'est pas traite (et ne pouvait gure l'tre) en grammaire (description des faits chez Wackernagel 2.1, p. 32). 35 L'upam pleine consiste en quatre lments; l'upamna ou objet auquel on compare qqch., ainsi le visage, Vupameya ou sujet de la comparaison, ainsi la lune, le terme "commun " (tertium), appel smnya, sdhrana ou dhorma, enfin la particule comparative (upampratipdoka ou dyotaka, upam tout bref, vadi). Il est notable que P. mentionne

34

plusieurs fois des faits de langue porte comparative, ainsi quand il dit que la particule cid (en vdique) est upomdrthe 8.2,101 ; de faon plus importante, quand il enseigne que les upamna se composent avec le tertium (appel samanyavoca na); au s. suivant figure le mot upamita = upameya. Cf. encore 2.2,24 vt. 12 cit plus loin. 36 Sur va comparatif, cf. BhtlingkRoth n 3: emploi des lexiques, de l'pope, du kvya, peut-tre par influence du va moyen indien (et, ventuellement, rgvdique?) = /Va, rfrences Debrunner Ai. Gr. 1 2 , p. 27 bas. Quant iva " pour ainsi dire", il est distingu de va "comme" KP. p. 784: le premier a sa place dans la Figure appele utpreks, qui comporte une "vision" directe, alors que iva "comme", ou yath, mme sens, suggre un tertium qui serait inappropri dans \'utpreks.

490 Louis Renou

leur seule prsence.37 Au contraire, la Comparaison indirecte est celle qui s'exprime, soit par le mot tulya ou un mot analogue, soit par un autre emploi du suffixe -vot, celui que dcrit P. 5.1,115. D'abord le cas de tulya : ici la proprit engendre par ce mot n'est pas directement " perue", elle est indirectement sensible, travers l'ide de similitude. Une telle upam affecte trois varits, " l e visage (de telle personne) est pareil (la lune)", o la similitude est par rapport l'objet compar; "ce (lotus, l-bas) est pareil (ce visage, ici)", o elle a lieu par rapport lachse quoi l'on compare; "ce (visage) et ce (lotus) sont pareils", o l'une et l'autre relations se prsentent la fois. Quant au cas du suffixe -vat, exemple brhmanavat " ( i l se conduit) comme un brahmane", la proprit commune ou tertium comparationis est une action verbale, la similitude est le propre de l'action (sous-entendue), d'o l'on " infre" la similitude des personnes agissantes.38 Le troisime et dernier type de Comparaison est celui qui se fonde sur un compos nominal (samsag). On s'attendrait ici rencontrer la vaste catgorie des tatpurusa porteurs d'une valeur comparative (P. 2.1,55 dj cit), mais c'est l une division de l'upamd " e l l i p t i q u e " , puisque le signe extrieur de la comparaison y fait dfaut. La Comparaison " pleine" est concerne, sous la forme directe, par l'emploi de tulya et termes analogues, placs au second membre du compos39; sous laforme directe, les poticiens font tat, d'une manire qui peut lgitimement surprendre, d'un type d'expression o la particule va forme compos avec le nom qui la prcde. Cet enseignement remonte, lui aussi, la Grammaire, P. 2.1,4 vt. 2: le compos ainsi form rvle sa singularit par la manire mme dont il est dfini : on nous dit en effet que ladsinence du membre antrieur y est maintenue, et que l'accent reste celui du nom isol.40
Au lieu de smarthya, Jagannatha (d'aprs J, by V. A. Ramaswami Sastri, p. 166) parle d'une laksan: mais en ce cas, la distinction entre upam expresse et upam implicite risque de perdre sa raison d'tre. 38 Le mot dcisif est kr/yd "action (verbale)", qui figure dj dans le Bhsyaad loc. Sah. Darp. 10.15-16, n 649-650 suit de prs KP. et explique ainsi la diffrence entre srauti et rthl (upam): les mots tulya et analogues sont "puiss" (visrmyant) aprs avoir servi de qualifiants Vupamna, ils sont ainsi incapables de rendre une notion supplmentaire, c'est--dire la similitude entre deux objets ayant une proprit commune. 39 Voire, d'un compos ncessaire (nitya), c'est--dire non rsoluble, prcise KP. (repris Hem. su. 114), alors que le vt. 2.1,4, n 2 ne disait rien de pareil. Y aurait-il l une doctrine non pninenne? C'est peu probable, vu que les textes de potique obissent en gnral au systme de P. 40 Le Bh.ad P. 2.2,18 in fin., qui en37

globe peut-tre un vt., ne fait que reprendre 2.1,4 vt. 2. Les commentaires de KP. ad loe, p. 557 se rfrent en outre P. 2.4,71. II s'agit apparemment d'une survivance vdique: tous les padaptha (cf. Whitney ad Ath. Prt. 1.82) combinent va avec le mot qui prcde, udadheh-iva remplace ainsi udadher va (qui devrait tre dcoup udadheh/iva); cf. aussi Weber ISt. 13, p. 5 sqq. De l l'habitude de considrer va, tantt comme mot indpendant tantt comme membre ultrieur. Noter aussi que la prsence de va affectait la finale des pragrhya RkPrt. 2.55. Tout ceci reflte une situation d'enclise archaque, mais pourquoi maintenue pour le seul val En Potique, il semble qu'on ait arbitrairement va compositionnel et va autonome, celui-ci plutt quand le lien avec le nom est plus lche, celuil quand va interrompt par exemple une squence au cas oblique forme d'une pithte, de va, du substantif affrent.

491 Grammaire et potique en sanskrit

10. A ct de la Comparaison pleine, la Comparaison " e l l i p t i q u e " . 4 1 Les sous-groupes sont peu prs les mmes. D u point de vue grammatical, q u i seul ici nous r e q u i e r t , nous avons citer l'ellipse du t e r t i u m comparationis : processus qui entrane une upomo indirecte, soit avec composition nominale (ex. omrtopom " s e m b l a b l e l ' a m b r o i s i e " ) , soit avec drivation secondaire (ex. visakolpa "analogue du poison " ) . L'ellipse du t e r m e auquel on compare l'objet donne lieu, c o m m e il est naturel, un compos: c'est la catgorie normale. Plus instructive (encore qu'troitement fonde sur le vykarona, comme toute cette doctrine de \'upomo) est l'ellipse de la particule comparative, laquelle aboutit l'emploi de plusieurs formations verbales relativement singulires (fort connues d'ailleurs dans le kvya): le dnominatif en " kyac" ,42 dans lequel le thme de base a valeur d'objet-transitif, ex. sutlyoti " il traite (quelqu'un) comme son fils", ceci suivant P. 3.1,10 qui pose dans son su. le terme UPAMNT; dans lequel aussi le thme de base a valeur de Locatif (dhra = adhikarana du vt. 1 ad P., mme su.), antahpuryati " il se meut comme dans le gynce". Le dnominatif en "kyo", suivant P. 3.1,11, ex. naryate " i l agit comme une femme" (P. reconduit ici, au tmoignage des commentaires, le mot UPAMNT).*3 L'absolutif en -am, appel "namul", o le thme de base est soit un objet-transitif (P. 3.4,25 et 29), type gharmmiuarsam "comme on voit le soleil ", soit un sujet de phrase, prthosamcram "comme Arjunase meut" (P. 3.4,43); l'emploi "comparatif" est assur par la mention UPAMNE chez P. 45. Enfin le compos nominal, ex. kminlgondapndu "ple comme la joue de l'amante".44 Suivent des cas o l'on reconnat une double ellipse, ellipse du tertium et ellipse de la particule comparative: c'est la formation dite en " kvip", ex. vidhavati (sur vidhu " lune"), " il se comporte comme la lune" : formation qui repose, sinon expressment sur P., du moins sur un largissement de P. 3.1,11 d au Bhsya, lequel mentionne l'ventualit d'un dnominatif sans affixe aprs une racine quelconque et cite asvati, gardabhati, o la nuance comparative (" il se comporte comme un cheval, comme un ne")
41 La notion d'ellipse en Potique est fuyante et multiforme, comme d'ailleurs, bien qu'avec des aspects trs diffrents, en Grammaire. Cf. Vm. 5.1,14 qui fait dpendre l'ellipse de l'usage. On a vu 12 des cas de pronom ellipse. Parmi les Dfauts de phrase, il en est un qui rside en l'absence de certains mots attendus (nynapada), cf. la strophe cite KP. s. 75, p. 339, Regnaud p. 163, o manquent asmbhih et ittham. Mais rien de tout cela ne ressemble la thorie du opa qui domine plus d'un dveloppement en Grammaire. 42 Sah. Darp. 10.19 = n 653, d. Kane5, p. 96, conteste que "kyac" soit un cas d'ellipse de la particule. II y voit une ellipse du dharma, car les affixes verbaux, dit-il, expriment d'eux-mmes la ressemblance: allu-

sion implicite la Grammaire! Hem. su. 115 et autres, suivent KP. 43 Le cas de "kyac" et de " kya" est cit pour la premire fois par Udbhata 1.36. 44 Sah. Darp. loc. cit. cite encore Yupam reposant sur le suffixe-/n-, autrement dit lan/n/gat(upama), qui s'appuie sur P. 3.2,79; c'est un cas d'ellipse de la particule, comme celui de la kangat (upam), qui remonte P. 5.3,98. Dans la chagat (upam), qui reflte le cas trait par P. 5.3,106, suffixe -i/o-, il y aurait ellipse de l'upamna. Ce dernier procd est trait tout au long dans Kuval. 1.7 qui allgue l'exemple bien connu en Grammaire, kkatliya "comme (les choses se passent pour) la corneille et le fruit du palmier",

492 Louis Renou

rsulte du su. mme. 45 En second lieu, dans des conditions analogues, apparat nouveau le procd compositionnel, ainsi rjokujara " un roi semblable un l p h a n t " , qui correspond la formation purusavyghra bien connue en grammaire, P. 2.1,56, qui fait tat, sinon d'une ellipse double, du moins de l'ellipse (APRAYOGE) du t e r t i u m . Cette mme ellipse du t e r t i u m , combine celle de l'objet auquel on compare, aboutit un procd ncessairement compositionnel, ex. kusumasadrsa "semblable lafleur " . L'ellipse de la particule et de l'objet compar donne un cas d'affixe " k y a c " , ex. sahasrayudhyati " il se comporte comme Sahasryudha". 46 Enfin les thoriciens les plus avancs croient discernerc'est le cas e x t r m e , et d'authenticit bien contestable, il faut l'avouer une ellipse des trois lments cits. Ceci donne un compos tel que mrganayan " (une fille) aux yeux de gazelle " , compos que KP. (y.) explique en fonction de P. 2.2,24 vt.12: cet endroit de la Grammaire il est fait tat d'un bahuvrhi avec chute du membre ultrieur, le membre antrieur consistant en un " u p a m n a " ,

saptamyupamnaprvapadasya bohuvrJhir uttarapadalopas ca, ex. ustramukha = ustramukham va mukham "dont la figure est comme la figure du chameau ". En l'occurrence, mrganayan reposerait sur une proposition telle que " mrganayane va cacae nayane
yasyh" .47

On voit que, sans comporter d'innovations, la thorie de la Comparaison en tant que Figure de style, telle que la dfinit Mammata, prend fond sur des catgories grammaticales, sur des stra ou des vt. de l'cole pninenne; c'est une mise en systme des donnes parses dans le vykarana. Les Traits postrieurs admettent tous, plus ou moins, cette classification complexe et les critres linguistiques qu'elle entrane. Quelques auteurs cependant protestent: laCitrammms(p. 27) fait valoir qu'un tel classement, reposant sur des principes grammaticaux, n'a pas sa place (no vyutpdyatm arhati) dans un Trait de potique; elle ajoute que, au surplus, des divisions de ce genre ne sont pas valables. Cette contestation isole pourrait trouver argument sur les Traits antrieurs KP., qui pour la plupart ignorent les indices de langue: c'est le cas au moins de Dandin, de Bhm., de Vm. Dandin, nous l'avons vu, ne connat qu'une enumeration de caractre lexical, qui par son ampleur prsente un assez vif intrt, mais qui n'est accompagne d'aucune doctrine morphologique. Udbhata se borne distinguer (1. 32 sqq.) Vupam pleine et Vupam elliptique, avec double ou triple ellipse, vingt et une varits en tout, mais sans entrer dans le dtail.48 Rudrata va un peu plus loin (8.4) en
D'aprs Sah. Darp. 10.10 = n. 655, la diffrence entre " kvip" et "kya" est que "kya " laisse une trace dans le verbe ( savoir, un affixe explicite), "kvip" non. 46 Sah. Darp. n 658 voit l une ellipse de l'upomeya. On peut se demander quelle est la diffrence entre sahasryudhly- et nriy- (prcit): c'est que, dans le premier de ces dnominatifs, l'upameya n'est pas le sujet de la phrase (sah . . .), mais l'objet implicite (tmnam), dans le second (qui lui est identique de notre point
45

de vue), il n'y a d'autre ellipse que celle de la particule (remplace par "kya"), Kane trad. p. 102. 47 Cf. Hem. su. 115, Jagann. op. cit., p. 164, Sah. Darp. 10.22. Kane op. cit., p. 104 rappelle que certains auteurs, justement inquiets de cette analyse, rejettent la trilopopam. Cf. aussi S.V. Dixit d. (1959) de KP., p. 214. 48 Vingt-quatre varits dans SKbh., vingt-sept Sah. Darp., trente-deux Kuval.

493 Grammaire et potique en sanskrit

distinguant d'abord entre somosa etpratyoya, comparaison compositionnelle et comparaison derivative, puis en sriant trois types de la premire, selon que l'upamna est compos avec le tertium (17) ou avec l'upameya (21), ou enfin l'upameya avec le tertium (19). Plus explicite est l'Agni Pur. qui (8.5 et 6) distingue deux upom, celle compos, celle hors de la composition ; il ajoute (7 et 8) que la premire est de trois espces suivant que l'lment mis en composition est Vupamdyotaka (c'est--dire la particule), Vupameya, ou les deux lafois. C'est le KP. qui, sur le plan grammatical, dveloppe en somme les virtualits de ce qui figurait l'tat de germe dans la littrature antrieure.49 11. Autre chapitre o apparaissent des traits de langue : celui de la "suggestion " considre comme facteur potique dominant. C'est l'objet du chapitre IV du KP. ; il n'y a rien de correspondant au chapitre V, d'ailleurs court, qui traite de laforme potique o la suggestion n'est qu'un lment secondaire. Rien non plus au chap. VI, qui concerne laforme infrieure, celle o la capacit suggestive fait dfaut. Aprs avoir dcrit le raso et ses environnements, Mammata pose qu'il est susceptible d'tre "suggr" (dyotyate, vyajyate) au moyen de trois indices formels : au moyen de certaines portions de motc'est--dire, pratiquement, soit du thme, soit plus souvent du suffixe (ou dsinence),au moyen de l'arrangement (racan), enfin au moyen des phonmes, su. 61. On reconnat la disposition ternaire, comme ci-dessus 4 et dans des conditions similaires. La question de l'arrangement et des phonmes sera traite par KP. propos des Figures, soit au chap. VIII ; elle ne porte qu'incidemment sur des points de grammaire. Au contraire, la question des " portions de mot" (podaikodesa) intresse des faits morphologiques et lexicologiques combins. Parmi eux, il en est qui ne nous concernent pas directement ici, ainsi lorsqu'on enseigne quejoyoti " il l'emporte" (p. 169) suggre le rasa Erotique, ce que n'et pas obtenu le terme plus simple, apparemment plus banal, sobhote " il est beau " ; on nous affirme p. 170 que poda " pas" suggre l'anxit amoureuse mieux que n'et fait dvra " porte". Mais c'est peine si l'on peut parler de qualification lexicale; c'est en vrit le contexte, l'ambiance de la strophe tout entire, qui amne prfrer tel mot tel autre, sans que le vocable slectionn ait plus de vertu intrinsque que le terme vinc. Cette remarque s'applique pareillement aux cas o la morphologie est en jeu. Telle dsinence, dans une str. donne, passe pour tre mieux apte qu'une autre suggrer quelque rasa (le rosa Erotique, comme dans tous les exemples des thoriciens, a un rle indment privilgi); entendez par l la dsinence, en tant
49 Les critres de langue font dfaut en effet dans les plus anciens Traits, ceux de Dandin, de Bhm., d'Udbhata; aussi chez Ruyyaka, qui est singulirement sommaire touchant V upa ma. Bhm. 2.31 se bornait citer le cas de yath et iva, puis le khysmya (donc, = P. 5.1,115) avec -vat (2.33); -vat est mentionn sommairement par Dandin 2.57 et Udbh. 1.37 (en valeur de karman), 38 (en valeur de Gn. et de Loc). Rien chez

Vam. Mais Dandin expose une remarquable liste de substantifs, d'adjectifs, de verbes, entranant une valeur "comparative", 2.57-65: cela, certes, se limite au lexique, sans entrer dans la doctrine grammaticale, mais c'est la matire premire du kvya, et, sous l'angle de la documentation, c'est plus proche de notre conception moderne de la stylistique que les analyses primes de la Potique sanskrite ultrieure.

494 Louis Renou

qu'elle exprime soit une relation casuelle ou temporelle, soit une personne, soit un n o m b r e ; il s'agit donc du contexte, non de l'aspect morphologique. On enseigne ainsi que le prsent kirati (p. 171, Regnaud p. 87) note un acte en train de se raliser [sdhy amona \ cf. provrttasyvirme dans la dfinition du prsent chez P. 3.2,123 vt.1), que le nom verbal atJta note un acte accompli (siddha: cf. les caractres bien connus de la nistho P. 1.1,26), si bien que le suffixe -ta- " s u g g r e " le caractre rvolu de l'acte dsign par le verbe; " o r , ces deux actes sont dans un rapport logique de cause et d'effet qui se t r o u v e i n t e r v e r t i grammaticalement, puisqu'en vertu des suffixes en question, l'effet aurait prcd la cause. Mais cette e r r e u r apparente a pour rsultat de suggrer la rapidit avec laquelle les blessures de l'Amour sont suivies de leurs consquences, et par cela mme dfaire natre le rasa e r o t i q u e " (Regnaud loc. cit.). De mme (p. 173), dans Hkhann ste bhmim " il se t i e n t grattant la t e r r e " , le participe j o i n t kste (l o l'on attend, parat-il, *likhaty sitah) suggre un acte qui se poursuit jusqu' apaisement de l'esprit (prosada): ceci, j o i n t l'Accus. bhmim (lo l'on attend parat-il, le Loe. bhmau), suggre que l'acte de " g r a t t e r " a lieu sans intention particulire (na hi buddhiprvakam).50 Cette indication sur le participe de continuit descriptive, en locution periphrastique, n'est pas sans valeur du point de vue linguistique; non plus celle du Locatif envisag comme cas " intentionnel " , ce qui voque sans doute la nimittasaptamJ souvent mentionne dans les commentaires grammaticaux. Mais, dans la strophe cite, de tels phnomnes manent d'une interprtation psychologique a poster i o r i , bien p l u t t que des formes grammaticales, qui n'ont aucun contenu expressif par elles-mmes. Les autres traits allgus sont de caractre plus lmentaire encore: si l'Instr. divasena " s u g g r e " l'ide que le f r u i t souhait est atteint (p. 179) " a u t e r m e de la j o u r n e " (et non " a u cours de la j o u r n e " ) , il y a bien l le rappel implicite de I' APAVARGE de P. 2.3,6, expression que justement la Ksikglosait par phalaprpti (comme fait KP. ici mme) : mais l'emploi serait peine digne d'tre relev si l'entourage de la strophe ne donnait quelque impulsion dgager cette nuance. De mme p. 180 (Regnaud p. 91), o le KP. note que le suffixe -ko- dans agoka " membre " suggre l'ide de compassion, conformment P. 5.3.76.51 Rarement la doctrine puise chez Patajali: pourtant p. 178 sur l'ordre des membres dans le dvandva, command par le prdhnya ou " p r i m a u t " relative des lments en prsence: enseignement qui mane de P. 2.2,34 vt. 4. Des donnes de ce genre n'ajoutent pratiquement rien notre connaissance de la grammaire, ni mme n'expliquent l'affectation stylistique dont telle ou telle formation peut bnficier dans le Regnaud dit trs bien: "tracer le sol, sens exprim par une construction qui ne permet pas d'admettre l'ellipse d'un rgime direct destin dsigner ce qu'on s'applique y tracer, indique un acte machinal rsultant du trouble prouv par la personne qui l'accomplit; au lieu que tracer sur le sol supposerait l'ellipse d'un rgime direct l'accusatif, exprimant ce qu'on y tracerait (des caractres, etc.), et impliquant par l
50

un acte accompli avec rflexion". 51 Importance relative des nuances d'affectivit chez P., telles que krose,
lipsym, sisi, prasamsym, non

seulement dans la thorie des modes du verbe (o cet apport affectif est pour ainsi la base de la description), mais ailleurs. La Potique n'a pas manqu de faire son profit de telle de ces notations, ainsi naturellement celle concernant le "manye" ( =
manyase) prahse de P. 1 4,106.

495 Grammaire et potique en sanskrit

kavya.52 Retenons toutefois comme de quelque valeur parce que lachse est inatteste dans la littrature grammaticale la remarque touchant l'emploi de ca r i t r : emploi dfini comme un tat d'quilibre (tulyayogit) entre deux actions simultanes, dont en ralit la premire est la cause et la seconde est l'effet: " e t ton esprit s'est t o u r n vers la fiert et nos ennemis ont t d t r u i t s " , c'est--dire " t o n me a m o n t r de la fiert et cela a suffi pour que nos ennemis fussent d t r u i t s " (KP. p. 181, Regnaud p. 92); comme dit Regnaud, " de cette suggestion rsulte la p r o m p t i t u d e avec laquelle le roi a ananti ses adversaires; et par l apparat un sentiment (bhva) en rapport avec le rasa h r o q u e " . Le cas est dj signal dans le Dhvany. loc. cit. Il rentre, de notre point de vue, dans les voies et moyens de l'expressivit linguistique. 53 Le caractre "suggestif" est attribu, entre autres, au suffixe d'abstrait -iman-, avecl'exemple taruniman " j e u n e s s e " (Regnaud p. 93). P. 5.1,122 fait dpendre d'un gana l'apparition du suffixe d'abstrait -iman-, autrement dit il s'agissait d'une formation relativement rare, s u r t o u t si on la compare aux autres drivs abstraits tels que ceux en -tva- et -ta-, voire en -y-. Mais ceci n'aurait pas suffi nous p e r m e t t r e de penser que -iman- comportait une nuance qualitative, si nous n'avions le tmoignage corroborant de Bhm. 6.54, qui recommande en posie l'usage de ce suffixe, alors que Vm. 5.2,56 carte certains emplois, peut-tre simplement parce qu'on en abusait. Le fait que le suffixe -iman- (en tant que suffixe secondaire) ait t rattach la srie des noms en -(/")yas- et -isthaet certains dnominatifs en -ayati a contribu accentuer le caractre concret de cette petite catgorie, qui l'origine tait peine distincte de la classe des noms en -man- et dpourvue de valeur affective; -Jyas- lui aussi est dclar particulirement souhaitable en posie, Bhm. loc. cit. 12. Le chapitre des Dfauts (chap. VII) est riche en notations grammaticales, souvent plus novatrices, en t o u t cas moins dconcertantes, que celles que nous venons d'examiner. Ici d'abord on r e t r o u v e la question des composs, qui forme vraiment l'pine dorsale de la thorie grammaticale en Potique. Il est vrai que KP. est plus libral en dtails que les autres textes. Certains des dfauts portant sur un mot ne se rencontrent qu'en composition nominale, samsagatam eva KP., s. 72 p. 266 et 284. Observation qui, de notre sentiment, est sans porte apprciable, car il est clair que les dfauts sont les mmes, qu'il s'agisse d'un mot simple ou d'un compos (le compos, sur le plan smantique, est un m o t comme les autres), ainsi le dfaut appel klista "(expression) difficile ou o b s c u r e " . T o u t ce qu'on peut dire est qu'un t e r m e mis en compos rend d'emble l'expression plus difficile, moins tolerable parfois l'ambigut: d'o cette catgorie spciale. Mais il y a un dfaut au moins qui est spcifique la composition nominale, c'est Yavimrstaviheymsa, p. 266 et 285: tel lment

II n'y a gure tirer non plus des remarques de l'Aucityavic. 20 sqq. concernant la "proprit" ou "convenance" (aucitya) de divers cas, genres, nombres, celle des adjectifs, prverbes, particules, temps verbaux.

52

On est l dans un domaine trop fluent. 53 Sur l'emploi de ca ritr, cf. J. Gonda Vk 5, p. 37 et notamment p. 41.

496 Louis Renou

de phrase qui devrait figurer comme prdicat (yidheya)54 demeure l'tat non-autonome ou subordonn (gunbhta), c'est--dire sous la forme d'un membre de compos. Ainsi mithymahiman " f a u x prestige" (cit aussi Regnaud p. 146), lo le mouvement de la phrase exigeait mithy mahim " m o n prestige (s'avra) i n a n e " ; dvityamaurvJ " d e u x i m e corde d ' a r c " (aussi Regnaud, ibid.), expression qui passe inaperue, alors qu'il fallait dire dvityam maurvim " (sa ceinture tait comme) une corde d'arc (qui et t paradoxalement) la deuxime (porte par A m o u r , lequel n'en a qu'une d ' o r d i n a i r e ) " ; olaksyajanmot, o et convenu alaksit janih "sa naissance est inconnue" ; omukt " non dlaisse" (Regnaud, ibid.), quand on attendait na mukt " e l l e n'a pas t dlaisse". 5 5 Il s'agit bien l d'un fait de style, mais fondement linguist i q u e : la tendance " resserrante" de la strophe, l'abus du compos nominal, le mouvement inluctable qui conduit l'expression en kvya revtir l'allure d'un stro versifi, t o u t cela aboutit masquer le prdicat (comme est masqu ou mme ellipse, bien souvent, le sujet mme de la phrase); il important au thoricien d ' a t t i r e r l'attention l-dessus, mme si les exemples allgus sont ingalement propres assurer la dmonstration. 13. Le mme dfaut apparat un peu plus loin (su. 74 p. 306) l'occasion, non plus du compos, mais de la phrase. Il s'agit ici encore d'un prdicat qui n'est pas mis en vidence, mais est laiss dans la pnombre, voire, compltement omis. Les exemples cits, qui sont relativement nombreux il est visible que Mammata a entendu donner quelque ampleur une thse qui n'avait t f o r m u le avant lui ni en Potique ni en G r a m m a i r e , concernent la phrase corrlative, cette structure majeure de la phrase sanskrite. Le KP. considre comme un dfaut l'absence de pronom corrlatif
54 Vidheya en ce sens n'appartient pas la Grammaire (dans le Bhsya, le mot ne signifie rien de plus que "ce qui fait l'objet d'une rgle, ce qui est instruire sous forme de vidhi"). En Potique, le terme s'oppose onuvdya (non attest en Grammaire), " ce qui exprime une rfrence quelque chose qui a t antrieurement mis". Le contraste vidhij anuvda se trouve dans les Nyyastra 2.1,62, mais il appert que cet emploi o'anuvd(y)a relve de la Mmms, cf. MT. Kosa, s.u.; anuvdya, glos uddesya, prptenvagata, quivaut sur le plan grammatical "sujet (de la phrase, en tant que dsignant qqch. de connu)", d'o vidheya "prdicat (en tant que la chose faire connatre)". Le Tantravrtt. (cit dans le c. moderne de KP., p. 285) parle des phrases corrlatives, o l'ondyat ( = anuvdyat) caractrise la proposition en yad, comme se situant en protase et dsignant la chose "acquise", par rapport la vidheyot, qui caractrise la proposition en tad, place en apodse et

dsignant la chose " raliser", yacchabdayogah prthamyam siddhatvarn copy andyat/ tocchobdayoga auttaryam sdhyatvam ca vidheyot. 55 Sah. Darp. n. 574 reprend ces exemples et souligne le cas de amuktj na mukt, qui illustre la fameuse thorie sur les deux espces de ngation, le prasajya = et le paryudsapratisedha. Un peu plus loin (p. 304), KP. cite l'exemple (aussi repris Sah. Darp.) svargagrmatikvilunthanavrthocchnaih kirn ebhir bhujaih " quoi bon mes bras qui se sont en vain gonfls des dpouilles du hameau cleste! " (on attendait: c'est en vain que . . .). Le rapport entre vidheya et anuvdya est galement en jeu dans lastr. nyakkrah KP. s. 74, p. 304 (repris Sah. Darp.), mais il s'agit ici d'un "ordretroubl" (le vykirna de SKbh. 1.23): on attend ayam eva nyakkrah, avec \'anuvdya (ayam) avant le vidheya (nyak). Ceci concerne d'ailleurs le style bien plutt que la syntaxe, comme le Dfaut similaire appel samkirna
KP. su. 75, p. 362.

497 Grammaire et potique en sanskrit

dans la squence attendue yah . . ./sah . . . : il d i t que le pronom asau (qu'on a dans un lment strophique tanoti yo 'sau subhage tavgatah) ne saurait remplacer sah, qu'il faudrait donc yah . . ./ sa gatah, cf. Regnaud p. 148. Car asau, est-il prcis, sert simplement dsigner un objet en tant que la chose laquelle on se rfre (anuvdyamtra), il n'a pas, dirions-nous, force corrlative. 5 6 Si la structure fondamentale yah . . . / sah . . . requiert la stricte corrlation, la structure inverse (moins bien accrdite o u , du moins plus faible en tant que systme corrlatif) n'exige pas aussi fermement le dmonstratif: on peut dire sdhu krtam . . . yad . . . " c'est une chose bien faite, ( savoir) que . . .", au lieu de tat sdhu krtam . . . yad . . . (Regnaud p. 150). Pourquoi ? C'est que le relatif a pouvoir-inhrent (smarthya) (d'assurer lui seul la valeur corrlative). O n constate en effet aisment que ya postrieur est plus f o r t que ya antrieur, en ce qu'il entrane souvent avec lui une valeur compltive, loin de couvrir t i t r e purement f o r m e l , comme ya antrieur, un simple " a d j e c t i f s y n t a x i q u e " (suivant l'expression de M. Benveniste). Cette notion de pouvoir-inhrent est, il est vrai, dangereuse manier; c'est une de ces abstractions que poticiens et grammairiens aiment voquer. Mammata observe qu'en vertu du smarthya il arrive que ya ou ta puissent manquer; on cite un exemple o l'un et l'autre pronoms font dfaut, utpatsyate/ 'sti marna ko 'pi samnadharm = ya utpatsyate/ tarn praty asti. . . (Regnaud, ibid.). On rentre ici dans une autre abstraction commode, celle de l'ellipse. Une double objection se prsente quant la ncessit de la rigoureuse liaison yah . . ./ sah . . .: (a) ne peut-on supposer que (dans la phrase dj cite) asau joue le rle de sah, de mme qu'on trouve ailleurs yah . . . / asya . . ., l o l'on comptait sur yah . . ./ tasya . . . ? Rponse : non, car alors, dans un cas tel que yo 'sau . . . sah . . ., le pronom sah serait superflu, ce qui n'est nullement avr; ensuite, si asau tait corrlatif, il viendrait en tte de la seconde proposition ; donc yo 'sau agit comme sa yah, pour indiquer la personne en tant qu' " inconnue". On notera incidemment que l'ordre des mots, pourtant si arbitraire dans le kvya, est une exigence de la Potique; sans doute y a-t-il l le souvenir d'un tat de langue plus rigoureux, tel celui de la prose vdique, o peu de libert tait laiss aux auteurs de changer la place d'un mot. (b) si l'on escompte une corrlation stricte, comment se fait-il qu'on ait telle strophe o tad (seul) rpond hyad-yadl Rponse: tad renvoie une entit unique, qui quivaut justement cette unicit qu'exprime (d'autre manire) le complexe yad-yad.51 Ces indications ont leur prix, mme si on les applique (comme c'a t srement l'intention de Mammata) la structure un peu relche qui est celle du sanskrit classique, ft-ce mme l'intrieur du kvya. 14. Dans ce mme cadre, la question de la phrase relative revient au su. 75 p. 347, afin d'illustrer le Dfaut appel abhavanmatayoga = abhavannistasambandha (Regnaud p. 169): la relation
56 A ce propos, KP. enumere judicieusement les nuances du tad non corrlatif: dsigner l'objet dont traite le contexte, prakrta; l'objet connu, prasiddha; l'objet dont le sujet parlant a l'exprience personnelle, anubhta. 57

De mme un peu plus loin (p. 349) yad.../ tadanm, au lieu eyadanm .../ t ou de yad . . .tad. Pour viter cette rupture d'uniformit, Mammata prfre corriger yad en ced. L'exemple yad-yadjtad est repris Sah. Darp., n 574.

498 Louis Renou

qu'on souhaite (entre relative et principale) n'a pas lieu. Dans l'exemple cit, il s'agit de dmons (ksapcrinm) par la puissance desquels (yesm). . . (et) par qui (yaih). . ., enfin dont (yesm)... : l'antcdent est au Gnitif, ce q u i , selon Mammata, ne s'accorde pas avec l'Instr. yaih, command par la syntaxe de l'apodose. \\ n'y a rien l que de rgulier notre sens; t o u t au plus estimerat-on que cette m i x t u r e de pronoms relatifs entrane de prime abord quelque obscurit. Toutefois le KP. n'hsite pas taxer de dfectueux le rapport ksapcrinm/yaih, ce q u i , si l'on gnralisait la remarque, conduirait rejeter nombre de structures prsentant une d isparit analogue, ne gardant comme correctes que celles o, par une sorte d'attraction (ici mal concevable), l'antcdent a t appari au cas grammatical du relatif. Chose curieuse, l'object i o n faite cette phrase prend appui, non sur un grammairien (la Grammaire, on le sait, se soucie f o r t peu des structures), mais sur Jaimini : la Mmms, t o u t bien considr, n'est-elle pas la vraie " s y n t a x e " du sanskrit, une vkyamlmms ou disquisition sur les propositions, t o u t comme la grammaire est une padammms, une disquisition sur les mots. 58 De fait, Jaimini 3.1,22 donne l'axiome gunnm ca parrthatvd asambandhah samatvt syt " i l ne peut y avoir de relation (entre des rites secondaires), tant donn que (lesdits rites) servent quelque chose d'autre ( = n'ont pas leur fin en eux-mmes) et qu'ils sont gaux (entre eux cet gard par rapport l'acte p r i m a i r e ) " . Transpos en langage grammatical, ceci signifie: quand il y a plusieurs subordonnants (yac-chabdanirdesya), ils servent quelque chose d'autre et ne sauraient tre en corrlation m u tuelle. Conclusion pratique, et d'ailleurs irralisable: ct de yesm . . . ksapcrinm, \ faudrait un yaih . . . ksapcribhih (d'autant plus que yaih est en corrlation avec taih au 4 e pda). Un autre exemple, assez diffrent, du mme Dfaut est encore donn p. 35O(Regnaud p. 170): la strophe cite commence par " c o u t e (le rcit de) t o u t ce qui (yad-yad) a t (atteint, obtenu ou conquis, samsditam) par telle et telle (chose respective, tena-tena): ainsi les flches (Nomin.) (furent obtenues) l'aide de l'arc (Instr.), par les flches (fut conquise) la tte de l'ennemi, etc.". La phrase parat irrprochable. Mammata t r o u v e y redire et pose une alternative: ou bien les mots "flches, arc, tte, etc." sont les rgimes de " c o u t e " et doivent tre pareillement l'Accus.; ou bien l'on admet que " c o u t e " n'a pas de mot particulier pour rgime, mais seulement le sens-global ou sens-de-laphrase, vkyrtha, et alors on attend une proposition autonome commenant avec " les flches" et la mise au N o m i n . de tous ces mots. Double solution qui nous parat spcieuse, sinon absurde, mais qui recle t o u t de mme un germe de ralit. Le poticien prconise la correction de yena yena . . . yad yad en kena kena . . . kirn kirn.59 Du moins, voit-on par des dtails de ce genre, par
58 Cf. nos tudes 6, p. 66. Les poribhs mmmsistes ont affaire au vkya (lequel fait partie, soit dit en passant, des six critres ou promano). Elles ont aussi une incidence grammaticale, cf. JAs. 1941-42, p. 121. [this volume, page 442]Sur l'influence de la Mmms sur la Potique, Kane, op. cit., p. 375. 59 " Flches " , etc. ne saurait ni tre

appos yo, ni tre inclus dans le sens du relatif (KP.). Exemple accessoire concernant la rupture d'uniformit dans le pronom: todvisrsto est corriger en oneno vi, la rfrence tant Siva, qui vient d'tre nomm: ceci confirme que tod est senti comme pronom de rfrence lointaine ou mme abstraite.

499 Grammaire et potique en sanskrit

l'excs mme du purisme syntaxique, l'intrt que prennent certains thoriciens lathorie de la phrase, thorie dont l'absence, constitue une grave lacune, la seule grave peut-tre, dans l'enseignement des grammairiens.60 Ainsi, sans aborder de front les problmes de langue, la Potique sanskrite ouvre tout de mme quelques perspectives nouvelles, du moins chez ses reprsentants les plus pntrants, Anandavardhana et surtout Mammata. On peut observer encore ceci : l o la Grammaire prescrit ou interdit ne connaissant en guise d'"options" (vikalpa) que des faits de pure indiffrence linguistique (par ex. il est indiffrent qu 'on forme sisvya ou susva, selon P. 6.1,30) , la Potique, intervenant qualitativement, indique certains choix, certaines prfrences, notamment dans le domaine de la Drivation nominale, qui de tous les compartiments du vykarana, a toujours t le plus voluant, le plus mobile. Elle connat des nuances expressives, ventuellement affectives, mme s'il lui advient trop frquemment d'attacher la vertu de "suggestion" tel ou tel lment grammatical qui, pris en soi, est parfaitement inerte. En dpit du caractre de bonne heure strotyp qu'ont revtu les enseignements de potique, on y dcouvre encore la recherche d'une certaine hirarchie dans les valeurs du langage, recherche qui demeure trs loigne des proccupations de la grammaire traditionnelle.
60

Les Dfauts portant sur un mot isol ne nous concernent pas ici, tant affaire de lexique, sauf le n 2 (su. 73, p. 268) dnomm cyutasamskrti, glos vykaranalaksanahino, "chappant aux rgles de grammaire", qui rpond au sabdocyuta de Bhar. trad. Man. Ghosh 17.94, au sabdahina de SKbh. 1.111 dont on nous dit qu'il peut se muer en Qualit s'il s'agit d'une imitation, anukorana, c'est--dire si l'on veut imiter le parler dfectueux d'un homme sans culture: tel pourrait tre le cas, souvent cit, de la phrase pasyoisa ca gav ti KP. su. 80, p. 412 (l'exemple gav iti est d'origine grammaticale, il est signal Ks. ad 1.1,16, justement comme cas d 1 " imitation", cf. la trad. Ojihara-Renou p. 63; sur le principe de l'anukarana en Grammaire, v. quelques rfrences dans Terminol. s.u. et Durgh. Introd., p. 134). Ainsi anunthate contrevient P. 1.3,21 vt.7, vu que la str. o cette forme figure n'implique pas de nuance prcative (sisi); moins pusillanime, Durgh. ad P. 1.3,21 parvient justifier la

forme; cf. de mme jaghne Sah. Darp. 7.4 SKbh. 1.20 Durgh. ad P. 1.3,28. II semble qu'en Potique l'incorrection grammaticale, le solcisme, n'appartienne jamais qu'au mot, non la phrase, car on nous dit (KP. p. 296) que le Dfaut de phrase rpondant la cyutasamskrti n'existe pas: ceci montre bien que la puret grammaticale tait affaire de mot, non de phrase. Tel Dfaut concernant une " portion de mot " (cf. ci-dessus 11) effleure un problme grammatical, ainsi le n 3 (nirorthaka) s. 74, p. 321 ; pourquoi le plur. drsm "des yeux ", s'agissant d'une seule femme? Ou encore le moyen kurute, l o manque toute nuance rtroflexe? A propos de drsm, la v. note assez finement que le pluriel est injustifiable, car dans une notion unitaire I n'est admissible que s'il y a diversit de fonctions (vypra), or ce n'est pas le cas quand il s'agit des "yeux". Inversement un singulier "gnrique" est plausible Vm. 5.1,17 (avec recours de la v. la notion de "genre" d'aprs la Mmms).

D. La Thorie des temps du verbe d'aprs les grammairiens sanskrits (I960)

Louis Renou

The last article is Renou's " La Thorie des temps du verbe d'aprs les grammairiens sanskrits," Journal asiatique (248,1960, 305-337). The related treatment of verbal moods is the subject of a more recent monograph by Rocher (1968a). Many of the rules discussed here occur on Plate III. 1. Rien1 n'est plus simple, en apparence au moins plus lmentaire, que lafaon dont Pnini dcrit les valeurs temporelles dans le verbe. En fait, une lecture plus attentive, un recours aux commentaires, aux grammairiens non-pninens, permettent de constater que, sur beaucoup de points, l'enseignement tait prcis et souvent mme pntrant. L'indice qui dsigne les formations temporelles ou, plus exactement, les dsinences affrentes est un indice L initial. 2 Les signes conventionnels et T marquent respectivement les temps secondaires imparfait LAN, aoriste LU, conditionnel LR3 et les temps primaires (ou mieux: non-secondaires) prsent LAT, parfait UT* futur priphrastique LUT, enfin futur (simple) LRT. A cette srie s'ajoutent les indices modaux : c'est l'optatif (avec l'indice des temps secondaires) LIN,5 puis le subjonctif

Abrviations: Bh. = Mahabhasya; BM. = Blamanoram; C(V). = Candra(vrtti); H. = Hemacandra (avec la Brhadvrtti); J. = Jainendra (avec la Mahvrtti); K. = Ksikvrtti; Kai. = Kaiyata; Kt. = Ktantra(avec le c. de Durgasimha); N. = Nysa; Ng. = Ngesa; P. = Pnini; SK. = Siddhntakaumudl; SK. = Sabdakaustubha; SK. = Sarasvatkanthbharana; s. = stra's; vt(t). vrttika('s). Nos investigations dans la littrature grammaticale ne sont pas compltes; il est peu probable que des faits importants se laissent dceler chez les postPninens; nous n'avons rien trouv de remarquable chez Skatyana, ni dans la Bhsvrtti ou la Prakriykaumud. Vopadeva a ceci de curieux qu'il supprime les rgles lmentaires, ne gardant que quelques curiosits; de mme le Srasvata, qui se limite
lumhi lunlhi ( = P . 3.4,2), yajati smo ( = 3.2,123), supto 'harn vilalpa ( =

3.2,115 v t ) , yvot karoti ( = 3.3,4), smorasi (avec yod, = 3.2,113). Il n'entrait ni dans nos vues ni dans nos moyens de faire intervenir les philosophes de la grammaire; la Sabdasaktipraksik a tout un prakarana sur les valeurs "la" du verbe qui transcende pour ainsi dire les donnes grammaticales tout en conservant les articulations de P. Au contraire, dans le Vakyapadya (3 me Livre, kla-

samuddesa), on a une exposition abstraite sur le temps grammatical, o les faits de langue sont peine dgags de la spculation. 2 Emanant de kla "temps," comme l'a not Ksh. Ch. Chatterji Techn. Terms 1 p. 9. Le klovisesa ou "spcification du temps " est l'une des deux raisons d'tre du " a(kra)," l'autre tant l'arthovisesa ou spcification du sens K. ad 3.4,77. Les autres indices en L sont trangers la thorie temps. 3 Ce sont les trois temps augment (bhtakaranavati) que Kt. 3.1,14 appelle rdha "transmis par la tradition," pensant aux noms qui les dsignent dans cette Grammaire. * Le parfait n'est pas, proprement parler, un temps primaire, mais il se comporte comme tel, tant pour les formes que pour l'emploi (cf. Brugmann 22. 3 p. 769), et tout d'abord en ce qu'il est dmuni d'augment; ni pour le vd. ni pour le "class.", P. ne reconnat un plus-que-parfait, cf. le silence ce sujet de Thieme Plusquamperfektum. 5 Le soi-disant prcatif n'est qu'une variante de l'optatif, du point de vue smantique aussi bien que formel ; P. le caractrise par UN ASISI 3.4,116, d'o la dnomination sirlin des Pninens, sis seul des non-Pni-

501 Thorie des temps du verbe

LET, enfin l ' i m p r a t i f LOT6 (ces d e u x - c i , avec l'indice p r i m a i r e ) . 7 Un tel ensemble traduit d'abord cette vrit que, dans Torganisation du verbe sanskrit, il n'existe pas de catgorie modale qui serait distincte par nature d'une catgorie temporelle, comme le Veda avait tent d'en constituer une8; dans la prose vdique de type " rcent" (celle mme que vise P.),9 aussi bien qu'en sanskrit ultrieur, l'expression verbale dans la phrase subordonne, c'est-dire dans le domaine par excellence du mode, se rpartit entre l'optatif, le futur (thoriquement aussi, le conditionnel) et le prsent. 2. La thorie des temps chez Pnini s'insre l'intrieur de l'expos sur ladrivation nominale (primaire).10 La chose se comprend de soi si l'on considre que les dsinences verbales, de mme que les suffixes primaires, s'attachent la racine (ou au thme verbal, DHTOH). Le paralllisme entre l'une et l'autre formations se marque dj dans la terminologie: dsinences et suffixes s'appellent pratyaya "ce qui s'attache, ce qui s'affixe qqch." 11 L'expression KARTARl v\se, dans le nom, la fonction " agentis " et, dans le verbe, la voix active (c'est-a-dire, non-passive) : de fait, les noms d'agent sont gloss uniformment par le verbe actif, karotti krt "qui fait", pacatti paco h "qui cuit." L'expression KARMAI)il dsigne, dans le nom, l'Accus. et, dans le verbe ou le nom verbal, la voix passive: le passif ne repose-t-il pas sur le transfert (au Nomin.) d'un objet direct qui, dans la phrase normale, c'est--dire la voix active, s'inscrivait l'Accus. ? Enfin le terme BHVE concerne les drivs nominaux "actionis" et, dans le verbe, il s'applique la catgorie de l'impersonnel.12
6 L'impratif, lui aussi, et une formation primaire, en dpit de dsinences spciales (la finale -u rpondant 1'-/ des temps primaires). Quand au subionctif, qui est mixte, les formes en sont claires l surtout o les dsinences primaires les caractrisent. P. use, l'occasion, d'indices diminus, LR qui englobe LR et LRT; Ll qui associe au parfait (UT) les morphmes d'aoriste dsigns gnriquement par Cil. 7

II est notable que P. n'ait pas senti le besoin de poser la catgorie de l'indicatif (dont les critres modaux sont purement ngatifs); s'il l'avait fait, il l'aurait dfinie sans doute un peu comme, dans le nom, il dfinit le Nomin. 2.3,46 (cf. Thieme JAOS. 76 p. 4), ce que C. 2.1,93 rsume par arthamtre " ce qui note le sens et rien d'autre".
8 Avec un subjonctif bien dfini, un optatif et un " injonctif " , ce dernier, il est vrai mixte entre le temps et le mode. 9 Celle des Br. expurge de ses archasmes, des Stra, des Up. en prose. 10 Chez les non-Pn inens, elle se dgage parfois du chapitre de driva-

tion, ainsi Vop. aun "kti-pdah" (chap. 25) qui traite des valeurs personnelles, abstraction faite des suffixes. 11 Autre interprtation du mot chez les Gramm., v. Termin, s.u. protyayo et JAs. 1941-42 p. 144 [see pages 457-458 of this volume]. Pro,krt est galement commun la morphologie nominale et verbale, ibid. 12 Ainsi podyate 'neneti podah " le pied est ce avec quoi on marche" CV. 1.3,7 = P. 3.3,18. Ceci ne veut pas dire qu'au sentiment des Gramm, les "actionis" reposent ncessairement sur un impersonnel; P. 19 les connat aussi notant un rgime verbal quelconque (autre que le rgime d'agent, cd. le Nomin.), avec cette restriction importante que le domaine dudit s.19 est celui d'une samjn, n. propre ou n. techn. Au dbut de 3.1,67, le Bh. est amen distinguer le bhdva not par un verbe et le bhva not par un suffixe : le premier indiquant temps, personne et voix, tant li un agent, assimilable un acte (kriyvat), le second tant dnu de ces spcifications et assimilable un objet concret (dravyavat).

502 Louis Renou

A quel endroit du troisime Livre (lequel t r a i t e de la drivation primaire) commence l'expos sur les valeurs temporelles? A l'occasion des suffixes valables au sens du " p a s s " (BHUTE P., atte Kt. et ailleurs), l'expression bhte tant complter en dhtvarthe bhte K. ad 3.2,84"quand le sens de la racine est celui du pass". Cette prcision, comme la plupart de celles qu'apporte K., remonte au Bh. : Patajali proposait d'abord de sous-entendre kole " q u a n d le temps est celui du pass", mais son interlocuteur observe qu'il n'y a point de rubrique gouvernante (adhikra) sur le temps chez Pnini. 13 Il propose alors de sous-entendre dhtau, mais l'objection est qu'une " racine v e r b a l e " est un mot et qu'un mot ne saurait tre prsent, pass ni futur. Il propose enfin dhtvarthe "sens de la racine," autrement dit kriyym " q u a n d l'action (verbale) est passe," et cette faon de f o r m u l e r est l'abri de t o u t e objection. 1 4 De telles indications locatives, en Grammaire, ne servent pas ncessairement dfinir un emploi " s m a n t i q u e " . Une spcificat i o n comme KARTARI ou BHVE (nous l'avons rappel) indique une catgorie trs gnrale ("quand il s'agit de . . . " ) . Les mentions BHUTE ou BHAVSIYATI marquent la nuance (ventuellement) passe ou future propre tel ou tel driv nominal, quels que soient les caractres qui, par ailleurs, distinguent telle formation de telle autre. On voit l'importance de ces critres temporels propos des suffixes oundika que P. voque au dbut du troisime pda: ces suffixes sont valables, nous d i t - i l , BAHULAM "diversem e n t , " notion qui peut s'entendre, videmment, sur plus d'un plan, mais que le grammairien asrement entendue aussi sous l'angle t e m p o r e l , puisqu'il pose au s. 2 BHUTE 'PI DRSYANTE " o n les rencontre galement au sens du pass", et au su. 3 BHAVSYATI " a u sens du f u t u r . " Le Bh. n'enseigne rien ce sujet dans son commentaire sur le s.l, mais K. glose vartamne 'rthe " a u sens du p r s e n t " , ce qui est sans porte pratique, mais a l'avantage d'tre logique avec le systme. 15
13 Cf. encore ci-dessus 10. P. 129). 15 vitait ces odhikra qui ne sont pas Toute une tranche de la Drivation strictement grammaticaux, ce qui primaire est ainsi subordonne aux n'empchera pas, plus tard, Kt. 3.1,10 valeurs temporelles: BHUTE vaut de de formuler un "kle" qui se justifie, 3.2,84 122 inclus (mais la K. ne ajoute le c., quand il y a doute si le l'allgue expressment que pour qualifiant appartient au verbe ou 85-87,102,104,105,110-122; cf. en l'objet (dravya); le temps, qui est un outre 3.3,2 et 140,141); VARTAMNE par nature, se divise, lorsqu'il rside vaut de 3.2,123 188 inclus (cf. K. dans un objet ayant forme matrielle, 3.3,131), mais K. ne reconduit le en prsent, pass, futur, du fait de terme que sous 123; BHAVISYATl vaut l'action (qui est elle-mme prsente, de 3.3,3 15 inclus (et en outre 136passe, future). 139), mais n'est cit par K. que pour 14 Une autre remarque dans Bh. ad 3-13 et 15; trisu klesu K. ad 3.3,16 (et vt.4 conduirait supprimer BHTE, 3.4,6) et suiv.; enfin BAHULAM (3.3,1) sous prtexte que les suffixes dsiimplique en partie au moins une gnent le pass par nature. La rponse valeurtemporelle, tout comme l'indice est qu'on a besoin que le sens pass de "transfert" -VAT(3.3,131 132135). soit spcifi dans certains cas d'espce, Pratiquement, on le voit, la ainsi danskumraghtinetslrsa(3.2,86 validit est restreinte; elle lserait et cf. 51 ), khuhan (76 et 87), sutvon davantage si l'on tait des listes les (103 et 132), susupvas (172 et 107), s. relatifs aux formes personnelles du anehas (3.3,1 et cf. 3.1,133), enfin dans verbe. Si bien qu'on a moins lieu de vd. agnim dodhnasya (3.2,106 et s'tonner de l'importance alloue par

503 Thorie des temps du verbe

Or, I y avait une srie de formations nominales qui, au moins pour l'expression du prtrit, traaient une ligne de force temporelle: ce sont les absolutifs en -tv, les noms verbaux en -ta- et -tavant-, les participes parfaits (le participe aoriste est obsolet et, d'ailleurs, n'avait jamais eu nettement le sens d'un pass). C'est l prcisment que P. accroche la thorie du verbe personnel. 3. La transition est marque par le participe parfait (3.2,106), le s. "vdique" 105 qui anticipe sur le su. "commun" 115 servant pour ainsi dire de prlude. Ce participe, tant au moyen (106) qu' l'actif (107), est envisag, non en tant que paradigme, mais bien plutt comme substitut (optionnel) du parfait personnel, agnim cikynah " il a empil (les briques de l'autel du) Feu " . Limite au Veda,16 cette disposition est ensuite (108) tendue la bhs, o toutefois on nous apprend qu'elle se limite trois verbes, sad-, vas-, sru-.^1 Enfin le s. 109 installe titre de formes les Gramm, l'expression du temps dans les suffixes primaires (d'ailleurs la rgle DHTUSAMBANDHE cidessous note 20 vient encore en limiter les effets). Elle s'explique par l'extension mme de la phrase nominale, qui transfre aux drivs primaires tout le poids de l'expression temporel le : tout driv primaire ("agentis") est plus ou moins, en sanskrit, un participe. 16 La mise en exergue d'un enseignement vdique n'est nullement sans parallles. Elle se justifie ici parce qu'il a t question du participe au 104et que lecas du participe revient dans 106 et 107 (qui commandent leur tour 108 et 109): or 106 et 107 dpendent de 105. 17 Ainsi, au 124, le participe prsent est substitut du LAT. Les exx. cits pour 106-108 sont bien des exx. o le participe joue le rle de prdicat (non pas 109, bien que K. donne pour contre-exemple ansvn par ex.: nsit/ nsnt/ nasa " i l ne mangeait ou ne mangea point "). Pourtant le s. 124 justement nous laisse dans le doute : si l'Accus, pacamnam est trait en substitut du LAT, s'il est prcis que le Nomin. pacan ne peut tre employ qu'au sens de pacati (il faut diredevadattah pacati, non d pacan [mais K. ad 3.3,14 autorise au participe futur karisyan devadattah]; l'interdiction est d'ailleurs ensuite attnue dj dans Bh. in fin., que reprend K.), on en dduira analogiquement que le jaksivn de K. 107 signifiera "ayant mang" tout autant que "j'ai (ou : la) mang". LesCi. 109, sous l'apparence de poursuivre le mme thorie, vise surtout instruire des formations " irrgulires". Les non-Pninens, qui conservent 108 en ignorant 105-107, considrent par la force des choses que ledit s. concerne, non le participe, mais le parfait: ainsi C. 1.2,73 pose qu'avec sru-sad-vas- le parfait s'emploie au sens d'un prtrit indiffrenci (englobant donc le proksya BM., ce qui rsulte de Bh. 108 in fin.), concurremment l'aoriste et au parfait, tout comme P. 105 enseignait le sens du prtrit indiffrenci pour le UT vdique. Il est vrai que C. 74 connat aussi un participe parfait substitut optionnel du parfait personnel, c'est--dire qu'il admet la validit non vdique de P. 107, possibilit qu'voque encore la Bhsv. J. 2.2,88 atteste que "selon certains" -vas- et -ana- sont usits aprs toutes racines aux lieu et place du parfait personnel. H. 5.2,1 est celui qui transcrit avec le plus de dtail cette situation assez trouble : il pose d'abord le parfait en valeur se prtrit pur avec sru- sad- vas-, concurremment aux temps requis par d'autres rgles, mais il ajoute que certains rclament -^osseul valable, non le parfait personnel; que d'autres admettent aussi l'imparfait (comme forme indiffrencie?). Ensuite (2), que -vas- et -ana- valent seulement dans le domaine du parfait personnel, -vas- au sens "agentis", -ana-, en outre, en emploi passif et impersonnel; les exx. son tirs de verbes divers. H. ajoute encore que -na- manque parfois, que "selon certains" -vas- lui-mme n'est valable qu'aprs sru- sad- vas- (-ana- tant

504 Louis Renou

toutes-faites (nipatana) trois participes parfaits, soit qu'il les entende aussi comme substituts optionnels de l'indicatif (ce q u ' i n d i quent les exemples, ainsi upasedivn kautsah pninim " Kautsa a t l'lve de P."), soit (ce qui devait tre la vritable intention de P.) qu'il attire, l'attention sur certaines anomalies morphologiques. T o u t cela ne laisse pas de crer une situation un peu t r o u b l e , que les commentateurs e x p l o i t e r o n t en des directions diverses. 4. C'est cet instant que Pnini i n t r o d u i t le verbe personnel. D'abord le LU (110), c'est--dire l'aoriste considr comme notant un pass pur (ou mieux: indiffrenci). Sitt aprs, le LAN ou imparfait (111), un peu plus loin le UT ou parfait (115), qui e x p r i ment des aspects spcialiss du p r t r i t ; dans l'intervalle et aprs, sont indiqus quelques autres procds pour rendre le p r t r i t sans s o r t i r du verbe personnel. 18 Suit la thorie du prsent ou LAT (123), quoi s'annexent aussitt les substituts du prsent personnel, c'est--dire les participes (jusqu' 133) ; on peut, lato sensu, y rattacher les suffixes dcrits ensuite, tels que -tr- (134 et suiv.), dont les fonctions sont, en grande partie, celles d'un duratif, donc, indirectement au moins, d'un " p r s e n t " . D'autres suffixes viennent aprs, qui sont sans support t e m p o r e l , mais ont en commun avec la srie prcdente d'appartenir la classe des " a g e n t i s " . Le troisime pda fait alors retour aux valeurs temporelles, et plus particulirement l'expression du f u t u r : celle-ci concerne, o u t r e certaines formes du verbe personnel, des drivs nominaux en - t u - (infinitifs), en -oka- (10) et quelques autres. Le f u t u r p r o p r e m e n t d i t n'est allgu qu'au su. 13 (et encore, de manire indirecte, cf. 14 ci-dessous), c'est--dire aprs qu'auront t dcrites d'autres expressions personnelles ayant valeur ou faisant fonction de f u t u r (4-9). Suit presque immdiatement (15) le f u t u r priphrastique: formation d'origine nominale, qui donc nous achemine assez naturellement l'expos de nouveaux drivs primaires, qui sont, cette fois-ci, des " a c t i o n i s . " Le r e t o u r la thorie des temps du verbe se produira plus loin (131): il s'agit ici, derechef, du f u t u r personnel, mais orient vers des nuances modales : prcis en effet par le conditionnel (139-141) qui est morphologiquement un temps, syntaxiquement un m o d e , combin avec l'optatif, lequel apparat d'abord t i m i d e m e n t aux s. 134, puis 143, plus fermement partir de 147, encore que ses caractres fondamentaux ne soient formuls qu'au s. 161. Les six aspects de l'optatif sont ensuite transfrs l'impratif (162), qui en possde trois

tout fait rejet); selon d'autres, on aurait -vas- partout, mais non -na-. Les potes, dit SK. 3095, usent "diversement" c'est--dire librement de -vas- (et de -na-1), exx. tasthivmsam, adhijagmusah (donc: hors de la fonction de prdicat). BM., rappelant que d'aprs Bh. -vas- est vdique (comme -ano-), admet implicitement l'usage de -vos- dans la langue non vdique. Il ressort de tout cela que les Gramm, ultrieurs ont exploit P. 3.2,106 sqq. pour baucher un emploi gnral du participe parfait,

perdant de vue le rle de prdicat. Cf. encore nos Etudes de gr. skte ( propos de CV.) 1 p. 98. 18 P. part de la fonction, non de la catgorie; de l s'explique d'ailleurs qu'il juxtapose les drivs nominaux nuance (accessoirement) temporelle et les formations de participe ou de verbe personnel. Ainsi, dans le nom, part-il des kraka ou modalits de recton casuelle (lie au verbe), pour aboutir aux dsinences (vibhakti) aptes exprimer ces modalits,

505 Thorie des temps du verbe

autres t i t r e singulier (163). Le pada s'achve par l'expos du prcatif et du d-prcatif (autrement dit, du prohibitif), 5. L se t e r m i n e la description des formes personnelles : elle comporte, on le voit, un passage nuanc des valeurs temporelles aux valeurs modales. Cette manire de prsenter les faits rpond, en somme, ce que les textes littraires o n t nous apprendre sur la dperdition du mode aprs l'poque des mantra vdiques (l'impratif tant part, en tant qu'il est par nature impropre la subordination): disparition rapide du subjonctif (et de pseudoinjonctif nuance modale), restriction de l'optatif aux emplois d'ventuel ou de p r e s c r i p t i f donc, des emplois trs faiblement modaux si bien que l'indicatif, sous ses divers aspects, deviendra le mode par excellence, entranant dans son sillage des rsidus d'optatif. Seul dborde sur le pda4 de l'AstdhyyT un fait de langue curieux ( peine attest d'ailleurs en littrature), savoir la rptition (mredita) d'un impratif, type lunhi lunJhty evyam lunti " il coupe de faon rpte," " il coupe et recoupe". 19 Enfin les s. 6 8 n'intressent que le domaine vdique: 6 constatant l'indiffrence au temps des trois prtrits personnels ; 7 et 8 dcrivant l'emploi du subjonctif; autrement dit, la notion de mode l'emporterait en expressivit, dans le Veda, sur celle de temps, ce qui ne laisse pas d'tre exact. La thorie pninenne est ainsi encadre par les faits vdiques, le s. terminal 3.4,6 ne faisant que reprendre, sous une autre perspective, le s. 3.2,105 par lequel cette description avait commenc.20 II fallait bien inscrire l'enseignement 2-5 cette place insolite de la Grammaire, puisqu'il est cens dpendre du DHTUSAMBANDHE1. L'indicatif (prsent) est intress ce mme emploi (selon 3 et 5.), titre facultatif, puisqu'on peut dire chondo 'dhlsva (etc.) ou chando 'dhlte (etc.). . . ty evyam adhite " il tudie continment la mtrique" (littralement: comme qq'un dont on dirait " il tudie, il tudie. . . "). Mais c'est certainement l'impratif qui est typique (cf. JAs. 1959 p. 70 pour les quelques rfrences littraires connues), l'indicatif n'est qu'un subrogat. Les formules en unhi-lunhi s'apparentent aux composs tels que jah'tstamba (Wack. 2.1 p. 315), ehiyavom (p. 328) et surtout utpacanipac (ibid.). ce DHTU0 pour le s. 6 parat hautement artificielle (et d'ailleurs les exx. de K. ne comportent que des verbes isols, comme ody mamara " il est mort aujourd'hui ", exception 3.2,115); elle se comprend bien, en revanche, dans le cas cit par Bh. (1 ad vt. 2), agnistomayjy asya putro janit "celui qui clbrera un Agnistoma aura un fils", o ydj)'n, qui devrait signifier "ayant clbr" (3.2,85), reoit une acception temporelle conforme celle, du verbe voisin (janit): cas intressant s'il est linguistiquement authentique d'une concordance de temps, sur un plan non morphologique, mais syntaxique. Bh. ajoute que c'est le qualifiant (en l'occurrence, le driv primaire) qui, du fait de son caractre secondaire, se plie au temps du qualifi ou verbe 20 personnel; ce n'est pas l'inverse La q u e s t i o n se pose d o n c d e la c o m (raisonnement repris N., SK. 2824 et patibilit mutuelle de ces deux rgles ailleurs); BM. voque le primat de encadrantes. K. est formelle: 3.2,105 l'acte. Toutefois H. 5.4,41 cite une vaut pour un pass non spcifi (oviseseno); 3.4,6 vaut dans le cadre du phrase (du Sisupla) o l'on voit un DHTUSAMBANDHE 1, c'est- -dire l futur (yaih plvayisyanti) s'adapter au prtrit qui suit (sa dadarsa) au point o sens du verbe est conditionn par de perdre le sens futur (ex. attest sa " relation avec un verbe " voisin, padntarasambandhe, dit CV. 1.3,128 aussi SK. 2.4,224). Le Bh. tend cette sorted'" attraction" temporelle aux un peu diffremment. L'adduction de
19

506 Louis Renou

6. L'nonc des valeurs fondamentales requiert chez P. un minimum de mots: l'aoriste s'emploie [BHTE] 3.2,110, l'imparfait [BHUTE] ANADYATANEm, le parfait [BHUTNADYATANE] PARO'KSE 115, le prsent VARTAMNE 123, le f u t u r simple [BHAVISYA Tl] 3.3,13, l'autre f u t u r [BHAVISYA Tl] ANAD YA JANE 15; enfin le conditionnel est dcrit par un "actionis" KR/Y'ATI PATIAU 139, avec reprise de [BHAVISYATI] ou (140) BHUTE. Avec ce laconisme contraste l'ampleur relative donne aux nuances modales, y compris celles du subjonctif (bien que le subjonctif soit vdique et, partant, sujet une formulation abrgeante). Certes, la concision est l'un des traits majeurs de l'AstdhyyT. Elle a l'avantage, en l'occurrence, de souligner le ct rudimentaire des emplois temporels en sanskrit commun, dans cette langue mal dfinie qui englobe d'un seul mouvement l'tat vdique (archasmes exclus) et l'tat "classique." 21 Mais, si l'on fait entrer en jeu des faits, des opinions, des tendances que rvle la littrature grammaticale post-pninenne, on s'aperoit que lathorie des temps tait relativement toffe et qu'elle a conserv plus d'un trait intressant. 7. L'aoriste est considr comme temps du pass (sans plus), par opposition l'imparfait et au parfait: cela, parce que ces deux derniers comportent des spcifications qui manquent l'aoriste, non parce que l'aoriste pninen indiquerait le " procs pur et drivs secondaires, considrant que goman, misa ct de sJt, pourra signifier " il tait possesseur d'une vache", donc "I possdait. . .", alors que, pris en soi, le suffixe -mont- dsigne le temps prsent, TAD ASYSTY ASMIN 5.2,94. Il y a l, bien entendu, de notre point de vue un raisonnement spcieux. D'ailleurs le vt. 2 prcise lui-mme que le changement de temps ne porte pas sur le suffixe en tant que tel, mais affecte le driv en tant que mot complmentaire (upapada) du verbe; autrement dit, c'est la phrase seule qui reoit la notation temporelle. La remarque est assez fine. Bh, analyse en consquence l'ex. prcit (en agnistomayjin) "quelqu'un, ayant clbr l'A0, . . .qqch. se passeraen lui. . . en qui?. . . en celui qui aura un fils. . . quand?. . . quand il y aura une clbration d'A faite par lui". 21 Nous avons vu que l'intrusion du Veda, dans lathorie des temps, se bornait deux passages, fidles l'un et l'autre au principe du BAHULAM qui domine toute la description vdique : (a) selon 3.2,105 (s. non comment Bh.), le parfait vd. s'emploie au sens du prtrit (pur), ce qui, d'une part, largit l'usage du sanskrit commun (115), d'autre part fait opposition, par avance, 3.4,6 (ci-aprs); les exx. choisis, ohom...
dadarsa, aham. . . tatna (K., mais

emprunts Bh. 106-107), sont en effet conus en fonction de 115. Cette manire d'noncer les faits peut sembler sommaire; mais I faut avouer que, compte tenu des intentions de P. en matire de vdisme, elle reflte bien l'impression globale que laissent les mantra; (b) selon 3.4,6, ce sont les trois formations personnelles du prtrit qui sont utilises "pourtous les temps" (K.; s. non comment Bh.). Extension un peu surprenante, dont K.avoulu peuttre limiter les consquences en ajoutant dhtusambandhe. Mais cette adjonction est ici inoprante (cidessus n. 20). Tout aussi bien les exx. retenus sont ceux o n'existe aucune " relation " extrieure, ce sont akarot okorom en fonction de prsent (ou d'imparfait?); avrnta, d'aoriste; mamra, de prsent (mais aussi bien, d'aoriste), les temps normaux tant valides optionnellement (K.). On retiendra ce tmoignage sur l'arbitraire temporel prvalant dans la langue des mantra et des yajus. Reste concilier 3.2,105 et 3.4,6 pour ce qui est du parfait : ce sont deux perspectives lgrement diffrentes d'une mme ralit.

507 Thorie des temps du verbe

s i m p l e " , au sens o Meillet Introd. 7 p. 250, dfinissait l'aoriste indo-europen. En ralit, bien que l'aoriste soit affect au p r t r i t en gnral, les Grammairiens ont t amens, pour en dlimit e r le domaine par rapport aux autres formations du pass, lui allouer le pass " r c e n t " , voire le pass " d ' a u j o u r d ' h u i " : de l, le nom o'adyatan(vibhaktih) "formation-dsinentielle relative au (pass) d'aujourd'hui " que donnent l'aoriste K t . et d'autres grammairiens. A u s. 110, Bh. se demande si l'aoriste, du fait mme qu'il est instruit sans spcifications, ne f o r m e pas une exception-entravante (apavda) l'emploi de l'imparfait; autrement dit, si, dans opma payah " nous avons bu du lait " , l'imparfait attendu par 111 (car il s'agit d'un fait d'hier, comme on peut le supposer) ne sera pas ncessairement vinc par l'aoriste selon 110; mais il ajoute qu'il ne s'agit pas l d'une " e n t r a v e " , disons plus clairement, d'une opposition d'o rsulterait un choix ncessaire: si l'on emploie l'aoriste apoma, c'est qu'on n'a pas voulu rendre un pass antrieur aujourd'hui, qu'on a choisi de rendre un pass indiffrenci. 22 C'est reconnatre l'importance de la vivaks, de I'" i n t e n t i o n " , dans le domaine des fonctions grammaticales : on n'est jamais tenu d'employer une forme "spcifie". C'est reconnatre aussi (de notre point de vue) que l'aoriste, t o u t en s'adaptant particulirement noter l'aujourd'hui, demeure une forme gnrale, propre noter t o u t vnement dont l'emplacement temporel n'est pas prcis. Dira-t-on, ajoute encore Bh., qu'en disant " nous avons bu du l a i t " , on sait trs bien quel jour cet vnement s'est situ (seul le f u t u r , si l'on admet cette object i o n , aurait le d r o i t d'tre "gnral ")? La difficult se rsout par la vivaks: t o u t dpend si l'on veut employer une forme gnrale
22

Le principe de la vivaks (Termin. s.u., Introd. Durgh. p. 131), relativement vit dans Bh., se dveloppera chez les Gramm, ultrieurs, tant il offre de commodits, au moins pour les faits d'emploi et de syntaxe. A propos de la thorie des temps, cf. surtout CV. 1.2,81 ; 3,3 Kt. 3.1,16 H. 5.2,5 et ailleurs; H. va jusqu' introduire la vivaks dans le s. mme, en spcifiant qu'on a l'aoriste l o il n'y a pas intention de diffrencier. A suivre CV., on conclurait que des tranches entires de la thorie s'expliquent ce qui veut dire, dans la vraie pense de C, s'ludent en faisant appel la vivaks. Le principe se fonde videmment sur le VA pninen, dont l'extension n'est nulle part aussi sensible que dans le chapitre des valeurs temporelles, mais quel nonc de grammaire rsisterait s'il fallait en soumettre l'application la "volont" que peutavoir le sujet d'employer tel le forme plutt que tel le autre? Un autre procd, non incompatible avec le prcdent, est de considrer que la position morphologique prcde la constitution de la

phrase, c'est--dire a lieu sans tenir compte de la phrase; il y a l un paralogisme, qui permettra K. 3.3,131 de dire que svah karisyati " il fera demain " est correct, karisyati se situant sur le plan formel, l'addition de svas sur le plan du sens-de-laphrase : c'est le conflit entre le padaet le vkya-samskra (Termin, s.uu.). Ace passage, K. raisonne ainsi: celui qui estime que, dans laforme gocchdmi (employableselon leditsu.au sens de "j'irai dans un avenir proche"), c'est purement et simplement le sens du prsent qu'on a, et que la notion d'un autre temps rsulte de la phrase; or, au moment o l'on forme un mot, on ne met pas en exercice le temps connu par la phrase; (nous rpondrons que) ce n'est pas un interprte tel que celui-l, au courant du sens de la phrase, qu'est destin le problme ici entrepris (mais un tudiant ordinaire, auquel on a donc faire savoir que gacchmi signifie bien "j'irai "). Le prsent s., prcise N., est fait pour les mandabuddhi, les gens l'esprit lent.

508 Louis Renou

pour noter un fait particulier, ou si l'on prfre recourir une forme spcialise, t e l l e l'imparfait. 8. La discussion qui s'engage autour de la racine vas- " passer la n u i t " (110 vt. 3) claire bien le sens qu'il faut donner au [BHUTE] du s. Le v t . en question ajoute la thorie de l'aoriste le cas de la racine vas- sous deux conditions: (a) si l'on a dire " j ' a i pass la nuit (dernire tel e n d r o i t ) " , donc [kva bhavn usitaht] aha m amutrvtsam, la nuit tant arrive son t e r m e ; ou encore (b) " j ' a i pass la nuit (sans d o r m i r ) " . Le premier de ces deux emplois constitue une exception l'imparfait: on attend en effet l'imparfait pour noter un procs qui n'appartient plus l'aujourd'hui 2 3 ; si l'on a ici l'aoriste, c'est bien parce que l'aoriste concerne un p r t r i t qui est d'aujourd'hui mme ou qui confine l'aujourd'hui. 2 4 On rejoint alors une valeur linguistique connue: l'aoriste de constatation, dont l'usage privilgi a lieu au cours du dialogue, quand les personnages, bien entendu, parlent d'vnements qui leur sont survenus eux-mmes ou dont ils o n t t les tmoins. Dans le second exemple (b), s'il y a " v e i l l e c o n t i n u e " (vt.4, jgaranasamtatau), le sujet qui d i t " j ' a i pass la nuit ( tel endroit, s a n s d o r m i r ) " 2 5 croit avoir affaire aujourd'hui et emploie l'aoriste. C'est donc bien par leur opposition rciproque que les Grammairiens dfinissent imparfait et aoriste. 26 9. Ds lors, contrairement aux apparences, c'est l'imparfait qui est laforme non-marque, n'tant ni actuel (comme l'est l'aoriste), ni hors la vue (comme nous verrons qu'est le parfait). Ceci rpond assez fidlement aux tendances littraires, o nous voyons que l'imparfait ne peut gure se dfinir sinon ngativement, soit par rapport l'aoriste, soit par rapport au parfait. Il est un cas toutefois o l'imparfait, d'aprs les Grammairiens (111 v t . 2), empite sur le t e r r i t o i r e du parfait: il dsigne en effet un vnement du pass hors la vue, condition qu'il s'agisse d'un fait notoire 2 7 et qui (par sa date: arunadyavanah sketam " les Grecs assigrent Sketa") 2 8 pouvait tre peru du sujet parlant: on souligne ainsi, par anticipation, que le parfait sera appropri
L'aujourd'hui, disent les ce. depuis K., va du lever normal au coucher normal, en incluant la demi-nuit de part et d'autre, ainsi H. 5.2,7; lever normal signifie quatrime veille Kai. rtrivisese K. est une mauvaise leon, bien que connue de N. (rien dans la Padamaj.). 24 Pour dsigner un acte englobant hier et aujourd'hui (ou laissant le choix entre ces deux temps,) l'aoriste prvaut sur l'imparfait, adya hyo 'bhuksmahi " nous avons mang aujourd'hui (et) hier" (peu importe si Bh. Ill vt.1 tire cette disposition d'un artifice technique, savoir du mot ANADYATANEL compris comme compos possessif). De mme, le futur non spcifi prvaut sur le futur priphr. quand il y a mixture entre l'aujourd'hui et le demain. La forme la plus gnrale l'emporte.
25 23

tenu d'user de l'imparfait (Bh.). C'est bien par rapport l'aoriste que l'imparfait, appel temps du nonaujourd'hui chez les Pninens, est dnomm hyastani "temps de l'hier" chez Kt. et H. ; c'est aussi parce que lesexx. grammaticaux serrent au plus prs les confins de la temporalit. 27 P. LOKAVIJTE, H. 5.2,8 khyte. 28 CV. rajeunit l'exemple (que K. maintient par tradition tel qu'il l'a trouv dans Bh. : souci d'autonomie, du moins en matire d'exemple, chez les Gramm, htrodoxes) en ojayaj japto(\\re : guptol) hnn, cf. Liebich d. Ksratar. p. 266 (JAs. 1932,1, p. 152); H. loc. cit. orunat siddharjo 'vantln. II demeure au moins ceci dans l'usage littraire que l'imparfait fonctionne comme une sorte de doublure (plus rare) du parfait, tout en dbordant largement surl'aparo'ksa.
26

S'il a dormi, ft-ce un instant, il est

509 Thorie des temps du verbe

aux faits lointains, on marque aussi que l'acte not par l'imparfait n'est pas un acte indiffrent; dans le choix, apparemment libre, voire arbitraire, que font les auteurs, en l i t t r a t u r e , entre les diverses expressions du pass, l'imparfait convient aux vnements de caractre " n o t o i r e " , ceux auquels le sujet porte, a p r i o r i , quelque intrt. Les contre-exemples prcisent ce p o i n t : " le soleil s'est l e v " , udagd dityah (aoriste: fait d'exprience); " Devadatta a fait une n a t t e " , cokra katam devadattoh (parfait: fait hors la vue, mais nullement n o t o i r e ) ; " Vsudevaatu Kamsa jaghna kamsam kilo vsudevoh (parfait " m y t h o l o g i q u e " : fait hors la vue, n o t o i r e , mais loign dans le temps; kilo souligne la nuance " d i t - o n communment"). 2 9 Pnini t r a i t e cet endroit (112-114) d'un procd occasionnel pour rendre le sens du pass au moyen du f u t u r grammatical ; si cet enseignement succde celui de l'imparfait, c'est qu'il concerne galement le pass du non-aujourd'hui, et qu'au surplus il comporte l'imparfait t i t r e de variante, soit ncessaire (113), soit optionnelle (114). C'est le f u t u r qu'on t r o u v e dans une phrase du type abhijnsi devodatto kasmresu vatsymah " t e souviens-tu (que) nous rsidions (en ce temps-l) au Kasmr?". Peu i m p o r t e en ce moment l'origine de cette phrasologie singulire, qui en l i t t r a t u r e est quasiment inatteste: s'agit-il d'un f u t u r au sens d'un ventuel du pass? Ou d'un t o u r direct ( quoi fait penser l'ellipse de yod) " t e souviens-tu (de ce que nous disions alors, savoir:) nous allons rsider au K.?". En t o u t cas, la prsence de yod suffit rtablir la norme, c'est--dire l'imparfait (. . . yat kasmresv avasma), t o u t comme elle rtablit l'optatif (forme attendue en contexte modal) dans 3.3,147-151 155168. 30 10. Climax des emplois temporels, le parfait pninen est le temps du pass " h o r s la v u e " . Bh. se demande si c'est le " t e m p s " qui est hors la vue, mais kle n'tant pas rubrique-gouvernante (ci-dessus 2), mieux vaut suppler le DHT0H de 3.1,91 et ent e n d r e que c'est la " r a c i n e " donc le " v e r b e " qui est hors lavue. Objection : " r a c i n e " n'est qu'un mot, comment un m o t serait-il sous la vue ou hors lavue? DHT0H signifie: le sens not par la
29

Kt. 3.1,16 introduit ici la vivaks : l'imparfait est valable si l'on n'entend pas traiter l'vnement comme tant hors la vue, tout en maintenant l'ide d'un temps historique. N. et SK. prcisent qu'un contemporain de Krsna pouvait seul dire l'imparfait " i l a tu Kamsa ' '. 30 Mais Sr. 2.34,5 c i t e smarasi yad upakarisyati au sens de upakuruthh [sic]. Avec ou sans yad, le futur (114) reprend ses droits (concurremment l'imparfait) quand le verbe affect requiert lui-mme un autre verbe sa suite (SKNKSE; Vop. 25.30 libelle tout autrement, anekasmrye)] autrement dit (K.) s'il n'est qu'un signe (Iaksana) annonant une chose signifie (laksya), note par un verbe ultrieur (lequel sera donc ventuellement entran au futur). L'ex. de SK. est plus pertinent que celui de Bh. ou

de K. : " t e souviens-tu que nous rsidions Prayga [simple "signe" de ce qu'on a dire ensuite], que nous faisions l'ablution du mois de Mgha [chose rellement signifie, chose qui seule devrait dclencher le souvenir]?" abhijnsi devadatta prayge vatsymah (/avasma), tatra mgharn snds/mah (asnma). En fait, prcise Bh., la notion de "chose requise" vaut des deux cts : le second verbe requiert le premier, comme le premier requiert le second. La position 'KNKSest rare chez P.; on la retrouve propos d'une phrase (ou d'un verbe) complmentaire (par la prsence duquel ou de laquelle la rgle a son plein effet) 3.4,23 8.1,35; 2,96 et 104. Sur les attestations littraires (rarissimes) du futur de "souvenir", v. ma Gramm, scte p. 462 JAs. 1959 p. 73.

510 Louis Renou

racine....Quel est ce sens? C'est I " ' a c t i o n " (kriy): c'est donc l'action qui sera hors de vue. Ne risque-t-on pas alors d'avoir le parfait dans " il a cuit (un mets) h i e r " (o l'on attend l'imparfait selon 111, hyo 'pacat), car il s'agit l d'une action non point vue, mais infre, qu'on ne peut m o n t r e r l'instar d'un objet mis en boule 31 ? Bh. adopte la thse finale selon laquelle ce qui est hors la vue, ce n'est pas l'action elle-mme, mais les instruments (sdhana) permettant de l'effectuer 32 ; un sodhana, en tant que totalit des " q u a l i t s " , mais distinct d'elles, peut tre atteint par voie d ' i n ference; il est capable d ' e x p r i m e r un tat sous la vue ou hors la vue. Suit une discussion sur la question de savoir si PARO'KSE " hors la v u e " vise qqch. qui a eu lieu il y a cent ou mille ans, qqch. qui est spar (du sujet parlant) par un mur ou une natte, enfin ( r e t o u r au paro'ksa t e m p o r e l , entendu sous forme attnue) qqch. qui a eu lieu il y a deux ou trois jours (seulement). Bh., commesi souvent, ne dcide pas: ce qui importe, conclut Kai., est qu'il s'agisse d'un acte du pass non-d'aujourd'hui, ralis par des instruments sis hors de la porte des sens.33 11. Corollaire e m p i r i q u e : la premire personne se t r o u v e exclue, sauf, ajoute Bh., au cas o le sujet parle de quelque vnement qui lui est arriv pendant qu'il dormait ou quand il tait en tat d'brit: supto 'ham (ou : matto 'ham) kila vilalpa " il parat que j'ai parl, tant endormi ou ivre". Il se peut d'ailleurs que, mme veill (et non ivre), on ne peroive pas le temps : ainsi le grammairien Skatyana, assis sur le chemin, ne vit pas la caravane passer [distraction des grammairiens !]. Les objets des sens sont intrumentaux pour la perception quand ils sont attels par la conscience; s'il n'y a pas conscience, il y a un tat de fait " hors la vue". Autre cas, plus inattendu, o le parfait est licite, bien qu'a la premire personne (vt.1): le cas d'une dngation absolue (atyantpahnava), exemple " il n'est pas vrai que je sois all au Kaliga" no (. ..) kal'mgnjagma (entendez: je n'ai pu commettre le crime qu'on m'impute au K, n'ayant jamais mis les pieds en ce pays).34
31 pindlbhta : mme " mise en boule" on ne peut voir cette action car, tant donn le caractre instantan de ses lments constitutifs, il n'y a pas l de " mise en boule " (vritable) (Ng.). 32 Kai. glose " les lments pourvus d'un pouvoirexpressif", mais Bh.gunasamudya ensemble des qualits. C'est le sdhana, ajoute K., qui cause le fait que les gens croient en le caractre sous-la-vue (pratyaksa) d'une action, alors que, prise en soi, toute action est paro'ksa. De mme H. 5.2,12 : toute ide verbale est paro'ksa, les sdhana seuls sont susceptibles d'tre sous-lavue. Analogue J 2.2,95. Une autre objection est donne Bh. : quel temps employer lorsqu'on a vu sur le chemin l'eau de riz (ayant servi la cuisson)? On ne sait d'abord, prcise Kai., si ce riz a t cuit ou offert-en-sacrifice ou battu, mais aprs on ralise qu'il s'agissait d'une cuisson et l'on dit " il a cuit (du riz) " . Les instruments de

l'action tant demeurs hors la vue du sujet, on devra dire papca; de mme encore, si sont hors la vue les particularits nes de l'action, tels les grondements et sifflements (du chaudron). En dernier sur le parfait dans les textes littraires (kvya), JAs. 1959 p. 70. 33 Ces derniers mots sont complts par Kai. 34 Prcision fournie par H. 5.2,11 ( c ) ; il s'agit du meurtre d'un brahmane (acte explicable au pays Kaliga); une dngation partielle comme " j e n'ai pas tu. . . " auriat entran l'imparfait, Noter que dans l'tmastuti Rgv. 10.48 (-49), la premire personne du parfait est peu prs exclue, mais on la trouve dans les emplois dngatifs n para jigye, na. . . va tasthe " i l n'est pas vrai que j'aie. . . " : cas d'atyantpahnava. Cf. Valeur du parfait p. 83 85 87.

511 Thorie des temps du verbe

C'est la ngation d'un fait situ sous la vue (pratyaksa), comme dit Kt. 3.1,16. II n'existe gure de catgorie ferme, du moins dans le domaine de lafonction. On ne peut donc s'tonner si Pnini ouvre la voie l'expression du hors la vue par l'imparfait, non seulement dans le cas dont nous avons trait 9, mais encore au s. 116, o il autorise l'imparfait, concurremment au parfait, lorsqu'il y a pour terme "adjacent" (upapada) la particule ha ou sasvat, ex. ti ha (ainsi dans un pass du non-aujourd'hui, sis hors la vue)". 35 La mme disposition s'tend ensuite (117) la phrase interrogative, en tant qu'elle se rfre un pass proche36 (encore que situ hors lavue):on peutdoncdire y aja (ou : aya jad) devadattah " Devadatta a-t-il sacrifi ( date rcente)?" (les conditions tant les mmes que pour le su. 116). Nous en conclurons que le parfait tait mieux adapt exprimer un temps lointain, mais que, en phrase interrogative (ce qui revient dire: en discours direct), l'usage tait indcis.37 Nous voyons bien, dans la linguistique contemporaine, l'intrt qu'il y a distinguer, quant l'expression temporelle, entre la narration (ou la description) et le discours.38 12. Fidle au principe de partir des fonctions, Pnini poursuit la notation du prtrit en traitant de l'indicatif prsent avec sma. Cet emploi concorde avec celui du parfait (118),39 ou bien traduit un aparoksatva (APARO'KSE 119), autrement dit concide avec l'imparfait. Double validit qui reflte assez bien l'usage classique, o nous voyons le prsent avec sma doubler le prtrit de narration, sans comporter de rfrence l'actuel ou de participation au discours.40 Mais le prsent (sans sma) figure lui-mme pour noter un pass non spcifi (c'est--dire un pass de type aoristique), lorsqu'il
II s'agit du ssvat des Br, signifiant " p e u t - t r e " ou "certes", non du sasvat des mantra "toujours " (que connat seul Amara); cf. Minard Trois nigmes 1 5896 sur cette volution de sens inattendue, quoi l'on peut comparer dvit passant de " de deux manires" "assurment", ou fr. toujours au sens de "encore" ou " d u moins". H. 13 enseigne l'imparfait seul (avec ha et sasvat) dans ledomaine rpondant P. 112. 36 Pass proche (alpakle 'pi Sr. 2.34,3) signifie : moins de cinq ans (N. r se rfrant aux Naiyyika); yugntar H. 13. 37 H. 14 met au crdit de la vivaks (ci-dessus n. 22) l'usage de formes varies en une mme phrase, comme anvanaisJt / nyaksipat \ proce, pour un pass hors la vue. Ceci se rencontre en effet abondamment en littrature, surtout peut-tre dans le kvya jaina (que H. a en vue); le mme auteur cite encore abhaisit / ayuyutsayat \ yuyutsaymcakre. S K. ad 115 rappelle que les potes peuvent
35

cakra (ou : ti hkarot), sasvac cakra (ou : sasvad akarot) " i l a fait

user de l'aoriste si, mme dans le domaine du UT, ils entendent n'exprimer qu'une tranche de pass indiffrencie. Durgh. cite ici un aoriste en valeur de parfait. 38 Nous avons not propos du kvya l'intrt qu'offre la distinction entre narration (impersonnelle) et discours, JAs. 1959 p. 2 et passim. Cf. pour le franais mme (dans la rpartition des temps du pass), l'expos magistral de Benveniste BSL. 54(1959) p. 74. 39 Kai. justifie la reconduction de PARO'KSE dans 118: si elle n'avait pas lieu, ANADYATANE (qui lui est associ) cesserait de valoir; d'autre part, il fallait une rgle indpendante APARO'KS CA119 afin de rvler la validit de Vanadyatana. Noter que sma sert aussi entraver l'optatif attendu, au profit de l'impratif, selon 3.3,165-166.
40

Sur le prsent avec sma, en dernier JAs. 1959 p. 71 ; mentionn Ganaratnam. p. 15.

512 Louis Renou

s'agit de rpondre, soit avec nonu " c e r t e s " (120: okrsJh katam devadatta/nanu karomi bhoh "as-tu fait la natte? O u i , je l'ai f a i t e " ) , soit avec no seul " non " ou nu seul " c e r t e s " (121)41 (no karomi bhoh ou aham nu karomi, concurremment nkrsom, aham nv okrsom

" non, je ne l'ai pas faite", " o u i , je l'ai faite"), une interrogation qui a t pose l'aoriste. C'est la un nouveau tmoignage (cf. 11) du souci que prend P. relever les emplois propres au discours, un tmoignage aussi de latendance qu'ont certaines particules modifier laforme du verbe qu'elles accompagnent: c'est ce qui se produisait tout l'heure avec smo. On notera ici que les particules plus neutres (no ou nu) admettent l'option, la plus appuye (nanu) commande un choix strict. C'est enfin le prsent (122) qui rend le sens du pass nond'aujourd'hui (donc, en filiation directe avec le s. 119) lorsque la particule adjacente est pur,42 ex. vosontha pur chtrh " les coliers rsidaient ici (antrieurement aujourd hui)". La restriction par rapport 118 est que, si la phrase comporte la fois sma et puro, 118 s'applique par prfrence 122; que, d'autre part, l'aoriste est admis titre optionnel dans le champ du s. 122 (ovtsuh...).*3
II s'agit du n Bhtl.-Roth sous 1) g), issu du n ou n des mantra; c'est le nu protivacane de Ganaratnam. p. 8-9. H. 17 et 18 connat nonu kurvantom, nu k, no k ct de nanu karomi, nu k, na k (par application de P. 3.2,124); emploi fictif. J. 2.2,100 considre ces deuxs. comme inutiles, l'emploi du prsent s'expliquant par le (dsir de noter) le non-achvement. 42 Emploi connu de Ganaratnam. p. 16. H. 5.3,7: si l'on a la fois pur (ou : yvat) et svah, on a le prsent. En fait, cet auteur (5.2,15) semble d'abord ignorer le prsent, instaurant avec pur (et tad ! Cf. " purdau ' ' dans la teneur de s.) les temps du pass; toutefois le prsent cumule avec ces temps selon 16, le prsent vaut seul avec smo pur comme avec ha sma, ha sasvat, ha sma pur (ce dernier conglomrat tant attest dans la prose vd., cf. un ex. de la Taitt. Samh. cit dans Bh. 122, o figure ha sma pura ssvat avec verbe au prsent). Enfin on peut se poser la question de savoir si ces temps passs "de remplacement" sont identiques aux temps de base, ainsi H. 15 rappelle que certains auteurs rejettent l'aoriste avec pur si le sens est celui d'un pass hors la vue. Ces temps valent chacun en son domaine propre, enseigne K., qui admet l'imparfait, que CV. 1.2,81 donne aussi, mais simplement en fonction de la vivaks.
41

Bh. ne cite d'exemples (pour le prtrit) qu'avec l'aoriste. 43 Dans la prose vd., pura s'utilise en effet avec le prsent pour noter des vnements typiques du pass (mais engageant l'actuel) selon Delbrck Ai. Syntax p. 278 Idg. Syntax 2. p. 266. Plus frquent est sma pura (ou plus exactement ha smo pura) qui, avec un prsent, note un pass indiffrenci (Ai. Syntax p. 502 Idg. Syntax loe. cit. ; ci-dessus, n. 42). Comme on l'a montr depuis longtemps, l'habitude d'avoir un prsent avec sma au sens du pass dcoule de l'emploi plus ancien avec sma puro. On peut se demander si la reconduction 'ANADYATANE122 qui n'est acquise que par l'artifice du Saut de Grenouille est vraisemblable, car elle aboutit crer un aoriste du non-aujourd'hui, cd. oppos aux valeurs habituelles de l'aoriste. En outre, les exx. de sma pur donns par K. 118 sont mal venus; on attendrait des exx. avec sma seul, comme en ont CV., Bhsv., SK. et autres. Toutes ces dispositions refltent les lments de discussion de Bh. 118 vt. 1, qui proposait d'abord d'entendre le prsent avec sma pur au sens d'un pass pur (non au sens d'un pass d'aujourd'hui), vt. qui sera rejet ensuite (vt. 2), la reconduction tant dclare valide aussi bien avec pur qu'avec smo.

513 Thorie des temps du verbe

Pareil flottement doit correspondre un certain usage : la langue hsitait, en prsence de puro, entre deux tendances : maintenir le temps requis par le contexte, ou imiter l'emploi avec sma, c'est--dire gnraliser le prsent, considrant que la particule suffit voquer la nuance temporelle exacte. 13. C'est alors seulement (123) qu'est introduite la thorie du prsent (indicatif) s-qualits: l'acte qu'il note se place " e n ce moment m m e " (samprati Kt. 3.1,11 ; sati "dans l'tant" H. 5.2, 19), d'o l'appellation vartamn donne au temps prsent (Kt.), ou encore bhovant(c\u\ est un nom des anciens Matres, d'aprs Kai.). Deux propositions additionnelles dans Bh. : a. L'indicatif prsent note galement un acte qui, ayant t entrepris (dans le pass), n'a pas cess encore (au moment o l'on parle), hdhlmahe " nous sommes en train d'tudier" : c'est, dit Sk., la valeur "suggre" (dyotya) du prsent. Bh. observe que cette addition est inutile, car cet acte inachev au moment o l'on parle n'est autre qu'un acte " prsent". D'ailleurs la cessation a lieu de toutes manires, car lorsqu'on dit " Devadatta est en train de manger", il advient ncessairement qu'il ne ou cause ou boive en mangeant. 44 On peut donc gloser VARTAMANE par "entrepris et non encore achev" (comme fait K. d'emble), sans avoir postuler aucune valeur additionnelle; b. L'indicatif prsent sert noter aussi un acte permanent, ainsi " les montagnes se tiennent (immobiles)" tisthanti parvat h, proposition qui ne comporte aucun fractionnement temporel (et suppose donc un additif au moins mental l'nonc de P.). Rponse : on peut concevoir des fractionnements temporels dans cette locution, ainsi " les montagnes se tiendront, se sont tenues".Mais est-ce en usage?Ce n'est pas affaire d'usage. Sur terre, les actions des rois passs, futurs et prsents sont le lieu (adhikarana) o se situe "se tenir" ; on dit " les montagnes se tiennent" quand ont lieu les actions des rois de maintenant; " . . . se tiendront" quand auront lieu les actions des rois venir, etc.Cette discussion aboutit, si nous comprenons bien, dgager l'ide d'un temps relatif, ide trs souhaitable en matire syntaxique et quoi P. luimme avait song en posant son DHTUSAMBANDHE (ci-dessous, n. 21). Quant la situation du prsent grammatical, elle se trouve largie par l'admission d'un temps "gnral ", admission d'autant
Dans "nous sommes en train montagnes se tiennent (debout)" d'tudier," le fait d'tudier n'est pas (b), H. ne voit qu'une application du interrompu par l'acte (ventuel) de mmes prsent de lachse "en manger, ou plutt ce dernier est trait cours " ; si l'on dit tasthuh, sthsyonti comme partie intgrante de l'tude en cette formule, c'est qu' "on se (Kai.). K. rsume grossirement ce d- rfre aux actions qui ont pris place veloppement en glosant VARTAMANE ou prendront place en prsence des par rabdho 'porisamptas ca, ce qui montagnes." Dj CV. 1.2,82, citant semble exclure toute autre nuance tisthanti girayoh, s'appuyait sur le de prsent. Plus prcis est SK. 2151 prsent "en cours"; de l vient, sans (c.) qui tient compte du prsent doute, le silence de K. Selon J. " permanent." H. 19 considre part 2.2,101 le temps prsent est celui le cas d'une phrase ngative, jlvam na qui va depuis le commencement (d'un mrayati " (on) ne fait pas prir un acte) jusqu' sa non-cessation, c'est tre vivant," o c'est l'ordre formul un temps intermdiaire, tanmadhyam ngativement (le niyama) qui est "en kdlam. cours et non achev." Mais, dans " les
44

514 Louis Renou

plus utile que ni I' " a c t u e l " de P., ni le " non-achev" de Bh. n'y prparaient. Mais Bh. soulve une objection c o n t r e la notion mme de " p r s e n t " : on ne saurait d i r e valablement " les fleuves coulent vers la m e r " , sy andante sa rita h sgarya, car l'actuelnon plus que le pass ou l'avenirne c o m p o r t e pas de mouvement. A u t r e ment d i t , ajoute Kai., l'instant seul est peru, en l'occurrence l'instant actuel, qui ne saurait r e c o u v r i r l'acte de " s e m o u v o i r " . Rponse de Bh. : cet acte a lieu en vue d'un certain but, qui est la cause mme de I'effectuation de l'acte; c'est en considrant un tel but q u ' o n pourra employer l'expression " i l marche". 4 5 14. Reste v o i r l'expression grammaticale du fait venir. Elle est dvolue par privilge au temps grammatical appel bhavisyant (nom des anciens Matres selon Kai.) chez les non-Pninens, mais simplement LRT, comme nous avons v u , chez P. La valeur en est dcrite au dbut du t r o i s i m e pda, c'est--dire propos des suffixes (primaires) dsignant une chose venir (3). Mais P., selon un procd qui lui est assez familier, t r a i t e d'abord du prsent employ au sens du f u t u r (4).46 De mme, en effet, que certaines particules inclinent l'indicatif p r s e n t f o r m a t i o n m i n e m m e n t instable vers le sens du p r t r i t , quelques autres, nommment yvat et pur,47 l ' o r i e n t e n t noter une chose venir, ainsi yvad Kai.: si l'on considre l'enchanement des actes consistant en le cumel des instants (notion d'instantanit, de provenance sans doute bouddhique). SK. utilise l'exemple pacati " il cuit": la racine pac- vise un faisceau d'actes, commenant avec l'ide de " mettre au feu " et circonscrits par le but unique qu'on veut atteindre. BM. note que la notion de prsent ne fait que qualifier le sens du verbe, non pas l'agent ou autres instruments qui, eux, peuvent appartenir au pass. S K. se demande aussi dans l'hypothse o le prsent consiste en un acte "en cours" pourquoi on dit " l'tmon est, tait, sera"? Rponse: dans cette essence unique qu'est l'tmon, une diffrenciation se manifeste, qui dpend des conditions particulires et repose sur telle ou telle activit. Bh. voit encore une autre manire d'chapper l'objection prcdente: il existe certes un prsent, mais on ne le peroit pas plus qu'on ne peroit la marche du soleil; aussi peu qu'on voit se modifier chacune des fibres de lotus en train de brler en masse, aussi mal voit-on se modifier l'entre-en-acte en apprhendant un un les instants dans l'afflux des kraka ou lments qui participent l'achvement d'une action verbale. Seuls lesTribhva(Kai.: Yogin's pour
45

qui existe une ralisation dans les trois temps) peroivent une telle modification. Kai. prcise encore : de l'ide verbale ayant forme d'acte, ide comportant plusieurs instants, on infre qu'il existe une valeur de prsent, parce que ces instants ne se manifestent pas tous ensemble. Et Ng. : en saisissant chaque instant, on est hors d'tat de percevoir l'entre enacte qui se modifie par la diffrenciation des instants. Mais, si tnu que soit chaque instant, leur total n'est-il pas perceptible aux sens? Rponse: l'acte est accessible aux sens, soit comme partie, soit comme totalit; en l'occurrence, le temps prsent s'accrdite si l'on impute dans l'ide verbale une totalit, l'acte en question comportant une pluralit d'instants perceptibles par inference. Durgh. 115 connat un prsent d ' " imputation-secondaire," qui, en fait, est fonction d'un verbe voisin.
46

ventuellement, du futur priphr.

(vt.). 47 Ces mots yvat et pur posent un petit problme, que les commentaires grammaticaux n'aident gure rsoudre. Il s'agit de particules, prcise P. (NIP ATA), donc probablement depuro au sens de "avant peu," yvat au sens de "cependant" (ou simplement, l'un et l'autre au sens de "assurment"). Les attestations litt-

515 Thorie des temps du verbe

bhukte ou pura bhukte au sens de bhoksyate " i l mangera". Mme o r i e n t a t i o n , mais optionelle seulement, 48 aprs les adverbes interrogatifs kad et karhi (5), aprs le pronom ko et ses appartenances (KfMVRTTE, 6), ex. kad bhukte ou kad bhoksyate " q u a n d mangera-t-il ?". C'est l'exact pendant du temps prsent employ au sens du pass (12); c'est aussi une vidence nouvelle du rle attnuateur des particules, permettant l'usage du temps le moins marqu, alors mme qu'existe une nuance dsidrative implicite (LIPSYM 6, ex. kataro bhiksm dadti " lequel de vous deux d o n nera l'aumne?", aumne que le sujet interrogeant "a dsir d'obtenir"). La participation effective est voque encore au s. 7, lequel admet (hors de la phrase interrogative) l'emploi du temps prsent pour noter un fait venir qu'on inscrit comme actuel afin d'encourager (protshayati K.) celui dont on dcrit l'acte ("celui qui donnera la nourriture, il ira au ciel " , yo 'annam dadti sa svargam gacchati, LIPSYAMNASIDDHAU, quand on russit atteindre ce justement qu'on dsire atteindre).49 L'un et l'autre verbes sont mis au prsent ( titre, il est vrai, optionnel), tout comme, dans l'Atharvaveda magique, nous voyons l'aoriste employ pour noter des vnements venir dont la ralisation est cense d'ores et dj acquise. 15. Ainsi accde-t-on aux valeurs modales du prsent gram matical quand il s'agit de noter un fait venir: elles sont poses concurremment au futur grammatical. En premier lieu (8), on a affaire une action ( venir) qui est le signe (LAKSANE)50 d'une rairessont rares (JAs. 1959 p. 79 sur pur). BM. renvoie Amara o yvat a plusieurs sens; de mme Ganaratn. p. 11. Pour SK. 2783, qui sans doute a vu juste, ces particules suggrent la certitude (niscaya). Mais CV. 1.3,3 (et d'autres) voi(en)t dans ce pur la conjonction "avant que . . ." (celle de P. 8.1,42) et de fait N. ad 8.1 r42 voque 3.3,4: ex. pur vidyotate vidyut "avant qu'ilne fasse un clair," conjonction bien connue (Bhtl.Roth I) c)) avec prsent au sens futur. En ce cas, yvat pourrait tre aussi la conjonction subordonnante qu'enseigne P. 8.1,36 "jusqu' ce que . . . " (mais CV. ad. loe: yavodbhukte tato vrajati " il mange d'abord, puis s'en va"); Sr 2.34,4 pur karti et yvat k = karisyati. Il faut se tenir au NIPTA de P. 48 Ceci dpend de la vlvaks (ci-dessus, n. 22) N., principe que CV. 1.3,3 tend tout un groupe de rgles relatives au futur. C'est en prenant ce s. pour exemple que Kt. 3.1,17 nonce prayogatah dans le s. mme: le c. constate que si, au lieu de dire yvad bhukte, on voulait employer le futur, il suffirait de ne pas poser yvat. C'est reconnatre la sollicitation qu'exercent les particules. Plus gnralement, c'est " l'usage" qui doit enseigner la spcification du temps; mais on aboutit alors une notion qui, en Grammaire, est tout aussi destructive de rgles que peut l'tre celle de vivaks ou de samskra (n.22). 49 Traduit selon la lettre, mais K. distingue l'homme [le donateur] qui russira gagner (le ciel) et l'homme [le donataire] qui souhaite obtenir (la nourriture). SK. 2786 dveloppe K.: le sujet parlant encourage le donateur en exprimant l'espoir que, partir du riz objet dsir le succs cherra au donateur, sous la forme du svarga. Noter que rien n'indique dans le s. mme la prsence d'une phrase incluant une proposition relative. 50 Nous avons dj rencontr le "signe" propos d'actions corrlatives (ci-dessus, n. 30). Laksana chez P. dsigne un signe matriel, servant reconnatre qqch., indice (ainsi 2.3,21) ou but (ainsi 2.1,14) ou, plus abstraitement, cause (1.1,62; 2,65). Le terme sert indirectement dcrire le Loc. absolu (2.3,37, "ce par l'tre de quoi il y a signe d'un [autre] tre"). Mais il a aussi une acception purement lexicale = "au sens de," ainsi 3.4,16.

516 Louis Renou

autre action (connecte), note par un impratif, ex. " s i le matre vient, apprends la mtrique" updhyyas cedgacchati, atha tvam chando 'dhsva; autrement dit. la venue du matre est le "signe" (nimitta N., hetu H. 5.3,11) de l'exhortation apprendre (adhyayane praisata N.). Nous constatons l la tendance des phrases hypothtiques renoncer l'expression prcise que conseille leur intention propre, cela parce que la conjonction (ced) dispense en quelque sorte de noter le mode attendu (les autres types d'hypothtiques seront abordes plus loin, 22). On remarquera que rien dans l'nonc de P. n'imposait de croire que le s. se limitait ce typede phrases: c'est laglose des commentaires qui nous l'apprend. Le s. 9 est une variante mineure de 8, instituant qu'il y a choix libre entre prsent et optatif pour noter un acte situ dans un avenir relativement proche,51 ainsi updhyyas ced gacchet (ou : gacchati, aussi d'ailleurs gamisyati ou ganta), atha tvam chando 'dhsva "si le matre vient arriver (dans un certaindlai)...". Ici apparat pour la premire fois l'optatif, dont la description ne sera reprise que vers la fin du pda en cours, d'abord dans des emplois "concurrents", ensuite seulement dans des emplois autonomes. Les. 9 correspond ce qu'on adit ci-dessus (12), touchant le prsent valeur de prtrit proche. Jusqu'ici donc nous avons des valeurs " venir" notes par des temps autres que le futur grammatical. Ce dernier ne sera instruit qu'au s. 13 : encore l'enseignement s'y combine-t-il avec celui d'une expression " intentionnelle". En effet, poursuivant ladescription entame aux s. 10-12, P. se proccupe d'abord de montrer que le futur-est parallle l'infinitif ou au datiffmal des noms d'action, ex. natam draksymlti vrajati " il va voir la comdie", littralement " i l va (donnant pour motif sa marche cet lment de discours:) j'ai l'intention de voir...". 52 Mais, au cours de la mme rgle, P. introduit le temps futur " pur et simple " (suddhe K.) avec cette formulation soustractive quoi il incline, SESE CA,53 c'est--dire " les dsinences du futur sont valables aussi, pour noter un fait avenir, dans les cas restants", soit "dans les cas non couverts par le KRIYYM KRIYRTHYM 10". Les exemples des com51 muhrtd rdhvakllne B M : il s'agit, sans doute, d'un dlai suprieur de peu un muhrto. 52 A u t r e m e n t d i t , " ti" n ' e s t pas prouv par les Gramm, comme nous l'prouverions, c'est--dire interrompant le sentiment d'une subordination; draksymlti est pour eux un quivalent pur et simple de drastum. Ex. analogue K. 3.3,156, l'exemple normal (K. 14) tant orjoyisyamno vosati " i l habite (l) dans l'intention d'acqurir (des biens)". On aurait tort, en effet, dans bien des cas, de voir dans ti autre chose qu'un outil grammatical, avec lequel le discours direct est pur formalisme. 53 SESE se dit presque toujours des emplois " restants", ceux qui n'ont pas t couverts par la description antrieure: ainsi, la voix active 1.3,78,

le bahuvrihi 2.2,23, le Gnitif 2.3,50, les suffixes rdhadhtuka 3.4,114, etc.: ceci pargne une dfinition, qui risquerait d'tre longue. Une acception diffrente du mot ne se prsente qu'une fois, 7.4,60 "(qui subsiste) seul". Il n'empche que laformulation de P. est un peu cryptique; SK. 2193 cherche la clarifier en posant asotym satym co (kriyym kriyrthym) "qu'il y ait ou non action faite en vue d'une (autre) action " ; mais un tat pdagogiquement satisfaisant n'est atteint que chez un grammairien tel que H. qui nonce par un s. distinctif bhavisyan 5.3,4. L'avantage du libell pnlnen est, comme nous le disions, d'impliquer ce qu'on pourrait appeler une vision historique des faits. J. 2.3, donne deux s. distincts, /rt (11), puis sese (12).

517 Thorie des temps du verbe

mentaires confirment bien qu'il s'agit du futur en son emploi le plus ordinaire, karisyati " il f e r a " (sans spcification de dlai ni d'intention). Le Bh. s'interroge sur l'interprtation de ce SESE (CA) : le vt. 1 propose d'y voir une allusion au s. 10 (Kai. : le s. 10 cessant de valoir dans 13): s'il n'y avait pas SESE, ajoute le vt. 2, le futur serait entrav par l'application de 3.1,94 (plus prcisment, par l'infinitif BM.). Le point de vue dcisif est le suivant: la portion LRJ u s. veut dire que le futur vaut dans le champ de 10; la portion SESE dont on pourrait aussi bien faire un su. distinct que le futur est valable "ailleurs" (CA tant superflu). Laissons de ct cette technique et retenons seulement que Pnini fait sortir en somme le temps futur de l'emploi " intentionnel ", ce qui concorde avec les conditions historiques : le futur normal du sanskrit commun n'est qu'un futur "des cas restants", une sorte de privation du caractre intentionnel. 16. Sitt aprs est pos le futur priphrastique (15), dfini comme un futur du non-aujourd'hui, ANADYATANE: c'est la rplique exacte de l'imparfait (10). Et de mme que l'imparfait trouvait son terrain privilgi dans l'vnement " d ' h i e r " , ce futur a le sien dans l'vnement " d e demain " (svah kart " il fera demain " ) , d'o son nom de svastan " le temps de demain ", nom que connaissent dj les vtt. 54 A ce propos s'engagent deux vrttika destins largir l'emploi du futur priphrastique: a. Ce futur remplace le futur simple quand il s'agit d'une action causant un chagrin (par i devane): "quand donc marchera cette (femme malade), qui pose ainsi les pieds ?" yam nu kad ganta yaivampdau nidadhti. On redoute une attente vaine; l'expressivit du contexte, la nuance implicite de "chagrin ", permettent d'employer un temps qui n'et convenu normalement que si l'acte s'inscrivait dans un avenir spcifi par des mots tels que "demain " ou "aprs-demain" 55 ; b. L'autre vt. n'est qu'une interprtation diffrente (et contourne) de l'exemple prcdent: du fait mme du temps qui passe, l'ide surgit d'une comparaison, "elle est pareille quelqu'un qui marche", d'o "elle ne marchera pas (vraiment)" ganteveyam ganta, neyam gamisyati. Sous la pression de l'usage, ajoute Kai., le futur priphrastique s'accrdite ainsi dans une valeur secondaire (soumise la prsence virtuelle d'une particule comparative).56
54

Dveloppant Bh.r Ng. note que, selon certains, il est incorrect d'avoir le futur (simple) au sens du non-aujourd'hui, ceci n'tant valable qu'au participe (pourquoi ce privilge du participe? est-ce simplement parce qu'il n'y a pas de participe rpondant au futur priphr.? Le dcalage entre indicatif et participe est, en tout cas, un dtail retenir). Ex.: svo 'gnin dhsyamnena " par lui devant placer lesfeux demain". Bh., qui cite cet ex., obtient ce rsultat en scindant le su. 15 en ANADYATANE (a) et LUT (b): peu importe.
55 N. montre qu'on est bien dans le champ d'action de la svastani, mais SK. 2.4,13 dit adyatanrtha aro-

mbhah. L'emploi de ganta est indpendant du sens de la phrase, poursuit N., rappelant que la causeefficiente dans la formation d'un mot est son sens propre, non le sens de la phrase o il est appel figurer (cf. n. 22 ci-dessus). Pour Ng. aussi, on est dans le domaine du non-aujourd'hui, 56 II n'y a point de comparaison possible avec un verbe personnel, dit Bh. De ce vt. on peut retenir qu'il at senti un lien entre "comme" et la ngation: ceci rappelle l'emploi rgvd. bien connu de no particule tantt ngative, tantt comparante; avec raison, sans doute, Vendrys BSL. 46,1950 p. 10, enseignait t'identit foncire des deux emplois.

518 Louis Kenou

17. Un dernier faisceau (le plus massif) de rgles temporelles est celui qui commence au s. (3.3,)131 : concernant essentiellement le futur, il s'agrge sans peine au groupe antrieur. 5 7 Toutefois, le s. initial (131) aborde, de faon plus large, la question du pass proche et de l'avenir proche, notions qui comp o r t e n t t i t r e optionnel l'emploi du prsent grammatical. C'est la tendance bien connue du prsent empiter sur les zones limitrophes, sur le pass proche par continuation ou reconduction (ayam gacchmi " j e viens d ' a r r i v e r " ) , sur le f u t u r proche par anticipation (esa gacchmi " j e vais (y) a l l e r " ) . Pninidit, non pas "(l'affixe en vigueur est optionnellement) le p r s e n t " , mais " ...est t r a i t comme un prsent". 5 8 Cette manire de s'exprimer sert, d'aprs Bh., empcher que les affixes ici prescrits ne s'attachent qu' la racine pure et simple (il faut qu'ils puissent s'attacher aussi un thme verbal); mais la motivation relle du -VAT est de p e r m e t t r e d'intgrer dans la rgle les substituts du prsent, tels que le participe. C'est ce que K. rsume en disant que -VAT marque la similitude absolue 59 : quelle que soit la qualification avec laquelle les affixes o n t t prescrits en valeur de prsent (entre 3.2,123 et 3.3,1), c'est--dire quelle que soit la base, quel le membre de compos accol, quelles les conditions d'emploi ce sont ces affixes mmes et non d'autres qui sont ici valables. Il est permis de penser que les grammairiens avaient s u r t o u t en vue le LATM3, mme si quelques exemples supplmentaires sont donns avec le participe (prsent: gacchantam eva mm viddhi "sache que je viens d ' a r r i v e r " ) ou avec un driv primaire comme alamkarisnu " q u i a rcemment o r n ; qui va bientt orner ". 60 18. A partir du s. 132, le f u t u r grammatical est en j_eu_(Bh.), d'abord avec la nuance semi-modale d ' " e s p o i r " (SAMSAYAM); le verbe notant l'acte espr figure soit, comme il est normal, au f u t u r (updhyyas ced gomisyati/ ete vykaranom adhyesymahe

"si le matre venait (dans un avenir espr), nous apprendrions la grammaire" (cf. le contre-exemple gamisyati K.); soit, par une sorte de subrogation, au prsent ou l'aoriste; le vt. 1 exclut nommment l'imparfait et le parfait, parce que l'nonc en -VAT implique qu'on a affaire un temps "gnral ", non un temps parti57 La transition est l'expos sur le suffixe KHAL (et ses dpendances) 126-130, suffixe qui est proche du verbe, en ce sens que, par ex., Isatkaro bhavat confine (sat) kriyate ou karisyate bhavot. Dans CV. 1.3,106, khal joint yuc, est englob sans discontinuit dans la description du futur: c'est pousser l'extrme la disposition de P.

non la constatation impersonnelle, 60 N. discute une fois de plus sur l'opposition qui peut se produire entre pada et vkya-samskra (cidessus, n. 22): il faut appliquer le temps requis par la phrase au moment mme o l'on forme le mot, sinon l'on obtiendrait le groupement fautif svah karisyati au lieu de svoh kart (groupement que semble admettre pourtant K. elle-mme, cite n. 22, et plus tard H.). CV. 1.3,106 explique que ce qui est proche du prsent n'est autre qu'un prsent, car les fruits de l'acte y sont reconduits; le futur de mme, parce qu'on met en uvre (au moment o l'on parle) les moyens propres atteindre ces fruits. Incidence grammaticale de la doctrine philosophique du kormanl

On a donc affaire l'un de ces s. d'extension ou, pour mieux dire, de transfert (atidesa, Termin, s.u.), dont l'interprtation est souvent d'autant plus malaise retrouver que les ce. la chargent d'enseignements latents. 59 Donne reprise SK. 2.4,187 H. 5.4,1 Prakr. Kaum. 3.2,131 SK. 2789, e tc. Noter, est-ce un hasard?, que les ex. sont emprunts au discours,

58

519 Thorie des temps du verbe

cularis.61 Nous sommes donc en plein flottement temporel, ce qui n'a rien d'tonnant puisque ces phrases nuance d' "espoir" sont, si l'on en juge par les exemples des commentaires, des hypothtiques, correspondant celles des s. 8 et 9 dj vus.62 On notera, une fois de plus, le soin que prend P. viter de donner directement la structure de la phrase hypothtique; en matire de syntaxe, ce qui lui importe est la nuance psychologique appelant telle ou telle structure. Le s. 133 est une annexe au prcdent: le voisinage de l'adverbe ksipram (ou d'un mot analogue) entrane le retour exclusif du futur, autrement dit du temps normal.63 Pareillement (134), l'optatif est seul possible si la nuance d' "espoir" est inscrite dans la phrase mme par un mot prcis (les commentaires citent les formes verbales samse, avokalpaye, kmaye, non suivies de yod). De notre point de vue, c'est 132 qui est une exception 133-134, du point de vue des grammairiens 133-134 sont des exceptions 132. Le vt. 2 (poursuivi par 3) connat un emploi du temps "achev " pour noter un acte qu'on admet par avance comme ralis64 ; au lieu de dire "s'il pleut, le riz sera abondant" devas ced vrstah sampatsyonte slayah, on dira ainsi "...est abondant" sampannh. Le riz tant encore en pi I let, on s'attend bel et bien qu'il soit battu (Bh. ad vt. 4). L-dessus, Bh. (vt. 5) dveloppe la thse suivant laquelle toutes les dsinences valent au sens du prsent quand il s'agit d'un verbe d'existence: le sujet parlant sait quoi s'en tenir
61 Id SK. 2790; la Bhsv. ajoute le futur priphr. et l'optatif et cite un ex. littraire plutt hardi, mam upyamsto "(j'espre qu'elle) m'pousera". H. 5,4,2 donne: aoriste, verbal en -to-, prsent, les deux futurs. 62

Q u e l sens d o n n e r SAMS "espoir", par rapport hSAMBHVANA (154) "supposition", se demande le vt. 3? Le vt. 4 tablit qu'sams fait partie intgrante de sarn, le premier terme tant susceptible ou non de ralisation, le second tant seulement ralisable (si tel est bien le sens de [on]abh'mJta, mot que Kai. rend par " capable (ou incapable) d'tre atteint par l'appropriation des causes '') ; Vsams est pradhrit " rifie par la pense qui se dit: puisse cela m'arriver!", le sam est apradhrita (si telle est bien la leon) = " indtermin par nature". Notable est l'emploi possible du verbal en -ta- comme substitut du futur, soit, dans le cadre syntaxique cit, gatah et adhitavantah. K.: cet emploi (comme, dj, celui de l'aoriste) montre que, dans le domaine de la subordination, les formes verbales sont rgies par l'intention gnrale de la phrase et perdent

pour ainsi dire leur valeur propre: ainsi l'imparfait fr., au sens futur, en proposition hypothtique. 63 H. 5.4,3 connat le prsent, l'aoriste, les deux futurs; il imagine un conflit avec P. 15 (coexistence de svah et slghram), d'o rsulte la prvalence du futur simple. Le vt. 1 note, quant lui, le conflit ventuel entre 133 et 134:134 l'emporte en vertu du principe "ce qui est nonc aprs l'emporte (sur ce qui est nonc avant)", l'une des paribhs fondamentales de la Grammaire. Pour K. l'nonc LRT vise indiquer que le futur simple couvre aussi le domaine du futur priphr.: tel est le risque qu'entranent les enseignements implicites, d'abolir par endroits les rgles les mieux tablies. SK. met mme cette disposition en s. (2.4,191): pour un acte non ralis, on emploie les mmes affixes que pour un acte ralis, quand il y a sarpbhvan (glos saktisraddhna "croyance en une possibilit de ralisation "), avec l'ex. (que donne aussi H. 5.4,4) "s'il faisait effort en temps utile, le succs se produirait", samaye cet prayatno 'bht, udabhvan vibhtayah.
64

520 Louis Renou

quand il a vu une fois un puits; mme s'il n'observe pas de dnivellation, il n'en ralise pas moins en pense l'existence permanentede ce puits et il d i t " // y a (ici) un puits". 6 5 En fait, t o u t e dsinence est employe avec le temps qui convient, pourvu qu'elle ne soit pas enseigne t i t r e optionnel ou que l'usage inverse ne soit pas acquis (par une autre rgle): personne ne dira " il fut un p u i t s " kpo 'bht, quand il faudrait dire " il y a un p u i t s " kpo 'sti.66 L'activit des sens donne simplement un contact, c'est la conscience qui dcide. Q u e l qu'un voulant aller Ptaliputra d i t " s u r le chemin menant P il y aura un p u i t s " yo 'yam adhvgantavya ptaliputrd etasmin kpo bhavsyati (futur simple); quand il aura march et envisagera un temps (prcis) du non-aujourd'hui, il dira " il y aura l un puits ( d e m a i n ) " bhavit (futur priphr.); quand il aura atteint l'endroit, " il y a un puits " asti (prsent); le point une fois dpass, " il y avait un puits " abht (aoriste); la nuit une fois coule, " il y avait l un p u i t s " sh (imparfait); lachse oublie, " i l y eut l un p u i t s " babhva (parfait). Les dsinences sont ce qu'est l'activit des sens ; c'est l'activit de la conscience qui dcide qu'on a affaire un prsent. 19. Une rgle d'apparence dconcertante est 135, enseignant aoriste et f u t u r (simple) aux lieu et place de l'imparfait et du f u t u r priphr. quand il y a rendre un procs continu (KRiYPRABANDHA) ou une p r o x i m i t (dans le temps, SAMPYA). Exemples: a) " il adonn (ou : donnera) de la n o u r r i t u r e sa vie d u r a n t " yvajjvam annam adt (dsyatr, K. prcise la nuance en ajoutant bhrsam); b) yeyam paurnamsy atikrnt (ou : yeyam mvsygminl), etasym updhyyo 'gnJn dhita (ou : dhsyate) " lors de la pleine lune prcdente (ou : de la nouvelle lune prochaine), le matre a dispos (ou : disposera) les f e u x " . C'est donc une entorse faite l'expression normaledu non-aujourd'hui. Le SMPYA a cet intrt pour nous de dgager un aoriste de constatation c'est un peu ce que nous attendions, mais en vain, en lisant 3.2,111 sans le secours des commentaires. 67 D'autre part, le PRABANDHA tend heureusement le
65

Ceci repris (en forme de su.!) SK. 192, d'o suit qu'on dira indiffremment marge kpo 'bht, slt, babhvo, etc.; de mme H. 5.2,19 toutes dsinences sont valables avec les verbes d'existence. C'est se limiter la position premire de Bh., sans tenir compte du siddhnta. 66 Cf. Bh. ad 1.4,80 vt. 4 "nul ne dit pocoti pra, quand il faudrait dire pra pacati", d'o suit que la Grammaire n'est pas faite pour interdire ce qui par nature est tranger l'usage, prayuktnm eva laksanennvkhyant "car seules les choses en usage sont expliquer par voie de rgle" Kai. ad 1.1,24 fin. Sage prcepte, que les commentateurs ne semblent pas avoir toujours fidlement suivi ! 67 II est vrai qu'un aoriste de continuit semble contradictoire avec l'ide que nous nous faisons de ce temps: il faut y voir un aspect de l'aoriste notant un pass fait d'in-

stants rpts, allant jusqu' l'actuel. La rgle, une fois de plus, est soumise la vivaks CV. 1.3,106. Quant aux deux ngations qu'elle contient (NA et AN), Bh. observe que l'enseignement inverse se trouve par l mme impliqu, savoir que, en cas de futur ou de pass d'aujourd'hui, on devra employer les temps prescrits; ainsi, il n'est pas redouter qu'on ait le LRJ dans le domaine du LU, le LU dans celui du LRT. N. dit: si P. avait dit que, dans les circonstances nonces, le traitement serait conforme celui d'un pass ou d'un futur d'aujourd'hui, le prsent s. serait prescriptif et il y aurait confusion des temps; mais, en fait, le s. est prohibitif simplement: il prohibe les temps du non-aujourd'hui et eux seuls; ce point acquis, les temps normaux valent chacun en son domaine. Selon certains, dit encore H. 5.4,5, ce s.

521 Thorie des temps du verbe

d o m a i n e du f u t u r simple, d o n t nous savons assez par l'usage l i t t r a i r e q u ' i l est p a r f a i t e m e n t c o m p a t i b l e avec une datation rapp r o c h e ou avec l'ide d ' u n e c o n t i n u i t . Le s. suivant (136) d o n n e un a u t r e cas o le f u t u r (simple) remplace le f u t u r p r i p h r . a t t e n d u par 1 5 : c'est, savoir, si'l y a m e n t i o n d ' u n e l i m i t e en de de laquelle se situe l'acte, ex. " n o u s mangerons d e u x fois du riz en de de Kausmb sur la r o u t e

menant Ptaliputra" yo 'yam adhv gantavya ptaliputrttasya yad ovarom kausmbys tatro dvir odanam bhoksymahe (et non : bhoktsmahe). Cet empitement peut surprendre: il s'explique justement parce que la prcision temporelle est suffisamment marque dans la phrase. Au s. 136 s'annexe naturellement 137, qui contredit l'enseignement prcdent quand la limite est de caractre temporel (non plus spatial) yo yam samvatsara gml tatra yad avaram grahyanys tatra yukt adhyesymahe " nous tudierons ensemble dans l'anne qui vient (jusqu' la priode qui est) en de de l'AgrahayanT". 138 laisse le choix entre l'un et l'autre futurs quand ladite limite (de temps) est celle au-del de laque-Ile l'action se situe (tasya yatparam gra0). Le vt. 1 ad 136 prcise que ledit s. (videmment aussi 137 et 138) ne comporte pas la validit rcurrente de KRIY0135. 20. Intervient ce point le conditionnel (139). Le conditionnel est trait comme un temps du futur et appel cet endroit de la Grammaire par l'allusion aux phrases hypothtiques que contiennent les s. 132 134. Il se dfinit par le fait que le sujet passe outre l'action (KRIYTIPATTAU)6*: c'est l une manire indirecte, plus approprie peut-tre rendre l'intention psychologique, de noter ce que nous dsignons par " irrel ". Le Bh. propose d'noncer "outrepassement (non de l'acte, mais) des instruments de l'acte", puis il rejette cette modification, car tout outrepassement des moyens propres raliser un acte contient ncessairement t'outrepassement de l'acte lui-mme. On notera, une fois de plus, que Panini ne dfinit pas en termes grammaticaux la phrase irrelle; il renvoie seulement par avance (LINIMITTE) la mention HETUHETUMATOH156, laquelle ne dfinit pas davantage une structure, mais invite reconnatre l'existence d'une relation de cause effet. A nous d'insrer cette mention dans un cadre syntaxique; comme dit K., l'outrepassement se laisse comprendre par la phrase (non par les termes mmes d'un s.).69 Dans l'exemple "s'il appelait ( l'aide) Kamalaka, son chariot ne se renverserait pas" yadi komalakam hvsyan na sakatam parybhavisyat, l'appel fait K(ou plutt: l'appel que le sujet outrepasse l'occasion de faire K) est la cause (hetu), le non-renversement du chariot est l'effet (hetumant). Le

interdit le parfait et autres temps, mme s'ils sont prescrits par une autre rgle, en tant qu'indiquant spcialement le non-aujourd'hui. 68 Glos anabhinirvrtti K.; H. 5.4,9 du fait de l'autorit que revtent les notions de cause et d'effet, le sujet parlant, ayant compris qu'il doit se produire une non-ralisation de l'acte, dit "s'il appelait. . .". Noter que le conditionnel chez. P. est dfini au moyen d'un nom d'action, alors que les autres valeurs temporelles

sont dfinies par des participes ou des noms verbaux; c'est bien ce qui montre l'attache du conditionnel aux modes, car les modes eux aussi sont dfinis par des noms d'action indiquant des catgories psychologiques. 69 Toutefois, K. dit ity evamdikam, ce qui peut viser 157; pour 140, elle allgue la suite du vt. 1 l'optatif 152 "et suivnt(s) ". Durgh. 139 met l'emploi de syat au compte de la vivaks (ci-dessus, n. 22).

522 Louis Renou

sujet, dit N. (prcisant K.), ayant vu par des appels antrieurs rpts faits K que le char ne se renversait pas, ayant reconnu qu'un appel f u t u r K serait galement la cause du non-renversement (ayant donc vu l, comme d i t K., un rapport entre signe et chose signifie), comprend par quelque vidence extrieure (comme le fait que K va dans un autre pays, etc.) qu'il y a "outrepassem e n t " de l'appel, et prononce la phrase en question ; K. considre que si le cocher s'abstient ainsi d'appeler, c'est " par suite de quelque dfectuosit m e n t a l e " (yaigunya). Le s. 14070 transporte au pass le cadre prcdemment acquis, donc " s ' i l avait appel K, le chariot ne se f t pas r e n v e r s " . Ceci n'est pas une modification accessoire du s. 139, c'est bien p l u t t la reconnaissance du caractre eminent du conditionnel : on sait assez, par l'usage littraire, que ce mode est pratiquement rserv l'expression d'un irrel du pass. Ainsi dfini, le conditionnel (selon 140) est cens valoir, entre 143 et 152, en concurrence aux temps directement enseigns auxdits s., c'est--dire s u r t o u t l'optatif, parfois au prsent et au f u t u r : il suffit, pour qu'apparaisse le conditionnel, que l'acte soit conu comme " o u t r e - p a s s " . Ceci largirait singulirement la zone de cette f o r m a t i o n , la ferait dborder hors du cadre des phrases hypothtiques. De fait, le conditionnel dans la prose vdique n'est pas limit de telles phrases, il dpend moins d'une structure que d'une notion rendre. 7 1 Les commentateurs sont donc justifis enseigner le conditionnel entre 143 et 152; t o u t ce qu'on peut dire, ici comme ailleurs, est que ces emplois obtenus par le jeu de la " c o n c u r r e n c e " ont moins de poids que les enseignements sortant directement d'un s. ; ils sont sujets quelque automatisme. 72 Mais le principe en est difficilement contestable. 21. Le s. 142, faisant r e t o u r au temps prsent (LA7), instruit ce ternps lorsque la phrase comporte une ide de " b l m e " (GARHYM), l'exemple de base tant " s e peut-il que vous fassiez faire un sacrifice par un sdra (acte minemment blmable)?" api (ou \jotu) tatrabhavn vrsalam yjoyati/gorhmahe. A vrai dire, ce s. n'est valable que si l'on rejette l'hypothse que le s. VARTAMANE _A7s'appliquerait dj noter le prsent comme temps indiffrenci. Les particules api etjtu interviennent ici (selon le processus dj observ) pour autoriser par leur seule prsence une forme du verbe moins marque qu'elle l'et t en leur absence. 73
140 succde 139 uniquement parceque139esten dpendance des s. antrieurs, mais P. n'a pu manquer de raliser que le conditionnel du pass est incomparablement plus important que celui du futur. 71 II manque une tude sur le conditionnel, mais les indications de Del brck Ai. Syntax p. 365 pour la prose vdique (ce mode manque prsque totalement dans les mantra) montrent la souplesse de la formation au stade ancien; plus tard (Speyer, etc.) vient une sorte de normalisation, concidant avec une rarfaction. 72 En fait, 146 est exclu, ainsi que 151 (K.); cf. CV. 1.3,108 116 (bhte livisaye). Le conditionnel futur est donn
70

comme ncessaire (nitya) pour 143145,147-150,152; le conditionnel pass comme optionnel (v) pour 144, 145,147-150 (mais comme ncessaire pour 152); cf. Prakr. Kaum. 3.2,143 Ira1 bhavisyati nityam, bhte va. Voici quelquesexx. dans K., emprunts au type vrsalam yjayati( 21) :konma vrsalo yam tatrabhavn ayjayisyat 144, nvakaipaymi tatrabhavn nma vrsalam ayjayisyat 145 (plus loin K. renonce aux exx. de conditionnel), 73 Bh. ad vt. 1 note que le su. tait inutile, car il s'agit l d'une action non acheve, o le prsent est valable par dfinition. Dira-t-on queces. sert empcher les substituts du prsent (participe, etc.) d'y valoir? Non, vu

523 Thorie des temps du verbe

143 et 144sont des corollaires 142: il y a concurrence entre prsent et optatif si la phrase comporte Tinterrogatif kotham " c o m m e n t se peut-il que . . . ? " ; concurrence entre optatif et f u t u r s'il y a le pronom ko ou ses drivs (KIMVRTTE). Les mots i n t e r r o gatifs sont moins slectifs que les particules, ces dernires imposant un temps prcis, ceux-l permettant un choix (cf. dj une rpartition analogue aux s. 5 et 6, ci-dessus 14). 74 Suivent (145) deux nuances accompagnant l'emploi optionnel de l'optatif et du f u t u r : c'est, savoir, si le sujet considre l'acte, soit comme impossible (ou du moins, invraisemblable, sur le plan moral), soit comme intolerable: les exemples m o n t r e n t qu'on est toujours en face d'une variante 142, " j e ne crois pas possible ou tolerable que vous fassiez faire un sacrifice... " nvokolpoymi (ou : no morsoymi) tatrabhavn nomo vrsolom y joy et (ou : yjayisyoti) : le mot ANAVAKPT est rendu par osrad'dh C. 1.3,111, plus expressif que osombhvono K. et correspondant, d'ailleurs, au no sroddodhe de l'exemple affrent; de mme krodha C. est plus prcis que Vaksam K. glosant P. AM ARS A. La prsence d'un mot adjacent tel que kimkilo ou d'un verbe d'existence ( valeur attnue, vidyote nomo " s e peut-il que... ?") suffit selon 146 carter l'optatif: on en conclura que, dans t o u t ce dveloppement, l'optatif tait bien le mode normal, dont le f u t u r est un substitut conditionn par le contexte. 7 5 La p o r t i o n qui suit concerne l'optatif seul et ne nous intresse pas ici. 76 Le f u t u r reparat au s. 151, comme substitut de l'optatif (selon 150), l o l'ide prvalente est celle d'une chose " t o n n a n t e " (OTRKARANE), mais sans qu'il y ait de mot adjacent ( savoir, sans la conjonction yac co ou yotro prvue pour 150, quoi s'ajoute encore yodi),77 donc en subordonne implicite (comme le sont d'ailleurs les phrases cites sous 145 et. 146), ex. " i l serait surprenant qu'un sourd apprt la g r a m m a i r e " , scoryom / badhiro nomo

vykoronom odhyesyote. Le s. 155 introduit un futur optionnel quand l'ide est celle d'une chose qu'on "suppose" (SAMBHVANA), mais en fait, d'aprs les exemples cits, d'une chose qu'on prsume devoir se produ ire et mme sur laquelle on compte,78 ainsi aprs sombhvoymi, ovoque ces substituts sont souhaitables, ainsi api mrn yjayantarn posyo " me vois-tu en train de faire faire un sacrifice , . . ? " . 74 H. 13 ( = P. 143) donne ici "tous les temps ", ainsi kotham bhaksayet bhaksayoti ababhaksat abhaksayat bhaksaymcakra bhaksayit bhoksayisyati obhaksoyisyat. Au contraire, le s. suivant (H. 14) entrave toutes dsinences, il ne reste que bryt vaksyati avoksyat. De mme 145 (H. 15). Ganaratn. p. 12. 75 kimkila Ganaratn. p. 11-12. II y a cette restriction que les exx. sous 145 comportent, non seulement l'ide d'une chose non possible ou non tolerable, mais les formules expresses nvakalpaymi et analogues (tout comme 155 et cf. ci-dessus. 18) "je ne crois pas, je ne supporte pasque", sans l'intervention de la conjonction ya; qu'ensuite, d'aprs 147, la partieu le jtu suivie de yod maintient l'optatif (le vt. 1 adjoint yad yoda et yadi). On remarquera l'importance de la subordination, implicite dans tous ces schmas (depuis 145); c'est l surtout o apparat le libre choix des formes temporelles. 76 147 entrave le futur (H. 17). 77 Mais AYADAU est inutile selon le vt. 1, car avec yadi il n'existe pas de nuance d'tonnement, mais simplement de supposition. 151 vaut sorvesu kd/esu d'aprs SK. 2.4,208, mais entrave tous autres temps d'aprs H. 20. yac co cit Ganaratn. p. 10. 78 "Croyance en la capacit ou possibilit (d'agir), facult de discriminer une convenance dans les actes (faire)" K.

524 Louis Renou

kalpayami, sraddadhe (...bhujJta bhavan / bhoksyate bhavan " j e p r sume qu'il mangera" : toujours avec subordination implicite): si la conjonction yad intervient, on retombe sur le tour normal, c'est-dire l'optatif.79 22. Dj traite indirectement sous 132-134 (ci-dessus, 18), 139-140 (20), la phrase hypothtique est aborde de front au s. 156; laprotase est appele HETU "cause", l'apodose HETUMANT "effet", 80 termes qui font bien sentir le lien troit entre les deux parties du diptyque, mais n'accusent pas a priori l'aspect syntaxique. L'optatif y este le mode habituel, mais, s'agissant d'un fait venir car ce s. continue tre sous la rcurrence de BHAVISYATI comme tous les s. environnants le temps futur peut tre employ.81 Les exemples allgus montrent bien qu'il s'agit d'un " potentiel " corrlatif l'irrel de 139, "s'il fait appel Kamalaka, son chariot ne se renversera pas " yadi kam a la kam hvayen na sakatam parybhavet (ou : hvsyati... parybhavisyati): la thorie du futur s'achve sur cet encadrement de formules parallles, qui donne la description son style propre. Cette thorie est relativement longue, compare celle du prtrit ou du prsent. On notera toutefois que le futur n'est enseign, le plus souvent, qu' titre optionnel : concurremment au prsent 5-8, au prsent et l'optatif 9, de nouveau au prsent 131-132, au futur priphr. 138, au prsent et l'optatif 143, enfin l'optatif seul 144-145,155-156, sans compter quelques cas conditionns. Ainsi le futur pninen est essentiellement un temps facultatif, enclin empiter sur le mode ou se laisser empiter par lui.82 C'est bien l'impression que nous en donnent aussi les textes classiques. Cette fin du troisime pda marque aussi le terme de la thorie d'ensemble des valeurs temporelles. Demeure seul un emploi secondaire, un peu part, emploi qu'incitait placer ici la mention du prcatif (173 et 174): c'est la question du prohibitif. Elle occupe deux s. fort brefs, 175 et 176: on y apprend le choix que laisse la Grammaire entre l'aoriste (inaugment, 6.4,74) avec ma et l'imparfait (id.) avec ma sma, ce dernier doubl par l'aoriste (donc, ma krslt/m sma karot ou krst "qu'il ne fasse pas! " ) : la forme verbale la moins expressive choisit ladouble particule. Il est difficile
79 II y a prsent implicite dans le domaine de 153, le mot adjacent tant kaccid. Futur tmplicitedans ledomaine el e 154, le mot adjacent tant alam : ci encore, la particule entrane avec elle une forme verbale moins marque. 80 phola C. 1.3,120 SK. 2.4,213 et ailleurs; hetumant Smkhyakr. 10 "ayant une cause". 81 Les autres temps sont exclus K., mais SK. loc. cit., donne yyt ysyati yti aysit, avec les correspondants dans l'apodose. D'aprs CV., repris K., la rgle ne s'applique pas un emploi tel que varsatlti dhvati "tant donn qu'il pleut, il court", o il ne s'agit pas du futur et o, note CV.r iti entrave l'optatif. N. pose la vivaks (ci-dessus, n. 22), qui fait prfrer le

le phalo sont suggrs par iti, si bien que le prsent sufft. C'est un cas tout analogue celui dedraksymt, cidessus, 15. 82 Si le futur tend au mode (Delbrck Ai. Syntax p. 289 Idg. Syntax 2 p. 243 Brugmann 22.3 p. 784), peut-tre parce qu'il est l'indicativisation d'une ancienne forme modale, I arrive que l'optatif tende au prsent : ainsi cchet donn comme quivalent d'icchati 160 : c'est dans ce cas trs limit que les Gramm, ont aperu la vaste quivalence entre optatif et indicatif, quivalence qui d'ailleurs est surtout sensible aprs l'poque vdique. Un autre cas galement trs spcial atteste une concurrence entre impratif et indicatif (ci-dessus, n. 19).

prsent; enfin, SK. dit que le hetu et

525 Thorie des temps du verbe

de mesurer quel degr les textes confirment cette rpartition ; de toutes manires, et ds l'origine, l'aoriste est f o r t e m e n t prdominant. 8 3 23. Un regard, mme rapide, sur t o u t cet enseignement, porte croire que Pan ni a not exactement, bien que, en partie, de manire rudimentaire, les valeurs temporelles du verbe; que, ct de rgles trs comprhensives, il a pris soin de relever des dtails f o r t menus, souvent insaisissables pour nous (peut-tre simplement parce qu'ils appartiennent un usage oral ou familier dont nous n'avons plus de tmoignage). Cette disparit dans la prsentation des faits se retrouve, du reste, dans d'autres portions de la Grammaire. Mais l'cole pninenne, prise dans son ensemble, fortifie et l par les faits nouveaux qu'enregistrent les nonPninens (notamment Hemacandra, le plus novateur de tous, du moins dans le domaine qui nous a retenu ici), rtablit un certain quilibre. C'est ainsi que Patajali a conserv de prcieuses d o n nes, avec un arrire-plan philosophique, sur la signification du " p r s e n t " ; il nous a permis de voir plus distinctement quelles taient, selon Panini mme, les dlimitations entre les temps du pass. Il demeure, certes, une disproportion entre le f u t u r et les autres formations personnelles, mais il faut voir que le f u t u r est en partie un substitut de mode et que, au surplus, la plupart de ses emplois sont donns t i t r e de " c o n c u r r e n c e " . Peut-tre les commentateurs ont-ils pouss jusqu' l'arbitraire une tendance qui n'tait pas ce degr dductible des stra; ils o n t cherch, l comme ailleurs, galiser et balancer les enseignements, m u l t i plier les indices de reconduction. Il est vrai que, dans le chapitre de la syntaxe verbale, ils avaient des excuses: rien n'est plus flou, surt o u t postrieurement l'poque vdique, que l'usage des formes verbales dans la l i t t r a t u r e ; il faudra t o u t e la rigueur, l'ascse grammaticale du kvya, pour y m e t t r e quelque o r d r e . Parmi les observations retenir, notons le rle modificateur (attnuateur) a t t r i b u certaines particules. Le parti pris de m e t t r e en avant la fonction et non la catgorie, a t salutaire pour l'unit de l'expos et pour sa fermet. Il a permis de dgager plusieurs notations psychologiques, qui o n t t des facteurs de syntaxe. Il y a donc chez les grammairiens un ensemble de faits dont l'interprtation n'est pas toujours facile, mais qui enrichit notre connaissance de la langue, qui rend justice aussi cet instinct d'harmonie interne qu'on dirait avoir prsid sa s t r u c t u r a t i o n .
83

K. cite comme incorrect donc, bien tabli dans l'usage l'emploi de ma bhovatu, ma bhavisyati. Les Gramm. soucieux de conciliation le justifient en posant une particule m distincte du MAN de P.; cet artifice est repris, entre autres, par Kt. 3.1,22 SK. 2219 H. 5.4,39 Durgh. (ad 175). BM. ajoute que, dans mstu, ostu est lui-

mme une particule apparence (pratirpaka) deforme verbale. Vop. 25.27 met sur le mme plan m virarnsit/ m viramotu/ m viramsyati. Kt. admet aussi "selon certains " l'interversion sma karon m. Sur l'usage vd., v. BSL. 43,1946 p. 46, n. 2 JAs. 1959 p. 74; plus anciennement, Delbrck Ai. Syntax p. 358, 361, 501.

Bibliography

Aklujkar, Ashok(1969). "Two Textual Studies of Bhartrhari,"

Journal of the American Oriental Society 89, 547-563. Allen, W. S. (1953). Phonetics in Ancient India. London. Aufrecht, Th. (ed.) (1859). Ujjvaladatta's Commentary on the UndiStras. Bonn. Ballantyne, J. R., and the pandits of the Benares College (eds.) (1856). The Mahbhshya with Its Commentary, the Bhshya-pradpa and the Commentary Thereon, the Bhshya-pradpodyota, vol. I. Mirza-

pore.
Beal, Samuel (trans.) (1885). Si-yu-ki. Buddhist Records of the Western World, by Hsan Tsang, vol. 1, Boston. See pp. 6-7. (1911). TheLifeofHiuen-Tsiang, by Hwui Li, London. See pp. 7-10. Belvalkar, S. K. (1915). An Account of the Different Existing Systems of Sanskrit Grammar. Poona. Benfey, Theodor (1874). Einleitung in die Grammatik der vedischen Sprache. Gttingen. Bhandarkar, R. G. (1864,1877). " Review of Goldstcker's Pnini, His Place in Sanskrit Literature,1' Indian Antiquary 6,108-113. See pp. 72-78. (1872). " O n the Date of Patajali and the King in whose Reign He Lived," Native Opinion, 1864, and Indian Antiquary 1, 299-302. See pp. 78-81. . (1873a). " Note on the Above " [See Weber 1873c], Indian Antiquary 2, 59-61. See p. 84. (1873b). "Reply to Professor Weber "[Weber 1873c], Indian Antiquary!, 238-240. See pp. 84-85. (1876). "crya, the friend of the student, and the relations between the three Acryas," Indian Antiquary 5, 345-350. See pp. 86-87. (1883-1885a). "The Date of Patajali. A Reply to Professor Peterson," Journal of the Bombay Branch of the Royal Asiatic Society 16,199-222. (18831885b). "Development of Language and of Sanskrit." Journal of the Bombay Branch of the Royal Asiatic Society 16, 245-274. See pp. 87-93. (1883-1885c). " N o t e " [on Peterson 1883-1885], Journal of the Bombay Branch of the Royal Asiatic Society 16, 343-345. (1883-1885d). " Relations between Sanskrit, Pali, the Prakrits and the Modern Vernaculars," Journal of the Bombay Branch of the Royal Asiatic Society 16, 314-342. See pp. 94-101. (1889). "My Visit to the Vienna Oriental Congress," Journal of the Bombay Branch of the Royal Asiatic Society 17', 72-95. Also in N. B. Utgikar, ed., Collected Works of Sir R. G. Bhandarkar 1933, vol. 1, 332-360, Poona. See pp. 81-83. Bhattacharya, B. (1962). A Study in Language and Meaning: A Critical Examination of Some Aspects of Indian Semantics. Calcutta. Biardeau, M. (1964a). Thorie de la connaissance et philosophie de la parole dans le brahmanisme classique. Paris/La Haye.

527 Bibliography

(1964b). Bhartrhari VkyapadJya Brahmaknda avec la vrtti de Harivrsabha. Paris. al-Brun!. See Sachau (1887) and (1910). Birw, Robert (1966). Studien zu Adhyya III der AstdhyyJ Pninis. Wiesbaden. Bloch, Bernard (1949). "Leonard Bloomfield," Language25, 87-98. Reprinted in Sebeok 1966, 11,508-518. Bloomfield, Leonard (1927). " O n Some Rules of Pnini," Journal of the American Oriental Society 47, 61-70. See pp. 266-272. (1933). Language. New York. (1939). "Linguistic Aspects of Science," International Encyclopedia of Unified Science ( O t t o Neu rath, ed.) 1, no. 4, Chicago. Boehtlingk, Otto (ed.) (1844). Die Undi-affixe. St. Petersburg. (ed.)(1847). Vopadeva's Mugdhabodha. St. Petersburg. (1851). ber die Sprache der Jakuten. St. Petersburg. (1876a). " Ktjjanaoder Patagali im Mahbhshja," Zeitschrift der deutschen morgen/indischen Gesellschaft 29,183-190. (1876b). " Das Verhalten der drei kanonischen Grammatiker in Indien zu den im Wurzelverzeichniss mit s und n anlautenden W u r z e l n , " Zeitschrift der deutschen morgenlndischen Gesellschaft 29, 483-490. (ed., trans.) (1887). Pnini1 s Grammatik. Leipzig (the edition of the text first appeared in 1839; reprint Hildesheim 1964). (1893). " W h i t n e y ' s letzte Angriffe auf Pnini," Berichte ber die Verhandlungen der kniglich schsischen Gesellschaft der Wissenschaften zu Leipzig, Philologisch-Historische Classe 45, 247-257. See pp. 186-192. and Rudolf Roth (1855-1875). Sanskritwrterbuch, vols. 1-7. St. Petersburg. Bopp, Franz (18' (1816). ber das Konjugationssystem der Sanskritsprache. Frankfurt Main. urt am M. . (1827). Ausfhrliches Lehrgebude der Sanskrita-Sprache. Berlin. (18673). Glossarium comparativum linguae sanscritae in quo omnes sanscritae radices et vocabula usitatissima explicantur et cum vocabulis graecis, latinis, germanicis, lituanicis, slavicis, celticis comparantur. Berlin. Boudon, Pierre (1938). " U n e Application du raisonnement par l'absurde dans l'interprtation de Pnini (les jnpakasiddhaparibhs)," Journal asiatique 230, 65-121*. See pp. 358-391. Breloer, B. (1929). "Studie zu Pnini," Zeitschrift fr Indologie und Iranistik 7,114-135. Brough, John (1951 a). Selections from Classical Sanskrit Literature. London. (1951b). "Theories of General Linguistics in the Sanskrit Grammairians," Transactions of the Philological Society 27-46. See pp. 402-414.

528 Bibliography

(1952). "Audumbarayana's Theory of Language," Bulletin of the School of Oriental and African Studies 14, 73-77. (1953). " S o m e Indian Theories of Meaning," Transactions of the Philological Society 161 -176. See pp. 414-423. Bhler, Georg (1894). " T h e Roots of the Dhtuptha N o t Found in L i t e r a t u r e , " Wiener Zeitschrift fr die Kunde des Morgenlandes 8,17-42. Also in Indian Antiquary 23,141-154, 250-255. See pp. 194-204. Buiskool, Herman E. (1934). Prvatrsiddham: Analytisch onderzoek aangaande het systeem der TripdJ van Pn i ni's AstdhyyJ. Amsterdam. (1939). The TripdJ, being an English recast of Prvatrsiddham (An analytical-synthetical inquiry into the system of the last three chapters ofPnini's AstdhyyJ). Leiden. Burnell, A. C. (1875). Essay on the Aindra School of Sanskrit Grammarians. Mangalore. Burrow, T. (1955). The Sanskrit Language. London. Cannon, Garland H. (1958). " S i r W i l l i a m Jones' Persian Linguistics," Journal of the American Oriental Society 78, 262-273. Reprinted nSebeok1966,l,36-57. Cardona, George (1967). " Pan in i's Syntactic Theories," Journal of the Oriental Institute (Baroda) 16, 201-215. (1969). "Studies in Indian Grammarians I : The Method of Description Reflected in the Sivastras," Transactions of the American Philosophical Society 59, part 1, 3-48. Carey, W . (1806). A Grammar of the Sungskrit Language/ . . . / t o Which are Added Examples for the Exercise of the Student and a Complete List of the Dhatoos, or Roots. Serampore. Chatterji, K. C. (1934a). " T h e Siva Stras." Journal of the Department of Letters, University of Calcutta 24 (2), 1-10. (1934b). " T h e Critics of Sanskrit G r a m m a r , " Journal of the Department of Letters, University of Calcutta 24 (3), 1 - 2 1 . See pp. 287-297. . (1948). Technical Terms and Technique of Sanskrit Calcutta. (ed. and trans., 1957 2 ). Patanjali'sMahbhsya: Calcutta. Grammar. Paspashnika.

Chatterji, S. K. (1951). " A l - B l r n l a n d Sanskrit," AI-BJrnJCommemoration Volume, AH. 362-A.H. 1362, Calcutta, 83-100. Chomsky, Noam (1964). Current Issues in Linguistic Theory. The Hague. (1966). Cartesian Linguistics. N e w Y o r k / L o n d o n . Colebrooke, H. T. (1803). " O n the Sanskrit and Prcrit Languages," Asiatic Researches 1,199-231. See pp. 33-45. (1805). A Grammar of the Sanscrit Language, vol. 1. Calcutta. (1823). "Essay on the Philosophy of the Hindus," Transactions of the Royal Asiatic Society 1 : 1 9 - 4 3 , 92-118, 439-466,

549-579:2:1-39.

529 Bibliography

Dasgupta, Surendranath (1922). A History of Indian Philosophy, v o l . I. Cambridge. Delbrck, Berthold (1904-1905). " O t t o Bhtlingk," Indogermanische Forschungen 17,131 -136. Reprinted in Sebeok 1966, 1, 261 -268. Dove, Alfred (1881). " H u m b o l d t , " Allgemeine deutsche Biographie 13, 338-358. Reprinted in Sebeok 1966,1, 71-101. Edgerton, F. (1928). " S o m e Linguistic Notes on the Mmnsa System," Language4,171-177. (1929). The MJmns-Nyya-Praksa and London. or Apadev. N e w Haven

Eggeling, Julius (1887). Catalogue of the Sanskrit Manuscripts in the Library of the India Office, v o l . I, part I, Vedic Literature. London. Eliade, Mircea (1954). Le Yoga, Immortalit et libert. Paris.

Emeneau, M. (1955). " I n d i a and Linguistics," Journal of the American Oriental Society 75,145-153. Faddegon, Barend (1929). " T h e Mnemotechnics of Pnini's G r a m mar I : The Siva-Stra," Acta Orientalia 7, 48-65. See pp. 275-285. (1936). Studies on Pnini's Grammar. Amsterdam. FaTsang. See van Gulik (1956). Filliozat, J. (1937). " Une grammaire sanscrite du XVIIIe sicle et les dbuts de l'indianisme en France," Journal Asiatique 229, 275-284. .(1967a). " L o u i s Renou et son uvre scientifique," Journal Asiatique 2 5 5 , 1 - 1 1 . (1967b). " Bibliographie des travaux de Louis Renou," Journal Asiatique 255,13-30. Foster, Henry P. (1810). An essay on the principles of Sanskrit mar, P a r t i . Calcutta. Gram-

Frauwallner, Erich (1959). "Dignga, sein W e r k und seine Entw i c k l u n g , " Wiener Zeitschrift fr die Kunde Sd- und Ostasiens und Archiv fr indische Philosophie 3, 83-164. (1960). "Sprachtheorie und Philosophie im Mahbhsyam des Patajali," Wiener Zeitschrift fr die Kunde Sd- und Ostasiens und Archiv fr indische Philosophie 4, 92-118. (1961). "Geschichte und Aufgaben der W i e n e r Indologie," Anzeiger der sterreichischen Akademie der Wissenschaften, Philosophisch-historische Klasse 98, 77-95. Geiger, Bernhard (1909). " Mahbhsya zu P. VI, 4, 22 und 132 nebst Kaiyata's K o m m e n t a r : bersetzt, erlutert und mit einem A n h a n g , " Sitzungsberichte der philosophisch-historischen Klasse der kaiserlichen Akademie der Wissenschaften zu Wien 160, VII : 1-76. See pp. 209-259. Goldstcker, Theodor (1861). Pnini: His Place in Sanskrit Literature. London. Reprinted Varanasi 1965, Osnabrck 1966. (1874a). Patajali1 s Mahbhsya, a MS dated Samvat 1751, as reproduced by photo-lithography. London. (1874b). Patajali s Mahbhsya with Kaiyyatas Bhsyapradpa, an undated MS, as reproduced by photo-lithography. London.

530 Bibliography

Gopinatha Rao, T. A. (1916). Elements of Hindu Iconography, vol. II. Madras. Gray, J. E. B. (1959). " A n Analysis of N a m b u d i r i Rgvedic Recitat i o n and the N a t u r e of the Vedic A c c e n t , " Bulletin of the School of Oriental and African Studies 22, 499-530. (1968). Reviews of Biardeau 1964a and 1964b, and of Subramania Iyer 1965, Foundations of Language 4, 70-77. van Gulik, R. H. (1956). Siddham. An Essay on the History of Sanskrit Studies in China and Japan. Nagpur. See pp. 18-19. Hamilton, Alexander (1814). Terms of Sanskrit Grammar. London. Harle, J. C. (1963). Temple Gateways in South India. O x f o r d . Harweg, R. (1964). Kompositum und Kotolysot/onstext vornehmlich im spten Sanskrit. The Hague. Hsan Tsang. see Julien (1857-1858), and Beal (1885). Humboldt, Wilhelm von (1832). ber die Kawi-Sprache auf der Insel Java. Berlin. (1836). ber die Verschiedenheit des menschlichen Sprachbaues und ihren Einfluss auf die geistige Entwickelung des Menschengeschlechts. Berlin. See pp. 60-64. Hwui Li. See Beal (1911), Julien (1853), and Li (1959). Ingalls, Daniel H. H. (1960). " O n the study of the past," Journal of the American Oriental Society 80,191-197, I Tsing. See Takakusu. Jones, William (1807). " A Dissertation on the O r t h o g r a p h y of Asiatick W o r d s in Roman L e t t e r s , " The Works of Sir William Jones, vol. Ill, London, 253-318. Joshi, S. D. (ed. and trans.) (1967). Bhatta. Poona. (P.2.1.1). TheSphotanirnayaofKaunda

(1968). Patanjali's VykaranaMahbhsya : Samarthhnika Edited with Translation and Explanatory Notes. Poona.

, P. Kiparsky and J. F. Staal (in preparation). The Sanskrit Grammarians. Jouveau-DubreuM, G. (1937). Iconography of Southern India. Paris. Julien, Stanislas (1853). Historie de la vie de Hiouen-Thsang et de ses voyages dans l'Inde par Hei-Ii et Yen-thsong. Paris. (1857-1858). Mmoires sur les contres occidentales, traduits du sanscrit en chinois, en l'an 648, par Hiouen-Thsang, et du chinois en franais par M. Stanislas Julien, vols. M l . Paris. Keith, Arthur Berriedale (1 935). Catalogue of the Sanskrit and Prakrit Manuscripts in the Library of the India Office, vol. II, part I. London. Kielhorn, F. (ed. and trans.) (1866). Sntanava's Phitstra. Mit verschiedenen indischen Commentatoren, Einleitung, bersetzung und Anmerkungen. Leipzig. jibhatta, (ed. and trans.) (1868-1874). Paribhsendusekhara of Ngovols. 1-4. Trans, and re-edited 1960, Poona.

531 Bibliography

(1876a). Katyyona and Patojali: Their Relation to Each Other and to Pnini. Bombay. Reprinted Varanasi 1963, Osnabrck 1965. . (1876b). " O n the Mahbhsya" Indian Antiquary 5, 241-251. (ed.) (1880-1885). The Vykarana-Mahbhsya ofPatajaii, vols. 1-3. Bombay. Second edition 1892-1909, third revised edition 1962 (vol. 1), 1965 (vol. 2). Poona. (1883). " O n the grammarian Bhartrhari," Indian Antiquary

12,226-227.
(1885). " D e r Grammatiker Pnini," Nachrichten von der Kniglichen Gesellschaft der Wissenschaften und der Georg-AugustsUniversitt zu Gttingen 5,185-190. See pp. 103-105. Sa). Book Notice of Whitney's Roots, Verb-forms and Primary Derivatives of the Sanskrit Language, Indian Antiquary 15, 86-87. See pp. 155-157. (1886b). " Indragomin and other Grammarians," Indian Antiquary 15,181-183. (1886c). "The Gndra-Vykarana and the Ksik-Vrtti" Indian Antiquary 15,183-185. (1886d). "cryadeslya" or Notes on the Mahbhsya 1, Indian Antiquary 15, 80-81. (1886e). " Gnikputra and GnardTya" or Notes on the Mahbhsya 2, Indian Antiquary 15, 81-84. (1886f). " O n Some Doubtful Vrttikas" or Notes on the Mahbhsya 3, Indian Antiquary 15, 203-211. (1886g). "Some Suggestions Regarding the Verses (Kriks) in the Mahbhsya " or Notes on the Mahbhsya 4, Indian Antiquary 15, 228-233. (1887a). "The Authorities on Grammar Quoted in the Mahbhsya" or Notes on the Mahbhsya 5, Indian Antiquary 16,101-106. See pp. 106-114. . (1887b). "The Text of Pnini's Stras as Given in the Kasika-Vrtti, Compared with the Text Known to Ktyyana and Patajali" or Notes on the Mahbhsya 6, Indian Antiquary 16, 178-184. See pp. 115-123. (1887c). "Some Devices of Indian Grammarians" or Notes on the Mahbhsya 7, Indian Antiquary 16, 244-252. See pp. 123-134. Kiparsky, Paul (1970). " From paleogrammanans to neogrammarians," Essays in the History of Linguistics (Dell Hymes, ed.). Bloomington. , and J. F. Staal (1969). "Syntactic and Semantic Relations in Pnini," Foundations of Language 5, 83-117. Krishna Sastri, H. (1916). South-Indian Images of Gods and Goddesses. Madras. Kulke, H. (1969). Gidambaramahtmya. Eine Untersuchung der religionsgeschichtlichen und historischen Hintergrnde fr die Entstehung der Tradition einer sdindischen Tempelstadt. Wiesbaden.

532 Bibliography

Kunjunni Raja, K. (1963). Indian Theories of Meaning. Madras. |_anman, Charles Rockwell (1897). " Memorial Address," The Whitney Memorial Meeting: A Report, 7-27. Reprinted in Sebeok 1966,1.426-439. Lassen, Christian (1830). " b e r H e r r n Professor Bopps grammatisches System der Sanskrit Sprache," Indische Bibliothek 3.1.1-113. Lees, Robert (1963 2 ). The Grammar of English The Hague. Nominalizations.

Leitzmann, Albert (ed.) (1908). Briefwechsel zwischen Wilhelm vom Humboldt und August Wilhelm Schlegel. Halle a.S. Lvi, S. (1891). " N o t e s de chronologie indienne, Devnmpriya Asoka et Ktyyana," Journal Asiatique 181, 549-553. (1906-1908). " D e s prverbes chez Pn in i," Mmoires de la Socit Linguistique de Paris 14, 276-279. Liebich, Bruno (1891). Pnini. Ein Beitrag zur Kenntnis der indischen Literatur und Grammatik. Leipzig. See pp. 159-165. (1892). Zwei Kapitel der Kcik: bersetzt und mit versehen. Breslau. (1895). Das Cndra-vykarana. (1902). Cndra-vykarana. Gttingen. Einleitung

Stra, Undi, Dhtuptha. Leipzig. Candragomins

(1918). Candra-Vrtti, der Original-Kommentar zu seinem grammatischen Stra. Leipzig.

(1919-1920). Zur Einfhrung in die indische einheimische Sprachwissenschaft, vols. I-IV. Heidelberg. (1921). Materialien zum Dhtuptha. Heidelberg. Breslau.

. (1928). Konkordanz Panini-Candra.

(1930). KsJrataraginJ, KsJrasvmin's Kommentar zu Pnini's Dhtuptha. Breslau. Li Yung-hsi (trans.) (1959). The Life of Hsuan-Tsang, by Hui-Li. Peking. Lohmann, J. (1960). " b e r den paradigmatischen Charakter der griechischen K u l t u r , " Die Gegenwart der Griechen im neueren Denken. Festschrift H.-G. Gadamer, 171-187. Tbingen. Master, Alfred (1956). "Jonesand Pnini," Journal of the American Oriental Society 76,186-187. Matilal, B. K. (1966). " Indian Theorists on t h e N a t u r e of t h e Sentence (vkya)," Foundations of Language 2, 377-393. (1968). The Navya-nyya Doctrine of Negation. The Semantics and Ontology of Negative Statements in Navya-nyya Philosophy. Cambridge, Mass. McCawley, James D. (1967). " T h e phonological t h e o r y behind W h itney's Sanskrit Grammar, ' ' Language and Areas. Studies presented to George V. Bobrinskoy 77-85. Chicago. Misra, V. N. (1966). The Descriptive Technique of Pnini. The HagueParis. Mukherjee, S. N. (1968). Sir William Jones: AStudy in EighteenthCentury British Attitudes to India. Cambridge.

533 Bibliography

Mller, F. Max (1866,1870 2 ). A Sanskrit Grammar for Beginners. London. See pp. 138-139. . (1869). Rig-veda-prtiskhya, schen Phonetik. Leipzig. das lteste Lehrbuch der vedi-

(1883). India, What can it teach us? London. Ojihara, Yutaka (1958,1960). "S. 1.1.62 vis--vis de s. 1.1.56," or Causerie Vykaranique I (unfinished), Journal of Indian and Buddhist Studies 6, 3O5-32; 8, 370-369. (1959,1963). " A n t r i o r i t du Ganaptha par rapport au Straptha" or Causerie Vykaranique II, Journal of Indian and Buddhist Studies 7, 797-785 ; 1 1 , 852-846. (1961,1962). "Incohrence interne chez la Ksik" or Causerie Vykaranique III (unfinished), Journal of Indian and Buddhist Studies 9, 753-749; 10, 776-766. (1967a). " Jti 'genus' et deux dfinitions pr-patajaliennes" or Causerie Vykaranique IV, Journal of Indian and Buddhist Studies 16, 459-461. See pp. 425-431. (trans.) (1967b). La Ksik-Vrtti Paris. , and Louis Renou (trans.) (1960-1962). La Ksik-Vrtti (Adhyya I, Pda 1), vols. 1-2. Paris. Pandeya, R. C. (1963). The Problem of Meaning in Indian Philosophy. Delhi-Varanasi-Patna. Paranjpe, V. G. (1922). Le Vrtika de Ktyyana. Une tude du style, du vocabulaire et des postulats philosophiques. Paris. Pawate, I. S. (1935?). The Structure of the AstdhyyJ. H u b l i . Peterson, P. (1883-1885). " N o t e o n the Date of Patajal i," Journal of the Bombay Branch of the Royal Asiatic Society 16,181-189. Pillay, K. K. (1953). TheSucJndram Temple. Madras. Pischel, R. (1885). " D e r Dichter Pnini," Zeitschrift morgenlndischen Gesellschaft 39, 95-98. der deutschen (Adhyya I, Pda 1), vol. 3.

Pons, J. F. (1743). in Lettres difiantes et curieuses, crites des Missions trangres, par quelque Missionaires de la Compagnie de JESUS, X X V I Recueil, 222-227. Paris. See pp. 30-32. Pott, August Friedrich (1833-1836). Etymologische Forschungen auf dem Gebiete der indo-germanischen Sprachen, vols. 1-2. Lemgo. Raghavan, V. (1956). Sanskrit and Allied Indological Studies in Europe. Madras. Raja, C. K. (1957). " T h e SivaStras of Pnini," Annals of Oriental Research, Centenary number, Sanskrit section, 65-81. Rjarmasstri and Blasastri (eds.) (1871). Patajals with Kaiyyatas BhsyapradJpa. Benares. Mahbhsya

Rgnier, Adolphe (ed.) (1856-1858). " tudes sur la Grammaire vdique. Prtiskhyadu Kig-Vda," Journal Asiatique ser. 5: 7, 163-239, 344-407, 445-474; 8, 255-315 ; 482-526 ; 9, 210-247; 10, 57-111, 374-450, 461-474; 1 1 , 289-379; 12,137-220, 329-394, 535-593.

534

Renou, Louis (1931). Bibliographie

Vdique. Paris.

(1936). "Sylvain Lvi et son oeuvre scientifique," Journal asiatique 228,1-59. . (1937). " Review of Faddegon's Studies on Pnini's G r a m mar," Orientalistische Literaturzeitung 40, no. 5, 318-319. (ed. and trans.) (1940-1956). La Durghatavrtti de Saranadeva; Trait grammatical en Sanskrit du Xlle sicle, I n t r o d u c t i o n ; 1,1-3; II, 1-3. Paris. (1941-1942). "Les Connexions entre le rituel et la grammaire en sanskrit," Journal as/at/que 233,105-165. See pp. 434-469. (1942,1957 2 ). Terminologie grammaticale du sanskrit. Paris.

(trans.) (1948-1954). AstdhyyJof Pnini, vols. N i l . Paris. Re-edited w i t h t e x t of stras, 1961. (trans.) (1951). Sakara: Prolgomnes au Vdnta. Paris. (1953a). " Les Transitions dans la grammaire de Pnini " (tudes Pninennes I), Journal asiatique 241, 417-427. (1953b). " L e Vdachez Patajali" (tudes Pninennes II), Journal asiatique 241,427-464. (1955-1967). tudes vdiques etpninennes, vols. 1-16. Paris.

(1956a). " L'indianisme au Japon," Kratylos 1 , 99-103. (1956b). "Les undistra" (tudes Pninennes III), Journal asiatique 244,155-165. (1956c). " A m a r a s i m h a e t P n i n i " (tudes Pninennes IV), Journal asiatique 244, 369-377. (1956d). " Le Vdachez les grammairiens non-paninens" (tudes Pninennes W), Journal asiatique 244, 377-389. (1957). " G r a m m a i r e et Vdnta," Journal asiatique 245, 121-133. See pp. 470-478. (1957). See Wackernagel, J., and L. Renou. (1960). " La Thorie des temps du verbe d'aprs les grammairiens sanskrits," Journal asiatique 248, 305-337. See pp. 500-525. (1960-1962). See Ojihara, Yutaka, and Louis Renou. (1961). " G r a m m a i r e et Potique en sanskrit." tudes vdiques et pninennes 8, 105-131. See pp. 478-499. (1963). " S u r le genre du stra dans la l i t t r a t u r e sanskrite," Journal asiatique 251,165-216. (1969). " Pnini," Current Trends in Linguistics 5, 481-498. Rocher, Rosane (1964a). " ' A g e n t ' et ' O b j e t ' chez Pnini," Journal of the American Oriental Society 84, 44-45. (1964b). " T h e Technical T e r m ' hetu ' in Pnini's Astdhyy," Vishveshvaranand Indological Journal 2, 31-40. (I967). Review of Birw 1966, Journal of the American Oriental Society 87, 582-588. (1968a). La Thorie des voix du verbe dans l'cole paninenne (le I4ehnika). Bruxelles.

535 Bibliography

(1968b). Alexander Hamilton (1762-1824). A Chapter in the Early History of Sanskrit Philology. N e w Haven, Conn.

(1970). "New Data for the Biography of the Orientalist

Alexander H a m i l t o n , " Journal of the American Oriental Society 90, 426-448. Roth, Rudolph (ed.) (1852). jska's Nirukta, sammtden Nighantavas. linguis-

Gttingen.
Ruegg, D. S. (1959). Contributions l'histoire de la philosophie tique indienne. Paris.

Sachau, Edward C. (ed.) (1887). AI-BlrnVs India. An Account of the Religion, Philosophy, Literature, Chronology, Astronomy, Customs, Laws and Astrology of India. London. Reprinted Leipzig 1925. (trans.) (1910). Alberuni's india, An Account of the Religion, Philosophy, Literature, Geography, Chronology, Astronomy, Customs, Lows and Astrology of India, vols. I-II. London. See pp. 20-22. Sastri,G. N. (1959). The Philosophy of Word and Meaning: Some Indian Approaches with Special Reference to the Philosophy ofBhartrhari. Calcutta. Schiefner, Anton (trans.) (1869). Trnatha's Geschichte des Buddhismus in Indien. St. Petersburg. See pp. 23-26. Schlegel, August Wilhelm von (1820). " U e b e r den gegenwrtigen Zustand der Indischen Philologie," Indische Bibliothek 1,1-49. See pp. 51-52. (1832). Rflexions sur l'tude des langues asiatiques. BonnParis. See pp. 54-57. Schlegel, Friedrich von (1808). ber die Sprache und Weisheit der Inder. Heidelberg. Schwab, Raymond (1950). La Renaissance Orientale. Paris. Sebeok, Thomas A. (ed.) (1966). Portraits of Linguists. A Biographical Source Book for the History of Western Linguistics, 1746-1963, vols 1-2. Bloomington-London. Seymour, Thomas Day (1894). " W i l l i a m D w i g h t W h i t n e y , " American Journal of Philology 15, 271-298. Reprinted in Sebeok 1966,1, 399-426. Skld, H. (1926). Papers on Panini and Indian Grammar in General. Lund. Somasundaram, J. M. (1955). The University's Environs. Annamalainagar. Staal, J. F. (1962a). " A Method of Linguistic D e s c r i p t i o n : the O r d e r of Consonants According t o Pnini," Language 3 8 , 1 - 1 0 . (1962b). " Negation and the Law of Contradiction in Indian T h o u g h t , " Bulletin of the School of Oriental and African Studies 25, 52-71. (1965a). "Context-sensitive Rules in Pnini," Foundations of Language 1, 63-72. (1965b). "Euclid and Pnini," Philosophy East and West 15, 99-116.

536 Bibliography

(1966a). " Room at the Top in Sanskrit : Ancient and Modern Descriptions of Nominal C o m p o s i t i o n , " ndo-lranian Journal 9, 165-198. _(1966b). " I n d i a n Semantics, I," Journal of the American tal Society 8 6 , 3 0 4 - 3 1 1 . _(1967). Word Order in Sanskrit and Universal Grammar. drecht. OrienDor-

_(1969). "Sanskrit Philosophy of Language," Current Trends in Linguistics 5, 499-531. (1970a). " O r i g i n and Development of Linguistics in India," Essays in the History of Linguistics (Dell Hymes, ed.). Bloomington. (1970b). " Review of Cardona (1969)," Language 46, 502-507. Subramania Iyer, K. A. (1942). " T h e Conception of Guna among the Vaiyykaranas," New Indian Antiquary 5,121-130. (1948). " T h e Point of View of the Vaiykaranas, "Journal Oriental Research'lu, 84-96. See pp. 393-400. (ed.) (1963). Bhartrhari's Poona. (trans.) (1965). The VkyapadJya ofBhartrhari Chapter I. Poona. (ed.) (1966). Bhartrhari's VkyapadJya, Knda I. Poona. with the Vrtti, VkyapadJya, Knda III, Part 1 . of

(1969). Bhartrhari. A Study of the VkyapadJya in the Light of the Ancient Commentaries. Poona. Swaminathan, V. (ed.) (1965). Bhartrhari's MahbhsyatJk, vol. I. Varan asi. Takakusu, J. (transi.) (1896). A Record of the Buddhist Religion as Practised in India and the Malay Archipelago, by I-Tsing. O x f o r d . Trantha. See Schiefner. Thieme, P. (1935a). Pnini and the Veda: Studies in the Early History of Linguistic Science in India. Allahabad. _(1935b). " Bhsyazu v r t t i k a 5 z u Pnini 1.1.9 und seine einheimischen Erklrer. Ein Beitrag zur Geschichte und W r d i g u n g der indischen grammatischen Scholastik," Nachrichten von der Gesellschaft der Wissenschaften zu Gttingen, Philologisch-historische Klasse Neue Folge, 1, 5,171-216. See pp. 299-332. (1937-1938). " O n the Identity of the V r t t i k a k r a , " Indian Culture 4,189-209. See pp. 333-356. (1956). "Pnini and the Pniniyas", Journal of the American Oriental Society 7 6 . 1 - 2 3 . (1957a). "Pnini and the Pronunciation of Sanskrit," Studies Presented to Joshua Whatmough on his Sixtieth Birthday (Ernst Pulgram, ed.), 263-270. 's-Gravenhage. (1957b). " T h e I n t e r p r e t a t i o n of the Learned," Felicitation Volume presented to Professor Sripad Krishna Belvalkar (A. S. Altekar, ed.). 47-62. Baaras. (1958). Review of Renou 1957 2 , Gttingische Gelehrte Anzeigen 212,19-49.

537

Varma, S. (1929). Critical Studies in the Phonetic Observations of

Bibliography

Indian Grammarians. London. Second ed. 1961. Delhi. Vasu, S. C. (ed. and trans.) (1891). The AshtdhyyJof Pnini. Allahabad. Reprinted 1962, Delhi-Varanasi-Patna. Verbrg, P. A. (1950). "The Background to the Linguistic Conceptions of Franz Bopp," Lingua 2, 438-468. Reprinted in Sebeok1966, 1,221-250. Wackernagel, J. (1908). "Franz Kiel horn ( f l 9 Mrz 1908)," Nachrichten der K. Gesellschaft der Wissenschaften zu Gttingen, Geschftliche Mitteilungen 1,1-22. -and L. Renou (1957). Altindische Grammatik: Introduction genrale, 34-42,112-125. Gttingen. Waley, Arthur (1952). The Real Tripitaka and other pieces. London. Weber, Albrecht (1858). " Das Vjasaneyi-Prtikhyam," Indische Studien, Beitrge fr die Kunde des indischen Alterthums 4, 65-171, 177-331. (1862). " Z u r Frage ber das Zeitalter Pnini's," Indische Studien, Beitrge fr die Kunde des indischen Alterthums 5,1-176. (1873a). "Das Mahbhshyades Patajali," Indische Studien, Beitrge fr die Kunde des indischen Alterthums 13, 293-496. (1873b). "Remarks on Parts Xand X I , " [including on Bhandarkar 1872], Indian Antiquary 2, 57-59. (1873c). Professor Weber on Patajali, &c [reply to Bhandarkar 1873a], Indian Antiquary!, 206-210. (1877). To the Editor of the Indian Antiquary [reply to Kielhorn 1876b], Indian Antiquary 6, 301-307. Westergaard, Niels (1841). Radices linguae sanscritae. Copenhagen. Whitney, William D. (ed., trans.) (1862). The Atharvaveda Prtiskhya or SaunakJy Caturdhyyik. New Haven, Conn, (reprinted 1962, Varanasi). (ed., trans.) (1871). "TheTittirya-Prtiakhya with Its Commentary, the Tribhshyaratna," Journal of the American Oriental Society 9,1-469. (1874). Oriental and Linguistic Studies. New York. (1879,18892). Sanskrit Grammar. Leipzig. (1884). "The Study of Hindu Grammar and the Study of Sanskrit," American Journal of Philology 5, 279-297. See pp. 142-154. (1893). " O n Recent Studies in Hindu Grammar," American Journal of Philology 14,171-197. See pp. 165-184. Wilson, H. H. (1819.18322). Dictionary in Sanskrit and English. Calcutta. Windisch, Ernst (1917-1920). Geschichte der Sanskrit-Philologie und indischen Altertumskunde, vols. l-ll. Strassburg. (1921). "Philologie und Altertumskunde in Indien," or Geschichte der Sanskrit-Philologie und indischen Altertumskunde III, Abhandlungen fr die Kunde des Morgenlandes XV: 3,1-38. Woods, J. H. (1914,19272). The Yoga-system of Patajali. Cambridge, Massachusetts.

Of Names

Abhimanyu, 104 Abhinavagupta, 16 Abhyankar, K. V., 406, 411 Abul'aswad Addu'al, 22 AcyutaJallak, 43 Agatsya, 440 Ajtasatru, 23 Aklujkar, A.,16,392, 526 Allen, W. S., 401,526 Altekar, A. S.,536 Amarasimha, 42, 44, 50, 444, 534 nandapla, 22 nandsrama, 291 nandavardhana, 421-422, 499 Anguetil-Duperron, A. H., 30 Anubhtisvarpcrya, 30-31 pastamba, 202, 441, 443, 450, 459 pisali, 108-109, 304 Aristotle, x Asoka, 100,197,199-200, 532 svalyana, 288 Audumbaryana, 401, 528 Aufrecht, T., 62, 426, 446, 526 Blasarman Pgondiya, 41 Blasstn, 85, 309, 533 Ballantyne, J. R., 85, 526 Barnett, L. D., 392 Baudhyana.440,443,459 Beal.S., 4-5, 7,526,530 Belvalkar, S. K., 14, 21-22, 30, 317, 485, 526, 536 Benfey, Th., 139,143,150,193, 526 Benveniste, E., 497 Bergaigne, A. H. J., 453 Bhguri, 44 Bhmaha, 480-481, 487-488, 492-493, 495 Bhandarkar, R. G., x, xvi, 16, 70-102,115,138,155,401,435, 526,537 BhnujidTksita, 43 BhradvjaSstri, 314, 377 Bhrgava, 440, 450 Bhartrhari, xxiv, 15-17, 37,104,195, 286,296,310,315,319,332,392, 397, 401-404, 407-425, 526-527, 531,536 Bhattacharya, B., 392, 425, 526 Bhatt ,293 Bhattojidksita, xxvi, 33-34, 40-41, 44, 51,116,195,266,271,293,298, 308-309, 312-321, 326-328, 331, 411-412 Bhimaukla, 25 Bhoja, 480, 485 Biardeau, M, 15, 392, 425-426, 526, 530

al-irni, A. R., 20-22, 24, 34,106, 527, 535 Birw, R., 108, 286,481,534 loch,B.,264,527 Bloomfield, L, 264-266, 268, 270, 272, 274,298, 527 Bloomfield, M, 264 Boehtlingk, O., x, xiv, xvi, 52, 62, 70-71,85-86,116,151,162,182, 185-192,194, 210, 213, 234, 238, 245, 247, 266, 271-281, 425, 446, 465,489,512,515,527,529 Boplita, 44 Bopp, F., 50, 52-61, 63,102,138-140, 180,183,527,532 Boudon, P., 123, 357-391, 442, 527 Bradke, P., 437 Breoier, B;, 274, 527 Brhadratha, 80 Brhaspati,440 Brhaspati Mukuta, 43 Brough, J., xiv, xvi, 262, 392, 401-423,527 Brugmann, K., 264, 500, 524 Buddha, 6 Buddhisgara, 195 Bhler, G., xvi, 71,102,134,153, 193-204,286,446,528 Buiskool, H. E., 273, 357 Burgess, J., 70 Burneil, A.C., 21, 86,106-107,461, 528 Burnouf, E., 70 Burnouf, J. L, 60 Burrow, T., 401, 528 Bu-ston, 23 Caland, W., 350, 354, 453-454, 458-460 Camps, A., 30 Candra(gomin), xxiv, 21-22, 37, 44, 86,104,106,116-122,134,158, 195-196, 207, 437, 444, 446, 532 Candragupta, 80, 84 Cannon, G. H., xvi, 33, 528 Cappeller, C.,480, 482 Cardona, G., xiv, 185,274, 528 Carey, W., 33, 528 Carnap, R., 264, 404 Chakravarti, 429,469 Chao Pu-chu, 7 Chatterji, K. C, xvi, 274, 286-296, 469,500, 528 Chatterji, S. K., 21, 528 Chattopdhyya, K.,438 Chaturvedi, S. P., 438 Chzy, A. L. de, 30, 50, 357 Childers, R, C, 99

540 Index

Chomsky, N., 60, 62,140,528 Colebrooke, H. T., x, x, xvi, 33_45f 49_57( 82,102, 401, 528 Confucius, 14 Copilowish, 413 Cowell, E. B., 410 Dksyana, 113-114 Dksi,323 Dandin, 484-486, 492-493 Dasgupta, S. N., xvi, 395, 469, 529 De, S. K., 405-406, 440, 479, 487 Debrunner, A., 482, 489 Delattre, P., 417 Delbrck, B., 52,186,189,191, 435, 512,522-525,529 Deussen, P., 444, 465, 468 Devanandin, xxiv Dhanamjaya, 44 Dharandhara, 34 Dharmapla, 16-17 Dhrtasimha, 44 Dick, W.,33 Dirinaga, 15,529 Dixit, S.V., 492 Dove, A., 60, 529 Durgdsa, 426-427, 429 Durga(simha), xxiv, 44 Edgerton, F., 286, 422, 455, 469, 529 Edgren, A. H., 144,196-199, 201 Eggeling, J.,67,529 Eliade, M., xvi, 529 Emeneau, M. B., xiv, 33, 529 Euclid, 20 Faddegon, B., 273-284, 441, 529, 534 FaTsang,18-19, 529 Fausbll, M., 199 Felber, E.,207 Fick, A.,196 Filliozat,J.,xiv, 30,432, 529 Finot, L.,204 Firth, J. R., 401-402 Foster, H. P., 33, 53, 529 Foucher, A., 392, 411 Franke, O., 165,168-171,181 -184, 186,435,446 Frauwa!iner(E.,15,78,207f529 Gaastra, D.,447 Gadamer, H. G., 532 Glava, 37 Gangdhara, 44 Garbe, R., 441, 469 Gardiner, A. H., 403, 418 Garge, S., 472-474 Grgya, 37, 335

Geiger, B., 207-258, 529 Geldner, K., 452, 454, 461-462, 467 Ghate, V. S.,444 Ghosh, B., 435, 438 Ghosh, M., 479, 483,499 Gobhila.38,443 Gibbon, E.,85 Goldstcker, Th., xi, 70-79, 86, 92, 107,121,136,210,213,273,280, 332, 338, 370, 372, 380-384, 386, 390,401,435,461,526,529 Gonda, J., 484, 495 Gopinatha Rao, T. A., xvi, 530 Gotama, 88, 469 Gough.A. E.,410 Govardhana, 44 Gray, J. E. B., xiii, xiv, 392, 530 Grierson, G. A., 440 Gulik, R. H. van, 8,18, 529 Halde, du, 30 Halle, M., xiv Hamilton, A., 49, 54, 432, 530, 535 Haradatta (Misra), xxiv, 39, 42,115, 121,172,195,289,298,314-316 Haridksita, 41 Harivrsabha, 527 Harle, J.C.xiv, xvi, 530 Harweg, R., 185, 530 Heesterman, J. C, xiv Helaindha.44 Helrja. 45,109, 393-400, 425, 438 Hemacandra, 106,131,195,197, 204, 437, 486,492, 500, 503, 507-513, 518-521,523 Henry, V., 350, 354, 454 Herodotus, 4 Hillebrandt, A.,440 Hiranyakesin, 202, 441, 443, 459 Holtzmann, A., 146,148 Homer, 190 Hoogt, Van der, 460 Hsan Tsang, 4-10,18, 526, 530, 532 Humboldt, W. von, xii, 30, 52-53, 59-64,85,138,529,530,532 Hwui Li, 4, 7-10,13,18, 526, 530, 532 Hume, R. E.,465 Hymes, D., 531, 536 Indragomin, 106, 531 Ingalls, D. H. H., 34, 530 Inyanendra Sarasvat, 41 I Tsing, 5, 7,11-17, 392, 530,536 Jacob, G. A., 315 Jacobi, H., 199, 469, 479 Jacquemort, V., 58 Jaimini, 2, 3,88,286,469,498

541 Index

Jainendra, 500 Jayditya.33,133,196,266,271 Jayapla, 22 Jespersen, O., 418 Jinendrabuddhi, xxiv, 195-196 Jinendra-Pjyapda, 195 Jhalkkar, Bh.,198 Jones, W. r xv, 33-34, 40, 44, 49, 58, 530, 532 Joshi.S. D.,x, 54, 392,401,530 Jouveau-Dubreuil, G., xv, 530 Julien, S., 4-5, 7, 530 Jumaranandin, 30-31 Kaiyata, xxiv, 38, 70, 74-79,107-109, 115-122,126,136-137,195, 207-210,216-217,220-262, 267-272, 289, 293, 298, 300-322, 326,332,366-368,382-386, 389, 426, 431, 447, 461, 468, 500, 508-514,520,529,533 Kamsa, 509 Kane, P. V., 479, 491-492, 498 Knva, 454 Kaplan, A., 413 Krttikeya, 293 Ksakrtsna,109 Ksintha, 34 Ksyapa, 37 Katre, S. M., xiv Ktyyana, xxiv, 14, 37-38, 55, 78-79, 85-132,138,155,174,194, 209-221, 228, 230, 235, 251, 259, 267-272, 287,-295, 299-301 305-308,315,318,322,324, 331-357, 378, 385, 398, 403, 406, 437-441,446,452,464,527, 531-532 Kausika.443,458,463,466 Kautilya, 440 Kautsa, 349, 504 Keith, A. B., 262, 333-339, 347-348, 354-356,405,435,438, 447, 530 Kern, H., 84,199-200 Khdira, 443 Kielhorn, F., xii, xviii, 11-12,15-16, 21, 54, 70-71, 82, 85-86, 90, 102-134,138,151,155-157,195, 207-210,249/258,262,269, 287-291, 298,300,310, 318-319, 329, 335, 346-347, 357-358, 363-380,387, 389-405, 425, 430-431,435,446,530,537 Kiparsky, P., xi, xiv, 54, 61,185, 530-531 Kondabhatta, 406 Kramadsvara, 30-31 Krishnamacharya, V., 412

KrishnaSastri, H., xvi, 531 Krsna,509 Krsnapandita, 41 KsTrasvmin, 207, 532 Kudla, S. D.,267 Kuhn, A., 199 Kulke, H.,2, 531 Kumrila, 195, 286, 469 Kumbhakarna, 486 Kunaravdava, 113 Kunjunni Raja, K., 392, 414, 425, 532 Lanman.C. K., 140, 204, 532 Lassen, C , 53-54, 61,70, 78, 81, 138-139,532 La Valle Poussin, de, 392 Lees, R.,532 Leitzmann, A., 52, 59, 61-62, 85, 532 Leumann, E., 199 Lvi, S., 357, 392, 432, 436, 468, 532, 534 Liebich, B.,xi, 11,13,16, 53-54, 70, 78,138-139,158-186,188, 190-191,194,196, 207, 219, 266, 275,278,411,433, 435, 444, 448, 450,469,508,532 Li Ying-hsi, 7, 530, 532 Lohmann, J., 62, 532 McCawley, J.,140, 532 Mackintosh, J., 52 Mdhava,406,410 Maitreya Raksita, 39 Malayagiri, 195 Mammata, 439, 478, 493, 496-499 Mnava, 437, 450 Manicya Chandra, 35 Master, A., 33, 532 Matilal, B. K., 286, 532 Medinikara, 44 Meillet, A.,507 Menandros, 78, 81 Minard, A., 511 Misra, V. N., 274, 532 Misra, Kamalknta, 298 Misra, Prthasrathi, 286-287 Monier-Williams, M., xviii, 102 Mookerji, R.,435 Morris, 199 Muir, J., 82 Mukherjee,S. N.,33,532 Mller, F. M., 11,15, 33, 70-71, 85, 102,104,138-139,145,153-154, 197,443,465,533 Ngrjuna, 24-25 Ngojlbhatta, xxiv, 41, 70, 85,110-122 195,254, 267-269, 293, 298, 303,

542 Index

Nagojbhatta (continued) 308-332, 345-346, 358-391, 406, 411-413, 421, 426-431, 443, 500, 507-509,514,517,530 Nahahari Sstr Pendse, 318 Nam,103 Nanda, 23-24 Nrada, 288, 439 Nryanabhatta, xxiv Neisser.W., 461 Nooten, B. A. van, xiv Oberhammer, G., xiv Oertel, H., 437, 463 Ojihara, Y., xiv, xviii, 424-432, 481, 499, 533-534 Oldenberg, H., 58,199, 264, 441, 454, 461,468
Paik.T. S.,xv Palle, C , 439 Pandeya, R. C , 392, 425, 533 Pnini, passim Prnanda, 440 Paranjpe, V. G.,357,533 Patajali, x i , xvi, xvii, xxiv, 2, 3, 14-15, 28-29, 38, 55, 68, 70-134, 136,138,151,155,162,174,194, 207,209-223, 231-235, 239, 248, 250-274, 286-299, 305-396, 401-408, 425-437, 446, 448, 468-475,494, 502, 525-534, 537 Pathak, K. B.,435 Paulinus of St. Bartholomew, 30 Pauskarasdi, 111 Pawate, I.S.,286,533 Pyagunda, 373,379 Peterson, P., 87, 526, 533 Pillay, K. K., xvi, 533 Pingala, 440, 444, 447 Pische, R., 102-104,197, 533 Piyadasi,357 Plato, 190 Pons, J. F., x i i i , xiv, xviii, 30-33, 61, 357,533 P o t t , A . F.,62-63, 533 Ptolemy, 20 Pulgram, E., 536 Pulle,150 Punyarja.45,407-408,413 Purusottamadeva, xxiv, 44, 432, 447 Puspamitra(Pusyamitra), 79-85,103 Rabhasa Pla, 44 Raghavan, V., xiv, 432, 485, 533 Raja.C. K.,274, 533 RjarmasstrT, 85, 533 Rmacandra, xxiv, 39-40, 316

Ramachandran, R.,xiv Rmakrsna SstrI, 314 Rmnuja, 469 Rma SstrT Mnavalii, 317 Ramaswami Sastri, V. A., 490 Rantideva, 44 Regnaud, P., 491, 494-498 Rgnier, A., 70, 533 Renou, L.. 21, 54, 62, 68, 123,165-166, 273,357,403,424,426,432-525, 529, 533-537 Rhys Davids, 199, 392 Rocher, R., 49-50,108,185, 500, 534-535 Roth, H., 30 Roth, R., 70-71, 85,108,113,116,140, 489,512,515,527,535 Rudrata, 492 Ruegg, D.S., 15, 392, 425,535 Russell, B., 404, 413 Ruyyaka, 439, 479, 493 Sabarasvmin, 88, 286, 291, 443, 447, 452,456,458,469,473 Sachau.E., 20-21, 527, 535 Skalya, 37, 290, 303,321 Skatyana, 21, 37,108-109,105, 500, 510 Sketa, 508 Salisbury, E.E.,58,139 Samalvhana, 22 Sangama, 39 Sankarcrya, 42, 88, 286, 294, 442-443, 445, 461, 469-478, 534 Sntanava, 102, 530 Saranadeva, xxiv, 372, 432, 534 SarasvatT Kanthbharana, x x i i , 442 Sarvadhara Vmana, 44 Sarvavarman, xxiv, 22,196 Sasideva, 22 Sassetti, F., 30 Sastri, G . N . , 392, 535 Ssvata, 44 Stavhana, 22 Satyamurti, S. T., xiv Syana (-Mdhava), 39,183,195; 291, 308, 466 Scharbau, 468 Scharfe, H., 425-426 Schiefner, A., 23-24, 535 Schlegel, A. W . von, xii, 33, 49-62, 70, 85, 432, 532, 535 Schlegel, F. von, 49-50, 535 Schopenhauer, A., 30 Schroeder, L.,437 Schwab, R., 49, 58, 535 Scott, W . , 3 3 Sebeok, T. A., xviii, 33, 52-53, 58, 60,

543 Index

140-141, 264, 529, 532, 535, 537 Senart, E., 199, 456, 465 Seymour, T. D., 140-141, 535 Sieg, E.,298 Sinclair, B., xiv Sivabhatta, 41 Sivadatta Kuddla, 310, 318, 387 Sivadatta Sastr, 316 Sivaprasad Bhattacharyya, 478 Sisupla, 505 Skandasvmin, xxii, 449 Skld, H., 274, 284,435,535 Somasundaram Pillai, J. M., xiv, 2, 535 Speijer, J.S.,182 Spencer, G. J., 33 Spiegel, F. von, 150 Srinisvasan, C , xiv Staal, J. F., xi, 21, 50, 54, 78,137,140, 155,185,274,286,392,530,531, 535-536 St. John, 411 Strauss, O., 333, 411, 450, 456, 467, 469 Subhaga, 44 Subramania Iyer, K. A., xiv, xvi, 392-400, 530, 536 Sui-Shih, 14 Svmin, 44 Swaminathan, V., 392, 401, 536 Takakusu, J., 11-12,16, 530, 536 Trantha, 23-26, 401, 535, 536 T. Ganapati SstrT, 372 Thibaut, G., 444, 470, 472-473 Thieme, P., xiv, 30, 54, 274, 298-356, 432-441,473,500-501,536 Trapp, V., 408 Trenckner,V.,199 Trivedi, K.P., 316 Udayana, 24-26 Udbhata, 491-493 Ugrabhti,22 Ujjvaladatta, 62, 526 Upavarsa, 469 Utgikar, N. B., 526 Uvata, 334, 451-452, 456-457, 462 Vcaspati,44, 448 Vdava,113 Vaidikbharana, xxi, 462 Vaidyanathabhatta, 41, 380-381 Vjapyyana, 111, 398 VlmTki,288 Vmana, 33,196, 439, 480-499 Vmanendrasvmin, 41 Vararuci, 25-26, 44 Varma, S., 401, 469, 479, 537

Varsyayani, 113 Vasu,S.C.,xi,266,537 Vsudeva, 509 Vtsyyana, 88, 294, 440,469 VendryesJ.,517 Verbrg, P. A., 53, 537 Verhaar, J.W. M., xiv Vikramditya, 44 Vishva Bandhu Sstn, 443, 452 Visvantha Pancnana, 74 Vitthala, 318 Vopadeva, 33, 42, 49, 51-52, 426, 500-501, 527 Vydi, 44,111,398,402 Vyg h rapada, xvi, 2-3 Wackernagel, J., 71,102,166,285, 349, 432,435,437-438, 465, 469, 489, 505, 534,537 Waley, A.,5, 7, 537 Weber, A., 58, 70-71, 81, 83-86,102, 140,199, 335-338, 435, 438, 447, 490,526, 537 Wecker, O., 435 Westergaard, N., 54,138,155, 195-196,199,537 Whatmough, J.,536 Whitney, W. D., xii, 50, 70,139-196, 207,273,490,531,532,537 Wilkins, C.,33,49, 53, 57,59 Wilson, H. H., 49-50, 57-58, 82, 537 Windisch, E., 33-54, 60, 70, 469, 537 Winternitz, M.,435,470 Wittgenstein, L.,423 Woods, J.H., xvi, 456, 469, 537 Yj naval kya, 294 Yjnikadeva, xxi, 447, 455 Yska, 88, 94-95, 288, 399, 402, 414, 467, 535 Zachariae,Th.,199, 204

Index of Sanskrit Terms

Sanskrit alphabet: a, , , T, u, , r, r, j , e, o, ai, au, k, kh, g, gh, , c, ch, j , jh, , t, th, d, dh, n, t, th, d, dh, n, p, ph, b, bh, m, y, r, I, v, s, s, s, h

ak, 337 aka, 291 aksara, 107, 354-355, 452-454, 461-462 aghosa, 114, 336 aga, 162, 221, 223, 227, 251, 257, 362, 364,373, 377-378, 444,450-451, 457 aca,377,379(cf.ca) at, 122, 223, 225-226 an, 122, 250, 374 ati, 342 etc. (cf. tin) atidsa, 450, 518 add, 121 adyatanl, 114, 436 adhikarana, 124,166,186, 294, 464-465, 471 adhikaranya, 214 adhikara,'i28, 213, 221, 226, 235, 244, 248-249, 367, 373, 445, 472, 474, 502 an, 120 ana,103,288(cf. na) anabhidhna, 126-127 anit, 221 anitya, 226, 256, 326, 358, 371, 380 anudtta, 73,99,120,132, 245, 340-344,347, 349-353, 371, 438 anunsika, 73, 226, 228, 251, 311, 320, 336,375,387 anupasanhrin, 74 anubandha, 99,107-108,129-130, 241, 251 -252, 275-281, 311, 319, 326-327, 337, 358-360, 376,379 anuvrtti, 445, 472 anusvra, 326 anaikntika, 74 antaraga, 109, 225-227, 229, 234, 236-238,253-256, 344-345, 360, 364, 366, 368, 448, 477 antaragalaksana, 219, 364 antarangaparibhs, 360, 362, 364-365, 368 antodtta, 245, 247, 370-371 anyonysraya, 73 anvarthasamjn, 131 apavda, 217, 239, 257, 304, 343, 352, 360,375-376, 442-444, 450, 507 apdna, 186,166, 290,465 apratyaya, 325 (cf. pratyaya) apradhna, 107 abhinidhna, 466 abhinihita, 466 abht, 397 abhysa, 122,128, 369, 375-376 ayogavha, 114 avyayTbhva, 150,164,178-181, 191-192,464,481

avypti,219, 234 avyutpatti, 72 avyutpanna, 62,108, 229 as, 120 asarpa, 131-132 asarvanmasthna, 283 asavarna, 316, 321 (cf. savarna) asdhrana, 74 asiddha(tva), 207-259, 273, 335, 363, 364-366 asti,397 knks,415-416, 474 krti, 425-429, 431,470, 481 khyta, 342, 344, 349, 351, 354-355, 416 gama, 319 an,244,365-366 crya (desTya), 106 t, 121, 216, 223-224, 226 tmanepada, 7, 9, 215-216, 232, 278, 288 desa, 109, 216-219, 224, 249, 251, 257,343,360,380,449,457 dyudtta, 245 p,337 mantrita, 8,13,120, 341 -342, 349, 355 rdhadhtuka, 220, 223, 225-230, 232, 234-235, 238, 256, 288, 362-363, 380,387,454,516 slrlin, 161 in,360-361,367-368 it, 163, 231-232, 255-256, 368, 379, 383-384 in, 217, 234 it, 358 (anubandha) thuk, 382, 384 ni, 361 isti, 86, 450, 480-481 Isatsprsta, 299,303,307, 312 undi. 8,13-14, 62, 64-65, 76, 229, 296,437 utsarga, 217-219, 239, 348, 251, 304, 343, 352, 360, 375-376, 380, 442-443,450,472 udtta, 73, 99, 245-247, 340-344, 350-353,355,360,371,438 upadesaparampar, 298 upadh, 221, 240, 243, 250-251, 362 upadhmnlya, 114, 326, 336-337 upapada,452,472, 506, 511 upasamkhynavdin, 245, 247 upasarga, 117,121, 290, 338, 366, 456, 459

545 Index

upasarjana, 107, 248, 372, 444, 456, 459 uvan,238-241 t, 229, 251-252 th, 252-253, 362-364 sman, 462-463 ekadesa, 215, 217-218, 220, 224, 245, 254,323,328,365,443 ekayoga, 128 ekavacana, 13, 360, 463 ekasesanirdesa, 130 ekasruti, 344, 350,438 eknta, 358 aun, 107-108 aut, 349 kan,383 kap, 371-372 karaa, 166,186,464-465 karisyant, 464 kartr, 166,186-187, 291, 394, 464, 501-502 karmadhraya, 73,150,164,191-192, 362,464, 472, 488 karman, 124,166,186-187, 292, 464-465, 501 karmapravacanya, 331 kraka, 8,111,166,185-187, 273, 344, 398,464-465,504,514 krita, 114, 464 (cf. cekryita) kit, 117, 214-215, 222, 226, 230, 233, 240-242, 372, 374-375, 379-380,
383

277,280,290,304,311,362-366, 373,386-389,392,455 guru,303, 325 ghi, 117 ghu, 72-73, 75-77, 80,113-114, 208, 230,291, 376-377 ghosa, 463 as, 358, 360 asi, 358 ni, 358 e, 358, 360 it, 210, 220, 222, 230, 233, 241, 358, 372, 379 nip,247-248,250 ms, 247-248, 372 ca, 133-134 can,377-378 canap,130 caturtha, 336-337 caritrtha, 359, 366 carkarlta, 114 ciklrsita, 464 ein, 223, 232-233, 258-259 cit, 370 cucup, 130 cekryita, 114, 464 (cf. karlta) cpha,117 cli, 223, 378, 501 evi, 109,119 cha, 121, 491 japa, 354 jti, 398-399, 412-413, 424-431 jit, 335-337 jihvmlya, 114, 326,336-337 jus, 130 jnpaka, 110,124-125, 247, 253, 258, 329-331, 357-391, 442 japakasiddha, 329, 357-391 (cf. lokanyyasiddha) jnpakasiddhaparibhs, 357-391
jhit, 337

ku(tva),293,354, 378 kurvant, 464 krt, 62, 64, 234, 305, 328, 337, 372, 380,464 krta, 464 (past) krtya, 380, 464 krtvasuc, 385, 431 knit, 109,131, 212-213, 233, 241, 291,
372

ktv, 103, 238, 288, 364, 380 kyac, 109, 479, 491-492 kya, 479, 491-492 kramaptha, 438 kvip, 252, 364,491-492
khal. 518

it, 288, 368 nit, 368 tac, 370, 372 tp,344 ti, 72-73, 75-77,251,291,371,374,
386

khyun,118 ganaptha, 38, 54,124,159, 247, 481 guna, 109,128,130,133,138-139, 221-223, 228,230-231, 236-237, 240-242,244,248-249, 252-254,

thak.119 than,329

546 Index

da,386-389 dat, 382-384 da, 129 dp,248 dit, 386-389 du,114,336 dhak, 111,122, 296,403-404 na, 119, 374 namul, 92, 96, 292, 436, 479, 491 nal, 240 ni(c), 231-232, 238-239, 377-378 nit, 214-215, 241-242, 368 nini,491 nvi, 252-253, 362-363 nvul,118, 380 tan, 114, 336 tatpurusa, 73-74, 80,132,150,164, 340,362,270,464,481,483,488, : 490 taddhita, 117, 229, 250-251, 281, 321, 337, 368,374,450, 464, 483 . tadvadatidesa, 131 tamap, 233 tarap,233,431 tarm,213 tas(JJ), 431 tsi,232 tin, 7,13,336,338,342,415,484 tianta (padam), 7, 9,13, 283, 374 tithuk,382, 384 tip, 376 tuk, 207, 216, 218, 255-256, 328-330, 364-365 trc, 235, 291, 380 trtya, 114 trty, 245-246, 266, 336-337 thai, 240 thuk,382 drgha, 73, 277, 281, 300-302, 309-310,313,315,317-318, 321-322,326,329-330,354, 365-366,372, 274-376 dyu,114 dvandva, 74,150,164, 236, 266, 362, 437, 464, 483, 489,494 dvigu, 150,164,101-192, 464 dvitya, 114 dvity, 73,187, 331, 336-337 dvirvacana, 235, 244 dvivacana, 13, 339-340, 346, 463 dha, 385 dhtu,7-8,13,42, < 3,187,242,363,

366, 376, 380, 412-413, 454, 502, 505 dhtuptha, 7,13, 38-39, 42,124,154, 159,174-175,185,193-194,196, 198, 200-202, 209, 221, 238, 253, 326,376,450 dhi, 336-337 nah, 125-126, 210 na, 125, 288, 291, 379 napumsaka, 396, 463 nik, 375 nitya(tva), 103,114, 218, 220, 223, 225-226,229-231,241,250,256, 290,320-321,323,371,446 nipta, 457, 459, 460, 514-515 niptana, 113,120,125-126, 210, 249, 412,459,482,504 nimittasaptamT, 494 nisedha, 447 nisth, 465, 494 nuk,235, 375 nykha, 354 pacama, 335 pancam.121,233,301,339 pacamT, 464 (imperative) pada, 451-455, 462, 467, 491, 493, 507 padaptha, 34, 70,140, 438-439, 460, 465,490 para, 230-231 parasmaipada, 7, 9 paribhs, 37,41,123,316,319, 322-324, 330, 345-346, 357-391, 441-444, 452, 455, 459, 469, 473, 480 paroks, 114, 436,457, 510 paryudsa (pratisedha), 472, 496 pit, 220, 230, 233, 240, 328, 337 puk,362 punarkta, 281 pumlmga, 463 pumvad, 430-431 pumsa, 428 prvapaksa, 88, 95, 403, 471 prvapaksin, 89, 95, 99-100, 441, 471, 474 prvastra, 107-108 prakarsagati, 131 -132 prakrti, 76,109,113,120, 232, 251, 294, 377, 447, 450, 457, 470 pragrhya, 121, 339, 340, 438,490 pratipada, 256-257, 339, 376 pratipdana, 125 pratipdya, 125 pratisedha, 139, 217, 219, 223, 229, 246,249, 343, 361, 373, 377-378, 403,414,447,481,496

547 Index

pratyaga, 109,253,255 pratyaya, 73-74, 77-78,126, 223, 228-229, 231-232, 253, 257-258, 290,336,361,368,374,412, 457-458,493,501 pratyayasvara, 245 pratyakhynavdin, 245-247 pratyhra(grahana), 5, 25,129, 241,

256,276-282,311,320,325,
336-337,372,463,466 pratyudharana, 264 prathama, 114 pratham, 73,187, 336-337 pradesa, 383 pradhna, 450,459-460 (cf. sesin) pradhna, 164, 456 praslesa, 242 praslistanirdesa, 129-130 prasajya (pratisedha), 472, 496 prna, 463 prtipadika, 98,108, 229, 343, 361, 363,412,452,472 pluta, 73,121, 270, 303, 354, 406, 455, 466 pluti.133,299,437, 466 pvul,110 phakkik, 309,319,332 phitstra, 440, 446 bahiranga, 225, 229, 234, 236-238, 253-256,360,364-366, 368, 448, 477 bahirangalaksana, 211, 214, 219, 234, 364 bahuvacana, 13,120,463 bahuvnhi, 73-74,150,164-165,180, 191-192, 247-248, 266,292,340, 371-372, 424, 431, 450, 464, 472, 483,488,492,516 bha,72-73,75-77, 234, 242-243, 251, 253, 283, 291, 362-363, 374,452 bhavantl, 114 bhavisyati, 397, 502, 524 (future) bhavisyant, 114, 514 bhva, 501-502 bhvin, 336 bhta, 502-522 (past) mandkapluti, 128 matup,110,132,229 man, 525 mtra, 297, 302-303, 313-314, 326, 346,354,358, 387 mtrac, 128 mud, 336-337 mrdhanya, 279,283-284

yak,225-226 ya,235, 238, 374-375 ya,124 yandesa, 224 yan, 124 yuk,239 yut, 239, 243-244 yogavi bhga, 116-118,123,127-128, 224-226,240, 242-243, 361, 372, 474,480 repha, 281, 307, 312, 314, 326, 460 ru,282 la, 114 (opa) la(kra),13,5OO lat, 13, 79,114, 500, 503-504, 518, 522 lit, 13,114,120,162, 214, 243-244, 369,500-501,503-504,511 lut, 13,114,133, 343, 400, 517 Irt, 13,114, 500-501, 514, 516-517, 519-520 let, 13, 500-501 lot, 13, 500-501 la, 13, 225-226, 500, 504 li, 13,161, 380, 469, 500, 521-522 lu, 13,114,162, 223, 225, 243-244, 378,500,504 Ir,13,255, 500-501 laksana, 347, 350-351 laghu.121,362 lac, 110 liga, 75-76,396,400,463 luk, 213, 223, 228, 232-233, 258-259, 338, 345, 360, 362, 366, 374-375, 465 lup, 76-77, 223,258,338,465 luptanirdesa, 130 lumat, 223 lokanyyasiddha, 329 lopa, 110,114,122,128, 214, 216,

220-224,226-231,234,238, 242-243, 244-245, 248-251, 257,

259,291,343-346,363,371,378, 465,491-492 lyap,103,121,207,238-239,288,380 lyut, 288 vatup,291, 382-385 vanip, 229 vartamna, 513, 522 vkya, 344, 351, 355, 452, 473-474 vikarana, 225-226, 362,457 vikalpa,446, 499 vikra, 288, 323, 337, 376, 450, 455, 457 vikrti, 109, 294-295, 457 (cf. prakrti) vie, 242

548 Index

vit, 234-235 vipratisiddha, 242-243 vipratisedha, 257-258, 339, 341, 344, 441,444, 451 vibhakti, 245, 247, 292, 294, 452, 457-459,504, 507 vibhs, 108,117-119,163, 235, 266, 322,335,371,446,477 vivaks, 127, 507, 509,511,515, 520-521, 524 vivaksitatva, 212 vivrta, 299-300, 303,307, 312 visayavisayibhva, 232, 258 visarjanya, 326-327 vuk,238-243 vu.110,117 vrddha,122 vrddhi, 109,128,138-139, 221, 226-227, 240-242, 244, 250, 252-253, 277, 279-280, 290, 304, 363-364,386,389, 403-404, 455, 469 vyajana, 112,114, 244, 299, 336-337, 343,386-387,389,462-463 vyartha,359, 366-367, 370,384 vyavasth, 228, 322, 328, 448, 472 vyavast hitavikalpa, 446 vyavasthitavibhs, 132-133, 322-323 vysraya, 259 vyutpatti, 72-73 sap, 103, 220, 230, 376 sas, 385 siks, 73,401,447,461 sit, 113,121, 358-360, 372, 376-377 sista, 286-297, 425, 466 sisya.106,113 sisyaparampar, 308, 313 sesin,450 sn,230,281 snu,281 syan,103 slu, 223, 230, 258, 465 svastan, 114 sas, 114 sasth, 121, 355, 369, 471, 481, 488 samyoga, 73, 224, 228, 242, 283, 293, 299,325 samvrta, 299 sarphit, 70,140, 320-321, 335-336, 344,439 sam h i tapt ha, 335, 460 sakarmaka, 187 samkrama, 109 samkhy, 75,117,131, 382-385, 389

samjn, 71-78,108,113,117,119, 300-302, 309, 311-131, 316, 374, 386,389,403-404, 443-444, 470, 482,501 samjin, 404 san,235, 377 samdhyaksara, 114, 336-337, 461 saptaml, 386, 492, 494 saptm, 464 (opt.) sampradna, 166,186, 465 samprasrana, 117,124, 210, 219, 226, 228-229, 233-238, 244-246, 248-250, 252-257, 280, 362-363, 368-370, 455 sambandhavrtti, 128 sambandnuvrtti, 128 samnksara, 114, 336, 461 samnsraya, 231, 259 samsa, 103,116,132, 244, 266, 288, 290-291,320-321,361,370-371, 379,463-464, 483-484,488, 490, 493,495 sarvanman, 125,130, 210, 266, 472 sarvdesa, 359 sarvanmasthna, 243, 284 savarna, 112, 299-303, 306, 309, 311-312, 316-318, 321, 325-326, 333,341,365-366 sdhrana, 74 srvadhtuka, 220, 222-223, 225, 230-233,259,281,288,362,372, 387, 454 svarnya, 316-317, 323, 325-326 (cf. savarna) sie, 222, 232 sit, 337 siddha, 207-259, 341, 383, 387, 450, 478,494 siddhnta, 12, 88,95,308,315,320, 322-323,339-340 siddhntasthpaka, 308, 315, 320 siddhntin, 95, 99-100, 474 sip, 240-241 sim, 336-337 syut, 232 su,359 sut, 121,131 straptha, 159,437 sup, 7,13,118, 241-242, 244, 277, 283, 304,336-337, 374, 484 subanta(padam), 7, 9,13 stn(linga), 380-381, 428, 463 sthna,300,307,327,333,341,369, 443, 447 sthnin, 215-220, 227, 230-231, 248, 250-251,256,323,449,457 sthnivad(bhva), 214-216, 218, 231, 259, 378, 380

549 Index

sna,125 sparsa, 114, 335-336, 462-463 sprsta, 299 sphota, 405-414, 421, 467, 469-470 sya, 232 svara, 114, 293, 336-337, 343, 463 svarita, 128, 288, 341, 342-344, 350-351,353,355,438 hetu.291,465 hrasva, 73, 238, 249, 318,321, 328, 354, 455, 469

Index off Sutras

1.1.1, 128 1.1.2, 128,311,392 1.1.3, 113,128,386,388-389 1.1.5, 240-241,379 1.1.7, 293,325 1.1.9, 299-333,341 1.1.10, 299,325 1.1.11, 339-340,346 1.1.12, 340 1.1.14, 117 1.1.16, 438,499 1.1.17, 116-117 1.1.18, 116-117 1.1.20, 113,323,376-377,379 1.1.21, 433 1.1.23, 131,382-385,389-391 1.1.24, 520 1.1.26, 494 1.1.27, 125,130,210,266 1.1.28, 266 1.1.29, 109,266,270 1.1.30, 132,266 1.1.31, 266,270 1.1.32, 266-267, 269-270 1.1.33, 266-270 1.1.34, 266-271 1.1.35, 266-271 1.1.36, 92, 266-271 1.1.38, 337 1.1.45, 124,252 1.1.46, 251,383 1.1.47, 126,239 1.1.48, 337 1.1.49, 233,252,342,369,443 1.1.50, 300,311,314,328,443 1.1.51, 281,304,311-312,320, 327 1.1.52, 113,129,338,342-343, 359 1.1.54, 338,343,345 1.1.55, 129,358-360 1.1.56, 131,215-216,220,256, 337, 380 1.1.57, 217,227,230-231,259 1.1.58, 128,250 1.1.59, 378 1.1.61, 213,258 1.1.62, 223,231,252,257,363, 366, 443,515 1.1.63, 223,366 1.1.64, 386 1.1.66, 233,337-339,408,443 1.1.67, 233,337-339,345-346, 383, 408,443 1.1.68, 78-79,336-337,403,443 1.1.69, 240,302-303,312,316, 324-325, 391 1.1.70, 216,224,227,406 1.1.71, 311,320,336 1.1.72, 340,342,360-361,374

1.2.1, 173 1.2.2, 379 1.2.4, 220, 230 1.2.5, 240, 345 1.2.6, 240-242,244 1.2.10, 291 1.2.11, 128 1.2.21, 133 1.2.27, 302,307,313-314,316,324, 354 1.2.28, 316,324 1.2.32, 344,350-351 1.2.33, 344,350-351 1.2.34, 354 1.2.35, 291 1.2.37, 350-351,354,356,438 1.2.38, 109,350-351,353 1.2.39, 344,346,350-351 1.2.40, 350-351 1.2.41, 290 1.2.42, 132 1.2.43, 132,248 1.2.45, 363 1.2.46, 234,363 1.2.48, 430 1.2.50, 311 1.2.51, 77,311 1.2.53, 74-77 1.2.56, 77-78,456 1.2.57, 164 1.2.58, 396 1.2.59, 373 1.2.64, 111,481 1.2.65, 107,515 1.2.67, 481 1.3.1, 98 1.3.2, 80,320 1.3.3, 311,336 1.3.7, 130 1.3.9, 336,359 1.3.10, 130,228,443 1.3.11, 128 1.3.21, 499 1.3.25, 111 1.3.28, 215-216,499 1.3.29, 118 1.3.72, 288 1.3.78, 516 1.3.93, 133 1.4.1, 284 1.4.2, 109,129,237,243,257-258, 339,341,344-346,369,371, 441,443 1.4.6, 358 1.4.14, 283-284,364,374,451 1.4.17, 243,283-284,430 1.4.18, 243,284,362,374

551 Index

1.4.24, 471 1.4.30, 291,470 1.4.42, 18,284 1.4.43, 284 1.4.44, 133 1.4.45, 471 1.4.47, 133 1.4.55, 291 1.4.58, 117 1.4.59, 117 1.4.80, 520 1.4.89, 471 1.4.106, 494 1.4.109, 320,438 2.1.1, 109,182,344,394,483 2.1.2, 349 2.1.4, 244,490 2.1.11, 117 2.1.12, 117 2.1.13, 117,249 2.1.14, 515 2.1.16, 80 2.1.17, 133,487 2.1.18, 483 2.1.34, 483 2.1.37, 112 2.1.40, 483 2.1.48, 133 2.1.50, 470 2.1.51, 109,385 2.1.55, 490 2.1.56, 293,488,492 2.1.57, 393,483 2.1.58, 393 2.1.69, 345,483 2.1.72, 133 2.2.3, 292 2.2.4, 483 2.2.6, 483 2.2.8, 483,488 2.2.10, 483,488 2.2.11, 291,317 2.2.14, 133 2.2.15, 291 2.2.16, 488 2.2.17, 483 2.2.18, 490 2.2.19, 389 2.2.23, 268,450,516 2.2.24, 430, 488-489, 492 2.2.27, 127 2.2.28, 364 2.2.34, 268,483.488,494 2.3.6, 494 2.3.8, 331

2.3.16, 2.3.17, 2.3.18, 2.3.20, 2.3.21, 2.3.28, 2.3.37, 2.3.50, 2.3.69,

133 108 18 482 515 80 515 450,516 129

2.4.3, 437 2.4.9, 133 2.4.12, 336-337 2.4.14, 292 2.4.18, 133 2.4.23, 80,337 2.4.26, 289 2.4.32, 129 2.4.36, 380 2.4.44, 215 2.4.53, 289 2.4.54, 335 2.4.71, 257, 360, 366, 490 2.4.73, 220 2.4.75, 230 2.4.83, 487 2.4.84, 487 2.4.85, 129,130 3.1.2, 133 3.1.3, 245,430 3.1.4, 430 3.1.6, 128,376 3.1.7, 472 3.1.8, 109 3.1.10, 491 3.1.11, 491 3.1.22, 235 3.1.32, 377-378 3.1.38, 133 3.1.42, 348 3.1.48, 378 3.1.67, 225,501 3.1.68, 220,230 3.1.79, 223,230 3.1.80, 227 3.1.91, 509 3.1.92, 221 3.1.94, 131,380 3.1.95, 118,380 3.1.109, 115 3.1.118, 118 3.1.123, 348 3.1.124, 118 3.1.126, 118,133 3.1.127, 437 3.1.133, 110,380 3.1.143, 133

552 Index

3.2.1, 124,126-127,337 3.2.4, 504

3.2.6, 505 3.2.7, 505 3.2.8, 505 3.2.9, 504 3.2.10, 504 3.2.15, 504 3.2.16, 124
3.2.17, 124 3.2.21, 115 3.2.30, 133-134 3.2.48, 388 3.2.61, 252 3.2.64, 252,363 3.2.67, 234 3.2.74, 242 3.2.76, 488 3.2.79, 491 3.2.84, 502 3.2.85, 505 3.2.87, 488 3.2.88-144, Plate III 3.2.88, 436 3.2.97, 386-389 3.2.101, 388-389 3.2.105, 503-504,506 3.2.106, 503-504,506 3.2.107, 506 3.2.108, 503 3.2.109, 503 3.2.110, 504,506-508 3.2.111, 81,103,504,506-508,510, 520 3.2.112, 509
3.2.113, 500,509

3.2.161, 3.2.162, 3.2.163, 3.2.177, 3.2.188,

504 504 505 110 133

3.3.1, 62,393,502,518 3.3.4, 500

3.3.13, 506

3.3.14, 3.3.15, 3.3.16, 3.3.17, 3.3.18, 3.3.19, 3.3.30,

503 506 502 130 501 501 130

3.3.31, 437

3.3.32, 437 3.3.48, 130 3.3.90, 125,210 3.3.102, 482 3.3.119, 133 3.3.121, 118 3.3.122, 118,133 3.3.126, 518 3.3.130, 518 3.3.131, 502,518,524 3.3.133, 519 3.3.134, 80,519,521,524 3.3.135, 502 3.3.136, 80,521 3.3.137, 521 3.3.138, 521,524 3.3.139, 506,522,524 3.3.140, 506,522,524
3.3.142, 522-523

3.2.114, 3.2.115, 3.2.116, 3.2.117, 3.2.118, 3.2.119, 3.2.120, 3.2.121, 3.2.122, 3.2.123, 504, 3.2.124, 3.2.126, 3.2.131, 3.2.133, 3.2.138, 3.2.139, 3.2.141, 3.2.143,

509 500,503-506 511 126,210,511 511-512 512 512 512 512 78-79,81,84-85,494,500, 506,513,518 503 291 504 504 133 504 504 504

3.3.143, 3.3.144, 3.3.145, 3.3.146, 3.3.147, 3.3.150, 3.3.151,

522-523 523-524 522-524 523 509,522 522-523 509,523

3.3.152, 522

3.3.153, 524
3.3.155, 509,523-524

3.3.156, 3.3.168, 3.3.169, 3.3.173, 3.3.174,

521,524 509 380,389 524 524

3.2.134, 110,504

3.3.175, 524

3.3.176, 524 3.4.2, 500 3.4.6, 502,505-506 3.4.16, 515 3.4.25, 491

3.2.146, 110

3.2.147, 504

553 Index

3.4.27, 292 3.4.29, 491 3.4.40, 337 3.4.43, 491 3.4.45, 491 3.4.75, 62 3.4.77, 225, 345, 500 3.4.78, 7 3.4.79, 7 3.4.80, 7 3.4.82, 240 3.4.87, 220,223 3.4.93, 237,254, 364 3.4.113 225,362 3.4.114 , 223,362,516 3.4.115 230 3.4.116 500 3.4.117 230

4.2.7, 111,122 4.2.8, 122,296 4.2.9, 296

4.2.21 4.2.33 4.2.40, 4.2.42, 4.2.43, 4.2.55, 4.2.57, 4.2.58, 4.2.60, 4.2.67, 4.2.78, 4.2.81, 4.2.82,

119,127 403 124 124 119 127 127 127

4.2.59, 290

361 127 361 76 133 4.2.92; 450

4.1.1, 337,430 4.1.3, 396 4.1.7, 229,248

4.1.12, 4.1.13, 4.1.14, 4.1.15, 4.1.16, 4.1.17, 4.1.18, 4.1.20, 4.1.22, 4.1.28, 4.1.33, 4.1.39, 4.1.41, 4.1.44, 4.1.48, 248 248 107-108,248 118-119,250 119 128 128 247,430 111 247-248 437 110 247-248 396 271

4.3.11, 481 4.3.23, 482 4.3.53, 367 4.3.57, 292 4.3.112, 250 4.3.117, 117 4.3.118, 117 4.3.119, 116 4.3.132, 122 4.3.133, 122 4.3.143, 471 4.4.7, 329,330 4.4.17, 119 4.4.29, 133,134 4.4.35, 337 4.4.62, 374 4.4.118, 291 4.4.125, 127,437 5.1.2, 345 5.1.7, 133 5.1.16, 127 5.1.21, 108 5.1.22, 384-385 5.1.23, 382-385 5.1.36, 122 5.1.57, 117 5.1.58, 117 5.1.95, 437 5.1.115, 125,490,493 5.1.117, 109 5.1.118, 125 5.1.122, 495 5.1.124, 396 5.1.135, 437 5.1.136, 437 5.2.12,292

4.1.49, 92

4.1.62, 116 4.1.63, 425-426,430 4.1.74, 133 4.1.77, 430 4.1.82, 256,367 4.1.83, 367,428 4.1.87, 125 4.1.92, 367,428 4.1.95, 360,368 4.1.96, 133 4.1.98, 117 4.1.123, 133 4.1.162, 122 4.1.163, 122 4.1.165, 122 4.1.166, 122 4.1.167, 122 4.2.1, 126,367 4.2.2, 119

554 Index

5.2.39, 291 5.2.45, 127,361 5.2.51, 382-385,389-391 5.2.52, 382-385 5.2.53, 382-385, 389-391 5.2.65, 126 5.2.77, 127 5.2.86, 361,374 5.2.87, 361 5.2.94, 127,229,396 5.2.95, 127 5.2.96, 110 5.2.97, 110 5.2.101, 119 5.2.107, 127 5.2.108, 110 5.2.109, 110,229 5.2.112, 127 5.2.115, 127 5.3.2, 267,270 5.3.5, 120 5.3.7, 431 5.3.16, 173. 5.3.57, 80,431 5.3.65, 165 5.3.76, 494 5.3.92, 373 5.3.94, 506 5.3.106, 491 5.4.9, 425 5.4.10, 127 5.4.17, 385,431 5.4.20, 385 5.4.214, 71 5.4.25, 133 5.4.40, 119 5.4.43, 385 5.4.68, 116,370 5.4.91, 370 5.4.11, 5371 5.4.134, 431 5.4.141, 482 5.4.143, 482 5.4.145, 133-134,481 5.4.152, 372 5.4.154, 488 6.1.3, 112 6.1.6, 128 6.1.8, 230 6.1.9, 235 6.1.10, 230 6.1.11, 378 6.1.12, 345 6.1.13, 124 6.1.14, 292

6.1.15, 226,253,369 6.1.16, 210,369 6.1.17, 369-370 6.1.30, 499 6.1.32, 117 6.1.33, 117 6.1.36, 112 6.1.37, 120 6.1.45, 376-377 6.1.61, 122 6.1.62, 122 6.1.64, 85 6.1.66, 259,431 6.1.67, 234,242,252,363 6.1.68, 247,283 6.1.71, 207,217-218,328,364 6.1.72, 320-321 6.1.73, 255,329 6.1.75, 329-330 6.1.76, 330 6.1.77, 139,226,255-256,259,290, 316,338,342,345 6.1.78, 227,232,242,257 6.1.82, 247 6.1.84, 365,393 6.1.85, 217,254,364 6.1.86, 207,214-217,220,364 6.1.87, 217-218,237,242,254,312, 365-366 6.1.88, 252-253,363-364,366 6.1.89, 252-253,364 6.1.90, 133,224,226 6.1.90, 133,224,226 6.1.92, 304,311 6.1.94, 252,349 6.1.95, 365-366 6.1.97, 281,373,387 6.1.99, 122 6.1.100, 122 6.1.101, 207,217-218,255,300-332 364-365,367-368 6.1.108, 229, 244-245, 252-253, 344 363,366,368-369 6.1.109, 217-218 6.1.110, 242 6.1.113, 217 6.1.115, 120 6.1.122, 322 6.1.123, 322-323,406 6.1.124, 121,323 6.1.125, 121 6.1.127, 312,318,321 6.1.128, 303,312,321,323 6.1.132, 337 6.1.135, 122 6.1.136, 122 6.1.137, 121-122 6.1.138, 121-122

555 Index

6.1.144, 109 6.1.150, 121 6.1.156, 6.1.157, 6.1.158, 6.1.161, 6.1.163, 6.1.164, 6.1.165, 6.1.166, 6.1.168, 6.1.182, 6.1.184, 6.1.185, 6.1.197, 6.1.223, 122 122 340-342,347,430 245-248,371 107,117,370 117 117 110 245-247 245-247 130 341,350-351 117 343

6.4.34, 6.4.35, 6.4.36, 6.4.37, 6.4.41, 6.4.43,

222 219-222 220, 226 214,220,226-227 234-235 235

6.4.48, 214,220,226-227,235, 238-239 6.4.49, 235,238-239 6.4.51, 231-232,378 6.4.56, 121,238

6.4.62, 162,225,232,258
6.4.63, 239,243-244 6.4.64, 229, 234, 249, 255

6.4.71, 223,225,378 6.4.72, 223-226 6.4.74, 225-226,524

6.4.76, 6.4.77, 6.4.81, 6.4.82, 229 234,239-241,243,249 224 224, 227, 234, 239, 244

6.2.2, 117 6.2.82, 340-341 6.2.83, 340-341 6.2.107, 118 6.2.108, 118 6.2.121, 345 6.2.134, 438

6.4.83, 240-243, 257-258

6.4.88, 238-241,243,249 6.4.89, 240-241, 243 6.4.92, 238 6.4.98, 213,233,291 6.4.100, 199 6.4.101, 208,219-220,222 6.4.104, 213,223,232-233,258-259 6.4.105, 226,228 6.4.106, 112,223,228,230 6.4.107, 231-232

6.2.143, 118
6.2.144, 6.2.172, 6.2.175, 6.2.193, 6.2.197, 118 247 247,249 370 371

6.3.5, 119 6.3.6, 119 6.3.7, 119 6.3.14, 393 6.3.34, 125,430,482 6.3.35, 431 6.3.37, 430 6.3.40, 119 6.3.41, 430-431 6.3.42, 258,430 6.3.62, 130 6.3.83, 119 6.3.84, 290 6.3.97, 338 6.3.109, 99 6.3.111, 281,283 6.3.138, 255 6.3.139, 124,258 6.4.1, 221,373 6.4.6, 130 6.4.19, 252,255-256 6.4.22, 208-259,368 6.4.23, 213 6.4.24, 215-216,222 6.4.27, 221 6.4.33, 221

6.4.108, 230-232,234,259 6.4.109, 230-231,259 6.4.110, 109,223,230-232,259

6.4.111, 216,222,224,227 6.4.112, 230 6.4.113, 230 6.4.119, 208,219-222 6.4.120, 215-216 6.4.128, 244,253 6.4.129, 211,219,236,244,253 6.4.131, 234,236-237,249,252, 256-257, 368 6.4.132, 208-259,363-364,368 6.4.133, 229,244-247

6.4.134, 6.4.137, 6.4.140, 6.4.141,

6.4.142, 6.4.144, 6.4.146, 6.4.148, 6.4.149, 6.4.155, 6.4.157, 6.4.158,

244-245,247-248 245,248 234-235, 242-243, 257-258 349

130 371,374 248 229,248,250,430 250-251 251 251 248,372,470

556 Index

6.4.160, 6.4.163, 6.4.167, 6.4.172, 6.4.175,

372-373 251 374 374 211,219,229,236,244

7.1.1, 344 7.1.3, 389 7.1.8, 112 7.1.9, 329-330 7.1.10, 112 7.1.14, 266 7.1.15, 385 7.1,16, 268-272 7.1.17, 319 7.1.18, 107 7.1.25, 121 7.1.26, 258 7.1.27, 129 7.1.30, 373 7.1.37, 103,238,288,364,380 7.1.43, 348 7.1.48, 133 7.1.54, 258 7.1.58, 227 7.1.70, 366 7.1.72, 110 7.1.75, 360 7.1.81, 103 7.1.84, 359 7.1.90, 215 7.1.91, 242 7.1.96, 345 7.2.1, 139 7.2.2, 139,320 7.2.3, 139 7.2,4, 139 7.2.5, 139 7.2.7, 139 7.2.16, 133 7.2.17, 105,113 7.2.35, 234, 255-256, 368 7.2.70, 232 7.2.90, 373 7.2.95, 360 7.2.96, 360 7.2.98, 257,360,366 7.2.102, 373,387 7.2.114, 481 7.2.115, 232,240-242 7.2.116, 221,238,378 7.2.117, 250,254,337,368 7.3.15, 7.3.17, 7.3.19, 7.3.30, 111 111 118 373

7.3.44, 369 7.3.46, 369 7.3.54, 378 7.3.55, 378 7.3.56, 377-379 7.3,59, 378 7.3.59, 378 7.3.75, 121 7.3.77, 121 7.3.78, 288,373 7,3.79, 372-373 7.3.82, 130 7.3.83, 130 7.3.84, 139,223,230-232,240-242 387 7.3.85, 139 7.3.86, 222,227-228,252,363-364 373 7.3.95, 109,281 7.3.101, 118,276-278,281,372 7.3.102, 118,276-278,281 7.3.103, 278,373 7.3.104, 373 7.3.111, 358 7.3.117, 117 7.3.118, 117 7.3,119, 117 7.4.1, 378 7.4.2, 115 7.4.13, 281,369 7.4.25, 372 7.4.32, 109 7.4.41, 323 7.4.52, 161 7.4.59, 376 7.4.60, 369,376,516 7.4.79, 376,378 7.4.82, 376 7.4.83, 235,374-376 7.4.84, 375 7.4.85, 235,374-376 7.4.93, 376,378 7.4.94, 378 8.1.12, 8.1.16, 8.1.17, 8.1.18, 8.1.19, 8.1.28, 8.1.29, 8.1.36, 8.1.42, 8.1.51, 8.1.67, 8.1.73, 8.1.74, 396 284,341-343 341-343 341-343,347 341,343 342 343 515 515 110 120 120 120

557 Index

8.2.1, 213-214,254,268,324,335, 350-351 8.2.4, 342 8.2.5, 245 8.2.7, 364,430 8.2.8, 92 8.2.12, 120,122 8.2.18, 324,407 8.2.25, 222 8.2.26, 283 8.2.29, 283 8.2.31, 283-284 8.2.32, 283 8.2.35, 283 8.2.39, 282-284 8.2.41, 283 8.2.56, 322 8.2.66, 217,222,282 8.2.86, 302-303,316,324-325 8.2.101, 489 8.2.106, 113 8.3.1, 282 8.2.3, 282 8.3.4, 282 8.3.5, 110,131 8.3.6, 282 8.3.7, 274,282 8.3.13, 268-269,283 8.3.14, 330 8.3.15, 236-237,254,282 8.3.16, 282 8.3.17, 222 8.3.19, 222 8.3.34, 282 8.3.54, 284 8.3.57, 283 8.3.59, 217-218,232,234,283,337 8.3.78, 162-163 8.3.79, 163 8.3.108, 234 8.3.118, 120 8.4.1, 304-307,328,331 8.4.2, 304-305 8.4.3, 125 8.4.4, 108 8.4.7, 107-108 8.4.8, 108 8.4.17, 377 8.4.28, 121,290 8.4.29, 305 8.4.34, 378-379 8.4.40, 282 8.4.41, 268-269,283 8.4.42, 284 8.4.45, 335 8.4.47, 244

8.4.48, 8.4.53, 8.4.55, 8.4.56, 8.4.57, 8.4.58, 8.4.60, 8.4.62, 8.4.64, 8.4.66, 8.4.68,

111 268-269, 282-283 284 283 281 325 330 327-328 244 343-344, 350-351 345