Professional Documents
Culture Documents
General Editor
E.F. KONRAD KOERNER
(University of Ottawa)
Volume 48
Edited by
1987
This book was published with the financial support
of the Consiglio Nazionale delle Ricerche.
Preface page ix
General Programme of the Conference » xi
John Anderson, Gothic obstruents: The limits of reconstruction » 1
Françoise Bader, Structure de l'énoncé indo-européen » 13
Joëlle Bailard, Il s'en va où le français, et pourquoi? » 35
Giuliano Bernini, Attempting the reconstruction of negation pat
terns in PIE » 57
Pier Marco Bertinetto, Structure and origin of the «narrative»
imperfect » 71
Bernard H. Bichakjian, The evolution of word order:
paedomorphic explanation » 87
Joan L. Bybee and William Pagliuca, The evolution of future
meaning » 109
Theodora Bynon, Syntactic change and the lexicon » 123
Gualtiero Calboli, Die Syntax der ältesten lateinischen Prosa » 137
Andrew Carstairs, Diachronic evidence and the affix-clitic
distinction » 151
Thomas D. Cravens, The syllable and phonological strength: Gra
dient loss of gemination in Corsican » 163
William Croft, Hava Bat-Zeev Shyldkrot, and Suzanne Kemmer,
Diachronic semantic processes in the middle voice » 179
Gerrit J. Dimmendaal, Drift and selective mechanisms in mor
phological change: The Eastern Nilotic case » 193
Dorothy Disterheft, The diachronic relationship of morphology
and syntax » 211
Nils E. Enkvist and Brita Wårvik, Old English þa, temporal
chains, and narrative structure » 221
Thomas Fraser, The establishment of «by» to denote agency in
English passive constructions » 239
vi TABLE OF CONTENS
m
PREFACE
9.00 Registration
9.30 Official opening of the Conference (Aula Magna dell'Università)
10.00 Malkiel, Y.: integration of phonosymbolism with other categories of lan
guage change
10.40 Lazzeroni, R.: Considerazioni sulla cronologia relativa dei mutamenti fo
netici
11.20 Lüdtke, H.: Auxiliary verbs in the universal theory of language change
12.15 Cocktail given by the Rector in the Rector's reception rooms
SECTION I SECTION II
Chair: Martin Harris Chair: Elisabeth Traugott
15.00 Hock, H.H.: Yes, Virginia, syn 15.00 Neroznak, V.P.: Problems of lin
tactic reconstruction is possible guistic reconstruction: New per
spectives
15.35 Calboli, G.: Die Syntax der älte 15.35 Dezsö, L.: Historical typology of
sten lateinischen Prosa morphosyntax: Past, present and
future
16.10 Hopper, P.J.: Syntactic meaning 16.10 Bader, F.: Structure de l'énoncé
and syntactic change: Some obser indo-européen
vations
Break Break
17.00 Bichakjian, .: Typology and evo 17.00 Bynon, Th.: Historical syntax and
lution of word order the substance-structure dichoto
my
17.35 Bailard, J.: Il s'en va où le fran 17.35 Lehmann, W.P.: Time in histori
çais, et pourquoi?: The explana cal linguistics
tory value of pragmatic and syn
tactic functions in word order
change
XII GENERAL PROGRAMME
SECTION I SECTION II
Chair: Georges Lüdi Chair: Robert Jeffers
9.00 Cravens, Th.D.: The syllable and 9.00 Dietz, K.: Lehnwortphonologie
phonological strength: Gradient und englische Sprachgeschichte
loss of gemination in Corsican
9.35 Wanner, D.: On the persistence 9.35 Pagliuca, W. - Mowrey, R.: Arti-
of 'imperfect grammars': Clitic culatory evolution
movement from Late Latin to Ro
mance
10.10 Stehl, Th.: Sostrato, variazione 10.10 Shannon, Th. F.: The rise and fall
linguistica e diacronia of final devoicing
Break Break
11.00 Skytte, G.: Categoria e funzione 11.00 Dörum, H.: Disorder and regula
nella linguistica diacronica, con rity in linguistic change
particolare riguardo all'introdut
tore dell'infinito nelle lingue ro
manze
11.35 Bertinetto, P.M.: Structure and o- 11.35 Lubbe, H J . : Onset 5p-, st- and
rigin of the «narrative» imperfect sk- in Germanic languages
12.10 Parry, M.M: «Conditional para 12.10 Minkova, D.: The prosodic cha
digms» in Piedmontese and Ligu- racter of schwa loss in Old English
rian dialects and schwa preservation in Early
Modern English
Break Break
SECTION I SECTION II
Chair: Alberto Mioni Chair: Robert Stockwell
15.00 Luraghi, S.: Patterns of case syn 15.00 Dimmendaal, G.J.: Allomorphism
cretism in Eastern Nilotic languages: So
me dynamic historical processes
15.35 Posner, R.: Creolization and lin 15.35 Disterheft, D. : On the relationship
guistic change of syntactic and morphological
change
16.10 Campbell, L.: Pipil language 16.10 Carstairs, A.: Diachronic eviden
change ce and the affix-clitic distinction
Break Break
17.00 Dolde, K., Eichinger, L.: Syntak 17.00 Werner, O.: Neither uniform lan
tischer Wandel und Textverständ guage type nor agglutination are
nis am Beispiel von Grimmelshau- the aims of morphological change
sens Simplicius Simplicissimus
GENERAL PROGRAMME XIII
17.35 Price, G.: Substrata, superstrata, 17.35 Bybee, J.L. - Pagliuca, W.: The
adstrata and the early history of evolution of future meaning
French
18.10 Harris, J.: On doing comparati 18.10 Fleischman, S.: Aspectual featu
ve reconstruction with genetical res of tense oppositions in narra
ly unrelated languages tive discourse: A reversal of pre
dictable diachronic patterns?
SECTION I SECTION II
Chair: Nils Erik Enkvist Chair: Konrad Koerner
9.00 Ronneberger-Sibold, E.: A per 9.00 Wurzel, W.U.: Paradigmenstruk-
formance model for a natural turbedingungen: Aufbau und Ver
theory of linguistic change änderung von Flexionsparadigmen.
9.35 Lüdi, G.: Reflets de la variation 9.35 Krisch, Th.: The role of presup
du français à l'époque de la revo positions in Old Norse compara
lution dans un pastiche anti- tive constructions
revolutionnaire
10.10 Winters, M.E.: Syntactic and se 10.10 Traugott, E. - Dasher, R.: On the
mantic space: The development of historical relation between mental
French subjunctive and speech act verbs in English.
Break Break
11.00 Weigand, E.: Historical dialogue 11.00 Manzelli, G.: The Nostratic back
grammar exemplified by poetic ground of Hungarian nyár «sum
dialogues of «Kudrun» mer»
11.35 Croft, W. - Shyldkrot, H.B. - 11.35 Splett, J.: Bedeutung und Bedeu-
Kemmer, S.: Diachronic seman tungsindizierung im Rahmen der
tic processes in the middle voice Wortfamilien des Althochdeut
schen
12.10 Baron, N.S.: The role of compu 12.10 Panagi, O.: Bahuvrīhi, Ellipse,
ter in language change Synekdoche
Afternoon
Excursion to the Certosa di Pavia
SECTION I SECTION II
Chair: Simon Dik Chair: Winfred Lehmann
9.00 Maisch, D.L.: The grammaticali- 9.00 Galton, H.: From Indoeuropean
zation of social relationship: The perfect to Slavic perfect to Slavic
origin of number to encode defe preterite
rence
XIV GENERAL PROGRAMME
9.35 Manoliu, M.: Pragmatic strategies 9.35 Prosdocimi, A.L.: La legge di Sie
and linguistic change vers come species del genus silla-
bicità
10.10 Reiter, N.: Beobachtungen zur se 10.10 Hoenigswald, H.M.: A and
mantischen Entwicklung von Be the prehistory of Greek noun in
ziehungsadjektiven im Russischen flection
Break Break
11.00 Fraser, Th.: The establishment of 11.00 Rosén, H.B.: On «normal» full
by to denote agency in English root structure and its historical de
passive constructions velopment
11.35 Conradie, J.: Semantic change in 11.35 Jeffers, R.J.: Subordination in
modal auxiliaries as a result of Indo-European: The emergence of
speech act embedment grammatically signalled incorpo
ration
12.10 Enkvist, N.E. - Wårvik, .: Old 12.10 Bernini, G.: Attempting the re
English þā: A study in diachronic construction of negation patterns
text linguistics in PIE
Break Break
SECTION I SECTION II
Chair: Theodora Bynon Chair: Thomas Fraser
15.00 Hoppenbrouwers, : The insta 15.00 Wright, R.: The study of seman
bility of peripheral /./, / ø . / and tic change in Early Romance (La
/ . / in Dutch lects te Latin)
15.35 Longobardi, G.: Teoria dei para 15.35 Milani, : Note su / s / intercon
metri sintattici e linguistica stori sonantica nei dialetti greci antichi
ca: alcune asimmetrie tra il goti
co di Wulfila e l'originale greco
e la loro spiegazione
Break
16.30 Business Meeting of the Internat
19.30 Dinner at Collegio Nuovo
Closing remarks
*[ ] around names mean that the paper was presented by a colleague and/or di
scussed.
GOTHIC OBSTRUENTS: THE LIMITS OF RECONSTRUCTION
JOHN ANDERSON
University of Edinburgh
* This paper grew out of joint work on Gothic with Fran Colman. I am grateful to be
able to use here ideas that may have originated with her; she is, of course, not responsible
for any naughty bits.
2 JOHN ANDERSON
1 a. / k / b. / x / . / 3 /
kuni 'race' hawi 'hay' gawi 'region'
(ge-)lauk 'I-woke' þlauh 'I-fled' baug 'I-bent'
(us-)waikjan 'to-wake' hlahjan t-laugh' wagjan 'to-shake'
stiks 'point' slahs 'stroke' wigs 'journey'
These units are consistently separated in the orthography 1 . We have com
parative and graphogenetic evidence for the voiceless velar plosive in (1.a),
whose orthographic representation is based on the Greek uncial kappa. The
character of the fricative in (b) is perhaps a little less certain: the symbol is
used to transliterate the Greek «spiritus asper» (though also occasionally kappa
— Penzl 1950: 221) and is sometimes inserted in Greek loans medially bet
ween vowels in hiatus. This perhaps indicates that at least non-finally the unit
is not velar but already an aspirate, as in other Germanic languages, and as
is compatible with the use of the apparently Latin-based symbol <h>. But
the real problem is usually identified with (l.c).
Its etymological source turns up as voiced stop or continuant in other
Germanic languages. In some positions, at least, the stop value develops from
an earlier fricative, as in initial position in Old English, for instance (Hogg
1979, Anderson 1985). But that isn't very helpful in evaluating the state of
Gothic in this respect: even if all instances represented by <g> go back to
a fricative in some pre-Gothic stage, this does not, in the absence of a voiced
fricative/plosive contrast, determine the character of the Gothic unit; nor,
a fortiori, can we, in the absence of further evidence, reconstruct allophonies
for this unit. On this, see particularly Penzl 1950: 222. The use of a symbol
for the unit based on gamma is likewise unhelpful, given the lack of a
plosive/continuant contrast among voiced velars in both Gothic and Greek.
This is why I have used the eccentric phonemic symbol at the head of (l.c),
as neutral between / g / and / γ / . It seems to me that we have no way of
legitimating a choice between stop and continuant here. Thus, Voyles
(1981:10), for instance, is rather apologetic in opting for one phonemicisa-
tion rather than the other: «In instances of this type where two possible
analyses are so similar, we see no principled way of deciding between them».
1
I adopt here one of the usual romanisations of the Gothic alphabet (exemplified in
Wright 1910), but I include no editorial diacritics. The bulk of the Gothic we have is the substantial
remains of a fourth century translation (allegedly by Bishop Wulfila) of the Bible which is preserved
in a sixth century manuscript (Streitberg 1908). The alphabet is unique but clearly takes the Greek
uncial and to some extent the Roman and Runic alphabets as its model(s): see e.g. Wimmer 1887:
259-74.
GOTHIC OBSTRUENTS 3
2
On dependency phonology see e.g. Anderson and Jones 1977; Anderson 1980; Ewen
1977; Anderson and Ewen 1980, forthcoming; Durand forthcoming; Lass 1984: ch.ll; as well
as works referred to in the following.
4 JOHN ANDERSON
{C;V} ⇒{C.V;V}
where { ] enclose gestures and the discontinuous lines represent association
(here between gestures).
6 JOHN ANDERSON
| = /z/
|
{C;V}
{ }
|
| = /s/
{|C|}
and the redundancy in (8), rather than that in (6):
(8) {|l|}
|
|
{ C} ⇒ {V}
Thus, (8) as well as (5) applies to both / z / and / s / .
As the least marked obstruent, / s / is able to initiate clusters like those
in (9), apparently in defiance of the sonority hierarchy:
(9) speiwan «to-spit», steigan «to-ascend», skaidan «to-divide»
Such clusters also, as is well-known, behave as a unit under reduplication;
cf. (10.a) and (l0.b) with (l0.c):
(10) a. haitan «to-call» ⇒ haihait «I-called»
b. fraisan «to-tempt» ⇒ faifrais «I-tempted»
c. skaidan «to-divide» ⇒ skaiskaid «I-divided»
Whereas in (10.b) only the first consonant is reduplicated (as in (10.a)), in
(l0.c) the whole cluster is involved. This would make sense if the cluster in
examples like (lO.c) has the initial structure in (11):
(11) { J {|u|/d/l.u}
{|c|}
such that / s / gains a separate categorial gesture only by virtue of application
of (5) and (8). Thus, there is no infringement of the sonority hierarchy in
such clusters; and reduplication involves replication of the first consonant
GOTHIC OBSTRUENTS 7
in all instances: s + etc. are categorially simplex, such that a single gesture
is associated with two distinct articulatory gestures.
The / s / ≠ / z / contrast is also very generally neutralised. There is a con
trast only between sonorants:
(12) hazjan «to-praise», (ga-)marzjan «to-offend», maiza «greater» vs.
nasjan «to-save», lausjan «to-loose», galisa «silk»
Elsewhere we find a neutralised unit representable as in (13):
(13) { }
|
{C}
i.e. the minimal consonant, to which will apply the redundancies in (5) and
(8) to give (14):
(14) {|{C.V}
1|}
Mostly, the realisation of this is spelled with an <s>:
(15) sokjan «to-seek», smals «small», haihs «half-blind», andalaus «endless»
However, we do find <z> before a tautomorphemic voiced obstruent:
(16) huzd «treasure», azgo «ash» vs.
dis-dailjan «to-divide», us-bairan «to-carry-out»
The reahsational statement in (17) therefore seems appropriate for the unit
of (13/14):
(17) a. {C.V} ⇒ {C.V;V}/ {C;V}
b. ⇒{|C.V|}
if we take the spelling as our guide.
However, there is also evidence that in some cases the neutralisation is
derivative rather than initial/lexical. Consider alternations like that in the
strong neuters in (18):
(18) riqis «darkness» nom. sg. vs. riqizis gen. sg.
dius «wild beast» nom. sg. vs. diuzis gen. sg.
Between sonorants / z / and / s / are in contrast; compare (19) with (18):
(19) gras «grass» nom. sg. vs. grasis gen. sg.
agis «fear» nom. sg. vs. agisis gen. sg.
8 JOHN ANDERSON
That <h> represented an aspirate even in final position is perhaps suggested by its fre
quent assimilation to following consonants: iþ-þ rather than nih-þan, for instance. But this
phenomenon is limited to closed-class items. On the weakness of the traditional argument for
supposing that / 3 / devoices see Voyles 1981: 70, though / x / and / 3 / are even more generally
contrastive than he suggests. We find <z, b, d> retained in a minority of instances (Streitberg
1905-6: § 1) where by (20/24) we would expect <s, f, þ>. It is, I think, indeterminable whether
this represents (spelling or phonological) analogy or variable application of (20/24) (cf. Voyles
1981: 71).
10 JOHN ANDERSON
REFERENCES
ANDER
STRUCTURE DE L'ÉNONCÉ INDO-EUROPÉEN
FRANÇOISE BADER
E.P.H.E. IVème Section, Paris
1
De Brugmann, 1925, à Panhuis, 1983, en passant, essentiellement, par Watkins, 1963
et 1964, et Dressier, 1969. De nombreuses études portent sur l'ordre des mots, pour lequel
voir Schwentner, 1923, 1926, 1938, 1952, 1967 (lui ajouter Zimmer, 1976, avec la discussion
de Bader, 1978); des travaux particulièrement nombreux ont porté sur la phrase germanique,
ainsi Behaghel, 1932; Delbrück, 1911; Fourquet, 1938; Bacquet, 1962. Les recherches qui ont
porté sur la question de savoir si PIE. présentait l'ordre SOV, SVO, ou VSO dans le cadre
typologique des universaux linguistiques de Greenberg, 1963, sont d'origine américaine (Leh
mann, 1974; P. Friedrich, 1975; Miller, 1975), comme les critiques, justifiées, qui leur ont été
faites par Watkins, 1976; Strunk, 1977, discute de la valeur de la comparaison typologique.
Ma conclusion sera que les trois ordres ont chacun une fonction, que permet d'atteindre la
seule reconstruction i.e. interne.
2
Cet article n'est qu'un résumé; je n'y ai pas traité le problème de l'ordre de la princi
pale et de la subordonnée; ni le cas de tous les verbes initiaux; pour le problème des traduc
tions, voir note 13.
3
Laroche, 1959, 154.
14 FRANÇOISE BADER
4
Bader, 1975 ; 1981 a.
5
Voir Klein, 1985, 116-118, pour les rapports de *kwe indéfini, généralisant, inter-
rogatif.
ENONCE INDO-EUROPEEN 15
6
Bader, 1983 a, b, c.
7
Bader, 1979; 1982.
8
Bader 1982, 88-95; 1983 a et b.
16 FRANÇOISE BADER
9
Pour l'accentuation védique du vocatif, Macdonell, 1955, 564-66; pour l'accent du ver
be i.e. comme marque de dépendance, Meillet, 1933.
10
Pour l'atonie et la position enclitique du verbe, ainsi que sa tonicité subordonnante,
Delbrück, 1888, 35; 1911; Wackernagel, 1892, 93-102; 1877; 1893. Behaghel, 1929; Hirt, 1929,
293-369; Kurytowicz, 1958, 98; Minard, 1956, § 101 b; Renou, 1965, 230, entre autres.
11
Jucquois, 1970.
12
Wackernagel, 1877; Vendryes, 1958, 112-115.
13
Je me suis servie essentiellement des deux états de langue archaïques qui notent l'ac
cent, védique et grec, ainsi que du hittite, en raison de son archaïsme, et du germanique, parce
qu'il a été particulièrement bien étudié; j'ai laissé de côté le tokharien, malgré les bonnes don
nées fournies par Zimmer, parce que je les ai déjà discutées (Bader, 1978); sauf exception,
j'ai pris mes exemples védiques dans Macdonell, 1956, et hittites dans Friedrich, 1960, aux
quels je renvoie pour les traductions.
ÉNONCÉ INDO-EUROPÉEN 17
3.2 Ces vues seront d'abord illustrées par quelques exemples compa
rables à la phrase a) du louvite
1 2 3
louv. a-ta imrassa-dIM-unti pari tarauitta
R.V.2,12,13 a:
b:
R.V.1,85,12 a:
II. I 100
104
106
Ensuite, j'étudierai en détail: A. le segment initial; B. le segment final;
C. les variations de place des thèmes pronominaux qu'on peut trouver dans
l'un ou dans l'autre, en «tmèse» ou en «univerbation» avec le verbe.
A. Segment initial
4.1 Le problème spécifique de ce segment, en articulation neutre, ré
side dans l'ordre des éléments de la chaîne pronominale (et se retrouvera,
identique, dans l'articulation marquée, mais postposé au verbe). Il est fixe
pour une langue donnée, avec des constantes d'une langue à l'autre, mais
aussi des variations. Je partirai de l'anatolien, étudié magistralement par
E. Laroche 14, et je distinguerai trois groupes dans la chaîne hittite, du type
14
Laroche, 1957-58, 145.
18 FRANÇOISE BADER
15
Sur -san, -kan, -tar, etc., difficiles à cerner, voir Götze, 1933; Lee, 1966; Josephson,
1972.
16
C'est aussi, p.ex., l'ordre du v. slave, et du v. anglais: Vaillant, 1979, § 1487; Bac-
quet, 1962, p. 96.
17
«Now, give him to me»: Sturtevant, 1935, 65.
ÉNONCÉ INDO-EUROPÉEN 19
4.3 Dans une langue donnée, l'ordre de la chaîne peut présenter des
flottements: ainsi, outre les faits grecs, l'on voit le slave avoir une chaîne
de particules ze (gr. «de») li (particule interrogative) bo (gr. «gár»), ou li
ze bo, suivies, dans les deux cas, des référents dans l'ordre datif (enclitique)
- accusatif (semi-enclitique)18; et le hittite, avec préverbe ser-, ser-wa-ssi,
et ser-sit-wa 19: dans le premier cas, la particule de discours rapporté est à
la place usuelle des ligateurs; le second offre un ordre qui, en termes de gram
maire celtique, serait celui d'une «préposition conjuguée» 20 .
. Segment final
18
Vaillant, 1979, § 1487.
19
Friedrich, 1960, § 288.
20 Thurneysen, 1946, p . 272-276.
20 FRANÇOISE BADER
5.2 Le verbe peut être final, et alors atone ou tonique (cf. namete et
santi - accentué, parce qu'il est dans une relative -, § 3.2). Mais il peut se
trouver aussi dans le segment final ailleurs qu'à la finale absolue; il est:
a) antépénultième (et atone), en R.V. 4, 51, 10 b (§ 5.1), où l'impéra
tif de YAM- est suivi du locatif du pronom tonique de 1ère p.plur., et d'un
vocatif atone;
b) pénultième, tonique, ou atone, selon que le pronom final est lui-
même, inversement, l'un ou l'autre: R.V. 4,50, 11
a:
b:
Ces chaînes finales où s'accolent verbe et pronom, dont l'un est atone,
l'autre tonique, doivent être semi-enclitiques: vraisemblablement, dans une
langue à accent de hauteur, la voix tombait en fin de phrase, mais moins
que dans la portion atone de l'initiale.
6.2 Pour les préverbes21, la tmèse est le fait d'une articulation non
marquée (e.g. R.V. 2, 33, 7 d: ), l'uni
verbation, d'une articulation marquée, soit en parataxe (ainsi, louv. pari
tarauitta, § 1.2), soit en hypotaxe: en védique, la soudure du préverbe et
du verbe est normale en subordination 22 ; elle n'est pas encore accomplie
chez Homère, où on trouve à la fois II. IV 319 ka-
, La soudure de préver
be et verbe tient à l'inversion des rapports accentüels de l'un et de l'autre
en articulations neutre et marquée: dans la première, le verbe est atone, et
le préverbe porte l'accent à des endroits de la phrase qui appellent celui-ci;
dans la seconde, parce que le verbe est tonique, le préverbe, dans le système
supplétif qu'il forme avec lui, devient proclitique (comme les post- et pré
positions dans les syntagmes nominaux) 23 .
21
Kurylowicz, 1964, ch. VII; Watkins, 1963, 31-41; Lehmann, 1974, 232-35; Friedrich,
1975, 34-39.
22
R e n o u , 1952, 316.
23
Je laisse de côté les cas ambigus, pour lesquels on ne sait pas si on a affaire à un pré
verbe, ou une postposition nominale: Götze, 1963 (pour le hittite).
22 FRANÇOISE BADER
6.4 Dans les exemples grecs ci-dessus, verbe et postverbe sont acco
lés, ou séparés seulement par un enclitique (V 308); il peut en être de même
en v. angl. (wulfas atugan a stacan up) 25 , ou en sanskrit (R.V. 1, 85, 10
b: cid bibhidur ví parvatam), où, cependant, l'un et l'autre peu
vent être séparés par un mot (nom) tonique (R.V. 10,14,12 b:
26
. Et le postverbe en parataxe peut être sujet à une
tmèse inverse de la tmèse habituelle, avec verbe en tête et préverbe final:
hitt. KUB XXVIII 4 II 15 austa-ma-kan dKamrusipas nepisaz katta; la mê
me structure survit, mais vigoureusement, en hypotaxe, dans la principale
d'un système subordonné allemand (§ 12.2).
24
Vendryes, 1938, 239-49.
25
Bacquet, 1962, p . 8 1 .
26
Voir Macdonell, 1956, p . 285 p o u r la postposition des préverbes.
27
Je laisse donc de côté án; p o u r l'intéressante analyse d ' a r c , apseudēō-nan, Dubois,
1984, 161-62.
28 Friedrich, 1960, §§ 280; 330.
ÉNONCÉ INDO-EUROPÉEN 23
29
Delbrück, 1897, 521-24; Wackernagel, 1957, 348-313; L e h m a n n , 1974, 153-54; Sei
ler, 1952.
30
Dover, 1960, 14.
24 FRANÇOISE BADER
8.2 En hittite 31 :
a) en tmèse: KUB VIII 48 I 12 UL -war-an-kan tuētaza memiyanaz
kuennir;
b) en univerbation:
I o préverbée (en fin de phrase): III 4 III 17 sq. nu namma INA
ÍD
seha UL päun;
2 o postverbée: Madd. I 35 nu-war-an sannatti-ya le murināsi -ya-war-
an lē (articulation marquée où se rapprochent les verbes de première et se
conde phrases); le verbe est séparé de la négation par une chaîne, BoTU
23 A II 44 (qui a, de plus, un pronom final): parkunusi-ma-za UL kuit.
8.3 En sanskrit 32 :
a) tmèse: na vai nünam bhagavantas ta etad avedisuh; ou, en second
mot: tebhyo na sarvam iva pratipatsye;
b) en univerbation:
1 ° préverbée, p.ex. avec verbe composé, kim svāminam api na jānāsi,
et ;
2° postverbée: pathi yad ajāyata tad ahar jānāmi no.
8.4 En latin:
a) en tmèse: Cic., Phil. 8,12 non, si tibi antea profuit, semper proderit;
b) en préverbation (Cic, Mur. 84 sine gemitu hoc dici non potes).
8.5 La préposition de la négation est la place préférée de la négation,
en sanskrit, hittite, ou latin, par une tendance très largement suivie, qui se
manifeste ailleurs33: en tokharien (A ); en arménien,
Fon a des exemples de tmèse (Ezn. 187 ew «et non les cieux
ne tournent pas»); mais la préverbation est plus fréquente (duk' oč' mta-
nêkr «vous, vous n'entrez pas»); de même en vieux slave, tmèse (L. 6, 39
ne oba li se), mais plus souvent préverbation (Mat. 8, 10
ni I, comme en lituanien (tmèse: kur es biju
negājusi; mais préverbation normale: es tur neęsmu bijis). En gotique, la
31
Friedrich, 1960, § 279-283.
32 G o n d a , 1951; M i n a r d , 1936, § 20-35.
33 Tokharien: Krause, in Lehmann, 1974, 124; arménien: Jensen, 1959, § 500 (et lettre
du 28-3-1984 de Ch. de Lamberteire, que je remercie ici); slave: Meillet, 1928; lituanien: End-
zelin, 1923, § 836; germanique: Jespersen, 1917; Mossé, 1956, § 252; Fourquet, 1938, 249-51;
Bacquet, 1962, 178-182; 629-46; v.irlandais: Thurneysen, 1946, § 860.
ÉNONCÉ INDO-EUROPÉEN 25
négation de phrase est toujours préverbée (Mt. 6, 26 pei ni saiand nih snei-
pand); en vieil irlandais, elle n'est séparée du verbe que par une particule
conjointe (ni-dénat firtu úili), ou un réfèrent enclitique infixé (ni-n-fortéit),
comme p.ex. dans II. I 32 (§ 8.1).
34
Denniston, 1970, 282-85.
35
Pour les pronoms infixés du v. irlandais, Thurneysen, 1946, 255-70.
36
K r a u s e - T h o m a s , 1960, 162-63.
37
Friedrich, 1960, §§ 237; 2 3 9 .
38
Meillet-Vaillant, 1965, 292.
39 F o u r q u e t , 1938, 271-76.
26 FRANÇOISE BADER
40
J ' a i a b o r d é les problèmes q u e je reprends rapidement ici à plusieurs reprises; 1973
a, 43-48; 1973 b,93-109; 1974 a, 173-180; 1974 b,861-869; 1975 a; 1975 d; 1976,43-44; 1977; 1978.
41
Traduction de Perpillou, 1972, 180.
ÉNONCÉ INDO-EUROPÉEN 27
42
R e n o u , 1952, § 393.
43
P o u r l'inversion du sujet en védique, Jacobi, 1895; en germanique, Mogk, 1894; Ries,
1880.
28 FRANÇOISE BADER
be reste à la finale; le relatif, alors en tmèse, peut occuper une des deux
portions de l'initiale: atone (hitt. daliya-
44
nun nu-smas -us tehhun ); ou, plus souvent, tonique (R.V.
2,12,2 a: ) 45 .
12.4 Je fais figurer les données de cet exposé dans le tableau suivant
(a. = articulations; s. = segment)
44
Friedrich, 1960, § 334.
45
Pour la place du verbe dans les relatives grecques, voir Giseke, 1846.
46
F o u r q u e t , 1938, 28; p o u r la place du verbe en germanique, voir n o t a m m e n t Beha-
ghel, 1929; Delbrück, 1911; Hirt, 1929, 258-63.
47
Lewis-Pedersen, 1961, § 433; Thurneysen, 1946, § 513; bibliographie sur la loi de Ber
gin chez Dressler, 1969, 19 n o t e 109.
48
J'ai laissé de côté le problème du verbe médian, qui reste à étudier.
ÉNONCÉ INDO-EUROPÉEN 29
BIBLIOGRAPHIE
Jespersen, Otto. 1917. Negation in English and other Languages. Det Kö-
nigl. Danske Videnskabernes Selbskab. Hist.-Fil. Medd. I. 5.
Josephson, Folke. 1972. The Function of the Sentence Particles in Old and
Middle Hittite. Uppsala: Acta Universitatis Uppsaliensis.
Jucquois, Guy. 1970. Les postpositions du hittite et l'accent des préverbes
en indo-européen. Muséon 83.533-540.
Klein, Jared S. 1985. The Origin and Syntax of the Rigvedic Construction
yá (...)ká/í/ú... ca. Festgabe für Karl Hoffmann = M.S.S. 44.105-121.
Krause, Wolfgang & Thomas, Werner. 1960. Tocharisches Elementarbuch,
I, Grammatik. Heidelberg: C. Winter.
Kuryłowicz, Jerzy 1958. L'Accentuation des langues indo-européennes.
Wrociaw-Kraków: Wydawnictwo polskiej Akademii Nauk.
. 1964. The Inflectional Categories of Indo-European. Heidelberg: C.
Winter.
Laroche, Emmanuel. 1957-58. Comparaison du louvite et du lycien. B.S.L.
53.159-97.
. 1959. Dictionnaire de la langue louvite. Paris: A. Maisonneuve.
Lee, Dionys J.N. 1966. The hittite Particle -kan and others. .O.34. 1-23.
Lehmann, Winfred P. 1974. Proto-Indo-European Syntax. Austin & Lon
don: University of Texas Press.
. 1980. The Reconstruction of non Simple Sentences in P.I.E. in: P.
Ramat, ed.
Lewis, Henry & Pedersen, Holger. 1961. A Concise Comparative Celtic
Grammar. Göttingen: Vandenhoeck & Ruprecht.
Macdonell, Arthur Anthony. 1955. (4ème réimpr.). A Vedic Grammar for
Students. Oxford University Press.
. 1956. (3éme impr.). A Vedic Reader for Students, ibid.
Meillet, Antoine. 1928. Sur le caractère de la phrase négative en vieux sla
ve. R.E.S1. 8. 171-177.
—--. 1933. Sur l'accent védique. B.S.L. 34. 122-26.
. 1965. Le Slave Commun. 2ème éd. revue et augmentée avec le con
cours d'A. Vaillant, nouveau tirage. Paris: Champion.
Miller, D.G. 1975. LE. VSO, SOV, SVO or all the three? Lingua 37. 31-52.
Minard, Armand. 1936. La subordination dans la prose védique, Paris.
. 1936, Trois Enigmes sur les Cent Chemins. Paris: De Boccard.
ENONCE INDO-EUROPEEN 33
Mogk, E. 1894. Die Inversion von Subjekt und Prädikat in den Nordischen
Sprachen. I.F. 4.388-95.
Mossé, Fernand. 1956. Manuel de la langue gotique 2 . Aubier-Paris: Mon
taigne.
Panhuis, Dirk. 1984. Is latin an SOV Language? A Diachronic Perspective.
I.F. 89.140-159.
Perpillou, Jean-Louis. 1972. La signification du verbe eúkhomai dans l'é
popée, in: Mélanges de linguistique et de philologie grecques offerts à
P. Chantraine, Paris: Klincksieck. 167-182.
Ramat, Paolo (& al.) edd. 1980. Linguistic Reconstruction and I.E. Syn
tax. Amsterdam: J. Benjamins, B.V.
Renou, Louis. 1952. Grammaire de la langue védique. Lyon-Paris: IAC.
Ries, John. 1880. Die Stellung von Subjekt und Prädikatsverbum im He-
liand. Strasbourg.
Schwentner, Ernst. 1923. Bibliographie zur idg. Wortstellung. W.u.S.
8.179-83.
-----. 1936. W.u.S. 9.195-96 (avec l'aide de F. Maurer).
—--. 1938. W.u.S., N.F. 1.160 sq.
- — . 1952. Z.V.S. 70.122-24.
____. 1967. Z.V.S. 81.159-60.
Seiler, Hansjakob. 1952. Negation, den Begriff des Prädikats betonend. Stu
dia Linguistica. Revue de Linguistique generale et comparée, 6.79-91
Strunk, Klaus. 1977. Heterogene Entsprechungen zwischen idg. Sprachen.
Z.V.S. 91.11-36.
Sturtevant, Edgar H. 1935. A Hittite Chrestomathy. Philadelphia: Lingui
stic Society of America, Pennsylvania.
Thurneysen, Rudolf. 1946. (réimpr. 1966). A Grammar of Old Irish. Revi
sed and enlarged edition, translated from the German by D.A. Binchy
& Osborn Bergin. Dublin: The Institute for Advanced Studies.
Vaillant, André. 1977. Grammaire comparée des langues slaves, t.V. La Syn
taxe. Paris: Klincksieck.
Vendryes, Joseph. 1938. Traité d'accentuation grecque, nouveau tirage. Pa
ris: Klincksieck.
Wackernagel, Jacob. 1877. Der griechische Verbalaccent. K.Z. 23.457-70
= Kleine Schriften. 1058-71. Göttingen: Vandenhoeck& Ruprecht. 1969.
34 FRANÇOISE BADER
—--. 1892. Über ein Gesetz der idg. Worstellung. I.F. 1.333-436 = Kl. Schr.
1-104.
. 1893. Beiträge zur Lehre vom griech. Akzent. Programm zur Rekto
ratsfeier der Univ. Basel. 3-38 = Kl. Schr. 1072-1107.
—--. 1909. Akzentstudien I. G.N. 50-63 = Kl. Schr. 1108-1121.
—--. 1914 a. Akzentstudien IL G.N. 20-51 = KL Schr. 1122-1153.
. 1914 b. Akzentstudien III. Zum homerischen Akzent. G.N. 97-130 =
Kl. Schr. 1154-1187.
. 1925. J.P. Postgate, A Short Guide to the Accentuation of the An
cient Greek. I.F. 43 Anzeiger, 48-59 = Kl. Schr. 1188-1198.
. 1942. Indogermanisches - que als alte nebensatzeinleitende Konjunk
tion. K.Z. 67.1-5 = Kl. Schr. 257-261.
Watkins, Calvert. 1963. Preliminaries to a Historical and Comparative Ana
lysis of the Syntax of Old Irish Verb. Celtica 6.1-49.
. 1964. Preliminaries to the Reconstruction of I.E. Sentence Structure.
PICL 9.1035-1042.
. 1976. Towards PIE Syntax: Problems and Pseudo-Problems, Papers
from the Parasession on Diachronie Syntax. B. Steever & al. edd. Chi
cago Linguistic Society. 305-326.
Zimmer Stefan. 1976. Die Satzstellung des finiten Verbs im Tocharischen.
Paris-La Haye: Mouton.
BADI
IL S'EN VA OU LE FRANÇAIS, ET POURQUOI?*
JOËLLE BAILARD
University of California Los Angeles
There have been a number of studies during the last ten years suggesting
that French is undergoing a typological shift. Harris (1976a, 1978), for in
stance, argued that French is moving toward verb-first sentential word-order,
VSO or VOS, more probably VSO. In Bailard (1982), I suggested that the
basic sentence word order which is competing with the canonical SVO order
is not VSO but rather VOS. In both Harris' analysis and mine, though, it
was assumed that the emerging word order, just like the one it shows signs
of replacing, is a syntactically determined order, conform, that is, to the
word order typology described by Greenberg (1963). As a result in part of
the recent interest in discourse analysis, this prima facie hypothesis about
French basic word order has been brought into question. Lambrecht, in par
ticular, has suggested that there is in spoken French a «preferred clause pat
tern», to which we shall return, which is pragmatically, not syntactically
determined, and he has ventured that the next preferred word order in French
will also be a pragmatically determined one (Lambrecht, to appear). In that
view, then, French could be said to be more similar to Mandarin Chinese
than to English.
The purpose of this paper will be to discuss the relevance of proposi
tions with initial topic-phrases, as in (1) below, and of propositions with
final topics, as in (2), in relation to canonical sentences like (3) for the
characterization of the basic word order in contemporary French:
* This paper is for Harold Cohn. I wish to thank the LSA and the NSF for the travel
grant which made my participation in the ICHL-7 Conference possible. I also wish to thank
Sandra Thompson and George Bedell for their comments. Space limitations only prevented
me from dealing more adequately with some of these. A separate paper will be necessary to
do them justice.
36 JOËLLE BAILARD
hypothesis, one must be cautious then not to relate any claim about word
order change to his description of universals, which, while proposed for syn
chrony, is also assumed, sometimes implicitly, often explicitly, to have predic
tive value for diachrony as well, since, like any syntactic theory, it constitutes
in effect a prediction as to what may constitute a possible human language
and hence a possible linguistic change.
One prerequisite, as long as we retain Greenberg's notion of basicness,
before we try, to establish what specific typological shift, if any, is taking
place in French is to ascertain that the syntactic data selected may all be
taken to constitute equally valid evidence of the same phenomenon. It is
for this reason that a sentence like (4), while perfectly grammatical and as
frequent, if not more, as its counterpart in (3) with full NP's, will be ex
cluded from our primary data here. Similarly, sentences such as the ones
in (5)-(6), though very frequent also, will not be considered in this paper1 :
(4) Elles ne vont pas durer longtemps.
«They are not going to last long.»
(5) IIj Ij'a compris lui i .
«(As for him), he understood it.»
(6) Lui i , ili lj'a compris.
«HE / (As for) him, he understood it.»
Given only what we know at the moment, sentences like (5)-(6) cannot
be evaluated as evidence of syntactic change, because they contain pronouns,
and the accepted standard of comparison is a sentence where all the
arguments are nouns. These sentences can no more be used in comparison
to (l)-(3) than could a sentence containing another extraneous semantic
variable such as interrogation, as in (7):
(7) Combien de temps vont-ellesi durer ces illusionsi? (Butor)
How long are they going to last, these illusions?
How long are these illusions going to last?
'
It is taken for granted that questioning is a causal factor in word order
variation in many languages and, in the absence of evidence to the contrary,
considering the complications which fall out from the distribution of pro
nouns for the synchronic description of any state of the language (cf.
1
The previous examples, like much of the contemporary French data discussed here,
are from Larsson (1979), whose sources are primarily 20th century novelists and magazines.
38 JOËLLE BAILARD
2
These remarks should not be taken as a suggestion that the study of sentences with
pronouns is not necessary for an accurate picture of syntactic change. To the contrary. One
should keep in mind though, that by including them in the data consistently, we might have,
among other considerations, if their past and present frequency is similar, to reanalyze Classical
and 19th century French as SOV, in which case any attempt at analysis of current word order
change based on a previous SVO basic order would be pointless.
IL S'EN VA OÙ LE FRANÇAIS 39
not only at the beginning of a proposition, as in (16), but also to the right
of COMP, as in (17):
(17) ta ku de [neixie shu, mei yiben wo dou kan bu xia-que le].
he cry COMP those book every one I all read not down ASP
«He cried so much that, those books, I couldn't continue to read any
of them.»
An accurate syntactic and semantic description of topicalization in
French will have to take into account the fact that, as in Mandarin Chinese,
embedded topics do occur, as in (18):
(18) Mariéi dit que Jean;, il; ne 1¡' écoute jamais.
«Marie says that Jean, he never listens to her.»
To return to our main concern, it is of interest that, while Li and Thomp
son's objective is the same as Greenberg's in the sense that they are con
cerned with characterizing the unmarked sentence type in the language, they,
unlike Lambrecht (1984), do not extend the characterization they propose
for Mandarin Chinese to French. They suggest that French is different from
Chinese and, like English, a language with syntactic word order (Li and
Thompson, 1981: 19).
Yet, discourse-based studies such as the ones by Barnes (1983, 1984),
Larsson (1979), Harris (1985), suggest that, at the same time that we ought
to be able to give of contemporary French a description which distinguishes
it from Chinese, we also must try to make explicit how contemporary French
differs from a «pure» SVO type like English, or from what is taken to have
been the previous, equally pure, status of French, based, unfortunately on
ly on (selected) written records.
One of the several facts which we need to take into account for this
is that, in French propositions, there is often one constituent before the
sentence or after it, and in some cases both at once. Larsson (1979) gives
sentences (19)-(20) below as examples of multiple right-dislocations, and (21)-
(22) as examples involving both left-and right-dislocation silmultaneously:
(19) Ili les; sent passer ses dîners; Philippe i . (Rochefort)
«Philippe (he) gets to notice (them) his dinners.»
(20) Ellei leur; cache tout la viei aux hommes;. (Céline)
«Life (it) hides everything from (them) men.»
(21) Son enfant i , elle; l i 'adore, cette mère;. (Régula)
«Her child, this mother (she) adores him.»
IL S'EN VA OÙ LE FRANÇAIS 41
(22) Les miliciensi, ilsi leuri faisaient des agaceries, les Russes j .
«The militiamen, the Russians (they) were teasing them.»
With respect to (20), Larsson notes that the relative ordering of the last
two nominals may be determined in part by rhythm.
I have used the term «topic» so far indiscriminately to refer to some
proposition-final and to some proposition-initial constituents. But, as was
shown by Larsson, the two types of dislocation, when used to topicalize,
do not have exactly the same meaning. Some of the pragmatic differences
between left-side topics and right-side topics can be seen from her examples
in (23) through (26) below:
(23) C'est vrai qu'il y a des endroits aux Etats-Unis où les Noirs ont pas
le droit d'entrer? - Et les Indiens i , m'dame, pourquoi yi sont dans
des sortes de camps? (Delanoe)
«It's true that there are places in the United States where Blacks do
not have the right to go in? - And the Indians, Ma'am, why are they
in some kinds of camps?»
(24) *... - Et m'dame, pourquoi yi sont dans des sortes de camps, les In-
diensi?
(25) Ce briquet, justement, j'ai allumé tes cigarettes avec quand tu es venue
chez moi pour la première fois. (Beauvoir)
«This_lighter, precisely, I lit your cigarettes with (it) when you came
to my place for the first time.»
(26) *(Je lui ai montré un briquet en argent). J'ai souvent allumé ses cigaret
tes avec (çai) ce briquet i .
«(I showed her a silver lighter). I often lit her cigarettes with (it), this
lighter.»
Whether the nominal constituents are to the right or to the left of the
sentence, and despite their interpretive differences, these phrases all may
be said to be topics in the sense that they are «what the sentence is about»,
and the theme/topic distinction discussed by Dik (1978,1980) is not obviously
involved here.
With respect to right-side topics, there is no data at present on whether
Mandarin Chinese makes significant use of the construction, or whether what
follows a sentence in that language should best be analyzed as afterthought.
For French, though, as Larsson mentions, afterthought is not an appropriate
label for those right-side constituents, since they are frequently indispen
sable to the desired interpretation of the sentence. This is true, for instance,
42 JOËLLE BAILARD
3
Duranti and Ochs (1979) suggest a link between left-dislocation and turn-taking in
Italian. Such a discourse feature might also be relevant to the distribution of dislocation in
French.
4
For discussions of the grammaticalization of afterthought instead of topic as a
mechanism of word order change, see e.g. Hyman (1975) and Givón (1979).
IL S'EN VA OÙ LE FRANÇAIS 43
(28) Tous ceux qu'ya dans le quartier, c'est moi qui leur a donné des bouts.
(François, 1974).
«All the ones that there are in the neighborhood, it is I who gave them
cuttings.»
Compare (27)-(28) with the following:
(29) Ya quelque chose que je ne comprends pas là-dedans.
«There is something that I don't understand in there.»
(30) Tous les géraniums mauves qu(i)ya dans le quartier, c'est des boutures
que je leur ai données.
«All the mauve geraniums that there are in the neighborhood, it's cut
tings that I gave them.»
Sentences (27) through (30) are all example of constructions of what
Lambrecht gives as two of the important representations of his preferred
clause pattern, and which he suggests are the basis for the typological
classification of spoken French. Note that, while in (27)-(28) the relative
pronoun qui can only be interpreted as the subject of the embedded verb,
in the next two sentences, the relative pronoun form que shows that the
antecedent must be interpreted as the object of the lower verb. Assuming
here, only for the sake of the argument, that for a language to have as its
unmarked sentence-structure the topic-comment articulation (or, for that
matter the comment-topic one) should ever be considered a reason to posit
that, in that language, grammatical functions would have no reality, or at
best play a trivial role, it is not clear how such sentences might be construed
as evidence that French is moving from a syntactically determined word order
to a pragmatically determined one. In fact, whatever one's view of the
relevance of grammatical relations for French at all, it is not clear what,
if anything, might be concluded from the frequent occurrence of such
sentences with respect to word order typology, contrary to Lambrecht's sug
gestion, except that grammatical relations are necessary constructs, as has
been assumed in traditional analyses of word order change (cf e.g. Ven-
nemann 1975; Li and Thompson 1975).
For French, it would appear, study of the topic-initial and of the topic-
final constructions, as well as of the presentational and identificational con
structions, only seems to provide additional evidence, if need be, that gram
matical functions are necessary constructs for an exhaustive description of
such various unrelated structures as are found in contemporary French, just
as they have been for canonical French.
44 JOËLLE BAILARD
embedded topics. Lack of space prevents us from attempting this here but,
in order to give an explicit description of the form of French topicalized
sentences, we shall have at some point to determinate the hierarchical rela
tion between the Topic node(s) and COMP and also between the Topic and
S' and S" 5 .
With respect to interpretation, also, one question which still needs a
definitive answer is whether the ordering of topical constituents among
themselves must be taken for French as an indication of some distinction
in meaning, of scope or otherwise. We saw earlier that French right-side
topics are not exactly equivalent to their left-side counterpart, but with respect
to left-side topics alone, it does not appear warranted to assume cross-
linguistically that the leftmost topic is necessarily of a different kind, for
instance, generic, or has different scope automatically from its immediate
neighbor.
Thus, for Japanese, Kuroda (1965) suggested that there is no difference
in meaning between sentences like (33) and (34):
(33) John -wa nicyoo -ni -wa New York Times -o yomu
-TOP Sunday-on-TOP -ACC read
«As for John, on Sunday (he) reads the New York Times.»
(34) Nicyoo -ni -wa John-wa New York Times-o yomu.
Sunday -on-TOP -TOP -ACC read
For French, similarly, Larsson found no obvious hierarchical distinc
tion necessarily between left-side topics, and noted that it is not clear that
there is any difference in meaning between, for instance, sentences (11) and
(35):
(35) La liberté sexuelle et tout ça;, mes parents i , ilsi enj ont horreur.
«Sexual freedom and all that, my parents, they can't stand it.»
Whether the two positions convey a different meaning could be determined
only by comparing the freedom of occurrence of such sentences in context.
A distinct and more fundamental issue yet for us is whether such topical
constituents should be considered constituents within the sentence, and more
directly to the point here, as was suggested by the reference to predicative
5
On this topic, compare, for instance, Chomsky (1977, 1981), Banfield (1982) and
Reinhart (1983).
46 JOËLLE BAILARD
6
Harris (1984b), on the issue of statistics and linguistic change, mentions the observa
tion by Smith (1981) that typology «provides data which needs explanations... It provides no
explanation itself», related to the view of Lass (1980) that the distribution noted in the form
of a statistical law is the «primary explanandum», having «no explanatory force per se». Here,
on the other hand, we are assuming that, while it is, of course, true that statistics need to be
explained, typology is itself, as argued in Harris (1982, 1984a-b) one of the causal factors in
change, because of the tendency to analyze data based on analogy.
7
Among the notable exceptions to this prescriptive approach are Bauche (1929) and Frei
(1929).
48 JOËLLE BAILARD
final noun as subject itself could be due to a variety of factors, which joint
ly had the effect of making the structure of the French sentence opaque.
One of these factors was phonetic erosion. The verbal inflectional system
has become reduced, from a six-way opposition in Old French, to at best
a three-way opposition in the first (and most productive) conjugation. This
has been a result in part of the tendency in non-careful French not to pro
nounce the liaison consonant. Also, as noted by Harris (1978), the indefinite
pronoun on, which triggers 3sg agreement, is in the process of replacing
lpl nous, as in on aime ça (nous) «we like that», so that, in some dialects
at least, the conjugation suffixes for the first conjugation are in effect reduced
to two, zero in the non-2pl and /-e/ in the 2pl. As a consequence, as observed
by Sauvageot (1962), the subject clitics have now taken over the function
of verb agreement markers.
In such circumstances, for the rightmost constituent in the proposition
to become reinterpreted as the subject or as the object is therapeutic, since
it has the effect of making it possible again, as it was in canonical French,
to use linear order to mark grammatical functions, reducing syntactic opacity
within the sentence. This reanalysis of the syntactic data is consistent also
with the existence in V-initial propositions of two intonational contours,
one of which is associated with the right-topicalized sentence and the other,
the «integrated contour», with a simple predication. That French is an
oxytonic language itself may have been a contributing factor in the predilec
tion for right-dislocation as a topicalizing device earlier.
This earlier topicalizing function of right-dislocation would explain in
its turn why the spreading sentence structure was first noted with definite
NPs, since topics are by nature assumed to have an agreed-upon referent.
In Occitan, apparently, right-dislocation is still limited to definite NPs, sug
gesting that perhaps the same reanalysis which was first proposed by Harris
for French has not yet taken place in that language. In French, by contrast,
the scope of the V-initial strategy has been broadened to include also
sentences with a non-definite subject, by making use of ce/ça as the agree
ment marker, as in:
(44) Ça ne me déplairait pas une Porsche.
«A Porsche (that) would not make me unhappy.»
(45) (a) Ça fait toujours plaisir des fleurs.
(b) *Ça fait toujours des fleurs plaisir.
«Flowers are always pleasing.»
IL S'EN VA OÙ LE FRANÇAIS 51
8
For an analysis of the shift of ce from demonstrative to article, see Harris (1980).
52 JOËLLE BAILARD
as opposed to, say, socio-historical, then, the same cognitive reasons behind
the general preference cross-linguistically for the subject before object order
might in French now work likewise against the generalization of VOS, again
possibly in favor of the universally less marked VSO.
In conclusion, there seems to be little doubt that French is a mixed-
type language at the moment, with two predominant orders competing in
basic sentences, SVO and VOS. While the interaction of pragmatics and
syntax in typological change is not a new discovery, a wealth of new material
has recently been brought to our attention, and the main task confronting
us now will be to try to integrate the results of this research in functional
syntax into the overall study of linguistic change, much as was done earlier
in relation to the effects of sound change on typological change.
REFERENCES
BAILARD
ATTEMPTING THE RECONSTRUCTION OF NEGATION PATTERNS
IN PIE.
GIULIANO BERNINI
Istituto Universitario di Bergamo
1
The first essay on Indo-European negatives is Fowler 1896. For detailed treatment of
negative patterns in the proto-language see Brugmann-Delbriick 1897:519-540 of vol. 4,
Brugmann 1916:974-979, Hirt 1937: 69-80, and for the history of negative particles in the an
cient languages, mostly Greek and Latin, cp. Wackernagel 1928:248-313. For the three
reconstructed particles and their correspondences see Pokorny 1959: 756 f., 703, 757 f. respec
tively.
58 GIULIANO BERNINI
2
This seems to be true for verbal human languages, notwithstanding the possibility of
expressing negation by means of prosodic modifications in the form of change of distinctive
tones in some Niger-Congo languages as Ga and Mano (Dahl 1979: 82 with further references),
since this device must be rather regarded as morphological. Among human sign languages,
at least American Sign Language allows, beside lexicalization of various negative meanings
such as «not», «not yet» etc., the simple expression of sentence negation through a so-called
non-manual sign, actually a particular facial expression, accompanying the normal signing of
positive sentences (Liddell 1980:39). This device corresponds, mutatis mutandis, to a particular
intonational contour for negative sentences without overt negative morpheme in verbal languages.
3
In this respect the most important investigation on the typology of sentence negation
was carried out by Dahl (1979). For a preliminary overview of cross-linguistic data see Payne
1985 and for a functional approach Givón 1984:321-351. For ongoing investigation of negative
indefinites by the team of Professor Ramat in Pavia see Ramat-Bernini-Molinelli 1986.
PIE. NEGATION PATTERNS 59
4
. Streitberg 1910:217, who speaks of position before the verbal predicate without
further remarks. Behaghel (1924:68 f.) differentiates between verbal negation on the one hand
and quantitative negation of the verb and of non-verbal constituents on the other; in this se
cond case, ni must be posited only before the constituent to be negated if this precedes the
verb. (e.g. ni waihtai botida (Mk 5,26) «it profited him nothing»). This is clearly a special
case, pointing to an incipient process of negative incorporation, since it occurs mainly with
the two indefinite nouns manna «(a) man» and waiht(s) «(a) thing», and does not challenge
the main pattern of pre-verbal position of negation. For the early situation in Germanic see
Hopper 1975:39f., Ramat 1981:132f.
60 GIULIANO BERNINI
Ch. SI. ašte ne bi otu boga bylǔ sĭ, ne moglǔ bi «If he was not from god,
he could not (do it)», also ne prěžde poučati sę «not training oneself» with
ne before the verbal predicate, cp. Vaillant 1977:193); Albanian (cp.Skënder
nuk do ta turpëronte Brahimin «Skënder would not bring dishonor to him,
Brahim», see Lambertz 1959:835); Tocharian, where *mē substituted *ne
(cp. A. natsey-ac mä kumnäs «he won't fall into distress», puk wramäm
säpn-ä nåpak ā mäskanträ «everything is not as in the dreams», Krause
1955:38); Iranian with the only exception of Avestan (cp. Mn. Pers. soma
šâm na-xordîd «you did not eat supper» with strong stressed na (Mace
1971:50), Tadzhik ki on parda -buda «that it was not a curtain» (Comrie
1981:191 f.), O.Pers. yapā Gaumāta haya magus viþam tayām amāham naiy
parābara «for Gaumāta the Wizard did not dethrone our royal descent»
(DB I, 70-71 cp. Brandenstein-Mayrhofer 1964:85); some Indic languages,
notably Hindi (cp. uskā na jānā acchā hε «it is good for him not to go»,
liter. «to him not go good is», vo nani jāegā «he won't go», Bhatia 1978:16)
and Punjabi (cp. o kār naī giā «he did not go home», Bhatia 1978:16), but
also partly Nepali (cp. ma naja «should I not go» vs. ma khan-ñu «I don't
eat», Bhatia 1978:17); Hittite among the Anatolian subgroup (cp. šer-a-ššan
GADan pešiemi šu-uš LÚaš natta aušzi «I throw a cloth over it and no one
will see them», liter. «... and man not see» with natta «not», reported in
Lehmann 1974:54; see Friedrich 1960:147).
As one can easily see, this bulk comprises languages with clear word
order patterns (VSO as in Celtic; SVO as in modern Romance, Slavonic,
Albanian and partly Baltic; SOV both weak as in the Iranian group or more
rigid as in Indic), although languages with conflicting word order types as
Tocharian (classified as SVO with overlay of SOV features by Friedrich,
1975:9) are present too.
To this bulk we can also add the languages that do not continue the
Indo-European particle *ne as the one listed above do, but which behave
in the same way, i.e. Armenian, both Modern with SOV order (cp. vor gluxə
č-i uzum gərki vrajie bardzraenel «that she did not want to raise her head
from the book», liter. «that head-her not-was wanting book from-on to
raise» with či = (vo)č «not» + i «was», Comrie 1981:192 f.) and Old with
SVO order (cp. zi oc' goyr noc'a tełí yijavanin «for there was no place for
5
nuk( *ne-kwod with owing to conflation with Lat. numquam «never» (Hamp
1984:178).
PIE. NEGATION PATTERNS 61
them in the hostel», Schmitt 1981:176); SVO Greek, both Modern (.
«Ulysses does not eat fish»6) and Ancient (cp.
«he did not take any account (of it)» (Aesch, Prom. 232));
partly Albanian with SVO order (cp. s'kam pa kurr kësi nieri «I never saw
such a man», liter. «not I-have seen ever...», Lambertz 1959:157)7.
Among the languages not mentioned till now and which deserve par
ticular comments are Latin (e.g. si adgnatus nec escit, gentiles familiam
habento «if there is no close relative, distant ones shall take over the house»
(XII Tables), nee mi aurum posco nee mi pretium dederitis nec... «I don't
ask for gold, do not pay me any price, not...» (Annales, 186), non ego te
flocci facio «I don't care a bean for you» (Plaut., Curc. 713), Homeric Greek
(e.g. «for lowly men surely wouldn't beget
such (sons)» (Od. 4,64), :<but
your heart doesn't notice, Zeus!» (Od. 1,59) both reported in Lehmann
1974:126, 55 respectively; «Thetis did not forget her
tasks» (II. 1,495) ), Avestan (e.g.
«the karapans do not want to submit to orders and laws» (Gath. 51,14),
«I was not able to get hold of x v .» (Yt. 19,57)
both reported in Gonda 1951:69), Sanskrit (e.g.
táskarah «they perish not; no thief shall (can) harm them» (RV vi 28 3),
«no god,
no mortal could surpass what the mighty bull has done» (RV viii 96 2), tarn
«If you cannot explain this
(riddle) to me, your head will burst» (SB); all reported in Macdonell
1971:354, 362, 362 respectively).
Free ordering of negative particles is claimed for the four languages,
with two positions being actually and equally preferred: sentence initial and
before the verbal group (see Brugmann-Delbrück 1897:521 of vol. 4,
Brugmann 1916: 976 for all the languages at issue; Wackernagel 1928:259
ff. for Latin and Greek in particular; Bartholomae 1904:1071 for Avestan;
6
For cp. Schwyzer-Debrunner 1966:593,597.
7
Alb. s should be better regarded as another instance of reshaped IE *ne according to
the very interesting proposal of Hamp (1984: 179-181) who reconstructs an original emphatic
form *ne-kwid (parallel to *ne-kwod > nuk(ë)) which lost the negative element owing to the
morphonemic reassignments brought about by the intertwining of different cliticization
possibilities of *kwid and old optional contextual forms of topicalised personal pronouns with
suffixed -në as u(në) «I», ti(në) «thou» etc.
62 GIULIANO BERNINI
negative facts and sentence initial position is reserved for strong «affective»
reactions to previous statements. In this case ná means rather «it is not so
that...» (Gonda 1951:51-52).
Stress may also have a role in this regard, with proclisis correlating with
position immediately before the verb (cp. e.g. Skt. ná, Lat. nequeo vs. Pali
n'atthi «is not» (Gonda 1951:55), Goth. nist, MHG ih enmac with e = [ə]).
Typological considerations help in this case limiting the range of
hypothetical reconstruction and making at least plausible the further search
for single language (or language group) developments. Evidence from in
side as well as outside the Indo-European family makes unplausible the
equating of Vedic negative structures with Proto-Indo-European ones, but
this same evidence helps us also in the evaluation of tentative reconstruc
tions of Proto-Indo-European word order patterns, particularly of the claim
ed consistent SOV order with the almost automatic assumption of a sentence
final negative qualifier (notably Lehmann 1974).
In Dahl's sample of almost 250 languages (cp. his statistical tables, 1979:
103-104), verb final languages appear to share almost equally preverbal and
postverbal negative constructions with a slight preference for these latter
ones, found in 51 over 91 languages (56%) (cp. also Payne 1985: 225 for
similar statements, although not substantiated by explicit statistical coun
tings), but in the group of verb final languages with syntactic constructions,
only 2 over 23 (i.e. 8,69%) put the particle after the verb, while in the group
of verb final languages with morphological constructions postverbal negative
auxiliaries or negative suffixes are the great majority, ranging about 69%
the former (11 languages over 16) and 72% the latter (36 languages over
50). Furthermore, assuming a postverbal negative qualifier in a SOV Proto-
Indo-European involves reconstructing the ways of and the reasons for the
shift in position and in category of the old negative morpheme or the genesis
of a new negative morpheme, accounting for the preverbal particle(s) of the
historical languages.
4. We can now leave aside the case of new negative particles originating
from emphatic adverbial expressions deriving mostly from neuter Accusatives
with negative incorporation, as shown chiefly by Lat. nōn (if derivation from
*ne-oinom is correct, see Leumann-Hofmann-Szantyr 1977:452 and Wacker
nagel 1928:253), and also by Av. naēčīm, younger Av. naēčis (Bartholomae
1904:1034), Ved. nákim (and mákim), both with only two occurrences (Mac-
donell 1971:236,240) together with makis, all composed of th.e indefinite
64 GIULIANO BERNINI
pronominal stem *kwi-. This case, found also in languages which do not
continue IE *ne, cp. Anc. Gk. Mn. Gk. , is of minor typological
and diachronical significance since it is a simple matter of lexical substitu
tion which does not affect the position of the negative particles involved.
Of major significance are the factors enabling the genesis of new
negative morphemes (or words) or their shift around the verb which can
be sketched by considering some of the languages not yet subject to scrutiny
as regards positioning of negative particles and which show exceptional
behaviors running counter the overall trend to preverbal constructions.
Among these are all modern Germanic languages, which posit the negative
particle either after the finite element of the verbal group as SVO English
or Nordic languages in main TVX sentences (e.g. Norw. jeg har ikke sett
ham «I have not seen him» vs. Det er folk som ikke vil arbeide «There are
people who do not want to work», Sommerfelt-Marm 1981:53), or accor
ding to the distribution of topic and focus as German and Dutch (with its
offspring Afrikaans), which have TVX order in main sentences and SOV
in subordinate ones (e.g. Germ. Ich sehe das Haus (TOP) nicht (FOC) «I
do not see the house»). This type of negative constructions notoriously is
the result of discontinuous constructions with two particles, one before the
verb being lost, and a second one after the verb being originally an emphatic
expression, as one can find in ancient Germanic languages (e.g. OHG ni
zawêta imo es niawiht «he did not succeed» (Otfr. II, 5.12), see also Behaghel
1924:71-76, Ramat 1981:132 ff.) and today in Fr. ne V pas (e.g. il ne le dira
pas «he won't tell it») and in Welsh ni(d) V ddim (cp. nid wiLf i ddim yn
hoffi coffi «I do not like coffe», Bowen-Rhys Jones 1960:23) or Breton ne
V ket (. ne lavaro ket kement-se «he won't tell it», Pedersen 1913:213).
In colloquial usage both French (see Ashby 1981) and Welsh (see Bowen-
Rhys Jones 1960:23) often drop the first particle.
This development typically creates new negative words, either with or
without incorporation of negation, starting from words with originally
positive meanings (e.g. Germ. nicht < ni eo wiht, liter. «not ever (a) thing»,
Nord. ekke < *eint-gi(n) «one» with generalizing suffix, cp. Ramat 1981:112;
Fr. pas < lat. passum (Accusative) «(foot) step» as is well known; Welsh
dim, base of mutated form ddim, corresponding to Ir. dim glossed «quip-
piam, quicquam», cp. Lewis-Pedersen 1961 :§385). This evolution, which
incidentally goes along with the rise of new sets of non-referential pronouns
both with or without negative incorporation (e.g. Eng. no(ne), liter. «not
PIE. NEGATION PATTERNS 65
one», Fr. personne «nobody» < Lat. persona(m) «person»), is the result of
the non-referentiality acquired by indefinite noun phrases with focus func
tion in the scope of negation. Pragmatic factors are therefore at work in this
case, where the resulting postverbal construction becomes a kind of Adver
bial Operator in the sense of Bartsch-Vennemann (1972:133) harmonic with
the VO structure of the language and thus looses some of the markedness
character one would attribute to it on the sole basis of statistical regularities.
The fully character of negative morpheme acquired by these new par
ticles is proved by their usage as constituent negation, preposed to the word
or phrase they are intended to negate, as illustrated by known Germanic
and French examples and by Welsh dim/nid Blodwen ddaeth (on Gareth)
«it wasn't Blodwen who came (but Gareth)» (reported in Payne 1985:232).
Leaving aside the question of possible areal diffusion of such construc
tions, relevant only for Breton and for some Romance dialects on the
linguistic boundary to German (see Molinelli-Bernini-Ramat, forthcoming),
but otherwise chronologically and sociolinguistically unplausible, the dif
ferent diffusion of this construction shows the different effect of pragmatic
factors on different word orders, whereby strongly TVX languages as Ger
man show the most aberrant construction with negation on focus, while among
SVO languages only those which have known TVX order show fully gram-
maticalized discontinuous or postverbal constructions (notably French), the
other ones allowing them only as highly marked choices (notably Italian, cp.
mica). In VSO languages, the preverbal position does not appear to be serious
ly challenged by the pragmatic factors of topic-focus ordering, permitting only
the rise of weak discontinuous/postverbal constructions as in Welsh.
5. A case for the shift of the negative particle around the verb is
represented by some modern Indic languages, which compose the second
group of Indo-European languages with departure from the statistically
regular immediate preverbal position of our negative particle. Indic languages
have acquired almost all of the characteristics of rigid SOV languages ow
ing to their being since long time in contact with Dravidian languages. Among
the major Indic languages, Hindi and Punjabi maintain preverbal negation
(as examples reported above on p. 60 show) but Marathi and Sinhala,
among others, have postverbal constructions (e.g. Mar. gharāt konī āla nahī
«no one came home», liter. «home someone come NEG: was», Bhatia 1978:
17; Sinh. miniha gamətə yanne nææ «the man does not go to the village»,
liter. «man to-village go not», De Abrew 1981:12). Nepali (see above p. 60)
66 GIULIANO BERNINI
and Kashmiri (e.g. agar su vanihē ni me, bi mā vanihə: tas «if he had not
told me, I would not have told him», liter. «if he would-have-said not me,
I not would-have-said him» with partial overlapping of IE *ne, *mē, Bhatia
1978:26) come in between, showing the former postverbal negation in
declarative sentences and preverbal negation in conditional and imperative
ones, the latter just the reversed situation. In this case, the shift in the posi
tion of the inherited negative particle seems to be brought about by the overall
trend towards SOV due to contact, without appealing direct Dravidian in
fluence, since one should check the actual possibilities of cross-linguistic iden
tifications between negative morphemes of Dravidian and Indo-European
languages of the area. Interplay of affixation of negative particles to aux
iliary forms and positioning of these in relation to the main verb are more
likely to be at work, with the result of regular V-Aux order of SOV patterns.
6. The interpretation of the Indo-European data presented above has
shown that: (a) new negative morphemes arise through grammaticalization
of stylistic variants of emphasis expressions under the felicitous conditions
of a combined effect of grammatical and pragmatical order, this being most
easy in VX languages; (b) shift of negative particles around the verb can
take place inside a more general trend towards consistency of word order,
with verb forms with attached negative particles being actually moved.
This background, enriched by evidence from outside Indo-European
which points to a strong reduction of main verb forms in negative sentences
when auxiliaries become particles (cp. Estonian loen «I read», loed «you read»
etc. vs. ei loe «I, you, etc. do not read» (liter. «NEG read»); Finnish en lue
«I do not read» (liter. «NEG: 1SG read»), et lue «you do not read» (liter.
«NEG: 2SG read»), ei lue «he/she does not read» (liter. «NEG:3SG read»)
etc. vs. luen «I read», luet «you read», lukee «he/she reads», Comrie
1981:133,132) provides a number of checkpoints for possible reconstructions
not only of Proto-Indo-European sentence negation (with likely exclusion of
V-Neg ordering), but also of Proto-Indo-European negative morphemes, either
debatable and debated as Lat. nōn whose derivation from *ne-oinom presents
difficulties of phonetic development and of ordering and for which origin
from a reinforced particle *nō-ne has been recently proposed in Dunkel
1982-83:194, or of yet unknown origin as Arm. oč', Gk. où or *ne itself.
Evident should also be the feed-back action on reconstruction of general
Proto-Indo-European word order patterns, which receive a constraint on
likelihood of verb position.
PIE. NEGATION PATTERNS 67
REFERENCES
0. Introduction
The term «narrative» Imperfect (henceforth, NIPF, where IPF stands
for «Imperfect») refers to a peculiar usage of this tense, which is believed
to have emerged in relatively recent times (French grammarians also make
use of the term «IPF pittoresque»). The specific characteristics of this usage
lie, as is well known, in its aspectual features: in contrast to the imperfec-
tive value of the IPF in its normal usages, the NIPF is used in contexts where
we should expect to find a perfective tense, such as the Simple Preterite ( =
«Passato Remoto») or the Compound Preterite (= «Passato Prossimo»).
The purpose of this paper is twofold. First, I shall try to characterize
the aspectual value of the NIPF, showing that there is in fact a gradual shift
from the truly imperfective meaning of the normal IPF to the truly perfec
tive meaning of the extreme cases of NIPF. Second, I shall attempt to
demonstrate that the NIPF, although classified by grammarians only in
relatively recent times, has a more distant origin, and is deeply rooted in
the subtle and rich aspectual system of old Italian.
Finally, I should like to observe that although my analysis is likely to
hold also for other Romance languages, the examples will be drawn only
from Italian: my conclusions will therefore be prudently limited to this
linguistic domain.
which I call «continuous» IPF: for the moment, however, I refer exclusive
ly to the two aforementioned types which are standard in any textbook):
(1) Mentre andava a scuola, si accorse all'improvviso di non aver studiato
la lezione.
(As he was going to school, he suddenly realized that he had not done
his homework).
(2) L'estate scorsa mi veniva a trovare spesso.
(Last summer he used to visit me quite often).
Minor details aside, the «progressive» meaning conveyed by the IPF in (1)
can be described as follows: (a) the given process (going to school) was tak
ing place at a certain point in time tf (standing for «focalization point»);
(b) most importantly, the continuation of the process after tf is absolutely
u n d e t e r m i n e d . In other words, by the mere uttering of the
IPF in (1), the speaker does not commit him/herself to the belief that the
process of going to school was continued or (alternatively) was interrupted
after tf; in fact, he/she treats as barely irrelevant any question concerning
the temporal boundaries of the action (and this is true even if the speaker
happens to know how the given state of affairs in fact developed). As for
the «habitual» meaning conveyed by the IPF in (2), suffice it to say that
this usage of the IPF is defined by its allusion to a numerically unspecified
iteration of the event (in our case, the process of visiting).
In both cases we may say that the element crucial to the understanding
of the imperfective aspect carried by both the «progressive» and the
«habitual» IPF, is the notion of i n d e t e r m i n a c y , although
this notion has to be interpreted in two quite different ways; i.e., «indeter
minate continuation» (or, equivalently, «indeterminate duration») with the
«progressive» aspect, and «numerical indeterminacy» with the «habitual»
aspect. As a matter of fact, whenever such a characterization is violated
in the given context, the sentence is either ungrammatical or else deviates
in aspectual interpretation from the original imperfective value. Thus
sentence (3), corresponding to (1), is ungrammatical inasmuch as the Sim
ple Preterite (andò) is incompatible, because of its perfective meaning, with
the «indeterminate continuation» of the event after tf sentence (4), on the
other hand, corresponding to (2), is felt as «deviant» by any native speaker
inasmuch as it contains a numerical determination:
(3) * Mentre andò a scuola, si accorse all'improvviso di non aver studiato
la lezione.
NARRATIVE IMPERFECT 73
(As he went to school, he suddenly realized that he had not done his
homework).
(4) (?) L'estate scorsa mi veniva a trovare tre volte.
(Last summer he would visit me three times).
As it happens, sentence (4) is, as we shall see in a moment, a possible
manifestation of the NIPF. Notice, in this connection, that numerical deter
mination is quite compatible with perfective tenses, as is shown by (5). This
proves, therefore, that numerical specifications are the crucial element that
runs against the salient aspectual value of the «habitual» IPF:
(5) L'estate scorsa mi venne a trovare tre volte.
(Last summer he visited me three times).
The contrast between the IPF and the Simple Preterite may be observ
ed in a number of other instances, all of which yield similar results on
analysis. Thus, in (6) the IPF (in its «progressive» meaning) is interpreted
as temporally overlapping with the event expressed by the Compound
Preterite, whereas the Simple Preterite is interpreted as posterior. This is
another consequence of the aspectual difference between the two given tenses:
(6) (a) Quando arrivai, Luca preparava la cena.
(When I came, Luca was preparing dinner).
(b) Quando arrivai, Luca preparò la cena.
(When I arrived, Luca prepared dinner).
In (7) on the other hand we observe a fact which is relevant to the distinc
tion between the «progressive» and «habitual» aspects. The introduction
of an indication of «determinate duration», which is of course quite com
patible with the Simple Preterite and with perfective tenses in general (cf.
(7a)), bars the «progressive» interpretation of the IPF but not the «habitual»
one, as can be seen in (7b):
(7) (a) I due mangiarono in silenzio dall'inizio alla fine.
(Both people ate silently from the beginning to the end).
(b) I due mangiavano in silenzio dall'inizio alla fine.
(Both people used to eat silently...).
Indeed, the presence of an indication of «determinate duration», although
strictly incompatible with the definition of the «progressive» aspect as
described above, is not rejected by the «habitual» aspect, provided that the
number of iterations of the process remains indeterminate. In such cases,
therefore, each iteration has a determinate duration, whereas the total
74 PIER MARCO BERTINETTO
number of iterations is not specified. This confirms that the notion of «in
determinacy», which characterizes the imperfective aspect, must be inter
preted in quite different ways in the two cases, as noted above.
Of course, the IPF can be used in a number of ways other than those
considered so far. Without aiming at exhausting the inventory, I would like
to recall at least the following usages, which are among the most typical:
«oniric», «imminential», «ludic», «hypothetical», «prospective». Sentences
(8a-e) illustrate in turn each of these possibilities:
(8) (a) Gianni si mise a sognare ad occhi aperti: tutti lo consideravano
un genio, ed ascoltavano volentieri il suo parere...
(Gianni started to day-dream: everybody regarded him as a genius,
everybody respected his opinion...).
(b) Ancora un poco e il treno deragliava
(Another moment, and the train would have derailed)
(c) Giochiamo che tu eri la guardia ed io il ladro.
(Let's pretend that you are the policeman and I am the thief)
(d) Se veniva, mi faceva piacere.
(If he had come, I would have been happy).
(e) Mi ha detto che veniva il giorno dopo.
(He told me that he would come the following day).
Each usage could be analysed separately from the point of view of its tem
poral and aspectual properties. However this analysis would be far beyond
the scope of the present paper. I therefore limit myself to the observation
that although these usages do not all share the imperfective value (cf. the
«hypothetical» IPF, which is aspectually neutral), it can be shown that their
aspectual values derive from the basic imperfective character of this tense.
Further details are provived in Bertinetto (in press).
(9) L'illustre ospite si recava ieri in visita ufficiale presso il capo dello
stato. Più tardi si intratteneva in lungo e cordiale colloquio con il
presidente del consiglio, e a sera partecipava ad una cena di gala of
ferta in suo onore.
(The distinguished guest made an official visit yesterday to the presi
dent of the republic. Later on he had a long and friendly talk with
the prime minister, and in the evening took part in a dinner of state
given in his honour).
(10) «Quella sera stessa, alle dieci in punto, l'ingegnere Ribera batteva due
colpi discreti alla porta del signor Giacomo Puttini in Albogasio
Superiore. Poco dopo si apriva una finestra sopra il suo capo e vi com
pariva il visetto imberbe del sior Zacomo» (A. Fogazzaro).
(The same evening, at 10 o'clock precisely, the engineer Ribera knock
ed discreetely at the door of Mr. Giacomo Puttini in Albogasio
Superiore. Shortly afterwards, a window above his head opened and
the beardless face of sior Zacomo appeared).
(11) Al 30° Dossena scendeva sulla sinistra, stringeva al centro e mancava
di un soffio la conclusione.
(At the 30th minute, Dossena ran down the left side of the pitch, veered
to the middle and only missed scoring by a hair's breadth).
It should be emphasized that the usage of the NIPF is not as widespread
as the warnings of a few worried purists may have suggested. Its presence
in literary texts is quite limited (with the provisos I shall give later on), and
its usage in newspapers is nowadays minimal. The NIPF may be found
somewhat more regularly in some very special types of discourse, such as
speeches justifying awards for valour, and biographies:
(12) Incurante del pericolo, il nostro si gettava nelle acque turbolente dei
fiume, e dopo aspra lotta contro gli elementi riusciva a trarre in salvo
l'incauto giovane.
(Ignoring the danger, our hero dived into the river's turbulent waters,
and after a hard battle with the elements managed to save the careless
youth).
(13) Due secoli fa, nasceva a Bonn L. van Beethoven.
(Two centuries ago, L. v. B. was born in Bonn).
Now, the question is: On what grounds are we entitled to say that
sentences (9-13) illustrate the use of the NIPF? The obvious test for ascer
taining the presence of the NIPF consists, as is well known, in replacing
76 PIER MARCO BERTINETTO
the IPF by the Simple Preterite. Whenever this can be done without introduc
ing striking differences in meaning, we have good reason to believe that the
IPF has indeed been used in a perfective context (the reader can easily check
whether this is so in (9-13)).
A more principled answer may be derived from the analysis developed
in the preceding section. In fact, the NIPF shows up whenever the condi
tion of «indeterminacy», which is typical of the imperfective aspect, is not
fulfilled in the given context. Now, as we have seen, this condition may be
violated with the IPF in at least two ways: when the context provides an
indication of «determinate duration», and when it specifies the number of
iterations. The latter situation is illustrated in ex. (4) above, which exhibits
the type of NIPF deriving from violation of the conditions ensuring the ex
istence of the «habitual» IPF. The former situation, on the other hand, can
be illustrated by the following sentence, which shows that the second major
source of the NIPF lies (quite predictably) in the infringement of the condi
tion characterizing the «progressive» IPF:
(14) Il drappello marciava per due ore.
(The platoon marched for two hours).
Needless to say, this sentence would be perfectly acceptable, according to
our discussion of ex. (7b) above, even in the case of an «habitual» reading
of the IPF. However, if the context suggests a semelfactive reading, we
necessarily have an instance of NIPF.
Examples (4) and (14) are among the most striking illustrations of the
NIPF. In most cases, however, violation of the conditions for indeterminacy
is less explicit, and must be inferred through a careful examination of the
linguistic context. Thus, in (9), (11) and (12) the «narrative» reading of the
IPF is suggested by the presence of a series of events, each of which pro
vides a temporal boundary to the preceding one, thus preventing any kind
of «indeterminate continuation» of the process. As for (10) and (13), the
NIPF is the result of the combined action of a punctual adverbial and a
non-durative verb linked to it, which again precludes (at least in the given
contexts) the imperfective reading.
text reflects the beginning of the crisis which would soon affect the entire
aspectual system. There is no doubt, in fact, that the aspectual flexibility
of old Italian was lost rather soon. According to Wilmet (1973), the «ab
solute» IPF had been almost completely generalized in French by the 14th
century. I am not able to specify when the same process occurred in Italian,
but it is likely that it was not much later. Considerations such as this have
led several scholars (e.g. Tekavčič (1972:512)) to state that the NIPF is only
the final stage (up to now) of this crisis.
If this were so, however, it would be very difficult to justify examples
such as the following, which exhibit something very close to the NIPF. What
we observe here, indeed, is something quite similar to the «weak» type of
NIPF analysed above in (15-17); the examples come however from old texts:
(26) «Lo marito correndo al romore per aiutare la moglie, e '1 lume si
spense, si trovò nella detta acqua; ed essendo nell'acqua cominciò a
gridare: - Accurr'uomo -. Li vicini, udendo il romore, scendeano le
scale per sapere che fosse...» (Trecentonovelle).
(The husband, hearing the noise, ran to help his wife, but the light
went off, and he found himself in the water; he then started to shout:
- Help! -. The neighbours, hearing the noise, came downstairs to see
what was happening).
(27) «Mallio quando si vidde così assalito prese una lancia a due mani et
uccidea et abatteva cavalieri e cavalli...» (Fatti di Cesare).
(Mallio, when he realized that he was being attacked, took a spear
in his hands and started to kill and knock down knights and horses).
Furthermore, if this usage were really contrary to the aspectual system of
old Italian, we would presumably not find it in writers such as the Abbot
Cesari (first half of the 19th cent.), renowned as one of the most enthusiastic
purists of Italian literary history (notice, by the way, the use of the «Passé
Antérieur» si furono trovati in the main clause, an unmistakably archaic
choice):
(28) «I tre, allo scocco appunto dell'ora si furono trovati a casa il sig.
Giuseppe: sicché tutti e tre montavano insieme le scale, ridendo in
sieme di questa loro accuratezza; ed entrarono nella camera del
Torelli...» (Abbot Césari).
(The three men, at exactly one o'clock, met in front of the house of
Mr. G.; so that all three went up the stairs together, laughing at their
punctuality, and entered Torelli's room).
NARRATIVE IMPERFECT 83
5. Concluding remarks.
The view that I like to propose here is significantly different from the
one usually accepted. I do not consider the NIPF as a kind of extreme cor
ruption of the Italian (and Romance) aspectual system: rather, I view it as
a partial revitalization of the original flexibility of the language. Indeed,
the great majority of instances of NIPF which can be observed in literary
texts belongs, as noted in section 3, to the «weak» type of NIPF, i.e. the
NIPF which actualizes the sort of ambiguous merging of perfective and im-
perfective features illustrated in (15-17). Following Berrettoni (1972), I have
defined these examples as «live» aspectual metaphors, inasmuch as the basic
imperfective characterization of the NIPF is not completely lost.
Of course, the specific ratio of perfective /imperfective meaning injected
into each manifestation of NIPF varies from one case to another, and may
be calculated only through a careful examination of the context. However,
it is important to realize that the most genuine NIPFs are only the extreme
version of a kind of metaphorical usage which, as such, already existed in
old Italian, and was deeply rooted in the aspectual flexibility of the language.
If it is ever proved, by further research, that this kind of metaphorical
usage ceased to be exploited during an intermediate stage in the develop
ment of the language, this would only mean that Italian went through a
period of rigidification of its aspectual properties. I am nevertheless per
suaded that this development is not likely to have happened. Among the
cues that lead me to this belief, I would like particularly to cite the behaviour
of «verba dicendi», which traditionally enjoyed considerable aspectual
freedom, as was noted also by Dauses (1981):
(29) «Colui minacciava Carmignano e diceva: - Tu mi fai perdere questo
gioco; al corpo e al sangue che io te ne pagherò -. Carmignano allora
disse...» (Trecentonovelle).
(He threatened . and said: - You are making me lose; you will pay
for this in blood -. Carmignano then said...).
Furthermore, the received hypothesis that the NIPF is the most recent
84 PIER MARCO BERTINETTO
REFERENCES
BERNARD H. BICHAKJIAN
University of Nijmegen, The Netherlands
2. Existing explanations.
As Hawkins has pointed out, «there is no shortage of proposals in the
88 BERNARD H. BICHAKJIAN
is changing its word order from SVO to SOV through such a process
(1974:209). Recently, however, Li and Thompson's already-contested in
terpretation has been convincingly refuted by Sun and Givón, whose data
from oral and written «texts» and from language acquisition provide a «pro
file [that] is not consistent with a drift from SVO to SOV; if anything it
is consistent with continued and vigorous preservation of Mandarin as a
rigid SVO language» (1985:348). With the invalidation of what was thought
to be the Chinese evidence, the alleged change from SVO to SOV through
grammaticalization loses all empirical support. The African languages also
fail to provide an example of an SVO language becoming SOV through such
a process.
An SVO language may develop an SOV structure to express certain
grammatical notions (cf. also the VSO order used in certain cases by the
Germanic languages to express optativity), but the evidence seems to indicate
that in such cases, SOV is used for marked structures (cf. Sun and Givón,
1985: 348 for Mandarin Chinese and Sasse, 1977:113-4 for an African ex
ample).
There is also very little support for Sasse's presumption: certain incon
sistent SOV languages of Africa «waren vermutlich früher SVO-Sprachen»
(1977:109). There is no guarantee that the incongruous elements are ancient
relics; they could very well be innovations. The features that he mentions
(1977:108-9) are indeed those that Latin had introduced on its way from
Indo-European SOV to Romance SVO.
The evidence for the alleged changes from SVO to SOV in unmarked
situations being found wanting both as a result of grammaticalization or
of any other natural process (borrowing being, therefore, excluded), let us
return to the change from SOV to SVO. Sasse believes that grammaticaliza
tion may serve as one of the possible factors (1977:115). This may indeed
be so, but it should be noted that while this process introduces a right branch
ing phrase, the verb is recategorized as a postposition, a grammatical item
that is consistent with the SOV character of the language. From that point
on in the evolution, two subsequent changes may take place: either, as
Hyman points out, the postpositional phrase is moved to the left of the verb
and the consistency of the SOV pattern is maintained (1975:119), or the
postposition is moved to the left of its noun, becoming thereby a preposi
tion and introducing thereby a true SVO pattern. Since these two alternatives
are equally possible, it is obvious that grammaticalization cannot auto-
THE EVOLUTION OF WORD ORDER 91
matically explain the second. There must be another factor at work, but
which one?
Moreover, since, to my knowledge, the Indo-European languages have
never changed their word order through such a process, grammaticaliza-
tion cannot be considered to have a universal explanatory power (cf. also
Sasse, 1977:119). Rather, it seems to belong to the class of accidental changes,
whose occurence and explanation remain ad hoc.
I shall, for the convenience of the reader, illustrate the above rules with Latin
examples, although it should not be concluded that Latin was still consistently
applying all the phrase-structure rules of the protolanguage.
11.a Galli ... obsidionem relinquunt (Caes. B.G. V, 49i).
[The] Gauls [the] siege raised.
ll.b haec me vedisse certo scio (PL Mil. 299).
them me to-have-seen for sure I-know
I know for sure that I saw them.
12. mecum
me-with i.e. with me
13.a deorum sanguis
gods' blood
13.b magna verba
arrogant words
13.C Quod habuit id perdidit (PL Pers. 644)
what he-had that he-lost
he lost what he had
14. amaverat (am + a + v + era + t)
loved-had-he i.e. he had loved
15. rosas (= ros + a + m + s)
roses (Acc.)
16. dulciores (dulc + ior + es)
sweeter (Nom./Acc. PL).
17. melle dulcior
honey (Abl.) sweeter
sweeter than honey
18. senatus populusque romanus
senate people-and Roman
the senate and people of Rome
19. unus pluresve
one several-or
one or several
20. a undecim
one-ten i.e. eleven
20.b quinque et viginti
five and twenty i.e. twenty-five
96 BERNARD H. BICHAKJIAN
From the above rules and examples, it should be clear that the SOV
sequence is not an isolated order of constituents, but indeed one of the many
manifestations of a left-branching grammar. The study of word-order change
must therefore focus not solely on the switching around of verb and object,
but on the shift from left to right branching.
Had he decided to take the nature of the branching pattern into con
sideration, Martin Harris would not have asserted «that change in the do
main of word order ... is cyclical» (1984:183-4), nor would Gary Miller have
been led to argue in an otherwise interesting article that the Indo-European
proto- and derived languages have gone from VSO to SOV and thence to
SVO (1975). Such a sequence would imply that a once right-branching
language would have become left-branching and again right branching. Not
only does the bulk of the linguistic evidence point to Indo-European's be
ing a left-branching language (cf. structures 1-10 above), but, from an evolu
tionary point of view, Miller's hypothesis is tantamount to positing that
man's biological evolution went from a manlike primate (hominid) to an
apelike one (pongid) and then to modern man (homo sapiens). While the
zigzagging of minor (i.e. adaptive) features can certainly occur, major evolu
tionary processes (i.e. hierarchical changes) are not cyclical — in biology
and under normal circumstances mutatis mutandis in linguistics.
from the contingent changes, shown that the former proceed in the direc
tion of earlier-acquired features and strategies, and argued that this ongo
ing evolutionary process, which closely resembles man's biological evolu
tion should probably be traced back to the action of regulatory genes on
the cellular correlates of speech. I have called this the theory of linguistic
paedomorphosis.
In the present study, I shall not attempt to show that every single one
of the structures 1-10 given above is acquired earlier in its right- rather than
left-branching forms. I have no data showing, for instance, that children
could produce magis dulcis quam mel before melle dulcior or panem et
circenses «bread and games» before panem circensesque. I also know of
no comparative studies on the acquisitions of Germ. fiinf und zwanzig and
the corresponding Engl. twenty-five or the Engl. intelligent boy and the Fr.
garçon intelligent. I shall limit the discussion to two sets of crucially impor
tant structures: on the one hand left-branching nominal and verbal inflec
tion and their right-branching reflexes and on the other left-branching ver
bals and right-branching subordinates.
present indicative. However, when one takes into consideration the rest of
the conjugation with its gamut of tenses and its abundance of irregular forms
(cf. e.g. the third-pers.-sing. markers in the Sp. canta, cantó, digo, fue «sings,
sang, said, went») the situation changes: personal pronouns are acquired
halfway through the child's third year, whereas the mastery of the full con
jugation will go into the seventh or eighth year (cf. Ruke-Dravina, 1973 and
Slobin 1971).
The comparison between the acquisition of auxiliated and inflected ver
bal forms is vitiated by the aspectual value carried by one of them. The Fr.
j'ai fait «I did» is definitely acquired before je faisais «I was doing», because
the latter is marked while the former is not. Conversely, the ablauted Engl.
I did is acquired before the auxiliated I have done for the same reason (cf.
Cromer, 1974:221-4). Nevertheless, when the aspectual distinction is negligi
ble or nonexistent as between the Fr. futures je ferai «I shall do» and je
vais faire «I am going to do» the auxiliated form seems to be, if not ac
quired earlier, at least generalized sooner (cf. Grégoire, 1947:128).
became SVO, and in a few cases VSO. With postverbal râ-suffixed ac
cusatives and circumstantial complements, Standard or «Advanced» Ira
nian seems to be also entering into this second phase. In another step, which
may partially overlap the preceding, when they came to replace the late-
acquired inflectional markers, the earlier-acquired function words were given
a pattern consistent with the right-branching structures which at the clausal
level at least are acquired earlier than their left-branching counterparts. This,
for instance, was the case of the Romance languages.
It should be borne in mind that, while a steady shift from left- to right
- branching structures has been in progress, languages have always contain
ed some left-branching structures — this feature is indeed inherent — and,
at one time or another, these structures may have enjoyed the predilection
of archaizing writers or antiquity-loving grammarians, who have encourag
ed, or served to broaden their use. This was the case of German, which under
the impulse of humanistic grammarians received an iron-clad SOV order
in subordinate clauses (Behaghel, 1932:19-22), and where «a small set of
High German particles has come into use as postpositions» (Lehmann,
1971:21). I discussed in Bichakjian (1986) the French humanistic gram
marians' attemps to introduce vowel quantity into the French language and
the reasons of their failure. In an insightful article, Halle had pointed out
that, while the child always attempts to build the optimal grammar of the
language he is exposed to, adults introduce changes that may go against the
child's efforts (1962:64-5). The paedomorphic conception of language evolu
tion is somewhat different from Halle's, but in both approaches adults play
a similar role. Whereas languages proceed naturally towards ever-earlier-
acquired features, adults may seek to (re)introduce archaic features, which
are normally acquired late, containing thereby an added degree of difficul
ty, which will «enhance» the style. The obsession with manliness can take
many forms! Whatever the psychological needs of adult speakers, languages
naturally proceed towards early-acquired features, and for that reason gram
mars shift from left- to right-branching structures, non obstant the detours
or brakes imposed by conservative grammarians (cf. also mutatis mutandis
Givón's remark quoted in 2.4).
REFERENCES
0. Introduction
In many languages, the morpheme or morphemes used to indicate future
time have other uses that are not strictly temporal*. In recent work by
Ultan 1978, comparing approximately 50 languages, and by Fleischman 1982,
referring mainly to Romance and English, we find future morphemes describ
ed as indicating, along with their future meaning, one or more of the follow
ing: desire, intention, obligation, necessity, imminence, habitual, general
truth, characteristic behavior, command, polite request, supposition. If we
view this mad array in historical perspective, we would hope to find that
some of these uses line up in historical sequence, with one leading to another,
or that some of them represent retentions from earlier meanings. However,
as Fries 1927 pointed out, it is often difficult to know which of the many
uses of futures are retentions, or older meanings «glimmering through», and
which are later derivatives that might be common to all future morphemes.
We propose to present a solution to this problem based on cross-linguistic
study of the development of future morphemes. We will argue that the lex
ical sources of futures are limited — verbs indicating desire, obligation and
movement are the most common — and that futures having these flavors
have retained them from their lexical sources, and that other uses are com
mon extensions of future and can be found with future morphemes from
any source.
The morphemes that will be treated here are those that have as one of
their uses the expression of PREDICTION, i.e. the assertion by the speaker
* Work on this paper was carried out under a grant from the National Science Founda
tion (BNS 8318262).
110 J.L. BYBEE and W. PAGLIUCA
1
We choose British rather than American English because shall is rare or non-existent
in most American dialects, and because of the availability of extensive corpus-based studies
on British English.
2
The languages of Perkins' sample were randomly chosen, controlling for genetic and
areal bias. If one wants to arrive at valid statements concerning the relative frequency of a
given linguistic phenomenon in the languages of the world, then only samples so constructed
may be used; convenience samples, because they lack the controls, are inappropriate.
THE EVOLUTION OF FUTURE MEANING 111
(4) DESIRE:
Perkins' sample: Central Sierra Miwok, Serbo-Croatian,
Karankawa.
other languages consulted: English, Mandarin, Chukchi.
listed by Ultan 1978: Danish, Norwegian, Rumanian, Gallo-Ro
mance, Angevin French, Italian (occasionally),
Old Church Slavic, Modern Greek, Arabic,
Somali, Tagalog.
Heine and Reh 1982: Swahili.
Verbs of movement used for futures include both motion toward the
speaker (come) and motion away from the speaker (go), although the latter
appears to be more common. Languages with movement-derived futures are
listed in (5).
(5) MOVEMENT:
Perkins' sample: Southern Sierra Miwok (andative), Sonay
(come), Haitian, Logbara (go and come)
other languages consulted: Abipon, Arabic, English, French, Hausa, Kru
languages, Spanish, Tojolabal, Quechua.
listed by Ultan 1978: Cuna, Kwara ? ae, Bassa.
Heine and Reh 1982: Ewe (come), Acholi (come), Lango (come),
Lotuko (go and come), Duala (go and come).
The other verbal sources for futures are possession (have), existence
(be) or come into existence (become), and verbs indicating obligation (such
as shall). We believe that these sources are closely related. When predicates
such as have, be or become are used to indicate future they must be accom
panied by a form of the verb that contributes to the future sense: it is the
infinitive in many cases, which represents the action of the verb in the
abstract, as neither completed (as a past participle) nor in progress (as a
present participle). In Korean, the verb has a suffix that means «in order
to», which supplies part of the future sense. The role of the copula or marker
of possession is to attribute the infinitival predicate to the agent. This yields
the sense of OBLIGATION, SO that forms derived in this way follow a path
similar to verbs that originally meant «be obligated» or «owe».
Germanic languages are the only ones we have found that have a future
developed from a verb with the meaning «obligation», but a number of
languages have developed a future that originated in an obligation construc
tion, as shown in (6):
112 J.L. BYBEE and W. PAGLIUCA
(6) POSSESSION:
Perkins' sample: none
other languages consulted: Eastern Kru languages, Western Romance
languages (e.g. French and Spanish), Ukrai
nian.
COPULA: Ecuadorian Quechua, Korean
BECOME: German, Gothic, Cuna.
A general pattern is observable in the way that morphemes expressing
these three semantic notions gradually develop a sense of prediction. Both
DESIRE and OBLIGATION require a willful and animate agent, for they predicate
certain conditions on such an agent: DESIRE describes internal conditions
and OBLIGATION external, social conditions, MOVEMENT verbs require a sub
ject capable of movement. These original senses gradually weaken, and the
marker comes to signal prediction, and to be applicable in propositions with
any sort of subject. However, we would argue that the original sense of these
verbs is not lost entirely, but is rather retained in certain contexts, and hence
futures from different sources will have different shades or flavors of mean
ing. Thus a future derived from a verb meaning desire may have a sense
of will or willingness in certain contexts, a future derived from obligation
may give an obligation sense occasionally, and a future derived from move
ment may give the sense of being headed along a certain path, which gives
a meaning often labelled, erroneously we believe, as INTENTION.
We claim that when a future morpheme has any one of these senses,
it is a retention from the original lexical meaning of the verb from which
the future developed, and not a secondary development from the future
meaning. To argue for this hypothesis, we will examine the development
of futures from each of the three sources in English to demonstrate how
differences in their meanings are traceable to the sources from which they
arose.
more common, the prediction sense of will becomes more and more the cen
tral or dominant sense, and eventually this sense is possible even with human
agents (see example [1]).
The speaker is consulting with the wishes of the addressee, and, as Coates
points out, substituting will in this case is not appropriate, since it would
give the sense of questioning the speaker's own wishes. This use is also a
derivative of the original obligation sense of shall — obligation has a source
external to the agent (in this case the speaker) and the speaker here is check
ing this obligation with the person who is imposing it.
We noted above that an obligation verb was used to state the inten
tions of the speaker in earlier English, so the presence of many examples
such as (24) in Coates' sample supports our hypothesis:
INTENTION:
(24) and I shall get to London as soon as I can. (p. 186)
sort. The sense of prediction develops out of this meaning, and the modal
flavors described as being part of this construction, we will claim, are all
due to the original meaning «the subject is on a path moving toward a goal
(which may be an event, state or activity)».
2.5 We have argued in our discussion of will, shall and be going to,
that the differences in the uses of these future markers can be understood
as continuations of their original lexical meanings. Since obligation, desire
and movement are commonly occurring sources for future morphemes in
the languages of the world, we expect similar sequences of development to
118 J.L. BYBEE and W. PAGLIUCA
be repeated across languages. Assuming this is so, we can apply the following
reasoning to the reconstruction of the source of future morphemes whose history
is not known: if the future marker has a sense of obligation in some context,
then its lexical source was a verb that meant «to owe» or a marker of obligation
constructed with the copula or a marker of possession. If the future has a sense
of volition or desire, then its source is a verb of that meaning. The ability of a
marker to express intention or prediction is not helpful in reconstruction, since
these uses are shared by markers designated as futures whose origins differ.
Perhaps the most cogent illustration of the fact that these modal flavors
do not develop from the future meaning, but rather, when present, must
be interpreted as retentions, comes from Quechua. The inflectional future
paradigm in Quechua is heterogeneous — the first person singular and plural
exclusive inflections are based on an old movement morpheme (i.e. «going
to»); the second person is identical to the present, and the third person suf
fix appears to have evolved from an obligation marker. The evidence for
this is that in most dialects its reading is prediction, but in the dialect of
Cajamarca, it is used for obligation («must»), probability and future (Felix
Quesada, personal communication). The crucial point is that the movement-
derived first singular and plural exclusive can have readings with prediction
and intention but not readings with obligation or necessity. Only in the third
person — that is, the form derived from an obligation source — can obliga
tion or necessity be present as modal flavors.
(33) George n'est pas venu ce matin. Il aura oublié notre rendez-vous.
George didn't come this morning. He probably forgot our appoint
ment. (Fleischman 1982:132)
(34) ¿Serán las cinco, no?
It must be five o'clock, don't you think?
Similarly, the Dutch zullen (Kirsner 1969), the Korean be-future and the
Quechua third person future may all express probability. All of these futures
are derived from obligation constructions, as are the English modais in (35)
and (36) that can be used in this way:
(35) It must be five o'clock.
(36) He should be in Rome by now.
Notice that a substitution of will in these sentences introduces a stronger
sense of prediction and certainty. For this reason we suggest that the pro
bability reading of futures is specific to obligation-derived futures.
The other uses of futures that remain to be discussed constitute a third
category as shown in (37):
(37) 3. Senses that develop out of prediction and are consequently not
source-specific.
In this category fall the uses of futures as imperatives. Obligation-derived
futures may develop directly into imperatives since informing a second per
son agent that he or she has a certain obligation is functionally equivalent
in many cases to imposing that obligation, or issuing a command (as in the
Biblical thou shalt not). But other types of futures may also develop im
perative uses. Because commands necessarily refer to future acts and because
commanding is imposing a prediction on a second person, the marker used
for future prediction is semantically and pragmatically compatible with mak
ing commands.
The second group of uses in the third category are uses of the future
to express what are called «general truths» by Ultan, «predictability» by
Coates and «eternal truths» by Fleischman, as in examples (38) and (39):
(38) the antibodies are naturally occurring and over 95 percent of all reci
pients will have anti-A and/or anti-B in their serum. (Coates 1983:178)
(39) A l'égard des voleurs on ne sera jamais assez prudent.
When it comes to thieves, one can never be too careful. (Fleischman
1982:132).
These statements amount to predictions, though not necessarily predictions
120 J.L. BYBEE and W. PAGLIUCA
about future time. They are predictions based on past experiences and thus
apply to states of affairs that exist in the past, present and future. Coates
relates these predictions to statements of characteristic behavior, also bas
ed on past experience, as in (40):
(40) It is a fairly safe bet that one of the guests will want to take the empty
flask home; they make delightful lamp bases. (Coates 1983:178).
It is certainly a small step from prediction about future time to generalized
prediction, but often the «timelessness» of such statements may derive more
from the context than from the future marker. Note that the French sentence
in (39) contains the adverb jamais, and that all of the sentences cited have
indefinite subjects.
Occasionally, futures are cited as being used as indicators of habitual
action. To the extent that we have been able to document such usage, we
find that the future never constitutes the only, or even usual, means of in
dicating habitualness, and that the cases cited actually express predictabili
ty much as examples (38) and (40) do.
4. Conclusion.
What we have tried to show is that much of the meaning of grammatical
markers — particularly that which appears to be idiosyncratic and difficult
to analyze — may be attributable to the lexical sources from which they
derive. Since we find strong cross-linguistic tendencies to use the same small
set of lexical items to derive futures, and since the modal flavors of future
morphemes are also similar across languages, we hypothesize that the seman
tic changes leading to grammaticization and the changes during gram-
maticization are similar across languages. If this is so, it means that we can
specify common or frequently travelled paths of semantic change in much
the same way that we recognize common paths of phonological change, and
that these can be used to reconstruct the sources of grammatical markers.
LANGUAGE REFERENCES
ABIPON
Najlis, Elena Lidia. 1966. Lengua Abipona. Archivo de Lenguas Precolom
binas, Volumen 1. Buenos Aires: Centro de Estudios Linguísticos, Uni
versidad de Buenos Aires.
THE EVOLUTION OF FUTURE MEANING 121
ARABIC
Elbiad, Mohamed. Personal communication.
DAKOTA
Buechel, Eugene. S.J. 1939. A grammar of Lakota: The language of the
Teton Sioux Indians. Saint Francis, SD: Rosebud Educational Society.
HAITIAN
Hall, Robert A., Jr. et al. 1970. Haitian Creole: Grammar, texts and voca
bulary. Memoirs of the American Folklore Society, v. 43. Mamaroneck,
NY: Kraus Reprint Corp.
KARANKAWA
Gatschet, Albert S. 1891. The Karankawa Indians. Peabody Museum of
American Ethnology: Archaeological and Ethnological Papers, vol. 1,
no. 2. Harvard University.
KOREAN
Martin, Samuel E. 1960. Korean reference grammar. Research and Studies in
Uralic and Altaic Languages, Project no. 19. Cleveland: Bell and Howell.
Ramstedt, G.J. 1939. A Korean grammar. Helsinki: Suomalais-Ugrilainen
Seura.
Choi, Soonja. Personal communication.
KRU languages
Marchese, Lynell. 1979. Tense/aspect and the development of auxiliaries
in the Kru language family. Dissertation, UCLA.
LOGBARA
Crazzolara, J.P. 1960. A study of the Logbara (Ma'di) language. London:
Oxford University Press.
MANDARIN
Li, Charles N. and Sandra A. Thompson. 1982. A reference grammar of
Mandarin. Los Angeles: University of California Press.
MIWOK
Freeland, L.S. 1951. Language of the Sierra Miwok. Memoirs of UAL or
Indiana University Publications in Anthropology and Linguistics VI.
QUECHUA
Quesada, Felix. Personal communication.
SERBO-CROATIAN
Partridge, Monica. 1972. Serbo-Croatian practical grammar and reader.
Belgrade.
122 J.L. BYBEE and W. PAGLIUCA
SONAY
Prost, André. 1956. La langue Sonay et ses dialectes. Mémoires, no. 47.
Dakar, Senegal: Institut Documental D'Afrique Noire.
TOJOLABAL
Furbee-Losee, Louanna. 1976. The correct language: Tojolabal: a gram
mar with ethnographic notes. New York: Garland Press.
GENERAL REFERENCES
Bybee, Joan L. and William Pagliuca. 1985. Cross-linguistic comparison
and the development of grammatical meaning. Proceedings of the Con
ference on Historical Semantics and Word-Formation, Błazejewko,
Poland, March 1984, ed. by J. Fisiak. Berlin: De Gruyter.
Coates, Jennifer. 1983. The semantics of the modal auxiliaries. London:
Croom Helm.
Fleischman, Suzanne. 1982. The future in thought and language: Diachronic
evidence from Romance. Cambridge: Cambridge University Press.
Fleischman, Suzanne. 1983. From pragmatics to grammar: Diachronic reflec
tions on complex pasts and futures in Romance. Lingua 60.183-214.
Fries, Charles C. 1927. The expression of the future. Language 3.87-95.
Heine, Bernd and Mechthild Reh. 1982. Patterns of grammaticalization in
African languages. AKUP 47.
Kirsner, Robert S. 1969. The role of zullen in the grammar of Modern Stan
dard Dutch. Lingua 24.101-55.
Marchese, Lynell. 1979. Tense/aspect and the development of auxiliaries
in Kru dialects. Dissertation, UCLA.
Perkins, Revere D. 1980. The evolution of culture and grammar. Disserta
tion, SUNY Buffalo.
Scheffer, Johannes. 1975. The progressive in English. Amsterdam: North-
Holland.
Ultan, Russell. 1978. The nature of future tenses. In Universais of human
language, Volume 3: Word structure, ed. by Joseph H. Greenberg,
83-123. Stanford: Stanford University Press.
Wekker, H. Chr. 1976. The expression of future time in contemporary British
English: An investigation into the syntax and semantics of five verbal
constructions expressing futurity. Amsterdam: North-Holland.
SYNTACTIC CHANGE AND THE LEXICON
THEODORA BYNON
University of London
It is I think generally agreed that both syntax and lexicon are involved
in determining the form of a sentence and that, as regards the lexical aspect,
the decisive factor is the valency of the predicate, normally the verb. The
syntactic structure of the sentence and the valency of its verb may thus be
considered as being in a relationship of «mutual expectancy». There is less
agreement when it comes to specifying the respective domains of syntax and
lexicon and to describing the precise nature of the relationship between them.
In Generative Grammar the syntactic rules specify constituent structure and
linear order, the lexicon specifies the arguments of the verb together with
the semantic roles of these, and the link between the two is made via the
so-called lexical insertion rules. In traditional grammar on the other hand,
the separation of syntax and lexicon seems to be less sharp, and syntactic
structures tend to be described in conjunction with the classes of lexical items
with which they occur.
While therefore the synchronic aspect of this relationship has received
some attention, the question of how syntax and lexicon interact in the time
dimension has not so far been systematically explored. Any such investiga
tion will obviously have to take account of differences between theoretical
models. For clearly an analysis carried out within the framework of
Generative Grammar is likely to classify certain changes as «lexical» which,
in a more traditional framework, might be treated as «syntactic». Despite
these discrepancies, however, the interplay of lexical substance and syntac
tic form seems worth exploring, and might begin with a closer look at the
literature on what is loosely termed «syntactic change». It is true that some
diachronic developments can be, and in fact have been, described entirely
in syntactic terms without any reference to the lexicon. Word-order change
would appear to be a case in point. Another example of a change which
124 THEODORA BYNON
has been stated in strictly syntactic terms is the reanalysis of the phrase struc
ture rule S → NP + VP to S → NP + AUX + VP in the history of English.
This was put forward by Lightfoot as a prime example of «autonomous»
syntactic change «independent of considerations of meaning and use» (1979:
81-115; 153). But, as he himself showed, the new auxiliary constituent in
serted into the rule was the result of the grammaticalisation of the ancestors
of the modal verbs and we know that this process went hand in hand with
changes in the syntactic and semantic properties of these verbs in the lex
icon. Such an interplay between syntactic form and lexical substance would,
in fact, appear to be rather typical of what is traditionally termed «syntac
tic change» and, the issue being an empirical one, this should be open to
verification.
A cursory examination of the literature on syntactic change would seem
to confirm that change in the syntactic encoding of semantically like sets
of verbs constitutes a major — perhaps even the major — mechanism
of syntactic change (cf. Lakoff 1968, Ard 1977, Bynon 1983, Joseph 1983).
This would suggest that it is precisely at the intersection of syntax and lex
icon that a syntactic change takes root. The demise of the subjectless con
struction in English may serve to illustrate this type of change. It is a par
ticularly suitable example for this purpose because it is well documented
and has, even since Lightfoot's analysis of 1979 (229-39), given rise to two
book-length treatments (Elmer 1981; Seefranz-Montag 1983) and a number
of articles (Lightfoot 1981; Fischer and van der Leek 1983; Seefranz-Montag
1984). The change in question concerns the loss from the grammar of
English of structures of the type Me seems that...., Me rues of..., which
are associated with a specific lexical class of verb, and represents a discrete
episode in the syntactic change from SOV (or later TVX) to SVO word-order.
The term «subjectless construction» is generally used to refer to sentences
which lack a referential subject, the verb typically being impersonal as is
the case with weather verbs in many Indo-European languages (Latin pluit
«it rains», Old English hit snīwþ «it snows»). In Germanic other verb classes
were also construed impersonally, all apparently referring in a similar way
to events outside the control of the individual. These comprised verbs ex
pressing such physical sensations as hunger and cold (OE hine hyngreþ «he
is hungry»), mental experiences such as anger, shame, regret and desire, and
certain related concepts expressed through verbs like «to seem», «to hap
pen», «to behove» (see Elmer 1981: 29-44). One particular aspect of the
syntax of the above verbs has long been the topic of diachronic investiga-
SYNTACTIC CHANGE AND THE LEXICON 125
1
Jespersen's construed example sentence illustrating the historical development of the
subjectless construction ([1927] 1974:209) is precisely one in which the complement noun phrase
is in fact the subject and controls verb agreement (that is to say, the verb is not «impersonal»:
þam cynge (OBJECT) likodon peran (SUBJECT)
the king (OBJECT) likeden peres (SUBJECT)
the king liked pears (AMBIGUOUS)
he (SUBJECT) liked pears (OBJECT).
2
The preverbal noun phrase looks like the topicalised object of a transitive sentence
the normal unmarked order of which is SVO. But in the present case preverbal position of
the object-marked noun phrase is the unmarked order.
126 THEODORA BYNON
like a subject. It can for instance govern the deletion of a co-referential sub
ject in a conjoined clause (Elmer 1981:49):
Gode ofhreow þa and hraþe (Ø) cwaeþ to þam engle...
«God (DATIVE) rued then and quickly (he) said to the angel...»
Its historical development would also seem to single out the subjectless
construction as being isolated in the syntax as a whole. For it had its own
way of conforming to the new SVO target, either by developing a dummy
subject it, with shift of the experiencer noun phrase to postverbal position,
or by making the experiencer noun phrase into a full subject both in regard
to case-marking and verb agreement. The latter reanalysis resulted from the
fact that in Middle English the erosion of case distinctions had, where pro
per nouns and lexical noun phrases were concerned, led to structurally am
biguous sentences (see Table 1). Given the generalisation of SVO order in
the language as a whole, initial position and animateness would have
favoured reanalysis of the experiencer noun phrase as the subject, manifesting
itself in pronominal noun phrases acquiring nominative marking (I rue, etc.).
Table 1
SYNTACTIC PARADIGM OF rue AND ITS LEXICAL CLASS (after
Elmer 1981)
Canonical structure of the «subjectless construction»:
NP-dat/acc V-3sg Complement
OE: hreowan «to rue, make sorry, grieve»; sceamian «to feel/cause
shame»; ofþyncan «to cause regret», etc.
ting this position he is in line with tradition, which speaks of the demise
of «the subjectless construction», treating it is a continuing entity in the
time dimension. Elmer's unwillingness to postulate a syntactic split seems
to me to be correct, and by invoking analogical influence from semantical-
ly similar verbs he has, I think, pointed us in the right direction also with
respect to the rest of the syntactic paradigm (which he does not analyse).
For, looked at from the point of view of the grammar as a whole, it would
appear that all the variant constructions in the syntactic paradigm are, at
any given point in time, already «in the grammar» in the sense that they
are demanded by other normal personal verbs. If this is so, it implies that
we are not dealing with change in syntactic structure as such but rather with
the redistribution of lexemes between existing constructions. Elmer, however,
does not discuss these relationships except in broad statistic terms, stating
that in Old English the sentential complement type was the most frequent
and that with the nominative type preverbal experiencer constituted the un
marked order.
It would therefore appear that the diachronic developments under
examination did not in any way result in the creation of new syntax. What
did change syntactically, as an indirect result of the above developments,
was that from Late Middle English onwards every sentence must invariably
have an overt subject constituent in the nominative case and with which
the verb agrees. What is undoubtedly syntactic in this whole process is
the motivation of the change and the ensuing generalisation of SVO struc
ture.
In constrast to Elmer, whose theoretical position is fairly uncommit
ted, Fischer and van der Leek (1983) use for their analysis a recent variant
of Generative Grammar which, as already noted above, offers a fairly ex
plicit account of the intersection of syntax and lexicon. The solution they
propose is exemplary at least in that it seeks to account for both the syntac
tic variation at any one time and for the diachronic developments. They
interpret the syntactic variants as instances of parametric variation controlled
by the universal rule «move a», and the diachronic developments as chan
ge in the case-assignment rules. We shall consider these two aspects in
turn.
As shown in Table 2, the authors divide the syntactic paradigm (omit
ting, however, the it-variant) into three construction types and suggest that
each verb has all three types and that, furthermore, each type is associated
130 THEODORA BYNON
Table 2
3
Note that type (ii) is given with the less common word-order, presumably because the
more common order with the subject following the experiencer noun phrase poses problems
for configurationally defined subject and object categories. To get round this problem the authors
offer a highly technical solution in the form of a chain-government hypothesis (see pp. 358-60).
SYNTACTIC CHANGE AND THE LEXICON 131
As regards the diachronic aspect, Fischer and van der Leek assume that
when in Late Middle English the morphological case-system broke down,
this means that the language was losing the ability to assign lexical case in
the base. Construction type (i) could now no longer be generated, since a
verb could now case-govern only one noun-phrase, its object, the other noun
phrase receiving case from tense. In the absence of overt case-marking, the
co-existence of type (ii) and type (iii) variants for the same verb would now
be detrimental to communication, especially in the case of verbs such as
to like which may select two animate noun phrases. The solution adopted
by the language lay in generalising just one construction type per verb 4 .
The continued survival of Me thinks that... and similar expressions well
beyond the 16th century is accounted for by their being analysed by the
authors as a type (ii) construction, with the cause as the subject.
In Fischer and van der Leek's analysis, then, the restructuring is seen
as being essentially lexical in nature, English having changed from being
a language with lexical case (in the base) to one with syntactic case (in sur
face structure), a division which appears to correlate with the presence ver
sus the absence of a morphological case-system. There are a number of pro
blems which still require clarification, such as the possible absence of a sub
ject constituent in the lexical case stage compared with its systematic presence
in the syntactic case stage, and the assumption that despite the presence of
4
In the case of like we have the survival of type (iii) while Chaucer still had both
possibilities: It likes her «It pleases her» and She likes it «She likes it» (1983: 342). With please
on the other hand type (ii) has been generalised, except in such archaisms as If you please,
which now tend to be interpreted as type (iii).
132 THEODORA BYNON
lexical case the basic syntactic functions are configurationally defined at all
times - to the extent that the underlying structure in Old English of the cause-
subject type (him rued his sins) is given as His sins rued him (which, accor
ding to Elmer, is decidedly the rarer, marked order). But what matters for
our present purposes is the general principle rather than the technical detail.
The important element in their analysis is their claim that the loss of the
subjectless construction resulted from the loss of lexical case.
This latter statement would seem to capture in a generative framework
the same facts as does the frequently made observation in traditional terms
that, in the earlier stages of English, word-order and a viable case-system
together permitted a more direct mapping between semantic roles and mor-
phosyntactic form than is the case in present-day English. Old English was
thus closer to the ancient Indo-European languages which, it can be argued,
achieved a fairly direct matching of semantic structure and grammatical
form. Over the course of time this relatively direct and transparent rela
tionship would have become progressively obscured so that in Modern
English case is determined not by function but by position. Seefranz-Montag
(1984: 572) speaks rather appropriately of a grammaticalisation process
whereby argument types become grammatical relations.
The twofold role of case, both as a property of individual lexical items
and as a marker of syntactic relations, has long been a part of traditional
grammar also, although it appears to have received little systematic atten
tion. Thus for instance, of the two verbs meaning «to command» in Latin,
iubeo is said to govern the accusative and impero to govern the dative. At
the same time, in the syntax, the accusative is said to be the marker of the
direct object and the dative the marker of the indirect object. Now the
diachronic developments which affected the verbs of mental experience
discussed above would certainly suggest that we are dealing with a change
in case assignment in the lexicon, in view of the fact that the subsequent
fate of the object noun phrase (a label applied to the case which resulted
from the merger of dative and accusative in Late Middle English) in the con
text of verbs of mental experience appears to be quite distinct from the fate
of formally parallel object noun phrases elsewhere with verbs like to give,
to ask, to persuade. With these too, the comparable object noun phrase is
usually human and the choice of dative or accusative was originally lexical
ly determined, but here we never find a change to nominative marking. The
historical development would not therefore appear to depend on the case-
SYNTACTIC CHANGE AND THE LEXICON 133
5 It has been suggested (Wynn Chao, personal communcation) that the passive in ques
tion presupposes the construction type Give the dog the bone (as opposed to Give the bone
to the dog) in which order is the only overt marker of function.
SYNTACTIC CHANGE AND THE LEXICON 135
Winter has recently advanced the claim that we can reconstruct only
what is learnt, not what is generated (1984: 622 ff.). And since, for him,
syntax is generated it cannot be reconstructed. But if it is accepted that the
syntactic encodings and semantic role frames of verbs must be learnt in ad
dition to their phonological form and lexical meaning, there is then not much
left for «pure» syntax to cover other than the general conventions which
determine the relative order of constituents within phrases and the range
of the possible word-order variants together with their communicative func
tions. In fact the approach that I am here advocating is rather similar to
the neogrammarian position regarding syntactic change and reconstruction,
which was based on a comparison of the uses of the morphological cases
of related languages and successive language states. The difference is that
they saw case primarily in terms of syntactic functions (subject, direct ob
ject, etc.), and did not make systematic reference to the lexicon as the place
where case-frames/syntactic positions are assigned semantic roles. It would
appear, however, that it is these encodings rather than the syntactic func
tions directly which are vulnerable to reanalysis.
REFERENCES
Ard, W.J. 1977. Raising and word order in diachronic syntax. Indiana
Linguistics Club.
Bynon, Th. 1983. Syntactic reconstruction: a case study. Proceedings of the
XIIIth International Congress of Linguists, ed. by Shîro Hattori et al.
Tokyo. 244-58.
Elmer, W., 1981. Diachronic grammar: the history of Old and Middle
English subjectless constructions. Tübingen: Niemeyer.
Fischer, O.C.M. & van der Leek, F.C. 1983. The demise of the Old English
impersonal construction. Journal of Linguistics 19. 337-68.
Hopper, P.J. & S.A. Thompson. 1980. Transitivity in grammar and
discourse. Language 56, 251-99.
Jespersen, O. [1927], 1974. A Modern English grammar on historical prin
ciples. vol. iii. London: Allen & Unwin.
Joseph, B.D. 1983. The synchrony and diachrony of the Balkan infinitive.
Cambridge: Cambridge University Press.
136 THEODORA BYNON
Postscript:
The following have only just come to my notice:
Moessner, L. 1984. Impersonal constructions in Early Middle English. Studia
Anglica Posnaniensia 17. 29-38.
Krzyszpien, J. 1984. On the impersonal-to-personal transition in English.
Studia Anglica Posnaniensia 17. 63-69.
THEOD
DIE SYNTAX DER ÀLTESTEN LATEINISCHEN PROSA
GUALTIERO CALBOLI
Dipartimento di Filologia Classica e Medievale, Università di Bologna
Es ist kaum möglich, das ganze Material der Gesetze und der privaten
und staatlichen Kultus der Römer jetzt in Betracht zu ziehen 1 . Also werde
ich mich darauf beschränken, einige Bemerkungen aus einem Teil dieses Ma
terials, nämlich aus den Leges Regiae und Zwölftafelgesetzen und aus we-
nigen sakralen Schriften, den sogenannten carmina, zu behandeln 2 . Zuerst
scheint es mir, daß wenigstens von einem praktischen Gesichtspunkte aus
der Unterschied anwendbar ist, den Bruno Luiselli (1969: 123) erkannt hat,
nàmlich daB es in der ältesten lateinischen Prosa zwei verschiedene Stufen
des Ausdrucks gibt, eine niedrige und ärmliche der Gesetze und eine andere
hohe und erhabene, die in den sakralen Texten vorkommt. Dem metrischen
oder rhythmischen Wesen der carmina, die der zweiten, hõheren Stufe an-
gehören, sind schon eingehende Untersuchungen von berühmten Kennern,
wie Westphal, Peter, Norden, Thulin, Pasquali, Pighi gewidmet worden,
abgesehen von den Gelehrten der alteren Generation wie Friedrich Ritschl
und Otto Ribbeck, aber es ist den genannten Forschern nicht gelungen, jeden
Zweifel und Bedenken hinsichtlich des metrischen oder rhythmischen Cha-
rakters dieser carmina aus der Welt zu schaffen 3 . Nur eine Besonderheit
wurde von alien, wenn auch in verschiedenem Grad, angenommen, d.h. die
1
Dies würde alies betreffen, was «die Regelung des Verhältnisses vom Menschen zu den
hõheren Mächten und vom Menschen zum Menschen» umfaßt, wie E. Norden (1909 = 1958:156)
schreibt.
2
Zur Überlieferung und Authentizität der Zwòlftafeln, auch hinsichtlich der sprachli-
chen Besonderheiten und des griechischen Anteils, vgl. besonders F. Wieacker (1966: 294-353);
A. Watson (1975: 3-8; 177-186); zum Stil und Sprache der Zwòlftafeln vgl. G. Devoto (1944
= 1983: 95-97); F. Sbordone (1964); F. Wieacker (1966: 320-330).
3
Vgl. B. Luiselli (1969: 54-120), der die vorherige, diesbezügliche Literatur ausführlich
darstellt.
138 GUALTIERO CALBOLI
Erkenntnis, daß in diesen Texten ein deutlicher Parallelismus der cola und
commata vorhanden ist: selbst Luiselli (1969), der m.E. richtiggehend
alle Versuche einer metrischen Lõsung der Frage ablehnt, und besonders
die Auffassung ablehnt, wonach in diesen Texten ein metrischer Präzedenz-
fall zum Saturnier vorkomme, erkennt und betoni sogar bezüglich dersel-
ben Texte den oben erwähnten Parallelismus und hebt dieses stilistische
Merkmal in vielen carmina besonders hervor. In der Tat handelt es sich um
sehr alte carmina, wie die Begrenzung des templum in arce (Varro, ling. 7,8)
(deren Altertiimlichkeit neben modernisierten Teilen schon von Eduard
Norden (1939) aufgezeigt wurde) 4 , die lustratio agri (Cato, agr. 141, 2-3),
den ritus belli indicendi (Liv. 1,32, 5-14) und andere, von G.B. Pighi (1958)
gesammelte Texte 5 . Aber was bedeutet nun genau dieser Parallelismus?
Ich lasse natürlich das Problem der verschiedenen Schichten dieser Sprache
beiseite - die lexikalische Altertiimlichkeit der Zwòlftafelgesetze hoben
schon G. Devoto (1944 (= 1983): 73 ff.) und F. Wieacker (1966: 300
f.) hervor - und werde mich auf die syntaktischen Verbindungen be-
schranken.
Es ist das besondere Verdienst von Romano Lazzeroni (1959), erklärt
zu haben, daß dieser Parallelismus der Periodenglieder auf einer engen Ent-
sprechung der mehrgliedrigen Wortgruppen beruht, die statt einzelner Wörter
zum Ausdruck der betreffenden Dinge und Zwecke im Gebet benutzt wer-
den (vgl. Lazzeroni 1959: 26). Solch ein Verfahren, das analytisch die größ-
ten semantischen Einheiten in kleinere miteinander symmetrisch und paral
lel verbundene Stiicke zergliedert, findet nicht nur in der lateinischen Spra
che, sondern auch in sakralen Formeln des Umbrischen, Hethitischen, Lu-
wischen und Altindischen statt. Was das Umbrische angeht, zieht Lazzeroni
(1959) einige Stellen der iguvinischen Tafeln in Erwägung, nämlich das Ge
bet an Iuppiter Grabovius (Ig. Taf. VIa 23 f.) und an Fisovius Sansius (Ig.
Taf. VIb 10 ff.) und den Fluch gegen die totam Tarsinatem (Ig. Taf. VIb 58
4
Vgl. besonders E. Norden (1939: 6-9; 58 f.). Norden bemerkt, daB neben sehr alter-
tümlichen Vocabula, wie tescum, ollaner-ollaber, quirquir, cortumio, auch junge Flexionsfor-
men und Schreibarten, wie lingua (abl.), sunto, dixisse, rectissime, vorkommen. Er verteidigt
also die Hypothese, diese Formel würde «nicht lange bevor die Formel in das varronische Werk
Aufnahme fand» umgearbeitet (E. Norden 1939: 58).
5
Diese unkritische Ausgabe bietet das ganze Material dar, aber dieses Material muß in
den kritischen Ausgaben der verschiedenen Autoren nachgeprüft werden. Deshalb gebe ich
immer auch die Autoren und die Stellen an.
SYNTAX DER ÀLTESTEN LATEINISCHEN PROSA 139
6
Vgl. z.B. Ig. Taf. VIa 23 f. teio subocau suboco / dei graboui ocriper fisiu totaper
iiouina erer nomneper erar nomneper fos sei pacer sei ocre fisei / tote iiouine erer nomne erar
nomne «Te inuocaui inuoco Iouem Graboumm pro monte Fisio, pro ciuitate Iguuina, pro eius
[montis] nomine, pro eius [ciuitatis] nomine; fauens sis, propitius sis monti Fisio, ciuitati Igu-
uinae, eius [montis] nomini, eius [ciuitatis] nomini» (Übersetzung von R. von Planta 1897 =
1973: 571).
7
TS I, 1, 11d dive tvã 'ntariksãya tvã prthivyai tvã dem Himmel dich, dem Luftraum
dich, der Erde dich! Yt. 13, 153 imōm co zōm yozomoôoi ovom co osmõnom yozomoôoi «die
se Erde verehren wir, jenen Himmel verehren wir» (W. Krause 1922: 80 f.). vgl. auch H. 01-
denberg (1917): «In den eben erwähnten Beispielen tritt zugleich ein Zug auf, in dem sich wohl
eine álteste und zugleich häufigste Weise darstellt, aus den miteinander auftretenden Vorstel-
lungen eine charakteristische Figur zu formen: der P a r a l l e l i s m u s membro-
r u m» (S.5).
8
Vgl. z.B. KUB XV 34 II 6 ff. (ich gebe nur die Übertragung von Lazzeroni 1959: 131)
«E dentro nel paese di Hatti, preferito dagli dei, buono, florido, venite e porta[te] vita, salute,
lunghi anni, benessere (?) figli, [figlie nipoti] (e) pronipoti, l'amore degli dei, il [favore (?)]
degli dei».
9
Im ältesten Latein wird als kopulativische Konjunktion nur die enklitische Partikel
-que (niemals et) verwendet, und die Partikel -que bewahrt auch sptäter noch ihren archaischen
Charakter, vgl. dazu E. Norden (1939: 18-20); J.B. Hofmann-A. Szantyr (1965: 473 f.).
140 GUALTIERO CALBOLI
Viele andere Beispiele hat Elmo Lindholm (1931: 57-59) gesammelt, ein
Forscher, der in seinen «Stilistischen Studien» dem Gesetz der wachsenden
Glieder große wissenschaftliche Sorgfalt gewidmet hat, dessen Text ich wört-
lich zitieren möchte: «die sakrale Sprache ist in besonderem Grade durch
künstliche Klangmittel gekennzeichnet, wie Häufung von Synonymen, ge-
nau stilisierten Parallelismus, abwechselnd mit (bzw. verbunden mit) anwach-
senden Gliedern». Und weiter: «Die Àhnlichkeit der sakralen Prosa und der
[...] Gesetzessprache springt sogleich in die Augen, aber wir finden auch
eine große Verschiedenheit. Beide lieben Synonymenhäufung, um Deutlich-
keit und UnmiBverständlichkeit zu erlangen; Asyndeton, Alliteration, Iso-
kola und anschwellende Glieder sind in beiden gewähnlich. Aber in der
Rechtssprache beherrscht die Sachlichkeit die Wortstellung, der sakrale Ri-
tualstil dagegen wird von künstlich-stilisierten Rücksichten beherrscht; er
ist genau in Kola und Kommata gegliedert, Parallelismus und wachsende
Glieder werden zu einem Charakteristikum des Stils» (Lindholm 1931:
57-59)10. Lindholm und neuerdings Cesidio de Meo (1983: 143) vertreten
die Meinung, daB diese in der ältesten lateinischen Prosa vorhandenen Ten-
denzen später von der griechischen Rhetorik in Rom wiedererweckt und wei
terentwickelt worden sind. In der Tat drückt sich später C. Gracchus auf
folgende Weise aus: frg. 61 ORF Malc. 4 S. 196 quo me miser conferam?
quo vortam? in Capitoliumne? at fratris sanguine madet11. an domum?
10
Das Gesetz der wachsenden Glieder wird jetzt eingehend von Th.N. Habinek (1985:
175-187) wieder untersucht. In der Tat ist unleugbar, daB, wie Habinek (1985: 189) richtig be-
merkt, das erste Glied des Saturniers normalerweise langer als das zweite ist (vgl. die Samm-
lung der Saturnier von G.B. Pighi 1968: 264-276, und B. Luiselli 1967: 270-281). Die Verkiir-
zung des zweiten Kolons im Saturnier ist ein bekanntes Phànomen, das einer Art Katalypsis
entspricht, vgl. B. Luiselli (1967: 276 f.), der schreibt: «I secondi cola [...] hanno tutta l'aria
di essere accomunati da un'identica struttura che sembra chiaramente essere quella del primo
colon metelliano dabunt malum Metelli decurtato di una sillaba». Das Gegenteil geschieht im
Hexameter. So wiirde Habineks Ansicht nach (s. besonders Habinek 1985: 187-200) die Prosa
Catos von der metrischen Struktur der hexametrischen Dichtung in der Benutzung der wach
senden Glieder beeinfluBt. Ich mòchte lieber vorsichtig sein, aber teile im allgemeinen die ab-
schließende, etwas zurückhaltende, Bemerkung des amerikanischen Gelehrten (Habinek 1985:
200): «Sacrai language and Greek rhetoric contributed to the development of oratorical prose
(periodic and otherwise), but Roman poetry, too, deserves its place in any account of the de
velopment of Latin prose».
11
Sanguine madet ist die Lesung von M, während L sanguinem (-ne P 2 ) redundat gibt.
Der letzte Editor Teubnerianus, Kazimierz F. Kumaniecki, liest redundat, aber ich ziehe es
vor, der Meinung von M. von Albrecht (1971: 69 Anm. 85) zu folgen und madet als lectionem
difficiliorem anzunehmen.
SYNTAX DER ÀLTESTEN LATEINISCHEN PROSA 141
12
Enn. scen. 276 f. Vahlen2 Quo nunc me vortam? quod iter incipiam ingredi? / do-
mum paternamne? anne ad Peliae filias? Die zwei Stellen setzt schon Cicero in Beziehung,
indem er beide nebeneinander (die erste de orat. 3,214, die zweite de orat. 3,217) zitiert.
13
Vgl. Cic. Brut. 325 f.; Att. 12,6,1; Dion. Hal. comp. verb. 4, II S. 19,10 Usener-
Radermacher; E. Norden (1909 = 1958: 134-139).
14
Vgl. Cic. Brut. 100. Der Lehrer des C. Gracchus war Menelaus aus Marathus in Phoe-
nizien, ein Asiate also, wie der Lehrer seines Bruders, Diophanes von Mytilene (Cic. Brut. 104).
15
Vgl. Cic. de orat. 3,214 «[...] matremne ut miseram lamentantem videam et abiec
taml» quae sic ab illo esse acta constabat oculis, voce, gestu, inimici ut lacrimas tenere non
possent.
16
Das römische Publikum war gegen die Klausel sehr empfindlich, wie Cicero selbst be-
stàtigt, s. Cic. orat. 214 «patris dictum (spricht C. Carbo) sapiens temeritas filii comprobavit»
- hoc dichorio tantus clamor contionis excitatus est, ut admirabile esset. Quaero nonne id nu-
merus effecerit? Verborum ordinem inmuta, fac sic: «comprobavit filii temeritas»: iam nihil
erit. W. Kroll (1913: 183) schreibt diesen clamorem mehr der noch frischen Erregung iiber die
Vorgänge des Jahres 91 zu, schließt jedoch einen Einfluß der Klausel nicht aus. Vgl. auch A.
Primmer (1968: 93-99).
142 GUALTIERO CALBOLI
17
Der Plural harunce rerum ergo beweist m.E. das Streben zur Vollständlichkeit und
Unmißverständlichkeit. Wenn wir die Skala der Anapher von Chr. Lehmann (1984: 225) in
Betracht ziehen, entspricht die Wiederholung derselben Worte, wie rei ergo - rerum ergo der
ersten, niedrigsten Stufe der Grammatikalisierung.
18
Zur Grammatikalisierung vgl. Chr. Lehmann (1984: 36; 223-227; 246-248; 401-405).
SYNTAX DER ÀLTESTEN LATEINISCHEN PROSA 143
schen noch weit überwiegt. Ich mõchte hinzufügen, daB auch im oben er-
wáhnten Beispiel, d.h. in der von Cato (agr. 141) überlieferten lustrado agri,
der Relativsatz vorangeht. In den Leges Regiae und den Zwõlftafelgesetzen
finden sich insgesamt 18 Relativsätze oder Indefinita, die als Relativsätze
verwendet werden, und alle auBer einem Fall (s. Anm. 19) vorangestellt
sind 19. Nehmen wir jetzt die carmina in Betracht (die man alle in der
Sammlung von G.B. Pighi 1958 finden kann), so kann man feststellen, daB
die vorangestellten Relativsätze soweit überwiegen, daB auBer sehr wenigen
Beispielen im Gebet Scipios (Liv. 29, 27, 1 Divi divaeque, qui maria terras-
que colitis), in der evocatio deorum bei Macrobius (sat. 3,9,7-8; 10-11) und
in der devotio P. Deci Muris (Liv. 8, 9,4-8), alle Relativsätze vorangestellt
werden. Da es sich um die ältesten Texte der lateinischen Sprache han-
delt 20 , wird dadurch die Meinung von Christian Lehmann bestätigt, daB
«der indogermanische *kwi-/*kwo- Relativsatz ausschlieBlich oder primar
vorangestellt war» (Chr. Lehmann 1984: 369). Es geschieht auch nicht sel-
ten, daB ein im vorangestellten Relativsatz enthaltenes Wort wiederaufge-
nommen wird, z.B. Lat. quo die... eo die; Heth. kwēdani UDti... apēdani
UDti. Dieser Gebrauch ist Lehmanns Ansicht nach der indogermanischen
Rechts- und Sakralsprache zuzuschreiben. Ich stimme dieser Ansicht ganz
bei und mõchte zur Bestätigung aus der lateinischen Sakralsprache zwei ziem-
lich ähnliche Redewendungen anführen, die mit kleinen Veränderungen oft
vorkommen: quoius rei ergo... harunce rerum ergo (Cato, agr. 141, 2-3);
quod tibi fieri oportet... eius rei ergo (Cato, agr. 131-32). Sehr interessant
ist in dieser Hinsicht die im piaculum luci collucandi vorhandene Formel,
zumai wir hierbei, statt des vorangestellten Relativsatzes, einen in dersel-
19
(1-4) Leg. Reg. S. 8 Bruns (Fest. 189); (5) Tab. I 4 S.18 Bruns (Gell. 16, 10,5) Prole
tario civi quis volet vindex esto / cui quis Fγ civi Marshall; (6) Tab. II 2 S. 20 Bruns (Fest.
374) quid horum fuit unum ludici arbitrove, eo dies diffensus esto / Si quid horum fuat War-
mington; (7) Tab. II 3 S. 20 Bruns (Fest. 233); (8) Tab. III 3. S. 21 Bruns (Gell. 20,1,45) quis
endo eo in iure vindicit, secum ducito; (9) Tab. III 4 S.21 Bruns (Gell. 20,1,45); (10) Tab. V
4 S.23 Bruns (Vlp. frg. 26,1); (11) Tab. VII 7 S.27 (Fest. 369); (12) Tab. VIII lb S. 28 Bruns
(Plin. nat. 28,18); (13) Tab. VIII 8a S. 30 Bruns (Plin. nat. 28,18); (14) Tab. VIII 16 S. 33
Bruns (Fest. 162); (15) Tab. VIII 22 S. 33 Bruns (Gell. 15,13,11); (16) Tab. X 5a S. 36 Bruns
(Cic. leg. 2,60) homini mortuo ne ossa legito quoi pos funus faciat / quoi Müller Büchner
quo BA1 cuos AH (hierbei haben wir den einzigen Fall, in dem der Relativsatz nicht vorange
stellt wird); (17) Tab. X 7 S. 37 Bruns (Plin. nat. 21,7); (18) Tab. X 9 S. 37 Bruns (Cic. leg. 2,60).
20
M a n muß natürlich auch mit einer Modernisierung dieser Texte rechnen, aber die Hei-
ligkeit solite sie vor einer zu starken Veränderung bewahren (so schon G. Devoto 1944 = 1983:
73).
144 GUALTIERO CALBOLI
ben Weise vorangestellten Konditionalsatz antreffen, und sehr oft der mit
si oder ni eingeleitete Konditionalsatz in der alten Gesetzsprache begegnet:
Cato, agr. 139 (piaculum luci collucandi) Si deus, si dea est, quoium illud
sacrum est, uti tibi ius est porco piaculo facere illiusce sacri coercendi ergo
harumque rerum ergo, sive ego sive quis iussu meo fecerit, uti id recte fac
tum siet. eius rei ergo te hoc porco piaculo inmolando bonas preces precor,
uti sies volens propitius domo familiaeque meae liberisque meis. harunce
rerum ergo macte hoc porco piaculo inmolando esto. Auch die Beziehung
zwischen Relativ- und Konditionalsatz wurde schon von Christian Lehmann
(1984: 330-341) behandelt. Diese Wendungen durften natürlich nicht ohne
Unterschied benutzt werden, aber sie besitzen auf jeden Fall eine Verwandt-
schaft, die besonders darin liegt, daß beide «keine konkrete semantische Be
ziehung zum Hauptsatz oder einer seiner Konstituenten haben» (Chr. Leh
mann 1984: 331). Solch eine Verwandtschaft wird durch die Sprache der
Leges Regiae21 und der Zwòlftafelgesetze nochmals deutlich bestätigt. In
diesem Gesetzkorpus gibt es 36 Konditionalsätze, während der Relativsatz
etwas seltener vorkommt: ich habe nur 15 Falle gefunden22. Aber man muß
3 Indefinita, die als Relativ benutzt werden, hinzufügen: Tab. I S. 18 Bruns
(Gell. 16,10,5) proletario iam civi quis volet vindex esto; Tab. II S.20 Bruns
(Fest. 372, 13) quid horum fuit unum /si quid horum fuat vitium Warming-
t o n / ludici arbitrove reove, eo dies diffisus esto; Tab. III S.21 Bruns (Gell.
20,1,45) M iudicatum facit aut quis endo eo in iure vindicit, secum ducito
(hier quis = ni quis). Außerdem findet man auch 3 Koniunktionen relativi-
scher Herkunft: Tab. I S. 19 Bruns (Rhet. Her. 2,20) Rem ubi pacunt, ora
to; Tab. V S.23 Bruns (Vlp. frg. 11,14) Vti legassit super pecunia tutelave
suae rei, ita ius esto; Tab. VI S.25 Bruns (Fest. 180,9) Cum nexum faciei
mancipiumque, uti lingua nuncupassit, ita ius esto. Sehr selten, aber sehr
wichtig ist die Wendung si qui(s): Leges Regiae S.10 Bruns (Fest. P. 221)
Si qui hominem liberum dolo sciens morti duit, parricidas esto; Tab. VI
S.25 Bruns (Gell. 20,10,7) Si qui in iure manum conserunt eqs.
21
Zum Charakter und zur Datierung der sogenannten Leges Regiae vgl. E. Gabba (1966:
161) und besonders F. Wieacker (1966: 327 f.; 360 f.: zwischen den Leges Regiae und den Zwölf-
tafelgesetzen gibt es eine unleugbare Übereinstimmung; denn sie sind «mit dem formalen Auf-
bauschema des Zwölftafelsatzes völlig identisch» (Wieacker 1966: 327).
22
S. oben A n m . 19, wo ich alle Stellen angegeben habe, in denen sich entweder ein Re
lativsatz oder ein Indefinitum befindet. Die drei mit einem Indefinitum eingeleiteten Sàtze sind
auch dort von mir wiedergegeben worden.
SYNTAX DER ÀLTESTEN LATEINISCHEN PROSA 145
23
Vgl. J.B. Hofmann-A. Szantyr (1965 = 1972: 667).
24
Die Zergliederung der im Text angegebenen A . c . I . -Konstruktionen, in denen das re-
gierende Verbum zwischen den Akkusativ u n d den Infinitiv eingeschoben wird (se sierit testa-
rier, me sentio dixisse), ist eben ein Beweis, daB wenigstens in diesen Fallen der Infinitivsatz
kein Kolon bilden darf.
146 GUALTIERO CALBOLI
25
Vgl. G. Calboli 1983: 16 f.
26
T a b . I 1 S. 17 Bruns ( P o r p h . H o r . sat. 1,9,76); T a b . V 3 S. 23 Bruns (Vlp. frg. 11,
14). Subjektlose Sàtze treffen sich besonders in Bedingungssätzen, aber nicht nur dort (vgl.
dazu F. Wieacker 1966: 322 f.).
27
E . Fraenkel hat spàter dem P r o b l e m der Kola und der Gliederung des griechischen
und lateinischen Satzes groBe Aufmerksamkeit gewidmet (vgl. E . Fraenkel 1932; 1933; 1965)
und hervorgehoben, daB die abhängige Infinitivkonstruktion einmal «mit dem regierenden Ver-
b u m in syntaktischer Synaphie steht» (wie z.B. P l a u t . M e n . 538 dicam curarei; Ter. A n d r .
358 negat vidisse), ein anderes Mal «ein eigenes Kolon bildet» (wie z . B . : C. Papirius C a r b o
bei Cic. orat. 213 tu dicere solebas \ sacrarti esse rem publicam) (Fraenkel 1933: 326 f.; 1965:
50 f.). Aber Fraenkel berücksichtigt nicht in diesen Arbeiten die historische Entwicklung der
Sprache. Was unsere Frage anbelangt, möchte ich nur hervorheben, daß in den ältesten Zeiten
die erste Möglichkeit vorzuherrschen scheint, nämlich daB der Infinitiv mit dem regierenden
Verbum eine Einheit bildet.
SYNTAX DER ÄLTESTEN LATEINISCHEN PROSA 147
28
In dem Sinne, daB in der «Government and Binding Theory» von Chomsky (1981 und
1982) das Problem der anaphorischen Verknüpfung der Sätze eingehend behandelt wird. Diese
Theorie verwenden A. Bertocchi und C. Casadio (1983), u m ähnliche P h a n o m e n e im Lateini
schen zu erklären, aber s. auch A . Bertocchi (im Druck), die wichtige Einwände vorbringt.
29
DaB das Participium Praesens bis zum klassischen Latein wenig benutzt wurde, hob
schon J. Marouzeau (1910) hervor; dazu vgl. auch J . B . H o f m a n n - A . Szantyr (1965 = 1972:
384; 386-390).
148 GUALTIERO CALBOLI
LITERATUR
. 1932; 1933. Kolon und Satz: Beobachtungen zur Gliederung des anti-
ken Satzes. Nachrichten der Göttinger Gesell. der Wiss., Philol.-Hist. Kl.
(I) 1932, 197-213; (II) 1933, 319-354; jetzt in: E. Fraenkel, Kleine Beiträge
zur klassischen Philologie, l.Bd., Edizioni di Storia e Letteratura, Roma
1964, (I): 73-92; (II) 93-139.
. 1965. Noch einmal Kolon und Satz. Bayerische Akademie der Wiss.,
Philos.- hist. KL, Sitzungsberichte, 1965,2 (in Kommission bei C.H. Beck,
München).
Gabba, Emilio. 1966. Considerazioni sulla tradizione letteraria sulle origini
della Repubblica, in: O. Reverdin (publ.), Les origines de la république
romaine. Fondation Hardt, Entretiens sur l'Antiquité Classique, Tome XI-
II, Vandoeuvres-Genève, 135-174.
Habinek, Thomas N.. 1985. The Colometry of Latin Prose. University of
California: Berkeley.
Haudry, Jean. 1973. Parataxe, hypotaxe et corrélation dans la phrase latine.
Bulletin de la Société de Linguistique de Paris 68: 147-186.
. 1979. L'antéposition de la relative en indo-européen. La linguistique
15: 101-110.
Hofmann, Johann Baptist-Anton Szantyr. 1965., Lateinische Syntax und
Stilistik. C.H. Beck: München (Verbesserter Nachdruck 1972).
Krause, Wolfgang. 1922. Die Wortstellung in den zweigliedrigen Wortver-
bindungen. Zeitschrift für vergleichende Sprachforschung 50: 74-129.
Kroll, Wilhelm. 1913. M. Tullii Ciceronis, Orator, Als Ersatz der Ausgabe
von O. Jahn, erklärt von W.K., Weidmann: Berlin (Unveränderter Nach
druck 1961).
Lazzeroni, Romano. 1959. Contributo allo studio della preistoria del Car
men latino. Annali della Scuola Normale di Pisa, Serie II, Vol. 28: 119-139.
Lehmann, Christian. 1984. Der Relativsatz, Typologie seiner Strukturen,
Theorie seiner Funktionen, Kompendium seiner Grammatik. G. Narr:
Tübingen.
Létoublon, Françoise - Jean-Pierre Maurel. 1985. Passif et impersonnel, in:
J. Chocheyras, L. Dabène et autres, Autour de l'impersonnel. Ellug: Gre
noble, 7-33.
Lindholm, Elmo. 1931. Stilistische Studien. Zur Erweiterung der Satzglieder
im Lateinischen. H. Ohlssons: Lund.
150 GUALTIERO CALBOLI
ANDREW CARSTAIRS
University of Canterbury, New Zealand
Most linguists today will probably agree that bound morphemes can
be of three kinds: stems, affixes and clitics. What I want to talk about here
is the distinction between affixes and clitics. I want to suggest that the
historical development of the construction that Jespersen (1942) called the
«group genitive» in English and in certain other Germanic languages may
shed light on this distinction 1.
The group genitive construction is illustrated in (1):
(1) a. NP [that man]'s daughter
b. NP [that man upstairs]'s daughter
c. NP [that man we met]'s daughter
d. NP [that man we bumped into yesterday]'s daughter
It is clear that the constituent to which the -'s- gets attached (its host, in
Zwicky's (1977) terminology) is the noun phrase (cf. Janda 1980). Sometimes
the element immediately to the left of the -'s is the head of the host noun
phrase, as in (la), but sometimes it is a nonhead element, as in (lb), (lc)
and (1d); and there is no requirement that this non-head element should
itself be a noun. So it is quite usual today to regard the possessive -'s of
modern English as not an affix, to be described under the heading of in
flexional morphology, but rather a clitic, to be described (presumably) under
the heading of syntax. This is because inflexional morphology, like mor
phology in general, is generally regarded as being concerned by definition
with the behaviour of elements within words. Consequently, if we find some
1
This paper is a revised version of part of Carstairs (1981). Work on it was supported
in part by University of Canterbury Research Grant No. 573346 and in part by a Senior Student
ship from the Leverhulme Trust.
152 ANDREW CARSTAIRS
bound element (such as -'s) which enters into construction with constituents
larger than words, that element must be classified as a clitic.
What I have just said outlines one criterion for distinguishing between
clitics and affixes. We will call it criterion A. But clearly the importance
of the affix-clitic distinction will be more firmly established if we can find
further systematic differences between them. And in any case criterion A
alone will by no means always be adequate for determining whether a given
bound item is an affix or a clitic. Think of a language like Turkish or
Japanese in which modifiers consistently precede heads within the noun
phrase. In such a language, any bound element which always occurs at the
end of noun phrases can superficially at least be described equally well as
a clitic attached to the whole phrase or an affix attached to the head of the
phrase. Fortunately, further reasonably plausible criteria have indeed been
suggested; for example, Zwicky & Pullum (1982: 3) state:
Clitics exhibit a low degree of selection with respect to their hosts,
while affixes exhibit a high degree of selection with respect to their
stems.
The «high degree of selection» characteristic of affixes is really just the same
thing as grammatically or lexically conditioned allomorphy and, as such,
is too familiar to need much illustration; however, an Old English example
is provided in (2):
(2) Nom. Sg. Gen. Sg.
a. stãn stānes «stone»
b. tīd tīde «time»
c. sunu suna «son»
d. sweostor sweostor «sister»
Clearly, in any analysis of Old English one must recognise more than one
affix for the Genitive Singular; and stems and affixes select one another
in such a way that there is nearly always one and only one «correct» Genitive
Singular form for any noun. By contrast, the «low degree of selection»
characteristic of clitics can be illustrated by the modern English possessive
-'5. In Modern English prose and colloquial usage, the possessor noun phrase
must normally refer to a person, and there are further more or less vague
stylistic and register requirements: the possessor noun phrase must not be
too long and «heavy», and the group genitive construction is avoided
altogether in most forms of written English if the possessor NP is not «head-
final» (so that e.g. examples (lb-d) are unlikely ever to be written). But,
AFFIX-CLITIC DISTINCTION 153
provided that these criteria are satisfied, the shape of the possessive marker
is fixed: it will always be orthographically -'s, with phonetic allomorphs
/ s ~ z ~ I z / predictable on purely phonological grounds, just like the
allomorphs of the clitic version of the 3rd Singular copula 2 . Taking our cue
from this phonological predictability, let us amend the Zwicky-Pullum
criterion slightly, so as to say more precisely what we mean by «a low degree
of selection»:
The degree of selection that clitics exhibit with respect to their hosts
is limited to allomorphy entirely predictable on phonological or
semantic grounds.
Let us call this criterion B (or the Insensitivity Criterion). Phonologically
predictable allomorphy we have already illustrated; the sort of thing I mean
by semantically predictable allomorphy will be illustrated shortly.
By our criteria A and B, then, the Old English Gen. -es of (2a) must
be analysed as an affix and the -'5 of (la) must be a clitic. There is nothing
very controversial about that. But this analysis does have interesting im
plications if (as I suggest) the Old English -es is the historical ancestor of
the modern group genitive marker -'s. If we adhere to criterion B, then the
group genitive construction can arise only after the high degree of selection
between stem and affix found in Old English has disappeared through the
reduction of the range of Gen. Sg. alternants illustrated in (2) to a single
alternant (with or without phonologically predictable allomorphs); for it is
only after this high degree of selection has disappeared that the door will
be opened (as it were) for the Gen. -es to be reanalysed as a clitic rather
than an affix. Thus the history of English offers a chance to test the ap
propriateness of combining criterion B with criterion A for distinguishing
between affixes and clitics. If we find that the group genitive construction
arose before the disappearance of selection in Genitive marking, then we will
have pretty clear diachronic evidence that criterion B is inappropriate. But if
the two developments took place the other way round, then we have an ele
ment of support for our twofold criterion. So what are the historical facts?
The group genitive was not found in Old English at all; rather, where
in modern English one would say «King Malcolm of Scotland's daughter»
we find in Old English what Ekwall calls the «split genitive» construction:
2 This statement requires qualification in respect of the Plural: see the Appendix below.
154 ANDREW CARSTAIRS
3
In discussion at the ICHL it was suggested that, in southern Middle English dialects,
the developments may have occurred in the wrong order: a group genitive may have appeared
before the disappearance of Genitive Case-marking allomorphy. If so, Mustonoja and Jespersen
are wrong, and the implications for the explanation proposed here are serious. Unfortunately,
Visser's (1963) work on English historical syntax says nothing about the group genitive, so
to check the claim will require a fresh examination of all southern and Kentish Middle English
texts. Though this task is still to be done, it still seems worth putting forward here the pro
posal which motivates it.
AFFIX-CLITIC DISTINCTION 155
other than either is to Icelandic and Danish, within North Germanic; so,
other things being equal, we would not expect English and Danish to be
together on one side of an isogloss with Icelandic and German on the other
side. But this is precisely the situation with the group genitive construction.
German has no group genitive: it is out of the question to say, for example,
*des Königs von Englands Tochter or *der Kônig von Englands Tochter in
imitation of the English construction the King of England's daughter; rather,
one must say die Tochter vom König von England or die Tochter des Kônigs
von England. The same is true of Icelandic: Einarsson (1945: 110) confirms
that one can say afmselisdagur konungs islands og Danmerkur «the king
of Iceland and Denmark's birthday», but not e.g. *konung Islands og
Danmerkurs afmælisdagur. Danish, by contrast, does have a group genitive.
Jespersen (1934) reports that it has been frowned upon traditionally by
Danish normative grammarians, but it is nevertheless a lively feature of the
language. He quotes examples, overheard rather than elicited, cited in (3)
(where the Group Genitive marker -s is italicised):
(3) a. for dem der kommer-s skyld
for [those who come]'s sake
«for the sake of those who are coming»
b. et av de små børn som er her i
[one of the small children who are here for the
ferien-s fødselsdag
holiday]'s birthday
c. de som faldt på ærens mark-s efterladte
[those who fell on honour's field]'s descendants
Now it is not an accident, I suggest, that Danish resembles English also in
having only one remaining Genitive marker (which happens also to be -s).
What has happened, then, is that in Danish as in English the disappearance
of Case-marking allomorphy has opened the door to reanalysis of the one
remaining Genitive marker as a clitic rather than an affix, freeing it likewise
for attachment to whole noun phrases rather than to noun stems. This seems
consistent with the historical order of events too; a group genitive is reported
by 1500, at about the same time as -s had become the universal Genitive
marker in the Singular and had begun to be used in the Plural too, displac
ing or superimposed on the old Genitive Plural ending -a (Haugen 1976:
293-4). In both German and Icelandic, however, any such development was
156 ANDREW CARSTAIRS
4
Essentially the same causal link is proposed by Jespersen (1942: 297), though without
any explicit theory of the distinction between affixes and clitics.
AFFIX-CLITIC DISTINCTION 157
5
These data are based on the speech of two informants in Christchurch, New Zealand.
The two were an immigrant from the Netherlands and his daughter, who had recently spent
some time in the Netherlands. Their intuitions were confirmed by some native Dutch speakers
in the ICHL audience.
6
(5) is derived from one native-speaker informant in Christchurch, and was confirmed
by a native speaker (Dr Jac Conradie) at the ICHL.
158 ANDREW CARSTAIRS
ed» Singular than in the «marked» Plural. Yet, bizarre as it may seem, this
analysis does shed light on some otherwise rather mysterious facts about
the acceptability of group genitives whose head noun is Plural.
Zwicky (1975: 165) quotes Kruisinga (1932) to the effect that «such
groups as the queens of England never take a genitive suffix, although the
groups father-in-law and queen of England do». However, we cannot simply
say that a non-phrase-final Plural head (such as queens, in Kruisinga's ex
ample) blocks the addition of -'s. Consider the following:
(8) a. *my brothers-in-law's outrageous behaviour
b. those Portuguese men-of-war's dangling tentacles
(9) a. *the soldiers over there's uniforms
b. the children over there's uniforms
(10) a. *the priests of that parish's reaction to the problem
b. the women of that parish's reaction to the problem
The acceptability judgments are not as clear-cut as my distribution of
asterisks seems to imply. But, for me at least, there is a clear difference in
acceptability between the expressions in which the head of the group genitive
phrase has an s- Plural ((8a), (9a) and (10a)) and those in which it does not.
Where the head noun has an irregular Plural form, the fact that it is not
final in its noun phrase does not seem to matter. At first sight, phonological
factors may seem relevant: perhaps it is the proximity of sibilants (..5... 's)
in (8a), (9a) and (10a) which renders them unacceptable. But this cannot
be right, because similar proximity of sibilants is quite acceptable when the
first -s is not a Plural suffix: James Bond's girlfriends, Rose Macaulay's
novels, those people's uniforms.
We are now in a position to say what is wrong with (8a), (9a) and (10a).
The fact that the head of the possessor noun phrase has a Plural in -s im
poses a requirement that possession should be expressed inflexionally,
through the Genitive Plural affix -s'; but the fact that the head of the
possessor noun phrase is not phrase-final leads us to expect the phrase-final
clitic -'s as a possessive marker too. (8a), (9a) and (10a) therefore seem odd
because they contain double marking of possession, both on the head of
the noun phrase and at the end of it. (8b), (9b) and (10b) escape this oddity,
however, because the irregular Plurals men, children and women cannot be
inflected for Genitive Case, so the only marking of possession that these
phrase contain resides in the clitic -'5.
REFERENCES
THOMAS D. CRAVENS
Loyola College in Maryland1
0. Introduction.
Recent developments in multitiered phonology, and the efforts of a few
scholars to establish principles of relative phonological strength, present the
possibility for a coordinated approach which promises to contribute much
to the elaboration of a phonological framework suited to the study of sound
change. By offering a cogent description of the syllable and its constituents,
the multitiered approach provides the basis for analyzing common changes
such as assimilation, mora loss, and epenthesis in terms which emphasize
the fact that mutations are typically conditioned by prosodic context. Prin
ciples of relative phonological strength open the way to a theoretical
understanding of the way in which changes occur selectively within a
language.
Stated more ambitiously, if multitiered analysis allows a clear statement
of how a given change occurs, incorporation of the theoretical concept of
phonological strength can add another dimension to understanding, by enabl
ing a theoretical account of why the change occurs as it does. This will be
exemplified here in a cursory study of pandialectal gradient loss of conso
nant length at the surface in the Italo-Romance dialects of Corsica. All Cor-
sican data are from Dalbera-Stefanaggi (1978).
1
This paper is derivative of one presented at the 15th Linguistic Symposium on Romance
Languages, held in Boston in March of 1985. I thank discussants there as well as at Pavia for
comments, in particular John Hajek, Aldo Prosdocimi, Mario Saltarelli, Tom Shannon, Nigel
Vincent, Dieter Wanner, and Roger Wright; all shortcomings are my responsibility alone. I
am also grateful to the Linguistic Society of America and the U.S. National Science Founda
tion for making my attendance at ICHL 7 possible, and I especially thank Margaret Reynolds
of the LSA for kind assistance.
164 THOMAS D. CRAVENS
1.1 Vero.
In the village of Vero, slightly to the south and west of the center of
the island, voiced stops cannot lengthen at the surface, but voiceless stops
and the liquids and nasals have surface length of the Italian type. Table 1
illustrates with just / k / and / g / to represent the voiceless and voiced stops
(SI = Standard Italian, V = Vero; representations are phonetic unless other
wise indicated, though details of vowel quality are not given).
THE SYLLABLE AND PHONOLOGICAL STRENGTH 165
/k/ SI V
anlaut kasa kaza «house»
RS trek:ase trek:aze «3 houses»
V-V lakasa akaza «the house»
/g/
anlaut gola gola «throat»
RS treg:ole tregole «3 throats»
V-V lagola awola «the throat»
Table 1
The liquids and nasals at Vero differ from the stops in having length at the
surface not only phonemically in the word, and in RS, but also in absolute
anlaut 2 . Illustrating with the bilabial:
SI V
anlaut mam: a m:am:a «mother»
RS trem: am: e trem: am: e «3 mothers»
V-V lamam:a amam: a «the mother»
Table 2
2
One corroboration that [1:] in anlaut and RS is not underlyingly / l : / is the fact that
native speakers report that words such as [badi] «you dance» (Std. It. [bal:i]), [spada] «shoulder»
(Std. It. [spal:a]) contain «deux 1» — which Dalbera-Stefanaggi accepts as a native speaker
(Dalbera-Stefanaggi 1978:49) —, as opposed to the short [1] of [m:alu] «bad» (Std. It. [male]),
[m:eli] «honey» (Std. It. [mjele]), all with / l / . — It would appear that the same rule which
lengthens liquids and nasals in anlaut also affects in some way / p / , / t / , and / k / in that position.
The intervocalic consonants described as ambisyllabic and long in the present paper are transcrib
ed by Dalbera-Stefanaggi with a macron, so [p:] here is given as [p]. In the word, it is clear
that this represents a geminate: «ce que nous notons par une consonne forte, par exemple t,
équivaut à la géminée tt» (113). This consonant is also characterized as «tendue», however,
and Dalbera-Stefanaggi gives this description to anlaut-conditioned / p / , / t / , and / k / in Vero,
along with liquids and nasals. While the confusion is not serious in the case of the liquids and
nasals, since they can, and no doubt do, appear as phonetically long in all environments where
transcribed with the macron, the stops described as tendues in anlaut are clearly not phonetically
identical to ambisyllabic geminates in the word (similar realizations in closely cognate Sardi
nian are described as energiche by Loi Corvetto 1982:87). Whatever the actual distinction bet
ween intraword geminates and these phrase-initial stops, it would seem that the latter should
be interpreted, along with anlaut-lengthened liquids and nasals, as phonetic variants of generaliz
ed strengthening. This will remain undecided here, as it does not affect the analysis of RS.
It simply adds aniaut position as a strengthening environment (though it does raise interesting
problems for multitiered analysis, especially for that proposed by Chierchia 1982).
166 THOMAS D. CRAVENS
The voiced stops are most subject to loss of length at the surface, followed
by the voiceless stops, then nasals and / r / , and / l / is left as the most resis
tant to LL.
2. Syllabic Analysis.
The synchronic state of any of these dialects can be described in a
number of ways. So far, the most interesting account proposed for RS, in
that it provides a clear theoretical motivation for interword lengthening of
the sort found in [trek:ase], is that of Standard Italian resyllabification within
the multitiered approach, exemplified by Saltarelli (1983)3.
In this analysis, the so-called geminates are manifestations of a rhyme
structure containing an empty syllabic coda which is filled by rule. To give
an example in the word, [fa:to] and [fat:o] are distinguished at the syllabic
level by their CV structures. As would be expected, fato has the structure
CVCV, and fatto is construed as CVCCV, with, however, an unassigned
empty coda branching from the first rhyme (1).
3
Saltarelli's is not the first multitiered description of Italian geminates and RS. It is,
however, the clearest and most straightforward, and thus best suited for the purposes here.
See Vogel (1978, 1982), and Chierchia (1982) for alternative analyses of RS in the multitiered
approach, and Papa (1981) for an argument against the need for this account of RS. In fact,
the case of / l : / → [d] but RS and anlaut-Strengthened [1:] in Vero suggests that an analysis
which distinguishes underlying length from rule-derived length might be preferable to one which
posits a unique source for all long consonants. As the purpose of the present paper is to ex
amine the utility of strength principles rather than to arrive at the ideal multitiered description
of Italian and Corsican RS, the competing proposals will not be evaluated here. Regardless
of which version would best describe syllabic restructuring in Corsican, the problem of ac
counting for the differential effect of LL remains.
168 THOMAS D. CRAVENS
(2) ENC: Reassociate the empty node with the adjacent segment.
with the Italian reassociation specification:
(3) for Italian: Consonants reassociate to the right.
and finally the Italian surface phonetic rule:
(4) Bimoraic consonants are phonetically long.
This gives (5)
(5) R R
The same description applies to Italian RS. Lexical items such as tre which
cue lengthening in the phrase are interpreted as containing the empty node.
When juxtaposed to a following consonant in the breath group, the ENC
is actuated with its Italian specification, along with the Italian interpreta
tion of bimoraic consonants, and the result is the surface form [trek:ase] (6).
The empty node analysis is applicable to the Corsican dialects as well, though
here dialect-specific restrictions on surface length, such as those in (7) are
necessary.
4
The concept of breath group remains ill defined; see Napoli and Nespor (1979),
however, for a description of the minimal syntactic conditions for obtaining RS.
170 THOMAS D. CRAVENS
chical relationships. Martinet's push and drag chains come quickly to mind,
and more recently a number of others, including Zwicky (1972), Lass and
Anderson (1975), Hooper (1976), now Lass (1984) and Harris (1985), have
examined to some extent the promise of the concept of relative phonological
strength.
The most ambitious attempt to come to grips with implicational strength
is Foley (1977), a book which at best has been heavily criticized, at worst
misunderstood or simply ignored. At the heart of Foley's strength analysis
is the notion that underlying the phonological system of any language is
a universal network of strength relationships, held in tension by reciprocal
forces of phonological weakening and strengthening, manifested locally in
any language by the various degrees of mutation to which phonological units
are subject, Implicit is the assumption that sound change, whether allophonic
or phonemicized, affects at its incipience not an entire class, say, all voiceless
stops, or all velars, but rather that it begins typically with just one member
of the class. Flapping of American English / t / , but not of / p / and / k / is
a banal example of this.
A point to be emphasized, since nearly everyone who has commented
on Foley's theory has misunderstood it, is that the universal network of
strength units is abstract in the extreme. That is, the phonetic and phonemic
manifestations of universal strength relationships are language specific. In
the voiceless stop series, for example, one language might have / p / stronger
than / t / and / k / , while another might have / k / stronger than / p / and / t / ,
and so on. The universal is that relative strength relationships obtain in any
language, not that, say, bilabials will always be stronger than dentals and
velars, or any other permutation of this series.
A good amount of the criticism which Foley (1977) has received is well
taken: the presentation is often unclear, data are at times questionable, and
their interpretation is not always totally consistent. More fundamentally,
Foley's insistence that phonetics is irrelevant to phonological theory —
though he hedges on this considerably — leaves the theory with no empirical
base. 5 With no phonetic principles, there can be no a priori determination
that voiceless stops are expected to be stronger than voiced stops, geminates
5
«Phonology is not the study of sounds, but the study of rules. Phonological elements
are thus properly defined not in terms of their acoustic or articulatory properties, but in terms
of the rules they participate in» (Foley 1977: 5-6).
THE SYLLABLE AND PHONOLOGICAL STRENGTH 171
stronger than single stops, and so forth. That some strength relationships
are language-specific is evident from consideration of just a few languages;
the contention that all strength relationships are language-specific, however,
carries the theory to a level of abstraction that denies the articulatory and
acoustic reality of the speech chain. In their most straightforward manifesta
tions, at least, changes such as assimilations and intervocalic voicing and
spirantization are typically conditioned by articulatory factors.
If the theory is modified at its core, however, by assuming that phonetic
information should be central both as a heuristic tool in discovering strength
relationships, and as a motivating force for the existence of at least some
hierarchies, the concept of phonological strength becomes more attractive.
Within orders, for example, if phonetic information is admitted as a
heuristic and descriptive tool, and weakening is characteristically a movement
toward those qualities typical of vocality, i.e. aperture and/or voicing,
phonetic principles would assume that unless an overriding principle in
tervenes, geminates would be stronger than single stops, voiceless stops would
be stronger than voiced stops, and these would be stronger than spirants.
There are other principles which have been developed elsewhere, but
which are not relevant to the present discussion (Cravens 1984). The main
point here is the assumption that implicational strength hierarchies are poten
tially a useful theoretical tool, and that to a certain extent, hierarchies are
phonetically motivated. In the case at hand, the basic tenets of strength
phonology provide a theoretical means for understanding the ordered ar
rangement of resistance to LL in Corsican. If it is acceptable to view reduc
tion in consonant length (degemination) as a weakening, or conversely,
lengthening in RS as a strengthening, a first approximation of the relevant
strength relationships in the Corsican dialects can be expressed. Strength
values on the parameter of consonant lengthening can be assigned as in Table
5, enabling the panchronic generalization of (8), where S reads «relative
strength value».
Table 5
172 THOMAS D. CRAVENS
(10) Parameter L
This requires that the specification of the domain of the rule be modified
to reflect selective application. The generalization of (8) can be restated as
in (11) — (or in an improved version of (11)).
174 THOMAS D. CRAVENS
4. Conclusion.
This type of strength analysis is in many ways superficially similar to
Selkirk's (1984) proposal for indexing segments by relative sonority. But
there are some crucial differences. Whereas sonority relations apparently
are assumed to be constant across languages, i.e. phonetically universal
(Selkirk 1984: 112), the strength approach presented here is based on the
recognition that there exists not one universal hierarchy, but several net
works of hierarchical parameters, each consisting of a general categorial level
(say, all obstruents), this in turn branching into more closely defined
categories (all occlusives, all fricatives, etc.), and branching again until ex
haustive (e.g. all voiceless stops, all voiced stops). These are held in tension
in any language by both language-specific and universal relationships, which
in both cases are particular to the phonological process or rule in question.
Universally, voiceless geminates are stronger than voiceless simple stops with
regard to intervocalic voicing, for example. But in the same rule domain,
one language may voice / p / preferentially to / t / and / k / , while another
may, for example, voice / t / and / k / preferentially to / p / 6 .
Relative sonority is essentially a phonetic observation rather than a set
of theoretical principles. Inasmuch as it correlates well with a characteriza
tion of vicinity to vocality, the concept of relative sonority is of central in
terest in a principled determination of relative phonological strength, but
the notion of relative sonority is at most a component of strength analysis,
not an alternative to it. Without ancillary principles, relative sonority can
not be used to account for facts such as weakening of / k / preferentially
to / t / and / p / in peripheral Tuscany, or English flapping of / t / but not
6
Much of this interpretation is basically Irene Vogel's. She suggests (1982: 99-100) that
the strength hierarchy is composed of different levels, a universal level of broad categories,
which corresponds well to the sonority hierarchy used by Selkirk and others, and which may
well be definable in terms of relative sonority, and a language-specific level in which the relative
positions of members of each class can be expected to differ from language to language. In
this view, any language should show glides weaker than liquids, these weaker than nasals, these
in turn weaker than obstruents; within these categories, relative strength relationships are not
predictable, but discoverable in each language. This is an important step forward, but as has
been shown here, even this characterization is too powerful. There may well be universal hierar
chies corresponding to Vogel's for some rule types — intervocalic monomoraic weakening would
seem typically to follow the parameter obstruents ! spirants ! glides, for example — but am-
bisyllabic LL in Corsican demonstrates that liquids and nasals are not weaker than obstruents
in all rule applications.
176 THOMAS D. CRAVENS
7
If, as is claimed here, phonological weakening and strengthening are determined by-
phonetic conditions, the fact that no two languages have exactly identical surface phonetic
realizations may well account for diversity in arrangement of the language-specific level of
the strength scale. The Artikulationsbasis of American English, for example, is so different
from that of Italian (which itself differs by regional variety) that constant cross-linguistic con
gruence in relative strength within categories (/p t k/, /b d g/, /m n/, /r 1/) should not be
expected. This helps to understand the difficulty in establishing a language-universal ranking
of individual segments according to phonetic qualities such as sonority, and demonstrates the
utility of the theoretical concept of abstract strength units with phonemic and phonetic assign
ment determined, at least in part, language-specifically.
THE SYLLABLE AND PHONOLOGICAL STRENGTH 177
REFERENCES
Saltarelli, Mario. 1983. The mora unit in Italian phonology. Folia Linguistica
17.7-24.
Selkirk, Elizabeth. 1984. On the major class features and syllabic theory.
In: Language sound structure. Studies in phonology presented to Morris
Halle by his teacher and students, ed. by Mark Aronoff and Richard T.
Oehrle, pp. 107-136. Cambridge, MA: The MIT Press.
Vogel, Irene. 1978. Raddoppiamento as a resyllabification rule. Journal of
Italian Linguistics 3.15-28.
Vogel, Irene. 1982. La sillaba come unità fonologica. Bologna: Zanichelli.
Zwicky, Arnold. 1972. A note on a phonological hierarchy in English. In
Linguistic change and generative theory, ed. by R.P. Stockwell and R.S.K.
Macaulay, pp. 275-301. Bloomington: Indiana University Press.
CRAV
DIACHRONIC SEMANTIC PROCESSES IN THE MIDDLE VOICE
The middle voice is one of the most complex phenomena in the gram
mar of natural languages. Its intractability is reflected in the confusion over
terminology in the literature. On the one hand, the term «middle voice»
has been applied to a wide range of grammatical phenomena. On the other
hand, a large number of terms have been used for some of the same
phenomena by different people: e.g. mediopassive, pseudo-reflexive, quasi-
reflexive, agentless passive, deponent and neuter. In this paper, we will at
tempt to clarify the status of the middle voice and develop a diachronic
analysis of it, illustrating in the process a general model of grammatical
change.
We begin by calling MIDDLE VOICE or MIDDLE VOICE FORM the
morphosyntactic form in an individual language which expresses part or all
of a cluster of concepts which will be elucidated by the end of this paper,
but which roughly center around the simple or direct reflexive (i.e. coreferen-
tial agent and patient). We will call a MIDDLE USE any one of those con
cepts expressed by a middle voice form.
We make two observations concerning the nature of the middle voice,
both of which we believe apply to grammatical phenomena in general. First,
the middle voice phenomenon is both syntactic and lexical in character both
cross-linguistically and (often) in a single language. The middle voice is us
ed to express certain configurations of grammatical relations — compare
the standard transitive form in 1 to the middle form of the reflexive/
reciprocal in 2 and the (specialized) passive in 3 1 :
(1) Me felicitaron.
They congratulated me'.
(2) Se felicitaron.
'They congratulated themselves/each other'.
(3) Se organizaron los juegos.
T h e games were organized'.
One also finds the middle voice used for more unusual grammatical
constructions, such as the «presentative of an action» use of the middle form
in Spanish:
(4) Se habla mucho aquí.
There's lots of talking here'.
In addition to these syntactic uses, the middle voice form also appears
on certain lexical semantic classes of verbs, such as the class of verbs of
speech actions, particularly emotive actions, in Spanish:
(5) arrepentirse 'repent'
gloriarse 'glorify'
jactarse 'boast, brag'
quejarse 'complain, lament, moan'
decidirse 'decide'
negarse 'refuse'
Like other lexical processes, the use of the middle voice to indicate
semantic class is idiosyncratic: one finds a few semantically similar verbs
which do not take the middle form, for no apparent reason, such as ex
clamar 'exclaim'.
One also finds the middle voice used for lexical derivation, such as the
derivation of the term for 'borrow' from that for 'lend' in a number of
languages:
'borrow' 'lend'
Classical Greek
Somali (Cushitic) amahho amahh-di ('loan'; n.)
Amharic täbäddärä bäddärä
täbäwasä bäwasä
Fula (West Atlantic) wu'yake wu'yii
fe'y'yo fe'y'ya
'wash' 'dress'
Modern French se laver s'habiller
Latin lavor induor
Russian myt's'a od'et's'a
2
There are important pragmatic aspects to the middle voice phenomenon pertaining to
its syntactic uses outside the realm of grammatical relations, but these will not be discussed
in this paper (see Croft, Shyldkrot and Kemmer (in prep.)).
182 W. CROFT, H. SHYLDKROT, S. KEMMER
Old Icelandic
Classical Greek
Turkish
Dyola (West Atlantic)
Masai (Nilotic)
Amharic
Somali
Sidamo (Cushitic)
Malay (Indonesian)
Papago (Uto-Aztecan)
Yokuts (Macro-Penutian)
The semantics of these verb forms are closely related to the reflexive;
in most of these languages, there are also related active transitive verb forms
meaning 'wash [someone else]' and 'dress [someone else]'. However, there
are other verb classes whose relationship to the reflexive meaning is more
distant and obscure but nevertheless take middle voice forms in many
languages. Two of the more unusual classes are verbs of requesting,
particularly 'beg' or 'pray' (Table 3), and verbs of attacking and plunder
ing (Table 4), found only in languages with a broad-ranging middle voice
form:
We must select among the possible candidates for the relevant seman
tic parameter for the middle voice form. We considered two candidates:
valency and causality, and determined that causality was the relevant con
cept. Our analysis of causal structure as it pertains to grammatical and lex
ical semantics is based on Talmy (1972, 1976) and Croft (1985a). Causal
structure consists of fitting the participants of an action — that is, the
arguments in a clause — into a causally-linked chain of events so that their
184 W. CROFT, H. SHYLDKROT, S. KEMMER
causal relations to each other and to the event(s) can be defined. The easiest
way to visualize the causal chain is by representing participants in an event
as linked by arrows representing events in the causal chain, as in Figure 1 :
Initiator Endpoint
Action
In the indirect reflexive, the subject performs the action in such a way
that the outcome of the action indirectly affects himself. Thus, in a slightly
different manner, the subject is both the initiator and the endpoint, as in
Figure 3:
We do NOT wish to argue that the middle voice uses we have found
should all be subsumed under a general definition of situations in which
a subject is both the initiator and the endpoint of an action. That would
be contrary to both the historical and the typological evidence. First, not
every language with a middle voice form has all of the uses we describe as
middle uses. Second, in the one best attested case of the evolution of the
middle voice, that of the reflexive forms from Latin to Romance, the mid
dle voice uses were acquired gradually by the reflexive pronoun form. Third,
the middle voice in Romance and elsewhere has come to cover technically
non-middle uses, in particular the passive. Finally, it appears that the passive
form has sometimes spread to middle uses: the Latin -r form probably has
a passive source (Greenberg 1986), and the Sanskrit -ya passive gradually
gained some characteristically middle uses (Gonda 1951:1, quoting J.S.
Speyer).
Instead, it is essential to distinguish each of the middle uses from all
of the others as precisely as possible and to define the semantic relation
ships among all middle uses, as well as between middle and non-middle (i.e.
active and passive) uses. In this way we may then analyze the historical and
typological patterns that fall out from these uses. Although there is no space
here to outline the definitions of the middle uses and their semantic rela
tionships (see Croft, Shyldkrot and Kemmer (in prep.)), we will give a
diagram which roughly indicates the structure of the semantic domain, and
discuss the historical aspect of our analysis of voice.
The diagram in Figure 4 (ignoring the shading for the moment) sets
out the middle uses, and some related active and passive uses, in terms of
their semantic relatedness. For example, the indirect and causative reflex
ive are closely related to the simple or direct reflexive in that all three alter
the semantics of a normal transitive verb so that the subject is both initiator
and endpoint. However, all three differ in the directness of the subject's
involvement in the action: in the direct reflexive, the subject is the direct
initiatior and the direct endpoint; in the causative reflexive, the subject is
an indirect initiator but a direct endpoint; while in the indirect reflexive,
the subject is the direct initiator but an indirect endpoint (see Figures 2 and 3).
A theory of grammatical change based on the semantic analysis underly
ing Figure 4 consists of two parts: a theory of origins and a theory of spread.
Our theory of origins is based on direct historical evidence and indirect
typological evidence. The indirect typological evidence indicates that the mid
dle (and active and passive) voice forms cluster around certain «core uses».
DIACHRONIC SEMANTIC PROCESSES 187
[Comitative]
The core use for the middle voice is normally the direct reflexive. These core
uses are usually called «prototypes» by synchronic linguists. However, given
the direct historical evidence provided by Romance, in which the direct reflex
ive USE is the source of the middle voice FORM, it is more accurate to
describe the direct reflexive as a SOURCE USE. Our theory of origins is
that certain uses are source uses for morphosyntactic forms that later spread
to other uses which never serve as source uses. Presumably there is a fun
damental cognitive reason for the privileged status of source uses, but we
will not pursue that possibility here.
Our theory of the spread of a morphosyntactic structure to other uses
is that this spread is INCREMENTAL and SEMANTIC (see Croft 1985b).
New uses are gradually acquired by a form over time, and the order of ac
quisition of new uses is governed by the degree of semantic relatedness
of those uses. Direct semantic relations are indicated in Figure 4 by solid
lines directly linking two uses. The semantic constraint on the spread of a
form restricts the possible paths of change considerably: out of 300 pos
sible paths directly linking two uses, only 38 are semantically justifiable
and thus predicted to be possible historical changes. In addition, not all of
these paths are likely to be bidirectional, and further research may reveal
even more constraints on possible grammatical changes in the domain of
voice.
We may illustrate our hypothesis of grammatical change as incremen
tal semantic change with the evolution of the Latin reflexive to the Modern
French middle in Figure 5 (Kemmer 1985, based on data from Hatcher 1973
[1942]). The shading indicates the number of uses covered by the s- form
in Latin, Old French and Modern French. 3 The evolution of the Latin
reflexive fits our hypothesis: beginning with a direct reflexive form, it spread
gradually in different directions, but the spread was governed by the semantic
relations holding between the middle uses. It is important to note that the
active, middle and passive domains are not autonomous: all three are related
to each other, however distantly, through semantically intermediate uses.
The spread of the Latin reflexive has to be seen in terms of the concomitant
retreat of the active (unmarked) form and the passive(-r) forms from uses
such as verbs denoting cognition or change in body posture. This retreat
3 It should be noted that the use of the s- form for a lexical class, e.g. verbs of cogni
tion in Old French, does not imply that EVERY verb of cognition was in a middle voice form,
only that some number of them were.
DIACHRONIC SEMANTIC PROCESSES 189
has not occurred in English, where most of the lexical classes of verbs on
the diagram are in the active (unmarked) voice; this presumably reflects the
historical retreat and loss of the Indo-European mediopassive.
We have also included in Figure 5 a line indicating the range of the
Latin deponent form. It resembles the range of the Modern French se quite
closely. However, in Latin one must subtract the uses covered by the Latin
reflexive, so that the shape of the area covered by the deponent is more like
a doughnut. This state of affairs is also found in Russian, Turkish, Malay,
Fula and Acoma (Keresan), and reflects the simple process of renewal: the
source use of an older middle voice form is obscured by the reflexive serv
ing as the source of a newer middle voice form which then displaces the
older form at the core. (Of course, in the Latin case, it is probable that the
source of the deponent was the passive, not the reflexive; our comments
apply more appropriately to e.g. Fula).
We may conclude with some general observations about the structure
of the domain of voice which suggest more general hypotheses about seman
tic change. The uses in the domain of voice may be classified into three types,
indicated by the shading in Figure 4. The unshaded areas indicate source
uses: middle, passive and active. 4 The dark shaded area indicates MARK
ED MIDDLE uses. These are uses of the middle voice form for grammatical
constructions which have simple transitive counterparts, and which indicate
some marked alteration of the causal structure of the described event, e.g.
the causative reflexive. These uses are almost always marked in surface struc
ture in some way, often by a middle form, or by a separate form such as
the English 'each other' for reciprocals, or by a combination of a middle
form and another form such as the Latin inter + se for reciprocals or the
Modern French se faire construction for causative reflexives.
The lightly shaded area contains UNMARKED MIDDLE uses. These
are semantic classes of lexical items which are inherently «middle» in some
way or other. For example, in verbs of obtaining, the subject is inherently
an indirect benefactor by virtue of coming into possession of the thing ob
tained. As another example, certain intransitive actions such as nontransla-
tional motion ('twist', 'rotate') and change in body posture ('sit down', 'lie
down') could also be conceptualized as inherently reflexive transitive actions.
4 The reciprocal sometimes functions as a secondary source use in its own right,
spreading to naturally reciprocal actions such as 'fight [intr.]' or 'meet [intr.]'.
DIACHRONIC SEMANTIC PROCESSES 191
The unmarked middle uses are typologically sometimes active — that is,
unmarked — in form and sometimes marked as middle voice.
The status of unmarked middles reflects competition between what John
Haiman has called iconic and economic motivation (Haiman 1980, 1983).
Economic motivation, which can be roughly described as «don't mention
it if you don't have to», dictates that the middle causal structure of unmarked
middles need not be reflected in (marked) middle morphology, since it is
inherent in the event. Iconic motivation, roughly «if you mean it, say it»,
on the other hand, dictates that unmarked middle uses require marked middle
morphology since their semantics includes middle causal structure and the
marked middle morphology denotes that structure. This conflict in motiva
tion causes typological and historical variation in the expression of these
uses. In conclusion, we would expect to find that in a given semantic area,
a middle voice form will acquire the marked middle uses before it will ac
quire the unmarked middle ones, since the marked middle uses contrast with
normal active transitive uses while the unmarked middle uses do not generally
have such a contrast.
REFERENCES
DRIFT AND SELECTIVE MECHANISMS
IN MORPHOLOGICAL CHANGES: THE EASTERN NILOTIC CASE *
GERRIT J. DIMMENDAAL
Rijksuniversiteit Leiden
1. Introduction
Languages differ widely in terms of morphological complexity. It is
often assumed that, with regard to changes in morphological systems, there
is a tendency towards isomorphism, resulting in a minimization of allomor-
phy, without, however, causing homophony (Kiparsky's «levelling condi
tions»; Kiparsky 1972). This paradigmatic pressure may be counteracted by
syntagmatic modifications of a phonological or syntactic nature, which cause
the morphological component to become more complex.
This paper discusses morphologically defined allomorphs in Eastern
Nilotic languages from an historical point of view. Most languages belong
ing to this group have rich and complex morphological systems. Simplifica
tions in their respective morphologies can be accounted for on the basis of
traditional notions like analogical re-alignment. A common cause of mor
phological complications in these languages is paradigmatic displacement,
itself the result of phonetic attrition in most cases.
In one sub-group of Eastern Nilotic, the Teso-Turkana cluster, a cross-
categorial reinterpretation of allomorphic distribution has taken place, as
a result of which the form of morphological allomorphs has now become
predictable to a large extent, mainly on the basis of an innovated principle
* I would like to express thanks to Joan Bybee, Andrew Carstairs, Simon Dik, George
van Driem, Jaap van Marie and Robert Stockwell for their stimulating comments and sugges
tions.
ABBREVIATIONS:
A = aspect
= consonant PASS = passive (impersonal active)
3 = third person present or future PEN = Proto-Eastern Nilotic
3PA = third person past tense V = vowel
194 GERRIT J. DIMMENDAAL
Eastern Nilotic itself is one of the three primary branches of the Nilotic
language family, which belongs to the Nilo-Saharan phylum, one of the four
major language families of Africa (Greenberg 1963: 85-86).
The percentage of cognates in a Swadesh 200 wordlist between, for ex
ample, languages of the Teso-Turkana cluster and the Bari group is about 30%.
Languages belonging to the Teso-Turkana cluster share about 75% of
their basic items in the same wordlist. The major comparative study of
Eastern Nilotic is again Vossen (1982), who reconstructs phonological and
morphological properties of the common ancestor, Proto-Eastern Nilotic,
hereafter PEN, as well as subsequent historical developments. Reconstructed
forms referred to hereafter are taken from Vossen.
DRIFT AND SELECTIVE MECHANISMS IN MORPHOLOGICAL CHANGE 195
3. Allomorphism.
Ideally, all vowels within a word belong to one of the two harmony sets.
One of the consequences of this vowel harmony system is an (automatic)
alternation for many affixes and roots between a [-ATR] form and a
[ + ATR] form, as the following examples from Bari illustrate 1 :
singular plural
dak dak-an «pipe»
kurit kurit-ön «giraffe»
Thus, each vowel in the [ - ATR] set has a counterpart (in the vertical line)
in the [ + ATR] set. The actual systems are more complex; although in
teresting in themselves, these phonologically determined allomorphs are not
further discussed in the present paper.
Next to forms whose automatic alternation is conditioned by adjoin
ing sounds, there are variant forms of morphemes which are conditioned
by position (non-automatic alternation); these morphologically determined
allomorphs are the central issue of this paper. The form of the passive suf
fix in these languages, for example, not only depends on neighbouring
sounds, but also on the tense or aspect expressed by the verb. The follow
ing examples from Turkana illustrate this characteristic feature:
1 The transcription of examples is phonemic. Tones have not been indicated in these
examples. Several forms are subject to further (phonetic) modifications. Vowels before pause
in Turkana are de-voiced if only a single tone is associated with the vowel. They remain voiced
when the associated tone is complex (i.e. when the latter consists of a sequence of high plus
low, or low plus high tone).
196 GERRIT J. DIMMENDAAL
Each of these morphological allomorphs of the passive marker has its own
phonological allomorphs; however, the choice between, for example, -ai or
-o depends on which particular aspect marker precedes it. Similar complex
ities occur with nominal inflection and derivation in these languages. Many
Nilotic languages, including those belonging to the Eastern Nilotic branch,
have a rich number marking system for nouns, whereby either the singular
is the basic (uninflected) form, or the plural. The following examples from
Bari illustrate this system:
singular plural
plurative: kupö kupö-jin «large basket»
singulative: kuru-töt kuru «worm»
Nouns taking a number marker in the singular (but not in the plural) often
refer to entities which usually occur in large numbers (members of ethnic
groups, «worms», «flies», «mosquitos») or in pairs («ears», «breasts»). This
system does have a parallel in Indo-European languages, although in the
latter this is not expressed morphologically. Tiersma (1982: 835) notes that
«when the referent of a noun naturally occurs in pairs or groups, and/or
when it is generally referred to collectively, such a noun is locally unmark
ed in the plural». The local markedness (of singular forms in this case) in
Indo-European languages manifests itself in the sense that plurals are used
more frequently with such nouns; these nouns are also more resistent to
levelling2.
There is still a third type of number suffixation in Eastern Nilotic. In
languages of the Bari group there are certain nouns whose principal form
has a plural meaning, but these nouns take a singulative suffix in order to
indicate one item from a collection, and a plural suffix to denote a plurality
of collections, as in the following Bari examples:
«seeds»
«one seed»
«many kinds of seeds»
This system (as well as the singulative and plurative marking mentioned
earlier) is also found outside Eastern Nilotic. In the Non-Bari group of
Eastern Nilotic the third type has only been retained in a few nouns (com
pare the «special group» in Turkana as described in Dimmendaal 1983a:
240); elsewhere the unmarked (uninflected) form has disappeared, thereby
giving rise to a system of replacement, where both the singular and the plural
are marked, as in the following Turkana example:
e-cor-ot «ankle bell»
-coro-i «ankle bells»
3 The historical development of the stages through which these gender prefixes probably
have gone has been described in Dimmendaal (1983a: 222), and has been further analyzed in
Heine and Vossen (1983).
198 GERRIT J. DIMMENDAAL
4. Paradigmatic displacement.
There is considerable variation in morphological structure between
Nilotic languages in general, and between the three branches, Eastern,
Western and Southern Nilotic, in particular. In terms of Greenberg's quan
titative typology of languages (Greenberg 1960) the degree of fusion in
nominal and verbal forms is the highest in Western Nilotic languages, and
the lowest in Southern Nilotic, with Eastern Nilotic occupying an in
termediate position. The synthetic index is lowest in Western Nilotic and
highest in Southern Nilotic, again with Eastern Nilotic languages occurring
between the two extremes. Southern Nilotic languages are extremely ag
glutinative, whereas internal modification (Ablaut) of roots and stems is com
mon in Western Nilotic. Eastern Nilotic languages belong to the inflectional
type. This section discusses some of the historical processes that helped to
create, feed and disrupt the typological characteristics of Eastern Nilotic
languages.
An important historical process in Eastern Nilotic, and a frequent source
of new morphological allomorphs, is paradigmatic displacement, which has
been described as a dynamic historical process by Schuh (1980). The pro
cess (whereby a particular functional distinction is shifted from one posi
tion in the word to another position which precedes or follows) occurred
in verbal and nominal forms in various languages. The following paradigms
from Maasai and Turkana (contrasting active and passive paradigms) il
lustrate the innovation of a new morphological allomorph of the passive
in Turkana:
Maasai Turkana
Active ε-nap-ita ε-nap-it «(s)he carries (it)»
3-carry-A 3-carry-A
The final vowel of the aspect marker (*-ita) has been lost in active paradigms
in Turkana (and the other languages of the Teso-Turkana cluster). However,
in passive paradigms the allomorph of the passive marker after the aspect
marker -it is -ai (or -oi after [ +ATR] vowels). A comparison with Maasai
shows that the vowel a (or after [ + ATR] vowels) at one point was part
of the preceding aspect marker. The vowel is now an integral part of the
DRIFT AND SELECTIVE MECHANISMS IN MORPHOLOGICAL CHANGE 199
following passive suffix; in active paradigms the vowel never shows up, not
even when a plural suffix follows the aspect marker. One reason for the
complete loss of the vowel in active paradigms may be the fact that in passive
paradigms there is always a suffix following the aspect marker, whereas in
active paradigms there is only a following suffix in the plural. A further
consequence of the paradigmatic displacement in the passive paradigm is
the emergence of a new morphological allomorph of the passive, which alter
nates with, for example, the passive suffix -i (or -i) which is used immediately
after verb roots (as illustrated in section 3 above); the result is syncretism
because the form of the passive suffix also depends on the tense-aspect ex
pressed by the verb, i.e. the suffix is also an exponent of the verb form as
a whole. The final vowel ι or i of the passive markers is de-voiced before
a pause (due to tonal configurations, as illustrated in Dimmendaal and
Breedveld 1986). This de-voicing may be seen as an initial step towards com
plete loss (unless another suffix follows, to protect it from loss):
singular plural
Lotuxo
T e s o « d i g ! »
Turkana
(The prefix in these examples marks the verb class to which the verb belongs
in these languages; for a description of morphological verb classes in Nilotic
see Dimmendaal 1983b).
The original imperative suffix (*-a) is still attested in Lotuxo, but it
has disappeared as such in the singular in Teso and Turkana. However, the
vowel has been retained (next to Lotuxo) in Teso in the plural form, where
it has become part of the plural suffix. The original plural suffix (*-ta) (re-
200 GERRIT J. DIMMENDAAL
tained unaltered in Lotuxo) has become -ata in Teso, due to the paradigmatic
displacement. The number suffix -ata has been shortened to -a in Turkana.
This shortening rule does not seem to be due to phonetic attrition, because
in cases where a vowel is lost word-finally, the preceding consonant tends
to be retained in these languages; instead, the shortening seems to be the
result of a rule of proportional analogy. As the majority of verbal number
suffixes in Turkana (and Teso) has a structure -(C)V, and not -VCV, the
suffix -ata may have been remodelled by analogy with the predominant pat
tern, because of a general tendency towards paradigm coherence. The result
in Turkana is a displacement of the parameter of number suffixation from
the original suffix to a morpheme which functioned as an imperative marker
at a much earlier stage. The imperative marker (*-a) is also found outside
Eastern Nilotic, for example in the Southern Nilotic language Pokot:
kar-a shut (it)!»
-kar «shut (verbal root)»
(The presence of the word-final vowel ε in the plural form of «cow dung»
in Turkana depends on the dialect).
The PEN vowel *a in the root for «sandal, shoe» has been retained
as part of a number suffix in Turkana, whereas in the non-alternating form
(the plural in this case) it has been lost. In the case of «cow dung» the final
root vowel also became part of the following number suffix in Turkana,
but it subsequently changed from to between these alveolar consonants.
Several number suffixes can be reconstructed for PEN (Vossen 1982: 306).
However, due to processes of the type illustrated above (whereby a root final
vowel became part of a suffix whenever there was one in an alternating form),
DRIFT AND SELECTIVE MECHANISMS IN MORPHOLOGICAL CHANGE 2 0 1
new number suffixes have arisen. Sometimes the process gave rise to new
morphological allomorphs, as in the case of -at in the Turkana form for
«shoe» above, or to new instances of a particular suffix, which was already
there in the proto-language, as with the suffix -et (in the word for «cow
dung»), which has been reconstructed by Vossen (1982: 307) as *-εt. Because
there is not (or no longer) any phonological conditioning for the suffix -at
in Turkana (since the original vowel *a disappeared from the non-alternating
form) this suffix is no longer a phonological allomorph of some other suf
fix. Similar processes can be illustrated for other languages and sub-groups
of Eastern Nilotic. The result is the same: new number suffixes arise whose
form is no longer predictable on the basis of some adjacent segment because
the latter has disappeared in the non-alternating form.
Because of the number marking system in these languages, whereby
either the singular or the plural is marked with a number suffix, there is
a potential ambiguity with regard to each nominal ending, viz. whether it
ends in a number suffix or not. As there is a wide variety of morphological
and phonological allomorphs for number suffixation, this potential ambigui
ty occurs quite frequently. This often leads to reinterpretations of nouns
which originally ended in a number suffix as basic forms, and vice versa.
The following examples illustrate such innovations in Bari and Turkana:
singular plural PEN root
Bari mayat-ti mayat *-mayat «locust»
Turkana ε-maase ŋi-maase-i
Bari kurit kurit-ön *-kori «giraffe»
Turkana e-kori ŋi-kori-o
In Turkana the singulative suffix (which is still attested in Bari) has become
part of the root (*-ti > se) in the case of «locust», whereas in Bari the same
process of petrification occurred in the word for «giraffe» (*-kori-t > korit).
These processes of paradigmatic displacement (together with various
other changes, such as fusion rules for adjacent morphemes) have resulted
in considerable internal morphological diversification between Eastern Nilotic
languages.
5. A cybernetic counterpoise.
The various processes described above have resulted in new sets of suf
fixes in Teso-Turkana and elsewhere, some of which may still be called
202 GERRIT J. DIMMENDAAL
Roots with a -CVC structure in Turkana take a number suffix -in (or its
phonological allomorph -in). As with the examples from Bari above, such
roots correspond to random patterns in Bari (and other Eastern Nilotic
languages outside the Teso-Turkana cluster):
Bari Turkana PEN root
sg. ruk a-ruk *-rruk «hump»
pl. ruk-an ŋa-ruk-in
sg. kuŋu a-kuŋ *-kuŋ «knee»
pl. kuŋu-at ŋa-kuŋ-in
sg. munu a-muŋ *-munu «snake»
pl. muno-ö ŋa-mun-in
204 GERRIT J. DIMMENDAAL
Identical, or similar, criteria for number inflection are found in closely related
languages like Toposa. As the following Toposa examples show, criteria for
the selection of a particular number suffix marker may be the same for
singulatives and pluratives. Nouns with a root pattern -(C) VCVC take a
number suffix -a (or its phonological allomorph - or -o):
singular plural
«beard»
«crocodile»
«horn»
«Dreast»
(Note that the second consonant in the word for «horn» is followed by a
glide. The crucial feature here and elsewhere for mora counting is the
presence of vocalic nuclei and consonants in coda positions).
Several of the historical class shifts for nouns are shared by all languages
of the Teso-Turkana cluster, and therefore probably took place in Proto-
Teso-Turkana. The following set of -CVCV roots in Teso and Turkana takes
a number suffix -1 (or the phonological allomorph -i) according to the regular
(mora based) rules of these languages:
Teso Turkana
singular plural singular plural
DRIFT AND SELECTIVE MECHANISMS IN MORPHOLOGICAL CHANGE 2 0 5
ther (language internal) indication that the basis for number inflection in
Pre-Teso-Turkana was not based on mora counting. Additional internal
evidence comes from the structure of adjectives in these languages, which
share particular features with nouns, in that they, for example, take
singulative or plurative markers (unlike nominalized verbs, which are in
flected both in the singular and the plural (replacement)). Adjectives belong
to a closed lexical set: notions expressed by way of adjectives elsewhere in
Eastern Nilotic are often expressed by way of stative verbs in Teso-Turkana.
Adjectives do not have the regular patterning between root structure and
number suffix found in common nouns. Examples from Turkana:
singular plural
-ŋole -ŋole-ik «bald»
-kooki -kooki-ok «orphaned»
-ŋoror-i -ŋoror «last born»
singular plural
e-deke ŋi-deke-siney «disease»
a-pεga ŋa-pεga-siney «argument»
a-kipɔorɔ ŋa-kipɔro-sia «wet season»
a-kuan ŋa-kuari-sia «night»
Interestingly, more and more parts of the grammars of these languages are
affected by the mora counting principle. There is a clear drifts towards a
cross-categorial reinterpretation of morphological allomorphy, or in Sapir's
words, «there is unconscious selection on the part of the speakers of those
individual variations that are cumulative in some special direction» (Sapir
1921: 155). The mutations (historical changes) are not just random, they
have a direction. They are found in nominal, as well as in verbal morphology.
Languages like Turkana have at least four derivational suffixes which
are used productively with abstract nouns (compared to basically one suf
fix in a language like Maasai). The distribution of the four suffixes is almost
entirely based on mora counting:
206 GERRIT J. DIMMEND AAL
is (yet) unknown, neither does one know its target, one reason being that
the process of change may interact with other parameters of grammar. To
give one example, again deduced post hoc, languages of the Non-Bari group
have a system of verbal agreement marking which is cognate to the one found
in Western and Southern Nilotic languages. The agreement marking system
disappeared almost entirely from the Bari group of Eastern Nilotic. One
potential reason for this may have been related to a shift in word order in
the Bari group from an original (Proto-Eastern Nilotic) VSO (alternating
with VOS) to a SVO type, and from a system of potentially empty subject
positions to one where subjects are now obligatory. (Evidence for such a
shift is presented in Dimmendaal 1985). Languages like Bari have retained
the original plural suffix with imperative verb forms where subjects are ab
sent, as in most languages. Instead of adapting a wide range of verbal suf
fixes marking number, as in Teso-Turkana, Bari has almost completely aban
doned the suffixes (and person marking prefixes as well). Given the current
system of nominal number suffixation in Nilotic languages other than those
belonging to the Teso-Turkana cluster, there does not seem to be any deduc-
tive-nomological necessity (in the sense of Lass 1981) for a re-adaptation
of number marking in nominal forms, in other words, there is nothing
teleological about the changes in Teso-Turkana. Without denying the
relevance of functionalistic or pragmatic approaches towards language
change in general, it has to be concluded that they are not of any direct
relevance here. As an alternative attempt towards explaining the observed
restructuring in one sub-group of Eastern Nilotic, recourse might be taken
to the biological model in historical linguistics. This model has been pro
pagated, among others, by Stevick (1963). It provides a non-causal and non-
teleological way of thinking about mutations in grammatical systems, analo
gous to current ideas about evolution, as formulated by, for example. Mayr
(1968), who observes the following about mutations in biological systems:
«In view of the high number of multiple pathways available for most
biological processes and... the randomness of many of [them] ... causality
in biological systems is not predictive, or at best is only statistically predic
tive» (Mayr 1968: 54).
Mutations in biological systems are extremely common and frequent.
Accidental mutations may provide certain advantages, and therefore they
may be retained. Along similar lines, it might be argued that linguistic com
munities are not homogeneous, because of idiolectic and dialectic variation
between speakers. Mutations leading to class shifts for groups of nouns,
208 GERRIT J. DIMMENDAAL
has persisted again over a period of hundreds of years, without any sign
of loss or break down, probably because of a language-external (configura-
tional) pressure. These languages are spoken in an area where such systems
are extremely common. Aside from internal reasons (such as the functional
notion of redundancy), there is at least one external reason, the areal norm.
Any theory of morphological change claiming typological adequacy and
universal validity, and aiming at a proper understanding of what is natural
in morphological systems, should not only take into account «minimalistic»
language properties, such as constraints on allomorphy and preservation of
relevant contrasts, or paradigm economy (as defined by Carstairs 1983); it
should also leave space for language specific system adequacy, and the per
sistance of non-functional properties, in the same way as biological evolu
tion has left a place for such phenomena.
REFERENCES
DOROTHY DISTERHEFT
Department of English, University of South Carolina
1. Introduction
It has long been recognized that morphology preserves old syntax 1 .
Since the last century historical linguists have suggested that some of the
inflectional and derivational affixes of Proto-Indo-European had previously
been free lexical morphemes. However, the time depth at which such
reconstructions were made was so great that their status for the most part
remained speculative. More recently, however, interest concerning the
diachronic relationship of morphology and syntax has, quite appropriate
ly, been renewed. Most investigation in this vein can be attributed to the
impetus of Givón's celebrated archaeological excavation (1971), which ended
by paraphrasing the Chinese philosopher Lao Tse, and observed that «To
day's morphology is yesterday's syntax». Among other things, Givón argued
that earlier word order patterns determine the order of morphemes; that
agreement markers develop from pronouns and preserve earlier word orders;
that the placement of clitic object pronouns likewise preserves old word
orders, as in Romance; that inflectional and derivational morphology arises
from free lexical morphemes via an agglutinative stage. However, his handl
ing of much of the data is either idiosyncratic or marred by the syntactic
theory which he followed at the time; some of his claims have also been
invalidated. However, because Givón's view of this relationship continues
to exert strong influence, I will examine some of the subsequent revisions
to his claims. I conclude that, while his proposals about specific mor-
1 I acknowledge with gratitude the interest shown in these remarks by the conference
participants, both during the session and after. Several of my colleagues were eager for me
to incorporate a discussion of cliticization, but I ignored their advice because, while such pro
cesses are certainly relevant to the topic at hand, they are only a subset. They also involve
phonological effects and I wish to restrict my scope here.
212 DOROTHY DISTERHEFT
2 Timberlake (1977:153-8; 173n8) makes the case that pronouns in general do not
preserve old syntax. The loss of subject-to-object raising in Finnish was implemented with pro
nouns before full NPs. He cites similar cases from Yuman and Russian.
DIACHRONIC RELATIONSHIP OF MORPHOLOGY AND SYNTAX 213
at marks both NPs as subject, instead of only the sentence-initial one. The
prefix im is likewise difficult: it signals agreement with an oblique NP, of
which there are none in the sentence. The verb is an invariant form, agree
ing with neither NP, indicating location. He proposes that the synchronically
odd agreement can be explained by its development from an impersonal in
which the possessed NP was the original subject and the possessor an obli
que. When the latter was reinterpreted as subject, it was moved to initial
position and assumed subject marking, but the rest of the original construc
tion remained intact. Chung (1976) had similarly shown that among Polyne
sian languages, Tongan passives have been reanalyzed as active, but that
distribution of the earlier morphology is retained even though the basic syn
tactic structure has changed to an ergative one. For Anderson such reanalysis
involved a «rationalization of the syntax at the expense of complicating the
morphology» and while still critical of the assumption that «today's mor
phology can be taken reliably to be yesterday's syntax», he concluded that
there are definitely clear cases in which «today's syntax can be expected to
become tomorrow's morphology» (1980:67).
Anderson's revision of Givón's maxim was formalized by Cole et al.
(1980) in a study of the reanalysis of nonsubject NPs as subjects. They
concluded that the acquisition of behavioral properties (transformations)
precedes coding (morphosyntax). On the basis of diachronic developments
relating to passivized objects of verbs which control the dative case; dative
experiencers in Germanic and Georgian; and passive agents in Tongan,
they establish three stages in the transition to subjecthood: Stage A, no
subject properties are associated with the NP in question; Stage B, the
NPs develop behavioral properties of subjects, i.e. they participate in the
new transformations, but their morphology has not changed; Stage C,
the syntactic subject now has morphology consistent with its behavioral
properties. The principle which Cole at al. propose is certainly a valid
one and actually applies more generally than they argue, as I will show
presently. However, their argument is weakened by the fact that, with
two exceptions (Georgian and English), the languages they adduce as
evidence supporting this principle exhibit only one stage. Furthermore,
the analysis is circular: each language is assigned to a stage on the basis
of its morphological or syntactic properties, which are then used as witnesses
for the validity of the hypothesis. Most Germanic languages have one
stage: Proto-Germanic, Gothic, and Modern German are Stage A. Thus
in Modern German the dative experiencer (mir) of gefallen «please» lacks
214 DOROTHY DISTERHEFT
any properties whatsoever of subject (3) as does the dative object (uns)
of helfen «help» (4).
(3) Mir gefallen diese Damen
me(dat) they-please these ladies (nom)
«I like these ladies»
(4) Uns wird von der Polizei geholfen
us(dat) pass aux(3sg) by the police help(pptc)
«We are helped by the police».
Evidence from Gothic is complicated by interference from the Greek original.
It is clear, however, that dative experiencers do not control equi-NP dele
tion or conjunction reduction (p. 721). Although passivized non-accusative
objects would appear to have some subject properties (they appear in the
nominative case, control agreement and reflexivization), these passages are
literal translations of the Greek and may thus be discounted as evidence (p.
723n9). The status of Proto-Germanic is only inferred from comparative
reconstruction. Old Icelandic is transitional between A and B, whereas
Modern Icelandic is more fully B. In the earlier stages of Icelandic, the dative
experiencer of one verb pia «seem») controls reflexivization (p. 722),
whereas in the modern language more properties have been acquired by a
number of verbs. Both dative experiencers and passivized non-accusative
objects can be moved by subject-to-object raising (p. 724) and both are
operated on by equi-NP deletion. Old Swedish preserves the same stage:
dative experiencers do not control agreement, but do reflexivization (p. 726).
Modern Swedish, along with Danish and Norwegian, has reached Stage
because the experiencers now have nominative case. The same development
is seen in English. Old English accusative experiencers sometimes behave
like subjects (they may undergo subject-to-object raising, be equi-deleted
by under identity with a nominative subject) but retain accusative case (p.
729). The situation is similar in Polynesian where Samoan and Tongan have
reanalyzed passive agents as subjects in their shift to ergative morphology
(Chung 1978). In Tongan -e marks the subject, but the same marker is used
for the passive agent in Maori. Chung argued that the original function of
the -e was to mark oblique NPs as passive agent, as in Maori, whereas the
same morpheme is now attached to subjects in Tongan (and Samoan). Cole
et al. take this as evidence that the languages with -e marking the subject
are Stage because they have subject syntax but not morphology. Old
Georgian is Stage while the modern language has developed only minimal
subject coding (number agreement).
DIACHRONIC RELATIONSHIP OF MORPHOLOGY AND SYNTAX 215
documented that Irish (VSO), which has adjacent subject and object NPs,
lost most distinction (except for the mutation of the initial consonant of
a postposed modifier by lenition or nasalization) by Early Middle Irish. A
fair amount of superficial ambiguity has resulted — at least in the texts —
but there is absolutely no indication that Irish is shifting to SVO. Thus in
principle case loss or syncretism is not responsible for changes in the order
ing of constituents. If the two processes are related, it is the syntactic change
that would allow the morphological adjustment to take place.
REFERENCES
Ahlqvist, Anders, ed. 1982. Papers from the 5th International Conference
on Historical Linguistics (Current issues in linguistic theory, 21). Amster
dam: Benjamins.
Andersen, Henning. 1973. Abductive and deductive change. Lg. 49.765-93.
Anderson, John M. and Charles Jones, eds. 1974. Historical Linguistics.
Proceedings of the First International Conference on Historical
Linguistics. Amsterdam: North Holland.
Anderson, Stephen R. 1980. On the development of morphology from syn
tax. In Fisiak 1980.51-69.
Anttila, Raimo. 1972. An introduction to historical and comparative
linguistics. New York: Macmillan.
Bean, Marian C. 1983. The development of word order patterns in Old
English. London: Croom Helm.
Buck, Carl Darling. 1933. Comparative grammar of Greek and Latin.
Chicago: University of Chicago Press.
Chung, Sandra. 1977. On the gradual nature of syntactic change. In Li
1977.3-55.
. 1978. Case marking and grammatical relations in Polynesian. Austin:
University of Texas Press.
Cole, Peter, Wayne Harbert, Gabriella Hermon, S.N. Sridhar. 1980. The
acquisition of subjecthood. Lg 56.719-43.
Disterheft, Dorothy. 1981. Remarks on the history of the Indo-European
infinitive. FoLH 2.3-34.
. 1982. Subject raising in Old Irish. In Ahlqvist 1982. 44-53.
...... 1984. Non-final verbs in Hittite. KZ 97.221-7.
220 DOROTHY DISTERHEFT
DORO
OLD ENGLISH ÞA, TEMPORAL CHAINS,
AND NARRATIVE STRUCTURE
1. Recent studies of text and discourse have disproved the classic view
that Old English pa is an innocuous particle, one we should not worry about
in translation for instance. Apart from dictionaries (such as Bosworth &
Toller 1898) whose aims called for analyses even of particles, and from com
prehensive studies of various aspects of OE syntax (such as Andrew 1940,
Bacquet 1962 and Mitchell 1985), a succession of recent papers has thrown
new light more specifically on various important functions of pa.
Thus in a paper born out of fifteen years' teaching of Old English and
printed in the Mustanoja Festschrift, Nils Erik Enkvist pointed out that pa
should not be dismissed as meaningless. His hypothesis was that pa had tex
tual functions in signalling action in narrative texts (Enkvist 1972). Three
years later Robert Foster added that pa marked narrative units and thus
signalled narrative structure (Foster 1975). Enkvist read another paper on
pa at the Societas Linguistica Europaea Congress in Athens in 1982, which
has now appeared in the Fisiak Festschrift (Enkvist 1986). Here he tried to
relate þa, not only to action but also to grounding, making use of the new
developments in grounding theory (e.g. Hopper 1979, Hopper & Thomp
son 1980, Longacre 1981). His conclusions emphasized the multifunctionality
of þa: it is an action marker, but as actions tend to be foregrounded it also
becomes a grounding marker. When pa occurs with stative verbs or with
verbs with a low transitivity index in the sense of Hopper and Thompson
(1980), it dramatizes or highlights narratively important but stative condi
tions. Simultaneously it also indicates sequencing of events on the main story
line and marks the division of the narrative into units. It may well have been
1 The starting-points, basic methods, and goals of this paper were suggested by N.E.E.
The texts were analysed by B.W., who also contributed many methodological improvements
and observations of her own. The text was written jointly by both co-authors.
222 NILS ERIK ENKVIST - BRITA WARVIK
2 We define event as a change from one state into another, and action as an event caus
ed by an actor, often wilfully and consciously in pursuit of a specific goal.
OLD ENGLISH þA 223
(for instance, narrative), and styles (for instance, from the simple oral to
the elaborate written). Methodologically, such perspectives also raise the
usual questions as to optimal relative roles of structure and process in the
descriptions of discourse and of language change.
The present paper - a partial report of wider researches in progress -
will, however, only add a few notes to the points mentioned above. It will
discuss þa as a marker of narrative units, of sequencing, and of a speech
like narrative style. This time we shall be especially interested in chains of
statements linked by pa, by other adverbials of time, and by conjunctions,
notably and; and in their contributions to temporal text strategies.
Hopper 1979a; Tomlin 1985 uses the term «pivotal information». Whereas
salience was manifested largely in the ordering of elements within clauses
and sentences, and whereas main-line sequentiality marked those high-
transitivity actions that make up the main story line, foregrounding can be
a property not only of main-line actions but also of other elements, such
as low-transitivity actions or events, and features of setting, which are im
portant for and relevant to the proper comprehension of the story.
Foregrounding can thus include both main-line sequentiality and other
elements (clauses) that the adopted strategy regards as necessary for the
discourse. The French structuralists' distinction between kernels and catalysts
(Chatman 1969) is most closely akin to foregrounding and backgrounding.
Another distinction is that salience, in the sense adopted here, is more in
herent and less readily manipulable, whereas grounding is more accessible
to the text-producer's strategic choices.
In making up her mind as to what falls within main-line sequentiality
and what is outside it, and what is foreground and what background in a
specific text, and investigator can proceed in two ways. Either she will read
the text and base her classification on intuitive, introspective judgments.
Or she will try to analyse the text by certain explicit criteria (such as those
presented by Hopper 1979a and 1979b, Hopper and Thompson 1980, or
Tomlin 1985) and base her classification on pragmatic, semantic and syn
tactic considerations. (Of course all such explicit criteria, too, ultimately
depend on somebody's linguistic intuitions: so does everything in linguistics.)
In any case there is now widespread agreement that grounding classifica
tions are scalar rather than binary (cf. Hopper and Thompson 1980).
By chain we indicate a succession of narrative propositions which are
linked to each other temporally. A chain can be presented in iconic order,
first ρ then q, or in non-iconic order, q took place after p. In other words,
the order can be the same, or different, in fabula and sjužet. A chain
may be unmarked; or it may be marked with special temporal expressions
and conjunctions. The temporal expressions form an open class which varies
all the way from simple conjunctions (while, when) and adverbs (first,
next) through phrases (after an hour) to clauses, sentences and even sentence
sequences. The time expressions may be absolute (on January 3, 1986,
at 3 p.m.), text-related (three hours later, then); or speech-act related (ten
minutes from now; next week). As is well known, shifts between text-related
and speech-act related temporals may indicate shifts in point of view, for
OLD ENGLISH þA 225
The simplest kind of story would consist directly of moves, without episodes
and subepisodes. As the first move of a subepisode, episode or story usual
ly explains the setting of the action, it is often backgrounded. We have avoid
ed the term paragraph because of its associations with the marking of text
divisions in writing and typography.
3. After these conceptual and terminological preliminaries we can
finally go on to the role of OE pa in narrative structure as revealed in tem
poral chains. The only way of illustrating text strategies, including temporal
OLD ENGLISH þA 227
dyde cuþberhtus (9) swa his gewuna wæs. (10) sang his gebedu
on sælicere yðe. (11) standende oð þ swyran. (12) and syððan
his cneowa on õam ceosle gebigde. (13) astrehtum handbredum
to heofenlicum rodore; (14) Efne õa comon twegen seolas of
sælicum grunde. (15) and hi mid heora flyse his fét drygdon. (16)
and mid heora blæde his leoma beðedon. (17) and siððan mid
gebeacne his bletsunge bǽdon. (18) licgende æt his foton on
fealwum ceosle; (19) pa cuðberhtus ða sælican nytenu on sund
asende, mid soõre bletsunge. (20) and on merigenlicere tide
mynster gesohte; (21) Wearð pa se munuc micclum afyrht. (22)
and ádlig on ærnemerigen. hine geeadmette to ðæs halgan
cneowum. (23) biddende (24) pæt he his adl eallunge afligde. (25)
and his fyrwitnysse fæderlice miltsode; (26) Se halga òa sona and-
wyrde. (27) Ic ðinum gedwylde dearnunge miltsige. gif õu õa
gesihõe mid swigan bedíglast. oõ paet min sawul heonon siõige.
of andwerdum life gelaõod to heofonan; (28) Cuõberhtus õa mid
gebede his sceaweres seocnysse gehælde. (29) and his fyrwites gylt
forgeaf; (Ælfric's Catholic Homilies X, p. 83)
Text D. (1) Apollonius pa sopfice hyre arehte ealle his gelymp (2) and æt
pare spræcan ende him feollon tearas of ðam eagum. (3) Mid py
pe se cyngc pæt geseah, (4) he bewænde hine õa to ðare dohtor
(5) and cwæð. (6) «Leofe dohtor, pu gesingodest; mid py pe pu
woldest witan his naman and his gelimp, pu hafast nu geedniwod
his ealde sàr. Ac ic bidde pe pæt pu gife him swa hwæt swa ðu
wille». (7) Da ða pæt mæden gehirde (8) pæt hire wæs alyfed fram
hire fæder (9) pæt heo ær hyre silf gedon wolde, (10) ða cwæð
heo to Apollonio: (11) «Apolloni, soðlice pu eart ure. Forlæt į>ine
murcnunge and nu ic mines fæder leafe habbe, ic gedo ðe
weligne». (12) Apollonius hire pæs pancode, (13) and se cyngc
blissode on his dohtor welwillendnesse (14) and hyre to cwæð:
(15) «Leofe dohtor, hat feccan į>ine hearpan and gecig öe to į>ine
OLD ENGLISH þA 231
frynd and afirsa fram þam iungan his sarnesse». (16) Da eode
heo ut (17) and het feccan hire hearpan, (18) and sona swa heo
hearpian ongan, (19) héo mid winsumum sange gemægnde pare
hearpan sweg. (20) Da ongunnon ealle fra men hi herían on hyre
swegcræft (21) and Apollonius ana swigode. (22) Da cwæð se cyn-
ingc: (23) «Apolloni, nu ðu dest yfele, forðam fre ealle men heriaõ
mine dohtor on hyre swegcræfte and fru ana hi swigende tælst».
(24) Apollonius cwæð: (25) «Eala ðu goda cynge, gif ðu me gelifst,
ic secge fræt ic ongite fræt soõlice frin dohtor gefeol on swegcræft,
heo næfð hine na wel geleornod. Ac hat me nu sillan fra hear
pan; ponne wast pu fræt pu nu git nast». (26) Arcestrates se -
ing cwæð: (27) «Apolloni, ic oncnawe soõlice fræt fru eart on
eallum fringum wel gelæred». (28) Da het se cyng sillan Apollonige
pa hearpan. (29) Apollonius fra ut eode (30) and hine scridde (31)
and sette ænne cynehelm uppon his heafod (32) and nam fra hear
pan on his hand (33) and in eode (34) and swa stod, (35) fræt se
cyngc and ealle fra ymbsittendan wendon (36) fræt he nære
Apollonius (37) fræt he wære Apollines õara hæðenra God.
(38) Da wearõ stilnes and swige geworden innon õare healle.
(39) And Apollonius his hearpenægl genam (40) and he pa
hearpestrengas mid cræfte astirian ongan (41) and frare hearpan
sweg mid winsumum sange gemægnde. (42) And se cyngc silf and
ealle fre frar andwearde wæron (43) micelre stæfne cliopodon (44)
and hine heredon. (45) Æfter pisum forlet Apollonius pa hear
pan (46) and plegode (47) and fela fægera fringa frar forð teah,
(48) fre fram folce ungecnawen wæs and ungewunelic, (49) and
heom eallum pearie licode ælc frara fringa (50) õe he forð teah.
(Apollonius of Tyre, p. 24-26).
Text E. (1) Da het se cyngc scipa gegearcian and him æfter faran, (2)
hit wæs lang (3) ær õam þe ða scipa gegearcode wæron, (4) and
Apollonius becom ær to Tharsum. (5) Da sume dæge eode he be
strande. (6) pa geseah hine sum his cuðra manna (7) se wæs
Hellanicus genemnod, (8) se pe ærest pider com. (9) pa eode he
to Apollonium (10) and cwæð: (11) «Wes gesund, hlaford
Apolloni». (12) Da forseah he Apollonius cyrlices mannes gretinge
æfter ricra manna gewunan. (13) Hellanicus hine eft sona gegrette
(14) and cwæð: (15)»... (Apollonius of Tyre, pp. 10-12).
4. In this paper, then, our aim has been to present a more holistic
approach to narrative structure to illustrate the use of pa as one of a set
234 NILS ERIK ENKVIST - BRITA WARVIK
APPENDIX
Text (translation)
(1) He said (2) that at one occasion he wanted to find out (3) how far that land extended north
wards, (4) or whether any man lived north of the wilderness. (5) Then he travelled northwards
along the coast; (6) keeping all the way the waste land on the starboard and the open sea on
the portside for three days. (7) Then he was as far north (8) as the whalehunters go furthest.
(9) Then he travelled still northwards (10) as far as he could sail in another three days. (11)
Then the land turned east, or the sea into the land, (12) he didn't know which, (13) but he
knew (14) that he there waited for a wind from the west and somewhat from the north (15)
and sailed then east along the coast (16) as far as he could sail in four days. (17) Then he had
to wait for a due north wind, (18) because that land turned there directly to south, or the sea
into the land, (19) he didn't know which. (20) Then he sailed from there southwards along
the coast (21) as far as he could sail in five days. (22) Then there was a large river reaching
up into the land. (23) Then they turned back from the river, (24) because they didn't dare to
sail along the river for fear of hostilities, (15) because the land was all inhabited on the other
side of the river. (26) He had not met any inhabited land (27) since he had left his home, (28)
but all the way on his starboard there had been the waste land except for fishermen and fowlers'
and hunters, (29) who were all Lapps, (30) and on the portside he had had the open sea. (Orosius,
p. 14).
OLD ENGLISH þA 235
Text (translation)
(1) The above mentioned saint was accustomed (2) to go at night to the sea. (3) and stand in
the salty water up to his neck. (4) singing his prayers; (5) Then one night another monk watch
ed his going. (6) and with stealthy tread slowly followed his footsteps. (7) until they both
came to the sea; (8) Then Cuthbert did (9) as he was accustomed to. (10) he sang his prayers
on the waves of the sea. (11) standing up to his neck in the water. (12) and after that bended
his knees on the sand. (13) the palms of his hands stretched out towards heaven; (14) Indeed,
then there came two seals from the sea-bottom. (15) and they dried his feet with their fur.
(16) and bathed his limbs with their breath. (17) and after that by beckoning asked for his
blessing. (18) lying at his feet on the yellow sand; (19) Then Cuthbert sent the blessed creatures
back to the sea. with a true blessing. (20) and in the morning went back to the monastery;
(21) Then the monk was very much frightened. (22) and being ill. early in the morning, he
descended to the saint's knees (23) praying (24) that he would, drive his illness entirely away.
(25) and like a father forgive his curiosity; (26) The saint then soon answered. (27) I forgive
your error secretly, if you conceal the sight by your silence, until my soul departs from the
present life called to heaven; (28) Cuthbert then healed his spy's sickness with prayers. (29)
and forgave the sin of his curious step. {Ælfric's Catholic Homilies, p. 83).
Text D (translation)
(1) Apollonius then truly told them all his fortunes (2) and at the end of the speech tears were
falling from his eyes. (3) As the king saw that, (4) he then turned to his daughter (5) and said:
(6) «Dear daughter, you did wrong; as you wanted to know his name and his fortunes, you
have now renewed his old wound. But I ask you to give him whatever you want». (7) When
the daughter heard (8) that her father gave her permission to do (9) what she herself had wanted
to do, (10) then she said to Apollonius: (11) «Apollonius, truly you are one of us. Leave your
grief and now that I have my father's permission, I will make you rich». (12) Apollonius thanked
her for that, (13) and the king rejoiced in his daughter's kindness (14) and said to her: (15)
«Dear daughter, order the servants to bring your harp and call your friends to you and take
away the young man's grief». (16) Then she went out (17) and ordered the servants to bring
her harp, (18) and as soon as she began to play, (19) she mingled the harp's sound with a plea
sant song. (20) Then all the men started to praise her art of playing music (21) and Apollonius
alone was silent. (22) Then the king said: (23) «Apollonius you behave badly, because all the
men praise my daughter's art of playing music and you alone blame her by your silence». (24)
Apollonius said: (25) «Oh you good king, if you allow me, I say that I know that your daughter
has truly attempted music, but she has not learnt well. But order now the servants to give me
the harp; then you will know what you don't know yet». (26) The king Arcestrates said: (27)
«Apollonius I know truly that you are well learned in all things». (28) Then the king gave orders
to give the harp to Apollonius. (29) Apollonius then went out (30) and dressed himself (31)
and put a garland on his head (32) and took the harp (33) and went in (34) and stood in such
a way (35) that the king and everybody sitting around him thought (36) that he was not Apollonius
(37) but that he was Apollo the heathen god. (38) Then it became silent and quiet in the hall.
(39) And Apollonius took his plectrum (40) and began to move the harp-strings with strength
(41) and mingled the harp's sound with a pleasant song. (42) And the king and all those that'
were present (43) called out in loud voice (44) and praised him. (45) After this Apollonius left
the harp (46) and entertained them (47) and presented many pleasant things (48) that were
236 NILS ERIK ENKVIST - BRITA WARVIK
unknown and unusual to that people, (49) and all of them were very pleased with each of the
things (50) that he presented. {Apollonius of Tyre, pp. 24-26).
Text E. (translation)
(1) Then the king gave orders to prepare the ships and go after him (2) but it took a long time
(3) before the ships were prepared, (4) and Apollonius came before to Tharsus. (5) Then one
day he went by the sea. (6) Then he was seen by one of his men (7) that was called Hellanicus
(8) that came there first. (9) Then he went to Apollonius (10) and said: (11) «Good health,
lord Apollonius». (12) Then Apollonius scorned the common man's greeting according to the
rich men's custom. (13) Hellanicus soon again greeted him (14) and said: (15)»... (Apollonius
of Tyre, pp. 10-12).
TEXTS
Bately, Janet (ed.). 1980. The Old English Orosius. (Early English Text Socie
ty, S.S. 6) Oxford etc.: Oxford University Press.
Godden, Malcolm (ed.). 1979. Ælfric's Catholic Homilies. (Early English
Text Society, S.S.5.) Oxford etc.: Oxford University Press.
Gooiden, Peter (ed.). 1958. The Old English Apollonius of Tyre. (Oxford
English Monographs). Oxford etc.: Oxford University Press.
REFERENCES
Andrew, S.O. 1940. Syntax and style in Old English. Cambridge: Cambridge
University Press. [Reissued New York: Russell & Russell 1966.]
Bacquet, Paul. 1962. La structure de la phrase verbale à l'époque Alfré-
dienne. (Publications de la Faculté des Lettres de l'Université de
Strasbourg, Fascicule 145.) Paris: Les Belles Lettres,
Bosworth, Joseph and T. Northcote Toller. 1898. An Anglo-Saxon dic
tionary. Oxford: Oxford University Press.
Chatman, Seymour. 1969. New ways of analyzing narrative structure, with
an example from Joyce's Dubliners. Language and Style 2.1.3-36.
Chvany, Catherine V. Forthcoming. Foregrounding, «transitivity», salien-
cy (in sequential and non-sequential prose). [To appear in Essays in
Poetics.]
Enkvist, Nils Erik. 1972. Old English Adverbial pa-an action marker?.
Neuphilologische Mitteilungen 73.90-93.
OLD ENGLISH þA 237
THOMAS FRASER
Université de Lille III
I would like briefly in this paper to examine the choice of the preposi
tion by to introduce the agent of the passive sentence in English. This ques
tion is generally only alluded to in grammars of Old and Middle English,
if indeed it is mentioned. Only Jespersen (1914), Mustanoja (1960), Visser
(1973) and, more recently, Faiss (1977) go into the matter in any detail, with
Mustanoja and Visser especially giving numerous examples of the various
prepositions used in this function. The only study devoted specifically to
the question is Green's 1914 article. Most authors agree that, though ex
amples of by are to be found in early Middle English, and even in Late Old
English (the Old English examples are disputed by Visser), it is only in the
early Modern English period that the preposition finally became establish
ed in its present-day function. I will insist in this paper, however, more par
ticularly on the Old English period, for it is here that English began its search
for the ideal sign to express the relationship between agent and passive verb.
Although other Germanic languages do offer a certain variety of
prepositions to express agency (Old High German fon, duruh, mit; Gothic
fram, af, us, paírh), only English presents such a wide diversity of signs
- a fact which has led Green (1914: 519) for example, to conclude that: «It
is a fascinating but intangible pursuit to observe how later Anglo-Saxon and
Middle English present simultaneously several prepositions of radically dif
ferent basic connotation to render this one idea of agency». It is to be hoped
that a semantic analysis of the passive construction and of the various
prepositions involved will produce at least some tangible results.
I would like to deal first of all with the question of foreign influence,
since those authors who attempt an explanation of the problem generally
point to Old French par, which was used alongside de to introduce the agent.
While it is probable that certain Modern English prepositional phrases are
240 THOMAS FRASER
calques of their French equivalents (e.g. «by hundreds and thousands»; the
now obsolete distributive phrase «by year» = French «par an»; possibly
also the use of by with verbs expressing «beginning» and «ending» - «to
begin by doing», etc.), it is more likely that French and English, along with
other Romance and Germanic Languages, developed their prepositions in
dependently as solutions to a specific problem, namely the expression of
a dynamic relationship after the breakdown of the respective Latin and Ger
manic case systems. A more probable foreign influence is that of Old Norse,
insofar as the socio-linguistic situation in the Danelaw after the Scandina
vian settlement was such that the O.N. preposition af may well have en
couraged the use of the related O.E. preposition of as a rival to from in
the expression of agency. It is also probable that in O.E. versions of Latin
texts the translators used certain O.E. prepositions systematically as
equivalents of the Latin means of introducing the agent. Thus, for exam
ple, in the West Saxon Gospels, the Vulgate use of the preposition a/ab is
translated by O.E. from, Latin per by O.E. þurh, and a Latin prepositionless
agent is rendered by O.E. of:
1) Da wæs se hæ lend gelæd fram gaste on westen þaet he wære
fram deofle costnod. Mt. IV, 1. (Lat. a Spiritu, a diabolo)
2) Wa þam men þurh þone þe byð mannes sunu belæ wed. Mt.
26,24. (Lat. per quem)
3) wæs pæt scyp of pam ypum totorfod. Mt. 14,24. (Lat. jac-
tabatur fluctibus).
My main contention, however, is that the English Language, in the Old
and early Middle periods, was experimenting diverse semiological means
for the expression of the relationship between the passive verb and the
agent 1 . Examples 4) to 11) illustrate the main O.E. prepositions to be
found in this function.
4) Hit gelamp... pæt pa halgan apostolas... wæron gemartyrode
æt pam manfullan nero. ALS 29,117.
5) He wearp peh swipor beswicen for Alexandres searewe ponne
for his gefeohte. Or. 68,14.
1 The period under discussion offers many instances of the same phenomenon of ex-'
perimentation - for example for the expression of genetival relationships or to introduce the
infinitive after the loss of the inflexional ending.
«BY» IN ENGLISH PASSIVE CONSTRUCTIONS 241
2 The sign is the outward manifestation of the word, the significate (French signifié) is
the underlying lexical or grammatical meaning.
242 THOMAS FRASER
As Guillaume (1973: 130) puts it: «[cela vient] de ce que la sémiologie, pour
être opérante, doit épouser, reproduire suffisamment le psychisme, et donc
tendre à répéter en elle, à un degré suffisant, la cohérence qui s'est instituée
là».
Before looking at the O.E. expression of agency, I would like therefore
to offer a succinct analysis of the linguistic components involved in the
underlying system of langue. This analysis is based on Moignet's study of
the French verb, itself a development of Guillaume's dynamic theory of
language.
Any verb expresses on operation involved in time, that is an operation
which brings about a result; in other words, the semantic content of any
verb moves from an initial position, which can be called a causation, to a
final position which Moignet calls an effection; this movement is an opera
tion in the course of which the activity is transformed, by the development
of time, into a resultative passivity. In a figure:
Fig. 1
The causation referred to here is not necessarily the true cause of the event,
but rather the mental position from which the event takes its source. The
initial causation may be either expressed within the verb itself, in a language
such as Latin, for example, in which case the causation is internal, or else,
as in the modern European languages, it is expressed externally through an
independent set of pronouns; in this case the verb is said to be incident to
its external support. Similarly, on the right hand side, the end result may
be either arrived at internally, in which case we have an intransitive verb 3 ,
or else the effection can be carried on to an object lying outside the verb's
own lexical meaning, in which case the verb's incidence is brought to bear
resultatively on the object, which is situated prospectively in relation to the
verb. This can be summed up in the following figure:
3 «John drank», for example, announces that the subject is in the initial position of an
activity which at any instant is made up of a result.
«BY» IN ENGLISH PASSIVE CONSTRUCTIONS 243
Fig. 2
(The left-pointing arrow indicates the verb's active incidence to the external
support, i.e. the subject; the right-pointing arrow indicates the verb's passive
incidence to the object)4 In the case of the passive verb, things are slight
ly different, for here the subject is seen as the support of the grammatical
operativity (of the verb «to be») of the resultativity of the main verb (in
dicated by the past participle). In other words, the passive incidence of the
verb is seen as retrospective and is referred back to the support constitued
by the subject.
Turning now to the preposition, it can be seen that it too is the sign
of an operation which may link two parts of speech: an initial support, which
is optional, but which, if present, is modified by what follows, and an end
support, which is nominal in nature and compulsory, and which can be called
the modifier. This can be illustrated by the following figure:
4 For example in a sentence such as «John drank water», John is seen, from instant to
instant in the course of the action as «drinking», and the water is seen as «drunk».
244 THOMAS FRASER
12) peos lar me wæs seald næs na for mannum ac purh God sylfne
Bl.H. 185,31.
13) ac heo wearö gescild purh pone cristnan casere Arcaduisan &
purh pæt cristene folc. Or. 56,9
14) pæt -pr cweartern ... wearõ onliht sona wundorlice purh
god. ALS 35, 250.
15) pa wearö se haiga Thomas gewissod eft purh god. ALS 36, 255.
16) And wolde wip Alisaunder speke,
Forte ben rz hym awreke
Of a prince, Kyng Alisaunder (Bennet & Smithers) 5-7
17) po was pe castel of Gloucestre and pe toun also
poru Sir Roger of Clifford in pe Kinges hond ido. R. of Gl.
( & S.) 79-80.
In the two Middle English examples in particular, the agent can be viewed
not so much as the perpetrator of the action but rather as the in
termediary 5 .
The main preposition of agency in O.E. was from, this role being
gradually taken over towards the 11th century by of. Both were suited to
the purpose in that their semantic content, which we may call an operation
of «origination», coincided with the retrospective operativity of the preposi
tion itself as a category. There is, however, a difference in the focus of the
two prepositions, for whereas from focuses on the starting point from which
the verbal operation initiates, of underlines rather the conduction of the
operation itself. This may help to explain why of tended to supersede from
towards the end of the O.E. period. If we look at the use of the two preposi
tions in Alfred's translation of Orosius we find that the author systematically
uses from in conjunction with the kinetic passive in which the auxiliary is
weorðan:
18) ... wearõ Romeburg getimbred from twam gebroðrum. Or.
39,1.
19) æfter pæm fleame Hasterbai wearõ of slagen from his agnum
folce. Or. 95,17.
20) & hie ðær æfter hrædlice tide from pæm londleodum purh
seare ofsleane wurdon. Or. 29,17.
On the other hand, of tends to be used in the resultative or static passive
in conjunction with the auxiliary beon/wesan:
21) for pon pe Tarente seo burg wæs getimbred of Læcedemo-
nium. Or. 84,2
although there are exceptions in which from is found with wesen as in:
22) Seo ilce Bizantium wæs ærest getimbrad from Pausania Or.
64,4.
It would seem then that when the passive verb itself underlined the opera
tion rather than the ensuing result, the preposition from, insisting on the
initial point of the originator, was sufficient, and the «activity» of the end
support of the verb could be reduced to the starting point of the verb's event.
On the other hand, the preposition of suppletes the lack of operativity to
be found in constructions where the past participle is associated with the
be auxiliary 6 .
It would be my contention, then, that the replacement of from by the
preposition of as the main preposition of agency in early Middle English
should be seen in conjunction with the gradual decline of weorõan and the
temporary loss of the distinction between kinetic and static passives. As early
as the Blickling Homilies, and in the 10th and 11th century entries for the
Chronicle, of comes into prominence in this function:
23) Æpelstan wæs of Myrcum gecoren to cinge. ASC 924 C,D.
24) her wæs Olaf cing of slagen ... of his agenum folce. ASC 1030
25) Weorpian we eac pa clapas his hades, of pæm wæs ure ge-
cynd geedneowod. BI.H. 11,9
26) hwilc abbot pe pær coren of pe munecan pæt he beo geblet-
sad of pan ærcebiscop of Cantwarbyrig. ASC 675 E
From, on the other hand, with its insistence on the initial point of the ac-
6 1 have only come across one example in Orosius of the preposition of being used in
conjunction with the auxiliary weorõan:
Hu II æpelingas wurdon afliemed of S c i l i u m . Or. 1,25
This example is to be found in the table of contents of Alfred's translation, and it may possibly'
be explained by an attempt to gain greater expressivity in announcing an event to be recounted:
the event is underlined as dynamic by the auxiliary, the preposition and also the preverb a.
«BY» IN ENGLISH PASSIVE CONSTRUCTIONS 247
REFERENCES
8 The first recorded written example we have of the passive progressive dates back to
the end of the 18th century:
... a fellow whose uttermost upper grinder is being torn out by the roots by a mutton-fisted
barber.
Robert Southey, Life and Correspondence (1795) (quoted Visser)
It is probable that the form existed in the spoken language several decades earlier.
«BY» IN ENGLISH PASSIVE CONSTRUCTIONS 249
HERBERT GALTON
The University of Kansas
It should be obvious that I cannot deal with the entire problem of the
I.-E. perfect tense here, but only in its relation to its counterpart in Slavic.
There is, as is generally assumed, a single inherited perfect form in Com
mon Slavic: wädä, whose meaning is quite clearly that of a present «I know»,
and within Slavic there is no evidence to show that this was in any way felt
to denote the result of an antecedent event such as «I have seen (and therefore
I know)», which many have considered the proper function of the I.-E.
perfect to express. However, as has been very aptly observed by Warren
Cowgill (1974:563), every state - such as, for instance, the one with the
description of which we are dealing in the Slavic form - presupposes some
preceding process except for primeval chaos, and in such a broad sense, I
submit, resultativity is not a linguistic matter, but an ontological one.
The question has arisen as to what was the meaning of the category
Perfect in I.-E. Szemerényi (1980:274) opts for the original designation of
a state in the present (Zustand in der Gegenwart, his italics) resulting from
a past event, while subsequent developments could bear out the meaning
of a result characterizing the object of the verb all the way from the past
event affecting it right into the present.
This emphasis on the present can trace its pedigree all the way back
to Brugmann (1916: 768) and Delbrück (1897:3). Hirt (1928:270) likewise
stresses the original presential constituent in the meaning of the perfect,
possibly even going beyond his predecessors, to the point of seeing in the
(apparently facultative) reduplication a palpable expression of the intensity
of the action. We are not surprised at finding no trace of any such mor
phological means in the sole Slavic form attested; a derivation of the mean
ing «I have seen, I have seen -> I know» seems to us rather far-fetched.
We tend to agree here rather with Brugmann, who did not postulate the
252 HERBERT GALTON
noun, whose feminine gender was lost). In fact, the nominal origin of the
perfect had already been adumbrated by Szemerényi (1970: 306).
At a time when the I.-E. perfect led a much more untroubled existence
in comparative linguistics, the ending of wädä was universally attributed
to the medio-passive, I.-E. *-ai represented by Latin -i, Sanskrit -e (e.g.
tutudī, tutude, cf. Vaillant 1966: 76, Stang 1942: 22). However, since then
doubt has been cast on the existence of a medio-passive voice for the I.-E.
perfect, since this expressed a state of the subject and would not seem to
require the formation of such a voice. The Systemzwang now appears to
us to have had its seat in the mind of the explorer rather than in the in
vestigated subject matter; where an I.-E. medium voice is reliably continued
in Slavic, this is done by means of the reflexive pronoun s as e.g. in bojitb
s «he is afraid» to Skt. Pf. bibhāya, Pres, bháyate, where such a secon
dary formal development probably occurred in pre-Slavic, cf. also Lith. bi-
jóti(s). It is in fact more likely that the *ë (which appears in Slavic as j a f
= / ä / ) continues an I.-E. suffix likewise indicating a state, even if one does
not go all the way with Perel'muter placing the perfect tense and suffixes
with a long vowel on the same footing, both of them conveying an identical
stative verbal category (1977:201).
It would, accordingly, be altogether incorrect to say that the I.-E. perfect
was replaced by a more explicit formation consisting analytically of two parts
in Slavic, because the two do not mean the same thing at all; the I.-E. perfect
denoted a psychic or physical state of the subject, whereas the Slavic perfect
has a rather specific meaning, as we shall see. The statai meaning of the
I.-E. perfect was continued by a special verbal formation characterized
precisely by a stem in -ä'- < *-ē- such as we find in its pure form in wädä,
which on account of its entirely aberrant nature was replaced by wäirib still
in the history of O.C.S., with an athematic ending -, but surviving to
this day say in the Russian or Slovak particle ved' «after all». This Slavic
(or Balto-Slavic) development (replacement of Pf.), then, is not very specific,
but the solution of the problem of such a presential perfect by creating an
entire verbal class capable of functioning at any temporal level is very much
in accordance with the overall enrichment of the I.-E. verbal system in Slavic.
The stem of the present tense in -i- (e.g. bbditb «he wakes», έγρήγορε,
inf. bbdäti) is in fact a hang-over from the old perfect, as has been
demonstrated by Kurytowicz (1964:79-84) and does not contradict the statai
meaning, as Aitzetmüller thought (1962:244). Of course, -i- is not specializ
ed as a statai index and the entire connection is purely historical. An ex-
254 HERBERT GALTON
planation that -ä'- is limited to the infinitive and preterite stem because of
its derivation from a perfect tense marker overrates the preterite value of
the latter, in view of recent (and not so recent) findings on the meaning of
the Í.-E. perfect. The protolanguage apparently did have a class of statai
verbs, even if it is impossible to ascribe to it a full gammut of conjugational
forms at several temporal levels. These levels only sorted themselves out in
Slavic and some other descendent languages, in their separate development.
That the Pf. was originally formed only of non-durative verbs, allegedly
because it refers specifically to the result of a previous event which must
be couched in a non-durative, i.e. punctual verb (representing the event as
a point on the time axis, followed by the result as the next point) is entirely
unproven, cf. the lists of Homeric Pf. tense forms given by Perel'muter and
others, like «I smell, am scented», «I seem» etc. A Slavic verb
like goritb «it burns». inf. goräti (Leskien's class IV B), continues in both
forms the I.-E. perfect, of which it still gives a typical semantic example:
«something is in a state of burning». The circular route assumed by Aitzet-
müller (1963:213), via a preterit in -ě- (= -ä'-) is perhaps not necessary, if
we do not attribute a preterital meaning to the I.-E. Pf.; there was thus no
«gap to be filled in», but an entirely creative approach on the part of Slavic.
This language needed a perfect tense within the framework of its rich
temporal system, to convey a very specific meaning: an event which took
place in the past, but is still relevant at the moment of utterance. It is really
this which matters, and not whether the event has left a palpable result
behind; in fact, the relevance may be merely contrastive. By the same token,
the temporal location of the past event itself recedes into the background,
i.e. it does not really matter between which points on the time axis it was
located. It did not just arise and was over with the next event, but stretched
in its consequences beyond this location all the way to the present - which
is something altogether different from a mere present state.
It can in fact be claimed that the beginning of the new form was not
markedly preterital; it very much stressed the present relevance and only
turned into a full-fledged preterite afterwards. This is in marked contrast
to the new perfect formations of some of the other descendent languages,
probably even clearer than in the Modern Greek perfect which Seiler (1952:
80,148) claims to have encountered only in «situation actualisée», i.e. link
ing it to the speech situation, for both types contain
from the outset a past element; in «situation non-actualisée», the aorist
substitutes for it - as in those Slavic languages which distinguish the two
FROM I.E. PERFECT TO SLAVIC PRETERITE 255
tenses. To render past events per se, Slavic had an aorist and an imperfect,
and the extraordinary care bestowed upon the last formation with its long
suffix to distinguish it from the former (and on whose explanation there
is still nothing like a consensus) sufficiently shows how concerned the Slavs
were with a precise representation of past events, for in addition, Common
Slavic, as reflected in O.C.S., also possessed a morphological future in the
past as well as a pluperfect, about which nothing more shall be said here.
The meaning of the perfect tense was quite specific and cannot be proved
to carry on any Indo-European category, for to do that we would have to
know much more about the state of the I.-E. verb in the very period when
Balto-Slavic cut its ties with the proto-language, possibly coming under some
foreign influence.
In the light of our definition of the Common Slavic perfect - referring
to an event in the past without precise location there, but remaining rele
vant at the moment of utterance - very little need be said about the aux
iliary part of the new formation, i.e. jesmb e.g. in jesmb kupovalb «I have
bought» (imperfective aspect, pv. il), for it is precisely this element
which contains, in addition to the index of the grammatical person, the signal
of the present relevance: «I am one who has done such and such». The shift
away from what many believe nowadays to have been the original meaning
of the I.-E. perfect - with due allowance for the reservation just made - is
considerable, also the new analytic compound was surely much clearer and
more transparent to the speakers. To capture its significance we must
dissociate it from the one which the subsequent preterite forms have acquired
in the better known Slavic languages, above all Russian, where precisely the
7-form is all that is left of it and has acquired a purely preterital (sometimes
resultative) meaning by itself, yet that is what the i-form at first does not
seem to have had at all. It could even be maintained that the inherited par
ticipial form in -us- was replaced in Slavic verb phrases almost entirely
(Vaillant 1966:85; as distinct from Baltic!) because of the strong past associa
tion of the former.
No suggestion has been made that the i-form which constitutes the se
cond part of the analytic perfect implied by itself any reference to the past;
it originated apparently as an adjective derived from a verb (Vaillant 1966:83
«proche encore de l'origine adjectivale»; cf. Ivanov 1981: 222; Kurytowicz
1970-72:323 for some reason or other proceeds from an original medio
passive formation). That the i-form originally had a specific resultative mean
ing as has been assumed by Belić (1935:31) will be hard to prove; Greek
256 HERBERT GALTON
perfect would altogether upset the clear dialectic opposition between the two
aspects - tertium non datur. That the most natural way of viewing the pre
sent is that of a relatively lasting state and that of a past event as something
setting in and then being over, possibly against the background of a conti
nuing state, are ontological and not grammatical matters; cf. also the original
meaning of αόριστος: «unmarked».
Grammatically aspectual is the division of preferably every verbal for
mation along the lines of this dichotomy. The Anatolian group, therefore,
never lost an aspect, because I.-E. did not have it (similarly already van Wijk
1929:240 ff.). I regret that I cannot fall in with the widespread belief that
the I.-E. present and aorist verbal stems mark an aspectual rather than tem
poral opposition, and that Slavic has simply best preserved an I.-E. state
of affairs (Shimomiya 1974:219) - its rôle was very much more creative than
that. It has been suggested that Meillet was influenced by Durkheim's
sociological school in postulating a relatively late origin of temporal as against
aspectual categories (Serbat 1976:208).
It will have been noticed, of course, that the notorious «completion
of action» criterion does not figure at all in my definition of the pv. aspect;
in the view received more generally among Slavicists nowadays, it has been
replaced by «complexivity», which in my scheme is merely a corollary of
an event's being placed between the preceding and the following point on
the time axis, which is what really matters. The actual completion of an ac
tion would not seem to call for a special morphological category, being the
normal course of events, also it is very easy to find in Slavic contrary ex
amples of actions which surely were completed (like «who painted this pic
ture?» or «shelled walnuts») being rendered in the ipv. aspect; for illustra
tions I beg to refer to my book and other articles.
But whatever view one may take of the semantic content of the aspec
tual opposition typified in Slavic, its existence beyond the purely noetic
sphere, as a formalized and grammatical opposition in I.-E. has also been
questioned by van Wijk (1929: 240 ff.), Stang (1943:12) and Adrados (1963:
224, 273, 327, 825). For instance, unlike Slavic, the addition of prefixes still
does not perfectivize a Greek verb (unlike Ossetic), a fact which surely in
one way or another continues an I.-E. tradition. This can at the same time
also serve as an example of how Slavic utilized I.-E. procedures creatively
to set up a morphological category of verbal aspect, it did not just continue
an L-Ε. state of affairs, only with greater consistency! From the beginning
of its written tradition we have a wealth of grammatical tenses denoting
FROM I.E. PERFECT TO SLAVIC PRETERITE 259
sition from the I.-E. perfect to the Slavic verbs of state is still visible in the
residuary form wädä «I know».
Instead, Slavic developed a new perfect tense unparalleled in Baltic,
to denote an event which took place at some unspecified period in the past,
but has remained in one way or another relevant at the moment of utterance.
This reference is expressed by the present tense of the verb «to be», fixing
the single reference point of this tense, whereas the i-form, serving as the
other part ot this analytic tense, apparently did not even involve the past
to begin with, but rather indicated an agent or at least subject. The mean
ing of the whole collocation, however, gradually shifted in the direction of
a past, in whose domain the event still deemed to be relevant was located.
This new Pf. came to stand alongside the simple past tense forms of
Slavic, aorist and imperfect, denoting respectively more or less point-like
or in any case localized events in the past as against continuing stretches
on the time axis. Where the Slavic conjugational wealth (including the com
plications of the verbal aspect) has at least been preserved, as in the Slavic
South-East, the perfect has nearly kept its original composition and func
tion. Where this is not so much the case, the present tense forms of «to
be» at least still function as auxiliaries; this is the position of Serbo-Croatian.
In the North (including Slovene), the scope of Pf. has swallowed up
the aorist and imperfect and it has thereby lost its specific semantic func
tion, becoming to all intents and purposes the sole past tense form of the
verb. The ancient auxiliary has therefore become a mere morpheme indicating
the person, and present relevance, if any, has to be marked in other ways.
This morphematic character is clearer as we move from Czech to Polish,
and in Russian the process has moved full circle, the ancient «participle»
in 1 having become the one and only expression of past tense without even
a morpheme to mark the person.
REFERENCES
JOHN HARRIS
University College London
naculars are similar in that they all sprang up during the seventeenth cen-
tury as a result of British colonial and trading activity on both sides of the
Atlantic. As to matters of internal history, the main investigative task is
to disentangle the various strands that are interwoven in the origins and
development of contact varieties of this type. In particular, it is necessary
to ascertain the relative contributions of the following important com-
ponents: degree of substratum input (west African languages in the case of
the Caribbean creóles, Irish Gaelic in the case of Hiberno-English), the poten-
tial intervention of linguistic universals, the superimposition of internally-
generated changes, and the possible retention of archaic metropolitan
features. The view that the phonological systems of these contact vernaculars
are inherited directly from substratum sources rather than from metropolitan
English is reflected in statements such as the following:
back [a] (e.g. Jespersen 1909); others that it was open front [a] (e.g. Dob-
son 1968). (Throughout this paper, phonetic symbols have their IPA values).
Lass (1976: ch 4), besides providing a summary and discussion of the rele
vant literature, adduces comparative evidence from present-day English
dialects in support of the second view. The object of this section is to show
how this position can be further strengthened and refined by incorporating
additional comparative evidence from Caribbean creoles and Hiberno-
English. I will focus on three illustrative varieties: Jamaican Creole (JC),
Hiberno-English (HE, here represented by Mid Ulster English) and, as a
familiar point of comparison, British «Received Pronunciation» (RP). (For
material on JC I've drawn on Cassidy & Le Page 1967 and Wells 1982.
A summary of HE phonology can be found in Harris 1985b).
In all three varieties, Middle English (ME) / a / shows evidence of a
split (phonemic in JC and RP, allophonic in HE) into short and long reflexes.
Ignoring for the moment certain special combinative developments, we may
sketch the outlines of this pattern as follows. The short nucleus appears
minimally before voiceless noncontinuants: RP [æ], JC [ä], HE [ä] in fat,
tap, etc. The long one appears minimally before voiceless fricatives: RP
[a:], JC [ä:], HE [a:] in pass, path, etc.
The evidence I wish to concentrate on here has to do with realisations
of ME / a / in the environment of historical velars. In this context, RP shows
the main developments just outlined. JC and HE, however, display the ef
fects of two special developments, both of which involve the addition of
some element of palatality to the syllable. These can be summarised as (i)
palatalisation of historical velars and (ii) front-raising of ME / a / . Evidence
suggests that both of these had occurred in the metropolitan language some
time before the spread of English to Ireland and the Caribbean. The first
development is unambiguously reported by seventeenth-century orthoepists.
The second, however, has not generally been recognised by authorities on
the history of English, in spite of the fact that there is evidence of it in
Early Modern rhymes and spellings. Comparative material from present-
day contact vernaculars suggests how this evidence might be better exploited
than has hitherto been possible.
points out, the possibility that the vowel's realisation was subject to variabili
ty, whether phonologically or sociolinguistically constrained, tends not to
be acknowledged, except to the limited extent that this is necessary to ac
count for conditioned sound changes which have left their mark on stan
dard dialects (rounding after / w / (as in swan) in the ancestor dialect of RP
being a case in point). The evidence relating to palatalisation in both
metropolitan and contact varieties does indeed support the contention of
Dobson (1968), Lass (1976) and others that, at the beginning of the Early
Modern period, ME / a / was some kind of front vowel. However, com
parative evidence based on studies of sound changes currently in progress,
particularly in northern United States cities (Labov et al 1972) and Belfast
(Milroy 1984, Harris 1985b: 177 ff), indicates that this is probably only part
of the story. There are good grounds for assuming that some of these changes
are extensions of patterns of variation which were already evident in Early
Modern English.
One such change, the results of which survive in a number of contact
Englishes but which has since been aborted in standard dialects, is the con
ditioned raising to mid position of ME / a / , particularly in the environment
of nonanterior consonants. In the light of the preceding discussion, it seems
clear that many types of Early Modern English had palatal allophones of
/k, g, ŋ/ in the environment of historically front vowels. We may next con
sider to what extent these consonants themselves exerted an influence on
the quality of neighbouring vowels. Given what we know about the
naturalness of certain sound changes, we might expect the high front tongue
position of palatals to have a front-raising effect on adjacent reflexes of
ME / a / . In fact, there's evidence to suggest that this is exactly what hap
pened. The front-preserving property of historical / k , g , ŋ / is illustrated by
the fact that, in postvocalic position, they prevented the otherwise general
backing (and subsequent rounding) of ME / a / after / w / , a change which
has left its mark on southern English and derivative dialects. Thus RP has
back < ME / a / in, for example, swan, swab, what but front / æ / in,
for example, quack, wag, twang. The nonlow feature of palatals was ap
parently responsible for encouraging, more so than other consonants, the
raising of neighbouring ME / a / . This influence is not often acknowledged
by authorities on the history of English (Dobson only mentions it in pass
ing (1968: 551)), although it seems to be well supported by spelling evidence
from the Early Modern period. The evidence takes the form of e-spellings
for ME / a / , which scholars have attempted to explain piecemeal in a number
COMPARATIVE RECONSTRUCTION 277
4. Conclusion.
The decision to perform comparative reconstruction on a contact ver
nacular and its lexical source language is independent of the position we
adopt with respect to the question of whether their relationship is ap
propriately described in terms of the traditional family-tree model. The lex
ical relationship alone is enough to supply the raw material necessary for
applying the comparative method, namely systematic sound correspondences
between the two varieties. On the other hand, our response to the genetic
affiliation question does have a bearing on the way we interpret the results
of the comparative endeavour, and in particular on our interpretation of
the changes which produced systematic divergence between the varieties in
the first place. That is, do we attribute individual points of divergence to
the operation of internal sound change or contact-induced restructuring?
If we take the view that the lexicon of a contact vernacular represents, not
inherited stock, but the outcome of wholesale borrowing from the superstrate,
then it should be clear that drawing the two varieties into the same comparative
exercise amounts to applying, on a lexicon-wide scale, principles usually
associated with loan-word phonology. Whether our comparison is restricted
to a small set of lexical borrowings or encompasses the entire lexicon, the
value of the enterprise is the same. By comparing forms in the donor language
on the one hand and the recipient or emergent language on the other, we
derive correspondences which shed light on structural details of the two
languages at the time of borrowing. In this way, we can learn a good deal
not only about the system(s) which conditioned the integration of loan-forms
into the recipient or emergent language but also, as the comparison under
taken here demonstrates, about the history of the donor language itself.
REFERENCES
Taylor, Douglas. 1956. Language contacts in the West Indies. Word 12.
391-414.
Thomason, Sarah G. & Thomas Kaufman. 1975. Toward an adequate defini
tion of creolization. Paper read at the International Conference on Pidgins
and Creoles, Hawaii.
Weinreich, Uriel. 1966. Languages in contact. The Hague: Mouton.
Wells, J.C. 1982. Accents of English. 3 vols. Cambridge: Cambridge Univer
sity Press.
Wyld, H.C. 1920. A history of modern colloquial English. Oxford:
Blackwell.
A(í)εí AND THE PREHISTORY OF GREEK NOUN ACCENTUATION
(RÉSUMÉ)
H.M. HOENIGSWALD
University of Pennsylvania, U.S.A.
The accent of αίεί is compatible neither with the idea that it is an ar
chaic (like Skt. dyávi) nor that it is a Greek columnar (like ηθει) locative
of the 5-stem: the former should be
*Αιρεσί > αίει could, however, represent the late IE stage, with oblique-case
oxytonesis extended to the locative, but previous to the limitation of that
accentuation to monosyllables in Greek. Klingenschmitt (MSS 33.78), to be
sure, suspects a true dative after the Indo-Iranian manner from the «thème
II» (*H2yew-; cp. Av. yavaē), not extant in the various Greek words for
«age», loc. «always» though perhaps surviving in
the completely isolated ου «not» (cp. French jamais, etc.; Cowgill Lg
36.347-50, with a reference to F.B.J. Kuiper). If this is so, the fact that
*H2oyu goes to rather than tends to confirm the impression that
intervocalic -y- behaves differently from -s- in this respect, as, perhaps, it
does in others. (Lejeune, Phon. hist. du mycénien, etc., 169).
[The complete version of this paper will appear in «Studies in Memory of
Warren Cowgill, 1929-1985», Berlin / New York: De Gruyter].
THE INSTABILITY OF PERIPHERAL
/ . / , / ø . / , AND / o . / IN DUTCH LECTS
COR HOPPENBROUWERS
Rijksuniversiteit Groningen
(2)
The front spread / . / e.g. may become /ei/ and by ongoing diphthongi-
zation /ei/ and even / æ i / . Following the path of vowel laxing, the / . /
may become / I : / and / ε : / by further lowering. for details about these shifts
see Andersen's (1972) study of diphthongization and Hoppenbrouwers
(1982).
1
Since the feature TENSE has often been subject to criticism, I prefer to use the term
PERIPHERAL to subdivide the primary class of long vowels. (See Ladefoged 1975: 245, and
Hoppenbrouwers 1982).
286 COR HOPPENBROUWERS
2 The data that have been used to draw area I are taken from Sassen (1953) and
(1967).
288 COR HOPPENBROUWERS
3 I am here leaving aside the question whether such loans occurring in dialects are taken
over directly from the original language or, more likely, from the (national) standard.
4 Notice that for standard Dutch this is the only context in which truly long vowels have
survived, and this only at the phonetic level.
5 Standard Dutch, which incidentally never has diphthongs in the context of following
/ r / , uses the form /-I:rə/ in all the cases.
290 COR HOPPENBROUWERS
II
The second example is taken from south-eastern Dutch dialects. In this
case the vowels in question are variants of West Germanic î, û, and cor
responding loans. Outside the dialects that have preserved those high vowels
(the long û eventually becoming /y:/) we find in area II diphthongs and
lower monophthongs. Both the diphthongal and monophthongal paths in
dicated in (2) are reflected here in certain dialect continua.
In this area the distribution of the variants is much more complex than
in the Groningen instance, and therefore the map is only concerned with
one of the vowels in question in only one of the phonological contexts. The
shaded part of area II includes the dialects with diphthongal variants of Wgm
î in the context of following / s / , the case characterized by the largest
geographical extension of the diphthongization.
Most of the dialects in this area use diphthongs with the same degree
of wideness as those of the corresponding standard Dutch forms. (8) lists
examples that in some of the dialects have the wide diphthong / æ i / , both
in this and other contexts.
(8)
To the east of, and even within the shaded area, there are once again some
dialects, marked with dots, that use the lowered peripheral / . / as in the
first example. The two dialect communities inside of II that still use the
peripheral vowel, form a dialect island surrounded by dialects with
diphthongs. To the east there are also four dialects with the intermediate
narrow /ei/. But just as in the Groningen case, the use of the unstable vowel
is restricted here to a small number of dialects.
To the west of II there is a large area using open monophthongs. In the
case shown on the map we only find / ε : / . In other contexts the higher / I : /
also occurs. (9)a and b presents instances in the context of following /va/.
(9) a [bh:v3] «stay», [sxrI:və] «write», [vrI:və] «rub»
b ε: ε: ε:
Such examples, which occur in the geographic continua starting with / i : / ,
reflect the monophthongal path in (2), which is the alternative way to avoid
peripheral /./, / ø . / , and / . / .
Broadly speaking, the variants of Wgm û, ui and corresponding loans
reflect a comparable development, thus underlining the phoneticity, the in
ternal linguistic character, of those processes. For more details see Hop
penbrouwers (1982).
INSTABILITY OF PERIPHERAL VOWELS 291
III
IV
The last example shows the unstable character of the peripheral /./,
/ ø . / , and / . / in a regiolect. It is spoken in the eastern part of the
Netherlands, in area IV on the map.
The use of the traditional dialect as an independent system has become
more and more restricted to the group of nonmobile older rurais. It will
be clear, however, that in the gradual shift towards the standard language,
in which the descendants of those NORs are involved, many features recall
the former dialects. Some of the supracommunal regiolect features prove
to be particularly resistent to the process of standardization.
Rule (12) is a good example of such a regiolect feature that may be traced
back to the dialect of the ancestors.
(12) STANDARD DUTCH REGIOLECT
/__r all contexts
/./ → [I:]
b /ø./ → [Λ:]
/./ → [U:]
Whereas in standard Dutch the effect of this phonetic rule is restricted to
the position of following / r / , in the eastern regiolect it applies in every con
text. To simplify matters, I will limit the discussion to certain aspects of
subrule (12)a, which in every respect is collapsible with (12)b, and As
counterparts of standard Dutch /./, the dialects at the beginning of the
regiolect continuum use numerous variants. Thus the instances in (13), which
are taken from Bezoen (1938), belong to the dialect of Enschede, situated
in the eastern part of area IV.
According to Bezoen (1938), the items in (13)a are pronounced with a fairly
monophthongal [e:], i.e. differently from standard Dutch where this
phoneme has a slightly diphthongal pronunciation. The items in (13)b form
part of a large group, and are pronounced with the phone that is generated
by rule (12)a. The variants in c, d, and e may be passed over as it will be
evident that the frequently used dialect forms in a, and b constitute the basis
of regiolect rule (12).
Speakers hailing from this region who try to speak the standard language
are generally not aware of this deviation from the norm: thus the present
Minister for Social Affairs, who is a native of the eastern region, may be
heard to say in debates or interviews broadcast by radio or television, that
[xI:n lu:nsfərhu:vIŋ] «no wage increase» can be allowed by the government.
Even in such official statements he uses nonperipheral (lax) vowels instead
of the peripheral standard forms.
Another well-known speaker, who also hails from this region, is a
popular weatherman. In a special program for Dutch tourists, which can
be received in camp-sites all over Europe, he regularly forecasts [dI:prεsis]
«depressions», accompanied with [fI:l rI:vεn] «much rain» or sometimes
areas of [hU:və drΛk] «high pressure» 7 .
The instances taken from different Dutch lects have shown that the in
vestigation into sound change may successfully be directed at the develop
ment of peripheral / . / , / . / and / ø . / They are a group of vowels that,
often depending on the context formed by the following consonant, tend
to shift, taking one of the two available paths. In combination with
geographic and socio-economic factors, those shifts may contribute to our
understanding of language change.
7 Such persistent regiolectal variables, which often are used below the level of social
awareness (see Labov, 1972), may easily become part of a new substandard or sociolect as
indicated in (3).
294 COR HOPPENBROUWERS
REFERENCES.
1. Introduction.
1.1 The paradox has often been noted that scientific accounts of
language change are rooted essentially in the antithesis of change, namely
synchronic description. Synchronic description provides us with points of
orientation comparable to a navigator's compass, linking different stages
of a language and allowing us to chart the course of change. Yet the pro
cedures which we follow, ironically, do not lead us to an understanding of
change but only of results and, at best, inductive generalizations from those
results. Change is thus reduced to the subtractive difference between the
grammars of fixed stages in the historical existence of a language. The step
wise, discontinuous conception of change implicit in the procedures of com
parison is at variance with our human apperception of change as something
continuous, and gives us only a secondary, indirect grasp of change as a
process. Dilthey understood this clearly in his sardonic characterization of
Platonic realism:
«A true Plato, who first of all fixes in concepts things which become
and flow, and then supplements the fixed concept with a concept of flow
ing». (Gesammelte Schriften V: 112. Stuttgart: Teubner, 1958).
Our procedure of comparison of fixed stages forces us to view change
as a residual and separate component grafted onto these stages and sensed
as something different from them. We would like to emphasize that we are
not saying that linguists have misunderstood change; what we are instead
saying is that Dilthey's observation is an apt characterization of the
philosophy underlying our procedures, and that in order to translate the
results of these procedures into a more natural continuous apperception of
change we are obliged to depart from the procedures themselves.
Generalizing upon these observations, we would suggest that what is
ultimately being recognized is the intrinsic limitations of structuralism when
296 PAUL HOPPER / JANICE MARTIN
1.2 Referentiality.
The notion of referentiality as it is understood by logicians is based
firmly in word or sentence level semantics. This restriction has resulted in
a great deal of attention being paid to mysteries such as:
(1) Alice wants to marry a millionaire,
with its well-known ambiguity. Does Alice have a particular millionaire in
mind or will any millionaire do?
A recent line of thought suggests that by starting with discourse rather
than with word or sentence level semantics we can by-pass this,question,
and turn our attention instead to the role which the NP in question plays
in the discourse. For example, is the indefinite NP in question subsequently
referred to as an individual, or never referred to again? If it is referred to
again, how many times does this occur? What is the typical distance bet
ween its first introduction and its next mention? This pragmatic approach
to referentiality, introduced through the pioneering work of Givón 1 (see
also Levy (1982)2 and Givón (1983)3, has now been investigated in a
number of languages (see Givón, ed., 1983; Hopper 1985 [to appear] 4 ), but
rather little work has been done on its diachronic implications. This paper
is an investigation of the history of the Modern English indefinite article
() from the point of view of its discourse referentiality. Its goal is to trace
quantitatively the development of the general indefinite article of Modern
English out of an Old English quantifier having a much more limited
discourse function.
1 «Logic vs. Pragmatics, with Human Language as the Referee». Journal of Pragmatics,
6.2:1981.
2 Levy, Elena, 1982. «Towards an Objective Definition of Discourse Topic». CLS 1982.
3 Givón, T., ed., 1983. Topic Continuity in Discourse. Amsterdam: Benjamins.
4 Hopper, Paul J. 1985. «A Discourse Function of Noun Classifiers in Malay». In C.
Craig, ed., Noun Classification. Amsterdam: Benjamins.
STRUCTURALISM AND DIACHRONY 297
5 Traugott, E.C. 1982. in Lehmann & Malkiel, eds, Perspectives on Historical Linguistics.
Amsterdam: Benjamins, 245-271.
6 Givón, T., 1981. «On the development of the numeral " o n e " as an indefinite marker»
Folia Linguistica Historica 1,2.
298 PAUL HOPPER / JANICE MARTIN
the hypothesis that when an edged out sum on the one side and zero on
the other, it did so at first mainly by encroaching on sum, with a relatively
slow encroachment on the domain of zero. Consequently the average referen-
tiality of an increased during this period. At the beginning of the modern
period, an moved increasingly into the domain of zero; this movement is
manifested as a decline in its average referentiality.
Our study suggests, then, that the English Indefinite Article began as
a referential presentative marker in competition and functional overlap with
sum. Through the centuries, Old English an passed through stages of decreas
ing referentiality and weakening of the presentative function.
4. Conclusions.
We conclude by pointing out that some general forces are at work here.
As forms widen their functional domain, the more specialized functions
which they once served are no longer catered to. This is true of the English
indefinite article, which is no longer exclusively presentative; the presentative
function is executed by a variety of different strategies, as exemplified in
the following, for example:
(2) At the door stood a /one/ a certain/this Mr. Smith.
The use of «this» as a presentative has been discussed by Wald
(1983)7; its appearance seems to be a relatively recent phenomenon, and
cannot be shown to be earlier than the 20th century.
Earlier in this paper we referred to Traugott's thesis that, in accordance
with her general semantic-pragmatic typology of change, the trajectory of
English a(n) has been from a propositional to a textual and perhaps expressive
meaning. Another way of formulating this typology is to speak of a develop
ment from «less personal» to «more personal» (Traugott's Hypothesis A
[253]). Insofar as this form of the hypothesis does not involve the acceptance
of the notion of «propositional meaning», or indeed of categorial differences
between «types» of meaning, we find it distinctly preferable, and completely
compatible with the results of our own study. The meanings of forms are all
to some degree negotiable, and all constitute a part of the set of resources
which speakers of a language have for constructing «texts» in the broadest
sense. Some of these resources are available with relatively little context, that
7 Wald, Benji. 1983. «Referents and Topic within and across Discourse Units». In Flora
Klein-Andreu, ed., Discourse Perspectives on Syntax. Academic Press.
302 PAUL HOPPER / JANICE MARTIN
is to say their meanings are relatively stable and fixed in advance of their use,
while others are very highly contextualized. The Modern English indefinite arti
cle is highly contextualized, since its uses range from specific to simply non-
anaphoric, from singular («a bird in hand», «a stitch in time») to expressive.
Probably, then, one way in which grammaticalization is to be com
prehended is as an evolution from lesser to greater contextualization, or from
relative textual autonomy to relative textual dependence. The more dependent
a form is, of course, the more «personal» it is, i.e. the more adaptable to the
speaker's goals, and the more negotiable its meanings. The more autonomous a
form is, the less flexibility and range of negotiability it makes available to users.
This brings us back to the difficulty stated at the beginning of the paper,
the problem of the relationship between synchrony and diachrony as a win
dow on change. It has often struck us that truly diachronic statements about
languages seem incompatible with synchronic ones. This is not always a con
tradictory kind of incompatibility, but rather more often one of salience. Facts
about languages which are important in studying change and reconstruction
are frequently trivial, or at best peripheral, when viewed synchronically.
Among them are minor paradigmatic irregularities; frozen idioms and pro
verbs; facts of variation; categorially ambiguous forms like Adverbs/Preposi
tions, Adjectives/Participles; and semantic-pragmatic subtleties of the kind
Traugott has shown to be so central in understanding change, such as the
deontic-epistemic distinction in modals, are not generally viewed as central
to the «rules» for generating «sentences». We may add to these quantitative
discourse data showing the distribution of some linguistic forms to be a mat
ter of statistical preference rather than binary categorial implementation.
Facts of this kind are more likely to be seen as exceptional and idiosyn
cratic, and to be described by special rules extra to the «core grammar».
In the case of discourse data they may even be relegated synchronically to
a special component outside the grammar as such. Structuralism biases its
adherents towards fixed categories, and forces them, if we may adapt
Dilthey's words, to «fix in categories things which change, and then sup
plement the fixed category with a category of changing».
This is, of course, by no means a new lament of historical linguists.
But ironically at the very time when the social sciences are vigorously debating
such issues as post-structuralism, post-modernism, and de-construction
linguistics appears to be making a reactionary turn in the direction of for
mal abstract models, and there is no better perspective than that of the study
of change from which to challenge this turn.
STRUCTURALISM AND DIACHRONY 303
0. In this paper I have two principal goals, one particular and a se
cond more general. In the particular, I propose to review selected facts about
inter-clause syntax (INTCL) in the ancient Indo-European (IE) text
languages; my purpose is to demonstrate how an analysis of those facts,
undertaken without certain preconceptions, might inform our hypotheses
about the nature of INTCL in the parent language, especially if pursued
in the context of our more general knowledge about devices available to
languages to mark structural and semantic relationships between clauses.
More generally, I will be using these necessarily selected comments on INTCL
in Indo-European to highlight what seems to me, at least, to be certain fun
damental misdirections in traditional and more contemporary scholarship
on syntactic reconstruction (SR).
1.1 Before proceeding, I should define certain terms that I will be us
ing here. INTCL refers to those structural devices of grammar that mark
an intimate semantic relationship between/among clauses (actually, proposi
tions), such that the underlying propositions so linked function in surface
structure as constituents of a syntactic construction larger than the clause,
and such that the correct semantic interpretation of each constituent pro
position is incomplete or unrecoverable without reference to the other(s).
In cases of INTCL, one or all of the constituent propositions may not be
manifested as full clauses in surface structure.
I have found it advantageous to work with the relatively neutral no
tion INTCL in other contexts 1 , and I use the term advisedly in the present
discussion. It is my intention here - to the degree possible - to avoid evoking
3 Hermann admits the possibility that subordinate clause marking by means of accent
is reconstructible for pIE.
4 Lehmann 1980 calls particular attention to Bonfante 1930 and Meillet 1937. Meillet
and others have, however, consistently pointed out that there is substantial evidence for the
reconstruction of a class of verbal-nouns (? infinitives) and verbal-adjectives (participles); the
reconstruction of such forms in prehistoric Indo-European would of course demand the positing
of reduced incorporated clauses (GSI, Type 2). For the purposes of the present discussion,
I will be concerned only with underlying propositions which maintain their integrity as full
clauses in surface structure. For discussions of IE non-finite verbals, see Jeffers 1975, 1976c,
1979, 1984.
METHODOLOGY IN SYNTACTIC RECONSTRUCTION 309
6 There is ample literature, both critical and supportive on other aspects of the
typological method in syntactic reconstruction, and there is no need to rehearse that discus
sion here; see, however, Jeffers 1976a, 1976b; Lightfoot 1979, 3.3.
METHODOLOGY IN SYNTACTIC RECONSTRUCTION 311
nuine sentence connectives10; and clitic particles, principally -(i)a and -ma
«or». I have discussed the functions of the lexicalized conjunctions elsewhere
(Jeffers 1976c). Let it suffice to say here, that these forms announce aspec
tual and adverbial relationships between their own and subsequent clauses
according to principles which have no obvious correspondence in general
IE 11 . Hittite relative words are based on the IE ku-stem, which I am ex
cluding for the sake of simplicity from the present exemplary discussion (see,
however, Jeffers 1986b). Most important to this brief survey of the reflexes
of IE *yo, however, are the sentence connectives and clitics.
3.1.3 Some scholars, notably Watkins (1963), assume that -ia is
originally a member of the same category of forms that produced Anatolian
sentence connectives; I have argued elsewhere however (Jeffers and Pepicello
1979), that -ia is to be distinguished from the class of pIE sentence connec
tives. An analysis of Hittite discourse features demonstrates that synchronic
-ia/a is functionally distinct from Anatolian (and by inference early IE)
sentence connectives like Hittite nu-12.
A genuine sentence connective, like nu-, serves to move the discourse
forward, and is appropriately translated «and then»; consider ii. and iii.
A sequence of three or more «nu-clauses» is commonplace.
ii. tak-ku ÌR URU Ha-at-ti IŠ-TU KUR URU Lu-ú-i-ya-az LÚ
URU
Ha-at-ti ku-iš-ki da-a-i-iz-zi na-an A.NA KUR
URU
Ha-at-ti ú-wa-te-iz-zi ...
If any Hittite man steals a Hittite slave from the country of Luwiyas, and
then brings him to the country of Hatti... (Laws I § 20)
iii. ták-ku DUMU.SAL-aš LÚ-ni ha-me-in-kán-za nu-uš-ši ku-u-ša-ta
píd-da-iz-zi ap-pí-iz-zi-na-at at-ta-aš an-na-aš hu-ul-la-
an-zi na-an-kán LÚ-ni tuh-ša-an-zi ku-u-ša-ta-ma 2-ŠU
šar-ni-in-kán-zi.
If a girl is betrothed to a man, and he pays the bride-price for her, and after-
10 The genuine sentence connectives include *no-, *so-, *to-, and *e-, reflected in Hitt.
nu-, šu-, ta-, and Luv. a, respectively.
11 takku (later, man) marks a conditional relationship; mán (later, mahhan), perfective;
kuitman, simultaneous/durative; kuit, causal; kuwappi, simple temporal (whenever).
12 In terms of distribution, sequences of sentence connective plus clitic are not uncom
mon, a pattern that would be inexplicably redundant if the forms did not serve different func
tions.
314 ROBERT J. JEFFERS
wards (her) father (and) mother violate it (i.e., the marriage contract), and
then take her away from the man, they return the bride-price in double the
amount. (Laws I § 29)
On the other hand, Hittite -ia (and its alternant -a) generally connects two
clauses in which the described actions or states are intimately connected in
time and space (i.e., where the two clauses describe two components of a
single situation), or where the connected clauses refer to parallel notions.
Consider iv. and v. (See also Friedrich 1959: 116).
iv. ták-ku LÚ DAM.GÀR URU Ha-at-ti ku-iš-ki ku-en-zi I ME
MA.NA KÙ.BABBAR pa-a-i pár-na-aš-še-e-a šu-wa-a-iz-zi.
If anyone kills a Hittite merchant, he pays one and a half pounds of silver
and gives his farm buildings in security. (Laws I § 5)
v. ma-a-na-aš SIG-at-ta-ri-ma nu-uš-ši 6 GÍN KU.BABBAR
pa-a-i LÚ A.ZU-ia ku-uš-ša-an a-pa-a-aš-pat pa-a-i.
When he 1 gets well, he 2 pays him six shekels of silver, and then he 2 pays
the physician his fee (Laws I § 10)
According to this analysis, then, Hitt. -ia marks a relationship between two
clauses which must be interpreted, structurally and semantically, as consti
tuents of a single syntactic construction.
In the context of this interpretation of Hitt. -ia, let us turn to a review
of some of the syntactic and semantic properties of the more well-established
reflexes of *yo to determine if there exists evidence of any structural and/or
functional correlation between these reflexes and their alleged Hittite cognate.
The form yo occurs most commonly as the stem/base for relative adjec
tives and pronouns, and for the so-called «relative conjunctions» of Indo-
Iranian and Greek 13. The «relative conjunctions» are complex, lexicalized
forms comprising yo plus one or more additional (probably adverbial) par
ticles, e.g., Skt. yathā (ya- < *yo-); Gk. ophra (o- < ho- < *yo-).
3.2.1 A review of a representative sample of situations in which a yo
form plays some role in marking a relationship between two clauses
13 Even when the actual term «relative conjuction» is not used, the assumption that these
lexicalized forms reflect an extension in the use of the relative stem/theme is standard. See,
e.g., Monteil 1963: 76-79, et infra.
METHODOLOGY IN SYNTACTIC RECONSTRUCTION 315
demonstrates that each of the clauses maintains its internal structural in
tegrity and its surface structure autonomy, (i.e., neither of the clauses «is
surrounded by material from the other»). (The examples given in vi.-xiii.
comprise such a sample. The particular sentences given in these items were
also chosen to exemplify other points that I will be making in the remainder
of this paper, and I will refer to them throughout).
Compare
Compare Latin
b. yá udŕcīndra devágopāh/
if brilliant-I. poss. the gods as guardian (we are)
14 The character of this syntactic pattern has, of course, been noted by earlier scholars,
but it has been attributed, for example, to the style and purpose of the texts that have come
down to us; see, e.g., Minard 1953: 2-3.
METHODOLOGY IN SYNTACTIC RECONSTRUCTION 319
tions») are typical and, at least in the later classical versions of those dialects,
standard (see xii.). Examples of relative clauses of purpose and condition
are given in viii. and ix., respectively. (Compare the corresponding
and similarly archaic Latin relative clause of purpose with pronominal qui,
viii. c).
I have suggested elsewhere (Jeffers & Pepicello 1979), - a claim reiterated
in Holland 1984 in a slightly different context - that the contemporary oc
currence of these two types of adverbial clause - those characterized by *yo-
based conjunctions and others with inflected *yo-forms - is probably best
understood as a consequence of the fact that a coreferential noun may or
may not be present in pairs of connected clauses whose relationship is one
of purpose, result or condition.
It is significant that whenever an adverbial clause is marked by a
stereotyped case form like Skt. yad (xi.) or yena (x.a.), or by a complex,
lexicalized *yo-based conjunction like Skt. yathā (xii.a.), the constituent *yo
simply marks the presence of some close, but undifferentiated interclausal
relationship (again cf. Hitt. -ia). If a complex *yo-based conjunction is the
marker of INTCL, the morphological entity attached to *yo- serves to refine
the semantic character of the particular relationship; the INTCL relation
ship may also be specified by means of a case marking (e.g., yena; x.a.);
but in the very common situation where *yod (Skt. yad) indicates a rela
tionship between two clauses, the nature of that relationship is determined
solely by context and, to some degree, by the relative order of the clauses
involved (see ..1., below)15. On the other hand, the occurrence in adver
bial clauses of an inflected form of *yo makes possible the marking of focus
in situations where a corefential noun is, in fact, involved in the bi-clausal
construction. (See also Pepicello 1978).
3.2.3 The facts discussed in 3.2.2 highlight certain problems with
attributing the notion «relative» to the full range of *yo-based forms
associated with INTCL in the ancient IE text languages. If the term «relative»
refers to corefentiality, its use with reference to complex, *yo-based lexical
conjunctions is clearly inappropriate, as the forms comprising this category
typically serve as markers of INTCL in constructions where no coreferen-
tial noun is involved. It is surely the case that the term has come to be used
with reference to such forms because of the presence in them of the consti
tuent *yo. The early texts, however, offer little evidence to support an
etymological interpretation (relative) for the form *yo. It is only in cases
where an inflected form of *yo occurs that coreference is a factor in the
inter-clausal relationship; moreover, the semantic information relevant to
coreference in those situations is carried by the case affixes whose «attach
ment» to *yo may well reflect some secondary reanalysis/restructuring, the
details of which remain obscure 16. The one property common to all oc
currences of *yo considered here is that it marks the relationship between
clauses which must be interpreted as constituents of a larger syntactic con
struction.
3.3.1 I have already briefly mentioned the role of relative clause order
with reference to the interpretation of certain adverbial clauses (3.2.2). It
is in fact the case that the order of any two linked clauses, in which a yo-
form serves as the marker of INTCL, is always determined by semantic con
siderations which are transparently pragmatic/iconic. yo-clauses which refer
to topics, conditions or causes come first (see vi., vii., ix. and xi.); those
which express purpose or consequence not surprisingly come second (viii.,
x., xii.). This is the case whether *yo appears as an inflected form, as the
base constituent in a complex, lexicalized conjunction, or in the stereotyped
*yod.
4.1 In light of these facts about the syntax and semantics of *yo, let
us return to the «structure-neutral» question I posed in section 3.0, and see
if some tentative response might be constructed at this point. We might begin
by considering the facts detailed in sections 3.1 -3.3 in terms of the embedding-
adjoining continuum for INTCL in the world's languages 17.
On the basis of the restricted situations reviewed here, we might sug
gest tentatively that constructs/morphemes, including at least the form *yo,
served as generalized markers of a close semantic relationship between clauses
comprising some larger syntactic construction in prehistoric Indo-European;
pairs of clauses linked by *yo, however, always retain structural autonomy
even when they share a coreferential noun phrase. The term «adjoining»,
16 The claim in Jeffers and Pepicello 1979 concerning the origin of IE relative pronouns
was premature.
17 I owe the term/notion «embedding-adjoining continuum» to Suzanne Romaine (1985).
METHODOLOGY IN SYNTACTIC RECONSTRUCTION 321
18 The similarities to Australian are striking. Consider these examples from Walbiri (Hale
1976) in which the form kutja- (prefixed to AUX) marks the interclausal relationship. Con
struction 14a includes a coreferential noun phrase; 14b does not.
a. yankiri-Įi kutja-lpa ŋapa ŋa-nu ŋatjulu-lu ø-na pantu-nu.
emu-ERG COMP-AUX water drink-Past I-ERG AUX spear-Past
«The emu which was drinking water, (and) I speared it».
«While the emu was drinking water I speared it».
b. rjatjulu-lu lpa-na kali tjantu-nu, kutja-0-npa ya-nu-nu njuntu.
I-ERG AUX boomerang trim-Past COMP-AUX walk-P-hither you
«I was trimming a boomerang when you came up».
19 This hypothesis is not inconsistent with facts relating to the occurrence of yo in its
function as a genitive case marker; see Gonda 1954a.
322 ROBERT J. JEFFERS
REFERENCES
F
CONSIDERAZIONI SULLA CRONOLOGIA RELATIVA DEI
MUTAMENTI FONETICI
ROMANO LAZZERONI
Università di Pisa
10 Cfr., per lo stato della questione e per la bibliografia, R. Caldarelli, Sulle più recenti
interpretazioni della cosiddetta Legge di Lachmann, Macerata, 1982 (= Quaderni Linguistici
e Filologici - Ricerche svolte presso l'Università degli Studi di Macerata, II).
11 Cfr. R. King, op. cit., p. 58 ss.; T. Bynon, op. cit., p. 118 ss.
12 Significative le parole di R. King, Lg, XLIX, 1973, p. 577.
13 Proceedings of the 11th International Congress of Linguists, Bologna, 1974, p. 369 ss.
328 ROMANO LAZZERONI
14 V. sopra, n. 9.
15 Spesso, per es., non si tiene conto di mutamenti, per così dire, pancronici, o, me
glio, condizionati dalla struttura di una lingua. Non si può affermare - l'osservazione è di H.M.
Hoenigswald, op. cit., p. 373 - che nelle lingue slave k ́ > s è anteriore a sr > str perché l'epen
tesi della dentale avviene anche con s < k. Per lungo tempo, infatti, il nesso sr non è esistito
nelle 11. slave, sicché st è semplicemente un allofono di s davanti a r; cfr. Id., Ancient Indo-
European Dialects, ed. H. Birnbaum - J. Puhvel, Berkeley - Los Angeles, 1966, p. 12.
16 Proceedings, cit., alla n. 13, p. 373.
CRONOLOGIA RELATIVA DEI MUTAMENTI FONETICI 329
34 Il caso più noto è quello della neutralizzazione di ā ed eā nella storia dell'inglese; cfr.
W. Labov, Il continuo, cit., p. 121 ss.
35 Linguistica e Filologia, Omaggio a B. Terracini, Milano, 1968, p. 151 ss.
36 Dialetto e società industriale nella valle di A n d o r n o (Supplementi al B A L I , I), Tori
n o , 1970.
336 ROMANO LAZZERONI
Si aggiunga che la reazione del sistema dominato alla pressione del si
stema egemone talvolta non è di semplice adeguamento, ma comporta lo
sviluppo di forme autonome 37 ; e allora, se i documenti non ci soccorrono,
l'interpretazione è ancora più difficile.
Si dirà che questi non sono mutamenti fonetici, ma sostituzioni di suo
ni trasportati da serie lessicali? A parte la dubbia legittimità della distinzio
ne - su ciò, in polemica con Ch. F. Hockett 38 ha scritto una splendida pa
gina U. Weinreich39 - resta il fatto che essa non può cogliersi quando il
processo è compiuto ed osserviamo soltanto i risultati finali.
37 Cfr. C. Grassi, art. cit., p. 153 ss.; G.B. Pellegrini, Protimesis, Scritti in onore di
Vittore Pisani (= Studi Linguistici Salentini, II), Lecce, 1969, p. 99 ss.; Id., ID, XLV (n.s.
XXII), 1982, p. 25 ss.
38 A Course in M o d e r n Linguistics, New York, 1958, p . 439 ss.
39 R o m P h , X I I I , 1960, p . 331 ss.
CRONOLOGIA RELATIVA DEI MUTAMENTI FONETICI 337
LAZZE
TIME
WINFRED P. LEHMANN
The University of Texas
Our discipline took its start in the 19th century when the learned world
was developing methods and theory to account for its physical surroundings.
Geologists were sorting out the configuration of the earth; chemists were
examining the inner structure of objects. Among other scientists anatomists
were scrutinizing, describing and analyzing the human body, including the
vocal tract, to determine its functioning in accordance with strict scientific
principles. Their successes deriving from application of their rigorous
methodology provided them with prestige among other investigators, who
sought similar successes in dealing with their topics of research, such as
language. To achieve such successes these further scientists (earlier referred
to as moral or historical, more recently as social, behavioral, or human)
adopted methodological principles developed and refined by the physical
scientists of their day.
The 19th century scientists were also concerned with combatting
previous explanations of their physical surroundings through catastrophes,
a concept differing from that of René Thorn, who might with some cir
cumspection better have chosen a different term. The previous explanations
are evident in the work of the early linguists, like Sir William Jones. As
one of his major priorities he sought out Indian and Chinese accounts of
a great flood, looking in this way to corroboration of a catastrophic event
that was held to have radically modified the course of the earth.
That catastrophe also affected man's relationship to time. Before the
flood man could enjoy virtually a millennium of existence, in contrast with
the 70 or 80 years which the psalmist allots his contemporaries and which
we expect for ourselves. By contrast, the new scientists assumed a steady
and comparable course through time for themselves and their physical sur
roundings, with states being maintained and changes taking place in the same
way through past ages as at present. Glaciers formed and receded, rivers
340 WINFRED P. LEHMANN
produced deltas, substances oxidized over specific periods in the past exact
ly as in the present. Greater or lesser quantities of water, debris or oxidizable
matter might be involved, but time elapsed then as it does now. To under
stand their surroundings scientists had to be concerned with the dimensions
of quantities and their components in variation but could confidently assume
that the earth proceeded on its course at the same pace two millennia, two
hundred, two thousand millennia ago as today. Substances are maintained
or changed at the same pace in Asia, Africa or the Americas, even on other
planets and on stars, as in Europe.
The principle so applied is known as uniformitarianism. In accordance
with uniformitarianism the physical world is everywhere homogeneous;
hydrogen in Britain is homogeneous with hydrogen in Australia, on the planet
Venus, and elsewhere. When changes involving hydrogen or other substances
take place, they occur everywhere in accordance with homogeneous prin
ciples (Christy 1983).
In carrying out scientific study involving time, the variables were sought
in substances. A scientist would identify an entity, whether a hydrogen
molecule, a crystal, or a human organ and determine possible modifications,
knowing that entities are open to change, or in the magic word of the day
to evolution, which might affect their form, their quantity, their relation
to their surroundings. Such changes were stated with reference to time, which
was regarded as an independent dimension.
Biologists also regarded time in this way. Organisms might be subject
to different processes than are lifeless substances, but in biological study
as well time was measured apart from the substances at hand. Biological
clocks tick on inexorably and steadily, whether the organisms existed in
paleolithic, pre-Cambrian or more recent periods.
When the less substantial attributes of organisms, such as language,
came to be examined scientifically, they were treated in accordance with the
same principles. Language was regarded as an organism. It could be examin
ed at a specific stage, as it was by Adelung, Humboldt, Becker and other
descriptivists of the past who are currently neglected. Or, language could
be examined in change, as by the preponderant number of linguists in the
19th century. Under either approach the established view of time establish
ed by the physical scientists was maintained. One examined linguistic material
for its elements, or substances — the sounds, forms, lexical elements, syn
tactic strings, stylistic sequences. All of these were held to observe similar
procedures, whether in the second millennium before our era or merely a
TIME 341
sively as did English from French and Latin, but the basic structure of
Japanese remained unmodified. Phonological and syntactic change clearly
took place at different rates of time in Japanese from those in English.
Internal social relationships have specific effects. Pastoral societies en
courage frequent meetings among speakers of a specific culture, leading to
uniformity of language in spite of sparse populations spread over broad
areas. Change takes place more slowly in such groups than in sedentary
cultures with self-contained groups having little intercommunication. Proto-
Indo-European in its pastoral society apparently changed more slowly than
did the language of feudal Europe. The pastoral nomadic situation in which
Arabic was maintained through three millennia may have led to similar
maintenance of the language with little change.
There has been much speculation about the situation in modern societies
who have developed means of rapid intercommunication by such means as
radio, moving pictures, and television (Labov 1965). The extensive society
so interrelated contrasts greatly with the small, self-contained groups in
medieval Europe, when multiple dialects flourished. Changes are apparent
ly carried out rapidly in such self-contained social groups, as the multiple
dialects of Chinese, the many languages of Papua and pre-Columbian
America attest, as well as the numerous dialects of German, French and
other European languages which developed during little more than a millen
nium.
Time then cannot be disregarded in relation to its surroundings, on the
assumption that change takes place at a comparable pace, whatever the type
of social organization or social relationships. In a short period Akkadian
underwent more changes than did the Semitic dialects of Arabia over millen
nia. And Japanese remained relatively unchanged during a period in which
English was thoroughly modified. Previous work in historical linguistics pro
vides a good number of accounts which we can examine in sketching various
models of language change. We need now to determine the scenarios and
the likely changes in each.
Through such an effort we will be able to suggest more credible causes
of language change than those proposed in the past. These were based on
selected conditions which make up only a part of the elements to be taken
into consideration. As examples we may recall the explanations for the Ger
manic consonant shifts cited by Prokosch (1939: 55-57).
1) «an expression of the impetuous character of the Germanic tribes
during their early history;...
TIME 345
male and female language may have an effect on change. The relatively great
differences between male and female Japanese may well bring about strict
attention to usage, which may lead to greater language stability.
External social relationships may have been most widely explored among
causes for change, and also noted for important conditions: the size of dif
fering communities and the prestige of their speakers. Early English, in its
contacts with Celtic, Scandinavian and Norman French, provides instruc
tive examples. The intimate contacts with Norman French speakers of higher
prestige may help account for the greater impact of French and Latin on
English than we find for Japanese, where the Chinese lexicon was transmit
ted by native Japanese speakers educated in China rather than by a domi
nant foreign social class in the country.
The attitudes of speakers to language must also have its effects. Do
they respect a standard? Do they maintain a traditional, religious language
and is this language related to their own, as is Latin to the Romance in con
trast with the Germanic languages? A contrastive study taking into account
the development and maintenance of a common Slavic religious language
in eastern Europe as opposed to the absence of a common Germanic religious
language in central Europe might illuminate some of the bases of stability
in the Slavic as opposed to the Germanic languages. Openness to other
languages, and social stratification may also affect language change. The
self-segregation of the Brahmans in India must have assisted in time stand
ing still for Sanskrit while the religious language of Europe developed into
Kitchen Latin and the somewhat more conservative Latin of medieval learned
discussion. Social stratification may also lead to change, possibly sporadic
change, as in the introduction of the third person plural as form of address
in parts of the area dominated by Latin culture.
To account for language change as we find it under these varied condi
tions with reference also to geographical alignments requires scrutiny of the
patterns of language use in society and the relationship of these with the
lapse of time. Through such scrutiny we will have the possibility of learning
how and why language changes. We will also contribute to other students
of culture in change, who like us have operated with a simplistic view of time.
«Time itself», as Hazan states in a study with a specific focus, «has
rarely been the principal object of anthropological consideration» (1984:
567). Like anthropologists, linguists have maintained «the conception of the
flow of time as irreversible», with little attention to it in relation to other
effects on linguistic communities. Nor have we dealt adequately with those
TIME 347
REFERENCES
Andersen, Paul Kent. 1983. Word Order Typology and Comparative Con
structions. Amsterdam: Benjamins.
Hawkins, John A. 1979. Grammatical Universais as Predictors of Word
Order Change. Language 55:618-48.
. 1983 Word Order Universais. New.York: Academic.
Hazan, Haim. 1984. Continuity and Transformation among the Aged: A
Study in the Anthropology of Time. Current Anthropology 25:567-78.
348 WINFRED P. LEHMANN
HELMUT LÜDTKE
Romanisches Seminar - Universität Kiel, BRD
1. Preliminaries
Language (in the sense of Fr. langage) should not be viewed as an in
strument of communication; we should rather say that language is (human
vocal) communication; in other words, it does not exist but it takes place.
Viewed as an object, language is a platonic idea, or a construct, to use more
recent terminology.
As far as its structure and functioning is concerned, language is cer
tainly neither a machine nor a living organism. Nevertheless, being a pro
duct of the human brain, it is organised along biological rather than on
mechanical principles; excepting, perhaps, the speech wave considered as
an extracorporeal phenomenon.
When a machine - or some parts of it - breaks down it will be repaired.
The history of a machine may be described as a cyclic sequence of two alter
nating states, i.e. functioning and repair. According to the followers of De
Saussure (and many other linguists who never explicitly confessed their struc
turalist orientation), the same would hold for any linguistic subsystem. It
eventually becomes dysfunctional and then gets either reconstructed or
replaced. In keeping with this mechanistic philosophy of mainstream struc
turalism, language change is viewed as an unending sequence of destruc
tion and subsequent re-generation.
I think the first linguist who questioned the validity of this view was
Walther von Wartburg in his 1931 booklet intitled «Das Ineinandergreifen
von deskriptiver und historischer Sprachwissenschaft» (Berichte der
Sächsischen Akademie der Wissenschaften 83/1). «Ineinandergreifen» is a
metaphor taken from the notion of cog-wheels. It implies that the two pro
cesses involved are not in chronological succession; which means that neither
one can be the cause of the motivation for the other one.
350 HELMUT LÜDTKE
SHRINKAGE
ACCRETION
MERGER
3. Corollaries.
One of the immediate consequences that derive from the above scheme
is that the organisation of naming through items by a giving language can
not be stable. Since the wheel of the triad always keeps turning, every item
is, in the long run, doomed to loose its identity either by elimination from
the repertoire through disuse or by merger with some syntactically adjacent
item.
Another consequence ensuing from the above assumption is that
presumably every item of every language stems from a former syntactic com
bination.
Up to now we have taken it for granted that items exist, cease to exist,
or come into existence. We must also look at the other side of the coin (cf.
chapter 1.): items exist, i.e. they have their individual identity, insofar as
they are identified by the hearer in actual communication, and stored in
his/her brain in order to be re-used when (s)he acts as a speaker.
Being or not being identified as an individual item — versus constituting
part and parcel of the phonic representation (or exponence) of a longer item
— is a matter of probability. The score will be highest for the bulkiest items,
i.e. those having greatest phonic size.
Size, however, is not the only factor that determines identification. The
others are frequency of use, distributional constraints, i.e. occurrence at a
certain place in the syntactic chain where a given item, or member of a given
class of items, may be expected to occur, and, in the latter instance, size
of the class to which the item in question belongs. Obviously, potential for
identification, i.e. the probability of being identified, is positively correlated
with phonic size, frequency, and distributional constraints; negatively, with
the number of commutable items.
The foregoing argumentation helps to account for some phenomena
of semantic change (e.g. such as go along with an increase of frequency)
and for a bundle of processes described under the heading of «gram-
maticalisation», i.e. a gradual shift transforming lexemes into grammatical
morphemes. The rise and decay of morphological (sub)systems is partly ex
plainable as a complex process by which some lexemes gain a temporary
respite from loss of identity. This is brought about by compensating phonic
352 HELMUT LÜDTKE
4. Auxiliarity.
With respect to accretion in the diachronics of naming we may set up
the following distinctions:
— juxtaposition vs. subordination;
— enrichment: lexicon vs. accidence;
— accretion occurring in: verb phrase vs. noun phrase or between noun
phrases;
— anteposition vs. postposition of auxiliary to lexeme;
— replacement vs. reinforcement of decaying morphological subsystems by
auxiliaries;
— semantic ranges covered by auxiliarity (or, more specifically, by auxiliary
verbs) in different languages.
5. Auxiliary verbs.
The notion «auxiliary verb» may be given two different, though com
plementary, definitions, one of them being synchronic, the other diachronic;
viz.:
— items that occur in the same positions as verb lexemes do, while their
potential for combining with lexemes, including themselves, is like that of
grammatical morphemes;
— a cross-linguistic device of accretion by which certain verb lexemes are
substituted for decaying subsystems of verb accidence.
Auxiliary verbs are likely to occur wherever two prerequisites, a syn
tactical and a morphological one, are met, viz.:
— verb serialisation as a regular pattern;
— existence of uninflected (or only partially inflected) verb forms, e.g. in
finitive, participle.
True auxiliary verbs are fully integrated into the pattern of verb ac
cidence, which means that, on principle, they combine with any non-
AUXILIARY VERBS AND THEORY OF CHANGE 353
obsolescent verb lexeme in the language. Thus, they are factors in the Carte
sian product constituted by verb accidence. They are to be distinguished from
serially occurring verbs that express all sorts of modalities but whose distribu
tion is limited on semantic grounds.
SILVIA LURAGHI
Università di Torino
* I thank Michael Job and Harm Pinkster for their comments on earlier versions of this
paper.
1 I refer here to the reconstruction given in Szemerényi (1985: 191-195). On possible alter
native reconstructions, see below, section 2.2.
2 Functional syncretism should be kept apart from morphological syncretism. The lat
ter has its immediate cause in phonetic erosion; its result is the phonological identity of one
or more morphemes. The development leads to homophony; real merging of the morphemes
involved may also occur, if functional syncretism goes hand in hand with morphological syn
cretism. Coleman (1976) mentions three patterns of case syncretism in Latin. The first two
concern instances of mere homophony. I have doubts, whether one should really use the term
«syncretism» for such phenomena; see the discussion in Mignot (1978). The third pattern describ
ed by Coleman concerns instances of functional syncretism in conjunction with morphological
syncretism. It must still be noticed that, although functional syncretism often shows up together
with morphological syncretism, morphemes involved in functional syncretism need not be
phonologically identical, i.e. functional syncretism may result in allomorphic variation.
356 SILVIA LURAGHI
1.
1.1 Conditions on syncretism.
Functional elements such as cases are seldom used completely unam
biguously, for one function only. More often, they can be used with a cer
tain degree of freedom, which enables speakers to choose among different
formal means in order to convey the same meaning 3 . The formal means
(e.g. cases) involved can then be regarded as being characterized by partial
synonymy 4 ; i.e. they are felt as synonyms in those contexts in which they
can overlap freely. Free overlap of cases clearly constitutes a basis for their
merging (syncretism). The latter can take place under two general condi
tions: a) the degree of partial synonymy must be high enough to call for
a reduction of redundancy; b) functional merging and possible formal merg
ing of morphemes must not engender unacceptable ambiguity 5 .
7 For example, the locative forms of Sanskrit dame, Latin domi, Greek oikoi, with the
meaning «at home».
8 I follow here the terminology of Dik (1978). A typical example for this correspondence
in Indo-European languages may be the nominative case, taking almost exclusively the syntac
tic role of subject, i.e. the first argument of a predicate.
9 There is evidence for semantic syncretism even in languages relying on syntactic syn
cretism and viceversa. For example, in Latin (see below, section 3.2.1), syntactic syncretism
of the ablative, the locative and the instrumental can possibly be reduced to previous semantic
syncretism. The merging of the ablative and the instrumental must have happened at an earlier
stage than the merging of the two cases mentioned with the locative. In fact, formal merging
of the locative with the ablative / instrumental was not yet fully accomplished for the -ā- stems
358 SILVIA LURAGHI
in Classical Latin. See Brugmann-Delbriick (1911: 487). Syncretism of ablative, locative and
instrumental is also attested in Homeric Greek by the adverbial ending -phi-. This instance
of syncretism was contrary to the basic pattern operating in Greek, in that it engendered semantic
opacity. Consequently, the ending -phi- was increasingly emarginated, until it finally disap
peared. On the use of -phi- in homeric Greek, see Chantraine (1958: 234-241). There is no
agreement among scholars on the status of Mycenean -pi. Recently, Hettrich (1985) has pro
pounded that Mycenean -pi only marked the relations of ablative and instrumental. The func
tion of the locative would then have been taken over at a later stage, parallel to the develop
ment of the Latin ablative.
10 The bearing of lexical features in determining case marking has been emphasized by
many scholars; see, among others, Croft (1984), Deane-Wheeler (1984), Bybee (1985) and
Pinkster (1985).
11 Cf. Pinkster (1985: 178). A good example for this phenomenon is given by the follow
ing sentence: hėn. Phlégrāi Gigantes éstësan theoîs, «(the war) that Giants fought in Phlegra
against Gods» (Eur. Ion 988). Lexical features also explain the existence of 0 locatives, such
as Hittite takan «on the earth», Greek êmar, «by day»; see Neu (1980), Benveniste (1962).
12 See above, fn. 8. Dik (1978: 27) stressed that the instrumental case seldom takes the
position of an argument, as most predicates do not require instruments within their predicate
frame.
CASE SYNCRETISM IN INDO-EUROPEAN 359
For the purpose of the present paper, animacy appears to have par
ticular relevance among lexical features of the noun, in that it turns out to
be crucial in determining overlap among cases. In fact, as it will become
clear in the discussion in section 3. below, animacy plays a major role in
establishing the degree of acceptable ambiguity. In Indo-European languages,
animacy affects in particular the position of the dative case. This had already
been stressed by Brugmann-Delbrück, in a passage which is worth quoting,
because it also emphasizes, as a syntactic property of the dative, that of in
dicating the third argument of a three-place predicate, as opposed to its se
cond argument: «Ueberwiegend erscheinen Personalbegriffe im Dativ, wobei
die Person den Vorgang irgendwie aufnimmt, besonders wenn daneben noch
ein Akkusativ oder Genitiv vom Verbum abhängt, der dann das nähere Ob
jekt bildet» (1911: 548)13.
Moreover, lexical features of the verb also contribute to disambiguating
the semantic relation borne by a noun phrase, as well as its syntactic role.
This largely depends on the fact that predicates bear inherent information
about their predicate frame, concerning the number of arguments and the
semantic relations that these arguments must be assigned14.
To sum up, lexical features are crucial in determining the structure of
any given case system. This applies to our understanding of both the syn
chronic functioning of a case system, and its development in time.
2.
2.1 The Proto-Indo European case system.
Recent contributions to the reconstruction of the Proto-Indo-European
case system have increasingly made it clear that the classical reconstruction,
which one still finds in Szemerényi (1985), cannot constitute the basis for
an explanation of each specific case system attested in the daughter languages.
On the one hand, one finds limited morphemic distinction for some
cases in all Indo-European languages (e.g. ablative singular) or morphemes
which are etymologically connected but have a different status in different
languages (e.g. Ancient Greek suffix -phi as compared to Sanskrit endings
13 For exemplification from different Indo-European languages see, for instance, Jakob-
son (1936), Starke (1977), Luraghi (1986a) and Pinkster (1985).
14 Cf. Dik (1978: 25-32) and Pinkster (1985).
360 SILVIA LURAGHI
in -bhi-), so that one could easily conclude that the really innovative pat
tern was the Sanskrit one, in that it was built up by widening and regulariz
ing inflectional paradigms.
On the other hand, doubts on the correctness of the traditional
reconstruction also come from the discovery of a directive case in Old Hit
tite, indicating goal with verbs of motion, which was marked by the ending
-a, also attested in adverbs in other Indo-European languages15. The ques
tion could be raised, whether Proto-Indo-European should then be regard
ed as having had a bigger number of cases than was thought so far.
One wonders whether it makes any sense at all to talk about case syn
cretism, or whether one should better think to a process contrary to it, as
I mentioned above for Sanskrit, which one could propound for Hittite as
well, with regard to the directive case. However, one still must reckon with
well attested instances of morphological syncretism, as it is the case for the
Greek dative 16 .
The question of case syncretism can be reformulated as follows: Indo-
European languages must have inherited from the proto-language a set of
relational markers, partly definable as more or less independent particles,
partly integrated within an inflectional paradigm. These particles, whatever
their morphological status might have been, at least in late Proto-Indo-
European fulfilled the various functions that one singles out for each
reconstructed «case endings» on the basis of a comparison of the case systems
of the daughter languages. Moreover, the proto-language must already have
possessed a certain degree of semantic or syntactic overlap among those mor
phemes that the daughter languages have exploited in building up their own
inflectional paradigms.
For the purpose of the present paper, I will consider the eight cases
traditionally reconstructed for the proto-language, with the proviso that they
constitute a purely hypothetic system, a function, in the mathematical sense,
of the case systems of the historical languages 17 .
15 I refer here to the etymology put forward in Laroche (1970). Alternative proposals
have been made by other scholars: see for instance Shields (1982).
16 Chantraine (1973) has discussed throughly the morphological constitution of Greek
nominal paradigms. On morphological syncretism in Germanic, see Prokosch (1939: 231-241).
17 Of course, in a diachronic perspective, the evolution which resulted in the case
systems of the historical languages must be regarded as having started with an already
heterogeneous situation. As Serbat (1981a: 87) puts it, «l'histoire n'a pas pour premier chapitre
une 'Genèse'».
CASE SYNCRETISM IN INDO-EUROPEAN 361
18 See Greenberg (1963). On the typological evolution of Germanic, see Ramat (1981).
19 Plank (1978: 290) has stressed the importance of considering the interaction of alter
native and simultaneous use of coding devices as a prius to a study of case syncretism. On
the Greek locative, see below, fn. 26.
20 I discuss this possible instance of syncretism here instead t h a n in section 3 . , because
it is in fact unattested in historical languages.
362 SILVIA LURAGHI
two-place predicates taking the directive case as marker of their second argu
ment, i.e. verbs of motion. Syntactic syncretism between the directive and
the accusative could be achieved because the lexical features of the verb alone
may suffice in indicating the semantic relation underlying the noun phrase
which stands in the position of the second argument.
3. In order to exemplify semantic and syntactic syncretism I am now
going to take into account the case systems of Ancient Greek, Hittite, Latin
and Old Germanic. The instances of syncretism that I am going to survey
are the following21:
Ancient Greek: genitive, ablative → genitive;
dative, locative, instrumental → dative;
Hittite: dative, locative → dative/locative;
dative/locative, directive → dative/locative;
ablative, instrumental → allomorphs with the same function;
Latin: ablative, instrumental, locative → ablative;
Old Germanic: ablative, instrumental, locative, dative → dative.
21 Ch. Lehmann (1982: 107-112) has examined possible instances of interrelations among
cases, that may lead from «concrete» to «grammatical» cases, in a grammaticalization pro
cess. This can be interpreted as a source for case syncretism, mainly on semantic basis, An
analogous interpretation of the evolution from concrete to grammatical cases is given by
Kurytowicz (1949: 149).
22 O n case syncretism in Indo-European languages, and in particular on semantic sources
for syncretism, see Wackernagel (1926: 300-305) a n d Kurytowicz (1964: 189-197).
23 Morphological syncretism of the ablative and the genitive case also occurred in Balto-
Slavic. However, functionally speaking, things are not exactly the same as in Ancient Greek.
In Slavic, the function of the ablative has been taken over mostly by the prepositional genitive,
whereas the plain genitive can function as an ablative under much stronger lexical constraints
than it does in Ancient Greek. This is possibly due to the wide use of the partitive genitive,
as well as to the use of the genitive case ending as a marker of the animate accusative. See
Vaillant (1977: 37-80).
CASE SYNCRETISM IN INDO-EUROPEAN 363
24 I will be speaking of functions of cases in a rather loose way, referring to what one
can most likely reconstruct for Proto-Indo-European. The development of an ablative of
cause deriving from an ablative of origin or separation can best be observed in languages in
which the ablative and the instrumental were still being kept apart, i.e. Sanskrit (see Haudry,
1977: 141-142) and Old Hittite (see Melchert, 1977: 160 and 190). Partial syncretism of the
genitive and the ablative can be observed, in Ancient Greek, also with the adverbial ending
-then, limited to pronouns. The fact that with nominal stems -then only functioned as an ablative,
may be interpreted as an archaism, referring back to a stage at which genitive and ablative
were still being kept separated. See Chantraine (1958: 241-243) on the use of -then in Homeric
Greek.
364 SILVIA LURAGHI
which can be emploied with any kind of verbs, including those expressing
motion, while the distribution of the locative is in general more restricted25.
I will return to this peculiarity of the dative in section 3.1.3 below.
The complementary distribution of the dative and the locative in cer-
tain environments engenders a high degree of redundancy within the system,
in that the dative can be conceived as the animate counterpart to the locative.
Thus, merging of the two cases does not bring about ambiguity.
Syncretism of dative and locative is well attested in Hittite; in Ancient
Greek, although morphologically attested, it was functionally restricted 26 .
For the true locative, all Greek varieties, including Homeric Greek, resorted
rather to the prepositional dative. This made the dative case ending again
unexploited for inanimates.
25 On the use of the dative case with verbs in Indo-European languages, see Brugmann-
Delbriick (1911: 549-554); on possible use of the locative case with verbs of motion, see ib.:
514-516.
26 T h e use of the dative case without prepositions for indicating the locative was mostly
limited t o t o p o n y m s in Classical Greek. In H o m e r i c Greek, although m o r e widespread,
it already was constrained by a n u m b e r of lexical restrictions, see C h a n t r a i n e (1953:
78-81).
27 Notice, however, t h a t , for a particular set of verbs clearly indicating t h e end of m o -
tion (such as « t o sit d o w n » , « t o lay d o w n » ) , t h e locative was used instead of the directive in
Old Hittite, see Starke (1977: 47-66) a n d Luraghi (1986a: 33 a n d fn. 2). This must have held
already for P r o t o - I n d o - E u r o p e a n , see Brugmann-Delbriick (1911: 514-515).
28 I have analysed this instance of syncretism in terms of system internal economy in
Luraghi (1985).
CASE SYNCRETISM IN INDO-EUROPEAN 365
3.2.1 Latin,
The Latin ablative is the output of the locative, the ablative and the
instrumental. It can still be observed covering these three functions without
3.2.2 Germanic.
In the Germanic languages, the dative case as well merged with the
locative, the ablative and the instrumental.
The difference between the Latin and the Germanic patterns derives
from the syntactic twofold nature of the dative case. Syntactically, the dative
case can be conceived either as the third argument of a three-place predicate,
as opposed to the cases of satellites (ablative, locative and instrumental) as
was the case in Latin, or as the case of the non-closest complement of a
verb, as opposed to the accusative, as was the case in the Germanic languages.
In other words, whereas Latin made a distinction between arguments and
satellites, Germanic rather discriminated between central arguments (i.e. first
and second argument) and all other possible actants.
Having chosen the latter perspective, Germanic increasingly made use
of the prepositional dative for marking satellites. Without prepositions, the
dative case mostly functioned as a true, argumentai dative40. Therefore, the
Germanic dative took a different way from that of the Latin ablative, in
38 The ablative case, without prepositions, could also convey the meaning of manner
and cause in Latin. The semantic relations of cause, instrument and manner can be distinguished
as follows: a) cause: source of the action, with no implications about the existence of an agent;
b) instrument: the means through which an action is accomplished, with explicit reference to
the existence of an agent; c) manner: the way in which an action develops/is accomplished,
with specific reference to the action itself and to its participants. Manner is rather close to
instrument, in that the way in which an action is accomplished can be conceived as a means
for accomplishing it. Functional merging of markers of instrument/manner with markers of
cause can be explained through the general notion of conditioning factor, which I have discussed
above, section 3.1.6. See also Vester (1983: 46) on the difference between cause, instrument
and manner with regard to controlled or uncontrolled states of affairs in Latin.
39 See Luraghi (in print b).
40 A prepositionless dative functioning as a locative or an instrumental is seldom to be
found in early Germanic languages, see Ramat (1981: 143-144).
368 SILVIA LURAGHI
that, when used without prepositions, it was restored in its originary func
tion. The function of satellites was transferred to prepositional phrases,
within which the dative became the unmarked case 41 .
As a last remark, let me stress that, as I mentioned earlier, the Ger
manic branch of the Indo-European family is the one in which the tendency
to the shift from a synthetic toward an analytic type had the greatest
relevance, at least among the languages surveyed in the present paper.
Therefore, it might not constitute the best item for a study of case syncretism.
4. Conclusions.
In the course of this paper, I have examined functional case syncretism
in four branches of the Indo-European family. I have been arguing that:
a) functional syncretism derives from partial synonymy among cases;
b) synonymy can take its origin either from the semantic or from the syn
tactic overlap of cases;
c) instances of semantic syncretism are brought about by similarities in the
semantic relations underlying noun phrases marked by specific case endings;
d) syntactic syncretism is brought about by similarity in the syntactic posi
tion that noun phrases marked by specific case endings usually take with
respect to the predicate;
e) lexical features, and animacy in particular, play a major role in case syn
cretism, in that they contribute to determining the degree of acceptable am
biguity;
f) syntactic syncretism engenders semantic opacity within case systems and
results in an increasing use of prepositions.
REFERENCES
Serbat, G. 1985. Le syncretisme des cas. Paper read at the 3rd Colloquium
on Latin Linguistics, Bologna, 1-4 April 1985.
Shields, K. 1982. The Indo-European origins of the Old Hittite directive
case. JIES 10. 273-282.
Starke, F. 1977. Die Funktionen der dimensionalen Kasus und Adverbien
im Althethitischen. Wiesbaden: Harrassowitz.
Szemerényi, O. 1985. Introduzione alla linguistica indo-europea. Milano:
UNICOPLI, revised edn. of O. Szemerényi, Einführung in die
vergleichende Sprachwissenschaft, 1980 2nd edn. Darmstadt:
Wissenschaftliche Gesellschaft.
Valentin, P. 1975. Le group prépositionnel allemand en grammaire du
signifié. BSL 70. 253-274.
Vester, E. 1983. Instrumental and manner expressions in Latin. Assen: van
Gorcum.
Wackernagel, J. 1926. Vorlesungen über Syntax, 1st vol. Basel: Birkhäuser.
INTEGRATION OF PHONOSYMBOLISM
WITH OTHER CATEGORIES OF LANGUAGE CHANGE
YAKOV MALKIEL
University of California, Berkeley
1 For a rather detailed critical summary of this earliest synthesis of Sapir's thinking on
generalities consult Malkiel, 1985a. For bibliographic clues and other hints to Sapir's later con
version to phonosymbolic experiments see an earlier piece from my pen (1977a: 511-29) as well
as R. Jakobson's endorsement (1979: 182).
PHONOSYMBOLISM AND CATEGORIES OF CHANGE 377
ray Structure A. Then, in every second paragraph or so, they would list some
annoyingly disobedient, unruly words, which refused to be fitted into the
scheme; Bloomfield's celebrated «residues» (1933).
There is no need to imply that the discussion of these special cases
necessarily appeared in small print, was relegated to a footnote, or was com
pressed into a single sentence of indefinite length. Some virtuoso performers,
e.g., Gottfried Baist, in the two successive editions of his sketch of a historical
grammar of Spanish, managed succinctly to state each regular cor
respondence in a half-line, without at all bothering to illustrate each point,
then devoted the entire remainder of each relevant paragraph to a leisurely
analysis of recalcitrant etymologies (1889: 684-714; 1904-6: 878-915). Despite
this whimsical disproportion even Baist formally recognized the — to him,
dull — law and order.
At rare intervals, a scholar might succeed in recognizing, within such
a sizable residue, a modest amount of lexical equations that would embolden
him, in a moment of triumph, to formulate a new, narrow-scope Lautgesetz
capable of reducing the number of vexatious exceptions at the cost of com
plicating the architecture of Structure A. But a typical scholar of that period
and persuasion would not in his wildest dreams have thought of erecting
rival structures (call them B, C, or D) for those members of the residue (bor
rowed words, lexical blends, Lallwörter, and so on) that could not, by any
stretch of one's imagination, be smoothly squeezed into Structure A: These
«bad boys» were doomed to remain an amorphous pile, a mass of second-
rate material, and an embarrassing reminder of the vaguely-felt inherent
limitations of the method applied.
Antoine Meillet was not an average member of that generation of
scholars. Of course, he knew how to manipulate — imaginatively and with
assurance, at that — the, by then, classical sound correspondences where
needed. In other contexts he, independently, knew how to see and look
beyond. Take his searching review of the memorable monograph by the
Swedish Romanist Ernst G. Wahlgren, which revolved around the transmis
sion, through semilearned conduits, of words like Lat. medicus «doctor,
physician», into Old French, as mire, illustrative of d > , as against (a) the
total loss of ancestral intervocalic -d- in vernacular words (uādit «he walks»
> [il] va) and (b) its faithful preservation in strictly learned lexical items:
mŏdu «manner» > mode. Without actually taking side in the issue, once
hotly debated (for details see Malkiel, 1977b: 69-85), Meillet, a relative out
sider to Romance inquiries, wondered whether such originally semilearned
378 YAKOV MALKIEL
to be the best, having been thoroughly brought up to date (1980); in structuralist reinterpreta
tions of earlier findings (e.g., by E. Alarcos Llorach [1961]); and in monographic investiga
tions, of which those by Amado Alonso (1954: 93-95) happen to be the most circumstantial.
4 This observation does not hold for Arabic ingredients of Hispano-Romance, as was
demonstrated by, among others, A. Steiger (1932: 136-43).
5 To be sure, it is more common to see ancestral -er- transmuted into medieval and/or
modern -ar- (cf. mercåtu > Fr. marché «market»); but, given the transitional state of waver
ing, an occasional move in the opposite direction causes little surprise.
380 YAKOV MALKIEL
coase, It. cucire, Fr. coudre beside couture, Sp. coser, etc. Contamination
with suere would adequately explain the rise of in çurcir, especially if
one remembers that the paradigm of suere did not lack members marked
by ū, such as the p.ptc. sūtus. But what of the word-initial ç? To be sure,
one could charge it to the agency of consonant assimilation alone — i.e.,
to the anticipation, at the start of the word, of word-medial As/. But
the rapid generalization of ç in this word, in lieu of expected s-, clamors
for richer motivation. The social disfavor in which darning — the im
pecunious person's escape from a minor predicament — has been held
might have produced the climate of contempt or disgust which favored ç-
over s-.
The words to be mentioned now have in common three features: an
involved biography, a controversial etymology, and a predominantly un
complimentary meaning. Zafio «rough, coarse, crude, uncouth», orig. çafio
(. 1500), is an item of Moorish background and indeed may represent a
blend of two Arabic adjectives, even though safîh suffices to account for
it; the change in stress and the choice of the word-final -io segment are no
insuperable hindrances 6 . Çafio teaches us where to look for a plausible
source of word-initial As/, with Latin falling short of offering any suitable
model. The rendering of Ar. s- by ç was normal, and the association of Moors
with preminently negative features, in the few lexical units of that ilk that
involve value judgments, causes little surprise. Our second witness is sutano,
older çutano «so-and-so», typically used not in isolation, but as the last
member of a series: Fulano, Mengano y Zutano («X,Y, and Z»). The word
may be a straight Arabism, or a Romano-Oriental hybrid; noteworthy is
the fact that the older form, çutano, was used in keen rivalry with ci-tano,
-trano, -crano (and with sicrano, seclano in Portuguese), to say nothing of
Perengano7. The two outstanding features of zutano's eventual triumph
over its rivals are: the speakers' preference for As/ over / s / thus again
manifested, and their keener pleasure at having in this context pretonic
rather than some front vowel.
Finally, zurdo, anciently çurdo «left (-handed)», an unflattering
characteristic according to the old scale of values, has been hazily associated
6 The Arabic intruder could easily have been sucked into the maelstrom of Hispanic -io
< -ǐdu formations, which among adjectives outnumbered those in -ίο.
7 The most noteworthy etymological conjectures, including those by A. Steiger and L.
Spitzer, have been summarized by Corominas (1957: 893), who gravitates toward the latter's view.
PHONOSYMBOLISM AND CATEGORIES OF CHANGE 381
with ill-defined pre-Roman languages, and with Basque; the wavering be
tween -rr- and -rd- will cause little surprise among those familiar with the
vicissitudes of the words for «left»: izquierdo ~ esquerro, and a certain
affinity to lerdo «slow, sluggish, dull», again a word of non-Latin ancestry,
makes itself felt. Even though competing forms such as Gal. xurdo, Ptg.
churdo, and even churro ~ churdo, all of them endowed with some such
vaguely negative meaning as «dirty, vile, base», plus Bearn, (i.e., Gascon)
sourrou «stingy», were available in abundance, Spanish speakers, significant
ly, selected not / s / , nor / š / , nor indeed /tš/, but — once again — solely
As/, later /Θ/, followed by ás their favorite device for carrying such a
thoroughly unpleasant meaning.
One more point can be appropriately made here before we take up the
next situation: Speakers visibly — and audibly — enjoy doubling and trebl
ing phonosymbolic effects. For instance: repetition of a syllable-initial con
sonant in consecutive syllables is, as is yet to be shown in detail, a powerful
phonosymbolic device almost amounting to a language universal and peculiar
also to children's talk. The vowel may but need not be repeated, as shown
by It. chiacchiera «gossip». With these bits of information, consider the
following triad of Spanish words: zozobra «worry, anxiety», lit. «capsiz
ing, foundering, sinking»; zozobrar «to be in jeopardy, to worry, fret», lit.
«to capsize, founder, sink», used also as a transitive verb: «to wreck, sink»
(a business); finally, the adj. zozobroso means «worried, anxious, restless».
Etymologists apprise us of the Catalan origin of this Castilian word, a con
jecture perfectly compatible with the maritime (or navigational) frame of
reference. The Catalan etymon is the compound sots «down» (from Lat.
subtus) plus sobre «up»; so the more widespread order of components in
the cross-linguistic binomial «up and down» (cf. R. vverx i vnix, G. herauf
und herunter, etc.) has here, for a change, been inverted. The Catalan verb,
sotsobrar, was and has remained a technical term; its bold semantic expan
sion in Castilian remains unexplained. While the rendering of word - medial
Cat. -ts- by Sp. -ç- is understandable, the change of OSp. soçobra into
çoçobra (eventually spelled zozobra) on the assumption of consonant
assimilation may well be one factor, but need not represent the whole story.
It is advantageous to bracket (a) the echoing effect of consonant repetition,
(b) the generalization of the ç - in preference to the s-; (c) the presence of
a back-vowel (if not of u, then at least of o); and (d) the bold projection
of the meaning and, as a result, of the word's actual use far beyond the
382 YAKOV MALKIEL
8 I am adopting here the corpus established by Corominas, who is at his best in handl
ing material of Catalan background (1957: 245a). Rodíguez-Marín (1923: 23-50) provides in
sight into genuinely Castilian altibaxar and baxialto.
9 Conversely, if we adopt the var. strambus as our starting point (a hypothesis favored
by Meyer-Lübke and, later, endorsed by Corominas), we simply push back into Antiquity the
addition of the intercalary nasal; as also happens if we opt for scambus, in harmony with Diez's
preference. See Corominas (1957: 816b - 818b).
PHONOSYMBOLISM AND CATEGORIES OF CHANGE 383
derivational suffixes 10.) Take Sp. aguaza «aqueous humor, sap», transpa
rently from ancestral agua «water». One widely used derivative, aguazal,
is best rendered by «pool, puddle»; another, even more important, aguacero,
is tantamount to «(heavy) shower». There are further in existence one ad
jective, aguazoso (alternating with aguanoso) «wet, watery, soaked»; plus
a triad of verbs, one basic: aguazar «to make marshy, to flood», refl. «to
become marshy», and two secondary: enaguazar «to flood» and desaguazar
«to drain, empty». Aside from the controversial side-issue of aguanoso,
which need not detain us today, the situation, as seen from a language
historian's vantage point, appears relatively simple: Either he agrees to
operate with well-documented Lat. aquātiō «fetching of water, water(ing),
rains, watering-place», a decision made easy by the independent existence
of Arag. aguazón, which audibly echoes the oblique case aquātiōne; or he
musters the courage to reconstruct a (plausible, but unrecorded) adjective,
to wit, *aquāceu(s); or else he assumes a blend of the two, not improbable,
since either was pronounced /akwatsju/ in most provinces of the Empire.
So far, so good; it now happens that Spanish also boasts a rival
derivative, namely aguacha «foul, stagnant water» (note the starkly
derogatory overtones, absent from the semantic gamut of «neutral» aguaza).
From the noun aguacha, there must have branched off the verb aguachar
«to flood», distinctly less common than enaguachar «to spill water over,
soak with water, flood», «to upset the stomach of»; refl, «to become upset»
(from excessive eating or drinking). A more messy tableau is hardly con
ceivable. As a noun, aguachar, paralleling abovementioned aguazal, is tan
tamount to «pool, puddle». One could go on and on; there is no dearth
of other lexical items, in part confined to dialect speech, like the self-
explanatory adj. aguachento, or the verbs aguachinar and enguachinar, of
which the latter presupposes an apheresized stem variant guach-, etc. But
how can the relation of aguaza to aguacha be described — not loosely, but
in a technically cogent manner?
On this point, the authorities are singularly evasive, whether they stage
their attacks on the elusive problem from the citadel of phonology or from
10 This confusing tangle of issues has become conventionally known under the tag of
«Old Spanish sibilants», ever since J.D.M. Ford's Harvard dissertation, submitted in the late
'nineties. For one provisional attempt to cut a swath through this jungle see Malkiel (1971:
1-52). A major consideration to be added to the earlier inquiry is the fact that, among the
erratically-developed suffixes, -eza straddled Castilian and Galician-Portuguese, whereas OPtg.
-açom, -aço, and -iço clashed with OSp. -azón, -azo, and -izo (plus analogues and variants).
384 YAKOV MALKIEL
11 Even though, in all too numerous instances, Corominas, over the years, has overin
dulged the assumption of transmission of Latin lexical units through Mozarabic (cf. his remarks
on zambo, 1957: 816b), he controlled this temptation in his long article on agua (1954: 57-58b).
In analyzing (coll.) guachapear «to splash with the feet», «to botch, bungle», (fig.) «to clat
ter, clank», he went to the opposite extreme of positing a blend of (a)guacha with chapotear
(vars. chapalear, chapullar) intr. «to splash», tr. «to moisten, sponge», which he traced to
an onomatopoeic root chap-. So presented, the explanation is misleading, because it leaves
out of our reckoning the essential -pear element traceable to pie «foot» and painstakingly in
vestigated first by myself (1954:1-22, 65-95) and later by H. Bursch in his Bonn dissertation
directed by H. Meier (1978). Despite this qualification, Corominas' comment, if felicitous,
throws interesting light on word blends in phonosymbolism (or, strictly, morphosymbolism).
Take aguachirle «cheap wine; slosh, slipslop; trifle», a compound easily decomposable into
agua «water» and (coll.) chirle «tasteless, insipid» (also «sheep manure»); it is best understood
on the assumption of its overlap with aguacha.
PHONOSYMBOLISM AND CATEGORIES OF CHANGE 385
12 The number of relevant examples can be easily increased; contrast Sp. hambre, orig.
fambre «hunger», from fame (via *famine), with Ptg. fome; or costumbre «habit», from
cōns(u)etūdine (on the assumption of a blend with the suffix -ūmen), with Ptg. costume. I
have elsewhere sought to show that the change in the paradigm of the inchoative verbs from
-esco to -eço could have been conditioned by the disinclination of Lusophones to avail themselves
of forms reminiscent of the (characteristically Castilian) velar insert.
PHONOSYMBOLISM AND CATEGORIES OF CHANGE 387
13
One is tempted to cite enx- from ex- contaminated with an anticipatory nasal; or tres-,
a merger of tras- and es- with some additional influence from tres- «three», «thrice» (echoing
bes-, biz-, «twice») for good measure.
14 C. Michalis', the future Sra. de Vasconcelos', etymological notes here appealed to
(1873: 86-91) served, along with C. Joret's 1874 monograph and with A. Darmesteter's lengthy
critical reaction to it, as the three starting points for the protracted discussion of the Paleo-
Romance sibilants.
15 In this tabulation so- and its variants echo Lat. sub- either actually documented or
apt to be projected onto the level of Late Antiquity and the early Middle Ages.
388 YAKOV MALKIEL
16 There would be little point in delving here into C. Michalis' errors of which the most
serious was her attempt to bracket doublets due to borrowings from Old French (chanzón
beside canción, chanciller beside canciller) with the other instances of proliferation. Some of
her etymologies have also proved to be untenable, e.g., the derivation of zurdo from sǔrdus
«deaf».
PHONOSYMBOLISM AND CATEGORIES OF CHANGE 389
posais 17 ; but progress, on the whole, has been distinctly slower than might
have been predicted — precisely as a consequence of certain scholars' reluc
tance or refusal to systematize our grasp of phonosymboiically orchestrated
sound change. If anyone entertains doubts about the validity of the special
shift «word-initial s- ) ch-», because it is alien to normally-developed words,
on the order of sagīmen ) saín «grease, fat», sella > siella (later silla) «sad
dle, stool», sex > seis «six», sĭccu )seco «dry», soc(e)ru ) suegro «father-
in-law», südat > Sp. suda alongside Ptg. sua «he sweats», let such a skeptic
focus his attention on a formation with downright onomatopoeic implica
tions, whose phonosymbolic effectiveness no one would want to question.
Take Lat. sibilare «to hiss», with a by-form sīfiläre which can be, and often
has been, explained as a borrowing from some outside source, including
Italic, except that the unique raison d'être of such a borrowing, in the absence
of any cultural advantage, can indeed have been the aforecited factor of
effectiveness. From sībi-, sîfi-lāre Spanish has inherited three verbs: first,
silbar «to whistle»; second, chiflar «to hiss, gulp down», «to whistle, sing»
(said of birds), plus, reflexively and only in the familiar register, «to become
unbalanced», witness chifladura «craze»; and, third, chillar «to shriek,
screech, creak, squeak, hiss, scream» (figuratively said of colors); (in New
World dialects) «to squeal», i.e., «to turn informer»; among hunters, «to
utter a fox call». Observe that the second and the third products are given
away as phonosymboiically colored not only by the representation of
ancestral s-, before the high front vowel, via ch- / c / , but also by a bizarre
rendition of the ancestral word-medial -bil- or -fil- sequence through /fl/
or /λ/, counter to expectation; in contrast, the -lb- cluster of silbar involves
an entirely innocuous metathesis of syncopated parental -b'l-. Silbar is, prac
tically, «lautgesetzlich»; nothing of the sort can be asserted of either chiflar
or chillar. But this is not all: The semantic development of silbar has been
slow and imperceptibly slight; conversely, those of chiflar and chillar have
been saltatory, to the point of wildness — in unfettered substandard and
dialect speech even more so than in relatively polished literary Spanish.
If now a critic unhappy about encountering this state of affairs wished
to extricate himself from any ideologically embarrassing commitment to the
17
Disappointing, despite a few valuable lexical data and bibliographic clues, is the treat
ment of the issue by Menéndez Pidal (1941: § 37-2bc), who recognizes as causative factors
behind ç ~ ch, etc. solely consonant dissimilation and dialectal provenience of certain lexical
items, without being in a position to identify the dialect at fault.
390 YAKOV MALKIEL
assumption of common origin for silbar, chiflar, and chillar, the strongest
probabilistic rebuttal of any such denial is the astonishing degree of
parallelism between the derivational offshoots from the three verbs at issue:
silbato «whistle, whistling crack» is echoed by chiflato; silbido «whistling,
whistle, hiss» boasts even two counterparts, namely chiflido «whistle, whistl
ing sound», and chillido «shriek, scream». Identity of suffix despite strong
differentiation of meaning can be demonstrated with silbón «name of a bird»
(«widgeon»), chiflón (in New World dialects) «cold blast of air», «rapids»,
«slide of loose stones» (in a mine). A masculine postverbal in -o is also shared
by the three verbs: silbo «whistle, hiss, whiz», chiflo «whistle» (beside more
widely used chifle, also more richly nuanced on the side of meaning), and
chillo (hunter's) «fox call» or «hare call». A parallel series of feminine
postverbals has a wider semantic range: silba «hiss(ing)», chifla «whistle,
whistling, hissing» (accompanied by the New World phrase estar de chifla
«to be in a bad humor»), and chilla «fox call, hare call»; etc. 18 .
Even this superficial inspection of the record of sibilare in Spanish suf
fices to bring home to the beholder the involvement of phonosymbolism
in the intricate issue of channels of transmission. For almost exactly a cen
tury and a quarter — ever since the appearance of Auguste Brachet's slim
book on the subject of French doublets and of the supplement to that pam
phlet — Romance scholars have intermittently concerned themselves with
the study of doublets and triplets (or, to put it in Italian, of allòtropi), the
two peaks of this once highly promising line of curiosity being C. Michaelis'
mistitled 1876 book on Spanish doublets and, shortly thereafter, U.A.
Canello's monograph-length counterpart dealing with doublets in Italian 19.
18
Because chiflar and chillar were each allowed to develop a number of secondary,
idiosyncratic meanings, it stands to reason that each of the three verbs here under scrutiny
should, at least, have a few suffixal derivatives peculiar only to itself. If one can bracket, by
stretching one's imagination, silbador(a) «whistler, hisser», (b) «whistling, hissing» with chillador
«shrieking, squeaking, screaming», thorough isolation seems to prevail in each of the follow
ing cases: silb-oso (adj.) «whistling»; chifi-adera and chifl-ete «whistle» (on chifladura see above);
and chill-ería «shrieks, screams, outcries»; «screaming and scolding». The situation is, actual
ly,even more complicated, because speakers have isolated the opening syllable chi- as the one
most heavily endowed with phonosymbolic substance and experimented with its repetition:
chichear «to hiss», chicheo «hissing». Furthermore, the colloquialisms chufleta «joke, jest»,
chufletear «to joke», and chufletero «joker, jester» appear to involve vocalic variation (i/u)
on chiflar which, being a joke of sorts, revolves around the idea of «jest».
19 My own sketchy statements on this, now half-forgotten, frontier of knowledge (1973b:
749-51; 1977c: 191-6) are in urgent need of drastic revision and expansion.
PHONOSYMBOLISM AND CATEGORIES OF CHANGE 391
20 I prefer not to overreach myself by stringing the many infelicities that, unfortunate
ly, mar Menéndez Pidal's treatment of this elusive problem in four passages of his oft-revised
textbook of historical grammar (1941: §§37-2, 72-2, 126-2, and 128-4). His comments on the
plausibility of certain shrewd theories advanced by other scholars: A. Castro, A.M. Espinosa-
padre beside that scholar's son, V. García de Diego, J. Jud, Fr. Krüger, H. Schuchardt, A.
Thomas, M. de Unamuno, and others invariably deserve attention, even though none seems
to have thoroughly convinced him. The author complicated his task by interpreting too broad
ly — after the fashion of the Madrid School — what it chose to call «acoustic equivalence»
and by dragging into the discussion the partial interchange of / s / and / š / , later / x / , which
may involve a separate phenomenon, of distinctly older vintage. Particularly unfortunate, in
my view, is the attribution of the nasal infix to the pressure of cum, in, nōn.
PHONOSYMBOLISM AND CATEGORIES OF CHANGE 393
the shift ) were the first regular sound changes undergone by the word
after its initial Latinization; the shape it then assumed must have been
*sompoña. The vowel dissimilation - ó > a - ó it underwent next pertains
to the jurisdiction of «general accidents»; and only the last brush stroke,
namely the change of \*sa- to za-, requires the separate assumption of a
phonosymbolic process.
If, at the end of the road, one is asked whether any independent proof
of phonosymbolic action is available, the answer could well be that cir
cumstantial evidence to that effect indeed exists. That evidence includes:
(a) the broad involvement of the realm of music and sound; (b) the inor
dinately lively semantic development within colloquial Spanish; (c) the ex
istence of independent phraseological parallels in languages only remotely
kindred, like English (fiddlesticks!) and Russian (dudki!), quite apart from
the fact that both the English compound and the Russian diminutive (from
dudit' «to sip through a straw»?) have certain constituents in their make-
up that also, once more independently, smack of phonosymbolism 21 .
In similar, though not strictly identical fashion, one could, step by step,
analyse the metamorphosis of ancestral sepelire «to bury», after its local
blend with subbulhre «to foam or bubble slightly» into reflexive za(m)bullir
«to dive, duck, plunge» — at present a popular term of competitive athletics,
but once used perhaps of crabs, or small fish, or ducks who, by diving, tem
porarily bury themselves under water and, in so doing, emit or release some
bubbles at the surface. This time, the medieval midway form sobollir is,
fortunately, extant. Again, only a minority of intermediate steps can be writ
ten off as phonosymbolic; specifically, the intercalation of the nasal, which
tends to add a semi-comic touch, plus the change of s- to ç-, later z-. The
remainder of the changes, though fairly heterogeneous this time, require
no such additional assumption.
Fresh insights into the working of phonosymbolism in a whole cluster
of interrelated languages are simultaneously provided by the Romance
chapter in the involved biography of ciccum — clearly, an old rustic word
whose original referential meaning hardly emerges with any degree of ac
curacy from the mutually contradictory definitions furnished by ancient
21 For additional documentation and analysis of certain crucial points see, among my
own earlier papers, above ail, (1949: 183-232) and (1984: 27-46).
394 YAKOV MALKIEL
22 I am here falling back on the information provided by Ernout and Meillet (1959-60:
119a) as well as Walde and Hofmann (1938: 211), in addition to standard dictionaries.
23 Françoise Skoda's recent monograph (1982) — cross-linguistic even though slanted
in the direction of Ancient Greek — and weighty critical reactions to it, on the order of Jean-
Louis Perpillou's constructive review (1984: 158-62) — are only the latest links in a long chain
of investigations, which included book-length studies from the pen of J. André (1978) and,
so far as consonant gemination rather than reduplication is concerned, from those of A. Graur
(1929) and A. Martinet (1937). For a bird's-eye view of my own thinking (until recently) see
Malkiel (1985b: 1-25).
PHONOSYMBOLISM AND CATEGORIES OF CHANGE 395
its provenience 24 , was, from the start, literally cut out for phonosymbolic
effects.
To the Latinist, ciccu had a short nuclear i, much as floccus at the start
had, and later kept, its short o 25 . But, at least, some Romanists would be
happiest if allowed to assume a folkspeech change of ciccu into *cïccu, since
J, in contrast to ĕ, was a vowel exceptionally resistant (if not entirely im
mune) to shifts, and since a measurable advantage accrued from the preser
vation of the high front vowel in a word suggestive of «smallness» or «next-
to-nothingness», if we may here appeal to what can be facetiously called
«Jespersen's Law». Should critics tag our assumption as arbitrary, it can
be shown that alternative solutions invoked in certain concrete cases yet to
be mentioned border on the fantastic.
We are now almost off on our cruise aboard the boat of phonosym-
bolism — one more brief comment is needed before we can weigh anchor.
A disyllabic word in Latin with a lengthened consonant (not infrequently
a -cc- or a -tt-, but also an -ss-, rarely a -bb-) as its midpoint was very fre
quently (although, to be sure, by no means exclusively) apt to be an adjec
tive, fairly often one qualifying or ridiculing the physique of an individual.
Characteristic examples of a less than flattering description include: crassus
«thick, fat» and (late) grossus «thick» (which were doomed to merging on
a regional scale, witness Fr. gras «fat»), as well as their semantic opposite
flaccus «flabby», plus, to cite a near-opposite, *plattus «flat», a transparent
adaptation of Gr. platys , whose arrival had been prepared by preëx
istent planus «level, even», as shown by Fr. plat, It. piatto, Sp. chato, against
the background of (obs.) plain, piano, llano. Still within the same sphere
one clearly recognizes gibber and gibbus «hunchbacked», with a solidly
reconstructed quasi-apophonic by-form *gubbus, known from It. gobbo,
sgobbare and from Sp. agobiar26. Neither It. brutto «ugly» nor (O)Fr.
brut, -e «brute, raw» can be smoothly traced to recorded brütus, as numerous
reputable scholars over the years have attempted to do, with or without ap
peal to all sorts of «tricks»; only by positing the existence of the by-form
*brüttus does one, at one blow, remove all difficulties from his path 27 . The
list of other primary adjectives displaying this configuration includes: cassus
«empty, vain»; cossus «with a wrinkled skin»; mattus «drunk»; and siccus
«dry» 28 .
Under this set of circumstances it no longer seems too bold to argue that
cïccu, better still, *cïccu sounded so suggestive to speakers as to have easily
lent itself to all sorts of reinterpretations. OSp. chico was from the start
a far more intimate word than pequeño, designating the «tot», the «Knirps»,
the «maljutka» in the family, by virtue of the consensus of such potentially
phonosymbolic features as (a) word-initial ch- in lieu of ç-; (b) the high rather
than the mid-high nuclear vowel; (c) the intervocalic surd, echoing a lengthen
ed intervocalic consonant in the parent language; and (d) disyllabicity. But
Fr. chiche «cheap» (i.e., essentially, «small») can also have descended from
*cicca, with subsequent generalization of the feminine form, if one allows
for the reduplicative effect of *ciche ) chiche, a process that would make
sense in this particular context. While these conjectures, by definition, are
hypothetical, I submit that they are less hazardous, in narrowly linguistic
and broadly cultural terms, than several previously advanced suppositions.
We have, somewhat haphazardly, touched upon the issue of lexical bor
rowings here, actually a whole web of problems falling under the rubric of
cultural diffusion. Let me try to establish one important dichotomy at this
eccessiva»; while Bruno Migliorini, toiling in tandem with Aldo Duro, was
distinctly more cautious in his wording: «Prob[abile] adattamento antico
del lat. brütus «bruto» (1964: 74b) 29 . I am willing to go off in an entirely
different direction, arguing that the by-form *brüttus, with phonosymbolic
lengthening of the word-medial consonant, should be traced all the way to
Antiquity, on the strength of OFr. brut, brute, in lieu of expected bru, *brue.
An advantage of this hypothesis is that it also enables us to grasp It. tutto,
-a, OFr. tot, -e (mod. tout, -e) «all, whole», forms which contrast dramatical
ly with Sp. Ptg. todo, -a, a faithful echo of ancestral tōtu, -a. Since the
Latinity of Luso- and Hispano-Romance is notoriously more archaic than
that of Italy, the failure of the innovation *tõttus to have reached it causes
no surprise 30 . On the semantic front, the qualifications of brütus for suc
cumbing to the influence of phonosymbolic gemination are too obvious to
invite comment. As regards the expression of «totality» 31 , it suffices to
observe the lively gesticulation normally accompanying the categorial state
ment: Tutti! Tutte! among spontaneous speakers of Italian to realize what
emotions any exciting affirmation of totality can release. In this instance,
then, phonosymbolism has indeed been at work, but there is no point in
assuming any dialect borrowing as a triggering force.
The situation is wholly different in Central Europe. It is a truism that
word-initial d- in Low German, no less than in Flemish-Dutch and in English,
29 The idea of *brüttu here championed is not entirely new; it was pioneered, without
any attempt at justification, by C. Merlo (1941), then more carefully advanced by M. Regula
(1965:740), who, rather pointlessly, started out from *brütidu, allegedly patterned on hispidus,
horridus, etc. As if to complicate matters, P. Fiorelli (1967: 23) traced the adverbial phrase
di brutto «unexpectedly» to ex abruptō. Cortelazzo and Zolli (1979: 172a) list all three hypotheses
without taking side. Clearly, if one were to take Regula's proposal seriously, it might become
unavoidable to allow for *tõtidī as the starting point for tutti!
30 There exists a considerable body of literature on the innovative type / t o t : u / ; prac
tically all of the conjectures played with by C . Nigra, E . Richter, V. García de Diego were
indignantly rejected by an impatient Meyer-Liibke (1930-35: §8815), w h o in the end turned
against his o w n earlier thinking a n d halfheartedly accepted only Georg Ebeling's identifica
tion of nūllus as a possible model. Meyer-Lübke's private preference was for «eine A r t D o p
pelung», *tototus, for which he cited OIt. tututto a n d OSard. (Logudoro) totta(t)ta as witnesses,
sailing in the wake of G. G r ö b e r . F o r e a r m e d with the assumption of p h o n o s y m b o l i s m , one
posits, as the first step, plain consonant lenghtening and, as the second step, additional repeti
tion of the key syllable, perhaps via reduplicative tutti, tutti.
31
It m a y in this context prove rewarding to collate in detail the findings of earlier
Romanists with the fresh insights that an inspired E d w a r d Sapir gained in his 1930 m o n o g r a p h
on totality.
PHONOSYMBOLISM AND CATEGORIES OF CHANGE 399
32 It is arguable that, once dumm and dunkel had been accepted and semantically ad
justed, they paved the way for dumpf(ig, -icht) «damp, stifling, hollow, gloomy», not used
as an adjective before the New High German stage; it is in apophonic relationship to Dampf
«steam» (Kluge and Mitzka, 1963: 147b). If the prototype *dhumbos reconstructed with the
help of Goth, dumbs involved a variant of an ancient chromonym marked by a wedged-in
nasal, another manifestation of the phonosymbolic effect of such inserts may here have been
at issue. Without questioning the derivation of Dudelsack «bagpipes» from Turk, duduk «flute»,
one is tempted to single out the built-in phonosymbolic effectiveness of that word (and of G.
dudeln «to play the bagpipes», Dudelei «piping») as a factor that may have preconditioned
the borrowing process. Berlinese duft(e) «good, okay» has been traced to Yidd. tofl (from
H. tob); the leap from t- to d- may be credited to an allusion to Duft «smell, aroma» (Kluge
and Mitzka, 1963: 146a); however, that association — in itself, slightly far-fetched — might
not have jelled were it not for surrounding local cases, such as toll > doll.
400 YAKOV MALKIEL
REFERENCES
-----. 1946a. La parola come testimone della storia. Napoli: Raffaele Pironti.
-----. 1946b. La parola come mezzo d'espressione. Napoli: R. Pironti.
1950. Colonizzazioni nell'antico Mediterraneo occidentale alla luce
degli aspetti linguistici. Napoli: Libreria Editrice Liguori.
-----. 1951. Grammatica storica della lingua francese; aspetti e problemi.
Napoli: Liguori.
Bloomfield, Leonard. 1922. Review of Jespersen (1922). American Journal
of Philology 43.370-3.
-----. 1927. Review of Jespersen (1924). Journal of English and Germanic
Philology 26.444-6.
-----. 1933. Language. New York: Henry Holt & Co.
Boas, Franz. 1911. Introduction (1-83) to Part 1 of Handbook of American
Indian Languages. Washington: Government Printing Office (= Smithso
nian Institution, Bureau of American Ethnology, Bulletin 40).
Brachet, Auguste. 1868. Dictionnaire des doublets, ou doubles formes de
la langue française. Paris: A. Franck.
-----. 1871. Supplément. Paris: A. Franck.
Bursch, Horst. 1978. Die lateinisch-romanische Wortfamilie von
*INTERPEDARE und seinen Parallelbildungen. Romanistische Versuche
und Vorabeiten 52. Bonn: Romanisches Seminar der Universität.
Canello, U.A. 1878. Gli allotropi italiani. Archivio glottologico italiano 3.
285-419.
Castro, Americo. 1914. Review of Hanssen (1913). Revista de Filología
Española 1. 97-103, 181-4, at 102.
Corominas, Joan. [1954-57]. Diccionario crítico-etimológico de la lengua
castellana. 4 vols. Madrid: Gredos, & Bern: Francke.
Cortelazzo, Manlio, & Paolo Zolli. 1979. Dizionario etimologico della lingua
italiana 1. Bologna: Zanichelli.
Darmesteter, Arsène, 1874. Review of Joret (1874), Romania 3. 379-98.
Devoto, Giacomo. 1966. Avviamento alla etimologia italiana. Dizionario
etimologico. Firenze: Felice Le Monnier.
Ernout, Α., and A. Meillet. 1959-60. Dictionnaire étymologique de la langue
latine; histoire des mots. Rev. 4th edn. Paris: Klincksieck.
Espinosa, Aurelio Macedonio. 1930. Estudios sobre el español de Nuevo
Méjico, I, tr. A. Alonso & Á. Rosenblat. Biblioteca de dialectología
hispano-americana 1. Buenos Aires: Instituto de Filología (at 182).
402 YAKOV MALKIEL
DERRY L. MALSCH
University of Oregon
Table 1
Their model assumes three, possibly cyclic, stages that characterize three
different relationships that hold between form and use. In the first, purely
semantic, stage, form and meaning completly determine usage. Using pro
noun systems as an example, this stage would characterize a system in which
singular number indicates only a singular referent, and where plural number
indicates only a plural referent. It would also characterize a system in which
first person refers only to speaker, second person only to addressee, and
third person only to bystander or non-participant in the speech event. In
the second stage an implicature develops relating structurally determined
usage to pragmatic usage. Again, exemplifying this scheme with reference
to pronoun systems, an implicature from usage 1 to usage 2 would allow
the use of a plural pronoun to refer to a singular addressee with respect.
Usage 2 could in turn have an effect on form. Just such a feedback loop
410 DERRY L. MALSCH
I. IL III. IV.
Proto-T. Old T. Inter. T. Modern T.
Singular niin nii nii nii
Singular + nii + yir niir niir or niinkaL
Respect
Plural niim nii + yir niim + kaL niinkaL
Table 2
The plural form of Old Tamil is a realization of stage 2 from table 1 : nii + yir
indicates both semantic plurality (usage 1) and pragmatic respect (usage 2)
in singular reference. But in Intermediate Tamil the historical reflex of niiyir
is niir and has become an honorific. All sense of plurality is lost and plural
reference is now marked by a new form in the paradigm (actually the original
plural form niim with added plural suffix kaL). The present state of affairs
in Tamil shows a repetition of stage 2: the (new) semantic plural of In
termediate Tamil has again been pragmatically extended to indicate social
relationship.
Given this framework, the question remains: what in fact is the nature
of the implicature that enables usage 1 to be extended to usage 2? One
possibility is the generalization of the notion of power as defined by Brown
and Gilman. If power were the relevant social variable defining the im
plicature, then there should be a correlation between societies stratified by
a power differential and the use of plural pronouns to show respect to a
power superior. There are in fact a number of languages which have
developed pronouns of respect on the base of plural forms within the con-
THE GRAMMATICALIZATION OF SOCIAL RELATIONSHIP 411
Table 3
Table 4
1978) in which the verbal prefix on-, meaning physical motion away from
the speaker, is used to express deference.
In addition to plurality, third person forms, or the use of deictic
markers, many languages, including some familiar ones like German and
French, as well as others like Slovakian utilize an impersonal construction
to express deference. Thus in Slovakian one would say Kam sa ide? «where
is one going» rather than Kam idešl «where are you going» in respectful
questioning. And in Sinhalese, plurality is not used at all to express deference;
rather, the demonstrative pronoun mee is used as a prefix on the second
person singular pronoun. Like the other linguistic devices discussed above
this serves to remove the addressee from the usual role as participant in the
interaction and to reconstruct him or her as a part of its context.
Moreover, in a number of languages pragmatic extension to plurality
and/or third person pronominal reference is insufficient to express the re
quired deference to an addressee. That is, the use of pronouns themselves
may be seen as tying the addressee too closely to the interaction. In such
circumstances, names are often substituted for pronouns, and where there
is a distinction between surname and given name it is the surname that takes
on the function of a social deictic. While names can be viewed as transfor
ming the addressee into a non-participating referent, in a manner parallel
to the use of third person pronouns, the substitution of an individual's name
for second person forms can still undermine the addresse's right to nonin-
volvement by focusing on his or her individuality. It is because of this that
names themselves are often proscribed in deferential interaction. Thus in
languages like Djapu, cited above, personal names cannot be used either
in address or reference when an avoidance relationship obtains. And in
Bengali and Tamil, one of the ways in which a wife shows deference to her
husband is never to use his given name. As a result there is often further
symbolic displacement of an addressee by the use of titles and/or kin terms.
This serves to protect the autonomy of the addressee by placing him or her
within well-defined social roles that bring to the speech event well-defined
constraints on the behavior of the speaker.
Kin terms particularly are used in many languages as both vocatives
and substitutions for pronouns, particularly when age is a salient aspect of
one's social identity. Thus, deference may be shown by addressing any in
dividual as old as or older than the speaker as either mother or father, as
in Yoruba and Hopi; as elder brother or elder sister, as in Turkish; or as
THE GRAMMATICALIZATION OF SOCIAL RELATIONSHIP 417
REFERENCES
MARIA MANOLIU-MANEA
University of California. Davis.
1
Roughly speaking, the difference between conversational and conventional implicatures
may be stated as follows:
Conversational implicatures are cancellable and nondetachable. Let us consider the ut
terance (i) John has three books, which, by the Gricean's «cooperative principle» has
the implicature (ii) «John has only three books», (ii) is suspendable by mention in a
IF-clause, as shown by (iii) John has three books, if not more. Now, let us consider
the ironic interpretation of (iv) Peter is a genius, i.e. (v): «Peter is an idiot». Any
equivalent of (iv) performed in a context in which it is mutually known that (iv) is very
definitely false will share the ironic reading (5): see (vi) Peter is a great brain (or) (vii)
Peter is a mental prodigy, which means that the conversational implicature (v) is not
detachable.
420 MARIA MANOLIU-MANEA
The difference between «same» and «self» can then be described in terms
of the variables linked by the «unexpected identity»: «same» carries an ex
pectation concerning the relation between two variables, χ and y, while «self»
expresses «identity with oneself» (see Coseriu, 1955), i.e. a relation between
the values of the same variable; viz. the value taken by a certain variable
does not correspond to the expected value (for more details see Manoliu,
1983). In less cryptic terms, one may state the difference under discussion
as follows: «same» expresses «repetition», while «self» expresses «coin
cidence». As means of denial, the pronouns of identity belong to a special
category of «insinuating morphemes» which cannot be defined semantical-
ly without considering their pragmatic value, i.e. the conventional implicature
they carry.
2 AIA is the colloquial form for aceea «that» - feminine singular or «neutre», i.e. the
form having the function of a pro-phrase.
3 TOT may also be a sentence connective, expressing a special type of contextual
equivalence, i.e. it introduces the denial of an expected event as a reason for introducing another
event, as a normal consequence, if not as an alternate of the denied event, in a chain of conse
quences, such as «if non Ρ (because of Q), then Z»:
see, for example, Hai sã ne uităm la telewizor, TOT nu mai putem pleca pe ploaia asta...
«let's watch TV, since we cannot go out any longer because of the rain».
422 MARIA MANOLIU-MANEA
In (7) the focus is on the identity between the room in which Peter lives
and the room occupied by John. In (8) TOT emphasizes the temporal con
tinuity, denying the expectation (9) «Peter and John no longer live there».
In other words, (7) and (8) are different as to the scope of negation: in (7)
it is the relation between two arguments which falls into the scope of nega
tion, while in (8) the predicate is in the scope of negation, i.e. «it is not true
that living in room z at time η-l is no longer true for time n». In short,
TOT... ACELA asserts the continuity of the process, by denying the con
ventional implicature of an expected discontinuity, while ACELAŞI asserts
an identity of arguments, by denying an expected non-identity. Situational-
ly, the continuity of processes (or events) entails the permanence of the ob
jects involved (in our case, Peter, John and the room where they live). In
other terms, contextually, the conventional implicature conveyed by TOT
entails the conventional implicature conveyed by ACELAŞI. This entailment
is thus a conversational implicature which has favored a change in the seman
tic content of TOT in such a way that it may convey the same conventional
implicature as ACELAŞI. The conversational meaning of the adverbial TOT
in contexts such as (4), where the NP is a predicative noun, becomes a value
of its conventional meaning.
As a means asserting a «permanence», TOT has various advantages
over ACELAŞI:
(a) Unlike ACELAŞI, which can modify only nouns and behaves like an
adjective, TOT can focus on various constituents:
(i) personal pronouns: e.g. (10) vorbeşte tot el «it is again he who
speaks», and not (11) *vorbeşte acelaşi el, lit. «speaks the same he»;
ROMANIAN PRONOUNS OF IDENTITY 423
(ii) adverbial phrases: e.g. (12) tot acolo «in the same place» and not
*acelasi acolo «the same there», (13) tot atunci «at the same moment»
and not *acelaşi atunci, lit. «the same then», etc.
(in) prepositional phrases: (14) vorbeşte tot cu Petru «it is again Peter
to whom he speaks» (where the repetition refers to an argument; comp.
(15) vorbeşte din nou cu Petru «he speaks again with Peter», where
the repetition refers to the predicate);
(iv) and adjectives: e.g. (16) Dunărea e tot albastră «the Danube is
still blue».
(b) TOT is invariable, while ACELAŞI agrees in gender and number with
its head-noun (it is therefore morphologically more complex):
see (17) tot rochia aceea - aceeaşi rochie «the same dress» (fem. ,sg.)
tot constumul acela - acelaşi costum «the same suit» (masc. sg.)
tot pantofii aceia - aceiaşi pantofi «the same shoes» (masc. pl.)
tot rochiile acelea - aceleaşi rochii «the same dresses» (fem. pl.).
The demonstrative following the noun behaves, of course, as an adjective.
The replacement under discussion is not yet a «fait accompli», since
there are various contexts where TOT and ACELAŞI do not carry exactly
the same conventional implicature: comp. (18) and (19):
Comp. (18) s-a imbrăcai cu aceeasi rochie
«she put on the same dress»
and (19) s-a imbracai tot cu rochia (nu cu pantalonii)
«she put on the dress again (and not the slacks)».
In (18) the denial concerns the relation between the members of the same
class, the class of «dresses»; in (19) the denial concerns the relation between
two classes, the dresses and the slacks.
Only with a NP followed by a demonstrative, TOT may convey a similar
implicature as acelaşi, i.e. the unexpected identity between two values of
the same variable:
Cf. (19) s-a imbracai cu aceeaşi rochie
«she put on the same dress»
TOT is compatible both with acela «that» and acesta «this», as shown by
(21) and (22):
424 MARIA MANOLIU-MANEA
(21) ieri te-ai imbracai tot cu rochia aceea «yesterday you put on that
dress again» (the dress is not on the hearer at the moment of speaking).
But, in combinations with pronouns, adjectives and adverbs, TOT is
now the only alternative for expressing an unexpected identity between two
variables, as sown by (10), (11), (12), (13) or (16) 4 .
Another phenomenon which may be considered as an indirect proof
for the lexicalization of the new conventional implicature of TOT (... acela)
is the fact that its homophone, the adjectival TOT (- TOATA-TOTI-TOATE)
«every» is no longer used with the value of a universal quantifier in com
bination with nouns characterized by the feature - Uniformity, as its cognates
are used in other Romance languages.
see (23) Fr. tout enfant doit se coucher tôt «every child must go to bed
early»,
but (24) Rom. *?tot copilul trebuie sa se culce devreme, somewhat unac
ceptable, with an archaic flavor,
rather (25) a. fiecare/ orice copil trebuie sa se culce devreme,
with more specific universal quantifiers, such as fiecare «every» and orice
«any», if not (25 b), with the plural form and the definite article -i.
(25) b. copiii trebuie să se culce devreme
«children have to go to bed early»
4 In old Romanian, the implicature under discussion was conveyed by the reflexive - Ş1
(Lat. SIBI): e.g. atunceşi «the same time», acoloşi «the same place», which is also a compo
nent of the pronoun of identity acela şi.
ROMANIAN PRONOUNS OF IDENTITY 425
and
(27) fetita chiar plînge, -i glumai «the little girl is weeping indeed,
seriously!»,
where chiar is a VP modifier.
chosen by the realm of events does not coincide with the candidate favored
by our expectations (arising from our common beliefs).
Because of this conversationl meaning, when focussing a NP, the
argumentative chiar is accompanied by - ŞI, which brings in the meaning
of «aditional argument» 5 :
(33) chiar şi Petru a venit «even Peter came»
The adverbial ŞI originates in Lat. SIC «so» and has the meaning of «also»:
see (34) spune-i şi lui «tell him also»
In Romanian, chiar alone cannot play this triple role, since it always precedes
the entire modified constituent (see chiar cu regina in (32.b)) and the definite
article is always cliticized and follows the noun (see (36) copilUL lit. «child»
«the»).
The spread of CHIAR as a means of denying expectations concerning
the identity has been favored by various factors:
(a) CHIAR is invariable, while însuşi has a rather complex mor
phology:
see the paradigm of însu§i in (37).
(37) însu§i «himself» - însaşi «herself» - înşişi «themselves» (masculine)
- inseşi (or însele) «themselves» - feminine»; însumi «myself -
masculine» - însami «myself - feminime»; însuti «yourself - masculine»
- msati «yourself - feminine»; inşine «ourselves»; înşiva «yourselves-
plural».
5 The same argumentative value can be expressed in this case by another phrase, name
ly pîna si (see Lat. paena ad): e.g.pisi Petru a venit «even Peter came», where pina brings
in the idea of «the attainement of a limit with an accompanying perspective». See the meaning
of pîna when used alone in contexts such as umple sticla pîna la gura «fill the bottle to the
rim», or astept pîna la patru «I shall wait till four o'clock». For a similar evolution of a space-
time preposition to an «insinuating word» see Sp. hasta: e.g. hasta Pedro vino «even Peter came».
ROMANIAN PRONOUNS OF IDENTITY 427
In the everyday speech (even in the speech of cultivated persons, and in the
written language of the press), confusions between these forms occur rather
frequently (see Iordan et al., 1967: p. 133: e.g. (38)însaşi(forînse§i) cuvintele
acestea formează o poezie, Flacãra, 16 (1957), p. 22 «these words form a
poem by themselves»).
( b ) Another factor favoring chiar is the fact that însuşi is governed by
more restrictive constraints:
(i) însuşi still preserves its «reflexivity» (given by the reflexive mor
pheme -şi (see Lat. sibi):
comp. (39) s-a împăcat chiar cu el «he made peace precisely with him»
and (40) s-a împacat cu el însuşi (or s-a împăcat cu sine însuşi) «he made
peace with himself».
(ii) The choice of articles after însuşi is more restricted than after chiar:
comp. (41) a spus aceasta chiar (a) mamei (definite article)
(b) unei mane (indefinite article)
«he said that precisely to the/a mother»
and (42) a spus aceasta (a) însaşi mamei (definite article)
(b) *însaşi unei mane (indefinite article)
(c) Chiar is capable of modifying any class of constituents with a similar
pragmatic value, insinuating the fact that an expectation of non-identity or
non-coincidence has been cancelled (see 26-32).
(d) Since CHIAR precedes the entire NP (or PP), it spread is also in
line with the present tendency to prepose noun determiners, which
characterizes contemporary colloquial Romanian:
comp. (43) standard Romanian: i-a spus mamei,
«he told to (my/his) mother»
lit. «to her has said-he mother-dative-definite article-dative»
and
(44) Colloq Rom. i-a spus lu mama (same meaning as (43)) lit. «to her
has said-he TO (invariable LU) mother (zero case marker)».
The on-going changes presented in this paper are not at all exceptional
if one considers the entire history of Romance pronouns of identity, which
have known almost cyclic changes precisely because of the pragmatic im
plications of their semantic area. The few changes under discussion are not
only meant to advocate the appropriateness of the Gricean model of im-
plicatures in accounting for semantic changes, but also, and mainly, to bring
428 MARIA MANOLIU-MANEA
REFERENCES
6 For a synthesis on the «final causes and language change» see Coseriu, 1982.
NOTE SU / S / INTERCONSONANTICA
NEI DIALETTI GRECI ANTICHI
CELESTINA MILANI
Università di Verona
1.2.1 Nei casi qui considerati /C1/ è un'occlusiva velare sorda men
tre /C 2 / è una nasale labiale. Si considerano solo gli esempi certi (e qua
si), rimandando per la problematica di ogni esempio ai dizionari etimologi
ci di Boisacq, Frisk, Chantraine che si devono considerare sempre citati.
1
Del greco si dà la translitterazione non la trascrizione fonetica.
430 CELESTINA MILANI
Con Gk. si intende il greco antico da Omero in poi, con PGk. si designa
la forma greca originaria ricostruita, con PIE la forma indeuropea.
Esempi: Gk. aikhmé ( PGk. aiksma, Aeol. aíkhmã «punta della spada».
II. 6.320, «spada»//. 12.45, cfr. Lit. iëszmas «spilla», OPruss. aysmis «spil
la», v. Hesych. aîkloi'hai gõníai toû bélous e (con altro vocalismo) iktéa a-
kóntion Hesych., cfr. Cypr. ikmaménos Lejeune 1972: 62, 66 n. 8, Adra
dos 1973: 115.
Gk. amukhmós < PGk. *amuksmós «scalfittura» Theocr. 24.126, cfr. Gk.
amússõ, ámugma, Hesych. amukálai-hai akídes ton belôn para tò amús-
sein; però cfr. arnukhé: analogica a amukhmós (o viceversa?). Gk. īõkh-
mós < PGk. «tumulto della battaglia» Il. 8.89 cfr. Gk. iöké, To
xis, Hom. Paliõxis, Corynth. wiókei.
Gk. lakhmós (PGk. *laksmós «battito» Antim. 54, cfr. Gk. laktismós «bat
tito del piede», láx «con il tallone» Boisacq 1950: 555, ma Frisk 1970: 82
e Chantraine 1968-80: 619: non accennano a / -sm-/. Gk. lakhmós ( PGk.
*laksmós «lana» Od. 9. 445 (mss.) per lakhnós Eust. 1638. 39. Gk. mukh-
mós {*muksmós «gemito» Od. 24.416, cfr. mukáomai, mugmós, v. Chan
traine 1968-80: 718s, s. 2 múzõ.
Gk. plókhmos < PGk. *plóksmos «ricciolo» Il. 17.52, cfr. Gk. plókamos,
plékõ, Skt. praçnah «lavoro a intreccio, turbante», OHG flahs «lino», OE
fl «lino», Lat. plectõ.
Si aggiunge a questo paragrafo Gk. rhõkhmós ( PGk. *rhõgsmós «fendi
tura» Il. 23.420, cfr. rhegnumi, rhôgas (. plur.), rhõgmós Schwyzer 1959:
493; rhõkhmós presenta varie grafie cfr. Liddell-Scott 1940-68: 1579.
na» Od. 9.445, cfr. Avest. varəsa- «lana», OSI. vlasu «lana», OSI. vlakno
«cappello», Lat. lana Benveniste 1935: 101s.
- GK. lúkhnos <PGk. *lúksnos «fiaccola» Od. 19.34, cfr. Avest. raoxšna
«brillante», OSI. luna (le/ouksna), OPruss. (plur.) lauxnos «stelle», Lat.
lūna (louksna), Prenest. losna «luna» Risch 1981: 501. Gk. pákhnê < PGk.
*páksnã < PIE *pagsnā «brina» Od. 14.476, cfr. Gk. pégnumi, págos Chan-
traine 1933: 192, Schwyzer 1959: 327. - Gk. pelíkhnè < PGk. *peliksnā «taz
za» Alcm. 19 cfr. . pe-ri-ke MY Ue 611 pelikes, Aeol. pelíka, Ion. Att.
pelike.
- Gk. tékhnë < PGk. *téksnã < PGk. *tektsnā < PIE *tekþna, cfr. Sk. táksan-
«carpentiere», Avest. tasan- «artigiano», Gk. téktõn «carpentiere, costrut
tore», lat. texõ Thurneysen 1887: 157 n. 3, Saussure 1892: 90 ss, Chantrai-
ne 1968-80: 1117; cfr. Locr. tekna Buck 1955:60.
1.2.3 Ci sono poi i casi di Gk. mórphnos e ptérnè in cui / C 1 / non
è una velare e / C 2 / è una nasale dentale. Cfr. Gk. mórphnos «di colore
scuro» Il. 24.316, un tipo di aquila Arist. Hist. anim. 618b.25; da PIE
*morgwsnos Walde 1897:478, Hirt 1901: 221ss. cfr. ON myrkr «scuro», OE
mierce «scuro», OS mirki «scuro»; oppure da PIE *morqwsnos cfr. OSI.
mrakǔ «oscurità», etc.: Solmsen 1897:24ss. cfr. Gk. morphe, dubbio Pisa
ni 1939-40: 497 ss. per cui vedi Belardi 1950: 214. - Gk. pternë Il. 22.397,
ptérnă Lycophr. 500 «tallone»; da PGk. *ptérsnā cfr. . pte-no KN Sd
«due predellini (del carro)»; cfr. PIE *pērsnā, Skt. parsnih «tallone», Avest.
pāšna-, Hitt. paršna-, paršina- «alto della coscia», Got. fairzna «tallone»;
OHG fersana «tallone», OE fiersn «tallone», lat. pērna «coscia» Benveni
ste 1954:41ss.
1.2.4 Si considerano in questo paragrafo i casi in cui / C 1 / è un'oc
clusiva velare sorda e / C 2 / è un liquida laterale una vibrante: - Focese
mukhlós < PGk. *mukslós < PIE *mughslos; in focese significa «asino ri-
produttore», cfr. Hesych. mukhlós skoliós okheutes lagkēs, moikhós. Phō-
keîs dè kaì ónous toùs epì okheίan pempoménous, v. anche Hesych. muk-
loi. hai perì taskélēkai en toîs posi epì notou tôn ónõn mélainai grammaí... ;
cfr. Hesych. múskloi · hoi skolioí (múskloi ( PGk. *mukslo-); cfr. Lat. mülus
(*mukslos), Alban, musk «muletto» ((PIE * mughslos), etc.; cfr. Nieder
mann 1902: 101 ss.
- Gk. ablēkhrós Il. 5. 337 (a- protetica Wackernagel 1910: lss.), blēkhrós
Alc. 16: da PGk. *mlāksros, cfr. Lat. flaccus (*mlākos), Lit. blaka «pun
to smagliato della tela»; è da mettere in rapporto con Dor. bláx, blākós,
432 CELESTINA MILANI
2 Più certo di questi casi è diphthéra, cfr. . di-pte-ra PY Un 1318, di-pte-ra3 PY
Sb 1315; Saussure 1892:91 richiama Gk. dépsō con chiusura di / e / in / i / come in histië cfr.
Schwyzer 1959: 351; la sibilante di depsō può spiegare l'aspirata / p h / ma la formazione resta
singolare; Pisani 1939-40: 513 pensa a diph- + tera; si è pensato a un neutro in /-tar/ come
néktar, hiktar, divenuto tema in / - ã / come hēméra/hêmar cfr. Chantraine 1968-80:80.
/ S / INTERCONSONANTICA NEI DIALETTI GRECI ANTICHI 433
2.1 In questa parte della presente ricerca si studiano i vari casi in cui
la sibilante interconsonantica, evidentemente ridottasi ad approssimante la-
/ S / INTERCONSONANTICA NEI DIALETTI GRECI ANTICHI 435
BIBLIOGRAFIA
DONKA MINKOVA
University of California, Los Angeles
early change — Late Old English up to 1250 — are less numerous and fall
into several clearly defined groups. Such data is easily identified and isolated,
and is available in the standard literature. By restricting this study to un-
controversial early cases of loss, one avoids the overwhelming bulk and in-
conclusiveness of later evidence. I shall try to interpret the early data in terms
of a non-linear phonological framework, making the analytical approach
also «isolative» in a sense.
The earliest, and most frequently cited instances, of schwa loss follow
several familiar patterns 1 :
1. Elision in hiatus. «There can be no doubt that weak vowels were
often elided before another vowel [or -h DM] in ordinary OE speech» (Sweet
1922: § 374). A statement to the same effect comes from Luick (1921-40/64:
§ 452), according to whom after the 10th elision became «more frequent
and general as a consequence of the acceleration of the speech tempo: already
before Orm's time each post-tonic -e disappears before a following vowel
or a -h in a weakly stressed word in the spoken language (emphasis DM).
And further: «this phenomenon is .... so regular, that it must have had a
purely linguistic foundation [eine rein sprachliche Grundlage]». So, for Late
Old English and Early Middle English the phenomenon is well documented;
some relevant early instances are 2 :
— sum sar(e) angeald; on þæm gold(e) ongitan; gyted(e) hine Beowulf (Sweet
1922) — all from Old English.
— mōstic, todeld it, mādim, mighti (Luick 1921/64: § 452). The same pat
tern, loss of schwa in hiatus, accompanied by cliticization of a following
pronoun or preposition, is frequent in the Peterborough Chronicle: heng-
1
The paper will not cover OE/EME «functional» loss, e.g. tō dæg, æt hām, for *tõ
dæge, *æt hāme (Horn-Lehnert 1954: 306), nor co-occurrences of the type ... þa hwile fulwiht
sio ... and ... þa hwile fre fulwihte sio (9th c. Dahl 1939: 14). These are all instances of
purely morphological alternation confined to nouns, usually as part of prepositional phrases.
For a survey of the data and an evaluation of its importance cf. Minkova 1983: 62 ff and 1984
b: 452-3.
2
I have not included cases of «pre-accentual» loss in Old English (Luick § 451). This
group, as isolated by Luick, covers the loss of the vowel in prefixai be-, ge- and also in ne,
be, to, me, sche, e.g. binnan «within», gede { geeode «went», nis «is not», þoþre, «the other»,
toffren «to offer», malon «me alone», shaligte «she alighted», etc. Their treatment is essen
tially the same as that of the instances discussed below; I am not taking them into considera
tion because of the later (18th c , according to Luick, ibid.) reversal of the process, and the
ultimate reinstatement, and even strengthening, in the case of the definite article, of the «weak
schwa».
SCHWA DELETION IN ENGLISH 447
ed up, smoked heom, makod he (wk. pret. sg.), pined heom, bebyred
him, læd him (pret. pl.) (Clark 1970: LXV).
— mid sac and mid socne (10th c, Dahl 1936: 15); chele 7 þrisst, sune 7
mone, hete 7 niþ, kare 7 serrge, depe 7 derne, lefe 7 dere, etc. — the
latter are examples from the Ormulum (c. 1200), where the consistent
metrical pattern makes elision of -e in the first word of the pairs
obligatory. (Burchfield 1956).
2. Schwa loss occurs also early (in the course of the 12th c.) in non-
major class words: pen/pan (OE þenne/þanne), when/whan (OE
hwenne/hwanne), sōn (OE sōne), ges (OE gese), as well as in the inflected
forms of pronouns: än, mm, hir, your, etc. (Luick 1921-40/64: § 454; Jor
dan 1934/74: 152). Here belong also words of French origin which would
be unstressed in combination with a proper name, e.g. sir, dam, frēr (ibid.
§ 461).
—þon hast a moder fair and fre. (134)
She is yhote dame Candace:
Fair and brigth is hir face —
(Kyng Alisaunder)
or:
— (And leuen al his muchele kare) (152)
To dame Siriz pe hende.
— «God pe iblessi, dame Siriz! (161)
Ich am icom to speken þe wiz...
(Dame Sirith)3
the secondary stress and simultaneous loss of final -e (Luick op. cit. §
459 4 .
This, in very broad outline, is the picture we get of the loss of final
schwa in Late Old English and Early Middle English, before the process
was generalized and began to affect all lexical items in the language 5 .
Unlike post-1250 data of across-the-board loss which is frequently in
conclusive, there can be no doubt about the validity of the above evidence;
all three groups above include instances of graphically marked loss, i.e.,
no final <-e> appears in the sources. In the cases of graphic preservation
of <-e>, as in some of the instances of loss in hiatus or the examples from
kyng Alisaunder and Dame Sirith, there is a consensus among scholars con
cerning the loss, based on unequivocal metrical evidence, as in the Ormulum.
(On this cf. also Burchfield 1956, Strang 1970: 244).
Once the dating problem and the problem of inconclusive evidence is
out of the way, we can turn to the issue of causality. The classroom wisdom
here is a noncommittal appeal to a balanced interaction between the lack
of phonetic distinctiveness of the final unstressed vowel and the functional
«overloading» of the grammatical morpheme which it represents. What has
been generally overlooked, however, is that all groups above share some
important characteristics beyond the now trivial phonological neutraliza
tion of the final vowels and their weakened function as inflexional markers.
These are:
— The loss is not confined to any particular segmental environment 6 .
4
Instances of syncopation of the medial syllable in originally trisyllabic words (e.g. kin-
dom, OE cynedom, balfull, OE bealuful, hergong, OE heregang, etc.) are not quoted in this
list only because the focus in the paper is on the loss of final -e. It will become apparent, however,
that the rhythmic principle accounting for the syncopation and for the loss is the same in both
cases.
5
As mentioned above (cf. note 2), one set of examples of schwa loss I have not attemp
ted to cover in this paper are the so-called «contracted» forms: ME poþre «the other», nis
«is not», malone «me alone». These are all instances of the Early Schwa Deletion Rule (below)
operating in its (b) form. The modern contractions hasn't, we've, I'll etc. are also instances
of the (b) form of the rule, after it was generalized to affect [-as] and [-əd] terminal nodes.
6
Koziol (1937:308) claims that «within each individual category — weakly stressed,
polysyllabic, disyllabic, native and borrowed words, loss occurs earlier after voiceless stops
and fricatives, as well as after liquids or nasals, than after voiced stops and fricatives». The
evidence adduced in support of this claim is very selective and fragmentary; my own textual
data (Minkova 1983:51-101) does not warrant this conclusion.
SCHWA DELETION IN ENGLISH 449
7
According to Jordan (1934/74:141) loss after long first syllables in trisyllabic words
precedes loss of schwa in unsyncopated trisyllabic words with a short first syllable. This state
of affairs is assumed also in Wright and Wright (1923 § 141, but cf. § 153). I have disregarded
this temporal sequence for two reasons: (a) it is only valid for the South; in the North the
situation «is simplified by the fact that here every final -e disappeared already in the 13th c.»
(Jordan, ibid.); and (2) «words of the form x x x always had a long first syllable» (Luick,
op. cit. § 509). This comes as a consequence of an earlier development, 11th c, whereby the
majority of the trisyllabic forms with a short stressed syllable undergo syncopation of the medial
syllable (Luick, op. cit. § 513).
8
For a statistical account of the distribution of loss in hiatus across the various word
classes, cf. Minkova 1978.
450 DONKA MINKOVA
I: Intonational Phrase
P: Phonological Phrase
: Clitic Group
W: Word
I have replaced theC1,C2,C 2 C 3 , and C 4 labelling in the original to avoid
confusion with the Clitic Group 9 . The branching is, of course, entirely ar
bitrary, though not so the hierarchical arrangement.
Phonological rules refer to phrases of a particular size, regardless of
the syntactic category they represent. An important property of the hierar
chy is that 'rules applying across a given «boundary» necessarily apply across
all weaker «boundaries», and rules applying before or after a given «boun
dary» necessarily apply before or after all stronger «boundaries»' (Hayes
1985:6). Thus, in this set of data, the rule deleting / ə / in [w ælmesse] (where
-e > 0) applying at the right edge of the W(ord) domain, will apply at the
right edge of the higher domains, the Clitic Group, the Phonological Phrase,
etc. I am also assuming and applying the provisions spelled out in the
typology of domain-sensitive rules proposed by Selkirk 1980. According to
this typology, domaine juncture rules «must know whether the segments
[they] involve belong or not to distinct domains of the type Dj included
within the domain D i ... A rule of this type scans the terminal string includ
ed within the Di domain and performs the structural change only if the
segments of the terminal string of the phonological representation can be
factored into smaller prosodic domains in the way specified in the struc
tural description» (Selkirk 1980: 111-2).
Since the phrasal phonology rule defined above allows us, in fact re
quires us, to scan the terminal string within the domain of application, we
can attempt a structural description of the terminal nodes as constituents
which enter into binary strength relations. If we specify the W S relations
on the constituent nodes of the examples in groups 1-3 before and after dele
tion has applied we get:
9
The Clitic Group is defined as «a single content word together with all contiguous
grammatical words in the same syntactic constituent» (Hayes 1985:12). The Phonological Phrase
is formed from one or more Clitic Groups. It is not necessary for the purposes of this study
to examine the details of the syntactic diagnostics for phonological phrasing.
SCHWA DELETION IN ENGLISH 451
452 DONKA MINKOVA
(Boldface W nodes indicate the position and the relations of the deletable
syllable).
An examination of the deletable nodes in 1-6 allows us to formulate
the following rule:
The Early Schwa Deletion Rule:
If within the Clitic Group a W node is filled by a syllable of the follow
ing configuration:
(i.e., crucially non-branching rhyme, with schwa in the peak position), and
either
10 I am assuming, with Luick (op. cit. § 456), that deletion follows upon the suppres
sion of the secondary stress in words of this type.
π Same as 5. An account of the first stage of the rule is not intended here, but one cold
suggest that the OE loss of schwa in - is concomitant with the weakening of the second
element of the compound and the elimination of the word boundary within it.
SCHWA DELETION IN ENGLISH 453
(a) is adjacent to one or more W nodes in the terminal string (as in 1,2,3,5
and 6), or
(b) is immediately dominated by a higher W node (as in 3,4,6),
then this node is deleted.
The (a) and (b) parts of this rule have two important properties in
common:
(1) They both make reference to the Clitic Group. This I regard as the
unifying principle behind the entire process of early schwa loss. The Clitic
Group is the only domain which would correctly predict the deletion in ex
ample (2) above, by stipulating adjacency of W's on the terminal string where
disyllabic «heads» are involved.
(2) They both provide that the deletable W node must be attached to
other material to the left and must have a non-branching rhyme, and a peak
filled by [-3]. At first blush this may sound trivial. But if we bear in mind
that since Late Old English times this has been a sufficient specification for
the deletion to occur, it appears that schwa loss is not a consequence of a
segmental rule but of a general prosodic constraint in English which bans
syllables of the above configuration from appearing adjacent to another W
node of the same configuration or word finally.
It also appears, at least preliminarily, that any violations of the above
constraint fall easily into clear-cut groups:
— Proper names, native and foreign: Anna, Birma, China, Donka, Erica etc.
— Words borrowed into English after 1500 and /or retaining their foreign
flavor: gamma (1450), mesquita (1477), naphtha (1500), dilemma (1523),
saga (1709), ballerina (1892), tuna (1884), marijuana (1923), rumba (1937),
canasta (1948), etc. Only hosanna and mandragora are recorded from
before the 12th
— British Southern Early Modern English [-r] vocalizations: number, er
ror, potter, summer, etc.
— In casual speech [ou] can surface as [-ə]: fellow, pillow, window.
A discussion of these groups does not belong here, yet let me stress that
in spite of the apparent violation that they represent, they are still subject to
one of the constraints noted above: English disallows two adjacent W nodes
wherein both have the above configuration:
This observation has been made
about Modern English, but to the best of my knowledge it has not been
regarded as a more general historical constraint. It is a matter of relative
454 DONKA MINKOVA
12
The [-] vs. [-ən] selection does not have to be analyzed as «avoidance of hiatus».
It is that, of course, but it is also a constraint in English on the co-occurrence (word medially)
of two syllabic nuclei, the first of which is []. The same principle would be applicable to the
appearance of the «linking» and «intrusive» [-r-].
13
Bell and Hooper 1978:8 write: «Within the section [roughly the phonological word],
VV sequences («hiatus») are not permitted by about one-half of the world's languages...».
SCHWA DELETION IN ENGLISH 455
e.g. heþene → hepen), monosyllabic adjectives of the type gōd, dëf, smal,
strong mark the syntactic distinction between attributive and predicative,
and the morphological distinction between singular and plural, much
longer (Mosse 1952/68: 64, Samuels 1972b)14. It can be suggested that in
the inflected forms of these adjectives (with, final -e) there would be no
pressure for deletion arising from W W adjacency15. I am not sure this is
the right explanation, but at least it opens a line of reasoning thus far
unavailable to us.
14
Also see Cable 1983:63, who writes: «Although adjectives of two syllables were
generally uninflected, adjectives of three syllables often took an inflectional -e.» This made
a fourth syllable that sustained the naturally alternating rhythms of English prose and the pat
tern of the iambic pentameter: the sémelièste mán.
15
One issue of which I am aware, will not be addressed here, namely the subsequent
history of the schwa. The prosodic rule, as stated here, does not predict that deletion will later
generalize to all word-final positions. By deleting one of two adjacent W's the later rule will,
as a matter of fact, create undesirable stress clashes and thus run counter to our vague intui
tions and expectations that an evenly spaced rhythmic alternation is an innate feature of the
language. On a theoretical account of this state of things, cf. Hayes (1985:84).
I cannot here defend the prosodic hierarchy approach as compared to the boundary ap
proach to junctural phenomena. This is a theoretical issue to which my analysis cannot con
tribute; both approaches would suit the data and the choice of one over the other perhaps
does not matter. What matters is that the rule should be formulated within the boundaries
of a prosodie domain. This is the only formulation under which the apparently random in
stances acquire some coherence.
The Early Schwa Deletion Rule makes reference to the Clitic Group. It might be interesting
to investigate the possibility of confining the operation of the rule within the domain of the
metrical foot. I suspect the results will be similar to the ones obtained here.
I have not gone into the theoretical implications of «flattening» of metrical structure,
which appears to be what is going on here. For a discussion of this notion and its application
to the metrical treatment of German compounds, cf. Giegerich 1984: 189ff.
Finally, on the advantages and disadvantages of the Early Schwa Deletion Rule: it sug
gests a reason why the whole thing started, namely the configuration of the W node undergo
ing it. Also, it tells us something about fast or casual speech during the period. Being a diagnostic
456 DONKA MINKOVA
REFERENCES
for phrasing, the rule as formulated here supports an observation (Cable 1974:90) that the
absence of a x x pattern is «not especially surprising». It appears to be «a natural result
of the linguistic system». Thus the rule can be used in other areas of historical research, the
study of Old and Middle English meter.
SCHWA DELETION IN ENGLISH 457
Dahl, Torsten. 1936. Form and function. Studies in Old and Middle English
Syntax. Acta Jutlandica, Aarsskrift for Aarhus Universitet. VIII. 1-258.
Giegerich, Heinz. 1985. Metrical phonology and phonological structure: Ger
man and English. Cambridge: University Press.
Hayes, Bruce. 1985. The prosodic hierarchy in meter. MS. UCLA.
Horn, Wilhelm and Martin Lehnert. 1954. Laut und Leben. Englische
Lautgeschichte der neueren Zeit. Berlin: Verlag der Wissenschaften.
Jordan, Richard. 1934/74. Handbook of Middle English grammar: Phono
logy. Translated and revised by Eugene J. Cook. The Hague: Mouton.
Koziol, Herbert. 1937. Der Abfall des nachtonigen -e- im Mittelenglischen.
Beiblatt zur Anglia. 48.306-309.
Luick, Karl. 1914-1940/1964. Historische Grammatik der englischen
Sprache. Stuttgart: Bernhard Tauchnitz.
Minkova, Donka. 1978. Unstressed final -e in the Ormulum. English studies.
Sofia University. 162-80.
Minkova, Donka. 1983. Phonological and morphological problems of the loss
of final -e in Middle English. Unpublished Ph.D. thesis, Sofia University.
Minkova, Donka. 1984a. Early Middle English metric elision and schwa dele
tion. English historical linguistics: Studies in development. CECTAL confe
rence papers series No. 3, ed. by N. Blake and Ch. Jones. Sheffield. 56-67.
Minkova, Donka. 1984b. On the hierarchy of factors causing schwa loss
in ME. Neuphilologische Mitteilungen. LXXX 4: 445-454.
Mosse, Fernand. 1952/68. Handbook of Middle English. Translated by
James A. Walker. 8th printing 1983. Baltimore: The Johns Hopkins
University Press.
Samuels, M.L. 1972a. Linguistic evolution. Cambridge: University Press.
Samuels, M.L. 1972b. Chaucerian final -e. Notes & Queries. Dec. 1972.
445-8.
Selkirk, Elizabeth. 1980. Prosodic domains in phonology: Sanskrit revisited.
Juncture, ed. by Mark Aronoff and Mary-Louise Kean. 107-131. Anma
Libri. Stanford University.
Strang, Barbara. 1970. A history of English. London: Methuen.
Sweet, Henry. 1922. An Anglo-Saxon reader in prose and verse. 9th ed. Ox
ford: Clarendon Press.
Wright, J. and E.M. Wright. 1923. An elementary Middle English gram
mar. 2nd ed. Oxford: University Press.
ARTICULATORY EVOLUTION
1
It is not particularly difficult to find evidence for the gradual erosion of fricatives.
Consider, for instance, the behavior of intervocalic / õ / in South American Spanish dialects
in which this fricative is in the process of eroding: native speakers report that remnants of
decaying / o / — elevation of the tongue in the attitude of the fricative, without contact —
persist even in those lexical items in which «deletion» would otherwise be assumed. Menéndez-
Pidal, in describing the same process in Iberian Spanish (1968: 99ff.) notes that this fricative
is generally very «weak», which may indicate how long a history an eroding fricative can have
on its way to 0. Nor is Spanish unique: in not a few grammars where the author attends to
phonetic detail, remarks on the weakness or strength of fricatives transcribed in IPA symbols
are common qualifiers on supposed universality. So, too, in careful records, we don't always
find the stops of a given language to be phonetically alike — differences in the extent of aspira
tion, voicing, and affrication seem to be usual rather than exceptional. To some extent, the
regularity and discreteness assumed for segmental configurations across languages is a direct
result of phonemic analysis, in which phonetic differences are routinely regularized away. When
care is taken in recording phonetic detail in changes in progress, the existence of «stages» such
as those we indicate in 3b and 3d becomes obvious. Thus, in a meticulous cross-dialectal study
of sound changes in Quechua, Quesada (1985) provides evidence for intermediate «steps» in
the reduction of [s] and other obstruents, including the uvular stop [q] which, in «spirantiz-
ing» as opposed to «voicing» dialects, reduces first to an uvular affricate, and then to a fricative.
2
Lass (1984:178), who has long been concerned with issues related to our discussion
so far, is an exception.
462 W. PAGLIUCA - R. MOWREY
Thus, it is not as if, for (b), a transitional glide is introduced between the
still-full apical fricative and the palatal vowel, but rather that the transition
is the fricative and vowel in coproduction at that locus. Necessarily, then,
we cannot regard the alphabetic symbols in (b) as representing, either tem
porally or substantively, the «same» configurations as in (a). That is, (b)
is a temporally-compressed descendant of (a), as (c) is of (b).
3
References for 9-14: For 9 and 11, Zwicky 1972 and Ohala 1974a, who provides ar-
ticulatory analyses. 10, adduced by Vago (1976) as an instance of an acoustically-based change,
is described in more detail in Pagliuca and Mowrey 1980, as is 11, for which cf. also the analysis
in Javkin 1978. For more discussion of the descriptive problem posed by monophthongiza-
tions like 14, cf. Pagliuca and Mowrey 1980.
466 W. PAGLIUCA - R. MOWREY
(14) ai > e
We must reject feature-and-rule notation as being inadequate to express
the relatively obvious temporal and substantive reduction processes we have
discussed. Because of the classificatory nature of their reference, such feature
specifications, like the alphabets they expand, are incapable of expressing
directly the phonetic content, and hence the phonetic nature, of a host of
diachronic phenomena.
3. The solution to the problem of representation does not lie in the
selection of new classificatory features which stack up in the familiar serial
segmental manner: such a system would remain incapable of expressing the
timing details inherent in the gradual coalescence of neighboring articulatory
gestures so common in historical change. In short, simply choosing new
features would maintain the notational myth of discrete segments which
mysteriously vanish when people open their mouths to speak.
Having rejected the segmental framework in favor of one which allows
us to express timing relationships in greater detail, we are left with the pro
blem of choosing the events whose timing should be represented. Candidates
are in 15.
(15) a. Constriction points and degrees of constriction of the various ar
ticulators.
b. Spatial configurations of articulators relative to a reference point.
Acoustic events.
d. Neuromuscular events.
Of these four, only neuromuscular events are capable of directly reflec
ting temporal and substantive reduction. For example, the reduction in 16.
(16) ρ > pf > f > h > 0
can be described as in 17.
(17) a. gradual loss of orbicularis oris superior activity { ρ > p f > f}
b. gradual loss of orbicularis oris inferior activity { f > h }
c. gradual loss of laryngeal opening (posterior crico-arytenoid activi
ty) {h }
or schematically as in 18.
ARTICULATORY EVOLUTION 467
(18) ρ pf f h 0
OOS
OOI
PCA
This schema may be over simplified (that is, underspecified as to the
number of neuromuscular events occurring) but it is at least capable of show
ing the series as reductive, which feature specifications cannot do.
Any claim that the acoustic similarity of two sounds is the basis for a
language-internal change entails a number of predictions about their behavior
4
Rounding in the vowel of full is historical, i.e. is a retention in Cockney.
Sivertsen's conventions: indicates a non-syllabic (glide) segment; — = retracted.
5
For more discussion, cf. Pagliuca 1982.
470 W. PAGLIUCA - R. MOWREY
ciencies inherent to the human perceptual system which would account for
changes of the magnitude [x] > [f], [k] > [p] or > [w]. The «jumps» in
volved are — articulatorily and auditorily — too large. It is true that learners
do not have direct access to many of the details of articulation of the language
they are acquiring, and must rely heavily on auditory information. But rather
than concluding from this that the natural locus of diachronic phonetic
change is therefore auditory perception — i.e. in misperception or error —
the success with which fine dialectal detail is copied should suggest to us
that the link between perception and articulation is far more subtle and ef
ficient than we tend to give it credit for being, and that we should be wary
of untenable conclusions about the lack of sophistication of the human
perceptual and motor systems.
REFERENCES
6
We thank Joan Bybee for reading and commenting on this paper, and the Linguistic
Society of America and the National Science Foundation for a travel grant to WP.
472 W. PAGLIUCA - R. MOWREY
REBECCA POSNER
Oxford
way as an attributive adjective and is incorporated into the NP, the boun
dary of which is marked by LA. My contention is that non-use of a com
plementizer in French creoles may simply be a generalization of earlier
Romance tendencies, in which the link between clauses need not be marked
overtly, but that the use of LA as a relative marker may reflect a more
unusual change, in which the relative clause is seen as an integral part of
its head NP.
In Bickerton's scenario, however, the two processes would not be so
different. For him «core» creole would have no overt complementizers: the
creoles I have been examining are assumed to have been influenced, at their
formation, by the superstrate language, to a greater or less degree, depen
ding on the exposure of language learners to superstrate data, if only in the
form of a stable pidgin. One result might be that the superstrate complemen
tizers would be used in some contexts and varieties: one can well imagine
that more salient French qui would resists erosion more than que.
Bickerton's (1984) table setting out the putative ranking of various
creoles, shows Haitian «nearer the bioprogram» than Mauritian, while Réu
nionnais, which is much more like popular French than creoles proper, is
much further away from the bioprogram. Presumably the creoles of Mar
tinique, Guadeloupe and Louisiana would lie between Haitian and Mauri
tian. The proportion of slaves to masters at the time of creolization is of
crucial importance in this ranking so that the language of slaves (e.g.
maroons) left to their own devices would be closer to the bioprogram. One
would predict that creoles in societies where masters were more in evidence
would make more use of complementizers than those in which the bioprogram
had full rein. Developments like the use of LA as a relative marker could
be seen as a compensatory device in conditions in which the QU - relative
marker is regularly omitted in some structures.
A more traditional story than Bickerton's - and for me a less convinc
ing one - is that abrupt creolization was followed by gradual decreolization.
This story envisages the insertion, by adult African speakers, of their masters'
lexical items into their native grammatical structures. The resultant unstable
jargon would provide the input to language learners of a new generation,
for whom the creole would be the mother tongue. Subsequent decreoliza
tion would depend on the degree of contact with the superstrate language:
creoles on the right of Bickerton's table would have undergone extensive
decreolization. One result would be the adoption of the superstrate com
plementizers.
478 REBECCA POSNER
I repeat that the evidence about creole complementizer use is most simply
explicable in terms of «normal» gradual linguistic change, and requires no
such appeal to a catastrophic creolization process.
However, the Haitian LA relative marker does point to a more radical
transformation, with French là interpreted as part of a quite different gram
matical framework. That substrate had a hand in this reinterpretation is
eminently plausible, but «biogrammar» could also have a rôle to play.
What I am suggesting is that at the moment of creolization there was
a change of «type» - in the sense of «the idea speakers have about what
their language is like» (Coseriu 1968). Let me use the Coserian hierarchical
schema to illustrate, spatially, my contention.
Universais
«top-down» type / «bottom-up»
changes system changes
norm
N speech
conditions, the «ideas» about language that a speaker may have, have been
acquired only from the innate bioprogram («universais») and not from any
tradition. Change comes right from the top, in other words.
Such a circumstance could, surely, result only from a catastrophic social
upheaval in which children acquiring language have access solely to the most
rudimentary speech data: such are the conditions which Bickerton envisages
for the formation of pure creole. It is to be doubted whether these condi
tions, in their extreme form, could ever obtain. But it is worth pursuing the
question of whether what Bickerton (1984) calls a «weaker reflex» of the
«internal program» had a hand in, for instance, the emergence of Romance
from Latin, or the linguistic transformation that took place in the Middle
French period.
In this perspective it will be important to define what the possible
«typological» variations can be, and whether some features are always
«packaged» with others, as Chomsky's (1981) hypothesis about parametric
setting implies. I have suggested that one way modern Romance differs from
Latin is that relative clauses are seen as similar to subordinate and com
pletive clauses, which are «hinged» to other elements of discourse by the
all-purpose QU complementiser. Incidentally, «free» relative clauses are
treated more like interrogatives than headed relatives (Posner 1985). Creole
tends on the other hand to omit overt complementiser «hinges» and to treat
relative clauses as integral parts of their NP head (making, it seems, no use
of «free» relatives). Whether these differences can be linked with, for in
stance, morphological (e.g. part-of-speech marking or syntactic (e.g. word-
order) features), remains to be investigated.
REFERENCES
Ahlqvist, Anders, (ed.) 1982. Papers from the 5th ICHL. Amsterdam: Ben
jamins.
Andersen, Henning. 1973. Abductive and deductive Change. Lg. 49. 567-593.
Baudet, Martha M. 1981. Identifying the African Grammatical Base of the
Caribbean Creoles: A Typological Approach. In Highfield & Valdman
1981. 104-117.
Bickerton, Derek. 1981. Roots of Language. Ann Arbor: Karoma.
-----. 1984. The language bioprogram hypothesis. The Behavioural and Brain
Sciences 7.2: 173-221.
480 REBECCA POSNER
This paper summarises a version which will appear as a book in its own
right (Prosdocimi «Sillab.»), wherein there is a chapter on the (non)reality of
reconstruction and on the meaning of reconstruction units which are not pho
nemes (or morphemes, lessemes, etc.) of natural languages, but comparative
segments as well as reconstruction units. They are units which have the syste
matic configuration and aspects of natural language units, but which are not
so by nature. The distinction, irrelevant in certain cases, becomes essential
for others: vocalism, sonantism and laryngeals of the Indo-European Recon
struction belong to these other cases. It is evident that as far as syllabicity
goes, vocalism which was and - despite the big shock of the «New Look»
of consonantism, is implicit and still at least in our eyes, the centre and punc
tum dolens of Indo-European reconstruction problems (and for good reasons).
From certain evidence it would seem that in the presence of a vocalised
sonant, the apophonic vowel (e/o) can come out as zero, associating the
phenomenon to accentual covariation. Also the more general problem of
the nature of apophonic e/o is posed, i.e. of vocalism in a nutshell. In our
thesis what matters is that it is a vowel of vocalisation and not different
- if not for reasons of chronology and/or functionality - to the vocalisation
which gives a in Latin (etc.). The vocalism e/o is colourable by laryngeals,
with indifference towards e vs. (cf. e/oH2 > -ã vs. e/os > -os).
With colourable vocalism, I have associated the Schwebeablaut (Pro
sdocimi 1984 «Latino»). With non-colourable vocalism, I associate the so
nant level expressed by symbols r etc.
We may therefore have:
1 pekwtos sprētus perk/prek-
2 cards gnātus prk
The type perk/prk justifies, amongst other things, verbal formations such
as prsko: perksk- of Italic-Latin, Celtic, etc. This takes back the vocalisa-
484 ALDO LUIGI PROSDOCIMI
tion e vs. a, reduced er vs. r (i.e. to the former r with vocalic support) to
an insertion factor in the vocalic system in as much as in morphological func
tion, rather than, or besides, chronological or purely chronological. The dou
ble level of vocalisation is connected to the question of the vowel a, which
I would suggest in terms different to Lubotsky's (at this congress): a, a
typologically probable but not necessary vowel, exists as a/å; when a/a)e/o,
there is space for new vocalisation.
The question is vast and here only aspects or fragments are dealt with.
There is, however, the practical question of notation. I will use e - as com
monly used where vocalisation may not be relevant. I will use (C)e/o where the
lack of distinction in timbre may be relevant. Where the possibility of non-
vocalisation may be relevant, I will use Ce/o in alternation with Co (the tet
ter's greater neutrality to indicate a vocalisation position or vocalic potentia
lity, which can be explained either as e/o (pok w -tos), or as a different vowel,
such as a of Latin canis). With > I mean «changes to» and/or production.
1
Henceforth I shall use a few conventional forms of the «Classic Sievers» and «Sievers-
Edgerton» type, etc. Since the phenomenon is particularly clear in Vedic - where the case -i(y)a
is particularly evident-, I shall use the Indian form (easily identifiable for -y- and -a-) in alter
native to -ijo- of the normal reconstruction without particular expedients, except where there
is the danger of causing ambiguity or confusion. For the same reason I will not worry about
putting forward a monster such as vrkiH-, and not wlk w jH-, where -iH- alone and nothing
else is under discussion. I will not use asterisks where it is obvious that I am dealing with
reconstructed forms, etc.
As far as apophonic vocalism is concerned, the matter is very delicate, because within
the analysis that I give below, at least in certain positions its very nature is put in question.
I will use e and according to common use (-jos and -jeHi) reserving the use of a convention
e/o or V where relevant. In some cases I shall use to signify a position with vocalization
or vocalic potentiality, which could be explained either as e/o (pokw-tos) or as a different
vowel, whether a in Latin or ï in Indian (on the difference, see §§ 3.12).
SYLLABICITY AND SIEVERS' LAW 485
An interactive right to left formulation, i.e. from the end of the word: cf.
çvabhih and not + çumbhih: «Expanded Sievers».
1.2.1 «Expanded Sievers» allows numerous exceptions, among which
is the same classical Sievers' Law (cf. Schindler 1977, reformulation accor
ding to Prosdocimi «Sillab.»):
1. Word initially:
2 «Bei den -men- Stämmen wurde / m / in der folge /CmnV/ nicht silbisch sondern
schwand (Typus ved. ásman-: ásnah)» (Schindler 1977). As I note, this class was put forward
by Schmidt (Kritik der Sonantentheorie) as an anti-consonantal argument. The types raçman-,
dra(g)hman-, açman- in the weak forms ought to give -CmnV and therefore *-CmnV > -Cama,
while on the contrary they give (instrumental): raçmá, dra(g)hmã, ráçnã. It is possible that
the variation in -m-/-n/ is due to stress variation, but the oscillation here in the sense that
in this sequence two nasals are reduced to one - obviously for tautosyllabic reasons or for being
- at the point of juncture equivalent to one nasal only. This is the same rule as pet-(t)ro-,
se(d)-(t)la- (Saussure MSL 1889 = Recueil pp. 420-432) as applied to nasals: -VCM//NV- =
-VC(M)NV-; or -VC//MNV- = -VCM(N)V- from which Cn-/Cm-. The same goes for krn(w)-wa
and krn(w)-má. grabh-nï-ma (unless we have -iH- as morph) and not *grabhãna < -- < -nH-
exemplifies application of syllabicity rule in initial position: -Vbh//n0H//mé.
486 ALDO LUIGI PROSDOCIMI
3 I leave aside the second diphthongal element in syllable final position as possibly pos
ing problems regarding the consonantalism (cf. krnwa-) of even greater diachronic complexity
- theoretically to the point of a structure with open syllables only. In syllable final position
then a diphthong for phonetic reasons results: Ceje > CejØ. Syllable initial position is the con
ditioning force, for which reason from now on, unless indicated, «sonant» will be synonymous
with sonant in initial position and at the centre of the syllable.
SYLLABICITY AND SIEVERS' LAW 487
against the general rule of cooccurrence: Sievers is a rule following (as will
become clear, it is not categorical) the one which changes consonantal sonants
to vocalic sonants
4 The phenomenon was recognised by myself and ascribed to Sievers' Law as early as
1965 and 1967. The phenomenon was taken up by Prosdocimi 1977 «i.e. io/i» and 1983
«Venetico» pp. 176-178. The explanations here are to be modified and/or completed by the
explanation given in this article and in Prosdocimi «Sillab».
5 The orthographic rules are complicated by the syllabic punctuation wherefore -VjV-
is -V.i.i(i)V-, extremely rarely -ViV-. This spelling (Lejeune/Prosdocimi) was phonetically inter
preted as -Vi//jo-. I maintain now that this spelling indicates strong syllabic caesura -V//j-,
which strengthens -j- to -jj- from which we have as the result -V1 jo-/-Vijo-. There is a confir
mation in Messapian (Prosdocimi 1985, «Messapico» and «Sillab.»): Cjo- is always -Ce- (with
-e- which can be shown to be short) and -Vjõ- is always -Vja- where -j- is equally written in
an unmarked manner (-hi- or J'-) indicative of accentuated fricativeness = consonantism =
syllabic caesura -V//j-.
6 «Bei den Komposita mit ja-stammingen Vorgliedern steht in der Kompositionsfuge
nach schwerer Silbe und nach Nebensilbe -i, nach leichter und nach wechselnden Auslautgruppe
-ja (andi-laus, aglaiti-waurdei gegenüber wilja-halfrei)». (Seebold 1972 p. 76).
For Germanic (as for Baltic) the question remains of why a polysyllabic presents the same
syllabic behaviour as that of Sievers. This is not an objection to our thesis, but it is however
a problem. I retain that the explanation is found in the fact that polysyllabicity, whence «Neben
silbe» is secondary with respect to a phonematic structure still without vocalisation in agree
ment with its zero vocalism or vocalism reduced phonetically to zero in atomic position.
494 ALDO LUIGI PROSDOCIMI
phonetic caesura -V//Cjo-, with the only exception of - -jo- due to the
7
special nature of H in relation to the previous vowel . That shows that the
vowel after j , or in covariation after i, is not a vowel equivalent to an e/o
vowel. The use of the term apophony or samprasärana (Lejeune) is not an
explanation but a name or a label.
2.1.3 Baltic presents similar phenomenological behaviour: -is (-ys)
already referred to Sievers by Sommer (followed by Nagy 1970) through
an -ijas with contraction - for which there is no reason and which does not
take account of -is- - is explainable by linking it to the Venetic phenomenon,
with Sievers' restrictive phenomenon (cf. the supnos type) and not with
nonrestrictive Venetic.
-ys could correspond to Messapic phenomenology (see note 5), with dou
ble vocalisation which is not, however, the -ijo- of Sievers but an intermediate
stage between -ijo- and i. so that ° in i° leaves a trace in lengthening. Even
if the precise thread is not determined, it is obvious that the Messapic 'e
and the Lithuanian -ys are something special, which allow vocality of -j-
> -i- with subsequent special vocalic status.
Summing up, the 3 way scheme given at 1.5. is confirmed and it must
be completed so that the relevant features in the syllabic caesura which unifies
-VCjo- from Venetic with Sievers' -VCCjo- may appear. It also renders as
recognisable -i- (or other substitutes like Lithuanian -y- and Messapic -'e-)
and -io- as outcomes of the same conditioning.
Simplifying:
The fundamental result is that I and To are realisations of the same basic
phonetic reality. The basic reality is conditioned by phonetics and syllabic
caesura. This in its turn is morphological (variously scaled in the applica
tion of phonetic or morphological caesura), for which a single morph ap
pears differently for reasons of syllabic phonetics conditioned by mor-
phonology. That has fundamental support in the feminine where the
phenomenon has been expanded in inverse proportion to the masculine.
The prediction pertaining to the masculine comes true and it allows the
explanation of the morphological unity of the stems in -I and -jā (Brugmann,
Grdr. II, 2, pp. 286-7) by means of the difference due to phonetics affected
by syllabic caesura. A crux of the (description of the) morphology of feminine
stems in -jã and -I is thus resolved in various languages and, hence, in the
reconstruction.
496 ALDO LUIGI PROSDOCIMI
The outcome //Cf explains the origin of the genitive, originally a form
derived from stems in -o- via -joH (thus semantically a possessive) which
then entered in the paradigm. The fixing of -ï in the genitive with respect
to -jā (and -jã) which ought to coexist statistically is easily explained. The
insertion in the paradigm in / / C o brings the caesura -VC//jeH to V//CJH
> V//CiH > V//Ci (an analogous reason of variations in the position of
the system is valid for the syllabic division of the vrki;"-Flexion 2.2.3.).
In the unity of the paradigm remains the Greek -jã, which, with respect
to -jā, appears preferentially after -CC- (cf. participles) with the following
distribution:
8 Examples in De Simone St. Etr. XLVI, 1978, especially pp. 237-4. Interpretation in
Prosdocimi 1985 «Messapico» and 1986 «Sillab».
SYLLABICITY AND SIEVERS' LAW 497
9 On this see Sommer 1914 in «Abh. Leipz.» 30, 4, taken up again in Seebold 1972,
pp. 103 ff. - regarding Sievers, where, as in other cases, negative results are arrived at breaking
up the overall picture of things. Our thesis puts things back together again and explains the
three outcomes whether for the masculine or the feminine here. It also explains the three feminine
forms in Germanic where the first premise leading us to the third outcome -ja-, present at the
same time with -i- and -jā-, in ylgr (-kwjãs) can be reasonably put forward as explicable and
proven.
498 ALDO LUIGI PROSDOCIMI
dev/Δ vr//klh
because morphological because phonetic
*deiw//jH wl//kjH
derived from deiw//o- because this derivative marker
(wlk(o)- + -jH-) turned into
a 'feminine' morpheme (wlko- : wlkjH)
(Kuryiowicz) therefore with
syllabicity according to the parallel:
wl-ko-s
wl-kjH-s (for *wl-keH-)
10 The stress structure different between the two categories (of the type devi, oxytone
and baritone and of the type vrkïh always oxytone in the nominative), noted by Panini and
by others, was assumed by Lanmann «discoverer» of the two inflexions (cf. Wackernagel Ai. Gr.
III p. 163) to be explanatory. It is not an explanation but an element that must fit in to the
picture of things and must be explained. It is not stress that establishes the segmental form.
The opposite is true in as much as the suffixes (instrum., and dat. etc.) which are normally
oxytonic are not here. Evidently this is because vocalisation of -j- created a special condition.
The proof is in the genitive pitúh < *pitros (§ 2.2.6.) which, having r-, bears the stress (udat-
ta) the rest of the inflexion which has stress on the final syllable. The types pitróh and not
*pitfoh (udatta) and not +pitròh (svarita) represent the normality of a paradigm conforming
to -trV- and not to -trV-: -tuh < -tfs < -tfoS vs. +-tre/os as exceptional is explicative.
SYLLABICITY AND SIEVERS' LAW 499
2.2.5 Lithuanian feminine -ė between -I- and -ja; masculine ys between -is
and -jas.
Our hypothesis also explains the feminine forms of Lithuanian in a
500 ALDO LUIGI PROSDOCIMI
general picture in harmony with the masculine. The case of -ė between -I-
< TH and -jã < -joH, as a result of //Cj/iH0 as a result of a covariate status
between j/i with subsequent vocalisation - likely a stage such as -ïa- parallel
to -io- ) -I- in Lithuanian -ys (vs. Venetic -is § 2.1.). Lithuanian has, unlike
other languages, Venetic included, a short-list even for the masculine: -ys
< -īs, between -īs and -70s, can be explained as a result of //CioS being an
alternative to //CIØs (Lith. -is, Venetic -Cïs).
If, for Gothic and Lithuanian, the heuristic iter has separated the rele
vant phenomology, the general picture agrees with and confirms the prin
ciples if not all the details of our explanation.
On these premises, detached from Sievers in view of the rule of which
Sievers is a species, besides Germanic and Lithuanian, the other languages are
to be revised, such as Greek and Celtic (on these, see Lindemann 1982 in Et.
Celt, for a presumed genitive type vrkïh. On all this in Prosdocimi «Sillab.»).
2.2.6 A piece of general evidence: O. Ind. genit. pitúh < pitfs < pitfe/os.
e/o vocalism and stress.
The details, besides being known, are evident: why not *pitárs or
* pitras? How is it that (cfr. note 12) there is udatta on the reduced grade?
The explanation comes from application of the rules of syllabicity (//ter),
therefore for the genitive //tros, that is for Sievers //tre/os. This is the same
as wl-kiH-e/os (note also the stress) but with the vocalisation rule in Venetic
where //Cjos ) //CiØs, here //Cro/es > //Crs > -Cúh:
n The case of trisyllabic Indrah, i.e.; Indrah, is not an objection because we have -a-
of the paradigm. Rather it is a confirmation because it corresponds to the case of -i/jos which
has been generalized with respect to the only sporadic type -Cis (Venetic and other languages).
This gives us the ratio: Indrah: pituh = //Ci/jos: //Cis. For the attribution of stress, see § 3.3.
SYLLABICITY AND SIEVERS' LAW 501
Whatever -trV- may be, it is not only found in pituh but also in other
cases where pitróh, matróh is counted as trisyllabic. Despite alternative or
anti-sonantal hypotheses, (e.g. J. Schmidt 1895 «Kritik» p. 159 ff.) joining
*pitfoS > *pitfs > pitúh with the trisyllabic count of the aforesaid cases con
firms the old interpretation dependent on Böthlingk's Devanagari forms such
as pitróh, matróh. The fact that the type pitré may never be trisyllabic is
not an objection but a piece of a complex mosaic about which we detail
a few elements: pitfoS as vrkias implies the accent on the sonant and not
finally. In the case of -tfe/os the sonantization involves e/o > 0 in Venetic
Cïs. That related to swopnos/supnos vs. kwon/kuon and not *kun but rather
dhïh < *dhiHs and not *dhieH (and //tfoS > -trs ~ //kiH0s > -kías in
as much as between/and -e/os H is placed: §§ 1.5; 2.2.3.) - shows the par
ticular nature of the apophonic vowel e/o or, at least, a particular kind of
behaviour involving covariation with a preceding sonant and, both, with
fixing of the stress (§ 3.4.).
We recognise an analogous explanation for pitúh < //tros in the third
plural of the perfect -
//Cròi of the middle, with parallel and the rule //Crós
of //kwon; //kuon but not ϯ kun.
The graduation -tr(0)s, never f-tre/os, -treus and -tréi and never ϯ-tréi,
indicates - with vrkias - the history of stress; the original opposition -tfV-
~ -trV- is fixed where vocalisation is constant in order to be fixed as a con
ditioned phonetic transformation .The paradigmatic accentua
tion follows where it is not fixed or paradigmatically fixed ( )
up to the point of preventing vocalisation (pitré). That means that the ac
cent follows or is in covariation and does not precede vocalisation. That
is to say it is subsequent i.e. the nature changes (cf. Lehmann 1952 «PIE
Phon.»?) with respect to the accent that provoked the reduced grades. A
reciprocal confirm supporting the picture of things is pita, ~ mata
but , German bhr^h- (Verner): 3.4.
3.1 The rule -CC-, associated with vocalisation of H12, also func-
12 In a work in press (1984 «Latino» 2.2. and ff.) I have shown as fas as gnãtus, status,
gnotus, sprëtus are concerned, that in proximity to a sonant H selects vocalisation position
according to the rule of two consonants and that the quantity of the resulting vowel depends
on the position of vocality with respect to
In relation to consonantal behaviour, I have also included wH and jH (see above for
jHo > -ia) with the inborn peculiarity with respect to other sonants, of vocalising to i and w.
502 ALDO LUIGI PROSDOCIMI
c)
The alternation in Greek between ara — r, and ana ~ na, etc. fits in completely.
The following may also fit in perfectly: O.Ind. -ir in as much as conditioning
of the sort V ~ C predisposes matters towards a different syllabic caesura. Moreover, as for
a < nH, there are the specific problems relating to the outcome of the sonants in Indian (cf.
Schmidt 1895 'Kritik')·
Also fitting in - in fact proof of the matter in as much as it explains the unexplained
- is the alternative vocalisation as i/ī of H in Old Indian I > 0 H, ï < H0.
SYLLABICITY AND SIEVERS' LAW 503
type. However, linked to the result th from tH, there is a further implica
tion to be foreseen for all the laryngeals in themselves and in relation to
the nature of the consonant given its method of articulation: the
monophonemic laryngeal may modify it or if of the same nature 13, it may
disappear without leaving a trace.
The second may be illustrated by d (lat. dõ) in which which
voices (glottalizes: cf. pibami) a voiceless step is voiced (or glottal) in nature
and in monophonemic contact with a voiced (or glottal) stop ought to be
absorbed without leaving a trace. The prediction comes true: the past par
ticiple *dita (except for a doubtful result in Prakrit) does not exist, but we
do have -tta- in composed forms (including the exceptional participle dat-
ta 14 with preceding vowel, evidently because -VdHt- permitted syllabicity
Vtt-) to give the result -tt-. The example is of excep
tional importance because H is absorbed to the point as not vocalize
into nothing as instead happens when it is absorbed in a different form
13 I have proposed (1984 «Latino») that the numbered or conventional forms Hi, H2,
H3 correspond to the three grades of articulation. Independently of this proposal, a fact of
this type is certain for H3 (cf. Gk. ) which voices (or glottalizes) pibami.The confirma
tion, as one will find in the main text, is the fulfilled reduction as dH >
14 The question widens to take in the following dadmas, dadhmas vs. dadima dahima:
and infractions of Bartholomae's Law of the type dhattas and not *daddhas, etc. It appears
reasonable to attribute everything to the double possibility of d/dH, dh/dhH and to paradigmatic
restructuring {da- is not doubling of the present and dad- also functions as a CVC root. On
this see elsewhere).
504 ALDO LUIGI PROSDOCIMI
ty and as far as stress is concerned, either the rule of swopnos: supnos (stress
on the syllable with apophonic vowel, whatever its origin may be) or if it
is not the same, the rule *pitr0s, vrkias ( 2., 2.3., 2.2.4.)
REFERENCES
15 Cf. G. Schmidt in 87, 1973, pp. 36-83. The question - Grassmann included -
merits reconsideration in light of possible or gH (the latter is less probable given the rules
of compatibility between dh and g).
SYLLABICITY AND SIEVERS' LAW 505
NORBERT REITER
Freie Universität, Berlin
Der Inhalt von krug ist selbst schon reichlich allgemein. Wahrschein
lich war er in sehr alter Zeit noch um einiges spezieller; denn krug = krçgz
wird mit anord. hringr «Ring» zusammengebracht.
Die Etwicklung im Slavischen hat eine Parallele im gallisch-germa
nischen Bereich. Deutsch rund geht in letzter Instanz auf lateinisch rotun-
dus zurück, das sich zu rota «Rad» stellt. In beiden Fällen scheint die Ent
wicklung ihren Ausgang von kreisrunden Gegenständen genommen zu ha
ben, dem Rade bei den einen, dem Ring bei den anderen.
Man sieht, jedes dieser drei Beispiele hat seine individuelle Note, und
so ist es bei allen anderen Fällen. Das Russische hat von dieser Art etwa
80 Paare, gezählt allerdings nur solche, an denen ein Adjektivum auf -ovyj
beteiligt ist. Die Beteiligung der -ovyj-Adjektive spielt für die Aufklärung
des Vorganges insofern eine Rolle, als dieser Typ der jüngste in der russi
schen Sprachgeschichte ist. Hierzu einige Erläuterungen:
Das Russische verfügt zur Bildung sog. Beziehungsadjektive über ein
knappes halbes Dutzend von Suffixen. Ich nenne die wichtigsten:
1. --, wie in dužnyj, allgemeinslavisch, mindestens seit tausend Jahren,
von allen einschlägigen Suffixen am häufigsten gebraucht. Das Russi
sche verfügt über schätzungsweise 25.000 Adjektive mit --, Suffix
-- hat keine spezielle Semantik, es dient lediglich zur Bildung von Ad
jektiven.
2. -'anyj, wie in ledjanój, ebenfalls allgemeinslavisch, außerhalb des Rus
sischen vertreten als -en-, sehr alt, doch weitaus weniger gebraucht. Das
Russische hat Überschlagsweise 300 Wörter mit diesem Suffix. Es gilt
als seit langem schon unproduktiv. Es hat eine speziellere Bedeutung.
Sie wird gewöhnlich mit «gemacht aus...» angegeben.
3. -ovyj geht auf die u-Deklination zurück, ist nach dieser Herkunft das
jüngste der Suffixe. Es ist etwa im 12. Jh. aufgekommen und nur im
Ost- und Westslavischen vertreten. Es hat keine spezielle Semantik. Darin
ist es dem Suffix -- gleich. Das Russische verfügt bzw. verfügte über
insgesamt an die 3.200 Adjektive auf -ovyj. Durchgesetzt hat sich -ovyj
etwa im 15./16. Jh., stark gebraucht wurde es im 18. Jh. und später bis
in unsere Tage.
Die eben skizzierten historischen Umstände der drei wichtigsten Adjek
tivsuffixe des Russischen sind für die Beurteilung der semantischen Ent
wicklung der Adjektive im Russischen insofern von Belang, als im Falle
510 NORBERT REITER
von Adjektivdubletten diejenige auf -ovyj, und das ist gewöhnlich die
jüngere, sich semantisch mit dem Grundwort deckt, während die ande
re auf -1- oder -'anyj, und das ist die ältere, semantisch vom Grund
wort abgerückt ist. Bei den im Russ. bekannten etwa 80 Paaren ist das
zu ca. 70% so. Bei den verbleibenden 30% ist es umgekehrt, doch zeigt
sich, daß in diesen Fällen das -ονχ/-Adjektivum das ältere und das auf
-- das jüngere, hinzugebildete ist.
Für die Erklärung des Phänomens haben also die Dublettenpaare mit
verschiedenalten Suffixen hochrangige methodische Bedeutung. Dieses Phä
nomen ist es, das in der sowjetischen Sprachwissenschaft «okačestvlenie ot-
nositel'nych prilagaternych» gennant wird, und damit wollen wir uns nun
näher befassen.
«Okačestvlenie» oder - wie ich es übersetzt habe - «Vereigenschaftung»
verrät durch sich selbst die theoretische Grundlage, auf der der Vorgang
sowjetischerseits erklärt wird. Sie besteht in der Annahme, daß die Haupt
wortkategorien ontische Gegebenheiten mitteilen, Substantive also «Substan
zen», Adjektive «Eigenschaften», Verben «Tätigkeiten». Demnach wird mit
«okačestvlenie» oder «Vereigenschaftung» gemeint, daß sich ein Adjekti-
vum semantisch zu einem wahren Eigenschaftswort entwickele und am En
de seiner Entwicklung einen ontisch anderen Tatbestand signalisiere als zu
Anfang. Mit dieser Annahme gerät das Problem in eine Sackgasse. Mit «Ei
genschaft» ist wissenschaftlich Endstation. Es bietet sich keine Möglichkeit
mehr, übergreifende Zusammenhänge aufzudecken. Das alles nur, weil man
sich bei der Einschätzung der Hauptwortkategorien von traditionellen Vor
stellungen leiten läßt, aber die sind falsch.
Die Wort kategorien signalisieren keine ontischen Tatbestände, vielmehr
handelt es sich dabei aller Wahrscheinlichkeit nach um kommunikations-
steuernde Zeichen, anders gesagt: Um Informationen zur Informationsver
arbeitung, die mit der durch ein Wort gegebenen Sachinformation grund
sätzlich nichts zu tun haben und daher einer jeden entsprechend den kom
munikativen Bedürfnissen beigegeben werden können. Daher ist es auch
möglich, daß ein Sachbestand auf verschiedene Weise angezeigt wird: sub
stantivisch, adjektivisch oder auch verbal. Deutsch eisern gibt keine andere
Sachinformation als das Eisen, das Laufen keine andere als laufen. Übli
cherweise treten die substantivische und die adjektivische, sodann die sub
stantivische und die verbale Anzeige zusammen auf. Es gibt aber auch eini
ge Fälle, wo alle drei üblich sind. Bei Farbbezeichnungen findet man das.
ENTWICKLUNG SEMANTISCHER INHALTE 511
Vgl.: der grüne Baum, ein lichtes Grün; die Bäume grünen. Auch im Slavi-
schen gibt es eine ganze Reihe solcher Tripletten.
Da die Wortkategorien mit den mitgeteilten Sachbeständen nichts zu
tun haben, ist auch die in der Slavistik übliche Zweiteilung der Adjektive
in sog. «Beziehungs-» und «Qualitätsadjektive» nicht gerechtfertigt, und
das sogar schon dann, wann man sich an die traditionelle Auffassung hält;
denn wenn Adjektive «Eigenschaften» zu signalisieren hätten, dann dürfte
es nur Qualitätsadjektive geben und wäre für «Beziehungsadjektive» gar kein
Raum, aber dazu versteht man sich in der Slavistik nicht, weil eben doch
gewisse Unterschiede zu bemerken sind. Aufzudecken, worin dieser Unter
schied in Wahrheit besteht, ist mein Anliegen hier.
Wörter wie ledovój, dolgovój, krugovój und unzählige andere i n d
Adjektive, und sie sind nicht adjektivisch verkleidete Substantive, wie oft
gemeint wird. Sie für verkappte Substantive zu halten, ist falsch, da sie den
selben semantischen Inhalt signalisieren wie das dazugehörige Substantiv,
von dem sie per Suffigierung formal auch abgeleitet sind. Ledovój signali
siert keinen anderen Sachbestand als lëd, nur geschieht es das eine Mal in
adjektivischer, das andere Mal in substantivischer Anzeige.
Obwohl an den drei genannten Beispielen festgestellt worden war, daß
jedes seine eigene Individualität hat, in e i n e r Hinsicht sind sie ein
ander aber gleich: Das eine der beiden Adjektive gibt den Inhalt des Sub-
stantivums in allen seinen Komponenten wieder, das andere aber nur einen
davon. Wir können es auch so sagen: Das eine Mal ist der Inhalt komplex,
das andere Mal einfacher. «Einfach» und «komplex» sind Skalenbegriffe.
Sie beziehen sich auf die Minus- bzw. Plusrichtung auf einer Skala, die wir
die «Komplexitätsskala» nennen können. So können wir sagen: Der seman
tische Inhalt von ledovój ist komplexer als der von ledjanój und dieser wie
der einfacher als der von ledovój.
«Einfach» und «komplex» beschreiben semantische Beschaffenheiten;
unter pragmatischem Gesichtspunkt könnte man von «allgemein» und «spe
ziell» sprechen. Das Speziellere hat einen geringeren Anwendungsradius als
das Allgemeinere. Das ist eine auch außerhalb des Sprachlichen geltende Re
gel. Je mehr Bedingungen an etwas gestellt werden, desto schwieriger wird
es, ein Exemplar zu finden, das alle gestellten Bedingungen erfüllt.
Und nun verbinden wir die systematische Feststellung mit der histori
schen Beobachtung:
512 NORBERT REITER
Von einem Dublettenpaar hat den einfacheren oder auch allgemeineren In
halt gewöhnlich das ältere Adjektivum, während das jüngere den komple
xeren, spezielleren signalisiert, der zugleich der des Grundwortes ist.
Demnach scheint die Entwicklung vom Speziellen zum Allgemeinen ei
ne Funktion der Zeit zu sein. Da nun aber - gerade auch am Russischen -
zu beobachten ist, daß die Entwicklung mal schneller, mal zögerlicher vor
sich gehen kann, d.h. die schnellere Entwicklung zum allgemeinen Inhalt
auch bei den jüngeren -ονyj/-Adjektiven auftritt und vornehmlich im 18. Jh.
beginnt, kommen wir zu dem Ergebnis, daß die Entwicklung nicht eigent
lich eine Funktion der Zeit, sondern der Anzahl der ein Adjektivum gebrau
chenden Menschen ist. Also: Je mehr Leute ein Adjektivum gebrauchen,
umso schneller entwickelt es sich semantisch vom Speziellen zum Allgemei
neren hin.
Auch das ist eine außerhalb des Sprachlichen beobachtbare Regel: Je
mehr Leute an einer Sache beteiligt sind, desto geringer wird die relative
Trefferquote. Die Treffer diffundieren, das arithmetische Mittel der Tref
ferwerte sinkt ab. Das etwa ist die Situation, in der sich die Semantik des
länger gebrauchten Adjektivums befindet. Die Einfachheit seines Inhaltes
entspricht etwa dem Absinken des Mittels der Trefferwerte. Die Entwick
lung hat zur Folge, daß die semantische Verbindung zwischen Adjektivum
und substantivischem Grundwort immer lockerer wird. Ledjanój hat mit
lëd, krúglyj mit krug nicht mehr viel zu tun. Sofern nun weiterhin Bedarf
danach besteht, den vollen Inhalt des Grundwortes auch adjektivisch zu sig
nalisieren, wird er durch die Bildung eines neuen Adjektivums befriedigt,
was umso besser geht, je mehr Suffixe zur Derivation bereitstehen. Ein sol
ches jüngeres Suffix ist im Russischen - ovyj, und so wird der Bedarf mit
Adjektiven dieses jüngeren Typs abgedeckt. Es entstehen ledovój, krugo-
vój, dolgovój usw.
Zurück zum «okačestvlenie». Dieser Begriff beruht nicht einfach nur
auf der traditionellen Auffassung von den Wortkategorien, sie wird dar
über hinaus durch eine optische Täuschung begünstigt. Sie besteht in einer
statistisch eindeutigen Zuordnung zwischen den Extrembereichen der Kom
plexitätsskala und den Wortkategorien. Das heißt: Hochkomplexe Inhalte
werden mehrheitlich substantivisch, die einfachen adjektivisch angezeigt. So
entsteht der Eindruck, als würden sich die Adjektive von der Dinghaftig-
keit weg zur wahren Eigenschaft hinentwickeln und erschiene somit «oka
čestvlenie» auch aus dieser Sicht gerechtfertigt, was es freilich nicht ist.
ENTWICKLUNG SEMANTISCHER INHALTE 513
LITERATURVERZEICHNIS
lej russ. jaz.ped. institutov Moskv. zony 25-26, okt. 1971. Moskva
(MGPI) 1973. 46-52.
Landsman, I.M., Process okačestvlenija ot substantivnych otnositePnych
prilagatePnych v russkom jazyke. AKD Taškent.
Potemkina, A.I., O perechode otnositePnych prilagatePnych v kačestven-
nye, in: Učennye zapiski Moskovskogo gos. ped. inst. inostrannych ja-
zykov im. M. Toreza. Moskva 1968, 4, I, 149-167.
Reiter, N., Die ovyi-Adjektive in Russland. In: Osteuropa-Institut an der
Freien Universität Berlin, Slavistische Veröffentlichungen, Bd. 61. Ber
lin 1986.
Werner, O., Zum Problem der Wortarten, in: Sprachgebrauch und Sprach
system. Festschrift für Hugo Moser. Düsseldorf 1975. Teil 2, 432-471.
REIT
A PERFORMANCE MODEL FOR A NATURAL THEORY
OF LINGUISTIC CHANGE
ELKE RONNEBERGER-SIBOLD
University of Freiburg, Fed. Rep. of Germany
0. Introduction
The basic idea of any natural theory of linguistic change is that languages
do not change randomly or due to certain entirely system-inherent evalua
tion criteria, but that languages are changed by their users in order to
facilitate communication. This means that the language users are constant
ly adapting their «communicative instrument», i.e. their language, to their
needs of production, e.g. the need for ease of pronunciation, and percep
tion, e.g. the need for unambiguity. (The needs of language acquisition, often
mentioned in this context, coincide with some of these.) This idea, which
is in fact an old one 1 , underlies not only Natural Phonology and Mor
phology in a narrower sense 2 , but also several other recent functionalist
contributions to the explanation of linguistic change, proceeding from very
different viewpoints, e.g. information theory and information processing
(Lüdtke 1980 and, less formalized, Werner 1977, 1984 and Ronneberger-
Sibold 1980), typology and statistics (Altmann/Lehfeldt 1973, 1980),
language universais research (the Cologne project, programmatically Seiler
ed. 1978), pragmatics (Givón 1979), formal logic (Vennemann 1975), «func
tional grammar» (Bossuyt 1983), notwithstanding their substantial theoretical
differences in other respects.
As part of the theoretical framework for such performance-based
1 For references to traditional work in this vein in phonology see Donegan/Stampe 1979,
126 and passim. For the topics to be discussed here, Jespersen 1894, 1941, Frei 1929, Zipf
1936, 1949, and of course functional structuralism - the Praguians and Martinet - are of special
interest. For a discussion see Ronneberger-Sibold 1980.
2 E.g. Stampe 1980, Donegan/Stampe 1979, Dressier 1980/1983 for Natural Phonology,
Hooper 1976 and Vennemann (in a diachronic perspective especially 1972) for Natural Generative
Phonology, Mayerthaler 1981, Wurzel 1984 on Natural Morphology, Dressier 1985 on Natural
Phonology and Morphology.
518 ELKE RONNEBERGER-SIBOLD
3 All operations referring to syntax (including congruence) and the respective needs are
excluded here. (But cf. Ronneberger-Sibold 1980, 134 ff).
A PERFORMANCE MODEL FOR LINGUISTIC CHANGE
519
Figure I: Production and perception of Engl, (these) houses and Lat. (hae) domus.
520
ELKE RONNEBERGER-SIBOLD
places the meaningful elements [house] and [PL], here called morphemes,
(which he has produced on higher levels of production not included in the
model) by the corresponding morphs / h a u s / and / - z / 4 . These could be
called his primary phonological intentions.
In English, this mapping operation is performed in the easiest possible
way, because to each morpheme corresponds just one morph (leaving for
the moment out of consideration the few completely irregular forms like
feet etc.)· For a speaker of Latin going to say domūs (Nom. PL), however,
the task is more difficult. (This is expressed by the fat lines in figure I,
denoting operations which are not performed in the easiest possible way.)
The speaker of Latin must first combine [PL] and [Norn.) into one so-call
ed portmanteau-morpheme, and then choose the corresponding morph
/-ūs/ from among a variety of allomorphs including
etc. 5
So we may note as a very important need on this level of production
the representation of one morpheme by one morph (syntagmatically and
paradigmatically). Note that this does not imply the converse, one morpheme
for each morph. This would mean unambiguity and is a need of the hearer,
not of the speaker 6 .
4 I take / - z / as the underlying form of the normal Engl. Plural suffix following the in
terpretation of the evolution of the Engl, plural rule by Keyser/O'Neil 1980. All informations
about diachronic and synchronic facts concerning Engl, plural formation as well as the phonetic
transcription used here are taken from Brunner 1960.
5 As in the case of many other mental functions, our main source of evidence for ope
rations of this type by the human brain are the cases where they do not function normally,
e.g. slips of the tongue (possibly under the influence of fatigue, alcohol etc.), aphasia, lan
guage acquisition, and of course language change, which always starts as an «error». (Another
important source of evidence is the adaptation of loanwords.) For all these situations characte
rized by reduced capacities for brainwork, planning errors e.g. in the choice of allomorphs
are typical. Normally, more regular allomorphs are substituted for less regular ones, e.g. mow-ed
for mow-n. The fact that this sort of errors occurs in situations where the speaker is just not
monitoring his production in favour of the hearer, is empirical evidence against the claim of
Mayerthaler 1981, 62, that production plays no role in the evaluation of morphological
naturalness.
6 This asymmetry between the speaker and the hearer is not properly expressed in the
often-cited formula «one meaning: one form», which suggests a biunique representation of
morphemes by morphs. Unambiguity becomes an aim for the speaker only indirectly as a result
of negative feedback from the hearer. So, if he is often misunderstood, because he has used
a particular ambiguous morph, he will tend to avoid this morph, but not ambiguity in general.
522 ELKE RONNEBERGER-SIBOLD
So, as is well known, in the diachronic development of the expressions of grammatical categories,
ambiguity is avoided, where it would handicap the hearer. (E.g. «subject» and «direct object»
referring to [ + human] nouns tend to be kept apart by morphology and/or preferred word-
order, whereas homonymy between endings of the nominal and verbal paradigms (e.g. Engl.
/-z/) does not cause any reaction of the system because there is in most cases enough disam
biguating context. This difference would be inexplicable if both, absence of allomorphy and
unambiguity were needs of the speaker or of the speaker and the hearer. (That they cannot
both be needs of the hearer alone is shown in note 5, above). Our assumption is that the need
for «1 morpheme →· 1 m r p h » (absence of allomorphy) is shared by the speaker and the hearer,
whereas unambiguity is a need of the hearer only.
7 This is in fact not entirely true, because it is blocked in the genitive plural for the se
cond underlying / - z / . This is a slight morphological constraint.
8 The difference lies only in the source of the obligation to do so and in the degree to
which these operations serve the need for ease of pronunciation.
A PERFORMANCE MODEL FOR LINGUISTIC CHANGE 523
speaker has to produce all of them, if he does not want to infringe upon
the norm (in the sense of Coseriu (1969) 1971). For the hearer, however,
the distinctive features are of course much more important than the non-
distinctive ones 9.
The basic need of the articulation level aims of course at ease of pro
nunciation, which comprises in fact several needs concerning the structure
and arrangement of segments and the length of utterances (cf. Ronneberger-
Sibold 1980, 186 ff.) For the sake of brevity, these needs are subsumed here
under the heading «ease of pronunciation».
The hearer's first task on his way back from sound to meaning is the
recognition of the smallest distinctive segments in the sound stream, i.e. the
identification of the taxonomic phonemes. In actual speech processing this
must be a very complex operation, where segmentation and identification
are interlaced. Also, this level may be left out in many cases by immediate
recognition of whole syllables or morphs, and the respective strategies may
depend to a certain degree on language structure. We cannot go into these
details here.
Clearly, this task is the easier to perform, the more each segment perceiv
ed resembles the ideal sound shape of the phoneme it has to be matched
with and only that one: In other words, the phones should be unambiguous
in relation to the taxonomic phonemes. In our examples this need is fairly
well satisfied, because there are no extrinsic allophones.
Unfortunately, the identification of the taxonomie phonemes does not
lead the hearer all the way back to the level of systematic phonemes. This
is due to possible neutralizations on the part of the speaker during mor-
phonological planning. So, the hearer will e.g. identify both sibilants in
[hauziz] as the taxonomie phoneme / z / . To identify the first one as the
systematic underlying / s / , he has to recognize the whole morph / h a u s / im
mediately connected with the morpheme [house]. The recognition of mor
phemes, however, is a complicated operation, which cannot be treated in
detail here. We just retain the needs for unambiguous phonemic represen
tations and the need for unambiguous morphs. As shown by the dotted lines
in figure I, the first one is not satisfied in the case of /hauz/, the second
one in the case of Engl, / - z / and Lat. /-ūs/.
10 For the most fundamental conflict between speaker and hearer (ease and necessity
to be understood) cf. Martinet 1955, 94 ff., and Liidtke 1980, 5. For conflicts with respect
to the structure of phonemic systems, cf. Martinet, loc. cit., with respect to processes and rules
in Natural Phonology Donegan/Stampe 1979, 129 ff., for conflicts within morphology Wurzel,
1984, 205, between morphology and phonology Martinet 1960/1963, 165 and Werner 1977,
1984 (mnemonic vs. articulatory ease), Mayerthaler 1981, 43 and Wurzel, loc. cit. (phonetic
naturalness vs. morphological naturalness) and Vennemann 1972 (phonetic analogy vs. con
ceptual analogy). Enumerations and explanations of various conflicting needs on different levels
of production and perception are to be found in Givón 1979, 268 f. and, in the older literature,
very explicitely in Jespersen, especially 1941, and Frei 1929. (This list is of course far from
being exhaustive.) All of these authors refer (explicitly or implicitly) to conflicts between dif
ferent performance needs.
π Krahe 1969, 24.
A PERFORMANCE MODEL FOR LINGUISTIC CHANGE 525
realized in some stage of Middle English with the pluralmorph /-iz/ after
all consonants. This was easy to plan on both levels but longer to pronounce
than the Modern Engl, variants [-s] and [-z].) Only a very complete satisfac
tion of all three needs together will inevitably result in difficult recognition
of the phonemes. This is symbolized by the converging arrows with an in
verted head and the symbols for conjunction in figure II. To a certain ex
tent this impediment to phonological processing can be compensated by the
higher levels of processing, but if also this «reserve» is used up, the absolute
limit of effort reduction for the speaker is reached: Communication will
be disturbed. In this sense, the hearer's need for unambiguity has an ab
solute priority over all other needs.
Note that the symbols in figure II denote purely implicational and con
junctive relations, namely: The more you satisfy the need at the tail of the
arrow, the less the need at its head will be satisfied, unless the decrease of
effort is compensated by an increase with respect to one of the other con
junctively connected needs.
This is a very typical constellation of relations among different needs
- though by no means the only possible one. There are several other types
of relations, just as there are many other needs which cannot be explained
here, such as the needs concerning the inventories of linguistic units, the
length of utterances, and the order of words and morphemes. (Cf.
Ronneberger-Sibold 1980, 177 ff.) Rather than going into these details, we
will now focus on the question of how this type of performance-based model
can contribute to the explanation of linguistic change.
the number of competing possible developments, but at the same time one
has to be aware of the fact that the goal will never be entirely reached 12.
So, of two explanations accounting for the same facts, the one is to be prefer
red which leaves less possibilities «that it might have been otherwise» (Lass
1980a, 9). But the fact that even the best explanation leaves a rest of in-
determinism should not keep us (as suggested by Lass) from accepting it
as an explanation. (Cf. the discussion between Lass 1980b and Dressier 1980.)
12 This may be due to the practical reason that we are unable to know all the facts which
have played a role in a given linguistic change, or to the more principled reason of human
freedom to behave in a non-causal way, or to both.
13 E.g. Dressier 1985, 168 ff. on palatalization in Italian and Polish, Wurzel 1982 on
examples from Germanic languages, and Hooper 1976 from Spanish.
14 Quantitative answers to this type of questions are provided by statistical typological
investigations like e.g. those of Altmann/Lehfeldt 1973, where the limits for certain mor
phological properties are determined for different classes of languages.
A PERFORMANCE MODEL FOR LINGUISTIC CHANGE 527
15 On the advantages and disadvantages of different language types for the different per
formance needs cf. Ronneberger-Sibold 1980, 137 ff. and, more concisely, 1976.
16 This explains, by the way, the logically unfounded but nevertheless often observed
affinity between inflection and allomorphy: Both rely on a tendency to expend the necessary
effort on the morphosyntactic planning level rather than on the lower levels. This comes very
close to the remark by Skalička (1951)/1979, 47, that (put in our words) the disregard of the
need for «1 morpheme → 1 morph» syntagmatically speaking (by the formation of portmanteau-
morphemes) facilitates the same disregard paradigmatically speaking (allomorphy).
528 ELKE RONNEBERGER-SIBOLD
on the articulation level in the strong group all those would be accepted which
create new distinctions or strengthen already existing ones. In the weak group,
on the contrary, all sorts of neutralizing changes would be accepted. This
is exactly what happened in the ancient Germanic languages with their two
vowel systems, an expanding one in stressed and a more and more reduced
one in unstressed position, and with the well-known consequences for their
inflectional systems. (Cf. Sigurd 1961.)
REFERENCES
H.B. ROSÉN
Hebrew University, Jerusalem
mulated the properties of a root on grounds of the fact that the sequence
of the radical elements with an addition whatsoever can form a syllable,
a feature that is due to a particular restriction providing a limited selection
of the medial root element to a class of phonemes that may or may not serve
as syllable centres, that is the sonants. No such restriction is prevalent in
Semitic, where every phoneme admitted at the extremity of a root is also
admissible as its medial element. No sonant phoneme, in fact, exists in
Semitic, as a consequence whereof, Semitic has no morphological Abstufung
and no zero degree can be spoken of there in contrast to a Vollstufe. The
morphological mechanisms vested in Semitic more on so-called internal
change than on affixation, as they are in Indo-European, are also a result
of this fact, since in Semitic there is an interplay of monosyllabic and
disyllabic bases of one and the same root. On the other hand, every one
of the three radical elements in Semitic can be constituted by one single non-
vocalic phoneme only, while, as we know, Indo-European roots may have
composite radicals initially and finally, subject to the condition, however,
that these groupings should not infringe on syllable structure, that is that
an initial element should be of a sonority-ascending sequence towards the
root centre, while the opposite should be the case for the root-final compo
nent, the central radical alone, the one preceding which all apophony takes
place, never being composite. This feature allows the distinction of roots
otherwise of identical or very similar phonological composition still to be
distinct, such as the two Verbal roots *g-n- «give birth» and *gn-... «know»
distinguished solely by strength of the relative position of the nasal, being,
in one case, the medial element itself and subject to apophony, in the other
case preceding the medial radical and not partaking in an apophonic
mechanism, since it constantly occupies a syllable-explosive position follow
ing the root initial occlusive. Analogous statements may be made, for in
stance, with reference to pairs such as *sw-d- «be pleasant» and *s-wd-
«sweat», *st-r- «be solid» and *str- «spread out» and the like.
It is obvious that I am not presenting IE. roots in a fashion consistent
with Benvenistian root-theory: principally, we are using the term «root» for
what would include by that theory the racine, the suffixe and the élargisse
ment, and we are not ascribing any meaning-preserving function to a posited
alternation of a thème II. This is not the occasion to set out in detail what
could be considered as the advantages of our terminological and structural
approach; it is only our terminological usage I intended to make clear. What
is most important appears to be the fact that our presentation allows a
ON «NORMAL» FULL ROOT STRUCTURE 537
ι Ammer 1950-2: 213: «Die idg. Wurzel besteht aus mindestens zwei Elementen, d.i. der
zweimaligen Aufeinanderfolge von Konsonant und Vokal».
2 Kuryiowicz 1935: 121: «La definition phonétique, plus féconde que la définition séman
tique, est la suivante: la racine est la partie du mot... qui comporte 1) la consonne initiale ou
le groupe consonantique initial, 2) la voyelle fondamentale, 3) la consonne finale ou le groupe
consonantique final». Cf. also Rosén 1961: 196 (= 1982:153).
3 Cf. Rosén 1968: 369f. (= 1982: 241f.).
538 H.B. ROSEN
of radicals in common can be considered to be the same: that is, if the two
interchanging radicals are cognate phonemes, phonemes which stand in some
relation to each other by virtue of a phonological rule 4 . Conditions of that
sort seem not to be prevalent in Indo-European, and consequently, any
theory of root determinatives has as little to stand on in Indo-European as
it has in Semitic. What we are concerned with are roots of reduced volume
or those that were such at least originally. They are found in both language
families that concern us here, and their status and history gives ground to
some typologically comparative observations.
The first and, as it seems, most important one, concerns the status of
these roots in the semantic framework of the language. These roots prevail
in the names of essential concepts of life 5 . In Semitic, verbal morphology
is such that it is only with considerable difficulty that biradical bases can
be maintained in inflection, and we shall see that many verbs undergo pro
cesses by which they can adapt themselves to the requirements of inflec
tion; that is why in Semitic we have parts of the body, Heb. yåd, Akk. qãt-,
«hand», Heb.. «tooth» ś-p-, «lip», ρ- «mouth», r-š «head», p-n- «face»,
c
-n- «eye», d-m «blood», š-t «the behind», to which we may also add šem
«name», as an individual characteristic of the human; verbs that may pro
perly be described as semantically primitive have a basically biradical root
structure: t-η- «cause to be, give», -k-1- «eat», - - «seize», Heb. Ar. -r-?-
«see», Aram, h-z- «id.». This group is matched in Indo-European by verbs
which describe the elementary physical activities, functions and positions:
with no initial occlusive *X-d «eat», (in zero-degree *d- in dens etc. «tooth»),
*-y- «go», *-r- «move» (intr.), *X-q- «move» (trans.), *X-q- «see», further
*st- «be positioned», *p-X- «drink», *s-d- «sit» with its causative *si- «set»,
(?) «march» etc., Slav, xod-), *k-i- " *dr- «run», . ■
«speak», *gw-m- «go», *s-q- «go after», *su- or *pu- «give life to», *g-n-
(give) birth», *m-r- «die», *p-t- «move fast», *l-u- «wash», *t-g- «cover»,
«touch (?)», and also by some nouns such as *p-d- «foot», -s- «hand»
enlarged by a probable - suffix as in χειρ, kesseras, hasnam; further
more, there are again the semantically «primitive» notions as *X-s- «be»,
-u- «exist». Then there are the basic kinship terms: «father»,
«mother», bcen «son», ?ah «brother», ham «father in-law», d-o-d «male
relative in non direct blood relation», cåm «head of a people», metonymically
«people», sår with a feminine suffix -at «sanctioned concubine»; not only
does this selection effectively illustrate the categorization of the desert-Semitic
family unit, but it also shows — through the absence of corresponding IE.
counterparts — that the IE. family was not organized by blood relation
ship, but by social and economic functions, the terms in question being nor
mal derivatives of verbal roots denoting the mutual functions of the members
of the family, such as pater, bhratr - etc.; the economic nature of this seman
tic field in IE. is underlined by the existence of a verb with «deficient» root:
d-m- «build a house».
We may add primitive objects of nature: m- «water», «tree», cf.
in IE. *w-d- with no initial occlusive «water», *dr-u- «big tree»; some time-
measuring units are of similar structure: y-m «day», š-n- «year» 6 .
Basic numerical terms have structures that do not correspond to nor
mal root patterns: Aram, had, IE. *s-m- or *Xoi- enlarged by -n- or -k-
«one», Sem. *Θη-, IE. *dw- «two», IE. tr-i- «three», also Semitic «six»
and IE. non-apophonic «ten» may be added, underlining the essential
dissimilarity of the respective bases of the numerical system. But Semitic
has also a root *m-? «hundred», a value for which IE. already has a designa
tion based on some full radical lexical entity.
Pronouns are radically deficient in both language families: Sem. š- or
h- «third person» or anaphoric, demonstrative IE. X-i-, q-, *gh-, *k - etc.,
and the same goes for adverbial notions such as C-I «above», m-η- «out of»,
etc. in Semitic, *X-n «within», *X-k «without» and the like in IE.
The integration of these terms in the morphological systems of the
language does not present any difficulty as long as they can be made to fit
into normal derivational and inflectional patternings. While this is essential
in Semitic for verbal and nominal lexemes alike, it pertains, in IE., to ver
bal stems alone. The mechanism of this adaptation have to be such that a
deficient and less voluminous root assimilates itself to a full and normally
structured one. Where the supplementing elements are taken from, we will
shortly examine in the following.
The root is optionally expanded 7 by an element often alternating with
zero, so that such an extension can be interpreted as a phonological feature
finitive or verbal noun d-åcat), while verbs whose final radical element is
a vowel, probably itself stemming from a root expansion (the so-called ver
ba IIae infirmae), have an infinitive of feminine shape (e.g. qånå «bought»:
-qno-t inf.) and compulsorily transport the thus emerging consonantal ex
pansion to the non-substantival 3rd person fem. sg. form of the perfect:
qånå-t-å (only the final -å is the normal ending of this form).
As to the analogous Egyptian feature, it consists in the occurrence, in
the Coptic infinitive, of a morphophonemically unstable -e, which goes back
to an Egyptian -it or -ut, e.g. Sahidicjese «lift» < Eg. ts-t; it is not immaterial
to note that the verbs belonging to this class show, in pre-Coptic Egyptian,
an expansion of the bi-radical base by means of consonantal lengthening
rather than by a vocalic extension 10. One also finds, as root pattern nor
malizes, morphemes functioning as characterizers of parts of speech (e.g.
Middle Hebrew t-r-m «contribute» involving the nominalization morpheme
t- and the original bi-radical root r-m «elevated», in the causative «lift, bring
up as an offering») or of verbal valence (e.g. Heb. š-k-n «dwell», trans,
«settle», involving the causative morpheme 5-, probably from another dialect,
and the original biradical k-η «be (stable)», cf. Ar. kana, dealt with above,
p. 540) n ; in Indo-European, aspectual morphemes can be observed in this
phenomenon: διδάξαι shows a consonant cluster transposed («verschleppt»)
from the incohative present διδάσκειν. If we use these findings for the pur
pose of explaining several not altogether lucid root variations in Indo-
European, we can include in this category not only the well-known
phenomena of univerbation (soluo (*se-luo, fressen (ver-essen, ,
nidus <*ni-sd-os, cf. Nest)13, but also some verschleppte nasals,
not only inside verbal paradigms, as in the familiar type of iungere, but also
taken over into nominal forms such as , panth (Lat. passus indicates
that the origin of does not involve n) or coniunx14; m- expansions
may be explained as hailing from a deverbative nominalizing suffix as in
(for which, admittedly, other explanations may also be of-
REFERENCES
THOMAS F. SHANNON
German Department, University of California, Berkeley
1 For present purposes, sonorants and vowels are disregarded and only obstruents are
considered. The longstanding controversy over the relation between tense/lax (fortis/lenis) and
the voicing distinction is also left aside.
546 THOMAS F. SHANNON
2 Hence, the voicing opposition is usually said to be neutralized here. However, this
claim has recently been disputed; cf. Fourakis (1984) and literature cited there on this issue.
3 To save space, sources are omitted except for languages where FD was found; essen
tially they are those of Ruhlen (1975) and Maddieson (1983). Besides the languages mentioned
in the text, the following were also investigated: Baltic - Lithuanian; Germanic - Danish, English,
Icelandic, Norwegian, Swedish, Yiddish; Romance - French, Italian, Portuguese, Rumanian,
Spanish; Celtic - Breton, Irish, Welsh; Greek; Albanian; East Armenian; Basque; Finno-Ugric
- Estonian, Finnish, Hungarian, Lapp; Caucasian - Abkhaz, Avar, Circassian, Georgian, Kabar-
dian. There was not enough information available on Breton, Basque, or Lithuanian to decide
whether they have FD: although they all neutralize voicing in final obstruents, my sources do
not indicate whether they have MP alternations. This list is, I believe, the most complete one
compiled to date, although it is surely not exhaustive. The author welcomes further informa
tion from interested linguists.
4 Judging from the data in Zolkhoev (1980), other Altaic languages, viz. Buriat,
Kalmyck, Mongolian, Tuvinian, and Yakut may also have FD. They neutralize the voicing
opposition finally, and there are even MP alternations; but it is not clear whether these are
due to FD or a voicing rule.
FINAL DEVOICING 547
5 Although in many languages final sonorants are devoiced, either partially or optionally,
this devoicing always appears to be purely phonetic. I know of no instances where a phonemic
distinction is neutralized.
6 This claim may in fact be empirically false, however. Goossens (1977) claims that
fricatives devoice first in certain Dutch dialects.
548 THOMAS F. SHANNON
tions. Further relevant data can be found in (West) Frisian, where Tiersma
(1979:166) finds that «at the beginning of this century, voiced obstruents
regularly occurred following short vowels and rising diphthongs, and were
regularly devoiced after a falling diphthong. When preceded by a long vowel
or sonorant consonant, there was free variation between the voiced obstruent
and its voiceless counterpart». Unfortunately, I know of no similar syn
chronic restrictions.
supraglottal pressure associateci with them would impede voicing, which re
quires greater subglottal pressure. Although this approach takes over where
Stampe's leaves off and succeedes in motivating certain facets of devoic-
ing, e.g. the lack of voiced velars in many languages (due to the greater im-
pendance of voicing associated with articulations farther back in the oral
track), it runs into problems with FD. It would seem to predict that velars
should devoice most readily, then dentals, then labials; but as was noted
earlier (cf. Dinnsen 1980), there is no empirical evidence in support of this
claim. Also, the proposal by itself provides no explanation for why devoic-
ing typically occurs in final position. Finally, this theory does not account
for the devoicing of final stops prior to fricatives. In balance, although this
approach offers motivation for obstruent devoicing in general, it does not
explain the specifics of FD.
For this we must look to the functional-perceptual explanation proposed
by Parker (1980, 1981). Here the following steps (table 1) are hypothesized
in the development of FD.
Table 1
I V C V
[-voi] [-voi]
II V C V
[-voi] [ + voi]
III V C
[-voi] [ + voi]
IV V C-
[-voi] [ + voi]
V V -1
[-voi] [-voi]
The language ideally starts out with CVCV word structure, initial stress and
voiceless stops; subsequently intervocalic stops naturally voice, after which
the final unstressed vowel reduces to schwa (stages II and III). With stage
IV the schwa is lost, thereby causing the final voiced stop to be unreleased.
Consequently, the final stop is in an «acoustically unstable environment»,
since it has lost its most salient perceptual cue, the release, and is in danger
of being confused with vowel-final words. Hence speakers devoice the stop,
voiceless final stops being perceptually more salient than voiced ones. Thus
speakers as it were «take unconscious action to preserve the intelligibility
of the speech signal» (136).
550 THOMAS F. SHANNON
7 A nice example of this comes from a personal experience. Asked to produce a word
with a final voiced stop in his native language, an Indonesian friend produced an unreleased
stop, to my ear of indeterminate quality. When I asked him to repeat, he pronounced the same
sound, this time giving it a popping voiceless release after a brief pause. Finally, when I asked
whether the word in question really ended in a voiced stop, he said yes and, apparently making
an extra, conscious effort, pronounced a fully voiced released stop (normally, final stops are
unreleased in Indonesian and Javanese). This anecdote shows how final stops can be confused
and how careful pronunciations can even produce (sharpened) voiceless stops for voiced ones!.
552 THOMAS F. SHANNON
Hence the two sides may work together. Eventually a sound change may
occur and the language will acquire FD as a synchronic rule, if MP alterna
tions result.
In fact, the process can continue: once the final obstruent is voiceless,
it may eventually be lost (cf. Chen & Wang 1975). We observe here roughly
the following chain of events, using a dental stop for illustration (everything
holds ceteris paribus for fricatives):
Apart from the possible sharpening to increase perceptual salience (6-7), there
is a continuous minimization of energy expenditure, leading to the gradual
weakening and loss of the obstruent through progressive curtailment of
various component gestures: labial release gesture (4-5), glottal voicing
gesture (4-5), supraglottal place gesture (7-8), and finally glottal place gesture
(8-9).
like to point out a few difficulties and then propose a somewhat different
account 8 .
First of all, observe that unlike such rules as umlaut in Germanic, FD
did not become opaque because its phonetic motivation, the conditioning
environment, was eliminated; rather it simply failed to apply to potential
new inputs resulting from apocope. Another way of saying this is that FD
became unproductive in Yiddish. But why? In other languages it has remain
ed persistent, productive, and exceptionless9: when potential new inputs to
the rule arise, they regularly undergo it. In German, for instance, English
loan words (Job, Smog, Jazz) are subject to FD. Moreover, when a final
schwa is elided, the resultant form shows devoicing: böse [bö:ze] ~ bös
[bö:s], habe [ha:ba] ~ hab9 [hap]. FD can also be transferred in language
contact situations. As Parker (1985) notes, some younger North Frisian
speakers are applying FD in their use of the dialect, which normally does
have final voiced obstruents. Furthermore, when Germans learn English,
they have a notoriously difficult time mastering the final voicing distinc
tion. Finally, most German (and Dutch) dialects have experienced apocope
to one degree or another, but few have lost FD as a result. All in all, FD
seems to be a very persistent rule. Why wasn't it in Yiddish?
King's answer here is to my way of thinking not fully satisfactory:
apocope was just added to the grammar and for some reason FD did not
apply to its output. In his theory (cf. table 2), King would somehow have
to block FD from applying after apocope in Yiddish (a), and yet let it apply
after apocope in Dutch and German (b), where FD remained productive.
Table 2
a) b)
sing,
sing. pl. sing.
sing, pl
Pl
tag tags tag tags
FD tac -— Apo -— tag
Apo tac tag FD tac tac
In order to handle the otherwise hard to explain counterfeeding order in-
8 The account give here is based on the more detailed work in Shannon (to appear a),
with the crucial difference that I now acknowledge the influence of apocope. Cf. that article
also for a more in depth critique of King's analysis.
9 Turkish is the only language I know of that seems to have FD and yet allows some
apparent exceptions to it.
554 THOMAS F. SHANNON
volved in Yiddish, one could always make the claim that rules are added to
the end of the grammar, but this is not motivated by any independent prin
ciples of the theory, thus clearly ad hoc, and does not really shed any further
light on the matter. Besides, one would still need rule ordering or deriva
tional constraints to prevent FD from applying. The question still remains:
why didn't FD just remain a productive, persistent rule in Yiddish and app
ly to all new inputs? Insisting on a counterfeeding order here is tantamount
to claiming that the rule was no longer productive — which is patently true
—, but this does not explain how and why it became nonproductive. I am
not saying that it is a priori impossible for FD not to apply to the output
of apocope, but simply that the question of how and why this happened
is an extremely crucial one and must therefore be addressed more seriously.
One final, albeit weaker argument is that the analogical levelling in Yid
dish violates the Semantic Transparency Hypothesis (Hooper 1977). Accor
ding to this hypothesis, the semantically unmarked singular should form
the basis for analogy, but here it was apparently the marked plural form
which reshaped the singular. Of course, this argument is only as strong as
the principle invoked, which I will not try to defend. However, I believe
that my account obviates this objection.
My proposal is two-fold. First, I now agree with King that apocope
could have and probably did introduce final voiced obstruents, or something
that could have been so interpreted. The question is how. As I see it, the
problem with King's account lies in a view of phonology in which rules are
too monolithic and abstract: all kinds of sound relations are squeezed into a
single, supposedly homogeneous class, and sound changes are just further
members of this same type. Recent work in phonology, on the other hand,
has shown the importance of distinguishing between different rule types;
the typology used here is from Lineli (1979). Lineli distinguishes between
morphophonemic rules proper (MRPs), which are adjustments in morpholo
gical operations, and all other phonological rules, which include phonotactic
rules (PhtRs) and articulatory reduction rules (ARRs). The former define pos
sible careful pronunciations, while the latter basically map these forms onto
alternate reduced forms (probably as a function of pragmatic factors such
as register and tempo). MRPs and PhtRs are viewed as phonetic planning
rules — they jointly define careful pronunciations —, whereas ARRs 10 are
10 These seem to correspond to what Bartsch & Vennemann (1982) call «realisa-
tionsphonologische Regeln».
FINAL DEVOICING 555
4. Conclusion.
I feel there are a few lessons to be learned from the preceding discus
sion of the rise and demise of FD. Firstly, the underlying motivation for
linguistic naturalness, or its opposite markedness, must be sought; although
this is often ignored, these concepts are ultimately explicanda, not final ex
planations (cf. Shannon, to appear b). Secondly, in order to understand
language in general and language change in specific, we must look more
carefully at what really goes on in linguistic communication. With respect
to sound change, this means that we cannot continue to theorize in over
simplified terms of «rule addition/loss». Rather, as John Ohala (1983:189)
has aptly put it, phonology must seriously concern itself with «mind, mat
ter, and manners».
REFERENCES
Booij, G.E. 1981. Generatieve fonologie van het Nederlands. Utrecht: Spec
trum.
Chen, Matthew & William S.-Y. Wang. 1975. Sound change: actuation and
implementation. Language 51: 255-281.
Coates, Richard. 1980. Time in phonological representations. Journal of
Phonetics 8: 1-20.
Dinnsen, Daniel A. (ed.) 1977. Current approaches to phonological theory.
Bloomington: Indiana University Press.
. 1980. Phonological rules and phonetic explanation. Journal of
Linguistics 16: 171-191.
. & Fred Eckman. 1978. Some substantive universais in atomic
phonology. Lingua 45: 1-14.
Donegan, Patricia & David Stampe. 1977. The study of natural phonology.
In Dinnsen (1977): 126-173.
Fasold, R. & W. Wolfram. 1970. Some linguistic features of Negro dialect.
In Fasold, R. & R. Shuy (eds.), Teaching Standard English in the inner
city. Washington, D.C.: Center for Applied Linguistics, 41-86.
Fourakis, Marios. 1984. Should neutralization be redefined? Journal of
Phonetics 12: 49-60.
Goossens, Jan. 1977. De tweede Nederlandse auslautverscherping. Tijdschrift
voor Nederlandse Taal-en Letterkunde 93: 3-23.
Hamm, Josip. 1975 2 . Grammatik der serbokroatischen Sprache.
Wiesbaden: Harrassowitz.
Hooper, Joan Bybee. 1977. Substantive principles in natural generative
phonology. In Dinnsen (1977): 106-125.
King, Robert D. 1980. The history of final devoicing in Yiddish. In Her
zog, Marvin et al. (eds.), The field of Yiddish, fourth collection.
Philadelphia: Institute for the Study of Human Issues, 371-431.
Koschmieder, Erwin. 1977. Phonationslehre des Polnischen. München. Fink.
Lees, Robert . 1961. The phonology of modern standard Turkish. The
Hague: Mouton.
Lencek, Rado L. 1982. The structure and history of the modern standard
Slovene language. Columbus: Slavica.
Lindblom, Björn. 1983. Economy of speech gestures. In Mac Neilage (1983):
217-245.
558 THOMAS F. SHANNON
SHAN
ON THE HISTORICAL RELATION BETWEEN MENTAL AND
SPEECH ACT VERBS IN ENGLISH AND JAPANESE
1. Introduction.
In his book, Res cogitans: an essay in rational psychology (1972), Zeno
Vendler discusses a number of correspondences between mental verbs like
know, believe, wish, and speech act verbs like claim, report, promise1.
Mental verbs name certain attitudes and states of mind. Speech acts verbs
name an action performed in saying something; a subset of them, called
performative verbs, not only name the action but can have the force of that
action, given the right linguistic, social, and psychological conditions of
satisfaction (cf. Searle 1979, 1983).
Vendler's purpose is to reveal that the semantic organization of verbs
of «speech and thought» is very similar. Although his approach is essen
tially synchronic, he notes in passing that there is a certain amount of what
he calls «leakage» of items across the two categories 2 . For example, the
speech act verbs maintain, agree, conclude, guess, diagnose, distinguish are
said to be leaking in the direction of mental verbs, and mental verbs such
as estimate, judge, recognize, identify, decide, choose, forgive are said to
ι This paper arose out of Elizabeth Traugott's several stimulating conversations with
David Olson, who introduced her to Vendler (1972). She is grateful for financial support for
this study from the John Simon Guggenheim Foundation, and from the Center for Advanced
Study in the Behavioral Sciences (with partial funding under NSF Grant BNS 76-22943). Both
authors are deeply grateful to Judith Hochberg for her help in collecting the English data,
to S.-W. Kuroda and Mary L. Pratt for comments on an earlier version of this paper, and
to Mayumi Ohira for native speaker intuitions concerning Modern Japanese. Any errors of
fact or interpretation are the sole responsibility of the authors.
2 Vendler classifies the verbs in question into verbs of mental state (e.g. know), mental
action (e.g. discover), and speech act verbs (e.g. claim) and shows that these classes have aspectual
properties similar to those exhibited elsewhere in the lexicon by state, process, and achieve
ment verbs (note the difference between *I am knowing that X, I am discovering that X, and
562 ELIZABETH C. TRAUGOTT - RICHARD DASHER
2. Etymological sources.
We turn now to a demonstration that the chief source domains for men
tal and speech act verbs are essentially the same. The discussion will be limited
to performative speech act verbs, but appears to apply equally well to speech
act verbs in general. The source domains in question are terms for i) mental
states and actions, ii) visual perception, iii) spatial relations, and iv) vocaliza-
I promise/ am promising that X). In addition, Vendler points to cross-class semantic cor
respondences, e.g. between the speech act verbs promise, undertake, guarantee, vow, mental
action verbs like decide, choose, and mental state verbs such as want, mean, all of which in
volve commitment toward a course of action or state of affairs. The present paper is concern
ed primarily with the dichotomy between mental and speech act verbs; both mental state and
mental action verbs are included in the category mental verb.
MENTAL AND SPEECH ACT VERBS 563
tion. There are some differences in the extent to which terms are drawn from
these four domains, both with respect to the field (mental or speech act verb)
and the language in question. There are also differences in the extent to which
other source domains are drawn from. However, the general picture that
emerges is of striking overall similarity 3.
3 The study of English mental and speech act verbs relies substantiallyon lists available in
Vendler (1972) and Fraser (1975), from etymologies provided in the American Heritage Dictiona-
ry of the English Language, and on Traugott's (1985) study of the origins of speech act verbs.
Equivalent mental and speech act verbs in Japanese were obtained from dictionaries and from
Martin (1975: 996f), and their usage was confirmed by consulting a native speaker. Data con-
cerning the earlier meanings and derivations of Japanese items come primarily from attested
uses cited in the Nihon Kokugo Daiziten (1974) and Morohasi (1955). Reconstructions of pre-
literary Japanese meanings and etymologies, which are in many cases spurious, were avoided.
564 ELIZABETH C. TRAUGOTT - RICHARD DASHER
To point out that mental verbs originate in terms for mental states and
actions no doubt appears to be trivial at first. It is when we see the similarity
in the sources for speech act verbs and consider the question of directionality
of change from the one category to another that the data become interesting.
Among speech act verbs that derive from terms for mental states and actions are:
Cog. admonish, comment, mention, remind < IE *men- «think» (cf. mind)
acknowledge, note, notify, recognize, diagnose < IE *gno- «know»
kotowaru «refuse» < OJ kotowaru «discern, judge, discriminate»
Eval. accuse, excuse < Lat. causa «cause/reason/lawsuit»
appraise, praise < Lat. pretiare «value» [ultimately < pretium «price,
that which is opposite/equal»]
hihan suru «criticize» < Mid 13th Jpns «judge, evaluate» (used as
speech act verb since later 19th C.)
Aff. consent < Lat. sentire «feel»
agree, disagree < Lat. ad + gratus «to + pleasing»
plead < Lat. placitum «legal action, agreement» < piacere «to please»
hagemasu «encourage, urge» < O J «cause to be enthusiastic» (caus.
of OJ hagemu «become emotional»?)
+ Jp. suru
4 The list would be greatly expanded if terms were included which are not obviously
locative but which can only be fully understood in terms of such spatial relations as transfer
of X from Y to Z, e.g. those derived from give and receive (cf. Jackendoff 1983).
566 ELIZABETH C. TRAUGOTT - RICHARD DASHER
Among those derived from terms for plane and shape are understand and:
surmise < Lat. super + mittere «send off, throw»
suppose < Lat. sub + ponere «under + put»
learn < IE *leis «furrow»
plan < Lat. planus «flat» (cf. ground plan, map)
(note that in several of these the root expresses location; plane is expressed
by the prefix). So far, no mental verbs involving plane or shape have been
noted in Japanese.
Among speech act verbs derived from spatial terms are:
cite, solicit < Lat. ciere «put in motion, move»
suggest < Lat. gerere «carry»
assent < IE *sent- «go, head for»
admit, commit, dismiss, permit, promise, submit < Lat. mittere «send»
conjecture, object, reject < Lat. jacere «throw»
command, demand, commend, countermand, recommend < Lat.
manu dare «into hand give»
concede < Lat. concedere «go away, withdraw»
insist, restore, state < IE *sta- «stand» (state < N for status, rank)
assert < Lat. asserere «join oneself to» < IE *ser- «line up»
exonerate < Lat. exonerare «free from burden»
iihukumeru «convince, give (a person) to understand» < EMidJ iu
«say» + hukumeru «insert, enclose, include»
iiharu «insist < EModJ compound of iu «speak» + ham «stretch»
and, from plane and shape:
propose < Lat. pro + habere/ponere «in front + have/put»
hypothesize < Gk. hypo + tithenai «under + place»
submit < Lat. sub + mittere «under + send»
inform < Lat. informare «give shape to»
reply < Fr. replier «fold back»
abrogate, direct, rank, reckon, rule < IE *reg- «move in straight line»
yakusoku suru «promise» «tie into a bundle» + Jp.
suru «do»
5 Relevant here is also the category of «delocutive verbs» (Benveniste 1971) in which
various locutions such as formulae of greeting are converted into verbs of saying, cf. welcome,
which means «to say «Welcome!»».
568 ELIZABETH C. TRAUGOTT - RICHARD DASHER
6 Simon Dik (p.c.) pointed out that ask oneself, say to oneself are in part mental, and
perhaps could be counted as cases of speech act verbs leaking toward mental verbs. The semantics
of the reflexive speech act verb construction and its history deserve special study.
MENTAL AND SPEECH ACT VERBS 569
show changes within the history of English. In the following examples mental
verb and speech act verb are abbreviated as MV and SAV respectively.
find: appears from OE on as MV; does not appear as SAV
until 1400, and then only in legal contexts (e.g. find
guilty means «determine and declare guilty»)
guess: appears in the meaning «form a judgement» from 1340;
the first instances cited in the OED with any
resemblance of SAV function date from two hundred
years later.
Other verbs show the shift in French:
agree: originally meaning «please» (12th C), Fr. agréer readily
acquired the meaning of receiving favorably, therefore
welcoming; from an expressive speech act meaning
«welcome» it developed the meaning of «allow». The
original Lat. source, adgratare «please» shows this
originated in a MV of the affective type.
distinguish: Lai. distinguo meant «separate mentally»; most Fr. and
even Eng. uses are difficult to establish as clearly per
formative.
Thus, these items actually underwent a shift from mental verb to speech
act verb in their historical development. In so far as a verb like agree now
has peripheral meanings of the mental verb type, they are relics, not innova
tions. We have found no examples of the shift from speech act verb to mental
verb in Japanese.
Indeed, the so-called «leakage» from mental verb to speech act verb
is well-attested and has a long history. Consider the following examples:
assume: first used in the 16th to mean «take upon oneself/
adopt», this verb had the concrete sense of «taking on
employment or an employee», and the mental sense «to
suppose»; the first clear SAV sense of «claim, pretend
that» cited in the OED is from 1714.
blame: the very first entries in the OED from the 13th attest
to the fact that this has always been used as a SAV in
Eng.
recognize: although clearly mental in origin (cf. cognize), this has
been used with the SAV meaning of «admit» from 11th
Fr. times on; Souter (1949) cites a 4th Lat. exam
ple of cognoscere with the same meaning.
570 ELIZABETH C. TRAUGOTT - RICHARD DASHER
4. Conclusion.
This particular study of the relation between mental and speech act verbs
leads to a number of conclusions. For one, the hypothesis that propositional
meanings give rise to interpersonal ones and not vice versa was used to ques
tion Vendler's claims, and the evidence supported the hypothesis. This shows
once more that there are powerful regularities in semantic change of a far
more specific sort than the «extension of meaning, metaphoric shift,
metonymie shift, amelioration» or «pejoration» we hear so much about in
earlier treatments of semantic change.
Secondly, the hypotheses that propositional meanings give rise to in
terpersonal ones and that invited inference can be a factor in semantic change
were both developed in connection with the history of grammatical forms
on the borders between syntax and morphology (conjunctions, articles, turn
takers, etc.). The present study shows that regularities and the processes that
lie behind them are by no means limited to grammatical forms but are rele
vant at the lexical level as well (cf. Traugott, in preparation).
Thirdly, the evidence considered here suggests that as far as lexicaliza
tion of metalinguistic repertoires is concerned, «having in mind» seems to
be more fundamental than «asserting that». This is further supported by
the fact that the number of IE roots identifying mental states is fairly substan
tial, and also by the fact that a large number of mental verbs are native
(e.g. know, hold, think, believe, hope, fear, deem, mean, blame, wonder,
etc.), but many Old English speech act verbs have been replaced by terms
572 ELIZABETH C. TRAUGOTT - RICHARD DASHER
REFERENCES
Kuroda. S.-Y. 1973. Where epistemology, style, and grammar meet: a case
study from Japanese. A Festschrift for Morris Halle, ed. by Stephen R.
Anderson and Paul Kiparsky. New York: Holt, Rinehart and Winston.
Lyons, John. 1977. Semantics. Cambridge: Cambridge University Press.
Martin, Samuel. 1975. A reference grammar of Japanese. New Haven: Ya
le University Press.
Morohasi, Tetuzi. 1955. Dai kanwa ziten [Great Chinese-Japanese dictio
nary]. 13 vols. Tokyo: Taisyuukan Syoten.
Nihon kokugo daiziten [Unabridged dictionary of Japanese]. 1974. 20 v.
Tokyo: Shogakukan.
Paez Urdaneta, Iraset. 1982. Conversational «pues» in Spanish: a process
of degrammaticalization? Papers from the Fifth International Conference
on Historical Linguistics, ed. by Anders Ahlqvist. Amsterdam: Benja
mins.
Searle, John R. 1979. Expression and meaning. Cambridge: Cambridge Uni
versity Press.
. 1983. Intentionality: an essay in the philosophy of mind. Cambridge:
Cambridge University Press.
Souter, Alexander. 1949. A glossary of later Latin to 600 A.D. London:
Oxford at the Clarendon Press.
Sweetser, Eve Eliot. 1984. Semantic structure and semantic change: a co
gnitive linguistic study of modality, perception, speech acts, and logical
relations. Ph.D. Dissertation, University of California, Berkeley.
Traugott, Elizabeth Closs. 1982. From propositional to textual and expres
sive meanings: some semantic-pragmatic aspects of grammaticalization.
Perspectives on historical linguistics, ed. Winfred P. Lehmann and Ya-
kov Malkiel. Amsterdam: Benjamins.
Traugott, Elizabeth Closs. 1985. Speech act verbs: a historical perspective.
The First International Roman Jakobson Conference, New York, Octo
ber 10-13th.
. In preparation. On regularity in semantic change.
Vendler, Zeno. 1972. Res cogitans: an essay in rational psychology. Ithaca:.
Cornell University Press.
ON THE PERSISTENCE OF IMPERFECT GRAMMARS:
CLITIC MOVEMENT FROM LATE LATIN TO ROMANCE*
DIETER WANNER
University of Illinois at Urbana-Champaign (USA)
* This paper constitutes a bare bones summary of essential points dealing with the topic
of Clitic Movement; the source study is Wanner 1986: Ch. 7,9. All bibliographical details and
review of existing literature are treated in there. A short published account of one case of Old
to Modern Romance development is available in Wanner 1982. — The research presented in
this study has been supported in part by grants from the Research Board of the University
of Illinois at Urbana-Champaign, and by a Fellowship for Individual Study and Research by
the National Endowment for the Humanities, 1983/84; both contributions are gratefully
acknowledged, as well as the travel subsidy from the Scholar's Travel Fund (University of Il
linois) for attending the Pavia meeting.
ι Cf. Wanner 1979 for a reconstruction of the development of word stress place from
Latin to Romance in this sense. Many instances of incomplete lexical diffusion can be reinter
preted in this way, as well as instances of linguistic interference (cf. Wanner 1980).
576 DIETER WANNER
2 Some essential sources for such ideas are Rosch 1978, Gleitman and Wanner 1982,
Lakoff 1982; Coseriu 1961, 1974; Andersen 1973, Lass 1979, Givón 1979, 1984. A broader
treatment of the question of historical change in this context is contained in Wanner 1986:
Ch. 9.4.
CLITIC MOVEMENT FROM LATE LATIN TO ROMANCE 577
of V(l) so that a predicate such as insistere «to insist» is found only with
the (a) type expression; cf. () vs. *(3b). insistere is thus a V(not-cm) as
compared to potere as V(cm).
(3) a. Perché insisti a invitar/o? «Why do you insist on inviting him?»
b. * Perché io insisti a invitare? «id.»
For Italian in its illustrative function as a Modern Romance idiom, the class
V(cm) comprises the following groups of predicates:
(4) a. causative fare, lasciare
b. perception vedere, sentire; guardare, ascoltare
(5) a. MODal potere, dovere, volere, sapere
b. ASPectual cominciare, finire, continuare, solere, etc.
c. CONative cercare, provare, tentare, etc.
d. MOTion andare, venire, tornare, etc.
e. SUBject embedding sembrare, parere, bisognare, etc.
f. OTHer (speaker oriented constituency)
The causative and perception predicates of (4) form a group apart with regard
to the placement properties of the embedded subject NP or subject clitic;
they shall not be treated here separately in much detail. For the true V(cm)
predicates of (5), a summary description can attribute to them a number
of variable scalar settings on relevant parameters:
(6) a. Normalcy of admission of + CM strings of type (lb/2b):
decreasing from MOD to ASP/CON/MOT to SUB to OTH
b. Exhaustiveness of V(cm) group with regard to each class:
decreasing from MOD to ASP/CON/MOT to SUB to OTH
Frequency of + CM string use per single V(cm):
decreasing from MOD to ASP/CON/MOT to SUB to OTH
d. Acceptability of + CM string types with regard to style levels:
decreasing from low to high style, and from spoken to written ex
pression
e. Ease of processing of +CM string types:
decreasing with specifity of V(cm) meaning,
decreasing with «transformational complexity» of + CM string (con
cerning various dislocation operations)
f. Geographic distribution of + CM string acceptability:
absence of + CM in Modern French, presence of + CM in all other
Romance language and dialects
Regardless of what might be considered a correct formal linguistic analysis
578 DIETER WANNER
3 Some treatments of CM in terms of formal grammar are found in Kayne 1975, Quicoli
1976, Strozer 1976, 1981, Rivas 1977, Rizzi 1978, Lujan 1979, Suñer 1980, Napoli 1981, Wan
ner 1982, Zubizarreta 1983. — The two additional classes CAUS(ative) and PERC(eption) verbs
fit in the above schema (6) in two ways. With regard to the embedded subject expression (e.g.
mi in (i)), CM is categorically present in all dimensions, thus highest position in (6a-f). Con
cerning an embedded object pronoun (e.g. la in (ii)), CM conforms to a high degree of realiza
tion, parallel to MOD(als). For modern Italian, the subject pronoun aspect is a problem separate
from CM proper; in Old Italian/Old Romance, both aspects were still unified and belong under
the same heading of highest level of CM realization,
(i) mi lasciarono partire «they let me leave»
(ii) sentiamo Carlo eseguirla su un vero fortepiano! «let's hear Carlo play it (e.g. la sonata)
on a true fortepiano».
CLITIC MOVEMENT FROM LATE LATIN TO ROMANCE 579
4 The texts studied for Old Italian comprise the Novellino, Bono Giamboni, Libro de'
vizi e delle virtudi, and the Tristano Riccardiano, all from the 13th century (all quoted after
Segre and Marti 1959). — In modern Romance, no elements may intervene between V(l) and
inf(2) other than a complementizing preposition (or an incorporated clitic element).
580 DIETER WANNER
practically all instances of V(l), inf(2) and cl(2) combinations. The typical
string form (9b) instead of expected (9a) is avoided only in those cases where
the + CM string would produce either an incompatible clitic cluster (cf.
(10)), or where V(l) has a pronominal object of its own (cf. (11)), or where
V(l) has the form of a complex verbal expression (e.g. V + NP; cf. (12)).
(9) a. - C M string: V(l) inf(2) - cl(2)
b. +CM string: cl(2) - V(l) inf(2)
(10) a. estiatia mente dirallegrartidel poco (Giam. 47) «and keep in mind
to appreciate the little which...»
b. *ti + ti under all conditions
(11) si teme di offendermi «he fears (for himself) to offend me»
(12) io avrò balia di mandarli (Nov. 829) «I'll have power to send them».
Only from the 13th century on in French, and from the 14th on in Italian
and other Romance languages can one find arrangements of -CM type (9a)
outside of the forced situations described by the classification of (10) to (12).
These rare examples expand however rapidly to constitute a sizeable por
tion of all potential strings, producing from the 16th century on a picture
of fully optional CM realization 5 .
5. What looked like a good example of a categorial phenomenon of
its own standing in Old Romance, a coherent process of moving the clitic
from an infinitive to its governing verb — hence Clitic Movement — lost
ground gradually, providing an instance of the lost generalization of general
interest here. Since a direct attribution of the evolution to «rule loss or reduc
tion» does not explain anything here, the study of the origin of Old Romance
CM from Latin will prove to contain an answer to the question of the reason
for this surprising change: CM was not an independent process neither in
Old Romance nor in Modern Romance, and it takes its driving force from
a convergence of a small number of other processes independent from clitic
syntax in the area of complementation. Being accidental and heterogeneous
in its origin, CM remains fragmented even in the period of its apparent
categorial status (Old Romance), paving the way for the modern anecdotal
presence of this string type. CM is a lost opportunity with regard to
generalizations, an interesting example of irrelevance of formal grammatical
notions.
5 This stage is reached e.g. in the 15th century text of the Lettere by Alessandra Ma-
cinghi Strozzi; cf. Wanner 1981b.
CLITIC MOVEMENT FROM LATE LATIN TO ROMANCE 581
[-CM] characterizes a Late Latin string where subj(2) and V(l) are separated,
and where subj(2) stands closer to inf(2) than to V(l); cf. (20). +CM
manifests contact between subj (2) and V(l) such that V(l) separates subi(2)
from its original anchor inf(2); cf. (21). ± C M foresees contact between
subj(2) and V(l) but without inf(2) being spaced from its subj(2) by V(l);
cf. (22).
(20) a. string: ... subj(2) ... inf(2) .. V(l).. (typically = Basic (14a)).
b. quem se illuc missurum2 ... testatus est2 (Peregr. 17.1)
«whom he affirmed ... to be ready to send there»
(21) a. string: ... subj(2) V(l) ... inf(2) .. (typically = Restructured (16a))
b. illud autem vos volo1 scire2, dominae (Peregr. 3.8)
«This however I want you to know, myladies»
(22) a. string: ... V(l) subj(2) ... inf(2)... (typically = Extraposed (15a))
b. putat 1 se maximum peccatum incurrisse2 (Peregr. 49.2)
«he believes to have incurred very great sinfulness»
The two classifications + CM and ± CM together constitute the source of
proto-Romance precursors of CM strings where a subordinate clause pronoun
appears associated with the governing verb rather than with its infinitive.
7. A study conducted on ten (Late) Latin texts between the 1st and
6th centuries — from the Cena Trimalchionis to the Peregrinatio ad loca
sancta to Gregory of Tours, and others more 7 — establishes the relevance
of this scenario on the basis of frequency and theoretical considerations.
The interpretive results show the increasing numerical dominance during the
period from the first to the sixth century of Extraposition strings with ± CM
sor to Romance CM composed of the same elements — even though in somewhat different
function —, but at the same time it can be shown (Wanner 1986: Ch. 4,5) that a non-emphatic
Latin non-subject pronoun (inlcuding a subject accusative pronoun of the a.c.i.) exhibits
behavioral traits which are closer to a Romance clitic object pronoun than to a free strong
pronoun — this latter being the original condition of a Latin pronoun. The subject accusative
pronouns of interest here are proto-clitics in a proto-CM situation.
7 The texts are the following: Two of more classical orientation, Plautus, Persa (2nd
cent. ), and Bellum hispaniense (1st cent. ); three of more spontaneous stamp, Cena
Trimalchionis (1st cent. AD), papyrus letters from Karanis (letters to and by CI. Terentianus,
2nd cent.), and Peregrinatio ad loca sancta (end 4th cent.); four texts of non-Biblical Chris
tian Latin, letters by Cyprianus, and letters by correspondents of Cyprianus (both 3rd cent.),
sermons by Augustinus (4/5th cent.), and Acta martyrorum christianorum (6th cent.); finally
a later historical text, Gregory of Tours, Historia Francorum (end 6th cent.). Details are found
in Wanner 1986: Ch. 7.
CLITIC MOVEMENT FROM LATE LATIN TO ROMANCE 583
8 Latin object pronouns as proto-clitics can be grouped as obeying one of three place
ment options (referring to a simple clause): placement into second position (2ND) as the
autochthonous IE condition, placement contiguous to the verb (VB) as the newer (proto-)
Romance trend, or no identifiable placement pattern, i.e. free pronoun placement either due
to undetected emphatic status of the pronoun or due to the general lack of categorial adherence
to any and all of these placement patterns in Latin. For VB placement as the relevant proto-
Romance vector, both preverbal and postverbal linear position are admissible within certain
bounds, the indigenous solution being preverbal position leading into basic Romance proclisis
of the pronoun, with defined contexts of encliticization. Details again in Wanner 1986: Ch. 5.
584 DIETER WANNER
for su.aci pronouns corresponds to a 90% -CM status for the ob.aci pro
nouns. The schematic string (24) visualized the crucial difference consisting
in the peripheral place of su.aci — leading to V(l) contiguity — and the
protected position of ob.aci as a necessarily internal constituent of the in
finitival clause. The example in (25) shows both pronoun types in the same
utterance in their typical place.
(24)
(25)
perception verb constructions, even in Modern French which does not know
CM otherwise.
(27) a. on lui a fait abjurer sa conviction
«one made her swear off her convinctions»
b. on le verra interpréter son meilleur rôle
«one will see him interpret his best role»
*on le voudrait convaincre du contraire
«one would like to convince him of the opposite»
d. on voudrait le convaincre du contraire «id.»
10. An entire battery of forces and conditions has been at work to
produce the Old Romance situation of para-categorial CM observance. Brief
ly listed, they are the following:
(28) a. Extraposition of the infinitival complement; in Latin grammar
b. Juxtaposition of V(l) and subj(2) in Extraposed and Restructured
strings; Latin
Bonding of V(l) and cl(2) under verb based (VB) placement for pro-
toclitic pronouns; Late Latin and chiefly Old Romance
d. Semantic and pragmatic ease of V(l) construal of cl(2) for core
predicates; Latin, Late Latin, and Romance
e. Analogical extension of V(l) dependence from su.aci to ob.inf; Late
Latin and Romance
f. Code switch from Late Latin to Romance for full manifestation of
CM phenomena.
These component factors are sufficient to produce the ultimate effect of
Old Romance CM; thereby, CM is a natural result of Latin syntactic trends
which manifested themselves more prominently in the lower register speech
situations reflected somehow — even though only very fragmentarily — in
the texts investigated in the comprehensive study, i.e. texts of known proto-
Romance tendencies in metachronic and teleological perspective 9 .
9 The texts investigated here are all Latin; they are clearly distinct from Romance not
only through their chronological position, but more importantly through the auctorial intent
of writing in Latin. As such, these texts will never be a mirror of true spontaneous speech =
proto-Romance of the day; they can only reflect some features which may pass through the
contrcl net of grammatical/rhetorical education of the writer. The transition from Latin to
Romance cannot be documented in its realistic gradualness, only in the abrupt code switch
of incipient Romance writing (8th/9th cent.). For some interesting discussion of such problems,
cf. e.g. Reichenkron 1965 and Wright 1982.
586 DIETER WANNER
REFERENCES
Rizzi, Luigi. 1978. A restructuring rule in Italian syntax. in: Recent transfor
mational studies in European languages, ed. J.S. Kayser, 113-158. Cam
bridge: MIT Press (Linguistic Inquiry Monograph, 2).
Rosch, Eleanor. 1978. Principles of categorization. in: Rosch, Eleanor and
B. Lloyd (eds). Cognition and categorization, 28-48. Hillsdale, NJ:
Erlbaum.
Segre, Cesare and Mario Marti (eds). 1959. La prosa del Duecento. Milan:
Ricciardi. (Letteratura italiana; Storia e testi, 3).
Sorrento, Luigi. 1951. Sintassi romanza. Ricerche e prospettive. Varese-
Milan: Cisalpino (2nd ed.).
Strozer, Judith A. 1976. Clitics in Spanish. Unpubl. Ph.D. diss. University
of California, Los Angeles.
Strozer, Judith A. 1981. An alternative to Restructuring in Romance syn
tax. in: Proceedings of the 10th anniversary Symposium on Romance
Linguistics, ed. H. Contreras and J. Klausenburger, 177-184. Seattle:
University of Washington, Dept. of Romance Languages and Literatures
(Papers in Romance, Supplement 2).
Suñer, Margarita. 1980. Clitic promotion in Spanish revisited. in: Contem
porary Studies in Romance Languages, ed. F.H. Nuessel, Jr., 300-330.
Bloomington: Indiana University Linguistics Club.
Ulleland, Magnus. 1960. Alcune osservazioni sulla legge Tobler-Mussafia.
Studia Neophilologica 32:53-79.
Wanner, Dieter. 1979. Die Bewahrung der lateinischen Haupttonstelle im
Romanischen. Vox Romanica 37:1-36.
Wanner, Dieter. 1980. Diachronie abstraite et interférence superficielle. in:
Actas del XVI Congreso Internacional de Filología y Lingüística
Románicas (Palma de Mallorca, April 1980; to appear).
Wanner, Dieter. 1981a. Clitic placement from Old to Modern Italian: Mor-
phologization of a syntactic rule. in: Cressey William W. and Donna J.
Napoli (eds.). Linguistic Symposium on Romance Languages, IX,
331-348. Washington: Georgetown University Press.
Wanner Dieter. 1981b. The language of the Lettere by Alessandra Macinghi
Strozzi. Papers in Romance 3:161-178.
Wanner, Dieter. 1982. A history of Spanish clitic movement. in: Proceedings
of the 8th Annual Meeting of the Berkeley Linguistic Society, ed. M.
Macaulay et al., 135-147. Berkeley: Berkeley Linguistic Society.
590 DIETER WANNER
OTMAR WERNER
Universität Freiburg i. Br.
1 Cf. summarizing Ineichen 1979, Werner 1984a. - I should like to thank Dr. Janet
Harkness for numerous corrections of my English text.
2 Cf. e.g. Mayerthaler 1980, Wurzel 1982, 1984.
3 Cf. from Humboldt 1836 to Schleicher 1850.
592 OTMAR WERNER
focal point of linguistic theory. It was reduced by some to the observation of analogy: what
are the «laws» of these morphological regulations after the destructions which the sound-laws
had brought about? 4 Other scholars (Finck 1909, Sapir 1921) greatly enlarged the number of
morphological types and gave certain languages as ideal examples of these types. Since, how
ever, most languages show features of different ideal types, it seemed to be more legitimate,
to study the different mixtures of typological features and their diachronic change, instead
of looking for language-types 5.
In spite of this, the idea that these actual mixtures are to be understood as states of transi
tion from one ideal type to another one, was never completely relinquished. And the question
why morphological development moves in one direction or another, was looked for in connec
tion with other linguistic levels; this especially, after a new language typology, based on syn
tactic criteria, became dominant 6. So Vennemann 1974 developed a sort of a pendulum-model
with the languages gradually wandering from the extreme of isolation to the extreme of inflec
tion and back again without end, under the influence of case-marking by word-order or by
nominal inflection.
In «Natural Morphology», however, the interrelation between phonological change and
morphological analogy is - again - the dominant idea:
7 Ideas of the following sketch are also contained in Werner 1984b, concentrating on
the different possibilities of morphological change, and in Werner 1984c, more materialbound;
here we emphasize the consequences for language-typology.
594 OTMAR WERNER
12 Ramat 1985 works with a similar scale of Germanic verb forms from agglutination
to suppletion, which latter is surpassed by former participles now separated from the verb in
flection (like the adv. während). His polarity is primarily «grammar ֊ lexicon»/«regularity ֊
idiosyncracy» and its psychological implications. In our scale the aspect of «isolation/separa
tion ֊ fusion/condensation» and its economic implications is in the foreground, with «regularity
- irregularity» parallel to a high degree. It seems to me it would be worthwhile to scrutinize
the parallelism of these parameters more closely than possible here (cf. footnote 14 below).
600 OTMAR WERNER
ble and hesitation, the remodelling after a simpler pattern will be more ap
pealing; i.e., analogy to a simpler technique or a new composition will take
place.
Since we have both in the frequency and in the morphological devices
not just two extreme poles but gradual scales the movements of analogy
can also go back to more or less isolated and regular stages; they need not
aim at full isolation or a simple 1:1 agglutination. We can compare, for ex
ample, MHG/NHG wir ziehen - zogen, a fairly high-frequency verb, which
is left unchanged, and MHG kiesen - kuren, which has gradually become
rare and is nowadays even archaic, mostly because of the existence of the
synonym wählen «to choose, (s)elect, vote»; here we have analogies like
küren - koren (still a strong verb with root-inflection, but now with uniform
ed consonantism and phonetically more related vowels) or küren - kürten
(now a weak verb with a simple pret. suffix) or - today more common, if
ever used - wir küren - haben gekürt (with a periphrastic perfect to express
past tense, i.e., still more decomposed into two words).
So, if an inflectional form has been moved by sound-change too far
in the direction of condensation and irregularity - too far according to its
frequency - or if a condensed and irregular form has lost of its former fre
quency, then an analogy will take place: In other words this form will be
substituted for by an inflectional technique more or less further left (in our
figure). Alternatively there might be a jump to the very beginning of our
scale, to the combination of separated/ «isolated» elements. In the case of
our German perfect, the elements which are combined to express a new in
flectional category are highly inflected in themselves: haben/sein are cases
of suppletion with suffixes, and a past part. like ge-kür-t has both prefix
and suffix; gezogen has in addition a high-degree of root inflection. That
means that even in the case of new combinations, we not necessarily have
just isolated units; they can be built up of elements which are highly inflected
themselves. All stages of our scale can coexist - nothing is necessarily unified,
not even «the beginning» (on the very left of our figure). And it is by no
way universally given how far the jump back in the development is made:
whether the analogical new form just abolishes root-consonant variation,
makes the root-vowel variation more regular, produces a regular suffixed
form without root-variants, or starts with a new combination of two forms.
All these possibilities correspond to the fact that frequency of usage is a
scalar and varying factor, too.
THE AIM OF MORPHOLOGICAL CHANGE 601
13 Poitou 1984 discusses and criticizes current «Natural Morphology», too֊ likewise
mostly under aspects of language economy; we hope to be able to compare his points more
closely at another opportunity.
14 Moves from weak to strong verbs are discussed by Ramat 1985; in his treatment, it
is more a change to higher regularity, an analogy to existing patterns, than a change to more
condensation. So he gets into difficulties about, how to place these processes into the scale:
is it parallel to «strong - weak» a move to more grammaticalisation, or is it, contrary to this,
a move towards suppletion (169) (cf. footnote 12 above).
602 OTMAR WERNER
ble that some phonological changes can, by chance, also produce more mor
phological uniformity and simpler inflection: the two OHG verb classes
heizan - hiaz and loufan - liof, for example, have coalesced in MHG into
one class with the pret. forms hiez, lief.
On the whole, however, sound-change would lead to condensation and
irregularity; and it is analogy, that produces more separation and regulari
ty. The decision whether morphological forms remain or change in one of
the two directions is made on autonomous morphological grounds.
(2) We cannot simply and unconditionally declare certain morphological
types to be desirable and others to be undesirable. All of these possibilities
as they exist in reality, have their advantages and disadvantages; it is not
mere chance that we have in the German verb quite a number of different
techniques alongside one another to express past tense: wir reden - haben
geredet, - redeten, sagen - sagten [za:ktən], nennen - nannten, nehmen -
nahmen, ziehen - zogen, gehen - gingen, sind ֊ waren. It depends mostly
on the relative frequency in usage, whether, for example, the advantages
of condensation are great enough to put up with the disadvantages of ir
regularity. Consequently, which solution is the most appropriate differs from
verb to verb.
Since frequency is a factor which changes between different groups of
speakers, on different occasions and in the course of time, the tendencies
to move in one of the two directions are highly variable. So we cannot ex
pect that every change in frequency immediately triggers a morphological
change. There is also a certain wholesome general tendency to keep to the
traditional form or to accept the new form created by sound-change; a cer
tain disproportion between average-frequency and morphological device
must be achieved (a «quantum», so to say), before an analogical jump is
made - and this happens on individual occasions first, later on it is possibly
accepted by the community and by normalization.
(3) Phonological reduction and thus morphological condensation goes
on, more or less, steadily and would eventually lead all forms to dissolution
or to suppletion - unless stopped and recommitted to an early type on the
scale of development. This «go back to Go» can happen at any stage of
the development: very early, after cliticalisation only, from the stage of ag
glutination, from any stage of inflection or from the final stage of supple
tion. And the jumps back can be short or long.
With this in mind we cannot expect the morphology of a language to
assemble around just one morphological type; on the contrary, a mixture
THE AIM OF MORPHOLOGICAL CHANGE 603
isolation to condensation
(5) A later insight of Natural morphology is that it need not be 1:1 ag
glutination; any device which is well-established and typical of a language
can become the aim of morphological change15, e.g. root-inflection in Ger
man noun-plurals, since there are already many nouns functioning with this
feature. This view would mean that there is no absolute universal aim in
morphology, but only a language-individual aim, which is determined by
the number of items which already follow a certain morphological device.
Here we find, it is true, that a numeric criterion is decisive, that of type-
frequency; but this type of frequency is restricted to data within the linguistic
system, it does not take the data of language use into consideration, which
seem to me to be crucial to explain the structure of the system. And type-
frequency and token-frequency are, on the whole, inversely proportional
to each other.
So this reasoning can provide a certain explanation of why analogy very
often aims at this morphological device, if analogy takes place; it is an in
dication that for a while there can be certain groupings around a certain
type. It is not an explanation, however, for why there are usually different
morphological devices alongside one another and why even stages of higher
irregularity are preserved in individual word-forms or even produced in dif
ferent ways. In addition to the observation of temporary trends to unifica
tion within certain limits, we need an explanation for the more fundamen
tal and permanent trend to diversification and mixture.
Understanding of the elementary economic polarity between separa
tion and condensation as directed by usage seems to me primary; with this
in mind we can ask further, why certain preferences or preponderances of
different morphological devices can be found in individual and temporary
language-developments. Such a wider model helps us to avoid asking wrong
questions, like: Why don't we have just one uniform language type? What
morphological technique is the really «natural» one? Why are most languages
kept so long in a state of undesired mixture and irregularity?
I would suppose that the ideal of a dynamic system in everchanging
balances between different polarities according to economic principles are
not only useful and enlightening in the domain of morphology but in all
other fields of language.
REFERENCES
MARGARET E. WINTERS
Southern Illinois University - Carbondale
0. Introduction.
In this paper I would like to consider some of the questions raised by
the subjunctive mood in French, and especially those pertaining to its con
trast with the indicative. These questions (perhaps more so than the various
answers proposed to them) have repercussions in the study of change in the
subjunctive, from Latin to early Romance, and from Old French to the con
temporary language. What will be proposed here is the application of a
recently emerging theory concerning the nature and organization of language
to the evolution of the meaning of the subjunctive. I will begin with a brief
restatement of some question about this mood, present the theory of
cognitive/space grammar, and then show some of its power in accounting
for the data.
* This paper and the research project it represents would not be possible without the
help of Geoffrey Nathan, who has given me encouragement, reactions, and suggestions at every
step. I would also like to thank Jo Rubba for listening and reacting to my work.
608 MARGARET E. WINTERS
2. Cognitive/Space Grammar.
During the last five years, research has been published within a
theoretical framework called space grammar (Langacker 1982, 1983) and
cognitive grammar (Lakoff 1982, ms.)· Although there are some differences
between these theories, they lie mostly in the emphasis placed on one or
another facet of language processing. I will accordingly treat them as a single
theory. Underlying it are two very basic assumptions. First, there is no im
portant division between human language use and other cognitive processes;
they are not to be sharply separated (Langacker 1983: 6). As a result, the
same kinds of strategies used in other kinds of cognition are at work in
human linguistic functioning. Second, the basis of language is semantic rather
than syntactic. Language is inherently symbolic, and hence meaningful at
all levels of organization (Langacker 1983: 6). Phonology and syntax are
both symbolizations of meaning and are to be understood in terms of an
underlying semantic organization (for an overview, cf. Winters and Nathan
1985).
As a result of the above conceptualization of language, one can also
say that every unit is meaningful and there is no division other than posi
tion along a continuum between what are usually divided into lexical items
and grammatical morphemes. It is on these bases that the view of the sub
junctive as mere «servitude grammaticale» should be rejected, even in con
texts where the speaker has no choice of mood. Larger grammatical units
(like compound lexical items or idioms) can be viewed on two levels, that
of the construction, where they are perceived as broken down into their com
ponents, and that of the unit, the whole as meaningful in itself. Langacker
(1982:28) states that language processing is done on both these levels, and
(1983:50) that it is inaccurate to analyze one to the exclusion of the other.
For the subjunctive in more or less automatic contexts, then, one can
argue that the trigger (that is, the verb, conjunction, or expression in the
main clause) is meaningful, as is the subjunctive form itself, and that there
is further meaning to be found in the collocation of the two (for a related
view,.cf. Nordahl 1969: 248-9 and Imbs 1953: 19). In a sentence like
6 II est possible qu'il parte.
for example, in addition to the meaning of each word (or morpheme) as a
separate item, the constructions «il est possible» and «qu'il parte» have
meaning, as does the whole sentence. In addition, part of the meaning of the
whole sentence is the juxtaposition of the two smaller constructions, as a unit.
610 MARGARET E. WINTERS
To use a cliché, the whole is greater than the sum of its parts, or, in
the context, the whole is more meaningful than the sum of its parts 1:
7 il est possible qu'il parte
unit: «subjectivity»
One futher notion extensively explored by cognitive/space grammar is
the general cognitive strategy of prototype effect. Studied first as part of
psychology and general cognition (notably by Rosch 1973, 1975), the effect
is seen in the way in which human beings classify items as belonging to some
mental set. Rather than depending on a finite list of criteria for member
ship in the set, they compare items (that is, anything which can be classified:
animate and inanimate objects, colors, grammatical constructions, uses of
mood) to prototypical members of the set, those members which best ex
emplify the class they are illustrating. A prototypical bird, for example, is
a robin or a sparrow, not an ostrich or a kiwi, and a prototypical green
has a certain balance of blue and yellow in its components which makes
it look totally green and not green verging on another color.
In addition to looking at prototype criteria for set membership,
cognitive/space grammar also studies the internal structure of sets by means
of what Lakoff (ms.: 72-82) calls the radial structure. Central to a given
set, and having special cognitive status as such, are prototypical members
against which others are compared. Radiating out from them in various ways
are extensions, items in the set which differ from the central, prototypical
members by one or more than one feature. Other, even less prototypical
members are extensions of extensions. The overall design is a set whose
members are directly or indirectly related to the central members, but which
may not at all be related directly to each other. One can think of variations
on physical things: a prototypical cup has a handle, is used with a saucer,
and is usually a container for hot liquids to be drunk. But a nonprototypical,
but contextually recognizable cup may be handleless, have no saucer, con
tain solids, or even be equivalent, as an abstract measure, to two four-ounce
sticks of butter.
compound question
SAAD
passive negative
I would hold that the uses of the subjunctive form a cognitive set as
contrasted to the uses of the indicative in a given language at a given time.
What constitues each set is a matter for research, as is the internal, radial
structure of the sets.
4. Examples.
4.1 At the simplest level, triggers of the subjunctive (which, as was
said above, contain part of the meaning of the whole expression) may transfer
because of analogical extension from the set of subjunctive uses to the set
of indicative uses, or vice versa. A change now taking place in French is
the use of «après que» with the subjunctive:
11 Je ne repartirai qu'après que vous m'ayez donné votre réponse.
Prescriptive grammar still, however, dictates the indicative (Grevisse 1961:
201-2). This question has been discussed extensively (Grevisse 1961, Barral
1967, Wilmet 1976) and there is general agreement that the change to the
subjunctive in the second clause is by analogy with the conjunction «avant
que» which has always governed the subjunctive. Since they are both ex
pressions of temporal sequencing and, possibly, because of their rather
similar prosodic shape, their syntactic and semantic behavior are now perceiv
ed as closer together.
This is not to argue, though, that all expressions with «que» form a
class, or are in the process of forming a class, regardless of the meaning
of the trigger; «parce que», «puisque» show no sign of becoming subjunc
tive triggers in French just because they have the same prosodic shape and
morphological structure as other one and two-syllable prepositions and
adverbs combined with «que» to form conjunctions. They are, in fact, pro
totypical instances of triggers in the indicative set.
4.2 The following examples show how items on the periphery of a
given set may transfer to another set. In each case they are extensions of
THE DEVELOPMENT OF THE FRENCH SUBJUNCTIVE 613
the central meaning of their set and, with very little reinterpretation, have
become instead extensions of meaning in the other set.
4.2.1 As was mentioned above, the verb «espérer», which in Old
French generally governed the future indicative and in modern French an
indicative tense, either future, or past, or present, shows more signs now,
even in the affirmative, of being used with the subjunctive. The usual forms
are the following2:
12 a c'est mes espoirs/Que courtoisie me ferez
«it is my hope that you will do me courtesy»
b Car j'ay espoir que Dieu ma guerison/Ordonnera
«For I hope that God will order my cure»
13 J'espère que ce café vous a plu/plaira/plaît.
There are, in addition, sentences of the following kind:
14 J'espère que ce café vous plaise
In the older, indicative form the semantic extension is from knowledge, a
form of certainty which takes the indicative, to a non-judgment about cer
tainty, and then extended to hope, with very little volition involved. With
a diachronic shift of emphasis towards volition, and therefore more uncer
tainty about the outcome of hope, comes a shift to the subjunctive. The
distance from the central, certain-knowledge meaning of the indicative set
has made this shift, still very much optional, a possibility. In addition, Old
French allowed only the future with «espérer», so that the lack of sure
knowledge about the verb in the sentential object was marked by the mean
ing of that tense. With the extension of the tense expression to the past and
present as well, other forms of marking the meaning of the relationship bet
ween «espérer» and another verb are becoming more attractive to the speaker
of modern French.
4.2.2 Similar analyses can be made for the verb «croire» in the affir
mative and «savoir» in the negative. Contrary to the usual use in modern
French, «croire» and its near equivalents «cuidier» and «penser» (cf. Jensen
1974: 53-4 on their differences) are found with the subjunctive in Old French,
especially in the second and third persons where the belief of some one else
2 Examples from Old French come variously from Jensen 1974, Ménard 1973, and
Moignet 1973. For modern French, examples have been fabricated or taken from the Larousse
de la grammaire 1983.
614 MARGARET E. WINTERS
4.2.3 With negation, the verb «savoir» is collocated with the subjunc
tive in Old French:
18 a Mais nus ne set que ce soit il.
«But nobody knows that it is he»
b Ne sorent que il fuissent frere.
They did not know that they were brothers»
In modern French, the indicative is found:
19 a Mais personne ne sait que c'est lui.
b Ils ne savaient qu'ils étaient frères.
As with «croire» above, the move from one set to another is based on the
strength of identification with a core meaning from a distance. Not know
ing is seen in Old French as a kind of doubting, while in modern French
it has become an extension of the same set as «croire»-type verbs, via seman
tic movement from certainty to knowledge and then, via negative knowledge
(knowledge of something's not being so) to not knowing:
20 Old French: doubt → not knowing
Modern French: certainty → knowledge → negative knowledge →
not knowing
4.3 The last example concerns verbs and expressions of evaluation,
such as the modern French «je regrette», «il est content», which are now
used with a subjunctive for the verb designating what is being evaluated.
Some standard examples are:
21 a Je regrette qu'il pleuve aujourd'hui.
b II est bien que tu sois prête à partir.
Elle est contente qu'il nous ait vus.
In Old French the indicative was much more the rule, with the subjunctive
being used to express the fact that the process being evaluated was a possibili
ty or hypothetical rather than a reality:
22 a De ce li poise qu'il ne vient.
«For this reasons he was sorry that he wasn't coming»
b Forment li plest qu'ira chacier.
«It pleases him greatly to be going hunting»
23 Certes ce poise moi molt fort/Que je li doie doner mort.
«Certainly I am very sorry that I must kill him»
In 22 are instances of the indicative, while 23 has one example which con
forms to modern use. What has happened is that the meaning underlying
the collocation of the trigger and second verb has shifted. In Old French
the focus is on the opposition of reality to non-reality or hypothesis, not
necessarily with reference to something envisaged as unreal, but where the
reality may simply be in doubt. In modern French, on the other hand, the
subjectivity of the judgment governs the choice of mood.
The above discussion about evaluative verbs provides an instance of
a shift, not in the relationship between extensions of prototypical meanings
616 MARGARET E. WINTERS
5. Conclusion.
A firm statement of what exactly this prototype (or prototypes) in
modern French is for the meaning of the subjunctive must wait for further
research. Based on some of the above discussion, I would posit non-reality
(but not necessarily unreality) as an Old French prototypical meaning of
the subjunctive, with a shift toward greater emphasis on subjectivity in
modern French. It will be a question for further analysis whether there is
or ever has been one central meaning of this mood or several, an analysis
which ultimately will need to consider carefully Middle French and Classical
French as intermediate stages, to look back to Latin and sideways to the
other Romance languges.
The present paper is obviously preliminary. To summarize my proposal,
I am working within a theory of grammar based in semantics, with syntax
as symbolic of meaning and therefore meaningful in itself. By using a radial
model of the internal structure of sets, syntactic (and therefore semantic)
change can be studied by consideration of the relationship of a given con
struction to prototypical members of the set to which it belongs.
Diachronically, sets change, first through rearrangement of members within
a set and then through membership or non-membership in the set. Those
set members at the periphery, that is, those which are related to the pro
totypical members through semantic extension, are most vulnerable to
change. Changes occurring first at the periphery, however, may ultimately
result in a different shape for the set itself and thus for perceptions by native
speakers about what constitutes prototypicality.
THE DEVELOPMENT OF THE FRENCH SUBJUNCTIVE 617
REFERENCES
ROGER WRIGHT
University of Liverpool
1979 I pointed out that lexical diffusion theory demolished the phonetic
evidence that had bolstered the view that Late Latin and Early Romance
were totally separate languages. Esa Itkonen in Helsinki, Michael Richter
in Dublin, Marc van Uytfanghe in Ghent and others have been working
simultaneously on Romance historical linguistics and Late Latin texts.
Romance historical linguists quite often now consider together Late Latin
and subsequent Early Romance documentary evidence without erecting an
artificial barrier between the two; e.g., among others, Timo Riiho's work
on Spanish prepositions, Mario Saltarelli's paper on syntactic diffusion at
the Stanford conference, Dieter Wanner's paper to this conference and
Suzanne Fleischman's book on the future. But there are still scholars who
ignore the Late Latin evidence entirely; for example, Saussol wrote an in
teresting book on the use of the Spanish copulative verbs estar and ser in
the Poema de Mío Cid, and then included in the book's title the phrase
«Origins of their functions», as if nobody in Spain had talked at all until
the twelfth century, or as if their Latin etyma STARE and SEDERE had
never existed.
The Romance Reconstructionists are of course right to be wary of some
of the Mediaeval Latin texts: texts written in England, Ireland or Germany,
because these were probably written by native speakers of languages other
than Romance. In addition, texts written after the ninth-century educational
reforms in France and much of Italy, or after 1080 in most of Spain, are
likely to be written by authors who did regard Latin as a conceptually en
tirely distinct language from their own spoken Romance; but texts from
before those dates in Romance-speaking areas were probably written by
speakers of Old Romance, who were using the only written mode they knew
for recording their language (cp. Wright, 1982). This was not a direct
transcription of speech any more than modern written French or English
is a direct phonetic transcription of spoken French or English. This lack
of direct correspondence means that some phonetic and morphological
reconstruction is necessary, but it is even in these fields a good idea to look
at texts and see if genuinely attested spellings appear in a chronological order
compatible with the order required by the simplicity metric. And in lexical
studies, the appearance of a new word in written form is precisely the
evidence we need (as when Moralejo discovered the source of Spanish
jerigonza written in a seventh-century text as Ihericuntina lingua).
In many respects the language of Late Latin texts seems to us archaic.
But there is no reason to assume that the apparent coexistence of an attested,
SEMANTIC CHANGE IN EARLY ROMANCE 621
eye has semantically real psychological boundaries in a way that other parts
of the face have not. For in Spanish, for example, the word for «mouth»
used to mean «cheek» (boca), the word for «chin» used to mean «little
beard» (barbilla), the word for «eyebrow» used to mean «eyelids» (ceja),
the word for «nose» used to mean «nostril» (nariz/ces), and the word for
«cheek» used to mean «jawbone» (mejilla), but the word for eye has not
changed (ojo). And now, at last, I'll turn to semantic change directly.
Semantic change in lexical items is often intelligently discussed by
etymologists whose primary focus is on the phonetic: e.g. Malkieľs paper
to this conference; but in general it is an underdeveloped branch of historical
linguistic theory. This is not because it is in itself a small subject. Countless
lexical items have acquired new meanings. Yet historical linguistic textbooks,
while occasionally mentioning it, rarely even give it a separate chapter (and
even these conferences average only about one paper per conference on the
topic). Writers on semantic theory, at least within the European tradition,
have been more likely to discuss change; Ullmann, Vilches Acuña, Waldron
and Baldinger, for example; but specifically diachronic semantic principles
seem not to have been elaborated in the last twenty years. Perhaps this is
because of Gilliéron's famous motto, which is certainly true, that each word
has its own history. Since historical linguistics, like other branches of
linguistics, has recently been dominated by believers in universals, the
generally accepted presumption that there are no universals in semantic
change may have led some linguists to assume that it is of no interest. This
is to put the cart before the horse. There can be no objection to looking
for universals, but there are strong objections to the assumption that if
universals are not found the whole investigation must therefore be pointless.
Roger Lass's realization that in diachronic linguistics there are no rigorous
ly predictive principles was a great step forward; no matter that many Euro
pean linguists have always taken this for granted anyway, it is an advance
for an instinctive seeker after clear universals to admit that there may not
be any. In diachronic semantics, no scholar has seriously been tempted to
claim that there are.
How and why does the modern Spanish word verdugo, «executioner»,
come from a mediaeval word meaning «sapling»? Schulte-Herbrüggen's ar
ticle will tell you how. It is hardly a predictable development. Nobody wants
to predict that the English word sapling will mean «executioner» in 700 years
time. But the change is explicable if we understand the difference between
sense and reference, if we know a little semantic theory, and, above all, if
SEMANTIC CHANGE IN EARLY ROMANCE 623
(whoever is going on a journey and comes into someone's lands); the older
meaning of «put together» is also attested. Since there seem to be no uses
of plicare in these texts, the evidence supports the view put forward by Cor
ominas that the direct development of these words was that of allegar from
ADPLICARE, which could in Imperial Roman times be used of ships com
ing into dock (not of folding sails); we could add now that the subsequent
backformation of llegar is not surprising, since many Old Spanish verbs have
two forms, one with a prefix a- and one without. Corominas (1980) quotes
Classical uses, the early fifth-century Peregrinatio Aetheriae, and eleventh-
century Mediaeval Latin uses, but the seventh-century ones would have
helped him considerably.
Spanish querer, «want» and «love», from Latin QUAERERE, «seek»,
is copiously attested meaning «love» in the seventh-century Visigothic prayer-
book, where the congregation are described simply as te querentes (as op
posed to those who hate the Lord: e.g. Oracional 765, Ne obliviscaris,
Domine, vocem querentium te, ut superbia eorum qui te oderint...)՛, it can
also mean «seek» (117: et querere et invenire), or both at once. Spanish dom
ingo «Sunday», comes from the adjective DOMINICUS: in seventh-century
texts this can still be a general adjective connected with any lord, any
dominus, but there are also six uses as a noun meaning «Sunday» in the
Visigothic prayerbook, and eight in the full phrase die dominico՛, we can
suggest that the new nominal use had arrived without implying that the old
more general and adjectival use had dropped from speech yet. We also find
feria used for «weekday», as in Modern Portuguese feira, developed away
from the original «holiday». Other new Christian meanings include missa
as a noun meaning «mass» in Oracional 550; sermo meaning «sermon» as
well as the older and less specialized «speech»: communicare and commumo
with both new specialized and older general meanings; whereas there is no
sign yet, for example, of the development of incensum from a participle
meaning «burnt» to a noun meaning «incense». It has also proved possible
to use this documentary evidence (Wright, in press) to argue that a word
often thought to have undergone a semantic change did not in fact do so:
Mediaeval Spanish ladino, meaning «vernacular», comes from LATINUM;
but rather than changing from meaning «Latin» to meaning «Romance»,
this lexical item always meant «vernacular», in the Roman Empire, in Isidore,
in the ninth-century Christian writers of Moslem Spain, and in Old Spanish.
Scholars have been looking at the usage of Isidore since his own lifetime,
626 ROGER WRIGHT
of course: the work of Sofer and Fontaine, for example, is essential. All
I want to do today is to point out that both general and Romance students
of diachronic semantics have got documentation available for a long period,
documentation which they tend to ignore but which might well be more
helpful than reconstruction. Rothwell has pointed out that discussing seman
tic change in Old French from a basis of theoretical reconstruction has often
led to published analyses that can be simply disproved by actually looking
at Old French texts; in essence I want to move that argument five centuries
earlier.
REFERENCES
1 Wir vernachlässigen, daß es im Isländischen noch zwei weitere Klassen von auf Kon
sonant endenden einsilbigen Feminina gibt, die vō- und die jō-Deklination. Sie haben die glei
chen Flexive wie die ō-Deklination, zeigen aber vor diesen die Elemente -v- bzw. -j-, vgl. stöð
«Station» - G.Sg. stöð-v-ar und skel «Muschel» - G.Sg. skel-j-ar usw. Sie lassen sich leicht
in die hier erörterten Zusammenhänge einordnen. (In der schließlich favorisierten Variante IV
der PSB haben sie wie der Typ vél im Lexikon das Merkmal [ar/N.A.Pl.] und zusätzlich
[v-Erweiterung] bzw. [j-Erweiterung], wodurch ihr Status im Flexionssystem gut erfaßt wird).
PARADIGMENSTRUKTURBEDINGUNGEN 631
Variante I
Als faktisch einfachste Variante könnte man sich PSB der folgenden Art
vorstellen, die die Flexionsparadigmen der vier Klassen erfassen:
(I) (1) fem. i-Dekl.: [Ø/N.Sg., ar/G.Sg., 0 / D . A . S g . ,
i r / N . A . P 1 , a/G.Pl., um/D.Pl.}
(2) fem. ō-Dekl. [ 0 / N . S g . , ar/G.Sg., 0/D.A.Sg.,
einsilbig: ar/N.A.Pl., a/G.Pl., um/D.Pl.}
(3) fem. kons. [ 0 / N . S g . , ar/G.Sg., 0/D.A.Sg.,
Dekl. 1: ur/N.A.Pl., a/G.Pl., um/D.Pl.}
(4) fem. kons. [ 0 / N . S g . , ur/G.Sg., 0/D.A.Sg.,
Dekl. 2: ur/N.A.Pl., a/G.Pl., um/D.Pl.}
Es ist leicht zu sehen, daß diese PSB im wesentlichen nichts anderes
ausdrücken als die Zusammengehörigkeit der mit bestimmten Flexionsen
dungen gebildeten Formen zu einem Paradigma 3 . Die Formen sind rein
konjunktiv zusammengefaßt; das Paradigma erscheint nur als die Summe
seiner Formen. Alle Formen einschließlich des N.Sg. werden als für den Auf-
2 In «gehobener Sprache» erscheinen z.T. im D.A.Sg. von einsilbigen Feminina wie höll
«Schloß», jörð «Erde» und laug auch Formen auf -u. Vgl. dazu weiter unten.
3 Allerdings nehmen diese PSB nur auf die Flexionsendungen und nicht auf die Vokal
wechselerscheinungen Bezug, die wir hier der Übersichtlichkeit halber außerhalb der Betrach
tung lassen. Für die Fakten vgl. Kress 1963: 62 ff.
632 WOLFGANG ULLRICH WURZEL
Variante II
Diese Variante der Fomulierung von PSB geht im Unterschied zur Variante
I davon aus, daß die Paradigmen der verschiedenen Flexionsklassen sich nicht
notwendigerweise in allen Formen unterscheiden, sondern auch bestimmte
von der jeweiligen Flexionsklassenzugehörigkeit unabhängige gemeinsame
Flexionsformen haben können:
(II) (A) Substantiv: { um/D.P1} 4
(B) Substantiv auf Konsonant bzw. schweren Vokal:
{a/G.Pl.}
(C) einsilbige Feminina: [ø/D.A.Sg.)
(1) fem. i-Dekl.: {ar/G.Sg., ir/N.A.Pl.}
(2) fem. ō-Dekl.: {ar/G.Sg., ar/N.A.P1}
(3) fem. kons. Dekl. 1: (ar/G.Sg., ur/N.A.Pl.)
(4) fem. kons. Dekl. 2: {ur/G.Sg., ur/N.A.Pl.}
Anders als bei Variante I gibt es hier zwei verschiedene Typen von PSB.
Die PSB (A) bis (C) spezifizieren auf der Basis von unabhängigen syntakti
schen und phonologischen Eigenschaften der Grundform der Wörter, d.h.
des N.Sg., wie «Substantiv», «Femininum», Silbenzahl und Auslautsegment
solche Flexionsformen der Paradigmen, die über die einzelnen Flexionsklas
sen hinaus gelten. Diese flexionsklassenübergreifenden Eigenschaften brau
chen dann nicht mehr in den PSB (1) bis (4) zu erscheinen, die sich jeweils
auf eine bestimmte Flexionsklasse beziehen. Da die PSB des erstgenannten
Typs von Eigenschaften der lexikalischen Grundform ausgehen, ist im Un
terschied zur Variante I die Form des N.Sg. hier bereits vorausgesetzt und
4 Eine solche generelle Formulier???ng der PSB (A) ist nur möglich, wenn wir annehmen,
daß bei einsilbigen Substantiven auf schweren Vokal wie á «Fluß» der Vokal des Flexivs -um
durch eine phonologische Regel getilgt wird, vgl. den D.Pl. á-m. Diese Annahme ist jedoch
streng genommen nicht haltbar, weil eine solche Regel, die ohnehin für Adjektive und Verben
nicht gelten würde (vgl. blá-r «blau» ֊ D.Pl. blá-um und fá «bekommen» - 1. Ps. Sg. Präs.
Ind. vid-fá-um), nicht einmal alle einschlägigen Substantive erfassen würde, vgl. z.B. vé «Hei
ligtum» - D.Pl. vé-um und snjó-r «Schnee» - D.Pl. snjó-um «Schneefällen». Es wäre also ei
gentlich für den D.Pl. neben -um noch ein Flexiv -m anzusetzen. Wir wollen die PSB (A) hier
dennoch in der aufgeführten Form stehenlassen und kommen später auf die Problematik zurück.
634 WOLFGANG ULLRICH WURZEL
wird nicht erst durch die PSB spezifiziert. Damit kommt in dieser Variante
der gegenüber den übrigen Flexionsformen ausgezeichnete spezifische Sta
tus des N.Sg. als lexikalischer Grundform, d.h. als «Lern-und Merkform»,
klar zum Ausdruck. Weiterhin werden - wie bereits gesagt - über die Einzel
paradigmen hinausgehende Generalisierungen über den Aufbau der Para
digmen der Sprache getroffen und insofern wird zwischen generellen und
spezielleren Flexionsformen unterschieden. Auch die PSB der Variante II
setzen noch partiell die traditionelle Flexionsklasseneinteilung voraus, wenn
sie sich auch so weit wie möglich schon auf unabhängige Eigenschaften der
Wörter beziehen.
Was die Erfassung, d.h. die Erklärung und Vorhersage, von Sprachve
ränderungen betrifft, so stellt sich Variante II nur in einem Punkt angemes
sener als die Vorgängervariante dar. Auch hier geben die PSB nicht an, wie
potentielle abweichende Paradigmen angepaßt werden, und die Frage, wes
halb den PSB entsprechende Paradigmen die Flexionsklasse wechseln kön
nen, bleibt unbeantwortet. Variante II erlaubt jedoch begründete Vorher
sagen darüber, wie Neuwörter mit ihren gegebenen unabhängigen Eigen
schaften einige ihrer Flexionsformen bilden: Die PSB (A) bis (C) besagen,
daß im Isländischen neue Substantive im D.Pl. das Flexiv -um erhalten, daß
neue Substantive auf Konsonant bzw. schweren Vokal im G.P1. das Flexiv
-a zugewiesen bekommen und daß neue einsilbige Feminina im D.A.Sg. ohne
Flexiv bleiben, so daß also z.B. das entlehnte Wort kók die Formen D.A.Sg.
kók, G.P1. kók-a und D.Pl. kók-um bilden mußte. Für die übrigen von der
Flexionsklassenzugehörigkeit abhängigen Formen des G.Sg. und des N.A.P1.
legen sie dagegen nichts fest; die entsprechenden Formen bleiben nach wie
vor unerklärt 5 .
Variante III
Die Variante III unterscheidet sich von ihren Vorgängervarianten prinzipiell
darin, daß bei ihr im Paradigma existierende implikative Zusammenhänge
zwischen den Flexionsformen nachvollzogen werden. Vgl. die Formulie
rungen:
() D [0/D.A.Sg.]
(D [i'r/N.A.Pl. ] [ar/G.Sg.]
(2) [ ar/N.A.Pl. ] [ar/G.Sg.]
(4) [ ur/G.Sg. ] D [ur/N.A.Pl.]
Die PSB (A) bis (C) in Variante III drücken die gleichen faktischen Gege
benheiten aus wie ihre Entsprechungen in Variante II, allerdings indem sie
den implikativen Charakter der Zusammenhänge zwischen den Eigenschaf
ten der Grundform und den anderen Flexionsformen explizit machen (-K
bedeutet «auf Konsonant endend», - V «auf schweren Vokal endend», # Σ #
steht für «Einsilbigkeit»). Eine völlig andere Form als ihre Pendants in Varian
te II haben dagegen die PSB (1) bis (4). Sie nehmen generell keinen Bezug auf
die traditionellen Flexionsklassen mehr, sondern konstatieren statt dessen im-
plikative Zusammenhänge zwischen Flexionsformen. Diese implikativen Zu
sammenhänge beziehen sich zwar rein faktisch auf bestimmte Flexionsklassen,
haben aber - was wichtig ist - für das gesamte Flexionssystem Gültigkeit; bei
spielsweise zeigen eben alle isländischen Substantive mit -ir im Ν./Α.P1. ent
sprechend -ar im G.Sg. usw. Die PSB (3) von Variante II hat kein Gegen
stück in Variante III, weil zwischen den beiden relevanten Formen keine im-
plikative Beziehung existiert. Vgl. die Implikationen zwischen den variierenden
Flexionsformen der vier in die Betrachtung einbezogenen Flexionsklassen:
form zu erlernenden «Kennformen» wie z.B. dem N.A.Pl. auf -ir uns sich
implikativ ergebenden Formen wie z.B. dem G.Sg. auf -ar beim Typ mynd
sowie dem flexivlosen D.A.Sg. der einsilbigen Feminina.
Wörter miteinander konkurrierender Flexionsklassen mit den gleichen
unabhängigen syntaktischen und/oder phonologischen Eigenschaften wie
hier die einsilbigen Feminina auf Konsonant können dann im Lexikon ge
nerell durch die Angabe der differenzierenden «Kennformen» repräsentiert
werden, so der Typ mynd durch [ir/N.A.P1], der Typ vél durch [ar/N.A.P1]
und der Typ vík durch [ur/G.Sg.]. Man beachte, daß der Typ geit, bei dem
es keine implikative Beziehung zwischen den Formen des G.Sg. und des
N.A.Pl. (in welcher Richtung auch immer) gibt, durch die beiden Kennfor
men [ar/G.Sg., ur/N.A.PL] zu repräsentieren ist. Die übrigen, im Lexikon
nicht spezifizierten Formen ergeben sich redundanterweise aufgrund der PSB.
Auch im Lexikon brauchen damit keine Merkmale von traditionellen Fle
xionsklassen (wie i-oder ō-Deklination) mehr zu erscheinen 6 .
Was leisten nun die PSB in dieser Form für die Erklärung von auftre
tenden und zu erwartenden Sprachveränderungen? Wie die PSB der Varia-
ten I und II determinieren sie, daß Neuwörter ins Flexionssystem eingepaßt
und abweichende Paradigmen an dieses angepaßt werden. Wie die PSB der
Variante II legen sie fest, in welcher Weise Neuwörter aufgrund ihrer unab
hängigen syntaktischen und/oder phonologischen Eigenschaften bestimm
te Flexionsformen bilden; das Gesamtparadigma von Neuwörtern können
aber auch sie nicht erklären.
Zusätzlich zur Variante II erfassen die PSB jedoch aufgrund ihres im-
plikativen Charakters die mögliche Reihenfolge des Auftretens neuer For
men bei stufenweise verlaufenden Klassenübertritten. Die konsonantische
Deklination 2 unterscheidet sich in zwei Formen von der i-Deklination, vgl.
vík- vs. mynd-ar im G.Sg. und vík- vs. mynd-ir im N.A.Pl. Damit gibt
es theoretisch gesehen drei Wege des Übergangs vom Typ vík zum Typ mynd:
den Wechsel beider Formen zugleich, den Wechsel über G.Sg. -ar/N.A.Pl.
-ur und den Wechsel über G.Sg. -ur/N.A.PL -ir. Doch die PSB (4) besagt
ja, daß -ur im G.Sg. auch -ur im N.A.Pl. impliziert, und schließt damit die
Kombination -ur/-ir aus. Deshalb erfolgt der Übergang vom Typ vík zum
6 Man vgl. dazu die Angabe von «Kennformen» in Gebrauchswörterbüchern, die aller
dings in der Regel unreflektiert und daher inkonsequent erfolgt. So geben z.B. alle isländi
schen Wörterbücher von jedem Substantiv G.Sg. und N.P1. an, obwohl oft eine der beiden
Formen zur Erschließung der Deklinationsklasse ausreichen würde und in Fällen wie penni über
haupt keine «Kennform» notwendig wäre.
PARADIGMENSTRUKTURBEDINGUNGEN 637
Typ mynd direkt oder aber über -ar/-ur (den Typ geit), was die Fakten ein
deutig erweisen. So schwankt gegenwärtig z.B. das Substantiv mörk «Wild
mark» zwischen -ur/-ur und -ar/-ir, und Wörter wie eik «Eiche», steik «Bra
ten» und tìk «Hündin» schwanken zwischen -ur/-ur und -arAur. Ein
Schwanken zwischen -ur/-ur und -urAir gibt es aber nicht. Man kann also
sagen, daß Variante III den Verlauf von Klassenübertritten erfaßt. Nach
wie vor wird jedoch nicht erklärt, weshalb solche Übertritte überhaupt statt
finden und wodurch ihre Richtung bestimmt ist.
Variante IV
Die letzte zu diskutierende Variante der Fomulierung von PSB zeichnet sich
gegenüber den anderen dadurch aus, daß sie auf Markiertheitsverhältnisse
Bezug nimmt. Sie unterscheidet zwischen unmarkierten und markierten Fle
xionsformen und damit auch zwischen unmarkierten und markierten Para
digmen:
(IV) (A) [ + Subst ] d [um/D.P1.]
d [a/G.P1.]
() D
d [0/D.A.Sg.]
(Q
D
d
) D
([ir/N.A.P1.] ) [ar/G.Sg.]
eine andere N.A.P1. - Form spezifiziert ist, und daß ein Substantiv mit -ir
im N.A.P1. im G.Sg. das Flexiv -ar erhält, wenn nicht eine andere G.Sg.
- Form spezifiziert ist. Damit benötigen die Substantive des Typs mynd im
Lexikon keine Flexionsklassenspezifizierung mehr; sie bekommen sie durch
die PSB «automatisch» zugeschrieben. Anders die drei übrigen Typen: Die
Wörter des Typs vél müssen im Lexikon das Merkmal [ar/N.A.Pl.] spezifi
ziert haben, ihnen kann also nicht der N.A.P1. auf -ir zugewiesen werden.
Die erste Default-Implikation der PSB (1) ist blockiert. Da aber, weil der
mittlere Teil der PSB fakultativ ist, die unabhängigen Merkmale den G.Sg.
auf -ar auch direkt implizieren, bekommen diese Wörter ebenfalls ihre kor
rekte G.Sg. - Form auf -ar zugewiesen. In völlig gleicher Weise erhalten die
Wörter des Typs geit ihre Flexionsformen zugeschrieben, nur daß es hier
das Lexikonmerkmal [ur/N.A.Pl.] ist, das die Zuweisung von [ir/N.A.P1.]
blockiert. Auch diese Substantive bekommen dann aufgrund der auch di
rekten Implikation von den unabhängigen Merkmalen zu [ar/G.Sg.] ihre
G.Sg. - Form. Die Wörter des Typs vík schließlich sind im Lexikon durch
das Merkmal [ur/G.Sg.] gekennzeichnet. Für die Behandlung dieser Wör
ter ist wichtig, daß (entsprechend einem allgemein angenommenen Prinzip
der Regelordnung) im Falle der möglichen Anwendung zweier sich wider
sprechender Regeln die jeweils speziellere Vorrang gegenüber der allgemei
neren hat 7 . PSB (1) gilt für alle Feminina auf Konsonant, PSB (4) aber nur
für die Substantive auf -ur im G.Sg., eine echte Teilklasse davon. Deshalb
bekommen die Wörter vom Typ vík durch PSB (4) das Merkmal
[ur/N.A.Pl.]. Weil damit auch die N.A.P1. - Form spezifiziert ist, kann PSB
(1) auf diese Substantive überhaupt nicht angewandt werden. (Hätte dage
gen PSB (1) Vorrang gegenüber PSB (4), dann würde für solche Wörter in
korrekterweise [ir/N.A.Pl.] spezifiziert).
Die Formulierung der PSB als Default-Implikationen, durch die sich
die Variante IV von den vorher diskutierten unterscheidet, ermöglicht es,
den unterschiedlichen Status von Flexionsparadigmen innerhalb eines Fle
xionssystems hinsichtlich ihrer Markiertheit in recht angemessener Weise wi
derzuspiegeln: Bei den auf Konsonant endenden einsilbigen Feminina des
Isländischen ist in Übereinstimmung mit den bereits erwähnten sprachhi
storischen Fakten der Typ mynd, also die i-Deklination, unmarkiert, die an
deren drei Typen sind entsprechend markiert. Die Unmarkiertheit der i-
7 Für die Praktizierung dieses Prinzips in der Morphologie vgl. z.B. Anderson 1982:
593f., 606ff. Kiparsky 1982: 9 ff. und Zwicky 1985: 3 ff.
PARADIGMENSTRUKTURBEDINGUNGEN 639
8 Zu den hier relevanten Fakten s. Noreen 1923: 269 für das Altisländische und Kress
1963: 84 für das Neuisländische. Zur Unterscheidung zwischen unmarkierten (stabilen) und
markierten (instabilen) Flexionsklassen und deren Begründung vgl. Wurzel 1984: 116 ff.
9 So sind z.B. die phonologischen Merkmale [ + hoch] und [ + niedrig] nicht in einem
Segment miteinander kombinierbar.
640 WOLFGANG ULLRICH WURZEL
G.Pl. statt des erwarteten Flexivs -a das Flexiv -na aufweist. Was PSB (A)
betrifft, so wurde bereits (in Anmerkung 4) darauf verwiesen, daß für den
D.P1. neben -um noch ein selbständiges Flexiv -m vorzusehen ist. Dieses Fle
xiv erscheint ganz regulär bei den Feminina auf schweren Vokal, vgl. á
«Fluß» - D.P1. á-m und bru «Brücke» - D.P1. bru-. Das Isländische ver
fügt also noch über eine zusätzliche PSB (A'), die Feminina auf schweren
Vokal das Merkmal [m/D.PL] zuweist10. Diese PSB kann sich aber nicht
in gleicher Weise auf die ebenfalls vorkommenden Neutra auf schweren Vo
kal beziehen, da bei diesen z.T. -m und z.T. -um erscheint, vgl. tré «Baum»
- D.P1. trjá-m (mit Vokalwechsel) und vé «Heiligtum» - D.P1. vé-um. Die
vereinzelten Neutra auf -m im D.P1. wie tré sind also Ausnahmen zu PSB
(A); sie müssen im Lexikon durch [m/D.P1.] spezifiziert sein, damit sie nicht
das ihnen regulär eigentlich zukommende Merkmal [um/D.Pl.] durch PSB
(A) erhalten können. Fälle dieser Art, die gar nicht so selten sind 11, bilden
faktisch einen Übergangsbereich zwischen PSB mit strikten und Default-
Implikationen, und es ist nicht klar wie sie plausibel zu klassifizieren wä
ren, so daß uns auch in dieser Hinsicht die Annahme, daß alle PSB einheit
lich Default-Charakter haben, angemessener erscheint12.
Es sind aber dennoch ihrem Aufbau nach zwei verschiedene Typen von
PSB zu unterscheiden, wie man leicht sehen kann. Die PSB (A), (B), (C)
und (1) stellen jeweils einen implikativen Zusammenhang zwischen unab
hängigen Eigenschaften der Grundform der Wörter und unmarkierten For
men her, während PSB (4) einen implikativen Zusammenhang zwischen ei
ner im Lexikon spezifizierten markierten Form und einer weiteren markier-
10 Die PSB (A') ist spezieller als PSB (A), weil sie sich auf Feminina auf schweren Vo
kal bezieht, während sich PSB (A) auf alle Substantive bezieht. Sie hat demzufolge Vorrang
in der Anwendung; vgl. weiter oben.
11 Solche Fälle sind natürlich keineswegs auf das Isländische beschränkt. So gilt im Rus
sischen nahezu durchgängig, daß Substantive im Instr. P1. -ami/-jami haben, doch erscheint
in zwei Fällen obligatorisch und in drei Fällen fakultativ statt dessen -mi; vgl. det-i «Kinder»
- det'-mi, ljud'-i «Menschen» - ljud'-mi; dver' «Tür» - dver'-mi, doč' «Tochter» - dočer'-mi
und loŠad' «Pferd» - loŠad'-mi. Im Deutschen haben alle Maskuline auf -e außer dem Einzel
fall Käse im Plural -n usw. usf.
12 Damit besagt dann allerdings PSB (4), zu der es keine Ausnahmen gibt, auch nur
noch, daß normalerweise keine Substantive mit G.Sg. - ur/N.A.Pl. -ir vorkommen; sie sind
nicht mehr grundsätzlich ausgeschlossen. Ein Übergang von G.Sg. -ur/N.A.Sg. -ur zu G.Sg.
-ar/N.A.Pl. -ir über G.Sg. -ur/N.A.Pl. -ir kommt deshalb nicht vor, weil die erste Stufe der
Entwicklung keinen Abbau morphologischer Komplexität in Termen der Lexikonrepräsenta
tion darstellt.
PARADIGMENSTRUKTURBEDINGUNGEN 641
13 Es sei denn, sie sind mit Derivativen (Suffixen) gebildet, die bestimmten Flexionsklas
sen zugeordnet sind.
642 WOLFGANG ULLRICH WURZEL
14 Es gibt im übrigen auch Übertritte von der (in Anm. 1 erwähnten) vō-Deklination zur
i-Deklination, z.B. ist dögg «Tau» übergetreten und ör «Pfeil» befindet sich im Übergang.
15 Man beachte, daß in Variante III die i-Deklination, die o-Deklination und die kon
sonantische Deklination 2 durch jeweils ein Lexikonmerkmal und die konsonatische Deklina
tion 1 durch zwei Lexikonmerkmale spezifiziert sind. Bei diesen Spezifizierungen läßt sich im
Unterschied zu Variante IV weder der Übergang der übrigen Klassen zur i-Deklination noch
der Übergang der konsonantischen Klasse 2 zur konsonantischen Klasse 1 erklären.
PARADIGMENSTRUKTURBEDINGUNGEN 643
schaften der Struktur von Flexionsparadigmen erarbeitet, die wir kurz als
Regelhaftigkeit (Variante I), Bezugnahme auf unabhängige Eigenschaften
der Wörter (Variante II), implikativen Aufbau (Variante III) und Einbezie
hung von Markiertheitsverhältnissen (Variante IV) charakterisieren können.
Die schließlich daraus resultierenden PSB der Variante IV spiegeln nicht nur
die Zusammengehörigkeit der Flexions formen zu einem einheitlichen Para
digma wider, sondern sie erklären auch, weshalb im Bereich der Flexions
paradigmen einer Sprache ganz bestimmte morphologische Veränderungen
eintreten und andere theoretisch mögliche nicht eintreten16, wodurch dann
entsprechend dem oben formulierten Kriterium die Annahmen über den Pa
radigmenaufbau ihren spekulativen Charakter verlieren und bei weitem plau
sibler werden. Verallgemeinernd läßt sich sagen, daß sich hier wie in vielen
anderen Bereichen der Grammatik die entscheidende Bedeutung der Sprach
veränderung für die Erarbeitung einer angemessenen und verifizierbaren
Grammatiktheorie zeigt. Die Notwendigkeit der Erklärung von Sprachve
ränderungen bedingt einerseits zusätzliche Anforderungen an die Gramma
tik, doch sie führt andererseits zu Einsichten in das Wesen grammatischer
Strukturen, die einem bornierten Synchronismus schlicht verschlossen blei
ben müssen.
LITERATURANGABEN
16 Die PSB erklären aber natürlich nicht alle möglichen Veränderungen in der Flexions
morphologie; vgl. dazu speziell die durch die Tendenz zur Systemangemessenheit bedingten
Veränderungen, s. Wurzel 1984: 81 ff.
644 WOLFGANG ULLRICH WURZEL
Comrie, . 60, 66, 410, 475 Dobson, E.J. 272, 274, ff., 469
Conradie, J. 157 Donegan, P. 517, 524, 548
Conti, G. 538 f., 542 Doria, M. 326
Čop, . 433 Dover, K.J. 23
Corominas, J. 375, 380, 382, 384, Dressler, W.U. 13, 28, 208, 517, 523,
625 526
Cortelazzo, M. 398 Dubois, L. 22
Coseriu, E. 421, 428, 478, 523, 576 Dunkel, G.E. 66
Cowgill, W. 251, 257, 284 Durand, J. 3
Cravens, T. 171, 621 Duranti, A. 42
Croft, W. 181, 183, 186, 188, 358 Durkheim, A. 258
Cromer, R. F. 101 Duro, A. 398
Cummings, Α. 48
Ebbinghaus, E.A. 8
Dahl, Ö. 58, 62 f. Ebeling, G. 398
Dahl, T. 447 Eckman, F. 547
Dalbera-Stefanaggi, M.-L. 163 ff. Edgerton, F. 484 f.
D'Ans, A.-M. 476 Edgren, A.H. 488
Darmesteter, Α. 387 Eichner, H. 499
Dauses, Α. 79, 81, 83 Einarsson, S. 155
De Abrew, . . 65 Einstein, A. 341
Deane, P. 358 Ekwall, E. 153
Debrunner, A. 61 Elmer, W. 124 ff., 132 f.
Delbrück, . 13, 16, 23, 28, 51, 61 Endzelin, J. 24
f., 251, 308, 358 f., 363 f., 366 Enkvist, N.E. 221, 228
De Meo, . 140 Ernout, A. 394 f.
de Serpa Leitão, Α. 469 Espinosa, A.M. 392
Denniston, J. D. 25 Espinosa hijo, A.M. 392
Densusianu, . 424 Ewen, . 3 f.
De Simone, . 596
Devoto, G. 137 f., 143, 397 Faine, J. 476
Diez, F. 382 Faiss, . 239
Dik, S. . 41, 357 ff., 568 Fanciullo, F. 333
Dillon, M. 59 Fant, G. 469
Dilthey, W. 295 Fasold, R. 547
Dimmendaal, G.J. 197, 199, 206 ff. Feuillet, J. 256
Dinnsen, D.A. 547, 549 Fick, Α. 434
Disterheft, D. 215 f., 218, 474 Finck, N:. 592
648 INDEX OF NAMES
Harris, M.B. 35 f., 40, 46 f., 49, 51 Humboldt, W. von 340, 591
f., 96, 217, 608 Hyman, L.M. 42, 89 ff., 216
Hatcher, A.G. 188
Haudry, J. 142, 363, 365 Imbs, P. 608 f.
Haugen, E. 155, 407 Ineichen, G. 591
Haviland, J.B. 411 Ingold, T. 347
Hawkins, J.A. 57, 87, 93 Ingria, R. 167
Hayes, B. 449 f., 455 Iordan, I. 331, 427
Hazan, . 346 Isačenko, Α. 260
Head, . 408, 413 Itkonen, E. 620
Heine, . 111 Ivanov, V.V. 255, 307
Heisenberg, W. 341
Hermann, E. 308 Jackendoff, R. 565
Hermon, G. 213 ff. Jacobi, H. 27
Hettrich, H. 358 Jakobson, R. 3, 87, 359, 376
Heubeck, Α. 437 Janda, R. 151, 156
Hill, J.H. 415 Jansen, F. 291
Hill, K.C. 415 Javkin, H. 465
Hirschbühler, P. 475 Jeffers, R.J. 305, 308, 310, 313, 319
Hirt, H. 16, 62, 251 f., 431, 434 f.
Hochberg, J. 561 Jensen, F. 613 f.
Hockett, Ch. F. 336 Jensen, H. 24
Hoenigswald, H.M. 327 f. Jespersen, . 24, 125, 151, 154 ff.,
Hoffman, O. 439 239, 272, 275, 341, 375 f., 465,
Hofmann, J.B. 62, 139, 145, 147, 517, 524
394 f. Job, M. 355
Hogg, R.M. 2 Jokinen, U. 475
Holland, G. 318 f. Jones, . 3
Hooper, J.B. 170, 219, 454, 517, Jordan, R. 447, 449
526, 554, 556 Joret, Ch. 387
Hoppenbrouwers, C.A.J. 285 f., 290 Joseph, B.D. 124, 218
Hopper, P.J. 59, 221, 223 f., 296, Josephson, F. 18
307 Jucquois, G. 16
Horn, W. 446 Jud, J. 392
Horning, A. 374 Justus, C.F. 309
Huang, G.T.J. 36
Hughes, J.P. 274 Kalmar, І. 223
Hulst, H. van der 4 Karlsson, F. 629
650 INDEX OF NAMES
Li, C.N. 36, 38 ff., 43, 89, 93, 217 Maurei, J.-P. 146
Liddell, S.K. 58 Mayerthaler, W. 517, 521, 524, 528,
Liddel, H.G. 430, 434 591, 597
Lieber, R. 629 Mayr, E. 207
Lightfoot, D. 49, 87, 124, 212, 219, Mayrhofer, M. 60, 599
267, 310 f., 341, 575 Mc David, R.I. 275
Lightner, T. M. 435 Meier, H. 384
Lindblom, . 469, 550 f. Meier, J. 332
Lindeman, F.O. 584, 589 f., 600 Meillet, A. 16, 24 f., 267 f., 308, 342,
Lindholm, F. 140 377 f., 394 f. 435, 439
Lineli, P. 554 Melchert, H.C. 363, 366
Lodge, G. 184 Ménard, P. 613, 616
Loi Corvetto, I. 165 Menéndez Pidal, R. 378, 384, f.,
Longacre, R.E. 221 389, 391 f., 461
López Pereira, J.E. 623 Merlo, . 398
Lord, C. 216 Meyer-Lübke, W. 325, 374, 382,
Lubotsky, A. 584 384, 398
Lüdtke, Η. 517, 524, 595 Michaëlis (de Vasconcelos), . 387
Luick, . 446 ff., 452, 456 f., 390
Luiselli, . 137 ff. Migliorini, . 398
Lunt, Η. 546 Mignot, X. 355
Lujan, M. 578 Miller, D.G. 13, 89, 96
Luraghi, S. 359, 364 ff. Miller, G.A. 469
Lyons, J. 565 Milroy, J. 275 ff.
Minard, Α. 16, 24, 318
Macdonell, A.A. 16, 22, 61 ff. Minkova, D. 446, 448 f., 454
Mace, J. 601 Mitchell, . 221 f.
Maddieson, I. 546 Mitzka, W. 399
Malkiel, Y. 202, 333, 376 ff., 383, Mogk, E. 27
394, 622 Moignet, G. 242, 608, 613
Mańczak, W. 353 Molinelli, P. 58, 65
Manoliu-Manea, M. 421 f. Monteil, P. 314
Marchese, L. 116 Moralejo Laso, A. 620
Marm, I. 64 Morohasi, T. 563
Marti, M. 579 Moscati, S. 538
Martin, R. 420 Mossé, F. 24, 455
Martin, S. 563 Moulton, W.G. 8, 546
Martinet, A. 170, 394, 517, 524 Mowrey, R. 465, 468
652 INDEX OF NAMES
North Frisian 546, 553 ff., 378, 392, 394, 395, 538, 540, 542,
Fula 180, 181, 183, 190 543, 546, 564, 566, 567, 591, 596,
600, 602 (s. also grec, greco,
Ga 58 griechisch)
Gascon 381 Classical Greek 180, 182 f., 184 f.
Georgian 546, 213 f. Homeric Greek 61
German 64, 65, 94, 103, 112, 155, Katharevusa 621
213,354,373,381,399,542,546, Modern Greek 111, 212
547, 553, 591, 593-597 passim, griechisch 147 (s. also grec, greco,
599-601 passim, 603, 604 (s. also Greek)
allemand, deutsch, plattdeutsch, Gugadje 412
tedesco) Gujarati 470
Middle High German, 63, 436, Gungu Yimidirr 183, 411
552, 594-596 passim, 600-602
passim Haitian 111, 476-479
Modern German 134 Hausa 202
Old High German 64, 239, 256, Hebrew 399, 535, 538, 540, 541, 542
343, 431, 439, 595-597 passim, 602 Classical Hebrew 181, 621
Germanic 1, 2, 5, 6, 59, 62, 64, 65, Middle Hebrew 541, 542
90, 154, 158, 217, 239 f., 445, 526 hethitisch 138 f., 142 f. (s. also Hit
f., 529, 546, 553, 584, 593, 597, tite)
599 f., 603 f. (s. also germanique) Hindi 58, 60, 62, 65
North Germanic 155 Hispanic 380
Old Germanie 362, 367 f. Hittite 17 ff., 22, 24, 28, 60, 138 f.,
Proto-Germanie 196, 213 f., 529, 142 f., 216, 312 ff., 358 ff., 431
597 (s. also hethitisch)
West Germanie 217, 187 f. і 416
germanique 27 (s. also Germanie) Hungarian 399, 546
Gothic 1-, 58, 59, 63, 112, 213 f.,
239, 399, 430, 431, 436, 439, 600 Icelandic 155, 341 ff., 546 (s. also
(s. also gotique) isländisch)
gotique 24 f. (s. also Gothic) Modern Icelandic 214
grec 17 ff., 23, 25 (s. also greco, Old Icelandic 181 ff.
Greek, griechisch) Indian 584, 591
greco 326, 329, 429-439 (s. also grec, Old Indian 537 (s. also altindisch)
Greek, griechisch) Indic Languages 60, 65
Greek 2, 3, 57, 58, 61, 64, 66, 213, Indo-European 57-66 passim, 251,
254 ff., 258, 284, 314 ff., 354, 357 305-321, 374, 524, 535-543 passim,
660 INDEX OF LANGUAGES
Teso 194, 199, 204 umbro 329, 336 (s. also umbrisch)
Teso-Turkana Cluster 193-206 Uto-Aztecan 182
Tocharian 58, 60 (s. also tokharien)
Tojalabal 111 Venetic 591 ff., 600
tokharien 24 (s. also Tocharian) vieil anglais 22 (s. also English)
Tok Pisin 475 vieil irlandais 24, 25 (s. also Irish)
Tongan 213 f. Vietnamese 417
Toposa 194, 204
Torinese 335 Walbiri 321
Tunen 182 Welsh 59, 64 f., 354, 546
Turkana 194, 201, 206
Turkish 182, 190, 399, 546, 547, 553
Tuscan 378, 384, 397 Yakut 546
Tuvinian 546 Yiddish 399, 546, 552-556 passim
Yokuts 181 f.
Ukrainian 112, 546 Yoruba 416
umbrisch 138, 142 (s. also umbro)
Uralic 58 Zuñí 62
INDEX OF SUBJECT MATTER
— Markers 99 Clitic
Semantic and Syntactic Value of — Group 450, 453 ff.
— 357 ff. — Movement 575-587
— Syncretism 355-368 — Pronouns and Particles 52,
— Synonimy 356, 358 151-161, 313
— System 355 Cliticization 593 ff.
Causality 183 Cognitive/Space Grammar 607,
Change 478, 556 609-612, 616
Abductive — 478 Complementation 147
Analogical — 528 Infinitival — 578-581 passim, 584
Contingent — 99 Complementizer 474-479 passim
Deductive — 478 Condensation 599-604 passim
Directionality of — 562, 568-571, Consonat Shift
601-604 Germanic — 344 f.
Explanation of — 47 High German — 461
Inherent — 98 Contact 65 f., 459, 468, 470, 473 ff.,
Morphological — 193-209, 474, 477, 530, 553, 603
591-604, 621, 630-643 passim — Vernacular 268-279
Motivations for —344 ff., 525-530 Core Grammar 474, 476
— of Type 478 Coreferentiality 319
Phonetic/Phonological—271 -279, Correlation 94
285 ff., 325-337, 373-406, 459-471, Covariation 493, 496
592-604 passim, 621 Creoles 473-479 passim
Rate of — 343 Creolization 268, 473-479
Semantic — 351, 507-514, Cronologia relativa 325-337
561-572, 607-616, 619-626
Social — 407 Decreolization 477
Syntactic — 35-53, 123-135, Deference 407, 416 f.
211-219, 474-479, 575-587, 621 — Markers 413
Theory of Language — 49, 349 Degemination 171
ff., 517-530 Dependency Phonology 3 f.
Therapeutic — 50, 208 Determination
Universal Laws of Language — Sentence — 222
350 Determiners
Clauses Noun (Phrase) — 476 f.
Adjoint — 321 Detransitivizing Phenomenon 184
Adverbial — 318 f. Diachronic
INDEX OF SUBJECT MATTER 667