Isabelle Bril. Clause Linking and Clause Hierarchy

Clause Linking and Clause Hierarchy
Studies in Language Companion Series (SLCS)

This series has been established as a companion series to the periodical
Studies in Language.
Editors
Werner Abraham
University of Vienna
Elly van Gelderen
Arizona State University
Editorial Board
Bernard Comrie
Max Planck Institute, Leipzig

and University of California, Santa Barbara
William Croft
University of New Mexico
sten Dahl
University of Stockholm
Gerrit J. Dimmendaal
University of Cologne
Ekkehard Knig
Free University of Berlin
Christian Lehmann
University of Erfurt
Brian MacWhinney
Carnegie-Mellon University
Marianne Mithun
University of California, Santa Barbara
Heiko Narrog
Tohuku University
Johanna L. Wood
University of Aarhus
Volume 121
Clause Linking and Clause Hierarchy. Syntax and pragmatics
Edited by Isabelle Bril
Clause Linking
and Clause Hierarchy
Syntax and pragmatics
Edited by
Isabelle Bril
CNRS-LACITO
John Benjamins Publishing Company

Amsterdam/Philadelphia
TM
The paper used in this publication meets the minimum requirements of

American National Standard for Information Sciences Permanence of
Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data

Clause linking and clause hierarchy : syntax and pragmatics / edited by Isabelle Bril.
p. cm. (Studies in Language Companion Series, issn 0165-7763 ; v. 121)
Includes bibliographical references and index.
1. Grammar, Comparative and general--Clauses. 2. Grammar, Comparative and general-Syntax. 3. Pragmatics. I. Bril, Isabelle.
P297.C546 2010
415--dc22
2010034699
isbn 978 90 272 0588 9 (Hb ; alk. paper)
isbn 978 90 272 8758 8 (Eb)
2010 John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. P.O. Box 36224 1020 me Amsterdam The Netherlands
John Benjamins North America P.O. Box 27519 Philadelphia pa 19118-0519 usa
Table of contents
List of contributors
Editors introduction. The syntax and pragmatics of clause linkage
and clause hierarchy: Some new perspectives
Isabelle Bril
vii
1
part i. Syntactic terminology and typological methods

Clause linkage and Nexus in Papuan languages
William A. Foley
27
Capturing particulars and universals in clause linkage: A multivariate analysis

Balthasar Bickel
51
part ii. Clause-chaining, converbs, masdars, absolutive constructions, etc.

Specialized converbs and adverbial subordination in Axaxdr Akhvakh
Denis Creissels
105
Finite and non-finite: Prosodic distinctions on Budugh verb stems

Gilles Authier
143
Converbs and adverbial clauses in Badaga, a South-Dravidian language

Christiane Pilot-Raichoor
165
Coordination, converbs and clause chaining in Coptic

Egyptian Typology and structural analysis
Chris H. Reintges
203
part iii. Subordination, informational hierarchy and referential hierarchy

Informational and referential hierarchy: Clause-linking strategies
in Austronesian-Oceanic languages
Isabelle Bril
269
Comment clause: Crossing the boundaries between simple

and complex sentences
Zygmunt Frajzyngier
313
Deixis, information structure and clause linkage in Yafi Arabic (Yemen)

Martine Vanhove
The role of the Berber deictic and TAM markers in dependent
clauses in Zenaga
Catherine Taine-Cheikh
Deixis and temporal subordinators in Pomak (Slavic, Greece)
Evangelia Adamou
Correlative markers as phoric Grammaticalised Category Markers
of subordination in German
Colette Corts
333
355
399
421
part iv. Informational hierarchy and TAM markers

functions in clause-linkage
Focus, mood and clause linkage in Umpithamu
(Cape York Peninsula, Australia)
Jean-Christophe Verstraete
451
Clause chaining and conjugations in Wolof: A typology of

parataxis and its semantics
Stphane Robert
469
Pragmatic demotion and clause dependency: On two atypical

subordinating strategies inLoTogaand Hiw (Torres,Vanuatu)
Alexandre Franois
499
Tense-mood concordance and clause chaining in Mankon

(a Grassfields Bantu language)
Jacqueline Leroy
549
Clause dependency relations in East Greenlandic Inuit

Nicole Tersis
581
Coordination and subordination: ma in Bulgarian dialectal Greek

Eleni Valma
603
Author index
619
Language index
623
Topic index
625
List of contributors
Evangelia Adamou
Lacito-CNRS, Centre Andr-Georges
Haudricourt
7, rue Guy Mquet
94800 Villejuif, France
adamou@vjf.cnrs.fr
Denis Creissels
Universit Lumire- Lyon 2
Laboratoire Dynamique du Langage,
14, avenue Berthelot
69363 Lyon, France
Denis.Creissels@univ-lyon2.fr
Gilles Authier
Institut National des Langues et
Civilisations Orientales
2, rue de Lille
75007 Paris, France
gilles.authier@gmail.com
William Foley
Department of Linguistics
The University of Sydney,
NSW 2006, Australia
william.foley@arts.usyd.edu.au
Balthasar Bickel
Institut fr Linguistik
University of Leipzig
Beethovenstrasse 15
04107 Leipzig, Germany
bickel@uni-leipzig.de
Isabelle Bril
Lacito-CNRS,
Centre Andr-Georges Haudricourt
7, rue Guy Mquet
ibril@vjf.cnrs.fr
Colette Corts
tudes Interculturelles de Langues
Appliques
Universit Paris 7- Denis Diderot
7, rue Watt
75013 Paris, France
colette.cortes@free.fr
Zygmunt Frajzyngier
Department of Linguistics, Box 295
University of Colorado,
Boulder, CO 80309-0295, USA
Zygmunt.Frajzyngier@Colorado.EDU
Alexandre Franois
Lacito-CNRS,
7, rue Guy Mquet,
francois@vjf.cnrs.fr
Jacqueline Leroy
Universit Paris-Descartes and
Lacito-CNRS,
7, rue Guy Mquet
leroy@vjf.cnrs.fr
lacito-CNRS,
7, rue Guy Mquet,
raichoor@vjf.cnrs.fr
Nicole Tersis
sedyl-celia-CNRS,
7, rue Guy Mquet,
nicole.Tersis@vjf.cnrs.fr
Chris H. Reintges
Laboratoire de Linguistique FormelleCNRS,
Universit Paris 7-Denis Diderot
175, rue du Chevaleret
75013 Paris, France.
chris.reintges@linguist.jussieu.fr
Eleni Valma
lacito-CNRS,
7, rue Guy Mquet,
evalma@vjf.cnrs.fr
Stphane Robert
llacan-CNRS
7, rue Guy Mquet,
robert@vjf.cnrs.fr
lacito-CNRS,
7, rue Guy Mquet,
cath.tainecheikh@gmail.com
Martine Vanhove
llacan-CNRS
7, rue Guy Mquet,
vanhove@vjf.cnrs.fr
Department of Linguistics
University of Leuven,
Blijde-Inkomststraat 21 - PO Box 3308
B-3000 Leuven, Belgium
jean-christophe.verstraete@arts.
kuleuven.be
Editors introduction
The syntax and pragmatics of clause linkage
and clause hierarchy: Some new perspectives
Isabelle Bril
LACITO (Laboratoire des Langues et Civilisations Tradition orale)
1. Presentation
This volume is the outcome of a research programme (20032007) conducted by
linguists specializing in a wide array of language families, from varied theoretical
backgrounds. We thankfully acknowledge the financial support of the Fdration de
Typologie et Universaux Linguistiques of the CNRS (Centre National de la Recherche
Scientifique).1 The research project was coordinated by the editor of the present volume.
This collection of articles explores clause-linkage strategies in a cross-linguistic
perspective. It concentrates on issues generally relating to coordination and subordination, with a greater emphasis on subordination, marked by a variety of constructions
such as clause-chaining, converbs, masdars, correlative constructions, specific types of
conjugations or verbal inflectional morphology, T.A.M. markers, as well as informational hierarchy and referential hierarchy strategies.
The choice of topics addressed was guided by their being comparatively less studied in the existing literature. This volume provides further documentation on such
morphosyntactic phenomena from slightly different angles and perspectives; in particular, it explores the interaction between syntax, pragmatics and semantics in the
architecture of complex sentences. These new data are analysed in the light of current
debates relating to the typology of coordination and subordination.
2. Previous studies
Only over the last two decades, clause-linkage or clause-dependency and its related
syntactic categories, coordination, subordination, and co-subordination (Olson 1981;
. The editor is grateful to Jean-Michel Roynard for his help in editing the volume, and
to Margaret Dunham who translated various articles and corrected the final version of the
volume. Their invaluable help is much appreciated.
Isabelle Bril
Foley & Van Valin 1984) have given rise to a wealth of studies from various theoretical perspectives (inter alia, Dik 1997; Culicover & Jackendoff 1997; Van Valin &
LaPolla 1997; Cristofaro 2003; Bril & Rebuschi 2006; Rebuschi 2003 etc.). Some recent
publications have focused on specific construction types cross-linguistically, such
as clause chaining (Longacre 1985), converbial constructions (Knig & Haspelmath
(eds) 1995), adverbial constructions (van der Auwera (ed.) 1998), coordination
(Sag et al. 1985; Munn 1993; Johannessen 1998; Progovac 1998; Haspelmath 2004;
Godard & Abeill 2005). Typological studies have also focused on clause-linking in
its various aspects (inter alia, Lehman 1988; Haiman 1988; Comrie 1989 and various contributors in Shopen (ed.) (1985, 2007): Keenan on relative clauses (1985),
Noonan on complementation (1985, 2007), Haspelmath on coordination (2007),
Thompson, Longacre & Hwang on adverbial clauses (2007), etc.
3. Aims
This collection of studies aims to bring new insights to a domain which has a long research
tradition. Each of the eighteen chapters presents an in-depth study of clause-linkage and
clause-relationships, in often lesser known and lesser documented languages.
The case-studies are based on first-hand data collected by the authors. A sample
of 23 languages (and a survey of 17 others), from 12 different language families, are
analysed (see Appendix 1). Though far from exhaustive, this sample enlarges the scope
of previously available research.
Among the questions addressed are the following:
What types of clause-linking structures, and what levels and degrees of hierarchy
are distinguished in a given language?
What is the range of morphosyntactic devices used for clause-linking and more
specifically for subordination? For instance finite verb forms, masdars, converbs,
T.A.M markers, specific conjugations, case-marking systems, demonstratives and
referential devices, informational hierarchy devices, etc.
What categorical and functional domains do these morphosyntactic devices originate from?
Some more general theoretical and methodological questions are also addressed:
Are coordination, subordination and co-subordination universal syntactic categories?

Are there any clause-linking hierarchy universals?
How should these notions be defined so as to have some cross-linguistic validity?
What set of criteria could help define them?
Are there mismatches between form, function and meaning of clause-linking

devices?
Are there areal clause-linking phenomena?
Which functional and semantic types of clause-linking tend to be grouped or
distinguished? Are they comparable?
The notion of language universals has been the centre of recent debates; some doubts
have been expressed as to the possibility or even the relevance of presupposing universal constructions and categories, or any universal conceptual structures or formal categories (Dryer 1998; Croft 2001, 2003; Haspelmath 2007; Frajzyngier & Shay
2003; Evans & Levinson 2009). Despite such scepticism (see Newmeyer 2007 for a less
sceptical approach and some counter-arguments), cross-linguistic typological studies
and in-depth case-studies of (lesser known) languages contribute to (i) comparing
and refining the understanding of syntactic constructions or categories, (ii) assessing
their variability, and (iii) distinguishing language-specific and areal features or constructions from more invariant ones. This in turn leads to revising definitions and
to proposing more refined sets of criteria. This approach is the main guideline of the
volumes contributions.
4. S
ome properties of coordination and subordination
and some distinctive tests
Coordination is generally distinguished from subordination by a number of tests
and properties with variable cross-linguistic validity. Among the subordination tests,
to which coordination reacts negatively, are the following (summarised in Yuasa &
Sadock 2002; and Haspelmath 2004 for instance):
i.
Permutability of the clauses without any logico-semantic change (i.e. only additive
coordination allows it, other coordination types do not);
ii. embeddedness;
iii. possible pronominal cataphora (coreferential with a NP in the following clause);
iv. possible extraction (Ross 1985 [1967]).
Among common features distinguishing main clauses from subordinate clauses are:
i.
Imperviousness to the illocutionary force of the matrix clause and disjunct illocutionary scope (see Foley & Bickel this volume); this is in contrast with conjunct
illocutionary scope found in some clause-chaining or converbial constructions, in
which the dependent clause falls under the scope of the illocutionary operators in
the main clause;
Isabelle Bril
ii. T.A.M dependency of the dependent clause on the main clause (found for instance
in the medial verb forms in clause-chaining), while unconstrained tense marking
can be found both in coordinate or subordinate constructions;
iii. unequal assertive clauses status (with possible non-assertive status for some subordinate clauses, through various morphosyntactic and pragmatic devices);
iv. deranked, unasserted clauses, possibly displaying non-finite verb forms (vs.
co-ranking coordinate clauses), nominalised clauses, participial forms, casemarkers and adpositions;
v. possible restrictive focalisation of subordinate clauses (with one restriction: a
term may not be focused within the dependent clause);
vi. Use of topic markers as indicators of subordinate clauses (Papuan, Oceanic
languages);
vii. Use of case-markers functioning as topic markers and projecting a case functional
head above the subordinate clause (Foley, this volume).
Subordinate clauses as islands
Subordinate clauses are impervious to the illocutionary force of their main clauses.
They are islands, their features cannot percolate up to the level of the main clause,
nor can the main clauses Inflectional features (tense, mood or polarity) move down
into them (see Foley, this volume). Besides, they usually are presupposed statements,
which accounts for some other features investigated in this volume, which are related
to the pragmatic structuring and informational structure of complex clauses, and also
involve constructions based on the contrast between presupposition vs. assertion.
Distinctive features among subordinate clauses: nexus and juncture layer
Other distinctive features among subordinate clauses involve the layer at which the
subordinate clause operates (as developed in Foley & Van Valin 1984; Van Valin 2005,
Bickel this volume). They may operate (i) at the predicate-verb layer (as ad-V clauses),
(ii) at the clause layer (ad-Clause), (iii) at the whole sentence layer (ad-Sentence) as
detached, topic subordinate sentences for instance, or (iv) they may operate beyond
the sentence, at utterance level. The layer at which they operate then determines their
specific syntactic functions (argument function in complement clauses, modification,
adjunction), or specific discourse functions when subordinate clauses operate beyond
the sentence (see the contributions by Tersis, Robert this volume).
Properties and structure of coordination
In contrast with subordination, coordination is usually considered to be a logically
structurally symmetric relation, in that if x is coordinated with y, then y is coordinated with x. Although coordination may contain some logical and formal symmetry,
at least in some of its instances or at some abstract level, this does not mean that it is
syntactically or semantically unconstrained (Progovac 1998): for instance, some coordinate constructions are subject to causal relations and readings, and are thus sensitive
to order (as in I was angry and he left the house vs. he left the house and I was angry).
Conjuncts have also been considered as having co-ranking status, but a co-ranking
analysis of coordination has been challenged by numerous cases of morphosyntactic
asymmetries between conjuncts (Johannessen 1998), not to mention the many cases of
form-function-semantic mismatches (Yuasa & Sadock 2002). In Johannessens analysis
of coordination, the conjuncts are in a hierarchic specifier-complement configuration;
the first conjunct (in VO languages) stands structurally apart, while the conjunction
heading the other conjunct(s) (i.e. the complement) forms a structural unit (Johannessen 1998; Progovac 1998). The conjunctive head may be transparent in allowing
symmetric marking of the non-initial conjuncts, or it may assign different morphosyntactic features to the complement conjuncts; these may be different finite properties on VPs; or they may be different case-marking on NPs, either default cases or cases
selected by the conjunctive head (as with comitative coordinators) (Sag 2005; Bril &
Rebuschi 2007:1012). Asymmetric features resulting from the properties of the coordinator itself provide evidence of some hierarchical structure in the coordinate phrase.
Cross-linguistic studies thus show that conjuncts with symmetric properties are just
one possible option of coordinate constructions.
5. Outlook of content
The volume is subdivided into four parts devoted to more specific topics relating to
clause-linkage; however, the various contributions within each part interact more than
the subdivisions suggest.
Part I presents some theoretical reassessment of terminologies from syntactic and

typological perspectives (Foley), as well as the quest for typological methods based on
statistical methods and on sets of variables allowing comparability (Bickel). These are
illustrated by case-studies in various languages, mostly Papuan and Tibeto-Burman.
Part II deals with issues and morphosyntactic strategies relating to the syntax and
semantics of clause-chaining, conjunctive conjugations, converbial constructions,
masdars.
Part III centers on issues relating to the interaction between syntax, pragmatics and the semantics of clause-linking strategies and subordination, mostly in
relation to (i) informational hierarchy and the contrast between presupposed vs.
asserted propositional contents (Bril, Vanhove) (ii) to referential hierarchy (based
on deictics or anaphorics) (Bril, Vanhove, Taine-Cheikh, Adamou), and (iii) to
correlative constructions (Corts).
Isabelle Bril
Part IV presents insights in the clause-linking and subordinating functions of some

T.A.M. markers and conjugation systems, which occur via informational hierarchy
(Verstraete), via situational dependency effects between clauses (Robert), or via
the backgrounding effects and lack of illocutionary force of specific aspect and
mood forms (Franois). Complex verbal inflectional categories and conjugations
are also shown to be at work in the syntax (Leroy) and discourse functions (Tersis)
of clause-linkage.
The origin and evolution of clause-linking morphemes or strategies is a topic of

investigation in many contributions (see Akhvakh, Greek, Coptic Egyptian, Oceanic
languages, Yafi Arabic, Zenaga, Pomak). (Poly)grammaticalisation of adpositions,
demonstratives, verbs, etc. is a frequent process which gives rise, among other things,
to clause-hierarchy and subordinating morphemes.
Part I. A reassessment of terminology and typological methods

A. A reassessment of the theory of nexus
W. Foley (Chapter 2) presents a revision of the theory of nexus (first developed in
Foley & Van Valin 1984), which distinguished three categories of nexus, subordination,
coordination and cosubordination, and which is now reduced to only subordination
and coordination.
Clause-chaining and cosubordination
Clauses in a cosubordinate nexus were defined as being in a dependency relationship
for a specific inflectional category or operator such as tense, mood or illocutionary
force, a dependency which did not occur in coordinate nexus. But some analyses of
clause chaining (in Papuan languages) and converbial constructions (of central and
south Asian languages, Haspelmath & Knig 1995), which were formerly identified
as prototypical cases of cosubordinate nexus, show that illocutionary force, the highest peripheral operator or I feature, need not be shared across the clauses; hence they
cannot exemplify cosubordinate nexus. Foley thus proposes a revised theory of nexus
based on functional categories like I (inflectional features) and their projection IP
(Inflection Phrase) and restricted to two nexus, subordination and coordination. The
cosubordinate nexus is re-analysed as a type of coordinate nexus that differs from normal clausal coordination in the type of constituents coordinated. Clause chaining is
thus distinguished from standard coordinated clauses by the nature of the coordinated
constituents, S versus IP. In clause chains, S constituents are coordinated under a single
IP node; the verbs in the coordinated S constituents are non-finite, they have no intrinsic I feature specifications, tense is only indicated on the final clauses verb and it has
scope over the preceding medial or dependent clauses. In Foleys view, the inflectional
I categories of the final clauses verb belong to the structure as a whole, rather than to
the final clause only.
Assessment of I-features: Polarity, illocutionary force, mood, tense

Foley also points out cross-linguistic differences in the treatment of polarity as either
belonging or not to I features. Thus, in some Papuan languages like Usan, negation in
the final clause is an I feature with scope over all coordinated S constituents under it,
while in Tauya, the scope of negation only spreads to clauses sharing the same subject.
On the other hand, illocutionary force is an I feature in Tauya with scope over the
whole series of coordinated constituents. The conclusion is that polarity is a lower
level I category, while illocutionary force remains a feature of the highest IP projection.
Other Papuan languages show variation in other I features like mood and tense.
In clause chaining constructions, many Papuan languages inflect the verbs in medial
clauses for realis vs. irrealis mood, while the verb of the final clause bears the full
inflectional possibilities of tense and illocutionary force. Foley thus points out that
it cannot be assumed that clause chaining always corresponds to the same types of
structures across languages; minimal S constituents may be coordinated using this
structure, but so can fully specified, essentially independent IP constituents. As these
features do not have the same status cross-linguistically, ranking them on a gradient
proves a useful undertaking.
B. Statistical methods for typological comparability: A multi-variate approach

Some of the syntactic criteria (listed in paragraph 4 above) stand more as statistical rather
than as absolute features. This is the main thrust of Bickels contribution, also pointed
out in various contributions where form-function mismatches occur, and where clearly
assigning a given construction to either the coordinate or the subordinate type often
proves uneasy: in Papuan clause-chaining (Foley), some converbal constructions in
Akhvakh (Creissels), Budugh (Authier), Badaga (Pilot-Raichoor) and Coptic (Reintges).
B. Bickel (Chapter3) thus raises the question of cross-linguistic comparability of structures which are at best similar, but never identical. Mismatches are due to the fact that
language-specific structures are analysed with terminologies whose definition is based
on other languages, or are theory-dependent. Thus, due to the amount of structural
variation in feature functions, and to the difficulty in establishing robust comparative
notions and criteria proving impervious to language specific variation, some other
method is needed.
Bickel advocates the use of statistical methods standard in other disciplines for
the analysis of diversity, and the recourse to a multi-variate approach based on sets of
variables (or parameters) in order to capture variation and probabilistic assessment of
clusters and to establish typological prototypes. Taking the pattern of adjoined (nonembedded) clause linkage across languages as a sample case, he decomposes features
of various structures of adjoined clause linkage into sets of variables (parameters) that
allow precise measurement of cross-linguistic similarities and differences, as well as
the discovery of typological patterns based on statistical techniques.
Isabelle Bril
This study, based on two dozen languages, isolates a set of twelve analytical
variables (ranging from the scope of illocutionary force operators to extraction constraints) that are applied to a representative selection of clause-linkage structures. The
analysis supports Foleys view that cosubordination is not a distinct prototype, while
there is a cross-linguistic prototype of subordination characterised by disjunct illocutionary scope, local tense scope, flexible positions, and with less probability a ban on
question formation or focusing inside the dependent clause. Furthermore, there is a
cross-linguistic cline between more vs. less tightly constrained types of clause adjoining, specifically between three types of coordination-like structures varying according
to the extent to which tense marking and tense scope is constrained. Finally, while
a tentative prototype of subordination seems to emerge from this pilot database, no
coordination prototype does.
Part II. Converbs, masdars, clause-chaining, conjunctive conjugations

Converbs, masdars, clause-chaining constructions and conjunctive conjugations are
the focus of Part II.
A. Converbs
Three papers are concerned more or less centrally with converbial strategies in Daghestanian (Creissels, Authier), Dravidian languages (Pilot-Raichoor) and Coptic (Reintges).
They generally follow Nedjalkovs (1995) distinction between specialised and generalpolyfunctional converbs.
D. Creissels (Chapter4) discusses the rich converbial morphology of Northern Akhvakh,
which he deems to be an ancient feature among Nakh-Daghestanian languages. The
term converb is used after Nedjalkovs definition (1995), to refer to non-autonomous
verb forms, different from infinitives, masdars/verbal nouns or participles, in that that
they do not occur in complement clauses or in relative clauses. If specialised converbs may be defined as essentially marking adverbial subordination, the multipurpose general converb occurs in constructions analyzable as clause coordination, but
it has two other main functions in clause-linkage: (i) it may specify the manner of an
event encoded by an independent verb form, (ii) it may encode an event viewed as the
first stage of a complex event whose second stage is encoded by an independent verb
form. Some of its syntactic properties are thus interpretable as coordination, while
others are interpretable as cases of subordination. Thus, in many constructions using
the general converb, the type of coreference found between full NPs and pronouns
is commonly considered incompatible with coordination; other facts such as linear
order, embedding, relativization, and negation are also incompatible with coordination. Another particularly strong argument in favour of subordination is that, in some
complex constructions, the general converb in Akhvakh may show external suffixal
agreement, i.e. controlled by an NP outside the clause headed by the dependent verb
form and controlled by the main verbs S/P agreement.
Converbal clauses usually precede the main verb, but the reverse order is also possible though not equally usual for all converbs. The order main verbconverb is more
common for the Purposive converb than for the Conditional converb (the decisive
factor seems to be discursive); it is excluded for the Immediate converb. An interesting
use of the Conditional converb, apart from its use in condition clauses or in clauses
with concessive meaning, is its occurrence in insubordinate constructions (Evans
2007) with optative meaning.
B. Converbs, masdars and prosody

G. Authier (Chapter5) studies the specific case of bare verb-stems in Budugh (Daghestanian). Bare verb-stems are uninflected for tense or mood, and can be used (i) either
as subordinate forms, as participles, as verbal nouns/masdars, as adverbial sequential
converbs, or (ii) as independent, finite non-indicative verb forms with modal value
(especially injunctive). Their uses are distinguished by stress: (i) they carry falling
stress on the first syllable when independent and finite, (ii) they carry rising stress on
the last syllable (like nouns) when dependent-subordinate. Falling intonation is thus
a marker of syntactic finiteness and pragmatic completeness. Budugh also displays
a clause-chaining device using sequential converbs characterised by an initial rising
stress and which falls in between these two main types, as well as Masdarswhich occur
both in independent (sentence-final) and embedded (subordinate) syntactic positions.
Masdars can acquire the initial falling stress typical of finite verbs; they are also used
as insubordinate forms in independent but non-finite clauses with non-assertive or
non-indicative modality, or in deliberative clauses (often extraposed and afterthought
like). When the extraposed non-finite form is stranded and acquires syntactic autonomy,
this triggers a stress shift. A new, insubordinate form, segmentally homonymous to
the Masdar, is then used as a finite debitive mood.
C. Converbs and other subordinating strategies
C. Pilot-Raichoor (Chapter 6) analyses various subordinating strategies in Badaga
(Dravidian): converbs, nominalised verb forms and a quotative strategy. The main
findings are that (i) none of the clause dependency strategies are specific to a single
semantic function, (ii) subordinate clauses are fully autonomous in terms of their
argumental and tense-aspect settings (with the exception of some clauses headed by
a Contextual converb), and are impervious to the inflectional features of the matrix
clause (see Foley this vol.).
Again, converbs subdivide into a polyfunctional converb, used in constructions
ranging from clause-chaining to modifying functions, and other semantically specialised
converbs expressing adverbial meanings. Both relativization and nominalization are
Isabelle Bril
based on an adjectival participle form, used with a nominal or an adverbial head in

one case, and derived into a nominalised form of the verb in the other case. Case
markers and postpositions specify the semantic relation of the subordinate clauses.
Word order, prosodic and pragmatic factors also contribute to the interpretation of
the subordinate clauses.
Apart from clause-chaining, the polyfunctional Contextual converb (Pc) may have
contextual modifying functions with implied adverbial meaning (cause, purpose,
manner, concession, etc.); clauses headed by the Pc converb form are non-finite and
syntactically anchored to the next verb/predicate. This is a frequent cross-linguistic
pattern for sequential constructions (see Wolof this volume). On the other hand,
Specialised converbs are semantically constrained; they head adverbial clauses and
express circumstances (relative time), polarity or modality. While the Pc form contains no overt marker and no explicit semantic information on clause relations, Specialised converbs do. This is reminiscent of the central distinction made by Verstraete
(this volume) between inferential and compositional coding devices.
D. Clause chaining and conjunctive conjugations

C. Reintges (Chapter7) analyses conjunctive conjugations in Coptic Egyptian. While
the most frequent pattern for clause coordination is symmetric with awf and, there
are also asymmetric clausal conjunctive patterns, using the Conjunctive and the Inferential verb conjugations (interacting with tense, aspect and mood), and the Converbal
Relative tenses which occur in main and in embedded clauses.
The Conjunctive conjugation is primarily an asymmetric clause chaining device,
ambiguous between coordination and subordination: the initial clause contains the
controlling verb with tense/aspect specifications and illocutionary force, while the
less specified conjunctive verb form occurs in the successive chain-medial and final
clauses (see Foley, Bickel, Robert, this volume for similar facts). But the functions of
the Conjunctive conjugation may extend to various types of subordinate clauses with
a wide range of semantic relations. The Inferential conjugation, has modal(-evidential)
semantics and expresses purpose, consequence, and evidentiality.
Clause chains with Conjunctive verbs display the syntactic behaviour and semantics
of standard coordination. In other cases, a coordinative interpretation is excluded by
selectional restrictions and/or by the tense/aspect specification of the controlling verb.
Some Conjunctive clauses thus display properties typical of subordinate clauses, such
as embedded complements of some verb types (manipulative, knowledge, volition,
intent), as oblique clausal modifiers, or in the apodosis of condition clauses. Connectives and subordinating conjunctions may occur to disambiguate the semantics of
Conjunctive clauses. In some restricted contexts, Conjunctive verbs may also occur in
chain-initial position, as insubordinate forms, possibly resulting from the deletion of
a volition or intent verb. Though the more common complementation strategy makes
use of finite subordinate clauses and infinitives, the Conjunctive conjugation may be a
marked alternative for complementation, restricted to some verb types and excluding
perception and discovery verbs which almost exclusively select the converbal relative
tenses (see Dixon 2006 for the role of clause chaining in complementation).
The syntactic pattern of different-subject vs. same-subject conjunctive clauses also
plays a role, compositionally (see Verstraete this volume on this notion); differentsubject conjunctive clauses have an illocutionary force different from that of the initial
conjunct, the different subject is contrastively focused and the spreading of the illocutionary force from the chain-initial clause to the conjunctive clause is blocked; thus the
construction is no longer interpreted as coordinate, but as subordinate with adverbial
purpose or reason meaning.
The Converbal Relative tenses are absolute finite verb forms which contain a tense/
aspect particle, with person agreement manifested in the coreferential pronoun. They are
primarily subordinate forms, although they also occur in asymmetrical clause coordination and complex predicates. They occur (i) in restrictive relative clauses, (ii) in predicative adjuncts (modifying the main verb), (iii) in temporal adverbial clauses (expressing
simultaneity, precedence and subsequence, relative to speech time); (iv) in information
packaging constructions (constituent questions, declarative focus sentences).
The last type, the Inferential conjugation (or inferred evidential) is yet another
asymmetrically coordinating verb conjugation, expressing consequence, goal, and
encoding inference based on (non-)observable facts (see Verstraete this volume for
the notion of encoded inference). These conjugations, especially the Converbal Relative tenses, also play a role in information structure and focus marking, which sets
them apart from pragmatically neutral declarative clauses.
Part III. The syntax, pragmatic, semantic interface

The role of informational hierarchy (topic or focus) strategies and markers in the construction of clause-hierarchy, has received somewhat less attention cross-linguistically
(since Haimans 1978 seminal article, see also Lakoff 1984). As have the origins of informational hierarchy markers: coordinators (Bril), demonstratives (Frajzyngier, Bril, Vanhove, Taine-Cheikh), verbs (Vanhove), or other morphosyntactic domains recruited to
express informational hierarchy, such as absolute constructions, case-markers (Foley,
Bickel, Bril), specific conjugation paradigms (Reintges, Robert, Leroy), tense-aspect
morphemes (aorist, perfect or imperfective aspect, see Robert, Taine-Cheikh, Franois, Valma), or mood morphemes (Bril, Verstraete, Franois, Leroy, Tersis).
The contributions in Part III investigate the interaction of pragmatics, semantics and discourse with the syntax of clause-linkage: one main aspect is the role of
informational hierarchy and referential hierarchy strategies. It is the central topic of
contributions by Bril, Frajzyngier, Vanhove, and is also studied in some contributions
Isabelle Bril
in Part IV (Verstraete, Robert, Franois). Clause-linking and subordinating strategies

based on demonstratives, referential hierarchy and correlative constructions are the
main focus of Adamous and Cortss contributions.
A. Informational hierarchy, referential hierarchy and clause-linkage

I. Bril (Chapter8) focuses on the functions of informational and referential hierarchy
strategies in the architecture of complex clauses in mainly Austronesian languages.
Informational hierarchy and its markers (topic and focus morphemes), structure clauses
as subordinate via the central contrast between presupposition vs. assertion. Topics
are presupposed frames for some other assertion, while foci are asserted restrictors
(Krifka 2007), restricting some presupposed propositional content to a specific asserted
variable. On the other hand, referential hierarchy and its markers (endophoric demonstratives, deictic and definite markers) are another strategy for marking clauses as subordinate, via the contrast between already backgrounded/referential clause vs. asserted
clauses. Different deictic grades play distinct functions in clause-linking, as in Takia
where the medial deictic (dx2) is used for clause sequencing, while the proximal deictic
(dx1) appears in consecutive and deductive clause sequences. The languages surveyed
tend to use these two distinct strategies (referential and informational hierarchy) as fullfledged subordination strategies, and to recategorise topic or focus markers into subordinators. Such recategorisation generally combines with clause order changes (as in
Manam or Korafe), variations in scope and syntactic domain, as well as prosodic changes.
Paths of evolution leading from coordinators to topic markers and subordinating
devices, or from endophoric demonstratives to topic or focus markers and conjunctions are more specifically discussed.
Among other subordinating strategies are the backgrounding effects of reduplicated verbs, the use of mood or aspect markers (as in Takia), or the use of absolute
constructions creating adsentential subordination and fulfilling similar clause topic
functions. In Roviana, for instance, adverbial subordinate clauses display neutral case
marking and occur in sentence-initial position as adsentential foci/topic clauses; similar strategies using ergative absolute case-marking for adsentential adverbial topic
clauses also occur in Papuan languages (Foley, this volume) and Kiranti languages
(Bickel 1999)). It is thus argued that informational hierarchy and referential hierarchy
strategies and their markers are inherent to the syntactic architecture of the complex
clause between main and subordinate clauses, rather than being a peripheral level
added to the syntactic level.
Z. Frajzyngier (Chapter9) investigates the grammaticalisation of the comment clause
in Wandala (Central Chadic, Afro-Asiatic). Comment clauses are marked by w and
always follow the element on which they are a comment (a noun phrase, an adverbial
phrase or a clause); but the form w itself belongs to the preceding clause or phrase.
Comment clauses include for instance comments on topicalised noun phrases, clausal
complements of a noun phrase, complement clauses of verbs of saying in epistemic
and deontic modality, as well as temporal and conditional apodoses, comments on a
reason clause, and even afterthought clauses. In other languages lacking a dedicated
grammaticalised morpheme to serve this function, distinct tense and aspectual systems may be used as comment clause markers. Thus, the comment clause confirms the
assumption that speakers operate with different motivations, on the coding of various
functional domains.
M. Vanhove (Chapter 10) focuses on the functions of two polyfunctional particles in
Yafi Arabic (Yemen), ra, and ta, whose origins are respectively a verb see, look, and a
demonstrative. Both are used as deictics, topic markers, focusing particles, and clause
coordinating and subordinating devices. Even though other subordinating constructions
and markers are also available, topicalisation and focusing strategies have become the
preferred clause-linking strategies, especially in causal, relative and complement clauses.
Depending on its syntactic scope, ra developed (i) a deictic function as a presentative, (ii) an assertive function as a copula, (iii) an aspectual function as a perfectresultative verb clitic, and has also become (iv) an informational hierarchy marker
as a contrastive NP focus particle, and a contrastive topic marker. When this focus
particle has scope over a clause, it carries explicative meaning which led to its reanalysis as (v)a causal/explanatory clause subordinator. The particle ta is also used as (i) a
presentative, (ii) a copula, (iii) a contrastive NP focus particle, (iv) a subordinator in
relative, complement clauses, and (v) causal clauses. Both markers thus illustrate the
reanalysis of deictic items (though of a different kind) as discourse and clause-linking
particles, but ta has a larger range of subordinate functions. As a focus marker, ra
marks the hierarchy between a presupposition and an assertion. When used in clauselinkage, the clause focused by ra is interpreted as the cause or explanation for the
other event. Following Verstraete (this volume), Vanhove points out that the function
of ra as a sentence-focus particle is based on a mechanism of encoded inference;
since the presupposed element is not retrievable in the discourse context, it forces the
inference of an explanatory relation with the preceding clause.
C. Taine-Cheikh (Chapter11) studies the functions of ad in Zenaga (Berber, Mauritania),
whose deictic origin explains most of its uses (as presentative, copula, relative pronoun,
injunctive particle) (see Vanhove for similar facts), and whose functions are found in
most Berber languages. However, ad in Zenaga shows some features which diverge from
Berber when used as a conditional and a quotative particle. These divergent evolutions
of ad are analysed in their interaction with T.A.M. markers (the Aorist in particular),
with the structure of simple and complex clauses, and with discourse constraints in
topic position. Ad thus has referential, pragmatic (as a focus marker) and syntactic
Isabelle Bril
functions in clause-linkage. Demonstratives may have deictic, endophoric reference,

as well as reference to imaginary worlds, which accounts for the polygrammaticalisation of ad as a complementiser, quotative marker, consecutive or goal subordinator,
condition marker, optative or injunctive marker.
The other central discussion bears on the different functions of the Aorist combined with ad vs. its functions without ad. As in Berber, Wolof and some Oceanic languages (see Robert & Franois, this volume.), the Aorist is the neutral form in the
T.A.M system; it is not anchored in speech time and expresses habitual or potential
events. In complex clauses, it depends on another clause or verb with tense-aspect
specifications to provide its situational anchor. Thus, alongside sequential clause
chains, or sequences of events running counter to the expected state of affairs (equivalent to but), the Aorist also occurs in subordinate clauses expressing the purpose or
consequence of the first event. It is also used in main clauses following generic temporal subordinate clauses, conditional protases, or relative clauses with generic, usual, or
potential meanings. Unless preceded by ad which confers some situational reference to
the Aorist, a verb in the Aorist is barred from fronted or sentence initial position. Thus
ad has an anchoring function, which confers referential autonomy upon the Aorist
and enables it to stand in initial position in prohibitions and orders, or in conditional
clauses or in the fronted protasis of conditional clauses where the fronted ad + p clause
is the frame for the q clause Governing verbs expressing orders, requests or wishes are
also regularly marked by ad +Aorist.
B. Deictics as conjunctive and correlative markers

Demonstratives are a well-known source of conjunctive markers and strategies (Diessel
1999), as is illustrated in various contributions (Bril, Vanhove, Taine-Cheikh). Some
further aspects are investigated in Pomak (Adamou), as well as in German, in relation
to correlative constructions (Corts).
E. Adamou (Chapter12) focuses on three deictic suffixes in Pomak (Greece) used to
form temporal subordinate conjunctions. The deictic suffixes locate the event in relation to the speech time and situation (proximal, medial and distal), and their choice
depends on the event type encoded in the clause and its reference (present, past, future
and habitual). They stand in contrast with the free temporal subordinator lacking any
deictic suffix which indicates that the event is not anchored in the discourse situation.
The medial and distal deictic suffixes also have temporal reference relative to some
past event different from the time and discourse situation.
C. Corts (Chapter 13) addresses the status and the -phoric properties of correlative
markers in modern German subordinate clauses. These correlative markers originate from
demonstratives and determiners, and retain some of their etymological determinative and
phoric properties in their demarcative or conjunctive functions. Their semantic and
pragmatic functions vary with the type of subordinate clause and complex sentence in
which they occur; their meaning and function result from the complex interaction and
interdependent relations of the binding determinative markers and the assertive markers
which integrate the sentence into a pragmatic and textual whole. The complex sentence
is thus not a mere concatenation, nor an addition of two simple sentences; it results from
intricate constructions which must be considered from a holistic viewpoint.
Part IV. T.A.M. strategies and informational hierarchy

The fourth part of this volume focuses on clause-linkage marked by T.A.M markers,
inflectional verbal categories and specific conjugations. T.A.M morphemes interact
with clause-linkage and subordination via illocutionary force and assertion. Some
moods and tense-aspects (such as the aorist) are devoid of any assertive or illocutionary force, which results either in generic or gnomic propositional content, or in
unanchored propositions which then stand in need of some other clause, sentence or
text to be anchored by some situational locator (Robert), some IP in the structure of
complex clauses (Foley), and from which it receives its specifications in terms of illocutionary force, assertive force, spatio-temporal anchoring, and thus its interpretative
features in relation to the syntactic structure, the informational hierarchy structure,
the argumentative pattern and the semantics of clause relations.
The types of clause structure considered in part IV are basically related to systems
of instructional procedure, as the choice of a given form then sets off an interpretative
procedure as to the syntactic and semantic clause relation. Verstraete develops a threeway mechanism of instructional procedure (compositional, inferential and encoded
inferential) that accounts for the various types of clause-linkage.
A. Mood, informational hierarchy and clause-linkage

J.-C. Verstraete (Chapter14) examines the system of mood and focus in clause-linkage
in Umpithamu (Paman language, Australia). Umpithamu is relatively poor in markers
specifically encoding clause linkage, but uses markers from the domains of mood and
information structure. Such markers contribute to clause linkage in three distinct ways.
The first is through a mechanism of compositional encoding, as with the purposive relation which is encoded by the combination of a verb marked for potential
mood with a general syntactic schema of argument sharing (same subjects) without
any cross-referencing pronoun (thus dependent on the main clause). In compositional
encoding strategies, the mood marker provides semantic specification in a more general syntactic schema of clause linkage, and encodes the interclausal relation jointly
and compositionally with this syntactic schema.
The second mechanism is inference: the marker merely provides a basic semantic
prerequisite for the encoding of interclausal relations, which must be enriched by
Isabelle Bril
inferential strategies based on world and discourse knowledge, as well as prosody. This
occurs with conditional clauses where the verbs in both clauses are marked for potential mood with their own cross-referencing pronouns, and the causal relation between
clauses is left to inference.
The third mechanism is encoded inference, as in explanatory relations where the
use of a focus marker invokes a presupposition which, when not found in the sentence,
forces the search for some link in the discourse context and forces the inference of an
explanatory relation with the preceding clause.
The central distinction is thus between clause-linking structures with encoded
vs. inferred interclausal relation. This correlates with the use of specialised markers of
interclausal relations vs. non-specialised markers like potential mood and information
structure markers, which, though belonging to other distributional domains, may contribute to interclausal relations (purposive or conditional for instance).
Prominence being by definition a relational concept, its potential relevance for
clause linkage is obvious. Thus, the focus marker (an ergative marker on transitive subject NPs) has some procedural and instructional function. Marking an argument as
prominent in a particular clause invokes a link with something beyond this clause, like a
presupposition for focal prominence. It thus instigates the search for some link between
the presupposition (in relation to which the ergative NP fills out a variable) and the
preceding clause(s), and leads to inferring some explanatory relation between them.
The marker itself does not encode any explanatory relation, instead it encodes some
inferential procedure, and is thus encoded inference. It illustrates a third way in which
non-specialised markers can contribute to clause linkage and how a focus marker may
function as a mechanism for clause linkage rather than discourse linkage, also relying
on inference.
B. Conjugations, informational hierarchy and clause-linkage

S. Robert (Chapter15) discusses the complex inflectional verbal system (conjugations)
involving tense or aspect forms in Wolof (Niger-Congo, Senegal), in relation to clauselinkage, subordination and informational hierarchy. The various conjugations, with
restrictions on their combination, express predictable semantic effects: succession,
contrast, causality, consecution, explanation or strengthened assertion. The system
generally dispenses with coordinating or subordinating morphemes.
One of these conjugations is the Aorist (or Null Tense), which occurs in narrative
clause chains depending on some previous sentence or context for its temporal specification (thus creating situational anaphora); it also occurs in generic statements,
in wh-questions, injunctions and hypothetic clauses. Clauses with an Aorist are not
anchored in speech-time and lack temporal and modal specification. The Aorist is
fundamentally a dependent mode and a marker of subordination (for complement
clauses, consecutive or purpose clauses), without any other subordinating morpheme.

As a dependent mode, it is in need of some locator (another clause or a main clause)
whose nature is the variable determining the various degrees of dependence displayed
by the Aorist clauses, ranging from discourse coherence to embedding. Its semantic
interpretation also depends on the temporal and epistemic status of the locator clause:
if perfective, the Aorist clause has consecutive meaning; if the locator is irrealis, the
Aorist clause is purposive. If the locator is the verb of a previous clause, the Aorist
clause functions as a complement clause.
Narrative clause chains with the same conjugation may occur for all conjugations, though they are much less frequent than the succession of Aorist clauses.
Chains of Aorists express successive events in narratives, while clause chaining with
two Perfects expresses successive and resulting events anchored in speech-time and
discourse. Clause chains express (1) temporal succession (for Perfect and Aorist only)
or (2) cumulative or contrasting assertion (all other cases), depending on whether the
argumentation of the clauses is convergent or divergent.
The Presentative conjugation expresses simultaneity or immediate sequence
between the event and the speech act; it refers to some unexpected process with detrimental and modal meaning (surprise, warning). When used in a protasis, combined
with an Aorist in the apodosis P2, it refers to some unexpected event and expresses
discordance (and yet). An Aorist with the Imperfective suffix in an apodosis tends to
indicate temporal concomitance with the protasis.
The Verb Focus conjugation is another type of clause chaining whose semantics
depends on position: a Verb Focus in the protasis (P1) and an Aorist in (P2) expresses
causal relation, with (P1) as the cause; while a Verb Focus in the apodosis (P2) is the
explanation of P1. It may also have contrastive or corrective meaning. In clause chaining, the Verb Focus marks dependence based on the usual contrast between a presupposition and some additional and asserted information. As in Umpithamu (Verstraete)
and in some Oceanic languages (Bril, Franois), informational hierarchy is thus part of
clause linking strategies.
C. Pragmatics, T.A.M dependency and subordination

A. Franois (Chapter16) studies the semantic and syntactic effects of certain T.A.M.
markers on the syntax of clause dependency, and in relation with information hierarchy, in Hiw and Lo-Toga (Oceanic, Vanuatu). In spite of their wealth of subordinators, these two languages show a great propensity to dispense with them and instead
use two T.A.M markers for clause dependency, the Subjunctive and the Background
Perfect. The Subjunctive lacks illocutionary force, the Background Perfect lacks informational focus. Such features account for their strong affinities with subordination,
backgrounding function and informational hierarchy. Each one marks the clause in
which it appears as subordinate to another main clause, without any conjunction.
Isabelle Bril
The Subjunctive clause refers to some virtual state of affairs and lacks illocutionary force. It requires an anchor point, another clause or predicative operator with some
T.A.M. specification and illocutionary force, to constitute a complete utterance. It may
also appear on its own in spontaneous subordination when marking the protasis of
a conditional sentence without any conjunctive marker (similar to should I hear this,
Id be very angry).
Similarly, clauses in the Background Perfect encode presupposition and require
some other clausal anchor with asserted informational focus to form a valid utterance.
While grounded in discourse pragmatics, these two TAM-based strategies are a routinised device for clause subordination (occurring for instance in relative clauses with
a backgrounded event). While they are both compatible with subordinators, they tend
to function as a subordinating strategy on their own.
The Aorist also occurs in sequential clause chains depending on some other clausal
anchor, and in clauses with generic, prospective, optative, imperative and conditional
semantics, or in complement clause of optative verbs. Even though the Aorist and the
Subjunctive show some functional overlap, the Subjunctive is preferred when the subordinate clause is explicitly irrealis or generic (as in Zenaga and Wolof).
D. T.A.M dependency and clause-chaining

J. Leroy (Chapter17) explores strategies of clause-chaining and tense-mood concordance
in Mankon (Grassfields Bantu, Cameroon). Four verbal constructions, the successive,
exhortative, non-future consecutive and future consecutive conjugations, their combination and clause linking functions are analysed. Although these four conjugations
are central to the syntaxof complex sentences, optional coordinating or subordinating
morphemes may appear in some types of complex sentences. The main focus is on the
perfective positive conjugation, which is the most complex and common in texts.
Various strategies are used for complex sentences, some use conjunctions combined
with specific conjugations, such as |t| in order to (+ future consecutive and exhortative), others lack conjunctions and use conjugations which occur in the non-initial
clauses of complex sentences; the third type involves chronologically ordered clause
chains which differ from the preceding types in that the clause chains obey strict tensemood concordance rules involving three specific conjugations (non)-future consecutive,
exhortative or successive. Semantic relations and hierarchy are further refined by auxiliaries
expressing sequence (then), simultaneity (also), or a time span between events.
N. Tersis (Chapter18) discusses the verb inflection system which marks subordination
in Tunumiisut (Eskaleut, Eastern Greenland). Clause subordination is characterised
by the use of verbal inflectional morphology and by a general lack of subordinating
conjunctions. Clause-chaining in narratives is also marked by verbal morphology.
The markers most commonly found in subordinate clauses are the attributive,
theconcomitant, the causative/effected, and the conditional. The attributive -ti- is a
dependent conjugation, used in relative and complement clauses (of thought, perception, and declarative verbs), and also used at paragraph level as indicating some
situational dependency on a preceding sentence. (ii) The concomitant -ttu- expresses
concomitant events in temporal, purpose or causal adjunct clauses (and in some complement clauses). (iii) The causative (or effected) marker refers to a past action relative
to the time of speech, or preceding another action, it may also express causal-explicative
relation. (iv) The conditional (or non-effected) occurs in conditional or hypothetical
clauses and also refers to an action occurring after another one, with cause-effect
relations. The causative and the conditional verb forms display structural and formal
similarity to possessive noun phrases.
In the unmarked order, adverbial clauses (marked by the concomitant, causative and conditional verbal markers) occur before the main clause; the reverse
order [main/subordinate clause] is found with complement clauses (of perception,
thought, declarative verbs) or with dependent clauses expressing a cause-effect
relation or a purpose.
Tunumiisut also has clause chains in narratives with only subordinate verbal markers, not depending on any main clause with the indicative marker, but pragmatically
linked to and dependent on some preceding utterance. In a given story, only 31% of
clauses are independent, 69% are dependent clause chains anchored in some initial
clause. Dependence thus reaches beyond the clause into the textual and discourse level.
Some complex sentences may thus display two effected/causative verb forms and
one concomitant verb form, all of them depending on a clause locator in the indicative,
found much earlier in the paragraph. The causative verb form encodes explicative or
background information up until the main clause containing the major information.
E. T.A.M markers and subordinators

E. Valma (Chapter19) analyses the functions of ma as a contrastive coordinator but
and as a subordinator meaning when, if in dialectal Greek from Bulgaria, as well as
the origins of these now homonymic markers. Synchronically, when ma functions as
a temporal and hypothetical subordinator, it reacts positively to the various syntactic
tests of subordination (clause permutation, extraction, focalisation, pronominal cataphora); while it reacts negatively to these tests when used as a contrastive coordinator.
The role of tense-aspect markers (referring to a state, a process or an event) in both
clauses further disambiguate its meanings when used as a subordinator. The aspectual
system makes use of two themes: the present (or imperfective), and the aorist (or perfective). The use of the aorist in the subordinate and the main clause triggers the temporal interpretation of the subordinate clause: ma then marks one event as anterior
to another one. When ma is followed by an aspectual form, it is open to hypothetical
interpretation and functions as a vericonditional junctor. The use of ma as a constrastive coordinator seems to result from contact with Turkish.
Isabelle Bril
References
Bickel, Balthazar. 1999. From ergativus absolutus to topic marking in Kiranti: A typological
perspective. Proceedings of the 25th Annual Meeting of the Berkeley Linguistics Society,
3849.
Bril Isabelle & Rebuschi, Georges (eds). 2006. Coordination et subordination: typologie et modlisation [Faits de Langue 28]. Paris: Ophrys.
Comrie, Bernard. 1989. Language Universals and Linguistic Typology, 2nd edn. Chicago IL: The
University of Chicago Press.
Cristofaro Sonia. 2003. Subordination [Oxford Studies in Typology and Linguistic Theory].
Oxford: OUP.
Croft, William. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspective.
Oxford: OUP.
Croft, William. 2003. Typology and Universals, 2nd edn. Cambridge: CUP.
Culicover Peter W. & Jackendoff, Ray. 1997. Semantic subordination despite syntactic coordination. Linguistic Inquiry 28(2): 195217.
Diessel, Holger. 1999. The morphosyntax of demonstratives in synchrony and diachrony. Linguistic Typology 3: 149.
Dik, Simon.C. 1997. The Theory of Functional Grammar, 2 Vols. Berlin: Mouton de Gruyter.
Dixon, R.M.W. 2006. Complement clauses and complementation strategies in typological
perspective. In R.M.W. Dixon & Alexandra Y. Aikhenvald (eds), Complementation: A crosslinguistic typology, 148 [Explorations in linguistic typology 3]. Oxford: OUP.
Dryer, Matthew. 1998. Are grammatical relations universal? In Essays on Language Function and
Language Type, Joan Bybee, John Haiman & Sandra A. Thompson (eds), 115143. Amsterdam: John Benjamins.
Evans, Nicholas R.D. 2007. Insubordination. In Nikolaeva Irina (ed.), Finiteness, 366431.
Oxford: OUP.
Evans, Nicholas R.D & Levinson, Stephen C. 2009. The myth of language universals: Language
diversity and its importance for cognitive science. Behavioral and Brain Sciences 32(5):
429492.
Foley, William.A. & Van Valin Jr., Robert D. 1984. Functional Syntax and Universal Grammar.
Cambridge: CUP.
Frajzyngier, Zygmunt & Shay, Erin 2003. Explaining Language Structure through Systems Interaction [Typological Studies in Language 55]. Amsterdam: Benjamins.
Godard, Danile & Abeill, Anne (eds). 2005. La syntaxe de la coordination [Langages 160]. Paris:
Larousse, Armand Colin.
Haiman, John. 1988. Inconsequential clauses in Hua and the typology of clauses. In Haiman,
John & Sandra A. Thompson (eds.), Clause Combining in Grammar and Discourse. 4969.
Amsterdam/Philadelphia: Benjamins.
Haspelmath, Martin. 2004. Coordinating constructions: An overview. In Coordinating
Constructions [Typological Studies in Language 58], M. Haspelmath (ed.), 339.
Amsterdam: John Benjamins.
Haspelmath, Martin. 2007. Coordination. In Language Typology and Linguistic Description, Timothy
Shopen (ed.),151. Cambridge: CUP.
Haspelmath, Martin. 2007. Pre-established categories dont exist: Consequences for language
description and typology. Linguistic Typology 11(1): 119132.
Haspelmath, Martin. & Knig, Ekkehard (eds). 1995. Converbs in Cross-linguistic Perspective:
Structure and Meaning of Adverbial Verb Forms, Adverbial Participles, Gerunds [Empirical
Approaches to Language Typology]. Berlin: Mouton de Gruyter.
Johannessen, Janne. B. 1998. Coordination. Oxford: OUP.
Krifka, Manfred. 2007. Basic notions of information structure. In Interdisciplinary Studies of
Information Structure 6, Caroline Fery & Manfred Krifka (eds). Potsdam: University of
Potsdam.
Lakoff, Robin. 1984. The pragmatics of subordination. Proceedings of the Berkeley Linguistics
Society 10: 481492.
Lehman, Christian. 1988. Towards a typology of clause linkage. In Clause Combining in Grammar
and Discourse [Typological Studies in Language 18], John Haiman & SandraA. Thompson
(eds), 181225. Amsterdam: John Benjamins.
Longacre, Robert. 1985. Sentences as combination of clauses. In Language Typology and Linguistic Description, Timothy Shopen (ed.), 235286. Cambridge: CUP.
Munn, Alan. 1993. Topics on the Syntax and Semantics of Coordination. Ph.D. dissertation,
Maryland University, College Park.
Newmeyer, Frederick J. 2007. Cross-linguistic formal categories. Linguistic Typology 11(1):
133157.
Olson, Michael. 1981. Barai Clause Junctures: Towards a Functionnal Theory of Interclausal
Relations. Ph.D. dissertation, Australian National University.
Progovac, Ljiljana. 1998. Structure for coordination (Part I & II). Glot International 3(7): 36 &
3(8): 39.
Rebuschi, Georges. 2003. Towards a theoretical treatment of left-dislocated subordinate clauses. In P. Lafitteren sortzearen mende-mugako biltzarra, i: Gramatika gaiak
[coll. iker 141] Jesus Mari Makazaga & Beat Oyharabal (eds), 395416. Bilbao:
Euskaltzaindia.
Ross John. 1985 [1967]. Constraints on Variables in Syntax. Ph.D. dissertation, MIT. (Published
as Infinite Syntax. Dordrecht: Reidel).
Sag Ivan A., Gazdar, Gerald, Wasow, Thomas & Weisler, Steven. 1985. Coordination and how to
distinguish categories. Natural Language and Linguistic Theory 3: 117171.
2005. La coordination et lidentit syntaxique des termes. In Godard, Danile & Anne
Abeill (eds.), La syntaxe de la coordination. Langages 160: 110127.
Shopen, Timothy (ed.). 1985. Language Typology and Linguistic Description. Cambridge: CUP.
Shopen, Timothy (ed.). 2007. Language Typology and Linguistic Description. Cambridge: CUP.
Thompson Sandra A., Longacre, Robert & Shin Ja Hwang. 2007. A typology of adverbial clauses.
In Language Typology and Linguistic Description, Timothy Shopen (ed.), 237300. Cambridge: CUP.
van der Auwera, Johan. 1998. Adverbial Constructions in the Languages of Europe. Berlin: Mouton de Gruyter.
Van Valin Robert.D. 1984. A typology of syntactic relations in clause linkage. Proceedings of the
Berkeley Linguistics Society 10, 542558.
2005. Exploring the Syntax-Semantics Interface. Cambridge: CUP.
Van Valin Jr., Robert.D. & LaPolla, Randy J. 1997. Syntax. Structure, Meaning and Function.
Cambridge: CUP.
Yuasa, Etsuyo & Sadock, JerryM. 2002. Pseudo-subordination: A mismatch between syntax and
semantics. Journal of Linguistics 38(1): 87111.
Isabelle Bril
Index: Languages studied

Axaxd6r6 akhvakh (Daghestanian)
Badaga (South-Dravidian)
Belhare (Sino-Tibetan)
Budugh (Daghestanian)
Bulgarian (Slavic, Indo-European)
Chantyal (Sino-Tibetan)
Coptic egyptian (Afroasiatic)
German (Germanic, Indo-European)
Greek (Attic, Indo-European)
Hiw (Austronesian, Oceanic, Vanuatu)
Lotoga(Austronesian, Oceanic, Vanuatu)
Mankon (Bantu)
Nlmwa (Austronesian, Eastern Oceanic, New Caledonia)
Nepali (Indo-European)
Newar (Dolakha, Sino-Tibetan)
Pomak (Slavic, Indo-European)
Tunumiisut (Eskimo-Aleut, Eastern Greenland Inuit)
Umpithamu (Paman, Australian)
Wandala (Central Chadic, Afro-Asiatic)
Wambule (Sino-Tibetan)
Watam (Papuan)
Yafi arabic (Yemen, Semitic, Afro-Asiatic)
Yimas (Papuan)
Zenaga (Berber from Mauritania, Afro-asiatic)
Other languages surveyed in this volume

Ajie (Eastern Oceanic, New Caledonia)
Amele (Papuan, Madang-Adelbert Range, N. Coast New Guinea)
Bargam (Papuan)
Berber (Afro-Asiatic)
Burushaski (Indic)
Cmuh (Eastern Oceanic, New Caledonia)
Fijian (Eastern Oceanic)
Fore (Papuan, Kainantu-Gorokan, interior New Guinea)
Godi (Kru)
Hua (Kainantu-Gorokan, interior New Guinea)
Iaai (Eastern Oceanic, New Caledonia)
Jawe (Eastern Oceanic, New Caledonia)
Kabyle (Berber from Algeria, Afro-asiatic)

Kadazan (Austronesian, Borneo)
Kte (Papuan, Huon, N. Coast New Guinea)
Kaulong (Western Oceanic, New Britain)
Kewa (Papuan)
Kiranti (Tibeto-Burman)
Korafe (Papuan, Binanderean, S. New Guinea)
Manam (Western Oceanic, PNG)
Mianmin (Papuan)
Nemi (Eastern Oceanic, New Caledonia)
Newari (Tibeto-Burman)
Nyelyu (Eastern Oceanic, New Caledonia)
Roviana (Oceanic, Solomon Islands)
Russian (Indo-European)
Sobei (Western Oceanic, Irian Jaya)
Shilha (Berber from Morocco, Afro-asiatic)
Suena (Papuan)
Swahili (Benue-Congo)
Takia (Western Oceanic, PNG)
Tamazight des At Seghrouchen (Berber from Morocco, Afro-asiatic)
Tauya (Papuan, Trans-New-Guinea)
Tawala (Western Oceanic, PNG)
Timugon murut (Austronesian, Borneo)
Tombunuo (Austronesian, Borneo)
Tonkawa (Amerindian isolate)
Turkish (Turkic)
Usan (Papuan, Madang-Adelbert Range, N. Coast New Guinea)
Yupik (Eskimo-Aleut)
part i
Syntactic terminology
and typological methods
Clause linkage and Nexus

in Papuan languages
William A. Foley
University of Sydney
Interclausal relations in Papuan languages and in particular their prototypical
clause chaining structures have long presented serious descriptive problems.
These have been analyzed variously as instances of subordination, coordination,
and even a third unique type of relationship, cosubordination. This paper
argues that clause chaining structures are actually a type of coordination, but
distinguished from familiar types of coordination by the type of constituent
coordinated, S versus IP. The parametric variation found in clause chaining
constructions across Papuan languages is in turn accounted for in terms of
the types of functional heads of verbal inflections, negation, mood, tense,
illocutionary force, which head the individual IPs conjoined in clause chains.
This paper presents a revision of the theory of clause linkage, in particular the theory
of nexus, first developed in Foley & Van Valin (1984) and restated in Van Valin &
La Polla (1997) and Van Valin (2005). The original theory proposed three categories
of nexus, the traditional ones of subordination and coordination and a new type,
cosubordination. Subordination and coordination were distinguished along the traditional lines of embedded versus non-embedded. For our purposes here, we will
define an embedded clause as one which functions as a constituent, either core or
oblique (Andrews 2007; Foley 2007), of another clause, the main or matrix clause.
Conventionally, grammarians have called embedded subordinate clauses which function as core arguments complements, and those which function as oblique constituents, adverbial clauses, but in our view this is not the most perspicacious terminology
because it obscures their overall similarity, a similarity clearly brought out in the
structure of many Papuan languages. For that reason, in this paper we will refer to
both types simply as subordinate clauses and note the level of embedding, core versus
oblique. Clauses linked in a coordinate nexus are not in an asymmetrical relationship of embedded versus matrix clause, but rather are joined at the same level, strung
along rather like beads on a string. Designating a clause by the exocentric category
William A. Foley
S(Bresnan 2001), we can represent the contrast between subordinate and coordinate
nexus as Figure 1:
Subordinate Nexus
SSSS
Coordinate Nexus
Figure 1. Traditional Nexus Types
Olson (1981) and Foley & Van Valin (1984) introduced a third type of nexus to
the traditional two, a type they called cosubordination. This was distinguished from
coordination in that clauses in a cosubordinate nexus linkage were in a dependency
relationship for a particular inflectional category or operator like tense or mood, a
dependency which did not hold for coordinate nexus. This inflectional dependency
somewhat parallels the dependency that a subordinate clause has on its matrix clause,
although the nature of the dependency is semantic for cosubordinate clauses, but
structural for subordinate clauses. But keeping this difference in mind, both types
could be characterized as [+dependent]. On the other hand, clauses linked in either
coordinate nexus or cosubordinate nexus are not embedded, but co-ranked, so they
can both be classified as [embedded] in contrast to subordinate clauses; the sole contrast between coordinate and cosubordinate nexus is in the behavior with respect to
inflectional verbal categories like tense, mood or illocutionary force. For coordinate
nexus each clause is separately specified for these, but in cosubordinate nexus there
is a single specification for these, either in the initial or final clause and every other
clause in the linkage takes its specification for such features from them, as in these
examples from the Amerindian isolate language Tonkawa (Hoijer 1949):
(1)
a.

tekekee:k a:pa-w e:-ta ke-yae-w,

in.that.bush hide-imp and-sr 1sg.o-watch-imp
Hide in that bush and watch me!
b. tekekee:k a:pa-ta ke-yae-w

in.that.bush hide-sr 1sg.o-watch-imp

Hide in that bush and watch me!
Note that in the (a) example with coordinate nexus, both clauses are specified as
commands with the imperative suffix w on the verbs in each clause; further, the
clauses are linked by an explicit conjunction e:- and to which the suffix ta is
added, indicating that the subjects of the two clauses are coreferential. In the (b)
example illustrating cosubordinate nexus, the indication of the sentence as a command is marked only once, by the suffix w on the verb of the second clause; the verb
of the first clause is simply affixed with the suffix indicating coreferential subjects
between the two clauses. Yet the initial clause is also a command: the scope of the
imperative suffix spreads backward across the nexus linking the two clauses to apply
to the initial clause as well as the clause in which the verb is overtly marked with w
IMP. In addition no coordinating conjunction is used; this is typical of cosubordinate
nexus. In coordinate nexus each clause is individually specified for verbal inflections
like tense, mood and illocutionary force (following current conventions in generative grammar, I will call these I features, short for inflection), while in cosubordinate
nexus, all clauses are under the scope of the I features of the fully inflected verb in
the initial or final clause (hence in a loose sense, all clauses are cosubordinated to
the I features, although not truly embedded in the precise way we defined the notion
above). The contrast between coordination and cosubordination may be represented
as Figure 2:
Coordinate Nexus:
(S)I (S)I (S)I (S)I
Cosubordinate Nexus: (S S S S)I

Figure 2. Coordinate versus Cosubordinate Nexus
Over the past two decades or so, it has become increasingly obvious that there
are some serious difficulties with the theory of nexus and particularly problematic is
the notion of cosubordination. Foley & Van Valin (1984) proposed a set of what they
termed peripheral operators, here renamed the verbal I features, to which clauses in
cosubordinate nexus were subject: tense, mood, illocutionary force. But evidence has
been mounting that the scope relationships of these need not be the same. Examples
will be presented below, but the basic point is that clauses might be separately specified for tense inflection, but be under a single illocutionary force marker. In terms of
Figure 2, this would entail that the clauses are in a coordinate nexus with respect to
tense, but a cosubordinate nexus with respect to illocutionary force, hardly a happy
conclusion if nexus is to be taken, as it should be, as a structural relationship, for normally clauses should not be able to bear contrasting structural relationships to each
other. Constructions which have often been identified as prototypical examples of
cosubordinate nexus like clause chaining constructions in Papuan languages and the
converb constructions of central and south Asian languages (Haspelmath & Knig
1995) continue to provide examples in which illocutionary force, the highest peripheral
William A. Foley
operator or I feature, need not be shared across the clauses and hence by definition
they cannot exemplify cosubordinate nexus:
(2) a.

Tauya
(Trans New Guinea; MacDonald 1990)
tepau-fe-pa
yate fitau-a-nae?
break-tr-sr go
throw-2-q
Did you break it and go away? or
You broke it and did you go away? or
Did you break it before going away?
b.

Newari
(Tibeto-Burman; Genetti 2005)
mun biskut ar-i
do-an
ch
yer-a
r
3sg.erg biscuit buy-inf finish-cvb house come-3sg.pst q
When he finished buying the biscuits, did he come?
Note that in the clause chaining example from the Papuan language Tauya (2a), the
illocutionary force question suffix nae can have scope over both clauses, as in the first
translation, what would be expected from cosubordinate nexus. But crucially it does
not need to: it may have scope over the final clause, whose verb is affixed with nae,
leaving the first clause as a statement, as in the second gloss; and even more surprisingly in the final gloss, only over first clause, whose verb is unaffixed for interrogative
illocutionary force, leaving the final clause as a statement, in spite of the fact that
its verb actually hosts the interrogative suffix! The second two translations are not
compatible with an analysis of cosubordinate nexus, but instead suggest coordination. A similar effect obtains in English in sentences like do you work two jobs because
you need the money? in which the interrogative illocutionary force has scope over the
second clause even though it is realized formally in the first by subject-finite verb
inversion. In the Newari example of (2b), the question particle only has scope over the
final clause; the initial clause is again an assertion. Yet the construction involved is a
converb one, which elsewhere robustly shows all the features of cosubordinate nexus,
typically shared I features of the verb in the final clause across the preceding clauses.
Cosubordinate nexus has all the features of a mirage: sometimes it appears clearly;
other times it vanishes into the familiar territory of the traditional notion of coordination. This raises serious questions about its viability as a theoretical construct.
The notion of cosubordination was developed in the early 1980s, well before the
rise of a rich theory of functional categories like I and their projections. These innovations, particularly the notion of I and its projection IP, actually are quite central
to the revised theory of nexus we will present here. In Lexical Functional Grammar
(Bresnan 2001) there is a distinction between lexical categories, noun, verb, adposition, etc. and the phrase types they project, NP, VP, PP, respectively, and functional
categories like Iwhich do not typically correspond to independent lexemes, but are
more commonly inflections, like the verbal inflections of tense, aspect or mood or the
nominal inflections of definiteness or case. Still these functional categories like I or D
(for Determiner) or K (for Case) can project phrases like IP, DP or KP. In other words,
functional categories like I can be the heads of phrases (e.g. IP) as much as lexical
categories such as N can function as the heads of NPs. In this system of endocentric
phrase structure, i.e. a head of type X projects a phrase of type XP, so that N projects
NP and I projects IP, there is feature percolation of inflectional specifications of the
head to the phrasal category projected by it. Consider the following clause structure
from the Papuan language Yimas:
(3)
S
[Class: III]
[Num: PL]
NP
ADJ
kpa
big
numpray
pig
[Class: III]
[Num: PL]
pu-mal
-die
[Class: III]
[Num: PL]
The big pigs died.
Grammatical relations are indicated in Yimas by affixes to the verb; specifically for an
intransitive verb like mal- die, its subject is indicated by a prefix, which for third person subjects in addition must specify their gender class assignment and number. The
subject of mal- die is an NP kpa numpray big pigs, headed by a noun which belongs
to noun gender class III and is plural. This noun projects an NP in (3) and its features
of class and number percolate from the head noun to the phrasal level (indicated by
the bent arrow); the whole NP is now a syntactic constituent belonging to class III and
bearing plural number (feature matrix associated with the NP node). The features of
this NP and the subject agreement prefix on the intransitive verb are the same, so this
sentence is grammatical. If they clashed, the sentence would beungrammatical.
The syntactic category S is the odd man out in this framework. It is not endocentric like the other phrasal categories, but exocentric; in other words it lacks a projecting head. Consequently in (3) both daughter constituents of S, the NP and the V,
contribute equally to the semantics of the S node. The sentence is grammatical because
the features carried by the NP [Class: III; Num: PL] and the verbs subject prefix are
the same so that they unify together with no conflict to produce a structure like (4).
William A. Foley
(4)
V: die
SUBJ:
NP:
pig
ADJ:
big
Class: III
Num: PL
Note that both constituents, NP and V of (3) contribute equally to produce (4). The
verb provides the grammatical relation being predicated and its schematic features of
class and number, although these permit essentially an infinite number of possible NPs
which could fulfill this function. The NP provides the specific details of the participant
which does function as subject.
This theory of phrase structure, endocentric and exocentric, and associated
notions of projections are central to our revised theory of nexus. In essence we will
return to a traditional claim of two types of nexus, subordination and coordination,
distinguished by the type of phrasal configuration that they are in. The former notion
of cosubordinate nexus will be re-analyzed as a type of coordinate nexus that differs
from normal clausal coordination in the type of constituents coordinated. But firstly,
let us look at subordinate nexus in more detail. As discussed earlier, clauses in subordinate nexus are in an embedding relationship, with the subordinate clause functioning
as a constituent of the main or matrix clause. They may function as one of three types
of constituents, arguments (core), adjuncts (oblique) or modifiers of these two, corresponding to the traditional categories of complement, adverbial subordinate clause or
relative clause. The close interrelationships of these three types of embedded clauses is
strongly brought out in many Papuan languages, in that all three have the same structures, as in Fore (Scott 1978):
(5) a.

na-kib--ka-na
i-i-e
eat-likely-3pl.a.sbd-ref-3sg.a talk-3sg.a-decl
He talks about how they will eat.
b. a-ka-kib-i-pa
me-ki-i-e

3sg.o-see-likely-3sg.a.sbd-top get-likely-3sg.a-decl

If he sees it, he will get (it).
c.

a-egu-t--ti
wa:n-
3sg.o-hit-np-1sg.a.sbd-all go-2sg-sq
Are you going to where I hit him?
d. mi-nt-i
kuma:-ta-sa kana-i-e

be.at-rp-3sg.s.sbd village-loc-abl come-3sg.s-decl

He came from the village in which he stayed.
(5a) corresponds to a complement clause construction in traditional grammar terms,

the what is talked about of i- talk, while (5b) is an adverbial clause, specifically a
conditional clause, and (5c,d) are relative clauses, headless and headed respectively.
The overall unity of these examples is demonstrated by their structural realization:
all have subject markers drawn from a single set, which is used only in subordinate
clauses, and further all take typical markers of NPs, either case suffixes like ka REF, e.g.
concerning, ti ALLative and sa ABLative or the topic marking suffix pa. This last
fact is particularly salient for the syntactic properties of subordinate nexus: subordinate clauses in Fore and other Papuan languages are always embedded as complements
within a particular phrase type such as NP, PP or DP. Consider the structure of the
subordinate clause of (5b) in (6):
NP
(6)
TOP
N
IP [TNS: LIKELY]
S
[TNS: LIKELY]
V
a-ka- - i
3.-see -3..

- ib
-pa
(In a more fully articulated version of Lexical Functional Grammar, the N node in
(6) would be omitted due to the Principle of Economy of Expression (Bresnan 2001),
but we retain it here for ease of exposition. If Economy of Expression did apply to
(6), the NP node would become that which is unable to host I features). Because this
is a finite clause inflected with the tense/mood specification of LIKELY, an I head
is present and this in turn projects an IP. Because the marker of tense/mood is a
bound affix, it must be realized as a suffix to the verb, between the verb root and the
William A. Foley
subject agreement affix, but nonetheless it projects the dominating phrasal category
IP. Because IP is an endocentric phrase type, the features of the I head percolate to
the IP node, but from here they can go no farther, as the next dominating node is N,
a category not compatible with the verbal inflectional features of the I node. Essentially, this makes the subordinate clause an island: neither can its features percolate
up to the level of the matrix clause nor can the I features of the matrix clause move
down into it through the mismatching NP node. This accounts for the oft noted fact
that subordinate clauses are typically impervious to the illocutionary force of their
matrix clauses; they are usually presupposed statements.
Having said this, some putative subordinate clauses in English and presumably other languages seem to contradict this claim. Consider the example quoted
above, do you work two jobs because you need the money?; this sentence is ambiguous between three readings, and in two of these the clause beginning with because,
traditionally analyzed as an adverbial subordinate clause, is within the scope of the
interrogative illocutionary force: is it true you work two jobs and is that because
you need the money? and I take it you work two jobs, but do you do that because
you need the money?. As this paper specifically concerns Papuan languages, a
full consideration of this issue is beyond its scope, but it seems that a fruitful
approach would be to query whether these types of adverbial clauses with conjunctions like because, if, when, although, etc. are embedded at all and hence instances
of subordination in the restricted terms defined here. Note that many of them
function only elliptically, if at all, as the heads of phrases: ???if/when/although the
party. This renders them ineligible to project a phrase within which an IP could be
embedded under its complements node. Unquestionably this is related to another
systematic difference between these adverbial clauses in English and subordinate
clauses in Papuan languages. In the latter, constituent NPs within subordinate
clauses cannot be relativized (MacDonald 1990), presumably due to a constraint
against stacked embedded clauses, but this is perfectly possible in English: do
you work two jobs because you want to make up the money that your wife lost on
the horses?
Many Papuan languages have a contrast between finite and nonfinite subordinate
clauses, and in both cases the dominating syntactic node is that of an NP. Finite subordinate clauses are embedded under an IP projected by the I head bearing the I verbal
inflectional features, while nonfinite subordinate clauses lack these I features and hence
simply correspond to a S constituent undominated by an IP. Yimas istypical:
(7) a.

nonfinite
tantaw-am-kia-r-awt-\an ma\ckrm
tma-mp-kra-k
sit-eat-night-nfn-sg-obl binding(v.dl) v.dl.o-3dl.a-cut-irr
While (he) was sitting and eating, they both cut the two bindings.
IP
[TNS: IRR]
S
NP
I
[TNS: IRR]
NP
CASE
OBL
tantaw-am-kia-r-awt- -\an
sit-eat---
ma \ckrm
binding(.)
tma-mp-kra..-3.-cut-
-k
b. finite

ya-mpu-na-pay-kulana-tay-\c-mp-n

v.pl.o-3pl.a-dur-now-walk-see-prs-VIII.sg-obl

ya-mpu-na-wayk-n
v.pl.o-3pl.a-dur-buy-prs
While they are walking around looking at (the goods), they are buying them.
[TNS: PRS]
IP
[TNS: PRS]
V
NP
N
[TNS: PRS]
IP
CASE
-mp
VIII.
I [TNS: PRS]
ya-mpu-na-pay-kulana-tay
-\c
.-3.--now-walk-see-
-n
ya-mpu-na-wayk .-3.--buy
-n
William A. Foley
These are subordinate clauses which express events essentially simultaneous with the
events expressed in the matrix clause. In Yimas there is a choice between a finite and a
nonfinite subordinate clause to express this meaning. Both constructions are expressed
as nominalizations, e.g. NPs suffixed with the oblique case suffix n ~ nan. The nonfinite structure has no head noun; the verbal complex is suffixed with the nonfinite
suffix r(u) and a suffix marking the number of the subject of the clause. This last is
drawn from a set of subject markers used solely in nonfinite constructions of all sorts.
No other pronominal agreement affixes for core arguments are possible in nonfinite
constructions, in contrast to finite constructions which have full agreement possibilities for all core arguments. The structure of finite complements is more complex and
like the Fore example in (5c) is essentially a relative clause. The suffix mp is a number
and gender class nominal suffix for VIII.SG and denotes an obligatorily missing noun
of this class, pucm part, time. This suffix functions as the head of the relative clause,
as is typical of relative clauses in the language (see Foley (1991: 413433) for further
discussion), so the embedded clause can be more accurately be paraphrased as at the
time that they are walking around looking at (the goods). Furthermore, as the verb of
the subordinate clause is a fully inflected one, with the required I feature of tense for
a finite verb, it takes the normal pronominal agreement prefixes for core arguments
in contrast to the truncated agreement pattern of the nonfinite verbs (compare the
subject marker awt of the nonfinite verb in (7a) which simply marks its number with
the much richer agreement array of the finite verb in (7b).
The use of a topic marker illustrated by the Fore example (5b) is a very common
mode of indicating subordinate clauses in many Papuan languages, particularly those
which like Fore belong to the Trans New Guinea family, a fact that was first noted
by Haiman (1978). The actual syntactic status of this topic marker varies somewhat,
although it always diagnoses a maximal XP projection impervious to the percolation
of I features from the embedded subordinate clause. In some languages like Tauya
(MacDonald 1990), it patterns very much like a case marker:
(8) a.

fofe-a-te-ni
yate fitau-e-a
come-3sg-dr-erg go throw-1/2sg-decl
Because he came, I went away.
b. yau-pa-ra tu-ane-e

see-sr-top give-2pl.fut-imp

If you (PL) see (him), give (it) to him!
In such languages the node dominating the embedded clause is either an NP as in (6),
or, if we take the topic marker as a case functional head K, then it would project a KP
above the subordinate clause:
(9)
a.
(= (8a))
KP
b.
[CASE: ERG]
K
KP
[CASE: ERG]
(= (8b))
[CASE: TOP]
[CASE: TOP]
fofe-a-tecome-3-

-ni
yau-pasee--
-ra
In Tauya, the topic marker can co-occur with an overt case marker, indicating the
possibility of case stacking, as in Australian languages (Nordlinger 1998):
(10) a.

mei fofe-a-te-ni-ra
yate fitau-e-a
here come-3sg-dr-erg-top go throw-1/2sg-decl
Because he came here, I went away.
b.
KP
[CASE: ERG]
[CASE: TOP]
KP [CASE: ERG]
S
XP
mei
here

K
[CASE: TOP]
K
V
fofe-a-tecome-3-
[CASE: ERG]
-ni
-ta
The percolation of [CASE: ERG] from a lower KP node to the upper KP node is permissible: they are nodes of the same category and hence able to host the same type of
features, e.g. CASE, as long as their specifications are not contradictory. Unlike other
case specifications, say accusative, topic is compatible with ergative.
But in still other languages like Usan (Reesink 1987), the topic marker seems to
belong to the category of Determiner. The topic marker eng in Usan is quite clearly
the same as the proximal deictic eng this one, composed of the stem e- here plus a
William A. Foley
specifying suffix ng. In Usan eng is used to mark subordinate clauses, those functioning
as adjuncts (i.e. adverbial clauses) or NP modifiers (relative clauses):
(11) a. ye gigi di-em
eng

1sg first come.up-1sg.fp top

tp susu ir-amei
path wrong go.up-1sg.fp
When I came up first, I took the wrong path.
b. munon emi bau-ori

eng ye me ge-au

man bow take-3sg.fp top 1sg neg see-neg

I didnt see the man who took the bow.
The topic marker eng in languages like Usan is a functional head D that projects a DP.
DPs are a common areal feature of languages of the Madang region, possibly due to diffusion from Austronesian languages; elsewhere among Papuan languages they are rather
rare. DP is another phrasal category that cannot host I verbal inflectional features, so the
subordinate clauses are again islands with respect to the I features of the matrix clause:
(= (11a))
(12)
[TNS: FP]
IP
S
DP
NP
D
[TNS: FP] IP
S
NP ADV
ye
1

I [TNS: FP]
V
ADJ
I [TNS: FP]
V
gigi
di-em
first come.up 1.
eng tp
path
susu
ir- -amei
wrong go.up 1.
Now consider the following contrasting sentences in Usan:

(13) a.

ye namanimun gumat-em eng big-ine

me is-au
1sg letter
write-1sg.fp top put-1sg.dr neg go.down-neg.tns
I didnt mail the letter that I wrote.
b. ye namanimun gumat big-ine

me is-au

1sg letter
write put-1sg.dr neg go.down-neg.tns

I didnt write and mail a letter. = I didnt send a letter.
Note the differential behavior of the negative me in these two sentences. In the first
example the negation fails to spread into the subordinate clause, i.e. the subordinate
clause remains a positive statement (the facts of NEG-transportation in languages like
English in sentences like I dont think that John is the thief, in which the clause following that is actually under the scope of negation again might suggest that that complement clauses of verbs of saying or thinking are actually not embedded. For an analysis
suggesting this is in fact true at least with direct quote complements of verbs of saying in some languages see Munro (1982)). In the analysis of subordination presented
here the failure of negation to spread into Usan subordinate clauses is to be expected
because the dominating phrasal node DP provides a barrier to the spread of any I features like tense, mood or polarity from one clause to the other:
(14)
IP
S
I [TNS: NEG]
[POL: ]
DP
[TNS: FP]
[POL: +]
IP
S
NP
NP
D
I [TNS: FP]
[POL: +]
ye namanimun gumat- -em

1
letter
write 1.
eng
me -au
.
The fact that the negative polarity does spread in (13b) strongly indicates that this is a
different type of nexus relationship; indeed (13b) illustrates coordinate nexus, but at
the S level, not the IP level, i.e. it is a single IP projected by a single I head dominating
a string of coordinated S constituents:
(15)
(= (13b))
[TNS: NEG]
[POL: ]
IP
S
S
NP
ye
1
NP
namanimun gumat
letter
write
I
S
big-ine
put-1.
isgo
[TNS: NEG]
[POL: ]
me -au
.
William A. Foley
(15) illustrates the classic structure of clause chaining so well attested in languages of
the New Guinea region. As is well known, in such structures the verbal inflectional I
features of the verb of the last clause typically have scope over the preceding medial
or dependent clauses. Verbs in medial clauses are commonly stripped down inflectionally in comparison to final verbs, as a reflection of this scope dependency. But,
in fact, the inflectional I categories of the verb of the final clause do not belong to it,
but rather to the structure as a whole, as in (15); they merely appear on the final verb
in the sentence because it is the closest verb capable of hosting them. The verb of the
final clause is actually at the same level as all those medial verbs preceding it; more
precisely, it is just one more dependent verb which takes its I feature specifications
from the dominant IP node projected by the I head of the whole sentence. The verbs
in the coordinated S constituents are strictly speaking nonfinite, as they themselves
have no intrinsic I feature specifications.
Example (13b) is an instance of what was analyzed in Foley & Van Valin (1984)
and Van Valin & La Polla (1997) as cosubordinate nexus. We are now re-analyzing the former cosubordinate nexus as simply coordinate nexus. This was foreshadowed in Foley (1986) where cosubordinate nexus was defined as coordinate
but dependent, but the notion of dependence remained undertheorized. Here
dependence is simply taken as being the complement of a single I head. What
really distinguishes clause chaining structures or the former cosubordinate
nexus from standard coordinated clause structures is simply the nature of the
constituents being coordinated, S versus IP. Compare these Kewa examples
(Franklin 1971):
(16) a. nip pu-la

pare n
paal na-pa
3sg come-3sg.prs but 1sg afraid neg-be.3sg.prs

He is coming but Im not afraid.
b. n rka-mo gaa l-a

1sg stand-dr talk say-3sg.pst

I stood up and he talked.
(16a) is a coordination of two IPs, linked by a coordinating conjunction pare but;

note that the verbs are both fully inflected for the I feature tense and that the scope
of the negation is confined to the second clause. (16b) is a coordination of S constituents under a single IP node, a clause chaining structure; tense is only indicated
on the verb of the final clause. The verb of the first clause has no tense marking, but
is taken as past tense; it is a stripped down medial verb form simply registering the
switch of the referents of the subjects between the two clauses (I he). The difference can be represented as follows:
(17)
a.
IP
(= (16a))
[TNS: PRS]
[POL: +] IP
S
NP
I [TNS: PRS]
[POL: +]
NP
npu pu-la
3 come 3.
[TNS: PRS]
[POL: ]
IP
CONJ
I [TNS: PRS]
[POL: ]
AJT
pare n paal
but 1 afraid
pbe.
na- -a
- -3.
b.
(= (16b))
IP
[TNS: PST]
I [TNS: PST]
NP
NP
n
1
rka-mo
stand-
gaa
talk
V
l-a
say 3.
This re-analysis of the former cosubordinate nexus as just coordinate nexus, but
with coordination of S constituents rather than IP begins to provide an explanation for
the differential behavior of I features across Papuan languages. The examples of (17)
suggest a binary contrast between the coordination of multiple S constituents under a
single IP projection from a single I head, the sole place for the specification of the I features and the coordination of multiple IP constituents, each with their own I head and
independent specification of I features. These may indeed be the prototypical extreme
cases and were the basis of the original typology of Foley & Van Valin (1984), but there
are in between types that languages often exploit. For instance, Usan treats negation as an
I feature which must have scope all coordinated S constituents under it (example (15)).
In Tauya (MacDonald 1990), on the other hand, this spread of negative scope is only
possible when all clauses share the same subject (although the scope of the negative need
William A. Foley
not spread). When the subjects between the clauses are different, negation in the final
clause can never have scope over the preceding medial clauses:
(18) a. ne fofe-pa wate pofei-a-a

3sg come-sr neg talk-3sg-decl

He didnt come and talk. or he came and didnt talk.
b.

ne fofe-a-te
ya-ni
wate tu-e-a
3sg come-3sg-dr 1sg-erg neg give-1/2sg-decl
He came and I didnt give it to him.
*He didnt come and I didnt give it to him.
This differential behavior of negation in Tauya is in sharp contrast to illocutionary

force, which although marked on the final verb is an I feature which always has a reading in which it has scope over the whole series of coordinated constituents (although
other readings are also possible: see example (2a) and discussion thereafter). This indicates that polarity is a lower level I category which may be independently specified for
coordinated constituents with same subjects and must be so for those with different
subjects, while illocutionary force remains a feature of the highest I head and hence
highest IP projection:
(19)
(= (18b))
[IF: DECL] IP
IP
I
[IF: DECL]
[POL: +] IP
S
[POL: ] IP
I [POL: +]
NP
ne
3
fofe-a-te
come-3-
I [POL: ]
NP
ya-ni
1-
tu-e
give-1/2
wate
-a
Other Papuan languages show variation in other I features like mood and tense,
particularly the former. Many Papuan languages inflect the verbs in medial clauses in
clause chaining constructions for mood, typically a realis versus irrealis contrast, while
the verb of the final clause bears the full inflectional possibilities of tense and illocutionary force, the features of the final I head. Watam is typical of this pattern; verbs in
medial clauses are marked for realis versus irrealis:
(20) a. min amba-r-a sag-ri

3pl eat-r-ev go-past

They ate and then went.
b. min am-(m)be sa(g)-nan

3pl eat-irr
go-fut

They will eat and then go.
The verbs of the medial clauses are marked r realis when the tense of the whole sentence, i.e. the main I head is past or present, and they are marked with mbe irrealis
when the tense is future or the illocutionary force is imperative. Amele (Roberts
1990) and Bargam (Hepner 1995) are other Papuan languages which behave similarly. Languages like Watam, Amele and Bargam all require I heads in medial clauses
for which mood is indicated, but this cannot be independently specified from the I
features of the main I head and the top dominating IP node. The mood inflection
possible in the lower I heads is strictly determined by the tense and illocutionary
force of the dominating IP node, which in turn are projected from the main final
Ihead:
(21)
IP
(= (20b))
[TNS: FUT]
[MOOD: IRR]
[MOOD: IRR]
IP
I
[TNS: FUT]
IP
IP
I
NP
min
3
ameat
[TNS: FUT]
[MOOD: IRR]
V
-mbe
sa(g)go
-nan
The final coordinated verb here has and can have no overt marking for mood as the
tense suffix for the sentence as a whole usurps its position; there is only one suffix slot
for tense-mood-illocutionary in the language.
A few languages do seem to allow mood inflection to differ between the clauses in
a clause chaining structure. Mianmin (Fedden 2007) is one such language, although
the data are still inconclusive as to whether the inflectional category involved is
tenseor mood; in our view it is the latter and that is how we will analyze it here. A
William A. Foley
disjoint reading of mood is possible in clause chaining structures if the clauses have
different subjects:
(22) un-e-a

eil- a-nan-omab-io-be
3sg go.prf-3sg.m-mv 3pl pig-m 3sg.m.o-kill.prf-aux.prf-irr-2/3pl.a-decl
He will go and they will kill a pig. or He has gone and they will kill a pig.
In the first gloss of (22) the mood-tense suffix omab irrealis/future has scope
over both clauses in the coordination, but in the second it only has scope over the
second. Whether we analyze omab as irrealis mood or future tense, the fact remains
that on the second reading the first clause is not either of these, but realis or past tense.
Note that there is no overt mood-tense inflection in the first clause to indicate realis/
past, yet that is an available reading. This means that the first clause must have the
possibility of its own I head with mood-tense specification, even though it iscovert:
(23)
(= second gloss of (22))
[IF: DECL]
IP
I
[IF: DECL]
IP
[MOOD: R] IP
S
IP [MOOD: IRR]
I [MOOD: R]
NP
un-e-a
go-3-
[MOOD: IRR]
NP
NP
eil-
pig-
a-nan- -io3..
-kill- --
-omab
-be
The category of tense can be complicated because tense morphology can be used to
mark both absolute and relative tense. Absolute tense, the deictic anchoring of the time
of the event reported in the sentence with respect to now, the time of the speech event,
is always a feature of the highest I head and hence percolates to the top dominating IP
node. On the other hand, verbal forms inflected for tense can be used to signal relative
tense, i.e. a sequential or simultaneous relationship between the events expressed in the
coordinated clauses in the clause chaining construction. In these cases the tense inflected
forms function rather like aspectual markers or temporal suffixes in other Papuan languages. Korafe and Suena (Farr 1999) are examples of such languages; past tense verbs
indicate a sequential relationship between the events denoted by the clauses, while present
tense forms express simultaneous events. Consider these examples from Suena:
(24) a. pot-ena
bam-ia

give-1sg.pst go-3sg.ip

I gave it to him and then he went.
b. gi
pupi-nona pu bam-ia

spear get-1sg.prs pig go-3sg.ip

While I was getting a spear, the pig went away.
In both these examples the overall tense of the sentence is immediate past (today), as
indicated by the inflection for tense on the final verb, which, of course, is simply the
formal realization of the tense feature for the highest I head. The verbs of the medial
clauses are also inflected for tense, past versus present respectively, but these tense specifications are interpreted with respect to the absolute tense of the final verb, determined
by percolation from the main I head to the top dominating IP node, the absolute tense
of the whole sentence. The past tense on the medial verb means events that are past with
respect to the immediate past of the whole sentence, hence earlier in time or a sequential relationship between the time of the event of the medial clause and that of the final
clause. Present tense of the medial verb indicates events at the same time as the immediate past of the whole sentence, or a simultaneous temporal relationship between the
events of the two clauses. In essence these relative tense inflections function like aspect
in other languages, such as past perfect versus past progressive in English, although
Korafe and Suena do have other ways to express aspect such as serial verb constructions. We can treat tense inflection on the verbs of medial clauses in languages like
Korafe or Suena as specifications of tense in lower I head positions, but clearly those
must be interpreted in line with the overall scope of the absolute tense of the top dominating IP node (which in turn comes from the tense feature of the main I head of the
sentence); hence they will almost by definition correspond to relative tense:
(25)
(= (24b))
[TNS: IP] IP
IP
I [TNS: IP]
[TNS: PRS] IP
I [TNS: PRS] S
S
NP
gi
spear
IP
V
pupi- -nona
get 1.
NP
pu
pig
bamgo
-ia
3.
William A. Foley
Illocutionary force has been regarded as the most peripheral operator, the feature
least available to lower I heads. This generally seems to be true. Many researchers in
Papuan languages have remarked that the scope of illocutionary force is generally
over the whole sentence. There are two exceptions, however. One concerns content
questions involving wh-words, as in this Kte example (Schneuker 1962):
(26) ohe wena yu-ha-pire
gogo de-ye
2dl where be-sim-2dl.dr bell
sound-3sg.np
Where were you two while the bell rang?
Note that the first clause is in this clause chaining structure is a content question,
where were you two but the second is a statement, the bell rang; the illocutionary force of the two clauses is clearly different. But crucially the interrogative force
does not follow from an I feature, a verbal inflectional question marker, but from
a phrasal argument constituent within one of the S constituents. In terms of the
analysis presented here, such a phrasal argument within an S constituent, regardless of its status as a question word, cannot project an I feature specification to
conflict with whatever is illocutionary force of the verb of the final clause, because
the content question semantics is an inherent lexical feature of the question word,
not a functional I head, an inflectional category. So it is not possible for the lexical
semantics of this question word argument to percolate up to the maximal IP head
node. The actual sentence minus the content question argument is overall presupposed, i.e. has a neutral, perhaps declarative illocutionary force, albeit not an assertive one (a fact that sharply distinguishes content questions from polar ones, hence
their systematic crosslinguistic differences): given that you two were somewhere
and then the bell rang; so where was that; and this is the realization of the illocutionary force of the main I head and hence the top dominating IP and the sentence
as a whole:
(27)
The other exceptional case concerns sporadic uses of interrogative and imperative
or hortative illocutionary force for the main I and hence the top dominating IP, but
the medial clauses are neutral or declarative statements of presupposed information.
The opposite pattern in which the presupposed information is in the final clause is also
found. The examples in (2) are illustrative, as is (28) below from Kte (Suter 1992):
(28)

su
kpeue-me
natsa-ndzepie
banana ripen-3sg.seq.dr 1pl o.tell-2pl.hort.fut
Let us know when the bananas are ripe!
literally The bananas will ripen and you let us know!
The exact analysis of cases like (28) is not entirely clear, and the ideas presented here
are preliminary and speculative, requiring more extensive future research, but it does
seem possible that a illocutionary force feature of the highest I head and hence the
top dominating IP of the sentence need not apply to medial clauses, which remain
presupposed statements. In this way, illocutionary force in such cases behaves behaves
like negation in Tauya. But even more striking is the possibility exemplified by the
last gloss of the Tauya example in (2a): in that example the interrogative illocutionary
force only has scope over the preceding medial clause did you break it before going
away?, i.e. the final clause you go away is a presupposed statement, what is being
queried is whether you broke it first. The data lead to a conclusion that any clause can
be presupposed in a coordinate nexus, regardless of the dominant illocutionary force
feature, even the clause which bears the interrogative or imperative illocutionary force
marker! This last fact makes any analysis with lower I heads bearing distinct illocutionary force features implausible because it would require the verbs of final verbs in
examples like (2a) to bear two conflicting illocutionary force specifications, one overt,
e.g. interrogative or imperative and the other covert, declarative. Note further that no
clear examples of conflicting overt illocutionary force operators are forthcoming; no
examples, for instance, of clause chaining structures with interrogative in the medial
clause and imperative in the final one, or vice versa. Nor are the so-called declarative
clauses ever really assertions, but they are typically presupposed, taken for granted,
background information, such as the sentence minus the content question word in
(27) above. Sentences like the following one from Fore (Scott 1973) might be seen to
contradict this claim:
(29) na-me-gnt-
na-ku-w-e
1sg.o-give.2sg.a.fut.dr-1sg.seq-imp eat-fut-1sg.a-decl
Give me something and then I will eat it.
Here the final clause is quite arguably an assertion and the first clause an imperative,
clearly two distinct illocutionary force features in a clause chaining structure. But this
exception is more apparent than real. While (29) is indeed a clause chaining structure,
William A. Foley
it is a coordination of two independent IPs, each with completely specified I heads.

Example (29) has the structure of the Kewa example in (16a) and (17a). Fore differs
from Kewa in lacking coordinating conjunctions and the familiar coordination pattern
of independent sentences. It only has two ways of joining clauses together, subordination, exemplified in (5), and clause chaining. One cannot assume that clause chaining
always corresponds to the same types of structures across languages; minimal S constituents may be coordinated using this structure, but so can fully specified essentially
independent IP constituents.
All this evidence suggests that presupposed statements in clause chaining structures have no real independent specification for illocutionary force at all, i.e. they
are not asserted, questioned, commanded, exhorted, etc., and hence whatever their
formalstatus appears to be cannot conflict with whatever the illocutionary force of
the sentence is as a whole. Their apparent inflection as declarative or interrogative
or whatever is quite likely a superficial morphosyntactic constraint requiring inflections to be realized in particular formal ways, e.g. final suffixes in Trans New Guinea
languages, initial subject-finite verb inversion in English, but the presupposed clauses
actually have no true illocutionary force at all. The illocutionary force of a sentence is
always determined by the illocutionary force specification of the main I head, which
in turn percolates to the top dominating IP, and the place of that head is determined
by the formal properties of the language, suffixation on the final verb in Trans New
Guinea languages versus position of the finite verb in English; its position by itself does
not determine which clauses are presupposed or within its scope. Finally, as we have
seen, both subordinate clauses and coordinate clauses in clause chaining structures
in Papuan languages can be presupposed. What the division of labor is between these
two types of nexus in these languages with respect to presuppositions and the scope of
illocutionary force is another major research question awaiting exploration.
Abbreviations
1
2
3
III, V, VIII
a
abl
adj
ajt
all
aux
ben
first person
second person
third person
Yimas gender classes
subject of a transitive verb
ablative case
adjective
adjunct
allative case
auxiliary
benefactive
caus
cvb
decl
dep
dl
dr

dur
erg
ev
fp
causative
converb
declarative
dependent verb
dual number
different referent of
subjects
durative
ergative case
echo vowel
far past tense
fut
hort
if
imp
inf
inst
ip
irr
kp
loc
m
mv
neg
nfn
np
o
obl
pst
future tense
hortative
illocutionary force
imperative
infinitive
instrumental
immediate past tense
irrealis
case phrase
locative
masculine gender
medial verb
negative
nonfinite
near past tense
object of a transitive verb
oblique case
past tense
prf
pl
pol
prs
q
r
ref
rp
s
sbd
seq
sg
sim
sr
subj
tns
top
tr
perfective
plural number
polarity
present tense
question marker
realis
referential case
remote past tense
subject of an intransitive verb
subordinate
sequential
singular number
simultaneous
same referent of subjects
subject
tense
topic
transitive
References
Andrews, Avery. 2007. The major functions of the noun phrase. In Syntactic Description and
Language Typology, Timothy Shopen (ed.), 132223. Cambridge: CUP.
Bresnan, Joan. 2001. Lexical-Functional Syntax. Oxford: Blackwell.
Farr, Cynthia J.M. 1999. The interface between syntax and discourse in Korafe, a Papuan language of Papua New Guinea. Pacific Linguistics C148.
Fedden, Olcher Sebastian. 2007. A Grammar of Mian, a Papuan Language of New Guinea. Ph.D.
dissertation, University of Melbourne.
Foley, William. 1986. The Papuan Languages of New Guinea. Cambridge: CUP.
Foley, William. 1991. The Yimas Language of New Guinea. Stanford CA: Stanford University Press.
Foley, William. 2007. Toward a typology of information packaging in the clause. In Syntactic
Description and Language Typology, Timothy Shopen (ed.), 362446. Cambridge: CUP.
Foley, W. & Van Valin Jr., Robert D. 1984. Functional Syntax and Universal Grammar. Cambridge: CUP.
Franklin, Karl J. 1971. A grammar of Kewa, New Guinea. Pacific Linguistics C16.
Genetti, Carol. 2005. The participial construction of Dolakh Newar: Syntactic implications of
an Asian converb. Studies in Language 29(1): 3587.
Haiman, John. 1978. Conditionals are topics. Language 54: 564589.
Haspelmath, Martin & Knig, Ekkehard (eds). 1995. Converbs in Crosslinguistic Perspective.
Berlin: Mouton de Gruyter.
Hepner, Mark. 1995. Tense, aspect and modality in Bargam. Language and Linguistics in
Melanesia 26: 132.
William A. Foley
Hoijer, Harry. 1949. Tonkawa syntactic suffixes and anaphoric particles. Southwestern Journal of
Anthropology 5: 3755.
MacDonald, Lorna. 1990. A Grammar of Tauya. Berlin: Mouton de Gruyter.
Munro, Pamela. 1982. On the transitivity of say verbs. In Studies in Transivity, Paul Hopper &
Sandra A. Thompson (eds), 301318. New York NY: Academic Press.
Nordlinger, Rachel. 1998. Constructive Case: Evidence from Australian Languages. Stanford CA:
CSLI.
Olson, Michael L. 1981. Barai Clause Junctures. Ph.D. dissertation, Australian National University.
Reesink, Ger. 1987. Structures and their Functions in Usan [Studies in Language Companion
Series 13]. Amsterdam: John Benjamins.
Roberts, John R. 1990. Modality in Amele and other Papuan languages. Journal of Linguistics
26: 363401.
Schneuker, Carl L. 1962. Kte Language Handbook. Madang PNG: Lutheran Mission.
Scott, Graham K. 1973. Higher levels of Fore grammar. Pacific Linguistics B23.
Scott, Graham K. 1978. The Fore language of Papua New Guinea. Pacific Linguistics B47.
Suter, Edgar. 1992. Satzverbindung im Kte. MA dissertation, University of Zrich.
Van Valin Jr., Robert D. & La Polla, Randy J. 1997. Syntax: Structure, Meaning and Function.
Cambridge: CUP.
Van Valin Jr., Robert D. 2005. Exploring the Syntax-Semantics Interface. Cambridge: CUP.
Capturing particulars and universals

in clause linkage
A multivariate analysis*
Balthasar Bickel
University of Leipzig
Cross-linguistic variation in adjoined clause linkage is higher than what is
allowed by universal concepts like coordination or subordination which entail
sets of strictly correlated properties. This chapters uses statistical techniques to
uncover probabilistic correlations and clusters in a pilot database. For this, a set
of variables is developed that ranges in coverage from the scope of illocutionary
force operators to extraction constraints and that allows both detailed qualitative
analysis of language-specific clause linkage structures and large-scale quantified
measurement of the similarities of such structures within and between languages.
The study tentatively suggests that there is a prototype of subordination which is
closer to and-like than to chaining constructions, and that there is a continuum
between more vs. less tightly constrained types of converb and chaining
constructions, but no general prototype of cosubordination.
1. Introduction
The analysis of individual, language-specific structures normally starts with a set of
terms that are defined, or assumed to be defined, in a cross-linguistic way, and are
taken as such from field manuals, handbooks, formal theories, or reference grammars
*Versions of this paper were presented at the International Symposium on the Grammar and
Pragmatics of Complex Sentences (LENCA 3) in Tomsk, June 29, 2006, at the International
Conference on Role and Reference Grammar in Leipzig, September 30, 2006, at the Syntax
of the Worlds Languages conference in Berlin, September 26, 2008, and as a guest lecture at
the University of Zrich, December 8, 2008. I thank all audiences for stimulating questions.
Many thanks also go to Robert Van Valin for discussing issues of focus and extraction with
me and to Zarina Molochieva and Alena Witzlack-Makarevich for discussing the Chechen
and Russian data with me. I am also indebted to Volker Gast and the two non-anonymous
reviewers Jeff Good and Michael Cysouw for very helpful comments on an earlier draft. This
chapter is dedicated to the memory of Mickey Noonan (19472009).
Balthasar Bickel
of other languages. A case in point is the term clausal cosubordination, which was
introduced by Olson (1981) and Foley & Van Valin (1984) and is defined by conjunct
illocutionary scope: a clause that is cosubordinate to a main clause obligatorily falls
under the scope of illocutionary operators in the main clause. An example of this is
the medial form or converb construction, as it is found for example in the Papuan
language Amele or the African language Swahili:1

(1) Amele
(Trans-New Guinea: Madang; Papua New Guinea; Roberts 1988)
ho busale-e-b
dana age gbo-ig-a
fo?
pig run.out-ds-3s man 3p hit-3p-t.pst q
Did the pig run out and did the men kill it?

(2) Swahili
(chain)
(Niger-Congo: Bantu; Bickel 1991)
je, u-li-baki
nyumba-ni u-ka-tayarisha ch-akula ch-etu?
q 2s-pst-stay home-loc 2s-seq-prepare VII-food VII-our
Did you stay home and prepare our food?
(chain)
In both cases, the interrogative marker in the main clause (final fo in Amele, initial
je in Swahili) has scope over both clauses so that the only possible reading is one in
which the speaker inquires about the truth value of both propositions.
The definition sets cosubordination apart from coordination, where the scope
of such markers does not necessarily extend over both clauses, and also from subordination, where it is impossible to have conjunct scope (cf Foley & Van Valin
1984; Tikkanen 1995; Van Valin 2005; among others). When one takes the term
cosubordination further to the field, however, one quickly runs into structures
that look very similar to the data in Amele or Swahili, but do not entirely fit the
definition. Such structures are found for example in South Asian languages, such
as Belhare:2
. Where constructions figure in the pilot study described in Section4, I include in brackets
the (relatively arbitrary) identification label used in the Appendix. Glossing follows the Leipzig
Glossing Rules (http://www.eva.mpg.de/lingua/resources/glossing-rules.php), with the addition of add additive (focus), ass assertive, b B gender (in Chechen), cess cessative, conc
concessive, cond conditional, decl declarative, dep dependent, ds different subject, f.pst
far past, hort hortative, ill illocutionary, j J gender (in Chechen), pred predicate marker,
purp purposive, ptcl particle, seq sequential, ss same subject, temp temporary (aspect), t.pst
todays past, v V gender (in Chechen), w.pst witnessed past, and y.pst yesterdays past. Roman
numerals indicate noun classes.
. Data without a source specification are from my own fieldnotes.
Capturing particulars and universals in clause linkage
(3) Belhare
(Sino-Tibetan: Kiranti; Nepal)
a. khar-e
ki jutta
-in-ghutt-he-ga
i? (chain)

[3sS]go-pst seq shoes[nom] 3sA-buy-bring.for-pst-2sP q

Did she go [there] and buy you shoes?
or Did she buy you shoes when she went [there]?
(presupposing either she went or she bought)
b.

ne-e
yu-sa mundhupt-he i?
here-loc sit-cvb [3sS]chat-pst q
Did he sit here and chat (with you)?
or Did he chat with you when sitting here? (presupposing either
he chatted or he sat)
(cvb)
In these structures, the scope of the interrogative marker in the main clause (i), is indeterminate: depending on the context of utterance, the sentences may be interpreted as
having conjunct or disjunct scope. This indeterminacy can be found both with finite
(3a) and nonfinite (3b) forms (cf Section3.3 on finiteness). The same pattern can also
be observed in the Indo-European (Indo-Aryan) language Nepali:

(4) Nepali
yah -era
khn
kh-yo?
(chain)
here come-cvb food[nom] eat-3sM.pst
Did he come here and eat?
or Did he eat after coming here? (presupposing either he came here or he ate)
Data similar to these can be found in many other languages of South Asia (e.g. in
Kathmandu Newar: Hale & Shrestha 2006; Dolakha Newar: Genetti 2005; Burshaski:
Tikkanen 1995; Sanskrit: Tikkanen 1987; or Pali: Bickel 1991), and also in Papuan
languages even in languages of the same family as Amele, e.g. in Tauya:

(5) Tauya (Trans-New Guinea: Madang, Papua New Guinea; MacDonald 1990:226)
tepau-fe-pa yate fitau-a=nae?

(chain)
break-prf-ss go throw-2=polar.q
Did you break it and go away?
or Did you go away after breaking it? (presupposing either you went away or
you broke it)
Other examples from Papuan languages include Hua (Haiman 1980:400), Usan (Reesink
1987:297f), Kte (Suter 1992:25ff), and Korafe (Farr 1999).
The question that arises is how to analyze structures like (3)(5). There is a number
of possibilities. First, one could posit a second analytical term (cosubordination 2),
defined without a scope constraint. But this would miss the fact that the structures
are so similar to each other that one reading of cosubordination 2 (namely the one
with conjunct scope) is the sole reading of cosubordination 1. Second, one could
Balthasar Bickel
try and argue that in Belhare and similar languages, one reading really reflects
cosubordination while the other reflects something else presumably subordination,
with disjunct scope (Bickel 1998). While this may be a viable solution in some cases, at
least in Belhare and Nepali, I am not aware of any independent evidence for assuming
structural ambiguity: the possible readings can only be resolved pragmatically, and
it seems unjustified to posit differentiated syntactic representations for this (at least
under a parsimonious approach to syntax that does not try to resolve in the syntax
what can just as well be left to pragmatics). Third, one could revise the definition of
cosubordination, for example by defining the term without any syntactic constraint
on illocutionary scope (as is done by Bickel 1991 or Croft 2001). However, this may
not solve the problem once and forever because ultimately, we can base the definition
on any property we want (e.g. non-assertion, finiteness, tense scope, extraction possibilities, etc.) and always run into the same problem when analyzing other languages:
if we define cosubordination without a scope constraint, the term would no longer
capture the distinctive properties of cosubordination in Amele and Swahili, and we
would miss again the overall similarity between these structures and those in the
other languages. Similarly, if we define a notion like subordination via non-assertion
(Cristofaro 2003), we will run into structures that look very similar to subordinate
structures, yet are asserted (e.g. with imperatives in an although-clause such This is true,
although dont expect examples!, cf. Green 1976; Lakoff 1984; Takahashi 2008, among
many others). Any property that is picked as definitional will favor one type and make
it the model for others. The fundamental problem is that there is no non-arbitrary
choice (cf. Croft 2001): should Amele provide the model, or Tauya, or English? Any
answer seems wrong. Finally, one could follow Lazards (2006) or Haspelmaths (2007)
suggestions and keep issues of language-specific analysis free of comparative notions:
we could set up entities like Amele e-construction or Nepali era-construction,
describe their properties and leave it to typological research to compare these entities
on the basis of some comparative concept like cosubordinationcomp, defined without
regard to the language-specific details and independently of their analysis. While this
may seem to solve the problem of how to classify language-specific constructions, it
does not address (and is not intended to address) the comparative problem of just what
the right definition of the comparative concept might be. Yet comparative notions
are often of critical help in language-specific analyses and when positing constructional entities, and it is one of the great steps of progress that typological and theoretical knowledge increasingly informs such descriptive work: for example, without
the publication of Foley & Van Valin (1984), the issue of scope properties would have
had little chance of being addressed in descriptive grammars. In fact, as many early
descriptions in the American structuralist tradition testify, any attempt at describing
languages purely in their own terms risks missing important analytical questions.
Moreover, unless the analysis is coupled with an explicit metalanguage of description,
the range of properties that are taken to be relevant for a given construction remains
arbitrary (or even opportunistic, as Croft (2001) puts it). To most empirically-minded
linguists, however, such a descriptive metalanguage is the more appealing the more
it is informed by knowledge of typological variation and not just by meta-principles
such as elegance in theory design. But then, we are back to the problem of defining the
right comparative concepts for analyzing language-specific structures.
Underlying all these solutions and their problems is a general attempt to try and
reduce the observed diversity here between languages like Amele and Swahili on the
one hand, and languages like Belhare, Nepali or Tauya on the other hand to one or
two universal structures or comparative concepts. In this paper, I propose an alternative, based on standard methods used in other disciplines when confronted with
diversity: this alternative consists in measuring instead of reducing diversity. I describe
the general ideas behind this in Section2. In Section3 I review some of the key structural properties that lead to the diversity in clause linkage noted above and in general.
Section4 presents ways in which the diversity can be measured and discusses crosslinguistic and possibly universal patterns emerging from this, based on a pilot database
of 69 constructions from 24 languages.
2. Multivariate analysis
When confronted with diversity, most other disciplines try to measure it. The key to
making this possible is that structural similarity needs to be understood as what it is:
structures S1 and S2 are similar iff they are identical in some variables (also known
as properties, parameters, or features) A...M, but different in other variables N...Z.
Therefore, we need to decompose terms like cosubordination (or subject, word,
sentence, antipassive etc.) into sets of variables that capture all dimensions A...Z in
which any given pair of structures may be identical or different whether between languages or within languages. I call such decompositions multivariate analysis, extending the use of the term from its statistical meaning of simultaneously analyzing entire
sets of variables to the development of these sets itself.3
The set of variables must be large enough so as to capture all known variation,
and in principle could extend to the minutest phonetic differences. Obviously, practical choices in research interests and time budgeting dictate upper limits, as always.
The choice of variables is determined by similar questions of research planning, but
if one subscribes to standard principles of economy in theory design, variables need
. For an earlier proposal moving in a similar direction, but assuming pre-defined ideal
types, see Lehmann (1988). For more general discussion of the multivariate approach, see
Bickel (2007).
Balthasar Bickel
to be logically independent of each other and, if one subscribes to empiricist principles of theory design, variables should also be developed inductively (e.g. using the
autotyp method of Bickel & Nichols 2002) rather than exclusively on the basis of a
priori assumptions about the nature of grammar. Ideally, the set of variables is large
enough to capture the full range of known diversity, so the logically possible combinations of all levels in this set allow a precise description of each known structure. Full
development of such a system of variables is clearly a long-term goal and must proceed
in tandem with progress in the detailed analysis of many different languages.
Two kinds of variables are relevant in multivariate typology: structural and denotational variables. Structural variables are defined in syntactic or semantic terms and their
levels capture specific syntactic or semantic properties, e.g. properties like conjunct
scope of illocutionary operators or conditional. Denotational variables are defined
in terms of extra-linguistic stimuli or contexts, to which language-specific structures
may respond in the same or in different ways: e.g. a narrative context may elicit one
kind of structure in one language and another structure in another language. In the
following, I will limit myself to structural variables. Within these, I will furthermore
mostly concentrate on aspects of syntax and issues of semantic scope. I have nothing
to say in this paper on the semantic relations between propositions or usage patterns,
although there is no doubt that the relevant variables are important for understanding
the distribution of clause linkage structures in the languages of the world.
3. S
ome variables in clause linkage, with particular attention
to adjoined structures
In this section, I review some of the better-known ways in which clause linkage structures differ from each other within and across languages. To keep the scope of the
discussion manageable, I focus on adjoined clauses and disregard clauses that are
subcategorized by main clauses (i.e. that are embedded in the sense of complement
clauses). The results of the discussion are summarized in Section4, where the variables
are applied to a pilot database.
3.1 Illocutionary scope

As noted in the introductory discussion, a key variable is the scope of illocutionary operators4 like question or imperative markers that occur in the main clause.
. Throughout this chapter, I use the term operator for any grammatical category that takes
scope over some other linguistic object. Thus, illocutionary force markers are operators, while
for example person markers are not.
In principle, each of these operators could have their own scope properties, but for
present purposes I simplify matters by collapsing them.
A first possibility is for structures not to impose any syntactic constraint. This
was illustrated by the introductory examples in (3) through (5). If a structure constrains the scope, there appear to be at least three options. First, in some structures, the
scope is always conjunct. This is what was illustrated by the Amele and Swahili data
in (1) and (2). Second, some clause linkage structures obligatorily impose disjunct
scope so that only one of the linked clauses can be in the scope of the illocutionary
marker. This behavior is traditionally associated with the notion of subordination
(e.g. Tikkanen 1995) and is instantiated here by an example from Belhare (and its
English translation):

(6) Belhare
ne-e
yu-a=naa
mundhupt-he i?
dem-loc [3sS]sit-sbjv.pst=top [3sS]chat-pst q
When he was here, did he say something? (or was he silent?)
or Did he say something when he was here? (or later only?)
but not Was he here, and did he say something?
(sub)
In structures involving the topic marker =naa in Belhare (or when in English), only
one of the linked clauses can be questioned; the other clause must be interpreted
as presupposed.
A further option is for main clause illocutionary scope to be limited to its clause
Icall this local scope in the following. Structures with local scope often correspond
to what is traditionally called coordination, but translations do not always involve
coordination in English and linked clauses may show asymmetrical dependencies that
one would normally take to indicate subordination in traditional terms. Obviously, the
variable of illocutionary scope is logically independent of variables having to do with
symmetry or dependency. The following structure in Amele involves two independent
clauses, linked by the conjunction gba but:

(7) Amele
(Roberts 1987)
ho busale-i-a
gba dana age gbo-i-ga
fo?
pig run.away-3s-t.pst but man 3p hit-3p-t.pst q
The pig ran away but did the men kill it?
(but)
In Tauya, local scope is associated with topic clauses, more akin to adverbial subordination in English:

(8) Tauya
(MacDonald 1990)
nen mei momune-i-nani=ra pofei-ti nen=tu-e=nae?

3p here sit-3p-ass=top
talk-conj 3p=give-2=polar.q
They sat here and/but did you talk to them?
or Since they sat here, did you talk to them?
(topic)
Balthasar Bickel
In some languages, local scope optionally extends to dependent clauses, a pattern

which will be called extensible scope (as opposed to local, conjunct, and disjunct
scope). For the structure in the following example from Usan, Reesink (1987: 239)
notes that it is not automatically clear whether the interrogative marked on the final
verb also applies to the preceding predicate:

(9) Usan
(Trans-New Guinea: Madang; Papua New Guinea, Reesink 1987)
munon iyau wrm-a

um-orei
iyo?
man dog him.hit-3sDS die-3sREM.PST q
Did the man hit the dog and it died?
(chain.swr)
A similar pattern has recently been noted by Molochieva (2008) for the NakhDaghestanian language Chechen and by Forker (2009) for Hinuq, another language of
the same family. In Chechen, clause linkage with the general-purpose converb in -na
allows main scope illocutionary markers to scope over the main clause or over both
the main clause and the dependent clause, but not over the dependent clause alone:
(10) Chechen

(Nakh-Daghestanian; Good 2003)
Maliika tyka-na=a j-ax-na ca-j-ea-r=ii?

m.(j).nom store-dat=ss j-go-cvb home-j-come-w.pst=q
Did Malika come home, having gone to the store?
(presupposing having gone to the store)
or Malika went to the store, but did she come home ?
or Did Malika go to the store and come home?
but not: Did Malika go to the store, having come home?
(chain)
Structures with extensible scope are almost as flexible as structures with unconstrained
scope, except that extensible scope requires the main clause to always fall into the
scope of main clause illocutionary operators.
3.2 Scope of negation, tense and other main clause operators

While the scope of illocutionary markers is particularly prominent in differentiating clause
linkage types, the scope of other main clause operators can also provide fundamental distinctions, and each of these can in principle be subject to their own scope constraints.
One important such operator is negation. Turkish converb clauses in -Ip, for
example, have been noted to have conjunct negation scope, while converb clauses in
-Ip de or -ince have local scope, i.e. are limited to the main clause:
(11) Turkish
a.

(Johanson 1995)
ev-e
gel-ip
el-ler-in-i
yka-ma-d.
(chain)
house-dat come-seq hand-p-3sPOSS-acc wash-neg-pst[3s]
He did not come home and [did not] wash his hands.
b. ev-e
gel-ip
de el-ler-in-i

house-dat come-seq and hand-p-3sPOSS-acc

yka-ma-d.
wash-neg-pt[3s]
When he came home, he did not wash his hands.
c.

ev-e
gel-ince
el-ler-in-i
yka-ma-d.
house-dat come-when hand-p-3sPOSS-acc wash-neg-pst[3s]
When he came home, he did not wash his hands.
As the data in (11bc) show, -Ip de and -ince block the scope of the negation marker
(-ma) in the main clause. This contrasts with -Ip alone in (11a), where the scope of the
negation marker is conjunct.
Converbs in Puma, a Kiranti language, impose disjunct scope:
(12) Puma

(Sino-Tibetan: Kiranti; Schackow et al. in press)
gaph mu-so kama p%-mu-e-min

talk do-sim work neg-do-1p-neg
Chatting, we do not work.
or We work without talking.
but not: We neither work nor talk.
Burshaski converb constructions do not constrain the scope of negation, and sentences like the following can be understood with either conjunct or disjunct scope:
(13) Burshaski
(Tikkanen 1995)
khiruman sis
majt-ar
n-u-nin
some
people mosque-dat cvb-3hum.pSBJ[-go]-cvb
nimaz ay--=-am.
prayer neg-do-dur=aux-3hum.pSBJ
(chain)
Having gone to the mosque some people do not pray (but read)
or Some people do not pray after getting to the mosque (but after getting up)
or Some people do not go the mosque and do not pray.
While I am not aware of a structure with negation having extensible scope of the kind
discussed earlier for illocutionary force, negation operators allow for yet another possibility, not attested for illocutionary force: some clause linkage structures require the
scope to extend exclusively to the dependent clause. This is well-known from studies
of complement clauses, where the phenomenon has been dubbed neg-transport (e.g.
Horn 1989) and is exemplified by expressions like I dont think that p which are regularly interpreted as I think that p is not the case. The following examples illustrate this
Balthasar Bickel
for adjoined structures in Belhare (in contrast to the converbs in the related language
Puma, which imposes disjunct scope, as in (12)):
(14) Belhare
(Bickel 1993)
a.

taw-a=lo
kam
n-cokg-att-u-n.
(com)
[3sS-]come-pst.sbjv=com work[nom] neg[-3sA-]do-pst-3sP-neg
He didnt keep working until he came.
(i.e. he worked but not until he came.)
b.

ya- his-sa
la--um--ni.
around- look-cvb walk-[3nsS]neg-walk-npst-neg
He didnt look around while walking.
(i.e. he walked without looking around)
(cvb)
In this case, the negation marker has scope over the dependent clause; the main clause
must be interpreted as affirmative. This type of transported scope is so far unattested
with other operator categories. Indeed, with regard to illocutionary operators, Belhare
lo(k)-constructions have local scope, as illustrated by the following examples:
(15) Belhare
a.

tupt-u-=lo
khem-t-u-.
understand[sbjv]-3sP-1sA=com listen-npst-3sP-1sA
Ill listen so that I understand.
b. tupt-u-=lo
pr%sta ka-lur-a!

understand[sbjv]-3sP-1sA=com clear 1sP-tell-sIMP

Talk to me clearly so that I understand!
(com).
(com)
Here, the difference between the illocutionary markers in the main clause has no
impact on the interpretation of the dependent clause. The fact that the same constructions show different scope behavior under negation and under illocutionary operators
confirms that these are independent typological variables.
Illocutionary force and negation are the best known operators whose scope beha
vior differentiates between clause linkage types. Other operators are not well-studied
in this regard. One exception is tense and status (realis/irrealis) markers. Especially
for Papuan languages, the scope behavior of these markers has often been noted to
differ across clause linkage structures (Foley & Van Valin 1984; Foley 1986). Amele
e-structures illustrate conjunct tense scope (cf (1), which shows conjunct illocutionary
scope of the same construction):
(16) Amele
(Roberts 1988)
ho busale-e-b
dana age gbo-ig-a.
pig run.out-ds-3s man 3p hit-3p-t.pst
The pig ran out and the men killed it.
(chain)
b. ho busale-e-b
dana age gbo-gbag-an.

pig run.out-ds-3s man 3p hit-3p-fut

The pig will run out (not: ran out) and the men will kill it.
(chain)
a.

Here, the temporal interpretation of the dependent clause strictly depends on the tense
choice in the main clause. This contrasts with Belhare ki-clauses, where the dependent
clause is not necessarily within the scope of the main clause tense marker. The scope is
primarily limited to the main clause, but can optionally be extended into the dependent
clause. Accordingly, a sentence like the following can have different interpretations,
depending on context:
(17) Belhare

khimm-e n-ta-ch-u
ki mun-n-dhup-chi.
house-loc 3nsA-reach[sbjv]-d-3sP seq chat-3nsS-chat[npst]-d
They will reach home and chat.
or When they reach home, theyll chat.
or They reached home and now they will chat.
(chain)
For a similar construction in another Sino-Tibetan language of Nepal, Chantyal,

Noonan (1999:412) notes: Because of the context in which this was said, we understand the converb to be referring to a time future relative to the moment of speaking;
however, had the sentence been uttered after he had arrived in his home village, but
before he met his father, we would understand the clause as referring to an event past
relative to the moment of speaking:
(18) Chantyal

(Sino-Tibetan: Tamangic; Noonan 1999)
na tBem- Bya-sir bw-ra

dBo-wa
Bin.
I house-loc go-seq father-dat meet-nom be.npst
I will go home and see my father.
or Having gone home, I am to see my father.
(chain)
Tense markers whose scope is not conjunct but extensible are also characteristic of some
Indo-European participle constructions. Participles in Ancient Greek, for example, are
attested both within and outside the scope of main clause tenses:
(19) Ancient Greek
a.

pollakho d me
epskhe
often
ptcl 1sACC stop.3sIMPERFECT
lgo-nt-a
metax.
talk-ipfv.act.ptcp-acc.s in.the.middle
(part.coni.)
[The oracle] has often stopped me when I was in the middle of talking.
(Plat. Apol. 40b)
b. eg`
er
hs e epist-men-os.
(part.coni.)

1sNOM speak.1sFUT ptcl well understand-ipfv.med.ptcp-nom.s

I will speak out because I understand it well.
(Herod. Hist. IX 42)
In (19a), the time reference of the participial clause coincides with that of the main
clause; in (19b), by contrast, the participial clause makes a present tense assertion
while the main clause refers to the future.
Balthasar Bickel
Tense markers with local scope, i.e. scope that is restricted to the main clause,
are also attested. This is mostly the case when both the dependent and the main
clause are marked for their own tense. The other scope types observed with illocutionary force and negation, viz. disjunct or transported scope, never seem to occur
with tense markers.
3.3 Finiteness and marking possibilities

At first sight, it seems plausible to analyze every kind of non-local scope constraint as
resulting from the fact that a dependent clause is not fully finite, i.e. lacks some operator specification of its own. However, as we saw in the preceding, nonfinite dependent
clauses do not always lead to scope constraints (cf the data in (18) and (19) above).
And, finiteness is determined not only by operator categories (i.e. categories with a
certain scope behavior) but also by referential categories like person marking (which
does not define a specific scope over other linguistic objects). Thus, the variable of
finiteness is logically independent from variables of scope.
Finiteness is traditionally defined by the range of categories that can be expressed
by inflectional forms in those main clause types that allow the maximal number of
categories. Dependent clauses can deviate from this by allowing more or less categories. If there are less categories, the verb form heading the dependent clause is called
nonfinite. The case where a dependent clause expresses more categories than a main
clause does not fit any traditional term (but see the Wambule data in (30) below for
an example).
Finiteness is relevant for clause linkage on its own, but there are two additional
aspects: first, dependent clauses vary as to exactly which categories can be marked and
which categories are not allowed. To the extent that categories are realized by inflectional morphology, this follows from finiteness; however, many such categories, e.g.
tense or illocutionary force, can also be realized by independent elements (words, clitics, particles) and therefore, possibilities here can be independent of finiteness. Second,
some linkage types require that two clauses contain the same degree of finiteness and
the same range of expressible categories, while other linkage types do not require this.
In the following, I concentrate on the first issue, i.e. on marking possibilities of dependent clauses in their own right. The second issue will be taken up in Section3.4.
3.3.1 Illocutionary force marking

An issue that is often critical is whether or not dependent clauses allow the marking of illocutionary force operators. In many languages conjunct illocutionary scope
entails that illocutionary markers cannot themselves appear on dependent clauses.
This is the case for example in the Amele constructions that were first illustrated by
(1) and (16):
(20) Amele
(Roberts 1988)
*ho busale-e-b
fo dana age gbo-ig-a?
pig run.out-ds-3s q man 3p hit-3p-t.pst
Did the pig run out and did the men kill it?
(chain)
But such a constraint on illocutionary marking is not universal. Turkish dependent

clauses in -Ip show conjunct scope when the illocutionary marker appears on the finite
final clause (as in 21a), but nothing bars such markers from appearing on the dependent clause (as in 21b):
(21) Turkish
(Bickel 1991; Johanson 1995)
a.

gel-ip
ana
bir ey syle-di
mi?
come-cvb 2sDAT one thing say-pst[3s] q
Did he come and say something?
(chain)
b.

otur-up mu konu-tu-lar?
sit-cvb q speak-pst-[3]p
Did they speak (while they were) sitting? (asking about the
circumstance of speaking)
(chain)
And similar possibilities are known from the Papuan language Fore:
(22) Fore
a.

(Trans New Guinea: Kainantu-Gorokan; Scott 1978)
kana-a-k-t
a-ka-us=.
come-3sPRES.ds-dep-1dAS 3sP-see-1dA=q
Is he coming and we see it?
(chain)
b. na-m-e-g-nt=
na-ku-w-e.
(chain)

1sP-give-2sFUT.ds-dep-1sAS=imp eat-fut-1s-decl

Give me something and then I will eat it. (i.e. Give me something to eat!)
The sentence in (22a) illustrates the fact that main clause question markers have conjunct scope, while (22b) shows that this basic structure does not ban the occurrence of
at least imperative markers on the dependent clause.
A number of languages allow illocutionary marking on dependent clauses only if
it matches the marking in the main clause. In Belhare, for example, ki-clauses (of the
kind illustrated before by (3) and (17)) can be marked by an imperative, but only if the
same mood is also marked on the main clause:
(23) Belhare
caw-a ki khar-a!
eat-imp seq go-imp
Eat and go!
(chain)
In Chechen, illocutionary marking on the dependent clause is possible in some but

not other structures. For example, the marking is blocked in clauses marked by the
sequential converb in -na, or the concessive converb in -na-shehw that is derived from
Balthasar Bickel
this. In each case, the interrogative clitic (=ii) must occur on the main clause (24a,c)
and cannot be attached to the dependent clause (24b,d):
(24) Chechen
(Good 2003; Molochieva 2008)
a.

Maliika tyka-na=a
j-ax-na ca-j-ea-r=ii?
m.(j).nom store-dat=SS j-go-cvb home-j-come-w.pst=q
Did Malika come home, having gone to the store?
(presupposing having gone to the store)
or Did Malika go to the store and come home?
but not: Did Malika go to the store, having come home?
(chain)
b.

*Maliika tyka-na=a
j-ax-na=ii ca-j-ea-r
m.(j).nom store-dat=SS j-go-cvb=q home-j-come-w.pst
Intended: Did Malika come home having gone to the store?
(presupposing Malika came home)
(chain)
c.

ahw
toex-na-shehw j-axa-r=ii
2sg.erg hit-cvb-conc j-go-w.pst=q
Zaara
tyka-na?
Z.(j).nom store-dat
Did Zara go to the store, even though you hit her?

not Did Zara go to the store, and is it even the case that you hit her?
(chain)
d. *ahw
toex-na-shehw=ii j-axa-ra

2sg.erg hit-cvb-conc=q j-go-w.pst

Zaara
tyka-na?
Z.(j).nom store-dat
Intended: Did Zara go to the store, even though you hit her?
(chain)
But illocutionary marking is allowed in clauses headed by the temporal converb in -cha:
(25) Chechen
(Molochieva 2008)
a.

Muusa
ca
v-ea-cha,
naan-na
m.(V).nom home V-come-when mother-dat
xaza
xiiti-r=ii?
beautiful seem-w.pst=q
Was (his) mother happy, when Musa came home?

(presupposing either mother was happy or Musa came home)
not Did Musa come and was his mother happy?
(cvb.temp)
b. Muusa
ca
v-ea-ch=ii,
naan-na

m.(V).nom home V-come-when=q mother-dat

xaza
xiiti-ra?
beautiful seem-w.pst
Was (his) mother happy, when Musa came home?

(presupposing mother was happy)
(cvb.temp)
The possibility of illocutionary marking is again independent of scope behavior: both

the concessive and the temporal converb are limited to structures with disjunct scope,
as shown by (24c) and (25a). Yet the concessive converb bans illocutionary marking
(24d), while the temporal converb doesnt (25b). This shows that one cannot define
a universally applicable notion of subordination which would entail both disjunct
scope and a ban on illocutionary marking on dependent clauses. Instead, scope behavior and the possibility of illocutionary marking are independent variables.
3.3.2 Tense and status marking

Another type of operator that is often relevant for distinguishing clause linkage types is
tense or status (realis/irrealis) markers. When discussing tense scope in Section3.2, we
noted that a lack of tense markers in the dependent clause does not necessarily entail
conjunct scope. In other languages, such as Amele, however, a ban on tense marking
does go together with conjunct tense scope (Roberts 1988). This suggests that, like in
the case of illocutionary force, scope and marking possibilities are independent variables in the typology of tense as well.
In some languages, tense marking in dependent clauses is possible, but at the same
time subject to constraints on tense choices across clauses (in the spirit of Latin consecutio temporum rules). One example is Belhare lo(k)-constructions of the kind illustrated by (14a) and (15). The dependent clause must be in the subjunctive, but within
this, it can be either in the past (-a) or nonpast (zero-marked). The choice must match
the tense choice in the main clause, however (Bickel 1996:92f):
(26) Belhare
a.

(Bickel 1996)
nam kus-a=lo
kam cog-he-a (*cou--a)
sun [3sS]set-sbjv.pst=com work do-pst-1sS (do-npst-1sS)
I worked till the sun set.
b. nam ku=lo
kam cou--a
(*cog-he-a)

sun [3sS]set[sbjv]=com work do-npst-1sS (do-pst-1sS)

I will work till the sun sets.
(com)
(com)
Another example comes from Hua, a Trans-New Guinea language from the Papuan
highlands (and further Papuan examples are discussed by Foley in this volume):
(27) Hua
a.

(Trans-New Guinea: Kainantu-Gorokan; Haiman 1980:421)
fu=mo d-mi-sa-ga-da
u-gu-e.
pig=top 1sP-give-fut-3pDS-1sAS go-fut-1sDECL
They will give me pork and then I will go.
b. *fu=mo d-mi-sa-ga-da
u-e.

pig=top 1sP-give-fut-3pDS-1sAS go[nfut]-1sDECL
Intended: They will give me pork and so I went,
i.e. I went because they will give me pork.
(ds-chain)
(ds-chain)
Balthasar Bickel
In one type of clause linkage, marked by an agreement paradigm that also signals
switch-reference (here in the form of the third person, different subject marker -ga),
tense marking is allowed in the dependent clause only if it matches the tense marking of the main clause. This is the case in (27a), but not in (27b). In another type of
clause linkage in Hua, marked by an agreement paradigm without switch-reference
coding (here in form of the third person plural marker -ma), no such constraint is
imposed. Dependent clauses freely allow tense marking, regardless of main clause
tense choices:
(28) Hua

fu=mo d-mi-ga-ma-da
u-e.
(topic)
pig=top 1sP-give-fut-3pSUB-1sAS go[nfut]-1sDECL
They will give me pork and so I went, i.e. I went because they will give
me pork.
Unconstrained tense marking in dependent clauses is traditionally associated with

either coordination and subordination, while the presence of constraints has sometimes been suggested as diagnostic of chaining and converb constructions (e.g. Foley
1986). However, in a number of languages, including languages from the Trans-New
Guinea family, structures traditionally classified as chaining freely allow tense marking. A case in point is Korafe, where the different subject marker in the following
examples establish past tense reference in the dependent clause. The main clause can
be independently marked as present or past tense:
(29) Korafe
a.

(Trans-New Guinea: Binanderean, Papua New Guinea; Farr 1999)
mut-eno
er-ira-re.
(chain)
give.1s-seq.realis.1sDS ipfv-go.dur.prs.3sIND-current.relevance
I gave it and he is currently going.
b. mut-eno
i-sira.

give.1s-seq.realis.1sDS go.dur-f.pst.3sIND

I gave it and he went (two or more days ago)
(chain)
In Wambule, a language from the same Sino-Tibetan branch as Belhare, tense marking
is allowed exclusively on chained, dependent clauses. Main clauses, by contrast, cannot
be marked for tense.5
. This is likely to result from a regular process of nominalizing main clauses, following a
general trend characteristic of the Sino-Tibetan family at large (Bickel 1999b). The nominalizer was eventually reanalyzed as an illocutionary marker, with some kind of assertive
or affirmative function.
(30) Wambule
a.

(Sino-Tibetan: Kiranti; Nepal; Opgenort 2004)
saiso
am kam pa-si tum-nu-ma-kho lwa-nu-mei. (chain)
yesterday dem work do-inf finish-2s-pst-seq go-2s-ass
You finished doing this work yesterday and you went.
b. na
hep
ja-ma-k
tya

previously cooked.grain eat[1s]-pst-seq from.now

iskul di--m.
school move-1s-ass
I ate cooked grain before, and now I will go to school
(chain)
Both sentences in (30) contain dependent clauses marked as past tense, but in (30a),
this combines with past tense reference in the main clause, while in (30b), the adverb
tya from now on suggests future tense reference in the main clause.
3.4 Categorical symmetry

Issues of finiteness and marking possibilities are not only relevant with regard to the
dependent clause on its own, but also in relation to the marking possibilities of the
main clause. The core issue is wether or not the dependent and the main clause allow
the same range of categories, i.e. whether the construction is symmetrical or asymmetrical in this regard. The difference between symmetrical and asymmetrical clause
linkage was first emphasized by Haiman (1980, 1985a,b) in work on Hua and other
Papuan languages; a related concept of balanced vs. deranked linkage was developed
by Stassen (1985) and Cristofaro (1998, 2003).
Hua constructions like those in (27), for example, are asymmetrical in that the
dependent clause allows anticipatory subject marking, while the main clause does
not allow this, but instead allows marking of illocutionary force. This contrasts with
symmetrical constructions such as those in -roh and -reh, which signal two iterative
events in alternation:
(31) Hua
(Haiman 1980)
p-mi-roh
de-reh.
(alter.iter)
3pO-give-1sALTER.ITER eat-3pALTER.ITER
I gave them some, they ate; I gave them some more: they ate; and so on.
Note that the verb forms in this example are not fully finite in Hua: they lack tense and
illocutionary force marking, and the agreement paradigm is greatly reduced (Haiman
1980). What matters for symmetrical linkage is that conjuncts allow the same range of
categories to be marked, not whether they are finite or not.
Balthasar Bickel
The contrast between symmetrical and asymmetrical constructions is often associated with clause linkage types based on conjunctions vs. affixes. In Amele, for example,
constructions involving dependent verb morphology and switch-reference of the kind
illustrated by (1), (16), or (20) is asymmetrical. Conjunctions like fo or or gba but, by
contrast, require symmetry in the range of categories expressed (also cf (7)):
(32) Amele
a.

(Roberts 1987, 1988)
Fred um
ho-i-an
fo gbee gbila ho-i-a?
f.
yesterday come-3s-y.pst or neg today come-3s-t.pst
Did Fred come yesterday or today?
(or)
b. ud weg-i-me-ig
fo eed
weg-i-me-ig

sago weave-pred-ss-2p or bamboo weave-pred-ss-2p

fo fal-do-ig-a!
or fence-3sP-2sA-imp
Weave some sago or bamboo and fence it!
c.

*ija ja hudo-o-min gba ugba sab mane-i-a

Is fire open-ds-1s but 3s
food cook-1s-t.pst
(but)
d. ija ja hud-ig-a
gba ugba sab mane-i-a.

Is fire open-1s-t.pst but 3s
food cook-1s-t.pst

I lit the fire but she cooked the food.
(but)
(chain)
In (32a), both conjuncts are finite and tense-marked, in (32b) both conjuncts are nonfinite and marked for same subject, enclosed by a repetition of fo. The ungrammatical
sentence in (32c) shows that the conjunction gba is not compatible with an asymmetrical choice of categories, e.g. switch-reference marking in one conjunct but tense
marking in the other. Only a symmetrical choice, as in (32d), is accepted.
While such an association of symmetry with conjunctions is common, it is not
universal. In Belhare, for example, some conjunctions license asymmetrical, while others require symmetrical category choices. The conjunction ki allows the subjunctive
category, as shown by (25b) above a possibility that is not given for main clauses. This
contrasts with the (enclitic) conjunction cha, which is used in a symmetrical constructions for expressing alternating events. In this case, both conjuncts must include verb
forms from the same paradigm:
(33) Belhare
ka=na
ten-he-=cha
phed-he-=cha.
1sNOM=top hit-pst-1sA=add drive.off-pst-1sA=add
I hit him and drove him off.
(and)
Asymmetrical clause linkage allows the expression of different ranges of categories.

What is still required, however, is that conjuncts are of the same type, e.g. both verbheaded clauses. This is different in what I call here constraint-free linkage, where it
is not required that the part of speech of conjuncts be matched. This is typical for
sentential topic clauses (Marchese 1977; Haiman 1978; Bickel 1993, 1998, 1999a), such
as those of Usan:
(34) Usan
a.

(Reesink 1987:283ff)
munon iyau wrm-or

eng um-orei
iyo?
man dog 3sP.hit-3sF.pst top die-3sF.pst q
Given that the man hit the dog, did it die?
b. yr-b eng ye-nipat

ur
big-r.

come-ss top 1sP-step.over.ss money put-pIMP

If you come, step over me and put your money (in the basket).
c.

(topic)
(topic)
munon eng, wonou man soau

is-orei.
man top 3sPOSS garden landslide go.down-3sF.pst
As for the man, his garden went down in the landslide.
In (34a), eng marks a finite clause as providing the background against which the main
clause question is to be understood. In (34b), the same conjunction marks a samesubject converb as providing the relevant background, leading, as is often the case with
topic clauses (Haiman 1978), to an interpretation as a conditional. In (34c), finally, the
framework is provided by a bare noun (munon man), in a typical topic construction.
There is no reason to assume that the semantic contribution of the conjunction varies
across these three cases; in all instances the conjunction defines the framework within
which the rest of the clause is to be interpreted, fairly close to Chafes (1976) definition of a topic as a spatial, temporal, or individual framework within which the main
predication holds.
3.5 Focus or questions in linked clauses and extraction possibilities

In many languages some types of linked clauses ban the occurrence of question words.
It is sometimes claimed that such constraints can be derived from general extraction
constraints (Ross 1967; Chomsky 1973) or from information structure constraints
(Foley & Van Valin 1984; Van Valin 1995), and that on either account, the constraints
differentiate between universally defined types of clause linkage (coordination vs. subordination). However, it seems best to treat question and focus possibilities as a variable that is independent of extraction possibilities, and not to tie any of these to other
variables of clause linkage such as illocutionary scope behavior. The cross-linguistic
variation is just too big for any reductionist approach. There are three pieces of evidence for keeping the variables distinct.
First, as argued by Foley & Van Valin (1984) and Van Valin (1995), a ban on questions in dependent clauses cannot be reduced to extraction constraints because such
a ban may obtain even without extraction, i.e. where question words remain in situ.
Balthasar Bickel
Foley & Van Valin (1984) and Van Valin (1995) demonstrate this with data from the
Siouan language Lakhota, and a similar situation obtains in Tauya:
(35) Tauya
a.

(MacDonald 1990)
ne-ni we
tu-a-te
yau-i=ne?
(chain)
3s-erg who[nom] [3sP]give-3s-ds [3sP]see-3p=parametrical.q
Who did he give it to? and they saw him
(Who did he give it to when they saw him?)
b. *we
mei fofe-a-nani=ra

who[nom] here come-3s-ass=top

Intended: *if who did come here ...
(topic)
Example (35a) is based on the same chaining construction that was illustrated earlier
in (5). In these constructions, in situ question words are possible inside the dependent
clause (headed by tuate he gave it to someone and ). This is in minimal opposition
to topic clause constructions of the kind observed earlier in (8) and illustrated here
by (35b): in these constructions, the dependent clause does not allow question words.
This constraint is imposed by the type of clause linkage, and does not result from any
extraction constraint since question words are never extracted in the language.
Second, the domain of possible extraction sites may be different from the domain
that defines where question words can appear, and this suggests that these are independent variables. In Chechen, for example, converbs in -na and -alc allow in situ
question formation, just like their Tauya counterparts in (35a):6
(36) Chechen
a.

(Good 2003)
Maliika hu iec-na ca
j-ea-ra?
m.(j).nom what buy-cvb house j-come-w.pst
What did Malika buy and came home?
(chain)
b. mila
ca
j-all-alc
Ahwmad irs

who.nom home j-come-until a.nom happy

d-ol-ush v-a-ra?
d-be-cvb v-aux-wpst
Ahmed was happy until who came home?
(cvb)
But the same constructions block extraction in the form of relative clauses:
(37) Chechen
a.

Zaara
koch
ec-na
ca
j-ea-ra.
Z.(j).nom dress.nom buy-cvb home j-come-w.pst
Zara bought a dress and came home.
(Molochieva 2008)
(chain)
. and since they also allow conjunct scope of illocutionary force operators, as shown in (10),
the construction qualifies as cosubordinate in Role and Reference Grammar (cf Good 2003).
b. *Zaara(-s) ec-na
ca
j-ea-cha
j-olu
koch

Z.(j).(-erg) buy-cvb home j-come-cvb j-aux.ptcp dress

Intended: the dress that Zara bought and came home.
(chain)
This is in direct contrast to Tauya, where the switch-reference marked dependent

clauses illustrated by (35a) above do allow extraction in the form of relative clauses:
(38) Tauya
a.

(MacDonald 1990)
e fenaa-ni
ya-tu-a-te
te-pa yate-e-a.
dem woman-erg 1sP-give-3s-ds get-ss go-1s-decl
That woman gave it to me and I took it away.
b. ya-tu-a-te
te-pa yate-e-na
fenaa

1sP-give-3s-ds get-ss go-1s-nmlz woman

the woman who gave it to me and I took it away
(chain)
(chain)
Example (38b) is a relative construction based on (38a), where the sole argument of
the dependent clause (fenaa the woman) is extracted as the head.
Third, the possibilities of in situ questions or focus formation are independent of
other properties of clause linkage. For example, while Tauya bans in situ questions in
conditional clauses (as noted in (35b) above), other languages allow this in otherwise
very similar clause linkage constructions. This is so, for example, in Amele and Belhare
conditional clauses as well as in when and while clauses:
(39) Amele
ugba sab man-i
j-igi-an
ja in hude-e-b fi.
s/he food fry-pred eat-3s-fut fire who open-ds-3s if
She will cook and eat if who will light the fire?
(40) Belhare
a.

sa-a
ya=m-phekt-a-k=naa
who-[s]erg call=3nsA-call-sbjv.pst-2=top
-khatd-at-ni-gak=phe?
neg-go-pst-neg-2=irr
You hadnt gone if who had called you?
(Roberts 1987)
(cond)
(Bickel 1993)
(sub)
b. sa-a
lap-yukt-u=lo
m-phig-he?
(com)

who-[s]erg [3sA-]grab-hold-3sP=com 3nsA-pick-pst[-3sP]

They picked the fruit when who was holding down [the branch] ?
c.

hene yu-sa n-lur-he-ga?

where sit-cvb 3sA-tell-pst-2sP
While where sitting did he tell you?
(cvb)
Direct translations into English typically fail because a question word in English if
and when clauses is interpreted as an echo question. But in the original, these are
genuine questions.
Balthasar Bickel
Structures with no illocutionary scope constraint often freely allow question and
focus words in dependent clauses as well. This is shown by the earlier example from
Tauya in (35a) and the following data from various languages:
(41) Kte
(Trans-New Guinea: Huon, PNG; Schneucker 1962)
ohe wena yu-ha-pire gogo de-ye.

2d where be-sim-2dDS bell
sound-3sn.pst
Where were you when the bell rang?
(chain)
(42) Belhare
a.

(Bickel 1993)
laitar hene lept-he-ga

ki s%lai
am-t-u-ga?
(chain)
lighter where throw-pst-2sA seq matches light-npst-3sP-2sA
Where did you throw the lighter so that you have to use matches?
b. han-na=cha tha n-tog-u-n-ga

ki

2s-erg=add know neg-know-3sP-neg-2sA seq

emgari mun dhup-ka?

how talk talk-2sNPST
Since you dont know it either, how can you talk [about it] ?
(43) Nepali
(chain)
(Clark 1963:169)
tyo alch-le
ke
gar-kana kam-era khncha?
dem lazy.fellow-erg what do-cvb earn-cvb eat-3sNPST
How does that lazy fellow earn a living?
(chain)
In some languages, there are also structures with extensible illocutionary scope that allow
question words. This is the case in Chechen na-constructions, as noted in (37) above.
While these data showed that in situ question (and, thereby, focus) formation is
independent of other clause linkage variables, the following show that the same is true
of the possibilities for question word extraction. Some languages, like English, block
such extraction from adverbial clauses, while other languages allow it. The following
example from Latin shows extraction of an object argument (permulta many) from a
conditional clause (marked by nisi if not) into the pre-clausal position that is generally
used in the language for forming relative clauses:
(44) Latin
permultai
[[quaei
many.n.NOMp rel.n.acc
orator
a natura
nisi
orator.NOMs by nature.ABLs if.not
haberet]
non multum
have.3sPST. sbjv neg much.n.ACCs
a magistro
adiuvaretur]
by teacher.ABLs help.3sPST.sbjv.pass
(Cic, de Or. 1, 126)
many [properties] such that if the orator didnt have them by nature, he couldnt
be much helped by a teacher. (Literally: *many properties [whichi couldnt be
helped by a teacher [if the orator didnt have ti by nature]])
Other languages distribute possibilities in a different way. In Tauya, for example, we

noted in (38) that extraction is possible in switch-reference-marked chaining constructions, but it is not possible with topic clauses (of the kind illustrated in (8) and
(35b) above):
(45) Tauya
(MacDonald 1990:273, 298)

e fanaa yau-e-nani=ra
ne-ni
dem woman [3sP-]see-1-ass=top 3s-erg

wawiya ya-tu-a=a.
(topic)

mango 1sP-give-3s=decl

I saw the woman and she gave me a mango.
b. *yau-e-nani=ra
ne-ni wawiya

[3sP-]see-1-ass=top 3s-erg mango

ya-tu-a-na
fenaa
(topic)

1sP-give-3s-nmlz woman

Intended: the woman who I saw and who gave me a mango.
a.

The preceding examples all involved extraction in the form of relative clauses, but also
the possibilities for afterthought extraction (i.e. a special type of right-dislocation) are
sometimes subject to variation across types of clause linkage. In Belhare, for example,
ki-clauses (as exemplified in 3, 17 and 42) regularly support the extraction of arguments into post-verbal positions, whereas topic clauses marked by naa (cf 6 and 40)
do not:
(46) Belhare
a.

chokt-he
ki,
[3sA-]point.with.finger-pst[3sP] seq
n-celi-a,
2sPOSS-unmarried.female.agnatic-erg
doko-ep=phu
cha-da-he.
basket-loc=rep [3sS-]curl.up-come-pst
When she pointed with her finger at him, your celi, that is,
[he] came and curled up in [her] basket.
(chain)
b. u-tak-a
ya phekt-u=naa

3sPOSS-friend-erg call [3sA-]call-[sbjv]-3sP=top

-khatd-at-ni.
neg-[3sS-]go-pst-neg
When his friend called him, he didnt go.
(sub)
c. *ya m-phekt-u=naa,
u-tak-a,

call [3sA-]call[-sbjv]-3sP=top 3sPOSS-friend-erg

-khatdatni.
neg-[3sS-]go-pst-neg
When they called him, his friend, that is, he didnt go.
(sub)
Balthasar Bickel
As we have seen in (40) above, Belhare naa-clauses allow in situ question formation.
This confirms again that the possibilities of question formation are a variable independent of the variable regulating extraction constraints.
3.6 Focus marking on dependent clauses

In the preceding we concentrated on question and focus formation inside dependent
clauses, but focus markers can of course also occur attached to the edge of such clauses,
with a scope that typically includes the entire clause but does not extend beyond it.
The typical functions of such markers lie in restricting, adding, or contrasting propo
sitions. Especially when tightly integrated into the main clause syntax, dependent
clauses allow such focus markers. This is well attested in many Eurasian converbs, as
amply illustrated by Haspelmath (1995), but it also extends to some structures that
are less integrated into main clauses. This is exemplified by Belhare ki-constructions
(cf. the example in (3) above):
(47) Belhare
cama ca-he
ki=cha
r%ksi u-he.
food [3sS]eat-pst seq=add liquor [3sS]drink-pst
He drinks liquor even after the meal.
(chain)
The possibility of focus marking on a dependent clause seems to be independent of

the possibility of interrogative or focused constituents within the dependent clause.
Belhare ki-clauses allow both focus markers on the clause, as just shown in (47), as well
as question words inside the clause, as shown earlier in (42). But this is not universal:
Chechen na-clauses allow question words inside the clause, as shown in (36a), but the
same clauses do not allow focus markers attached to their edges, regardless of whether
the clause contains a question word or not:
(48) Chechen
a.

(Zarina Molochieva, p.c.)
Maliika
hu iec-na(*=a)
ca
j-ea-ra?
m.(j).nom what buy-cvb(=foc) house j-come-w.pst
b. Maliika tyka-na=a
j-agh-na(*=a) ca
j-ea-ra.

m.nom store-dat=ss j-go-cvb(=foc) home j-come-w.pst

Malika went to the store and then came back home.
Whether or not focus markers are allowed on dependent clauses therefore seems to be
again an independent variable. The possibility of focus marking perhaps depends on
the exact semantics of the relevant focus markers, but in some cases it seems to depend
on the form of the dependent clause. While, as just noted, Chechen na-clauses ban
focus markers, -cha-clauses allow them, although the difference in meaning between
these converbs is minimal:
(49) Chechen
(Good 2003)
Maliika
hu iec-cha=a
ca
j-ea-ra?
m.(j).nom what buy-when=foc house j-come-w.pst
(cvb.temp)
The possibility of purely formal constraints is further illustrated by Burshaski, a language for which Tikkanen (1995:514f) notes that its converbs cannot host any focus
particle. This is in contrast with Indo-Aryan languages of the same region, like Hindi
or Nepali, where such particles freely attach to semantically similar converb clauses.
Especially in work on European languages, it has been suggested that focus
assignment and topic-comment articulation in general constrains the scope of illocutionary force and similar operators (e.g. Jacobs 1984, 1991, and many others since).
Preliminary evidence in Bickel (1993) and Schackow et al. (in press) suggests that this
may not be universal. For example, from all we know, the presence or absence of the
restrictive focus clitic =a does not resolve the scope ambiguity of the illocutionary
force operator in Puma:
(50) Puma
(Schackow et al. in press)
risiwa=cha
mu-so(=a)
shamanic.rhythm[nom]=add do-cvb(=foc)
m%-ta-a=ku,
bura-ci?
3pS-come-pst=nmlz old.man-ns[nom]
Did they come and play the drum? (conjunct)
or: Did they play the drum while coming?
(only converbal clause in scope of question)
or: Playing the drum, did they come?
(only main clause clause in scope of question)
While more research is needed to firmly establish the nature of =a as a focus marker,
it seems best to assume for now that choices in focus marking and in illocutionary
scope are in principle independent of each other. If focus and illocutionary scope are
structurally linked in a language, this may necessitate positing distinct structures
for example one structure with focus on the main clause and local illocutionary or
negation scope, as opposed to a construction with focus on both clauses and conjunct
illocutionary or negation scope.7
3.7 Clause position

The position of dependent clauses and main clauses has traditionally been taken as a
key variable in clause linkage types. In many types, the position is fixed, as either initial
. Thanks to Volker Gast for drawing my attention to this possibility.
Balthasar Bickel
or final. An example is the Amele construction that was already illustrated in (1), (16)
or (20). The dependent clause in this construction must always precede the main clause:
(51) Amele
(Roberts 1988)
ho busale-e-b
dana age gbo-ig-a.
(chain)
b. *dana age gbo-ig-a

ho busale-e-b.

man 3p hit-3p-t.pst pig run.out-ds-3s

The pig ran out and the men kill it.
(chain)
a.

The same condition obtains for Usan clauses marked by the conjunction eng (cf. 34
above):
(52) Usan
(Reesink 1987)
wau e-b igo-iner

eng unor
child cry-ss be-3suncertain.fut top mother

mni u-t-i-b-.
yam 3sP-give-s-fut-3s
(topic)
If the child is crying, his mother will give him yam.
Dependent clauses marked by end because, by contrast, can appear both before or
after the main clause, salva veritate:
(53) Usan
(Reesink 1987)
a.

ya itum der
igm-a igo-i
urigerm-a end
rain night come.down be-3sDS be-cess light-3sDS because
irumban sig boru.

mud
very bad
Because it has been raining all night until daybreak, it is very muddy.
b. irumban sig boru, ya ib dr-a

end.

mud
very bad rain big come.down-3sDS because

It is very muddy; thats because a big rain came down.
(cause)
(cause)
While such flexibility as with Usan end-constructions are traditionally associated with
subordination, it can also be observed in some languages with constructions that are
functionally closer to narrative chaining uses. An example is Belhare ki-constructions. In most cases, ki-clauses appear before their main clause (cf 3, 17, 42, 46a, and
47), but they can also follow the main clause, as in the following example:
(54) Belhare
a.

ca-ma=na ca-yau-t-u,
t%r% he-lle
eat-inf=top [3sA-]eat-ipfv-npst-3sP ptcl where-dir
le
kina?
[3sS-]direct seq
It (the cow) is eating, but after having turned towards which direction?
(chain)
b. p%scim i-ba
khat-ke,
lotthi tekap

west one-hum [3sS]go-temp stick hold.on

cou-se
ki
[3sA]do-prf[3sP] seq
He is going west, after having taken hold of the (walking) stick.
(chain)
For many constructions, positional flexibility is limited, however, in that the dependent clause must still be adjacent to the main clause. This is so in Belhare. By contrast, Chechen converbs in -na, which cover a similar range of narrative sequential
uses as Belhare ki-clauses, are not subject to an adjacency constraint. The following
data show some possible salva veritate permutations of two converb clauses and one
main clause:
(55) Chechen
a.

Maliika tyka-na=a j-agh-na zhejna=a

iec-na
m.nom store-dat=ss j-go-cvb book.nom=ss buy-cvb
ca
j-ea-ra.
home j-come-w.pst
(Good 2003)
(chain)
b. Maliika tyka-na=a j-agh-na ca

j-ea-ra

m.nom store-dat=ss j-go-cvb home j-come-w.pst

zhejna=a
iec-na.
book.nom=ss buy-cvb
c.

Maliika ca
j-ea-ra
tyka-na=a
j-agh-na
m.nom home j-come-w.pst store-dat=ss j-go-cvb
zhejna=a
iec-na.
book.nom=ss buy-cvb
Malika went to the store, bought a book, and came back home.
(chain)
(chain)
The example in (55c) shows that a converb clause need not be adjacent to the main
clause that it refers to: here, the sequential converb suffix -na situates the event of book
buying into a direct relation to the event expressed in the main clause (coming home,
here initial), not in the immediately adjacent clause (going to the store, here the second clause). Such behavior is traditionally taken to be associated with subordination
and can often be observed with what translates English adverbial clauses.
3.8 Layer of attachment

Adjoined clauses are not embedded in the sense of being arguments of main clause
predicates. But they can attach to various levels inside a main clause. While more finegrained distinctions are certainly possible, what seems to be the most important difference is between (a) clauses adjoining to a predicate or verb (ad-V), (b) clauses
adjoining to entire clauses (ad-S), or (c), clauses that are adjoined to some higher level
and appear as detached from main clauses in the sense of Knig & van der Auwera
Balthasar Bickel
(1988, 1990), Knig (1995) and others working on European clause linkage patterns.8
A further possibility is for clauses to be adjoined not so much to a sentence but to an
utterance, as in the case of speech-act modifying clauses like frankly speaking.
Of all these distinctions, the one that seems to be of widespread relevance is that
between ad-V and ad-S constructions. The traditional term adverbial subordination
conflates the two, but many languages make sharp distinctions (Bickel 1991, 1993,
1998). Ad-V clauses perform the function of adverbial modifiers; ad-S clauses, by contrast, provide general frameworks for the main clause (as in if, when or general topic
clauses) or sequences in an event chain. Ad-V clauses typically behave like ordinary
adverbial constituents, often case-marked like NP constituents and entirely transparent to the assignment of case to other constituents of the main clause. The critical
result of all this is that ad-V clauses can be center-embedded:
(56) Belhare
a.

Dhankuta him-yakt-a-lok=to
khar-e.
D.[loc] [3sS-]stumble-ipfv-pstsbjv-com=foc [3sS]go-pst
He went to Dhankuta stumbling.
(com)
b. pit-chi-lo
ap-khat-ket.

cow-ns-com [3sS] come.on.the.same.level-go-temp

She is passing by with the cows.
The dependent clause in (56a) (also cf. 14a, 15, and 26) is marked by the same comitative case suffix as the NP in (56b). The NP Dhankuta in (56a) appears with a zero
allomorph of the locative case rather than the regular overt locative in -e. This version of the locative is licensed by the main clause verb (and cannot be licensed by the
dependent verb him- stumble), which suggests that the dependent clause does not
interfere with case and semantic role assignment. This is different with ad-S constructions, which block case assignments and therefore center-embedding:
(57) Belhare
a.

u-chom
pok=naa
Dhankuta kha-yu.
3sPOSS-desire [3sS]rise[sbjv]=top D.[loc] [3sS]go-npst
(sub)
b. *Dhankuta u-chom
pok=naa
kha-yu.
(sub)

D.[loc] 3sPOSS-desire [3sS]come.up[sbjv]=top [3sS]go-npst

If he wants, he will go to Dhankuta.
Because of the intervening ad-S clause u-chom pok=naa if he wants, the main
clause verb cannot assign case to the first NP (Dhankuta) in (57b). Here, the ad-S
. also cf. the notion of left-detached and right-detached position in Role and Reference
Grammar (Van Valin & LaPolla 1997; Van Valin 2005).
clause is a topic clause, but the same blocking effect obtains if the ad-S clause is a
sequential chain:
(58) Belhare
a. *a-tak
[s%pp%i mai-mat-pir-he

1sPOSS-friend[nom] all[nom] 1s-[3sA]narrate-benef-pst

ki] khar-e.
seq [3sS]go-pst
(chain)
b. [s%pp%i mai-mat-pir-he
ki]

all[nom] 1s-[3sA]narrate-benef-pst seq

a-tak
khar-e.
1sPOSS-friend[nom] [3sS]go-pst
My friend told me everything and went off
(chain)
Nominative case on atak my friend in (58a) can only be assigned by the verb form
khare went, but this is impossible because of the intervening ki-clause. If this clause is
to be included, it must precede the entire main clause (atak khare), as in (58b).
In the absence of case, patterns of semantic role assignment may be the only reflex
of ad-S attachment. This is illustrated by Amele constructions of the kind exemplified
earlier in (1), (16) and (51):
(59) Amele
a. *dana age [ho busale-e-b] gbo-i-ga.

man 3p pig run.out-ds-3s hit-3p-t.pst

The men, the pig having run out, killed it.
(chain)
b. [ho busale-?e-b] [dana age gbo-ig-a].


The pig ran out and the men killed it.
(chain)
Sentence (59a) is ungrammatical because dana age the men cannot be assigned a
semantic role: the NP belongs to the main verb (gboiga they killed it), but role assignment by this verb is blocked by the intervening dependent clause (ho busaleeb the pig
ran out and), which itself has no extra role to assign. The sentence becomes grammatical in (59b) where the dependent clause is not center-embedded but precedes the main
clause containing both dana age and the main verb.
3.9 Other variables

The preceding discussion suggested a number of variables that are logically independent of each other. There are other variables. Some of these, such as the difference
between adjoined and subcategorized (argumental) dependent clauses are outside the
scope of this survey. Another variable that has received great prominence is syndesis,
i.e. the question of whether clauses are linked by affixes, independent conjunctions
Balthasar Bickel
and/or by intonational means. This is clearly an important issue, but it derives by and
large from the general morphological and phonological structure of the language and
is perhaps not as relevant to syntactic issues as has traditionally been assumed.
An important class of variables relates to issues of cross-clausal coreference. This
includes constraints on backward anaphora (which has often been taken to be diagnostic of clause linkage types), control and raising patterns (which is critical in distinguishing various complement clause types), and the presence of reference-tracking
devices such as switch-reference morphology, cross-clausal reflexivization or logophoric pronouns. A survey of the relevant variables would take us far beyond what can
be covered in this chapter.
Another issue that goes beyond the scope of this survey is the nature of the interpropositional relation that is realized by a clause linkage construction. Developing
variables for this is a tall order because especially temporal relationships are tightly
connected with the aspectual system of a language, and reference grammars typically
do not provide sufficient detail in order to understand these connections.9
4. Typological patterns
A set of typological variables like the one discussed here raises two questions: (i) Are
some structures more similar to each other than to others so that they define crosslinguistic type clusters? (ii) Which variables are correlated, forming statistical implicational universals? In the following, I discuss methods that allow answering these
questions, based on standards in other disciplines. To illustrate the methods, I apply
them to a pilot database of adjoined clause linkage structures.
4.1 A pilot database

I sampled languages based on what analyses I had readily available from an earlier
survey (Bickel 1991), and this leads to a bias towards Eurasia, Papua New Guinea and
Africa. Therefore, any results are preliminary and will have to be tested against a larger
and better balanced database. The pilot database contains data on 69 constructions
from 24 languages and codes each construction for eleven variables, plus one variable
registering the set of markers used. For ease of reference, each construction is identified by a short label. The full database is reproduced in the Appendix.
Within each language, constructions are differentiated by unique combinations
of markers and the levels across all variables. The set of variables largely follows the
. Another topic that is is difficult to survey in reference grammars is verb gapping, i.e.
whether pro-verbs are allowed or not, as in John cooked dinner, and so did Harry.
discussion in Section3, but I exclude negation scope because relevant data was not
available with sufficient analysis. I also use a simplified version of the finiteness variable. The following definitions of levels were applied, listing only levels that are actually
attested in the database:
ILL-scope: The scope of illocutionary operators in the main clause is
conjunct: extends to the main clause and the dependent clause
disjunct: extends to either the main or the dependent clause but never to both
local: is limited to the main clause
extensible: extends to either the main clause alone or to both the main clause
and the dependent clause, but never to the dependent clause alone
constraint-free: is not regulated by the clause linkage type
T-scope: The scope of tense or status operators in the main clause is
conjunct: extends to the main clause and the dependent clause
local: is limited to the main clause
extensible: extends to either the main clause alone or to both the main clause
and the dependent clause, but never to the dependent clause alone
Finiteness: The dependent clause is headed by a verb form that is
finite: at least as many categories must be marked as in main clauses
nonfinite: only fewer categories are allowed
any: either the same range or less categories can be marked
ILL-mark: Marking of illocutionary force operators in the dependent clause is
ok: allowed
banned: not allowed
harmonic: allowed but only if it matches the marking on the main clause
T-mark: Marking of tense or status operators in the dependent clause is
ok: allowed
banned: not allowed
harmonic: allowed but subject to constraints based on the tense or status choice
in the main clause
Symmetry: The range of categories that can be expressed on linked clauses is
symmetrical: must match
asymmetrical: can be different
free: can be different and can even include elements of different type (different
parts of speech, clauses and NPs, etc.)
WH: Question words and constituent focus inside dependent clauses are
ok: allowed
banned: not allowed
Extraction: Extraction of elements of dependent clauses is
ok: allowed
banned: not allowed
FOC: Focus marking on dependent clauses is
ok: allowed
banned: not allowed
Balthasar Bickel
Position: The position of the dependent clause vis--vis the main clause with
which it enters a dependency relation is
fixed: post-main: is fixed and is always after the main clause
fixed: pre-main: is fixed and is always before the main clause
flexible-adjacent: can be before or after the main clause but must be
adjacent to it
flexible-relational: can be before or after the main clause and can be
separated from the main clause by other dependent clauses
Layer: The dependent clause adjoins
ad-v: to the predicate and can be center-embedded
ad-s: to the clause and cannot be center-embedded
detached: to the clause but is separated syntactically and intonationally
For current purposes I treat all these variables as unstructured multinomial variables.
This is a simplification since in some cases, there may be an underlying structure for
example, local and extensible scope are arguably more similar to each other than
either is to disjunct or conjunct scope; or the levels banned, harmonic, and ok
could be modeled in terms of degrees of restrictiveness, i.e. as a rank variable.10 I leave
exploration of such possibilities to future research.
4.2 Prototypes and scales

A key question is to what extent there are cross-linguistic clusters or prototypes of
clause linkage patterns that is, types of structures to which language-specific constructions bear similarity to a quantifiable degree. To find out, we need methods for
measuring similarity between individual constructions and for aggregating the similarities of all constructions at once.
4.2.1 Methods
A standard measure of similarity used in many fields is the inverse relative Hamming
distance (also known as the Gower coefficient; see Kaufman & Rousseeuw 1990): the
distance or dissimilarity between two constructions is equal to the proportion of different levels in all non-empty variables. For example, according to the pilot database
in the Appendix, the difference between the Amele but-construction and chainconstruction is 5/8 = .625 since there are eight nonempty (non-NA) variables and
five of them have different levels. The distance between the Amele but-construction
and the Belhare and-construction is smaller, viz. 3/8 = .375, and this captures the
intuition that these constructions are relatively similar to each other. Some structures may even be identical. For example, from all we know, Chantyal and Burshaski
. Thanks to Jeff Good and Volker Gast for reminding me of this.
chain-constructions have identical properties in all non-empty variables; this results

in a relative Hamming distance of 0.
Since the database contains 69 constructions, computing Hamming distances of
all pairs of constructions results in a matrix of 69 (692 1) = 2, 346 pairs, similar in kind to
distance charts found in road maps (except that there can be 0-length distances). The
challenge is now to find patterns in these 2,346 pairs by searching for clusters of elements that are closer together across all pairwise distances.11 One way of going about
this is by plotting all distances at once in a graph. However, distances in a 69 69
matrix can be faithfully plotted only in 68-dimensional space, which is not visually
accessible. All reductions to two- or three-dimensional space are non-trivial and
necessarily impose some degree of distortion. For example, if we have three elements
with the distances dist(A,B) = 1, dist(A,C)=1, and dist(B,C) = 2/3, this can be faithfully plotted in two-dimensional space (cf Figure la). If we now add an element D
with distances dist(A,D) = 1, dist(B,D)=1/3, and dist(C,D) = 1/2, this immediately
becomes impossible. For example, if we try to respect the distance of D to C (which
is 1/2) and B (which is 1/3), we necessarily violate the distance to A, which is 1. One
popular class of solution in typology and psychology, Multidimensional Scaling, seeks
a geometrical arrangement with an optimal balance between minimal distortions and a
minimal number of dimensions. Often, however, no good balance can be found and
all reasonably low-dimensional solutions (two- or three-dimensional ones) are heavily
distorted (cf. Cysouw 2007).
B
0.167
B
0.334
D?
0.334
0.084
D
0.667
(a) Simple graph
0.167
0.084
0.667
0.25
C
(b) Split graph
Figure1. Representing all pairwise distances between four items in two-dimensional space
. Except where noted, all computations here and in the following were done in R (R Development Core Team 2009), with the additional packages cluster (Maechler et al. 2005), vcd
(Meyer et al. 2009a), entropy (Hausser & Strimmer 2009b), and minet (Meyer et al. 2009b).
Balthasar Bickel
An alternative solution, which I adopt in the following, consists in simultaneously plotting competing distances in what is known as a split graph (Bandelt & Dress
1992). This is shown in Figure lb, where the lines show the possible paths one can take
while respecting the true distances: each true distance between elements is the sum
of the indicated (rounded) lengths on the shortest connecting path. As a result, the
less distances (i.e. dissimilarities) are in conflict, the less there are splits and elements
cluster into clear groups. If there is more conflict, there are more splits, such as the
parallelogram in Figure lb, and the evidence for clusters is weaker: in Figure lb, for
example, there is no particular cluster except for a trend towards separating B and D
from all else (since dist(B,D) is smaller than any other distance in the example). Split
graphs of this kind can be efficiently computed by an algorithm known as the NeighborNet algorithm (and all split graphs here and in the following are produced with
the program SplitsTree4 implementing this algorithm; see Bryant & Moulton 2004,
Huson & Bryant 2006).12
A general problem with similarity analysis (whether done computationally or by
hand) is that it works best if data are complete, with no missing values: as soon as one
variables lacks a value (marked as NA in the Appendix) for a given construction, the
construction can no longer be compared to these other constructions with regard to
this variable, even if the other constructions have a value. As a result, a considerable
part of the information in the database cannot be exploited in the analysis. Dissimilarity methods are known to be fairly robust and not very sensitive to the amount of
information used. But it is clear that they can only be used as a heuristic and need to
be backed up by other methods (cf. Section4.3).
4.2.2 Results
Traditional conceptions of clause linkage would lead one to expect two or three clearly
distinct clusters, representing coordination, adverbial subordination, and possibly
cosubordination or chaining. The split graph in Figure2 casts doubt on the this.13
There is some degree of clustering of structures in the top left region of the graph
that one might loosely associate with adverbial subordination, including some topic
clauses (e.g. in Godi or Nepali) as well English or German when and if clauses
(labeled as sub in the graph and the Appendix). But this excludes ad-V structures,
. http://www.splitstree.org/. Split graphs are usually applied to genetic data, both biological and linguistic. Distances then represent language change. For another application to
typology, see Cysouw (2008).
. The graph follows exactly the same principles as the toy graph in Figure lb, but now
applied to the whole dataset in the Appendix. For readability, the numerical distances are left
off the edges of the graph, but the relative length of the edges is proportional to these distances.
which are dissimilar in many regards and are therefore placed in a different region, the
lower right.
One cluster that emerges as somewhat more pronounced is a group of and-like
structures on the left side of the graph. Interestingly, these structures share a significant
proportion of properties with topic-constructions in Tauya and also with detached
finite clauses in German. Moreover, they are closer to the when, if and topic constructions in the top left region than to the chaining-like structures that are placed at
the other end of the graph on the right. The positioning of and-constructions closer
to topic, when and if clauses than to chaining-like structures seems to be mostly
caused by the fact that these structures all have flexible position, symmetrical category
marking, and involve finite clauses, whereas chaining-like constructions are mostly
nonfinite and asymmetrical.
Amele.cond
Tauya.chain
Fore.chain Newar..Dolakha..chain
Godie.topic English.fin.sub
Belhare.purp
Yup.ik..Central..sub
Belhare.cvb
Nepali.topic
Nepali.cvb
Belhare.sub
Chantyal.chain
English.and
Lenakel.chain
Usan.chain.swr
German.sub
Hua.ds.chain
Swahili.and
Burushaski.chain
English.although
Chechen.cvb
Hua.ss.chain
German.and
Belhare.inconseq
Russian.and
Nepali.chain
Belhare.and
Kate.chain
Kate.serial
Chechen.and
Amele.chain
Amele.but
Turkish.chain
Swahili.and.INF
German.cause. Amele.or
Swahili.chain
detached Tauya.topic Kate.TOP
Godie.chain
German.cond.
Wambule.chain
detached
Hua.alter.iter
Belhare.chain
German.cvb Chechen.chain
Greek..Ancient..part.coni.
Tauya.inconseq
English.go.and
Chechen.cvb.temp Turkish.serial
Usan.cause
Toura.chain
Amele.purp
Hua.cond English.to.ing.detached
Hua.inconseq German.purp
Korafe.chain
German.cvb.adV
Hua.topic
Russian.cvb.detached
English.cvb.detached
Usan.topic
English.cvb.adv
Belhare.com
Russian.cvb.adV
Swahili.topic
Figure2. A NeighborNet representation of clause linkage patterns based on the relative

Hamming distances between the constructions in the pilot database. Nodes represent
individual constructions and are identified by the language name and the constructional
label given in the Appendix
The right side of the graph contains various converb and chaining structures.
The way these are arranged suggests a possible continuum between two sub-clusters,
arranged by the extent to which illocutionary force scope is constrained: the cluster
at the top right of the graph contains structures with less such constraints (e.g. Fore
or Hua different subject chains) than the cluster in the mid-right region (e.g. Amele,
Swahili or Turkish chains). Interestingly, Belhare and Wambule chains (both Kiranti),
but no others, are closer to and-coordination, in line with their finiteness.
Balthasar Bickel
4.3 Correlations
Clusters such as those emerging from the similarity analysis in the preceding result
from co-varying variables, where particular levels on one variable (i.e. specific
constructional properties) are associated with particular levels on other variables.
Indeed, traditional cross-linguistic notions can in fact be thought of as property
bundles in which properties are strictly correlated and fully entail each other.
For example, a notion like subordination entails (at least) disjunct illocutionary
scope, the availability of restrictive focus, and a ban on extraction and question
formation. The evidence discussed in Section 3 suggests that the variables surveyed are all logically independent of each other. Therefore, there is no reason to
expect exceptionless correlations (also cf Lehmann 1988). However, there may be
statistical trends, and it is such trends that underly the possible clusters that we
observed in the similarity analysis in Figure2. Thus, to the extent that there are
probabilistic clusters based on specific kinds of subordination and coordination,
as suggested by Figure2, we can expect corresponding correlations of the properties defining these clusters.
The difference between traditional property bundles and probabilistic correlations
is the same as the difference between absolute and statistical (or empirical) implicational universals. As property bundles, traditional cross-linguistic notions have the
form A B (where A and B are properties), and they suffer from the same validation
problem as absolute universals (Bickel in press): no language sample can guarantee
that the universal is without exceptions because we cannot survey all languages that
have ever been or will ever been spoken. The only available route to justification is
logical deduction. Apparent counter-examples (e.g. constructions that show disjunct
illocutionary scope but allow questions, such in Amele (39) or Belhare (40)), then
need to be re-analyzed (e.g. by positing suitable underlying structures with different
scope properties), or the definition of the notion (the property bundle) needs to be
revised. Both options bring us back to the discussion in the introductory section and
the problems associated with finding universal definitions of property bundles. Probabilistic correlations have the same structure as statistical universals, and so they can be
validated in the same way, by statistical evidence and significance testing.
4.3.1 Methods
With eleven variables there are 11(112 1) = 55 possible pairs of variables. Given these
possibilities, we need a heuristic technique to find those pairs with substantial correlations. One method for finding these would simply consist in performing statistical tests for each pair of correlations, but this leads to well-established problems with
spurious success rates that arise from multiple testing (known as familywise errors
in the statistics literature). An alternative heuristic is based on information theory

and has been developed for exploring associations between genes.14 The key idea is to
estimate what is called the mutual information between two (categorical) variables,
which can be informally thought of as the extent to which knowledge of the frequency
distribution in one variable (i.e. knowledge of the actual counts of each level defined
by the variable) allows prediction of the frequency distribution in the other variable.
Estimates of mutual information can be normalized by the frequency distribution in
one of the variables, and this yields an estimate of the extent to which the frequency
distributions in one variable can be predicted by the frequency distribution of the
other variable. This can be represented as (A|B), i.e. the probability of the frequency
distribution of A given the frequency distribution of B. (A|B) captures the intuition
behind a classical implicational universal of the form B A and, conversely, (B|A)
is the equivalent of what is traditionally written as A B, and if (A|B) is about the
same as (B|A), we find what is traditionally written as A B.15
When computing implications between many pairs of variables in this way, one
risks detecting artifacts caused by chains of associations. For example, if C predicts
B and B predicts A, then C predicts A redundantly. What we are interested in is the
strongest of these associations, i.e. the ones with the highest probability (X|Y). A
powerful method to achieve this is suggested by Margolin et al. (2006) for bioinformatic purposes and consists in removing all implications with the lowest values
on (X|Y) given an association of X and Y via some other term. Assume that the
implication B A is associated with (A|B) and that for the implications C A and
. I am grateful to Korbinian Strimmer for drawing my attention to this technique as well

as to Margolin et al. (2006) and Hausser & Strimmer (2009a).
. Formally, the mutual information (MI) of variables A and B is defined in terms of the
Shannon entropy (H) of A and B and the entropy of the joint probability distributions of A and
B:MI(A;B) = H(A) + H(B) H(A,B). The probability (A|B) is then defined as MIH((AA;)B ) . The entropy
H itself describes the extent to which the probabilities of the levels of a variable are skewed. For a
variable A with levels ai {a1k} and associated level probabilities pai , H(A) = pai log( pai ).
H(A) is zero if there is a total bias towards a single level, e.g. with pa1 = 1, pa2 = 0,and pa3 =0;
it reaches its maximum in uniform distributions, e.g with pa1 = .33, pa2 = .33, and pa3 = .33.
The entropy of the joint probability distribution of A and B is H(A, B) = pai ,bi log( pai ,bi ).
The probabilities pai and pai ,bi can be estimated by the relative level counts ( pa1 = NN(a(a1 ) ) etc.)
1k
in the database, but this method (the Maximum Likelihood method) is known to be unreliable when individual level frequencies are relatively low, as is often the case in typological data
with small sample size. Alternative methods are discussed by Hausser & Strimmer (2009a),
and in this paper I use the James-Stein-type estimator developed there and implemented in
the R package entropy.
Balthasar Bickel
CB we observe that (A|C) < (B|C). Then, C A is an artifact of the stronger

implications and we set (A|C) = 0, i.e. we effectively remove the link:
B
A
B
C
Once implications and their strengths are established, the question arises as to
which particular levels cause them. The mutual information method only shows which
variables are likely to be correlated (e.g. ILL-scope WH), but for analyzing clusters, it is more important to know which exact levels (properties) are behind this (e.g.
local illocutionary scope question words banned). A standard solution rests on
the analysis of the Pearson residuals of each cell in a contingency table defined by the
variables of interest (e.g. the cell at the intersection of local and ok in a contingency
table defined by the variables ILL-scope and WH). Informally, the Pearson residual
describes the extent to which the observed frequency of a given cell in a contingency
table deviates (positively or negatively) from what can be expected under the null
hypothesis of no association, i.e. what one would expect in this cell if the proportions
of level counts were constant across rows and columns (e.g. so that the overall .4 vs. .6
proportion of banned vs. allowed question words in the data is the same for all levels
of the ILL-scope variable, i.e. for local scope in the same way as for conjunct scope
and all others).16 Adopting methods developed by Zeileis et al. (2007) and Meyer etal.
(2006), I use a permutation test to establish which residuals exceed what one would
find under random reshuffling of cell counts. If one finds the observed residuals in less
than 5% (or 1%) reshufflings, they point to statistically significant associations of specific levels (at what is called a 5% or 1% significance level). Note that it is possible that
while the overall table shows a significant association (because the total sum of residuals is high), no individual residual may be significant or several (or all) are significant
(because the residuals spread uniformly over the cells). In this case, it is impossible to
establish which specific properties are responsible for the overall correlation without
expanding the dataset considerably, or by collapsing distinctions in the definitions of
the variables.
. Formally, the Pearson residual is the relative contribution of a cell in a contingency

n
m
table to the 2 sum of the table and is defined as m where n is the observed count and the
expected count.
possible
ok
Extraction
blocked
(a) ILL-mark extraction
Illocutionary scope
local exten disju const
c
raint- onjunct
sible nct
free
Illocutionary force marking

h ar m
bann
ed
onic
Constituent question or focus

banned
ok
(b) ILL-scope WH
Figure3. Residual analysis: box heights are proportional to the residuals and box width to
the number of datapoints, the direction of boxes from the baseline signals their sign (raising
boxes for positive, falling boxes for negative residuals), and shading indicates that residuals are
individually significant under a permutation test, with light grey at a 5% and dark grey at a 1%
significance level (see Meyer et al. 2006). The order of levels in the plot is arbitrary (alphabetical).
The plots in Figure3 illustrate the method for two of the implications found, ILLmark extraction and ILL-scope WH. Figure3a suggests that if a construction
allows illocutionary marking on dependent clauses only if the marking matches that in
the main clause (harmonic ILL-mark), then it is likely (at a 1% significance level, dark
shading) that constituents can be extracted. In other words, extracting constituents
is significantly facilitated by harmonic illocutionary force marking. Structures illustrating this include English go and-constructions, where extraction is possible and
which allow, say, imperatives if they apply to both conjuncts (cf the beer I went to the
store and bought and Go and buy beer!; Stassen 1985; Lakoff 1986). Another example
is Belhare ki-constructions (cf. (23) for illocutionary harmony and (46a) for extraction). In Figure3a, no other residuals are significant. Figure3b shows three significant
associations. First, there is a highly significant association (at a 1% significance level,
dark shading) between local illocutionary scope and a ban on constituent question
formation, i.e. question formation is banned more often when scope is local than when
it is conjunct, extensible, or constraint-free. The same trend can be observed if scope is
disjunct: although not statistically significant on its own, the positive residuals suggest
that structures with disjunct scope appear to ban question formation more often than
what one would expect under the null hypothesis of no association. Second, constraintfree scope is negatively associated with a ban on question formation, i.e. if there is
Balthasar Bickel
no constraint on illocutionary scope it is unlikely that question formation is banned

(as is the case for example in Belhare ki-construction, as shown by by (42) and (54)).
This is mirrored in the third association, which is the positive association between
constraint-free scope and allowed question formation. This association is individually
significant only at a lower (5%) significance level (light shading), which suggests that
there is somewhat less evidence for clause linkage with constraint-free scope explicitly
allowing question words than there is evidence for the finding that this type of clause
linkage tends to lack an explicit ban on question words.17
4.3.2 Results
Table1 summarizes the findings on all correlations which have an MI-based probability (X|Y) that is not set to zero after removing all artifacts. Where there is evidence
from unequal probabilities, the implications are formulated in a directed way ();
where the probabilities are identical or near-identical, the correlations are formulated
as two-way implications () The table also reports the results from the residual analysis, where * indicates significance at a 5% and ** at a 1% level. No entry means that
none of the residuals was significant, i.e. the pilot database does not allow identifying
which specific properties are responsible for the overall association of the variables. In
those cases where an additional (but non-significant) trend in the same direction as
that of a significant association can be discerned (as was the case with disjunct scope
in Figure3b), this is added in brackets.
Most of the findings in Table1 receive straightforward theoretical interpretations
and are good candidates for genuine universals that deserve testing against larger and
less biased datasets.
One set of associations concerns scope and marking possibilities. First, there
are direct links between these two properties (ILL-scope ILL-marking, conjunct
T-scope banned T-marking, local or extensible T-scope allowed T-marking).
Although these links are not logically necessary, many languages seem to conventionalize the pragmatic principle that a category value is maintained as long as there is no
explicit marking against this, and, vice-versa, that the explicit marking of a category
blocks the scope of other markers of the same category. Supporting evidence for this
comes from the association of conjunct tense or status scope with ad-V, but not ad-S
structures (conjunct T-scope ad-V): ad-V structures are often center-embedded
. This reflects a fundamental descriptive problem: if there are no constraints on question

formation in a language, linguists are naturally tempted not to discuss question formation at
all in that language but then we dont know whether question formation is allowed or simply
not studied.
within the main clause and thus a category value needs to be maintained only for a
shorter time (in processing) than in ad-S structures, which cannot be center-embedded
(by definition).
T-marking, but not ILL-marking, possibilities are also associated with finiteness (banned T-marking nonfinite, freely or harmonically allowed T-marking
finite). This may reflect a universal trend to the effect that tense and status have a
higher probability of being realized through verb morphology than illocutionary
force, which is more commonly realized through clitics and particles on the sentence level. While many theories assume that illocutionary force is on a more external
(peripheral) syntactic shell than tense and status (e.g. Foley & Van Valin 1984; Van
Valin 2005), there is in fact no clear motivation for this because just like illocutionary
force, tense and status are properties of entire propositions or even entire texts, and
there is no sense in which they are properties of predicates. Perhaps this an arbitrary
fact of our language faculty, but at any rate, the issue calls for further research if the
trend observed here is robust.
Table1. Implications detected by analysis of the relative mutual information between

variables in the pilot database (see Appendix) and by permutation tests of the
Pearson residuals
Implication
(X|Y)
(Y|X)
ILL-scope ILL-mark
ILL-scope WH
.09
.18
.05
.08
ILL-scope FOC
ILL-scope position
.22
.23
.06
.20
T-scope T-mark
.29
.26
T-scope symmetry
.21
.17
T-scope FOC
T-scope layer
ILL-mark extraction
T-mark finiteness
.11
.09
.22
.31
.05
.06
.12
.28
symmetry finiteness
.28
.28
symmetry ILL-mark
position layer
.13
.12
.12
.06
Statistically significant associations

of properties
local (or disjunct) banned**
constraint-free ok*
constraint-free flexible-adjacent**
disjunct flexible-relational**
conjunct banned**
local (extensible) ok**
local symmetrical (or constraintfree)**; extensible constraint-free*
conjunct ad-V*
harmonic possible**
banned nonfinite*
ok (or harmonic) finite*
constraint-free any (or finite)*
symmetrical finite**
symmetrical ok (or harmonic)*
Balthasar Bickel
Marking possibilities and finiteness are furthermore connected with regard to

symmetry: if a clause linkage type requires the same amount of marking, this involves
typically finite structures and allows independent or harmonic ILL-marking, and
vice-versa (symmetrical finite, symmetrical freely or harmonically allowed
ILL-marking). If a clause linkage type does not select a specific type of dependent
(e.g. allows NPs as much as clauses), it imposes no constraints on the kind of morpho
logy (e.g. finite) when the dependent is a clause (constraint-free symmetry any
or finite).
Scope properties of illocutionary force are moreover linked to question formation (local or disjunct ILL-scope banned question formation, constraint-free ILLscope allowed question formation). Structures with local ILL-scope often seem
to block dependent clauses from any modulation of illocutionary force, assimilating them to presupposed, taken for granted topic expressions, as Foley (1986) has
noted for Papuan languages. In turn, the flexible kind of illocutionary scope behavior
found in chaining-like constructions significantly often allows question formation
(cf Section3.5).
Some of the correlations in Table1 are responsible for the clusters that emerged
from the similarity analysis in Figure2: the cluster around topic, when, and if clauses
noted in the upper left of the figure is characterized by the association of disjunct ILLscope, local T-scope, finite dependent clauses, symmetry, and flexible-relational position (disjunct ILL-scope flexible-relational position, local T-scope symmetrical,
symmetrical finite). There is in addition a statistically non-significant trend for
disjunct ILL-scope to be associated with a ban on question formation or constituent
focusing (disjunct ILL-scope banned WH, but see e.g. (39) or (40) for counterexamples). The properties of symmetry and finiteness are shared by the adjacent
cluster of and-constructions, but in addition, these structures are characterized by
flexible (constraint-free) illocutionary scope, flexible-adjacent position and allowing
ILL-marking and question formation (constraint-free ILL-scope flexible-adjacent,
symmetrical freely or harmonically allowed ILL-marking, constraint-free ILLscope allowed question formation). Harmonic ILL-marking characterizes specifically Belhare ki-constructions (chaining) and English go and-constructions, which
show up as closely related in Figure2 (lower left). This type of and-like structure also
allows extraction (harmonic ILL-marking allowing extraction).
5. Conclusions
In this chapter I have argued that cross-linguistic diversity in clause linkage is far too
big as to be amenable to traditional typologies which seek to define a couple of universal
notions that entail sets of properties without exception (cf Lehmann 1988). What is
needed instead is an approach similar to what is standard in other disciplines studying diversity: sets of variables capturing the variation and probabilistic assessment of
clusters and correlations. Methods for doing this are readily available and have a solid
and well-tested mathematical foundation.
Applied to a pilot study of 69 constructions from 24 languages, this approach
has revealed a set of possible implicational universals that now need to be tested
against larger datasets, with systematic control of genealogical inheritance and areal
spread factors, before they can be fully explained. The approach proposed here also
allows examining the evidence for probabilistic clusters, or typological prototypes
of constructions. The pilot database casts doubt on universal cluster concepts close
to traditionally expected structures like coordination, subordination, or perhaps
cosubordination. Instead, there is tentative evidence for a specific prototype of subordination, which tends to associate disjunct illocutionary scope, local tense scope,
finite and symmetrical clauses, and flexible position (and, with slightly less probability, also a ban on question formation or focusing inside the dependent clause).
This cluster is in close neighborhood not to chaining structures but to and-like constructions, with which it shares symmetry and finiteness, but differs by having more
constraints on illocutionary scope, illocutionary marking and question formation
or constituent focusing inside the dependent clause. Chaining constructions do not
form a tight cluster but are spread out in a continuum of structures with more vs. less
tightly constrained scope properties.
The advantage of a multivariate approach as suggested here is that it brings typology closer to language-specific analysis. Instead of being separated enterprises (as
proposed by Lazard (2006) or Haspelmath (2007)), typological survey work is identical to language-specific analysis and consists in detailed descriptions of properties.
To the extent that we learn more about language-specific properties, the better can
we define fine-grained typological variables. And such fine-grained variables provide
in return the ideal questionnaires for fieldwork. No fieldwork can proceed out of a
theoretical vacuum. But what this chapter has shown is that none of the received theoretical concepts in clause linkage fills this vacuum with the right questions: it makes
little sense to debate whether a particular construction is subordinate or not because
such a notion is far too coarse to be universally applicable. But it makes sense to ask
whether the construction has disjunct illocutionary force scope, or whether it allows
question formation. Such questions are indeed much closer to the level of granularity that is needed in field linguistics. In return, only by answering such questions in
great detail can we develop full datasets, with a minimum of missing data. And this
is a precondition for establishing and understanding implicational universals, crosslinguistic prototypes and continua.
Balthasar Bickel
References
Bandelt, Hans-Jrgen & Andreas Dress, 1992. A canonical decomposition theory for metrics on
a finite set. Advances in Mathematics 92, 47105.
Bearth, Thomas, 1986. Larticulation du temps et de [aspect dans le discours Toura. Bern: Lang.
Bickel, Balthasar, 1991. Typologische Grundlagen der Satzverkettung. Zrich: ASAS.
Bickel, Balthasar, 1993. Belhare subordination and the theory of topic. In Ebert, Karen H. (ed.)
Studies in clause linkage, 2355. Zrich: ASAS.
Bickel, Balthasar, 1996. Aspect, mood, and time in Belhare. Zrich: ASAS.
Bickel, Balthasar, 1998. Converbs in cross-linguistic perspective [review article of Haspelmath and
Knig, eds., Converbs, Berlin: Mouton de Gruyter 1995]. Linguistic Typology 2, 381397.
Bickel, Balthasar, 1999a. From ergativus absolutus to topic marking in Kiranti: a typological
perspective. Proceedings of the 25th Annual Meeting of the Berkeley Linguistics Society
3849.
Bickel, Balthasar, 1999b. Nominalization and focus constructions in some Kiranti languages.
In Yadava, Yogendra P. & Warren W. Glover (eds.) Topics in Nepalese linguistics, 271296.
Kathmandu: Royal Nepal Academy.
Bickel, Balthasar, 2007. Typology in the 21st century: major current developments. Linguistic
Typology 11, 239251.
Bickel, Balthasar, in press. Absolute and statistical universals. In Hogan, Patrick Colm (ed.)
The Cambridge Encyclopedia of the Language Sciences. Cambridge: Cambridge University Press [pre-print available at http://www.uni-leipzig.de/~bickel/research/papers/
universals_cels_bb.pdf].
Bickel, Balthasar & Johanna Nichols, 2002. Autotypologizing databases and their use in fieldwork.
In Austin, Peter, Helen Dry, & Peter Wittenburg (eds.) Proceedings of the International
LREC Workshop on Resources and Tools in Field Linguistics, Las Palmas, 2627May 2002.
Nijmegen: MPI for Psycholinguistics.
Bryant, David & Vincent Moulton, 2004. Neighbor-Net: an agglomerative method for the construction of phylogenetic networks. Molecular Biology and Evolution 21, 255265.
Chafe, Wallace L., 1976. Givenness, contrastiveness, definiteness, subjects, topics, and point of
view. In Li, Charles N. (ed.) Subject and topic, 2755. New York.
Chomsky, Noam, 1973. Conditions on transformations. In Anderson, Stephen R. & Paul Kiparsky
(eds.) A Festschrift for Morris Halle. New York: Holt, Rinehart and Winston.
Clark, T.W, 1963. Introduction to Nepali. Kathmandu: Ratna Pustak Bhandar (Reprint).
Cristofaro, Sonia, 1998. Deranking and balancing in different subordination relations: a typological study. Sprachtypologie und Universalienforschung 51, 342.
Cristofaro, Sonia, 2003. Subordination. Oxford: Oxford University Press.
Croft, William, 2001. Radical construction grammar: syntactic theory in typological perspective.
Oxford: Oxford University Press.
Cysouw, Michael, 2007. Building semantic maps: the case of person-marking. In Miestamo,
Matti & Bernhard Wlchli (eds.) New challenges in typology: broadening the horizons and
redefining the foundations., 225248. Berlin: Mouton de Gruyter.
Cysouw, Michael, 2008. Generalizing scales. In Malchukov, Andrej & Marc Richards (eds.)
Scales (Linguistische ArbeitsBerichte 86), 379396. Leipzig: Institut fur Linguistik.
Farr, Cynthia J., 1999. The interface between syntax and discourse in Korafe, a Papuan language
of Papua New Guinea. Canberra: Pacific Linguistics.
Foley, William A., 1986. The Papuan languages of New Guinea. Cambridge: Cambridge University
Press.
Foley, William A. & Robert D. Van Valin, Jr., 1984. Functional syntax and universal grammar.
Cambridge: Cambridge University Press.
Forker, Diana, 2009. Converbs in Hinuq and some remarks on linguistic concepts. Ms. MaxPlanck-Institute for Evolutionary Anthropology, Leipzig.
Genetti, Carol, 2005. The participial construction in Dolakha Newar: syntactic implications of
an Asian converb. Studies in Language 29, 3587.
Good, Jeff, 2003. Clause combining in Chechen. Studies in Language 27, 113170.
Green, Georgia M., 1976. Main clause phenomena in subordinate clauses. Language 52,
382397.
Haiman, John, 1978. Conditionals are topics. Language 54, 564589.
Haiman, John, 1980. Hua: a Papuan language of the Eastern Highlands of New Guinea.
Amsterdam: Benjamins.
Haiman, John, 1985a. Natural syntax. Cambridge: Cambridge University Press.
Haiman, John, 1985b. Symmetry. In Haiman, John (ed.) Iconicity in syntax, 7395. Amsterdam:
Benjamins.
Hale, Austin & Kedar P. Shrestha, 2006. Newar: Nepal bhdsd. Muenchen: LINCOM.
Haspelmath, Martin, 1995. The converb as a cross-linguistically valid category. In Haspelmath,
Martin & Ekkehard Konig (eds.) Converbs in cross-linguistic perspective, 155. Berlin:
Mouton de Gruyter.
Haspelmath, Martin, 2007. Pre-established categories dont exist: consequences for language
description and typology. Linguistic Typology 11, 119132.
Hausser, Jean & Korbinian Strimmer, 2009a. Entropy inference and the James-Stein estimator, with application to nonlinear gene association networks. Journal of Machine Learning
Research 10, 14691484.
Hausser, Jean & Korbinian Strimmer, 2009b. entropy: entropy and mutual information estimation. R package, http: //www. R-project. org/.
Horn, Laurence R., 1989. A natural history of negation. Chicago: The University of Chicago Press.
Huson, D.H. & D. Bryant, 2006. Application of phylogenetic networks in evolutionary studies.
Molecular Biology and Evolution 32, 254267.
Jacobs, Joachim, 1984. Funktionale Satzperspektive und Illokutionssemantik. Linguistische Berichte
91, 2558.
Jacobs, Joachim, 1991. Focus ambiguities. Journal of Semantics 8, 136.
Johanson, Lars, 1995. On Turkic converb clauses. In Haspelmath, Martin & Ekkehard Konig
(eds.) Converbs in cross-linguistic perspective, 313347. Berlin: Mouton de Gruyter.
Kaufman, Leonard & Peter J. Rousseeuw, 1990. Finding groups in data: an introduction to cluster
analysis. New York: Wiley.
Knig, Ekkehard, 1995. The meaning of converb constructions. In Haspelmath, Martin &
Ekkehard Knig (eds.) Converbs, 5795. Berlin: Mouton de Gruyter.
Knig, Ekkehard & Johan van der Auwera, 1988. Clause integration in German and Dutch conditionals, concessive conditionals, and concessives. In Haiman, John & Sandra A. Thompson
(eds.) Clause combining in grammar and discourse. Amsterdam: Benjamins.
Knig, Ekkehard & Johan van der Auwera, 1990. Adverbial participles, gerunds and absolute
constructions in the languages of Europe. In Bechert, Johannes, Giuliano Bernini, &
Claude Buridant (eds.) Toward a typology of European languages, 337355. Berlin: Mouton
de Gruyter.
Balthasar Bickel
Lakoff, George, 1984. Performative Subordinate Clauses. Proceeding of the 10th Annual Meeting
of the Berkeley Linguistics Society 472480.
Lakoff, George, 1986. Frame-semantic control of the Coordinate Structure Constraint. Proceeding
of the 20th Annual Meeting of the Chicago Linguistics Society 152167.
Lazard, Gilbert, 2006. La qute des invariants interlangues: la linguistique est-elle une science?
Paris: Champion.
Lehmann, Christian, 1988. Towards a typology of clause linkage. In Haiman, John & Sandra
A. Thompson (eds.) Clause combining in grammar and discourse, 181226. Amsterdam:
Benjamins.
Lynch, John, 1978. A grammar of Lenakel. Canberra: Pacific Linguistics (PLB55).
MacDonald, Lorna, 1988. Subordination in Tauya. In Haiman, John & Sandra A. Thompson
(eds.) Clause combining in grammar and discourse, 227246. Amsterdam: Benjamins.
MacDonald, Lorna, 1990. A grammar of Tauya. Berlin: Mouton de Gruyter.
Maechler, Martin, Peter Rousseeuw, Anja Struyf, & Mia Hubert, 2005. cluster: cluster analysis
basics and extensions. R package, http://www.R-project.org/.
Marchese, Lynell, 1977. Subordinate clauses as topics in Godie. In Mould, Martin & Thomas
J. Hinnebusch (eds.) Papers from the 8th Conference on African linguistics, 157164. Los
Angeles: University of California.
Margolin, Adam, Ilya Nemenman, Katia Basso, Chris Wiggins, Gustavo Stolovitzky, Riccardo
Favera, & Andrea Califano, 2006. ARACNE: an algorithm for the reconstruction of gene
regulatory networks in a Mammalian cellular context. BMC Bioinformatics 7, S7.
Meyer, D., A. Zeileis, & K. Hornik, 2006. The strucplot framework: visualizing multi-way
contingency tables with vcd. Journal of Statistical Software 17, 148.
Meyer, David, Achim Zeileis, & Kurt Hornik, 2009a. vcd: visualizing categorical data. R package, http://www.R-project.org/.
Meyer, Patrick E., Frederic Lafitte, & Gianluca Bontempi, 2009b. minet: mutual information
network inference. R package, http://www.R-project.org/.
Molochieva, Zarina, 2008. Scope properties of Chechen converbs. Handout of paper given at the
Syntax of the Worlds Languages III conference, Berlin, September 2528, 2008.
Noonan, Michael, 1999. Converbal constructions in Chantyal. In Yadava, Yogendra P. & Warren
G. Glover (eds.) Topics in Nepalese linguistics, 401420. Kathmandu: Royal Nepal Academy.
Olson, Michael L., 1981. Barai clause juncture: toward a functional theory of inter-clausal relations.
Ph.D. thesis, Australian National University.
Opgenort, Jean Robert, 2004. A Grammar of Wambule. Leiden: Brill.
Pilhofer, Georg., 1933. Grammatik der Kte-Sprache in Neuguinea. Berlin: Reimer.
R Development Core Team, 2009. R: a language and environment for statistical computing.
Vienna: R Foundation for Statistical Computing, http: //www. r-project.org.
Rappaport, Gilbert C, 1984. Grammatical function and syntactic structure: the adverbial participle
of Russian. Columbus, Ohio: Slavica.
Reed, Irene, Osakito Miyako, Steven Jacobsen, Paschal Afcan, & Michael Krauss, 1977. Yupik
Eskimo grammar. Fairbanks: Alaska Native Language Center, University of Alaska.
Reesink, Ger P., 1987. Structures and their functions in Usan. Amsterdam: Benjamins.
Roberts, John R., 1987. Amele. London: Croom Helm.
Roberts, John R., 1988. Amele switch-reference and the theory of grammar. Linguistic Inquiry
19, 4564.
Ross, John R., 1967. Constraints on variables in syntax. Ph.D. thesis, MIT, Cambridge, Mass.
Schackow, Diana, Balthasar Bickel, Shree Kumar Rai, Narayan P. Sharma (Gautam), Arjun Rai,
& Martin Gaenszle, in press. Morphosyntactic properties and scope behavior of subordinate
clauses in Puma (Kiranti). In Gast, Volker & Holger Diessel (eds.) Clause-combining in
cross-linguistic perspective. Berlin: Mouton de Gruyter [pre-print available at http://www.
uni-leipzig.de/~autotyp/download/schackowetal2009puma.pdf ].
Schneucker, Carl L., 1962. Kate language handbook. Madang: Lutheran Mission.
Scott, Graham, 1978. The Fore language of Papua New Guinea. Canberra: Pacific Linguistics
(PL-B47).
Stassen, Leon, 1985. Comparison and universal grammar. Oxford: Blackwell.
Suter, Edgar, 1992. Satzverbindung im Kate. Masters thesis, University of Zrich.
Takahashi, Hidemitsu, 2008. Imperatives in concessive clauses: compatibility between constructions. Constructions 2.
Tikkanen, Bertil, 1987. The Sanskrit gerund: a synchronic, diachronic, and typological analysis.
Helsinki: The Finnish Oriental Society.
Tikkanen, Bertil, 1995. Burushaski converbs in their South and Central Asian areal context.
In Haspelmath, Martin & Ekkehard Konig (eds.) Converbs in cross-linguistic perspective,
487528. Berlin: Mouton de Gruyter.
Van Valin, Robert D., Jr., 1995. Towards a functionalist account of so-called extraction constraints. In Devriendt, Betty, Louis Goossens, & Johan van der Auwera (eds.) Complex
structures: a functionalist perspective, 2960. Berlin: Mouton de Gruyter.
Van Valin, Robert D., Jr., 2005. Exploring the syntax-semantics interface. Cambridge: Cambridge
University Press. Van Valin, Robert D., Jr. & Randy J. LaPolla, 1997. Syntax: structure,
meaning, and function. Cambridge: Cambridge University Press.
Zeileis, Achim, David Meyer, & Kurt Hornik, 2007. Residual-based shadings for visualizing
(conditional) independence. Journal of Computational and Graphical Statistics 16, 507525.
Label
but
chain
or
cond
purp
and
chain
inconseq
sub
cvb
purp
com
Language
Amele
Amele
Amele
Amele
Amele
Belhare
Belhare
Belhare
Belhare
Belhare
Belhare
Belhare
conjunct
conjunct
disjunct
disjunct
-me, -V
fo ~o
fi
nu
finite
nonfinite
finite
finite
-lo(k)
-si
conjunct
conjunct
finite
nonfinite
nonfinite
conjunct
-sa
constraintfree
constraintfree
local
extensible nonfinite
extensible finite
extensible finite
local
extensible any
banned
ok
ok
ok
banned
harmonic
ok
banned
ok
ok
banned
banned
Finiteness ILL-mark
extensible any
local
conjunct
local
T-scope
-kone
local
-naa, -hu(go/ disjunct
do/cha), =cha
ki(na)(hu)
constraintfree
constraintfree
local
gba
=cha...=cha
ILL-scope
Marker
asymmetrical
asymmetrical
asymmetrical
constraint-free
asymmetrical
symmetrical
asymmetrical
constraint-free
symmetrical
asymmetrical
symmetrical
Symmetry
harmonic asymmetrical
banned
banned
banned
ok
ok
ok
ok
ok
ok
banned
ok
T-mark
ok
ok
ok
ok
ok
ok
NA
NA
ok
NA
NA
NA
WH
banned
banned
banned
banned
banned
possible
banned
NA
NA
NA
NA
NA
Extraction
Appendix: Pilot database on clause adjoining (NA = information not available)
ok
ok
ok
ok
ok
ok
banned
NA
NA
NA
NA
NA
FOC
flexible-relational
flexible-adjacent
flexible-adjacent
flexible-adjacent
flexible-relational
flexible-adjacent
flexible-adjacent
fixed:pre-main
flexible-relational
fixed:pre-main
fixed:pre-main
fixed:pre-main
Position
ad-V
ad-V
ad-V
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
Layer
Bickel (1993),
(14a), (26),
(40b), (56)
Roberts (1987,
1988), (7),
(34cd)
Roberts (1987,
1988), (1), (16),
(20), (32b),
(51), (59)
Roberts (1987,
1988), (32a)
Roberts (1987,
1988), (39)
Roberts (1987,
1988)
Bickel (1993),
(32)
Bickel (1993),
(3a), (17), (23),
(42), (46a), (47),
(54), (58)
Bickel (1993)
Bickel (1993),
(6), (40a),
(46bc), (57)
Bickel (1993),
(3b), (14b), (40c)
Bickel (1993)
References and
examples
Balthasar Bickel
disjunct
local
constraintfree
conjunct
disjunct
local
disjunct
local
disjunct
constraintfree
disjunct
local
-nach, -alc
etc.
-lie, -cha
tqa
and, but
if, when etc.

-ing
although
to, -ing
-ing
-ki
cvb
cvb.temp
and
and
go-and
fin. sub
cvbdetached
although
to.ing.
detached
cvb-adv
chain
sub
cause.
detached
purp
Chechen
Chechen
Chechen
English
English
English
English
German
German
German
German
German
German
English
Fore
cond.
detached
cvb-adv
and
disjunct
-na, -(u)sh, -i
chain
Chechen
English
English
extensible
-si, -r
chain
Chantyal
um zu, ohne zu disjunct

-nd
disjunct
und
constraintfree
wenn
local
weil, wenn
weil
and
constraintfree
NA
n(V)---(i)n
Bunshaski chain
local
conjunct
conjunct
local
local
local
conjunct
conjunct
local
conjunct
conjunct
conjunct
conjunct
local
local
conjunct
local
conjunct
finite
nonfinite
nonfinite
finite
finite
finite
nonfinite
finite
nonfinite
nonfinite
finite
nonfinite
finite
finite
finite
nonfinite
nonfinite
nonfinite
ok
banned
banned
ok
banned
ok
banned
ok
ok
banned
banned
banned
harmonic
ok
ok
ok
banned
banned
NA
NA
symmetrical
symmetrical
symmetrical
asymmetrical
asymmetrical
asymmetrical
asymmetrical
asymmetrical
symmetrical
asymmetrical
ok
banned
banned
ok
ok
ok
symmetrical
asymmetrical
asymmetrical
symmetrical
constraint-free
symmetrical
banned
asymmetrical
ok
banned
harmonic symmetrical
banned
asymmetrical
ok
ok
ok
banned
ok
banned
banned
banned
possible
banned
banned
banned
banned
banned
NA
NA
banned banned
ok
banned
banned banned
ok
banned
banned banned
banned banned
banned possible
NA
NA
banned banned
banned banned
banned banned
banned banned
ok
ok
ok
ok
ok
ok
NA
NA
banned
ok
ok
ok
ok
banned
ok
NA
ok
ok
ok
ok
ok
ok
NA
ok
ok
banned
NA
banned
fixed:pre-main
flexible-relational
fixed:post-main
flexible-adjacent
flexible-relational
fixed:post-main
fixed:post-main
flexible-adjacent
flexible-relational
fixed:pre-main
flexible-relational
fixed:post-main
flexible-adjacent
flexible-adjacent
flexible-adjacent
flexible-relational
flexible-relational
flexible-relational
flexible-adjacent
flexible-adjacent
detached
ad-V
ad-V
detached
ad-S
detached
ad-V
ad-S
ad-S
detached
ad-S
detached
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
Scott (1978),
(22)
Lakoff (1986),
etc.
Tikkanen
(1995), (13)
Noonan (1999),
(18)
Molochieva
(2008), (10),
(24), (36a), (37),
(48), (55)
Molochieva
(2008), (36b)
Molochieva
(2008), (25), (49)
Molochieva
(2008)
Ross (1967)

conjunct
disjunct
disjunct
conjunct
constraintfree
local
yi
n%
-men, -nt
-rohI, -rehI
-ga
chain
topic
part.coni.
alter.iter
ds-chain
topic
inconseq
cond
ss-chain
chain
chain
serial
topic
chain
chain
topic
cvb
Godi
Godi
Greek
(Ancient)
Hua
Hua
Hua
Hua
Hua
Hua
Korafe
Kte
Kte
Kte
Lenakel
Nepali
Nepali
Nepali
disjunct
conjunct
constraintfree
disjunct
constraintfree
mutsa
m-era
bhane
-era
conjunct
local
local
conjunct
constraintfree
-ra, -me, -ku, ... constraintfree
-mana
-si
-o, -i,
-ma
disjunct
-nd
cvb
German
ILL-scope
Marker
Label
Language
nonfinite
nonfinite
nonfinite
nonfinite
nonfinite
NA
nonfinite
nonfinite
nonfinite
nonfinite
nonfinite
finite
extensible finite
conjunct nonfinite
local
conjunct
local
conjunct
local
local
conjunct
local
local
conjunct
conjunct
banned
banned
banned
banned
NA
banned
banned
banned
banned
banned
NA
banned
banned
banned
banned
banned
banned
banned
Finiteness ILL-mark
conjunct nonfinite
local
finite
conjunct
T-scope
symmetrical
asymmetrical
constraint-free
asymmetrical
asymmetrical
Symmetry
ok
ok
ok
ok
ok
banned
banned
banned
ok
banned
ok
ok
constraint-free
asymmetrical
asymmetrical
asymmetrical
asymmetrical
asymmetrical
asymmetrical
asymmetrical
asymmetrical
asymmetrical
asymmetrical
constraint-free
banned
banned
ok
banned
banned
T-mark
Extraction
NA
banned
NA
NA
NA
ok
ok
NA
ok
ok
NA
ok
banned
banned
ok
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
banned NA
ok
NA
NA
NA
NA
banned banned
WH
ok
ok
NA
ok
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
ok
FOC
flexible-relational
flexible-adjacent
flexible-adjacent
flexible-adjacent
NA
fixed:pre-main
fixed:pre-main
fixed:pre-main
fixed:pre-main
fixed:pre-main
fixed:pre-main
fixed:pre-main
fixed:pre-main
flexible-adjacent
flexible-adjacent
flexible-relational
flexible-relational
flexible-relational
Position
ad-S
ad-V
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
ad-S
Layer
fieldnotes
fieldnotes, (43)
Pilhofer (1933),
Suter (1992),
(41)
Pilhofer (1933),
Suter (1992)
Pilhofer (1933),
Suter (1992)
Lynch (1978)
fieldnotes, (4)
Haiman (1980),
(31)
Haiman (1980),
(27)
Haiman (1980),
(28)
Haiman (1980)
Haiman (1980)
Haiman (1980)
Farr (1999), (29)
Marchese (1977)
Marchese (1977)
(19)
References and
examples
Balthasar Bickel
inconseq
topic
chain
serial
chain
chain.swr
topic
cause
chain
sub
Tauya
Tauya
Toura
Turkish
Turkish
Usan
Usan
Usan
Wambule
Yupik
(Central)
Swahili
Swahili
Swahili
Swahili
Tauya
cvb-adv
cvb.
detached
chain
and.inf
and
topic
chain
constraintfree
disjunct
-kho
NA
disjunct
end
le
-Ip
local
conjunct
conjunct
conjunct
=ra
eng
local
-na
extensible
local
kana
na
ki
-pa, -te/-fe/
-tefe
NA
conjunct
conjunct
conjunct
disjunct
constraintfree
-a, -v
-a, -v
and
constraintfree
constraintfree
disjunct
disjunct
-en
chain
Russian
Russian
Newar
(Dolakha)
Russian
finite
any
any
nonfinite
nonfinite
nonfinite
nonfinite
finite
nonfinite
nonfinite
nonfinite
finite
nonfinite
nonfinite
nonfinite
nonfinite
finite
nonfinite
local
NA
local
conjunct
conjunct
conjunct
conjunct
NA
local
conjunct
conjunct
local
conjunct
conjunct
conjunct
conjunct
local
NA
banned
NA
NA
banned
NA
banned
banned
ok
ok
ok
banned
banned
ok
banned
ok
banned
banned
ok
NA
ok
ok
NA
ok
ok
banned
banned
banned
NA
ok
banned
banned
ok
banned
harmonic
banned
banned
ok
banned
constraint-free
asymmetrical
asymmetrical
constraint-free
asymmetrical
asymmetrical
asymmetrical
asymmetrical
constraint-free
asymmetrical
asymmetrical
asymmetrical
symmetrical
asymmetrical
asymmetrical
asymmetrical
asymmetrical
symmetrical
asymmetrical
banned
NA
banned
NA
NA
NA
NA
possible
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
possible
NA
NA
NA
NA
banned banned
NA
NA
NA
NA
NA
ok
banned banned
banned banned
ok
NA
NA
NA
NA
NA
NA
NA
NA
ok
NA
NA
NA
NA
NA
ok
NA
ok
ok
ok
ok
flexible-relational
flexible-adjacent
flexible-adjacent
fixed:pre-main
flexible-adjacent
flexible-adjacent
fixed:pre-main
fixed:pre-main
fixed:pre-main
fixed:pre-main
fixed:post-main
fixed:post-main
flexible-adjacent
flexible-relational
flxed:pre-main
flexible-relational
fixed:pre-main
flexible-adjacent
flexible-adjacent
Genetti (2005)
ad-V
detached Rappaport
(1984)
ad-S
fleldnotes, (2)
ad-S
fleldnotes
ad-S
fleldnotes
ad-S
fleldnotes
ad-S
MacDonald
(1990), (5),
(37a), (40)
ad-S
MacDonald
(1990)
ad-S
MacDonald
(1990), (8),
(35b), (45)
ad-S
Bearth (1986)
ad-S
fleldnotes
ad-S
Johanson (1995),
(11a), (21)
ad-S
Reesink (1987),
(9)
ad-S
Reesink (1987),
(34ab), (52)
ad-S
Reesink (1987),
(53)
ad-S
Opgenort
(2004), (30)
ad-S
Reed et al.
(1977)
ad-S
ad-S

part ii
Clause-chaining, converbs, masdars,

absolutive constructions, etc.
Specialized converbs and adverbial

subordination in Axaxdr Akhvakh
Denis Creissels
Universit Lumire (Lyon2), Fdration Typologie et Universaux

linguistiques CNRS
Northern Akhvakh, like other Nakh-Daghestanian languages, has a large
inventory of converbs specialized in the expression of various semantic types of
adverbial subordination: locative, temporal, conditional, concessive, similative,
gradual, and purposive. After outlining the general characteristics of Akhvakh
morphosyntax, this paper describes the morphology and the uses of the
individual converbs, and discusses the possibility of reconstructing their origins.
Attention is drawn to the high proportion of etymologically transparent converbs,
which suggests that the particular type of morphosyntactic organization
characteristic of Nakh-Daghestanian languages favors grammaticalization
processes that do not modify the overall organization of the adverbial
subordination system, but result in a constant and relatively rapid renewal
of the inventories of specialized converbs.
1. Introduction
Akhvakh (aWali mici, Russian axvaxskij jazyk) belongs to the Andic (sub)branch
of the Northeast Caucasian (or Nakh-Daghestanian) family.1 Like the other Andic
languages, Akhvakh has no writing tradition.2 According to Magomedova & Abdulaeva
2007, Akhvakh has approximately 20 000 speakers. Four dialects are traditionally
. The other Andic languages are Andi, Bagvalal, Botlikh, Chamala, Godoberi, Karata, and
Tindi. None of them has a particularly close relationship to Akhvakh. Andic languages are
traditionally grouped with Avar and Tsezic languages into a single branch of the Northeast
Caucasian family. The other branches of the Northeast Caucasian family are Lak, Dargi (or
Dargwa), Lezgi, Khinalug (sometimes considered a marginal member of the Lezgi branch),
and Nakh.
. The transcription used in this paper departs from the API conventions on the following
points: alveolar voiceless affricate c; palato-alveolar fricatives (voiceless) and (voiced);
palato-alveolar affricates (voiceless) and (voiced); lateral voiceless affricate L; the macron is
used for long vowels and strong consonants.
Denis Creissels
recognized. One of them is designated as Northern Akhvakh, whereas the other three
are grouped under the label Southern Akhvakh.
Northern Akhvakh is spoken in four villages of the Axvaxskij Rajon in the
western part of Daghestan (Tadmagitl, Lologonitl, Kudijab-Roso, and Izani), in recent
settlements in the lowlands of Daghestan, and in Axaxdr near Zaqatala (Azerbaijan),
where I carried out fieldwork on Akhvakh. Axaxdr Akhvakh (henceforth AD
Akhvakh) is a variety of Northern Akhvakh very close to those spoken in the Axvaxskij
Rajon of Daghestan (henceforth AR Akhvakh), presented in Magomedbekova 1967
and Magomedova & Abdulaeva 2007.3 The Southern Akhvakh dialects are each
spoken in their own village (Cegob, Tljanub and Ratlub), all situated in the amilskij
Rajon of Daghestan.
The analysis of the Akhvakh specialized converbs proposed in this paper is based
on texts collected in Axaxdr between June 2005 and April 2008.4
In accordance with Nedjalkov (1995), the term converb is understood here as
referring to non-autonomous verb forms (i.e. verb forms devoid of the ability to head
independent clauses) that depart from other types of non-autonomous verb forms
(infinitives, masdars, participles) by the fact that they do not occur in either complement clauses or relative clauses. This broad definition follows the Altaicist tradition
from which the term converb originally stems. In the description of Daghestanian
languages, a more restrictive definition according to which converbs are non-finite
verb forms specialized in adverbial subordination (Haspelmath 1995a) may be
problematic for the following two reasons:
In the verbal system of Daghestanian languages, the distinctions morphologically

expressed in verb forms may show no simple correlation with finiteness as a
feature of predicative constructions (Section2.4.6 provides a brief discussion
of the morphological correlates of finiteness in Akhvakh; for a more detailed
discussion, see Creissels (2009)).
. Judging from Magomedbekova 1967 and Magomedova & Abdulaeva 2007, the variety
of Akhvakh spoken in Axaxdr does not differ significantly from the varieties of Northern
Akhvakh spoken in the Axvaxskij Rajon, and this judgment was confirmed without any
reservations by Indira Abdulaeva, co-author of the Akhvakh-Russian dictionary and a native
speaker of Northern Akhvakh herself, who spent one week in Axaxdr while I was carrying
field-work there in April 2008. I have been able to find no precise information as to when
Akhvakh migrants began to settle in Axaxdr, but in Axaxdr there are still old people who
were born in Daghestan, and whose parents migrated to Azerbaijan at the end of the second
world war, when the economic situation in Daghestan was particularly difficult.
. The analysis presented here has benefited from the comments and suggestions of Bernard
Comrie, Misha Daniel, and Isabelle Bril. Any remaining faults are of course my own.
The reference to adverbial subordination in the definition of converb is not problematic for the specialized converbs that constitute the main topic of this paper, but it
may be problematic for the general converb, which at least in some languages occurs
in constructions analyzable as instances of clause coordination (see Section3).5
The distinction between contextual and specialized converbs, put forward by

Nedjalkov in a paper that appeared in 1990 in Russian and was later published in
English in Haspelmath & Knigs volume on converbs (Nedjalkov 1995), has been
used among others by Haspelmath (1995b) for Lezgian: contextual converbs leave
the precise nature of the semantic link between the clauses open, whereas specialized
converbs carry quite specific adverbial meanings. Given the typological homogeneity
of the Nakh-Daghestanian language family, it is not surprising that this distinction
proves useful for the description of Akhvakh too. As pointed out by Bisang (1995:156),
generally speaking, the distinction between contextual and specialized converbs must
certainly be conceived of as gradual rather than discrete. However, in Akhvakh, there
is a clear-cut distinction between multipurpose converbs undertaking a variety of
functions, and converbs specialized in particular types of adverbial subordination.
This paper is organized as follows. Section 2 outlines the main characteristics
of Akhvakh morphosyntax: clause structure (Section2.1), nouns and noun phrases
(Section2.2), verb inflection (Section2.3). Sections 3 and 4 briefly present the general converb and the progressive converb respectively. Section 5 discusses the general
properties of specialized converbs. Sections 6 to 12 are devoted to individual converbs or semantic groups of converbs: the locative converb (Section 6), temporal
converbs (Section7), conditional and concessive converbs (Section8), the similative converb (Section9), the gradual converb (Section10), the explicative converb
(Section 11), and the purposive converb (Section 12). Each section or subsection
includes not only illustrations of the use of the converbs, but also a discussion on the
possibility of reconstructing their origins.
2. General remarks on Akhvakh morphosyntax

2.1 Clause structure
Akhvakh clause structure is characterized by flexible constituent order. Case marking and gender-number agreement between the verb and its core arguments are
. For a general discussion of the notion of converb, see Haspelmath (1995a), Nedjalkov
(1995), Knig (1995), Bisang (1995), van der Auwera (1998), Bickel (1998), Ziga (1998),
Haspelmath (1999), Ylikoski (2003).
Denis Creissels
consistently ergative. In contrast, assertive agreement (see Section2.4) follows a split

intransitive pattern.
Arguments whose identity is recoverable from the context can freely be omitted,
and unexpressed arguments receiving an arbitrary interpretation are common too.
The causative is the only valency-changing mechanism systematically expressed
via verb morphology or grammaticalized periphrasis.
2.2 Nouns and noun phrases

Three semantically transparent agreement classes of nouns are distinguished in the
singular: human masculine (M), human feminine (F), and non-human (N).6 In the
plural, the distinction masculine vs. feminine is neutralized, resulting in a binary
opposition human plural (HPL) vs. non-human plural (NPL). Noun morphology
shows only frozen vestiges of gender prefixes.
In canonical NPs, the head noun is in final position and is inflected for number
and case. Noun dependents other than adjectives show no agreement marking, and
the agreement morphology of attributive adjectives is reduced in comparison with AR
Akhvakh and the other Andic languages.7
In the absence of a head noun, the last word in the NP, whatever its nature, is
marked for gender, number and case.
Number inflection for nouns is irregular and involves considerable free variation.
The absolute form of nominals (used in the extra-syntactic function of quotation
or designation and in S or P roles) has no overt marking. Case suffixes may attach to a
stem identical with the absolute form, or to a special oblique stem. In the singular, the
formation of the oblique stem is very irregular and involves considerable free variation. In the plural, the formation of the oblique stem is more regular. In particular,
oblique stem markers expressing class distinctions (M su, F/N Si, HPL lo, NPL
le ~ li) are used more systematically in the plural than in the singular.
Case inflection includes the following cases:
ergative (de),
dative (La),
. The only exceptions to the semantic rule of class assignment are de person and mike
child, which in the singular trigger N agreement, whereas the corresponding plural forms do
and mikeli regularly trigger HPL agreement.
. In AR Akhvakh, all noun dependents in canonical NPs optionally take class suffixes
agreeing with the head noun, but in the data I collected in Axaxdr, noun dependents
other than adjectives never occur with agreement marks in canonical NPs, and the suffixal
agreement of adjectives never occurs in classes other than HPL.
genitive ( or -li),8
comitative (kena),
purposive (tana),
five series of spatial cases encoding different spatial configurations,9 with three
cases in each series: essive i or e, lative a, and elative u(ne).10
Personal pronouns have irregular inflection but show the same case distinctions as
nouns, and the distinction between three spatial cases also applies to locative adverbs.
There are two possible constructions for NP coordination: either NP1kena NP2,
where kena is the comitative case suffix (also used for comitative or instrumental
adjuncts), or NP1la NP2la, where la is an additive particle (glossed add) also
found in contexts in which it corresponds to English also, in turn, or even. See
Section3 for more details on this particle.
2.3 Adjectives
Like verbs (see Section3.4.2), adjectives divide into those which obligatorily carry
a class agreement prefix, and those devoid of it. Like nouns, they cannot bear TAM
inflection and fulfill the predicate function by combining with the copula godi or with
the verb bikuruLa be.11
In Axaxdr Akhvakh, adjectives in the role of noun dependent or in predicate
function do not show suffixal inflection, whereas nominalized adjectives (i.e. adjectives
occurring as the last word of a noun phrase) are inflected for plural and take suffixed
. In principle, the zero-marked genitive characterizes M and HPL NPs, whereas -li is used
with F, N or NPL NPs, but this rule is not very strict, and variations are observed.
. The Akhvakh system of spatial cases departs from the typical Daghestanian pattern in that
one of the series (the g series) is a default series that does not encode a particular spatial configuration, and spatial configurations tend to be encoded by combining NPs showing default
spatial case marking with locative adverbs showing parallel spatial case inflection, rather than
via traditional case marking. Moreover, the characteristic consonant of the g series is not
always apparent, due to morphophonological processes (for example, aha-Si-ga to the town
can optionally be pronounced aha-Sa).
. Magomedbekova (1967) identified u as ablative proper, and une as perlative, but in AD
Akhvakh, these two endings are in free variation.
. In Akhvakh, non-verbal predications involving neither the copula nor the verb bikuruLa
be are exceptional in statements. By contrast, the omission of the copula regularly occurs
in questions.
Denis Creissels
class marks.12 In the absolute form, the class marks suffixed to nominalized adjectives
are M we, F je, N be, HPL ji, NPL re, whereas in combination with overt case
markers, the class markers suffixed to nominalized adjectives are identical to the
oblique stem markers found in the case inflection of some nouns (M su, F/N Si,
HPL lo, NPL le ~ li).
2.4 Verb inflection

Independent verb forms are inflected for TAM, polarity, and gender-number agreement;
TAM and polarity are conjointly expressed by portemanteau markers. Assertive
agreement (see Section2.4.3 below) is morphologically distinct from gender-number
agreement and occurs in one tense only.
In addition to the synthetic forms listed in Section2.4.1, AD Akhvakh has analytic
verb forms with the copula godi or the verb bikuruLa be in auxiliary function.
2.4.1 TAM-polarity marking in independent verb forms

The TAM/polarity inflections of verbs heading independent clauses include the
following possibilities:
pf (perfective): hpl iri, other classes ari

pf (perfective): hpl idi, other classes ada(-cl)
pf.neg (perfective negative): iLa (-cl)
ipf (imperfective): iri
ipf (imperfective): hpl idi, other classes ida(-cl)
ipf.neg (imperfective negative): iki
ipf.neg (imperfective negative): ika(-cl)
uw (past unwitnessed): m u-wi(di), f i-wi(di), n a-wi(di), npl ari-wi(di)
uw.neg (past unwitnessed negative): m iL-u-wi(di), f iL-i-wi(di), n iL-a-wi(di),
npl iL-ari-wi(di)
mir (mirative): m u-wa, f i-wa, n a-wa, hpl aji, npl ari-wa
mir.neg (mirative negative): m u-u-wa, f u-i-wa, n u-a-wa, hpl u-aji,
npl u-ari-wa
pot (potential): m/n -u-wa, f i-wa, hpl oji, npl uri-wa
imp (imperative): a
proh (prohibitive): uba
opt (optative): ala
opt.neg (optative negative): ubala
. In AR Akhvakh, attributive or predicative adjectives optionally show gender-number

suffixes.
This list of independent synthetic verb forms calls for the following remarks:
a. The forms labeled perfective and perfective carry distinctions in the involvement
of the assertor (i.e. the speaker in assertive clauses, the addressee in questions) in the
event referred to: in assertions, pf is typically used with transitive verbs involving
a 1st person A, or intransitive verbs representing controllable events involving
a 1st person S, whereas in questions, pf is typically used with transitive verbs
involving a 2nd person A, or intransitive verbs representing controllable events
involving a 2nd person S (see Creissels 2008a & 2008b for more details). This
distinction is neutralized in the negative.
b. The distinctions pf vs. pf and ipf vs. ipf seem to be historically related, since
the pf and ipf suffixes share the ending ri, contrasting with da common to the
pf and ipf suffixes. Moreover, pf and ipf share the possibility of being used as
participles, whereas pf and ipf are strictly finite verb forms. However, functionally, the choice between ipf and ipf in their use as independent verb forms clearly
plays upon aspecto-modal distinctions, and has nothing to do with distinctions
in assertor involvement, nor more generally with person distinctions: both forms
can express habitual meaning, but there is a clear tendency (particularly strong in
AD Akhvakh) to prefer ipf as the marker of habitual aspect, whereas ipf is used
in modal contexts where it is impossible to substitute ipf for it. In AD Akhvakh,
ipf is particularly common as a narrative tense (historical present), whereas ipf
is never used in this function.
c. uw (past unwitnessed) is typically used to refer to events known by hearsay. It
occurs in inferential contexts too, but this use is much less common. This tense
has no hpl form, and in contexts in which it could be expected to occur with hpl
agreement, Akhvakh speakers use the perfect (an analytic tense consisting of the
general converb of the auxiliated verb and the copula in auxiliary function).
d. mir (mirative) is a verb form encountered mainly (but not exclusively) in questions.
This form is particularly common (in fact, almost obligatory) in why-questions,
which suggests analyzing it as expressing surprise.
2.4.2 Gender-number agreement

Gender-number agreement of verbs involves both prefixes and suffixes, with two
different kinds of conditioning:
The presence of gender-number prefixes in verb forms involves no grammatical

conditioning. Verbs divide into two phonologically and semantically arbitrary
morphological classes, those that cannot occur without a gender-number prefix
indexing the S or P argument (i.e. to the argument encoded by an NP in the
absolute form), and those that never take such a prefix.
Denis Creissels
By contrast, the presence of gender-number suffixes indexing the S or P argument

is conditioned by the grammatical nature of the verb form. The rules governing
the presence and the phonological realization of gender-number suffixes in verb
forms are complex. In some verb forms, gender-number suffixes do not occur
at all; in others, obligatory gender-number agreement marks merge with TAM/
polarity markers; in a third group of verb forms, gender-number suffixes are
optional, and when they are present they may appear as distinct segments, or
merge with TAM/polarity markers.
The presence of gender-number prefixes or suffixes in verb forms depends therefore

on a complex combination of lexical and grammatical factors, but the agreement rule
itself is very simple, at least in the general case: when gender-number markers are
present in a verb form, they index the argument in S or P role, represented by an NP
in the absolute form. Exceptions to this rule are however observed in some complex
constructions in which the suffixal agreement of dependent verb forms may be
controlled externally. Such a construction is presented in Section3.
2.4.3 Participles
AD Akhvakh has four participles. Each one is characterized by a stem homonymous
with one of the independent verb forms listed above: perfective ada, perfective
negative iLa, imperfective ida, and imperfective negative ika.
2.4.4 Dependent verb forms
Strictly dependent verb forms include an infinitive u(ruLa), a general converb (M o(ho),
F e(he), N e, HPL i, NPL ere), a progressive converb (M ero, HPL eri, other classes
ere), and several specialized converbs whose description constitutes the topic of
this paper.
AD Akhvakh also has a verbal noun e, which is rarely found however with dependents treated as dependents of a verbal head. Contrary to the verbal noun (or masdar)
of most Caucasian languages, the verbal noun of Akhvakh tends to also behave like a
noun with respect to its internal syntax.
2.4.5 Morphological manifestations of finiteness
A striking particularity of Akhvakh is that there is no simple correlation between
the morphological characteristics of verb forms and the finiteness properties of
the predicative constructions they head (i.e. their ability to be used as independent
clauses with particular illocutionary values and/or to be involved in particular types
of complex structures).
Table 1. Suffixed class agreement markers in Akhvakh verb forms

Independent verb forms
Dependent verb forms
a. Suffixes expressing obligatory

class agreement with more than
two possible values
potential
past unwitnessed
mirative
general converb
b. Suffixes including an obligatory

HPL vs. other classes distinction, and
compatible with additional suffixes
optionally expressing class agreement
with other classes
perfective positive
imperfective positive
c. Suffixes including no obligatory

class agreement, but compatible
with optional class markers
perfective negative
imperfective negative
progressive converb
similative converb
d. Suffixes expressing an obligatory

HPL vs. other classes distinction,
but without the possibility of optional
class agreement with other classes
perfective positive
conditional converb
posterior converb
inceptive converb
e. Suffixes that never include class

agreement markers
imperfective positive
imperfective negative
imperative
prohibitive
infinitive
verbal noun
verbal locative
simultaneous converb
immediate converbs
imminent converb
anterior converb
non-posterior converb
concessive converb
gradual converb
explicative converb
purposive converb
In particular, there is no simple relationship between the way Akhvakh verb

forms behave with respect to class agreement with the absolutive argument, and
theirability to head independent clauses, as can be seen from the classification of AD
Akhvakh verb forms depending on the presence/absence of suffixed class markers
given in Table 1. In this table, verb forms are divided into those that have the ability
to head independent clauses, and those that are only found in complex constructions.
3. The general converb

This section provides a brief account of the main characteristics of the general converb
(glossed cvb). A more detailed analysis can be found in (Creissels Submitted).
Denis Creissels
The uses of the general converb can be classified into three broad types:
a. It may add manner specification to an event encoded by an independent verb
form, as in Example (1).
(1) a. jae qel-a
j-et-e
j-ni.

girl home-lat f-run-cvb.f f-go.ipf*13

The girl ran home. (lit. the girl went home running)
b. hu-be aqilo
b-eq-aj-e
g-x-ida.

dem-n intelligence n-work-caus-cvb.n do-oblg-ipf*

This must be done cleverly.
b. It may encode an event viewed as the first stage of a complex event whose second
stage is encoded by an independent verb form, as in Example (2).
ako-de rili b-i-e
q-ni.
wife-erg meat n-cook-cvb.n eat-ipf*
The wife cooked the meat and ate it.
b. mola-sW-e
tai-gunu gora b-e-e
hora-ge

Molla-m-erg pocket-el bread n-take-cvb.n lake-ess
(2) a.

gel-i
tk-ini.
inside-ess dip-ipf*
Molla took some bread from his pocket and dipped it into the lake.
c. It combines with the copula or the verb bikuruLa be in auxiliary function to give
analytic tenses semantically similar to the English perfect, as in Example (3).
(3) a. iWada-sW-e
lgi b-iqW-e
godi.

shepherd-m-erg sheep n-kill-cvb.n cop.n

The shepherd has killed a sheep.
b. iWada-sW-e
lgi b-iqW-e
b-ik-awi.

shepherd-m-erg sheep n-kill-cvb.n n-be-uw.n

The shepherd (reportedly) had killed a sheep.
The status of constructions that constitute the translational equivalent of English

clause coordination but involve non-autonomous verb forms, as in Example (2) above,
is notoriously difficult to establish with respect to the traditional notions of coordination
. In the example glosses, the asterisk signals verb forms involving morphophonological
processes, the underlying representation of which is given in an annex at the end of the article.
and subordination. In the literature, at least three different types of analyses can be
found for similar constructions:
They may represent instances of co-subordination, with coordinate (parallel)

and subordinate (asymmetrical) properties co-occurring within one and the
same sentence.
Depending on semantic factors, they may show variations in their syntactic
properties that justify analyzing them as instantiating coordination in some of
their uses, and subordination in others. Among Daghestanian languages, such
an analysis has been advocated by Kazenin & Testelec (1999) for Tsakhur, and by
Polinsky (2007) for Tsez.
In spite of their translational equivalence with clause coordination, they may
uniformly show syntactic behavior consistent with a subordination analysis, as
argued by Kazenin (2001) for Bagvalal, a close relative of Akhvakh.
In this respect, Akhvakhs situation is similar to that described by Kazenin for Bagvalal.
I have come across no obvious case where the use of the general converb would contradict the subordination analysis, whereas the texts I collected include many occurrences of the general converb in constructions in which the coreference relationships
between full NPs and pronouns are organized in a way commonly considered incompatible with coordination. Additional observations incompatible with a coordination
analysis can be made on linear order, embedding, relativization, and negation, and a
particularly strong argument in favor of the subordination analysis is that Akhvakhs
general converb may show double agreement, one of the agreement markers then being
governed by the S/P argument of the main verb. As already explained, Akhvakh verbs
may have two slots (a prefixal one and a suffixal one) for gender-number agreement.
The availability of the prefixal slot is lexically determined, whereas the availability of
the suffixal slot depends on the verb forms TAM value. Prefixal agreement is invariably
controlled internally, i.e. by the S/P argument of the verb. In most constructions, the
same rule applies to suffixal agreement, but in complex constructions, it may happen
that a dependent verb form shows external agreement, i.e. agreement controlled by an
NP that does not belong to the clause headed by the dependent verb form. In particular,
the suffixal agreement of the general converb may be controlled by the S/P agreement
of the main verb, as illustrated by Example (4). 14
. The analysis of external agreement in the converbal construction of Akhvakh constitutes

the main topic of (Creissels Submitted). External agreement is also observed in the participial
construction, in which verb forms in participle function agree both with their S/P argument
and with the noun they modify.
Denis Creissels
(4) mola rasadi w-uL-i

Wela-la
m-ne
Molla Rasadi m-die-cvb.hpl graveyard-illat hpl-go.prog*
ba-k-i
goli.
hpl-be-cvb.hpl cop.hpl
Molla Rasadi died, and they were going to the graveyard.
In this sentence, the masculine singular prefix of wuLi having died expresses agreement with the S argument mola rasadi, whereas the only possible explanation for the
human plural suffix is that it expresses agreement with the understood S argument of
the main verb.
4. The progressive converb

The progressive converb is characterized by a suffix ere, glossed prog, with optional
class agreement in classes M and HPL: the form ere can be used for all classes, and this
is the only possible form in classes F, N and NPL, whereas M and HPL agreement can
optionally be expressed via the forms ero and eri respectively.
Apart from its use in the formation of analytic tenses (see below), this converb
occurs exclusively in combination with the copula or the verb bikuruLa be in existential
function, or with one of the following three verbs: biuruLa remain, hariguruLa see,
and miunuLa find. Note that hariguruLa and miunuLa both imply an experiencer
in the dative case.
The only possible construction for the progressive converb is a control
construction in which the S/A argument of the progressive converb cannot be
expressed and is obligatorily identified with the absolutive argument of the main
verb, as illustrated by Example (5) to (8).
Note that, with the progressive converb of transitive verbs, if suffixed agreement
markers are present, they reflect the class of the A argument identified with the S
argument of the higher verb, whereas class prefixes reflect the class of the P argument.
(5) a. e rua-Si-ge La
e totokiro
b-ik-awi

one tree-n-ess on.ess one woodpecker n-be-uw.n

[rua-Si-ge or-ere].

tree-n-ess knock-prog
On a tree there was a woodpecker knocking on the tree.
b. huLe
[lga
k-no]
iWada
w-uk-uwi.

up.there.ess sheep.pl eat-caus.prog.m* shepherd m-be-uw.m*

Up there, there was a shepherd feeding his sheep.
(6) a.

di-La
mike [-re]
m-i-ani.
1sg-dat baby cry-prog* n-find-pf*
I found the baby crying.
b. di-La
hudu-we w--ani [qeleko b-iqW-ero].

1sg-dat dem-m m-find-pf* cock n-kill-prog.m

I found him killing the cock.
(7) a.

di-La
baa harigW-ari [ema q-ne].
1sg-dat wolf see-pf
cow eat-prog*
I saw the wolf eating the cow.
b. di-La
hudu-we harigw-ari [Suda
b-uq-ero].

1sg-dat dem-m see-pf
firewood n-cut-prog.m

I saw him cutting wood.
(8) mike [-re]
b-iW-ari.
baby cry-prog* n-remain-pf
The baby kept crying.
But, as already mentioned above, the progressive converb is mainly found in analytic
tenses expressing progressive meaning, as illustrated by Example (9).
(9) a.

hudu-sW-e Suda b-uq-ere

godi.
dem-m-erg wood n-cut-prog cop.n
He is cutting firewood.
b. hudu-sW-e Suda b-uq-ere

b-ik-awi.

dem-m-erg wood n-cut-prog n-be-uw.n

He was (reportedly) cutting firewood.
These analytic tenses quite obviously result from the grammaticalization of the
construction in which the progressive converb is subordinated to the copula or the
verb bikuruLa in existential predicate function. Note the following three differences
between the progressive tenses and the construction from which they originate:
The copula or the verb bikuruLa in auxiliary function must be immediately

preceded by the progressive converb of the auxiliated verb, whereas in the construction in which the progressive converb is subordinated to the copula or the
verb bikuruLa in existential predicate function, the verb phrase headed by the
progressive converb has no fixed place within the matrix clause.
In progressive tenses, with transitive verbs, the agreement of the progressive
converb can reflect the class of the P argument, whereas the progressive converb
of transitive verbs subordinated to the copula or the verb bikuruLa in existential
predicate function shows suffixal agreement with A.
In progressive tenses, with transitive verbs, the NP representing the A argument
can be in the ergative case, as in Example (9), but when the progressive converb of
transitive verbs is subordinated to the copula or the verb bikuruLa in existential
predicate function, the A argument must be in the absolute form, as in Example(7b),
since it is expressed as the S term of an intransitive predication.
Denis Creissels
I have no etymological hypothesis to put forward for this converb. Its syntactic distribution could suggest a participial origin, but there does not seem to exist any concrete
evidence supporting this hypothesis.
5. The specialized converbs: Introductory remarks

5.1 The expression of core arguments in converbal clauses
A general property of the specialized converbs described in the following sections is that
their arguments can always be expressed exactly in the same way as in an independent
clause, irrespective of the fact that converbs may behave very differently with respect to
agreement with the absolutive argument. This is illustrated in Example (10) by complex
sentences involving the negative form of the posterior converb (a), the immediate
converb (b), the imminent converb (c), and the conditional converb (d).
(10) a.

[mamaali-de reLa dan-iL-eSi]

imo-de
Mehmet-Ali-erg hand draw.away-neg-post father-erg
el-awi, w--a!
say-uw.n m-go-imp
As Mehmet-Ali insisted (lit. did not draw away his hand), his father said, Go!.
b. [Se
W-ikena] ka
gi-ga
tam-a!

water boil-immed dumpling.pl cooking.pot-lat put-imp

As soon as the water boils, put the dumplings in the cooking-pot!
c. [cari c-daSa]
am-e b-oil-ri.

rain fall-immin* mow-vn n-end-pf*

The hay-harvest was up just before it rained.
d. [me-de co-ge
La
qdiro gin-aj-e
m-i-ala],

2sg-erg door-ess on.ess sickle hang-caus-cvb.n n-be-cond*

isi ea m-da
golidi.
1ple away hpl-go.ipf* cop.hpl
If you have hung a sickle on the door, we will go away.
5.2 Missing arguments in clauses headed by specialized converbs

A missing argument in a clause headed by a specialized converb may be identified
with an argument of the main verb, but the identification of missing arguments in the
construction of specialized converbs is a purely pragmatic phenomenon, which does
not play upon syntactic constraints. Depending on the context, a missing argument in
the construction of a converb can always be identified a speech act participant, or with
any other salient referent.
For example, the sentence in (11) is given with the translation corresponding to
its interpretation in the context in which I found it, but the same sentence in different
contexts could equally be interpreted as While sleeping, I saw a ghost, While I slept,
he saw a ghost, While he slept, I saw a ghost, While hei slept, hej saw a ghost, etc.
The only constraint is that the missing argument of sleep must be masculine singular,
since the converbal form lkideSi shows masculine singular agreement; the interpretation of the missing argument of see is totally open, since the form harigWiri shows
no agreement marking, and there is no co-reference or disjoint reference constraint
between the missing argument of the converb and any of the main verbs arguments.
(11) [raLa lk-ideSi],
e b-aida Le-la
i-ada
at.night msleep-simult one n-white cloth-add wear-pf
otala harigW-iri.
ghost see-ipf
At night while sleeping, he saw a ghost wearing a white cloth.
5.3 Linear order in constructions involving specialized converbs

5.3.1 Converbal clauses preceding the main verb
Most of the time, in complex constructions involving a converb, the converbal clause
precedes the main verb. Converbal clauses in sentence initial position, followed by all
other phrases belonging to the construction of the main verb, are particularly common.
However, terms belonging to the construction of the main verb may precede the
converbal clause, as in Example (12).
(12) a. imaa-de
[mola-sW-e el-eroqe] gW-wi.

donkey-erg Molla-m-erg say-simil do-uw.n*

The donkey did as Molla had said.
b. hadi aka-S-e
[hadi-we w-kena]
dakumate b-e-e

dem woman-f-erg dem-m m-go.immed* document n-take-cvb.f

j-it-he
j-ni
wao-lo-lira
rasija-S-a.

f-go.straight-cvb.f* f-go.ipf* son.pl-hpl-adlat Russia-n-lat
This woman, as soon as this man went, took her documents and went
directly to visit her sons in Russia.
As illustrated by Example (13), an NP representing an argument shared by the two

verbs may precede the converb and at the same time show the case marker required by
the main verb. In sentence (a), beqidi do is in the absolute form required by the main
verb mni (< mnuLa go), and not in the dative case that would point out its status
as an argument of Likena (< LunuLa hear). In sentence (b), dede is in the ergative
case required by the main verb maqWida gWeda (< maqunuLa dig), and not in the
dative case that would point out its status as an argument of mialaqo (<miunuLa find).
Denis Creissels
In sentence (c), ekWasWe is in the ergative case required by the main verb gWije bikWari
(< gruLa do), and not in the absolute form that would point out its status as an argument of wuLideSi (< biLuruLa die).
(13) a.

b-eq-idi
do
[ha reqa
L-ikena] ea m-ni.
hpl-come-pf.hpl person.pl dem word.pl hear-immed away hpl-go.ipf*
As soon as they heard these words, the visitors went away.
b. de-de
[miidi m-i-alaqo] m-aqW-ida gWeda.

1sg-erg gold n-find-ant* n-dig-ipf* cop.n

I will dig until I find gold.
c. hudu ekWa-sW-e [w-uL-ideSi] wasijati gwij-e
b-ikw-ari,
do-cvb.n n-be-pf
dem man-m-erg m-die-simult* will
b-e-e
ai
o-x-a
l-e.
n-take-cvb.n money n-give-imp say-cvb.n
This man, when dying, expressed the will that I should give you the money.
Note however that a shared NP occurring sentence-initially may also have its case
determined by the dependent verb, being thus presumably a constituent of the
dependent clause, as in Example (14).
(14) a. [hudu-sW-a maina b-iW-iL-eSi]
aki imaa-ge

dem-m-dat car
m-find-neg-post wife donkey-ess

dik-aj-e
qel-a
j-eL-ari.
fsit-caus-cvb.f home-lat f-take.away-pf
As he didnt find any car, he took his wife home on his donkey.
b. [dene roo-la
w-deSi]
We Lo
b-eL-ide.

1sg forest-illat m-go.simult* dog together n-take.away-ipf.n

When I go to the forest, I will take the dog with me.
5.3.2 Converbal clauses following the main verb

Although converbal clauses usually precede the main verb, the reverse order is
possible too, as illustrated by Example (15).
(15) a. mene w-uL-ida gWido [du
imii Loda miqi-Si

2sg m-die-ipf* cop.m 2sg(gen) donkey three time-n.ess

le o-t-eSi].

fart n-let.out-post
You will die at the third fart of your donkey.
b. We-de car-awi
[b-e-alaqo].

dog-erg drink-uw.n n-be.full-ant

The dog drank until it slaked its thirst.
c.

oda b-ikW-ide [hage-se-lo-gu

rc-ala].
good n-be-ipf.n here-adjz-hpl-el ask-cond
It would be better if you asked local people.
d. de-de
hute gW-re
gWeda [qimate Saloo b-ik-utana].

1sg-erg so
do-prog* cop.n price
up
n-be-purp

I am doing so in order to keep the price high.
This alternative order is not however equally common for all converbs. Some of
them (for example, the immediate converb) never follow the main verb in the data
I collected. The semantics of the individual converbs probably play a role, since for
example the order main verb converb is more common for the purposive converb
than for the conditional converb, but the decisive factor seems to be the discourse
function of the converbal clause.
6. The locative converb (or verbal locative)

The locative converb is marked by the suffix iS (glossed vloc verbal locative)
followed by one of the three spatial case suffixes i (essive, glossed ess), a (lative,
glossed lat), or u(ne) (elative, glossed el). Its meaning can be rendered as at/to/from
the place where V-ing occurs. In other words, the locative converb shows exactly the
same case inflection as locative adverbs, and a clause headed by the locative converb
is entirely equivalent to a locative adverb with respect to its relation to the main verb.
Examples (16) to (18) illustrate the use of the locative converb in the essive, lative, and
elative cases, respectively.
(16) a. de-de
thi [b-ikW-iS-i]
b-iS-ada.

1sg-erg gun n-be-vloc-ess n-put-pf

I put the gun in its place.
b. keda ree-Si
dene [e o-x-iS-i]

w-oq-ada.
two year-n.ess 1sg tea n-sell-vloc-ess m-work-pf*
I worked two years in a tea-house.
(17) a. w--a
[gora o-x-iS-a].

m-go-imp bread n-sell-vloc-lat

Go to the bakers.
b. mola [ako-de qhula gW-S-a]
w-wi.
Molla wife-erg food do-vloc-lat* m-go.uw.m*
Molla went to the place where his wife was preparing the meal.
(18) a. waa [taduk-iS-une] kokor-iLo.

boy msit-vloc-el move-pf.neg

The boy did not move from the place where he was seated.
Denis Creissels
b. dene [bainaci-de Se
car-iS-une]
w-oq-ero
gWido.

1sg deer-erg
water drink-vloc-el m-come-prog.m cop.m

I am coming from the place where the deer drinks.
As illustrated by several of the preceding examples, the locative converb does not by
itself imply a particular temporal or aspectual meaning, but it constitutes the usual way
of referring to places dedicated to professional activities (tea house, bakers, etc.).
Etymologically, a possible explanation for the locative converb is the contraction
of a synonymous sequence consisting of a participial relative clause headed by the
imperfective participle (suffix ida) modifying mia place (or another noun with the
same meaning, or perhaps a locative adverb) inflected in one of the three spatial cases.
Example (19) illustrates the synonymy between the locative converb (a) and the participial construction which constitutes its probable origin (b).
(19) a. bainaci-de Se
car-iS-i

deer-erg water drink-vloc-ess

at the place where the deer drinks
b. bainaci-de Se
car-ida
mia-Si

deer-erg water drink-ipf place-n.ess

at the place where the deer drinks
7. Temporal converbs
7.1 The temporal reference of the temporal converbs
The temporal converbs express temporal relations between the event they denote and
the event denoted by the main verb, but do not specify past, future or habitual reference by themselves: this precision entirely depends on the tense of the main verb,
and has no incidence on the morphological characteristics of the temporal converbs.
Example (20) illustrates the use of the same simultaneous converb, marked by the suffix
ideSi, in sentences with different temporal reference (a: future, b: habitual, c: past).
(20) a. ut-e
[eropula-ge tadak-ideSi], hirida knopka ic-uba!

2pl-erg plane-ess
hplsit-simult red
knob push-proh

When you will be sitting in the plane, dont push the red knob!
b. [-deSi]
guLa
mene, [m-aqW-ideSi] guLa
mene,

sow-simult* cop.neg.m 2sg
q-nuLa ugu w-oq-uwa?

eat-inf* why m-come-mir.m
n-dig-simult* cop.neg.m 2sg
You are not here when we sow, you are not here when we dig the ground,
why are you coming to eat?
c.

[eada Suda
b-e-o
w-oq-ideSi],
b-al-ri
back firewood n-take-cvb.m m-come-simult n-begin-ipf*
imii le o-t-uruLa.
donkey fart n-let.out-inf
While he was returning with the wood, the donkey started farting.
7.2 The posterior converb

The posterior converb (glossed post) is formed by means of a suffix showing the
following three variants: HPL idiSi, other classes eSi (non-alternating stems) ~ adeSi
(alternating stems). It expresses that the event encoded by the main verb follows the
event encoded by the posterior converb.
(21) a. [ili-gune minaS-idiSi],
-la
ese-kena b-eq-ari.

house-el separate-post.npl bear-add 1ple-com n-come-pf

When we left the house, the bear came with us too.
b. [rala-le-ge
La-se
kama harigW-eSi], di-La
Loda
bone.pl-npl-ess on.ess-adjz belt
see-post
1sg-dat all
hula biil-ri.
thing understand-pf*
When I saw the belt that was on the bones, I understood everything.
The meaning of posteriority can be emphasized by putting qedo after immediately

following the posterior converb, as in (22).
(22) a. [roo-S-a
gu-eSi]
qedo, e ka rua-S-a

forest-n-lat mreach-post after one big tree-n-lat

Lada er-iri.
on.lat climb-ipf
After arriving at the forest, he climbed a big tree.

b. [hudu ekWa w-deSi]
qedo, mola rasadi w-u-o
w-ni.

dem man m-go.post* after Molla Rasadi m-fall-cvb.m m-go.ipf*

After the man had left, Molla fell down.
The variations in the formation of the posterior converb are reminiscent of those in
the formation of the variant of the perfective that can be used as a participle, which
suggests that this form may result from the contraction of a sequence consisting of the
perfective in participle function followed by a noun in the essive case, probably riSi at
the moment. As illustrated in (23), a meaning exactly identical to that of the posterior
converb can be expressed by a construction in which riSi is modified by a relative
clause with the verb in the perfective.
Denis Creissels
(23) a. dene uku-S-a

w-oq-eSi
When I arrived at school

1sg school-n-lat m-come-post
= dene uku-S-a
w-oq-ada ri-Si

1sg school-n-lat m-come-pf moment-n.ess
b. isi uku-S-a
b-eq-idiSi
When we arrived at school

1ple school-n-lat hpl-come-post.hpl

= isi uku-S-a
b-eq-idi
ri-Si

1ple school-n-lat hpl-come-pf moment-n.ess
c. wao-de dene -deSi
When the boy called me

boy-erg 1sg call-post*
= wao-de dene -da

ri-Si

boy-erg 1sg call-pf* moment-n.ess
d. wao-de isi ab-idiSi

boy-erg 1ple call-post.hpl

= wao-de isi ab-idi
ri-Si

boy-erg 1ple call-pf.hpl moment-n.ess
When the boy called us
This analysis is supported by the fact that forms with the suffix eSi ~ adeSi ~ ideSi
do not appear in Magomedova & Abdulaevas dictionary, which however includes
many examples of sentences in which the same meaning is carried by forms with
the suffixes a(da)riSi and idiriSi. These suffixes, in which riSi at the moment is
still clearly recognizable, quite obviously represent a less advanced stage in the
grammaticalization process:
a(da)riSi eSi
dariSi
deSi
idiriS i
idiSi
7.3 The negative form of the posterior converb

The posterior converb is characterized by the complex suffix iL-eSi. One of the possible
meanings of this form is to denote a habitual or virtual event whose consequences are
described by the main verb. In this use, the negative form of the posterior converb is
practically synonymous with the negative form of the conditional converb.
[doroi t-L-eSi],
atali el-ike.
yeast put-neg-post* dough rise-ipf.neg.n
When one does not add yeast, the dough does not rise.
b. [kWani goL-eSi]
televizor b-eq-ike.

light cop.neg.n-post TV.set n-work-ipf.neg.n

When there is no light, the TV does not work
(24) a.

However, the negative form of the posterior converb may also refer to specific past
events, and when this is the case, it more commonly expresses a causal rather than
strictly temporal meaning (as V-ing did not happen ).
(25) a.

[hudu-sW-a maina b-iW-iL-eSi]

aki imaa-ge
dem-m-dat car
m-find-neg-post wife donkey-ess
dik-aj-e
qel-a
j-eL-ari.
fsit-caus-cvb.f home-lat f-take.away-pf
As he didnt find any car, he took his wife home on his donkey.
b. [qaali qel-a
b-eq-iL-eSi],
rai
ea r-ol-ari

dragon home-lat n-come-neg-post young.pl out npl-set.out-pf

-di-li-da
ila
eqed-ruLa.
ana-npl-gen-int mother look.for-inf*
As the dragon did not come back home, the young set out to look for
their mother.
The choice between the two possible interpretations of the posterior converb depends
entirely on the context. The form itself gives no clue as to the correct choice. For
example, in (24a), maina b-iW-il-eSi is interpreted as As he didnt find any car ,
but in a different context, maina biW-il-eSi could equally mean When one doesnt
find any car .
As illustrated by Example (26), the etymological hypothesis put forward for the
positive form of the posterior converb can also account for the negative form.
(26) di-La
maina b-iW-iL-eSi
1sg-dat car
m-find-neg-post
As I didnt find any car
< di-La
maina b-iW-iLa
ri-Si
1sg-dat car
m-find-pf.neg moment-n.ess
lit. At the moment when I found no car
7.4 The simultaneous converbs

Simultaneity is commonly expressed by means of a converb characterized by the suffix
ideSi, glossed simult. The event denoted by this converb is presented as a temporal
frame within which the event denoted by the main verb occurs.
(27) a. [dene roo-la
w-deSi]
We Lo
b-eL-ide.

1sg forest-illat m-go.simult* dog together n-take.away-ipf.n

When I go to the forest, I will take the dog with me.
b. [miqo-ge do
harigW-ideSi] naS-ero
w-uk-uwi.

road-ess person.pl see-simult insult-prog.m m-be-uw.m

Whenever he saw people on the road, he kept insulting them.
Denis Creissels
Similarly to the posterior converb, this form probably results from the contraction of
a synonymous sequence consisting of the imperfective in participle function followed
by riSi at the moment.
waa uku-S-a
w-oq-ideSi
boy school-n-lat m-come-simult
While the boy was arriving at school
= waa uku-S-a
w-oq-ida
ri-Si

boy school-n-lat m-come-ipf moment-n.ess
b. il-e
rili q-deSi
(28) a.

1pli-erg meat eat-simult*

While we were eating the meat
= il-e
rili q-da ri-Si

1pli-erg meat eat-ipf* moment-n.ess
Here again, this hypothesis is supported by the fact that the suffix ideSi does not
appear in Magomedova & Abdulaevas dictionary, which however includes many
examples of sentences in which the same meaning is carried by verb forms with the
suffix idariSi.
An apparently synonymous converb formed by means of a suffix adada sporadically occurs in some of my texts Example (29), but most of my informants do
not use it productively.
(29) [w-oq-ero
w-uk-adada] e ekWa w-oq-ari.
m-work-prog.m* m-be-simult* one man m-come-pf*
A man arrived while I was working.
This converb seems to result from the combination of the perfective suffix ada with the
intensifying particle da. However, this etymological analysis is semantically problematic.
7.5 The inceptive converb

The inceptive converb (from the moment when V-ing began/begins onwards), glossed
incep, is characterized by the suffix -ariSoa. The main verb denotes an activity, and
the event encoded by the inceptive converb marks the initial boundary of the time
interval within which this activity takes place.
(30) [kW b-iqW-ariSoa] [raiSe
t-loqo],
light n-cut-incep whiteness throw-ant*
r-eq-ere
r-ikW-ari erepula-di.
npl-work-prog npl-be-pf plane-pl
From the moment when the light was cut up to dawn, the planes kept flying.
The probable origin of this converb is again the contraction of a synonymous construction involving an inflected form of ri moment modified by a relative clause,
as indicated in (31).
(31) kW b-iqW-ariSoa
light n-cut-incep
From the moment when the light was cut onwards,

= kW b-iqW-ada ri-S-u
ea
light n-cut-pf moment-n-el away
7.6 The immediate converbs

Two forms can be designated as immediate converbs (as soon as V-ing occurred/
occurs), and are both glossed immed. Their characteristic suffixes are ikena and ula.
The event encoded by an immediate converb is presented as immediately preceding
the event encoded by the main verb, both events being conceived of as telic.
(32) a.

[qel-a
gu-ikena]
ako-ga el-iri
home-lat mreach-immed wife-lat tell-ipf
As soon as he arrived at home, he told his wife
b. [duna gWS-ikena], Salao he-ada.

world get.lit-immed up
rise-pf

I rose at dawn (lit. as soon as the world got lit).
(33) a.

[mola ea w-la],
ki-sW-e
ka qor-e
Molla away m-go.immed* judge-m-erg finger thrust-cvb.n
Wa-gune ada-be q-ni.

butter-el little-n
eat-ipf*
As soon as Molla went away, the judge thrust in his finger and ate
a little butter.
b. [hudu aka
hida-sqe s or-ula],
uduka mike-ge

dem woman there-dir turn-immed eagle child-ess

La
goc-awi.
on.ess nbump-uw.n
As soon as the woman went away, the eagle attacked the baby.
ikena probably contains a second formative cognate with the suffix of the comitative
case kena, but I have no hypothesis to propose with respect to its first formative.
ula probably results from the combination of the short variant of the infinitive
suffix u with the additive particle la and, also, in turn, even. This etymological analysis is supported by the semantic proximity between the comitative suffix and the
additive particle.
Denis Creissels
7.7 The anterior converb

This converb may correspond to the English before V-ing occurs/occurred, as in
Example (34), or to until V-ing occurs/occurred, as in Example (35). Its characteristic
suffix, glossed ant, is alaqo or aloqo (commonly articulated alao or aloo).
(34) me-de [isi b-eq-alaqo] ila-li
kuqala gWij-a!
2sg-erg 1ple n-come-ant house-gen work.pl do-imp
Do the housework before we come back!
(35) a. [w-uL-alaqo] hidi-ike
du
oSe.

m-die-ant* forget-ipf.n 2sg(gen) kindness

Until I die I will not forget your kindness.
b. [dene w-oq-alaqo] hage w-ot-a!

1sg m-come-ant here m-stay-imp

Stay here until I come!
miqWalaqo, lit. until it grows, until it is sufficient (anterior converb of miqunula

grow, be sufficient), has been lexicalized with the non-temporal meaning enough.
(36) di-be
[m-iqW-alaqo] oda Le gWeda.
1sg(gen)-n n-grow-ant good cloth cop.n
I have good clothes in sufficient quantity.
I have no etymological hypothesis to put forward for this converb.
7.8 The imminent converb

The imminent converb expresses the meaning just before V-ing occurs/occurred. Its
characteristic suffix, glossed immin, is idaSa or idaSoqe.
(37) a.

[iWada gu-idaSa],
-S-e
elada lgi
shepherd marrive-immin bear-n-erg another sheep
b-e-e
m-wi.
n-take-cvb.n n-go.uw.n*
Just before the shepherd arrived, the bear took another sheep away.
b. [mola lapatka-la b-e-o
w-oq-idaSoqe], koti

Molla spade-add n-take-cvb.m m-come-immin* mouse

e ko kati gWeda mia-S-a
gel-a
m-wi.

one small hole cop.n place-n-lat inside-lat n-go.uw.n*
Just before Molla brought a spade, the mouse went into a place where there
was a small hole.
The immediate converb of mnuSa go, whose literal meaning is just before going,
has postposition-like uses with the meaning before a particular length of time has
passed (within).
(38) [ituda lida o m-daSa]

gW-da
mike-la
five
six day n-go.immin* make-pf* child-add
[ituda lida o m-daSa]
maktaba-S-a-la
m-da-be.
five
six
day n-go.immin school-n-lat-add n-go.ipf-n*
A child born (lit. made) within five or six days will certainly go to school within
five or six days too.
Etymologically, this is probably a complex suffix with the imperfective suffix ida as
its first formative. Consequently, a possible origin for VidaSa is the contraction of a
sequence Vida riSa, with the imperfective participle Vida modifying the allative
form of ri moment.
As for idaSoqe, it probably results from the addition of the similative particle oqe
like, as to idaSa.
7.9 The non-posterior converb

The suffix of the non-posterior converb, illustrated by Example (39), is ileda, glossed
npost. It expresses that the event encoded by the main verb occurs so quickly that the
event encoded by the converb does not occur beforehand (and possibly does not occur
at all, as in sentence (c)).
(39) a. [mola-sW-e q-Leda], ako-de -S-e-da
e

Molla-m-erg eat-npost* wife-erg ana-f-erg-int one

s kWa car-iri.
spoon drink-ipf
Before Molla started eating, his wife ate a spoonful [of soup] herself.
b. [e balahe-Si-ga
w-ux-iLeda], tadi-ga
w-u-a!

one misfortune-n-lat m-fall-npost down-lat m-come.together-imp*

Calm down before something unpleasant happens to you!
c. [keda gali w-Leda],
kebiLa e de-S-e
Le

two
harigW-e rc-ini
see-cvb.n ask-ipf*
step m-go.npost* second one person-n-erg cloth
Before he walked two steps, a second person noticed his cloth and asked
The non-posterior converb of muqunula close (an eye) occurs in the idiom e keha
muqileda lit. before closing an eye in the twinkling of an eye.
The suffix ileda probably results from the combination of the negative form of
the general converb (ile) with the intensifying particle da, lit. without even V-ing.
However, the fact that ileda does not express class agreement provides evidence that
it has grammaticalized as a distinct suffix.
Denis Creissels
8. Conditional and concessive converbs

8.1 The conditional converb
8.1.1 The positive form of the conditional converb
The conditional converb is marked by a suffix ala (with the long variant alasige
~ alasqe) or ala, glossed cond.
Example (40) illustrates the use of the conditional converb referring to conditions
whose realization is considered in the future.
[hu-ji
baza-ga
m-a-ij-ala],
dene-la w-da
gWida.
dem-hpl market-lat hpl-go-hpl-cond 1sg-add m-go.ipf* cop.m
If they go to the market, I will go too.
b. [ha reqa
bea-La
L-ala],
maSeq-ide.

dem word.pl mountain-dat hear-cond be.vexed-ipf.n

If the mountain hears these words, it will be vexed.
(40) a.

The same construction is used in conditional sentences referring to the present, if the
speaker is not in a position to know whether the condition is realized or not.
[ha-be rili m-i-ala]
keto hali?
dem-n meat n-be-cond* cat where.ess
If this is the meat, where is the cat?
b. [mene hute adada w--ala], du-Si-gune

2sg thus mad
m-be-cond* 2sg(gen)-n-el

eko-S i-ga
ugu t--awa?

anothers.property-n-lat why put-neg-mir.n*
(41) a.

If you are as mad [as you pretend to be], why dont you put [wheat] from
your [bag] into [the bag] of other people?
In conditional sentences with present reference but mentioning a non-realized condition, the same form of the conditional converb is used in the subordinate clause.
The main verb is in the analytic form imperfective of the auxiliated verb + perfective of
bikurula be. This form is similar to the future, from which it differs by the use of the
perfective of be instead of the copula.
(42) [du-ge heLar-i-se
ilo di-ge heLar-i
b-ikW-ala],
2sg-ess in.the.mouth-ess-adjz tooth 1sg-ess in.the.mouth-ess n-be-cond
ha qada-Si-ge-la
ea b-eq-u
t-da
b-ikW-ada.
dem moment-n-ess-add out n-take.out-inf throw-ipf* n-be-pf
If the tooth that is in your mouth were in my mouth, I would have it out
immediately.
Example (43) illustrates the contrast between conditional sentences referring to a condition that may come true in the future, and a construction in which the conditional
perfect (conditional converb of bikurula be preceded by the general converb of the
auxiliated verb), combined with a main verb in the form perfective of the auxiliated
verb + imperfective of be is involved in the expression of the same counterfactual
meaning as English conditional clauses in the past perfect.
(43) a.

[jae j-eq-ala],
me-de Wi gW-da gWeda?
girl f-come-cond 2sg-erg what do-ipf* cop.n
If the girl comes, what will you do?
b. [jae j-eq-e
j-ikW-ala], me-de Wi gW-da b-ikW-ada?

girl f-come-cvb f-be-cond 2sg-erg what do-ipf* n-be-pf

If the girl had come, what would you have done?
Example (44) provides another illustration of the conditional perfects use.

(44) [ii q-da-sW-a
o-x-e
b-ikW-ala], he-lao
each ask-pf-m-dat* n-give-cvb.n n-be-cond today-until
b-iW-ida
b-ikW-iLa-be.
n-remain-ipf n-be-pf.neg-n
If I had given to every person who asked for it, there would be none of it
left now.
More generally, analytic forms of the conditional converb are used to express aspectual
shades of meaning. Note that, in the conditional, the use of miunula find, be found,
be in auxiliary function (rather than bikurula be) is particularly common.
[w-x-o
w--ala], u-t-a
w-nuLa.
m-go.oblg-cvb.m* m-be-cond* m-loosen-imp m-go.inf*
If he must go, let him go.
b. [hu-we L-ro
w--ala], w-bala!

dem-m be.afraid-prog.m* m-be-cond* m-go.opt.neg*

If he is (being) afraid, let him not go!
(45) a.

The conditional converb may also be involved in constructions expressing concessive

meaning.
(46) a.

[beko-de qali iWi-ala],

ama
iWi-ike.
snake-erg skin change-cond character change-ipf.neg.n
Although the snake changes its skin, it does not change its character.
b. [hatuda-la ka w-uk-ala], warani-gu ka w-uk-iko.

so.much-add big m-be-cond* camel-el big m-be-ipf.negM*

Tall as he may be, he is not taller than a camel.
Denis Creissels
Another possible use of the conditional converb is the expression of unrealized

wishes, in sentences resembling the first part of conditional sentences whose main
verb would be missing: they are headed by the conditional converb, but no main
verb is present. This use of the conditional converb can therefore be analyzed as an
instance of insubordination (Evans 2007).
(47) a.

di
ima w-uk-ala!
1sg(gen) father m-be-cond*
If only he could have been my father!
di-be
ska b-iqW-e
b-ikW-ala, du-be-la
1sg(gen)-n leg n-cut-cvb.n n-be-cond 1sg(gen)-n-add
mici b-uq-e
b-ikW-ala!
tongue n-dry-cvb.n n-be-cond
b. mitaqali
du-ga ddi w-u-ada
mia-S-e,

in.the.morning 2sg-lat across m-remain-pf* place-n-ess
This morning instead of meeting you, (lit. at the place where I met you)
I would better have cut my leg, and my tongue would better have dried!
It is tempting to break down the conditional suffix ala into a imperative and la
additive particle, since the protasis of conditional sentences may be an imperative
clause, as in (48).
(48) heresi-kena di-La
ahidiSila gWij-a, du-La
aada
lie-com
1sg-dat testimony do-imp 2sg-dat ten
turui o-x-uwa.
rouble n-give-pot.n
Give a false testimony for me, and Ill give you ten roubles.
This hypothesis is consistent with the fact that the conditional suffix ala and the
imperative suffix a show the same phonological interaction with the verbal stem.
However, the fact that the HPL marker ij can be inserted before the conditional
suffix, whereas no agreement mark can be inserted before the imperative suffix, is
evidence against this hypothesis.
As regards the variant ala, it seems reasonable to assume that it results from the
contraction of an analytic form containing mala, a conditional converb of munula
find, be found, be in auxiliary function.
8.1.2 The negative form of the conditional converb

There are two variants of the conditional converbs negative form. One of them results
from the insertion of the negative marker il between the stem and the conditional suffix ala, the other one, marked by a suffix iilala, lends itself to the same etymological
explanation as the ala variant of the positive conditional suffix.
[jae j-eq-iL-ala],
me-de Wi gW-da gWe-da?
girl f-come-neg-cond 2sg-erg what do-ipf* cop.n
If the girl does not come, what will you do?
b. [hu-be di
aqilo-Si-ga b-eq-e
m-i-iL-ala],

dem-n 1sg(gen) mind-n-lat n-come-cvb.n n-be-neg-cond*

hu-su
aqilo-Si-ga-la
b-eq-ike.
(49) a.

dem-m(gen) mind-n-lat-add n-come-ipf.neg.n
If it did not come to my mind, it will not come to his mind either.
Example (50) illustrates the possibility of a contextual interpretation of the conditional

converbs negative form except that .
(50) [warani Lib-L-ala]
koa aza b-ikW-iLa.
camel.pl be.afraid-caus.neg-cond* bad pain n-be-pf.neg
Apart from the fact that the camels frightened me, nothing bad happened.
The meaning whether or not can be expressed in Akhvakh by putting the negative
form of the conditional converb immediately after the positive form of the conditional
converb of the same verb.
(51) du-La
u-bi, [di
raqoli b-ol-ala
b-ol-iL-ala]?
2sg-dat what-n 1sg(gen) back n-ache-cond n-ache-neg-cond
What does it matter to you, whether my back is aching or not?
8.2 The concessive converbs

Two suffixes, both glossed conc, can be used to form a concessive converb: alala
(resulting from the adjunction of the additive particle la to the conditional suffix ala),
and alotola ~ alotona.
(52) a.

[w-oq-alotola] egaza e
b-ol-iLa.
m-work-conc* nothing profit n-occur-pf.neg
Although I worked, I got no result.
b. [b-ix-alotola] imii zor-e

b-ol-ere
b-ikW-iLa.

n-hold-conc donkey slip-cvb.n n-move.forward-prog n-be-pf.neg

Although we held it, the donkey slipped and did not move forward.
An etymological analysis of this suffix can be imagined on the basis of the observation
of its variations. In addition to the forms alotola and alotona I have found in AD
Akhvakh, the variants erotona and erotana also occur in Magomedova & Abdulaevas
dictionary. The crucial observation is that the last two syllables of the variant erotana
are identical with tana, the suffix of the purposive case. It seems therefore reasonable
Denis Creissels
to consider that the suffix of the concessive converb was originally a complex suffix
with tana as its second formative, and that, given their semantic proximity, the forms
alotona and alotola developed as hybrids of erotana and alala. Concerning the
first formative of the complex suffix erotana, the hypothesis of a relationship with
the verbal noun is contradicted by the selection of the short allomorph of alternating
stems, but I have no other etymological hypothesis to put forward.
9. The similative converb

This converb can be translated as in the same way as . It is characterized by the suffix
eroqe ~ ereqe, glossed simil.
(53) a.

[uni
gW-roqe] eLa-la
gWij-a!
yesterday do-simil* today-add do-imp
Do today as you did yesterday!
b. [di-gune ai
q-roqe] de-de-la
maina q-da.

1sg-el money ask-simil* 1sg-erg-add car
ask-pf*

As he asked me for money, I myself asked him for the car.
Etymologically, the suffix of the similative converb quite obviously contains the
similative particle oqe like as its second formative. As for the first formative of this
suffix, the selection of the short allomorph of alternating verb stems contradicts an
otherwise possible etymological analysis according to which the similative converb
would originate from the verbal noun, and points rather to a relation with the suffix
of the progressive converb ere.
10. The gradual converb

This converb, marked by the suffix da ~ daSe (glossed grad), expresses the
meaning rendered in English as in proportion to , the more , the more .
[hudu-we l-daSe]
taS-ari.
dem-m dance-grad* get.tired-pf
The more he danced, the more he got tired.
b. [heaS-da] w-u-o
w-da
gWida.

rise-grad m-fall-cvb.m m-go.ipf* cop.m

The higher you rise, the more likely it is that you will fall.
(54) a.

I have no suggestion of any possible etymology for this converb.
11. The explicative converb

11.1 The positive form of the explicative converb
The explicative converb (because ) is marked by the suffix erogu ~ eregu, glossed
explic. The identification of this form as a converb is somewhat problematic in the
sense that it often occurs in constructions that could be analyzed as involving complementation rather than adverbial subordination (for example, with lrula be afraid,
or reci grula thank). However, Example (55a) shows that it also occurs in contexts
unambiguously involving adverbial subordination.
(55) a.

[mola makWa-o
w-uk-erogu], keda reLa-Si-kena
Molla be.hungry-cvb.m m-be-explic* two hand-n-com
q-ne
b-ik-awi.
eat-prog* n-be-uw.n
Because Molla was hungry, he ate with both hands.
b. dene L-ro
gudi
[mike resaSero-ge di-ga,

1sg be.afraid-prog.m* cop.m baby beauty-ess 1sg-lat

aqilo-ge
du-ga eS-enogu].

intelligence-ess 2sg-lat resemble-explic*
I am afraid that the baby will resemble me as regards beauty, and you as
regards intelligence.
c.

[-de
qalaqala ea eq-erogu]
dene damaSil-ri.
bear-erg often
out look-explic 1sg be.surprised-pf
I was surprised that the bear often looked outside.
Morphologically, the explicative converb can be identified as the elative form of the verbal noun, and the meaning it expresses is consistent with its morphological make-up,
but there are three reasons for treating it as a converb distinct from the inflected form
of the verbal noun it originates from:
in the inflection of the verbal noun, the original oblique stem formative ro tends
to be replaced by the standard oblique stem formative Si, whereas the suffix of
the explicative converb is never found with Si instead of ro or re;
the verbal noun is rarely found with its core arguments in the ergative or
absolute cases, and tends to behave syntactically like ordinary nouns, whereas
the explicative converb consistently shows verbal behavior with respect to its
modifiers.
the explicative converb has a synthetic negative form (see Section11.2), whereas
the verbal noun only has an analytic negative form.
Denis Creissels
11.2 The negative form of the explicative converb

The negative suffix il can be inserted between the verb stem and the suffix of the
explicative converb.
[me-de ai
o-x-iL-erogu]
azi
b-eq-ere
godi.
2sg-erg money n-give-neg-explic complaint n-take.off-prog cop.n
He is complaining that you do not give him money.
b. hu-Si-tana reci
gW-re
gWeda, [imaa-ge La

dem-n-purp thanks do-prog* cop.n donkey-ess on.ess

w-uk-iL-erogu
[b-uq-ideSi]].
(56) a.

m-give-neg-explic* n-disappear-simult
The reason why I am thanking [God], is that I was not sitting on the
donkey when it disappeared.
Also note that erogu can attach to the negative form of the copula.
(57) mola [qel-i
egaza goL-erogu]
damaSil-ri.
Molla home-ess nothing cop.neg.n-explic be.surprised-ipf
Molla was surprised that there was nothing at home.
12. The purposive (or supine) converb

12.1 The positive form of the purposive converb
The purpose meaning expressed by this converb (in order to ) can also be expressed
by the infinitive, and it is always possible to replace the purposive converb by the infinitive without changing the meaning.15 The difference between the two forms is that
the infinitive has a variety of uses, whereas the purposive converb is more specialized,
occurring exclusively in adverbial purpose clauses.
The suffix of the purposive converb is utana (glossed purp).
(58) a.

raLaqu k-i
car-i
qedo, [lk-utana] relWa t-ni.
at.night eat-cvb.hpl eat-cvb.hpl after hplsleep-purp bed put-ipf*
At night after eating and drinking, they made the bed in order to sleep.
. For a general discussion of the possible historical connections between purposive and
infinitive, see Haspelmath 1989.
b. [hudu mia-S-unu Seni b-e-utana], odoli

dem place-n-el water n-take-purp every.day

e-e
jae azaho-ga
zor-da.
one-one girl dragon-lat throw-caus.ipf*
In order to (be allowed to) take water from this place, they throw a young
girl to the dragon every day.
The sporadic occurrence of synonymous forms with the suffix urulatana confirms
that, etymologically, utana is a complex suffix consisting of the infinitive marker
u(rula) and the purposive case marker tana. However, within the frame of a synchronic analysis at least, this converb cannot be described as an inflected form of
the infinitive, since the infinitive cannot combine with any other case suffix. The
explanation is that the status of tana as a case suffix is probably recent. In AD
Akhvakh, tana occurs in the same morphosyntactic slot as the other case suffixes,
but in AR Akhvakh, tana is a postposition combining with nouns in the dative case.
In addition, although a precise etymological analysis of the infinitive suffix urula
does not seem possible, a connection with the dative case suffix la is probable.
Consequently, historically, the purposive converb results from the evolution of a
construction involving the derived nominal form in the dative case that gave rise to
the infinitive of present-day Akhvakh, followed by the postposition from which the
purposive case suffix originates.
12.2 The negative form of the purposive converb

The purposive converb has a negative counterpart marked by the complex suffix ilukutana, which results from the contraction of ile bikutana, i.e. the suffix of the
negative form of the converb ile followed by the purposive converb of bikurula
be in auxiliary function.
(59) [kWet-iL-ukutana] mola-sW-e
Le rua-S-a
run.away-neg-purp Molla-m-erg calf tree-n-lat
kar-aj-e
b-ik-awi.
be.fixed-caus-cvb.n n-be-uw.n
Molla had tied the calf to a tree so that it would not run away.
12.3 The emergence of a new purposive converb

When the infinitive is used in adverbial purpose clauses, but not in its other uses, it
may combine with le, the reduced form of the general converb ele of elurula say, as
in Example (60).
Denis Creissels
(60) a. odaka harigW-eSi, [e-be b-ix-uruLa] l-e

hWara-ga

duck.pl see-post one-n n-catch-inf say-cvb.n lake-lat

gel-a
kas-iri.
inside-lat jump-ipf
When he saw the ducks, he jumped into the lake to catch one of them.
b. rili ab-e
b-iS-ada isto-Si-ge
La
[a

meat wash-cvb.n n-put-pf table-n-ess on.ess fire

kon-uruLa] l-e.
light-inf
say-cvb.n
I washed the meat and put it on the table to light the fire.
In such constructions, le is optional in the sense that it can always be deleted, and its
deletion does not modify the meaning, but its presence underscores the purposive
function of the infinitive. Syntactically, this use of le can be grouped together with
other complementizer-like uses of the same converb, either in its full form ele or in its
reduced form le Example (61).
(61) a.

el-uba [de-de el-iLa]

l-e!
say-proh 1sg-erg say-pf.neg say-cvb.n
Dont say that I didnt tell you!
b. j-eW-ere
gida [mike b-ikW-ida gWeda] le.

f-rejoice-prog cop.f baby n-be-ipf cop.n say-cvb.n

I am happy that we are going to have a baby.
Consequently, in the present state of the system, there are arguments for describing
this use of le as a particular aspect of the infinitives syntax. But the same facts could
equally be described by positing a second purposive converb characterized by the
complex suffix urulale.
13. Conclusion
The main conclusion that can be drawn from this survey of AD Akhvakh specialized
converbs is the high proportion of etymologically transparent converbs. In addition, the
available data on AR Akhvakh suggests that there are interesting differences between
the specialized converbs of these two Akhvakh varieties, despite their remarkable
similarity in most areas of the grammar. Since massive recourse to rich inventories
of specialized converbs to express various semantic types of adverbial subordination
is probably ancient among Nakh-Daghestanian languages, this observation confirms
that, in languages of this type, grammaticalization processes such as those recognized
above for AD Akhvakh are constantly at work, resulting in a relatively rapid renewal
of the inventories of specialized converbs. A possible consequence of this situation is
the proliferation of synonymous forms. We have seen for example that, in AD Akhvakh,
a construction involving the infinitive and a converbal form of elurula say is grammaticalizing into a new purposive converb, in spite of the fact that two forms (the
infinitive and the purposive converb) are already available as heads of adverbial
purpose clauses.
Abbreviations
1pli
1ple
1sg
2pl
2sg
add
adjz
adlat
ana
ant
caus
com
comp
conc
cond
cop
cvb
dat
dem
dir
el
erg
ess
explic
f
gen
grad
hpl
illat
1st person plural inclusive

1st person plural exclusive
1st person singular
2nd person plural
2nd person singular
additive
adjectivizer
adlative
anaphoric pronoun
anterior converb
causative
comitative
complementizer
concessive converb
conditional converb
copula
general converb
dative
demonstrative
directional
elative
ergative
essive
explicative converb
feminine singular
genitive
gradual converb
human plural
illative
immed
immin
imp
incep
inf
int
ipf
lat
m
mir
n
neg
npl
npost
oblg
opt
pf
pl
post
pot
prog
proh
purp
sg
simil
simult
uw
vloc
vn
immediate converb
imminent converb
imperative
inceptive converb
infinitive
intensifying particle
imperfective
lative
masculine singular
mirative
non-human singular
negation
non-human plural
non-posterior
obligative
optative
perfective
plural
posterior converb
potential
progressive converb
prohibitive
purposive case/converb
singular
similative converb
simultaneous converb
past unwitnessed
verbal locative
verbal noun
Denis Creissels
Annex
Underlying representation of verbal forms involving morphophonological processes,
signaled by an asterisk in the example glosses
(1) j-ni < |j-v-iri|,16 g-ux-ida < |gWi(j)-ux-ida|
(2) q-ni < |q(b)-iri|, tk-ini < |tk-iri|
(4) m-ne < |b--ere|
(5) w-uk-uwi < |w-ikW-uwi|, k-no < |k-a(j)-ero|
(6) -re < |a(b)-ere|, m-i-ani < |b--ari|, w--ani < |w--ari|
(7) q-ne < |q(b)-ere|
(8) -re < |a(b)-ere|
(10) c-daSa < |ca(b)-idaSa|, b-oil-ri < |b-oila(j)-ari|, m-i-ala < |b--ala|,
m-da < |b-v-ida|
(12) gW-wi < |gWi(j)-awi|, w-kena < |w-v-ikena|, j-ithe < |j-ita(j)-ehe|, j-ni <
|j-v-iri|
(13) m-ni < |b-v-iri|, m-i-alaqo < |b--alaqo|, m-aqW-ida < |b-qW-ida|, w-ulideSi < |w-il-ideSi|
(14) w-deSi < |w-v-ideSi|
(15) w-ul-ida < |w-il-ida|, gW-re < |gWi(j)-ere|
(16) w-oq-ada < |w-eq-ada|
(17) gW-S-a < |gWi(j)-iS-a |, w-wi < |w-v-uwi|
(20) -deSi < |i(b)-ideSi|, m-aqW-ideSi < |b-qW-ideSi|, q-nula < |q(b)-urula|,
b-al-ri < |b-ala(j)-iri|
(21) biil-ri < |biila(j)-ari|
(22) w-deSi < |w-v-adeSi|, w-ni < |w-v-iri|
(23) -deSi < |a(b)-adeSi|, -da < |a(b)-ada|
(24) t-l-eSi < |t(b)-il-eSi|
(25) eqed-rula < |eqeda(j)-urula|
(27) w-deSi < |w-v-ideSi|
(28) q-deSi < |q(b)-ideSi|, q-da< |q(b)-ida|
(29) w-oq-ero < |w-eq-ero|, w-uk-adada < |w-ikW-adada|, w-oq-ari < |w-eq-ari|
(30) t-loqo < |t(b)-alaqo|
(33) w-la < |w-v-ula|, q-ni < |q(b)-iri|
(35) w-ul-alaqo < |w-il-alaqo|
(37) w-oq-idaSoqe < |w-eq-idaSoqe|, m-wi < |b-v-awi|
(38) m-daSa < |b-v-idaSa|, gW-da < |gWi(j)-ada|, m-da-be < |b-v-ida-be|
. v represents the unspecified nasal vowel which constitutes one of the allomorphs of the
root of mnula go.
(39) q-leda < |q(b)-ileda|, w-u-a < |w-i-a|, w-leda < |w-v-ileda|, rc-ini <
|rc-iri|
(40) w-da < |w-v-ida|
(41) m-i-ala < |b--ala|, w--ala < |w--ala|, t--awa < |t(b)-u-awa|
(42) t-da < |t(b)-ida|
(43) gW-da < |gWi(j)-ida|
(44) q-da-sW-a < |qa(b)-ada-su-ga|
(45) w-x-o < |w-v-ux-o|, w--ala < |w--ala|, w-nula < |w-v-urula|, l-ro <
|li(b)-ero|, w-bala < |w-v-ubala|
(46) w-uk-ala < |w-ikW-ala|, w-uk-iko < |w-ikW-ika-we|
(47) w-uk-ala < |w-ikW-ala|, w-u-ada < |w-iW-ada|
(48) gW-da < |gWi(j)-ida|, m-i-il-ala < |b--il-ala|
(49) lib-l-ala < |li(b)-a(j)-il-ala|
(52) w-oq-alotola < |w-eq-alotola|
(53) gW-reqe < |gWi(j)-ereqe|, q-reqe < |qa(b)-ereqe|, q-da < |qa(b)-ada|
(54) l-daSe < |li(b)-daSe|, w-da < |w-v-ida|
(55) w-uk-eregu < |w-ikW-eregu|, q-ne < |q(b)-ere|, l-ro < |li(b)-ero|, eS-enogu
< |eSerogu|
(56) w-uk-il-eregu < |w-ikW-il-eregu|, gW-re < |gWi(j)-ere|
(58) t-ni < |t(b)-iri|, zor-da < |zor-a(j)-ida|
References
Bickel, Balthasar. 1998. Converbs in cross-linguistic perspective. Review article on Haspelmath&
Knig (eds) 1995. Linguistic Typology 2: 381397.
Bisang, Walter. 1995. Verb serialization and converbs differences and similarities. In Converbs
in Cross-linguistic Perspective, Martin Haspelmath & Ekkehard Knig (eds), 137188.
Creissels, Denis. 2008a. Person variations in Akhvakh verb morphology: Functional motivation and origin of an uncommon pattern. Sprachtypologie und Universalienforchung 61(4):
309325.
Creissels, Denis. 2008b. Remarks on so-called conjunct/disjunct systems. Paper delivered at the
conference Syntax of the worlds languages III. Berlin.
Creissels, Denis. 2009. Participles and finiteness: The case of Akhvakh. Linguistic Discovery 7.1.
106130.
Creissels, D. Submitted. External agreement in the converbal construction of Northern Akhvakh.
In Clause Combining in Cross-linguistic Perspective, Holger Diessel & Volker Gast (eds).
Evans, Nicholas. 2007. Insubordination and its uses. In Finiteness, Irina Nikolaeva (ed.), 366421.
Oxford: OUP.
Hasplemath, Martin. 1989. From purposive to infinitive a universal path of grammaticalization.
Folia Linguistica Historica 10(12): 287310.
Denis Creissels
Haspelmath, Martin. 1995a. The converb as a cross-linguistically valid category. In Converbs
in Cross-linguistic Perspective, Martin Haspelmath & Ekkehard Knig (eds), 155. Berlin:
Mouton de Gruyter.
Haspelmath, M. 1995b. Contextual and specialized converbs in Lezgian. In Converbs in
Cross-linguistic Perspective, Martin Haspelmath & Ekkehard Knig (eds), 41540. Berlin:
Mouton de Gruyter.
Haspelmath, Martin. 1999. Converbs. In Concise Encyclopedia of Grammatical Categories, Keith
Brown & James Miller (eds), 110115. Oxford: Elsevier.
Kazenin, Konstantin I. 2001. Deepriastnye konstrukcii. In Bagvalinskij jazyk (grammatika,
teksty, slovari), Kibrik, Aleksander E. (ed.), 554594. Moscow: Nasledie.
Kazenin, Konstantin I. & Testelec, Jakov G. 1999. Problema razgranienija soinenija i
podinenija predikacij: Bessojuznaja konstrukcija. In lementy caxurskogo jazyka v
tipologieskom osveenie, Alexander E. Kibrik (ed.). Moscow: Nasledie.
Knig, Ekkehard. 1995. The meaning of converb constructions. In Converbs in Cross-linguistic
Perspective, Martin Haspelmath & Ekkehard Knig (eds), 5796. Berlin: Mouton de Gruyter.
Kwon, Nayoung & Polinsky, Maria. 2007. What does coordination look like in a head-final
language?. In Asymmetric Events? Coordination and Conceptual Structure [Converging
Evidence in Language and Communication 11], Barbara Lewandowska-Tomaszcyk (ed.).
Magomedbekova, Zagidat M. 1967. Axvaxskij jazyk (grammatieskij analiz, teksty, slovar).
Tbilisi: Mecniereba.
Magomedova, Patimat & Abdulaeva, Indira. 2007. Axvaxsko-russkij slovar. Maxakala:
Dagestanskij Naunyj Centr Rossiskoj Akademii Nauk.
Nedjalkov, Vladimir. 1995. Some typological parameters of converbs. In Converbs in Crosslinguistic Perspective, Martin Haspelmath & Ekkehard Knig (eds), 97136. Berlin:
Mouton de Gruyter.
Polinsky, Maria. 2007. The relationship between coordination and subordination: Clause chains
in Tsez and beyond. Conference on Languages of the Caucasus, Leipzig.
Tikkanen, Bertil. 2001. Converbs. In Language Typology and Language Universals, Vol. 2,
Martin Haspelmath, Ekkehard Knig, Wulf sterreicher & Wolgang Raible (eds),
van der Auwera, J. 1998. Defining converbs. In Typology of the Verbal Categories: Papers
Presented to Vladimir Nedjalkov on the Occasion of his 70th Birthday, Leonid Kulikov &
Heinz Vater (eds), 273282. Tbingen: Max Niemeyer.
Ylikoski, J. 2003. Defining non-finites: Action nominals, converbs and infinitives. SKY Journal
of Linguistics 16: 185237.
Ziga, Fernando. 1998. Nomina sunt odiosa: A critique of the converb as a cross-linguistically
valid category. Ms, University of Zrich.
Finite and non-finite

Prosodic distinctions on Budugh verb stems
Gilles Authier
Institut National des Langues et Civilisations Orientales

In Budugh, a small Daghestanian language spoken in Azerbaijan, verb-stems,
defined as verb forms uninflected for tense, can be used either as dependent of
nouns (participles) and other predicates (verbal nouns or hereafter masdars,
and sequential converbs) or as finite non-indicative verb forms, that is with
modal use, especially in a variety of injunctive nuances. The main distinction
between dependent and independent uses is prosodic: dependent verb forms
adopt the prosodic features of nouns, while the same forms, if syntactically
independent, share the same prosodic pattern as indicative or more complex
modal forms which are segmentally marked (suffixed) as such.
1. Introduction
Budugh1 is an unwritten language belonging to the Lezgic branch of the Daghestanian
family.2 Today it is spoken as a first language by at most 200 speakers, all inhabitants
of a single village of north-eastern Azerbaijan. Despite generalised bilingualism in
Azerbaijani, which can probably account for the recent development of a 9-vowel
system (, i, u, , o, , e, a, a)3 and vowel harmony, it preserves typically Daghestanian,
and more specifically Proto-Lezgic features. In particular, gender-number agreement with S/P (Single argument or Patient) nouns is infixed on discontinuous or
1. I am immensely indebted to my friend and main informant Adigoezel Hadjiev for teaching
me his rare language; he provided most of the data for this paper. I also wish to express my gratitude to Timur Maisak, Yura Lander, Lena Kalinina, and to Isabelle Bril for very helpful comments on a first version of this paper. Remaining errors are, however, my sole responsibility.
2. This branch comprises core languages of Lezgian, Tabassaran, Aghul, Tsakhur, Rutul,
Kryz, and Budugh, all located along the watershed-line of the Great Caucasus range and the
peripheral Udi (to the south) and Archi (to the north).
3. /a/ is pronounced centrally except in contact with back consonants and in initial position;
while /a/ is pronounced more open and back.
Gilles Authier
bipartite lexical stems of synthetic verbs.4 These two features give rise to introflected, partially apophonic paradigms. Person is expressed by free pronouns, and is
not indexed on verbs.
The present paper will address the original interaction of syntax and prosody associated with bare verb stems.5 It will be shown that the opposition between finiteness
and non-finiteness can be unmarked segmentally, and left to interpretation from the
syntactic context and intonational contour of the clause; this is of some importance for
the question of finiteness formulated as follows: how are dependent and independent
verb forms morphologically related in a given language?6
The first and second sections outline the verbal categories expressed mor
phologically on verb-stems, and give an overview of parts of speech and word
orderin Budugh. The third and fourth sections present the prosody of imperfective
verb stems determined by the syntactic domains in which they are found, showing the relationship between segmentally homonymous forms and the correlation
of stress distinctions with syntactic function. Sections five and six propose a classification of perfective stems and illustrate the dependent and independent uses of
these forms.
2. Verb stem structure

The basic frame of any Budugh lexical verb stem consists of two or three original elements: one or two preverbs, and a root. Verb stems can also display:
negative marking infixed after the first preverb;

voice distinction, signalled by apophony and vowel harmony;7
infixed and partially vowel-harmonic indexation of S/P;
aspect marking, in the form of affixed sonorants and vowels or vowel harmony.
4. Verbal predicates are also represented by compound verbs, which are not relevant here.
5. I define verb stems in Budugh as minimal existing verb forms from which other forms are
straightforwardly derived through morpheme suffixation.
6. On this topic, see the recent detailed description by N. Evans of a reanalysis process which
he calls Insubordination in Finiteness (Nikolaeva, 2007), and in the same volume the insightful
study by Kalinina & Sumbatova (2007) on the independent use of formally non-finite forms in
some Daghestanian languages remotely akin to Budugh.
7. Since Budugh causative and anticausative voices are irrelevant to the problem of dependency, they will not be addressed here, cf. Authier (2009, to appear).
All verbs in Budugh are minimally made up of a root (one consonant plus a sonorant
element in some but not all verbs), preceded by one or two preverbs. The first preverbs
originally had spatial meaning; the second preverbs are synchronically opaque, usually
labelled expressive, and probably originating from a process of incorporation affecting serialised verbs. Following the cliticisation of these locational adverbs (in the first
preverbal slot) and of the incorporated serial verbs (in the second preverbal slot), a
derivational pattern resulted which at some stage affected all the languages central to
the Lezgic sub-family.8 As an extreme result of this trend, all Budugh verbs are bi- or
tripartite.9 Gender and number are now indexed inside the verb stem, just after the
first or second preverb. The root itself has either a single obstruent, or an obstruent
followed by a sonorant, and a vowel which may mark transitivity.
As in most Daghestanian languages,10 Budugh verbs have the gender and number
of their Patient or Single argument cross-referenced in a pre-root (post-preverbal) slot.
The gender-number agreement system comprises six gender-number classes of agreement: human masculine (m), human adult feminine (F), Animate (A, which includes
not only all animals and plants, but also non-adult human females and quite a few
motivated categories of abstract nouns, especially those meaning a speech act), inanimate (n for neuter), non-human plural (npl), and human plural (hpl). On the majority of verb forms agreement for m, n and npl is unmarked (zero). Nouns expressing
S and P are in the unmarked absolutive case. The Agent noun of a transitive predicate
is marked with the ergative case, and is not indexed on verb forms, except for person
and number in injunctive forms. Gender and number, which were originally marked
by initial prefixes (as is still the case in Avar for instance), have often become trapped
between root and preverb and sometimes are suprasegmentally marked, through vowel
harmony processes. The following examples show indexation of P and S respectively:
(1) malla-cr lem
votu-ri (2) malla-cr lemr
tu-ri.
Mullah-erg donkey a.beat.prs
Mullah-erg donkey-pl npl.beat.prs
Mullah beats the donkey.
Mullah beats the donkeys.
8. This full-fledged derivational system with two preverbal slots is found in Tsakhur, Rutul,
Aghul, Tabassaran and Budugh. Kryz verbs may have only one preverb if any: obviously the
two preverb slots have become conflated in Kryz, where some preverbs have locational origin
while others do not. Budugh is unique in having all its verbs, without exception, preverbed.
Kryz retains many unpreverbed verbs. It is less clear whether peripheral Lezgic languages Udi
and Archi once had a similar system of preverbs. Note that Khinalug has its own system of
preverbs with almost no formal cognates in the surrounding Lezgic languages.
9. Unlike other closely related languages like Kryz, Aghul or Rutul, which have preserved a
rather large proportion of root verbs, without any trace of preverbation.
10. This system of agreement has been entirely lost in only three Daghestanian languages:
Udi, Aghul and Lezgian (Haspelmath 1994), all members of the Lezgic branch.
Gilles Authier
(3) malla vxhici (4)

Mullah m.go.nar
Mullah went.
lem
v`xhc.
donkey a.go.nar
The donkey went.
Finite indicative, as well as semantically complex converbal forms always bear a suffix
or a copula and the former may also add a past-tense marker in final position. For
instance, the following form zblqlavini can be analysed as follows:
(5) zblq-
-lavi-ni
causative.cry.hpl progressive-Past
(was/were (at that moment) being caused to cry)
-
z-
b-
l-
q-
la-
vi-
ni
1st
2nd
Human Plural Imperfective (causative) Simultaneous Copula Past
Preverb Preverb Gender/Class aspect
Root
converb
Cognitive accessibility of such an intricate form probably proceeds in two steps: first
the stem, a semantic compound of preverbs and a root, segmentally or non-segmentally
marked for valency, aspect and gender/number of the Absolutive argument must be
recognised and retrieved from the competent speakers mental dictionary; then the
Tense-Mood ending, which is produced with minimal allomorphy, is identified.
Affixation in Budugh is thus of two types:
infixes evolving into suprasegmental and non-concatenative formatives on the

stems;
suffixes after the stems.
The paper will focus on the unsuffixed forms or verb stems. All verbs have two series
of verb stems and forms, perfective and imperfective. Broadly speaking, imperfective
aspect can be marked by a sonorant (r, l, n) immediately prefixed to the root, while
perfective aspect can be marked by a low vowel (-a, -e, -o, -) or a sonorant immediately suffixed to the root.
Apophony and suppletion also occur, and the verb stem paradigms are in fact
often intricate, with few classes of verbs. An example of non-straightforward aspectual derivation involving a change of preverbs is the verb be:
Be
m/n/npl
f
a
hpl
pf
sa-xh-a
sa-rxh-a
soxho < *sa-v-xh-a
sabka < *sa-b-xh-a
ipf
yi-xh-ar
yi-r-xh-ar
yuxhor < *yi-v-xh-ar
ybkar < *yi-b-xh-ar
An example of suppletive paradigm is the verb go:

Go
m
f
a
n/npl
hpl
pf
vi-xhi
vi-r-xhi
vxh < *vi-v-xhi
vidki < *vi-d-xhi
vibki < *vi-b-xhi
ipf
a-ar
a-r-ar
oor < *a-v-ar
a--ar
a-b-ar
3. Word order, parts of speech and stress position and nature

Budugh clearly has distinct word-classes, associated with specific inflectional morpheme compatibilities and syntactically determined stress properties. Grammatical
relations are indicated by case suffixes and indexation of S/P on the verbs. Only verbs
can bear Tense, Mood, and Aspect inflection. Conversion of verbs from predicates to
arguments (Masdars, Infinitives), attributes (Participles) and adverbs (Converbs) is
rich, inflectional,11 and possible for all verbs without restriction.
With the exception of a few particles, semantic word-classes bear stress. Substantive nouns and other arguments (substantivised attributes) have final or stem-final
stress (though this is only faintly audible on absolutive Patients in their pragmatically
neutral preverbal position). But on verbs, stress position and nature is determined by
syntactic function:
finite verb-forms have initial falling stress;
subordinate verb-forms are non-finite, having final rising stress like nouns;
an intermediate prosodic characterisation is the initial rising stress which appears on
the sequential form used for co-subordination.
Dependent verb forms retain their internal verbal syntax, and their arguments in the
embedded clause are marked in the same way as in independent clauses. But in the
valency of the matrix clause, they fill argument, attribute or adjunct positions, inflect
for case, and crucially undergo a stress shift, thereby sharing the stress pattern of
genuinenouns.
11. On inflectional category-changing morphology, see Haspelmath (2002).
Gilles Authier
Budugh is a strict GN (genitive-N), AN (adjective-N), SOV language, with ergative case marking, as shown below:
mall-co rij the mullahs daughter;
qus mall the old mullah;
mall-cr rij vataci The mullah beat the daughter.
Word order can be only slightly modified for pragmatic purposes. As in Azerbaijani
or in Kryz, focused constituents tend to occur in preverbal position, while topics are
clause-initial: ugo rij mall-cr vataci It was the mullah himself who beat his daughter
(here mall is focused and rij is topicalised).
Since Budugh is a strict verb final language, initial stress indicates de facto the end of
a finite (independent) or sequential (suspensive) clause: falling intonation is a marker of
both syntactic finiteness and pragmatic completeness. Once verbs are converted into syntactic attributes, arguments, and adjectives named respectively participles (perfective &
imperfective), verbal nouns (perfective infinitives vs. imperfective masdars)12 and converbs their stress shifts to the last syllable of the stem, indicating non-finiteness. This
suprasegmental feature serves to discriminate between otherwise homonymous forms.
On unsuffixed verb stems, rising and falling stress must thus be added to the list
of verbal morphological devices. Basically, a prosodic feature system is the main device
affected to part-of-speech-changing. The same segmental bare stems are used as independent (finite) and as dependent (subordinate) predicates, provided that their pitch
accent shifts from the first to the last syllable. Budugh makes productive use of this
pattern, both series of aspectual verb stems perfective and imperfective show their
dependent and independent functions through stress. There is no further distinction
between the formal-functional series: gender-number cross-referencing is marked in
the same way on both finite and non-finite homonymous forms.
4. Prosody and syntax of imperfective verb stems

Given that all nouns in Budugh are characterised by final stress and rising prosody,
while all finite non-interrogative verb forms show initial stress and falling prosody, all
verb stems have two variants, distinguished by the position and nature of their stress.
Since the classification of imperfective stems is slightly less complex than for perfectives,
we will address them in this order.
12. Masdar and infinitive are convenient labels justified by the respective ability of these
forms to inflect for case. Perfective and imperfective aspects have no necessary link with either
one of these deverbal categories: in contrast to Budugh, the Lezgian infinitive is morphologically imperfective, while its masdar is morphologically perfective, cf. Haspelmath 1994.
All imperfective stems have thus two accentual forms and syntactic interpretations:
Table1. Finite and non-finite uses of the imperfective stems
Non-finite: Masdar/IPF participle
Finite: Debitive mood
slay
m/n(pl)/A/hpl
sork
srku
hang tr.
m/n(pl)
f
a
hpl
sobork
anx
anx
onx
abonx
sborku
nxu
nx-u
nxu
bonxu
anxn
nxan
onxn
nxon
hang (anticausative)
m/n(pl)
f
a
hpl
go
m/n/npl
f
a
hpl
carry (causative of go)
m
f
a
hpl
anxn
abanxn
nxan
banxan
ar
ar
or
abr
ar
rar
or
bar
i
ir
i
ri
r
bi
ib
The oxytonic variant of the imperfective stem is a verbal noun or Masdar, while
the homonymous form with falling stress on the first syllable is the debitive mood,
used as a finite predicate in both independent and dependent clauses.
Following Evans (2007), we have shown in Authier (to appear) that the former
Allative-Purposive case-marker -u present on the Imperfective stem of transitive verbs
explains their use as noun attributes (participles) and as an argument in the valency of
matrix verbs (Masdar),13 to which we shall now turn.
4.1 The Masdar as a verbal noun with S/P argument function

The Masdar is strictly speaking a verbal noun, and as such it bears rising word-final
stress. It can be inflected for most of the 15 cases of the Budugh declension; these
13. For the relationship between participles and masdars in the related language Kryz, see
Authier (2007).
Gilles Authier
forms do not concern us here. A non-inflected, or Absolutive form of the Masdar is

used in nominal predicates like:
ari-vi,
(6) dx urzr
laki alql
pis-i.
early m.rise.msd good-cop much m.lie.msd bad-cop
It is good to get up early, sleeping for a long time is bad.
ari-vi,
(7) dard aqu
ac cidir
si
aram-i.
worry n.tell.msd good-cop 3-n hidden n.do.msd forbidden-cop
It is good to tell ones worries, to hide them is bad.
(8) sordobut vodo

ov-i!
berries gather.msd enough-cop
Enough with gathering berries!
(9) vn Amad-o bada ar
ov-i.
2 a.-ad
near m.go.msd enough-cop
You should not go to Ahmad anymore.
(10) duman-ca xab-ar araxhr-n
igiduval-i.
enemy-in hand-el m.escape.msd-also bravery-cop
Escaping from ones enemys hands is bravery as well.
(11) a-cber susundakir cre sib
lazm-i.
3-pl rec.subel apart hpl.do.msd necessary-cop
It is necessary to separate them.
oxuts
(12) Zo kda-pie kul
irv,
kul
vi.
1.ad work-job house.a a.build.msd house.a a.demolish.msd cop
My job is to build a house, and to demolish a house.
Used as the argument of a number of verbs (finish, stop, manage, know (how), see
and want),14 the Absolutive Masdar controls Neutral agreement on these verbs. This
is a case of subordination, since the Masdar retains separate indexation of its own
Absolutive argument:
(13) rij-eri da-ra
qala

baarmi si-rni.
girl-erg mountain-pl.in head.in a.climb.msd managing n.do-impf
The girl was able to climb to the summit of the mountains.
(14) rij-eri mektub yezmi s
baarmi si-rni.
girl-erg letter.a writing a.do.msd managing n.do-impf
The girl can write a letter.
14. With want and know (how) the Masdar has an obligatorily coreferent subject which, if
expressed, is in the Dative case -z.
(15) lem-ildir
anrmi yuxhr yotu-ci.
donkey-erg braying a.be.msd n.cut-nar
The donkey stopped braying.
(16) rij-ez
kul-ca
or
yka-ci.
girl-dat house-in a.go.msd n.want-nar
The girl wants to go home.
(17) hil ye-z abr
yka-ci.
now 4-dat hpl.go.msd n.want-nar
Now we want to go.
(18) tse
ma rq
ye-z htsar-dad.
goat.a how a.reach.caus.msd 4-dat know-neg.prs-n
We do not know how to make the goat reach [that place].
The verb see can have its verbal complement in the Masdar form, and accordingly
take Neutral gender-agreement rather than agree with the Absolutive argument of
the subordinate verb (here alam boot which is Animate):
(19) u-n-uz suncu pinei-ri
qul-nu alam serv
idxa-ci.
3-h-dat one.obl cobbler-erg m.sit-seq boot.a a.sew.msd(n) n.see-nar
He saw a cobbler who was sitting and sewing boots (literally: cobblers sitting).
4.2 Raising the S/P argument of the Masdar in complement clauses

Alternatively, the same main verbs and a few others may agree with the S/P argument
of the Masdar. This Long-Distance Agreement phenomenon is characteristic of a sort
of verbal serialisation commonly found in Daghestanian languages:15
(20) gada-r aye
aldrmi s
baarmi s`-rn.
boy-erg stone.a lifting a.do.msd managing a.do-impf
The boy was able to lift a stone.
(21) gada-ri ug-unda idir kitab suxuts ar
s`-ravini.
boy-erg self-pl.in sister book.a read.msd learning a.do.prog.pst
Brother was teaching his sister to read books.
(22) Murad-r kitab oun
kyan sxn-c.
m.erg
book.a a.take.msd at.heart a.forget.caus-nar
Murad forgot to take a book.
(23) ali-ri xor
kyan o-ci.
Ali-erg dog a.bring.msd at.heart a.bring-nar
Ali remembered to bring the dog.
15. See for instance Polinsky & Potsdam (2001).
Gilles Authier
(24) malla-cz lem

uvot
hvatsar-da-b.
Mullah-dat donkey a.beat.msd a.know-prs.neg-a
Mullah doesnt know how to beat the donkey.
(25) zo rij ug-u duxoruz orun
yrka-cni.
1.ad girl self-ad son.dat f.take.msd f.want-nar.pst
He wanted to take my daughter for his son.
(26) ali-z ma kyan snx
yko-ci.
a.-dat tale.a at.heart a.forget.msd a.want-nar
Ali wants to forget the tale.
(27) kul-ca
barxuda-cz u-cber
aborot
ybka-ci.
house-in master-dat 3-hpl.abs hpl.kill.msd hpl.want-nar
The house-master wanted to kill them.
With verbs meaning stop or finish, the matrix verb, as a rule, agrees with the Absolutive
argument of the Masdar:
vaqa-ci.
(28) ayel-cir gitara orot
child-erg guitar.a a.strike.msd a.hold-nar
The child stopped playing the guitar.
(29) ayel-cir kitab suxuts vaqa-ci

/ pasar s`r-c.
child-erg book.a read.msd a.hold-nar finished a.do-nar
The boy stopped reading the book.
4.3 The Masdar as an imperfective participle

The imperfective stem (of any verb) or Masdar can also be used as the dependent of a
noun, that is as a participle, in relative clauses:
(30) yin ybkr
cga-ca
1pl hpl.be.ipf.prt place-in
To the place where we were.
(31) ad unkn
kul-ca
3-a a.remain.ipf.prt house-in
In the house where she stays.
(32) sa-b keppi, xhim
ar
kovsa
one-a thick smoke.n n.go.ipf.prt ember.a
A thick and smoking ember.
odox
(33) hec-ir
kul
woman-erg neg.a.demolish.pf.prt house.a
A house which a woman does not destroy (will stand for ever)
It often has a prospective, deontic value:

(34) u-ulur dan soul-
vaxt-i.
bird-erg corn eat-ipf.prt time-cop
It is time for the bird to eat corn.
(35) ad xhn suxh
cga-ca (vuq-a)
qxutsu-ri.
3-m grass mow.ipf.prt place-in meadow-in find-prs
They find him at the place for grass-mowing =the meadow.
(36) va rq
el-i
zox.
2.in a.reach.caus.ipf.prt word.a-cop 1.apud
I have something to communicate to you.
These participles can also be used without a head noun:

(37) biqab-ca soul
yeci.
plate-in eat.ipf.prt n.put-nar
He put food (cf. French manger, English things to eat/eatables) on a plate.
Since participles are unoriented, this gives rise to an interesting homonymy between
Action nouns (Masdar stricto sensu) and Agent nouns (free participles); this bare form
fills the absolutive case slot as the Subject of intransitive verbs or the object of transitive
controlling verbs. Such headless relative clauses are often lexicalised:
(38) za xakila araxhr
laki-vni.
1.in behind m.come.ipf.prt many-cop.pst
Many were courting me.
(39) vez
irq
za kl-imer da-d.
5.dat see.ipf.prt 1.in arm-pl cop.neg-n
What you see are not my arms.
(40) filan
eher-ce sa-r azan
yuts -vi.
a.certain city-in one-m call.to.prayer give.ipf.prt-cop
In a certain city there is a (call-to-prayer-giver =) muezzin.
The same absolutive unmarked form, but with the semantic role of a Stimulus, headless participle constructions are used for complementation with the verb see. The
controlling verb takes gender agreement with the Absolutive participant of the subordinate clause:
(41) gada-z (ug-unda) ada uba-ca ar
rqa-ci.
boy-dat self-pl.in father Quba-in m.go.ipf.prt m.see-nar
The boy saw his father going to Quba.
(42) padah-cr ug-o
rij-eri dogru yu
dxa-c.
king-erg self-ad girl-erg true say.ipf.prt?/msd? n.see-nar
The king saw that his daughter had told the truth.
Gilles Authier
Note that the fact that the Imperfective participle is not formally distinct from the
Masdar is a rather trivial phenomenon typologically, since complement and relative
clauses are more often related in languages than not, differing only in the nature of
their head (noun or verb).
4.4 Disembedding the Masdar: The debitive mood

Used in deliberative clauses,16 masdars are often introduced as an afterthought, and
postposed to their matrix verb. In this externalised position, the Masdar can acquire
the initial falling stress typical of finite verbs. Since the matrix verb is finite and
final marking the end of a self-standing, independent clause, the extraposed non-
finite form becomes stranded, and forcibly acquires syntactic autonomy, which
implies a shift and change in stress placement. A new, insubordinated form, though
segmentally homonymous with the Masdar is consequently used as a finite debitive
mood. This process is expected with Masdars ending in -u, since they are supposed
to derive from a case-marked, purposive form, which in itself may well not require a
matrix predicate.17 A similar use with modal debitive value seems to have extended
by analogy to all imperfective stems, including those not ending in -u, and indeed, all
Masdars can be used both in independent (sentence-final) and embedded (subordinate) syntactic position.
Independent use, finite form:
(43) tse halma srku!

goat this.way pv.slay.deb
A goat should be slain this way.
(44) malla-cr lem
votu!
Mullah-erg donkey(a) a.beat.deb
Mullah should beat the donkey.
(45) lem
or
donkey a.go.deb
The donkey should go.
16. A similar finite deliberative use of a formally non-finite form is Spanish A ver se Lets
see if See Authier to appear.
17. Other Lezgic languages have another finite form (Future) derived from a purposive form
ending in z or -s, that was originally a dative case-marker.
(46) zo hec-a
icin hihanz
`drq.
1.ad woman-in face.a nobody.dat a.neg.see.deb
No one should see the face of my wife.
(47) (ye-z tse havatsar-da-b)
: ma / srku?
4-dat goat a.know-prs.neg-a how (a.)slay.deb
We do not know how a goat should be slain.
(48) mall-cr lem
`
/ lemr
i.
Mullah-erg donkey a.go.caus.deb donkey.pl npl.go.caus.deb
Mullah should bring the donkey(/s).
(49) vn gerek yxta cga-ca ar,
har cga va-z rqi.
2 must faraway place-in m.go.deb every place 2-d n.see.deb
You have to go and see faraway places.
Subordinate use:
(50) mall-cz lem

ykaci
/ ykoci.
Mullah-dat donkey a.go.caus.msd n.want.nar n.want.nar
Mullah wants to bring the donkey.
(51) mall-cz lem

htsar-da-d / hvatsar-da-b.
Mullah-dat donkey a.go.caus.msd know-neg.prs-n / a.know-neg.prs-a
Mullah knows how to bring the donkey.
(52) ye-z tse (ma) sork
hvatsar-da-b.
4-dat goat how (a.)slay.deb a.know-prs.neg-a
We do not know how to slay a goat.
(53) ye-z tse havatsar-da-b
ma sork.
4-dat goat a.know-prs.neg-a how (a.)slay.msd
We do not know how a goat should be slain.
A finite debitive form can also be subordinated: By means of the complementiser ki,
(54) vn i-vi
ki, kul
rve-ni
ad xutsu?
2 what-cop ki house.a a.build-seq 3-a a.demolish.deb
Who are you to build a house and then demolish it?
(55) a-n-z
hatsar-da-d-ni
ki, bolu rij i
s`.
3-h-dat know.ipf-neg.n.pst ki big girl what a.do.deb
He did not know what to do with his older daughter.
(56) zaz yka-cki
rij `rq.
1.dat n.want-nar+ki girl.a a.see.deb
I want to see the girl.
Gilles Authier
(57) vaz yka-cki,

z-n srqar.
2.dat want-nar+ki 1-also m.die.deb
You want me to die also.
(58) ali-z yka-ci
ki, melim-cir ug kyan snxi.
a.-dat n.want-nar ki teacher-erg self at.heart m.forget.deb
Ali wants his teacher to forget him.
5. Morphosyntactic classification and perfective stem uses

Uninflected perfective verb stems, like imperfective ones, have either nominal
dependent uses or modal injunctive values, with the same stress distinction which
was observed in the imperfective between the Masdar and the Debitive mood. But,
depending on their phonological shape, perfective stems can be divided into three
broad classes, labelled augmented (the root-consonant is followed by a sonorant),
thematic (the root-consonant is followed by an low vowel), and radical (the rootconsonant is either bare or followed by high vowel). Each class of perfective stems
shows specific mapping for finite and non-finite uses. The following charts show that
the perfective verb stem can be used in imperative, sequential, optative, participles and
short infinitive functions with distinctive stress properties.
5.1 Augmented perfective class

These augmented verb stems can be used in imperative, sequential and short infinitive18 functions, with distinctive stress properties, while optatives and participles take
special suffixes:
sit
m/npl
f
a
hpl
take
m/n/npl
f
a
hpl
Imperative
Optative
Sequential
Infinitive
Participle
qul
qul-e
qul
aql
aqul-
qul
oql
lqul
qul
etc.
bqul
yin
rin
vin
bin
lqul
bqul
yin-e
etc.
yin
rin
vin
bin
alql
etc.
abql
eyn
ern
evn
ebn
18. There is also a long infinitive, bearing the dative marker -z.
eyin-
etc.
5.2 Thematic perfective class

Imperative
Optative
Sequential
Short Infinitive =Participle
be
m/n/npl
f
a
hpl
saxh!
srxh!
sxh!
sabka-da!
sxha
srxha
sxho
sbka
sxha
srxha
sxho
sbka
saxh
sarxh
soxh
sabk
keep!
m/n/npl
f
a
npl/hpl
qa
raq
vaq
baq
qa
raqa
vaqa
baqa
qa
raqa
vaqa
baqa
aq
araq
avaq
abaq
The Thematic perfective stems are used as (1) optatives, (2) short infinitives, (3) participles, (4) sequential converbs.
5.3 Radical perfective class

kill
m/n/npl
f
a
hpl
enter
m/n
f
a/hpl
npl
Imperative
Optative
Sequential
Short Infinitive =Participle
tw!
rtw!
btw!
barta-da!
to
rto
bto
barta
tu
rtu
btu
barta
at
art
abt
abart
w!
bw!
bo
bu
ab
rw!
barw!
ro
bara
ru
bara
ar
abar
These irregular and labialised-root (or third class/strong) verb stems are used as
(1) short infinitives, (2) sequential converbs, (3) participles.
5.4 Perfective participles

The perfective participles are non-finite verb forms used in relative clauses. They are
characterised by rising final stress as on nouns. Perfective verb stems ending in a
sonorant take a special participial suffix -i to be used as participles; -u on Cabialised
stems agree with an Animate (ABS):
(59) la skr-
uval-co cga-ca
that a.slay.pf-prt sheep-ad place-in
In exchange for that slain sheep
Gilles Authier
Thematic and radical classes use the verb stem:

(60) rij-eri ug-ur
yp
sii-ri.
girl-erg self-erg say.pf.prt do-prs
The girl will do what she has said.
(61) avat
doldom-ca sas
a.beat.pf.prt drum(a)-in sound
The sound of the beaten drum.
u
(62) vn yez
xurak sal!
2 4.dat bring.pf.prt bread eat.imp
Eat the food that you have brought us!
(63) v xazna
yobt
quqa-r uldur
5.ad treasure(a) a.cut.pf.prt forty-m robber
The forty robbers who have stolen your treasure
(64) vn vixh
rix-a
2 m.go.pf.prt way-in
On the way by which you went
In rare instances, the participle serves as predicate in Wh-interrogative main clauses:

(65) ulu rij heyen vxh?
this girl where a.go.pf.Interr
Where has this girl gone?
5.5 The short infinitive

The short infinitive is unmarked in all three classes; it is rare but attested as object-like
complementation or as intransitive subject, that is, filling the unmarked-absolutive
constituent slot:
(66) zn rx-a arxh
sundovor-un
yumou!
1 way-in m.descend.inf one.h.adel-even proh.tell
Do not tell anybody that I have gone away!
ari-vi.
(67) ulca ibkn-n
ivovor
outside hpl.remain.inf-even 3.not.h.adel good-cop
Even remaining outside. is better than this.
(68) za til
kbsa-ci, kbs
az-i?
1.in finger.a a.bite-nar a.bite.inf what.dat-cop
I bit my finger, (but) to what purpose did I bite it?
The short infinitive tends to form a single phonological word when used in very common constructions with the verb be (able) immediately following, the subject being
expressed in the Adelative case:
(69) andovor
ebxr
sbka-da-b
=ebxirsabkadab.
3.hpl.adel hpl.sleep.inf hpl.be-neg.nar-hpl hpl.sleep.can.neg.nar
They could not fall asleep. (lit. from them were not (to) sleep)
5.6 Short sequential converbs

All perfective stems can be used as sequential converbs, which are characterised by
initial rising stress. This coordinating strategy is parallel to the Azerbaijani converb
in Ib, but the Budugh sequential, contrary to the perfective general converbs of
Azerbaijani or Akhwakh (Creissels, this volume) is usually the bare stem. It is used in
clause-chaining with same subjects (S/A/E):19
(70) ari
ton-a qosiye srga
kal srxha-ci.
old.wife oven-in inside f.burn.seq coal f.be-nar
The old woman burned in the oven and became coal.
(71) v-n andar ugugunda-zn yin
to!
2-also 3pl.erg refl.pl-cum take.seq bring.opt
Let them also take you with them.
(72) niana yin
vxhe!
sign
n.take.seq m.go.opt
Let him take the sign, and go.
u!
(73) lnc
vin
sword.a a.take.seq a.bring.imp
Take the sword and bring it!
(74) xillda yiqiye ka

rkin-ci.
horse.in back.in m.stick.seq m.remain-nar
He remained stuck on the back of the horse.
(75) middei-cir hec
rtu
cergir-ci.
enemy-erg woman f.kill.seq f.throw-nar
The enemy killed the woman and threw her away.
19. Note that coordinative conjunctions exist in Budugh like in other Lezgic languages, but
are all Arabic loans associated with formal registers, and not often used.
Gilles Authier
The case-marking of the shared subject (Agent or Single argument: note the accusative alignment, or Experiencer), is most frequently controlled by the valency of the
sequential verb form:
(76) middei-cir furi-mber sbkir
vxhi-ci.
enemy-erg man-pl hpl.slay.seq m.go-nar
The enemy slew the men and went.
(77) Malla-cr lem
vata
vxhi-ci.
Mullah-erg donkey a.beat.seq m.go-nar
Mullah beat the donkey and went.
(78) rij vxh lemer
ta-ci.
girl a.go.seq donkey.pl npl.beat-nar
The girl went and beat the donkeys.
The unmarked or short Sequential form is also often used with the verb finish which
agrees either with the Ergative Agent of a transitive verb:
(79) dide-ri
gada szin
pasar
srxha-ci.
mother-erg boy m.wash.seq finished f.be-nar
Mother finished washing the boy.
or with its Absolutive Patient:

(80) gada-r kitab sxo
pasar
s`r-c.
boy-erg book.a read.seq finished a.do-nar
The boy finished reading the book.
Note there is an alternate, suffixed form with -nV which seems to be used when nonshared arguments are inserted between the two verb forms:
(81) ari yxhar-a ki, vn virxhi-ni xhad-ri si`i.
good n.be-evt ki 2 f.go-seq water-pl npl.do.deb
Youd better go and make a stew.
(82) ara
sa-b mek btu-na a-dr
tona
ra-ci.
old.wife.in one-a kick.a a.kick-seq 3-abs.f oven.in f.push-nar
(He) kicked the old woman and pushed her into the oven.
5.7 Optative forms

The optative, like all non-interrogative finite moods, is characterised by initial falling
stress and always has a low vowel. Thematic stems are identified as optatives only if
they show this initial falling stress:
(83) malla-cr lem

i`vata.
Mullah-erg donkey a.beat.opt
Let Mullah beat the donkey!
(84) ari-n-a
ramat, itan-ca nalat sxho.
good-h-in mercy devil-in curse.a a.be.opt
Mercy to the righteous, and a curse to the devil!
The optative is widely used in subordination with purposive meaning. It is postposed

and is often, but not obligatorily, introduced by ki or other particles like heter (for
wishes) after all sorts of verbs, including want. The matrix clause ends with this
optative-controlling particle:
(85) zn ayel-ber a-n-o yvaci
heter baqa.
1 child-pl 3-h-ad give-nar so.that hpl.keep.opt
I gave the children to her to keep.
(86) rij-ez
ykan saxha-cki,
Amad kua-ca ro.
girl-dat n.wish n.be-nar+ki Ahmad street-in m.lead.out.opt
The girl wanted to make Ahmad go out in the street.
(87) ada xhad altu-lni ki, sadke edil sxha.
on
water pour-impf ki a.little soft be.opt
She would pour water on it so that it should become a bit softer.
(88) diderbez
yka-ci
ki, ug-unda ayelber nua sbka.
mother.pl.dat n.want-nar ki self-pl.in child.pl glad hpl.be.opt
It is the mothers wish that the child be glad.
The Augmented stems must add the low vowel to their bare perfective stem in order
to express optative modality:
(89) hatsar-ndar hadatsar-ciber
ar
sbir-e!
know-hpl.erg neg.know-hpl.abs teaching hpl.do.pf.opt
Let those who know teach those who do not know.
Radical verb stems form their optative by lowering the high vowel of the verb stem
(u > o, > , i > e):
(90) imeri
yodortu-ra
ki, lm vo bada `xh.
3.nhpl.erg neg.leave-evt ki death.a 2.ad near come.opt
They will not permit any eyes to come near you.
(91) a
gre
uval e
v`xh?
what.in considering sheep today a.go.opt
Why should a sheep go today?
Gilles Authier
(92) mall vxhe! (93) lem

v`xh!
Mullah m.go.opt
donkey a.go.opt
Let Mullah go!
Let the donkey go!
6. Conclusion
We have shown that, although endowed with rich segmental morphology, a language
like Budugh makes a very central use of stress or intonational distinctions to discriminate between the different syntactic functions of lexical verbs. The segmentally
unmarked verbal forms or stems that we have presented are very common in natural
Budugh discourse. One can find sentences only made up of such forms with different
accent patterns and syntactic functions:
(94) hzebtir
bu
so!
a.leave.imp a.go.out.seq a.freeze.opt
Let her go out and freeze!
The use of these forms gives much concision to Budugh discourse. Predicates tend
to be more often subordinated than in European languages, because subordinationmarking is very light, especially for clause-chaining, using sequentials.
Nevertheless, given their polysemy, labelling these unmarked forms is not
straightforward. It is stress pattern and stress position which help distinguish between
verb stems used as modifier or arguments (final stress), and verb stems used as finite
forms and as head-predicates (initial stress). Once recognised, such stress patterns
clarify the situation. Sequential verb forms fall in between.
Finiteness and non-finiteness are marked by stress pattern, i.e. stress types and
positions, and are not linked to verb stems segmental forms. Bare stems have not
only dependent uses: most perfective and imperfective stems also have modal values;
these finite modal values might be the result of some kind of insubordination,20 but
since not much is known about the diachrony of Budugh, we reserve our judgement
on this hypothesis.
In any case, the fact that the same verb-forms in Budugh can be found in dependent and independent clauses forces us to discard the current definition of finiteness as
a morphological feature. Kalinina and Sumbatova (2007) have presented arguments to
dissociate the assumed link between main clause status and finiteness, allowing certain
types of main clauses to be non-finite, provided that they take default non-assertive or
non-indicative modality. This is confirmed by the Budugh data: in the language, conversion by accent pattern shift is a productive feature in the verbal paradigm, although
the languages grammar is otherwise characterised by rich inflectional morphology.
20. See Evans (2007).
Fairly strict word order rules, along with specific prosody associated with purely syntactic categories, make it unnecessary to write stress position and nature (falling or
rising) in text transcriptions. Word order rules and stress properties of word classes
also provide sufficient explanation for this remarkable distribution: there is no reason
to assume any segmental origin for the functional stress shift leading to this state of
affairs. It replaces category-changing deverbal derivation which, as in most Daghestanian languages, is very scarce. Indeed, such a syntactic stress system may well be an
inherited feature.
Abbreviations
1
first person singular
2
second person singular
3
third person
4
first person plural
5
second person plural
a
animate noun or agreement
abs absolutive
ad
adlocative
apud apudlocative
caus causative
com comitative
cop copula
dat dative
deb debitive
dir directive
erg ergative
evt eventual mood
f human feminine noun or
agreement
gen genitive
hpl human plural noun or agreement
if
conditional converb
imp
imperative
in
inessive
el
elative
inf
infinitive
ipf
imperfective
m human masculine noun
or agreement
msd masdar
npl
non-human plural
obl
oblique
opt
optative
nar narrative tense
pf
perfective
prog progressive
proh prohibitive
prs
present
pst
past
prt
participle
rec
reciprocal
refl reflexive
seq
sequential converb
subel subelative
References
Authier, Gilles. 2007. Actance et changements de valence en Kryz. In Le Sujet, AnadDonabdian
(ed.), 1131. Paris: Cahiers de Linguistique de lINALCO.
Authier, Gilles. 2007. Participes et Masdar en Kryz: Une syntaxe mixte. Nominalisations. Faits
de Langues 33: 153165. Paris: Peeters.
Gilles Authier
Authier, Gilles. 2009. Grammaire kryz. Paris: Peeters.
Authier, Gilles. to appear. From adlocative case to debitive mood without desubordination. In
Tense, Mood and Aspect in Daghestanian Languages, Gilles Authier & Timur Maisak (eds).
Evans, Nicholas R.D. 2007. Insubordination. In Finiteness, Irina Nikolaeva (ed.). Oxford: OUP.
Haspelmath, Martin. 1989. From purposive to infinitive A universal path of grammaticalization.
Folia Linguistica Historica 10.
Haspelmath, Martin. 1994. A Grammar of Lezgian. Berlin: Mouton de Gruyter.
Haspelmath, Martin. 2002. Understanding Morphology. London: Arnold.
Kalinina, Elena & Sumbatova, Nina. 2007. Clause structure and verbal forms in NakhDaghestanian languages. In Finiteness, Irina Nikolaeva (ed.). Oxford: OUP.
Polinsky, Maria & Potsdam, Eric. 2001. Long distance agreement and topic in Tsez. Natural
Language and Linguistic Theory 19: 583646.
Converbs and adverbial clauses in Badaga,

a South-Dravidian language*
LACITO (Laboratoire des langues et Civilisations Tradition Orale),

Fdration Typologie et Universaux Linguistiques, CNRS
This paper analyzes complex sentence constructions in Badaga, a Dravidian
language spoken in Southern India. After identifying the strategies used to
express clause dependency, mainly based on converbs, adjectival participles and
a quotative element, it examines their use and distribution in different semantic
functions: temporal, causal, purposive etc. A few sample cases of the simultaneous
use of these strategies in complex sentences will be given. Noticeable findings are
that (i) there is a clear distinction between a polyfunctional dependent contextual
converb and several specialized converbs, constructing more autonomous
adverbial clauses, and (ii) none of the clause dependency strategies is specific to
a single semantic function. Crucially, word order, and prosodic and pragmatic
factors contribute to the interpretation of dependent clauses.
1. Introduction
In Badaga,1 as in the other South Dravidian languages, the construction of complex
sentences involving several clauses is based on linguistic devices which differ significantly from the familiar use of conjunctions in European languages.
*I wish to thank the two reviewers, P.S. Subrahmanyam and B. Bickel, as well as I. Bril for their
insightful comments, corrections and advice.
1. The Bad.aga language, spoken by about 200000 people in the Nilgiri Hills, is situated at
the junction of the three southern states of India, which have Tamil, Kannada and Malayalam
as their main languages. The South-Dravidian languages share typological features such as:
nominative-accusative syntax, suffixing morphology, consistent head-final order and finite/
non-finite verb distinction. As for other Nilgiri languages, the subclassification of Badaga into
the South-Dravidian group is problematic. Still officially and erroneously considered as
a dialect of Kannada in the Census of India 2001, this old lingua franca of the Nilgiri has to
be grouped not only with Kannada [one of the two main subgroups of SDr.] but also with
Kodagu, Kota and Toda [which belong to the other, Tamil-Kodagu, subgroup]. [] Badaga is
not closer to any one of the SDr. languages (Balakrishnan 1999:5354).
In South Dravidian languages, which are characterized by some asymmetry

between main clauses and dependent clauses, at least four distinct morphosyntactic
strategies are available to construct multi-clause complex sentences:
A.
B.
C.
D.
Subordination, with adverbial participles heading the dependent clause

Relativization of the clause using an adjectival participle
Substantivization of the verb heading the clause
Quotative strategy
The first strategy (A) has been fairly well investigated in typological studies under the
general investigation of converbs, but the others, notably the nominalizing strategies
(B and C) involving the use of adpositions and similar elements have less frequently
been considered in the set up of complex sentences.
This paper aims at characterizing these various types of clause dependency and
instanciating their actual uses and distribution in complex sentences. The analyses
will be carried out in reference to the relevant discussions in Dravidian linguistics
and current typological proposals. Thus the study of the converbal strategy will be
subdivided into contextual converbs and specialized converbs according to the distinction elaborated by I. and V. Nedjalkov in the 1990s (1995, 1998) and used by many
authors in typology, notably Haspelmath and Knig (1995).
Sections2 to 4 describe the morphological and syntactic properties of these strategies. Section 2 explores the converbal strategies. Analysis of the converbal strategy
supports the clear distinction between a polyfunctional contextual converb, used in a
variety of constructions, ranging from clause-chaining to modifying functions, and the
other types of converbs, semantically specialized, whose range of functions is limited
to the expression of circumstances. Paragraph 2.1. presents the most neutral strategy,
clause-chaining with contextual converbs, and sketches its morphosyntactic properties. Paragraph 2.2. presents the various specialized converbs based on their relevant
semantic domain. Section3 will focus on strategies B and C, both based on an adjectival participle form, used with an adverbial or nominal head in one case and derived
as a nominalized form of the verb in the other case. In these constructions, the use of
case markers and postpositions specifies the semantic relation of the dependent clauses.
Section4 will briefly show that the uses of the quotative, derived from the verb say,
basically stem from the previous morphosyntactic strategies. Section5 starts from the
semantic functions, i.e. temporal, causal, purposive etc. of the dependent clauses and
examines alternative ways of expressing them. The last section introduces the internal
structuring and hierarchies resulting from the simultaneous use of various strategies in
lengthy complex sentences.
Before coming to the description of these strategies, some general characteristics of the verbal clause in Badaga should be briefly mentioned. First, two important
roperties should be highlighted. Whatever the actual categorial status (finite, nonp
finite, nominalized) of the verb heading a clause:
i. the expression of its argument structure is very little altered
ii. the verb form keeps some markings characteristic of verbal categories,2 such as
aspect, tense, mood or polarity (i.e. no verb root form can head a clause)
Secondly, a brief account of the basic verb forms involved in the construction of complex sentences is necessary to grasp their systemic relationships.
Badaga verbs, subgrouped into four classes following the variants of their stem
formatives, are inflected along two paradigms, type for class 1 verbs, the most
numerous and regular, and type for class 2-3-4 verbs. The following table with the
verb md.u to do (cl. 1) gives an idea of the structural relationships between the basic
finite, non-finite and nominalized verb forms.
The core of verbal morphology is based on the opposition of three distinct stems,
traditionally called: Past, Non-Past and Negative in Dravidian linguistics. In spite
of these labels,3 the stems do not refer directly to the localisation of the events in reference to the speech time. To avoid any misunderstanding, it is preferable to refer
to them as S1 (Non-Past) and S2 (Past), following Emeneau 1967. Table1 below
shows that the polar distinction between two affirmative and one negative forms is
prevalent and consistent throughout the morphology of non-finite and nominalized
2. This restriction calls attention to the fact that, contrary to a nominal bare root lexeme
which can fill a syntactic noun phrase slot (or a predicate slot) without any modification or
additional categorial specifier, a verbal lexeme requires additional categorial specifications
(minimally an intonation pattern for 2nd pers. sg. Imperative) to fill a major syntactic slot, be
it predicative or not. In this perspective, a verb root cannot form a syntactic word.
3. This stem distinction reflects an unusual encoding of time into the verb forms. In the absence
of a stem morph, the verb remains notional, not representing any real event. The presence of a
(tense) stem morph turns it into an event inserted into a world where time flows. However,
except in cases where the realization of the event is explicitly denied (negative stem), the real
status of the event is not always directly accessible. The affirmative stem formatives merely indicate a preceding (S2) or following (S1) lapse of time associated with the verbal notion in
the representation of the event. In this sense, the stem morphs function more like aspectual
than tense markers, but they do not exactly match the usual Perfective/Imperfective distinction,
notably because the S2 stem does not entail achievement. The precise status of an event in a
given sentence is the result of a complex calculation which depends not only on the components
of the verb form (which do not give direct clues: thus, contrary to expectations, V+S1(NonPast) + aux. neg. ille gives a Realis Negative (present or past) and V+S2(Past)+person sfx.
with a stative verb like to be can give a sense of present idde I am here now), but also depends
on contextual elements (adverbs, other clauses, discourse settings, etc.).
V1Nz md.-uv-adu
> md.du
Substantivized verb
V2Nz md.-id-adu
Nominalized
P1 adj md.-uv-a
> md.
P2 adj md.-id-a
Adjectival Participle
Adv (Simultaneous)
md.-uv-ane
> md.ne
P1
Pc md.-i
P2 Adv (Anterior)
md.-id-ade
S1 (Non-Past) md.-uv-
T3 Prospective
md.-uv-e
Contextual Converb/
Adverbial Participle
Positive
T1 Present
md.-i n-e
Non-finite
T2 Past
md.-i d-e
S2 (Past) md.-i(d)-
Declarative mood
Stems
Polarity
Finite
Table1. Main forms of the verb md.u to do
(Infinitive md.-a_)
PNEG
(PadvNeg)
md.-d-e
VNeg Nz
md.-d-adu
PNeg adj
md.-d-a
Bare stem md.Unspecified
(+ periphrastic
forms)
Negative md.-dNegative
Negative 0
md.-e
forms (while, for the finite verb, a number of periphrastic negative forms have been
added to negate dissymmetrically, i.e. without a one on one correspondence the
synthetic affirmative ones).
Like other Nilgiri languages, Badaga presents the particularity of having finite
Present-Future tense (T1) forms as well as Past tense (T2) forms built on S2, the
so-called Past stem -i(d-), actually a kind of Realis mood stem, while, in this case,
the S1, Non-Past stem uv- is used for a Prospective tense restricted to modal
irrealis meanings.
Among the non-finite verb forms, we have unified the coding of P[articiple] for all
the syntactic functions of modifier, whether of a verb or a clause: PAdv[erbial], or as a
modifier of a noun phrase: PAdj[ectival], and distinguish it from the encoding of the
substantivized forms of the verb (V.Nz) which can fulfill the major functions of a noun.
In these non-finite forms, the regular morphological stem distinction is encoded
by a superscript1 for S1 Non-Past, 2 for S2 Past andNeg for Negative stem. The infinitive stands apart as it is nearly exclusively used as a component [verb root+a_Aux] of
modal auxiliary forms.
2. Converbal strategies
The notion of converb as a nonfinite verb form whose main function is to mark
adverbial subordination (1995: 3) was firmly established in Haspelmath & Knigs
volume in 1995. In the same volume, Nedjalkov proposed distinguishing three semantic
types ordered by decreasing specificity of meanings: (1) specialized converbs have
one or two meanings of the adverbial type; (2) contextual converbs have three or more
adverbial meanings that are realized under certain conditions; (3)narrative converbs
express a coordinative connection that advances the narration (1995:106). Keeping
this typological frame for the final evaluation, we will present the data in Badaga,
which clearly distinguish only two types of converbs, following the commonly used
dichotomy between contextual converbs, with indeterminate and contextually variable
meaning, and specialized converbs, with specific adverbial meaning.
The Dravidian converbal strategies present many similarities with the use of converbs in other language families, notably Turkic languages (Johanson) and Burushaski
(Tikkanen), as described in Haspelmath & Knig 1995. In Indian linguistics, converbs
were identified as an areal feature by Masica (1976: 108140), and Subrahmanyam
(2006) gives a more recent account of their functions.
In Badaga, as in other Dravidian languages, some constraints on and characteristic features of the construction of complex sentences perfectly fit the typological set up favoring the development of converbal strategies: (i) only one main
verbal predicate per sentence and (ii) a general absence of coordinative conjunctions for clauses.
More precisely, in Badaga, it follows from (i) that, except for the main, morphologically finite verb, which comes at the end of the sentence, all4 other verb forms
involved in a multi-clause construction are deranked5 from the characteristic verb
function (sentence head-predicate), and recategorized as non-finite forms (participles) or substantivized forms. The non-finite forms heading a dependent clause are
a type of converb, lacking subject agreement. In Badaga, the contextual converb has
a single representative, the Pc form,6 while several adverbial participle forms come
under the category of specialized converbs.
2.1 The contextual converb (Pc)

The contextual converb, coded Pc7 here, belongs to the same category as other adverbial participles, in the sense that, syntactically, it depends on a verb or clause. However,
it differs from them by its morphological systemic relations: it does not enter into
the regular stem opposition, being only marked as an affirmative converb (md.i, as
opposed to a negative one: md.-d-e, see Table1), and by its larger functional distribution. Apart from its clausal use, the Pc converb enters a number of periphrastic
constructions with tense/aspect/mood auxiliaries; it is also needed to form lexical
4. Except of course the verb forms introduced by the quotative.

5. The term deranked, introduced in typology by Stassen 1985, is based exclusively on the
inability for a verb form to occur in independent declarative clauses and is used in The World
Atlas of Language Structures (Haspelmath et al. 2005, online version 2008) in opposition to
balanced. For our purpose, the notion of deranking appears more relevant than the notion
of non-finiteness, as it can cover both participle forms (non-finite) and nominalized forms
(which cannot be said to be morphologically and syntactically non-finite).
6. In Nedjalkovs article, the Tamil equivalent to the Badaga Pc form is assigned to the narrative type. However, although it is certainly the converb with the least specific meaning,
it does not always express a coordinative connection that advances the narration. As the
interpretation of the Pc converb depends essentially on the context and co-text, the label
contextual converb seems more appropriate.
7. Structurally identical to the Past stem: verb root + S2 morph, the Pc form (often called
Past Participle in Dravidian tradition) differs from it by its status of free standing phonological word. With verbs of class 1 for which the S2 morph has a vocalic allomorph -i, there is
no obvious change: aku to put, S2 stem: ak.i-, Pc aki. However, for verbs of classes 2, 3 and 4
which have a consonantal S2 morph, the Pc form requires a final epenthetic vowel -u to form
a word. Ex. class 2 ari to know, S2 morph -d-, S2 stem: ari.d-, but Pc: aridu, cl. 3 (-t-): bi to
keep, S2 stem bi.t-, Pc bitu; cl. 4 (-nd-): mede to fly, S2 stem mede.nd-, Pc: medendu.
compound verbs.8 Alongside these predicative uses, the Pc form is a common derivative for adverbs,9 but rarely turns into a case marker or postposition10 in Badaga.
From a typological point of view, it can be said that the use of converbs in Badaga
has many affinities with other language families, especially Turkic languages, concerning the various levels of their uses.11
Moreover, as stressed by Bisang (1995) for Tamil converbs, the Badaga Pc converb, with its wide variety of clausal as well as non-clausal uses, is also very similar, in
its functions, to verb serialization in its broad and narrow senses (Bisang 1995:145),
except that it differs by a marked syntactic dependency due to the overall asymmetry
between finite and non-finite verbal forms in Dravidian languages.
2.1.1 Clause-chaining with the contextual converb

The most common strategy for complex sentences is to use Pc converbs in all clauses
except the last, as in the following pattern:
V1PC, V2PC , VnPC Vfinite
The converb basically expresses a syntactic relation of dependency on a clausal head: a

main verbal or nominal predicate or another dependent clause head which follows it,
whether immediately or not. Its precise function depends on the overall structure of
the complex sentence. The semantics of the dependency relation is left unspecified by
the Pc form itself, but may be constrained by various structural and prosodic factors.
As in the other languages of this type, the distinction between coordination and subordination is often blurred. On the one hand, semantically, they resemble coordinate
clauses, joined at the same level, strung along rather like beads on a string as Foley
(this volume) presents them. On the other hand, they are syntactically subordinate,
dependent on the specifications of some higher predicate. Though the Pc converbs
carry some minimal (relative) tense and polarity specifications, they lack agreement
features and modal specifications.
8. Ex. ottu b (carry.Pc_come) to bring; kan.d.(u)-id.i (see.Pc catch) to find; bitu nod.u (keep.
Pc see) to compare.
9. Ex. tirigi again < Pc of tirigu to turn. The Pc of to become -agi is the most common
derivative for adverbs: sandosa happiness > sandosavagi happily, neja true > nejavagi really.
10. The Pc of agu to become is added to the dative case suffix -ga and grammaticalized into
a postposition -gagi meaning for, in order to, for the sake of, concerning, among.
11. Thus, the four levels of integration identified by Johanson 1995 for Turkic languages can
also be applied to the Badaga Pc converbs: 1. full predications (each has a first actant of its
own); 2. Construction with a shared first actant; 3. (semi-)lexicalized converb constructions;
4. Periphrastic constructions where the lexical verb is turned into a converb while the (aspectual) auxiliaries carry the inflectional markers.
The Pc converb can be used in any temporal context,12 past (1), present (2) or
future (3):
(1) Jgi mane.ga bandu, tindu, origi
but.t.a.
Jogi house.dat come.pc eat.pc sleep.pc pfv.3
Jogi came home, ate and slept.
(2) Bla Mda avasara hgi hat.t.iga
kad.edu banna.
Bela Mada haste go-pc village-dat leave-pc come:T1:3m
Bla Mda hurries up and leaves for the village.
(3) [nanga hosa raja ettuvo].
desa
pura.ga
tandura udu
we
new king takeT3.1pi country whole.dat drum beat-pc
kaida [] akuvo.
notice
put.T3.1pi
[We will select a new king.] Throughout the country well beat the drums
and put up notices.
The Pc converb can be linked both to modal verbs (4) and to nominal predicates (5):
(4) hav-e,
ni eddu
sed.e ettu!
snake-emph you get.up.pc hood take.imp
Snake, raise your hood!
(lit. Having got up set your hood, translation from the above sentence)
(5)

ajji
eddu
ajjaga
ed.e.
old lady get.up.pc old man.dat place
The old lady having got up, the place is for the old man.
(Proverb meaning that things go to the appropriate person, but by turns)
The Pc converb can be repeated to express intensity of the action:

(6) gubisi od.i od.i od.i ondu ed.e oge jari
sparrow run.pc run.pc run.pc one place loc slip.pc
buddu but.t.a.
fall.pc pfv.T2.3
The sparrow ran, ran, ran, at one point, it slipped and fell down.
2.1.1.1 Argument constraints

Ellipsis of coreferent arguments is a prevailing device in Badaga. (Pro-) nominal
expression of the arguments occurs only when necessary, to avoid ambiguity (see
(12) below).
12.
Generally speaking, there is no sequence of tense phenomena in Badaga.
Syntactically, it should be noted that there is no absolute constraint on the identity

of the subject across the various clauses. When the verbs are intransitive, free chaining
can occur:
(7) m hdu, dri ella tgi
sudda tu.
rain beat.pc road all wash.pc clean become.T2.3n
It rained and all the roads were washed clean.
There are certainly more constraints on the chaining of transitive clauses. However,
P.S. Subrahmanyam summarizes the situation: the general rule (originally formulated
by Panini for Sanskrit) that the verb(s) that take the past participle and the finite
verb should have the same agent holds []. But there are important exceptions for
it in Dravidian. (Subrahmanyam 2006:217). Three sets of conditions under which
the agent need not be the same are given: (i) when the sentence refers to a change
in the climate, situation, etc. (ex. The rain stopped and sunshine came); (ii) when
the participle clause expresses the cause for the action of the main verb (ex. The
houses collapsed because of the earthquake) and (iii) when the main clause refers
to a stretch of time that has elapsed after the action denoted by the past participle
took place (ex.two months have passed since I visited Hyderabad). Furthermore, the
constraints, which are often heavier with human agents, vary among the Dravidian
languages and that pragmatic considerations play a crucial role in interpreting a sentence (ibid. 2006:218).
Along these lines, it can be said that Badaga also shows some constraints on
matching semantic roles and syntactic encoding between the dependent clause(s) and
the main clause. Thus, in experiential and other oblique constructions, when the common topic is an agent in one clause but not in the other, the conflict is solved either by
using a nominative experiential verb (to feel) which allows placing the shared subject
in initial position (8b) or by displacing the case marked (dat) experiencer into its own
clause (ex. 9 to Sevana surprise came):
(8) a.

b.

jli.ya
ktu
Sevana ciriya
that news.obl hear.pc Sevana surprise
Sevana
jli.ya
ktu
ciriya
Sevana that news.obl hear.pc surprise
Hearing that news Sevana felt surprised.
pat.t.a.
feel.T2.3
pat.t.a.
feel.T2.3
(9)
jli.ya
ktu
Sevana.ga ciriya tu.
that news.obl hear.pc Sevana.dat surprise become.T2.3n

/*Sevanaga jliya ktu
Hearing that news, Sevana became surprised.
On the whole, subject or maybe more exactly topic continuity across clauses is preferred, but in Badaga the same agent rule has more exceptions than those mentioned
in Subramanyam. Especially concerning the narrative use of the Pc converb, where
many irregularities appear. However, in these cases, an explicit mention of the proper
subject/topic, ex.: ivaka vs. dod.d.a Gaud.a in (10), tends to clarify the situation.
(10) ivaka atte endu dod.d.a Gaud.a hen. n. a koricina.
they so say.pc Great Gauda girl.obl call.T1.3m
So they decide and the Great Gauda calls his daughter.
In some cases, a nominal clause may even follow a converbal agentive clause as in (11):
(11) avaka nattaru ellava seti
ondu innuru
they relatives all.obl gather.pc one two.hundred
munnuru
mane.
three.hundred house
They have gathered all their relatives and [the village has now] some two
three hundred houses.
In sentence (12), the change of subjects between ava she, the demoness in the first
clause, ava and Kariabetta in the second clause (sande d.i) and Kariabetta alone
(represented in the a, 3sg. pers. index, devoid of gender specification, on the final
auxiliary but.t.a) creates an ambiguity which is clarified by the next sentence where the
demoness is case marked (obl) as an object.
(12) ava allinda
bandade,
kariabet.t.a.g
ava.g
sande
she there.abl come.P2adv Kariabetta.dat.add she.dat.add fight
d.i
koddu but.t.a.
araki.ya
kodda.
play.pc kill.pc pfv.T2.3 demoness.obl kill.T2.3
After she arrived, she and Kariabetta had a fight and [he/she] killed
[him/her]. He killed the demoness.
2.1.1.2 Temporal constraints

There are also probable constraints on the homogeneity of the temporal setting of the
whole sentence. P.S. Subrahmanyam was the first to draw attention to this particular
point. After stressing that the Dravidian participle does not denote absolute past time,
but that it denotes only an action done prior to another action denoted by the finite or
non-finite verb of the clause that follows, he notes the ungrammaticality of a sentence
like they came [pc] yesterday and will go [finite fut] tomorrow since the participle
refers to past time while the finite verb refers to future time with respect to the time of
the utterance of the sentence in contrast to a sentence like they will come [Pc] tomorrow and leave [finite fut] the day after which is grammatical since the participle also
refers to future time like the finite verb (2006:217). In other terms, the whole sentence
(Pc(s) and main predicate) should be situated in the same temporal space indifferently past, present or future considered from the time of utterance.
To summarize, the characteristics of the Pc converb in Badaga do not fundamentally

differ from what has been observed in other Dravidian languages.13 The Pc converb offers
very flexible means of chaining clauses. It allows a variety of readings. The events may
be sequential (1, 5, 10) or simultaneous (2, 3). Subject/topic continuity is preferred,
but change of subject is allowed (without the use of a same/different subject device). The
converb clause may be semantically restrictive, with a possible causal reading as in (7, 8,
11), see Section5 below for other readings. In the more tightly linked case, the converb
clause and its superordinate predicate refer to a single event and the converb falls under
the scope of its governing predicate for modality (4) and negation (see below ex. 15).
2.1.1.3 An alternative strategy: Independent clause chaining

Although converbs are extensively used for clause-chaining, in some cases the simple
juxtaposition of independent sentences can fulfill the same function.
With affirmative clauses, simple chaining of independent clauses, with a specific
prosodic contour, seems to be an open alternative option for expressing a sequence of
events which would be coordinated in other languages:
(13) hattine,
kudane,
it.t.ane,
mad.ila hid.utul.o,
Mdi!
climb.T1.1 pluck.T1.1 throw.T1.1 lap.obl catch.impPolite Madi
I will climb up, pick [some fruits] and throw [them]; catch [them]
in your lap, Madi!
(14) brivi,
kul.ivi,
sadda md.a bd.a!
come.imp2p sit.imp2p noise do.inf oblig neg Cp. Use of pc in (4)
Come, sit down [and] dont make any noise!
To express a series of negative injunctions, there seems to be no other option than to

chain independent clauses, as in (15) with the Obligative Negative forms (at.t.du ille,
kattarsodille, etc.):
(15) ta atteta
nir
at.t.mbadu sopa aki
nir
at.tdu
ille;
he so.emph water pour.oblig soap.obl put.pc water pour.oblig neg
man.d.e kattarsodille; kaluga met.t.u ikkodille;
gad.d.a
hair
cut.obl.neg foot.dat sandal put.obli.neg chin
bisodille,
atte make iddubud.du.
shave.oblig.neg so like be.pfv.oblig
He should bathe like that; he should not bathe with soap; he should not
cut his hair; he should not wear sandals and he should not shave his chin; he
should live like that.
13. However, at least one difference should be noted. There is no trace in Badaga of the use
mentioned for Tamil in Bisang (1995:157) as a complement to verbs of perception in alternation with infinitive clauses, such as I saw Kumar driving a car. In Badaga only a nominalized
form of the verb can occur in such cases, see ex. (41).
The sentence sopa aki nir at.t.du ille is of particular interest as it does not support
the sequential interpretation which could result from its literal translation having put
soap, he should not bathe. In this case, the use of the converb form (aki) forces the
interpretation that the two syntactic clauses describe the unitary event take bath with
soap under the scope of the negation he should not take bath with soap. The same
constraint occurs in (4) where a bi-clausal sentence appears as the translation of a
single English verb raise (your hood).
2.1.2 Disjunction
Disjunction is usually expressed by two juxtaposed clauses:
(16) ksa
kattara.v
ille.y
endu kan.d.id.i!
boy.oblA study.T1.3n.interr not.to.be.interr quot find.imp
See if the boy is studying or not!
~
h embadu ille.
(17) nag
ama h embadu ille
1
what.dat-any he yes say.N z not.to.be no say.N1z not.to.be
On any matter, he says neither yes nor no.
See below, Examples (28) and (29) for other instances of disjunction in dependent
clauses.
2.2 Specialized converbs

While the contextual converb Pc does not carry any explicit information about the
semantics of the relation between the dependent clause and the one to which it is attached,
there are other participial forms which perform the same syntactic function of integrating several clauses into one sentence, but with additional semantic information in the
domains of time (paragraph 2.2.1), polarity (paragraph 2.2.2) and modality (paragraph
2.2.3), all of them compatible with various aspectual specifications (paragraph 2.2.4).
2.2.1 Adverbial time clauses (Padv)

Badaga differs from the surrounding literary languages by having a regular subsystem of participial forms, based on three basic stems. Two affirmative stems give rise
respectively to the Non-Past adverbial participle (P1Adv), formed with the suffix
-ane attached to the stem 1 nd..uv-ane > nd.ne, ba.pp-ane, gi.v-ane, and the Past
adverbial participle (P2Adv), with the suffix -ade attached to the S2: nd..id-ade, ba.ndade, gi.d-ade. These forms denote relative time reference.
2.2.1.1 Simultaneity (P1adv)
hasututu.
(18) ama hit.t.u md.du.na nd.ne
ama.ga hot.t.e
1
1
he food do.N z.obl see.P adv he.dat stomach be hungry.T2.3n
When he saw the food being prepared, he felt hungry
2.2.1.2 Anteriority (P2adv)
(19) miccal
bandade

Naradaru avaka mra hi
but.t.ra.
lightning come.P2adv this Naradaru they three go.pc pfv.T1.3ph
After the lightning struck, these three people, Naradaru and them, left.
2.2.2 Negative adverbial clause: Negative Participle (Pneg)

The negative participle is morphologically akin to the other specialized participles, the
suffix -e is attached to the Negative stem: nd..ad-e, bar.ad-e, gi.yad-e. It fulfills the same
function of heading a dependent clause in which it negates the event without any temporal
specification. Depending on the context, it can express manner (20, 21) or cause (22, 23).
(20) d.de,
n
pd.a.va
ette kattare?
read.PNEG you this lesson.obl how learn.T1.2
Without reading, how will you learn the lesson?
(21) dayamad.i ni hit.t.u tinnade illiga
ba!
Please
you food eat.PNEG here.dat come.imper
Please come here without eating!
(22) ena egodu
endu ariyade
sapena iddu but.t.a.
what say.Oblig quot know.PNEG quietly be.pc pfv.T2.3
Not knowing what to say, she remained quiet.
It can also occur with a nominal predicate:

(23) me illade hanja.
rain be.PNEG poverty
Without rain, there is poverty
However, as shown in Table1, this participle form (PNEG) is also the negative equivalent
of the contextual converb (Pc) and as such it also covers functions from which the
other specialized converbs are excluded. Notably, it occurs in some of the most integrative uses of the Pc, such as its inclusion in periphrastic constructions (24), as well as
in combination with other specialized converbs, for example in (29, 70) below where
it occurs with the Conditional converb.
(24) ama barade
ibbad.enda
he come.PNEG be.N1z.abl.
Because of the fact that he did not come
[= Neg. Perfect]
2.2.3 Modality: The conditional (Hyp)

Morphologically, the conditional participle is formed with the suffix -le/-ale/-le
attached to the stem 2 of the verb. Historically, this suffix is derived from the conditional form le of the verb gu to become, which is also the origin for the adversative
particle le but. As in other languages, conditional sentences can express a wide array
of meanings (see Thompson, Longacre & Shin Ja Hwang 2007:255262 for a typological approach). Whatever the precise meaning of the construction, the Hyp marker can
be used to mark conditions in predictive (25), actual (26, 28), hypothetical (27), or
counterfactual (36) contexts as well as to express other meanings.
(25) karad.i bandale nangava koddu but.t.ara.
bear come.Hyp we.obl kill.pc pfv.T1.3n
If the bear comes, it will kill us.
(26) ama asaga
mad.ile gana.
he so.much do.Hyp enough
If he does that much, it is enough.
hogaku.
(27) ama bandale nanga hat.t.iga
he come we
village.dat go.Pot
If he comes, we may go to the village.
The same marker is used to express alternate possibilities. It should be noted here that
the clitic14 -, which usually cannot coordinate clauses, is grammatical in this conditional disjunctive use:
(28) idu ninaga sedaleyu
enaga sedaleyu ena?
this you.dat join.Hyp.foc me.dat join.Hyp.foc what
What does it matter if this thing belongs to you or if it belongs to me?
Negative conditions are expressed with the Negative participle (PNeg: -ade) followed
by the independent form of the conditional participle (ale).
hidiyade aleyu
(29) avakaga hid.ataleyu
na enna
.
they.dat like.pc.Hyp-foc like.Pneg Hyp-foc I me.obl
gelasava gidane.
work.obl do.T1.1
Whether they like it or not, I will do my work.
2.2.4 Note on aspectualized converbs

The basic converb forms described in the preceding sections, contextual or specialized,
can also carry various aspectual specifications. However, although aspectualized converbs are semantically specialized, they do not form a distinct category of specialized
14. The clitic -, and its equivalents in other Dravidian languages, is a very polysemous
marker. Emeneau (1980:199200) identified five distinct meanings: also, and (when repeated),
even, all, any (any element of the whole) which can be grouped under additive [add], totalizing and indefinite meanings [any], all with a shade of emphasis dominant in its focalizing
meaning [foc]: also and even.
converbs because the aspectual specifications they introduce are marked by auxiliaries
which themselves must be inflected in accordance with one of the preceding converb
categories: Pc, Padv, Pneg or Hyp.
Morphologically, the aspectualized forms of the converbs are built on the Pc form
of the lexical verb followed by an aspect auxiliary, such as the Progressive: V.Pc_ul..lu
(<be with), the Perfective: V.Pc_bud.u (<to leave) or the Perfect: V.Pc_iru (<to be)
for the three main aspectual distinctions. These auxiliaries can appear either in their Pc
converb form, respectively un.d.u > -n.d.u for the Progressive contextual converb (Pprog),
but.t.u > -t.t.u > t.u for the Perfective contextual converb (Ppfv) and iddu > iddu > -idu,
for the Perfect contextual converb (Ppf), or they can occur in a specialized converbial
form expressing simultaneity (ane), anteriority (ade) or condition (-ale).
Aspectual specifications can occur in several combinations with various shades
of meaning.
2.2.4.1 The progressive aspect (Prog)

The progressive aspect can appear as a simple Pc form (30) or combine with other
specifications, see below (35).
(30) ksu attun.d.u erigira.
this child cry.Pprog come.down.T1.3n
This child is coming down crying.
2.2.4.2 Perfective aspect (Pfv)

The Perfective aspect of the Pc form appears in (31) with a sequential reading, and in (32)
with a causal interpretation, in (33) it combines with an anteriority converbal form.
hgi
(31) hen.n.u kicc.a btut.u
bud.ali!
girl
fire.obl put.Ppfv go.pc pfv.optative
When the girl will have set up the fire, let her go!
(32) a
sile enaga id.utut.u
na enaga adu beku ende.
that sari me.dat appeal.pc.Ppvf I me.dat that want say.T2.1
That saree having appealed to me, I said I want it.
(33) ima bandu but.t.ade
ava ena mad.ida?
he come.pc pfv.P2adv she what do.T2.3
After he arrived, what did she do?
2.2.4.3 The perfect aspect (Pf)

In dependent clauses, the Perfect aspect frequently conveys durative meaning as
in(34), and freely combines with other specifications. For instance, Example (35)
shows the combination of a Progressive form (-un.d.-) with the Perfect aspect (-iru)
inflected as a simultaneous converbal form (P1adv: ibbane) to give durative meaning.
In Example(36), the combination of the Perfect aspect with the conditional converbal
form conveys irrealis meaning.
(34) modalu nanga Todand.u ayya
Tnriya n.d.iddu,
first
our
Todanad grand-father Tuneri.obl rule.Ppf
Formerly, our Todanad ancestor had been ruling Tuneri and
(35) aga had.udun.d.ibbane,
innu jamanoge iva ena mad.idiya?
then sleep.Prog.Pf.P1adv yet time.loc she what do.pc.Pf.T1.3f
Then, while [they] were sleeping, a little time later, what did she do?
(36) ni ninne
bandidale
amana nd.iraku.
you yesterday come.pc.Pf.Hyp him.obl see.pc.Pf.Pot
If you had come yesterday, you would have seen him.
Semantically, as with any other sentence level adjunct, adverbial or postpositional

phrases, the Padv specialized converbs specify or restrict the situation for which the
main predication takes its truth value. The temporal converbs express relative tense;
they specify situations preceding (P2adv) or simultaneous (P1adv) with the occurrence of the main predication. The negative converb (Pneg) simply indicates, as a Pc
form without further semantic specification, that the non-occurrence of some event
(/activity/state) is linked to the event described in the following predicate. The conditional converb (Hyp) restricts the situation in which the main predication can occur.
Syntactically, the specialized adverbial converbs differ from the contextual Pc converb in their ability to construct an autonomous actancial and spatio-temporal space.
They behave like any ordinary sentence adjunct phrase, and can often be replaced by
a simple temporal adverb: aga then, hinde after or a postpositional phrase, ex. hinde,
adu.ga hinde, huliya kan.e [after, that.dat after, tiger.obl cant see] After, after that, [we]
did not see the tiger [again], see also below paragraph 3.2. The position of the adverbial clause is fairly flexible. In some cases, a PAdv can even be inserted in the middle
of a clause; see the position of bendade hinde after it will be cooked between the object
my brother and the verb Ill call of the embedded quoted sentence (emba tit.t.a a
plan saying/planning that: ) in the following example:
(37) besi un.d.u,
enna tammana
bendade hinde
koracine
cook.caus.Pprog my y.-brother.obl cook.P2adv.after call.T1.1s
emba
tit.t.a ava akidave.
say.P1adj plan she set.Pft.3f
As she was cooking (the pieces of meat), she was planning that:
after it will be cooked, I will call my younger brother.
It should be noted here that the time referred to by the P2adv is posterior/future as
compared to the speech situation when she makes her plan, but, as expected for a P2,
anterior to the predicate on which it depends (koracine). This example corroborates
P.S. Subrahmanyams remark (paragraph 2.1.1.2) that, contrary to a Pc form which is
temporally bound to the verb/predicate on which it depends, the specialized converbs

can construct independent spatio-temporal situations (see notably their discourse
uses in (91)).
When a specialized converb clause occurs in the middle of a clause chain, it triggers a pause, or hiatus, like in Turkic languages.15 Typically, the specialized converb
closes the sequence of events already narrated and determines the situation relevant
to the event described in the main clause, yet to occur: grammatically, the preceding Pc forms are dependent on the Padv and the whole sequence remains opaque to
the tense/aspect/mood and agreement values of the final main predicate. The [P1adv]
ujjone in the complex sentence (99) is a clear example of this function.
Similarly, the participants of the clause headed by a specialized converb are, grammatically, fully independent from those of the main predicate.
In particular, there is no subject/topic continuity constraint and a different subject
is entirely permissible in a dependent clause headed by a specialized converb and the
following clause or main clause (see ex. (33), (35)), even when both verbs are transitive
(araki koricone vs. hen.n.u endara) as in:
(38) [arakii]
tannai
tammanai
koricone,
demoness her[refl/hon] younger.brother.obl call.P1adv
hennu ena endara
i
adu enna tamma
alla.
.. j
j
this girl
what say.T1.3n that my y.-brother is not.
ninna tammana
ni bet.t.i but.t.e endu.
your y.-brother.obl you cut.pc pfv.1s quot
When [the demoness] calls her [own] younger-brother, what this girl says
[is that]: that [=those cooked pieces] is not my younger-brother. [It is] your
younger-brother [whom] you have cut into pieces.
(Due to a magical trick done by the girl, the demoness has killed and cooked
her own brother instead of the girls brother she intended to enjoy eating with
her demon-brother)
However, as can be seen in other Examples ((21), (22), (25), (64), (75)), nothing prevents a specialized converb and a main predicate from sharing common arguments.
3. Nominalizing strategies
As noted at the beginning of this article, two strategies (B and C) can be used to form
an embedded clause, which can function as a nominal constituent: nominal derivation
(or substantivization, encoded Nz) and relativization with an adjectival participle as
15. Johanson (1995: 330331) used this term for Turkic languages where he noted that, in
pluripredicate sentences, some types of converbs were used to signal tight connections while
other types indicated looser connection, signaling a hiatus in the clause chain.
predicate. In both cases, the verbal forms are derived from the basic morphological
subsystem of Past, Non-Past and Negative stems as displayed in Table1.
3.1 The two ways of nominalizing clauses

Both strategies, relativization and nominalization, are based on the adjectival participle forms: the suffix -a is added to the three stems; for instance, on the root bar- to
come, the following adjectival participles are formed: bapp-a (P1adj), band-a (P2adj)
and barad-a (PNEGadj). On this basis, a 3rd person neuter pronominal derivative
-du is suffixed to produce the substantivized forms of the verbs:16 bapp.a-du (N1z),
band.a-du (N2z) and barad.a-du (Nzneg). Inside the nominalized clause, the encoding
of the arguments is not modified.
3.1.1 Substantivized verbs

Substantivization of the verb is a simple way of embedding clauses. The description
of real ((39), (41)) or virtual ((40), (42)) events is thus packaged into referential units
which can be used as core arguments ((39)(41)) or as adverbial clauses ((42), (43))
and can be chained to express a coordinative (39) or a disjunctive (40) sequence of
events. In the following example, the substantivized forms of the verbs are used to
recall the salient narrated events:
(39) i
varalaru besagu Naradaru egina:
this event
totally Naradaru tell.T1.3
Vis.n.u bandadu, ivaka hodadu, panja miruda
Vishnu come. N2z they go. N2z offerings
kot.t.adu, kusu ut.t.idadu,
avakerad.a
give.N2z child be.born.N2z they.both
itte
so

ottu odadu, i
arakigai,
Vis.n.u Lacmi
bring.pc go.N2z this demoness.for Vishnu Lakshmi

erad.a avadara
etti
bandadu.
both appearance take.pc come.N2z
Naradaru tells the whole of the story: The coming of Vishnu, their departure,
the gift of sacred offerings, the birth of the children, how both of them were
taken to the demoness and the fact that Vishnu and Lakshmi assumed new lives.
16. Similar nominalized forms of the verb can be made with 3rd pers. pronominal derivatives: -ma 3m bapp.a-ma the man who comes, barad.a-ma the man who does/did/will not
come; -va 3f bapp.a-va the woman who comes; -vaka 3pl band-a-vaka the people who
came. Derivatives in 3rd pers. neuter -du, such as bapp-a-du can also mean the thing/animal
which comes. Here we use the term substantivized to distinguish the cases where it is the
verbal notion itself which is turned into a nominal form: bapp-a-du the fact of coming, the
coming, barad.a-du the fact of not coming, band.a-du the fact of having come.
In (40), the narrator expresses the idea that the traditional Hindu way of honoring
the gods is not followed by the Badagas:
h
(40) nanga tenge
od.apad-o,
bere pujiya md.d-o,
1
1
we
coconut break.N z-dub other ritual do.N z-dub. flower
akod-o
ondu.n-u ille.
put.N1z-dub one-any be not
We do not offer coconuts, do not perform rituals or offer flowers.
Lit. Breaking coconut, doing puja, putting flowers, anything [like these]
does not exist
Note that the expression of a sequence of negative embedded clauses is not straightforward. Here the affirmative substantivized forms of the verbs are apposed to ondu.n-u
(one-any), the subject of the negative existence verb (ille). They are chained with the
use of the clitic -o (whose basic meaning is to express doubt -dub), repeated to carry
the alternative meaning or. Substantivized clauses can be used as a complement (41)
or an adverbial clause, marked for case ((42), (43)), here with the dat = for:
hvaduna
(41) Jogi ni a
kusa
nod.ida.
Jogi you that boy.obl beat. N1z.obl see.T2.3s
Jogi saw you beating the boy.
ha
(42) enaga a
musoduga ase.
me.dat that flower.obl smell.N1z.dat pleasure
I like smelling these flowers.
(43) enga Ramana pad.aga

bappaduga
sari
mad.ideyo.
we Rama.obl movie.dat come.N1z.dat alright do.T2.1pexc
We persuaded Ram to come to the movie.
3.1.2 Relativization
The use of an adjectival participle (PAdj) enables the construction of clauses which
are roughly equivalent to relative clauses, insofar as they are typically used to modify a
nominal head. However, two specific properties of the PAdj should be noted to understand the use of this strategy to construct subordinate adverbial clauses.
First, as with the small set of true adjectives (dod.d.a big, kunna small, ol.l.eya
good, etc.) in Badaga, the adjectival participle is strictly confined to the function of
modifier inside a phrase whose head is (a) exclusively a noun in the case of true adjectives and (b) never a verb. To fill sentence constituent positions, such as subject/topic
or main predicate functions, these adjectives need to be pronominalized, with -du
(ex. dod.d.a-du the big one) as seen in the preceding section for the derivation of an
adjectival participle into a substantivized verb form (cf. paragraph 3.1.1) and other
pronominal derivatives ( ma, va, vaka, etc., cf. note 16). This property distinguishes
true adjectives from other quality terms, such as color terms (keppu red, kappu black
etc.) which behave like abstract nouns.17
Second, while true adjectives can only modify a noun head, the participle adjectives accept a wider range of lexical heads. Notably, they can be syntactically dependent on a subset of lexemes which could broadly be called adverbs or adverbials. They
are polyfunctional lexical units18 which (a) cannot fill the core argument positions
of a verb, i.e. are not nouns, (b) can be the main predicate of a sentence, but do not
have verbal morphology, i.e. are not verbs. The properties of the lexemes included here
under the cover term adverbs are not homogenous and cannot be detailed here. We
will consider only a few of them which are relevant for the construction of adverbial
clauses: a few spatio-temporal and manner adverbs.
The following sentence illustrates the two types of relative clauses, an ordinary
one where the Padj is attached to a noun (porul. thing) and another where the Padj is
attached to the adverbial make as, like:
(44) na tappa
porul.a
hottogi,
na egida
make md.ire-ya?
I give.P1adj thing.obl take.go.pc I say.P2adj as
doT1.2s-interr
Will you take the things I shall give [you] and will you do as I told you?
The ability of these adverbs to head a relative clause must be paralleled with their use
as postpositions. In both cases, a variety of constructions can be observed.
The postposition can be preceded by a bare noun, an oblique case noun or a dative
case noun:

maduve hinde after the mariage

[marriage behind/after]
suga.da mle after happiness
[happiness.obl on/after]
samudara kere.gu mle dod.d.adu
sea
lake.dat on big.3n
The sea is larger than a lake
The same type of lexical item can also head an adverbial temporal clause whose predicate is a Padj, see for instance, ex. (100): appanella satta mle After all my parents have
17. Cp. dod.d.a mun.d.u keppu [big shawl red] the big shawl is red vs. keppu mun.d.u dod.d.adu
the red shawl is big; i keppu this red one vs. i dod.d.adu this big one) and idu1
apat.t.i2 keppu3 ille4 this one1 is not4 so2 red3 vs. idu1 apat.t.i2 dod.d.adu3 ille4 this one1 is not4
so2 big3.
18. The lexeme mle, basically a term of the spatial domain, roughly meaning up, above can
be taken to briefly illustrate the polyfunctionality of these items, such as verb modifier: mle
iru ! Stay upstairs! , noun modifier: mle kade the above side, sentence localizing adverb: aga1,
mele2, obba3 enda4 then1, upstairs2, someone3 said4 , which can also be case marked in a
localizing adverbial phrase: melanda from above (cf. 46) or postposition: pustaga1 pet.t.i2 mele3
hadade4 The book1 is4 on3 the table2 (see also 58).
.
died, (cp. with the independent sentence: appanella sattaru All my parents died).
When these adverbs head a clause, the variety of their constructions differs slightly
from nominal ones. The verb form of the dependent clause can be:
i.

most frequently, as already mentioned, an adjectival participle Padj, cf. ex. (53):
bgu1 da2 mle After the dawn1 has come2 [become.P2adj]
ii. but also sometimes a substantivized form of the verb, Nz, with a dative suffix,
cf.ex. (54):

kattale1 apadu.gu2 mundad.u3 Before3 darkness1 has come2 [become.N1z.dat]
iii. or an adverbial participle, Padv:

gan.d.anuravu1 apane2 munde3 Before3 I became2 a husband1 [become.P1adv]

tiricide mle ima kta After he had finished, he asked [finish.P2adv]
Among all these constructions, the use of a postposition with an oblique nominal and
with a Padj clause are closely related as both these strategies are widely used in this
language to carry out modification function.
The use of an adverb/postposition with a nominalized form of the verb is closely
related to the use of case markers on nominalized adverbial clauses, described below
in paragraph 3.2).
Type (i) and (iii) constructions are also frequently used with manner adverbs/
postpositions, see above ex. (44) and below ex. (92): egidenge as I said [tell.P2adj.like],
ex. (91): egonenge [tell.P1adv.like], as well as the use of a nominalized form with a case
suffix ex. (90): aradodunoge [know.N2z.loc].
Finally, it should be noted that when two or more predicates fall under the scope
of the adverb/postposition, the ones preceding the last one take the Pc converb form.
See the use of hoi [go.Pc] inside the relative clause, whose regular predicate is an adjectival participle (seda) and the head an adverb (indad.u behind, after):
hindad u, []
(45) itte avaka Kattale Bikkega hoi
seda
.
so 3ph Kattale Bikke.dat go.pc attain.P2adj after
erad.a alli kul.udu ondu pujiya mad.iyara.

two there sit.pc a
puja do.T1:3ph
So, after they had left for Kattale Bikke and reached it, [], both of them sat
there and performed a puja (religious ceremony).
The adverb/postposition heading an adverbial clause therefore has the same closing
effect as the Padv form, making the predicate included in it opaque to the spread of
the tense/aspect/ mood features of the finite predicate on which it depends. As in the
case of the Padv constructions, the arguments may also be autonomous (ex.44) or
shared (ex. 45).
3.2 The use of case and postpositions in dependent clauses

Once a clause has been turned into a nominalized constituent, it becomes possible to
mark it with the same cases and postpositions as any other noun phrase. The forms
of the cases/postpositions are not changed in such clausal functions, but the range of
meanings seems to be slightly reduced.
Time localization is the most frequent use of the Locative marker: -/-ge.
hennu ml.anda
(46) hal..laga bappa
jma.nge,
..
1
river.dat come.P adj time.loc this girl upon.abl
bandu but.t.ara.
come.pc pfv.T1.3n
By the time they came to the river, the girl had arrived from up there.
(47) miccalu banda
od.atane.oge ima had.udun.n.ana.
2
lightning come.P adj moment.loc he lie down.pc.prog.T1.3m
As soon as the lightning struck, he laid down [and died].
The Ablative marker (-enda/-nda) is most often used to express a cause:

(48) appara od.idudenda enaga kalu ja
marici but.t.a.
much run.N2z.abl me.dat leg cramp got.pc pfv.T2.3
Because I have run too much,/Having run too much, I got a cramp in my leg.
(49) ninga kepa
muru endavu ningaga savu
you ask.P1adj three abl.any you.dat death
bappaduga aga.
come.N1z.dat become.neg [=PotNEG]
From the three [causes] you mention, death cannot come to you.
But Ablative markers can also convey other meanings, such as time reference (50):
(50) ninga banda
jena enda, ondu makkavu ille
you come.P2adj jour abl one child=any is not
Since the day you came, no child has been [born to us]
The Dative marker (-ga) is generally used to express a goal:

hatti.ge
(51)
mvan.n.u suvaduga dr.
ille.
..
this village.loc mango buy.N1z.dat who.any not.to.be
In this village, there is nobody to buy mangos.
But it can also appear with other meanings. In the following example, the verb anju to
be afraid of regularly case-marks the source of a fright with the Dative, irrespective of
its clausal (52a) or phrasal status (52b):
(52) a.

na bappaduga
anjide.
I come.N1z.dat fear.T2.1s
I was afraid of coming.
b. Kalla huliga
anji
moraga hat.t.ida.

Kalla tiger.dat fear.pc tree.dat climb.T2.3

Kalla climbed up the tree fearing the tiger.
Various postpositions can also mark the dependent clause, expressing for instance a
spatial or temporal localisation, mle upon/after or mundad.u in front of/before:
(53) bgu da
mle oraguvad-?
dawn become.P2adj after sleep.Obl-dub
Should one sleep after dawn?
(Proverb)
(54) ama kattale apadugu

mundad.u banda.
he darkness become.N1z.dat before
come.T2.3
He came before it was dark.
4. Quotative strategy
The quotative markers derived from the verb ennu to say offer another powerful
device to construct complex sentences. They appear with various morphological specifications: converb (endu: quot(.Pc), ex. (55), (56)), conditional (endale: quot.hyp,
ex. (56), (58)), adjectival (emba: quot.P1Adj, ex. (57)), nominal (embadu: quot.N1z,
ex. (59)), fulfilling the same functions as seen in the preceding sections. The quotative
markers typically occur with verbs like egu to say, nenacu to think, ke to ask, ari to
know, but not with the verb ennu itself (cf. ex. (32)).
(55) nn
sattane endu gina.
me-add die.T1.1 quot say.T1.3m
He says I shall die also.
(56) savudu
endale,
ette savudu
endu egina.
die.Oblig quot.hyp how die.Oblig quot say.T1.3m
He says if we must die, how should we die?
hidana
(57) illigata
emba
pointunoge illiga
bandiya.
here.dat.emp go.pf1.3m quot.Padj point.loc here.dat come.T1.3f
She comes to the place (point) where he has come here precisely.
As shown above, embedding with the quotative is a substitute for the conjunctive use
of adverbs such as how (ette endu in 56), why (eka endu) or where (with a
locative phrase in (57): emba N.loc).
The conditional form of the quotative (endale) is frequently used to report the
content of a saying, thought, etc. In this case, the regular order [clause quot V] can
be reversed [V quot clause].
(58) tanna appa ena nencida

endale:
idu enna
mele
she.obl father what think.T2.3 quot.hyp this me.obl on
gava bipile.
love keep.RNeg
What her father thought is: this (girl) has no affection for me.
In many occurrences, the literal meaning of the quotative say > think is retained, with
a frequent shade of intentionality think so that.
(59) embattaru raja makka, ondu rajana kot.ega
raja
86
king sons
one king.obl palace.dat king
ille
embaduga
ittitte tandru hududare.
1
not.to.be quot.N z.dat so
drum beat.pf.3ph
The 86 princes have beaten the drums to announce that one of the royal palaces
has no king.
(60) Nellitore ibba ava
kobbadu
endu itte devaru
Nellitore be.P1adj people.obl kill1.Oblig quot so god
bandudane endu ava ondu.nu ariya.
come:pf.3m quot she one.any know0:neg.3
[The demoness] thought that the god had come with the intention of killing the
people who were living in Nellitore, [but actually] she did not know anything.
5. Alternative ways of expressing similar semantic relations

In the preceding sections, we identified the different strategies used in Badaga to build
complex sentences. In this section, we will consider their different semantic relations.
5.1 Temporal relations

As seen at the end of paragraph 2.1.1.2 the Pc converb can implicitly express various
temporal relations: from a simple iconic sequence of events to readings requiring strict
anteriority (5), sometimes emphasized with Perfective (31) or Perfect (36) aspect, as well
as simultaneity (2), (3), (4) which can also be stressed by Progressive aspect (30), (35).
The specialized converbs basically express relative time, simultaneity, P1adv (18)
and anteriority, P2adv (12), (19), and the same relation can be obtained with nominalized expressions involving the concomitance of both events (at the time of (46), as
soon as (47)) or their sequentiality (after (53)). However, the nominal expressions,
marked with case or a postposition, differ from the converbs because, in addition to
indicating the relative time of occurrence of one event in relation to another, they also
convey a more precise time setting for these events according to relative (from Vi) or
absolute external references before darkness (54) after dawn (53).
5.1.1 Time/condition
As has been noted for other languages (Thompson et al. 2007:257), expressions of time
and condition are semantically very close. Thus in some cases a conditional marker
appears where temporal reading is required, as in the course of a narration (61):
hanna
(61) a
nod.ile appara aval ara.ta.
..
that fruit.obl see.Hyp much craving become.T1.3n.Emph
When [she] saw that fruit, [she had] much craving [for them].
In other instances, it seems that the difference in the degree of expectability (ibid. p.
258) between when and if clauses is neutralized by the use of the Perfective aspect
in an if clause.
(62) had.i.ya taradut.t.ale,
ima sattudane.
door.obl open.pc.pfv.Hyp he die:Pf.3m
When [they] opened the door, he was dead.
In many cases, both readings are acceptable:

(63) ama ane
kan.d.ale anjina.
he elephant see.Hyp be afraid.T1.3m
If/when he sees elephant, he is afraid.
5.2 Causal relations

As with temporal relations, causal reading may be fully contextual, notably in bi-partite
sentences where the initial Pc clause appears to restrict the conditions of occurrence of
the second clause (8), (9), which can itself contain more than one clause (7) or be a nominal predication (11). A causal reading is also possible with the P1adv specialized time
converbs (18). However the most explicit way of expressing a cause is to use a nominalized construction with the Ablative suffix (48). A negative cause can be expressed
with either of these devices, Pneg alone (23) or integrated into a nominalized expression with an ablative suffix (24).
5.2.1 Cause/time
In many cases, both temporal and causal readings are possible:
(64) ava itte ibbane
appara siggada.
she thus be.P1adv much shy.become.T2.3
While/Because sitting like this she was feeling embarrassed.
(65) nn
bandadenda mane bsalgi
tridega
you-foc come.N2z.abl house bright.adv seem
From your coming/Because you have come, the house seems bright
5.2.2 Cause expressed by the quotative

An alternative expression of cause is to use the quotative, usually in its Pc form endu,
but sometimes in more complex constructions. Thus a sentence like Desiring the car,
the boy got angry, which clearly has the causal meaning Because he desires the car,
has been translated with two distinct constructions:19 Pc endu in (66a) and a substantivized form of the quotative embadu, with a postposition -gi meaning for in (66b).
(66) a.

a
kusu karu beku endu
kopa pat.t.itu.
that child car want quot.pc anger feel.T2.3n
Lit. The boy felt angry because/thinking he wanted the car.
b.

a
kusuga
karu beku embadugai
kopa bandutu.
that child.dat car want quot.VN1z.for anger come.T2.3n
Lit. The boy got angry for/because of the fact that he wanted a car
Desiring the car, the boy got angry.
5.2.3 Cause/purpose
The use of a dative marker (67), or its equivalent, as in the preceding Example (66b),
to express a cause or a reason may seem strange at first, as it is also regularly used to
express purpose (see below 83), but it turns out to be a fairly common typological
feature, due to the fact that both provide explanations, or accounts, for the occurrence
of a given state or action (Thompson et al. 2007: 250). However, in Badaga, there is no
trace of any different marking to signal the unrealized/irrealis status of the purpose
clause versus the realized/realis status of the reason clause (ibid. p.251). In both cases,
the nominalized verb is formed on the stem 1 (Non-Past).
(67) na asaga a
mundadun ibbaduga siggu.
I so much people in front.obl be.N1z.dat shy
I feel embarrassed to stand before so many people.
(cp. with 64)
Finally, we should mention that the simple juxtaposition of two independent sentences,
here in reverse order, may also express a cause:
(68) avve, satakana ba,
amme
attiya!
mother quickly come.imp young sister cry.T1.3f
Mother come quickly, Sister is crying!
19. Independently from the fact that the impersonal verb beku be wanted, requires the
experiencer to be case marked as dative. In the first sentence (66a), kusu, the subject agreeing
with the verb (pat.t.itu), is omitted in the embedded clause karu beku, while in (66b), the
experiencer kusu.ga, marked in the dative and placed in initial position, is shared by both
clauses, but the second clause has its own subject kopa bandutu: anger came to the boy
Temporal, cause and condition clauses are semantically close in the sense that they all
restrict the world setup in which the event of the main clause occurs.
5.2.4 Concessive clauses

The circumstances which condition the event denoted in the main clause can be
emphasized in a concessive clause. To mark a concessive clause explicitly, the clitic -
(polysemous, see note 14) is added to various adverbial clauses. In this case, it takes the
meaning of even, though. In general, the clitic - is added to conditional clauses, but
it can also occur on Pc converb clauses (71).
(69) ama sangat.a le.y-,
nad.abaduga mud.adara.
he illness become.Hyp-foc, walk.N1z.dat Pot
Though he is ill, he can walk.
(70) avaka barade aleyu
ninga bari.
they come.Pneg Hyp-foc you.pl come.imp2p
Even if they dont come, you, come!
(71) Bel..la ondu gelacaga ariciy,
gelca sikkule.
Bella one work.dat search.pc-foc work get.RNeg
Though Bella searched for a job, he did not get one.
In some cases, a simple contextual Pc converb may require a concessive reading:

(72) hettu
barad.i
kattu
pot.t.i.
give birth.pc childless lady learn.pc ignorant
[Though she has] given birth, [she is like] a childless lady,
[though she has] learnt, [she is] an ignorant.
(Proverb 531)
A restrictive meaning only if, unless can be obtained by adding the emphatic particule
-t to the conditional converb:
(73) kareman.iya
tandale-ta [] n Kariyabet.t.ana
bead.necklace.obl give.Hyp-emph I Kariabetta.obl
kan.d.u bannane.
see.pc come.T1.1
Only if you give me this necklace, shall I bring [the news] to Kariabetta.=
Unless you give me this necklace, I wont bring the news to Kariabetta
5.3 Manner clauses

Manner clauses, which specify how an action is carried out, are usually expressed by
a contextual Pc converb closely connected to the predicate it modifies. Negative converbs (20), (21) and Progressive Pc are also common (30). Manner clauses occur frequently with movement predicates (74), but not necessarily (75).
(74) rn.i nad.adu bandiya.

queen walk.pc come.T1.3f
The queen comes walking.
(75) kusu bukka odin.d.
origi
but.t.ara.
boy book.obl read.Pprog- sleep.pc pfv.T1.3n
The boy will fall asleep reading the book.
A nominalized expression may sometimes fulfill a similar function:

(76)
ksu madilun kul.un.d.ibbadun
se.
that child lap.loc sit.Prog.be.N1z.loc happiness
Sitting in her lap the child was feeling happy.
5.3.1 Manner/comparison
A more explicit way of expressing manner, alongside comparison, is to use adverbials
with the meaning like, as in a variety of constructions. These adverbials can be suffixed to an adjectival or an adverbial (79) participle as well as to a substantivized form
(82) of the verb. Two elements: mke as in (44) and enge, are used for simple manner
or comparison clauses.
hadade.
(77) enaga negabange ai
.
1
me.dat laugh.P adj-like become.pc be.T1.3n
I feel like laughing.
(78) Kunnurama and[u]

iddange-ta
idda.
Coonoor.he those days be.P2adj-like-emph be.T2.3
The man from Coonoor is as he was in those days.
(79) i
bukku oduvanenge
ille.
this book read.P1adv-like is not
This book is not (as) readable.
Sentences ending in -enge are common in proverbs:

(80) ane
bga
kabb[u] it.t.nge.
[Proverb meaning: lost forever]
elephant mouth.dat sugarcane throw.P2adj-like
It is like having thrown sugarcane into an elephants mouth.
It should be noted however that a distinct adverbial vid.a in comparison with, instead
of, (rather) than is used for evaluative comparison:
(81)

Nlagiriyo ibba
ellavavid.a Bad.ugaru janatogeta
jasti.
Nilgiri.loc be.P1adj all.obl.than Badagas population.loc.emph much
The Badagas, among the [local] populations, are more than all the
[others] living in the Nilgiri.
(82)

ni i
kas.t.a
pat.t.u kaiyoge murupadavid.a na katti tanne.
you this difficulty feel.pc hand.loc cut.N1z.obl.than I knife give.T1.1
Instead of suffering and/by cutting [the firewood] with your hands,
I will give you a knife.
5.4 Purpose clauses

Purpose clauses are taken here in their broad typological meaning: a relation []
such that one of the linked events (the one coded by the main clause, or the main
event) is performed with the goal of obtaining the realization of another one (the one
coded by the purpose clause, or the dependent event) (Cristofaro 2008, ch. 125). The
most common device to express a purpose clause is to use a substantivized form of the
verb suffixed with a dative case marker, see Examples (43), (51), (59) and below.
(83) appa sattu hda
tapptu
jli-da
//
father die.pc go.T2.3 wrong become.T2.3n matter-addr
kptduga
ninaga obbat
idde-da.
protect.N1z.dat you.dat alone.emph be.T2.1-addr
[Your] father is dead. Things are going bad. I am alone to protect you.
(84) Jogi bukka ododuga
nenacana.
Jogi book.obl read.N1z.dat think.T1.3m
Jogi wants to read a book.
An alternate way to express purpose or goal is to use the Pc form of the quotative endu.
This strategy expresses modal specifications and often conveys a shade of intentionality and decision.
(85)

hpadu
Jogi tanna manega
endu nenacana.
1
Jogi his
house.dat go .Oblig quot think.T1.3m
Jogi wants to go home.
Lit. Jogi thinks he should go home which was given as a translation of the
above English sentence.
(86) enga Rama pad.aga

ba
endu sari
mad.ineyo.
we Rama cinema.dat come.imp quot alright do.T1.1pex
We persuaded Ram to come to the movie.
[Cp. 43]
A less common way of expressing purpose is to use the adverbial enge. This type of
adverbial clause is semantically close to indirect causative clauses.20
(87) Ranga gelacakarana danava edonenge egida.
Ranga servant.obl cattle.obl chase.P1adv-enge say.T2.3
Ranga told the servant to chase the oxen out.
20. Note that this is a marginal use. There are other devices, morphological as well as periphrastic, to express direct and indirect causativity.
(88) ama avva

sangaga
sebanenge
mad.ida.
he she.obl association.dat join.P1adv-enge do.T2.3
He made her join the association.
In some cases, a simple Pc contextual converb may carry a purpose meaning:

(89) irudoge iga mora tori
bandidde.
night.loc now face show.pc come.pc.Pf.T2.2
Now, in the night, you have come to show your face.
[about the moon]
5.5 Discourse functions

In this section, we will present two distinct functions of adverbial clauses at discourse
level. The clause morphological devices and syntactic properties are the same as those
presented hitherto, but the semantic functions of these adverbial clauses are different.
5.5.1 Discourse comment adverbial clauses

This first type of adverbial clause differs from the preceding uses in that it does not
modify or semantically depend on the main predicate, but pertains to the whole sentence. It introduces a comment at the discourse level.
(90)

n araddunge, modalu bdiy ogutun.d.u

iddaru.
I know.N1z.loc before ash.loc wash.pc.Pprog be.T2.3ph
To my knowledge [/as far as I know], before they used to wash the
clothes with ashes.
(91) nanga egonenge

palav.uru kat.t.i
nanga bad.ugaru iddun.d.iddare.
we tell.P1adv-enge many village build.pc our Badagas be.Prog.be.T1.3p
As we were saying, our Badagas have built many villages and live there.
(92) na mundadu egidenge
I before
tell.P2adj.enge
As I said before
~
(93) h, na obbata
baki
idde,
make torira.
yes, I alone.emph remain.pc be.T2.1, like show.T1.3n
Yes, I am the only one to have survived, as it seems!
See also in (101) below the use of atte embadiddale if we think so/imagine that
Among these discourse functions, let us also mention some specific uses of the Pc
form bandu from the verb to come. In the course of a narration, bandu can be used,
not to refer to a real movement, but to trace back to some participant or object the
X in question or to return to the main story after a digression (Bandu Coming
back to the story).
(94)
ondu pt.o bandu
[] maisuruga banda.
that one photo come.pc
Mysore.dat come.T2.3
One of these photos in question came to Mysore.
5.5.2 Discourse linkers

The function of adverbial clauses as discourse linkers, to ensure the cohesion between
sentences or paragraphs, is a widely used feature in languages with oral traditions. The
Tail-Head linking strategy (95), (96) is particularly common.
(95) ondu jena me hda.
me hvane,
gubisiya
mane
One day rain fall.T2.3 rain fall.P1adv sparrow.obl house
id.udu
but.t.a.
destroy.pc pfv.T2.3
One day it rained. During the rain, the sparrows house got destroyed.
(96) aga hadiya
tarada
.kake hadiya
taradu,
then door.obl open.T2.3 crow door.obl open.pc
gubisiya
[] enda
sparrow.obl [] say.T2.3
Then it opened the door. Having opened the door, the crow said
to the sparrow [].
(97) avaka aricin.d.u
bandaru. tanna ammeya nod.odugagi,
they search.Pprog come.T2.3 their sister.obl look for.N1z.for
sole dari
bappane,
ondu koi egira
forest through come.P1adv one cock tell.T1.3n
They started searching [for their sister]. As they went through the forest in
order to find their sister, one cock spoke.
At the end of this survey of the main semantic functions of converbs and adverbial
clauses, it appears quite clearly that none of the morphosyntactic devices presented
in the preceding Sections 2 to 4 is specific to a single function. The Pc converb
virtually appears in all the functions enumerated under Section 5 and, as suggested by its name, its interpretation depends essentially on the syntactic, prosodic
and semantic contexts. At the other end of the spectrum, the temporally specialized
converbs appear to be more semantically constrained. The functions of the quotative
are also quite varied, but in Badaga, despite its grammaticalization, the quotative
forms of the verb ennu frequently retain semantic values linked to human thought
processes. The semantic functions of the nominalized clauses are sometimes more
straightforwardly interpretable depending on their components, but their structural position remains crucial (ex. V.Nz.dat: adverbial clause vs. component of a
periphrastic modal form). One of the findings of this typological survey of clauselinking in Badaga are the similarities between cause and purpose constructions,
as evidenced by the parallel alternative of the morphosyntactic devices used in
both functions: V.N1z.dat and clause+endu as in (66ab) for cause function and
(84)(85) for purpose function.
6. Instances of multi-clause complex sentences

To balance the analytic approach adopted in the preceding sections, we will now present a few excerpts from narratives to show how the various devices are used and combine to construct complex sentences.
As said at the beginning, converbs are frequently used to express simple sequential clause chains. There is no specific limit to the number of converbs which can occur
in a chain. However, even with the use of simple Pcs, the structure of the sentence
is not flat. In the following sentence (98), verb order is iconic with the sequence of
events (go break carry give), but from a structural point of view, three distinct
sequences can be identified. Between the first (a) Korangu hogi and the last (c) kel.aviga
kot.t.a, the middle sequence is made up of two successive events murutu break and
bring (expressed by a lexicalized compound verb hottu ba carry. Pc come) which are
closely linked both thematically (savude) and prosodically. Also of note is that though
it appears trans-clausally, the basic SOV word order is respected: Korangu (S) savude
(O) kel.aviga (Dest.) kot.t.a (V).
(98) korangu hgi, jsi savude ella murutu hottu
bandu
monkey go-pc small firewood all break.pc carry.pc come-pc
kel.aviga
kot.t.a.
old woman.dat give-T2.3s
The monkey went away, cut some small firewood, brought it back and gave it
to the old lady.
Contrary to the Pc converb whose main function is to simply tie events together, the
specialized converbs set up a delimited clausal domain (with, if required, their own
arguments). In the following sentence (99), the sequence of events get up, wash their
face, brush their teeth is interrupted by the specialized simultaneous suffix -ane on
the converb ujjone which, in this case, delimits the only domain relevant for the main
clause their teeth were chattering.
(99) orakadu, [ella eddu,
moga togi,
< hallu ujjone],
morning all get up.pc face wash.pc tooth brush.P1adv
ivaka hall ella ki ki.. endara
>
their tooth all ki ki sound.T1.3n
In the morning, they all got up, washed their faces and when they were
brushing their teeth, all their teeth were chattering [lit. sounding ki..ki..].
In the next instance (100), the sentence is structured by nominalizing devices. Three
independent events occurred: [the parents died], [the girl took a decision] and [Siva
came], each one has its own arguments and spatio-temporal setting with a sequential
linkage between the first two clauses and a simultaneous linkage between the last
two clauses.
(100) [appanella satta

mele] [{nanu buddu sattu bud.du} emba
Father all die.P2adj upon I.too fall.pc die.pc pfv.Oblig quot.PAdj

kit.t.aga akagod.a ],
[sivaperuman bandu but.t.a].
decision.dat put.dur.moment Siva
come.pc pfv.T2.3
After all my parents have died, at the moment I took the decision
I too should fall [into the fire] and die, Siva arrived.
In this last extract (101), several devices are used conjointly to construct a lengthy
sentence made up of ten clauses. Two sets of participants, one man and the villagers, are involved in an imaginary situation, introduced by a conditional converb (etti
un.d.le), suffixed by the clitic which carries indeterminacy meaning. The concrete
situation is the following: a big field needs to be cultivated, it is presented using a
predicate in real declarative mood (gsina) embedded in a conditional form of the
quotative (embadiddale) which articulates the text around the problem (cultivate a
big field) and its solution (call the villagers for help).
(101) adu avaka ottumeya torcira.
vel..lameya
etti
un.d.ley
that their solidarity show.T1.3n agriculture.obl take.pc Prog.Hyp.any
avaka ondu hola gvane
obba
ondu dod.d.a hola gsina
1
they one field work.P adv someone one big
field work.T1.3m
hat.t.iyge ondu hatta
iddre
atte embadiddale
village.loc one ten.person be.T1.3ph so quot.N1z.become.Hyp
hattava

koraci avaka amana ton.ega gelaca gsi
that ten person.obl call.pc they he.obl help work work.pc
avakaga han.a ley ille

bede ley avaka md.ida gelacaga
they.dat money some be not grain some they do.P2adj work.dat
bede ley kot.t.u avakella santsavgi iddu bandaru.
grain some give.pc they.all happy
be.pc come.T2.3ph
That will show their solidarity. In case they have to undertake [etti un.d.ley]
some agricultural (work), when they have to cultivate [gvane] a field, if it is the
case (/let us imagine) [embadiddale] that someone should cultivate [gsina] a
big field and that there are [iddre] some ten people in the village, he will call
[koraci] those ten people and they will do [gsi] the work for his help (= they
will help him to do the work); for them, there wont be [ille] any money, [but]
some grain ; to those who have done [md.ida] the work, he will give [kot.t.u]
some grain and they will go back happy [iddu bandaru].
7. Conclusion
This study has aimed at exploring the strategies used in Badaga to integrate clauses
expressing adverbial meanings (temporal, causal, conditional, etc.) into complex
sentences. In this concluding section, a few points should be stressed concerning the
semantic, syntactic and typological properties of the constructions studied.
From the semantic point of view, a clear result of this study is that several devices
can convey similar semantic relationships between dependent and governing clauses,
as discussed in paragraph 5. We will now focus on the relations between the morphosyntactic devices available and their semantic uses in Badaga.
To start, it may be useful to distinguish between overt and implied strategies
marking adverbial relations. The strategies devoted to adverbial clauses are specialized converbs (paragraph 2.2) and various nominalizing strategies (paragraph
3) which can be accommodated to express adverbial relations. The latter strategies
are based on an adjectival predicate and have an explicit marker: a case marker or a
postposition/adverbial lexeme specifying the semantic relation. The adjectival clause
can be substantivized (with a derivative pronominal suffix added to the adjectival participle) or may depend on a nominal head (typically a noun of the spatial, temporal:
jma time or manner domains) or may be directly dependent on the postposition/
adverbial marker which express the semantic relation to the main clause. The specialized converb constructions (Padv) and the nominalizing strategies (developed
on a Padj) can be grouped together as proper instances of adverbial clauses. On the
other hand, one finds complex sentences and dependent clauses making use of contextual converbs (Pc) which have some implied adverbial meaning (cause, concession, purpose), since there is no overt marking of the semantic relation. As stated
in paragraph 2.1.1, the contextual converb is also, even more frequently, used for
simple clause-chaining and can thus be considered as expressing basically nothing
more than the syntactic dependency of the verb form. In other words, even though
the contextual converb expresses syntactic dependency, it is not inherently a marker of
adverbial subordination. At this point, it may be fruitful to re-assess the distribution
of the Badaga converbs in Nedjalkovs tripartite semantic frame given in paragraph 2.
The two types of converbs found in Badaga, specialized converbs (Padv) and the Pc
form, would be positioned at the two extremes of the typological scale: the specialized
converbs having one definite meaning, with possible extension to adjacent semantic
domains, as shown in paragraph 5, while the Pc form would have a basic function of
chaining clauses, with a possible extension towards contextual semantic meanings.
Specialized converbs
Contextual converbs
Padv
Figure1. Typological distribution of Badaga converbs
Narrative converbs
Pc
Structural, pragmatic and prosodic factors play an important role in the interpretation of adverbial clauses and in the contextual extension of their basic functions. In
many respects, the Badaga data fit in with a number of typological tendencies.
Thus, as noted in Thompson et al. (2007), (i) the before clauses behave differently from the when and after clauses. In Badaga, only the nominalizing strategy can
be used in this case. (ii) There is some affinity between purpose clauses and futurity
(unrealized p.250, future tense p.253). In Badaga (Example 4950), the construction
NVz.ga is used for purpose clauses as well as for the Potential mood.
From a syntactic point of view, Nedjalkovs typological definition of converbs
should be slightly broadened: A converb is semantically related to another verb form
(either finite or another non-finite form) (Nedjalkov 1995: 445). The Badaga data
clearly show that a converb can be syntactically and semantically dependent on any
other clause head. This clause head may be the main predicate of the sentence: a finite
verb form or any other type of main predicate (a noun, a nominal phrase or even in
some case an adverbial phrase). The clause head may also be a subordinate clause:
another non-finite verb form or any other type of embedded clause (relative, adverbial
clause, etc.).
A second critical point which needs clarification is the contrast between true
adverbial clauses and the Badaga Pc converb constructions. Adverbial clauses are preasserted units (i.e. their content cannot be questioned) and are fully autonomous in
terms of their arguments (though they may be shared with the main predicate) and
of their tense-aspect settings. They stand in a certain kind of relation (time, cause,
condition, etc.) to the main predicate and may be compared to the syntactic islands
analyzed by Foley (this volume) which are also usually presupposed statements and
impervious to the inflectional features of the matrix clause.
The constructions involving a Pc converb are more problematic.
Roughly speaking, the Pc strategy appears to be mere grammatical marking
for non-finiteness, signalling that the clause headed by the Pc form has to be syntactically anchored in the next verb or predicate. The form is minimally specified:
its S2 component does not bring any precise tense/aspect/mood meaning, but simply indicates the (potential) positive meaning of the event. But, even this polarity
meaning can be adjusted contextually to the tightest cases of syntactic and semantic
dependency, not only in auxiliary constructions (see paragraph 2.2.4) and lexical
verb compounds, but also when the verb-Pc and the subsequent finite verb form
denote a single event (cf. ex. (4), (15)). In these cases, the mood and the negative
polar meanings of the finite verb (/auxiliary) spread to the Pc verb. In the other
cases, the Pc form simply refers to a sub-event (/activity/state) which is simultaneous or anterior to the verb/predicate to which it is linked. It signals both syntactically and semantically that the sentence/sequence/narration is not finished and that
the situation described in the Pc headed clause is part of a complex situation or

scene which should be considered as a whole. The entire unit ends when an adverbial clause marker occurs (Padv or nominalized adverbial clause) or when the main
finite predicate is uttered. As mentioned previously, such syntactic and semantic
units generally entail some constraints on the arguments (cf. paragraph 2.1.1.1),
typically some shared subject/topic and temporal settings (cf. paragraph2.1.1.2),
a cohesive time span regarding the reference (speech) time (i.e. *yesterday Pc
tomorrow V) . These constraints are more regular on isolated (or elicited) sentences
than within a narration where the Pc forms sometimes merely express a coordinative connection that advances the narration as noted by Nedjalkov. It is this ability to construct a grammatically and semantically unified sequence/scene which
makes the functioning of the Pc forms somehow analogous to that of serial verbs
in other language types, notably in cases when a sequence #V-Pc Vfinite #has
(i)a shared subject, (ii)common tense/aspect/mood/illocutionary force meanings
and (iii)describes a single event. Conversely, it should be noted that this syntactic
linkage, which forcibly brings two (or more) distinct events/situations into a unified scene, probably favours various semantic readings of the Pc headed clause such
as cause (ex. 7).
Two other points would require further study. Firstly, the internal hierarchy of
dependent clauses: their structure is neither flat, nor fully hierarchical. Syntactic
markings, such as Adverbial Participles, confine the scope under which the converbs
are dependent, but here again the linear order of the clauses, as well as prosodic factors, interfere with syntactic marking. A wider discourse approach to the question
could be helpful.
The last problematic point is the complex interaction between negation and the
converbal strategy. This is a question which has already surfaced in other languages
using converbs, see for instance Tikkanen (1995) for Burushaski.
Let us now conclude on a few typological comments. From an areal point of
view, it appears that Badaga uses the same strategies as the other neighbouring
South-Dravidian languages to express adverbial clauses. However, its use of the contextual converb seems to be submitted to fewer constraints than in the other languages.
From a broader typological perspective, the massive use of converbs is evidently
a feature which includes the Dravidian languages in a wide area covering southern,
central and eastern Asia. However, Bisang (1995) drew attention to a different line of
research, based on indeterminacy of the verb form. In his typological study, which
included the Dravidian language Tamil, Bisang noted some functional parallels
between the use of converbs and serial verbs (as for the Badaga Pc). He pointed out
that one of the fundamental features allowing such functioning is the indeterminacy
of the verb form, which is greater for verb serialization than for converbs. However,
the Badaga Pc converb is a very minimally specified verb form, indicating nothing
more than syntactic dependency. If we retain this criterion, it would also be interesting, typologically speaking, to consider other types of underspecified verb forms
found in other language families, such as sequential forms in some African language as
suggested by Bisang (see for instance Wolof, Robert this volume). A typological study
of the various strategies using, full or partial, verb form indeterminacy in the construction of complex sentences would certainly help clarify the shady spaces between
coordination and subordination, between full and reduced predication as well as to
identify their distinct morphological setups and constraints.
Abbreviations
1, 2, 3
Index of person
1pi, 1pex 1st pers.pl. inclusive/
exclusive
3m, 3f, 3n, 3rd pers.sg. masc., fem.,
neuter
3ph
3rd pers. plural human
abl
Ablative case marker
add Additive meaning of the
clitic -
addr
Clitic of address
any Indefinite meaning of the
clitic -
dat
Dative case marker
dub
Doubt clitic
dur
Durative aspect marker
emph
Emphasis clitic
foc Focalizing meaning of the
clitic -
hyp Conditional adverbial
participle
imp
Imperative mood
impPolite Polite Imperative mood
interr
Interrogation clitic
loc
Locative case marker
n1z, n2z Substantivized verb on
S1/S2
neg Negation; 0neg : Tenseless

(absolute) negative ; rneg:
Realis negative
obl Oblique case marker,
covering accusative and
genitive functions
Oblig
Obligative mood
opt
Optative mood
p1adj, p2adj Adjectival participle
on S1/S2
1
2
p adv, p adv Adverbial participle
on S1/S2
pc
Contextual participle
pf
Perfect aspect
pfv
Perfective aspect
pot
Potential mood
prog
Progressive aspect
quot
Quotative < ennu to say
t1 Present future tense =
imperfective
t2 Past tense =
realized achieved
-
Insistence clitic
References
Balakrishnan, Ramaswami. 1999. Badaga. A Dravidian Language. Annamalainagar: Annamalai
University.
Bisang, Walter. 1995. Verb serialization and converbs differences and similarities. In Converbs in Cross-linguistic Perspective, Martin Haspelmath & Ekkehard Knig (eds), 137188.
Cristofaro, Sonia. 2008. Purpose clauses. In The World Atlas of Language Structures Online,
Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds). Munich: Max
Planck Digital Library.
Emeneau, Murray B. 1967. The South Dravidian languages. Journal of the American Oriental
Society 87: 365413.
Emeneau, Murray B. 1980. Language and Linguistic Areas. Stanford CA: Stanford University
Press.
Haspelmath, Martin, Dryer, Matthew S., Gil, David & Comrie, Bernard (eds). 2005. The World
Atlas of Language Structures. Oxford: OUP.
Haspelmath, Martin, Dryer, Matthew S., Gil, David & Comrie, Bernard (eds). 2008. The World
Atlas of Language Structures Online. Munich: Max Planck Digital Library.
Haspelmath, Martin & Knig, Ekkehard (eds). 1995. Converbs in Cross-linguistic Perspective.
Johanson, Lars. 1995. On Turkic converb clauses. In Converbs in Cross-linguistic Perspective,
Martin Haspelmath & Ekkehard Knig (eds), 313347. Berlin: Mouton de Gruyter.
Masica, Colin. 1976. Defining a Linguistic Area: South Asia. Chicago IL: The University of
Chicago Press.
Nedjalkov, Vladimir P. 1995. Some typological parameters of converbs. In Converbs in Crosslinguistic Perspective, Martin Haspelmath & Ekkehard Knig (eds), 97136. Berlin: Mouton
de Gruyter.
Nedjalkov, Igor V. 1998. Converbs in the languages of Europe. In Adverbial Constructions in
the Languages of Europe, Johan van der Auwera & Dnall Baoill (eds), 421456. Berlin:
Walter de Gruyter.
Subrahmanyam, Prakya S. 2006. The role of the conjunctive participle in the Indian linguistic
area. Indian Linguistics 67(14): 215235.
Thompson, Sandra A., Longacre, Robert E. & Shin Ja Hwang. 2007. Adverbial clauses. In Language Typology and Syntactic Description. Complex Constructions, Vol.2, Timothy Shopen
(ed.), 237300. Cambridge: CUP.
Tikkanen, Bertil. 1995. Burushaski converbs in their South and Central Asian areal context.
In Converbs in Cross-linguistic Perspective, Martin Haspelmath & Ekkehard Knig (eds),
Coordination, converbs and clause

chaining in Coptic Egyptian: Typology
and structural analysis*
Chris H. Reintges
CNRS/LLF & University Paris 7 Denis Diderot, Fdration Typologie

et Universaux linguistiques CNRS
Coptic Egyptian (3rd13th c. CE) possesses a large variety of coordinating
constructions. The most central pattern for symmetric clause linkage involves the
coordinating conjunction awf and. Symmetric awf-coordinations at different
levels (phrasal, clausal, discourse paragraph) differ systematically in meaning.
In the domain of asymmetric clause coordination the language employs special
verb conjugations. The focus of this study is on converbs and conjunctive verbs.
Coordinating converbs belong to the paradigm of relative tenses, which formally
distinguish a large family of focus-sensitive constructions from pragmatically
neutral declarative sentences. The conjunctive covers a broad spectrum of
semantic relations between clauses, crossing the traditional distinction between
coordination and subordination.
1. Introduction
This chapter presents a typologically oriented study of clause linking and chaining in
Coptic Egyptian, the latest descendent of the Ancient Egyptian language (Afro-Asiatic;
from around the third to the thirteenth century CE).1 As with many languages of the
African continent, nominal and verbal/clausal coordination are distinguished on a
*I am indebted to William A. Foley, Martin Haspelmath, Jacques van der Vliet, and Ewa
D.Zakrzewska for their detailed written comments on an earlier version of the present chapter.
I am particularly grateful to the volumes editor Isabelle Bril for different kinds of help and
advice. The remaining shortcomings are entirely my own responsibility.
. The modern term Coptic derives from Middle Arabic qubt, itself a corruption of the
Greek adjective (ai)gypt(ios) Egyptian. Ancient speakers of the language called themselves
n-rm-n-kme the people (rm) of Egypt (kme lit. the black country) and used the abstract
noun t-mnt-rm-n- kme what belongs to the people of Egypt to refer to their language
(Crum 1939:110a).
Chris H. Reintges
orphological basis. The situation is, however, complicated by the fact that the invenm
tory of native linkage devices is enriched by the massive influx of Greek connectives,
which are insensitive to the syntactic category of the coordinands. As a result, Coptic
phrasal and clausal coordination also exhibits features of the non-differentiating type
common to European languages.
Coordinating constructions can be identified on the basis of their symmetry.
Symmetric coordination involves a relation between two or more elements of syntactically equal status, which are combined into a larger structure, whilst entertaining the
same syntactic and semantic relationship with the surrounding structural elements. In
structurally symmetric clause coordination the equality of the coordinated clauses is
reflected by the fact that usually either of them could stand alone as an independent
sentence in place of the whole coordination. Consequently, we cannot say that one
clause is the head and the other clause the dependent constituent of the coordinative
construction. Symmetric clause coordination contrasts with structurally asymmetric
(or unbalanced coordination), where two or more coordinands are of syntactically
unequal status. Despite the coordinative semantics of the larger structure, only one
coordinand can stand alone as an independent sentence, while the other coordinands
are structurally dependent. It is therefore always possible to identify one clause as
the head and the other clause as the dependent constituent of the entire coordinative construction (see Dik 1968: 5254; Schachter 1977: 90; Haiman 1983b: 8687;
Johannessen 1998:39; Huddleston, Payne & Peterson 2002:12751276 paragraph 1.1.,
12991300 paragraph 2.2.3.; Haspelmath 2004:34, 34, 2007:56, 46; Bril & Rebuschi
2007:1013, for representative views on symmetric and asymmetric coordination).
The most central and most frequent pattern for symmetric clause linking involves
the coordinating conjunction awf and, which is operative at the phrasal, clausal and
discourse level. Example (1) is an example of a binary clause coordination with identical subjects. In contradistinction to English and related languages, there is no ellipsis
of the same subject in the coordinated clause. When the first clause of a binary coordination has a nominal subject, the following clause has a coreferential third person
subject pronoun. If, on the other hand, a personal pronoun occurs in the first clause, it
recurs in the coordinated second clause.2
. In most Coptic grammars, it is tacitly assumed that the corresponding letters of the CopticGreek alphabet and and and indicate a contrast in quantity though not in quality.
However, distinctive vowel length was already being lost in Egyptian Koine (i.e. the variety of
Greek spoken in Egypt during the Ptolemaic and Roman periods) by the mid-second century
BCE. It is therefore hard to see how vowel quantity could have been re-introduced into the
Copto-Greek alphabet by the time it replaced Demotic writing, unless it was part of the
Egyptian phonological system. Stressed closed syllables contain short vowels, while stressed
open syllables have open vowels instead, e.g. kT /kft/ to build vs. kT /kfte/ to turn;
Coordination, converbs and clause chaining in Coptic Egyptian typology
(1) Symmetric awf -coordination involving subject-identical clauses
a.
The identical subject in the first clause is a noun phrase
ontfs a
tfwn awf
really perf def.m.sg-lord rise.abs and
a=f-wfnh
e-Simfn
perf=3m.sg-reveal.abs to-Simon
The Lord has really risen and revealed (himself) to Simon.
(Luke 24, 34)
p-toeis
b. The identical subject in the first clause is the 1st sing. personal pronoun ti

(Budge, Miscellaneous Coptic Texts 266, 1314)

ti=r
khria m-pei-sa
n-e
(pres)1sg=do.nom need link-dem.m.sg-beautiful link-tree
awf ti=na-fit=f
and 1sg=fut-carry.pron=3m.sg
I am in need of this beautiful tree and will carry it away.
A hallmark of asymmetric clause linkage is the presence of a syntactically dependent

verb form in the coordinated clauses. As a result, the initial and non-initial conjuncts
of the coordination are distinguished not only in terms of linear order, but also in
terms of morphological marking. The concern here is with coordinative converbs and
conjunctives, which, despite their structurally asymmetric syntax, are not only interpreted as coordinative in the semantics, but also exhibit the grammatical behaviour of
standard symmetric coordination.
Coordinative converbs belong to the paradigm of relative tenses, so called because
their morphological derivation involves the merger of a tense/aspect/mood formative with a relative particle. This functionally specialized relativization morphology
is allowed and/or grammatically required in a broad range of information packaging
constructions, including constituent questions, declarative focus sentences, relative
clauses, predicative adjuncts, temporal adverb clauses, and asymmetric clause coordination. Predicative adjuncts, which modify the main verb and one of its arguments,
provide a typical context for the adverbial-subordinative use of relative tenses as converbs. The relative particle e- (glossed as rel) that introduces the predicative adjunct
ea=f-r lle having become blind in (2a) recurs as a clause linkage device in the
coordinate structure in (2b).
Unstressed open syllables, on the other hand, have short vowels, e.g. the /ke/ in k /kelfl/
pitcher, jar. There is some evidence that the digraph N represents a word-final velar nasal,
e.g. N /a/ I. Greek loan words are given in their original orthography. A more comprehensive
discussion on Coptic phonology is offered in Reintges (2004a: Chapter1).
Chris H. Reintges
(2) Converbal relative tenses used in predicative adjuncts and in asymmetric

clause coordination
a. Predicative adjunct introduced by the relative particle e

(Eudoxia 36, 2223)

a=uhe
ero=f
ea=f-r
lle
perf=3pl-find.abs prep=3m.sg rel-perf=3m.sg-do.abs blind
e-pe=f-al
snau
at-def.m.sg=poss.3m.sg-eye two
They found him (Diocletian) having become blind in both eyes.
b. Asymmetrically coordinated clause introduced by the relative particle e

(Shenoute, Leipoldt IV 27, 2627)

a
p-nute
nau
e-pe=u-mkah
perf def.m.sg-god see.abs to-def.m.sg=poss.3pl-distress
ea=f-ti
n-hen-re
n-Jakf hn-hrakhl
rel-perf=3m.sg-give.abs prep-indef.pl-son to-Jacob through-Rachel
God saw their distress and he gave children to Jacob by Rachel.
The conjunctive plays a pivotal role in the grammar of clause combining and interclausal relationships. As an anaphoric tense category, it has no inherent temporal,
aspectual or modal features of its own. Rather, the exponent of asymmetric clause
linkage occupies the structural slot of the tense/aspect particle. Accordingly, conjunctive verbs receive a temporal and aspectual value by being anaphorically related
to a controlling verb in the initial conjunct. (3) is another instance of binary clause
coordination where the conjunctive verb in the second clause assumes a future tense
interpretation from the main verb in the first clause.

(3) Binary coordination with conjunctive verbs
(Testament of Isaac 233, 1314)
se=na-mere
p-ios
n-t-anakhfrsis
3pl=fut-love.nom def.m.sg-life link-def.f.sg-seclusion
et-waa
n=se-apotasse
m-p-kosmos
compREL-be.holy.stat conj=3pl-renounce.abs prep-def.m.sg-world
They (the hermits) will love the life of holy seclusion and will renounce the world.
The opposition between a complete verbal form in the initial conjunct and a less
complete verbal form in all following conjuncts is reminiscent of clause chaining
structures with medial verbs in the languages of Papua New Guinea and the African
continent (see Foley this volume; and, among various others, Haiman 1983a; Longacre
1985:238239, 263283; Givn 1990:864874; Haspelmath 1995:2027; Van Valin &
LaPolla 1997:448454).
In a seminal article Longacre (1985) advances a typological distinction between
co-ranking and chaining languages, where chaining languages employ structural
patterns for clause linking that are radically different from those used in co-ranking
languages. While co-ranking structures typically involve symmetrically coordinated

clauses, it is simply impossible to link two or more clauses of equal status in chaining structures. Approached from this angle, Coptic holds an intermediate position
between co-ranking and chaining language types.
The conjunctive conjugation instantiates an essentially co-subordinative nexus
type, in which several clauses are placed in sequence, but where none of them are
embedded in another. There are, however, other instances of conjunctive clauses that
function as complements to manipulative verbs, verbs of knowledge, and verbs of volition and intent, to mention three representative classes of complement-taking verbs
(Noonan 1985:110132; Dixon 2006:2733). Consequently, with subordinate clauses,
they share the important property of being embedded into the syntactic structure of
the main clause. As structurally ambiguous nexus types, conjunctives and to a lesser
degree converbal relative tenses pose an analytical challenge not only for the traditional division between coordination and subordination, but also for the more refined
taxonomies of clause linking and chaining in current typological studies.
2. Main typological characteristics

As of yet, Coptic has not received a lot of attention in linguistic typology, with The
World Atlas of Language Structures being a notable exception. Before we turn to the
discussion of specific constructions, we must first establish some factual and descriptive background. This section provides some basic information about the language
and its main typological characteristics. The focus of this section is on the division of
labour between alternating verb stems and free-standing tense/aspect/mood particles.
Another topic which will be explored is the correlation between flexible word order
and information packaging.
2.1 The Coptic language

Coptic Egyptian (not to be confused with Egyptian Arabic) is the indigenous language of
late-antique and medieval Christian Egypt. The language is actually a dialect cluster with
at least six regional varieties, two of which gained supra-regional importance: Sahidic
(from Arabic as-S
a d Upper Egypt), and Bohairic (from Arabic al-buhair, a province southeast of Alexandria), the vernacular of the Delta and Lower Egypt, the latter of
which presently functions as the liturgical language of the Coptic Orthodox Church.
Sahidic is renowned as the classical dialect of Coptic for its early records and its
rich literature. It was probably the first Egyptian vernacular into which the Scriptures
were translated (Orlandi 1986:121). Prior to the Arabic conquest of Egypt in 641 CE,
Sahidic was the predominant literary idiom. Its hegemony was broken by Bohairic from
the eighth and ninth centuries onwards. By the eleventh century Bohairic had replaced
Chris H. Reintges
Sahidic as the official church language and had become the sole representative of Coptic Egyptian. In the course of the Islamisation and Arabisation of Egypt in the early
Middle Ages, Coptic and Greek were replaced by Arabic in all public sectors. By the
eleventh century Coptic no longer existed as a spoken vernacular, but continued to be
used as an ecclesiastical language. The appearance of Coptic grammars, glossaries and
text editions in Arabic in the thirteenth century shows a revived cultural awareness and
interest in the ancient vernacular, but also reveals the rather fragmented knowledge of
the language. So far, attempts to revitalize the language have not had any lasting results.
Coptic represents the most recent form of Ancient Egyptian, which has the longest
written tradition of any language: the earliest records date back to the third millennium
BCE. A widely held view in Egyptological linguistics is that the language history comprises two macro-stages with distinct typological features. On the one hand, there is
Earlier Egyptian, which includes Old Egyptian (27502400 BCE) and Middle Egyptian
(21501750 BCE). On the other hand, there is Later Egyptian, which consists of Late
Egyptian (1350950 BCE), Demotic (950 BCE200 CE) and Coptic (3001300 CE)
(Loprieno 1995). This diachronic-typological model does, however, not very well
accommodate the structural differences between Coptic and its Late Egyptian and
Demotic forerunners. In Reintges (2001, 2004b) I therefore propose to trace the nonEgyptian features of Coptic grammar to linguistic change through intensive language
contact: Greek superstratum influence manifests itself not only in massive lexical
borrowing of the native word stock, but also in the restructuring of Egyptian syntax
according to the Greek model. Coptic may therefore be classified a bilingual language
variety with two parent languages, Egyptian and Greek.
2.2 Derivational categories: The verbal grade system

The basic principle of Coptic derivational morphology, which it shares with Ancient
Egyptian and the neighbouring Semitic languages, is that of the root-and-pattern
(Reintges 1994). Verb stems are derived from relatively abstract lexical representations
(roots) by superimposition of templatic patterns. Roots can occur in four and at most
five mutually exclusive stem patterns or grades, depending upon their semantic and
grammatical appropriateness. The different verb stems of the underlying root are formally distinguished by means of vowel alternations and syllable structure; see Table1.
Table 1. The four grades of transitive-active verbs (kt to build)
Grade
Form
Semantics
Valence
Absolute state
Nominal state
Pronominal state
Stative
kft
ketkot=
kt
Eventive
Eventive
Eventive
Stative
Transitive verb with prep. object

Transitive verb with direct object noun
Transitive verb with pronoun object
Intransitive verb with affected subject
The verbal grade system has a semantic and a syntactic dimension. The semantic
dimension relates to situation aspect or Aktionsart (see Comrie 1976; Smith 1991; and
much related research). The first three grades the absolute state, the nominal state,
and the pronominal state convey eventive meaning and are used for the description
of actions, activities and accomplishments. The stative grade, on the other hand, designates states resulting from prior events or states irrespective of their origin.
Aside from marking a core distinction in the aspectual domain, the grade system encodes information that is relevant for syntax. In distinguishing the direct object
from other arguments or adjuncts, the absolute and the nominal state grades manifest
different case-marking patterns. The nominal state represents a head-marking pattern
in the sense of Nichols (1986), where the two constituents, the verb and its complement, must be adjacent to each other.

(Mark 14, 58)
(4) The head-marking pattern of the nominal state grade

n-kit
awf ti=na-ket
ke-wa
n-at-mu^
and 1sg=fut-build.nom other-one as-neg.pfx-build with-hand
hn-omnt n-hou
in-three
link-day
And I will build another (i.e. temple) not as a hand-made one in three days.
The corresponding dependent-marking pattern is represented by the absolute state, in

which the direct object is encoded as a prepositional phrase. Since the original locative meaning of these prepositions has been lost beyond recognition, they function as
semantically empty case markers (glossed as prep, which stands for prepositional object
marker) (see Reintges 2004a: 217224 paragraph 6.3.1 for a more detailed discussion).3
(5) The dependent-marking pattern of the absolute state grade

(Budge, Coptic Martyrdoms 207, 1112)
a=i-kft
na=i
n-u-ma
perf=1sg-build.abs for=1sg prep-indef.sg-place
n-fpe
mayaat
link-dwell.abs reflex.1sg
I built me a dwelling place (all) by myself.
The stative functions as a detransitivizing voice, which eliminates the agent and
aligns the patient argument with the subject function (Polotsky 1987/1990:203205
. The absolute state stem kft to build contains the lexically specified theme vowel /f/, which
must bear stress, while the nominal state allomorph ket- to build has a default vowel /e/ or the
schwa //, which cannot bear stress. In the nominal state grade, the verb and its complement
form a single domain for stress assignment, with the main stress being located on the direct
object noun. The interested reader is referred to Reintges (2004: 201206 paragraph 6.3.1,
218219 paragraph 6.3.1.1) for a more detailed analysis of Coptic verb morphology.
Chris H. Reintges
paragraphs 24; Layton 2000:129 paragraph 168a; Reintges 2004a: 228 paragraph6.3.3).
With an extremely rich and largely unpredictable morphology, the stative is markedly
derivational in character. Yet, as a detransitivizing voice, it shows the earmarks of an
inflectional category.

(Luke 6, 48)
(6) The detransitivizing stative grade

ne=f-k7t
kalfs
comp pret=3m.sg-build.stat well
because it (the house) was built well
te
From the perspective of major syntactic categories, eventive stems are less finite and,
in a sense, less verbal, than their counterparts in earlier language stages. This is why
they have traditionally been analyzed as infinitives (Stern 1880: 154 paragraph 318; Till
1961:122 paragraph 253; Polotsky 1987/1990, 197196 paragraph 40; Layton 2000:125
paragraph 160). Indeed, absolute and nominal state stems are readily available for nominalizations, e.g. p-kft the act of building, edification, the building (Crum 1939:122b).
More importantly, however, alternating verb stems are no longer compatible with the
exponents of tense, aspect and mood.
2.3 Inflectional categories: Tense/aspect/mood particles

Coptic belongs to a small fraction of 139 languages out of a sample of 1062 in which
the exponents of temporal, aspectual and modal categories are encoded by means of
separate words or particles (Dryer 2005: Chapter69). Tense/aspect/mood (tam) particles furnish a broad range of conjugation patterns (i.e. patterns in which verb stem can
appear) and can be subdivided into two positional classes of preverbal and presubject
particles. Preverbal particles are morphologically invariant and appear at the edge of
the verb phrase constituent.

(7) The preverbal future particle na
(Psalm 28, 11)

na-krine
n-n-laos
def.m.sg-lord fut-judge.abs prep-def.pl-people
The Lord will judge the nations.
p-toeis
Members of the large and diversified group of presubject particles are placed at the
left edge of the clause, leaning on the subject noun phrase. The postponement of the
enclitic particles provides evidence for the proclitic status of presubject particles,
which are attached by the phonology to the immediately following subject constituent
(Muysken 2008:4041).

(8) The presubject perfect particle a
(Mena, Martyrdom 4a, 12)
a
te=f-sfne
de fl
n-ne=f-kees
perf def.f.sg=poss3m.sg-sister pcl carry.abs prep-def.pl=poss.3m.sg-bone
His sister carried his bones.
Presubject particles may undergo allomorphic variation, depending on the nominal

or pronominal status of the subject. As we can see from the contrast between (9a)
and (9b), the base form is selected in the context of pronominal subjects. When a
nominal subject is to follow, the tam-particle is lengthened, and contains the suffix -re
(Polotsky 1960:399, 418419).

(9) a.
Base form a= of the habitual aspect particle

awf a=f-fk
aro=s n-sep-snau mne
and hab=3m.sg-go.abs to=3f.sg on-time-two daily
And he would go to her twice a day.
b. Lengthened form are of the habitual aspect particle

(Hilaria 6, 16)
(Psalm 36, 11)
are p-dikaios
n
ht=f
hab def.m.sg-righteous be.merciful.nom heart=poss.3m.sg
The righteous one is merciful
Dryer (1992:99, paragraph 3.5) distinguishes auxiliary verbs and tense/aspect particles on the basis of the presence and the absence of verbal inflections, respectively. On
this view, inflected tense/aspect particles would qualify as auxiliary verbs, while the
corresponding uninflected forms would be uninflected particles. There are, however,
several reasons why a binary opposition between inflected auxiliaries and uninflected
tense/aspect particles would not give us the descriptively adequate results. To begin
with, the impoverished agreement of presubject particles seems to be a morphological innovation of the Sahidic dialect, which is not shared by all neighbouring dialects. Moreover, the alternation between short bases and lengthened allomorphs has
not been paradigmatized throughout the entire class of presubject particles. From a
diachronic perspective the emergence of inflected particles in Sahidic is all the more
surprising as neither the finite verb of VSO nor the auxiliary of Aux-S-V-O clausal patterns display any kind of agreement with the immediately following subject.
2.4 Word order variation and information structure

The languages basic word order is SVO. Canonical SVO clauses involve a minimal
amount of syntactic structure and morphological marking and are productively used
in out-of-the-blue contexts, i.e. they provide felicitous answers to questions like Whats
going on? or What happened next?.
(10)

Basic SVO order in out-of-the-blue contexts

(Mena, Miracles 10b, 3334)
a
t-kamaule
mise
n-u-eere
n-shime
perf def.f.sg-she-camel deliver.abs prep-indef.sg-girl link-woman
The she-camel delivered a daughter.
Left-dislocation encodes a pragmatic relation of aboutness between a referent and a

proposition vis--vis a given discourse. The topicalized noun phrase or its equivalent
Chris H. Reintges
is placed in the left periphery and connected to the associated comment clause via a
resumptive pronoun.
(Mena, Martyrdom 4b, 69)
(11) Left-dislocation
m-p-toeis
p-aggelos
de
def. m.sg-angel pcl link-def.m.sg-lord
a=f-wonh=f
e-p-arkhiepiskopos
perf=3m.sg-reveal.pron=3m.sg to-def.m.sg-archbishop
The angel of the Lord revealed himself to the archbishop.
Subject-verb inversion is an information-packaging construction used when the subject corresponds to presentational (new information) focus.4 The informational status
of the postverbal subject is marked by the focus particle nki, while its grammatical
role is marked by a cataphoric subject pronoun.
(12) Subject-verb inversion
(Eudoxia 40, 2728)
awf a=f ti
eow nki p-ran
and perf=3m.sg-take.nom glory foc def.m.sg-name
m-pe=ntois
Jsus pe-khristos
link-def.m.sg-poss.1pl-lord Jesus def.m.sg-Christ
And the name of our Lord Jesus Christ was glorified.
The positional freedom of the subject and, to a lesser degree, the direct object and
adverbial modifiers is largely determined by the information load that these constituents have in the discourse.
3. Patterns of nominal and verbal/clausal coordination

According to the Word Atlas of Language Structures, 125 languages out of 301 distinguish nominal and verbal/clausal coordination on a morphological basis, as opposed
to the majority of 161 languages, in which the same marker is employed for the conjunction of both noun phrases and clauses (Haspelmath 2005: Chapter64). Coptic is
correctly identified as belonging to the differentiating type. Nominal coordination is
coded by locative and comitative prepositions, while there are functionally specialized
. There is a strong tendency for presentational foci to be hearer-new. This results in the
frequent occurrence of indefinite and quantified noun phrases in subject-inverted clauses.
The novelty condition on inverted subjects does, however, not exclude definite noun phrases.
Definite noun phrases and proper names are admissible as presentational foci names if the
relation their referents enter into is novel with respect to the preceding stretch of discourse.
dependent verb conjugations for the expression of asymmetric VP/clausal coordination. The most central pattern for symmetric VP/clause linking involves the coordinating conjunction awf and, which also has a marked use as a nominal coordinator. The
inventory of native coordinating devices is enriched by the massive influx of Greek
function words, which retain most of their structural properties, in particular, their
category-insensitivity with respect to the two coordinands (Reintges 2001, 2004b).
Accordingly, Coptic clause coordination also exhibits features of the non-differentiating type common to European languages. This section examines the distinctive grammatical properties of nominal and verbal/clausal coordination and its morphological
markers. The discussion is confined to coordinative patterns for which a native origin
is established.
3.1 Nominal coordination

In the Word Atlas of Language Structures, Coptic Egyptian is classified as a withlanguage, in which an originally comitative marker has been grammaticalized into a
nominal coordinator (Stassen 2005: Chapter63). This view is, however, not entirely
accurate, since the comitative strategy only applies to phrasal coordination, while a
locative preposition hi on, at operates at the sub-phrasal level and connects determinerless, bare nouns.5 On the semantic side, hi-coordination creates conceptual
units like flesh and blood and abominations and crimes (cf. Mithun 1988:332).6
. The coordinating function of the preposition hi is clearly distinguished from its original
locative function. Coptic prepositions come in two forms, one used with nominal objects,
called the nominal state, and another used with pronouns, called the pronominal state. Since
bound pronouns cannot receive stress, the pronominal state allomorph comprises additional
phonological material (Till 1966: 109 paragraph 233; Mallon: 157 paragraph 311; Layton
2000:162 paragraph 200). In the case of hi, this material is supplied by noun incorporation:
(pron. st.) hiff= < hi + ff= back, dorsum (Reintges 2004a: 101 paragraph 3.2.1.1). Crucially, the pronominal state hiff= is excluded from the domain of nominal coordination,
since coordinated pronouns are marked by the comitative preposition mn.
. For Stassen (2004:8), such conventionalized pairings are more likely to be encoded by
zero-marking than other, less predictable NP-coordinations. The situation is markedly different in Coptic, where the locative strategy performs this semantically restricted linkage
function. This does not seem to be a parochial feature of Coptic Egyptian, however. In Iraqw
(Cushitic, Tanzania), the linker nee and is productively used to coordinate inherently linked
nominal expressions. Yet, while the Coptic preposition hi is restricted to bare noun coordination, the Iraqw coordinator nee displays a much broader syntactic distribution. See Mous
(2004) for a detailed discussion.
Chris H. Reintges
(13) Bare noun coordination by means of the locative preposition hi

a. hi-coordinated nouns determined by an indefinite article


a u-[sarks
hi-snof]
I indef.sg-flesh prep-blood
I am someone of flesh and blood.
b. hi-coordinated nouns determined by a definite article

n-[ote
hi-anomina] m-pe-mto
def.pl-abomination prep-crime in-def.m.sg-presence
eol m-p-nute
pcl link-def.m.sg-god
the abominations and crimes in the presence of God
(Eudoxia 40, 15)
The comitative preposition mn/nmma= with enters into a paradigmatic opposition with the locative preposition hi, combining determined noun phrases, personal pronouns, and proper names into a complex noun phrase (Shisha-Halevy
1986:21 paragraph 1.1.1 (3); Ernst 1994:95; Layton 2000:56 paragraph 65 (ii), 109
paragraph(145).
(14) Phrasal coordination by means of the comitative preposition mn
a.
Coordination of indefinite noun phrases
ti=na-ti
n-hen-smu
men-hen-sahu.
1sg=fut-give.abs prep-indef.pl-blessing with-indef.pl-curse
I will give blessings and curses.
b. Coordination of definite noun phrases
(Deuteronomium 11, 26)
(Till, KHML I 83, 1718)
pf=k
pe
p-eow
men-t-tayo.
poss=2.m.sg cop.m.sg def.m.sg-glory with-def.f.sg-glory
Yours (is) the glory and the honour.
c.
Coordination of definite noun phrases
ma-tuto=k
(Luke 23, 39)
nmma=n
give.imp-save.pron=2m.sg with=1pl
Save yourself and us!
d. Coordination of proper names
(Till, KHML II 22, 35)
pai
pe
Antfnios pai
pe
Pahfm
o
dem.m.sg cop.m.sg Anthony dem.m.sg cop.m.sg Pachomius great
men-Petrfnios men-Thefdfros pai

pe
Apa enute
with-Petronius with-Theodore dem.m.sg cop.m.sg Apa Shenoute
This (is) Anthony, this (is) Pachomius the Great, Petronius and Theodore
(and) this (is) Apa Shenoute.
On the syntactic side, the prepositional character of the comitative construction is

maintained, in particular, the discontinuous constituency of the coordinated noun
phrases (Stassen 2000:18; Haspelmath 2007:31).7
(15)

Discontinuous constituency of coordinated NPs

(Mark 16, 8)
nere u-stft
amahte mmo=u men-u-tortr
pret indef.sg-trembling grasp.abs prep=3pl with-indef.sg-fright
Trembling and fright grasped them.
The domain of the clausal connective awf and is extended to include some marked
uses as a nominal coordinator, which thus provides an exception to the overall categorysensitivity of native coordinating devices. A case in point is layered NP coordination,
where the conjunction awf connects a series of participant pairs, each of which contains
the comitative preposition men (Shisha-Halevy 1989:54 paragraph 2.1; Ernst 1994:95).
(16) The layered coordination construction A awf B mn C mn D
(Mark 6, 3)
m-pai
an pe
p-hame
p-re
m-Maria
neg-dem.m.sg not cop.m.sg def.m.sg-carpenter def.m.sg-son link-Mary
awf p-son
n-Iakoos men-Ifs men-Iuda men-Simfn?
and def.m.sg-brother link-Jacob with-Joses with-Judas with-Simon
Is this one not the carpenter, the son of Mary and the brother of Jacob, Joses,
Judas, and Simon?
Coptic provides yet another example of a language that switches to different coordinators when the coordinands are understood to represent a list (Payne 1985:24). As
a rule, NP coordination is monosyndetic and involves a single coordinator per coordinand (Haspelmath 2004:45). There are only a few attested examples of bisyndetic
coordination, in which the coordinands are bound together by the two coordinators
awf and mn (Till 1966:190 paragraph 376; Layton 2000:109 paragraph 145).
. According to Stassen (2000: 27) languages differ with respect to the degree to which
the string NP1with-NP2 is reanalyzed as a plural noun pattern. Even though the comitative
preposition mn changes semantically to become a conjunction marker, it does not adopt the
coding properties of a coordinative construction, with the nominal complex NP1 mn-NP2
triggering third person plural agreement. Examples like (i) are marginal.

(i)
Plural agreement with comitative NP coordination
(Hilaria 5, 3031)
awf a=f-ti
na=u n-u-ma
and perf=3m.sg-give.abs to=3pl prep-indef.sg-place
e-tre=u-wfh
nht=f ntos mn-p-diakonos
to-caus.inf=3pl-stay.abs in=3m.sg she with-def.m.sg-deacon
And he gave them a place to stay, her and the deacon.
Chris H. Reintges
(17) Bisyndetic coordination A awf mn B
(Pistis Sophia 18, 1112)
pi-nok
ete
ntof pe
n-woein
rel (pres) he cop.m.sg dem.m.sg-great link-light
awf men-p-mystrion
m-p-tiu
m-prohgumenos
and with-def.m.sg-mystery link-def.m.sg-five link-instructor
He is the great light and the mystery of the five instructors
Nominal coordination covers some middle ground between a semantically coordinative and a syntactically locative or comitative strategy.
3.2 Structurally symmetric verbal/clausal coordination

In symmetric coordination, the coordinands are equal at the syntactic level, i.e. they
are full clauses of the same type. This is reflected in the fact that the coordinated
clauses are not differentiated in terms of syntactic function, but only in terms of linear
position. A second indication of the equality of the coordinands is that the order of the
coordinated clauses can be reversed without altering the truth conditions or having
any significant effect on the well-formedness of the larger structure. Symmetrically
coordinated verb phrases and full clauses in Coptic are typically bound together by the
coordinating conjunction awf and. Despite their symmetric syntax, awf-coordinated
clauses exhibit a considerable amount of asymmetry in terms of their interpretation.8
3.2.1 Asyndetic clause coordination

Although coordinate constructions are usually marked by a coordinator, they do not
have to be. In fact, coordination represents a syndetic variant of paratactic clause
combining. As Longacre (1985:239) and Haspelmath (2007:78) have pointed out,
asyndetic coordination provides further evidence for the category-sensitivity of
. The coordinating conjunction awf and has been grammaticalized out of a petrified
imperative jw or r-w Add! < w to place, add. The coordinating function of r-w is
first attested in Late Egyptian and Demotic (see Mller 1888: 95; Erman 1933:87 paragraph
196; Spiegelberg 1924:36 paragraph 55; see also Steindorff 1951:200 paragraph 415; Westendorff
1977:14). The Demotic coordinate construction in (i) involves the ellipsis of the main verb
tw=j I gave in the first conjunct.

(i)
Coordinative use of the imperative r-w in Demotic Egyptian

(Book of the Dead, pap. Bibliothque Nationale 149, 2, 3233)
tw=j
tp-ntr
n-n-ntrw
q
r-w
give=1sg offering-god to-def.pl-god-m.pl bread aug-add.imp
mw n-n-i>-w
water to-def.pl-spirit-m.pl
I gave a gods offering to the gods and bread and water to the spirits.
c oordinative strategies, since the coordinated entities must be clausal in the broadest
sense, i.e. including predicate/VP coordination.9
Asyndetic coordination in Coptic is stylistically marked: the very absence of coordinator and linkage markers implies a tighter unity between the juxtaposed clauses
(Layton 2000:260 paragraph 335b).
(Hilaria 5, 14)
(18) Asyndetic clause coordination
p-diakonos de a=f-eire
m-pe=s-wf
def.m.sg-god pcl perf=3m.sg-make.abs prep-def.m.sg=poss3f.sg-wish
a=fti
m-toot=s
m-p-nomisma
perf=3m.sg-take.abs from-hand=poss.3f.sg prep-def.m.sg-solidus
a=u-wfm
mn-ne=u-eru
perf=3pl-eat.abs with-def.pl=poss.3pl-recip
m-pe-hou
et-mmau
on-def.m.sg-day comp.rel-there
The deacon fulfilled her wish, took the solidus (a golden coin) from her (and)
they ate together on that day.
The linear order of asyndetically coordinated clauses generally reflects the temporal sequence of events. Even though temporal inferences are very strong, the event
expressed in asyndetically coordinated clause is often understood as a consequence of
the event expressed in the preceding clause.
3.2.2 Syndetic clause coordination with awf

Functionally, Coptic VP coordination behaves like a verb phrase, although it does not
have the syntactic structure of one. Consequently, the coordinated VPs are modified
. Some isolated instances of asyndetic noun phrase coordination can be found in the
context of enumerations (Layton 2000:109 paragraph 145(a)).

(i)

Marked asyndetic noun coordination

(Shenoute, Leipoldt III 206, 13)
hen-tfm
n-ht
hen-mnt-sokj
indef.pl-hardness link-heart indef.pl-nominal-foolish
hen-mnt-at-sown
indef.pl-nominal-neg-know.abs
hen-nokj
m-mnt-ass
a
p-nute
indef.pl-great link-nominal-impious perf def.m.sg-god
nt=e
eol n-ht=u
bring.pron=2f.sg pcl from=3pl
hen-pornia
hen-mie
indef.pl-fornication indef.pl-strife
(Concerning) acts of hardheartedness, foolishness, ignorance, fornication, strife,

(and) great impiety, God has delivered you (woman) from them.
Chris H. Reintges
by a single tam-particle. (19) features the negative perfect mp-, whose past tense
reference and negative polarity extends to the entire VP coordination.
(19) VP coordination with awf
(Shenoute, Wess. 9, 87a, 3033)
mn
wa n-wot
te
mp=f-fms
awf
be.not one link-single comp neg.perf=3m.sg-drown.abs and
ffte
eol
wipe.abs pcl
There is no one such that he did not sink and get wiped out.
In structurally symmetric awf-coordination, the coordinated full clauses could be

used as an independent sentence in place of the whole coordinate structure. The
coordinated awf-clause is often subject-identical with the first clause, but differentsubject clauses are also very common. This is illustrated in (20) and (21), respectively.
Subject identity is therefore not a definitional criterion for structurally symmetric
awf-coordination.
(20) Same-subject awf-coordination
(Apophth. Patrum, Chane n0 1, 1, 34)
a=f-fpe
hn-u-amelia
perf=3m.sg-become.abs in-indef.sg-negligence
awf a=fte
pe=f-ahe
eol kakfs
and perf=3m.sg-spend.nom def.m.sg=poss.3m.sg-time pcl badly

He became negligent and spent his time badly.
(21) Different-subject awf-coordination
(Sahidic Vita of Pachomius 90, 1517)
nere Thefdfros hmoos m-p-we

n-u-kui
pret Theodore sit.abs in-def.m.sg-distance link-indef.sg-small
er
p=f-ho
rel(-pres) def.m.sg=poss.3m.sg-face
ude-ne=f-pat
e=f-reime
between-def.pl=3m.sg-knee rel(-pres)= 3m.sg-weep.abs
awf ner hah on hn-ne-snu
reime
and pret many pcl among-def.pl-brothers weep.abs
Theodore was sitting at a short distance with his head between his
knees weeping and many of the brothers were also weeping.
The coordinated clauses need not have entirely parallel syntactic structures. The prologue of the Gospel of John is particularly instructive, since the Sahidic translation preserves the chiastic structure of the Greek original. In the first conjunct, p-ate the Word
represents the new information focus, which triggers subject-verb inversion. It recurs in
the second conjunct as a left-dislocated topic constituent directly following awf.
(22) Same-subject awf-coordination with different word orders

nki
(John 1, 1)
p-ate
hn-te-howeite
ne=f-oop
in-def.f.sg-beginning pret(-pres)=3m.sg-exist.stat foc def.m.sg-word
awf p-ate
ne=f-oop
and def.m.sg-word pret(-pres)=3m.sg-exist.stat
nnahrn-p-nute
in.front.of-def.m.sg-god
In the beginning was the Word, and the Word was with God.
Awf-coordination may operate at the discourse level, connecting a new paragraph to

a preceding thematic unit.
(23) Use of awf as a paragraph marker
awf e=i-thefrei
mmo=u mn-ne=u-karpos
and rel(-pres)=1sg-look.abs prep=3pl with-def.pl=poss.3pl-fruit
eis ftow n-re
m a=u-ei
m-p-we
look four link-child small perf=3pl-come.abs from-def.m.sg-distance
And while I was looking at them (the trees) and their fruit, behold, four young
fellows came from a distance.
Coordination at different levels differs in meaning. In general, phrasal and clausal

coordination imply a closer association between the coordinands than higher level
awf-coordination, which provides an instruction for the addressee to open a new discourse file and to close the preceding one.
3.2.3 Semantic asymmetries in awf-coordination

As many researchers have observed, symmetric clause coordination can (and in certain cases must) receive an asymmetric interpretation, which diverges from standard
logical connection by being sensitive to coordinand permutations. When the semantic
connection between the coordinated clauses involves a relation of temporal succession
and/or a relation of cause, consequence or concession, the order of the coordinands is
no longer interchangable, suggesting that the coordinands are not equal at the semantic level. For this reason, the constructions concerned are not cases of prototypical
symmetric coordination (see Johannessen 1998:233, 241242; Huddleston, Payne&
Peterson 2002: 12991300 paragraph 2.2.3; Blakemore & Carston 2005: 570572;
Bril& Rebuschi 2007:810 for relevant discussion).
A case in point for semantic asymmetry is the left-subordinating awfconstruction
in which the first conjunct is interpreted as a subordinate adverbial clause. (24)illustrates a special case of the conditional implicature, where the initial imperative politeue
mn-pe-Khristos Lead (your life) with Christ! has a conditional interpretation, while the
Chris H. Reintges
conjoined clause awf f-na-nahm=k and he will save you has the semantic interpretation of a consequent clause. The implicature is If you lead (your life) with Christ, he will
save you provides a reason for complying with the directive Lead (your life) with Christ.
This example is entirely parallel to left-subordinating and-constructions in English,
e.g. You drink another can of beer and Im leaving (Culicover & Jackendoff 1997).
(24) Conditional reading of awf-coordination
(Teaching of Antonius 3738)
politeue
mn-pe-Khristos
awf f-na-nahm=k
administer.imp with-def.m.sg-Christ and 3m.sg-fut-save.pron=2m.sg
Lead (your life) with Christ and he will save you!
A closely related reading of awf-coordinated clauses is that of a concessive construction in which the left conjunct of awf specifies a nevertheless or despiterelationship between the coordinated clauses. According to Knig (1995: 6869),
conditional and concessive clauses share the property of presupposing a general
connection between two types of situations, but while concessive clauses have a factual interpretation, conditional clauses are interpreted hypothetically. Furthermore,
concessive constructions imply that two facts are in conflict with general expectations about cause and effect. The intended concessive reading may be reinforced by
adding the Greek discourse particle ketoi (< ka toi and yet, even though) to the
initial conjunct.
(25) Concessive reading of awf-coordination (Apophth. Patrum, Chane n0 210, 55, 12)
ketoi ne-wnta=n
hah mmau pe
pcl pret-have=1pl many there cop.m.sg
awf ne=u-rfe
mmo=n an
and pret=3pl-suffice.abs prep=1pl not
And even though we had many (things), they were not enough for us.
Although awf-coordination is symmetrical in terms of syntactic structure, one also

finds an impressive variety of non-canonical types, ranging from awf-clauses that
follow a major thematic break, and hence lack a first conjunct, to awf-clauses in
which the first conjunct is semantically interpreted as a subordinative conditional or
concessive clause.
3.2.4 A note on gapping

The connective awf is consistently used in gapped coordination, in which the coordinated clause is structurally incomplete: the main verb and the tense/aspect particle are
elided so that there is a gap in the middle of the clause. The deletion of the subject and
the verb in gapped coordination is only possible when the coordinated clauses have
entirely parallel structures.
(26) Gapped coordination

a.
Gapped coordination of the infinitival verb
(Mark 12, 17)
ke
ti
na-p-rro
m-p-rro
give.nom poss.pl-def.sg.m-king pcl to-def.sg.m-king
awf na-p-nute
m-p-nute
and poss.pl-def.sg.m-god to-def.sg.m-king
Give to the king what is to the king and to God what is to God!
b. Gapped coordination of the subject and the finite verb

(Praecepta Pachomii 115)

t-kathkei
de n-t-nsteia
e=f-na-r
def.f.sg-instruction pcl link-def.f.sg-fast rel=3m.sg-fut-do.nom
t-wi
hm-pe=f-i
awf
def.f.sg-one in-def.m.sg=poss.3m.sg-house and
t-kete
hm-pe-i
m-pe=f-r
def.f.sg-other in-def.m.sg-house link-def.m.sg=poss.3m.sg-colleague
(As for) the instruction of fast: he shall make the first one in his house
and the other in his colleagues house.
Each remnant, i.e. each constituent flanking the gap in the second conjunct generally
contrasts with its counterpart in the first conjunct. In other words, the remnant in the
coordinated clause must allow for a contrastive focus interpretation for gapping to
apply (see Fry & Hartmann 2005 and the references cited therein).
4. Asymmetrically coordinating verb conjugations

In asymmetric clause coordination the coordinands are of syntactically unequal status.
There is a single clause that can stand on its own outside of the coordinate structure,
whereas all the other clauses contained in it cannot be used as independent utterances.
In expressing structural dependency between two or more clauses, asymmetric coordination exhibits varying degrees of affinity with subordination, in which one clause is
embedded in or adjoined to the other. Asymmetries at the syntactic level may be manifest in the choice of verb forms. Many of the world languages have at their disposal
specialized verb conjugations to formally mark the asymmetric features of a coordinate
structure (see Foley, this volume and, among various others, Givn 1983:83; Longacre
1985:238239; Johannessen 1998:3435; Van Valin & LaPolla 1997:448454; FabriciusHansen & Ramm 2008:8).
Coptic provides yet another example of a language in which the initial and the noninitial conjuncts of asymmetric clause coordination are morphologically differentiated
Chris H. Reintges
by means of separate verbal paradigms. The two major asymmetrically coordinative

verb conjugations are represented by converbal relative tenses and the conjunctive. To
complete the picture, the conjunctive will be compared with the inferred evidential,
which, in addition to asymmetric clause linkage, marks consequence and goal and
hence fulfils purposes in narrative discourse other than event sequencing.
4.1 Converbal relative tenses

4.1.1 A paradigmatic split in the Coptic tense system
Tense is a deictic category, which grammaticalizes the location of a situation on
the time axis. The concern here is with deictic tenses, which take the present moment
as their temporal reference point. The linear ordering of events with respect to the
present moment yields three deictic tenses, to wit, the present, the past and the
future that locate the time of the event at, prior to or after this temporal reference
point (see Comrie 1985 and much related research).10
With more than twenty different conjugation patterns, the tense-aspect-mood
system constitutes one of the most complex areas of Coptic grammar. At the foundation of such a wealth of morpho-semantic distinctions is the subdivision of the present, future and perfect tense into basic and relative tenses. Relative tenses are derived
from basic tenses by adding a relative marker to the tense/aspect particle. The relative
marker e- constitutes the elsewhere form, while the allomorph nt- is restricted to a
single context, that of the perfect tense marker a.
The class of relative tenses can be further subdivided into independent and converbal relative tenses. However, while the paradigms of relative and converbal tenses
are identical in the present and the future, morphological differentiation takes place
only in the perfect tense: the independent relative perfect is formed with the marked
allomorph nt-, whereas the corresponding converb is marked by the default form
default form e- (Reintges 2004a: 309 paragraph 8.1.5.1). Table 2 presents the inflectional paradigms of present, future and perfect tenses.
. The tripartite tense system of Coptic Egyptian allows for more fine-grained distinctions
in the past domain. In Reintges (2004a: 27175 paragraph 7.3.5) I analyze the preterit copula
ne- as the spatio-temporal marker of a secondary deictic center with respect to which temporal
interpretations are made. Events can then be interpreted as coinciding with, preceding or
following this newly established vantage point, which furnishes a triplet of preterit tenses.
These are the preterit present ne=f-sftm he heard, was hearing (present-in-the-past), the
preterit past (pluperfect) ne-a=f-sftm he had heard (past-in-the-past) and the preterit
future ne=f-na-sftm he was going to hear (future-in-the-past).
Table 2. The inflectional paradigms of Coptic deictic tenses (sftm to hear)

Present
sg
pl
1
2m
2f
3m
3f
1
2
3
Basic
Relative
Converb
ti=sftm
k=sftm
te=sftm
f=sftm
s=sftm
tn=sftm
tetn=sftm
se=sftm
e=i-sftm
e=k-sftm
ere-sftm
e=f-sftm
e=s-sftm
e=n-sftm
e=tetn-sftm
e=u-sftm
e=i-sftm
e=k-sftm
ere-sftm
e=f-sftm
e=s-sftm
e=n-sftm
e=tetn-sftm
e=u-sftm
Future
sg
pl
1
2m
2f
3m
3f
1
2
3
Basic
Relative
Converb
ti=na-sftm
k=na-sftm
te=na-sftm
f=na-sftm
s=na-sftm
tn=na-sftm
tetn=na-sftm
se=na-sftm
e=i-na-sftm
e=k-na-sftm
ere-na-sftm
e=f-na-sftm
e=s-na-sftm
e=n-na-sftm
e=tetn-na-sftm
e=u-na-sftm
e=i-na-sftm
e=k-na-sftm
ere-na-sftm
e=f-na-sftm
e=s-na-sftm
e=n-na-sftm
e=tetn-na-sftm
e=u-na-sftm
Perfect
sg
pl
1
2m
2f
3m
3f
1
2
3
Basic
Relative
Converb
a=i-sftm
a=k-ssftm
are=sftm
a=f-sftm
a=s-sftm
a=n-sftm
a=tetn-sftm
a=u-sftm
nt-a=i-sftm
nt-a=k-sftm
nt-are-sftm
nt-a=f-sftm
nt-a=sftm
nt-a=n-sftm
nt-a=tetn-sftm
nt-a=u-sftm
e-a=i-sftm
e-a=k-sftm
e-are=sftm
e-a=f-sftm
e-a=s-sftm
e-a=n-sftm
e-a=tetn-sftm
e-a=u-sftm
Relative tenses must be selected in restrictive relative clauses, which contain no

relative pronoun or subordinating complementizer besides the relative marker to
formally mark the core relativizing functions of attribution and subordination. The
relative marker itself does not encode any nominal functional features of the relative antecedent; all these features are present on the resumptive pronoun inside the
embedded clause.
Chris H. Reintges
(27) Relative clauses with the relative perfect
m-pa-toeis
t-irn
def.f.sg-peace link-def.sg.m.poss.1sg-lord
[RC nt-a=f-taa=s
na=i]
rel-perf=3m.sg-give.pron=3f.sg to=1sg
the peace of my Lord, which he has given to me
Relative particles can readily be identified with finite relative complementizers, albeit
with an interesting twist. Due to their reanalysis as a focus-sensitive morphology, relative tenses are not simply subordinate verb forms, but display a broad syntactic distribution across different clause types and can also be found in main and embedded
contexts. Constituent questions and declarative focus sentences are two cases in point.
(28) The selection of relative tenses in sentences with marked information structure
a. Constituent question with relative perfect


nt-a
u
fk
e-pe=k-ht?
rel-perf what come.abs to-def.m.sg=2m.sg-heart
What has come into your heart?
b. Declarative focus sentence with relative perfect

(Budge, Coptic Homilies 14, 1921)

mmon, alla nt-a=u-stot=f

no
but rel-per=3pl-prepare.pron=3m.sg
m-p-diaolos
mn-ne=f-aggelos
for-def.m.sg-devil with-def.pl=poss.3msg-angels
(Is the Purgatory prepared for us?) Not at all! It is rather prepared
for the devil and his angels.
The reader is referred to Reintges (2003, 2007a, b); Green & Reintges (2004); and
Reintges, LeSourd & Chung (2006) for a comparative syntactic analysis of this special
relativisation morphology.11
. Hausa, a Chadic (Afroasiatic) language spoken by more than 30 million people in

northern Nigeria and southern parts of Niger, displays a comparable paradigmatic split of the
imperfective and perfective paradigms into a general and a relative class. Jaggar (2001:162
paragraph 6.4 Footnotes4 and 5) recently proposed to replace the conventional term relative
by the semantic cover term focus, which provides an appropriate semantic characterization for the syntactic environments in which relative aspects are selected. Appealing though a
uniform analysis in terms of focus marking may appear, further research should clarify how a
semantic focus analysis carries over to relative clause constructions, in which relative marking
is obligatory, or conditional and temporal adjunct clauses, where it is optional (see Reintges &
Green 2004:185186 for preliminary observations).
4.1.2 Adverbial-subordinative versus coordinative converbs

Relative tenses occur as converbs in predicative adjuncts that modify the main verb
and one of its arguments without actually forming a complex predicate or serial verb
construction.12 Coptic predicative adjuncts are clausal modifiers, which presupposes
that they have their own subjects. The connection between the secondary predicate
and the modified argument is established via a coreferential pronoun. (29) is a typical
example of an object depictive, where the direct object noun e-u-agelei a herd acts as
the controller of the converbal relative perfect ea=s-ei having approached.
(29) Object depictive with converbal relative tense

a=i-kft
a=i-nau
e-u-agelei
perf=1sg-look.up.abs perf=1sg-see.abs prep-indef.sg-herd
n-o
ea=s-ei
m-p-we
link-antelope rel-perf=3f.sg-go.abs from-def.sg.m-distance
I looked up (and) saw a herd of antelopes having approached from a distance.
Schultze-Berndt & Himmelmann (2004:67) contend that in genuine depictive predication, it is generally not possible to have tense and mood marking as well as the
expression of all verbal arguments. Nonetheless, converbs are included in their typological study of depictive and resultative predicates (see, in particular, their discussion
on pages 98101). One might raise a principal objection to the analysis of adverbialsubordinative relative tenses as converbs on the grounds that converbs cannot be
readily analyzed as a sequence of a verb and a complementizer (Haspelmath 1995:4).
However, in Coptic relative tenses, the relative marker and the tense/aspect are fused
to form an internally complex conjugation base. In other words, relative marking is an
integral part of the inflectional paradigms of Coptic verbs.13
. According to Bisang (1995), there is no direct correlation between canonical morphological type (synthetic, analytic, agglutinative) and serial verb versus converb languages. He
writes that it would, however, be too simplistic to assign so-called isolating languages to the
serial languages and the agglutinative and inflectional languages to the converb languages
(idem, p.138). Coptic Egyptian is a language of the analytic/isolating type, yet makes productive use of various types of converbs.
. To account for the presence of relative tenses in predicative adjuncts, one may capitalize on the focus role of depictive predication. Depictives make a predication that is partially independent of the primary predication and are often semantically more specific than
the main verb. In this respect, they generally contribute to the new information conveyed
by the entire construction (Winkler 1997: Chapter6). Alternatively, one might argue that
the depictive predicate functions in much the same way as a restrictive relative clause, narrowing down the predication conveyed by the main verb, to the exclusion of other possibilities
Chris H. Reintges
The presence of converbs in non-initial conjuncts represents a marked typological feature of asymmetric clause coordination. When connecting two or more clauses
in series, converbal relatives are not adverbial in any sense, but must be analyzed
as coordinate (see, among various others, Polotsky 1987/1990:254 paragraph 28;
Shisha-Halevy 1988:116 paragraph 28.1; Layton 2000:343344 paragraph 428; and
Reintges 2004a: 312 paragraph 8.1.5.3).
(30) Asymmetric clause coordination with converbal relative tense

(Eudoxia 42, 12)
mnnsa-p-eou
de m-pe-Khristos
after-def.m.sg-god pcl link-def.m.sg-Christ
a=f-ei
aro=f
nki Mikhael p-arkhaggelos
perf=3m.sg-make.abs to=3m.sg foc Michael def.m.sg-archangel
n-m-pwe
ea=f-tsao=f
link-def.pl-heaven rel-perf-teach.pron=3m.sg
After the glorification of Christ, Michael, the archangel of heaven,
came to him (Constantine) and instructed him.
Coordinative converbs differ semantically from adverbial-subordinative ones in terms

of temporal interpretations. In secondary predication, converbal relative tenses combine deictic with anaphoric temporal reference and locate an event with respect to
the present moment as well as with respect to the event denoted by the main verb.
Coordinative converbs, on the other hand, convey a deictic tense value and extend the
narration along the same time-line as the main verb in the initial conjunct. Thus, the
temporal interpretation of converbal relative tenses is directly related to their adverbialsubordinative or coordinative use.
It is also possible, although not very common, to combine independent and converbal relative tenses in a coordinate construction. The clearest instances involve the
perfect tense, in which the two patterns are morphologically differentiated. (31) is an
example of a rhetorical question introduced by the Greek interrogative particle m.
The presence of the independent relative perfect nt-a=f-fk he came in the first
conjunct is motivated by the focus-sensitive nature of this sentence type (Reintges
2003:375). By contrast, the converbal relative perfect ea=f-eine and he brought in
the second conjunct has an elaborative function and adds supplementary information
for the purpose of discourse.
(Schultze-Berndt& Himmelmann 2004:92). I leave a more precise characterisation of the

focus structure of Coptic relative clauses and predicative adjuncts for future research.
(31) Cooccurence of syntactically independent and converbal relative perfect

m nt-a=f-fk
ehrai e-t-pe
nki Pakhfm
o
Q rel-perf=3m.sg-go pcl to-def.m.sg-heaven foc Pachomius great
ea=f-eine
epest n-ne=f-entol
rel-perf=3m.sg-bring pcl prep-def.pl=poss.3m.sg-rule
tekas e=n-e-sftm
ero=u nt=n-aa=u
so.that rel=1pl-prep-listen.abs to=3pl conj=1pl-do.pron=3pl
Did Pachomius the Great go up into heaven and bring down his rules,
so that we might hear them and obey them?
To conclude, coordinative converbs introduce an asymmetric feature into standard

symmetric coordination in the sense that they manifest special verb conjugations not
used in independent clauses.14 In other words, the relative marker serves as the exponent of asymmetric clause linkage rather than as a special type of inflectional morphology that registers the presence of different types of semantic focus in information
packaging constructions.
4.2 The conjunctive conjugation

The conjunctive occupies a special position in the Coptic tense/aspect/mood system.
Apart from its clause linking function, it has no temporal, aspectual or modal value
of its own. Consequently, conjunctive verbs receive a temporal interpretation by being
anaphorically related to a controlling verb with fully specified tense/aspect marking in
the initial clause of the coordinate structure.
4.2.1 Paradigmatic irregularity

The inflectional paradigm of the conjunctive conjugation in Standard Sahidic displays
a considerable degree of allomorphic variation in various person, gender, and number
distinctions. The paradigmatic irregularity results from the interaction of two parameters, one of which concerns the distribution of the alternants n- and nte- and the other
the variable position of the schwa //. There are two allomorphs nta- and ta- for the
first person singular, where the short form ta- involves the deletion of the morphemeinitial syllabic nasal n-.15 Table 3 presents the corresponding forms in the non-standard
variety of post-classic Sahidic and in Bohairic Coptic (Till 1942; Kahle 1954:160163
. I am indebted to Martin Haspelmath for clarifying this issue.
. The first person singular form nta- would be more correctly represented as nt=a-,
where the vocalic suffix =a is an allomorph of the first person singular suffix pronoun =i.
Chris H. Reintges
paragraph 138; Steindorff 1951:171 paragraph 366; Mallon 1953:119 paragraph 249;
Layton 2000:276 paragraph 351; Reintges 2004a: 296297 paragraph 8.1.1.4.1).
Table 3. Intra-dialectal variation of the conjunctive paradigm
Standard Sahidic
Late Sahidic
Bohairic
2m
nta-sftm,
ta-sftm
-sftm
nta-sftm,
ta-sftm
nte=k-sftm
2f
3m
nte- sftm
n=f-sftm
3f
n=s-sftm
nte=n-sftm
2
3
nte=tnsftm
n=se-sftm
sg
pl
nte=k-sftm,
te=k-sftm
te-sftm
nte=f-sftm
te=f-sftm
nte=n-sftm,
te=n-sftm
te=tne-sftm
nt=u-sftm,
t=u-sftm,
n=su-sftm
nte-sftm
nte=f-sftm
nte=s-sftm
nte=n-sftm
nte=tn-sftm
nt=u-sftm
On the basis of interdialectal comparison, it can be concluded that the conjunctive paradigm in Standard Sahidic has two base morphemes n- and nte-, which cannot be derived from each other. The base morpheme nte- in Standard Sahidic lies at
the root of the more streamlined paradigms of Bohairic Coptic and the non-standard
idiom of Late Sahidic. The elision of the syllabic nasal onset n-, which is restricted to
the first singular form ta- in Standard Sahidic, affects the entire paradigm of the nonstandard variety (see also Polotsky 1944:1011).16
4.2.2 Conjunctive verbs and clause chaining

Longacre (1985:238, 264267) distinguishes co-ranking structures from clausal chaining in terms of the syntactic dependence of the non-initial conjunct. In co-ranking
structures, each conjunct can stand alone as an independent sentence in place of the
coordination. By contrast, there is only a single instance of an autonomous sentence in
. The inflectional paradigm of the Coptic conjunctive represents a language-internal innovation. The first attestation of the base morpheme n- is found in two bilingual Greek and
Demotic magical formularies from the first centuries CE that already contain some Old Coptic
glosses (Spiegelberg 1924:70 paragraph 140; Johnson 1976:181).

(i)
The innovative conjunctive formative n- (London/Leiden Magical Papyrus V33, 3)

xr=w
r
imi
n-i-ir=k-wnm
say=3pl Horus come.imp conj-aug-aux=2m.sg-eat.inf
They said Horus, come and eat!
clause chaining. Series of conjunctive clauses display the hallmarks of chaining structures. The initial clause of the chain contains a fully specified verb form with complete
tam-markings, while conjunctive verbs without inherent tense/aspect specifications
are consistently used in all subsequent clauses.
(32) is an example of a medium-length clause chain, in which a sequence of four
same-subject conjunctives receives a future time interpretation from the controlling verb
f=na-ale he will ascend in the initial clause. In covering a sizeable stretch of discourse
such clause chains exceed an average complex sentence in a European language and
compare well in distribution and length with a typical paragraph (Longacre 1985:264;
Haspelmath 1995:22).
(32) Medium-length conjunctive chain with initial future tense

mnnsf=s on f=na-ale
etn-u-e
n-stauros
after=3f.sg pcl 3m.sg=fut-ascend.abs on-indef.sg-wood link-cross
n=f-mu
ha-p-tr=f
conj=3m.sg-die.abs for-def.m.sg-all=poss.3m.sg
n=f-tfwn
m-pe=f-meh-omnt
n-how
conj=3m.sg-raise.abs on-def.m.sg=poss.3m.sg-nominal-three link-day
n=f-fl
n-amnte n=f-fei
conj=3m.sg-destroy.abs prep-Hell conj=3m.sg-carry.abs
n-t-mnt-rfme
n-toot=f
m-p-tate
prep-def.sg.f-nom-man from-hand-poss.3m.sg prep-def.m.sg-enemy
After that he (Jesus Christ) will ascend on a cross (lit. a wood of cross) and will
die for the whole (world) and rise on the third (lit. his third) day and destroy
Hell and take all humanity away from the hands of the enemy.
The Coptic facts support the typological correlation between verb-object order and
posterior clause chaining. The basis word order is SVO and the controlling verb with
complete tam-specifications precedes all conjunctive verbs (Stassen 1985:101; Givn
1990:881). It is worthwhile pointing out that the sequencing of conjunctive clauses has
a semantic correlate in iconicity, with clause order mirroring the chronological order
of events (Haiman 1983a: 120).17
. According to Johannessen (1998:5556), there is a significant correlation between general

word order and asymmetric or unbalanced coordination: a language that is head-final (SOV)
has a normal final conjunct and a deviant first conjunct. Similarly, a language that is headinitial (SVO, VSO) has a normal initial conjunct and a deviant final conjunct. The situation
is different in Coptic, where the final conjunct order does not exist. Borrowed particles and
function words may appear in the postverbal domain, but never in the absolute clause-final
position. Thus, consider (i), in which the Greek enclitic particle de occurs after the verb and
the direct object, but before the inverted subject NP and the locative prepositional phrases.
Chris H. Reintges
Givn (1990:889) contends that the position of the clause-linking morpheme provides another feature of interclausal grammar that correlates with word order typology. In verb-object languages, the connecting morpheme appears clause-initially in
the following clause, often as a prefix on the clause-initial word. Conjunctive clause
chaining provides a potential counterexample to this claim. To begin with, conjunctive
morphology conforms entirely to the languages analytic morphological type. The conjunctive base morphemes n- and nte- are functional particles rather than morphological prefixes. More importantly, however, conjunctives are not in any sense structurally
deficient clauses. Accordingly, conjunctive clauses can be modified by the same range of
interclausal connectives as standard symmetric coordination. In a similar vein, adverb
preposing and left-dislocation are applicable to conjunctive clauses. When following
a clause-initial conjunction or a left-dislocated topic constituent, the conjunctive formative appears in clause-second or clause-third position. In other words, there is no
requirement for conjunctive morphology to occur in absolute clause-initial position.
4.3 Delimiting the conjunctive from converbal relative tenses

Haspelmath (1995:23) observes that medial verbs in a clause chaining structure behave
like converbs in that they (i) cannot stand alone as independent sentences, (ii) generally do not express tense and mood, and (iii) frequently depend on the superordinate
verb for subject reference. The distinction between adverbial-subordinative converbs
and clause-chaining medial verbs is therefore a moot point. In this section I consider
ways to delimit the conjunctive from related constructions, in particular, converbal
relative tenses and infinitives, paying particular attention to the definitional criteria of
adverbial subordination and non-finiteness. This syntax-based distinction is further
supported by semantic criteria, in particular, the different temporal interpretations of
converbal relative tenses and conjunctives.
4.3.1 Adverbial-subordinative status

The definitional criterion of adverbial subordination is primarily intended to exclude
participial modifiers as well as masdar and verbal noun formation from the crosslinguistic study of converbs (Haspelmath 1995: 7). The adverbial status of converbs
is particularly clear in depictive predication, even though their occurrence may be
(i)
Postverbal placement of Greek enclitic particles
(Luke 8, 33)
a=s-ti
pe=s-woi
de nkji t-agel
perf=3f.sg-give.nom def.m.sg=poss.3f.sg-way pcl foc def.f.sg-herd
etn-t-fme
epest e-t-lim
on-def.f.sg-slope pcl to-def.f.sg-lake
The herd (of demons) rushed down the slope to the lake.
licensed by certain types of main verbs. Verbs of perception and discovery, for instance,
typically function as the primary predicate of objective depictives, as seen in (33).
(33) Object depictive with converbal relative present
k=nau
ero=i
e=i-fne
(pres-)2m.sg=see.abs prep=1sg rel(-pres)=1sg-become.sick.abs
You find (lit. you see) me sick.
Outside the context of secondary predication, converbs may function as the counterparts of temporal when/while-clauses in the familiar European languages (Nedjalkov
1995: 99). When a temporal connective and/or a durational or temporal location
adverb are added, the semantic relation between the converb and the superordinate
clause is specified to a large extent (Knig 1995:7581). The Greek conjunction en
hoson while in (34) marks the temporal overlap between the event denoted by the
converbal relative present and the main clause event.
(34) Temporal clause with converbal relative present
(Till, KHML I 143, 1214)
e=n-ate
en hoson de
mn-ne=n-eru
in so.far pcl rel(-pres)=1pl-talk.abs with-def.pl=poss.1pl-recip
a=f-ei
ehun nki p-hgemf n
perf=3m.sg-come.abs pcl foc def.m.sg-governor
n-te-khfra
et-mmau
link-def.f.sg-district comp.rel-there
While we were talking with each other, the governor of that district came in.
According to Haspelmath (1995:8), coordinative converbs are not adverbial at all, but
this represents a less central or less typical use.
(35) Coordinative converbs
(Budge, Coptic Martyrdoms 210: 2326)
p-ran
de n-te-heneete
et-mmau
pe
def.m.sg-name pcl link-def.f.sg-monastery compREL-there cop.m.sg
Erte e=no
n-u-ht
n-wft
Erte rel(-pres)=1pl-do.stat in-indef.sg-heart link-single
mn-ne=n-eru
hi-u-sop
with-def.pl=poss.1pl-recip in-indef.sg-time
ere
t-irin
oop
hn-t=n-mte
rel(-pres) def.f.sg-peace happen.stat in-def.f.sg=poss.1pl-midst
e=n-oop
hn-u-esykhia
mn-eru
rel(-pres)=1pl-happen.stat in-indef.sg-ease with-recip
e=n-ti
eou m-p-nute
rel(-pres)=1pl-give.nom praise to-def.m.sg-god
The name of that monastery (was) Erte and we lived together with one
another in harmony and peace was in our midst and we were at ease with
one another and praised God.
Chris H. Reintges
Coptic has a closed class of aspectual verbs like kjf to remain, stay and lo to stop,
finish, which describe, respectively, the ongoing state and the termination of an event,
while the event itself is denoted by the following converb (Noonan 1985:129; ter Meulen
1995: Chapter2). Since converbs form a complex predicate with the aspectual verb,
they function syntactically as complement clauses rather than clausal modifiers, which
are adverbial.
(36) Converbal relative tenses used as aspectual complements
a.

With aspectual verb of continuation kf to remain

nne=i-kjf
e=i-kk
ahu
neg.fut=1sg-remain.abs rel(-pres)=1sg-strip.off.stat naked
I would not stay naked.
b. With the aspectual verb of cessation lo to finish


a=i-lo
e=i-fne
perf=1sg-stop.abs rel(-pres)=1sg-become.sick.abs
e-pa-hpar
at-def.m.sg.poss.1sg-liver
I stopped being sick in my liver.
As subordinative verb forms, conjunctives are also excluded from occurring in main
clauses. The conjunctive manifests an essentially co-subordinative nexus type, in which
the linked clauses have a coordinative interpretation in the semantics, but are realized
as dependent clauses in the syntax.
(37) Embedded bipartite coordination with a conjunctive verb
(Ephesians 6, 22)
pai
ent-a=i-tnnou=f
arf=tn e-pei-hf
dem.m.sg rel-perf=1sg-send.pron=3m.sg to=2pl for-dem.m.sg-thing

te
e=tetn-e-eime
e-pe=n-wf
comp rel=2pl-prep-know.abs about-def.m.sg=poss.1pl-news
awf n=f-parakalei
n-ne=tn-ht
and conj=3m.sg-comfort.abs prep-def.pl=poss.1pl-heart
This one (Tychicus) I sent to you for this matter that you would know
about our whereabouts and he would comfort your hearts.
There are other instances of conjunctive clauses that behave like subordinate structures. A case in point is conjunctive complement clauses to manipulative verbs.
(38) Conjunctive complements to manipulative verbs
(Till, KHML I 5, 2324)
a
p-rro
keleue
n=se-tfks
perf def.m.sg-king order.abs conj=3pl-drive.in.abs
m-p-hermetarion
ha-p-tetrapylfn
prep-def.m.sg-hermetarion under-def.sg.m-fourgated.archway
n-t-polis
n=se-eie
m-p-gennaios
ero=f
link-def.f.sg-city conj=3pl-hang.up.abs prep-def.m.sg-noble on=3m.sg
The king ordered that they drive in the hermetarion (a torture instrument)
under the fourgated archway of the city and that they hang the noble one on it.
Simplifying matters somewhat, converbal relative tenses are adverbial-subordinative

forms, bearing in mind that coordinative uses and complex predicate formation are
by no means excluded. The primary function of conjunctive verbs is that of clause
combining and chaining, with possible extensions to different types of subordinate
clauses. A more detailed discussion of conjunctive argument and adjunct clauses will
be offered in Section5.2.
4.3.2 Scale of finiteness

The role of finiteness as a definitional criterion for converbs has been controversially
discussed in the literature. Haspelmath (1995:3) defines converbs as non-finite verbs
forms, whose primary function is to mark adverbial subordination. Finiteness is
not included in Nedjakovs (1995:97) definition of a converb as a verb form which
depends syntactically on another verb form, but is not its syntactic actant, i.e. does not
realize its semantic valencies.
The task of distinguishing converbs from related constructions in terms of finiteness is not an easy one, since finiteness and non-finiteness do not represent crosslinguistically stable categories of verbal inflection. The traditional notion of finiteness
goes back to Latin grammar, which divided the verbal paradigm into forms determined by person and number (verba finita) and personally undetermined forms
(verba infinita), where the latter class includes participles, gerunds and supines (Nikolaeva 2007:1). Following this tradition, many descriptive grammars view finiteness as
an inflectional category of the verb. Although in Latin the finite/non-finite distinction
was originally motivated by the presence versus the absence of person and number
agreement, other inflectional categories were subsequently taken into account, most
prominently tense (Huddleston 2002:8889 paragraph 1.8.1).
From a functional-typological perspective, Cristofaro (2007:9192) argues that
although finiteness and non-finiteness are manifested in different constructions in different languages, the various properties that are regarded as distinctive for finiteness
occur in predictable correlation patterns from one construction to the other. Following
Chris H. Reintges
Lehmann (1988:200) and Haspelmath (1995:5), I consider that (absolute) finite and
(absolute) non-finite forms are just two extreme points on a scale of desententialization. Absolute finiteness and non-finiteness, on their part, are connected to the presence or absence of tense and person agreement.
Shifting the attention back to Coptic grammatical categories, converbal relative
tenses are absolute finite verb forms, since they contain a tense/aspect particle, with
person agreement being manifest in the coreferential pronoun. The criterion of non-
finiteness is more delicate to assess for the conjunctive conjugation, where the exponent of asymmetric clause linkage appears in the structural slot of the tense/aspect
particle. Conjunctive verbs exhibit temporal restrictions in the sense that a coordinative interpretation is sometimes excluded by the tense-aspect specification of the
controlling verb. This generally shows that conjunctive verbs are not entirely devoid of
temporal features. Finally, they are distinguished from (absolute) non-finite infinitives
by the following syntactic properties:
i.
Omissibility of the subject. There are no infinitival constructions in which the

presence of an overt subject is grammatically required, even though the semantics
requires there to be an understood subject. The identity of infinitival subjects
is generally recoverable from the subject of the superordinate clause. In samesubject conjunctives, on the other hand, there is no subject ellipsis. Rather, the
identical subject must be overtly expressed by a coreferential pronoun.
ii. Prepositional complementizers. Most infinitivals, apart from the complements of aspectual and modal verbs, are introduced by a prepositional complementizer. The most common subordinator is the directional preposition
e- to. Although conjunctive clauses are modifiable by a range of subordinating conjunctions, they are in complementary distribution with prepositional
complementizers.
iii. Nominalization. Infinitives are readily available for nominalization, e.g. p-kft
the act of building, edification, the building (Crum 1939: 122b). Conjunctive
verbs, on the other hand, cannot occur in any kind of nominalization structure.
The key differences between conjunctive verbs and infinitives can conveniently be
illustrated with examples like (39), in which the two forms are coordinated (ShishaHalevy 1986:203204 paragraph 7.2.5.1).
(39) Coordination of conjunctive and infinitival clauses

alla e=f-roeis
awf e-f-hareh
but rel(-pres)=3m.sg-guard.abs and rel(-pres)-3m.sg-watch.out.abs
e-tm-paht=f
n-u-daimonion
awf
to-neg.aux-throw.pron=3m.sg to-indef.sg-demon and
n=f-fpe
ha-t-eksusia
conj=3m.sg-become.abs under-def.sg.f-power
n-u-pneuma
n-tate
link-indef.sg-spirit link-enemy
But if he (the god-fearing man) guards (himself) and watches out not to submit
himself to a demon or fall under the power of a hostile spirit.
Conjunctive verbs may but need not agree with the controlling verb in polarity. To initiate
a switch from positive to negative polarity, conjunctive verbs are modified by the negative
auxiliary verb tm to do not. This is also the standard pattern of negation for infinitives.
(40) Coordination of converbal relative present and negated conjunctive

(Acts of Andrew & Paul 200, 8990)
nim pe
p-rfme
e=f-na-kft
who cop.m.sg def.m.sg-man rel=3m.sg-fut-watch.abs
nsa-p=f-re
e=f-k
n-emtf
after-def.m.sg=poss.3m.sg-son rel(-pres)=3m.sg-go.stat to-depth
n=f-tm-othei
ero=f ?
conj=3m.sg-neg.aux-help.abs prep=3m.sg
Who (is) the man who will watch his son drowning (lit. go to the depths) and
will not help him?
In outranking infinitives on a scale of finiteness, the conjunctive provides a potential

counterexample to Givns (1990:875) typological generalization that in the clausechaining type, sequentially-conjoined clauses are the least finite of all clause types [his
emphasis, CHR].
4.3.3 Temporal ordering relations

The temporal interpretation of converbal relative tenses is in accordance with
Nedjalkovs (1995: 131) typological generalization, stating that if a language has a
converb with the fundamental meaning of succession, then it also has a converb with
the fundamental meaning of precedence. The existence of the latter usually also presupposes the existence of a converb with the fundamental meaning of simultaneity.
Converbal relative tenses have a reference point that is located at, before, or after the
present moment and in addition locate a situation at, before, or after that reference
point. Depending on their inflectional tense, they establish temporal ordering relations of simultaneity, precedence and subsequence.
(41) Converbal future with successive interpretation
e-u-toi
(Hilaria 3, 1718)
e=f-na-kr
a=s-he
perf=3f.sg-find.abs prep=indef.sg-ship rel=3m.sg-fut-sail.abs
e-u-polis
te
Sarala
to-indef.sg-city comp Sarala
She (Hilaria) found a ship ready to sail to a city called Sarala.
Chris H. Reintges
(Eudoxia 36, 2223)
(42) Converbal perfect with precedence interpretation
a=uhe
ero=f
ea=f-r
lle
perf=3pl-find.abs prep=3m.sg rel-perf=3m.sg-do.abs blind
e-pe=f-al
snau
at-def.m.sg=poss.3m.sg-eye two
They found him (Diocletian) having become blind in both eyes.
(43)

Converbal present with simultatenous interpretation

(Hilaria 12, 67)
t e
a=uhe
ero=s
rf e=s-onh
comp perf=1pl-find.abs prep=3f.sg pcl rel(-pres)=3f.sg-live.stat
because we found her (Hilaria) at last alive
In expressing a succession of events, the conjunctive has inherent future orientation,

where the time expressed or implied in the coordinated clause is later than the time of
the superordinate clause. In contrast to converbal relative tenses, the notion of simultaneity and anteriority is not part of its temporal semantics.
(44) Sequential-consecutive interpretation of conjunctive verbs
(Hilaria 9, 1314)
a=s-paht=s
n=s-rime
hab=3f.sg-throw.pron=3f.sg on-def.sg.m-earth conj=3f.sg-weep.abs
etm-p-kah
ante=s-herp
p-kah
hn-ne=s-rmjowe
term=3f.sg-drench.nom def.sg.m-earth with-def.pl=poss.3f.sg-tear.pl
She (Hilaria) used to throw herself on the ground and weep until she drenched
the ground with her tears.
There is, however, a single context in which conjunctive verbs must be interpreted as
contemporaneous with the controlling verb. This is when conjunctive verbs are combined with aspectual verbs of initiation to form a complex predicate. Since the complex predicate designates a particular stage of a single event, the time reference of the
conjunctive must be identical to that of the aspectual verb.
(45) A conjunctive verb combined with an aspectual verb of initiation

e-i-na-tfwn
ta-okt
mmo=i
rel=1sg-aux-raise.abs conj.1sg-strangle.abs prep=1.sg
I will raise and strangle myself (to death).
The converbal relative present is compatible with the stative grade, in which case it
denotes the condition or state attained by the referent of the controlling argument.
(46) Converbal relative present with the stative
alla ti-er
hote e-fk
ehun
but (pres)=1sg-make.nom fear to-go.abs pcl
e-p-topos
e=itahm
to-def.m.sg-shrine rel(-pres)=1sg-defile.stat
But I am afraid of entering the shrine (of the holy Apa Mena) being impure.
The stative grade is generally excluded from the event-oriented conjunctive. Conjunctive verbs may, however, appear in the inchoative construction, which is formed with
the existential verb fpe to happen, become. In this context, they combine periphrastically with a secondary stative predicate (Funk 1977:25).
(47) Combination of conjunctive and stative in the inchoative construction

(Matthew 9, 17)
alla e-a=u-net
rp n-rre
e-askos
n-rre
but rel-hab=3pl-put.nom wine link-new to-wineskin link-new
n=se-fpe
e=u-wot
m-p-snau
conj=3pl-become.abs rel(-pres)=3pl-be.healthy.stat prep-def.m.sg-two
Unless one puts new wine into new wineskins, the two are in a healthy condition.
Converbal relative tenses combine with stative stems in the present tense and may
assume a range of relative-temporal interpretations, whereas conjunctives are more
semantically restricted. In particular, they cannot be stativized and are incompatible
with simultaneous present tense and anterior past interpretations.
4.4 Delimiting the conjunctive from the inferred evidential

In the Coptological research tradition, the conjunctive has been studied together with
yet another asymmetrically coordinating verb conjugation, labelled the inferred evidential in the present study (Shisha-Halevy 1986:195 paragraph 7.2.1.1.5.1; Polotsky
1987/1990: 163164 paragraph 38; Depuydt 1993: 8586; Layton 2000: 276 paragraph 350).18 As suggested by the terminology, the inferred evidential has a semantic
component besides its coordinative semantics that is not present in the conjunctive.
In marking consequence and goal, it fulfils purposes different from event sequencing in the production of narrative discourse, but rather encodes inference based on
(non-)observable facts. As an introspective evidential, the inferential also indicates
the speakers belief that the inference being made is particularly robust and cogent.
Consequently, the speaker takes responsibility for the truth of his or her assertion (see
Aikhenvald 2004:174176 for the typological distinctions in inferred evidentials).19
. The traditional nomenclature of the inferred evidential as a conjunctivus futuri future conjunctive (Stern 1880: 284 paragraph 450) reflects the alleged semantic affinity of the two asymmetrically coordinative patterns. Sterns terminology has been revived in Laytons (2000:283284
paragraph 57) reference grammar. However, as already pointed out by Polotsky (1987/1990:163
paragraph 38), the inferential does not have a more futurate orientation than the conjunctive.
The alternative term finalis, which is, for instance, found in Tills grammar (1966:157156 paragraph 311), does not capture the evidential semantics of this verb conjugation.
. Contrary to what is stated in many Coptic reference works (inter alia, Steindorff 151:166
paragraph 352; Polotsky 1987/1990:160 paragraph 31; Layton 2000:276 paragraph 276, N.B.),
there is no evidence for a paradigmatic gap in the first person singular (Reintges 2004a: 324
Chris H. Reintges
The semantic differences between the conjunctive and the inferred evidential can
conveniently be illustrated with the example of mixed coordination, which involves
a sequence of different clause types. The coordinative constructions in (48) and (49)
consist of an initial directive speech-act (in the imperative mood) followed by a statement (or prediction) introduced by a conjunctive and inferential verb form, respectively.
Within such mixed coordinations, the coordinating device has wider scope than the illocutionary component (Huddleston, Payne & Peterson 2002:13321333 paragraph 3.5).
(48) Mixed coordination with initial imperative and different-subject conjunctive clause

(Eudoxia 68, 1011)
tenu ke tfwn
maro=n nta-tit=e
interj pcl raise.imp opt-1pl conj.1sg-take.pron=2f.sg
etm-pe=f-taphos
to-def.m.sg=poss.3m.sg-tomb
Now, then, rise, let us go and I take you (woman) to his (Jesus Christs) tomb!
(49) Mixed coordination with initial imperative and different-subject inferential clause

(Ephesians 5, 14)
tfwn=g
p-et-nkotk
=lo
raise.imp=2m.sg def.m.sg-compREL-sleep.abs conj.2m.sg-cease.abs
ute-n-et-mowt
tare pe-Khristos
r
from-def.pl-compREL-die.stat infer def.m.sg-Christ do.nom
woein ero=k
light to=2m.sg
Raise yourself, you who is sleeping, and cease from among the dead and
Christ will appear to you!
paragraph 8.2.4.1). In a system based on a primary division of the source of knowledge into
the speaker and the other speech participants, inferred evidence is intrinsically related to
the speaker. When the speaker was a knowing participant in an event, either as a voluntary
agent or conscious experiencer, the knowledge of that event is normally direct and evidential
markers are therefore often omitted. First person singular inferentials do, however, occur in
those contexts where the speaker distances himself from his own actions and takes the stance
of an outside observer (see Aikhenvald 2004:15756, 219233 for a more detailed discussion
on first-person effects in evidentials).

(i)
First person singular inferred evidential
awf fi
p-rou
tar=i-ei
nsf=k
and take.imp def.m.sg-care infer=1sg-go.abs behind=2m.sg
hn-u-rae
e-men
fsk
with-indef.sg-joy rel-not.be hesitate.inf
And take care and I shall fetch you with joy without delay.
In mixed coordination with same-subject conjunctives, the directive force of the initial
imperative clause is spread out to the non-initial conjunctive clause. As a result, the
entire coordinative structure expresses a single directive speech-act.
(50) Mixed coordination with initial imperative and same-subject conjunctive clause

fei
n-ne=k-al
ehrai
lift.imp prep-def.pl=poss.2m.sg-eye pcl
-nau
e-n-kolasis
conj.2m.sg-see.abs prep-def.pl-punishment
Lift your eyes and look at the punishments!
The consequent implicature renders the inferred evidential an illocutionary island,

impermeable to the directive or interrogative force of the initial conjunct. The mixed
coordination in (51) is interpreted as a sequence of a directive speech-act and an assertion. The speaker assures the addressee about a particular result or outcome once he
complies with his directive.
(51) Mixed coordination with initial imperative and same-subject inferential clause

(Sirach 6, 18)
pa-re
tin-te=k-mnt-koui
def.m.sg.poss.1sg-child since-def.f.sg=poss.2m.sg-nominal-young
sftm
n-te-so
tare=k-he
e-u-kharis
listen.imp to-def.f.sg-teaching infer=2m.sg-find.abs prep-indef.sg-grace
a-te=k-mnt-hllo
until-def.f.sg=poss.2m.sg-nominal-old
My child, listen to the teaching from your childhood onwards and
(I assure you) you will find grace up until your old age!
This concludes my discussion of the inferred evidential, which generally shows that
clause coordination interacts in complex ways with other domains of verbal semantics,
such as modality, illocutionary force and evidentiality (see Reintges 2004a: 324327
paragraph 8.2.4 for further discussion).
5. Conjunctive verbs and interclausal nexus relations

Foley & Van Valin (1984: Chapter6) acknowledge three types of interclausal nexus
relations: coordination, subordination, and cosubordination. The three types of interclausal nexus are defined in terms of two syntactic parameters [ dependent] and
[embedded]; see Table4 for further illustration.
Chris H. Reintges
Table 4. A ternary distinction of nexus relations

Coordination
[ embedded]
[ dependent]
Subordination
Cosubordination
[+ embedded]
[ embedded]
[+ dependent]
[+ dependent]
Foley & Van Valin (1984:241 (6.4), 42, Figure10).
Cosubordination behaves syntactically like standard coordination in that the conjoined clauses are neither modifiers nor arguments of the controlling verb. It is syntactically distinguished from coordination in that the conjoined clauses depend on a
superordinate clause in some feature, for instance, tense, aspect and illocutionary force.20
In involving coordinative semantics and syntactically dependent syntax, the cosubordinative nexus type has independently been observed to challenge the traditional contrast
between coordination and subordination (see, among various others, Haiman 1983a:
121123; Roberts 1988:5053; Givn 1990:864865; Van Valin & LaPolla 1997:448
454; Yuasa & Sadock 2002:99; Haspelmath 2004:3337, 2007:4748; Foley, this volume).
The Coptic conjunctive is an essentially cosubordinate nexus type, where several
clauses are placed in sequence, but none of them are embedded in another. The picture is, however, complicated by the fact that there are other instances of conjunctive clauses that function as dependent constituents within a larger structure. As subject
and object argument clauses or as oblique clausal modifiers, subordinative conjunctives
belong to the core or the periphery and share with standard subordinate clauses the
property of being embedded into the syntactic structure of the superordinate clause.
5.1 Cosubordinate nexus

5.1.1 Same-subject conjunctive clauses
Cosubordinate constructions with only same-subject clauses are generally not used to
simply narrate events. Rather, the coordinated clauses describe parallel events that are
. In Van Valin & LaPollas (1997: 454) semantically oriented approach to interclausal
nexus relations, the dependent status of cosubordinated clauses is defined in terms of operator
dependence, i.e. shared tense and illocutionary force. Subordinative nexus, on the other hand,
is characterized by structural dependence and subsumes subject and object complement
clauses on the one hand, and clausal modifiers (relative clauses, adverbial clauses) on the other
hand. Foley (this volume) presents various types of empirical evidence, showing that clauses
may be separately specified for tense inflection, but share the same illocutionary marker. This
would entail that they are coordinate with respect to tense and aspect, but cosubordinate with
respect to illocutionary force. If nexus is a structural relationship between linked clauses,
this is certainly not a warranted result, since conjoined clauses should not be able to bear
contrasting structural relationship to each other (idem, p.29).
fused into a single complex event, as in (52a), or events that are contrasted with one
another, as in (52b) (see Blakemore & Carston 2005:570571, 580581).
(52) Binary coordination with same-subject conjunctive clause
a.
Coordinated clauses describe a complex event
(1 Corinthian 15, 32)
mar=n-wfm
n=tn-sf
te
tn=na-mu
opt=1pl-eat.abs conj=1pl-drink.abs comp 1pl=fut-die.abs
gar n-raste
pcl in-tomorrow
Let us eat and drink, since we will die tomorrow!

(Psalm 36, 21)
b. Coordinated clauses describe contrasted events

ti
eto=f
are p-ref-r-noe
hab def.sg.m-nominal-do.nom-sin take.abs for=3sg.m
n=f-tm-taa=u
conj=3m.sg-neg.aux-give.pron=3pl
The sinner borrows and will not pay (back).
Due to the ban on ellipsis in conjunctive clauses, the identical subject must be
overtly expressed by means of a coreferential pronoun. Occasionally, the subject
noun phrase of the initial conjunct recurs in the following conjunctive clause. Such
verbatim repetitions have a stylistic flavour and underscore the truth value of each
individualconjunct.
(53) Binary coordination with same-subject conjunctive clause and repetition
of the identical subject
ere hen-psykh
na-ti
hu awf nte hen-psykh
rel indef.pl-soul fut-give.nom profit and conj indef.pl-soul

fk
e-rat=f
m-p-nute
hn-to nim
come.abs to-foot=3m.sg prep-def.m.sg-god in-purity each
e-rat=f
m-p-nute
hn-to nim
to-foot=3m.sg prep-def.m.sg-god in-purity each
The souls will benefit and the souls will come to God in all purity.
The coreferential subject pronoun in the conjunctive clause is anaphorically related

to the subject of the preceding clause, even when that subject is left unexpressed.
Implicit subjects are semantically implied by predication, but not overtly encoded in
the syntax. A case in point is mixed coordinative structures with an imperative clause
as the initial conjunct. The pronominal subject of the conjunctive verb in the second conjunct recovers the number and gender features of second person imperative
subject (Steindorff 1951:172 paragraph 369; Shisha-Halevy 1986:192193 paragraph
7.2.1.1.1; Layton 2000:279 paragraph 352; see also Zanuttini 2008).
Chris H. Reintges
(54) Mixed coordination with an initial imperative and same-subject conjunctive

clause
fk
-tok=k
e-u-son
come.imp conj.2sg.m-attach.pron=2sg.m to-indef.sg-brother
ef=r
hote n-ht=f
m-p-nute
rel(-pres)=3m.sg-make.nom fear in-heart=poss.3m.sg prep-def.m.sg-god
Go attach yourself to a god-fearing brother!
Cosubordinating constructions comprising an initial infinitival and a non-initial conjunctive conjunct are very common. The understood subject of an infinitival clause
may have a generic interpretation, corresponding to an impersonally used third person plural pronoun in the normal course of events. In (55), on the other hand, the
generically interpreted infinitival subject is rendered by a second person singular masculine pronoun in the same-subject conjunctive clause.
(55) Mixed coordination with an initial infinitival clause and same-subject
conjunctive clause
(Shenoute, Chass. 104, 2831)
u-noe
pe
wfm
p-oik
n-u-rfme
indef.sg-sin cop.m.sg eat.nom def.m.sg-bread link-indef.sg-man
-tm-r
pe=f-hf
conj.2m.sg=do.nom def.m.sg=poss.3m.sg-thing
It is a sin to eat a persons bread and not do (lit. and you do not do) his work.
The reverse situation does not occur, where the first conjunct contains a pronominal
subject that is cataphorically related to a subject noun phrase in the second conjunct.
The exclusion of cataphoric reference provides prima facie evidence for the essentially coordinative nature of the conjunctive verb conjugation (Roberts 1988:5657;
Haspelmath 1995:25; Kwon & Polinsky 2008: 9193).
5.1.2 Different-subject conjunctive clauses

Despite their ample attestation in series of subject-identical clauses, conjunctive
verbs are commonly used to connect different-subject clauses. Very often, differentsubject conjunctives involve a change in grammatical role: the referent of the different subject has already been introduced in the initial conjunct, albeit with a less
prominent grammatical role (Layton 2000:279 paragraph 352; Reintges 2004a:298300
paragraph8.1.4.2). The upgrading of referents is exemplified in (56), where the main
discourse referent Apa Mena is encoded as a possessor noun in the initial conjunct and
aligned with the subject function in the following conjunctiveclause.
(56) Binary coordination with different-subject conjunctive clauses and

upgrading of a discourse referent
e=nti
mmo-f
e-p-topos
n-Apa
Mna
rel(-pres)=1pl-carry.abs prep-3m.sg to-def.m.sg-shrine link-Apa Mena
aru n=f-kharize
na=f
m-p-talko
maybe conj=3sg.m-give for=3m.sg prep-def.m.sg-healing

We are bringing him (the possessed boy) to the shrine of Apa Mena and
perhaps he will grant him healing.
In binary coordination structures where the conjuncts have entirely parallel syntactic
structures, the different subject is contrastively focused.
(57) Binary coordination with structurally parallel conjuncts and contrastive
interpretation of the different subject
(Psalm 29, 5)
p-rime
na-fpe
e-ruhe
def.m.sg-weeping fut-become.abs at-sunset
nte p-tell
fpe
e-htfwe
conj def.m.sg-rejoic become.abs at-dawn
The weeping will happen at sunset and the rejoicing at dawn.
In different-subject contexts, the spreading of the illocutionary force from the chaininitial clause to the conjunctive clause is blocked. Consequently, the resulting construction is not interpreted as a coordinate structure, but instead as a complex sentence
with an adverbial clausal modifier (Schachter 1977:99100). Example (58) comprises
a sequence of a question followed by a declarative conjunctive clause with a purposive
interpretation.
(58) Mixed coordination with an initial constituent question and a different-subject
conjunctive clause (Besas Life of Shenoute, [Pap. Brit.Mus. 10820, folio 2 rct.])
e=n-na-kjn
rfme tfn
n-te=k-he
rel=1pl-fut-find.nom man where link-def.f.sg=2m.sg-kind
n=f-ti
so
na=n n=f-trypha
conj=3m.sg-give.nom teaching to=1pl conj=3m.sg-nourish.abs
mmo=n n-ne-graph
et-waa ?
prep=1pl with-def.pl-scripture comp.rel-pure.stat
Where shall we find someone of your kind to teach us and nurture us with
the holy scriptures?
As conjunctive clausal chains grow in complexity, consecutive action is less likely to

be attributed to a single discourse participant. At some point in the clausal chain,
a different-subject conjunctive introduces a topic switch without interrupting the
sequencing of events as such. Example (59) features an above-average chain of eleven
Chris H. Reintges
conjunctive clauses, which encompass an entire thematic paragraph (Longacre 1979).

The contextual parameter under which the series of events is grouped together is a
catalogue of precautionary measures for launching an imperial mission to the Holy
Land. The presentation of narrative information (a temporally ordered sequence of
events) is interwoven with itemized information (lists). (SSC and DSC stand for samesubject and different-subject conjunctives, respectively.).
(59) Longer-than-average conjunctive clause chain corresponding to an entire
discourse paragraph
(Eudoxia 62, 418)
fk
hn-u-kep
go.imp in-indef.sg-hurry
Go quickly
(SSC) -sote
na=s
n-ftow
n-stratlats
conj.2m.sg-prepare.abs for=3f.sg prep-four link-general
and make ready for her four generals
mn-ftou n-ta
m-matoi
mn-ftow n-harma
with-four link-ten.thousand link-soldier with-four link-chariot
and forty thousand soldiers and four chariots
n-ne-stratlats mn-omnt n-t-rrf
for-def.pl-general with-three for-def.f.sg-queen
for the generals and three for the queen,
mn-u-nu
e-naf=f
eol hm-p-i
with-indef.sg-gold rel(pres-)-be.plenty=3m.sg pcl from-def.m.sg-house
with much gold from the house
m-pe=ktoeis
mn-hn-eiaat
n-hat
link-def.m.sg=poss.2m.sg-lord with-indef.pl-linen link-silver
of your lord and some silver linen,
(SSC) -taa=u
na=s
e-tre=sto
eol
conj.2m.sg-give.pron=3pl for=3f.sg to-caus.inf=3sg.f-spend.abs pcl
and give them to her to spend
e-p-topos
n-pe=ntoeis
Jsus Khristos ()
on-def.m.sg-place link-def.sg.m=poss.1pl-lord Jesus Christ
on the place of our Lord Jesus Christ ()
(DSC) nte ftow n-toi
moone
ha-t-h
n-ftou
conj four link-ship harbor.abs in-def.f.sg-front link-four
and four ships shall lie in port before four
(DSC) nte ftow fpe
hn-thalassa
conj four become.abs on-sea
and four (ships) shall lie on sea
ant=stfk
eol n-hf
nim
term=3f.sg-complete.abs pcl prep-thing each
until she accomplishes everything
ent-a
pe-toeis
too=u
na=s
rel-perf def.sg.m-lord say.pron-3pl to=3f.sg
that the Lord told her
m-p-hun
n-saf
n-hou
in-def.sg.m-inside link-seven link-day
within seven days
(SSC) n=se-moone
e-t-Hierusalm
conj=3pl-harbor.abs at-def.sg.f-Jerusalem
and they shall land at Jerusalem
(DSC) -smane
m-pe=n-eift
n-arkhiepiskopos
conj.2m.sg-notify.abs prep-def.m.sg=poss.1pl-father link-archbishop
and you shall notify our father the bishop
(DSC) n=f-synage
n-t-polis
tr=s
m-p-ow
conj=3sg.m-assemble
prep-def.f.sg-city entire=3f.sg today
that he assemble the entire city today,
eol te
a
p-rro
pe-Khristos
ei
pcl comp perf def.m.sg-king def.m.sg-Christ come.abs
because the king Christ came
e-te=n-polis
n-tei-u
to-def.f.sg=poss.1pl-city in-dem.f.sg-night
to our city in this night,
anok hff=t
ti-nu
e-prosphora
I
reflex=1sg (pres=)1sg-go.stat to-service
while I myself will go to the service
(SSC) nta-synage
conj.1sg-worship.abs
to worship;
(DSC) mnnsa nai
de tr=u
-r
after
dem.pl pcl entire=3pl conj.2sg.m-make.nom

pai
hn-u-spud
dem.sg.m in-indef.sg-haste
in addition to all these (orders) you shall do this swiftly,
(SSC) n-orp e-nai
tr=u
-tow
at-first at-dem.pl entire=3pl conj.2sg.m-send.abs
before all these you shall first send out
n-u-pentkontarkhos
mn-pe=f-taju
m-matoi
eol
prep-indef.sg-commander.of.fifty with-def.m.sg-fifty link-soldier pcl
a commander-of-fifty and his fifty soldiers
hn-t-hi
m-p-rro
in-def.f.sg-road link-def.sg.m-king
on the kings road
Chris H. Reintges
(DSC) n-orp n=f-smane

n-m-polis
mn-n-time
at-first conj=3m.sg-notify.abs prep-def.pl-city with-def.pl-village
and he shall notify the cities and villages
tekas nne=u-tarasse ()
comp neg.fut=3pl-worry.abs
that they not be troubled ().
The coordinative interpretation re-emerges as the default of same-subject and differentsubject conjunctive clauses, connecting different informational layers with one another.
5.1.3 Syntactic variation in conjunctive clauses

Despite their dependency on the controlling verb in the initial conjunct, conjunctives are by no means structurally deficient clauses. Accordingly, they permit the same
range of syntactic variation as independent clauses with complete tense/aspect specifications. The applicability of topicalization in conjunctive clausal chains is shown in
(60), where the left-dislocated direct object NP n-ekklsia the churches appears to the
right of the conjunction awf and to the left of the conjunctive verb.
(60) Unbounded dependency between left-dislocated and resumptive pronouns in a
conjunctive clause chain
(Eudoxia 40, 810)
alla e=u-e-orr
n-n-rpwe ()
but rel=3pl-prep-destroy.abs prep-def.pl-temple.pl
n=se-kaa=u
m-ma n-rm
conj=3pl-put.pron=3pl as-place link-urine
awf n-ekklsia
n=se-ko=u
and def.pl-church conj=3pl-build.pron=3pl
n=se-fi
prosphora nht=u m-pei-hou
pai
conj=3pl-carry.nom Eucharist in=3pl on-dem.m.sg-day dem.m.sg
But they shall destroy the temples () and turn them into latrines and they
shall build the churches and celebrate the Eucharist in them on this very day.
Left-dislocated topics enter into an unbounded anaphoric dependency with resumptive pronouns across two clause boundaries. Left-dislocated personal pronouns and
demonstratives function as contrastive topics and are restricted to same-subject conjunctive clauses. Thus, consider (61).
(61) Different-subject conjunctive clauses with a left-dislocated free pronoun

(Shenoute, Aml. I.3, 368, 1011)
e-tre=f-fpe
eol nht=ttn
to-caus.inf=3m.sg-become.abs pcl within=2pl
awf ntftn nte=tn-ti
naake mmo=f
and you.pl conj=3pl-give.nom pains prep=3m.sg
so that he (Satan) grows inside you and you are in travail of him
The topic position in conjunctive clauses may also be occupied by preposed adverbial
modifiers, which connect a series of events to a specific temporal or spatial frame.
(62) Adverb preposing in conjunctive clause chains
(Eudoxia 62, 1618)
mnnsa nai
de tr=u
-r
pai
after
dem.pl pcl entire=3pl conj.2sg.m-make.nom dem.sg.m
hn-u-spud
n-orp e-nai
tr=u
-tow
in-indef.sg-haste at-first at-dem.pl entire=3pl conj.2sg.m-send.abs
n-u-pentkontarkhos
mn-pe=f-taju
m-matoi
eol
prep-indef.sg-commander.of.fifty with-def.m.sg-fifty link-soldier pcl
hn-t-hi
m-p-rro ()
in-def.f.sg-road link-def.sg.m-king
In addition to all these (orders), you shall do this swiftly, before all these you
shall first send out a commander-of-fifty and his fifty soldiers on the kings
road ().
When the nominal subject corresponds to new information focus, it may undergo
subject-verb inversion and is placed at the end of the clause chain.
(63) Subject-verb inversion in conjunctive clauses
(Shenoute, Aml. I 3 402, 35)
a m-ma
n(e) et=f-na-parage
mmo=u
what link-place cop.pl comp.rel(-pres)=3m.sg-fut-pass.by.abs prep=3pl
n=f-fk
ehun ero=u
n=f-tm-n
or conj=3m.sg-go.abs pcl prep=3pl conj=3m.sg-neg.aux-bring.nom

p-ose
n-te=f-psykh
mayaa=f
def.m.sg-damage to-def.f.sg=poss.3m.sg-soul reflex=poss.3m.sg
nki p-at-na
m-matoi ?
foc def.m.sg-neg.pfx-merciful link-soldier
Which are the places such that the merciless soldier would pass them by
or enter them and not bring damage to his own soul?
Although conjunctive clauses have intact left and right peripheries to host displaced
constituents, they exhibit some reluctance towards pragmatically driven reordering
processes, such as topicalization and subject-verb inversion. Presumably, this tendency originates from the prevalence of same-subject conjunctives with anaphorically
linked subject pronouns.
5.1.4 Modification by interclausal connectives

To specify a particular relation vis--vis the preceding discourse, conjunctive verbs
are regularly modified by a broad range of native and borrowed interclausal connectives, with the effect that the resulting hybrid construction displays features of both
co-ranking and clause chaining structures.
Chris H. Reintges
coordinative coordination. Conjunctive clauses share with standard symmetric coordination the frequent occurrence of the linker awf and. Its presence
in conjunctive clauses is far from being redundant, even though the semantic
distinctions between syndetic and asyndetic conjunctives are subtle and call for
closer scrutiny in future research. It looks though as if awf-conjunctives are not
intended to simply narrate events, since they often occur in binary coordinations
where temporal and causal relations are not at issue. In (64) the awf-coordinated
clauses are syntactically alike, but semantically distinct in that the first conjunct
ascribes to the subject referent a particular property, while the second conjunct
attributes a particular activity to it.
(64) Bipartite coordination with awf and conjunctive verbs
(Psalm 36, 21)
are p-dikaios
n
ht=f
hab def.m.sg-righteous be.pitiful.nom heart=poss.3m.sg
awf n=f-ti
and conj=3m.sg-give.abs
The righteous one is merciful and gives.
The repetition of the coordinating conjunction awf between three or more conjuncts
conveys a strong sense of focus, which is not present in bipartite coordinations (Payne
1985:5). Thus, consider (65), in which the conjoined clauses are ranked on a pragmatic scale of the speakers disapproval or outrage at the actions reported therein.
(65) Multiple coordination with awf and conjunctive verbs
(John 10, 10)
p-ref-tiwe

me=f-ei
ete-laau
def.m.sg-nominal-steal neg.hab=3m.sg-come.abs for-anything
eimti tekaas e=f-e-hfft
awf n=f-fft
unless comp rel-3m.sg-prep-steal.abs and conj=3m.sg-slaughter.abs
awf n=f-tako
and conj=3m.sg-destroy.abs
The thief does not come for anything else but to steal and to slaughter and
to destroy.
Disjunctive coordination. Coptic provides a good example of a language where

the transfer of borrowed function words is widely used to explicitly signal the semantic relation between two or more clauses in a chain, which would otherwise have to
be inferred from the surrounding context (Thompson & Longacre 1985:204205).
The Greek conjunction or is commonly used to express a disjunctive relationship between two or more clauses (Layton 2000:283 paragraph356). (66) is an
instance of an open disjunction, in which the listed alternatives do not exhaustively
cover all the possibilities (unlike the more standard closed conjunction which
generally implies that the two alternatives are mutually exclusive) (Payne 1985:24;
Huddleston, Payne & Peterson 2002:12931297 paragraph 2.2.1).
(66) Open disjunction with the Greek conjunction and conjunctive verbs

r-an
u-thlipsis
taho=u
rel-cond indef.sg-affliction rise.pron=3pl
7 nte u-kindynos
tfwn etf=u ()
or conj indef.sg-danger rise.abs upon=3pl
If an affliction comes upon them (the hermits) or a danger rises up
against them ().
The correlative counterpart either or tends to emphasize the equality of

the conjuncts, which must be considered separately. Normally, correlative disjunction strengthens the exclusiveness implicature associated with closed disjunction,
meaning that only one alternative is considered to hold true. However, the open disjunction type is by no means excluded. In (67) the conjoined clauses are semantic
paraphrases of each other and the states of affairs reported therein must therefore both
obtain (Huddleston, Payne & Peterson 2002:13061307 paragraph 2.3; Haspelmath
2004:2425, 2007:1718).
(67) Open disjunction with correlative and conjunctive verbs
kom
(Matthew 6, 24)
n-toeis
mn
n-laau
e-r
hmhal
snau
not.be power link-someone to-make.nom servant link-lord two
7 gar f-na-meste
wa n=f-mere
wa
or pcl 3m.sg-fut-hate.nom one conj=3m.sg-love.abs one
7 n=f-kolt=f
n-wa n=f-kataphronei
or conj=3m.sg-entangle.pron=3m.sg to-one conj=3m.sg-despise.abs
m-p-ke-wa
prep-def.m.sg-other-one
It is not possible for anyone to become servant of two lords, because he either
hates one and loves the other or adheres to one and despises the other.
Adversative coordination. The Greek conjunction alla but has adversative

meaning, indicating a contrast between two coordinands. Contrary to awf, alla is
restricted to binary coordination (Huddleston, Payne & Peterson 2002:13121314
paragraphs 2.62.7). Very often, the contrastive relation in alla-coordinated clauses
involves a switch from positive to negative polarity, and vice versa.
Chris H. Reintges
(68) Adversative coordination with the coordinator alla and a conjunctive verb

(2 Timothy 2, 24)
u-hmhal
nte-p-toeis
n-e
an
indef.sg-servant link-def.m.sg-lord neg(-pres)-is.proper not
ero=f
e-mie
alla n=f-fpe
for=3m.sg to-fight.inf but conj=3m.sg-become.abs
e=f-thiu
nnahrn-won nim
rel(-pres) =3m.sg-be.humble.stat before-one
each
(As for) a servant of the Lord, it is not appropriate for him to fight, but he
should rather be humble towards everyone
Interclausal topic shift. Conjunctive clauses may be modified by rhetorical

conjunctions that specify a particular discourse relation. A case in point is the
Greek enclitic particle de, which signals the discontinuity of referents and actions
(Reintges 2004b: 7981). In (69) the newly introduced discourse participant is the
inverted subject the laymen or the impious ones.
(69) The topic-shifting particle de in conjunctive clause chains
(1 Corinthians 14, 23)
efpe ke eran t-ekklsia

tr=s
ei
if
pcl cond def.f.sg-church entire=poss.3f.sg come.abs
e-u-ma
n-wft
n=se-ate
tr=u
to-indef.sg-place link-one conj=3pl-speak.abs entire=poss.3pl
hn-u-aspe
n=se-ei
de ehun nki
in-indef.sg-tongue conj=3pl-come.abs pcl pcl foc
hen-idifts
hen-apistos
ne=u-na-too=s
indef.pl-laymen or indef.pl-impious neg-rel=3pl-fut-say.pron=3f.sg
an te
e=tetn-lobe ?
not comp rel(-pres)=2pl-be.mad.stat
If the entire church came together at a single place and spoke in tongues and the
laymen and the impious ones came inside, would they not say you are mad?
It generally appears, then, that the modification of conjunctive clauses by various

native and borrowed interclausal connectives is no less free than in standard coordinate patterns of the language.
5.1.5 Marked coordinative uses

In this section I consider two marked uses of the conjunctive conjugation. In the first
case, the initial conjunct is not an independent sentence, but rather a subordinate conditional clause. In the second case, a conjunctive verb appears without a controlling
verb in chain-initial position.
Conjunctive verbs occurring in consequence clauses. Conditional constructions of the type if p then q are expressed by a hypothetical clause (the protasis)
and a full consequence clause (the apodosis). What is common to all constructions
fitting the if p then q schema is that they convey a consequence implicature, asserting that the situation described in the apodosis follows in one way or the other
from the situation described in the protasis clause. In conveying a conditional
implicature, the conjunctive can be used as an apodosis tense, continuing a protasis
clause in the conditional mood (Steindorff 1951:173 paragraph 370; Shisha-Halevy
1986:206207 paragraph 7.2.6.1; Reintges 2004a: 307306 paragraph 8.1.4.5).
(70) Conjunctive verbs occurring in consequence clauses of conditional

constructions
(Budge, Misc. Texts 441, 32442, 1)
awf e=n-an-ei
eol e-p-tow
and rel=1pl-cond-come.abs pcl to-def.m.sg-mountain
et-hiol
n=tn-synage
mn-ne-snu
comp.rel-outside conj=1pl-gather.abs with-def.pl-brothers
hm-p-saton
mn-t-kyriak
on- def.m.sg-Saturday with-def.f.sg-Sunday
And whenever we come out to the outer mountain, we celebrate Mass
with the brothers on Saturday and Sunday.
Conditional sentences with conjunctive clauses are clearly different from conditionally interpreted coordinating constructions, since the initial conjunct has not only the
literal meaning, but also the adverbial-subordinative form of a protasis clause (see
above, Section3.2.3).
Chain-initial conjunctives. As a cosubordinating nexus type, conjunctive

verbs are restricted to the non-initial conjuncts of a clause chain. There are, however, some exceptions to this distributional restriction, with a conjunctive verb
appearing in absolute clause-initial position. According to some scholars (Stern
1880: 276277 paragraph 446; Steindorff 1951:174 paragraph 371), such cases of
chain-initial conjunctives actually involve the deletion of a verb of volition and
intent, as seen in (71).
(71) Conjunctive verbs in initial position
anok ho
ta-agonize
mn-ne-judai
n-ke-kui
I
reflex.poss.1sg conj.1sg-fight with-def.pl-Jews in-other-little
I myself (want to) fight with the Jews a little more!
Negated conjunctives may be coordinated with the exclamative particle hamoi would
that, which expresses the speakers fear and regret that the actualisation of the event
at hand is foreclosed for good (Shisha-Halevy 1986: 211 paragraph 7.3.3; Layton
2000: 282 paragraph 354; cf. also Palmer 1986: 119121 for relevant discussion on
evaluative particles).
Chris H. Reintges
(72) Negative conjunctive verbs occurring with the exclamative hamoi

(Shenoute, Aml. I 231, 12)
awf hamoi nte-tm
pei-ate
tfmnt
and pcl
conj-neg.aux dem.m.sg-word apply.abs
e-hah
nht=n
to-many among=1pl
And would that this saying not apply to many of us!
Not only do chain-initial conjunctives complete elliptic material, they also respond to
or echo conversational implicatures. A particularly clear case is the rhetorical question
Will salvation come to many? in (73), which challenges the implicature conveyed in
the preceding context, namely that everyone hopes (and expects) redemption for his
nearest and dearest.
(73)

Chain-initial conjunctives elaborating on a conversational implicature

(Shenoute, Chass. 167, 56168, 9)
malista hen-eiote
kata-pei-bios
especially indef.pl-fathers according.to-dem.m.sg-life
e=u-we
p-to
n-ne=u-re
rel(-pres)=3pl-want.nom def.m.sg-purity for-def.pl=poss.3pl-son
n-tei-he
on pe
u-son
n-u-son
in-dem.f.sg-manner pcl cop.m.sg indef.sg-brother for-indef.sg-brother
awf u-sfne
n-u-sfne
and indef.sg-sister for-indef.sg-sister
ara nte u-utai
fpe
n-hah ?
q conj indef.sg-salvation become.abs to-many
Especially, biological parents wish purity for their children, and, in the same way, a
brother for a brother and a sister for a sister. And so, will salvation come to many?
It generally appears, then, that conjunctive clauses are not always related to the actual
utterance they continue, but may also be related to the contextual implicatures associated
with the utterance (see Ifantitidou 2001:197 for a related view on evidential particles).
5.2 Subordinative nexus

As the languages most versatile clause combining and chaining device, the conjunctive covers a broad spectrum of semantic relations between clauses, crossing
the traditional distinction between coordination and subordination. Being in an
embedding relationship with an independent sentence, subordinative conjunctive
clauses display an overall structural similarity with standard subordinate clauses.
Accordingly, the entire clausal chain is interpreted as a single complex sentence.
Despite this structural uniformity, there are clear syntactic and semantic differences
between subordinate conjunctive clauses that concern the level of embedding: core
versus periphery.
5.2.1 Conjunctive clauses used as core argument clauses

According to Dixon (2006:39), clause chaining can only occasionally be exploited as
a complementation strategy. The Coptic conjunctive fits into this picture. Although it
represents an essentially cosubordinating nexus type, it is by no means excluded from
various types of subordinate contexts. A case in point is the complement construction of indirect speech, as exemplified in (74). The complement clause comprising
the reported speech is extraposed to the end of the main clause, while its direct object
status is indicated by the impersonal 3rd person feminine singular object pronoun =s
it, which is attached to the quotative verb tf say.
(74) Binary coordination with a conjunctive verb occurring as an indirect
speech complement
a=u-too=s
ete-wa
n-ne-snu
perf=3pl-say.pron=3f.sg about-indef.sg-one link-def.pl-brother.pl
te
a=f-tamio
n-n-ir
n=f-ti
comp hab=3m.sg-produce.abs prep-def.pl-basket conj=3m.sg-give.nom
maate ero=u
handle to=3pl
They said about one of the brothers that he made baskets and gave those handles.
What interests us here are clausal chains in which the controlling verb and the conjunctive verb are separated by a clause boundary. In a syntactic variant of the complement construction of indirect speech, the conjunctive clause is extraposed to the right
edge of the reportative clause. This is, indeed, the preferred position for heavy constituents. The absence of the finite subordinating complementizer te that is a characteristic feature of this construction.
(75) Indirect speech complement formed with a conjunctive verb (Till, KHML I 8, 67)
ta-sfne
ete-u mper-too=s
def.f.sg-sister for-what neg.perf.2f.sg-say.pron=3f.sg
n-ne=s-eiote
n=se-ei
nmma=s ?
to-def.pl=poss.3f.sg-father.pl conj=3pl-come.abs with=3sg.f
My sister, why did you not tell her parents to accompany her?
Prima facie evidence for the conjunctive as a marked alternative for clausal complementation comes from its restricted lexical distribution. Verbs of perception and
discovery, for instance, never select conjunctive complement clauses and are almost
exclusively used with converbal relative tenses. More importantly, however, for each
class of complement-taking verbs that is compatible with subordinate conjunctive
Chris H. Reintges
verbs, it holds true that there is at least one, more common complementation strategy
(finite subordinate clauses, infinitives).
Lexical classes of complement-taking verbs. Conjunctive clauses are particularly common with the diversified class of manipulative verbs, which encode
reported directives. In addition, members of this lexical class may provide information about the way in which the illocutionary act has been put forward as well
as the social status of the interlocutors (Noonan 1985: 125126). The Greekbased verb keleue to command, for instance, is used when the reported directive
is issued by a socially superior person, whereas the native verb while sops to
beseech designates a plea from a socially inferior individual to a superior one.
The group of manipulative verbs also includes the closely related causative and
permissive verbs like kf to let..
(76) Conjunctive direct object clauses to manipulative verbs

a.
The permissive verb kf to let
ka
n-re
m n=se-ei
place.imp def.pl-child little conj=3pl-come.abs
e-rat
awf mper-kfly
mmo=u
to-foot.poss.1sg and neg.imp-stop.abs prep=3pl
Let the little children come to me and do not hinder them!
b. The manipulative verb keleue to command
(Luke 18, 16)
awf a=f-keleue
n=se-ent=f
and perf=3sg.m-demand.abs conj=3pl-bring.pron-3sg.m
etm-p-ma
to-def.m.sg-tribunal
And he (the governor) ordered (that) they bring him (Apa Nahrow)
to thetribunal.
c.
The manipulative verb sops to beseech
(Hilaria 8, 1819)
mare=n-sops
n=f-kharize
opt=1pl-entreat.abs prep-def.m.sg-lord conj=3m.sg-grant.abs
m-p-talkjo
n-te-eere
m-p-rro
prep-def.m.sg-healing link-def.f.sg-daughter link-def.m.sg-king
Let us entreat the Lord that he grants healing to the Kings daughter!
m-p-toeis
With verbs of volition, intent or desire, conjunctive verbs may alternate with infinitives. Members of this class are characterized by an experiencer subject and a complement proposition that is asserted to be realizable in the nearby future (Noonan
1985: 121125; Dixon 2006: 31). With first person singular subjects, volitional and
desiderative verbs designate a promissive speech-act where the speaker commits himself to a particular action.
(77) Conjunctive direct object clauses to desiderative verbs (Mena, Miracles 10a, 1015)
anok gar e=i-wf
nta-pft
i
pcl rel(-pres)=1sg-want.abs conj.1sg-depart.abs
6nta-pfht
eol m-pa-snof
conj.1sg-shed.abs pcl prep-def.sg.m.poss.1sg-blood
etm-p-ran
m-pa-toeis
Jsus pe-Khristos
on-def.m.sg-name link-def.m.sg.poss.1sg-lord Jesus def.m.sg-Christ
I want to depart to shed my blood in the name of my Lord Jesus Christ.
Conjunctive complement clauses are less common with verbs of knowledge and acquisition of knowledge, which convey a semifactive sense. As a rule, these verbs presuppose the truth of the complement clause (see Kiparsky & Kiparsky 1970 and much
related research).
(78) Conjunctive direct object clauses to verbs of knowledge

(Sahidic Vita of Pachomius 253a, 2528)
e=u-e-eime
n=se-agfnize
rel(-fut)=3pl-prep-know.abs conj=3pl-fight
e-tre=u-fpe
n-thysia
m-p-nute
to-caus.inf=3pl-become.abs prep-sacrifice for-def.sg.m-god
They shall know (that) they (the holy fathers) fought to become a
sacrifice forGod.
The impersonal subject construction. Subordinate conjunctive clauses may

function as the core argument clause of the impersonal subject construction. As
with the complement construction of indirect speech, the argument clause itself is
extraposed to the right periphery of the matrix clause. The preverbal subject slot
is occupied by the impersonal 3rd person feminine singular subject pronoun =s it.
(79) Conjunctive subject clauses of impersonal adjectival constructions

(Mena, Miracles 78a, 811)
nanu=s
nta-fk
e-pa-i
be.good.pron=3.f.sg conj.1sg-go.abs to-def.sg.m.poss.1sg-house
ere
u-laau
n-toot
rel(-pres) indef.sg-something in-hand.poss.1sg
It is better that I go home, while I have something.
Conjunctive clauses can be found in impersonal existential sentences, which are

formed with the inchoative verb fpe to become, happen.
Chris H. Reintges
(80) Conjunctive subject clauses of impersonal inchoative constructions

nne=s-fpe
mmo=i ta-wft
neg.fut=3f.sg-happen.abs prep=1sg conj.1sg-worship.abs
n-ne=k-nute
n-ote
prep-def.pl=poss.2sg.m-god link-abomination
It will never happen to me to worship your abominable gods.
In view of the fact that conjunctive subject clauses are always located in an extraposed
position, they do not have the same syntactic status as complement clauses, but rather
appear as an inner peripheral argument (Foley & Van Valin 1984:252).21
5.2.2 Conjunctive clauses used as adverbial clausal modifiers

In principle, any independent clause can be linked to an adverbially used conjunctive clause. With asyndetically linked conjunctive clauses, it is often difficult to decide
whether they are syntactically specified as adverbial clauses or whether they receive a
purposive reading from the context (see also Thompson & Longacre 1985:175176).
Ambiguity between coordinative and purposive interpretations. The

semantic ambiguity of asyndetic conjunctives can conveniently be illustrated with
chaining structures where the independent clause contains a motion verb. Under
a coordinative interpretation the conjunctive verb simply describes a successive
event. Under the equally possible purpose reading, it expresses a motivating
. In contrast to direct object clauses, extraposed subject clauses have independent left
peripheries, as the following example of contrastive pronoun topicalisation illustrates.

(i)
Conjunctive subject clauses with left-dislocated pronouns (Testament of Isaac 237, 14)
awf e=s-an-fpe
anok ta-r
rfme
and rel=3f.sg-cond-happen.abs i
conj.1sg-do.nom man
ta-mu
ta-tfwn
eol hm-n-et-mowt
conj.1sg-die.abs conj.1sg-stand.up.abs pcl from-def.pl-comp.rel-die.stat
m-p-meh-omnt n-hou ti-na-tre

won nim
on-ordinal-three link-day 1sg-fut-caus.inf one each
er
p-meewe
m-pe=tn-ran
do.nom def.m.sg-remembrance link-def.m.sg=poss.2pl-name
n=se-epikalei
mmf=tn na=u
n-eift
conj=3pl-call.upon.abs prep=2pl for=3pl as-father
And when it happens (that) I become human and die and rise from the dead on
the third day, I will cause everybody to remember your name and call upon you
for them as a father.
event that is unrealized at the time of the main event (see also Layton 2000:278
paragraph 352 (b) N.B.).
(81) Semantically ambiguous conjunctive clauses
ude
mpe=i-ei
e=i-na-fk
e-u-ma
eneh
and.not neg.perf=1sg-go.abs rel=1sg-fut-go.abs to-indef.sg-place ever
u-sowhs
ta-too=s
or indef.sg-congregation conj.1sg-say.pron=3f.sg
hfs e-wnta=i
t-eksusia
te
ma
comp rel(-pres)-have=1sg def.f.sg-authority comp give.imp
na=i n-u-eif
ta-ale
ero=f
to=1sg prep-indef.sg-donkey conj.1sg-ascend.abs on=3m.sg
I never intended (lit. I never went) to go to a place or congregation and/to
speak as if I had authority Give me a donkey that I may ride on it.
In mixed coordinate structures, different-subject conjunctives are semantically interpreted as adverbial purpose clauses (see above, Section5.1.2.).
(82) Mixed coordination with an initial constituent question and a different-subject
conjunctive clause with purposive interpretation
(Luke 18, 18)
e=i-na-r
u
ta-klronomi
rel=1sg-fut-do.nom what conj.1sg-inherit.abs
m-p-fnh
a-eneh ?
prep-def.m.sg-life until-eternity
What shall I do to inherit eternal life?
Temporal and aspectual restrictions. To complicate matters even further,

there are also instances of conjunctive clauses in which the unavailability of a
coordinative reading is correlated with the tense/aspect of the controlling verb in
the initial conjunct. A particularly illustrative case is the present progressive constraint, which causes telic motion verbs to occur in the stative grade in present
tense sentences (Reintges 2004a: 216217 paragraph 6.2.3). A following conjunctive verb does not assume a present progressive interpretation, but rather functions
as an aspectual shifter, which provides an endpoint for the motion event that is
described in the initial present progressive clause. The most natural interpretation
of the same-subject conjunctive clause is that of an adverbial purpose clause.
(83) Same-subject conjunctive clauses following a stative verb in a present tense

sentence
(Mena, Miracles 27b:2325)
e=i-k
e-p-topos
rel (-pres)-1sg-come.stat to-def.m.sg-shrine
n-Apa Mna nta-ll
link-Apa Mena conj.1sg-pray
I am on my way to the shrine of Apa Mena to pray.
Chris H. Reintges
The conjunctive is incapable of providing a coordinative connection between two past

events in historical narratives such that the plot is advanced (Shisha-Halevy 1986:189
paragraph 7.1.2.4; Layton 2000: 277 paragraph 352b). Significantly, this does not
always trigger a purposive reading. Example (84) features a series of different-subject
conjunctive clauses. The particle de signals a thematic break, which is effected by the
shift both in narrative perspective (from discourse to dialogue) and from historical
past to present tense.
(84) Conjunctive clauses with a temporal shift from historical perfect to present
tense
(Shenoute, pap. Brit.Mus. 198, 82b: 2225)
a= u-net
p-Satanas
eol hn-t-pe
perf=3pl-throw.nom def.m.sg-Satan pcl from-def.f.sg-heaven
te
a=f-kfte
nsa-t-he
n-r
comp perf=3m.sg-seek.abs after-def.f.sg-manner link-do.nom
ne=f-ote
-aa=u
de
def.pl=poss.3m.sg-scandal conj.2m.sg-do.pron=3pl pcl
hff=k
awf -tok=u
reflex=poss.2m.sg and conj.2m.sg-complete.pron=3pl
hn-n-paraphysis
n=se-tit=k
ehun emau ?
in-def.pl-monstrous.growth conj=3pl-take.pron=2m.sg pcl there
Satan was cast out (lit. they cast out) of heaven, because he sought a way to
perform his (sexual) abominations and you yourself perpetrate them in sin
against nature, would you then be received (lit. would they receive you) inside
there (i.e. heaven)?
Disambiguation by final/purposive complementizers. The polarity-sensitive Greek final-purposive complementizers hfste, hfs such that, in order
that and mpfs, mpote that not, lest are widely used to explicitly signal the
adverbial-subordinative status of the following conjunctive clause (Shisha-Halevy
1986: 209 paragraph 7.3.1.1; Layton 2000: 281 paragraph 354). The affirmative
complementizers hfste and hfs take factive complements, stating the actual result
of the main clause event.
(85) Conjunctive clauses introduced by hfs such that and (affirmative) polarity
agreement
a=f-pfne
hfs n=f-r
perf=3m.sg-change.abs comp conj-3sg.m-make.nom
t-he
n-ke-rfme
def.sg.f-manner link-other-man
He changed such that he became like another man.
The negative counterparts mpfs and mpote introduce negative purpose clauses, in
which the superordinate clause typically expresses the idea that precautions are taken
so as to avoid an apprehension-causing situation in the negated purpose clause (see
Thompson & Longacre 1985:188; Lichtenberk 1995:298). Occasionally, the all-purpose
complementizer te may be stacked on top of the negative complementizer mpote.
(86) Conjunctive clauses introduced by the stacked complementizers te and mpote
that and (negative) polarity agreement
a=itoo=s
te
mpote n=se-kf
perf=1sg-say.pron=3f.sg comp comp conj=3pl-remain.abs
e=u-ti
woi
hn-t-heneete
rel(-pres)=3pl-give.nom course in-def.f.sg-monastery
I said it so that they (the monks from outside) not continue walking
around in the monastery.
There exists a high degree of functional overlap between coordinated and adverbially
subordinated conjunctive clauses. This suggests that co-subordinative and subordinative clause nexus involves the same peripheral level of structural embedding, namely
adverbial adjunction.
6. Conclusion
Coptic has a rich system of specialized syntax and morphology for the encoding of coordinate constructions of various kinds. It falls squarely within the languages of the African
continent, in which the conjunction of noun phrases is morphologically distinguished
from the conjunction of verb phrases and clauses. Nominal coordination is sensitive to
the phrasal status of the conjuncts: the comitative strategy applies only to phrasal coordination, connecting in/definite noun phrases and personal pronouns, while an originally
locative preposition is operative at the sub-phrasal level, connecting bare nouns.
The system of interclausal relations presents a picture of great diversity of symmetric and asymmetric patterns. Clause coordination by means of the linker awf
and is generally symmetric in terms of syntactic structure, but allows for asymmetric
temporal and consecutive readings in much the same way as the coordinating conjunctions in the well-studied European languages. In the domain of asymmetric VP/
clause coordination, the language has at its disposal converbal relative tenses and the
conjunctive. Both dependent verb conjugations are of considerable typological interest, since they bear a close relation to information structure. Coordinative converbs
belong to the paradigm of relative tenses, which represent a special type of inflectional
morphology that flags a broad range of focus-sensitive sentence constructions and sets
them apart from pragmatically neutral declarative clauses.
Chris H. Reintges
Thematic paragraphs formed with same-subject conjunctive clauses are maximally cohesive: the presence of conjunctive morphology on all chain-medial and
chain-final verbs signals action continuity, while referential continuity is encoded by
means of identical subject pronouns. As conjunctive clause chains grow in length, the
series of events described in them is less likely to be attributed to a single discourse
referent. Conjunctive clauses with different subjects interrupt referential continuity
without necessarily disrupting action continuity. Switch-reference in coordinate structures with only two clauses has observable semantic effects in that different-subject
conjunctive clauses assume an illocutionary force different from that of the initial conjunct. Consequently, the construction is no longer interpreted as a coordinate structure, but rather as a complex sentence with an adverbial purpose or reason clause.
Clause chaining bears a close relation to information structure, although the
conjunctive itself cannot be identified with a special type of focus-sensitive morphology on a par with converbal relative tenses. Nonetheless, topic and focus prominence
triggers a departure from the canonical SVO order. As an essentially cosubordinative
nexus type, the conjunctive coordinates and connects clauses of equal status. However,
it is equally suitable for various types of subordination, in which one clause is embedded in or syntactically dependent on the other. The distinction between the coordinative and subordinative behaviour of conjunctive clause chains is not an all-or-nothing
property. Instead, asyndetically linked conjunctive clauses can often be interpreted
either way. In other instances, the purposive reading of conjunctive clauses becomes
available when a coordinative interpretation is excluded by the temporal or aspectual
specification of the controlling verb. To specify or disambiguate a particular interclausal relation, conjunctive clauses may be modified by a broad range of native and
borrowed connectives and subordinating conjunctions. The resulting hybrid pattern
has features of both a co-ranking and a chaining structure.
Abbreviations
1, 2, 3
first, second, third person
abs
absolute state grade
aug
augment
aux
auxiliary verb
caus.inf causative infinitive
comp subordinating
complementizer
comp.rel relative complementizer
cond
conditional mood
conj
conjunctive
cop
pronominal copula
def
dem
f
foc
fut
link
neg
neg.aux
neg.perf
neg.hab
nom
definite article
demonstrative article
feminine gender
focus marker
future
nominal linker
negation
negative auxiliary verb
negative perfect
negative habitual
nominal state grade
nominal
hab
imp
indef
inf
infer
m
perf
pcl
pl
nominalizing affix
habitual aspect
imperative
indefinite article
infinitive
inferred evidential
masculine gender
perfect tense/aspect
particle
plural
poss
prep
pres
pron
recip
reflex
rel
sg
stat
possessive pronoun
prepositional object marker
present tense
pronominal state grade
reciprocal noun
(emphatic) reflexive
relative marker
singular
stative grade
References
Aikhenvald, Alexandra Y. 2004. Evidentiality. Oxford: OUP.
Bisang, Walter. 1995. Verb serialization and converbs differences and similarities. In Converbs
in Cross-linguistic Perspective: Structure and Meaning of Adverbial Verb Forms Adverbial Participles, Gerunds [Empirical Approaches to Language Typology 13], Martin
Haspelmath & Ekkehard Knig (eds.), 137188. Berlin: Mouton de Gruyter.
Blakemore, Diane & Carton, Robyn. 2005. The pragmatics of sentential coordination with and
Lingua 115: 569589.
Bril, Isabelle & Rebuschi, Georges. 2007. Coordination, subordination et co-jonction: Faits linguistiques et concepts. In Coordination et subordination: Typologie et modlisation [Faits de
Langue 28], Isabelle Bril & Georges Rebuschi (eds.), 113. Paris: Ophrys.
Comrie, Bernard 1976. Aspect: An introduction to the study of verbal aspect and related problems
[Cambridge Textbooks in Linguistics]. Cambridge: CUP.
Comrie, Bernard. 1985. Tense [Cambridge Textbooks in Linguistics]. Cambridge: CUP.
Cristofaro, Sonia. 2007. Deconstructing categories: Finiteness in a functional-typological
perspective. In Finiteness: Theoretical and Empirical Foundations, Irena Nikolaeva (ed.),
91114. Oxford: OUP.
Crum, Walter E. 1939. A Coptic Dictionary. Oxford: Clarendon Press.
Culicover, Peter W. & Jackendoff, Ray. 1997. Semantic subordination despite syntactic coordination. Linguistic Inquiry 28: 195217.
Depuydt, Leo. 1993. Conjunction, Continuity, Contingency. On the Relationships between Events
in the Egyptian and Coptic Verbal Systems. Oxford: OUP.
Dik, Simon C. 1968. Coordination Its Implications for the Theory of General Linguistics. Amsterdam: North-Holland.
Dixon, Robert M.W. 2006. Complement clauses and complementation strategies in typological
perspective. In Complementation: A Cross-linguistic Typology [Explorations in Linguistic
Typology 3], Robert M.W. Dixon & Alexandra Y. Aikhenvald (eds.), 148 Oxford: OUP.
Dryer, Matthew S. 1992. The Greenbergian word order correlations. Language 68: 81138.
Dryer, Matthew S. 2005. The position of tense-aspect affixes. In The World Atlas of Language
Structures, Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.),
282285. Oxford: OUP.
Erman, Adolf. 1933. Neuaegyptische Grammatik. Leipzig: Engelmann.
Chris H. Reintges
Ernst, Ralph. 1994. NP-Koordination im Koptischen und Neugyptischen. Lingua Aegyptia 4:
89115.
Fabricius-Hansen, Cathrine & Ramm, Wiebke. 2008. Subordination and coordination from different perspectives (Editors introduction). In Subordination versus Coordination in Sentence and Text: A Cross-linguistic Perspective [Studies in Language Companion Series 98],
Cathrine Fabricius-Hansen & Wiebke Ramm (eds.), 130. Amsterdam: John Benjamins.
Fry, Caroline & Hartmann, Katharina. 2005. The focus and prosodic structure of German right
node raising and gapping. The Linguistic Review 22: 69116.
Foley, William A. this volume. Clause linkage and Nexus in Papuan languages, 2750.
Foley, William A. & Van Valin Jr., Robert D. 1984. Functional Syntax and Universal Grammar.
Cambridge: CUP.
Funk, Wolf-Peter. 1977. Zur Syntax des koptischen Qualitativs. Zeitschrift fr gyptische Sprache
und Altertumskunde 104: 2539.
Givn, Talmy. 1983. Topic continuity in discourse: The functional domain of switch reference.
In Switch-reference and Universal Grammar [Typological Studies in Language 2], John
Haiman & Pamela Munro (eds), 5182 Amsterdam: John Benjamins.
Givn, Talmy. 1990. Syntax: A Functional-typological Introduction, Vol. II. Amsterdam: John
Benjamins.
Haiman, John. 1983a. On some origins of switch reference marking. In Switch-reference and
Universal Grammar [Typological Studies in Language 2], John Haiman & Pamela Munro
(eds.), 105128 Amsterdam: John Benjamins.
Haiman, John. 1983b. Symmetry. In Iconicity in Syntax [Typological Studies in Language 9],
John Haiman (ed.), 7393 Amsterdam: John Benjamins.
Haspelmath, Martin. 1995. The converb as a cross-linguistically valid category. In Converbs in
Cross-linguistic Perspective: Structure and Meaning of Adverbial Verb Forms Adverbial Participles, Gerunds [Empirical Approaches to Language Typology 13], Martin Haspelmath
& Ekkehard Knig (eds.), 155. Berlin: Mouton de Gruyter.
Haspelmath, Martin. 2004. Coordinating constructions: An overview. In Coordinating Constructions [Typological Studies in Language 58], Martin Haspelmath (ed.), 339. Amsterdam: John Benjamins.
Haspelmath, Martin. 2005. Nominal and verbal conjunction. In The World Atlas of Language
Structures, Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.),
Haspelmath, Martin. 2007. Coordination. In Language Typology and Syntactic Description, Vol.
II: Complex Constructions, 2nd edn., 151. Cambridge: CUP.
Huddleston, Rodney. 2002. The verb. In The Cambridge Grammar of the English Language,
Rodney Huddleston & Geoffrey K. Pullum (eds), 71212. Cambridge: CUP.
Huddleston, Rodney, Payne, John & Peterson, Peter. 2002. Coordination and subordination. In
The Cambridge Grammar of the English Language, Rodney Huddleston & Geoffrey K. Pullum
(eds.), 12731362. Cambridge: CUP.
Jaggar, Philip J. 2001. Hausa [London Oriental and African language library 7] Amsterdam:
John Benjamins.
Johannessen, Janne Bondi. 1998. Coordination [Oxford Studies in Comparative Syntax]. Oxford:
OUP.
Johnson, Janet H. 1976. The Demotic Verbal System [Studies in Ancient Oriental Civilization 38].
Chicago IL: The Oriental Institute of the University of Chicago.
Ifantidou, Elly. 2001. Evidentials and Relevance [Pragmatics & Beyond New Series 86].
Kahle, Paul E. 1954. Balaizah Coptic Texts from Deir el-Balaizah, 2Vols. London: OUP.
Kiparsky, Paul & Kiparsky, Carol. 1970. Fact. In Progress in Linguistics: A Collection of Papers,
Manfred Bierwisch & Karl Erich Heidolph (eds.), 143173. The Hague: Mouton.
Knig, Ekkehard.1995. The meaning of converb constructions. In Converbs in Cross-linguistic
Perspective: Structure and Meaning of Adverbial Verb Forms Adverbial Participles, Gerunds
[Empirical Approaches to Language Typology 13], Martin Haspelmath & Ekkehard Knig
(eds.), 5795. Berlin: Mouton de Gruyter.
Kwon, Nayoung & Polinsky, Maria. 2008. What does coordination look like in a head-final
language? In Asymmetric Events [Converging Evidence in Language and Communication Research 11], Barbara Lewandowska-Tomaszczyk (ed.), 87102. Amsterdam: John
Benjamins.
Layton, Bentley. 2000. A Coptic Grammar with Chrestomathy and Glossary [Porta Linguarum
Orientalium N.S. 2]. 1st edn. Wiesbaden: Harrassowitz.
Lehmann, Christian. 1988. Towards a typology of clause-linkage. In Clause Combining in
Grammar and Discourse [Typological Studies in Language 18], John Haiman & Sandra A.
Thompson (eds.), 181225. Amsterdam: John Benjamins.
Lichtenberk, Frantisek. 1995. Apprehensional epistemic. In Modality in Grammar and Discourse
[Typological Studies in Language 32], Joan Bybee & Suzanne Fleischmann (eds.), 293327.
Longacre, Robert E. 1979. The paragraph as a grammatical unit. In Syntax and Semantics 12
Discourse and Syntax, Talmy Givn (ed.), 115134. New York NY: Academic Press.
Longacre, Robert E. 1985. Sentences as combinations of clauses. In Language Typology and Syntactic
Description, Vol. II: Complex Constructions, Timothy Shopen (ed.), 235286. Cambridge: CUP.
Loprieno, Antonio. 1995. Ancient Egyptian A Linguistic Introduction. Cambridge: CUP.
Mallon, Alexis. 1953. Grammaire copte. Beirut: Dar el-Machreq.
Mithun, Marianne. 1988. The grammaticalization of coordination. In Clause Combining in
Grammar and Discourse [Typological Studies in Language 18], John Haiman & Sandra A.
Thompson (eds), 331359. Amsterdam: John Benjamins.
Mous, Maarten. 2004. The grammar of conjunctive and disjunctive coordination in Iraqw. In
Coordinating Constructions [Typological Studies in Language 58], Martin Haspelmath
(ed.), 109122. Amsterdam: John Benjamins.
Mller, Max. 1888. Zur Etymologie des koptischen awf: woh. Zeitschrift fr gyptische Sprache
und Altertumskunde 26: 9495.
Muysken, Pieter 2008. Functional categories [Cambridge Studies in Linguistics 117].
Cambridge: CUP.
Nedjalkov, Vladimir P. 1995. Some typological parameters of converbs. In Converbs in Crosslinguistic Perspective: Structure and Meaning of Adverbial Verb Forms Adverbial Participles, Gerunds [Empirical Approaches to Language Typology 13], Martin Haspelmath &
Ekkehard Knig (eds.), 97136. Berlin: Mouton de Gruyter.
Nichols, Johanna. 1986. Head-marking and dependent-marking grammar. Language 61: 56119.
Nikoleva, Irena. 2007. Introduction. In Finiteness: Theoretical and Empirical Foundations, Irena
Nikolaeva (ed.), 119. Oxford: OUP.
Noonan, Micheal. 1985. Complementation. In Language Typology and Syntactic Description,
Vol. II: Complex Constructions, Timothy Shopen (ed.), 42140. Cambridge: CUP.
Palmer, Frank R. 1986. Mood and Modality [Cambridge Textbooks in Linguistics]. Cambridge: CUP.
Payne, John R. 1985. Complex phrases and complex sentences. In Language Typology and
Syntactic Description, Vol. II: Complex Constructions, Timothy Shopen (ed.), 341.
Cambridge: CUP.
Chris H. Reintges
Polotsky, Hans Jakob. 1960 . The Coptic conjugation system. Orientalia 27: 392422.
Polotsky, Hans Jacob. 1944. tudes de syntaxe copte [Publications de la socit darchologie
copte]. Cairo.
Polotsky, Hans Jacob. 1987/1990. Grundlagen des koptischen Satzbaus [American Studies in
Papyrology 2829]. Decatur GA: Scholars Press.
Orlandi, Tito. 1986. Coptic literature. In The Roots of Egyptian Christianity [Studies in Antiquity
and Christianity], Birger A. Pearson & James E. Goehring (eds), 5181. Philadelphia PA:
Fortress Press.
Reintges, Chris H. 1994. Egyptian root-and-pattern morphology. Lingua Aegyptia 4: 213244.
Reintges, Chris H. 2001. Code-mixing strategies in Coptic Egyptian. Lingua Aegyptia 9: 193237.
Reintges, Chris H. 2003. Syntactic conditions on special inflection in Coptic interrogatives. In
Research in Afroasiatic Grammar II: Selected Papers from the Fifth Conference on Afroasiatic
Languages, Paris, 2000. [Current Issues in Linguistic Theory 241], Jacqueline Lecarme (ed.),
363408. Amsterdam: John Benjamins.
Reintges, Chris H. 2004a. Coptic Egyptian (Sahidic Dialect): A Learners Grammar [Afrikawissenschaftliche Lehrbcher 15]. Cologne: Rdiger Kppe.
Reintges, Chris H. 2004b. Coptic Egyptian as a bilingual language variety. In Lenguas en contacto: El testimonio escrito [Manuales y Anejos de Emerita XLVI], Pedro Bdenas de
la Pea, Sofa Torallas Tovar, Eugenio R. Lujn & Mara ngeles Gallego (eds.), 6986.
Madrid: Consejo Superior de Investigaciones Cientficas.
Reintges, Chris H. 2007a. Variable pronunciation sites and types of wh-in-situ. In The Copy
Theory of Movement [Linguistik Aktuell/Linguistics Today 107], Norbert Corver & Jairo
Nunes (eds.), 249287. Amsterdam: John Benjamins.
Reintges, Chris H. 2007b. Coptic relative tenses: The profile of a morpho-syntactic flagging
device. In Focus Strategies in African Langages: The Interaction of Focus and Grammar
in Niger-Congo and Afro-Asiatic [Trends in Linguistics 191], Enoch O. Aboh, Katharina
Hartmann & Malte Zimmermann (eds.), 185220. Berlin: Mouton de Gruyter.
Reintges, Chris H. & Green, Melanie. 2004. Coptic second tenses and Hausa relative aspects:
Acomparative view. Lingua Aegyptia 12: 157177.
Reintges, Chris. H., LeSourd, Phil & Chung, Sandra. 2006. Movement, wh-agreement, and
apparent wh-in-situ. 2006. In Wh-movement Moving on [Current Studies in Linguistics
42], Lisa Lai-Shen Cheng & Norbert Corver (eds.), 165194. Cambridge MA: The MIT
Press.
Roberts, John R. 1988. Amele switch-reference and the theory of grammar. Linguistic Inquiry
19: 4563.
Schachter, Paul. 1977. Constraints on coordination. Language 53: 86103.
Schultze-Berndt, Eva & Himmelmann, Nickolaus P. 2004. Depictive secondary predicates in
crosslinguistic perspective. Linguistic Typology 8: 59131.
Shisha-Halevy, Ariel. 1986. Coptic Grammatical Categories. Structural Studies in the Syntax of
Shenoutean Sahidic [Analecta Orientalia 53]. Rome: Pontifical Institute.
Shisha-Halevy, Ariel. 1989. The Proper Name: Structural prolegomena to its syntax -A case study
in Coptic. Vienna: VWG.
Smith, Carlota S. 1991. The parameter of aspect [Studies in Linguistics and Philosophy 43].
Dordrecht: Kluwer.
Spiegelberg, Wilhelm. 1924. Demotische Grammatik. Heidelberg: Carl Winter Verlag.
Stassen, Leon. 1985. Comparison and Universal Grammar: An Essay in Universal Grammar.
Oxford: Blackwell.
Stassen, Leon. 2000. and-languages and with-languages. Linguistic Typology 4: 154.
Stassen, Leon. 2005. Noun phrase coordination. In The World Atlas of Language Structures,
Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.), 258261.
Oxford: OUP.
Steindorff, Georg. 1951. Lehrbuch der koptischen Grammatik. Chicago IL: The University of
Chicago Press.
Stern, Ludwig. 1880. Koptische Grammatik. Leipzig: T.O. Weigel Verlag.
ter Meulen, Alice G.B. 1995. Representing Time in Natural Language. The Dynamic Interpretation
of Tense and Aspect. Cambridge MA: The MIT Press.
Thompson, Sandra A. & Longacre, Robert E. 1985. Adverbial clauses. In Language Typology
and Syntactic Description, Vol. II: Complex Constructions, Timothy Shopen (ed.), 171234.
Cambridge: CUP.
Till, Walter C. 1942. Koptische Kleinliteratur. Zeitschrift fr gyptische Sprache und Altertumskunde 77: 101111.
Till, Walter C. 1961. Koptische Dialektgrammatik: Mit Lesestcken und Wrterbuch. Mnchen:
C.H. Beck.
Till, Walter C. 1966. Koptische Grammatik (Sadischer Dialekt) mit Bibliographie, Lesestcken und
Wrterverzeichnissen, Leipzig: Enzyklopdie Verlag.
Van Valin Jr., Robert D. & LaPolla, Randy J. 1997. Syntax: Structure, Meaning and Function
[Cambridge Textbooks in Linguistics]. Cambridge: CUP.
Westendorf, Wolfhart. 1977. Koptisches Handwrterbuch. Heidelberg: Carl Winter.
Winkler, Susanne. 1997. Focus and Secondary Predication [Studies in Generative Grammar 43].
Yuasa, Etsuyo & Sadock, Jerry M. 2002. Pseudo-subordination: A mismatch between syntax and
semantics. Journal of Linguistics 38: 87111.
Zanuttini, Raffaella. 2008. Encoding the addressee in the syntax: Evidence from English imperative
subject. Natural Language and Linguistic Theory 26: 185218.
part iii
Subordination, informational hierarchy

and referential hierarchy
Informational and referential hierarchy

Clause-linking strategies
in Austronesian-Oceanic languages*
Isabelle Bril
LACITO, (Laboratoire des Langues et Civilisations Tradition Orale),
This paper analyses clause-linking strategies in mostly Austronesian languages,
with particular focus on the functions of informational and referential hierarchy
strategies in the architecture of complex clauses. Informational (topic, focus)
hierarchy and its markers, structure clauses as subordinate via the contrast
between presupposition vs. assertion. Referential hierarchy and its markers
(endophoric demonstratives and definite markers), are another subordinating
strategy based on the contrast between already referential/backgrounded
clause vs. asserted main clause. Paths of evolution leading from coordinators
or from endophoric demonstratives to informational hierarchy markers and to
subordinating conjunctions or constructions are more specifically discussed. It
is argued that informational hierarchy and referential hierarchy strategies are
inherent to the syntactic architecture of the complex clause.
1. Introduction
Austronesian languages display various clause-linking and subordinating strategies
and devices involving less finite or non-finite verb forms (reduplicated or nominalised
forms), tail-head cueing constructions, adpositional markers, distinct case-marking
strategies for main and subordinate clauses, modal dependency (irrealis, aorist),
aspectual dependency (see Franois, this vol.).
Here, the focus will be on the distinct functions of informational and referential
hierarchy strategies and their markers in the architecture of complex clauses. Although
these strategies belong to the domain of discourse for the former, and to reference
tracking for the latter, they are projected onto the syntactic level and are an intrinsic
*I am indebted to Robert D. van Valin for his critical and constructive comments on an earlier
version of this article. I am solely responsible for the remaining shortcomings.
Isabelle Bril
part of the architecture of clause-linking. Informational hierarchy and its markers

(topic and focus morphemes), structure clauses as subordinate via the pragmatic contrast between presupposed vs. asserted proposition; while referential hierarchy and
its markers (endophoric demonstratives and definite markers), structure clauses as
subordinate via the contrast between already referential vs. asserted information.
Section 2 provides some definitions on informational and referential hierarchy
systems, their markers and their clause-linking functions. Section3 illustrates some
of their frequent functions in various Oceanic languages. Section4 discusses a type of
evolution leading from coordinators to topic markers and subordinating devices based
on informational hierarchy. Section 5 analyses the complementary clause-linking
functions of coordinators, pragmatic (topic, focus) markers, and deictic markers in
Takia and Roviana. Section6 deals with case-marking and focus strategies. Section7
centres on the subordinating function of referential hierarchy (with demonstrative and
deictic markers). Section8 discusses the syntactic and pragmatic functions of clauseordering. Section9 concludes.
2. Th
e syntax of information and referential hierarchy in clause-linking:
Some definitions
Lambrecht (1994) defines information structure as a component of grammar, []
more specifically as a component of sentence grammar [] in which propositions
as conceptual representations of states of affairs are paired with lexicogrammatic
structures []. Information structure is thus a determining factor in the formal
structuring of sentences (Lambrecht 1994:3, 56). The structure of the clause is thus
conceived [] as a domain in which the different components of grammar syntax,
morphology, prosody, semantics, information structure compete and interact with
each other []. (ibid. 1994:12).
Van Valin (2005) expresses similar views and discusses the linking algorithms
between the various components and structural levels (syntactic, semantic, pragmatic).
2.1 Informational hierarchy: Focus, topic and their markers

In Lambrechts view, the important categories of information structure are: (i) the contrast between presupposition and assertion; (ii) topic and focus; (iii) identifiability and
activation (Lambrecht 1994:56).
When projected onto sentence level, information structure creates hierarchy via
the central contrast between the presupposed and the asserted clauses, expressed by a
variety of syntactic, morphological and prosodic markers developed below.
Topics are frames, they are definite and presupposed entities already present in
discourse and about which something is predicated (Lambrecht 1994). They have
v arious degrees of salience, as contrastive or detached topics. By contrast, the focus is

a restrictor, an operator with scope over the rest of the sentence (Krifka 2007).
Focusing usually involves the partitioning of the utterance between the focus and
the presupposition.1 The focus is the pragmatically non-recoverable element in an
utterance (i.e. the assertion) (Lambrecht 1994:207). Foci fall under various subtypes,
informative-completive foci which are generally in situ, such as answers to questions
(who wrote? My father wrote); the focus may also identify (contrastively or exhaustively) a restrictive variable (its x who did it).
Focus is marked by various devices, prosodic, syntactic (position, word order or
specific constructions such as clefts), and morphological (with dedicated focus markers
or focal operators, such as restrictive only, inclusive even, exclusive but, except, etc.).
In complex clauses, the focus clause is the asserted variable which restricts the
validity of the presupposed proposition. Focus clauses are syntactically more integrated and embedded than topic clauses, as shown by the inverted word order of the
main clause following the restrictive focus time clause (i) only when I heard him yell
did I realise that something was wrong, or by the use of that complementisers in clefts
(ii) its when I heard him yell that I realised something was wrong.
By contrast, topic-comment structure is a looser type of informational hierarchy. In the architecture of complex clauses, clauses containing the presupposition may
be detached to constitute the topic or frame, while the other clause is the assertion.
Thus, in because they wanted to know more, they started investigating the whole matter
(Lambrecht 1994:196), the topic clause is an adjunct of the main clause, in a relation
of ad-clausal subordination (van Valin 2005). Detached adjunct clauses generally indicate time, condition, cause, explanation.
2.2 Referential hierarchy: Demonstratives, deictic and definite markers

Referential hierarchy rests on distinct notions relating to anaphora and reference
tracking. In Oceanic languages, this is another frequent clause-linking strategy,
marked by endophoric demonstratives, deictic and definite markers which refer to
the propositional content of a preceding clause or sentence as informationally backgrounded; their evolution into subordinators (complementisers) is a common and
well-known process.
When a presupposed proposition becomes a discourse referent and serves
as an argument in another proposition, it may be linguistically designated
with the same expression type as an entity (with a personal or a demonstrative
pronoun.
(Lambrecht 1994:78)
1. Presupposed is defined as not part of the same assertion as the main clause (Givn,
1980:372).
Isabelle Bril
Referential hierarchy and its markers structure clauses around the contrast
between a clause with referential/backgrounded propositional content and an asserted
clause. Referentially backgrounded clauses marked by endophoric demonstratives and
definite markers are prone to embeddedness and syntactic subordination, as arguments or modifiers of another clause (see for instance Sections5.4, 7.4).
Another function of demonstratives and definite markers is found in clause-chains,
as in Takia (Section5.5.2) where they encode sequential and consecutive dependency
relations, or in correlative strategies.
2.3 Source of focus and topic markers

Clausal topics and clausal foci are relational constructions with functional differences:
topics are frames for some other assertion about them; foci are restrictor variables.
Despite such functional differences, some topic and focus markers have several common sources in Oceanic languages, the most frequent being coordinators or demonstratives (see Bril in press, Mithun 1988). Their functions and semantics are then
context and structure dependent, varying with their syntactic domain and scope. They
are distinguished by the grammatical construction or constructional schema (van
Valin 2005:131 sq.) onto which these markers are projected, and which contain morphosyntactic, logical-semantic, pragmatic and prosodic information which help discriminate their function.
Demonstratives may develop into three distinct directions, (i) as topic or focus
information hierarchy markers (see Section7.3), (ii) as correlative markers and (iii) as
full-fledged subordinators.
Among coordinators, sequential coordinators (and then) are a frequent source
of topic markers. Focal operators often originate (i) from additive (and also) and
sequential (and then) coordinators like xe in Nlmwa (Section 4.3), =ak in Takia
(Section5.1.3), =be in Manam (Section7.2), or from additive-inclusive focus sensitive
operators meaning too, also, even; (ii) from coordinators with constrastive-exclusive
semantics but (like nga in Tombunuo (Section4.1), ma in Tawala (Section6.2)), and
(iii) in a lesser degree from disjunctive or coordinators.
When used as information hierarchy markers with distinct prosody, conjunctive morphemes link (i) the salient (phrasal or clausal) topic and the comment, or (ii)
the focus and the presupposition (possibly with sequential, consecutive, inclusive or
constrative semantics). The prosody associated with information hierarchy signals the
hierarchy between presupposition vs. assertion, a feature only compatible with sub
ordination; coordinate clauses do not contain any presupposition.
Prosody further distinguishes topic from focus constructions: topic-comment
often correlates with rising-falling prosody, while focal entities carry modified stress
and salient prosody.
Topic, focus markers and demonstratives do not exhaust the types of informational hierarchy markers; position, clause ordering, T.A.M. markers are also widely
distributed, but they will not be the core of this study.
2.4 Topic and focus strategies as clause hierarchy markers

In contrast with coordinate clauses which can neither be detached nor focused, subordinate clauses may undergo pragmatic structuring. In some Oceanic languages,
the positional and morphosyntactic markers of informational hierarchy are the only
actual markers of subordinate clauses. Kiss (2001: 1447) proposes that languages
can be classified typologically on the basis of how closely they reflect the discourse
semantic structure of their sentences in syntax. In some cases, informational hierarchy strategies and subordination (with complement or adjunct function) may thus
be coextensive.
Besides, as pointed out by Lambrecht (1994:69), some conjunctive markers have
pragmatic properties revealed by their compatibility with focus or presupposition.
Thus, in contrast with because, since is incompatible with focus; this is illustrated by
the impossible use of since in a cleft construction, its because he had insulted me that I
struck him (* its since), or as an answer to a question, why did you hit him? Because
he insulted me (*since he insulted me). Some types of subordinators and subordinate
clauses thus entertain close links with pragmatic properties.
In the structuring of complex clauses, focus strategies create hierarchy by subordinating the truth condition of the presupposed clause to an asserted and salient
restrictor clause (informatively or exhaustively restrictive), which may specify the circumstances or conditions under which the presupposition holds true; they are syntactically more integrated.
Detached sentential topics involving the fronting of some adverbial clauses
(condition, cause or time frames) are more loosely connected, as in after she arrived
at the party, Kim saw Pat. In van Valins model of the layered structure of the clause,
detached clauses constitute adsentential subordination and are adjoined to another
clause in frame-comment constructions (van Valin 2005:192196). Detached sentential topics are outside the scope of main clause illocutionary force and of illocution
and negation operators (Foley this vol.; Foley & van Valin 1985); while peripheral
clausal subordination, in which the subordinate junct is a modifier occurring in the
periphery of a layer of the clause, may fall within the scope of main clause illocution,
as shown by Did Kim berate Pat because she kissed Chris? (van Valin 2005:192197).
Clausal subordination subdivides into ad-core and ad-clausal subtypes: in Kim saw
Pat after she arrived at the party, the after clause constitutes ad-core subordination,
it modifies the core saw Pat; while in Kim berated Pat because she kissed Chris, the
because clause constitutes ad-clausal subordination, it modifies the main clause.
Isabelle Bril
In Kiranti languages, Bickel (1993:2425) mentions cases of adsentential subordination marked by the same markers as nominal topics. In Belhare, these detached
finite topic clauses may also be focused by restrictive or additive focus particles.
Just as detached sentential topics are outside the scope of main clause illocution
and negation operators, detached lexical topics are not syntactic arguments of a clause,2
they may even be extra-clausal, as in other languages, you dont just have straight tones
like that3 (Lambrecht 1994:193).
2.5 Scope and domain

Topic and focus operate in different domains. Lambrecht (1994:215) defines the focus
domain as constituents whose denotata are capable of producing assertions when
added to presuppositions. These constituents may be full propositions and clauses.
Van Valin further refines the definition of the respective domains of topic and focus
as follows:
The domain of topic may be a constituent within the clause or a detached sentence
adjoined to another one; the domain of focus is defined with a subdistinction
between the potential focus domain: the syntactic domain in which focus
elements may occur and the actual focus domain: the part of the sentence that is
actually under focus.
(van Valin 2005:75).
Domain and scope are two essential notions to account for the varying functions of
morphemes. According to their scope and syntactic domain, demonstratives and deictics, for instance, may function as NP determiners, as topic or focus markers, and as
subordinators or clause-linking functors. Similarly, with variations in syntactic domain,
scope and prosodic specificities, some types of coordinators come to function as topic
or focus markers and may further evolve as subordinators (see Sections4; 5.3; 8.2).
3. I nformation and referential hierarchy markers in clause-linking:

Some case-studies
Presupposition, which is one of the key notions of information hierarchy falls under
two types: (i) the type found in existential, categorical propositions such as this
boy is small (often structured as topic-comment clauses in Oceanic languages), and
2. The syntactic arguments are marked by anaphoric pronouns (Lambrecht 1994:188).

3. Such detached lexical topics may not even be constituents (argument or adjunct) of the
clause with which they are pragmatically associated, they are syntactically autonomous,
extra-clausal elements. (Lambrecht 1994:193).
(ii) factual presupposition as in he is sorry that she didnt come (in which the subordinate clause contains the presupposition: she didnt come). By contrast with categorical propositions, thetic propositions such as its raining are sentence focus structures
containing no presupposition, and in which focus and assertion coincide (Lambrecht
1994:138, 213).
These logical types are often encoded by topic, focus and deictic markers which
occur as delimiters between presupposition and assertion. Topic markers commonly
occur in categorical judgements; they also have common clause-linking functions in
relative, time and conditional clauses, less frequently in cause-result clauses, explicative clauses, in purpose and complement clauses (of perception, cognition, volition,
command verbs). As pointed out by Haiman (1978) and Dik (1997), condition clauses
frequently correlate with presupposition and topic frames. Yet, although topics are
often presupposed, they cannot be equated with presupposition.4
Various case-studies will now illustrate how informational and referential hierarchy operate as markers of syntactic hierarchy in clause-linking.
3.1 Sobei (svo, Oceanic, Irian Jaya)

There are few instances of true subordination in Sobei, clause-linking is mostly marked by
coordinators or sequential linkers.5 The topic marker mau (re-)introduces an otherwise
inaccessible entity; it appears in categorical judgement clauses (1) structured as framecomment, in relative and time clauses structured in the same way; but conditional clauses
are juxtaposed or coordinated by map and then (Sterner & Ross 2002:185).
3.1.1 Categorical judgement clauses

(1)

Sobei (Sterner & Ross, in Lynch et al. 2002)

Yafu=mau=to bano
yafu.
fire=tpc=def enchanted fire
That fire was an enchanted fire.
(Sterner & Ross 2002:181)
3.1.2 Relative clauses

There are two types of relative clauses: one marked by the topic marker mau (2a),
which refers to unanchored determination, asserted for the first time, while the
other type, marked by enclitic demonstratives (dx1 =tei this or dx2 =sake that)
4. What is presupposed in a topic-comment relation is not the topic itself, nor its referent,
but the fact that the topic referent can be expected to play a role in a given proposition due to
its status as centre of interest. (Lambrecht 1994:151).
5. Such as -p and, sep and then, dep and, dop once finished, dosep so that, dedwop and
(+NP), and then, fetap and then, map then, so, consequently, dasa and so.
Isabelle Bril
or by definite markers (=te, =to), is used for already referential determination and
information (2b). In the first type, the head of the relative clause is delimited by mau;
in the latter type, the right boundary of the relative clause is marked by the deictic
(dx1 =tei this) or by an enclitic definite marker (=to that or =te this) (Sterner &
Ross 2002:169171).
(2) Sobei
a. E-yit
mefne=mau [ri-fos-e
Lea].

3sg.R-take woman=tpc 3pl.R-name-3sg Lea

He married the woman they call Lea. (lit. he took the woman,

they call her Lea)
b. U-be
nyo [w-ar-ema=to]
r-en
sifa?

2sg-poss stuff 2sg-bring-dir=def 3sg.R-be where?

Where is your stuff you brought here?
c.

[Use [dai dafu-n=ma

ri-fi]=tei ]
yam=te
matter two cross.sibling-3sg=fam 3pl-make=dx1 2pl=def
a-fei
tap.
2sg-make neg
Dont you do this thing (sexual intercourse) that these two cousins did.
(2002:175) [the demonstrative ma refers to a fact known to both speaker
and listener]
Relative clauses marked by mau are not embedded; while those marked by determiners and definite markers are more tightly integrated and embedded. Similar facts will
be shown for Nlmwa (Section4.3.2) and Takia (Section5.4.1).
3.1.3 Time clauses

Time clauses may be headed by a topicalised and relativised circumstancial noun samo
time (3). The relative time clause is marked at its rightmost boundary by an enclitic
demonstrative pronoun dx2 sake that and a definite marker =to that (or =te this)
marking the presupposition.
(3) Sobei
a. Map [samo=mau [tidut
er-enon
dimo waridon] sake=to],

so time=tpc together 3pl.r-stay house inside
dx2=def

So at the time that they were all together in the house,

Yusup=mau=to d-enon, e-rom moni sare.

Joseph=tpc=def 3sg-stay 3sg-see sleep play
Joseph was dreaming.
Without any time noun, clauses are structured as topic (time frame)-comment
clauses (3b):
b.

[[Ri-dus pamo=mau=to] [tat damos=mau]] [pamo=to tapse].

3pl-cut sago=tpc=def beat knead=tpc
sago=def neg
(When) they had cut down the sago palm, pounded, kneaded (it),
there was no sago.
(lit. having cut down, pounded, kneaded the sago palm, there was no sago
(it had magically disappeared).
The complex sentence in (4) illustrates the interaction of several markers: (i) sequential coordination with the enclitic coordinator =p; (ii) a relative clause headed by a
locative noun marked as topic by mau (pede=mau) and as presuppositional by the
deictic and definite markers sake=to; and (iii) a time frame marked as a topic clause by
mau (e-tasan-ewo=mau):
(4) Sobei
E-wo=p
e-wo=p
e-be6=ma
e-wo=p
[pede=mau
3sg-go=and 3sg-go=and 3sg-ctrst=fam 3sg-go=and place=tpc
[tema-n=to
e-fos-fe] sake=to] e-wo=p
father-3sg=def 3sg-call-? dem=def 3sg-go=and
[e-tasan-ewo=mau] mamuse.
3sg-view-dir= tpc empty
He went and went, he was the one (who) went and to the place his father had
named, he went and (when) he looked/looking around, they werent there.

While the English translation of e-tasan-ewo=mau mamuse requires a conjunctive

time clause or a participial form (when) he looked/looking around, it was empty,
Sobei actually uses a frame-comment construction.
3.1.4 Complement clauses with mau

Complement clauses may be marked as frame-comment clauses with contrastive
meaning.
(5)

Sobei
Map e-ski-i=mau
ri-orpar.
then 3sg-command-3pl=tpc 3pl-be.unwilling
Then what she commanded them (tpc), they were unwilling to do/
They were unwilling to do what she had commanded them to.
6. Mau (re-)introduces an entity in discourse (N and Proper N, but it does not appear on
pronouns), be marks contrastive topic (and is often attached to a pronoun).
Isabelle Bril
3.2 Kaulong (svo, Oceanic, New Britain)

Kaulong has various coordinators7 and few subordinators8 (Throop & Ross 2002:409).
The focus here will be on the functions of a definite postmodifier tin [NP tin] which
marks already referential information, as well as on the functions of the topic marker
and clause delimiter men. Men indicates contrast, a change of topic, sometimes a switch
in identifiable referent (it is absent from most indefinite NPs), a fronted adverbial phrase
(yesterday men) (ibid. 2002:39395). As a clause delimiter, it marks (conditional and
counterfactual) frame clauses and appears in categorical clauses and relative clauses.
3.2.1 Categorical judgement clauses structured as topic comment

(6) Kaulong (New Britain, Throop & Ross in Lynch et al. 2002)
Hiang ti-men a Susupa.
3sg.m dem-tpc p Susupa
This (man) is Susupa. (Throop & Ross 2002:393)

[tin men is fused as ti-men].

Most relative clauses are marked by men; the definite head is marked by tin.
(7) Kaulong
Po tin [masang a-pi
e Au e
kum ta-p
i
men]
3pl dem male attrib-place p Au and work ben-2sg there tpc
kahut ponval misan in to
yu.
card two
only be com pig
Only two of these men from Au and who are working for you have pigs.

(Throop & Ross 2002:396)
3.2.3 Conditional, counterfactual and time-frame clauses

Time clauses have no other explicit marking than being marked as fronted and backgrounded frames by men.
(8) Kaulong
[Po me lut
mata yu men] ku lek-val
akomen.
3pl come dance eye pig tpc irr fight-rec neg.abil
(When) they come (and) dance, they cannot fight with each other.

7. Ma and, u and, kuma and then, kama and so (consequence), takuma but, si and, but,
and then, and so, sini instead, while, va or (alternation).
8. Reported speech is marked by the locative preposition po, and cause clauses are headed
by the preposition epo about, with, for (Throop & Ross 2002:409).
Before clauses are expressed by using the perfective and the negation markers, they
are correlated to the main clause by the coordinator ma:
(9) Kaulong
[Li
lais
hiang li pi
om kur] ma hiang pir
3sg.fr coil.around 3sg.m go place neg pft and 3sg.m stand
e
mir
mir
to
li
mi
ehul
and stagger stagger com 3sg.fr inside plant.growth
uva
sangga].
shoulder Tahitian.chestnut.tree

Before it (python) had completely coiled itself around him, he stood and
staggered with it inside the plant growth of the Tahitian chestnut tree.
(lit. it had not yet coiled itself and he stood) (Throop & Ross 2002:391)
Condition clauses display an irrealis marker (ta)ku and are marked either as topic
frames as in the preceding examples, or as restrictor foci as in (10); Kha marks the
bracketed clause as a focused restrictor protasis for the following apodosis:
(10) Kaulong
Vut tin vala
men [ku hun kha] vut hun ma hiang tin
3sg.f dem woman tpc irr die foc 3sg.f die and 3sg.m dem
masang men9 ku in li hiang hun pet
kira titit-en.
man
tpc irr live go 3sg.m die follow back spouse-3sg
She, the woman, should she die (foc), (then) she dies and he, the man, he will
continue living (and) he dies following his wife.
4. From coordinators to topic markers and other clause-linking functions

In Austronesian languages, some coordinators often develop into topic or focus markers (see Bril in press); this occurs via topic maintainance (with and coordinators), via
topic switch (with contrastive but coordinators) or via correlative functions.
4.1 Borneo languages (Austronesian)

Various Borneo languages display such functional evolution; one such case is Timugon Murut, where am and is also a contrastive topic marker and a clausal frame
marker (11b).
(11) Timugon Murut (Borneo, Sabah)
a. Aku am timinutup
aku ak ra tunturing ti.

1sg and act.past-shut.self 1sg just obl ricebin this

As for me, I just shut myself up in this ricebin.

(Brewis & Levinsohn 1991:38)
9. Note also constrastive function of the topic marker men on the NPs men and women.
Isabelle Bril
b. Ba ginio am andu-on mu
ak i-ali
ti.

well that and marry-pat you.gen just pivt-younger.sibling this

Well, that being the case, youll have to marry this younger sibling.

(Brewis & Levinsohn 1991:37)
In Tombunuo, the sequential connector om and marks topic continuity whereas the
adversative-constrastive connector nga but signals topic shift, a fact noted in other
Austronesian languages (Bril in press).
(12)

Tombunuo (Borneo, Sabah)

Roraa ri nga pilion nio.
maiden that advs chose he
As for that maiden, he chose her.
(King 1991:58)
In Coastal Kadazan, sequential and additive om and and adversative nga but also
function as informational hierarchy markers: nga signals topic shift and contrast (13b),
while om signals topic continuity (13c). They are also used as correlative morphemes
between subordinate and main clause, marking continuity or contrast between events.
Compare the coordinating nga in (13a) and its correlative function in (13b) between a
time frame subordinate clause marked by nopo and the other clause.
(13) Coastal Kadazan (Borneo, Sabah)
a. Intang-ai
no dau i
Lonsibog nga poingodop.

look.at-ref.red pft 3sg def Lonsibog advs sleep

He looked at Lonsibog, but he was sleeping.
(Miller 1991:123)
b. Pihapak nopo ino do
duvo, nga kivaa do
tanak.

split
tpc dem indef two advs exist indef child

When the rock split into two, there were children.
(Miller 1991:128)
c.

Koiduai nopo10 dii

disido ii,
om pamanau no vagu.
release tpc
anaph 3sg.m anaph and walk
pft again
After he released it, he continued on his way.
(Miller 1991:118)
Nopo nga also appears between topic and comment in categorical predications:
(14)

Coastal Kadazan
Ngaan ku
nopo nga zi Landin.
name 1sg.n.pivt tpc advs def Landin
My name is Landin.
(Miller 1991:128)
10. Nopo combines the completive and anaphoric no and the incompletive and forwardlooking po, it marks the first part of the construction as the topic and points forward to the
comment which follows (Miller 1991:126127).
4.2 New Caledonian languages (Remote Oceanic)

Various New Caledonian languages display a similar use of coordinators as topic markers. Topic constructions often use discontinuous morphemes, a presentative marker
and a correlative coordinator indicating continuity as in Cmuh.
(15)

Cmuh (North New Caledonia)

>
Mepw a-ni ka o
t a
cuf li
h-nyu.
as.for. this and fut ass neu.art small reservation of-us
As for that, we kept it for us.
(Rivierre 1980:211)
Similar constructions occur in other northern languages: in Nyelyu, wam ka~kam

as for and/but, then (Ozanne-Rivierre 1998:57, 62); in Nemi, pmwavo ko~(x)
o as for and (Ozanne-Rivierre 1979:39); in Jawe, phwvo o; in Fwi, fwvo
o there is and. In Aji (South New Caledonia), w is both a coordinator and topic
marker.11 As a conjunctive marker, it may appear in sentence initial position meaning so, and then, it may also have an explicative meaning as, then (La Fontinelle
1976:225, 329330).
(16)

Aji (South New Caledonia)

Pani-\a
w, na vi-rru.
mother-poss.1sg tpc 3sg sow
As for my mother, she is sowing.
(La Fontinelle 1976:193)
4.3 Nlmwa (VS, VOA, New Caledonia, Bril 2002)

Like many Oceanic language, Nlmwa has a variety of coordinators and logical connectors (Bril 2002, 2004). Coordination, complementation and subordination are all
attested conjunctive types (Bril 2002).
4.3.1 Sequential and topic marker xe

The sequential coordinator xe (formerly ke) has developed several functions, as topic
marker and as conjunctive marker in some relative and (factive) complement clauses.
4.3.1.1 Sequential conjunctive marker xe
As a sequential marker, xe mostly appears sentence initially as in (17) where the first xe
takes the preceding discourse as backgrounded and topical, and further elaborates on
11. Loyalty Islands languages use similar constructions: in Drehu ame tre as forthen or
ame ke as for so (tre then or ke so, as) (Moyse-Faurie 1983:197, 201); in Iaai, haba
me as for and (OzanneRivierre 1976:133).
Isabelle Bril
it. The topic function of xe appears in the second occurrence of xe in (17), and marks
the clause yo axe-ve (what) you saw as the topic frame:
(17) Nlmwa
Na i
shumwiny mwa Pw-Hivic: xe yo axe-ve xe yo axe
and 3sg do.thus
seq Pw-Hivic conj 2sg see-dir tpc 2sg see
o
da?, fo
idaama-m?
instr what? there.is eye-poss.2sg
And Pw-Hivic then says: so, (what) you saw (tpc), what did you see it with?
Do you have eyes?
(Bril fieldnotes 1995)
4.3.1.2 Topic marker xe

Example (18) illustrates its use as a contrastive topic marker at constituent level:
(18) Nlmwa
Na hla
xe hla khu na,
na co
xe yo khuxi
and 3pl.fr tpc 3pl eat excrement but 2sg.fr tpc 2sg eat.tr
caa-m
mido.
food-poss.2sg taro
And as for them, they will eat excrements, but you, you will eat your taro-food.
In the architecture of main and dependent clauses, the neutral order is Main
Subordinate; in the reverse order, the subordinate clause is topicalised and marked
by xe. This holds for time and condition clauses. The only subordinate clause which
cannot be topicalised are purpose clauses which are in a logically iconic order and
are marked by the all purpose conjunctive marker me.12 Causal clauses also prefer
the logical (causeconsequence) order (19a), rather than the syntactic order (Main
Subordinate). In (19b), the consequence is topicalised with xe.
(19) Nlmwa
a. Puxe-t khl i
ye me kio i
haxa hma shaya.

cause illness prep 3sg conj neg 3sg almost a.lot work

Because of his illness, he almost does not work any more.
b. Kio i
haxa hma shaya (xe) puxe-t khl i
ye.

neg 3sg almost a.lot work (tpc) cause illness prep 3sg

He almost does not work any more because of his illness.
4.3.2 Relative clauses with xe

As in Sobei and Takia (see Section 5.4.1), there are two types of relative constructions based on informational and referential status. In Nlmwa, relative clauses with
12. In van Valins model, the fact that they cannot be topicalised is a sign that they are cosubordinate.
xe bring unanchored, new information; while relative clauses with a demonstrative

marker (20b) refer to known and referential determination (Bril 2001, 2002). This
function of xe originates from two formerly conjoined clauses13 and two assertions.
(20) Nlmwa
a. Na fhe pwaxi-n
xe
i
khl.

1sg bring child-poss.3sg conj 3sg be.ill

I bring his child who is ill.
(Bril 2001:262)
b. Na fhe pwaxi-n
bai
(i) khl.

1sg bring child-poss.3sg anaph 3sg be.ill

I bring his child who is ill.
(Bril 2001:262)
In contrast with relative clauses with xe (20a) which contain two assertions, relative
clauses with a demonstrative (20b) are prosodically and syntactically more integrated,
as proved by the optional deletion of the coreferent subject index in (20b), but not
in (20a). Furthermore, relative clauses with xe are restricted to factive clauses which
presuppose the existence of the entity (21a); if the referential status is uncertain,
hypothetical (21b) or under the scope of a negation (21d), the irrealis marker o then
appears. Compare with relative clauses referring to already referential determination
and marked by a demonstrative in (21c):
(21) Nlmwa
a. Fhe-dume
hele xe
caak.

bring-down.here knife conj be.sharp

Bring me a sharp knife (lit. a knife which is sharp;

I know there is one)
(Bril 2001:268)
b.

Fhe-dume
hele o caak.
bring-down.here knife irr be.sharp
Bring me a sharp knife. (lit. a knife which would be sharp;
if there is one)
(Bril 2001:268)
c.

Fhe-dume
hele bai
caak.
bring-down.here knife anaph be.sharp
Bring me the sharp knife
d.

Kio wa tuuli foliik o c diyawo

na na-t.
neg 2pl find thing irr very important loc inside-it
You did not find anything important in it.
(lit. you did not find something irr important in it)
(Bril fieldnotes)
(Bril fieldnotes)
13. In Old Fijian, the clause coordinator ka and, also, also plus headed a relative clause
(Milner, 1972:36); this function sometimes persists in Boumaa Fijian under the influence of
church language, but clauses are now juxtaposed (Dixon 1988:251257).
Isabelle Bril
4.3.3 Xe as a factive complementiser (perception or cognition verbs)

Whether in relative or complement clauses, xe and o appear contrastively as marking
realis vs. irrealis events. As a complementiser, xe marks the factive complements of
perception or cognition verbs; whereas o appears with verbs indicating uncertain events
(hak o not know if , fweneng o beware of , havava o hesitate to) (Bril 2000:448460).
While the topic function may be associated with a slight pause (after xe), there is no
pause between the head verb and the clausal complement. This function of xe also
originates from two formerly coordinate clauses.
(22) Nlmwa (Bril fieldnotes 1994)
a. Na i
u tl da mwaida ni k
jahoot xe

and 3sg pft hear up up.there in source river conj

i
tho-du-me.
3sg call-down-here
And he hears up there at the source of the river that he is

calling down here.
b. Na kaxaak xe
na Pum.

1sg be.sure conj 1sg go Pum

Im sure to go to Pum.
c.

I
khacac o i
.
3sg hesitate irr 3sg go
Hes unsure whether to leave.
(Bril 2001:264)
(Bril 2001:268)
4.3.4 Discussion
As a sequential coordinator, xe and (formerly ke) links clauses which are on the
same syntactic level and express some logical and sequential relation between
them. As a topic marker, xe delimits a frame and a comment clause. Its function as
a conjunctive marker in some factive relative clauses and complement14 clauses also
derives from originally coordinated clauses and assertions referring to independent
events: the adnominal relative clause was originally a coordinate clause (from I
met the woman and she is learning Nlmwa to I met the woman who is learning
Nlmwa). Similarly, the complement clauses (of perception or cognition verbs)
originally expressed the perceived or cognitively apprehended event as a loosely
correlated clause with additive or sequential semantics; this further syntacticised as
a verbal complement. Haiman shows that in Hua (Papuan) some types of complement clauses have different constructions:
14. Croft (2001:351353) points out that complement clauses often originate from coordinate or purposive subordinate clauses.
In Hua, a quote is treated as the object of a verb of saying (and, as an object,

precedes this verb), while a perception is not treated as the object of a verb of
perception (and, thus, follows this verb).
(1988:64)
Sequential linkers (and then, and so) order events in logical and time sequence
and often express causal, consecutive or purposive relations. As they contain some
implication, they are semantically asymmetrical, which accounts for their frequent cross-linguistic use as correlative markers in time, conditional and consecutive clauses.15 Correlative and implication markers are on the border line between
clause coordination and subordination, as in another word and I leave or you calm
down or I scream!
5. C
oordination, topic, focus markers and deictic strategies
in clause-linking: The case of Takia
The focus will now be on Takias16 clause-linking strategies, which epitomise some types
often found in Western Oceanic languages. Three markers will be analysed: (i) the coordinator and focus marker =(a)k, (ii) the topic marker man, (iii) and the demonstrative
markers dx2 an, dx1 en. Their clause-linking functions are based on informational or
referential hierarchy (see Bril in press). Takias clause linking types include (paratactic)
coordination, co-subordination, clause-chaining, and subordination for relative, complement and purpose clauses (marked by conjunctions or nominalisation).
5.1 Subordination
5.1.1 Adverbial clauses
Takia is described as having no conjunctively marked adverbial clause (Ross 2002:241):
adverbial clauses are marked by logical and semantic inferences,17 or by foregroundbackground strategies (see Verstraete this vol. for inferential encoding).
15. In Old French, apart from their coordinating function, both si if (in modern French)
and et and could link a main clause to (i) a time adverbial clause, (ii) a conditional clause,
(iii) a relative clause with a correlative morpheme and an anaphoric pronoun.
16. Takia belongs to the same North New Guinea cluster as Kairiru and Manam, though to
different linkages: Kairiru and Manam are closely related (Schouten linkage); Takia belongs
to the Vitiaz linkage.
17. Some clauses may be reversible without much syntactic change, and interpretation varies
with sequential order.
Isabelle Bril
5.1.2 Other subordinate clauses

Subordination appears mostly (i) for purpose clauses referring to events incomplete at
the time of speech (23a), and (ii) for complement clauses of verbs of desire or request
(23b); both are marked by the intentive predicate enclitic mood marker (w)o (in
order) to (Ross 2002:246).
(23) Takia (SOV, Western Oceanic, PNG, Madang Province)
a. O [ana w-ani wo] -palu
ya.

2sg food 2sg-eat int 2sg-come R

You came in order to eat.
(Ross 2002:246)
b. I ai [-au o] i-bol
a.

3sg 1sg 1sg-go int 3sg-speak R

He told me to go.
(Ross 2002:247)
The other main strategy is nominalisation with a possessive classifier (ane- or sa-); this
appears in negative existential clauses (24), in some negative complement clauses and
some relative clauses (see below).
(24)

Takia
[Ma dugo ma-bol
sa-n]
tia
ya.
1pl.exc what 1pl.exc-speak poss-3sg not.exist R
There is nothing we can say.
(Ross 2002:240)
5.2 Subordination and clause-chaining

Subordination and clause-chaining are marked by clausal enclitics expressing sequential, chronological (before, after) or simultaneous events (while), as well as causal,
explicative and conditional relations. These enclitics are former conjunctions, now
mostly used as mood indicators (=go realis dependent, =pe irrealis dependent) or
aspect markers (=do continuative, =gu18 completive). Cause-reason clauses are marked
by =ta (Ross 2002:244; 1993:6566). Loose clause-chaining and complementation are
marked by de (25) (Ross 2002:24345).
(25)

Takia
Pas i-gire
de Mait i-li-ag
a.
letter 1sg-write dep Mait 3sg-see-1sg R
Mait saw me writing a letter.
(lit. I wrote a letter and M. saw me)
(Ross 2002:243)
While the co-subordinators =go and =de link events with some natural connection,
coordination with =(a)k connects events with speaker-imposed logical relation (Ross
1993:5859).
18. After a vowel, they appear as =g, =p, =d.
5.3 The coordinator and focus marker =(a)k

Coordination is essentially asyndetic in Takia, except between clauses where it may be
marked by =(a)k19 which is cliticised to a clausal mood or aspect marking particle.
5.3.1 Clause coordination with =(a)k

=(a)k has no semantics of its own, the semantic relation is contextually inferable. It
mostly refers to simultaneous, parallel, alternative events; it is also used for reported
speech and for logical (causal) connections between events established by the speaker
(vs. events with some natural connection) (Ross 1993:5859).
(26)

Takia
Matu ulat uga=da=k
a
kris
tamol pein
aal=da.
senior work 1sg.do=ipf=ak dx2 Christian man woman 1sg.get=ipf
I do the work of an elder and I lead the Christians.
(Ross 1993:59)
Consecutive relations are marked by the conjunction akot and so, and consequently
(the contraction of ak and the locative marker ote yonder) (Ross 2002: 243). The
functions of =(a)k in clause-linking cross-cut the coordination-subordination dichotomy; it appears in constructions that are equivalent to both, including relative clauses
(Ross 2002:235).
5.3.2 Selective, informative and contrastive focus with =(a)k

The second important function of =(a)k is as a focus marker. It appears in polar questions involving some presupposition and is enclitic to the informative20 focal expression (Ross 2002:228).
(27) Takia
a. O
pein
u-le=k?

2sg.fr woman 2sg-see=ak

Have you met a girl?
b. Awo=k, ai i-le
ya=k

yes-ak 1sg 1sg-see R=ak

Yes, I have.
(Ross 2002:236)
Example (27c) is the answer to what could you cook?

c.

Malkouk=ak sa-d
anai i-nei
a=n.
white=ak
poss-3sg food 1sg-cook irr=def
I could cook white mens food.
(Ross 2002:228)
19. =ak is cliticised to a final consonant or after a pause between clauses, otherwise =k.
20. Informative/completive focus (in answer to a question for instance) contrasts with
restrictive/exhaustive focus.
Isabelle Bril
The other functions of =(a)k appear in relative and complement clauses, where it constrasts with the enclitic definite marker =n (see Section5.4).
5.4 S ubordinating strategies: =(a)k vs. the definite clausal enclitic =n

vs. nominalised subordinate clauses
In clausal subordination (relative, complement and some adverbial clauses), =(a)k
marks asserted and unanchored information. It also indicates a looser syntactic bond
than the definite enclitic =n (which also marks NPs as definite) and which indicates
referential, presupposed information and stronger syntactic integration.

A. =(a)k vs. =n construction
In relative clauses, the =(a)k vs. =n strategies are similar to those already encountered
in Sobei or Nlmwa. =(a)k asserts some unanchored, focal determination while the
more frequent =n strategy expresses referential and presupposed determination. They
also correlate with definiteness: a definite NP appears with =n relative constructions
(28a), while an indefinite NP appears with =(a)k constructions (28b).
(28) Takia
a. Ab
a
[o w-abi
ya=n] ai Mait da lo

house dx2 2sg 2sg-build R=def 1sg Mait com 3loc.in

mu-mado da.
1exc-stay ipf
The house that you built, Mait and I are living in it. (Ross 2002:230)
b. i aar
[parapar na ya=k] kaek ma-a=p

3sg canarium platform loc R=ak one 1pl.exc-take=irr.dep

so we take one canarium nut which is on the bed and ...

(Ross 2002:231)
B. Nominalised possessive construction

Another possible backgrounding and subordinating strategy is the use of nominalised
possessive construction in relative clauses.
(29) Takia
di-ani sa-n
an.
3pl-eat poss-3sg dx2
Those (things) for them to eat. (lit. that (which is) their eating)

(Ross 2002:231)
5.4.2 Complement clauses (perception, cognition, utterance, ability verbs)

A. =(a)k vs. =n
Again the difference between =(a)k vs. =n lies in the focal, asserted features marked
by =(a)k vs. the presupposed and referential information status marked by =n. In (30a),
the two clauses are asserted and the reported speech clause is more loosely concatenated; whereas in (30b), the complement clause is embedded, its propositional content
is marked as presupposed and referential by the definite enclitic =n. Subordination in
(30b) further appears in the modal opposition between realis and irrealis; the irrealis
wa- in (30b) has generic meaning and refers to a norm.
(30) Takia
a. Ago-go uya-n
a-k du-bol.

pro-r:d good-3sg R=ak 3pl-speak

Then/so, they say it is good. (lit. thus, it is good they say) (Ross 2002:242)
b. Bin [i-siti
wa-n] i-lo
a.

Bin 3sg-read irr-def 3sg-hear R

Bin has learnt/knows how to read.
(Ross 2002:246)
B. Nominalised complement clauses

Some complement clauses can also be nominalised with a possessive construction
(31), similarly to the relative clause in (29).
(31)

Takia
ai man [nek du-fun-ag a-mat ane-n]
a-moi.
1sg tpc just 3pl-hit-1sg 1sg-die poss-3sg 1sg-not.want
As for me, I just did not want to kill them.
(Ross 2002:248)
5.4.3 Reason clauses with =(a)k vs. =n

Reason clauses show the same twin constructions varying with asserted, unanchored,
focal information with =(a)k vs. presupposed and referential information with =n.
Though more data would be needed to confirm this, I interpret the clauses as equally
asserted and coordinate in (32a), while in (32b), the reason clause is presupposed,
backgrounded and marked as subordinate by =n.
(32) Takia
a. O u-rer
a=k nie-n
ta w-au na ya.

2sg 2sg-fear R=ak matter-3sg neg 2sg-go dur R
You were afraid and for that reason you didnt go.
(Ross 2002:247)
(my translation; Ross translates both sentences (ab) as because you
were afraid you didnt go, but suggests that the clause may be in
apposition to nie-)
b.

O u-rer
a=n
nie-n
ta w-au na ya.
2sg 2sg-fear R=def matter-3sg neg 2sg-go dur R
Being afraid, you didnt go. Or You didnt go for the reason that/because
you were afraid.
(Ross 2002:247) (my translation)
5.5 Clause linking and sequencing functions of dx2 an, dx1 e, en

Deictic markers have various clause-linking functions. The medial demonstrative dx2
an, and the proximal demonstrative dx1 e, en have pronominal and adnominal uses.
Isabelle Bril
In their adnominal use [NP an ~ e, en],21 they mark definiteness: mau an that taro;
am en this story (Ross 2002:240).
(33)

Takia
An goun tia
ya.
dx2 dog not.exist R
That is not a dog.
(2002:240)
As a clause-linking marker and endophoric reference-tracking device, the medial

deictic dx2 an has a heavy functional load (Ross 2002:224).
5.5.1 Recapitulative function of dx2 an in sentence-initial position

In sentence-initial position, dx2 an thus, so is recapitulative and cohesive, it refers
back to a preceding clause or sentence (Ross 2002:245).
(34) Takia
An misin a
[du-palu ya=n] i urat nek
ago fun
dx2 mission dx2 3pl-come R=def 3.fr work ctrst thus base-3sg
du-ga ya.
3pl-do R
Thus, the missionaries who came, they started their work in this way.

(Ross 2002:230)
5.5.2 Clause sequencing with dx2 an an dx1 en

dx2 an and dx1 en are primarily paratactic links between discourse entities. Clauses
are sequenced with dx2 an, while consecutive relations are marked with proximal
dx1 e(n) here, this expressing the immediate consequence of the preceding action
(Ross 2002:245).
(35) Takia
[Bog i-du]
an [i-fun-ai
de i-bol] an [suuro-n wog
eagle 3sg-descend dx2 3sg-kill-3pl dep 3sg-say dx2 beak-3sg canoe
lo i-sofute] en i-mat.
in 3sg-ram dx1 3sg-die

lit. the eagle descended, (dx2) he thought he would kill them, (dx2) his beak
rammed into the canoe, dx1 he died
The eagle descended thinking he would kill them, [but] his beak rammed into
the canoe and consequently he died.
(Ross 2002:245)
21. The demonstrative en signals the final boundary of the definite NP (and may only be
followed by a quantifier; e is used when one or more modifiers follow the definite NP (as in
relative clauses) (Ross 2002:224).
5.6 Clause-linking functions of =(a)k an and =(a)k en

When combined with =(a)k, the deictics dx2 an and dx1 en have other clause-linking
functions.
5.6.1 Complement clauses with =(a)k an

Complement clauses with =(a)k (36a) stand in contrast with those marked by the
definite marker =n (see ex. (30), Section5.4.2). The third possible construction combines =(a)k an (dx2) as in (36b).
(36) Takia
a. ai
a-lo a=k i i-palu
wa.

1sg.fr 1sg-hear R=ak 3.fr 3sg-come irr

I have heard he will come.
(Waters & Ross 2002:242)
b. I
i-palu
wa=k an ai
a-lo a.

3sg.fr 3sg-come irr=ak dx2 1sg.fr 1sg-hear R

I have heard that he will come. (lit. he will come and [that] I have heard)

(Waters & Ross 2002:242)
The first difference lies in clause order: in (36a), the clause containing the perception
verb precedes the clause referring to the event, while it follows it in (36b), thus placing the perceived event clause in the expected position of a clausal argument in a SOV
language. But the other main difference lies in the anaphoric function of dx2 an in
(36b), which, as a propositional anaphora, refers back to the first clause and suggests
a looser clause relation.
Again SOV Hua offers some interesting perspective; Haiman (1988:64) notes that
[] a perception is not treated as the object of a verb of perception (and, thus, follows this verb). There are thus two constructions of perception verbs in Hua; the first
treats the perception as the verbs object, the complement clause is then headed by a
relativised noun na thing and precedes the perception verb (37a).
(37) Hua (Papuan, Haiman 1988, SOV)
a. [Eva kuttana ripa
na-mo]
kgoe.

money theft
take(2sg.rel) thing-(nmz) I.saw.you

I saw you stealing the money.
(Haiman 1988:61)
But this is rare and

[] limited (it seems) to clauses which represent propositions whose validity
is presupposed. [] and are acceptable only if the fact is already common
knowledge, and what I am reporting as news is only the fact that I observed it.

(Haiman 1988:61)
The other construction makes use of the inconsequential marker mana, when the
propositional content refers to non-presupposed event where my observation
Isabelle Bril
validates the truth of what I describe. The perceived event then follows the
perception verb.
b. Ka-mana
navibo rgi abaivao.

look (3du.incons) in.there really they.are.not

The two of them looked (and saw) that they really werent in there.

(Haiman 1988:61)
[] The inconsequential form is then used where the complement is
not subjective, and, as a non-object clause, follows the verb of perception.

(Haiman 1988:64)
Haiman then compares all the possible constructions of the complement clauses of
perception verbs as follows:
This may, then, be the best explanation for the use of the inconsequential
construction for expression of the relationship between acts of perception and
the acts perceived. The construction is used because none of the other available
constructions is appropriate. They make unacceptable claims about the semantic
relationship between the two events in question: either by claiming that the event
perceived is already common knowledge (the relative clause construction), or by
claiming that the event perceived is a physical product of the act of perception
(the direct quote construction), or by claiming that the event perceived is purely
subjective (the -gasi gerund construction), or by claiming that the act of perception
precedes and/or is the cause of the event perceived (the medial construction).

(Haiman 1988:65) [My italics].
The chaining (medial) construction which signals a consecutive relation such as you
came up (and) I saw you implies that the event is prior to the perception; the reverse
order with the same construction, I looked at you and consequently you came up, is
unacceptable unless the intended meaning is that the perception is the cause of the
event. Subordinating strategies are thus highly sensitive to semantic factors such as
modality (realis, irrealis), reference and logical relations. The data in Takia might be
analysed from such a perspective.
5.6.2 Clause-linking with =(a)k and dx1 e ~ en

Clauses marked by =ak and the endophoric dx1 e ~ en express consecutive and deductive
relations (Ross 2002:245). But constructions with dx1 e are subordinate-like, and based
on referential strategy. In (38a), the rightmost boundary of the first clause is marked by
final dx1 e, followed by a pause, [=(a)k e| (pause)], the first clause is thus backgrounded,
presupposed and subordinate, similar to a participial clause. In (38b), the pause is before
en [(pause) =(a)k | en] and dx1 en is in the initial position of the second clause, it stands
as the propositional anaphora of clause 1, recapitulates its propositional content, and acts
as a correlative marker expressing causal-consecutive relation.
(38) Takia
a. ai22 pein
u-le
la
i=k e,| o
pein
ta

1sg.fr woman 1sg-see term R=ak dx1 2sg.fr woman indef

u-le
o.
2sg-see int

(lit. I have met a girl (given this dx1), you should meet a girl (too).

(as) I have (already) met a girl, you should meet a girl (too).

(Ross 2002:242) [ | marks a pause)]
b. Id
mala-d
y-of
da=k | en panu na

1incl.pl eye-1incl.pl 3sg-close ipf=ak dx1 village loc

t-au
wa.
1incl.pl-go irr
(lit. Our eyes are closing, given this we will go home)

(As) we are sleepy, so we will go home.
(Ross 2002:245)
The demonstratives position thus signals distinct informational functions: backgrounding subordination of the first clause with dx1 e in (38a) with =ak possibly
acting as a restrictor focus marker; and a looser type of clause-linking in (38b) with
clausal anaphoric dx1 en.
5.7 =(a)k combined with the topic marker man

The topic marker man indicates a switch of topic between already referential topics;
there is usually no intonation break after man. It is different from mere fronting without any other marker (Ross 2002:238).
(39) Takia
a. [ai
sa-g
boi
an (man)] i e
sip

1sg.fr poss-1sg servant dx2 (tpc) 3.fr dx2 ship

sa-n
tamol.
poss-3sg man
(As for) my servant, he is a sailor.
(Ross 2002:237)
Sentence-initial man indicates that the preceding utterance is presupposed and topical
(as for that, thus).
b. Gu=g
milae-n tina-n a. Man u-moi.

cont=r.dep long-3sg big-3sg R tpc 2sg-not.want

But its very long.
Dont worry about that.

(Ross 2002:245)
22. Independent pronouns signal a change of topic. When there is no topic switch, the reference to the current topic is only marked by affixes on the verb.
Isabelle Bril
In (40), the condition clause (marked by the reason marker ta and the irrealis dependent morpheme =p) is marked as a topic clause by man:
(40)

Takia
You nam gire mi-gane
ta=p
man | aar
saen da.
water inst oil 1pl.exc-do reas=irr.dep tpc canarium bad ipf
(lit. because we mix the oil with fresh water (tpc), the canarium goes bad)
(If) we mix the oil with fresh water, the canarium goes bad. (Ross 2002:244).
5.7.1 The contrastive conjunction ak man in coordinate clauses

The contrastive conjunction ak man but occurs in sentence or clause initial position, after an intonation break; it indicates a topic shift between otherwise coordinate clauses.
(41)

Takia
ai
nor
u-palu ya, ak man o tia
ya.
1sg.fr yesterday 1sg-come R ak tpc 2sg not.exist R
I came yesterday, but you did not.
(Ross 2002:242)
5.7.2 Cliticised =(a)k man in subordinate clauses

In contrast with the conjunctive ak man but in (41) above, in (42), the cliticised
=(a)k man has scope over the topic clause to its left. The Dx2 an in initial position
of the following clause of (42) is the propositional anaphora of the preceding frame
clause, a correlative marker linking the frame clause and the clause which contains
the main assertion.
(42)

Takia
Ta i-win
na i=k man | an dal na i-mul.
neg 3sg-win dur R=ak tpc dx2 path loc 3sg-return
(lit. he does not win [the race] tpc, (given) that he turns back)
(When/as) he does not win [the race], then he turns back.
(Ross 2002:244).
5.8 Discussion
Coordinate clauses with equal assertive/illocutionary force thus stand in contrast with
clause-linking strategies based on asymmetric assertive force, using either (i) informational hierarchy (topic or focus) markers, contrasting presupposition vs. assertion, or
(ii) referential hierarchy devices such as determiners or definite markers which may
function as subordinators or as looser correlative markers.
A further strategy involving a differential case-marking system will now be analysed in Roviana.
6. C
ase-marking and focus strategies in clause-linking: The case of
Roviana (Oceanic, Solomon Islands)
According to Corston (1996, 2002), the preferred clause-linking strategy in Roviana
is coordination; subordination is extremely limited. Relative and complement clauses
are marked by sapu (43b), but the latter are uncommon: the preferred strategy is to use
(i) an epistemic modal such as gina maybe rather than a matrix verb of cognition, or
(ii) a conjunctive expression composed of and coordinators and a quotative marker
indicating epistemic doubt as in (43a).
(43) Roviana (Corston-Oliver, in Lynch et al. 2002)
a. Mala hite si rau meke gua meke lopu ruku sa popoa.

afraid little abs 1sg and say and neg rain def place

I am afraid that it might not rain.
(Corston-Oliver 2002:496)
b. Matatagu se John sapu kote seke-i-a
(e) Zima

fear
abs John comp fut hit-tr-3sg.o erg Zima

se Maepeza.
abs Maepeza
John is afraid that Zima will hit Maepeza.
(Corston 1996:30)
Adverbial clauses are restricted to conditional and time clauses and are headed by
totoso while, when, beto after, pude if . Adverbial clauses have a number of specificities which set them apart from other subordinate clauses: (i) they occur in clauseinitial Focus/Topic position23 which is part of the subordinating strategy; (ii) they
are optionally marked by the focal marker si; (iii) they display neutral case marking
(not the absolutive/ ergative morphology found in main clauses); (iv) they have a
distinct intonational contour; (v) they never contain new information in core argument positions (although they may in other syntactic positions) (Corston-Oliver
2002:496).
Main clauses (as well as relative and complement clauses) thus display a split ergative system, while adverbial subordinate clauses have a neutral case system. In (44ab),
the subordinate clause is focused by si and the pronoun goi you is in the neutral
case24, while it is marked by the absolutive marker si (si goi) in the main clause. The
23.
Roviana has the basic order [(top)VAO or (top)VS].
24. Absolutive pronouns are marked by si, while ergative and neutral pronouns are marked
by the same paradigm. Proper nouns are marked as absolutive by se and as ergative by e. Other
quantified NPs are marked as absolutive by si, ergative is unmarked (Corston 1996:1213).
Isabelle Bril
focus and absolutive markers si are homophonous, and perhaps historically related,
with the focus marker possibly originating from the absolutive case-marker.
(44) Roviana
a. [Pude la goi
pa popoa taqa rau pa Solomone si]

if
go 2sg.neu prep place poss 1sg prep Solomons foc

kaqu vagi meresina si

goi.
must gather medicine abs 2sg

If you go to my place in the Solomons, you must get some medicine.

(Corston 1996:2627)
b. [Totoso koa goi
pa korapa tropic si] kaqu pezaku

when stay 2sg.neu loc inside tropic foc must wash.hands

lamo si goi.
always abs 2sg

When you stay in the tropics, you must always wash your hands.

Adverbial clauses expressing concomitant events (45) display the same neutral case
marking system; they are marked as backgrounded by a reduplicated verb. Compare
the neutral pronoun ri they in the subordinate clause, with the ergative (ri) and the
absolutive NP se Noki in the main clause:
(45) Roviana
[En-ene ri
karua] tutuvi-a
ri
kara se Noki.
red-walk 3pl.neu two
meet-tr-3sg 3pl.erg two abs snake
(As) they were walking along, they met Snake.

The neutral case marking system also appears in clauses which are not syntactically
subordinate, but which are presupposed and backgrounded frames as in (46a): this
points to another shared feature between information structure and subordination.
Compare with the two coordinate clauses in (46b) where the arguments of both clauses
are marked as absolutive as in independent clauses:
(46) Roviana
a. [Pa ngati seda
si habotu gami]
meke vivinei si

prep root frangipani foc sit
1pl.exc.neu and chat abs

gami
kara Granpapa.
1pl.exc two Grandpa
(As) we were sitting under a frangipani and/then Grandpa and

I were having a chat.
(Corston 1996:32)
b. [Pa ngati seda

si habotu si gami] [meke vivinei

prep root frangipani foc sit
abs 1pl.exc and chat

si gami
kara Granpapa].
abs 1pl.exc two Grandpa

We sat down under a frangipani and Grandpa and I had a chat.

(Corston 1996:32)
At constituent level, the focus particle si encodes contrastive topics as in (47a) or

focus (47bd) (ibid. 2002:486).
(47) Roviana (Corston-Oliver, in Lynch et al. 2002)
a. Arau si na
qua
ginani
si pa batu huda.

1sg.fr foc indef poss.1sg nom food foc loc head tree

As for me, my food is in the treetops.
b. Esei poza-mu si agoi?

who name-2sg foc 2.fr

What is your name?
c.

Na sa
si kote hena-i-a kohite veluvelu?
disj what foc fut eat-tr-3sg today evening
What are we going to eat this evening?
d. Totoso sa
si kote beto si goi?

time what foc fut finish abs 2sg

When are you going to finish?
Thus, neutral case marking, avoidance of newly mentioned arguments in adverbial

clauses, and the occurrence of adverbial clauses in sentence initial (Focus/Topic) position, delimited by the focus marker si (indicating the restrictor variable), are part of
the subordinating strategy contrasting presupposition vs. assertion, and are clearly
related to informational hierarchy strategies.
7. R
eferential hierarchy: Demonstratives and deictics as markers
of subordination and clause hierarchy in Tawala
The function of demonstratives in clause-linking and subordination will be further
investigated, based on data from Tawala (Western Oceanic). Eastern Oceanic languages such as Mwotlap (Vanuatu, Franois 2000) display similar clause-linking functions for demonstratives. Nlmwa uses them in relative clauses containing referential
determination and in cleft constructions to mark the presupposed propositional content. Endophoric demonstratives are a frequent cross-linguistic mechanism of clause
integration and subordination, optionally associated with a conjunctive marker, as in
Isabelle Bril
French: il lui a menti, /ceci ~ cela ~ ce qui/ narrangea pas la situation he lied to him, /
this ~ that ~ which/ did not improve the situation.
In Tawala, deictics, demonstratives and definite markers have topic marking functions (see Section 7.3.), as well as clausal conjunctive functions, mostly in relative,
time, conditional, cause, and explicative clauses. As in Takia (where Dx2 an marks
sequential events and Dx1 en consecutive events) proximal and distal deictics are used
for distinct clause-linking functions.
7.1 The polyfunctional coordinators po and ma

Tawala also displays polyfunctional conjunctions. The and conjunctions po and ma
are used for coordination and factive25 complementation. The contrastive coordinator
ma and, but has other functions (Bril in press): (i) as a correlative conjunction linking
adverbial clauses to main clauses, as in (48a) where the reduplicated verb tugala marks
temporal overlap and durative aspect, (ii) as a contrastive and restrictive topic marker
(48b), or as a focus marker at constituent level (48b) or at clause level (48c).
(48) Tawala (SOV, Oceanic, Milne Bay, Papuan Tip, PNG)
a. I-tugu-tugala ma hi-woe.

3sg-red-sit and 3pl-paddle

While he was sitting, they paddled.
(Ezard 1997:250)
b. Yaka a koyama uyahi-n-ei ma a luyagohana i-weme-ya.

so
his lying
at-3sg-abl and his life
3sg-take-3sg

So by his trickery he saved his life.
(Ezard 1997:250)
c.

Geka
ma polo a-matut-e-ya.
this/here foc pig 1sg-fear-trv-3sg
It was here that I was frightened by the pig.
(Ezard 1997:197, 250)
While ma and, but marks a change of subject or topic, po conjoins clauses with a close,
consecutive semantic relation and rarely introduces a new topic (ibid. 1997:243).
7.2 Subordination
Among adverbial clauses,26 cause clauses are marked by two postpositions uyahi-n-ei
and ugoli-n-ei (lit. at-it-abl) because of and occur in the cause-result order (Ezard
1997:237). Purpose of motion clauses such as he went to the river to bathe are generally
juxtaposed and follow the motion verb. Time clauses (Section7.4.1), are generally preposed to the main clause and are simply juxtaposed or correlated to the following main
25. Irrealis complementation is marked by ipa (irr mkr).

26. Prosodically, sentences begin with a higher onset and end with a lower coda than clauses.
clause by ma and, po and (with consecutive meaning), or they may be backgrounded

with Dx2 naka (Section 7.4.2). Complement clauses also display various strategies,
nominalisation, adpositional complements, conjunction with po and, the irrealis
marker ipa and the demonstrative Dx2 naka (Section7.4.3). Conditional clauses are
usually in sentence initial position (Section7.4.4).
7.3 Demonstratives as constrastive topic or focus markers

Our focus will now be on the functions of the medial demonstrative Dx2 naka27 that,
there as a pragmatic topic and focus marker, and as a conjunctive marker in relative,
condition, time and complement clauses. Its functions vary with its domain and scope.
When preposed to an NP, naka has focal function (49a), as shown by the contrast with
a neutral definite NP marked by na (49b).
(49) Tawala (Ezard 1997:144)
a. Naka bada-na i-woe-woe.

dx2 man-def 3sg-red-paddle

That man was paddling.
b. bada-na i-woe-woe.

man-def 3sg-red-paddle

The man was paddling.
All three deictic grades (Dx1 geka, Dx2 naka and Dx3 noka) have constrastive topic or
focus functions. Topic maintainance and prominence of NPs is marked by postposed
demonstratives, such as Dx3 noka in (50a) and Dx1 geka in (50b) (ibid. 1997:143144).
(50) Tawala
a. Ma [dobu-na noka] dobu banei duma-na.

and town-def dx3.tpc town big very-3sg

As for the town there, it is a big town.
(Ezard 1997:143)
b. Ma [meyagai geka]
[naka meyagai dewadewa duma-na].

and village
dx1.tpc dx2 village good
very-3sg

As for this village, that is a very good village.
(Ezard 1997:144)
Compare the topic function of postposed geka and the focal function of the following
preposed naka in (50b).
27. The three deictic grades are: proximal Dx1 geka this, here, medial Dx2 naka that, there,
and distal Dx3 noka that, over there (out of speakers or hearers sight). They are reflexes of
Proto-Oceanic *a/*na near addressee; *o/*no distant from both speaker and addressee (Ross
1988:100).
Isabelle Bril
7.4 The clause-linking functions of Dx2 naka

Dx2 naka has conjunctive functions in relative, complement and adverbial clauses.
7.4.1 Relative clauses with Dx2 naka that, there

The verbs of all relative clauses are marked as subordinate by reduplication. Relative
clauses may be juxtaposed and embedded as in (51):
(51)

Tawala
Kwasikwasi(-na) [pom
u-gima-gimal-e-ya]
i-tutuma.
machete(-def) yesterday 2sg-red-buy-trv-3sg 3sg-blunt
The bushknife that you bought yesterday is blunt.
(Ezard 1997:207)
Or they may be hierarchised as topiccomment clauses as in (52a b), with postposed

Dx2 naka28 marking the presupposed status of the information contained in the relative clause (ibid. 1997:208209).
(52) Tawala
a. Numa [hi-wogo-wogo-hi naka] | hi-lata
duma.

house 3pl-red-build-3pl dx2.tpc 3pl-grow very

As for the houses which have been built [impers. 3pl], they are very big.

(Ezard 1997:209)
b. Meyagai noka [a-ga-gale-hi
naka] | meyagai

village there 1sg-red-see-3pl dx2.tpc village

dewadewa duma-na.
good
very-3sg
As for the villages which I saw, they are very big villages. (Ezard 1997:209)
The relativised NP is often repeated in the main clause and indicates a resumed topic
as in (52b). Topic-comment relative clauses are a common strategy.
7.4.2 Time clauses with naka

Time clauses are relativised time noun clauses with houga time, they are optionally
topicalised with naka and set the time-frame for another clause.
(53) Tawala
a. [Ma houga-na to-ne-nae
(naka)] | houga i-apapoe duma.

and time-def 1pl.exc-red-go dx2.tpc time 3sg-bad very

As for the time we were going along, the time/weather was very bad.

(Ezard 1997:209)
28. The bracketing of clauses (64a-b) is Ezards, naka belongs to the relative clause. According to
Ezard, they must be interpreted as nominal despite their clausal structure, because of the presence
of the demonstrative topic marker naka. In my analysis, it is a topic-comment construction.
b. Houga-na hi-gele-geleta, pona-pona hi-nonol-i.

time-def 3pl-red-arrive pl-voice 3pl-hear-3sg

The time when they arrived, they heard voices.
(Ezard 1997:210)
In (53b), the topic time clause is marked by a prosodic rise, thus houga-na hi-gelegeleta cannot be an independent clause, by contrast with (54) where houga-na is a full
NP focused by naka (similar to (49a)):
(54)

Tawala
Naka houga-na hi-gele-geleta.
dx2 time-def 3pl-red-arrive
They were arriving at that time.
(Ezard 1997:210)
7.4.3 Complement clauses with Dx2 naka

Complement clauses with Dx2 naka is one of several complementising strategies used
for realis complement clauses of perception verbs (see, hear), cognition verbs (think,
know, believe, desire), utterance and quotation verbs (say, tell) (Ezard 1997:226). Subject complement clauses occur as topic clauses, as in (55) where naka is an anaphoric
and focal pronoun,29 not a complementiser.
(55)

Tawala
[O-guuguya
lawa uyahi-hi] naka ega i-ta-dewa-dewa.
2sg-red-preach person at-3pl
dx2 neg 3sg-irr-red-good
As to your preaching to the people, that is not good.
(Ezard 1997:221)
Object complement clauses (including those marked by naka) normally follow the
verb,30 and are thus distinct from core arguments31 which precede the verb [OV] (ibid.
1997:221). In (56) and (57a), naka displays the common evolution from cataphoric
demonstrative to complementiser.
(56)

Tawala
A-gale-i
[naka hi-buli-bulili].
1sg-see-3pl dx2 3pl-dur-run
I saw that they were running.
(Ezard 1997:231)
29. The bracketing of example (58) is Ezards, naka thus heads the second clause as a full
focal NP: As to your preaching to the people, that is not good.
30. Complementation with po also follows the verb. This reflects the Proto-Oceanic SVO
order rather than the new SOV order of NP arguments due to contact with Papuan languages
(Ezard 1997:221).
31. But some nominalised object complements occur before (like core-arguments).
Isabelle Bril
Compare the three complementising strategies: with naka (57a), juxtaposed (57b),
and with po and (57c).
(57) Tawala
a. Hi-i-wogatala [naka apo iyowai hi-na-bagibagi].

3pl-dur-plan dx2 fut how 3pl-pot-work

They were planning how they would work. (Ezard 1997:222)
b. Hi-i-wogatala [hi-na-bagibagi].

3pl-dur-plan 3pl-pot-work

They are planning to work.
c.

Hi-wiwogatala [po apo hi-na-bagibagi].

3pl-plan
conj fut 3pl-pot-work
They planned to work.
(Ezard 1997:224)
(Ezard 1997:248)
7.4.4 Conditional clauses with optional naka

Condition clauses are generally in sentence initial position and are optionally followed
by dx2 naka (58); counterfactual conditions are marked by the hypothetical marker
inapa or itapa if and are optionally topicalised with dx2 naka or correlated by the
consecutive marker yaka so.
(58)

Tawala
I-na-dumol-i
(naka) a-nae.
3pl-pot-calm-3sg dx2.tpc 1sg-go
If it is calm, Ill go.
(Ezard 1997:242)
7.5 Discussion
When used as conjunctive markers, demonstratives may express sequencing and consecutive functions (as in Takia), but they mostly function as subordinators (in relative,
complement or adverbial clauses). As adverbial clause markers, they appear in time, conditional, causal and explicative clauses. Their semantics are contextual and the semantics
of the clause complex are inferred and depend on other collocated morphemes (adverbs,
conjugations or T.A.M. morphemes). Some of these morphemes are not always fully
grammaticalised as conjunctions; some are endophoric demonstrative operators with
conjunctive functions when their scope extends over a clause or sentence.
8. The syntactic and pragmatic functions of clause-ordering

Clause ordering is another pragmatic and syntactic indicator. Sequential (in line) constructions following the logical order of implication are the closest to coordination;
they may dispense with connectors when the semantic relation is straightforward (as
in the doctor arrived, (and) he examined the patient); if not straightforward, specific
coordinators express the intended semantic relation (consecutive, adversative, counter-expectancy, etc.) as in the doctor arrived, (but) he did not examine the patient).
8.1 Some general facts

While the sequential and logical order of events is the neutral order for chronological
and causal coordinate relations, non-chronological and anti-causal order is frequently
used to express anteriority or concessive meanings. These coordinate clauses in inverted
order are often coupled with time adverbs like already to mark non-chronological
events (ex. I arrived and/but John had already left; they may also be coupled with
negation markers as in Manam (ex. I arrived and/but John had not left meaning I
arrived before John left). Similar clause ordering strategies are used for causal relations.
As shown below, Manam and Korafe for example use anti-sequential order and focusing strategies to reverse chronological or logical (causal) relations.
8.2 From sequential coordination to subordination in Manam

Compare the sequential order of coordinate clauses with the coordinator =be and
in (59a) in Manam, with the inverted order associated with the focus marker (=be)
in (59b). Reversing the sequential order and focusing with clause final, focal =be
creates pragmatic hierarchy which is interpreted as subordination (see Bril in press
for discussion).
(59) Manam (SOV, Oceanic, Papua New Guinea, Madang Province)
a. Nra
malpi n-gu
di-lba=be,
tgo u-pra.

yesterday work cl-poss 1sg 3pl.r-be.big=and neg 1sg.r-come

Yesterday I had a lot of work and so I didnt come. (Lichtenberk 1983:524)
b. Tgo u-duma-o, Bi u-lo=be.

neg 1sg-help-2sg Bieng 1sg-go=foc

I did not help you (because) I went to Bieng.
(Lichtenberk 1983:548)
Adpositional causal subordination is used when the cause is backgrounded information; the cause clause is then in topic, sentence-initial position and is marked by
the similative postposition bana like32 which is the only subordinator in Manam
(Lichtenberk 1983:548).
c.

a-sege-a--u-ru
bana ma u-soai.
2pl.r-not.like-tr-1sg-bf-du cause here 1sg.r-live
Since you two dont like me, I live here.
(Lichtenberk 1983:548)
32. The postposition bana like has various subordinating functions (causal, manner,
similative-comparison, conditional, and counterfactual irrealis conditional (Lichtenberk
1983:37275, 528529, 533).
Isabelle Bril
Inversion of the sequential and logical order disrupts the implication ((if) x then/so (y)
and creates asymmetry between clauses. This semantic asymmetry may be strengthened by informational hierarchy (topic or focus) devices or by referential hierarchy
devices (demonstratives), as in Korafe.
8.3 From adversative coordination to concessive subordination in Korafe

In Korafe (Papuan), the linkers avata that, but and amo that, are members of a set
of originally demonstrative Dx2 markers, with topic/focus and conjunctive functions.
As a clause-linker, avata but conjoins clauses with adversative relationships or heads
concessive apodoses; it also marks contrastive topics or focus. Functions vary with
position and scope: as a contrastive coordinator in (60a), avata but heads the second
clause; as a concessive subordinator in (60b) avata occurs in clause-final position of
the contrastive frame clause to which it belongs prosodically. The negation marker in
the second clause bars the expected implication and triggers a concessive reading.
(60) Korafe (Papuan, Cape Nelson, PNG)
a. Oja-da
kau-mo jarusa-da
kaugo-ri,
[avata

shrimp-gen kind-tpc crayfish-gen kind.like-be dx2.ct.frus

memyako
bek-ri].
small.red.dim true-be

Freshwater shrimps are like crayfish, but they are very tiny.

(Farr 1999:252)
b. [Nunda guka titifa-ghae
avata],
aimi
jo

3s.gen back spine.red-com dx2.ct.frus that.agt neg

ga-y-ae
e-raira.
spear-epen-not.do do-cust.3s
Although it (mangrove ray) has spines on its back, it does not sting
(people) with them).
(Farr 1999:252)
The logical inversion of clauses, the different bracketing and scope of avata, modify its
functions and readings. Compare with the simple additive coordinate clause in (61):
(61)

Korafe
Bosivara-mo mindafu, nunda tamo ingago-ri.
porpoise-tpc big
and 3s.gen body black-be
The porpoise is big, and its body is black.
(Farr 1999:114)
Amo, also originally a deictic (a that near addressee +-mo topic/focus marker) is used
as a topic marker in categorical predications (62):
(62)

Korafe
Rika-jawo-mo uufa-ri.
bird-name-tpc W.Wagtail-cop.3sg.fn
The name of (this) bird is (the) Willy Wagtail.
(Farr 1999:114)
It also marks condition clauses as presupposed subordinate frames (63a); compare

with coordination in (63b), expressing consecutive relation.
(63) Korafe
a. Namonde dubo-kot-arera
amo,
fakina-arera.

1pl.exc neck-think-fut.1pl-fn dx2.t/f strength.do-fut.1pl-fn

If we reason well, well become strong.
(Farr 1999:253)
b. Namonde dubo-kot-arera
, fakina-arera.

1pl.exc neck-think-fut.1pl-fn and strength-do.fut.1pl-fn

We will reason well and well become strong.
(Farr 1999:253)
Amo may also function as a complementiser, as in (64a), or in frame-comment relative

constructions (64b); clauses are prosodically separated by a pause:
(64) Korafe
a. Nu kote-tira
amo,
uvu kafuru-ri.

3sg think-aor.3s.fn dx2.t/f water deep.water-be

He thought that the water was deep.
(Farr 1999:279)
b. Gagara vare-da
a-ira
amo,
nan-da komana-ri.

girl
garden-loc go-aor.3s.fn dx2.t/f 1sg-gen friend-be

The girl who went to the garden is my friend.
(Farr 1999:281)
9. Final discussion and conclusion

This study of clause-linking in Austronesian-Oceanic (and a few Papuan languages)
has shown a variety of strategies (coordination, clause-chaining, subordinate clauses),
marked by conjunctive markers, less or non-finite verb forms (reduplication, nominalisation), adpositional markers (mostly for cause or concomitance), mood markers
(contrasting irrealis vs. realis markers) in clause-chaining (Takia), in conditional and
some complement clauses.
The focus has been on types of subordination based on informational hierarchy
(topic and focus markers), or referential hierarchy markers (articles, definite markers,
demonstratives), as well as on clause ordering.
Clauses involving sequential, logical and causal continuity are mostly expressed
by (syndetic) coordination, clause sequencers, and clause chaining strategies; while
less or non-finite clauses (such as tail-cueing strategies) are mostly used with resumptive, anaphoric and backgrounding functions (see Table 1 below). Case-marking is
another possible strategy: in Roviana, adverbial subordinate clauses display neutral case
marking (in contrast with the ergative-absolutive system of main clauses) and occur
in sentence-initial position as adsentential foci/topic clauses. Similarly, in Hua (Papuan), conditional clauses, which are presupposed, are in the nominative, and Haiman
(1980: 412) describes them as functional equivalents of Latin ablative absolute constructions (see Thompson & Longacre 1985: 200203).
Isabelle Bril
Similar facts occur cross-linguistically, in Dyirbal same reference is marked in

topic chains with absolutive case, sometimes using antipassive detransitivisation
(Dixon 1972: 71; cited in Dik 1997: 319). In Kiranti languages, adsentential topic
clauses (which are outside the main clause) are marked with the ergative absolute,
functionally equivalent to Latin absolute constructions (Bickel 1999:74).
The core function of ergative clauses in Kiranti is to signal a sentential
topic, a frame for the main assertion. The ergative case marker indicates the
source, the cause, as well temporal and spatial circumstances, it is used in tailhead linkages and has further been grammaticalised into a marker of causal
subordination.
(Bickel 1999:4045).
Conjunct participles and absolute constructions actually fulfil the same adsentential, topic function. Informational structure and their morphemes are thus used
as complex clause hierarchy markers and indicators of subordination, specifying the
referential and informational status of the propositional contents (in a frame
comment template, or a restrictor clause main clause template). Some of the informational structure markers often originate from a pool of connectors or coordinators:
additive or sequential connectors express clause sequencing, logical and topic continuity, while contrastive-adversative coordinators mark topic shift, counter-expectancy
and concessive values. In relative clauses (ad-nominal specifiers) and in complement
clauses, the use of former coordinators reanalysed as subordinators often arises from
the expression of some former additive asserted information, while the use of definite
or demonstrative markers generally refers to backgrounded and referential information
in the subordinate clause.
Referential hierarchy with endophoric demonstratives and deictic markers is
another source of subordinating devices via propositional anaphora and presupposition, prior to some other assertion. They serve as tracking devices, like the the former,
the latter, or as correlative markers, and they use the proximal-distal grades as distinct
pointing or selective devices (see Culiolis notions of pointage and flchage operations,
1990). Demonstratives also display pragmatic functions as focus or topic markers,
generally indicating topic shift or contrast.
In many Oceanic languages lacking subordinating conjunctions, informational
hierarchy strategies are the only markers of clause hierarchy; they are inherent to the
syntactic architecture of the complex clause, not a peripheral discourse level added to
the syntactic level. The mere fact that informational and referential hierarchy should
be marked by syntactic morphemes and devices such as coordinators, demonstratives,
clause order, case markers, is an additional indication that these levels are interrelated
and inseparable from the grammar of clause complexes. To quote Lambrecht: Pragmatically structured propositions are () paired with appropriate lexicogrammatical
structure. (1994:334).
rel, time,
cond comp
CONJ; NMZation+
postp. (cause,
location)
embedding
(perception);
NMZation +adpos.;
COMPtizers
seq, result contrast,

disj
Tawala
(PNG)
Table1. Summary
cond + focus
marker (cause)
rel, time, cond
time, purp.,
cond
rel, seq
cause-result,
comp
NMZation
mode mkrs
+
intention,
volition,
cognition
speech
+
seq contrast
disj
seq, simult
+
seq disj
juxtaposed &
postposed to
verbal head
volition cognition
speech
postpos. (cause,
result, manner)
Takia
(PNG)
Manam
(PNG)
rel, comp
time, cond
rel
seq
seq
Sobei
(Irian Jaya)
equat, rel
time, cond
preposition
(speech, cause)
seq, contrast
result disj
SS or DS
pronouns
Kaulong
(New Britain)
tail-linking
complementation
clause chaining
equat, rel,
comp
rel
definite demonstr.
deictics
topic mkr
Topic clauses
marked by
asyndetic Coordination
syndetic Coordination
CONJ (purp.,. subordination

time, condition)
COMPtizer
seq, simult
disj
seq
Nlmwa
(New Caledonia)

Isabelle Bril
Abbreviations
abs
abl
acc
act
advs
abil
agt
anaph
aor
art
ass
attrib
ben
bf
card
caus
com
comp
cm
conj
cont
cop
ct
cust
dep
deict
def
ctrst
delim
dem
dep; d
dim
dir
disj
du
dur
emph
erg
absolutive
ablative
accusative
actor voice
adversative
abilitative
agent
anaphoric
aorist
article
assertive
attributive
benefactive
buffer
cardinal numeral prefix
causative
comitative
complementiser
class marking suffix
conjunctive
continuous
copula
contrastive topic/focus
customary, gnomic
dependency marker
deictic
definite marker
contrastive topic
delimiter
demonstrative
dependent marker
diminutive
directional
disjunctive
dual
durative
emphatic
ergative
exc
epen
f
fam
fn
foc
fr
frus
fut
gen
ipf
incons
incl
indef
instr
int
loc
m
neg
neu
nm
nmz
nom
n.pivt
obl
p
pat
pft
pivt
poss
pot
psi
prep
pro

pron
R
R.D
exclusive
epenthetic insertion
feminine
familiar
finite
focus marker
free pronoun
frustrative
future
genitive
imperfective
inconsequential marker
inclusive
indefinite marker
instrumental
intentional, intentive
locative
masculine
negation
neutral marker
noun marker
nominaliser
nominative
non pivot
oblique
proper noun mrk
patient
perfect marker
pivot
possessive
potential
person indirect
preposition
intermediate clausal
proform
pronoun
realis
realis dependent marker
reas
rec
red
ref.red
refl
rel
seq
reason marker
reciprocal
reduplication
referential reduced voice
reflexive
relative clause marker
sequential
stat
term
t/f
tpc
tr
trv
val
stative
terminative
topic/focus marker
topic
transitive
transitiviser
valency-changer
References
Bickel, Balthasar. 1993. Belhare subordination and the theory of topic. In Studies in Clause
Linkage, Karen H. Ebert (ed.), 2355. Zrich: Arbeiten des Seminars fr Allgemeine
Sprachwissenschaft.
Bickel, Balthasar. 1999. From ergativus absolutus to topic marking in Kiranti: A typological perspective. Proceedings of the 25th Annual Meeting of the Berkeley Linguistics Society, 3849.
Bickel, Balthasar. 1999. Nominalization and focus constructions in some Kiranti languages. In
Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren W. Glover (eds), 271296.
Kathmandu: Royal Nepal Academy.
Brewis, Richard & Levinsohn, Stephen H. 1991. Topic and emphasis in Timugon Murut. In
Thematic Continuity and Development in Languages of Sabah [Pacific Linguistics C-118],
Stephen H. Levinsohn (ed.), 2943. Canberra: Australian National University.
Bril, Isabelle. 2001. Postmodification and the structure of relative clauses in Nlmwa and
other Kanak languages of New Caledonia. In Proceedings of the Second International
Conference on Oceanic Linguistics, Vol. 2: Historical and Descriptive Studies [Pacific
Linguistics 505], B. Palmer & P. Geraghty (eds), 261284. Canberra: Australian National
University.
Bril, Isabelle. 2002. Le Nlmwa, analyse syntaxique et smantique [LCP 16]. Paris: Peeters.
Bril, Isabelle. 2004. Coordination and inclusory constructions in New Caledonian and Oceanic
languages. In Coordinating Constructions [Typological Studies in Language 58], Martin
Haspelmath (ed.), 499534. Amsterdam: John Benjamins.
Bril, Isabelle. In press. Coordination, information hierarchy and subordination in some
Austronesian languages. In Converbs, Medial Verbs, Clause Chaining and Related Issues,
Azeb Amha, Christian J. Rapold, Sascha Vllmin & Silvia Zaugg-Coretti (eds). Frankfurt:
Frankfurter Afrikanistische Bltter 17.
Corston-Oliver, Simon H. 1996. Ergativity in Roviana, Solomon Islands [Pacific Linguistics
B-113]. Canberra: Australian National University.
Corston-Oliver, Simon H. 2002. Roviana. In The Oceanic Languages, John Lynch, Malcolm Ross &
Terry Crowley (eds), 467497. Richmond: Curzon Press.
Croft, William. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: OUP.
Culioli, Antoine. 1990. Pour une linguistique de lnonciation, Tome 1: Oprations et reprsentations. Paris: Ophrys.
Dik, Simon. 1997. The Theory of Functional Grammar, ed. By Kees Hengeveld. Berlin: Mouton
de Gruyter.
Isabelle Bril
Dixon, Robert M.W. 1988. A Grammar of Boumaa Fijian. Chicago IL: The University of Chicago
Press.
Ezard, Bryan. 1997. A Grammar of Tawala: An Austronesian Language of the Milne Bay Area,
Papua New Guinea [Pacific Linguistics C-137]. Canberra: Australian National University.
Farr, Cynthia J.M. 1999. The Interface between Syntax and Discourse in Korafe, a Papuan Language of Papua New Guinea [Pacific Linguistics C-148]. Canberra: Australian National
University.
Foley, William A. & Van Valin Jr., Robert D. 1985. Information packaging in the clause. In Language Typology and Syntactic Description, Vol. I: Clause Structure, Timothy Shopen (ed.),
282364. Cambridge: CUP.
Foley, William A. 2007. A typology of information packaging in the clause. In Language Typology and Syntactic Description, Vol. I: Clause Structure, 2nd edn), T. Shopen (ed.), 362446.
Cambridge: CUP.
Fontinelle (de la), Jacqueline. 1976. La langue de Houalou (Nouvelle-Caldonie): Description
phonologique et description syntaxique [TO 17]. Paris: SELAF.
Franois, Alexandre. 2000. Contraintes de structures et libert dans lorganisation du discours. Unedescription du mwotlap, langue ocanienne du Vanuatu. Thse de Doctorat en
Linguistique, Universit Paris-IV Sorbonne.
Givn, Talmy. 1980. The binding hierarchy and the typology of complements. Studies in
Language 4(3): 333378.
Haiman, John. 1978. Conditionals are topics. Language 54(3): 564589.
Haiman, John. 1980. Hua: A Papuan language of the Eastern Highlands of New Guinea [Studies
in Language Companion Series 5]. Amsterdam: John Benjamins.
Haiman, John. 1988. Inconsequential clauses in Hua and the typology of clauses. In Clausecombining in Grammar and Discourse [Typological Studies in Language 18], John Haiman
& Sandra A. Thompson (eds), 4969. Amsterdam: John Benjamins.
King, Julie K. 1991. Thematic continuity and development in Tombuono narrative discourse. In
Thematic Continuity and Development in Languages of Sabah [Pacific Linguistics C-118],
Stephen H. Levinsohn (ed.), 7592. Canberra: Australian National University.
Kiss, Katalin . (ed.). 2001. Discourse configurationality. In Language Typology and Language
Universals, Vol. 2, Martin Haspelmath, Ekkehard Knig, Wulf sterreicher & Wolfgang
Raible (eds), 14421455. Berlin: Walter de Gruyter.
Krifka, Manfred. 2007. Basic notions of information structure. In Interdisciplinary Studies of
Information Structure 6, Caroline Fery & Manfred Krifka (eds). Potsdam: University of
Potsdam.
Lambrecht, Knud. 1994. Information Structure and Sentence Form. Topic, Focus and the Mental
Representation of Discourse Referents. Cambridge: CUP.
Levinsohn Stephen H. (ed.). 1991. Thematic Continuity and Development in Languages of Sabah
[Pacific Linguistics C-118]. Canberra: Australian National University.
Lichtenberk, Frantisek. 1983. A Grammar of Manam [Oceanic Linguistics Special Publication
18]. University of Hawaii Press.
Lynch John, Ross, Malcolm & Crowley, Terry (eds). 2002. The Oceanic Languages. Richmond:
Curzon Press.
Miller John & Miller, Carolyn. 1991. Thematic continuity and development in coastal Kadazan narratives. In Thematic Continuity and Development in Languages of Sabah [Pacific
Linguistics C-118], Stephen H. Levinsohn (ed.), 105135. Canberra: Australian National
University.
Milner, George B. 1972. Fijian Grammar. Suva: Government Press.

Mithun, Marianne. 1988. The grammaticization of coordination. In Clause Combining in
Grammar and Discourse [Typological Studies in Language 18], John Haiman & Sandra
A. Thompson (eds), 331359. Amsterdam: John Benjamins.
Moyse-Faurie, Claire. 1983. Le drehu, langue de Lifou (Iles Loyauts) [Langues et Cultures du
Pacifique 3]. Paris: SELAF.
Ozanne-Rivierre, Franoise. 1976. Le iaai, langue dOuva (Nouvelle-Caldonie) [TO 20]. Paris:
SELAF.
Ozanne-Rivierre, Franoise. 1979. Textes nemi, Vol. 12 [TO 31, 32]. Paris: SELAF.
Ozanne-Rivierre, Franoise. 1998. Le nyelyu de Balade (Nouvelle-Caldonie) [LCP 12]. Paris:
Peeters.
Rivierre, Jean-Claude. 1980. La Langue de Touho. Phonologie et grammaire du cmuh [Tradition
Orale 30]. Paris: SELAF.
Ross, Malcolm D. 1993. Describing inter-clausal relations in Takia. In Topics in Descriptive Austronesian Linguistics, Ger P. Reesink (ed.), [Semaian 11], 4085. Leiden: Rijksuniversiteit
Leiden.
Ross, Malcolm D. 2002. Takia. In The Oceanic Languages, John Lynch, Malcolm Ross & Terry
Crowley (eds), 216248. Richmond: Curzon Press.
Ross, Malcolm D. 2004. Aspects of deixis in Takia. In Deixis and Demonstratives in Oceanic Languages [Pacific Linguistics 562], Gunter Senft (ed.), 1536. Canberra: Australian National
University.
Sterner, Joyce & Malcolm D. Ross. 2002. Sobei. In John Lynch et al. (eds), The Oceanic
Languages, 167185. Richmond: Curzon Press.
Throop, Craig & Malcolm D. Ross. 2002. Kaulong. In Lynch et al. (eds), The Oceanic Languages,
387409. Richmond: Curzon Press.
Thompson Sandra A. & Longacre, Robert E. 1985. Adverbial clauses. In Language Typology
and Syntactic Description, Vol. 2: Complex Constructions, Timothy Shopen (ed.), 171234.
Cambridge: CUP.
Van Valin Jr., Robert D. 2005. Exploring the Syntax-Semantics Interface. Cambridge: CUP.
Comment clause
Crossing the boundaries between simple
and complex sentences
Zygmunt Frajzyngier
University of Colorado
The purpose of this study is to demonstrate the existence of the category

comment clause whose range of use by far exceeds the traditional
comment-on-topic category. In addition to a topicalized noun, the comment
clause may have in its scope an element in focus, an adverbial phrase, and
most important, a wide variety of clauses. The distinction between the many
kinds of complex sentences and the simple sentence with a comment clause boils
down to what element is in the scope of the comment clause. The data for the
present study come from Wandala (Central Chadic), where the comment clause
is overtly marked.
The importance of this study transcends Wandala or Chadic languages, as it
provides an explanation, applicable cross-linguistically, for the formal syncretism
involving a variety of functions considered hitherto unrelated.
1. Introduction
1.1 The purpose and scope of the study
The purpose of this study is to demonstrate that there exists a type of dependent clause
that can be used in a variety of environments. I shall call it a comment clause.1 In
addition to a topicalized noun, the comment clause may have in its scope an element
in focus, an adverbial phrase, as well as a wide variety of clauses. The importance of
. The term comment clause is used in traditional linguistics in at least two senses. One is
the distinction introduced by the Prague School between the topic and comment. This distinction is fully compatible and fully subsumed under the expanded notion of comment clause as
proposed here. The other sense of the term comment clause is used in descriptive studies of
English for parenthetical clauses within another clause (Mathews 1997:61, Quirk et al.1985).
There is no connection between that understanding of the term and the category described
in this paper.
Zygmunt Frajzyngier
the category comment clause as defined in this study is that it can occur in simple or
complex sentences. The distinction between the simple sentence and the wide range of
complex sentences with a comment clause boils down to what element is in the scope
of the comment clause.
The present study is based on Wandala (Mandara), a Central Chadic language,
spoken in the Far North Province of Cameroon and in northeastern Nigeria.
The interest of the comment clause in Wandala is that it is used in functions that
in many languages have been analyzed as belonging to different functional domains,
such as clausal complementation and subordination, topicalization, temporal and
conditional apodosis, clausal complements of a noun phrase, and even afterthought
clauses. Here is a non-exhaustive sample of various contexts in which the comment
clause is deployed in Wandala. The comment clause is marked by the particle w,
whose structure and derivation is described below:

Complements of verbs of saying:
(1) b pll--tr,
w n kn dgy
3sg say one-gen-3pl com pres now comp
One among them says, thus

kw dyr dgy
kw dy--rw
dgy
exist knowledge-gen-1sg comp
I have my knowledge, behold.
Temporal apodosis:
(2)

md dm mlrw
m d-d m mVr w
hyp go-go hyp now com
And when he [the groom] goes,
lv
tr nn b1`mm
d z` dd
speech 3pl def all
pred go hl father
everything that they have to say is with the father.

Conditional apodosis:
(3)

Vm m cn-kr-ncn jbon` nw
Vm
m cn-kr-n-cn
jb n n w
assembly hyp hear-2pl-3sg-hear matter def dem com
Listeners, if you understand this matter,
'k-r gn
'k r g n
neg.ex over to dem
there is no greater [truth/goodness] than that.
Comment clause
Comment on a topicalized noun phrase:
(4)

ywlv r'k

y-w
lv--rw
'k
well 1sg-com word-gen-1sg neg.ex
Well, I have no words.
(5)

mb k`V kl w jy k`gy
m b k`V kl
w j-y-
k
g` y
hyp foc count calculus com 3sg surpass-1sg-vent neg to 1sg
If it concerns counting, nobody surpasses me.
Comment on a reason clause:
(6) m bn b1`m mrmy mndymy

m-
bn
b1`m r--my
m n dy--my
hyp-pred concern all
head-gen-1incl 1incl assc knowledge-gen-1incl

w grn mh gmy
w gr n m
hg--my
[not in the recording]
com well def 1incl happy-gen-1incl
Because we have all of this knowledge, it is good.
As far as we are concerned, we are happy.
[The expression hg--my is not clear to other speakers.]
An afterthought clause:
(7) lv nn y s`
nd-n-e
n kw [error] w
story def 1sg come tell-3sg-1sg dem
com
The story I came to tell

kt lv ndz
kt lv ndz
3sg want story past
He wants an old story.
Unlike in many languages, the marker of the comment clause in Wandala has no
epistemic or deontic function, nor does it serve to separate arguments of complex
sentences (for both functions of complementizers see Frajzyngier 1996).
1.2 Typological characteristics of Wandala

In order to appreciate the role and the form of the marker w, some information about
the forms of morphemes is in order. All morphemes in Wandala can have three forms
within the clause: the root form which for most morphemes ends in a consonant. If
the syllable or consonant cluster conditions so require, an epenthetic vowel may be
inserted after the consonant. The epenthetic vowel is high front in the environment of
Zygmunt Frajzyngier
palatal segments, high round in the environment of round segments, and high central
in other environments. The epenthetic vowel is e before a pause:
(8) y gdz` knnn zw btnmS tr mm rg dd
y
gdz` knn-n
zw
b t-n
therefore child c.foc-intens first of all foc 3pl-assc
mS tr
mm
r g dd
attach 3pl pred mother pred on to father
And therefore, as for the daughter, she is more attached to her mother
than to her father.
Morphemes have the root form when the material that follows them is the expected
follow up. The lexical or grammatical morphemes have the root form before determiners; transitive verbs have the root form before objects in perfective aspect; all verbs
have the root form before adverbs; and most morphemes have the root form before
prepositions; and in one type of the head-modifier construction, the head has the root
form. Only some constructions are illustrated here.
Root form of the source extension s (with epenthetic high-central vowel) before
the object:
(9)

pm tnbps` zrvn t` bc
pm t nb pw-s` zrv-n

t` bc
onom 3pl then pour-s sesame-dem pred t mat
Pum! They poured out sesame onto a mat.
Here is an illustration of the root form of the verb (with epenthetic high-front vowel)
before the nominal object:
(10)

ttsmd l
t ts
md-
l
3pl get-up people-gen man
People of the groom get up,
ts`b`9y mdmks
t s`
b`9y md-
mks
3pl come meet people-gen woman
they come to find the people of the bride.
In one type of the head-modifier constructions, the head has the root form:
(11)

y lv wndl nn fyr mlrw [error]

y lv
wndl nn f-y--r
ml-rw
well speech Wandala dem 3sg put-1sg-go-on elder-1sg
This Wandala speech is authorized by my boss.
Comment clause
All prepositions have the root form before their complements:

(12) no v tn kn njnnj twfk ordinater yndl vwndl
no v t-n
kn nj-n--nj
t wfk-
pres time t-dem c.foc sit-1sg-ep-sit pred t face-gen
ordinater y nd lv
wndl
computer 1sg speak language Wandala
Today, I am sitting in front of the computer, I am speaking the Wandala
language. (T is the preposition coding locative targets)
The root + a form characterizes morphemes when they are followed by the
nominal subject:
(13)

`bVysbVy dks
`bVy--s-bVy- dks
fall-go-s-fall-pb thing:q
Something fell out?
The root + is used in a variety of structures which involve connecting two elements
that do not constitute an expected, natural sequence. The vowel with a high tone is
otherwise a goal marker and the locative predicator, required when the predicate of a
locative predication is not inherently locative. One type of genitive construction where
the second noun is not the natural modifier of the first:
(14)

VV lv h`r1 [m added in analysis]

V--V-
lv-
h`r1 [m
finish:3sg-appl-finish-pb business-gen farm [and
When their farming finished
Intransitive verbs code the ensuing noun phrase as the object through the root
+ ending:
(15) t d` hl-t zrv tr nn t pm d-m bh
t d` hl-t-
zrv--tr
nn t p-m
d-m bh
3pl seq gather-t-go sesame-gen-3pl def 3pl pour-in pred go-in bag
And then they gathered their sesame in and poured it into a bag.
(the verb hl gather is intransitive)
The comment clause marker w always ends in the high tone vowel . The form w is
most likely a combination of two morphemes: the form w, which is the consonantal
component of the remote deictic, and the question word what:
(16) a.

bn kndg kbun
b n k nd-n
g` w k bw n
3sg say 3sg 2sg say-go-3sg to what 2pl two dem
He says, Why do you say the two of you?
Zygmunt Frajzyngier
b.

, kndnggdzrw mk9y
, k nd-n g w
gdz--rw
m
k9y
ah, 3sg say-3sg to what child-gen-1sg 1incl three
Why did you say [so] my son? There are three of us.
(17) `s-t-w
lrus -m
xx-
wndl
dem-dem-dem marriage pred-in land-gen Wandala
That is the marriage in the Wandala land.
The high-tone vowel added to the question word/demonstrative marker w indicates

an unexpected connection with the ensuing element.
Lexical items and grammatical morphemes consistently have the root form before
the marker w, as they do before all other determiners.
2. Functions for which the comment clause is deployed

The comment clause is syntactically dependent, as evidenced by the fact that it is
never the sole component of a sentence. Moreover, its dependency is within the
sentence rather than within the discourse. This property of the comment clause
is fully confirmed by all natural language examples in the data gathered, many of
which are included in this study. Moreover, the element on which the clause is a
comment always precedes the comment clause. Again, the evidence is provided by
all natural language examples in the data gathered as well as by all examples cited
in this study.
The comment clause is also pragmatically dependent, as evidenced by the fact
that it cannot be interpreted on its own. It has to be interpreted in connection
with the element that precedes it, not with the element that follows the comment
clause. The sections below describe the various syntactic environments in which
the comment clause occurs. The overall function of the clause preceded by w is
to serve as a comment on the preceding material, be it a clause, a noun phrase, or
an adverbial phrase.
The comment clause can be a complement of a verb of saying, a temporal and
conditional apodosis clause, or a complement of any fronted constituent.
2.1 Comment on a topicalized noun phrase

The comment clause marked by w follows the topicalized noun phrase. When
the comment clause occurs after the topicalized noun phrase (and only topicalized
phrase), the noun phrase preceding the comment-clause marker w ends in the vowel
a, marking a phrasal boundary. When the comment-clause marker w follows other,
Comment clause
non-topicalized constituents, it is preceded by the root forms of the morphemes that

precede it:
(18) y, nnw dc mn dd ddw
y, n n w dc m-n
dd dd w
well def def com then hyp-dem father father com
Sypk ddw mbrk
Sy-p
k
dd w -m
mb--r
k
become able-appl neg:pb father com pred-in house-gen-3sg neg
Well, as for that, if that is so, [it means that] the father was not able to be a
father at his home.
In the following example, the last word of the topic phrase kl ends in the vowel a:
(19)

mb k`V kl w jy k`gy
m b k`V kl
w j-y-
k
g` y
hyp foc count calculus com 3sg surpass-1sg-vent neg to 1sg
If it concerns counting, nobody surpasses me.
(20) nw dks`nn`w s mbn kSyn bddmy

n w dks` nn` w s m b n
def com thing def com only hyp foc intens
kSy-n b dd-my
3sg cool-3sg foc father-1incl
That, that thing, only God knows how to cool it.
(only God knows how to fix the parents behavior) (kSy-n cool + 3sg obj)
The comment marker may form a unit with the demonstrative n, where the demonstrative retains its final vowel. It appears that the division of labor between the two
markers is that the demonstrative marks the topic phrase and the form w anticipates
the comment clause. In the following example, the adverbial phrase is the topic:
(21)

n vc stw nw ygnk
n vc s-t-w
n-w
y g
nk
eh pres day dem-dem-dem dem-com 1sg contract marriage
On such and such day I am going to get married.
(22)

wy nw td krw

wy
n w t d
krw
pred yesterday dem com 3pl go:go Krawa
As for yesterday, they went to Krawa
(23) a.

y gdz dwln d` rn rr
y gdz dwl-n d` r-n
r--r
well young man-dem 3 go see-[pause] friend-gen-3sg
Well, the boy went to see his friend,
Zygmunt Frajzyngier
b.

r d r
r-
d
r
man-gen father poss-3sg
a friend of his fathers.
c.

tbkrd t`rw pwtr dz ddnr g`gyl

t b krd- t`r w pw-tr
d` z
3pl two all-gen 3pl com 3sg send:pl-3pl go side
dd -r
g`
gyl
father poss-3sg dest girl
Both of them, he sent them to the girls father.
Evidence that the comment-clause marker w is not a topic marker is provided by

its contrast with the demonstrative w, which topicalizes the first person pronoun. In
topicalizer function, the form does not have the vowel a:
(24)

ywlv r'k

y-w
lv--rw
'k
well 1sg-dem word-gen-1sg neg.ex
Well, I have no words.
2.2 Comment on a contrastive focus

Wandala has the contrastive focus marker knn. The comment on the contrastive
focus phrase may be preceded by the marker w. The importance of this fact is that
in some languages, e.g. English, complementizers share some of the functions of the
comment clause marker w and do occur in the comment-on-focus clause. In Polish
and Russian, complementizers do not occur in the comment-on-focus clause. Here are
examples from Wandala:
(25) y gdzar knn w mkb
y gdzar- knn w m k
now child-pl c.foc com hyp 2sg
trlv btr sn zmn
b-tr lv
b-tr s-n
zmn
talk-3pl speech 3sg say-3pl 3sg come-3sg time (F)
And now, with respect to children, when you talk to
them, they say it is modern times that brought it.
(26) e, bddtrnn kn txdm
e, b dd tr nn kn t xd -m
well 3sg say father 3pl def now 3pl bury pred-in
Comment clause
ftrnn` kn w
f--tr
nn` kn w
field-gen-3pl def c.foc com
And now, their father, who was buried in the field
bn kndg kbun
b n k nd--n
g` w k bu n
3sg say 3sg 2pl say-go-3sg to what 2pl two dem
He says, Why do you say you two?
2.3 Complements of reason clauses

Reason clauses can be marked by the sequence m- bn because. A reason clause
can be the first clause in a sentence. The complement of a reason clause may be marked
by the form w. The importance of this fact is again that English and other Western
European languages do not have a complementizer after a reason clause:
(27)

m bn b1`m mrmy mndymy

m-
bn
b1`m r--my
m n dy--my
hyp-pred concern all
head-gen-1incl 1incl assc knowledge-gen-1incl
w grn mh gmy
w gr n m
hg--my
[not in the recording]
com well def 1incl happy-gen-1incl
Because we all have this knowledge, it is good. As far as we are concerned,

we are happy.
(28)

mbyknvw 'kdbr
m b y-k-n-v-y
w 'k dbr
hyp foc refuse-2sg-3sg-appl-refuse com neg.ex solution (H,F)
Since you have refused him, there is no other solution.
(29)

mbhyknkw tgkdruw
m b hy-k-n-k
w tg
k d-ruw
hyp foc like-2sg-3sg-neg com nevertheless 2sg father-1sg
Although you do not like him, nevertheless you are my father.
The marker w can by itself mark a reason clause:

(30)

lv`nn nyn w
lv`
nn n y n w
matter def pres 1sg tell dem com
What I am telling you, is because
1b yykn ytpntp krr ckwng1

1b y kn y tp-n-tp
kr--r
ckwng1
because 1sg c.foc 1sg taste-3sg-taste fire-gen-3sg a little bit
I myself tasted its fire a little bit.
Zygmunt Frajzyngier
(31) gdzrm tpnw mn nzn mmgnw

gdzr--m
t p n w
child:pl-gen-1incl 3pl pour dem com
m n nz n
mmg n w
hyp dem what intens 3sg do
dem com
n`n twyntrn tftr rksnwr
n` n
t wy-n tr n t f-tr-r
ksw r
def intens 3pl want-3sg 3pl dem 3pl put-3pl-on idiot
head
The thing that makes our girls laugh around is that the men that the girls like
cheat on them. (ksw idiot, somebody who cannot talk)
2.4 Comment clause as a complement of verbs of saying

The de dicto complements of verbs of saying can be marked by the comment-clause
marker w, whose function subsumes that of the de dicto complementizer. In this, the
function of w corresponds to the function of the de dicto complementizer that in
English, que in French, chto in Russian, and dass in German (for the notion of de dicto
complementizers see Frajzyngier 1991):
(32)

e, bnw kt bk kmgnstrwya
e, b-n-
w kt b-k
e 3sg say-3sg-ep com 3sg want tell-2sg
He said, he wants you to say
(33) yndnenw db kwkySr tm h1r

y nd-n-y n w db kwky Sr
1sg say-3sg-1sg dem com because much work
t--m
h1--r
3pl-pred-in belly-gen-3sg
As I have said earlier, there is too much work in it [the world].
(34) b pll--tr
w n kn dgy
3sg say one-gen-3pl com pres now comp
One among them says, thus

kw dyr dgy
kw dy--rw
dgy
exist knowledge-gen-1sg comp
I have my knowledge, behold.
The comment-clause marker may precede any type of direct speech. While it is quite
common for de dicto complementizers in Chadic languages to precede direct or indirect speech (Frajzyngier 1996), this is not the case in Western Indo-European languages,
where a de dicto complementizer such as that in English cannot precede direct speech:
Comment clause
(35)

b` pll knw tSn`sk f dn

b pll kn w tSn` sk fd--
n
3sg say one c.foc com sorry pol lie-gen-2sg dem
The other said, Sorry, this is your lie.
(36)

b dd kV klw
b dd-
kV
kl w
3sg say man-gen counting units com
The one who can count all says:
9ywn9sk kg jw jbr
9y-w-n-9
sk kw g jw-
jb r
leave-imp:pl-3sg-leave pol 2pl take doubt-go type q
Leave that, please, why do you doubt?
(37)

b pll kn w
b pll kn w
3sg say one c.foc com
Another one said:
m kyr dks` nn s b y nnn

m ky-r dks nn s b y n-n-n
hyp where-q thing def then foc 1sg see-3sg-see
No matter where the object is, I will see it.
The marker w can precede a clause in the imperative, a function shared by complementizers in IE languages:
(38)

y 'llh nn Vb dly ccn

y 'l lh--
nn Vb
dly w
ccn
well sing song-gen-2sg def yet pred again com 1excl hear:imper
Sing your song again, so that we may hear it again.
(39)

, bS`ksw p pvg`nn
, b S`ks w p-w-p vg` nn
eh, 3sg say Sultan com dig-pl-dig grave def
And the Sultan said, Dig up this grave.
(p-w-p is a reduplicated imperative form)
The marker w can precede the imperative verb of the complement clause. The de dicto
complementizers of IE language cannot be used in a similar environment:
(40) bS`ksw dmbr ny'lmd
b S`ks w d-w
mbr
3pl say Sultan com go-pl.imp together
n y 'l md
d z` kr nn
pres 1sg send people pred go hl 2pl def
The sultan said, Go, I am sending people to your place.
Zygmunt Frajzyngier
The marker w alone may be a marker of de dicto complementation, even if the main
clause does not contain a verb of saying. This is the case with the first instantiation of
the marker w in the following example:
(41) dndv z d rw gyl` n
d ndv
z d
-r
w gyl`-n
3sg go ask
pred hl father poss-3 com girl-dem
hyn hy m d rw kt ynvy
hy-n-hy m d
-r
w kt
love-3sg-love but father poss-3sg com 3sg want 3sg
y-n-v-y
refuse-3sg-appl-refuse
He went to ask her father, saying that the girl loves him.
But her father wants to refuse him.
(42)

tsndvwlv kVna rgy w

t s
ndv--w
lv
kVn-a-r
g-y
3pl come ask-1sg-vent matter goods-gen-1sg to-1sg
They came to ask me about the matter of the goods
w 'k lv
kVn
com neg.ex matter goods
(I said) There is no matter of goods.
The only example of the comment-clause marker w occurring after a verb of perception is in an elicited sentence, hence its validity is somewhat doubtful:
(43) y n w t gy
1f
1sg see com 3pl prepare food
I saw that they prepared food.
The comment-clause marker w is not an obligatory component of the complements

of verbs of saying. Thus, it does not have to precede direct speech:
(44)

b` pll dd Sm
b pll dd-
Sm
3sg say one man-gen ear
The one who hears all said:
ily zrv m pll mb`1 mb`1 dm ywn

il-
zrv-
my pll mb`1-m-mb`1
dm yw n
sand-gen sesame-gen 1incl one fall-in-fall
pred go:in water dem
One of our sesame seeds fell into the water.2
. The use of the verb mb`1 here is a lexical error, as this verb codes a fall from a standing
position. It can be applied to people, trees, houses, etc. The verb vVy [bVy] fall should have
been used instead. In the subsequent description of the same event the speaker does indeed
use the verb bVy fall.
Comment clause
Compare a structurally identical fragment with the comment-clause marker w:

(45)

b dd kV klw
b dd-
kV
kl w
3sg say man-gen counting units com
The one who can count all says:
9ywn9sk kg jw jbr
9y-w-n-9
sk kw g jw-
jb r
leave-imp:pl-3sg-leave pol 2pl take doubt-go type q
Leave that, please, why do you doubt?
The comment marker w does not occur if the complement clause precedes rather
than follows the verb of saying, thus providing the evidence that in the contexts where
it appears, the form w marks the clause as a comment on the preceding material
within the same sentence:
(46)

sydttty yssns sbn

si y d ttty y s--n-s-s
b-n
then 1sg fut search 1sg come-go-3sg-s-come 3sg say-3sg
I will search, I will get it out, and bring it back, he said.
(47)

krn dlvmvgbn
k r
nd lv
m vg
b-n
exist person 3sg speak word in grave 3sg say-3sg
There is somebody talking in the grave, he said.
Most examples in which the comment-clause marker w does occur have some
material after the verb of the first clause and before the comment-clause marker w. It
is usually the subject of the clause. If the complement clause immediately follows the
verb of saying, the comment-clause marker w most often does not occur:
(48)

bn kndg kbun
b n k nd--n
g` w k bw n
3sg say 3sg 2sg say-go-3sg to what 2pl two dem
He says, Why do you say the two of you?
dd` rmvgkn bb

dd` r -m
vg kn b b
well father 3sg pred-in grave c.foc 3sg say foc

ki mk9ysk gdzrw
ki m
k9y sk gdz--rw
no, 1incl three pol child-gen-1sg
And his father in the grave says, No, we are nevertheless three, my son.
The comment-clause marker w does not occur if the complement clause uses
the deontic complementizer si then, therefore borrowed from Hausa or Fula.
Zygmunt Frajzyngier
The reason the form w does not occur is that its function is subsumed by the
complementizer si:
(49)

mVr betr set d l

mVr b tr si t d-
l
now (f) 3sg say 3pl then 3pl go-go travel
Now they say that they have to travel.
(50)

bn ky syd` sns` sbn

b n ky
si y d` s-n-s`-s
b-n
3sg say 3sg nevertheless then 1sg seq come-3sg-s-come 3sg say-3sg
He says, Nevertheless, I will get it out.
The comment-clause marker w is used if there is some other material between the
verb of saying and the complement clause, e.g. the subject of the main clause. The
comment-clause marker w is not used if the complement clause precedes the verb of
saying. If the complement clause has the deontic complementizer si the commentclause marker w is not used. This points out that the role of the comment-clause
marker w is thus to mark the clause as being a comment on the preceding material.
2.5 Comment on a relative clause

A comment clause can have a relative clause in its scope. The de dicto complementizers
of IE languages perform a similar function:
(51) knnz b dzt`rbdz`nw
kn nz bdz--t`r-bdz` n w
now what 3sg spoil-go-3pl-spoil dem com

gdzr mSn mm
gdzr
mS n mm
daughter 3sg attach assc mother
Actually, what spoils them is that the daughter is attached to her mother.
(52)

lv md mkllm mrw mkpe1

lv-
md mkllm-r
w
m-k
pe1
word-gen people denigrate-head com pred make-neg force
The words of denigrators are of no use.
2.6 Comment on an adverbial phrase

A comment clause can occur after an adverbial phrase. Complement clauses with a de
dicto complementizer cannot have adverbial phrases in their scope in IE languages:
(53) `st
w V-V
nr-w-ndz
nn
like.that com finish:vent-finish language-mouth-past def
Lit..Like that finished this story.
Comment clause
The adverbial phrase may be locative:

(54) m j dnn nnnw 'k rgfrv
-m
nj-
dn-n
n nn w 'k
pred-in life:gen world-dem pres def com neg.ex
-r
g f-rv
pred-on to put-heart
In this life here, there is nothing that surpasses patience.
2.7 Comment clauses after volitional verbs

A comment clause can follow a volitional verb. The role of the marker w is merely to
mark the comment clause rather than any deontic function. The deontic function is
coded by the preposition g` before the verb of the complement clause:
(55) y camnkn nz ktnrw
y camn kn nz kt-n r
w
well first of all c.foc what 3sg want-3sg person com
b g` nj-r-nj
n her-n-klp-r
foc to stay-3pl-stay assc peace (Ar)-assc-health (K.Ar)-nom
First of all, what one wants is for them to live in peace and good health.
2.8 Temporal and conditional apodosis clauses

The comment-clause marker w occurs in both temporal and conditional sentences,
but it marks only apodoses:
(56)

mh`r1` nn gdzr bmVVl dm ftrw

-m
h`r1` nn gdzr b m V-Vl d-m f--tr
w
pred-in farm def child foc hyp go-go go-in field-gen-3pl com
While farming, when the child goes to their field,
'l lhr
'l lh--r
3sg send song-gen-3sg
he sings his song.
(57) a.

md dm mVrw
m d-d-w
m mVr w
hyp go-go-vent hyp now com
And when he [the groom-to-be] arrives,
lv
tr-nn b1`mm
d z` dd
speech 3pl-def all
pred go hl father
everything that they have to say is to the father.
Zygmunt Frajzyngier
b.

dd`n`kn SypSy ddmbrw

dd` n` kn Sy-v-Sy
dd -m
mb--r
w
father dem c.foc mature-appl-mature father pred-in house-gen-3sg com
As for the father, once he is a father at his home,
mnd swe kn `gdzrw VpVy kVb

m nd-s we kn `gdz--r
w
hyp 3sg say-s what c.foc child-gen-3sg com
V-p-Vy
k Vb
3sg accept-appl-accept neg again
whatever he says, his child should accept, shouldnt she?
The marker w may follow the demonstrative n:

(58)

my'l lghnw
m y 'l lh n w
hyp 1sg send song dem com
When I sing the song,
krn dlvmvgbn
k r
nd lv
m vg
b-n
exist person 3sg speak word in grave 3sg say-3sg
there is somebody talking in the grave, he said.
(59)

l mvct`n dtty mks`

l m-vc -t`-n
d tty mks`
man hyp-time pred-t-3sg 3sg go search woman
The man, at the time when he will go to look for a woman,
nw ts Vl mnrr
n w -ts
-Vl m-n
r--r
dem com 3sg-get up 3sg-go hyp-assc man-gen-3sg
he gets up and goes either with his man
mS kt`rml
m Skt` r ml
or fellow 3sg another
or with one of his buddies.
The following sentence contains two instantiations of the comment-clause marker

w: one when it marks a temporal apodosis, and the other when it marks a comment
on the topicalized noun:
(60)

dd`n`kn SypSy ddmbrw

dd` n` kn Sy-v-Sy
dd -m
mb--r
w
father dem c.foc mature-appl-mature father pred-in house-gen-3sg com
As for the father, once he is a father at his home,
Comment clause
mnd swe kn `gdzrw VpVy kVb

m nd-s we kn `gdz--r
w
hyp 3sg say-s what c.foc child-gen-3sg com
V-p-Vy
k Vb
3sg accept-appl-accept neg again
whatever he says, his child should accept, shouldnt she?
The comment clause can function as a conditional apodosis:

(61)

bn mgyvgyn mgyrw
b-n m gy-v-gy-n
mgy--r
w
3sg say-3sg hyp cook-appl-cook-pause millet-gen-1excl com
He said, If our hungry rice (Digitaria exilis) is cooked
zn br ntrmrw tn

z-n b r
ntr m-rw
tn
1excl eat-3sg foc 1excl conj mother-1sg only (F)
we will eat it only with my mother.
(62) bpll n kn w mb n b jr bVy-s-bVy

b pll n kn w m-b n
3sg say one dem c.foc com hyp-foc intens
b jr bVy-s-bVy
foc true fall-s-fall

d-m h1 hy w
d-m h1-
hy w
go-in belly-gen river com
One among them said, If really [one grain] has fallen into the river,
sydttty yssns sbn

si y d ttty y s--n-s-s
b-n
then 1sg fut search 1sg come-go-3sg-s-come 3sg say-3sg
I will search, I will get it out, and bring it back, he said.
(63)

Vm m cn-kr-ncn jbon` nw
Vm
m cn-kr-n-cn
jb n n w
assembly hyp hear-2pl-3sg-hear matter def dem com
Listeners, if you understand this matter,
'k-r gn
'k r g-n
neg.ex over to-def
there is no greater [truth/goodness] than that.
The de dicto complementizers of IE languages do not mark either temporal or

conditional apodoses.
Zygmunt Frajzyngier
2.9 Comment clauses in counter-expectation predication

The comment clause can clarify the reason for a counter-expectation. Western IndoEuropean languages do not deploy the complementizer in the coding of this function.
(64) y y gr
well excl good
Well, it is good.

m w ycnncn lv` nn
m
w y cn-n-cn
lv`
nn
however com 1sg hear-3sg-hear matter def
However, it seems I have already heard about this thing.
mw gdzrVb ckw nd l y
m w gdz-r Vb ckw ndh ly
but com child-1sg still small 3 read lecture
But, my child [daughter] is still young, she is still in school.
3. Synthesis
The syntactic environments listed above do not have much in common in terms of the
categories that precede the form w. That indicates that the presence of the commentclause marker w is not triggered by the properties of the phrases or clauses that
precede it. In all cases, the marker w is followed by a clause, hence the evidence that it
is some type of clause marker. But not all clauses, even those in complex sentences, are
preceded by the comment-clause marker w. The fact that the presence of the marker
w cannot be predicted from the material that precedes it or that follows it, is the
evidence that it is a means for independent coding, a property shared by complementizers in other languages (Frajzyngier 1996).
The form w marks the ensuing clause as being a comment on what precedes
it. It is also a part of the preceding clause or phrase. The marker w indicates that
something else follows in the utterance and that it constitutes a comment on the
immediately preceding material.
4. Implications
The fundamental function of being a comment on a preceding element allows us to explain
a relatively large variety of syntactic constructions in which the comment clause occurs.
The comment clause is the only element that these constructions have in a common. It
explains rather well the complex sentence structures with the comment clause.
Comment clause
The widespread use of the comment clause puts into the question the fundamental notion of clause combining (Haiman & Thompson 1988), viz., does clause
combining represent an observation of the internal structure of some utterances or
is it an observation of the speakers motivation. Clause combining harkens to the
same assumptions that led to the notion of phrase structure rules, viz. combining
larger entities out of smaller entities. The use of the comment clause confirms the
assumption that speakers operate with different motivations, viz. on the coding of
various functional domains (Frajzyngier & Shay 2003). If the speaker wants to make
a comment on a noun or an adverbial phrase, the use of the comment clause results
in a simple sentence. If the speaker wants to make a comment on a proposition, the
result is a complex sentence.
Abbreviations
1
2
3
aff
appl
ar
assc
c.foc
com
comp
conj
dat
def
dem
ex
f
foc
gen
go
h
hyp
imper
in
hl
incl
First person
Second person
Third person
Affected
Applicative
Arabic
Associative
Contrastive focus
Comment marker
Complementizer
Conjunction
Dative
Definite
Demonstrative
Existential
Fula (Fulfulde)
Focus marker
Genitive
Goal
Hausa
Hypothetical
Imperative
Inner space
Human locative
Inclusive
k
Kanuri
m
Masculine
n
Noun, nasal consonant
neg
Negative
nomin Nominalizing
past Past time marker
pb
Phrase-final marker
pl
Plural
pol
Polite
pnct Punctual
pred Predicative marker
prep Preposition
pres Presentative
q
Question marker
refl Reflexive
rem
Remote
rq
Rhetorical question
s
Source
seq
Sequential
sg
Singular
t
Target
tog Together (multiple
participants)
tr
Transitive
vent Ventive References
Zygmunt Frajzyngier
References
Frajzyngier, Zygmunt. 1991. The de dicto domain in language. In Approaches to Grammaticalization, Vol.1 [Typological Studies in Language 19], Elizabeth C. Traugott & Bernd Heine
(eds), 219251. Amsterdam: John Benjamins.
Frajzyngier, Zygmunt. 1996. Grammaticalization of the Complex Sentence: A Case Study in
Chadic [Studies in Language Companion Series 32]. Amsterdam: John Benjamins.
Frajzyngier, Zygmunt & Jasperson, Robert. 1991. That clauses and other complements. Lingua
83: 133153.
Frajzyngier, Zygmunt & Shay, Erin. 2003. Explaining Language Structure through Systems
Interaction [Typological Studies in Language 55]. Amsterdam: John Benjamins.
Haiman, John & Thompson Sandra A. (eds). 1988. Clause Combining in Grammar and Discourse
[Typological Studies in Language 18]. Amsterdam: John Benjamins.
Matthews, Peter. 1997. The Concise Oxford Dictionary of Linguistics. Oxford: OUP.
Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey & Svartvik, Jan. 1972. A Grammar of
Contemporary English. London: Longman.
Acknowledgments
The present work on Wandala is supported by a grant from the National Science
Foundation to Zygmunt Frajzyngier and Erin Shay and by an award from the Jane and
Charles Butcher Foundation. Most of the data were gathered in Cameroon, where
I have been hosted over many years by the Institut de Recherche Agricole pour le
Dveloppement in Maroua, which has also provided me with much-needed institutional and logistic support. I am most grateful to its current director, Dr. No Woin,
for support during the years 20042008.
I am most grateful to Erin Shay for the careful reading of this paper, critical
and constructive comments, and many suggestions for improvement. Comments
by an anonymous reader and by Bernard Caron led me to revise substantially a
number of points discussed in this paper. To Marian Safran I am most grateful for
the editorial work.
Deixis, information structure and clause

linkage in Yafi Arabic (Yemen)
Martine Vanhove
LLACAN, Fdration Typologie et Universaux Linguistiques CNRS

Several particles in the Arabic variety spoken in the area of Yafi (Yemen) show
tight recurring links between deixis, informational hierarchy, and syntactic
hierarchy, both diachronically and synchronically. In the light of Roberts (1993,
2000) findings on focusing strategies, these links are discussed in detail for two
polyfunctional particles, ra, and ta, which can be used, among other things, as
deictics, topic markers, focusing particles and clause-linking devices (see Vanhove
1996 & 2004). The present paper shows how, in a language where subordinating
constructions and markers are also available, topicalisation and focusing
strategies may become the preferred strategies for clause-linking, especially in
causal, relative and complement clauses.
1. Introduction
This paper takes the case of the Arabic variety of Yafi (Yemen) to discuss a crosslin
guistically recurrent phenomenon, which is analysed in several papers of this volume,
namely a type of interclausal dependency, different from embedding, subordination,
or co-subordination, which is analysed by Robert (this volume) as the pragmatic
dependency of a focused clause on its discursive landmark.
Although dedicated clause linkage markers are not particularly scarce in Yafi
Arabic, spontaneous speech very frequently makes use of other strategies based on
information structure to signal semantic interclausal relations. In spontaneous oral
discourse, they may actually be more frequent than the syntactic strategies available
for some specific semantic relations.
In this paper I will describe and discuss the various functions of two particles,1 ra
and ta, whose origins are respectively a verb meaning see, look, and a demonstrative.
. The data used for this research were collected in the Yafi area in Yemen during two field
trips in 1994 and 1998. They consist of 45 oral traditional tales gathered from five female
speakers aged 20 to 50 and two male speakers aged approximately 40 and 50. Unfortunately
I have not been able to go back to Yemen since then and no further data could be gathered
or elicited.
Martine Vanhove
Both illustrate cases of reanalysis of deictic elements as discursive particles and clauselinking elements. Both came to express, among other things, a special type of interclausal
relation, i.e. with explanatory/causal semantics, co-existing with a dedicated causal subordinator lianna because.2 It will also be shown that one of the two particles, ta, encodes
not only causal adverbial clauses, but also relative and complement clauses.
To begin, here is a brief typological overview of the language to help understand
the following description and analysis. The Arabic variety spoken in Yafi, a mountainous district in central Yemen, belongs to the so-called -k dialect group and is located
in the latters most southernmost part (see Vanhove 1995a). The -k dialect group of
this central Semitic branch of Afro-Asiatic is characterized by the fact that the Perfective form of the verb is inflected, like South-Semitic languages, with a k- suffix, instead
of t- as in all the other Arabic groups. Yafi Arabic is a VO language in which both
VS and SV orders are possible, although the former is the most frequent. Subordinate
clauses (adverbial, completive and relative) follow the matrix clause, except for hypothetical clauses. Most are introduced by a subordinate marker, except completives for
which the marker is not compulsory. The verb in the subordinate clause, when there
is one, is a finite verb. After the quotative verb qal say, only direct reported discourse
can follow.
The verbal system is based on a binary aspectual opposition between the Perfective (inflected with suffixes indexing person, number and gender of the subject) and
the Imperfective (inflected with prefixes). Optional particles expressing the Perfect
(qa(d) or ra), Progressive (b or y) and Future (ba, a, (a) or ha) can be procliticized
to the verb (for further details, see Vanhove 1995b), and past tense reference can be
explicited with the auxiliary verb kan be.
Nominal (i.e. verbless) sentences (with Subject-Predicate word order) are very frequent: any noun (basic or derived), noun phrase, adjective or participle can be used
as a predicate, without the addition of a copula. Such nominal utterances are neither
marked for tense, nor for aspect, and are context dependent. For past reference, the
verb kan be may be inserted between S and P.
Yafi Arabic also has a wide range of topic and focus particles in assertive, interrogative and negative sentences, which are very commonly used, both in nominal and
verbal utterances.
. The present paper builds on two preliminary descriptions published in Vanhove (1996
and 2004). Partly based on more material collected during fieldwork in 1998, the analysis
provided here refines details and explanations, and corrects misinterpretations.
2. The particle ra
The origin of the particle ra goes back to the imperative form of a verb ra (cognate
form of classical Arabic raa) meaning look, see, which, unlike other Arabic varieties,
is still in use in the language as such, as in (1):
(1) ra-i
kef >atab-k
ri-t
il-fuxri
look.imp-f.sg how cut.pfv-1sg wing-gen.f def-cockroach
Look how I cut the cockroachs wing!
Very frequently, ra is also used in various contexts other than the imperative, and
it has become a polyfunctional particle. In the Arabic variety of Yafi, the particle ra
still retains traces of its verbal origin and cannot be considered as fully grammaticalized yet, in the sense that (a) its form most often varies with the number and gender
of the addressee: m.sg. ra, f.sg. ra-i, m.pl. ra-u, f.pl. raen, and (b) object pronouns canbe suffixed to it (these pronouns differ from the possessive set only in the
1st person singular).
Depending on its syntactic scope (pronoun, noun phrase, predicative nucleus,
clause), and on discursive parameters, the particle ra has different functions related
to either deixis, assertion, topic, focus, or clause linkage, i.e. presentative, copula, topic
marker, focusing particle, and inter-clausal marker.3
Such evolutions of visual perception verbs are not unknown to other Arabic varieties, mainly Maghribi. Moroccan Arabic (Colin 1948; Harrel 1962; Caubet 1992),
Algerian Arabic (Cohen 1924; Cohen 1984; Madouni 1993) or other Yemeni Arabic
varieties (Landberg 1909) for instance also underwent, to various extents, a grammaticalization process towards deictic, copular, aspectual, topic and focusing values. But to
the best of my knowledge, developments towards a clause-linking device have never
been reported. Nevertheless, some authors comments on the semantics of the various
constructions with the cognate forms of the particle ra imply that some potentiality
for such developments exists. Harrel (1962:215) for instance mentions the following
on Moroccan Arabic:
The use ofr is especially common when a sort of logical connexion or consequent
result is implied, e.g. xud taksi u rh iw6ssl 6k take a taxi and hell get you (s.g.)
there, dr6 b l-xadia tilifun, raha f6 -d-dar
give Khadija a telephone call, shes
athome.
. In addition, the particle is also a perfect aspect marker, a function which is beyond the
scope of this paper. It implies a direct grammaticalization path from the presentative function
to the perfect value (for further detail, see Vanhove 1995b).
Martine Vanhove
On the Yemeni Arabic variety of the Dathinah area, Landberg (1909:494) writes:
As for the meaning of all these particles, it should be observed that they do not
only signal a sudden or unexpected appearance of an item, or just the antithesis,
but also the consequence, the continuous situation or the raison dtre.4
The following sections will describe the different uses of the particle ra.
2.1 Presentative
One function of the particle ra is directly linked to deixis, as is also found in the other
above-mentioned Arabic varieties. When the syntactic scope of the particle is a pronoun,
a noun phrase or a relative clause, ra has the deictic function of a presentative with
scope over either an entity or a whole predicative relation, quite similarly to voici or voil
in French (whose origin is also the imperative form of the verb meaning see voir).
When ra simply has scope on a pronoun, and is followed by an independent verbal clause in the Imperfective, or by a coordinated clause, the sequence ra + pronoun
constitutes a nominal clause of its own. In this particular case, ra often highlights the
speaker himself, expressed as an object suffix pronoun:
(2) ra=ni
a-sir
a-xrog
pres.m.sg=obj1.sg 1sg-go.ipfv 1sg-go.out.ipfv
Here I am, I am going, I am leaving!
Example (3) is one of the rare instances when the particle remains invariable in the
3m.sg (there are several addressees, not just one):
(3) ell-u
aqq-at=kum [] min jom atte=hen
take.imp-m.pl gen-f=2m.pl [] abl day put.pfv.3m.sg=obj3f.pl
zag=i
baen ra=ni
wa
husband=1sg then pres.m.sg=obj.1sg and
la lames-ku=hen
wu la ag-ah
neg touch.pfv-1sg=obj.3f.pl and neg thing-f
Take yours! [] One day, my husband put them there and then here
I am and I am not able to touch them or anything.
When ra has scope over noun phrases or prepositional phrases (ex. 4, 5, 6), pointing out/referring to animate as well as inanimate, definite or indefinite entities, the
sequence ra (+ pronoun) + NP or PP constitutes a nominal clause. Unlike most
. Quant la signification de toutes ces particules, il est observer quelles ne dsignent pas
seulement lapparition subite et inattendue dune chose, ou seulement lantithse, mais aussi la
consquence, la situation continue ou la raison dtre. (Landberg commented upon several
other discursive particles, hence the plural form (these particles).
Maghribi Arabic varieties, in this case, pronouns cannot be suffixed to the particle,
except with indefinite nouns (4):
(4) dxul
ra=oh
rua
enter.imp.m.sg cop.m.sg= obj.3m.sg drizzle
Come in, its raining! (lit. here is rain)5
(5) qal
ha=l=ek
ra
ar-rab-eh
say.pfv.3m.sg dem=all=2m.sg pres.m.sg def-jug-f
He said: Take it, here is the jug!
(6) qal-ah
ma a
allah ikr
in-nabi
say.pfv-3f.sg what want.pfv.3m.sg God mention def-Lord
ra-i
maa=
i-nen 7wel-eh
pres-f.sg with=2f.sg def-two child.pl-pl
She said: God almighty, here you are with the two kids!
But when ra has scope over a relative clause (ex. 7), the object pronoun is obligatorily
suffixed to ra, and functions as the antecedent of the relative marker:
(7) qal-ah
ra=ah
i akal
say.pfv-3f.sg pres.m.sg=obj.3m.sg rel eat.pfv.3m.sg
el-lam-a u 6rib
el-maraq
def-meat-f and drink.pfv.3m.sg def-soup
She said: Here is the one who ate the meat and drank the soup!
2.2 Copula
In Yafi Arabic, nouns, prepositional phrases, participles and adjectives can also function as predicates of existence, property or location, without the need for a verb meaning be or a copula.
The particle ra may be used with the assertive function of a copula for both
nominal properties and locative utterances (not only for locative utterances ones as
inMoroccan Arabic, Caubet 1993:35). In my data, this occurs only when the subject
ofthe nominal utterance is a pronoun, expressed by the set of suffix object pronouns,
as when ra is used as a presentative. The word order is thus Copula Subject Nominal Predicate.
The use of ra does not only characterize or mark the property of a subject, as
a simple nominal thetic sentence would do, but, as in other Arabic varieties (Caubet
1992; Madouni 1993), the particle adds an important modal and discursive/pragmatic
. In this particular instance, the corresponding simple thetic sentence would be a onenominal constituent utterance: rua (lit. rain) its raining.
Martine Vanhove
component of intensity to the predicative relation, which can be glossed as S is truly,

really P:6
(8) qal-ah
ja ijal=i
ra-u=ni
bi
al-ah
say.pfv-3f.sg voc child.pl=1sg cop-m.pl=obj.1sg with state-f
She said: My children, I am (really) in a strange situation!
(9) qal-ah
qa=qul-ku=lu=kum
say.pfv-3f.sg perf=say.pfv-1sg=all=2m.pl
ra=oh
mulxof
cop.m.sg=obj.3m.sg mad.ptcp
She said: I already told you that he was (really) crazy!
2.3 Discursive hierarchy marker and clause linkage

The particles other functions specifically relate to informational hierarchy. Similarly to
the Australian language Umpithamu (Verstraete, this volume), but to a lesser extent, ra
can be used as a marker of discursive hierarchy covering both the functions of constrastive topic marker and of focusing marker. This is also the case for other Arabic varieties, although this terminology is hardly, if ever, used. Both Caubet (1992) and Madouni
(1993), within the framework of enunciative theory, mention the values of refocusing on
the initial constituent (recentrage sur le terme de dpart, i.e. topicalisation), and refocusing on the entire predicative relation (recentrage sur toute la relation prdicative).
2.3.1 Topic marker

ra can be used as a discourse marker of contrastive topic. Examples are scarce in the
data, and limited to 1st person pronouns. Instances are only found when ra is followed
by an object suffix pronoun which refers to the subject of a following imperfective verb
form (to which a future particle is cliticized in the following example). Pragmatically,
it signals that the speaker is pointing to herself and highlighting her own commitment
as the author of the action following a request from her interlocutor:
(10) qal-ah
tajjib ra=ni
say.pfv-3f.sg well top.m.sg=obj.1sg
ba=a-kudd=l=ek
maa bent=i
fut=1sg-bring.ipfv=all=2m.sg with girl=1sg
She said: Fine. As far as I am concerned, Ill send it to you with my daughter.
. The non intensive nominal utterance would use the independent subject pronoun followed by the nominal predicate: ani bi al-ah (1f.sg with state-f) I am in a strange state;
hu mulxof (3m.sg mad.ptcp) he is crazy.
2.3.2 Subject contrastive focusing

More frequently, ra is used as a contrastive focus marker, on the subject of a nominal
(11) or verbal (12) sentence, one of the values of argument-focus structure in Lambrechts terminology (see Lambrecht 1994:222). When the NP subject of a nominal
predicate is overtly expressed, no suffix pronoun is added to the particle:
(11) qal
arf insi
qal-ah
ani
say.pfv.3m.sg smell human.being say.pfv-3f.sg 1f.sg
asal-k
u ra
arf=i
insi
ani
wash.pfv-1sg and foc.m.sg smell=1sg human.being 1f.sg
He said: There is a smell of human being! She said: I washed and MY SMELL
(is a smell of) human being, mine! (as opposed to the smell of the man she is
hiding from the ogre).
(12) qal
la ra=ha
qal-ah=l=i
kaa
say.pfv.3m.sg no foc.m.sg=obj.3f.sg say.pfv-3f.sg =all=1sg so
He said: No! She is really the one who told me so! (and not her mother).
As a subject-focusing particle, ra can also be used in a cleft construction (as in

French): the focused subject is preceded by the particle and followed by one of the
relative markers (which is also used after the subject of the preceding clause to which
it is contrasted):
(13) qal
eh baqar-t=e
a walad-ah
say.pfv.3m.sg yes cow-gen.f=2f.sg rel give.birth.pfv-3f.sg
ra-i
tabi=i i walad
foc-f.sg veal=1sg rel give.birth.pfv.3m.sg
He said: No kidding, your cow calved! Its my veal that calved!
2.3.3 Predicative relation focusing

In order to understand the development of the particle ra as an inter-clausal marker,
I will refer to the definition of focusing and the analysis of its various uses and interpretations provided by Robert (1993, 2000 & this volume). She considers (this volume)
focus as a specific mode of identification: in a focused sentence, assertion consists
in the qualitative designation of an element whose existence is pre-asserted. []
the focused proposition consists of a split assertion involving a presupposition
(of the predicative relationship []) and a qualitative designation (of the focused
constituent []).
Such a definition corresponds, all things being equal, to Lambrechts approach which
considers focus as The semantic component of a pragmatically structured proposition
whereby the assertion differs from the presupposition, and in particular to his category
Martine Vanhove
of sentence-focus structure, i.e. the event-reporting or presentational type, in which

the focus extends over both the subject and the predicate (Lambrecht 1994:222), i.e.
the predicative relation.
As a discourse marker, ra quite often focuses the entire predicative relation, whatever the category of the predicate (noun, verb in the Perfective or the Imperfective).
It should be noted that there is no morphosyntactic difference with the use of ra as a
topic marker or as a presentative, in particular no word order difference, but predicate
focusing clauses are the sole instances where the object pronoun suffixed to ra may
refer to the object of the transitive verb (ex. 15, 23) or to the addressee (ex. 24), not only
to the subject (see Sections2.3.4 and 2.3.5 below).
For predicate focusing in Yafi Arabic, two of the four values identified by Robert
(1993) are directly relevant: (i) intensive predication, or focus on the truth-value or
assessment of the predicate, and (ii) explicative value. Both values are also quite frequent in the Maghribi Arabic varieties.
2.3.4 The intensive value

One of the two values expressed by ra as a focus marker of the predicative relation is
that of assertive intensity, meaning that the event or state described by the predicate
is truly the case, a value which seems crosslinguistically common for focus markers (Robert 1993), and in line with the intensive value of ra as a copula (cf. 2.2).
The intensity meaning only occurs when there is no discourse reference point, what
Robert calls the discursive landmark, in the discursive chain, i.e. in the preceding clause, or no preceding clause as in (12) below. Since focusing always involves
the presupposition of the predicative relationship together with the designation of the
focused constituent (see 2.3.3), and since there is no discourse reference point, the
intensity value in a sentence-focus structure stems from the fact that the speaker
is mentally comparing the focused state or process to itself or to all other possible
interpretations of the same state or process. The presupposed assertion may then
refer to either a supposed shared knowledge between the speakers, or to some unexpected event or doubt.
In Example (14), a nominal predicate occurs at the beginning of a dialogue. It
should be noted that if ra focuses the subject, in this context it does not serve to
identify the subject, but to emphasize the truth-value of the predicative relation (it
is indeed common knowledge that women have to be very cunning to counter male
supremacy):7
. Just like the subject focus conjugation of Wolof (for an explanation of the dissociation
between the marked element and its scope, see Robert, 1993).
(14) ra
ked
en-nisa
keden aim
foc.m.sg cunning def-woman.pl cunning immeasurable
qa aggez-en
ali ibn abi taleb
foc play.a.trick.pfv-3f.pl A. I. A. T.
Womens cunning is truly endless! They even played a trick on Ali Ibn Abi Taleb.
Example (15) below presents another syntactic construction where the direct object
argument of the predicate is attached to the focusing particle (=oh in ra=oh refers to
the object complement of the predicate, i.e. the man to whom the woman sold the mat),
thus encoding the focusing of the direct object complement.8 Note that the intensive
value is reinforced by the use of the scalar focus particle etta until; even (see Knig
1991). In this particular example, selling something was a totally unexpected event.
(15) a-si
riggal taraf il-madin-eh wu
1sg-meet.ipfv man end def-town-f and
ddi=ni
tijah
l-xobz-eh
etta
give.pfv.3m.sg=obj.1sg dem.f.sg def-bread-f until
ra=oh
bi
sin-eh bia-ku=l=eh
foc.m.sg=obj.3m.sg with mat-f sell.pfv-1sg=all=3m.sg
I found this man at the very end of the town and he gave me that piece of
bread, and I even sold him a mat!
Such constructions have not been reported for Maghribi Arabic where the particle is
more grammaticalized than in Yafi Arabic (agreement in gender and number of the
particle with the addressee is no longer possible, but pronouns are regularly suffixed
to the particle).
2.3.5 Focusing, clause linking and explanatory value

As Robert notes (this volume) [s]ince it involves a presupposition [], a focused
clause is a priori seldom used alone []: the focused clause is, most of the time, related
to a previous clause or sentence [].
In Yafi Arabic the particle ra as a sentence-focus particle is in fact most often
used in a series of juxtaposed clauses, but the preceding clause does not represent
the presupposed element, its internal landmark in Roberts terminology. In this case,
. Note that the Yafi particle ra broadens the polyfunctionality of object focusing markers
as compared with Wolof (West-Atlantic): in Wolof, the exclamative value associated with an
intensification of the verb is not attested for the complement focusing conjugation, which is
limited to the identification of the complement in Roberts data (1993, 2000).
Martine Vanhove
the focused clause has explanatory value, i.e. the focused clause is the causal source
of the preceding clause. Similarly to Umpithamu, the use of ra is based on a mechanism of encoded inference, inherent to the definition of the process of focus promotion, i.e. the pre-constructed (or presupposed) element is not found in the discourse
chain and forces the inference of an explanatory relation with the preceding clause
(Verstraete this volume). Thus, in addition to many other languages such as French,
Berber (Leguil 1987), Wolof (Robert 1993, 2000 & this volume), Oceanic languages
(Bril in press & this volume), or Umpithamu (Verstraete this volume), Yafi Arabic provides another interesting case of inter-clausal dependency marked by forms indicating
informational hierarchy and which can be used to indicate semantic dependency such
as causal relations between clauses.
The explanatory value of the cognate forms of ra in Maghribi Arabic is also very
frequent (it is called reassertive value by Caubet 1992, and Madouni 1993), and examples are often translated by causal clauses. Caubet (1992:145) insists on the pragmatic
value of these constructions and specifies that the examples she provides are examples
of reassertion with refocusing (of the type it is that/it is because that), where the
speaker goes back on what he has just said because he feels he was misunderstood or
that his intentions were misinterpreted.9
In Yafi Arabic, the explanatory value of ra is often found in dialogues, as
an answer to why questions, linking the focused clause to a question asked by
another speaker; ra is even compatible with an overt coordination marker as in
(17) below:10
(16) qal-ah
wa l=ak
wa riggal lamma inteh
say.pfv-3f.sg voc all=2m.sg voc man why
2m.sg
kaa migabal >al
ra-i=ni
ga-ok
so dusty
say.pfv.3m.sg foc-f.sg=1sg come.pfv-1sg
min end ahl il-magann-ah
abl loc people def-graveyard-f
She said: Whats the matter with you man, why are you covered with dust?
He said: Its because I am back from among the dead.
. Des exemples de rassertion avec recentrage (de type cest que/cest parce que), o
lnonciateur reprend ce quil vient de dire, parce quil estime avoir t mal compris, ou quon
a pu se mprendre sur ses intentions. (My English translation).
. Note that the structure of Example (17) is identical to that of the presentative use of
ra in (3). The difference is that in (4), the ra clause is joined to the previous one by the
consecutive particle baen then, which marks a temporal, not a causal, relation between the
two clauses.
(17) qal
le tajjar-i
ali=ja
say.pfv.3m.sg why be.late.refl.pfv-2f.sg on=1sg
qal-ah
say.pfv-3f.sg
ma l-jom ra=ni
wa la asa-ok
i
top def-day foc.m.sg=obj.1sg and neg find.pfv-1sg thing
He said: Why did you make me wait so long? She answered: Its because today
Icould not find anything at all.
In other clause-combining types the causal value of the focusing particle is compatible
with all verbal conjugations, with nominal predicates, and with all types of argument
structure.
In (18) below, in addition to the Perfect meaning that ra brings to the Perfective
conjugation, the particle also makes the causal relation between the two clauses explicit:
(18) qal-ah
ja ibn=i ma fi=
a-ah
say.pfv-3f.sg voc son-1sg neg in=neg dinner-f
ra=hom
ahar-u
ali=je l-miskin
foc.m.sg=3m.pl appear.pfv-3m.pl on=1sg def-poor
l-7ww7l wu ddi=h
xobz-7t=i
def-first and give.pfv.1sg=3m.sg bread-gen.f=1sg
She said: My son, there is no dinner because they showed up and
I gave my bread to the first poor man.
With the Imperfective aspect and the Future particle, no aspectual connotation is
involved, and the explanatory focusing value of the particle simply creates an explicit
causal link between the clauses:
(19)

qal
ja tra-u=l=i
il-arar-ah hinija
say.pfv.3m.sg voc let.imp-m.pl=all=1sg def-bag-f
here
lil
is-sub qad
ana a-gi=le=hen
until def-dawn already 1m.sg 1sg-come.ipfv=all=3f.pl
ela
l-adawi
ra-u=ni
a-sir
dist.m.pl def-medicine.pl foc-m.pl=1sg 1sg-go.ipfv
a-rqud=l=i
bi
bet allah bi=l-misgid
1sg-sleep.ipfv=all=1sg with house God with=def-mosque
He said: Keep my bag here until dawn, Ill come and pick up the medicine,
because I am going to sleep in Gods house, at the mosque.
(20) qal
a-bi
d-dar7hem
la=l-jom
say.pfv.3m.sg 1sg-want.ipfv def-dirham.pl all=def-day
ra
umm=i
a=tu->tol=ni
foc.m.sg mother=1sg fut=3f.sg-kill.ipfv=obj.1sg
He said: I want the money today otherwise my mother will kill me!
Martine Vanhove
The same explanatory relation is found with nominal predicates, when they are juxtaposed to a preceding clause with a continuative prosodic contour (usually without any
pause). Although the construction is similar to the copular and presentative values of
ra the prosody blocks these interpretations which are no longer salient, and the causal
interpretation predominates:11
(21) qal
gib-i=l=i
7rb-ah ma
say.pfv.3m.sg bring.imp-f.sg=all=1sg drink-f water
ra-i=ni
musafer
foc-f.sg=1sg travel.ptcp
He said: Bring me some water because I am a traveller.
(22) qal
eh qa=kadd-u=ni
la
say.pfv.3m.sg yes perf=send.pfv-3m.pl=obj.1sg all
end=e ra-i
maa=hum z7wag-ah
loc=2f.sg foc-f.sg with=3m.pl wedding-f
He said: Yes, they sent me to you because they are having a wedding.
Example (23) below provides yet another illustration of the pronominal direct object
argument of the predicate attached to the focusing particle. But this time, object
focusing brings an explanatory value to the entire clause, not an intensity value (see
Section2.3.4 above).12
(23) qal
ja mar-ah xert
allah al=e
say.pfv.3m.sg voc woman-f good-gen.f God on=2f.sg
ra-i=hom
ijal=i
bi=a-ebb=hom
foc-f.sg=3m.pl child.pl=1sg prog=1sg-love.ipfv=obj.3m.pl
He said: Woman, be reasonable, because I love my children.
(lit. MY CHILDREN I love them).
Example (24) is the sole example in my data where the object suffix pronoun on the
focusing particle refers to the addressee, a clear indication (in addition to its imperative
origin) that, as in Maghribi Arabic, the particle ra refers to the discursive space of both
co-enunciators (Caubet 1992:142). It is also a clear indication that even in the particular
instance of explanatory value, the function of ra goes beyond that of a mere syntactic
marker and discursive and pragmatic information is still part of its semantics:
. Note that the presentative example in (5) could also be interpreted semantically as an
explanation of the preceding clause: come in, because its raining. Because the two clauses
correspond to two major prosodic units, this favours the presentative interpretation.
. Again this is not reported for Maghribi Arabic or for the Wolof complement focus
conjugation.
(24) ddi=ni
ale=k
min al-xobz-eh
give.imp.m.sg=obj.1sg on=2m.sg abl def-bread-f
ra=ak
ma=i riggal taban
foc.m.sg=2msg with=1sg man sick
Give me, please, a piece of bread because I have (my) husband who is sick.
(lit. YOU I have13 husband sick)
2.4 Summary
The above description has shown a certain number of semantic and syntactic constraints
on and restrictions in the use of ra, depending on the syntactic scope of the particle. In
the deictic domain ra has a wide range of uses as a presentative with NPs, PPs and relative clauses, but with pronouns its use is limited to 1st persons. As a copula, ra is strictly
confined to pronominal subjects, but concerns both property and locative utterances.
As far as information structure is concerned, its role as a topic marker is marginal and
restricted to 1st person pronouns; as a focusing particle, it can be a subject-contrastive
focus particle or, when it has scope over the entire predicative relation, it has both an
intensity and explanatory focusing value. At clause-combining level, the latter has
directly given rise to its function as a clause-linking device in causal clauses.
The different functions (with their constraints) of the polyfunctional particle ra
are summarized in Table1.
Table 1. The functions of the particle ra
Deixis
Speaker Presentative
NP, PP Presentative
Copula
Property utterance
with pronominal S
1st person topic
pronoun
Causal clause
Locative utterance
with pronominal S
Focusing S, O, P &
addressee
Information
structure
Clause
linkage
Relative clause
Presentative
Intensity
focusing
Explanatory
focusing
3. The particle ta
In Yafi Arabic, the use of the particle ta as a clause-linking device is less frequent than
that of ra. Its lexical source being different from the latter, it is nevertheless interesting to discuss it in the light of similar developments in various languages worldwide.
. ma with + possessive pronouns is the common way of forming a possessive predicate

equivalent to a verb have.
Martine Vanhove
As in Oceanic languages for instance (see Bril in press & this volume), ta illustrates
the case where demonstratives are reanalysed as informational hierarchy markers and
subordinators, a frequent clause-linking strategy based on some referential hierarchy
between a backgrounded, already referential clause and an asserted clause (Bril in
press). The particle ta and its polyfunctional uses seem to be specific to the Yafi Arabic variety. To the best of my knowledge this has not been reported for other Arabic
varieties.
3.1 Demonstrative
Ta is basically a nominal modifier, namely an invariable demonstrative. It is far from
being the most frequent demonstrative in the language; still it may function either as
an adnominal demonstrative (25) or as a pronominal demonstrative, or subject of a
nominal clause as in (26), or of a verbal clause as in (27):
(25) ta al-mafati entra-en
foq al-atab-ah
dem def-key.pl put.refl.pfv-3f.pl on def-doorstep-f
These keys were left on the doorstep.
(26) ta mai hu
ban=e
dem neg 3m.sg son=2f.sg
This one is not your son.
(27) qal-ah
dri
e hu
ma ani ta
say.pfv-3f.sg ignore what 3m.sg top 1f.sg dem
nazal-ah
kabir-ah aqq-at=i
go.down.pfv-3f.sg big-f
gen-f=1sg
She said: I dont know what it is, but as for me, what fell upon me is big.
3.2 Presentative
Ta can also function as a presentative. Its use is syntactically constrained: the scope of ta
is never a noun phrase or a prepositional phrase, but always an entire clause, either nominal or verbal; in nominal utterances, ta is limited to locative utterances (28, 29). Contrary
to ra, no object suffix pronoun can be cliticized to ta (a trace of its verbal origin).
(28) qal
ta
ad ad-dafa
andu=kum
say.pfv.3m.sg pres still def-payment loc=2m.pl
He said: I am afraid you still have to pay (lit. there is the payment at you).
(29) qal
iglis-i
wulla galas-
say.pfv.3m.sg stay.imp-f.sg or
stay.pfv-2f.sg
qal-ah
ta
ale=h
summ
say.pfv-3f.sg pres on=3m.sg poison
He said: You must stay! She said: There is poison in it!
(30) qal-ah
ja ha uf
ta
hu
say.pfv-3f.sg voc pres see.imp.m.sg pres 3m.sg
ji-gi
l-ifri
ji-qul
arf insi
3m.sg-come.ipfv def-devil 3m.sg-say.ipfv smell human.being
a-qul
ani insi
ta
an14 hu
1sg-say.ipfv 1f.sg human.being pres comp 3m.sg
b=ji-nfit
i-arar
ta
hu
prog=3m.sg-blow.ipfv def-spark.pl pres 3m.sg
raq6d
wun hu
mhemm6d ta
hu
sleep.act.ptcp and 3m.sg calm.ptcp pres 3m.sg
had6d
u hidd
get.up.act.ptcp and get.up.pfv.3m.sg
She said: Look, here comes the devil! Hes gonna say: There is a smell of human
being. Ill say: I am a human being. Here hell be blowing sparks, here hell be
sleeping and will calm down, here hell be awake, and he is awake!
3.3 Copula
Ta is also used as a copula in nominal property utterances.15 The word order, different
from the one with ra, is Subject Copula Predicate. The predicate can be an indefinite
noun (31) or a clause (32) and, unlike ra, there is no constraint on subject category:
(31) u ji-qaf=ha
bi=r-ras
u
and 3m.sg-hit.ipfv=obj.3f.sg in=def-head and
ras=ha
ta nosf-en
head=3f.sg cop half-du
He hit it on the head, and its head was in two parts.
(32) qalen
salef=na
ta hi
bi=t-baqqas
umes
say.pfv-3f.pl custom.pl=1pl cop 3f.sg prog=3f.sg-pinch.ipfv sun.dim
They said: Our traditions are (that) Little Sun pinches.
. Example (35) shows that another presentative an can be added to ta. This particle will
not be discussed here, but it presents another interesting case where a deictic, fused with
a coordinating element, has developped into a focusing and a clause-linking element (for
further details, see Vanhove 2000).
. Although deemed typical of their language by the speakers themselves, the use of ta as a
copula is rare in my data. This usage is neither found in the neighbouring Arabic varieties, nor
in any other Arabic variety, as far as I am aware.
Martine Vanhove
3.4 Topic marker

Like ra, the use of ta as a topic marker in my data is restricted to 1st persons singular
and plural, expressed, for ta, with the set of independent subject pronouns. Ta occurs
in contexts where the speakers actions contrast with or do not take into account those
of the interlocutor (not in response to a demand as with ra, see Section2.3.1 above):
(33) ta ana a-bekker
ma ana s-suq.
top 1m.sg 1sg-leave.early.ipfv top 1m.sg def-market
As for me, I am leaving early, for the market.
(lit. as for me, I am leaving early, as for me, for the market).
(34) qalen
ta rena salef=na
ani
say.pfv-3f.pl top 1pl custom.pl=1pl second
They said: As for us, our traditions are different.
3.5 Explanatory focusing

Like ra, ta can also function as a focusing marker. In this case, it normally precedes the
predicate, and its scope is the entire predicative relation. But contrary to ra, ta only brings
an explanatory value to the clause, no instances of intensive meaning were found.16
In order to understand the semantic and grammatical evolution of this particle, it
is worth mentioning that, in some semantic and discursive contexts, the explanatory
value is also understood when the particle is a presentative. This is in line with Lambrechts category of sentence-focus structure, his event-reporting or presentational
type. In the dialogue in example (35) repeated below, the semantic interpretation of
the second speakers utterance is to be understood as a refusal, which she justifies by
mentioning that the food is poisonous: the witch refuses to stay for lunch because she
knows that the food she has given to her interlocutor is poisoned:
(35) qal
iglis-i
wulla galas-
say.pfv.3m.sg stay.imp-f.sg or
stay.pfv-2f.sg
qal-ah
ta ale=h
summ
say.pfv-3f.sg foc on=3m.sg poison
He said: You must stay! She said: There is poison in it!
Ta often combines with other focusing markers, which have different pragmatic or
assertive connotations. In the parenthetical clause below (ta qa hu a-alab makkar),
which explains why the crow had to come back, the scope of ta is the entire predicative
. This lack might be linked to the fact that ta, unlike ra, has no intensity value as a copula
and marks a simple thetic nominal utterance, but examples are too scarce to be sure.
relation and is the explanatory focus marker, while qa is an intensity focusing particle
whose syntactic scope is limited to the verbal predicate itself:
(36) il-lel-eh
e-anij-eh haa e-alab ta qa
hu
def-night-f def-second-f dem def-fox foc int.foc 3m.sg
a-alab makkar [] raga
il-urabi
def-fox cunning [] come.back.pfv.3m.sg def-crow-sing
lel-eh anij-eh
night-f second-f
The following evening that fox, because that fox was really cunning, []
the crow came back another evening.
In (37) the speakers statement explains why it would be difficult for her to avoid killing
her son, and mai is the intensive negative focusing marker:
(37) xalli
l=aqq
l=ban=
qal-ah
ta
leave.imp.sg all=gen.m all=son=2f.sg say.pfv-3f.sg foc
mai
a-rif=hom
qa=hom
in-eh
foc.neg 1sg-know.ipfv=obj.3m.pl int.foc=3m.pl similar-f
Spare yours, your son! She said: I am afraid I cant tell them
(= the step brothers) apart, they look alike so much!
3.6 Focusing, topic, and clause linking

Similarly to ra, when the focusing particle ta occurs in a series of juxtaposed clauses,
it can function as a clause-linking device with causal meaning:
(38) ga
qattab
ros=hum
u
come.pfv.3m.sg cut.int.pfv.3m.sg head.pl=3m.pl and
ji-baqqi
wai-eh min=hen qal
hae iaw-ah
3m.sg-let.ipfv one-f
abl=3f.pl say.pfv.3m.sg dem salary-f
ta qa=alama=ha=l=ah
bi=l-bet
foc perf=mark.pfv.3m.sg=obj.3f.sg=all=3m.sg with=def-house
He came to cut their heads and he left one of them (= sheep) (alive). He said:
This one is my reward, because he had marked one for himself in the house.
(39) qal=l=ah
ja hija
ja ta ana
say.pfv.3m.sg=all=3m.sg voc go.imp voc foc 1m.sg
add-ok=l=ak
ala bent es-sultan
give.pfv-1sg=all=2m.sg on girl def-sultan
He told him: Lets go, because I have arranged the wedding with the sultans
daughter! (lit. as for me, I have arranged).
Martine Vanhove
3.7 Relative marker

As compared with ra, the particle ta has two additional functions as far as clause-
linking strategies are concerned. Not surprisingly, one of them is to introduce a restrictive relative clause, similarly to the numerous languages, among them English, but also
Oceanic languages (Bril in press & this volume) to name but a few , where demonstratives and topic particles can be used as relative markers. In Yafi Arabic, the use of ta as
a relative marker is by far less frequent than the two other relative markers (which are
also demonstratives), i and a. Unlike Nlmwa (Bril 2002:332347, 2000:261270),
none of the Yafi relative markers encodes specifically new information on the head
noun or already referential NPs. All three are simply ad-nominal specifiers:
(40) ji-gi
jakul
il-lam-ah ta qa=al-ena
3m.sg-come.ipfv 3m.sg-eat.ipfv def-meat-f rel perf=boil.pfv-1pl
He comes and eats the meat that we had boiled.
3.8 Complementizer
The last function of ta is to introduce complement clauses, again not surprisingly,
considering the very frequent crosslinguistic development of demonstratives. But the
complementizing use seems to be incipient in Yafi Arabic and limited to the request
verb wasa ask.17 Note that it cannot be used in reported speech after the quotative
verb qal say. Because of the limited number of examples, it is impossible to know
whether, like some Oceanic languages (Bril in press), the complementizing function is
specifically linked to asserted information, although it is the case in the two examples
found in my data:
(41) dri
hu
aje
u ad=oh
ignore 3m.sg lose.act.ptcp and still=3m.sg
a=ji-i
wa mat
ta ani
fut=3m.sg-live.ipfv and die.pfv.3m.sg top 1f.sg
a-wsi=kom
ta
la awi
1sg-ask.ipfv=obj.2m.pl comp if go.back.pfv.3m.sg
ta-nkir-un=ah
2-repudiate.ipfv-m.pl=obj.3m.sg
I dont know if he is lost, still alive, or dead. As for me, I am asking you
to repudiate him if he comes back.
. The other request verbs can be followed by the complementizer inna, by no complementizer at all, or more rarely by one of the other relative pronouns.
3.9 Summary
The polyfunctionality of ta is somewhat different from that of ra. As a presentative,
ta is only used with full utterances (not with NPs or PPs), or as a copula with property utterances, but both with pronominal and nominal subjects. Its role as a marker of
information structure is narrower than that of ra: neither focusing of an argument, nor
intensity focusing are observed in the data, only explanatory focusing and, marginally
again, topic particle with 1st person pronouns. On the other hand, ta has more conjunctive functions and can be used to mark causal, relative and complement clauses.
Table 2 below summarizes the various functions (with their constraints) of the
polyfunctional particle ta.
Table 2. Functions of the particle ta
Deixis
Copula
Information
structure
Clause linkage
Adnominal
demonstrative
Property utterance
with pronominal S
1st person topic
pronoun
Causal clause
Pronominal
demonstrative
Property utterance
with nominal S
Explanatory focusing
Presentative
utterance
Relative clause
Complement clause
4. Conclusion
Although the two polyfunctional particles discussed in the present study share a
number of common properties and functions, they differ in several respects. Some
functions or values are exclusive to one of them: only ra can be used as an intensityfocusing particle, and only ta is a demonstrative and a marker of relative and complement clauses. As for their common functions, they do not always have the same
syntactic distribution and discursive constraints. Regarding their use as presentatives,
the two particles are specific to different morphosyntactic contexts: speaker, NP, PP
and relative clause for ra, independent utterances for ta. As a copula, ra is more
limited than ta as it can only be used with pronominal subjects, not with nominal
ones, while ra brings a pragmatic value of intensity to the truth-value which does not
exist with ta. Concerning focusing, because of its verbal imperative origin, only ra
can focus the addressee (cf. ex. 24 where the object pronoun suffixed to ra refers to
the addressee). In addition, only ra is used as an intensity-focusing marker. What the
particles do have in common is the general function of presentative, of topic marker
(limited to 1st person pronouns), explanatory focusing value, and use as a clauselinking device for causal clauses.
Martine Vanhove
Table 3 summarizes the comparison between the two particles.

Table 3. The functions of the particles ra and ta
cop
*imp dem pres prop
pr
ra +
ta
+
+
+
+
foc
prop
n
loc
pr
top s,o,p int expl caus rel comp

add
+
+
+
+
+
+
The properties shared by both particles are enough to show how deictic elements
with totally different origins, but which both have a deictic presentative function, may
become, via evolution as a copula, the preferred strategies for informational hierarchy
as well as for the expression of particular semantic and syntactic links between clauses.
But the grammaticalization paths of the particles are partly different. Leaving aside the
topic function which is too marginal to have played a role in the grammaticalization
process, the proposed scenarios are as follows.
For ra:
Imperative of see > presentative > copula > explanatory focus marker > causal
clause marker.
Note that according to Stassen (1997:92) the grammaticalization of verbs meaning
see as presentatives and copulas is crosslinguistically rare.
For ta, two different grammaticalization paths have to be assumed:
1. demonstrative > presentative > copula > explanatory focus marker > causal clause
marker.
2. demonstrative > relative marker/complementizer.
Readers of this volume will by now not be surprised that a causal relation between
clauses should be marked by focusing strategies and that informational hierarchy may
eventually evolve into syntactic hierarchy. Yafi Arabic provides just another example,
which proves that such grammaticalization processes also occur in languages which
already have dedicated syntactic markers of subordinate clauses. Nevertheless, whereas
deictic and informational hierarchy functions have been, in some particular contexts
of clause combining, reanalyzed as marking inter-clausal dependency, the clausal and
complementizer functions are not fully integrated as syntactic subordinators in the
language in the sense that discursive and pragmatic inferences are still part of the
semantics of the particles.
Abbreviations
abl
act
add
all
caus
comp
cop
def
dem
dim
dist
du
expl
f
foc
fut
gen
imp
int
ipfv
loc
m
ablative
active
addressee
allative
causative, causal
complementizer
copula
definite article
demonstrative
diminutive
distal
dual
explicative
feminine
focus
future
genitive
imperative
intensity
imperfective
locative
masculine
n
neg
o
obj
p
pass
pfv
pl
pres
perf
pr
prog
prop
ptcp
refl
rel
s
sg
sing
top
voc
noun
negation
object
object
predicate
passive
perfective
plural
presentative
perfect
pronoun
progressive
property
participle
reflexive
relative
subject
singular
singulative
topic
vocative
References
Bril, Isabelle. 2000. Postmodification and the structure of relative clauses in Nlmwa and
other New Caledonian languages. In SICOL. Proceedings of the Second International Conference on Oceanic Linguistics, Vol. 2: Historical and Descriptive Studies [Pacific Linguistics 505], Bill Palmer & Paul Geraghty (eds), 261284. Canberra: Australian National
University.
Bril, Isabelle. 2002. Le Nlmwa, analyse syntaxique et smantique. Paris: Peeters.
Bril, Isabelle. In press. Coordination, information hierarchy and subordination in some Austronesian languages. In Converbs, Medial Verbs, Clause Chaining and Related Issues [Frankfurter Afrikanistische Bltter], Azeb Amha, Christian J. Rapold, Sascha Vllmin & Silvia
Zaugg-Coretti (eds).
Caubet, Dominique. 1992. Deixis, aspect et modalit: Les particules h- et r- en arabe marocain.
In Actes du colloque: La deixis, Mary-Annick Morel & Laurent Danon-Boileau (eds), 139
149. Paris: Presses Universitaires de France.
Martine Vanhove
Caubet, Dominique. 1993. LArabe Marocain, Tome 2: Syntaxe et Catgories Grammaticales,
Textes. Leuven: Peeters.
Cohen, David. 1984. La phrase nominale et lvolution du systme verbal en smitique. Etude de
syntaxe historique. Leuven: Peeters.
Cohen, Marcel. 1924. Le systme verbal en smitique et lexpression du temps. Paris: Ernest Leroux.
Colin, Georges S. 196063. Sur la phrase dite nominale en arabe marocain. Comptes Rendus du
GLECS 5 (Decembre 1948): 49.
Harrel, Richard S. 1962. A Short Reference Grammar of Moroccan Arabic. Washinton DC: Georgetown University Press.
Knig, Ekkehard. 1991. The Meaning of Focus Particles. A Comparative Perspective. London:
Routledge.
Lambrecht, Knud. 1994. Information Structure and Sentence Form. Topic, Focus and the Mental
Representations of Discourse Referents. Cambridge: CUP.
de Landberg, Carlo. 1909. Etude sur les dialectes de lArabie mridionale, Vol. II: Datnah. Leiden:
Brill.
Leguil, Alphonse. 1987. Structures prdicatives en berbre. Travaux prsents pour le Doctorat
dEtat, INALCO Universit Paris 3.
Madouni, Jihane. 1993. Les valeurs de la particule r- dans un parler de Sidi-Bel-Abbes. Matriaux arabes et sudarabiques n.s. 5: 123136.
Robert, Stphane. 1993. Structure et smantique de la focalisation. Bulletin de la Socit de Linguistique de Paris 58(1): 2547.
Robert, Stphane. 2000. Le verbe wolof ou la grammaticalisation du focus. In Topicalisation et
Focalisation dans les langues africaines, Bernard Caron (ed.), 229267. Louvain: Peeters.
Stassen, Leon. 2003[1997]. Intransitive Predication. Oxford: OUP.
Vanhove, Martine. 1995a. Notes on the Arabic Dialectal Area of Yafi (Yemen). Proceedings of the
Seminar for Arabian Studies (Oxford, July 1994) 25: 141152.
Vanhove, Martine. 1995b. A propos du verbe dans les dialectes arabes de Yafi (Ymen). Dialectologia Arabica. A Collection of Articles in Honour of the Sixtieth Birthday of Professor Heikki
Palva, 257269. Helsinki: Studia Orientalia.
Vanhove, Martine. 1996. Les particules qad et ra dans un dialecte arabe de Yafi (Ymen). In
Proceedings of the 2nd International Conference of lAssociation Internationale pour la Dialectologie Arabe held at Trinity Hall in the University of Cambridge, 1014 September 1995,
243252. Cambridge: University of Cambridge.
Vanhove, Martine. 2000. Coordination, subordination, deixis et focalisation en arabe de Yafi
(Ymen). Oriente Moderno 19(1): 209223.
Vanhove, Martine. 2004. Deixis et focalisation: La particule ta en arabe de Yafi (Ymen). In
Approaches to Arabic Dialects: Collection of Articles Presented to Manfred Woidich on the
Occasion of his Sixtieth Birthday, Martine Haak, Rudolf De Jong & Kees Versteegh (eds.),
329342. Leiden: Brill.
The role of the Berber deictic and TAM markers

in dependent clauses in Zenaga
LACITO (Laboratoire des langues et Civilisations Tradition Orale)

Fdration de Typologie et Universaux Linguistiques-CNRS,
Paris III and Paris IV-Sorbonne Universities
The deictic ad fulfills many functions in Mauritanian Berber (demonstrative,
copula, connector, injunctive particle, etc.). Most of these uses are found more or
less in all Berber languages. However, Zenaga differs markedly on several points:
ad is not used as a preverb to express future; on the contrary, it is used as a
conditional and quotative particle.
The author analyses these divergent evolutions and proposes explanations
for the various grammaticalization phenomena involved by conjointly studying the
TAM (tense-aspect-mood) markers and sentence structure. The polygramma
ticalization of the deictic is partly linked to the semantic indetermination of the
Aorist and to some specific conditions which bear on fronted verbs.
1. Introduction
Zenaga is a Berber language, which belongs to the Chamito-Semitic (or Afro-Asiatic)
phylum; it is quite representative of this family, yet differs from the other languages on
certain points.1
A. Word order typology

Typologically, Zenaga is characterized by the order determined-determiner noun,
namely in complementation relations (v + o),2 for true determination (where the
1. The author wishes to thank Margaret Dunham for her accurate translation and Isabelle
Bril, Lionel Galand, Maarten Kossmann and Karl-G. Prasse for attentively reading this article
and for their stimulating feedback. She assumes entire responsibility for the reconstruction
proposed here (for an alternate analysis, see e.g. Prasse 2008) and is to be held solely accountable for the imperfections to be found in this final version.
2. Abbreviations: a Aorist, ann annexation state, cas case marker, do direct object, io indirect object, lit. literally, nipfv negative Imperfective, npfv negative Perfective, np noun phrase,
eterminer may be a pronominal affix) and for qualifying phrases. This is also the order
d
followed by sentences with non verbal predicates, where the subject is always placed at
the beginning of the clause, and which we define as the fronted or initial position.3
The case of the verbal predicate is different because the person marker obligatorily
associated with the verb stem fulfills subject function (compare y-rm he took to
t-rm she took and t-rm-m you (pl) took). When the person marker is explicitated by a lexeme, the two are coreferential and the lexemes position as compared to
the verb depends on the information structure of the sentence: this lexeme which
scholars of Berber usually prefer not to call a subject is either in fronted (pre-verbal)
position, or in non fronted (post-verbal) position.4
Zenaga
(1) y-rm
iDy
t-gz-tt=n=
3m.sg-please.pfv man[m.sg] f-daughter-sg=of= pr.3[sg]
A man pleased his daughter (The man pleased his daughter is less probable)
(2) iDy
y-rm
t-gz-tt=n=
man[m.sg] 3m.sg-please.pfv f-daughter-sg=of=pr.3[sg]
The man pleased his daughter
A fronted element is subject to specific constraints (in Zenaga, but also more generally
in Berber and Arabic): it constitutes the initial locator, the anchor point of the sentence
so cannot be entirely indefinite on the semantic-referential level and, in Zenaga, it is
sufficient (except in very particular cases, e.g. that of the lexeme mn male person,
o object, op orientation particle (op1 venitive, op2 allative), pr pronoun, rel relater, v verb,
vp verb phrase.
3. On the nature of the predicate in noun phrases and on its role in the evolution of the
verbal system in Semitic, see Cohen 1984 (in particular Sections1 and 2, pp. 1150). On word
order in Berber (which usually places the complement after what is complementized) and
on its importance for locating the various functions in verbal predicates, see Galand 1988:221
and 2002 [1964]: 303.
4. In post-verbal position, this lexeme is generally found in a modified form which is, among
others, that of the noun determiner in determination phrases. Galand (2002 [1964]: 287307),
followed in this by numerous scholars of Berber, dubbed this lexeme which bears the annexation state the explicative complement and, in contrast, the lexeme placed just before
the verb, the theme indicator, which is always in the free state. As far as we are concerned,
even though our point of view does not diverge much from Galands, we prefer not to use
these terms, especially because, in Zenaga, the lexeme coreferential with the person marker is
never in the annexation state. In fact, the distinction annexation state vs. free state is not
marked morphologically, even though agglutination phenomena have sometimes been found,
in particular when the noun determined is a numeral (Taine-Cheikh 2005a).
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
someone with indefinite meaning) that the lexeme explicitating the subject person
marker be in pre-verbal position for it to be considered definite.5 Thus the fronted element and the topic often coincide the latter necessarily being considered definite (at
least in Berber)6 , but neither the expression of the topic nor its placement in initial
position are obligatory. In fact, in sentences containing only comments (rhme in
French), the topic is simply implied, and, furthermore, the normal order topic
comment can be modified through recourse to morpho-syntactic and/or prosodic
means. The notions of topic and comment concern (following the tripartite organization proposed in Hagge 1982) the discourse-hierarchic level, a level closely related to
the semantic-referential level, without however coinciding with it.
B. Word classes
While there are two main syntactic classes in Berber, verbs and nouns, all lexemes, be
they verbs or nouns, are the more or less regular result (in particular due to the panchronic evolution of the forms), of the cross between a root and a given pattern. The
following examples are from Zenaga.7
5. There are no morphemes for marking definiteness or indefiniteness, compare (1) and
(2) where the most probable difference resides in the referentiel status of the lexeme which
explicitates the subject person marker.
6. In some languages such as French, an indefinite noun phrase can be topicalized and focalized (ex. Un pre, a se respecte) (fathers, they are to be respected), but this entails the nouns
taking on a massive, collective, global meaning (Berthoud 1994:161) (evidenced in English
by the use of the plural).
7. The phonology and phonetics of Zenaga today are the result of particularly complex
historical evolutions (Taine-Cheikh 1999a, 2003a and 2005b). As with other Berber varieties
(generally belonging to the northern group), there is a tendency for non geminate consonants
to become lax, and particularly to evolve towards spirantization of the voiced dental consonant d (so that d > ). But this lax tendency in Zenaga has also spread to the voiced fricatives
(so that z > ) and to the liquid consonant (l > j). Furthermore, in some contexts, the sibilants
s and z have become postalveolar.
We only partially adhere to the IPA transcriptions. (i) The glides are transcribed w and y.
(ii) Emphasis is noted by a subscript dot (except in the case of f where the dot is superscribed).
(iii) A dash below or above a letter indicates that the articulation is lax, which is characteristic
of some non geminate consonants (except in certain specific contexts).
The transcription we have adopted here is phonetic and phonological. The fact that [] can
be the articulation of either /d/ or /t/ (the distinction /d/ ~ /t/ is neutralized in certain contexts)
is one of the problems which makes it difficult to use an entirely phonological transcription.
Inversely, noting all of the phonetic variations would have made the facts difficult to grasp.
Modifications through contact are extremely numerous, in particular among morphemes
belonging to the same stress group (moreover, as such they are significant for the morphosyntactic
The root, made up of ordered root consonants (variable in number, but usually
two or three), is reserved for a given word family:
RM in yrm he took, pleased, trmS < trmt act of taking, pleasing
(once), nrmi s.o. who habitually takes, pleases, rmi pond; handle
and yTyrm he has been taken, pleased;
RM in yrm he opened (his mouth, a trunk, etc.), trmt act of opening (ones
mouth, etc.) (once) and yTyrm it was opened (for the mouth, etc.).
As for the patterns, which are made up of vowels and sometimes affixes, they
characterize:
basic forms, such as the pattern yc(c)c of the 3rd m.sg Perfective of yrm
and yrm;
morpho-lexical derivations, such as the pattern ncc(c)ic for m.sg habitual
nouns with the affix -n- in nrmi and the pattern c(c)ic for m.sg nouns
with the affix -- in rmi;
morpho-syntactic derivations such as tc(c)ct, the f.sg verbal noun in trmS
< trmt and trmt, as well as yTyc(c)c, the passive verb pattern with the
affix -Ty- of the 3rd m.sg Perfective of yTyrm and yTyrm.
This organization is less clear cut as concerns person pronouns and demonstratives,
especially for autonomous forms. The latter (the only ones susceptible of constituting
the stress bearing element of the syntactic group, like lexemes) appearing, more even
than clitic forms, as agglutinations of elements of various types.
As for the dependence markers, be they intra- or inter-phrastic, they are often
reduced to a single consonant, preceded or not by a vowel.
The pan-Berber interphrastic connection particles are very rare: the disjunction
or (Zenaga na, with or without -) can be used between two clauses as well as
two phrases, but the Berber coordinator of comitative origin d and, with is not
generally used between clauses (this is also the case for the Zenaga ).
The pan-Berber subordination particles are also limited in number. In Zenaga,
some of them stem from a prepositional phrase, others have hazy origins. In many
cases, however, one finds one of the following two elements: ad or y. Here, we will
concentrate on the first, as it is not uniquely used as a subordinating element.
C. Conditional markers
In Berber, the protasis of conditional phrases is regularly introduced by specific particles. These tend to vary from one speech variety to another (as well as being influenced
analysis) but, to facilitate the identification of these morphemes, some assimilations (indicated
by ) are not noted. Apocope is also indicated by , but the deleted phonemes (generally a vowel,
at times followed by the laryngeal consonant h) are given in parentheses.
by the hypothetical system namely potential vs. irreal), but they often contain the
nasal element m, thus mara, imer, xem, Me, mur, lMr, mr, etc. (see Tafi 1993).
One also finds a particle m (m / hm / h.m) in Zenaga, i.e. Mauritanian Berber,
but only in particular counter-factual hypothetical uses:
(3)

(h)m= t--
t-nzz-T
t-az.r.-a=ih
if= op1
2-come.pfv-sg f-morning-sg 2-find.a-sg=pr.do.1sg
If you had come (here) this morning, you would have found me
(but you didnt come).
The particle ad (usually pronounced ) is used in the case of standard conditional

clauses. Thus, in the following proverb:8
y
(4) =ki
y-Tyf
m
. u- T f=ti
ad=pr.do.2m.sg 3m.sg-free.a well free.imp= pr.do.3m.sg
If the well frees you, free it.
This particle ad, which fulfils many other functions in Zenaga and the other Berber
varieties, seems to have deictic origins. Here we will examine its role an uncommon
one as demonstrative in expressions of condition and will study the other devices
used to mark dependency between clauses.
Alongside the role of the tam markers, we will explore whether it is justifiable to
analyze conditionals as topics in Zenaga (a discussion initiated by Haiman 1978), and
to what extent the informational and/or discourse sentence structure may contribute
to explaining the genesis of conditional clauses in Zenaga.
This variety of Berber is in danger of extinction.9 Among its specificities, some
appear to be archaisms. However, Zenagas peripheral situation, where it has been in
contact with Arabic for centuries, also favors specific developments. Using convergence points with Arabic and other languages, we will attempt to assess the proportion
of innovations which may be ascribed to Zenaga, and the grammaticalization of ad
will be reconsidered in the light of more general evolutionary processes.
2. Deixis and polyfunctionality

This section is concerned with Zenagas demonstratives and with the various grammaticalized uses of ad. In other Berber varieties, some of these uses are expressed
by other forms: a (without d), d without a vowel (or with a vowel other than a). This
may indicate that, contrary to what the Zenaga data seem to demonstrate, it is not
8. Its meaning is: Do what you have to do, but dont linger if you want to avoid trouble.
9. The last speakers (who number just a few thousand) are moreover all bilingual Berber /
dialectal Arabic (h.assniyya).
a single morpheme with multiple functions (i.e. polyfunctional), but rather several
morphemes which accidentally merged in Mauritanian Berber.10 We do not claim that
all of its uses are necessarily pan-Berber, but the hypothesis of the deictic origin of the
particle ad enjoys sufficiently widespread agreement among scholars of Berber (see
Galand 1977; Vycichl 1992; Chaker 1997; Mettouchi 2002) for our study on Zenaga to
adopt it as a starting point.
2.1 Demonstratives
In Berber, the demonstrative system generally contains between two and four units,
but it is also quite often ternary, as in many of the worlds languages. Such is the case in
Zenaga, where pronouns, determiners and place adverbs (without movement) tend to
be organized around three degrees of proximity/distance.
In the following table, the forms in bold type are formally identical with the forms
of the particle ad which is found in conditional clauses. Their place in the deixis system
will help us understand the other uses of these deictics.
Table1. Deictic system (Zenaga)
Near
Clitic
m.sg pronoun
Neutral pronoun (sg)
f.sg pronoun
m.pl pronoun
f.pl pronoun
Place adverb
sg -
y
t
ni
tni
pl -i
Mid
Far
-i
i
ti
ni
tni
-n
n
n
tn
nn
tnn
n
The three series have several points in common, but the mid series is characterized
by the presence of the laryngeal consonant , while the farthest series is distinguished
by a final -n (instead of ), alongside the presence of the long vowel --, found only in
the proximity adverb .
The distinction /d/ vs. /n/ has wide scope in Berber. It is found in orientation
particles (op), which are obligatory with certain verbs movement and stative verbs
in particular and optional with others. The oral dental op corresponds to movement
towards the speaker (op1), while the nasal dental op corresponds to movement
10. Generally, for Berber scholars, ad is composed of a prop pronoun a and the deictic d.
However, for Prasse (2008:152 and sq.), the element d could, in some of its uses (namely as a
conjunction), be a variant of the preposition d in. This analysis, however, is difficult to apply
to Zenaga as its preposition in (g ~ gg ~ g ~ ) is never reduced to d.
away from the speaker (op2) (see Bentolila 1969). In Zenaga, the ops take on the
following forms:
op1 - / -d and its allomorphs - / -(h) / -d(h) / -dd(h)
op2 -n and its allomorphs -n / -n(h) / -nn(h).
Zenaga (examples with obligatory op)
(5) y-=ddh
y-Dyi=nn=dh
3m.sg-go.pfv=op1 3m.sg-leave.pfv=pr.do.3pl=op1
He came (here) [and] he left them (around) here.
(6) y-=nnh
y-Dyi=nn=nh
3m.sg-go.pfv=op2 3m.sg-leave.pfv=pr.do.3pl=op2
He went over there [and] he left them (somewhere) over there.
Contrary to the op forms, those of the demonstratives vary enormously from one language variety to the next, especially as concerns the mid distance, but the distinction /d/
vs. /n/ is used by most of the Berber deictic systems. Thus, among the Ighchan (Galand
1988:219), the demonstrative determiners include one form containing an oral dental consonant: -ad here (near) and two forms with nasals: -Na (relative distance) and
-aN there.
The mid series (containing a laryngeal consonant in Zenaga and -Na among the
Ighchan)11 seems to be often used for resumptive phenomena:
rby=i that child there, as opposed to rby= this child (here) and
rby=n that child there (far away), that child over there,
but also rby=i this child (the one which was just mentioned).
Given its absence from exophoric uses in certain varieties, it would seem that the
mid series only appears alongside the other demonstratives used anaphorically. Thus,
among the At Seghrushen, -din (in question) is distinct from -u (close) and -iN (far),
(see Bentolila 1981:75).12
In this latter variety, one should note that the oral dental consonant absent from
the close series is present, as in Zenaga, in the intermediate series. This example
shows that, beyond the numerous formal variations, the link with the basic distinction
11. Among the Ighchan, and more generally in Shilha, there are two distinct anaphora
however: -Na and -Li. It is difficult to specify the meaning of -Na within the system (argaz=Na
that man there not too far? As compared to argaz=Li the man in question), except in the
presence of a relative where the distinction becomes one of indefinite -Na vs. definite -Li
(Lionel Galand, p.c.).
12. The At Seghrushens variety is one of the dialects where the dental consonant /d/ is
absent from the near series but present along with /n/ in the mid series.
/d/ vs. /n/ appears sufficiently recurrent for it to be retained as a structuring principle
in Berber deixis.
The determiners are clitics, always suffixed to the noun. There is no distinction in
number or gender, with the exception of the close demonstrative , which has a
specific plural form (-i), i.e.:
(7) rby= / rby=i / rby=n m

. lliy
This child / that child / that child over there [is] white.
(8) rbn=i / rbn=i / rbn=n m
. lliy-i
These children / those children / those children over there [are] white.
(9) tarba.= / tarba.=i / tarba.=n m
. lliy-
This girl / that girl / that girl over there [is] white.
(10) trbn=i / trbn=i / trbn=n m
. lliy-i
These girls / those girls / those girls over there [are] white.
Pronouns agree in number and gender, thus their morphology is more complex
than that of the determiners.13
(11) / i / n maz.z.g
This one (m) / that one (m) / that one (m) over there [is] small.
(12) ni / ni / nn maz.z.g-i
These ones (m) / those ones (m) / those ones (m) over there [are] small.
(13) t / ti / tn maz.z.g-
This one (f) / that one (f) / that one (f) over there [is] small.
(14) tni / tni / tnn maz.z.g-i
These ones (f) / those ones (f) / those ones (f) over there [are] small.
In two out of three series, there is a specific form for designating objects (and some
animates assimilated to the set of non-animates, such as insects):
(15) a.

taK ?
what this.one?
Who is he? Who is this one?

b. taK y ?

what this?

What is it? What is this?
13. This seems to correspond to a universal tendency, see Diessels (1999:25): Pronominal
demonstratives are more likely to inflect than adnominal and identificational demonstratives.
(16) a.

i
maz.z.g
that.one small
That one is small.
b.
maz.z.g

this.one small

This one is small.
The neutral forms ( and especially y) are used exophorically, but they also
play an important endophoric role, often in contrast to the series with ad, as we shall
see below.
2.2 The copula ad

In Berber the invariable particle d is used for nominal predication. In Zenaga, this
particle, pronounced d or , is obligatory in all affirmative non verbal sentences.14
Contrary to other varieties (where, therefore, it is not a copula), the presence of a
subject is obligatory in this Berber variety. Apparently it is usually definite, and therefore often corresponds to a determined noun (by the demonstrative clitic or a (pro)
noun) or to a pronoun (independent person pronoun or demonstrative pronoun).15
The order is Subject + ad + Noun phrase.
Zenaga
(17) yim= / bba=n= / ntt / [cop] ngrum
.
This camel / his father / him / this one (m) is paralyzed.
(18) t-yim-t= / yum
.m
. a=n= / nttah / t [cop] t-ngrum
. -t
This she-camel / her mother / her / this one (f) is paralyzed.
(19) iym-n=i / ymn= / nhni / ni [cop] ngrum
. -n
These camels / their brothers / them / these ones are paralyzed.
In Zenaga, adjectival forms (e.g. m

. lliy white and maz.z.g small), like verb forms,
assume predicative function (see Taine-Cheikh 2003b). With the other non verbal
forms including those which express states or qualities such as ngrum
. (a) paralytic the presence of the particle ad is obligatory.
This Zenaga copula is formally identical to the near m.sg demonstrative. This case
is exceptional as in most Berber varieties the predicative elements form (generally d)16
14. A specific form, of verbal origin, is used with negation.

15. Noun predicate relative phrases are a particular case; the copula ad is also mandatory,
but, exceptionally, the relater appears (having the form ), which indicates the relativization
of the antecedent in subject function (Taine-Cheikh 2007:304).
16. Thus in Figuig: netta d yuma he is my brother (Kossmann 1997:241).
is distinct from the demonstratives (especially when the latter is reduced to a vowel).
For us the Zenaga data reflects an archaic state, but other hypotheses are also viable. For
example, it is also possible that the formal variation ad vs. d between Zenaga and other
language varieties could be explained by slightly different basic constructions, one containing the determiner prop a and not the other (cf. 3.2.3). In any case, the presence of
ad / d in nominal sentences constitutes an example of the grammaticalization of the
deictic, which happens frequently in the worlds languages.17
2.3 Presentatives
In clauses containing the copula ad, the subject, as we saw above, can be a demonstrative pronoun. When the second part of the predicate (the predicated element) is a
noun phrase, either definite or not, to which the predication auxiliary lends referential
meaning, the subject and the predicative noun agree in gender and number.
(20) y / / n [cop] burdllih / m
. i / ytb=n=k
This / that / that (over there) is a centipede / a baobab / your book.
(21) / i / n [cop] ah.md / ama=n= / yim m
. lliy-n
This one / that one / that one (over there) [is] Ahmed / his brother /
a white camel.
(22) t / ti / tn [cop] kumb / tyama=n= / tyimt m
. lliy-n
This one / that one / that one (over there) [is] Kumba / her sister /
a white she-camel.
(23) ni / ni / nn [cop] ymn= / iymn m
. lliy-nin
These ones / those ones / those ones (over there) [are] his brothers /
white camels.
In such Zenaga sentences, the presence of a subject pronoun is necessary, contrary to

the other Berber varieties, where the predicative particle can also be used as an identifier, e.g. in Figuig: d netta its him (Kossmann 1997:241).
However, when ad is followed by one of the neutral demonstratives, the resulting
group (which is a stress bearing group) is an invariable predication auxiliary, with presentative meaning. In this case, only the distance varies (depending on which neutral
demonstrative is used).
(24) -y burdllih / ama=n=k / kumb / iymn / tyman=k
This is a centipede / your brother / Kumba / camels / your sisters.
17. For a discussion on the grammaticalization processes leading to the use of the non verbal
copula, see Diessels (ibid: 143 sq.).
y
(25) - m
. i / ama=n=k / kumb / iym-n / t man=k
That is a baobab / your brother / Kumba / camels / your sisters.
y
(26) -n m
. i / ama=n=k / kumb / iym-n / t man=k
That (over there) is a baobab / your brother / Kumba / camels / your sisters.
2.4 Relative pronouns or determination props

As Lionel Galand aptly showed in his article Typologie des propositions relatives
(2002 [1984]: 219240), in Berber, relative clauses generally lack relative pronouns.
Beyond the use of a specific pronoun, various solutions are possible, such as: using a
non pronominal relater, marking the verb with specific modality (i.e. bearing a nasal
suffix, known as participial marking), word order modifications, or recourse to an
empty slot. The Berber varieties make use more or less in the same manner of one
or another of these different resources, depending on the function of the antecedents
representative in the relative (for Zenaga, see Taine-Cheikh 2007 and note 13). Use is also
often made of deictic particles which furnish props for determination relations and, in
particular, relative clause type determination. Comparing the various uses of the prop
wa which, in Tuareg (Galand 2002:233), is apposed after a noun or is used on its own:
amis wa naye camel [namely:] the one I-saw = the camel that I saw or wa
naye the one I-saw = the one that I saw,
amis wa-re camel this one = this camel or wa-re this one
amis wa n amar camel of the chief = the chief s camel or wa n amar the
chief s.
In Zenaga, this construction is only found with relative clauses and the demonstratives
themselves serve as emergency props. Generally, after a nominal antecedent, the
demonstrative is not necessary. For example in the following proverb, where dependence is marked by the simple presence of the participle suffix -n:
Zenaga
(27) mn
y-rm-n
i g vu
someone 3m.sg-take.pfv-ptcp finger in hand
y-rmti
koL
3m.sg-take.pfv=pr.do.3m.sg all
Whoever takes a finger, takes the whole hand.
The presence of a demonstrative seems to add referentiality, or at least some specification. The most frequent series in this case is the mid series with anaphoric (... in
question), or deictic (clitic or pronominal) meaning referring to an already mentioned
referent (or a referent accessible through the discourse context).
(28) y-a-n
t-gr-h
this 3m.sg-be in.pfv-ptcp f-calabash-sg
y-uz. ar.i=h
vu
3m.sg-find.ipfv=pr.do.3m.sg hand
Whatever is in the calabash, the hand will find it.
The close series is used in commentary context, as in the following examples:

(29) m-=g=t-n
y-ssg
who-the.one=pr.io.2.sg=pr.do.3-pl 3m.sg-be.preferable.pfv
Which one of them do you prefer?
(30) y-ttDy=nnh
allah ynh wr=Dya-n
mn
3m.sg-leave.ipfv=op2 Allah this=op2 neg=leave.npfv-3pl people
God forgives what man has not forgiven.
As we can see in these last two examples, the pronoun affixes and the op (usually suffixed
to the verb) precede the predicate in some relative clauses. This characteristic is shared by
many subordinate clauses. However, in the case of shorter satellites (op and pr.do), this
phenomenon (called satellite attraction) entails the insertion of as a satellite prop.18
Whence the presence of in (31b) when the op is added but not in (31a):
(31) a.

taK y y-u.ur.-an
what this 3m.sg-fall.pfv-ptcp
b. taK y d=dh y-u.ur.-an

what this rel=op1 3m.sg-fall.pfv-ptcp

What fell (around here)?
In (32) with as satellite prop, the presence of i is optional:

(32) (i)
nh wr-n
y-
that one rel=op2 neg-ptcp 3m.sg-be.in.nipfv
-ih ab.a
war cop courageous
He who is not at war is courageous.19
18. In (30), ynh could be the contraction of y + + nh. This simplification appears only
with po2 (with po1, one would have y + + dh > y ddh, cf. (31b). Other cases have been
noted, however, with y as a prop for pronominal satellites.
19. In this proverbial phrase (which means: it is when someone is in danger that one can see
whether they are courageous or not), the demonstrative i would not be used.
This specific use of as a quasi-relator is limited to precise conditions (and only for
certain types of relatives). Its integration in relative clauses (where it appears to form,
along with the satellite, the first stress producing group in the subordinate clause)
marks the tendency for to grammaticalize in this specific context.
2.5 Focalization
Focalization belongs to the domain of the discourse-hierarchic perspective. Intonation, which always plays a role in the organization of a message, can indicate focalization of an essential segment in a sentence (namely with stress and contrast). Various
means such as a change in word order (especially topicalization) and/or recourse to
reminder pronouns are specific syntactic processes generally used to highlight the
topic or comment (or even both).
Galand devoted several articles to the study of these operations in Berber (1957;
2002: 331355). He showed the specificity of rhematization (focalization of the
rheme comment) as concerns relative clauses and the important role played by the
determination props in both cases:
One then obtains a nominal sentence with two terms, most often following the
pattern comment + topic: (is) X this-that one or (is) X the one (who.../that...,
etc.). Thus the sentences that over there, its a dog and it is to the dog that I gave the
meat will be respectively rendered by (is) dog that and (is) dog that to (which) I
gave the meat. The opposite order, topic + comment is also found, with an optional
pause between the two terms of the clause: that to (which) I gave meat, (is) dog, a
turn of phrase which could be called deferred rhematization.
(2002:345)
Shilha (after Galand 1957:34 and p.c.).
(33) a.

(d) afLah. a i-krz-n

(is) farmer this 3m-till.pfv-ptcp
It is the farmer who tilled.
b. a i-krz-n
d- ufLah.

this 3m-till.pfv-ptcp (is) farmer.ann

That which tilled, it is a/the farmer.
c.

w-Na i-krz-n
d- ufLah.
the.one 3m-till.pfv-ptcp (is) farmer.ann
The one who tilled is the farmer.20
20. The annexation state form ufLah. is used after the predication auxiliary d.
The predication auxiliary d, present in the topicalized sentence in (33c), is not always
present in rhematization. In Chleuh, it is mandatory in deferred rhematization (or
restrictive focalization) as in (33b), but not in standard rhematization in (33a).
In Zenaga, the presentative y- is used in standard rhematization, in associa
tion with the deictic i ( ... in question).21 Thus, once again, one finds a predication
auxiliary (having for initial element the neutral near demonstrative), except that here
it is invariable. Therefore the commentaries tar.ba.-i the girl (there) and trbn-i
the girls (there) are not introduced, respectively, by the demonstratives f.sg t / ti
/ tn + and f.pl tni / tni / tnn + (as in t- tar.ba. its a girl), but
instead by the presentative y- in (35):
Zenaga
(35) a.

y- t-ar.ba-.=i r
knt wy-g
It is
the girl (there) with whom (past) I spoke.
b. y- t-rb-n=i
r
knt wy-g

It is
the girls (there) with whom (past) I spoke.
The various roles played by ad in predicative function (i.e. as a copula and a predication particle, in presentatives) are quite removed from its uses as subject demonstrative or determination prop. Nevertheless, all the various uses of ad can be quite easily
understood given the usual meanings of demonstratives.
However, ad does have other functions. For Galand, these uses correspond to
those of a modal particle, not those of a conjunction (2002 [1987a]: 252). We shall see
below to what extent this affirmation also applies to Zenaga.
3. The use of tam markers and dependency relations

3.1 Presentation of the verbal system
The verbal system in Berber is fundamentally based on an aspectual opposition bet
ween perfective vs. imperfective (in French accompli vs. inaccompli). The imper-
21. An example without ay- has been noted however, linked perhaps to the presence of
the relater y (formed by the determination prop ay + the relation particle ):
(34)

t-rb-n=i
y t-ah.-a
t-wiy-
t=iNy
f-girl-pl=those rel 2-can.pfv-sg 2-speak.ipfv-sg with=pr.3pl.f
Those are girls with whom you may speak.
However, this same sentence (albeit incomplete) with different intonation could mean Those
girls with whom you may speak (...).
fective form of non derived verbs is characterized by the gemination of a root

consonant (generally the second) or by prefixation of t(t) to the stem.
Table2. Perfective and imperfective forms (Zenaga)
Perfective (pfv)
Imperfective (ipfv)
yrm he took
ya. he caught scabies, he has scabies
yud.d.a. he suckled
ykkur he is (has become) active
yirmm he takes
yi-Dya. he (will) get(s) scabies
yttad.d.a. he is suckling
ytkuri he is becoming active
The forms used in negative contexts (after the negation wr not) often have several variants, either only for the perfective, or also for the imperfective.
Alongside these positive and negative verb forms in the indicative, one must also
add the imperative (imp) and Aorist (a) forms. These regularly share a single vocalic
pattern but, in the imperative, the 2nd person marker t- ...-[d] is absent: only the gender
(f) and number (pl) markers remain. This pattern, which in numerous Berber varieties
is identical to the Perfective for some of the forms, is completely distinct in Zenaga for
almost all verbs, due to the retention of the distinction between the vocalisms and .22
Table3. Verbal vocalic patterns (Zenaga)
imperative
aorist
perfective
negative perfective
rm
ad.d.u.
yrm
yad.d.u.
yrm
yud.d.a.
yrm
yud.d.u.
The form called Aorist, whose meaning is that of a neutral form, holds a special
place within the Berber system. Used on its own, it only rarely alternates with perfective and imperfective forms. It does however play a major role in the language, because
of its frequent use in dependent clauses or preceded by a particle.
As the most frequent particle is ad, in Zenaga and the other Berber varieties, we
will explore its different uses, both with the Aorist and with the other tam markers. We
will start by examining the uses of the Aorist without the particle.
3.2 Uses of the Aorist without ad

The Aorist without any particle is almost never used as the first verb in a sentence in
Berber. It is however used as a second (and subsequent) verb.
22. And even more so between a and i or a and u. For more details, see Cohen and
Taine-Cheikh 2000.
3.2.1 Verbs in juxtaposed clauses with sequential meaning

Pan-Berber coordination conjunctions are rare, and some varieties, such as Zenaga,
continue to use them sparingly, although some borrowings from Arabic are used on
occasion (e.g. yayr but). In the absence of a usual coordinator (the coordinator of
comitative origin, d in Zenaga being mainly used between NPs), juxtaposition
is the most commonly used method in the enumeration of processes, states, events
(unless it is to express an alternative). Compare (36) with na or to (37) and (38),
where the second (and third) verb phrases are simply juxtaposed to the first.
Zenaga
(36) bh
na llg
leave.imp or stay.imp
Leave or stay!
(37) a
vu=n=k
okkh
=i
y
hold.out.imp hand=of=pr.2sg carry.imp with=pr.1sg this
Hold out your hand [and] carry this with me!
.
(38) y-uw=ddh
umnn=
y-ig
of =-n
3m.sg-bring.pfv=po1 camel.pl=pr.3 3m.sg-put.pfv on=pr.3-pl
t-w y-ig
rgn=
i t-ntym-t
f-luggage 3m.sg-put.pfv womans.saddle[pl]=pr.3sg for f-woman-sg
(...) He brought his camels, put (the) luggage on them [and] put his
(womans) saddle for the woman. (...).
Examples (37) and (38) show that, in discourse as in narratives (the last example is
taken from a story), processes may be juxtaposed without any coordination marking.
In these two examples, the vps in apposition have the same tam markers: two imperatives in (37) and three Perfectives in (38).
In Berber, the juxtaposed verb forms (vp2, vp3, etc.) may be replaced by Aorists
in enumerations. This happens more or less frequently depending on the variety, and
more often with Imperfectives than with Perfectives: in this case the verb takes on the
same aspect-mood meaning as the first verb, often choosing the Aorist, and lending
particular stylistic nuances, chaining for example (Galand 2002 [1987b] and 2003).
In Zenaga, choosing the Aorist after an imperative or Imperfective (this cannot
be done with a Perfective) is quite rare, and apparently never mandatory. It is slightly
more frequent in proverbs, but this does not seem to be due to any archaisms. In fact,
replacement by the Aorist was accepted by our consultant whenever the conditions
bearing on the tam markers were met and the juxtaposition had sequential meaning.
Here are various examples where the verb in the Aorist expresses an action having
a logical link with the preceding one, and first of all the conclusion of the story the
jackal and the hare: the judge (a hedgehog) proves that the jackal has no rights to the
young heifer because his bull cannot have given birth, contrary to the hares cow.
(39) y-nn(h)
ssn=ti
t-kf-
3m.sg-say.pfv=pr.io.3 know.imp=pr.do.3m.sg 2-give.a-sg
t-rki-=n=
i t-rumbL
f-young.heifer-sg=of=pr.3 to f-hare.sg
He tells him: know it [and] give his young heifer to the hare!
At times, it is some habitual succession of events, or a chain of events which corresponds to semantically asymmetrical coordination:
(40) m
am=n=k
. uy butilimt t-az. r.-a
go.imp Boutilimit 2-find.a-sg brother=of=pr.2m
t-um
.m
. ugr-=i
2-come.back.a-sg=op1
Go to Boutilimit, find your brother [and] come back here.
It can also be a sequence of events running counter to the normal state of things. It is then
equivalent to an adversative coordinator (but, nevertheless), as in the following saying:
(41) ggn <Ty>uz. uz.z.ug-n
unnugur-n=ti
bard.pl <pass>beat.impfv-3pl hide.a-3pl=pr.do.3m.sg
The bards, one hits them [and] they hide it.
3.2.2 Main clause verbs depending on an event pattern

Quite frequently, the choice of the Aorist for the main clause verb is conditioned by
the nature of the preceding clause: often a temporal subordinate clause with habitual
meaning.
..
(42) og=r
y-f f u
y-bi
g mn
moment=in.which 3m.sg-be.thirsty.pfv 3m.sg-drink.a in water.pl
When he is thirsty, he drinks water.
It can also be the protasis of a conditional clause (see Examples (3) and (4)) or a specifying relative (43):
(43) mn
y-ttttr-n
dy-ykh
yx=n=
someone 3m.sg-want.ipfv-ptcp ad=3m.sg-disappoint.a marabout=of=pr.3
.
y-ssnkr=ti
of y[] wry-r(i)
allah
3m.sg-invoke.a=pr.do.3m.sg on this neg=3m.sg-want.npfv Allah
He who wants his marabout to disappoint him, invokes him for
something that God doesnt want.23
23. This proverb contains another one of the uses of ad + Aorist (after yttttr he wants)
which we shall examine in 3.3.1.
In the preceding examples, the subordinate clauses refer to some generic, usual, or
possible situation. If the subordinate clause corresponds to a temporally specified situation, the main clause verb is not in the Aorist. Thus, after a temporal subordinate
clause introduced by ll-r (lit. place where), the verb in the main clause is in the
perfective form:
(44) llr= -g
ar nwakT.
when=op1 come.pfv-1sg to Nouakchott
uz. r.-ag
adyan
find.pfv-1sg family.plpr.1sg
When (as soon as) I arrived in Nouakchott, I found my family.
The same holds true for the causal subordinate clause introduced by -r-y since
(lit. this where that):
(45) ry=t
t-uz.z.ar.- t-ssn-
all=n=
since=pr.do.3f.sg 2-see.pfv-sg 2-know.pfv-sg place=of=pr.3
Since you saw her, you know her place (the place where she lives).
3.2.3 Main verbs of complex predicates after auxiliaries with modal meaning
Some complex predicates are made up of an auxiliary verb (at times with incomplete
conjugation paradigms) and a main verb, where the person markers are coreferential.
The choice of the main verbs tam generally depends on the aspectual or modal meaning of the auxiliary. It is always in the Aorist after two auxiliaries with modal meaning.
yaha. expresses a possibility (he can, was able). The invariable impersonal Dyri
denotes a necessity (it is necessary that).24
(46) nkni bb= Dyri n-yg
wr-n
ffllh
us
year=this
must 1pl-ascend.a in.direction-of up
Us, this year, we must go to the countryside.
On the contrary, after auxiliaries like yum

. r already be and yugg almost have, the
main verb is in the Perfective:
(47) um
b-n
. ra-n
already.be.pfv-3pl
leave.pfv-3pl
They have already left.
(48) t-ugg-
t-u.ar.-a
2-almost.have.pfv-sg 2-fall.pfv-sg
You almost fell.
24. The etymology could be: dyri that he want, from the verb yr (A yri) want.
3.2.4 Dependent clause verbs with consecutive meaning

In 3.2.1, the verb in the Aorist (without ad) referred to a process sequentially linked
to a preceding process. However, at times, the verb in the Aorist expresses a goal or a
consequence of the first process. In this case, the two verbs no longer share the same
temporality nor, often, the same subject.25 The attraction of the pronominal expan.
sions (of - on + pr) to pre-verbal position marks dependency.
.
(49) siyigm=ih
of =kn
yi--g na<>n
raise.imp=pr.do.1sg on=pr.2m.pl tell.a.1sg news. pl<pr.1.sg>
Raise me so that I may tell you my story!
.
(50) lubb
ilh
of =km
yi--g t-llS
Loubbe! this.way on=pr.2f.sg tell.a-1sg f-story.sg
Loubbe! This way, so that I tell you a story!
.
(51) a
vu=n=k
of =ki
sllm-g
extend.imp hand=of=pr.2m.sg on=pr.2m.sg greet.a-1sg
Hold your hand out that I may greet you!
3.3 Dependent clause verbs preceded by ad

Several cases are possible, depending on the governed verbs TAM.
3.3.1 ad + Aorist

This construction, undoubtedly more frequent with some verbs than with others (see,
for Kabyle, Nat-Zerrad 2001:141), is found with two sets of operators. Denoting a goal
could be common to both, but this is less clear in the first case than in the second.
The following examples show many similarities with those in 3.2.4. They seem
to have exactly the same consecutive or purposive meaning, and only differ in
the presence of ad preceding the Aorist. Apart from Example (52), the particle ad
could play the role of satellite prop (pr or op), which was not necessary in pres.
ence of the preposition of .
(52) ilh
-b-g
this.way ad drink.a-1sg
This way, that I drink!
(53) ilh
dt
-ib-g
this.way ad=pr.do.3f.sg make.drink.a-1sg
This way, that I make her drink!
25. ilh is an adverb (borrowed from dialectal Arabic) not a verb, but it also expresses a process.
(54) rmi
r= t-angaL.
rmi
remove. imp of=pr.3 f-candy.sg remove.imp
.
kf (i)
dti
t-and.ug
give.imp=pr.io.3 ad=pr.do.3m.sg 3f.sg-taste.a
Remove the candy! Remove it! Remove it! Give it to her, that she taste it!
(55) y=h
dt
az. r.-ag
wait.imp=pr.do.1sg ad=pr.do.3f.sg find.a-1sg
Wait for me, that I find her.
This construction with ad and the Aorist is sometimes equivalent to the construction
with the subordinator hn so that (which probably contains the element ), but this
particle (always followed by the perfective or imperfective) is used to express the goal
specifically, contrary to ad.
.
.
(56) akf (i)i=
az. r.f i hn aS-g
tiyn
wait.imp=pr.do.1sg=op1 money so.that buy.i-1sg shoe.pl
Give me (here) money so that I (may) buy shoes.
Governing verbs expressing orders, requests or wishes (expressed openly or to

oneself) are also regularly constructed with the Aorist preceded by ad.
(57) tttr-g
=ih
y-iddug
want.ipfv-1sg ad with=pr.1sg 3m.sg-marry.a
I want him to marry me.
(58) yb-g
tttr-g
iywu.-ag
leave.ipfv-1sg want.ipfv-1sg ad ritual.cleansing.a-1sg
I am leaving, I want to do my ritual cleansing.
(59) y-nn(h)
ti
y-ar.z. i
3m.sg-say.pfv=pr.io.3 ad=pr.do.3m.sg 3m.sg-break.a
He told him (asked him) to break it.
ad + Aorist is used after certain verbs such as ynnh say (which then expresses a
demand, a request) for indirect speech, and especially for indirect orders. Compare
(59) to (60), where the order is expressed directly, using the imperative (and not
using ad):
(60) y-nn(h)
ar.z. i=h
3m.sg-say.pfv=pr.io.3 break.imp=pr.do.3m.sg
He said to him: break it.
When several orders are given, the particle ad is repeated before each Aorist, as in the
following example taken from a story:
(61) t-nnh
iy obby y-rmi
t-i-
3f.sg-say.pfv to slave ad=3m.sg-take.a f-calabash.sg
y-az.z.ig
kr=n=i
ad=3m.sg-milk.a a.little=of=milk
dti
ad=pr.do.3m.sg
y-okkh wr-n ya.=6n=

3m.sg-carry.a
towards husband=of=pr.3
She told (asked) a/the slave to take a calabash, to milk a little bit [and] to bring
it to her husband.
3.3.2 ad + yu(u)g or yum

. r in indirect yes or no questions
In Zenaga, ad is also used to indirectly ask a question, or to express doubts as to

a process. In this case, ad is followed by an auxiliary verb, frozen in the 3m.sg:
either the verb become, switch to a certain state (Dyu(u)g < adyu(u)g), or,
more rarely, the verb be already (Dyum
. r < adyum
. r). yu(u)g and yum
. r were
y
perhaps themselves, at the start, Aorist forms. In any case, D u(u)g and Dyum
.
r are always followed, here, by a verb in the perfective or in the imperfective. The
most frequent governing verb is ygDyh ask, question, interrogate, but examples
have been found after yuz. z. ar. look, wr yssn not know, yznzgm think, try
to understand.
(62) gDyi ar am=n=k
dyu(u)g
ask.imp about brother=of=pr.2m.sg ad+yu(u)g
Ask on the subject of your brother if
a.

y-=ddh
3m.sg-come.pfv=op1
he came.
b. y-nhy
y-tt=ddh

3m.sg-future.pfv 3m.sg-come.ipfv=op1

he will come.
(63) wr=Sn-g
ady-um
. r ybb-n
neg=know.npfv-1sg ad=yum
. r leave.ipfv-3pl
I dont know if theyre leaving.
This means of expressing indirect yes or no questions seems to be restricted to Zenaga,

and bears no relation to the direct yes or no question marker (taK or k- what ) which
appears in some contexts.
(64) k=y
t-sskkr-
what=this 2-do.ipfv-sg
What are you doing?
3.3.3 ad + Perfective or Imperfective in indirect speech

Another case, which also seems to be restricted to Zenaga, deserves attention: the use of ad
following the verb yinnh say. It is particularly frequent in the introductory phrase for all
stories. The verb in the subordinate clause can be in the Perfective (65) or in the Imperfective
(66), depending on the relation between the time of narration and the events narrated.
(65) inn-n
dti
y-uy
say.pfv-3pl ad=pr.do.3m.sg 3m.sg-exist.pfv
tyfaa y-m
. ur-n
chief 3m.sg-be.old.pfv-ptcp
People say that there was an old chief [...]
(66) Zn=
dt
tttr-g
ty=
ddug-g
say.imp=pr.io.3 ad=pr.do.3f.sg want. ipfv-1sg with=pr.3 marry.a-1sg
Tell her that I want to marry her.
The tam markers in the completive clause distinguish between indirect speech (say
that ...) and indirect orders (say about ...). The latter is characterized by the subordinate verb in the Aorist (see 3.3.1).
The use of quotative ad is common in discourse and narratives, but it is not generalized because direct speech, even in narratives, is very frequent, cf. Example (39).
The case raised by the use of this complementizer ad seems quite simple as compared to the other uses of ad. We shall determine in 4. whether it sheds light on the
other cases of grammaticalization.
One must specify however that this construction appears only after yinnh. Generally speaking, the subordinator y introduces completive clauses, after opinion or
judgment verbs such as yznzgm think (that), ponder and after emotion or modal
.
verbs such as yssug prefer (that), ynnum
. want (that), yuf.ra. force to do.
(67) nnum
. -g y ukty-g
want.pfv-1sg that stay.pfv-1sg here
I want to stay here.
3.4 ad in fronted position

The uses of the Aorist (with ad) which we shall now examine are (almost) always in
fronted position.
3.4.1 ad + the prohibitive and injunctive Aorist

Berber has specific forms for expressing orders. The imperative forms are very similar
to the Aorist,26 as we saw in 3.1.
26. Such formal similarities are found in many languages having a so-called Aorist form (for
example in Mwotlap and Wolof, see Franois 2003, Robert 1991, etc.). This cannot be simple
coincidence. In Berber, it does indeed correspond to a semantic proximity.
With prohibitives, i.e. negative orders, the verbal negation particle wr never
combines with the imperative.27 In many varieties, the imperative is replaced by the
Aorist in this case, i.e. the form which resembles it the most.28
(68) a.

tyi
b. wr=t-tyi-
eat.imp ad neg=2-eat.a-sg
Eat! Dont eat!
In the following proverb, the same verb is used successively as an order and as a
prohibition:
(69) gtawni gm
wy=n=
. ug
wise.man follow.imp words[sg]=of=pr.3
wr=t-gm
y-sskkr
. ug-
ad neg=2-follows.a-sg that 3m.sg-do.ipfv
The wise man, follow his words, not his actions (lit. what he does).
While the imperative, being reserved for hearers, is only used in the 2nd person,29
the prohibitive does not differ from negative orders which may concern other people,
including those who are absent.
(70) =ki
wry-uz.z.ur.
allah
ad=pr.do.2sg neg=3m.sg-see.a Allah
g ll=r
wry-ukiy
in place=in which neg=3m.sg-want.npfv
That God not see you in a place where He does not wish to see you!
Affirmative injunctions are also possible,30 but in Zenaga, the only spontaneous
example we were able to find is in a given context, as in the following proverb:
27. At least with imperatives having the same stem as the Aorist. In some varieties, including
Zenaga, there are in fact imperative forms i.e. lacking the marker t-[-] for the 2nd person
which share a stem with the Imperfective. These may combine with the negative particle.
28. In Zenaga, negation is then, exceptionally, associated with Aorist forms (see (68b) and
(69)), whereas the negative Perfective and the negative Imperfective are normally used after
the negation wr, see (63) and (81).
29. We do not have any examples of the imperative in the 1st person. The cohortative is
expressed by ad + Aorist, as is the injunctive.
30. In Berber, ad followed by the Aorist frequently serves to express positive orders and
also the optative (positive and negative). Here is an example of a wish taken from Ar Tuareg,
analyzed by Galand (2002 [1984]: 126): Ialla1, a2 di3 t4qqn5a4! By-God1, that2 you4 me3 attach!
= I beg you, attach me!. In this variety, however, true injunctions are indeed expressed by the
Aorist, but it is to the verb that the particle -et is suffixed: ak1 awedn2 i3kkl4et5 teyint6 -net7! that5
each1 man2 he3-take4 the-pot6-of-him7.
(71) mn
wr-n
y-uz. r.i
y
someone neg-ptcp 3m.sg-find.npfv this
i-nbb
dy-zzn
tT
3m.sg-say.ipfv ad=3m.sg-say.a truth
He who has not found something to say, that he speak the truth!
(72) is a variant of (43). It differs however by the presence of ad and the attraction of
the suffix pronoun ti to pre-verbal position, as well as in its intonation which, with ad,
is exclamative.
(72) mn
y-ttttr-n
dy-ykh
someone 3m.sg-want.ipfv-ptcp ad=3m.sg-disappoint.a
yx=n=
dti
y-ssnkr
marabout=of=pr.3 ad=pr.do.3m.sg 3m.sg-invoke.a
.
of =y[] wry-r(i)
allah
on=this neg=3m.sg-want.npfv Allah
He who wishes to be disappointed by his marabout, that he invoke him for
something that God does not want!
3.4.2 ad with future connotations

In some varieties, the future is expressed by a specific construction. In Zenaga, the
future is expressed by means of the verb ynhy be occupied which, as future
auxiliary, is conjugated and forms a complex predicate, along with the main verb.
In varieties such as Shilha, the future is expressed using the particle rad, a grammaticalized form of want + ad (Galand 1977: 302). In this Berber variety, the
particle ad thus accompanies the Aorist in certain types of clauses, be they independent (orders, wishes) or dependent (consequence or goal), in sum, a list more
or less identical to what we just saw for Zenaga (Galand 1988:221, 2245, etc.).
In many Berber dialects, however, the Aorist preceded by the particle ad (or
one of its variants) provides a means (non specific but regular) of expressing the
future.31
At Seghrushens Tamazight (after Bentolila 1981:173)
(73) ad azn-x
lflus
ad send.a-1sg money
I shall send the money.
31. Based on the distinction proposed by Galand (1977:302), ad + Aorist would serve, if not
to denote the future, then at least for future connotations.
Kabyle
(after Nat-Zerrad 2001:110)
(74) ad as-
t-
fke-
tameddit
ad pr.io.3 pr.do.3m.sg give.a-1sg (the.)evening
I shall give it to him in the evening.
Figuig
(after Kossmann 1997:269)
(75) ad y-acer
ad 3m.sg-steal.a
He will steal.
Among the Berber varieties using ad to refer to the future, some of them also use
the imperfective form (ipfv). Such is the case for the Figuig variety, where the
distinction in meaning added by the imperfective is that of habit or repetition.
Figuig
(after Kossmann 1997:269)
(76) ad i-ttacer
ad 3m.sg-steal.ipfv
He will steal habitually, he will become a thief.
3.4.3 ad in conditional clauses

The use of ad in Zenaga conditional clauses raises a classificatory problem. It is a construction seldom used outside Mauritanian Berber. Indeed, it seems to us that, as in
Arabic (see Peled 1992), the protasis and the apodosis make up a correlative structure,
with set word order. There is a dependency relation between the two clauses, but it is
a question of codependence rather than a relation of subordinate to main clause, in
particular where both of the conditional clauses verbs are in the Aorist.
In expressions of condition, the verb after ad is frequently in the Aorist. When the
verb in the apodosis is in the imperative (as in (4)) or, more usually, in the Aorist,
the condition takes on habitual or generic meaning, see (77) and (78a). With the
Imperfective, whether negative or not, the potential meaning appears to be dominant, see (78b). The following example the morale to a story which illustrates the
need for brothers to stick together illustrates its generic meaning.
(77) dt-ddug-m wry-aha.

mn
ad=2-agree.a-pl neg=3m.sg-be.able.[n]pfv someone
y-iullur.=kn
dt--r-m
3m.sg-be.unfair.a=pr.do.2m.pl ad=2-separate.a-pl

kyun g=n
y-u.uff
each=one in=pr.2m.pl 3m.sg-weaken.a
If you agree amongst yourselves, no one (can) do you harm. If you separate,
each one of you becomes (will become) weak.
The choice of the tam in the apodosis determines the meaning: general condition in
(78a) and goal in (78b):
(78) a.

..
wry-bi
y-uf f u
ad neg=3m.sg-drink.a 3m.sg-be.thirsty.a
Although he isnt drinking, he is thirsty.
b. wry-bi / y-bi
y-nhy
y-tfu

ad neg=3m.sg-drink.a/npfv 3m.sg-future.pfv 3m.sg-be.thirsty.ipfv

If he doesnt drink, he will be thirsty.
(78b) shows that the Aorist can be replaced by the negative Perfective in a negative
protasis. There is however another type of hypothetical system, where the (main) verb
in the protasis is never in the Aorist.
When conditional clauses make use of the referential situation, Zenaga ad is followed by yu(u)g (> Dyu(u)g), the frozen form of the verb become, switch into
the state of (the same one which appears in indirect questions as seen in 3.3.2).
In these factual type hypothetical systems, which may be implicative or explicative (Banys 1996:222), all forms of the perfective (pfv and npfv) and the imperfective (ipfv and nipfv) seem to be allowed, in both the protasis and the apodosis,
with various changes in meaning.
..
Dyu(u)g y-b
wryuf f u
adyu(u)g 3m.sg-drink.pfv neg=3m.sg-be.thirsty.npfv
If he drank, he is not thirsty.
..
b. Dyu(u)g y-b kn
y-f f u

adyu(u)g 3m.sg-drink.pfv past.3 3m.sg-be.thirsty.pfv

If he drank, [then] he was thirsty.
..
(80) Dyu(u)g y-zss
y-f f u
adyu(u)g 3m.sg-drink.ipfv 3m.sg-be.thirsty.pfv
If he drinks, [then] he is thirsty.
..
(81) Dyu(u)g wry-uf f u
wry-zissi
adyu(u)g neg=3m.sg-be.thirsty.npfv neg=3m.sg-drink.nipfv
If he isnt thirsty, he doesnt drink.
(79) a.

In Example (82), the lexeme explicitating the person marker (here bba-n- his father)
is placed between Dyg and the verb y6=y. This is the sign of strong solidarity between
ad and yu(u)g, and could be additional proof of the grammaticalization of Dyg into
a subordinating phrase. Indeed, in the habitual uses of ad, such a lexeme would be either
preposed to the particle ad (which is repeated in this case), or postposed to the verb.
(82) Dyg
bba=n=
y-6-y
y-6tti-iy
adyu(u)g father=of=pr.3 3m.sg-be.fast.pfv 3m.sg-be.fast.ipfv
If his father was fast, he will be fast.
We will now study the origin of the Zenaga conditional particles.
4. Deixis, referentiality and discourse structures

Numerous are the uses of ad in Berber. The deictic origin of the morpheme undoubtedly explains most of these uses (copula, presentative, relative pronoun, etc.), as comparable grammaticalization processes have been observed in many other languages.
We will now attempt to determine if such an origin can help us understand the other
uses of the demonstrative, especially those, only found in Zenaga, where ad seems to
function as a subordinator. That being said, given that the place of ad and/or of the
Aorist has played an important role in the analysis of our data, we shall first of all seek
to better understand the relation between topic and tam.
4.1 Topicality, determination and referential constructions

The placement of the topic and comment depends on the speakers choice and can
be modified according to the organization desired for a given sentence. The choice
of the fronted element (initial locator), however, belongs entirely to the informative structure of the sentence (except in the case of specialized focalization, which
calls upon insistence stress phenomena) and to the basis-addition relation: a
fronted element, which corresponds to the least informative segment, is followed by the commentary, which corresponds to the most informative segment
(Taine-Cheikh 1999b).
This definition of the topic and its distinction from the initial locator may be
imposed by the nature of a language. Given the syntactic order determined-determiner
which characterizes Arabic and Berber, the fronted element may be a more or less
determined argument (and not necessarily a highly defined topic argument). However,
as in many of the worlds languages, the presence in this initial position of an entirely
indeterminate element is problematic.32
In Arabic, an indefinite np may occupy initial position under certain conditions.
These conditions depend on the Arabic variety in question,32 but they all follow the
same rule: to be fronted, all nps must be minimally specified, either quantitatively or
qualitatively. The partially defined initial locator, which fails the test of amm as for
a test serving to express a change in topic in Classical Arabic , must therefore be
distinguished from initial topicalized nps (see Kouloughli 1994).
Berber scholars have not felt the need to distinguish, in an analogous manner, between simple prop nps and topicalized nps. It is possible that the nps in initial
. In any case, it can be the presence of a determiner relative, but in Classical Arabic the
diminutive form can suffice to make topicalization possible, and in Arabic dialects, this can
be carried out by the indefinite modality, for those which have acquired one (often by grammaticalizing the numeral one).
osition are topics33 on a regular basis. However, it would seem surprising to us that
p
the prosodic difference observed in Kabyle by Mettouchi (2006: 118122), between
the post-verbal subject (or explicative complement) and the right dislocated subject
(dislocation being indicated by an intonational break)34 should not similarly distinguish the pre-verbal subject (in initial position) from the indicator of the left dislocated
topic.3334
In the case of Zenaga, the np as initial locator is not followed by a pause, while nps
with topicalized meaning may be followed by a pause. We also believe that some nps in
initial position are too indeterminate to be topics. This is the case e.g. of mn someone,
which can only be fronted if it is determined by a relative clause. Thus one may make
a distinction between the fronted mn (an indefinite lexeme denoting a male person,
an individual, whence the meaning someone) and the noun phrase with topicalized
meaning which it constitutes along with the relative clause wr-n yuzr i and its expansion y inbb:
(71) mn
wr-n
y- uzr i
y
someone neg-ptcp 3m.sg-find.npfv this
i-nbb
dy-6zz6n
t6T
3m.sg-say.ipfv ad=3m.sg-say.a truth
He who has not found something to say, that he speak the truth!
While in Berber the question of the simple prop np may appear rather secondary,
the problem raised by the place of the Aorist in a sentence is clearly of the utmost
importance for understanding the role of ad. In fact, the two phenomena are probably
similar in their workings. The impossibility for an indefinite np to appear in initial
position without any qualitative determination is comparable to the impossibility for
a verb in the Aorist to become the first verb in a sentence without the presence of certain particles. Given that the Aorist is the neutral form in the system, it appears that
there is a direct relation between its indeterminate tam meaning and the fact that
it is excluded from fronted position, all the more so as this constraint, far from being
limited to Zenaga, or even Berber, tends to also be found in other languages having an
Aorist with similar meaning.
. This is the impression one gets, for example, from the study of Rifain by Lafkioui
(2002:26670). An intonational dislocation (a melodic peak followed, in principle, by a drop
in pitch), characterizes all topic indicators, but the notion appears to be reserved here for
initial nps with topicalized meaning (which, as specified, is not intrinsically defined by their
pre-verbal position, even if it is their most frequent position).
. In this case, we prefer to use the term post-comment.
The Aorist in Wolof for example, which is also called Zero aspect, shows many
points in common with the Berber Aorist.35 Robert (1996) suggests considering it a
situationally dependent form. And indeed, in all the cases where the Berber Aorist
is used on its own, it is always preceded by a verb or clause which establishes the
situational background necessary for referential construal.353637
The Berber Aorist does not seem to have modal meaning in the sense of indicating subjective implication: what modal meaning it may have is more due to the
fact that in itself it is not anchored in discourse time. When it instantiates a process
therefore, it only does so indirectly, through the aspect-tense determinations transmitted by the situational background in which it is inserted (whence its plasticity
in the case of sequential juxtaposition, its capacity to take on the same meaning as
that of the preceding process). However, processes are not always anchored in precise frameworks, e.g. enumerations (cf. (37)) where each event necessarily follows
a certain (logical) continuity with the preceding event.36 Of course this framework
may be very general and be left partially indefinite, either because it corresponds to
usual situations (cf. (42)) or because instead of being located in the real world, it is
located in a possible world (cf. (43)). This explains why only the Aorist is used in
main clauses following certain temporal subordinate clauses37 and in subordinate
clauses with consecutive or purpose meaning. One also understands why certain
meanings, more compatible than others with the fundamental indetermination in
meaning, can be attributed to the Aorist without a particle: habitual meanings or
potentiality and possibilities.
4.2 Deixis am Phantasma and tam particles

The syntactic uses of the Aorist without particle are distinctly different from its uses
preceded by ad. With the exception of dependent verbs with consecutive or final meaning (see 3.2.4, 3.3.1), to which we shall return below, a noteworthy change takes place
. The uses of the Aorist are not identical in the two languages (in Wolof it is often found in
proverbs and questions), but many of them (narrative Aorist, injunction, final or consecutive
subordinates...) are common to both (see Robert 1991, 1996). Furthermore, it is symptomatic
that in Wolof, in stories and historical tales, the Aorist is never found as the first verb form
(Robert 1996:154).
. The reasons behind the choice of Aorist without particle in Berber are not always quite
clear. In any case, it is not always a stylistically marked choice.
. In which case a certain co-dependency relation is set up between the two clauses: the
first is subordinate to the second through the use of a subordinator, and the verb in the main
clause (in a non-assertive form) depends on the framework set up by the subordinate clause.
in presence of the particle ad: the possibility for the Aorist to appear in initial position.
This fact is common to the three uses studied in 3.4 (prohibitions and orders; future
and, more generally, tam particle; conditional clauses). We will now attempt to understand the causes of these phenomena, starting with the use of ad as tam particle. Certain peculiarities in its uses deserve more attention.
4.2.1 Non chained forms in Tamazight

The notion of chaining was proposed by Bentolila in his study on At Seghrushens
Tamazight (Bentolila 1981:152 sq.). He draws attention to the fact that having recourse
to Aorists (without any particle) in sequential chains sets up a specific, tight, semantic link, between various actions, but actually, in this variety, the distinction between
chained and non-chained forms is particularly significant. At Seghrushens Tamazight
is indeed endowed with a rich paradigm of preverbal aspect-tense and modal particles,
both affirmative (ad, a, al, La, da-ad) and negative (ur, uLi). The structuration of the
verbal system is based on a tripartite distinction: positive vs. negative, real vs. unreal,
and chained vs. non-chained.
While in a Berber variety such as Zenaga, the Aorist is the only verb form reserved
for sequential dependence uses, in At Seghrushens Tamazight there are two chained
forms, the Imperfective preceded by al, alongside the Aorist on its own. All the others
are non-chained forms (the only ones susceptible of being in initial position), in particular those with ad followed by the Aorist or the Imperfective. Whereas the former
cannot be fronted, the latter may be used in enumerations (i.e. in the same contexts
as chained forms). This is where the semantic difference between the two groups of
forms appears most clearly: the first denote chained (or logically dependent) actions
while the others translate isolated, disjointed actions, almost excluding one another
(Bentolila 1981: 154). The presence of ad therefore endows the Aorist with its own
situational references (as do ad and La, for the Imperfective) the clause thus becomes
as independent as one with a verb in the imperative or Perfective while the Aorist
on its own, and al + Imperfective place the clause in a state of situational anaphora
(Robert 1996:160).
4.2.2 ad and verbal satellites
Satellites (op and pronoun affixes) are usually positioned after the verb, but in subordinate clauses, they are often attracted to preverbal position, especially in relative clauses
and in constructions having their origins in relative clauses (Galand 1988:220). They
are also displaced before the verb when it is preceded by negation or certain particles
such as ad (and ar for Ighchan).
The causes of this attraction are unclear, but as pointed out by Mettouchi
(2006: 118), this attraction corresponds to a reorganization: the verbal phrase loses
its role of pivot in presence of the negation ur or the particle ad to the extent that they
become the anchor points for the satellites.38 As for this pivot role, we believe it could
function not only on the morphosyntactic level, but also on the semantic-referential
level. We do not uphold that syntactic anchoring and referential anchoring are necessarily superimposable, but it is not impossible that they be parallel, both for the ad as
tam particle and for the other ad (starting with the relative prop demonstrative ad).38
4.2.3 The anchoring function of ad

Among the numerous functions of ad, two (3.2, 3.3) concern noun predication. In
Zenaga, the presence of ad between two noun phrases suffices, along with the appropriate intonation, to construct a noun predicate clause, where the first np is the subject
and the second the predicated element (see Example (17) and following). This construction shows differences from one variety to another, to the extent that some, such
as Tuareg, can directly predicate a noun (without d), while other varieties, namely
Kabyle (cf. (83)) and other central Moroccan varieties only allow noun clauses limited
to a single np preceded by d.
Kabyle
(83) irwin These are crazies.
(Galand 1988:218)
Syntactically, the two particle constructions (ad / d) + np (with or without a subject

np) are distinct and if the construction without a subject np does not derive from
the one having a subject np, it is preferable to attribute different origins to them.39
However, the role of the particle is similar in both cases, from a semantic-referential
perspective. It consists in anchoring the np by attributing it a referent in the real world
or in the discourse universe. This np can be definite, but is most often without determiners (see (17) and (83)) and, in Kabyle, the simple presence of the particle renders
a non definite np frontable.39
Given the similarities previously observed between the np and the vp in initial
position, the anchoring function of the demonstrative ad was perhaps transposed,
metaphorically, from the indeterminate predicative np on one hand, to the predicative
verb phrase with an indeterminate vp on the other hand. Be that as it may, ad confers
a type of referential autonomy upon the Aorist (the indeterminate vp) comparable to
that which it confers upon the indeterminate np, an autonomy which is manifested for
example in the possibility for ad + Aorist to occupy initial position.
There is nonetheless a characteristic reserved for vps with the particle ad: they
are anchored in the world of possibles, because the verb form retains a large dose of
. In Zenaga, satellites are placed after ad when the two particles are present.
. The copula ad, which establishes a link between the two nps, is probably of an endophoric
nature, while the identification d (translated as it is) could be, originally, of an ostensive type.
indetermination, despite the presence of ad as a tam particle. This is true of the Aorist
preceded by ad, which generally takes on possible meaning (cf. (75)), but also of the
Imperfective preceded by ad, which rather takes on iterative or habitual meaning (cf.
(76)). Given that the anchoring of the particle does not take place in the domain of
discourse referential, but rather in the referential of possible situations,40 one may
suppose that the deictic ad, which introduces a fictive reference, belongs to the type
am Phantasma (in the terminology of Bhler 1934:121140).4041
4.2.4 The two Tamazight futures

We saw above (4.2.1) that At Seghrushens Tamazight has, among the preverbal particles of the non-chained vps, a particle which Bentolila notes d-ad. This d-ad is a particle specialized in the expression of the future. It appears in certain conditions, in the
place of ad which has only future connotations (among other more general meanings
belonging to the domains of potentiality and possibility). d-ad thus removes ambiguities (compare (84ab) and adds certainty (compare (85ab)).41
At Seghrushens Tamazight
(84) a.

(after Bentolila 1981:173)
a=
i-awn
r.Bi
a=pr.do.2sg 3m-help.a God
(i) God will help you.
(ii) May God help you!
b. d-a=
i-awn rBi

God will help you.
(85) a.

ad
azn-x
lflus
ad
send.a.1sg money
I will send the money.
b. d-ad
aznx
lflus

I will send the money, be sure of it.
Given the meaning of the d preceding the modal particle (d adds stress, emphasis,
which could be rendered in English by really, truly, precisely, undoubtedly, etc.),
Bentolila sees in it the identification presentative morpheme d it is found in nominal
sentences such as d aryaz It is a man (1981:173, note 156).
If the element d has its origins in the predication particle d, one could object that
this implies that the preverbal pan-Berber particle ad has a different origin. However,
since identical phenomena have a tendency to take place repeatedly over time, in similar conditions, one could also suppose that the origin of the new future in Tamazight
. On the different types of referentials, see Descls and Guentchva 2006.

. Furthermore, d-ad obligatorily replaces ad after the interrogative ma, after is that, mr if ,
zGwis since, ani because and the coordinator la ... la.
simply confirms the explanation proposed for the old one. The first anchoring (with
an am Phantasma deictic) only having led to a modalized future, a second anchoring
(with an identification deictic) gave rise to a more assertive future. For the second
grammaticalization process a reflex of the first, but based on the realis referential
to have been able to take place, it was probably necessary that there first be an obliteration of the etymology shared by the preverbal particle ad and the predication particle
d through divergent morphologies (which appears to be a general tendency in Berber,
with just a few exceptions, such as Zenaga).42
4.3 ad, tam particle or subordinator?

After this study of ad as a future particle (and, more generally, a non-realis particle), let
us now see to what extent this analysis can be applied to the other cases of ad.
4.3.1 The ad of injunction and prohibition4243

Injunctive uses and the expression of wishes, especially in a positive form, are probably
the uses closest to the above studied case. This could be in relation to the fact that in
Zenaga a variety which has not grammaticalized ad as a future particle , these uses
are quite restricted (to express wishes, it is not the phrase with ad + Aorist which is
used, but rather the Perfective preceded by ay).
Be it as it may, all these uses of ad + Aorist, with or without the negation particle, are non-assertive. The modal interpretation, marked by specific intonation, of the
imperative or exclamative type, is compatible both with the definition of the Aorist as
a form with indeterminate meaning and with the etymological hypothesis of ad as an
am Phantasma deictic. Once again ad, by bringing determination to the underspecified Aorist verb form, renders it frontable.
One could consider that this ad anchors (situates) the vp in the realm of what is
possible, as would do an identifying presentative for a np in the referential domain of
discourse. However, given the resemblances of ad with French que, it could also be a
demonstrative in cataphoric relation with the following clause, i.e. a relative type ad.43
4.3.2 The pan-Berber ad and governed clauses
In the varieties other than Zenaga, ad only precedes governed verbs in the Aorist. Such
is the case, as presented in 3.3.1, for clauses with consecutive or final meaning and
. The Tamazight examples show that the preverbal particle has two variants (a in (84) and ad in
(85)), and that they are both different from the form d taken by the identification presentative.
. In French, the conjunction que is originally, in Low Latin, a pronoun stemming from
Latin relativizers. More generally, one notes that the Romance languages have [...] developed
a pronoun which globally announces the following clause, a cataphoric deictic (Ramat 1985),
based on an inferior correlative element (a WH-form) (Muller 1996:97).
for clauses governed by modal verbs (can, must) or verbs expressing a request, a wish,
or an order.
The term governing follows Vycichls analysis of ad as a subordinator in these
particular uses. Vycichl (1992) studies the origins of the particle ad zur Bildung des
Konjunktivs, des Optativs und des Futurum and claims one single origin for ad (as
a demonstrative pronoun, like German dass or English that). He considers that the
complementizer44 function could explain the other uses, both modal and temporal. He
does however note one important difference between German and Berber: the Berber
verb preceded by ad may be used independently, whereas the German dass generally
follows a verb.
Galand (1977:301) mentions an argument (also found in Vycichl) in favor of the
demonstrative origins for ad: the fact that, at times, the preverbal ad in Shilha takes
on the annexation form (wad, annexation form of ad) after prepositions, exactly like
the m.sg. demonstrative pronoun. He does not give ad the status of subordinator: the
use of ad + Aorist in these dependent clauses is comparable to that of the Aorist (with
or without ad) in juxtaposed clauses (see 3.2.1 and 4.2.1). Indeed, the latter are not
necessarily preceded by a pause (Galand 1988:224), so it seems justified to attribute
the logical dependence (with consecutive, finality and goal meanings), characterizing
the clause introduced by ad, to the choice of the Aorist and to parataxis. Ad could be the
tam particle found in independent clauses with the Aorist: in the language varieties
where ad has this use, the presence of the particle grants more autonomy than dependence to the Aorist (cf. 4.2).45
In this domain, Zenaga is similar to the other Berber varieties, but the Aorist
without ad is less frequent and, contrary to other varieties, using ad after modal verbs
expressing possibilities or wishes is impossible (see 3.2.3). However, as ad is never
used in Zenaga as a future particle in independent clauses, a conjunctive interpretation
seems more adapted to the specific case of this language variety. In fact, one could consider ad as a demonstrative in direct object function, which cataphorically announces
the subordinate clause, especially as this hypothesis is the only possible one for certain
specific uses in Zenaga.4445
4.3.3 ad as quotative: Its prototypical use?

When demonstratives have endophoric deictic uses, their grammaticalization as quotatives and, more generally, as complementizers, is frequent in the worlds languages. In
. The one which one finds e.g. after ira want, in the construction which gave rise, after grammaticalization of the modal verb, to the true Tachelhit future: iraad ign > rad ign he will sleep.
. So far, however, we have not found any studies specifying what exactly is added by ad in
these contexts.
a first stage, the deictic is a demonstrative pronoun belonging to the main sentence
but cataphorically referring to information contained in the following clause. In a subsequent stage, the deictic is reinterpreted as a subordinator and integrated into the
conjoined clause (Diessels 1999:123 sq.).
Variations between direct and indirect speech show that the use of ad as a quotative
marker is not yet systematic in Zenaga. There is no doubt however as to the fact that ad
regularly functions as a cataphora, which is the necessary condition for reanalyzing it
as a complementizer. See the chain of two verbs say in the following example:
(86) y-nn(h)
zzn wn
3m.sg-say.pfv=pr.io.3 say.imp someone ad
taK=t
g t-rb-n y-ttttr
what=to.pr.do.3f.sg in f-girl-pl 3m.sg-ask.ipfv
He said to him: tell (= ask) someone this: which is the girl he wants?
The second verb say, in the imperative, is followed by a demonstrative ad which refers,
by anticipation, to the interrogative clause. In the absence of such a chain, the presence
of this ad is entirely optional:
(87) y-nn(h)
taK ayaxty
3m.sg-say.pfv=pr.io.3 what reason
He says to him (= asks): what is the reason?
Only Zenaga has adopted ad as a complementizer of the verb say, both in narratives
(say that...) and in indirect speech (say of...), such that one may even find both subordinate clauses in the same sentence, without the second introducing verb (y6nnh in
the sense of ask) being repeated:
(88) y-nn(h)-n
ntt dti
3m.sg-say.pfv=pr.io.3-pl ad him ad=pr.do.3m.sg
wry-nig
kr =
ar..iy-n
neg=3m.sg-have.npfv thing ad=pr.io.3 sell.on.credit.a.3pl
He tells them he has nothing, [that he asks them] to let him have credit
This originally cataphoric use of ad after say can quite easily explain the other uses of
ad after governing verbs, where the particle shows numerous points in common with
complementizers in Indo-European languages. As for the semantic differences noted,
they have their origins in the choice of the introducing verb and, concurrently, in that
of the tam associated to the dependent verb.
If the verb is in the Aorist, the contents of the clause are not asserted by the speaker,
only considered (in the completive and consecutive clauses) or aimed for (in goal
subordinate clauses). After y6nnh say, the clause expresses the object of a request,
a wish, just as after other verbs expressing a wish, expectation or intention.
If the verb is in the Perfective or Imperfective, the contents of the clause are given
independent referential anchorage. Following y6nnh say, it is a case of indirect
speech.
If a verbal auxiliary, yu(u)g or yum
r, precedes the verb in the Perfective or Imperfective (3.3.2), there is some confrontation between the contents of the clause and
its meaning of adequation to the real world.46 The governing verbs belong more
or less to the same verbal sub-classes as in other languages: authorizing indirect
interrogation [they] concern knowledge of their object, or more precisely, the possibility to take as argument an object of knowledge (Muller 2001:165).
These constructions, which could all be based on a cataphoric ad, are very different,
at first glance, from the constructions found in the hypothetical systems. Nonetheless,
some constructions could be related to them.
3.4 Conditional clauses, topics and topicalization464748

For ad, a conditional marker in Zenaga, two grammaticalization paths can be considered, but in both cases, one must take into account the fact that the clause stating the
condition or hypothesis (If p) precedes the conclusion clause (q) and is thus regularly
placed in fronted position. This tendency, which was raised to the rank of rule for Classical Arabic by Peled (1992:139 sq.), is said to be a universal (Greenberg 1963:845,
Comrie 1986:83 sq.). This justifies the usual terms of protasis (If p) and apodosis
(q) which we have adopted for Zenaga where, as in Arabic, clause order appears to be
set, at least for those with the marker ad.47 Protasis-apodosis organisation particularly
when languages have partly or entirely rigid clause order led Haiman (1978) to consider that Conditionals are topics.48
3.4.1 ad as cataphora

The protasis p in conditional clauses could be the result of topicalization: the clause
introduced by ad which previously followed a governing verb, is fronted to constitute
. This succinct semantic interpretation of indirect interrogation (inspired by Muller 2001)

is compatible with the meanings of the auxiliaries yu(u)g and yum
r outside of this particular context.
. We have found a proverb where the clause introduced by the counterfactual marker m
is in second position, but the change in order goes along with a semantic change, where m
even if no longer introduces a conditional but rather a concessive clause (for comparable
phenomena in an Arabic dialect, see Vanhove 2002:771 sq.).
. He defines the topic in a plural manner: either (i) The topic is what the speaker is talking
about the comment is what he says about it, or (ii) The topic is the given or old information
in the sentence the comment is the new information (Haiman op. cit: 583).
the topic or the framework of the clause q. In this case, the origin of ad would then
be, as before, the cataphoric demonstrative ad (either followed or not by the verbal
auxiliary yu(u)g).
The fact that conditional and indirect interrogative sentences may have identical
subordinators (e.g. the English if, a frequent occurrence in languages) shows how
close these two types of complex sentences are (Muller 2001: 173). In Zenaga,
the presence of adyu(u)g at the head of factual hypotheses indicates a strong
link with indirect yes-no questions. In fact, the protasis, formally identical to an
indirect interrogative clause (with a verb in the perfective or imperfective forms
in both cases), can be considered as taking up one of the terms in the alternative.
While the protasis posits the topic as hypothetically in conformity with reality, the
apodosis furnishes the comment. The relation between p and q is rather varied
and may be e.g. of an implicative or explicative nature.
If ad is not followed by yu(u)g, it could be a case of topicalization (without any
particular formal marker) of the subordinate clause governed by a request verb
or the verb y6nnh say (after which all tams are permitted). The hypothesis of
the quotative nature of English if has been the subject of some debate. It has been
evidenced when a future tense marker exceptionally appears in the protasis, as in:
If itll definitely rain, (as X says), then Ill take my umbrella (Akatsuka 1986:340).
If the complement clause is topicalized (with its complementizer ad), the resumptive
form, once again, is in a non-assertive mood. The tams allowed in the protasis are in
fact those of the clauses governed by request, order or intention verbs (the Aorist and
sometimes the negative Perfective),49 not those of indirect speech found after the verb
say (Perfective and Imperfective, either positive or negative).4950
Contrary to the preceding case, the dependence of q as concerns p is formally
marked by the regular use of the Aorist in the apodosis. The relation between p and q,
much less varied, seems to always belong to the implicative type (in the wider sense
of the term). Topicalization seems to establish a tighter link between p and q: one of
initial and subsequent elements rather than one of topic and comment (thme/rhme
in French).50
. If the future auxiliary ynhy is used in the apodosis, then it is in the Aorist.
. Even though we do not take up the definition of topic as given or old information, we are
not far, in distinguishing two topicalizations, from the two topics proposed by Haiman (1978).
This distinction appears perhaps more clearly in Classical Arabic where, in conditionals, the
use of the connection particle fa characterizes the pragmatical type connection (structured as
topic comment) as opposed to the logical type (p implying q). See Larcher 2000:2012.
It is worth noting that the same logical dependency relation is established between
juxtaposed clauses, without there being any necessity to hypothesize topicalization.
We shall now turn to this hypothesis.
4.4.2 ad as anchor

A clause may serve as a framework when it precedes another clause comprising a verb
in the Aorist. It is either a sequential juxtaposition (3.2.1), or a main clause and its
subordinate clause (3.2.2).
The similarity between conditional and temporal clauses is frequent in the worlds
languages. It is true in Zenaga for certain time subordinate clauses which have the
same uses as protases introduced by ad. Thus the main clause in (42) and the apodosis
in (89) are strictly identical:
..
(42) og=r
y-f f u
y-bi
g mn
moment=in.which 3m.sg-be.thirsty.pfv 3m.sg-drink.a in water.pl
When he is thirsty, he drinks water.
..
(89) dy-uf f u
y-bi
g mn
ad=3m.sg-be.thirsty.a 3m.sg-drink.a in water.pl
If (= when) he is thirsty, he drinks water.
The difference between the two constructions lies in the way the framework is construed. In the time clause introduced by the conjunction og-r, with a verb in the
Perfective, here expressing a state, the event framework is instantiated by the present
with habitual meaning. In the protasis introduced by ad, with a verb in the Aorist,
the event is only considered. It could thus be a case (as in 4.2.3) where the Aorist is
anchored in the referentiality of possible worlds.
If the ad of the Zenaga conditionals is to be considered, originally, as an anchor,
it could even have functioned as a tam particle in the past, like the future ad in other
Berber varieties. In this case, the Zenaga conditionals would basically be made up of
two juxtaposed clauses, and the implicative relation between p and q would be marked
by word order, by the choice of the Aorist, and, lastly, by the obligatory presence of ad
in the protasis, a specific tam particle.
In his typology of conditionals, Comrie shows that in languages such as Turkish,
conditions may be expressed through just the verb form (1986:87). Therefore, Zenaga
would not be the only language to have adopted this strategy, even if it appears to be
an isolated case in Berber.
4.4.3 Conditional markers and topic markers

Informative hierarchy and topicalization play a very important role in expressing conditions in Zenaga. One could be tempted to consider ad as having become a topic
marker, as in some Austronesian languages (see Bril, in press). Despite some similarities in their workings, they are nonetheless two distinct cases. The uses of ad, in
Zenaga, are not, specifically, those of a topic marker, but the behavior of a demonstrative such as ad does indeed present some similarities with this type of marker.51 In her
inventory of lexical sources for conditional markers, Traugott classifies topic markers
and demonstratives in the same group, while distinguishing between the two.52
This proximity shows up very clearly in Classical Arabic, with the particles in and
inna, the latter being a particle with presentative meaning, serving to introduce topics
(definite or not), especially discourse initially.53

Classical Arabic
(90) inna raul=a=n
a
inna man=cas.acc=indf come.prf.3m.sg
There is a man (who) came.
However, the discourse particle inna is related to the conditional marker in which
is very similar to Zenaga ad. Indeed, in, also of deictic origin, may be followed by an
apocopized element (Apocopate), a subjunctive whose use, like that of the Berber
Aorist, is typically restricted to specific, usually dependent contexts. in can however
also be followed by the suffixal conjugation (Perfective), like i.54
Classical Arabic
(91) a. in t-ai=n

if 2-come.sbjv=pr.do.1sg
(after Ayoub 2003)

-ukrim=ka
1sg.honor.sbjv=pr.do.2m.sg
b. in i-ta=n
akram-tu=ka

if come.prf.2m.sg=pr.do.1sg honor.prf.1sg=pr.do.2m.sg

If you come to my house, I will honor you.51525354
The preceding examples are considered equivalent, but the change in tam is not completely neutral: the Perfective (faala) and the Apocopate (yafal) are two donation
moods of temporary admittance or, to take an expression used by Antoine Culioli,
of virtuality (du virtuel). In the Apocopate, in our opinion, there is modal meaning: a
. It would be interesting to study the role of d in the Kabyle phrase ma d + np as for ... (lit.
if cop...): formally it is the protasis of a conditional with a non verbal predicate, but serves,
apparently, to mark a change in topic (Example (12), Mettouchi 2006:122).
. Traugott (1985) notes five main lexical sources in the worlds languages: (i) modalities
(especially epistemic modalities), (ii) copula constructions, (iii) interrogatives (especially interrogative complements of verbs such as know), (iv) words which indicate that something is
already known or given (apparently mostly topic markers or demonstratives) and (v) temporal notions. The Zenaga ad could belong to (ii) and/or (iv).
. However, the topic is not focalized after inna, whereas the contrastive topic introduced
by amm as for is focalized (the operation is marked in particular by a prosodic break).
. i, the other Classical Arabic conditional marker, commutates with in in (91b), not
in (91a).
request, an implicit that P be on the part of the speaker: who asks the hearer to temporarily admit that P exists (Ayoub 2003:35).
This modal meaning attributed to the Apocopate in Arabic must also be attributed
to the Berber Aorist. The Aorist, which does not add any aspect-tense meaning in this
context, serves to orient the hearer to a non-instantiated interpretation of the process.
The reference introduced by ad being purely fictitious, one may, once again, consider
it an am Phantasma use of the deictic.
Zenaga, in contrast with Wolof,55 does not have a specific deictic at its disposal
for non instantiated conditional sentences, but the combined use of the Aorist and ad
allows a sort of conditioned assertion (see Ayoubs donn admis (allowed fact)) so
that the process is constructed in the world of possibles: One construes a fictitious reference point, making it possible to dissociate the discourse producer from the speaker
(or writer). The viewpoint is that of this fictitious reference point and may thus have
bearing both on the present and on the future (X is said to be in London at the moment
alongside it is said that X will be in London in a week. [...]. This makes it possible to say
something without taking responsibility for what one is saying. (Culioli, 1990:150).
The fact that, for Culioli, the notion of fictitious reference point applies to the
conditional and to the future, undoubtedly makes the use of ad more comprehensible
in both cases, especially if they are two distinct varieties.55
5. Conclusion
Even if it seems difficult to admit that ad (or a(d) or (a)d) has a single origin, so varied
are its uses in Berber, all of the particles could indeed have a common source in the
near masculine singular demonstrative.
Grammaticalization of demonstratives is a widespread phenomenon crosslinguistically (Diessels 1999) and many of the uses of the Berber ad are unsurprising (e.g. ad
as copula or predicative particle, relative pronoun, complementizer, quotative, consequence or goal subordinator, optative or injunctive particle).
Other uses, compatible with those just listed (especially the latter ones), may
appear more difficult to reconcile with each other. This is the case in particular of
the ad conditional marker in Zenaga and the ad future particle in most of the other
. In the ternary deictic system in this language where the suffix -i is endowed with identification value and the suffix -a with differentiation value , it is the element -u signifying
the absence of localization (in this case there is a gap between the space-time of the process
and the space-time of the utterance) which is used for hypotheses. This -u is also found in
the marker bu which corresponds to the hypothetical if and the temporal when situated in
the future (in the other temporals, when is rendered by bi or ba). See Robert 2006:167 sq.
ialects. As they are also rarer among cases of grammaticalization in languages in gend
eral, that was our main focus of attention.
It would seem that the role of ad, in its anomic uses, is partially similar to its role
in relative clauses. However, while reference, in the case of antecedent resumption, is of
the anaphoric type, it is of the cataphoric type in the Zenaga conditionals; the presence
of ad, and especially the presence of adyu(u)g, protasis initially, can be explained by
topicalization: of the completive introduced by the quotative ad on one hand, and of
indirect yes-no questions marked by adyu(u)g, on the other hand.
Formerly, the study of ad as a tam marker (in Berber varieties other than Zenaga)
showed the fundamental role played by the particle in the syntactic and referential
autonomy of the Aorist. We believe that this ad was a former predicative particle and
have suggested that its use as anchor in independent nominal sentences could explain
most of the uses of ad before the Aorist: its uses as future marker, injunctive, request
and prohibition marker, and even some of its uses in Zenaga conditionals.
The presence of ad in protasis initial position in Zenaga can thus be explained in
several ways (tam particle, as in prohibitive constructions, or subordinators of deictic
origins, as in indirect speech and indirect interrogatives). It probably corresponds to
an inflection of the general system due to the tendency, specific to Zenaga, to use ad
with the Perfective and Imperfective, and to use it as a conjunctive element. But on the
whole Zenaga behaves like the other Berber varieties when using the Aorist as a virtually instantiated form: followed by the verbal systems neutral form, ad still remains an
am Phantasma deictic. From this perspective, the originality of Zenaga lies in the fact
that it suffices, in this variety, to pose a fronted clause p and to add a commentary q to
construe a conditional sentence, the implication being entailed by the simple sequential juxtaposition.56
The ability for demonstratives to refer, in all possible ways and on all levels
including am Phantasma, to imaginary worlds largely explains the polygrammaticalization of ad. Thus one may say that ad is a good illustration of a fractal object (on
this notion, see Robert 2003).56
However, the ability for ad to play so many roles in Berber, including as an isolated element, is also due to the fact that Berber is a paratactic language, as has long
been stressed by Berber scholars (cf. Basset 1952:40, Galand 2002 [1987a]: 242). Thus
one must not be surprised that prosodic means play a large role in the language, even
though it has only been mentioned in passing in the present article. Without intonation, the structuration of conditionals between a fronted frame/topic clause and an
implicational clause or comment would not be perceptible. The use of ad would not be
. In the same volume, one may see, among other examples, the Modern South Arabic
deictic - and its multiple functions (Simeone-Senelle 2003).
sufficient to mark the protasis: confusion would be possible not only with injunctions,
but also with complement clauses (after a verb say for example).
References
Akatsuka, Noriko. 1986. Conditionals are discourse-bound. In On Conditionals, Elizabeth Closs
Traugot, Alice ter Meulen, Judith Snitzer Reilly & Charles A. Ferguson (eds), 333351.
Cambridge: CUP.
Ayoub, Georgine. 2003. Corrlation et rupture modales. Formes verbales et particules nonciatives dans les hypothtiques en arabe littraire. In Mlanges David Cohen, Jrome Lentin &
Antoine Lonnet (eds), 2945. Paris: Maisonneuve et Larose.
Banys, Wieslaw. 1996. Propositions conditionnelles: Coordination, subordination, connexion.
In Dpendance et intgration syntaxique. Subordination, coordination, connexion, Claude
Muller (ed.), 221226. Tbingen: Niemeyer.
Basset, Andr. 1952. La langue berbre. London: Dawsons.
Bentolila, Fernand. 1969. Les modalits dorientation du procs en berbre. La linguistique 1969
(1): 8596 & 1969 (2): 91111.
Bentolila, Fernand. 1981. Grammaire fonctionnelle dun parler berbre: At Seghrouchen dOum
Jeniba (Maroc). Paris: SELAF.
Berthoud, Anne-Claude. 1994. Indfinis et thmatisation. Faits de langues 4: 1618.
Bril, Isabelle. In press. Coordination, information hierarchy and subordination in some Austronesian languages. In Converbs, Medial Verbs, Clause Chaining and Related Issues Azeb
Amha, Christian J. Rapold, Sascha Vllmin & Sylvia Zaugg-Coretti (eds). Frankfurter
Afrikanistische Bltter.
Bhler, Karl. 1934. Sprachtheorie, Jena: Fischer.
Chaker, Salem. 1997. Quelques faits de grammaticalisation dans le systme verbal berbre. In
Grammaticalisation et reconstruction, 103121. Paris: Klincksieck.
Cohen, David. 1984. La phrase nominale et lvolution du systme verbal en smitique. tudes de
syntaxe historique. Paris: Socit de Linguistique de Paris.
Cohen, David & Taine-Cheikh, Catherine. 2000. propos du znaga. Vocalisme et morphologie
verbale en berbre. Bulletin de la Socit Linguistique de Paris 95(1): 269322.
Comrie, Bernard. 1986. Conditionals: A typology. In On Conditionals, Elizabeth Closs Traugott,
Alice ter Meulen, Judith Snitzer Reilly & Charles A. Ferguson (eds), 7799. Cambridge:
CUP.
Culioli, A. 1990. Pour une linguistique de lnonciation. Oprations et reprsentations 1. Paris:
Ophrys.
Descls, Jean-Pierre & Guentchva, Zlatka. 2006. Rfrentiels aspecto-temporels dans les textes.
tudes cognitives 7: 734.
Diessel, Holger. 1999. Demonstratives: Form, Function, and Grammaticalization [Typological
Studies in Languge 42]. Amsterdam: John Benjamins.
Franois, Alexandre. 2003. La smantique du prdicat en mwotlap (Vanuatu), Leuven: Peeters.
Galand, Lionel. 1957. Un cas particulier de phrase non verbale: lanticipation renforce et
linterrogation en berbre. In Mmorial Andr Basset (18951956), 2737. Paris: Adrien
Maisonneuve.
Galand, Lionel. 1977. Continuit et renouvellement dun systme verbal: Le cas du berbre.
Bulletin de la Socit Linguistique de Paris 72: 275303.
Galand, Lionel. 1988. Le berbre. In Les langues dans le monde ancien et moderne, III: Les langues chamito-smitiques, Jean Perrot & David Cohen, 207242. Paris: CNRS.
Galand, Lionel. 2002. tudes de linguistique berbre. Prsentation dun parler: Le touareg de
lAr [1974], 117143; Typologie des propositions relatives: La place du berbre [1984],
219240; Subordination rsultant de la relation: propos de la relative berbre [1987a],
241256; Les emplois de laoriste sans particule en berbre [1987b], 259271; Lnonc verbal en berbre. tude de fonctions [1964], 287308; Propositions relatives, rhmatisation
et thmatisation. Lexemple du berbre, 331355. Leuven: Peeters.
Galand, Lionel. 2003. Laoriste berbre, laspect et les valeurs modales. In Mlanges David Cohen,
Jrome Lentin & Antione Lonnet (eds), 23546. Paris: Maisonneuve & Larose.
Greenberg, Joseph H. 1963. Some universals of grammar with particular reference to the order
of meaningful elements. In Universals of Language, Joseph H. Greenberg (ed.), 73113.
Cambridge MA: The MIT Press.
Hagge, Claude. 1982. La structure des langues. Paris: PUF.
Haiman, John. 1978. Conditionals are topics. Language 54(3): 564589.
Kossmann, Maarten G. 1997. Grammaire du parler berbre de Figuig (Maroc oriental). Paris:
Peeters.
Kossmann, Maarten G. 2000. Esquisse grammaticale du rifain oriental. Paris: Peeters.
Kouloughli, Djamel E. 1994. Indfini et structure thmatique en arabe. Faits de langues 4:
16975.
Lafkioui, Mena. 2002. Lintonation et ses fonctions syntaxiques en rifain. In Articles de linguistique
berbre. Mmorial Werner Vycichl, Kamal Nat-Zerrad (ed.), 253281. Paris: LHarmattan.
Larcher, Pierre. 2000. Subordination vs coordination smantiques: lexemple des systmes
hypothtiques de larabe classique. Annales Islamologiques 34: 193207.
Mettouchi, Amina. 2002. La forme ad + aoriste en berbre (kabyle). In Articles de linguistique
berbre. Mmorial Werner Vycichl, Kamal Nat-Zerrad (ed.), 335347. Paris: LHarmattan.
Mettouchi, Amina. 2006. Sujet postverbal et tat dannexion en kabyle (berbre). Faits de Langues 27 [numro spcial Les langues chamito-smitiques (afro-asiatiques) Vol. 2]: 113129.
Muller, Claude. 1996. La conjonction que: Rection vs. dpendance immdiate et concurrence
avec que pronominal. In Dpendance et intgration syntaxique. Subordination, coordination, connexion, Claude Muller (ed.), 97111. Tbingen: Niemeyer.
Muller, Claude. 2001. Smantique de la subordination: linterrogation indirecte. In La smantique des relations, Andr Rousseau (ed.), 163177. Villeneuve dAscq: Universit Charlesde-Gaulle/Lille 3.
Nat-Zerrad, Kamal 2001. Grammaire moderne du kabyle. tajerrumt tatrart n teqbaylit. Paris:
Karthala.
Peled, Yishai. 1992. Conditional Structures in Arabic Clauses. Wiesbaden: Harrassowitz.
Prasse, Karl-G. 2008. Manuel de grammaire touargue (tahggart), VIIIIX, Syntaxe. Schwlper:
Cargo.
Ramat, Paolo. 1985. Typologie des langues. Paris: PUF.
Robert, Stphane. 1991. Approche nonciative du systme verbal (le cas du Wolof). Paris: CNRS.
Robert, Stphane. 1996. Aspect zro et dpendance situationnelle: lexemple du wolof. In Dpendance et intgration syntaxique. Subordination, coordination, connexion, Claude Muller
(ed.), 153161. Tbingen: Niemeyer.
Robert, Stphane. 2003. Polygrammaticalisation, grammaire fractale et proprits dchelle. In
Perspectives synchroniques sur la grammaticalisation. Polysmie, transcatgorialit et chelles
syntaxiques, Stphane Robert (ed.), 85120. Louvain: Peeters.
Robert, Stphane. 2006. Deictic space in Wolof. Discourse, syntax and the importance of
absence. Typological Studies in Language 66: 155174.
Simeone-Senelle, Marie-Claude. 2003. De quelques fonctions de - dans les langues sudarabiques modernes. In Perspectives synchroniques sur la grammaticalisation. Polysmie, transcatgorialit et chelles syntaxiques, Stphane Robert (ed.), 239252. Louvain: Peeters.
Tafi, Miloud. 1993. L expression de lhypothse en berbre. In la croise des tudes libycoberbres. Mlanges offerts Paulette Galand-Pernet et Lionel Galand, Jeannine Drouin &
Arlette Roth (eds), 21528. Paris: Geuthner.
Taine-Cheikh, Catherine. 1999a. Le znaga de Mauritanie la lumire du berbre commun. In
Afroasiatica Tergestina, Marcello Lamberti & Livia Tonelli (eds), 299324. Padova: Unipress.
Taine-Cheikh, Catherine. 1999b. Topicalisation, thmatisation et anaphore en arabe. In La thmatisation dans les langues, Claude Guimier (ed.), 247261. Bern: Peter Lang.
Taine-Cheikh, Catherine. 2003a. La corrlation de gmination consonantique en znaga (berbre de Mauritanie). Comptes rendus du Groupe Linguistique dtudes chamito-smitiques
34 (19982002): 566.
Taine-Cheikh, Catherine. 2003b. Ladjectif et la conjugaison suffixale en berbre znaga. In
Mlanges David Cohen, Jrome Lentin & Antoine Lonnet (eds), 661674. Paris: Maisonneuve & Larose.
Taine-Cheikh, Catherine. 2005a. Le rle des phnomnes dagglutination dans la morphogense
du chamito-smitique. Exemples de larabe et du berbre. In Linguistique typologique,
Gilbert Lazard & Claire Moyse-Faurie (eds), 288315. Villeneuve dAscq: Presses Universitaires du Septentrion.
Taine-Cheikh, Catherine. 2005b. Du rle de la quantit vocalique en morphognie. Rflexions
partir de larabe et du berbre de Mauritanie. Faits de Langues 26 [numro spcial Les
langues chamito-smitiques (afro-asiatiques) Vol. 1]: 4163.
Taine-Cheikh, Catherine. 2007. Les propositions relatives en znaga et la question des relateurs
en berbre. In XII Incontro Italiano di Linguistici Camito-semitica (Afroasiatica). Atti, M.
Moriggi (a cura di), 301310. Rubbettino: Medioevo Romanzo e Orientale.
Traugott, Elizabeth Closs. 1985. Conditional markers. In Iconicity in Syntax [Typological Studies
in Language 6], John Haiman (ed.), 289307. Amsterdam: John Benjamins.
Vanhove, Martine. 2002. Conditionnelles et concessives en Arabe de Yafi (Yemen). In Sprich
doch mit deinen Knechten Aramisch, wir verstehen es!. 60 Beitrge zur Semitistik. Festschrift fr Otto Jastrow zum 60. Geburtstag, Werner Arnold & Hartmut Bobzin (eds),
755777. Wiesbaden: Harrassowitz.
Vycichl, Werner. 1992. Der Ursprung der Partikel ad- zur bildung des Konjunktivs, des Optativs
und des Futurum im Berberischen. Frankfurter Afrikanistische Bltter 4: 7780.
Deixis and temporal subordinators

in Pomak (Slavic, Greece)*
Evangelia Adamou
LACITO (Laboratoire des Langues et Civilisations Tradition Orale),

Deixis expressed by three definite articles is a well-known characteristic of
the Pomak varieties, given the absence of grammaticalized definite markers in
most Slavic languages. In this paper, we present a practically unknown use of
deixis in Slavic languages, namely the use of deictic suffixes in the formation
of temporal subordinate conjunctions in Pomak. The deictic suffixes on the
temporal subordinator indicate the relation between the process and the
discourse situation; the choice of the deictic suffix depends on the type of event
encoded in the clause. The free temporal subordinator indicates that the event is
not anchored in speech time. This study is based on first-hand data of a nearly
undescribed variety of Pomak spoken in Greece.
1. Introduction
In a variety of Pomak spoken in the Xanthi prefecture of Greece, three suffixes specify
the location of an entity in relation to the speakers sphere (-s-), the addressees sphere
(-t-), or away from the speaker and the addressee (-n-). All three suffixes are used in
forming definite articles and demonstratives (see paragraph 3). Two of them, namely
-t- and -n-, are also used with temporal meaning when the space and time are different
from that of the discourse situation: -t- is no longer used for an entity in the addressees
sphere, but for an entity in a past time relative to the discourse situation while -n- is
used for all non-past time relative to the discourse situation and for absence of anchoring in the discourse situation.1
These uncommon uses allow us to understand the equally uncommon uses of the
two deictic suffixes that also form temporal subordinate conjunctions which anchor
*I wish to thank Isabelle Bril and Stphane Robert; their work and comments were fundamental for understanding the Pomak temporal subordinators. I also thank Zlatka Guentchva
and the two anonymous readers for their useful remarks.
1. Similar spatial-temporal uses are observed cross-linguistically, see for example the case of
Nlmwa in Bril (2002).
Evangelia Adamou
the event in the discourse situation (for the theoretical framework and terminology of
this analysis see Culioli 1971, 1978, 1990 and Robert 2006, as well as paragraph 4.2.1).
Agato/kugato when (past) is used for a past moment in relation to the discourse
situation. Agano/kugano when, whenever, indicates future moments in relation to
the discourse situation, as well as habitual events. Absence of these deictics indicates
absence of anchoring in the discourse situation; aga when (no anchoring) is thus used
in fictional narratives (see paragraph 4). Immediate anteriority is expressed by prefixing
li immediately, all the time to the subordinators presented above: li ga as soon as (no
anchoring), li kugato as soon as (past) and li kugano as soon as (future).
Table1. Semantics and uses of deictic suffixes in Pomak
Noun
-s-
-t-
-n-
(no deictic
suffix)
1. Here & now
1. speakers
sphere
x
1. addressees
sphere
2. past moment
in relation to
the discourse
situation
1. distal
indefinite
past moment
in relation to
the discourse
situation
2. Different space
& time
Subordination
2. future & habitual

in relation to the
discourse situation;
absence of anchoring
in the discourse
situation
future or habitual
absence of
in relation to the
anchoring in
discourse situation
the discourse
situation
(conditional)
A tripartite deictic system of definite markers is attested in other South Slavic

languages, such as Standard Macedonian, but the usage constraints are very different
from those of Pomak described in this paper (see paragraph 3.1.1). Standard Bulgarian
makes use of one definite article, based on the -t- suffix, while practically all the other
Slavic languages have no grammaticalized definite markers.
As far as the use of deixis in the temporal subordinators is concerned, Pomak
holds a special place within the Slavic languages. Among the closest related languages, Macedonian does not make any use of the three deictic suffixes in order to
form subordinating conjunctions. Bulgarian, on the other hand, uses the -t- suffix
(also used for the definite article) to form a number of subordinators, but those conjunctions do not enter a system of oppositions related to deixis. The data available for
the Rhodope dialects in Bulgaria show a use of the deictic suffixes in the subordinators similar to those described in this paper, but no study is available on their precise
meanings and uses.
Even though the grammaticalization of a definite article is considered an areal

feature of the Balkans, the extensive use of deixis observed in Pomak temporals cannot
be explained through language contact, despite the Pomak speakers trilingualism with
Greek and Turkish.
2. The sociolinguistics of Pomak

Pomatsko Pomak is the name used for the South Slavic variety spoken by Muslim
inhabitants of the Rhodope Mountains in Greece (cf. Map 1) who often migrated to other
cities or countries during the second half of the 20th century. This study2 focuses on a
Pomak variety spoken in a village in the Xanthi area, which I will refer to as Pomak1.3
The language in the village under study is still transmitted to children, contrary to
other traditional Pomak speaking villages in Greece, where a shift to Turkish has become
generalized. The majority of the speakers in this village are trilingual. The younger generation, men and women, have learnt Greek and Turkish at school, within the Minoritariste primary school educational system provided to the Muslim Minority since the
Lausanne Treaty (1923). Elder women are sometimes monolingual or, most frequently,
have basic communicational skills in Greek and Turkish (see Adamou & Drettas 2008).
Pomak is often described as a conservative South Slavic group since it has for
example preserved a case system, in contrast with the most closely related South Slavic
languages that developed an analytical system for those functions. This feature is
important since loss of the case system is one of the features distinguishing Bulgarian
and Macedonian from Serbian for example.
Naturally, this approach can be questioned given that on other points the system
can be described as an innovative one, as for example in the case of the overt expression
of deixis and its use for the temporal conjunctions described in this paper. Moreover,
Pomak also shares some Balkan Sprachbund properties, such as will future, subjunctive, dative/genitive merger (observed in the Pomak1 variety), postposed articles, and
2. Fieldwork was conducted in 2005, 2006, 2007 and 2009 by the author with the financial
support of the laboratory Lacito, CNRS, France.
3. Given the political pressure on Pomak speakers, within a context of linguistic shift to
Turkish, I decided not to mention the villages names, despite the obvious interest this would
present from a dialectological perspective. Pomak1 stands for the village of the Xanthi area
and Pomak2 for the village of the Evros area. I sincerely thank the speakers who agreed to
participate in this study.
Evangelia Adamou
to some extent evidentiality (see Adamou 2008; for evidentiality as a Balkan feature
see Friedman 2004).
The Pomak spoken in Greece has a special interest for the study of South Slavic
because it had little contact with modern Bulgarian, standard or dialectal, during the
greater part of the 20th century (except for the few years of Occupation during the
Second World War when education in Bulgarian was obligatory), and therefore did
not undergo any changes due to standardisation practices, as did the varieties spoken
in Bulgaria (Kanevska-Nikolova 2001).
Pomak varieties in Greece are still practically undescribed because, within a context of shift to Turkish, the Pomak language is subject to political and ideological conflicts that make fieldwork research extremely difficult to conduct. However, educated
Pomak speakers, most often in collaboration with Greek authors, have participated in
the publication of dictionaries, grammars, teaching methods for foreigners and other
materials such as folktales and songs (see Theoharidis 1995, 1996a, 1996b, 1996c; Rogo
2002; Kokkas 2004a, 2004b).
3. Deixis and noun modifiers

The Pomak variety under study has a three-term person-oriented deictic system, the
deictic centre being not only the speaker but the addressee as well (see Anderson &
Keenan 1985). Innovative temporal uses also occur for the definite articles and are
described in 3.1.2.
Table2. Deictic suffixes in Pomak
-s-t-n-
speakers sphere; here and now

addressees sphere; past moment related to the discourse situation
distal (not located in the participants spheres); future, present moment related
to the discourse situation; no anchoring in the discourse situation
These three suffixes form the definite articles and the demonstratives presented
below (see 3.1.2 and 3.2.2).
3.1 Definite articles

3.1.1 Definite articles in the South Slavic languages
Definite articles are an exception within the Slavic languages: they have only been grammaticalized in some South Slavic languages (namely Bulgarian, Macedonian and all the
non-standardized varieties spoken in Greece), while their grammaticalization in some
dialects of North Russia is controversial (see Breu 1994; Kasatkina 2008). From an areal
perspective, postposed articles are a well-known Balkan Sprachbund property shared

by Romanian, Aromanian and Meglenoromanian, as well as Albanian even though
their development is probably equally influenced by each languages history as well as by
their mutual reinforcement through multilingualism (Assenova 2002).
It is generally admitted that definite articles are grammaticalized from demonstratives (Lyons 1999). In the case of South Slavic, written sources show that, by the
13th century, the postposed Old Church Slavonic4 demonstratives were grammaticalized into clitic demonstratives, and then into clitic articles: see Glbov (1950), Svane
(1961, 1962), Mirev (1964) or more recently Dimitrova-Vulchanova & Vulchanov
(in press); for a discussion of the status in the modern languages see Elson (1976); for
a detailed diachronic model of definiteness in Bulgarian see Mladenova (2007).
The threefold definite article is traditionally considered a conservative feature in
the literature (Mileti, Koneski, Cyxun), following the analysis that the whole of South
Slavic went through a phase of grammaticalization of three definite articles, followed
by some varieties developing a single definite article, while others retained the three
definite articles. Divergent analyses exist, arguing convincingly that the three-term
definites are innovations in specific varieties and did not develop in all the South Slavic
languages (Mladenova 2007:319325 also citing Veleva-Bojadieva and aur).
From a synchronic perspective, the use of a one-term or three-term definite article
requires a different dialectological analysis than the traditional dialectal classifications
of South Slavic in terms of Eastern-Western dialects following the phonetic distribution of the jat. Among the modern South Slavic varieties that have a grammaticalized
definite article, the ternary deictic definite system is found in three geographic areas:
this is the case for some Macedonian dialects and is the norm for Standard Macedonian (see Table3); it is also the case for most of the Rhodope Mountains varieties (in
Bulgaria and Greece), as well as for some varieties spoken in Eastern Serbia (close to
the Bulgarian border).
All the other South Slavic varieties that have developed a definite article use a
one-term system, usually based on the -t- form: this is the case for Standard Bulgarian and Bulgarian dialects,5 Eastern Macedonian dialects and the South-West
4. Old Church Slavonic [OCS] refers to the eldest available Slavic documents, mostly ecclesiastical. OCS extends from the 9th to the 11th century and is classified as South Slavic.
5. The distinction between Bulgarian and Macedonian dialects depends on the authors
and their ideological background. Bulgarian linguists have long considered the whole of the
South Slavic varieties as Bulgarian dialects, despite variation and important structural differences; the demarcation line being with Serbian. Since the standardization of Macedonian,
and reinforced by the creation of an independent Republic, the term Macedonian has slowly
made its way to into international bibliographies. Within this context, where linguistics are
linked with politics, we choose to apply the term Bulgarian dialects for the varieties situ-
Evangelia Adamou
Slavic varieties spoken in Greece (Drettas 1990; Adamou 2006). Some varieties are
said to be in an unstable state, evolving from the ternary to the single term system (Kanevska-Nikolova 2006, and fieldwork notes Adamou 20052006 for Evros
Pomak2).6 The existence of two-fold definite systems is controversial (for a critical
analysis see Mladenov 1990).
Furthermore, it is important to note that even among the ternary systems, the
parameters and morphological distribution can differ. It would be interesting to know
whether some differences observed are due to different analyses or to true differences
in the systems. For example, Kanevska-Nikolova (2006:79) describes the Rhodope
definite system as what could be called, after Anderson & Keenans (1985) termino
logy, a distance oriented system (the speaker being its centre, in pragmatic or spatial
terms), while in Mladenova (2007:318) the Rhodope system is clearly described as
person-oriented. The most recent descriptions of Macedonian also centre its system
on the speaker (Minova-Gurkova 1997; Friedman 2002; and for a different analysis
Topolinjska 2006):
Table3. Deictic suffixes in Standard Macedonian definite article
-f/v-t-n-
close to the speaker

unmarked (no indication as to distance)
away from the speaker
The same split is found in the analyses of OCS demonstratives: for example,
in Vaillant (1964: 140) the system appears to be distance-oriented while in Feuillet
(1999:148) it is described as person-oriented. According to Marchello-Nizia (2006)
this sort of distinction could indicate different states of the language and is thus important to specify.
ated in Bulgaria, Macedonian dialects for those in FYROMacedonia - even though for Macedonian linguists, the Macedonian dialects go beyond these borders (into Albania, Bulgaria,
Greece). For the denomination of Slavic in Greece see Adamou & Drettas 2008.
Moreover, the terms language and dialect are in this context highly connected to political
and ideological factors. Therefore, we use varieties as a linguistic term for oral tradition
languages seen in their diversity. Nevertheless, it is obvious that from a linguistic point of view
those South Slavic standard languages and non-standard varieties are closely related and are
part of the same linguistic diasystem.
6. In Greece, in a village of the Evros area, the system has been reduced to the -t- definite,
while possessive uses of the -s- article occur occasionally. I consider this change to be a loss;
the speakers in their metalinguistic comments mention three articles, which are not found in
their oral productions.
3.1.2 Definites in Pomak1

The definite article7 is suffixed to the noun, to the adjective or to the tonic possessive
pronoun and the possessive adjective. Three deictic suffixes are used in Pomak1 to form
definite articles. The choice of the deictic suffix depends on two sets of parameters:
1. For here and now situations the system depends on whether the entity is considered as being part of the speakers sphere, the addressees sphere, or neither one.
In Anderson & Keenan (1985), this type of system is dubbed person-oriented.
Deictic systems are generally described in terms of close vs. distant: close to the
speaker, close to the addressee, away from the speaker and the addressee. Still, it is
broadly admitted that pragmatic parameters generally rule such systems (Levinson
2004). In order to account for their usage, the term personal spheres is preferred here
(Bally 1926; Marchello-Nizia 2006).
In Pomak1, even though in most cases the sphere could be considered spatial,
other examples show that the system is determined by pragmatic and discursive criteria as well. The objective distance can thus be manipulated following pragmatic criteria, but such uses will not be explored in detail here.

-s-: speakers sphere
(1)

e
mi
pane kinito-so!8
mod 1sg.dat fall.3sg mobile-def.s
My mobile phone will fall!
(The speaker has the mobile phone in his pocket)
(S, M1, 8, Pmk, Ell, Tur)9
This definite is also used with strong possessive meaning.

-t-: addressees sphere
(2)

buluz-at
ti
je
hubaf
t.shirt-def.a 2sg.dat is.3sg nice
Your t-shirt (your sphere) is nice.
(The addressee is wearing the t-shirt)
(S, F5, 15, Pmk, Ell, Tur)
7. Indefinite nouns have no articles; they may be preceded by the numeral one which is not
a fully grammaticalized indefinite article.
8. There are many loan words from both Greek and Turkish in Pomak1. I do not specify
their origin, the Pomak morphology and syntax being respected in most cases.
9. Each example is followed by a description of the context. I systematically specify if the
example is spontaneous (S), the data recorded or not, or extracted by questionnaire (Q). The
questionnaire consists in translating from Greek a sentence that was usually produced in
Pomak by a speaker, in natural speech, but that I didnt have the chance to record. The speakers gender (male M, female F) is followed by an identification number valid for this paper.
Also noted are the age and the languages spoken by the consultant: Pmk stands for Pomak,
Ell for Greek and Tur for Turkish. All data were collected by the author between 20052007.
Evangelia Adamou
-n-: distal (not located in the participants spheres)
The suffix -n- is used for objects that are not included in the participants spheres.
(3)

jela
nah kapuje-ne
come.imp.2sg to doors-def.dist
Come to the entrance door (elsewhere).
(The speaker is out of the house, the addressee is sitting at the balcony, and
the entrance door is on the side of the house, away from both of them and not
visible to either)
(S, M3, 13, Pmk, Ell, Tur)
One should also note that in the variety described here, the deictic system does not
seem obsolescent among younger speakers, contrary to other Pomak varieties which
tend to reduce it.
2. When the entities are situated elsewhere in space and time, the speaker adopts a
temporal set of uses for the definite articles. In such cases, only the -t- and -n- articles
are concerned, the -s- leading back to the spatial-temporal reading here and now.
Therefore, the addressees -t- article no longer concerns the addressees sphere
but the past, while the -n- distal article is used for entities in the future, habitual in
relation to the discourse situation as well as for situations with no anchoring in the
discourse situation (all non past and non here and now situations):
-n-: future moment in relation to the moment of utterance
(4) na sfadba-na
e
nadena-m terven-en fustan
at wedding-def.dist mod wear-1sg red-def.dist dress
At the wedding Ill wear the red dress.
(Q, F2, 30, Pmk, Ell, Tur)

-t-:
past moment in relation to the moment of utterance
(5) na sfadba-ta
beh
sas terven-et
fustan
at wedding-def.loc.pas was.1sg with red-def.loc.pas dress
At the wedding I wore the red dress.

-s-: speakers sphere
(6) na sfadba-sa
sam
/ ima mnogo insan
at wedding-def.s am.1sg / is.3sg a.lot
people
Im at the wedding. There are a lot of people.
3.2 Demonstratives
3.2.1 Demonstratives in the South Slavic languages
In data available for South Slavic, demonstratives and definite markers do not seem
to follow the same evolution within a variety and certainly not at the same speed. For
example, even though the demonstrative system in Pomak1 is a three-term system,
used in the same way as the definite system, in other closely related Pomak varieties
(Kokkas 2004a: 22), demonstratives form a twofold system based on a close-distant

spatial contrast,10 while the definite articles form a three-term person-oriented system.
This difference in speed is also observed in Nashta (Adamou 2006), where a single
definite article is in use, while demonstratives present traces of a former ternary system distal/proximal/unmarked. In the same way, Standard Bulgarian has a two-term
demonstrative system, the -t- suffix being used either as proximal (when contrasted)
or as unmarked as far as distance is concerned, while the -n- suffix indicates distance.
On the other hand, Standard Macedonian presents the same tripartite system in both
the demonstratives and the definite articles.
Mladenova (2007:318) also signals this lack of symmetry in deixis between articles and the pronominal and adverbial systems in her account of dialects (the so called
mixed systems).
3.2.2 Demonstratives in Pomak1

Demonstratives in Pomak1 also follow the double deictic system determined by space
and time. They have both exophoric and endophoric uses. Demonstratives forms differ with the nouns gender and number:
aisvak (m.), aisos (f.), aisva (n.) and aises (m/f.pl.) aiseva (n.pl) for close entities,
ainvak (m.), ainos (f.), ainva (n.) and aines (m/f.pl.) aineva (n.pl) for distant entities
and for some future, habitual time related to the discourse situation, or for a moment
with no anchoring in the discourse situation; and aitvak (m.), aitos (f.), aitva (n.) and
aites (m/f.pl.) aiteva (n.pl) for entities close to the addressee or for a past moment
related to the discourse situation.
Here are some examples:

speakers sphere
(7) aisos je
dd-va-sa
kota11
dem.s is.3sg grandfather-poss-def.s house
This (my sphere) is grandfathers house.

addressees sphere
(8) tsalo aitva e

jade-
all
dem.a mod eat-2sg
Youll eat all of this (your sphere).
10. iz (m.), isy (f.), iz, isy (n.) and isy (pl.) for close entities, inaz (m.), iny (f.), inaz, iny
(n.) and iny (pl.) for distant entities.
11. Note that the vowels /o/, /e/, /a/ when stressed are diphthongized [uo], [je], [ja], and,
when not stressed, are reduced.
Evangelia Adamou
distal
(9) ainos
lelka hi
zla
je
be bin
evro
dem.dist auntie poss.3sg.f took.prf aux.3sg five.thousand euro
That auntie of hers, she has got5 000 euros.
past moment in relation to the discourse situation
(10) dve gudini napre mandili-te
bea
terveni
two years forward scarves-def.loc.pas were.3sg red
oti
bee
aitos
moda-ta
because was.3sg that.loc.pas fashion-def.loc.pas
Two years ago, the scarves were red because that was the fashion.

3.3 Deictic comparison markers, quantifiers and relatives in Pomak1

The three deictic suffixes -s-, -t-, -n-, are also used to form comparison markers (kakso/
kakto/ kakno as) and quantifiers (aisolkos /aitolkos/ ainolkos this much, that
much) in a quite productive manner. Note that in other villages of the area, systems
differ: for instance kakno as, is also used for simultaneous duration (sharing the
two uses with Greek), while it is not attested in my corpus of Pomak1 and aga is given
in questionnaires. Deictic suffixes can also partially be encountered for manner (aisi
this way/aini that way).
Relative pronouns also follow the deictic system determined by space and time:
addressees sphere
(11) gjuluti-te -te
nosi-
paratiki sa
glasses-def.a which-def.a wear-2sg ugly
are
The glasses that youre wearing are ugly.
speakers sphere
(12) gjuluti-se -se
nose-m aisa jatse sa hubavi
glasses-def.s which-def.s wear-1sg now very are nice
The glasses that Im wearing now are very nice.
past moment in relation to the moment of utterance
(13) gjuluti-te
-te
noseh lani
glasses-def.loc.pas which-def.loc.pas wore.1sg last.year
beha
guljami

were.3pl big
The glasses that I wore last year were big.
future moment in relation to the moment of utterance

(14) gjuluti-ne
-ne
e kupe-m sa terveni
glasses-def.dist which-def.dist will buy.1sg are red
The glasses that I will buy are red.
4. Subordinate clauses
4.1 Subordinate clauses in the South Slavic languages
According to Cyxun (1981) and Mladenova (2007), the use of the particle to in relative
pronouns and conjunctions is an innovation, characteristic of the Eastern South Slavic
area: the Rhodopean dialects are the centre of innovation in regards to the introduction of overt definiteness (Mladenova 2007:243). In this perspective, the uses of
deictics as described in this paper should also be viewed as innovative for South Slavic.
Indeed, conjunctions with -to, deriving from the demonstrative, are found
relatively late in the written sources. Analyzing the Damaskin12 texts, Mladenova
(2007:242) signals a split between uses with or without to for relative pronouns and
conjunctions; the relative deto being the most frequent. Concerning the temporal conjunctions, she notes a 96.88% use of koga and with just one use of kogato when; the
latter being currently used in Standard Bulgarian. In Standard Macedonian, despite
the use of a ternary definite and demonstrative system, there is no use of deictic suffixes in the temporals such as the one observed in Pomak1; speakers of the standard
language use the temporal conjunction koga when.
4.2 Subordinate clauses in Pomak1

Together with Slavic subordinators, many loans from Greek and Turkish are used in
Pomak1. Subordinating conjunctions are indeed high in the borrowability hierarchy
in language contact settings (Matras 1998, 2007). In Pomak, some result from early
contact, and are well integrated in the language (e.g. Greek oti because), while some
result from recent contact (such as Turkish tam just as), and are thus unstable and in
competition with Slavic markers.
We will now focus on temporal subordinators based on these deictic suffixes.
12. Translated literary texts of a religious and didactic nature, in what is referred to as
Modern and, for some of them, Middle Bulgarian (17th18th century).
Evangelia Adamou
4.2.1 Deixis in temporal subordinate clauses

In Pomak1 temporal subordinators combine reference to the discourse situation and reference to the situation of the process through the use of the deictic suffixes as follows:13
1. Presence of a deictic suffix indicates the anchoring of the utterance in the discourse
situation: aga-to, gato or kugato when (past), aga-no gano or kugano14 when
(future), whenever. Absence of deictics indicates absence of anchoring in the discourse situation (aga). Thus, the temporal marker with no deictic suffix is used for
folk tales, jokes and other narratives, whether fictional or not. Aga seems to be taking over some conditional uses, too, while the form ga is specialized for causal usage.
2. The choice of the deictic suffix indicates the relation between the process situation
and the discourse situation.
The addressees -t- suffix, in aga-to, is used for a past event in relation to the discourse situation. The distal -n- suffix, in aga-no, refers to generic or habitual timeframes and future reference.
Table4. The temporal subordinating conjunctions in Pomak1
aga
agato/kugato
when
if
when
agano/kugano
when
whenever
no anchoring to the discourse situation (24, 25, 26)

(conditional) (34)
past moment in relation to the discourse situation
(15, 16, 17)
future moment in relation to the discourse situation
(21, 22, 23)
or habitual in relation to the discourse situation
(18, 19, 20)
The interclausal relation is determined both by the subordinators and the TMA
markers (on the importance of the complex TMA system in adverbial clauses in Bulgarian,
see among others Fielder 1985; Guentchva 1995).
13. This is not a unified analysis for all the Rhodope varieties using deixis in the temporal conjunctions. Observations based on short term fieldwork in a Pomak variety in
the Evros department in Greece, show a different system whose understanding requires
further research (fieldwork notes Adamou 20052006). Moreover, a quick look at data from
Siroka Laka (Bulgaria) in Sobolev (2001), doesnt seem to corroborate the uses observed in
Pomak1, even though deictic suffixes are very frequently used with temporal subordinators.
Comparison would be interesting with, among others, Kabasanovs description of Tixomir
(1963), a Rhodope variety in Bulgaria, spoken by Muslims very close to the Greek border.
14. Variation between kuga-no/to (based on the interrogative kuga) and aga-no/to (based on
the temporal subordinator aga) depends on the speakers age and possibly education; older
speakers, or younger ones with little education, seem to prefer aga-, although I have not conducted any precise sociolinguistic study on this topic. In any case, both forms are accepted by
the speakers when asked.
Note that the three temporal markers presented above are not specialized in a specific type of temporal interclausal relation: for example agato is used for anteriority,
posteriority, simultaneity, terminus a quo, while all three markers can express anteriority (for this terminology see Kortmann 1997).
To describe the temporal subordinator system of Pomak1, I will use the terminology developed by Culioli (1971, 1978, and 1990) on the notion of situational anchoring, relating the discourse situation (Sit0) to the process situation (Sit2), and as applied
by Robert (2006) to the study of deixis and subordination in Wolof. Culioli defines
three types of relations between Sit0 and Sit2, often, but not uniquely expressed by
TMA markers: the process situation may be different () from the discourse situation,
towards being identical (=) to it, or have no relation to it ().
Pomak1 temporal conjunctions express these relations as follows:
Discourse situation and process situation are different: Sit2 Sit0

When the space-time frame of the denoted situation (Sit2) is different from the spacetime of the utterance (Sit0), Pomak1 uses the deictic suffixes -n- and -t- to form the
temporal subordinators. The -t- suffix, close to the addressee, is used for the past in
relation to the discourse situation, while the distal -n- suffix is used for habitual and
future events in relation to the discourse situation.

-to: past moment related to the moment of utterance
(15)

aga-to
fprh
kota-ta
be
izgorjala
when-loc.pas arrived.1sg house-def.loc.pas aux.3sg burned.down.p.prf
When I arrived, the house had burned down.
(On a past event)
(16)

kuga-to
be
dulala
niki jeda merenda pa>uto
when-loc.pas aux.3sg came.p.prf Niki ate.3sg Merenda ice.cream
When Niki came, she ate a Merenda ice-cream.
(On a past event)
(17)

aga-to
bee
lani
aitus imae kropa kosa
when-loc.pas were.2sg last.year here had.2sg short hair
When you were here last year, you had short hair.
(On a past event)
-no: habitual, generic, future moment related to the time of utterance

(18) aga-no
stanva
sfadba
when-loc.hab.fut happen.3sg marriage
zbira
sa
tsalo selo-so
gather.3sg refl all
village-def.s
Whenever there is a marriage, the whole (of my) village gathers.
(On local traditions)
Evangelia Adamou
(19)

aga-no
umre bunno ni prave-t volta
when-loc.hab.fut die.3sg someone neg make-3pl walk
Whenever someone dies, they dont take a walk.15
(20) aga-no
dojde
njakuf politikos
when-loc.hab.fut come.3sg some politician
fe
dava
njakvo er>o
always give.3sg some public.work
Whenever a politician comes, he always offers some public work.
(Conversation about politicians and their politics about the villages)

Below are some examples of the use of aga-no/ kuga-no for future events:
(21)

kuga-no
dojde
e
[h]i
gu
onesa-m
when-loc.hab.fut come.3sg mod 3sg.f.dat 3sg.m/n.acc bring-1sg
When she comes, Ill bring it to her. (The speaker has a gift to give to a friend
who is not at the village at the moment)
(22) aga-no
mi
punarasto-t mufka deti-se
when-loc.hab.fut 1sg.dat grow.up-3pl a.little children-def.s
e
dojde-me nah parisi
mod come-1pl to Paris
When the (my) children grow up a little, well come to Paris.
(Future plans)
(23) aga-no
fprje-s
na platia-na
when-loc.hab.fut arrive-2sg at square-def.dist
fati
nah ljavana strana
seize.imp.2sg to left
side
When you arrive at the central square, take the left side.
(Instructions in order to locate a place in the village) (Q, F8, 21, Pmk, Ell, Tur)
The choice of the temporal subordinator in Pomak1 does not depend on the TMA
choice, even though some TMA are more frequent with a given subordinator, since
they do relate to past or future events. The following examples illustrate this: in 24 and
15. This term refers to a special social event that takes place in the afternoon, when unmarried girls walk in the main village street in front of the boys in order to make a potential
match. This walk takes place in two villages of the area and gathers young people from the
surrounding villages.
25 the speaker gives two possible versions, changing the TMA markers while maintaining the temporal subordinator agano.
(24) aga-no
ima
sfadba zbira
sa
selo
when-loc.hab.fut have.3sg marriage gather.3sg refl village
Whenever there is a marriage, the whole of the village gathers.
(25) aga-no
e
imalo sfadba
when-loc.hab.fut aux.3sg have.prf marriage
zbiralo
sa
e
selo
gather.prf refl aux.3sg village
Whenever there was a marriage, the whole of the village gathered.
Also see Examples15 and 16 illustrating the fact that the subordinators choice is not
determined by the TMA.
No anchoring of the situation of process in the discourse situation: Sit2 Sit0

When there is a break between the space-time of the process and the space-time of the
utterance, Pomak1 does not use deictic suffixes, but the free subordinator aga (kuga
is not attested in here). In this case, the process is not anchored in relation to the utterance. This subordinator is almost exclusively used in fictional narratives and tales (the
habitual agano subordinator is very seldom encountered):
(26) aga sa
vornal-i
dve gudini sjetne ud alamanie
when refl return.evid-3pl two years after from Germany
utilil-i
da plat6 -t jajtsa-na
go.evid-3pl to pay-3pl eggs-def.dist
When they returned, two years later from Germany, they went to pay the eggs.
(Nasradin joke)
(27)

aga utilil-i
po
udvode sretal-i
majka mu
when go.evid-3pl more out
meet.evid-3pl mother 3sg.m.dat
When they went a little further, they met his mother.
(Nasradin joke)
(28) aga je
fprjala
je
razvorzala tuval-an
when aux.3sg arrived.prf aux.3sg untied.prf bag-def.dist
When she arrived, she untied the bag. (Folk tale)
(S, F3, 7, Pmk, Ell)
The constant use of aga in tales is also confirmed by other corpuses of the area, such as
the tales published in Theoharidis (1995) and Kokkas (2004b).
Evangelia Adamou
Identification between the two situations: Sit2 = Sit0

When the space-time of the denoted situation is identified with the space-time of the
utterance, Pomak1 could, in theory, use the proximal to the speaker deictic suffix -s-,
but in practice no such usages were attested during my fieldwork. When asked, the
Pomak1 speakers say they cannot use agaso. An educated speaker from another village
of the area thought this was a logical possibility, making a clear parallel with the threefold
definite system, but still had difficulties in producing any such examples. However, in
the Rhodope dialects in Bulgaria, gasu seems to be still in use (p.c. Kanevska-Nikolova).
4.2.2 Deixis and immediate anteriority
Immediate anteriority is also marked differently according to the type of discursive
anchoring. Immediate anteriority is expressed by prefixing li to aga (becoming ga) for
situations with no anchoring in speech time, to kugano for a future event related to
the discourse situation or for some habitual event, and to kugato for past events.
Table5. Immediate anteriority subordinating conjunctions in Pomak1
li ga
as soon as
li kugato
as soon as
li kugano
as soon as
immediate anteriority in a moment

with no anchoring in the discourse situation
immediate anteriority in a past moment
in relation to the discourse situation
immediate anteriority in future moment
or habitual event with relation to the discourse situation
As an adverb li in Pomak1 means:

1. immediately, e.g. li ripnala she hopped immediately;
2. all the time, constantly, e.g. li ruka she was calling all the time.
Li is also a yes-no interrogative particle, as was already the case in Old Church Slavonic,
and as is commonly found in many other Slavic languages (there have been many studies on the meaning and placement of li in Slavic languages). Example in Pomak1:
hubeve li ste? Are you OK?
jala li si? Have you eaten?
In Bulgarian, li is used as a subordinator with a meaning close to the conditional

(Guentchva 1995).
Below some examples of immediate anteriority in Pomak1:
li kugato immediate anteriority-past
(29) li-kuga-to
gu
rekah-me stana
immediately-when-loc.pas 3sg.m/n.acc said-1pl became.3sg
As soon as we said it, it happened.
li kugano immediate anteriority-future/habitual

(30) li-kuga-no
sa
razbudi-
z[mi]
immediately-when-loc.hab.fut refl wake.up-2sg take.imp.2sg
ma
telefon
1sg.acc telephone
As soon as you wake up, call me.
(31) li-kuga-no
sa
je
radalo
immediately-when-loc.hab.fut refl aux.3sg was.born.prf
dete-no
davali
mu
sa
iker
child-def.dist were.giving.prf 3sg.m/n.dat aux.3pl sugar
As soon as the child was born, they were giving him sugar.

tam immediate anteriority, past moment related to the moment of utterance
t am (from Turkish) just as, as soon as16, is external to the native system relative to the
time of utterance. Thus, language contact, first through loans, but then finally through
conceptual pressure within a stable contact situation, might have triggered a change in
the Pomak system of temporal conjunctions.
(32) tam fprh
faf iskete z
ma
telefon
just.as arrived.1sg in Isketshe took.3sg 1sg.acc telephone
Just as I arrived in Isketshe (Gr. Xanthi), s-he called me.

5. Polyfunctionality
Some of the adverbial subordinators presented above are polyfunctional.
5.1 Conditionals
aga: from temporal to hypothetical
It is very common cross-linguistically for temporal subordinators to express conditional or hypothetical meanings (Traugott et al. 1986). In Pomak1, aga, also has hypothetical meaning, even though the most frequent subordinator is aku.
16. tamam at the same time and li tamam as soon as, are known to be used in Pomak but did
not figure in our corpus. Tamam is also used in Macedonian and Bulgarian.
Evangelia Adamou
(33) aga ni poje-m nema

da ida-m faf panepistimjo
if
neg study-1sg not.have.3sg to go-1sg in university
If I dont study, I wont go to the university.
(Q, F3, 7, Pmk, Ell)
aku conditional
is is the most frequent conditional subordinator in Pomak1, used for both realis
Th
and irrealis clauses.
(34) aku ita-t
da ida-t da sa
uto-t
faf panepistimjo-no
if
want-3pl to go-3pl to refl study-3pl in university-def.dist
trva da ima-t
hubavi kitape ut
imotiko-no
must.3sg to have-3pl good
books from primary-def.dist
If they want to go to study at the University, they must have good books
since primary school. (Q, F2, 30, Pmk, Ell, Tur)
(35) aku ne raboti tyljk-an
mi
if
neg work.3sg man-def.dist 1sg.dat
e
da dojda
nah selo
mod mod come.3sg to village
If my husband isnt working, he will come to the village.
(36) aku beh
znala
if
aux.1sg knew.p.prf
e
beh
rudila
drugo-no
dete po
napre
mod aux.1sg gave.birth.p.prf another-def.dist child more forward
If I had known, I would have had another child earlier. (Q, F2, 30, Pmk, Ell, Tur)
5.2 Causal
Ga: from temporal to causal
It is also common cross-linguistically for temporal subordinators to express causality.

In Pomak1, ga also has some causal uses, most probably derived from the temporal
aga after phonetic erosion. The use of ga presents the cause-effect relation as obvious
and has a strong discursive component. The causal clause with ga usually precedes the
main clause.
(37) ga
nema-me
tumafil ni hode-me nah deniz-en
since not.have-1pl car
neg go-1pl
to sea-def.dist
Since we dont have a car, we wont go to the sea.
(38) ga
ni te-
nema
da hode-me
since neg want-2sg not.have.3sg to go-1pl
Since you dont want to, we wont go.
oti: causal
The other causal subordinator oti, is probably a loan from the New Testament Greek
oti because. In causal clauses with oti, the main clause generally precedes the subordinate clause.
(39) aitus trva da je
letalo
here must.3sg to aux.3sg rained.prf
oti
sabalahin
be
jatse mokro pot-en
because in.the.morning was.3sg a.lot wet
street-def.dist
Here, it must have rained because this morning the street was wet.

(40) gu
ni it
oti
vika gjo
3sg.m/n.acc neg wanted.3sg because say.3sg supposedly
je
paratik
is.3sg ugly
She didnt want him because, she says, supposedly hes ugly.

(41) kopel-an
hi
je
dal
sindir
boy-def.dist 3sg.f.dat aux.3sg gave.prf necklace
oti
j
imala duomgyny
because aux.3sg had.prf birthday
The boy offered her a necklace because it was her birthday.

6. Conclusion
The study of the Pomak variety spoken in the Xanthi area (Greece) reveals an interesting use of deixis in the formation of temporal subordinators, indicating the anchoring
of the event in the time of utterance while the choice of the subordinator indicates
the type of event encoded in the clause. Absence of a deictic suffix indicates a break
between the discourse situation and the process situation.
This is an uncommon phenomenon among Slavic languages, even though it is
relatively common cross-linguistically, as this volume shows. More research is needed
Evangelia Adamou
on this phenomenon, both from a synchronic and a historical perspective. Thus far it
has received practically no attention in South Slavic studies.
FYRO
Macedonia
Bulgaria
Pomak
Albania
Greece
Turkey
Map 1. The traditional Pomak speaking area in Greece
Abbreviations
acc
accusative
aux
auxiliary
dat
dative
def.a
definite addressees sphere
def.dist
definite distal
def.s
definite speakers sphere
evid
evidential
f
female
gen
genitive
ell
Greek
imp
imperative
loc.hab.fut located in the habitual/future
of the moment of utterance
loc.pas located in the past
of the moment of utterance
m
mod
n
neg
poss
prf
p.prf
pmk
q
refl
s
tma
tur
male
modality
neutral
negation
possessive
perfect
past perfect
Pomak
questionnaire
reflexive
spontaneous
Tense, Mood, Aspect
Turkish
References
Adamou, Evangelia. 2006. Le nashta. Description dun parler slave de Grce en voie de disparition.
Munich: Lincom.
Adamou, Evangelia. 2008. Sur les traces dune dgrammaticalisation: Le mdiatif en pomaque
(Grce). Revue des tudes Slaves: Proceedings of the French delegation at the XIV International Congress of Slavists 78(12): 177189.
Adamou, Evangelia & Drettas, Georges. 2008. Slave. In Le patrimoine plurilingue de la Grce,
EvangeliaAdamou (ed.), 107132. Leuven: Peeters.
Anderson, Stephen R. & Keenan, Edward L. 1985. Deixis. In Language, Typology and Syntactic
Description, Shopen Timothy (ed.), 259308. Cambridge: CUP.
Assenova, Petya. 2002 [1st edition 1989]. Balkansko ezikoznanie. Veliko Trnovo: Faber.
Bally, Charles. 1926. Lexpression des ides de la sphre personnelle et de solidarit dans les
langues indo-europennes. In Festschrift Louis Gauchat, Franz Fankhauser & Jakob Jakob
(eds), 6878. Aarau: Verlag Sauerlander.
Breu, Walter. 1994. Der Faktor Sprachkontakt in einer dynamischen Typologie des Slavischen.
In Slavistische Linguistik 1993, Hans Robert Mehlig (ed.), 4164. Muenchen: Verlag Otto
Sagner.
Bril, Isabelle. 2002. Le nlmwa (Nouvelle-Caldonie): Analyse syntaxique et smantique. Louvain:
Peeters.
Culioli, Antoine. 1971. propos doprations intervenant dans le traitement formel des langues
naturelles. Mathmatiques et Sciences humaines 34: 715.
Culioli, Antoine.1978 [1983]. Valeurs aspectuelles et oprations nonciatives: La notion
daoristique. In Enonciation: Aspect et dtermination, Sophie Fisher & Jean-Jacques Franckel
(eds), 99114. Paris: EHESS.
Culioli, Antoine. 1990. Pour une linguistique de lnonciation. Oprations et reprsentations.
Paris: Ophrys.
Cyxun, Gennadij A. 1981. Tipologieskie problemy balkanoslavjanskogo jazykovogo areala.
Minsk: Nauka i texnika.
Dimitrova-Vulchanova, Mila & Vulchanov, Valentin In press. An Article Evolving. In DIGS Volume, Stephen R. Anderson & Dianne Jonas (eds). Oxford: OUP.
Drettas, Georges. 1990. Le dialecte bulgaro-macdonien de Xr. (Edhessa, Grce). Questions de
typologie. Bulletin de la Socit de Linguistique de Paris 85(1): 227265.
Elson, Mark J. 1976. The definite article in Bulgarian and Macedonian. Slavic and East European
Journal 20: 273279.
Feuillet, Jack. 1999. Grammaire historique du bulgare. Paris: Institut dtudes slaves.
Fielder, Grace E. 1985. Aspect and Modality in Bulgarian subordinate clauses. In The scope of
Slavic aspect, Michael S. Flier & Alan Timberlake (eds), 181193. Columbus OH: Slavica.
Friedman, Victor. 2002. Macedonian. Munich: Lincom.
Friedman, Victor. 2004. The typology of Balkan evidentiality and areal linguistics. In Balkan Syntax
and Semantics [Linguistik Aktuell/Linguistics Today 67], Olga Tomi (ed.), 101134.
Guentchva, Zlatka. 1995. Limparfait perfectif en bulgare. Modles linguistiques 16 (2): 7394.
Glbov, Ivan. 1950. Za lena v blgarskija ezik. Izvestija na narodnija muzej v Burgas 1: 171227.
Kabasanov, Stajko. 1963. Edin starinen blgarski govor. Tixomirskijat govor. Sofia: Izd. na
Balgarskata akademija na naukite.
Evangelia Adamou
Kasatkina, Rozalia. 2008. Artikli, predartikli i psevdoartikli v russkom iazyke. In Proccedings of
the Russian delegation at the XIVth International Congress of Slavists, 305321.
Kanevska-Nikolova, Elena. 2001. Govort na selo Momilovci, Smoljensko - polovin vek po-ksno.
Sofia: Medunarodno sociolingvistiesko druestvo.
Kanevska-Nikolova, Elena. 2006. Trojnoto lenuvane v rodopskite govori. Plovdiv: Univ. Izdatelstvo
Paisij Xilendarski.
Kokkas, Nikos. 2004a. Uchem so Pomatsko. Xanthi: Pakethra.
Kokkas, Nikos. 2004b. Uchem so Pomatsko B. Texts. Xanthi: Pakethra.
Kortmann, Bernd. 1997. Adverbial Subordination: A Typology and History of Adverbial Subordinators.
Levinson, Stephen. 2004. Deixis and pragmatics. In The Handbook of Pragmatics, Lawrence
R. Horn & Gregory Ward (eds.), 97121. Oxford: Blackwell.
Lyons, Christopher. 1999. Definiteness. Cambridge: CUP.
Marchello-Nizia, Christiane. 2006. From personal deixis to spatial deixis. In Space in Languages:
Linguistic Systems and Cognitive Categories [Typological Studies in Language 66], Maya
Hickmann & Stphane Robert (eds), 103120. Amsterdam: John Benjamins.
Matras, Yaron. 1998. Utterance modifiers and universals of grammatical borrowing. Linguistics
36(2): 281331.
Matras, Yaron. 2007. The borrowability of structural categories. In Grammatical Borrowing in
Cross-linguistic Perspective, Yaron Matras & Jeannette Sakel (eds), 3173. Berlin: Mouton
de Gruyter.
Minova-Gurkova, Liljana. (ed.). 1997. Makedonski jazik. Skopje: Prosvetno delo.
Mirev, Kiril. 1964. Za lennite formi v srednoblgarskite pametnici. Izvestija na Instituta za
Blgarski Ezik 11: 231234.
Mladenov, Maxim. 1990. Ima li v blgarskija ezik dvojno lenuvane? Blgarski ezik 40/30: 229231.
Mladenova, Olga. 2007. Definiteness in Bulgarian. Berlin: Mouton de Gruyter.
Robert, Stphane. 2006. Deictic space in Wolof. In Space in Languages: Linguistic Systems and
Cognitive Categories [Typological Studies in Language 66], Maya Hickmann & Stphane
Robert (eds), 155174. Amsterdam: John Benjamins.
Rogo, Ali. 2002. Pomakika dhimotika tragoudhia tis Thrakis. Xanthi: Tameion Thrakis.
Theoharidis, Petros. 1995. Pomakoi. Xanthi: Pakethra.
Theoharidis, Petros. 1996a. Grammatiki tis Pomakikis glossas. Thessaloniki: Aigeiros.
Theoharidis, Petros. 1996b. Ellino-Pomakiko lexiko. Thessaloniki: Aigeiros.
Theoharidis, Petros. 1996c. Pomako-Elliniko lexiko. Thessaloniki: Aigeiros.
Sobolev, Andrej. N. 2001. Bolgarskij shirokolykskij govor: Sintaksis, Leksika duchovnoj kultury,
Teksty. Marburg: Biblion Verlag.
Svane, Gunnar. 19611962. O sintaksieskom primenenii bolgarskogo lena v XIII veke.
A. ergedskie teksty. B. Dobrejovo evangelie. Scando-Slavica 7, 8: 233251, 2428.
Topolinjska, Zuzanna. 2006. Trojnot len: da ili ne? Juznoslovenski filolog 62: 715.
Traugott, Elizabeth Closs, ter Meulen, Alice & Reilly, Judith & Ferguson, Charles A. (eds) 1986.
On Conditionals. Cambridge: CUP.
Vaillant, Andr. 1964. Manuel du vieux slave. Paris: Institut dtudes slaves.
Vaillant, Andr 1977. Grammaire compare des langues slaves, Tome V. Paris: Institut dtudes
slaves.
Correlative markers as phoric

Grammaticalised Category Markers
of subordination in German
Colette Corts
Universit Paris 7-Denis Diderot, Fdration Typologie et Universaux
Linguistiques, CNRS
The present article deals with the status and definition of correlative markers in
modern German subordinate clauses, which retain phoric properties from
their etymological determinative nature, from which their demarcative or
connective functions proceed, characterising either the subordinate clause itself
or its relation to the complex sentence. The discussion shows that correlatives
are deep structure determination markers and that their semantic and pragmatic
functions vary with the type of subordinate clause and the type of complex
sentence with which they combine. The complex sentence is thus to be defined
not as the mere concatenation of two simple sentences, but rather as the result
of intricate constructions which must be considered from a holistic viewpoint,
because the whole sentence is more than the sum of its parts, where additional
meaning is essentially contained in the binding determinative markers and
assertive markers which build interdependent relations and set the sentence
within a bigger pragmatic and textual whole.
1. Introduction
This article explores the status and definition of correlative markers in modern
German subordinate clauses.1 Correlative markers are rarely indispensable for the
acceptability of a sentence, consequently their function is difficult to describe. There
is also a theoretical difficulty: while, from a diachronic viewpoint, the etymological
determinative nature of German correlative markers is taken into account or even
presupposed as a linguistic datum, in synchronic descriptions, on the other hand,
their determinative properties have long been neglected. Since our analysis concurs
1. The author is indebted to Isabelle Bril and Anne Daladier for stimulating discussions and
helpful comments.
Colette Corts
with observations made by historical linguists on the relations between determination and subordination, this paper will instead focus on the key determinative role of
correlative markers in different types of subordinate clauses in modern German and
will describe their various functions at phrasal, sentential and textual levels, from a
strictly synchronic perspective.
One tricky question is the semantic and pragmatic status of the correlative markers, which requires cautious analysis: what needs to be defined is the genuine function
and semantic value of a determination marker, which, in other contexts, can be used
as a deictic or phoric (anaphoric/cataphoric) element. It will be shown that the correlative markers of modern German share only some of the properties of deictics; correlative markers are phoric entities with demarcative and connective functions which
may characterise the subordinate clause itself or its relation to the complex sentence
as a whole. Relative clauses will be excluded from our study since a specific theoretical
framework would be required to account for the particular relation between the subordinate clause and its antecedent noun phrase.
The analysis will be couched in the dependency theory framework, and more specifically in Jean Fourquets model. This theoretical framework is relevant not only for
every phrasal component of the German sentence, as is generally admitted, but also for
the whole complex sentence considered as a matrix construction and as part of a text.
Jean Fourquet (1970), the editor of Lucien Tesnires posthumously published major
work Elments de syntaxe structurale (1954), developed Tesnires connexion model
into a two-level model of dependency, which, for every kind of phrase, combines the
lexicon-driven connexions with a phrasal category frame, and which considers
the interrelation between connexions and category as the fundamental scheme of
German syntax. In Fourquets views (1970), every syntactic phrase has two levels:
The first level is sub-phrasal, comprising a base (i.e. a basic lexical node imposing a valency frame on the phrase) and any convenient number of members (the
number of participants depends on the valency of the lexical node, be it a verb,
a (predicative) noun, an adjective or adverbs; the number of adjuncts is open).
The relation between a base and its member(s) is called a connexion, a term
borrowed from Tesnire, with whom Jean Fourquet had long discussions at the
University of Strasbourg. Following Tesnires terminology, a connexion is a
dependency relation, involving a valency driven actantial relation or a circumstantial relation; but Fourquet inserts the notion of connexion into a binary
connectional construction, proceeding from an initial connexion between the
base and the nearest member to its left, which is the basis for another connexion
involving the next member to its left, and so on until reaching the last constituent
of the phrase (See the continuous lines representing connexions (1) and (2) in
Figure (1)). In this manner Fourquet also accounts for the information contained
in the linear word order of the phrase members, which he considers part of the
syntactic and semantic structure of the sentence: for instance, das alte groe Haus
the old big house is represented by a different connexion scheme from das groe
alte Haus the big old house in Figure (1).
On the second level, the syntactic sub-phrase which comprises the interconnected base and its member(s) must be overarched by specific category markers (such as definiteness for the noun phrase, or tense, aspect and mood (TAM)
for the verb phrase), in order to form a complete syntactic phrase. Fourquets
categories anchor the phrase in the discourse situation, involving an assertive
mood selected by the speaker, in accordance with its semantic and pragmatic relevance within the sentence and the communication context. Fourquets categories (marked by a dotted line in Figure (1)) insert the phrase into the speakers
discursive and pragmatic process. These categories are the grammatical tools
developed by natural languages to express the metalinguistic and cognitive processes relevant for a specific communication situation. Determination is one of
the most important grammatical concepts underlying the categories in German
phrases: noun phrase determination is grammatically marked by definiteness
markers, verb phrase determination is marked, for instance, by tense, aspect and
mood (TAM), and adjectival/adverbial phrase determination is marked by comparative markers. To avoid possible misinterpretations of the highly polysemic
term category, henceforth I will refer to Fourquets categories as Grammatical Category Markers (abbreviated as GCMs).
In Jean Fourquets model, a well-formed connexion does not constitute a phrase

until it has been instantiated by the speaker as either determinate or indeterminate.
The connexion and the cognitive or pragmatic instantiation in discourse through
the Grammaticalised Category Markers are closely interconnected at every syntagmatic level in a kind of parallel processing,2 from the smallest constituent to its
insertion step by step into the whole sentence or text. Such a model is incompatible
with a split-level linguistic theory, because it connects the syntactic, semantic and
pragmatic modules (Descls 1992) rather than considering them as superimposed
separate levels. The connexion and its relation to the Grammaticalised Category
Markers are distinct but parallel and interdependent processes linking and structuring the discourse at every syntagmatic level, from phrase to sentence and text levels:
Fourquets syntactic model is both a connectional and a categorical syntagmatic
2. The notion parallel processing is borrowed from computer science.
Colette Corts
construction. In relation to the subordinate clause, the determinative Grammaticalised Category Markers not only contribute to the interpretation of every phrasal
component, but they also contribute to the interpretation of the whole sentence. Thus,
following Fourquet, not only the lexicon-driven relations define the syntagmatics of
the subordinate clause, but the categorical part of every phrase and of the whole
sentence builds up a syntagmatic relation of its own which anchors the clause in the
communicative and cognitive context.
das
groe
alte
connexion (2)
Haus
connexion (1)
Figure 1. A sample analysis of a noun phrase: das groe alte Haus (the big old house)
A grammatical category marker (catgorie) in Fourquets model fulfills two

functions:
On the one hand, within the phrase itself, the GCM has a bracketing function: the
phrase is bracketed to its right by the base constituent, and to its left by the the
GCM, which fulfills a demarcative function and helps the hearer recognise the
edges of the phrase.
On the other hand, the GCM also fulfills cognitive and pragmatic functions.
First, the GCM is a link with the communicative context and expresses the speakers viewpoint (about the instantiated reference, a specific discourse universe, the
presupposed knowledge required to understand the message, and so on). GCMs
may be considered as deictic or phoric markers (See the discussion under 1.3.
below) which anchor the phrase in the discourse.

Secondly, the GCM may have a variety of scopes and may refer to different
discourse levels: phrase, sentence or text.

Thirdly and consequently, the GCM acquires a connective function within
the text: a new subject, first presented as indefinite at the beginning of the text,
becomes common knowledge as the text proceeds and is then referred to as being
definite. This is why determinative morphemes often fulfilll a connective function and are so frequently found as subordinating or coordinating conjunctions
in German.
I will show that the correlative markers, as determinative markers of a subordinate
clause and of its relation to the complex sentence, have the same demarcative, phoric
and connective functions as the determiners of any other phrase type and I will defend
the thesis that the correlative markers of subordinate clauses are specific GCMs (in the
sense of Fourquets catgories) of the whole sentence and of its components.
2. T
owards a definition of the german subordinate clause
Correlative Grammatical Category Markers (CGCMS)
Historical linguistic work on German has shown that determination plays a major role
in the form of subordinate markers, be they subordinating conjunctions or correlative
markers. Curme (1922) writes: From the indefinite wer originate the interrogative and
the relative wer, in both of which the former indefinite force is still felt. (1922:190) In
opposition to wer, der can be used as a definite demonstrative pronoun, as a relative, or
as the correlative marker of a subordinate clause introduced by wer: Der in Example1B
below, Wer das sagt, der lgt The one who says this tells a lie is not an antecedent of
wer, but only the repetition of the subject, der is not necessary and may be dropped.
However, if wer and the correlative marker do not bear the same case, the latter is usually not omitted. Wer einmal lgt, dem glaubt man nicht, und wenn er auch die Wahrheit
spricht The one who told lies once will never be trusted again, even when he happens to
say the truth (Curme 1922:207). Curme points out three important properties of the
correlative marker: it is definite, repetitive and may often be dropped.
Few synchronic analyses of German subordination show genuine interest in the
functions of correlative markers and their determinative role in such constructions,
apart from Fabricius-Hansen (1981), Sonnenberg (1992) and Zifonun, Hoffmann &
Strecker (1997). Fabricius-Hansen (1981) and Sonnenberg (1992) argue that the syntactic models used to account for subordinate clauses do not suitably describe the
correlative markers.
The correlative markers of the German subordinate clauses will first be defined
from four usual points of view: morphological, semantic, syntactic and pragmatic.
2.1 Morphology
Eisenberg gives a definition with a short list of the correlative markers under consideration: Als Korrelate werden in der Literatur Ausdrcke unterschiedlicher Kategorie
bezeichnet, die in phorischer Funktion mit Nebenstzen sowie Infinitiv- und Partizipialgruppen auftreten. Die wichtigsten Korrelate sind es, Formen des Demonstrativums (das, dem), die Pronominaladverbien (darauf, deswegen) und bestimmte andere
Adverbien wie so und dann. (Eisenberg 1994:351) (In linguistic descriptions, what
are called correlative expressions belong to different morphological classes and are
used in phoric function with subordinate clauses and with infinitive or participial
phrases. The main German correlatives are es it, demonstrative forms (das, dem this/
that), prepositional pronouns (darauf, deswegen thereon, therefore) and so-called
adverbs like so so and dann then). Eisenberg insists that the so-called correlative markers belong to different morphological classes (pronouns, pronominalized
prepositional phrases and adverbs), which have a common phoric function analysed
under Section2.2.
Colette Corts
There are different types of correlative markers: the third person singular neutral
pronoun es, and the demonstrative markers das, den, dem, which can be used either as
pronouns (substitutes), as articles heading a noun phrase or as correlative markers. The
Pronominaladverbien (darauf, deswegen thereon, therefore) comprise a demonstrative pronoun (da(r)-/des this/that) and a preposition (auf/wegen on/because of ), and
they may behave as the pronominal substitutes of a prepositional phrase. But, when
used as correlative markers, they are no longer substitutes; as Eisenberg pointed out,
they have a phoric function in relation to the subordinate clause. As for the so-called
adverbs so and dann, in contrast with standard manner or temporal adverbs (such
as anstndig well-behaved or am 9. November1989), so and dann require some com
plementary information within or about the discourse to be interpreted conveniently
(see 2.2.). Consequently, they are not adverbs, but discourse-dependent elements, a
property they share with pronouns. They are pro-adverbials with phoric function.
Morphologically, all these constituents are pro-elements, i.e. lexically empty morphemes referring to the context or the co-text, and which may be used as phrasal substitutes (es it, deswegen therefore, so, dann) or which may, for some of them (das/den/des/
dem this/that), introduce a phrase. As lexically empty morphemes, they have nothing to
do with connexional syntagmatics, and may be considered as resulting from the grammaticalisation of correlative schemes (Rousseau 1984). Their functions in subordinate
clauses will be dealt with in 2.2., and their semantics in 2.3.: more evidence of their being
grammatical category markers (in the sense of Fourquets model) will be provided.
2.2 Syntax
The pronominal nature and the lexical emptiness of correlative markers in German
subordinate clauses also account for their syntactic behaviour, e.g. their negative reaction to a reliable syntactic test used to identify the function of subordinate clauses:
none of the correlative markers may function independently as a reply to a question,
unless related to some deictic gesture or to some complement in the discourse. In
Examples (1a) and (2a), the proposed answer cannot be interpreted without reference
to the context or the co-text:
(1) a.

Worauf bestehst du? *Ich bestehe darauf.

(What do you insist on? *I insist thereon)
(2) a. Wie benahm er sich? *Er benahm sich so

(How did he behave? *He behaved so)
The correct answer would require the combination of the pro-element used as a correlative marker with a subordinate clause as in (1b) and (2b):
(1) b. Worauf bestehst du? Ich bestehe darauf, dass du mich zum Arzt begleitest.

(What do you insist on? I insist that you accompany me to the doctors.)
(2) b. Wie benahm er sich? Er benahm sich so, wie zu erwarten war.

(How did he behave? He behaved as expected.)
As an answer to a w-question, the correlative marker forms a syntactic unit with

the subordinate clause. This shows that it has no syntactic function of its own.
Furthermore, the correlative marker shares the same syntactic function as the
subordinate clause with which it is coupled, as we shall see in Part2. The correlative marker is a pro-element, but not a substitute for the subordinate clause, since
the syntactic function of the subordinate clause can always be defined, even in the
absence of any correlative marker. In Fourquets model, the correlative markers
must then be considered as catgories, i.e. as grammatical category markers of
the subordinate clause. Part2 will give a syntactic analysis of the property of correlative markers, as a syntactically transparent, ancillary category marker of the
subordinate clause.
Two other important factors are the position of the subordinate clause, in sentence initial, medial or final positions, and the key role of some assertive particles for
information structure. These linguistic markers must be analysed carefully to define
the different types of subordinate clauses, especially the difference between bound and
unbound subordinate clauses and the corresponding correlative markers. This will be
developed in Part2.
2.3 Semantics and pragmatics

According to Eisenberg, phoricity is the property common to the pro-elements
called correlative markers. A phoric (anaphoric/cataphoric) entity is an indexical
morpheme linked to the context or the co-text. Most phoric markers are homonymous with deictic markers, their meaning is contextual. Let us now investigate the
differences between deictic and phoric functions, starting with some definitions
of deixis. Deictics are indexical material which present and structure information
within the communication frame and the cognitive background of the communication event (Blhdorn 1993, 1996). Blhdorn insists on the interaction between
communicative and cognitive aspects as well as on the processive character of discourse construction. During discourse progression, all the information building
steps need to be stored for the hearer and speaker to be able to process it and refer
to it at any time. This is why Blhdorn (1996) considers that phoric and deictic
entities share the same indexical nature, but only part of their semantic properties. Phoric and deictic indexical morphemes share the four following properties:
(i) endophora, or the ability to refer to other parts of the discourse (Referenzverhalten), (ii) stress, in order to attract the hearers attention (Aufmerksamkeitsteue
rung), (iii) construction of reference from the communicative origo-parameters
Colette Corts
(Origobezug) and (iv) the proximal/distal (Nah-Fern) opposition.3 On the basis of

Blhdorns definition, phoric elements refer not only to representations of events or
objects, but also to all parts of the sentence as components of the discourse and communication process. The relations between the correlative marker and the subordinate
clause or the complex sentence are thus best dealt with in terms of phoric relations,
participating in the information structure and discourse construction at the level of
the sentence or some of its subparts, or at text level. As such, they undoubtedly belong
to the class of modern German grammatical category markers.
2.4 Results
Morphologically, the correlative marker is a pro-element. Syntactically, it shares the
syntactic function of the subordinate clause with which it is coupled. On the semantic
and pragmatic levels, it is a determinative phoric element playing a pragmatic role in
the construction and information structure of the sentence or of parts of the text; it is
able to mark focus and stress, or to structure the hierarchy between backgrounded vs.
foregrounded information. It is not a phrasal substitute, but an indexical category referring to the subordinate clause or to its relation to the complex sentence. This confirms
that it can be analysed as a Grammatical Category Marker (Fourquets Catgorie).
Like the other Catgories and according to the context, a correlative marker
may fulfilll a demarcative and/or a connective function. As expected, these functions depend on the syntactic, semantic and pragmatic role of the subordinate clause.
Part2 will provide further evidence for the fundamental observation that the correlative markers of the subordinate clauses are specific Grammatical Category Markers
of the sentence and its components; it will show the functions of the various types
of Correlative Grammaticalised Category Markers (CGCMs), corresponding to the
functions of the various types of subordinate clauses.
3. Im vorliegenden Aufsatz wird () die Position bezogen, dass nachricht- ebenso wie
textbezogene Anknpfung sich, wenn man das Kognitionsgeschehen mitbercksichtigt, als
Spezialfall von situationsbezogener Anknpfung erweist. Die besprochenen Sachverhalte (die
Nachricht) und das materielle Zeichen (der Text) sind nmlich nichts anderes als Teilnehmer
der Kommunikationssituation neben dem Sender und dem Empfnger, die ebenso wie letzterer
in der Kognition der Kommunikationspartner reprsentiert werden mssen, um als Substrat
fr Anknpfungen verfgbar zu sein. Mit anderen Worten : Bei der phorischen Anknpfung
wird vorausgesetzt, dass im Laufwissen des Empfngers eine (eventuell noch unvollstndige)
Reprsentation derjenigen Sachverhalte und Sachverhaltsbestandteile vorhanden ist, zu deren
Reprsentation die im Zuge des aktuell stattfindenden Kommunikationsereignisses bereits
bermittelten Zeichenbestandteile anleiten sollten. Blhdorn (1996:132)
3. F
our types of subordinate clauses corresponding
to four types of complex sentences
In this article, only subordinate clauses in their canonical form (beginning with a subordinate conjunction and ending with a conjugated verb-form) will be dealt with,
excluding relative clauses whose complex relations between the antecedent nounphrase and the subordinate clause would take us far beyond the limits of this paper. In
modern German, subordinate clauses fall into four classes as shown in 3.1.
3.1 S yntactic and pragmatic functions of the German subordinate

clauses having no noun phrase as antecedent
German has developed two systems of Correlative Grammaticalised Category Markers (the d- correlative markers (Part3) and the so correlative markers (Part4)); but
before proceeding, following is a classification of subordinate clauses according to
their syntactic and pragmatic properties. This classification is based on a series of
tests devised by Oswald Ducrot (1972), which highlight the diversity of the relations
between the subordinate clause and the complex sentence on the syntactic and pragmatic levels. The results of Ducrots tests applied to a corpus from the Spiegel CD Rom
(from 1997 to 2000) are summarised in Table1.
Table1.
(Subordinate Clause =Q)
Type 1
Type 2
Type 3
Type 4
Q as answer to a w-question
Insertion of a negation just before Q
Insertion of a grade particle
(Gradpartikel) just before Q
(e.g. only, just)
Insertion of a modal particle just
before Q
Q as an echo question
Insertion of und dies, und zwar
and this
Insertion of nmlich namely
Q is in the scope of a global question
Q is in the scope of a global negation
Q is in the scope of a global
modalisation
Embedding of the complex sentence
into P (ex. P =er glaubt, dass
(P =he believes that))
+
+
+
+
+
+
+
+
+
+
+
+
+
[P [P, Q]]
[P [P, Q]]
[[P P] Q]
interpolation
Colette Corts
As Table1 shows, Ducrots syntactic and pragmatic tests lead to a clear distinction
between four classes of subordinate clauses, two classes (type 1 and 2) with mostly
positive tests, giving evidence for a particularly close relationship between subordinate clauses and matrix sentences, and therefore called bound subordinate clauses
(subordonnes lies) by Ducrot, and two other classes (type 3 and 4) of unbound
subordinate clauses (subordonnes non lies) which all test negative, thus evidencing a looser relationship between the subordinate clause and the complex sentence.
The bound subordinate clauses are syntactic constituents of the complex sentence
and the unbound subordinate clauses are connected to the complex sentence on a
semantic or pragmatic level.
Type 1 tightly bound subordinate clauses can be used to answer a w- question

introduced by was, wo(r)+ preposition (what, preposition+ what), they are syntactic constituents of the complex sentence with actancy role; the relation between
the complex sentence and the subordinate clause is syntactic and depends on the
main verbs valency.
Type 2 subordinate clauses are less tightly bound; they can be used to answer a
w- question introduced by wann, wie, warum (when, how, why), and are syntactic constituents of the complex sentence with a circumstantial role; the relation
between the complex sentence and the subordinate clause does not depend on the
main verbs valency but refers to the circumstantial frame of the main process.
Type 3 unbound subordinate clauses are not syntactic constituents of the complex sentence; the relation between the complex sentence and the subordinate
clause is pragmatic, argumentative and rhetorical.
Type 4 unbound subordinate clauses have a metalinguistic function and
are used to comment on the sentence (or on the parts of the sentence) in which
they appear.4
The type of correlative grammaticalised category markers (CGCMs) used in subordinate clauses confirms the results of the syntactic analysis summarised in Table 1,
which shows that the subordinate clause and the corresponding correlative marker
4. The syntactic and pragmatic classification of subordinate clauses in German into the four
subclasses presented in this chapter is highly compatible with Foley and Van Valins model
(1984). Our bound subordinate clauses correspond to what the authors call subordination
([+ dependent], [+ embedded]) with its two subclasses, one being characterised as core embedding (corresponding to our type 1) and the other as oblique embedding (our type 2). Our
unbound subordinate clauses correspond to what they call cosubordination ([+ dependent],
[ embedded]), and our proposals about the correlative and subordinate clauses of type 3 and
4 in German may be seen as a contribution for the exploration of the dependency relation
involved in cosubordination, which is still under discussion (See Foley, this volume).
always share the same syntactic function for type 1 and 2 bound subordinate clauses,
and the same pragmatic function for type 3 and 4 unbound subordinate clauses. As
a determinative element, the CGCM underlines the relation between the subordinate
clause and the complex sentence:
In type 1 tightly bound subordinate clauses with actancy function, the correlative
marker is a simple pronoun (es, das) sharing a subject or an object function with the
subordinate clause, as in example (3), or it may be a complex pronoun, combining a
d-marker and a preposition (darum, daran, etc.) if the subordinate clause functions
as a prepositional object, as in example (4). The CGCM shares the syntactic function of the subordinate clause, depending on the governing verbs valency; it helps
identify the function of the subordinate clause in the matrix sentence, especially if
the subordinate clause is an oblique case (particularly dative dem) or a prepositional
object (da(r) +preposition). This could be the formal justification for the use of
a CGCM on the syntactic level. But the CGCM cannot be reduced to such twofold marking of a syntactic function, since it may be dropped most of the time, as
pointed out by Curme (1922). It also has a pragmatic function, that of structuring
the information contained in the complex sentence: in example (3), the CGCM es
in a cleft sentence leads to the rhematisation of the subordinate clauses content. The
CGCM puts the content of the subordinate clause headed by dass into focus, where
it receives special stress as the main piece of information in the complex sentence. In
other examples such as (15), the presence of the CGCM and the subordinate clause
in sentence initial position topicalises the subordinate clause.
(3) (Type 1) Manche wissen es, viele ahnen es zumindest, dass Aktienbewertung und
Realittbei hunderten von Firmen nicht mehrin Sichtkontakt stehen.
Spiegel 2000, n05, p.77 (Some of them know it/that, many of them suppose it/
that at least: (that) for hundreds of business companies the value of the stock
option is not connected to the (financial) reality.)
(4) (Type 1) In der jetzigen Lage geht es darum, dass unser Land in einer sehr
schwierigen Situation eine handlungsfhige Regierung braucht. Spiegel 2000, n01,
p.116 (Nowadays the most important thing for us is that our nation, which
faces a very delicate situation, needs a capable government.)
With type 2 less tightly bound subordinate clauses with circumstantial function, the CGCM is a phoric element; it is the grammaticalised category marker
(GCM) of the circumstantial function of the subordinate clause ((5), (6), (7)),
sharing the same syntactic function as the subordinate clause with which it is
coupled. The CGCM can function as pre- or postdeterminer of the subordinate
clause which it precedes or directly follows (5), (6) and it may put into focus or
stress the information contained in the subordinate clause (5), (6), (7).
Colette Corts
(5) (Type 2) Das gegen Lassa-Fieber verfgbare MedikamentRibavirin hilft meist

aber nur dann, wenn es frhzeitig verabreichtwird. Spiegel: 2000, n0 3, p.18 (But
the Ribavirin medicine which is available for Lassa fever is in most cases only
efficient (then), when prescribed early.)
(6) (Type 2) Dann, wenn dem Schriftsteller das Vollkommene gelingt, hat er Zutritt in
jenes imaginre Reich der Kunst erlangt, wo der Kritiker nur noch der Beckmesser
ist und Kritik nur noch Besserwisserei. (Z95/502.00692 Die Zeit, 03.02.1995,
Nr. 06, Ressort: Feuilleton; Was wir lieben, wenn wir Joseph Roth lieben)
((Then) when the author reaches perfection (in writing), he has acceded to that
imaginative power of art where the critique writer is only a Beckmesser and the
critique is mere arrogance.)
(7) (Type 2) Dustin Hoffman war nicht nur deshalbbeeindruckend in dem Film
Rain Man weil er sich so tief in die Rolle einesAutisten versetzen konnte,
sondern weiler tatschlich schwere Probleme mit derWahrnehmung von
Wirklichkeit hat. Spiegel 1997, n0 4, p.38 (Dustin Hoffman was impressive in
the film Rain Man not just because he was able to impersonate so thoroughly
the character of an autistic person, but because he really has great difficulties in
perceiving reality.)
In type 3 unbound subordinate clauses, which are not a syntactic constituent

of the sentence (as shown by the negative tests presented in Table 1, especially
because they cannot answer a w-question), but which have a pragmatic and rhetorical function within the complex sentence, the CGCMs dann and so have a
different status. First, they cannot be preposed to the subordinate clause. Second,
if they accompany the subordinate clause in sentence initial position, the CGCMs
dann and so are always placed after the subordinate clause; they have a summarizing function and present the subordinate clause as an argument for the thesis
contained in the complex sentence as a whole (8). Their function is pragmatic, it
focalises the argumentative function of the complex sentence, within which the
subordinate clause provides the necessary background information. The CGCMs
found in type 3 subordinate clauses are argumentative markers which link the
subordinate clause to the complex rhetoric project carried out by the whole sentence. The operations encountered at the level of type 3 subordinate clauses and
the corresponding CGCMs structure the utterance, as confirmed by the frequent
use of assertive markers in the subordinate clause (ja, doch, etc. (9)).
(8) (Type 3) Sowohl in der Politik wie in den Schriften zur Philosophie wird
Aristoteles nicht mde, den Wert aller theoriegeleiteten, also allein auf Anschauung
gesttzten Disziplinen fr die menschliche Sozialisation zu bekrftigen: Obwohl
das hier erworbene Wissen doch theoretisch ist, so vollbringen wir doch unzhlige
Handlungen nach seinem Muster, indem wir nach seiner Magabe das eine
e rgreifen, das andere lassen, und vor allem mit seiner Hilfe alles Gute
erwerben. (Z04/404.03494 Die Zeit (Online-Ausgabe), 22.04.2004, Nr. 18,
Ressort: Feuilleton; Selige Apathie, S. 47) (In his Politics as well as in his
philosophical writings, Aristotle never ceases to insist on the importance, for
the socialisation of humans, of all theoretical disciplines, that is to say only
based on intuition. Even though the knowledge you get here is theoretical, [so]
we fulfilll a lot of actions patterned on it, we take this and leave that, and above
all we pick up all the best in conformity with it.)

(9) (Type 3) Die Kollegen aus dem Westen feixten ber seine rummelige
Altbauwohnung am Prenzlauer Berg, wo er doch die noble Dienstvilla in
Zehlendorfhtte beziehen knnen, die er BundesprsidentJohannes Rau berlie.
Spiegel 2000, n0 7, p.22 (His colleagues from the West were making fun of his
old and noisy apartment on Prenzlauer Berg, when he could have lived in the
magnificent service villa of Zehlendorf which he renounced for the sake of
PresidentJohannes Rau.)
Last but not least, in type 4 unbound subordinate clauses, which fulfilll a metalinguistic (commentative or corrective) function (10), the phoric d- morpheme
can never be used as a correlative marker, but the CGCM so may be used under
specific conditions, i.e.: if the subordinate clause is preposed and if the CGCM
links the main sentence to its communicative, cognitive and/or textual frame (11)
(see Part4 below).
The subordinating type 4 conjunction is always the indetermination w-marker which

may never be combined with the determinative d-marker, but may commute with it:
there is a clear syntagmatic opposition between the hypotactic w-comment and the corresponding paratactic d-comment in German; for instance in (10), the w-subordinate
clause (wofr alles spricht =as can be expected) is semantically very close to its paratactic
d-equivalent (und dafr spricht alles =(and) that can be expected), but they are syntagmatically incompatible.
(10) (Type 4) Wenn er brav ist, wofr alles spricht, wird er, einundsechzigjhrig, im
Jahr 2015 entlassen. Spiegel 2000, n0 11, p.176 (If he behaves well as can be
expected he will be released from prison in 2015, 61 years old)
(11) Was meine Identitt und die unterstellten Motive angeht, so sind sie fr die Bewertung
des Buches bedeutungslos. (Z96/608.04034 Die Zeit, 02.08.1996, Nr. 32, Ressort:
Dossier; Das Versagen der Kritiker, S. 9) (Concerning my identity and my
supposed motivation(s), (so) they are not relevant for the evaluation of the book.)
In fact, Ducrots tests provide information about the segment CGCM +subordinate
clause for all bound subordinate clauses (types 1 and 2) in which the CGCM is a
demarcative or discontinuous marker with which the bound subordinate clause can be
thematised, rhematised or focused for the sake of information structure.
Colette Corts
For type 3 subordinate clauses, Ducrots tests produce evidence for a looser syntactic and semantic link between the complex sentence and the subordinate clause (all
the tests are negative), but the possible correlative markers have pragmatic function.
The complex sentence formed with a type 3 subordinate clause is a pragmatic discourse unit: the subordinate clause refers to backgrounded information which lends
support to the demonstration contained in the complex sentence. The CGCM (which
cannot precede the subordinate clause) is a discourse connector which underlines the
pragmatic structure and the argumentative purpose of the complex sentence.
In type 4 subordinate clauses, d- correlatives never appear and so correlatives are
rare: the relation between the commentative or the corrective subordinate clause and
the complex sentence or any part of it is essentially marked by the relative pronoun
heading the subordinate clause. Type 4 subordinate clauses offer accessory knowledge
about the main clause or information structure. They provide comments with their
own illocutive purpose: the comment provided by type 4 subordinate clauses is pragmatically heterogeneous to the aim of the complex sentence.
Thus, to every kind of subordinate clause corresponds a special type of CGCM
with a specific status; this will be detailed in Part4 for the functions of d- correlatives
and in Part5 for the functions of the so-correlatives. Before that, let us briefly inquire
into the possible status of the CGCMs.
3.2 The possible status of the Correlative Grammatical Category Markers

The CGCM is a determinative pro-element with an ancillary, pragmatic (phoric,
demarcative and connective) function relative to the subordinate clause, in its relation to the complex sentence. It stresses, focuses, and organises information structure.
The CGCM, which has no syntactic function and no lexical content of its own, may
have a pragmatic function, based on its phoric property, and organises the information structure of the discourse from sentence to text.
From the syntactic point of view, the CGCM and the subordinate clause, which
share the same syntactic function, are not co-constituents of the sentence.5 They are
5. Elemente der Kategorie Korrelatverbindung sind Paare, bestehend aus einem

phorischen/ deiktischen Ausdruck und einer Infinitivkonstruktion oder einem Subjunktorsatz, die in der Kette diskontinuierlich und gegebenenfalls akzentuiert auftreten knnen.
Durch diese regelhafte Herleitung werden zwei Aspekte garantiert:

der deiktische / phorische Charakter des Korrelats bleibt erhalten und wird nicht etwa
durch eine eigene Kategorienzuordnung speziell fr die Verwendung als Korrelat berdeckt.
die Infinitivkonstruktion / der Satz bleibt ber die Regel als valenzgesteuertes Element
erkennbar. (Zifonun et alii 1997:14871488).
two interdependent parts of the same sentence constituent, but their interdependency
does not always have the same pragmatic function. Three possible solutions can
beimagined:
The CGCM can be a discontinuous constituent of the subordinate clause, especially in cleft sentences introduced by the pronouns es/das,
The CGCM can be a predeterminer or a postdeterminer of the subordinate
clause and serve as a horematic (from Greek horos border), demarcative
category,
The CGCM can have a connective function: dann/so.
Part4 (d- correlatives) and Part5 (so correlatives) will propose what seems to be the
most convenient solution for explaining the use of CGCMs in combination with the
different types of subordinate clauses.
Before this, let us comment briefly on the use of the pronoun es as a CGCM,
which is only used with type 1 subordinate clauses, and which must be strictly
distinguished from all other CGCMs. Es can never be used as a pre- or postdeterminer, located in the immediate neighbourhood of the subordinate clause; there is
always some distance between the phoric es and the subordinate clause (Examples
(3) and (12)).
(12) Es ist wahrscheinlicher geworden, da Lnder wie Irland und Spanien, die noch
hhere Zinsen haben, diese auf das Niveau von Deutschland und Frankreich
senken werden (Z98/809.05257 Die Zeit, 17.09.1998, Nr. 39, Ressort: Wirtschaft;
Die Enttuschung ist gro, S. 27). (It is rather to be expected that countries
like Ireland and Spain, which still have higher interest rates, will reduce them to
their level in Germany and in France.)
The CGCM es is not part of the subordinate clause, it is a discontinuous element

referring to the following subordinate clause and especially to its relation to the
whole sentence; its mobility with respect to the subordinate clause makes it an
eminently suitable tool for the thematisation or rhematisation of type 1 subordinate clauses with subject or object function (see Part2), which is a property of the
sentence as a whole. As an autonomous morpheme which may be used to build a
cleft-sentence, the CGCM es is not a phrasal marker (i.e. it is not a constituent
of the subordinate clause as an actantial phrase), but a sentential marker which
refers to the information structure of the complex sentence considered from a
holistic viewpoint.
Let us now detail the analysis of d- and so-Correlative Grammaticalised
Category Markers and the different contexts in which they fulfill all the functions
presented in 3.1.
Colette Corts
4. C
ombination of d-Correlative Grammatical Category Markers
and subordinate clauses
The d-correlatives only combine with types 1, 2 and 3 subordinate clauses. There are
no d-correlatives in type 4 subordinate clauses (see the comment on example (10) in
3.1. and see 5.2.)
To reveal some essential distinctions relevant for defining the functions of the
CGCMs, we will take advantage of a sentence-specific position rule. German is a V2 language characterised by the strict position of the constituents at the beginning of assertive
sentences: the conjugated verb form can only be preceded by a single constituent; on the
basis of that syntactic positioning rule, the status of the CGCM may be clarified, depending on whether the CGCM and the subordinate clause may appear together in sentence
initial position, just before the conjugated main verb ((13), (14), (15)), or not ((16), (17)).
4.1 Type 1 (subordinate clauses with actancy function)

Type 1 subordinate clauses introduced by dass that, ob whether or wie how, may be
followed by the CGCMs das (13) and da(r) +preposition (14), or preceded by them
(15). In all these examples, the unit comprising the subordinate clause and the CGCM
constitutes one syntactic constituent preceding V2: CGCM +the subordinate clause
in actancy function is a phrasal constituent of the sentence. On the contrary, the preposed subject and object subordinate clauses introduced by wenn and als (when) do
not constitute such a distributional unit with their CGCM das, because the wenn- subordinate clause precedes and the CGCM follows V2 (16).
(13) Das grausame Sterben von 118 Soldatenauf dem Grund der Barentssee aber
whltedie Untertanen auf. Dass Putin das U-Boot-Unglck aus der fernen
Sommerfrischeam Schwarzen Meer kommentierte, dass Admirale ihren
Oberbefehlshaber tuschenund dass dessen Kreml-Apparat wiederumdas Volk
hemmungslos belgt daswar zu viel. Spiegel 2000, n0 35, p.148 (But the cruel
death of 118 soldiers at the bottom of the Barents Sea shocked ordinary people.
That Putin commented the wreck of the submarine from his distant holiday
resort on the shores of the Black Sea, that some admirals lied to their Chief
Executive and that the Kremlin also in turn lied to the people that was
too much.)
(14) Herzog: Dass es Staaten gibt, die hier sehr genau aufpassen, darber gibt es
berhaupt keinenZweifel. Spiegel 2000, n0 25, p.445 (That some states are very
prudent on that matter, that is beyond any doubt).
(15) Daran, dass dasParken in der Stadt gebhrenpflichtig ist, hat man sich lngst
gewhntSpiegel 2000, n0 37, p.255 (That we are charged for parking in a town,
that is something we have long been used to).
(16) SPIEGEL: Warum dann noch die Prsentationim Fernsehen? Caball: Das ist
wie ein Vorspielen.Wenn Intendanten von Konzerthuserndie Sendung sehen,
kann das der Wegzum Ruhm sein. Spiegel 2000, n0 1, p.96 (Why still present
shows on TV? Mr. Caball: It is like being auditioned. When directors of
concert halls watch the broadcast, that may make you [young singers who
accompany me] famous.)
Type 1 subordinate clauses with an actancy role therefore subdivide into two subgroups, following their combination with the corresponding CGCMs; class 1A is the
prototypical actantial subordinate clause headed by dass, wie and ob; class 1B is the
non-prototypical actantial subordinate clause headed by wenn or als:
Class 1A subordinate clauses together with their CGCMs precede V2. They are prototypical in that they can fulfill the three basic functions governed by the main verb
of the sentence (subject, object and prepositional object), and they may combine
with the three CGCMs (es, das, da(r) + preposition). The d-CGCMs positioned
together with the subordinate clause as the first constituent of the sentence will be
considered as predeterminers when they immediately precede the subordinate
clause (15), and as postdeterminers when they immediately follow the subordinate clause (14). Class 1A subordinate clauses have a particular property: they may
be introduced by a categorising abstract noun-phrase like: der Gedanke, dass the
idea that, der Glaube, dass the belief that, and so on. The prototypical d-CGCM
of type 1A is a horematic marker (horos border) for the actantial subordinate
clause, delimiting it as a determined phrase within the complex sentence. The dcorrelative delimiting type 1A subordinate clauses is the last overarching grammaticalised category marker of the actantial subordinate clause and, as such, it has a
demarcative function within the actantial phrase CGCM +subordinate clause. In
example (15), for instance, the subordinate clause introduced by its d-CGCM must
be analysed in the same way as every other determined actantial phrase (Figure2):
TAM Daran, dass das Parken in der Stadt gebuehrenpflichtig ist, man laengst sich gewoehn
Daran, dass das Parken in der Stadt gebuehrenpflichtig ist,
Figure2.
Colette Corts
The non-prototypical class 1B subordinate clauses introduced by wenn and als

and preceding V2, on the contrary, are never immediately preceded or followed
by the CGCM das, which must follow V2 if the subordinate clause is in sentence
initial position (16), and precede V2 if the subordinate clause is not in sentence
initial position (17). They may have subject or object functions, but are never used
as prepositional objects; they may only combine with two of the CGCMs (es, das).
They also constrain the mood (the subjunctive II is frequently used with wenn
(17)) or the tense of the complex sentence. This last distinctive property confirms
that class 1B subordinate clauses are not prototypical actant phrases, although
they undoubtedly serve an actantial function.
(17) Es wre schn, wenn es ein besseres Beispiel gbe. Spiegel 2000, n0 41, p.131
(It would be nice if we/you could find a better example).
To conclude, type 1 complex sentences are the syntactic matrix within which the unit
CGCM +type 1 subordinate clause is an actantial phrase (with subject, object or prepositional object functions). With class 1A clauses, the relation between the subordinate
clause and the CGCM is tighter than with class 1B clauses: with class 1B, all the CGCMs
(es or das) are discontinuous constituents of the subordinate clause, while with class 1A
the d-correlatives are immediate constituents of the subordinate clause and have a demarcative function as a pre- or postdeterminer. Depending on whether it combines with
1A or 1B subordinate clauses, the CGCM has two distinct scopes: the segment CGCM
+class 1A subordinate clause or the matrix sentence as a whole (1B). Combined with a
class 1A subordinate clause, the CGCM is a phrasal marker; combined with a class 1B
subordinate clause, the CGCM is a sentential marker. With type 1 subordinate clauses,
the CGCM fulfills two distinct pragmatic functions conditioning the anchoring of the
sentence in the discourse and in the text. We must thus distinguish between a Determinative, Demarcative Phrasal d-CGCM for type 1A and a Discontinuous Sentential
d-CGCMs for type 1B. The analysis of type 2-CGCM will lead us to a similar distinction.
4.2 Type 2: circumstantial subordinate clauses

Based on their distribution, the CGCM and the preposed type 2 subordinate clause
also fall into two subgroups:
Type 2A are temporal subordinate clauses which may precede V2 jointly with
their CGCM (with a preposed CGCM (6) or with a postposed CGCM (18)).
Type 2B are subordinate clauses of cause (19) and goal (20), which may precede V2,
but their CGCM is never preposed; it must be either postposed or placed after V2.
(18) Sie knnen dasbeste Auto bauen.Aber wenn derjenige, der am Steuer sitzt, nicht
fahren kann oderwill, dann ntzt dasbeste Auto nichts. Spiegel 2000,
n0 1, p.31 (You may manufacture the best possible car. But if the driver is not
able or not willing to drive it, then the best car is useless.)
(19) Die Romane und Erzhlungen kreisen immer wieder um die Katastrophe des
Gedchtnisverlustes. Weil Literatur fr ihn ein einzigartiger Gedchtnisspeicher
ist, deshalb ist der Autor Lenz bis heute ein besessener Leser geblieben, wie seine
umfangreichen Essaybnde zeigen. (Z98/803.01603 Die Zeit, 19.03.1998, Nr. 13,
Ressort: Literatur; Leben ist Leiden, S. 54) (His novels and short stories always
deal with the disaster of lost memories. Because literature is for him an exceptional memory bank, therefore the author Lenz now still remains a passionate
reader, as shown by his numerous volumes of essays).
(20) Um die dunklen Orte in Deutschlandaufzuspren, hat Winfried Krling, Mitbegrnder der Anti-Licht-Kampagne, deshalbam Computer Satellitenfotos ausgewertet. Spiegel 1997, n0 5, p.157 (This is to list ill-lit places in Germany, (thats
why) Winfried Krling, the co-founder of the anti-light campaign analysed
computerised satellite pictures.
Besides the different positions of the CGCM relative to the subordinate clause,
there is another difference between 2A and 2B subordinate clauses: the temporal
subordinate 2A clauses may be preceded by a pronominal adverb or a prepositional
phrase (im Augenblick, wenn, als/damals, als/jedesmal, wenn (at the time when you
do something, formerly as, whenever), whereas this is impossible for 2B subordinate clauses (* aus dem Grund, weil, * zum Zwecke, um zu (*for the reason why, *
for the purpose in order to). Thus, temporal subordinate clauses (2A) form a unit
with the d- correlative and their CGCM is a horematic, demarcative marker of the
circumstantial subordinate clause. By contrast, the 2B examples attest to a looser
relationship between the CGCM and the subordinate clause. This is due to a difference of scope of 2A and 2B over the complex sentence, and this is why they are often
used in different text types:
The temporal subordinate clauses (2A) are linked to the tense, mood, aspect and
Aktionsart of the complex sentence; they have scope over the propositional content of the complex sentence interpreted as an event and they are mostly used in
descriptive or narrative texts. The 2A-CGCM is a phrasal marker: the determinative CGCM (like dann in ((6) Dann, wenn dem Schriftsteller das Vollkommene
gelingt,) is a horematic, demarcative marker delimiting the entire prepositional circumstantial temporal phrase: CGCM +type 1A (temporal) subordinate clause.
With cause and goal subordinate clauses (2B), the propositional content of the
sentence does not refer to an event (or a chain of events), but is conceived of as
a motivated action, based on a plurality of causalities and intentions (Ricoeur
1977). The explicative meaning of the complex sentence built with 2B subordinate
clauses results from a reinterpretation of the global sentence as a whole intentional
process, thus the CGCM is a sentential marker. Furthermore, as the 2B-CGCM
may not precede the subordinate clause in initial position, it is then a discontinuous Sentential postposed CGCM.
Colette Corts
The bound subordinate clauses (types 1 and 2) both subdivide in two subclasses; for
classes 1A, 2A, the CGCM is a d-demarcative Phrasal marker while for the other
classes 1B, 2B the CGCM is a discontinuous Sentential postposed CGCM.
It will be shown in 4.3 that unbound type 3 subordinate clauses are not compatible
with d-demarcative phrasal markers, but only with sentential postposed d-CGCMs.
4.3 Type 3 unbound subordinate clauses

According to Ducrots tests, type 3 subordinate clauses are not bound and therefore
never embedded. They have a specific pragmatic function in a special type of complex
sentence: they have argumentative function, and are used to defend the main clauses
thesis. The whole complex sentence is argumentative and is supported by the background information contained in type 3 subordinate clauses. The CGCM used in such
argumentative sentences is never a demarcative phrasal marker, it is a sentential marker
which stresses the argumentative relation between the subordinate clause and the sentence as a whole.
Hypothetical sentences with subordinate-clauses introduced by wenn and falls
(if, in case) are examples of type 3 argumentative subordinate clauses which are compatible with d-CGCMs. As shown in Table1, type 3 subordinate clauses are never used
to answer a w-question; this is a fundamental difference between temporal subordinate clauses, which may answer a w-question and thus belong to type 2, and hypothetical wenn subordinate clauses which belong to type 3 (21), (22).
(21) Wenn es um die Wiederherstellung der verfassungsmigen Ordnung
gegangen wre, dann htte Jelzin nicht nur Dudajews Palast, sondern auch
seinen eigenen Amtssitz bombardieren lassen mssen (Z95/501.00182 Die Zeit,
06.01.1995, Nr. 02, Ressort: Politik; Der Wahnsinn hat Methode, S. 1) (If it had
only been a matter of restoring constitutional order, (then) Jelzin should not
have had only the Dudajew palace bombed, but also his own official residence.
(22) Falls wir also berleben, dann ist alles, was ich tun kann, mein Leben auf
eine gute Weise zu leben. Ist das ein Traum? Jedenfalls ist es nicht einfach.
(Z02/206.03148 Die Zeit (Online-Ausgabe), 06.06.2002, Nr. 24, Ressort: Leben;
Ich habe einen Traum, S. 64) (In case we survive, then all I can do is to live my
life in a right way. Is that a dream? It is not easy anyway).
In such examples, the CGCM immediately follows the subordinate clause and cannot
precede it. Type 3 subordinate clauses and the connector dann never form a syntactic
constituent; the CGCM dann does not have a demarcative, but a connective, sentential
function: it marks a hypothetico-deductive relation between the unbound subordinate
clause and the complex sentence; this relation is based on the opposition between the
backgrounded information contained in the subordinate clause and the foregrounded
information contained in the main clause. The presence of various assertion markers
in the subordinate clause and that of connectors in the main clause provide further
evidence that the relation between the subordinate and the main clause is not syntactic
but pragmatic: the subordinate clause provides an argument presented as background
information to support the thesis defended in the main clause.
The functions of the d-CGCMs are summarised in Table2.
Table2. d- CGCM
Type 1 A Bound Subordinate Clause
(actantial)
Type 1 B Bound Subordinate Clause
(actantial)
Type 2 A Bound Subordinate Clause
(circumstantial)
Type 2 B Bound Subordinate Clause
(circumstantial)
Type 3 Unbound Subordinate Clause
(argumentative)
d-Demarcative Phrasal CGCM

(pre- (4, 15) or postdeterminer (13, 14))
(das/dar + Prep)
Discontinuous Sentential postposed CGCM
(das) (16)
d-Demarcative Phrasal CGCM
(pre- (5, 6) or postdeterminer (18))
(das/dar + Prep)
Discontinuous Sentential postposed CGCM
(7, 19, 20)
Connective Sentential postposed CGCM
(21, 22)
In Part5, we shall compare the function of the CGCMs dann and so in type 3
subordinate clauses.
5. C
ombinations between so-Correlative Grammatical Category
Markers and subordinate clauses
So is a very important morpheme used in subordinating conjunctions (sooft, solange,
sofern as soon as, as long as, as far as, to the best of ) or used in discontinuous comparative morphemes (so, wie/so, als ob as, just as, as if ), but it is also used as a CGCM.
By contrast with d-correlatives, so correlatives are not used with bound (types 1 and 2)
subordinate clauses, but only with unbound (types 3 or 4) subordinate clauses. Consequently, the opposition between d- and so-correlatives can only be observed in relation
with unbound type 3 (argumentative) subordinate clauses.
5.1 S o-Correlative Grammatical Category Markers

and unbound type 3 subordinate clauses
The so-CGCM can be found immediately after the following unbound subordinate
clauses:
Concessive clauses introduced by obwohl although ((8), (23)), or wenn auch even
though (24).
Hypothetical clauses introduced by wenn (25), (26) or falls (27) if, in case.
Colette Corts
(23) Obwohl er ein begnadeter Lyriker war und blieb, so gibt es doch im Sptwerk
Ausrutscher, die der junge Brecht, mit seinem untrglichen Gespr fr das
mot juste und den richtigen Vers, spielend vermieden htte. Z98/802.00733
Die Zeit, 05.02.1998, Nr. 07, Ressort: Feuilleton; Mann ist Mann, S. 41
(Even though Brecht was and remained a gifted lyric writer, (so) however
there is in his later work some negligence (of style) which he would easily have
avoided as a young man with his unfailing sense of the mot juste (right word)
and the right verse.)
(24) Und wenn es auch tricht wre zu glauben, da Demokratien keinen Krieg
miteinander fhrten, so steht doch fest, da sie sich im allgemeinen sehr viel langsamer auf Kriege einlassen als Diktaturen. (Z96/601.00303 Die Zeit, 12.01.1996,
Nr. 03, Ressort: Politik; Seid nicht so stur) (And even though it would be foolish
to believe that war would never break out between democratic states, (so) one
thing is however clear: (that) they would generally embark upon a war much
more slowly than dictatorships).
(25) Das jedenfalls scheint sicher: Wenn es Tollheit ist, so hats doch Methode wie
Polonius im Hamlet sagt. Spiegel 2000, n0 41, p.244 (One thing seems to be
sure: Though this be madness, yet there is method in it. as Polonius says in
Hamlet (Act II, scene 2, l., 203))
(26) Wenn es um einen Grundsatz geht, so ist der schon lange entschieden. (),
und in Grobritannien vor allem England pat nicht in ein vereintes Europa.
(Z95/507.03716 Die Zeit, 07.07.1995, Nr. 28, Ressort: Modernes Leben; Ein Tag
im Leben einer Insel) (If the debate is on principles, (so) things have been clear
for a long time. (), and within Great Britain, England in particular does not
belong to a unified Europe.)
(27) Gegenber Scharons Vorgnger Barak mag er sich gedacht haben, dass eine Show
des Widerstands nicht schaden knnte (). Nur: Falls dies das Kalkl war, so war
es eine Fehlkalkulation. Denn die Situation entwickelte eine eigene Dynamik und
wurde von den Israelis fr ihre Zwecke ausgenutzt. (Z02/204.02337 Die Zeit
(Online-Ausgabe), 25.04.2002, Nr. 18, Ressort: Politik; Frieden ist noch immer
mglich, S. 6) (Unlike Sharons predecessor Barak, he might have thought that
there would be no harm in showing some resistance () But: if this was the
plan, (so) it was a failed calculation, because the situation developed its (specific)
dynamics and was used by the Israelis for their own purposes.)
After a concessive subordinate clause introduced by obwohl, obgleich, obschon

although or wenn auch even if, even though, so is the only possible CGCM. Consequently, the contrast between so and dann cannot be observed in a concessive, but
only in a hypothetical context. Let us compare wenn so in (25), (26) with wenn
dann in (21) and falls so in (27) with falls dann in (22), and try to give a semantic
interpretation of the contrast between the CGCMs so and dann.
The hypothetical sentence is composed of two parts: the protasis (the initial
subordinate clause in Examples (25), (26) and (27)) which functions as the first proposition of a hypothetico-deductive period (i.e. as the foundation for the hypothetical
demonstration), and the apodosis presenting the conclusion built upon the protasis
in the main part of the sentence. To account for the contrast between the dann- and the
so-CGCM, we posit that German distinguishes two kinds of hypothetico-deductive
complex sentences corresponding to two different conceptions of the relation between
the protasis and the conclusion. We interpret this as the contrast between the factual and the processual relation to the protasis:
When using the dann-CGCM after a hypothetical subordinate clause introduced

by wenn or falls, the speaker builds a factual relation to the protasis, which
itself is considered as a factual component of the background information, as
a datum. A working hypothesis is thus suggested about a fact which may or
may not be true, but which is provisionally admitted by the speaker and used as
a foundation, for the sake of the deduction in the frame of the hypothetical argumentation. The use of the determinative dann-CGCM attests to the (at least provisory) commitment of the speaker, who presents the content of the subordinate
clause as a reliable fact for the sake of the cause-effect demonstration.
On the contrary, by using the so-CGCM, the speaker builds a processual relation
to the protasis on a different level from the cause-effect relation: what is at stake
in this case is the relevance of the protasis which induces the speaker to express
a parallel or usually contradictory statement contained in the main clause. This
explains why, in such cases, the CGCM so is often accompanied by the assertive
opposition marker doch yet, nevertheless, however (see (25)), and the speakers
distanciation is so obvious that it allows some polyphonic6 interpretation of the protasis (25), or even a rectification of the protasis through the apodosis (27). Using the
so-CGCM, the speaker tests and assesses the reliability and relevance of the protasis and confronts it with another statement issued in the apodosis, which could be
well-founded in the context of the protasis; the complex sentence marked by the
so-CGCM allows the evaluation and weighting of two compared assessments.
The contrast between dann and so as CGCMs in a hypothetical context can be understood as an opposition between a linear cause and effect deduction, from a fact to
a conclusion with dann, and a non-linear evaluative process with so, as Angelika
Redder (1987: 323324) writes: Die Nicht-Linearitt der Ausfhrungen durch so
(steht) in Opposition zur Linearitt der Ausfhrungen durch dann. With the CGCM
6. i.e. presented as possibly assessed by a plurality of potential speakers.
Colette Corts
dann, argument and conclusion are presented as co-oriented in a then/therefore- cause

and effect argumentation grounded in a causal chain of facts, while the CGCM so
refers to an evaluation process, assessed by the speaker, and which may be used in an
anti-oriented (concessive) however-context. The influence of the CGCM on the global
interpretation of the sentence is thus very important. The type 3 so-CGCM, like the
type 3 dann-CGCM, is a connective sentential marker, but it suggests another cognitive frame for the argumentative sentence construction: the argumentative frame of the
dann-CGCM is a linear hypothetico-deductive process based on a one-way-oriented
chain of facts, while the argumentative frame of the so-CGCM is an evaluation process
taking into account co-oriented as well as anti-oriented viewpoints. The so-CGCM
and the dann-CGCM are both sentential connective markers: while the dann-CGCM
connects facts, the so-CGCM connects viewpoints.
5.2 S o-Correlative Grammatical Category Markers

and type 4 subordinate clauses
Type 4 subordinate clauses have a metalinguistic function, which is heterogeneous to
the purpose of the sentence, but assessed by the speaker. d-CGCMs are never found in
type 4 subordinate clauses, only the CGCM so appears in a few of them (28).
(28) Zudem ist wissenschaftliches Wissen von Beginn an kumulativ, es baut
aufeinander auf. Die Relativitt ist also zwar eine Grundbedingung des
wissenschaftlichen Denkens, aber sie muss umgesetzt werden in systematisches
wissenschaftliches Arbeiten. Verifizierung und Falsifizierung sind daher nur selten
absolut, aber was Wahrheit angeht, so sind sie annherungsfhig. (Z02/209.04479
Die Zeit (Online-Ausgabe), 12.09.2002, Nr. 38, Ressort: Wissen; Das Wissen der
Welt, S. 34) (Scientific knowledge is from the beginning the result of a cumulative
build-up process. Relativism is then certainly a condition for scientific thinking,
but it has to be adapted to the needs of systematic scientific work. Verification
and falsification are therefore rarely absolute, but as far as truth is concerned (so)
it is appropriate that they come closer to it.)
In example (28), the subordinate clause announces a new topic or a new viewpoint. The CGCM so has focusing function, and plays a part on information and
text structure.
The so-CGCMs are only used as sentential markers, never as demarcative phrasal
markers. They focus on the argumentative relation between the Type 3 Unbound
Subordinate Clause and the interpretation of the whole sentence or on the relation
between the complex sentence and its communicative frame for the type 4 Unbound
Subordinate Comment. From a semantic and pragmatic point of view, they can be
considered connective markers because they play a significant role in discourse and
textual construction: they mostly appear in argumentative texts where they control the
progression of the text and its anchoring in the communicative context.
6. Conclusion
The correlative markers in German are grammatical category markers, i.e. horematic categories in Fourquets model, overarching the subordinate clause or
focusing its relation to the interpretation of the whole sentence, and enabling its cognitive and pragmatic instantiation in the discourse and communicative context. They
may have scope over either a phrasal constituent or the whole sentence, thus leading
to a distinction between Phrasal CGCMs and Sentential CGCMs.
Phrasal CGCMs are determinative, demarcative, phoric pro-elements, focusing

on or referring to a bound subordinate clause, with which they form a syntactic,
semantic and pragmatic unit (d-correlatives in class 1A, 2A).
Sentential CGCMs are discontinuous (es in 1A, es and d-correlatives in 1B,
d-correlatives in 2B) or connective (with unbound subordinate clauses 3 and 4),
phoric pro-elements, referring to the relation between the subordinate clause and
the complex sentence as a whole.
The semantic and pragmatic functions of the CGCMs vary with the type of subordinate clause and the type of complex sentence, as shown in Table3:
Table3.
es CGCM
d- CGCM
Type 1 A Bound
Subordinate Clause
(actantial)
Discontinuous
Sentential CGCM
(3, 12)
Type 1 B Bound
Subordinate Clause
(actantial)
Type 2 A Bound
Subordinate Clause
(circumstantial)
Discontinuous
Sentential CGCM
(17)
d-Demarcative
Phrasal CGCM
(pre- (4, 15) or
postdeterminer (13,
14)) (das/dar +Prep)
Discontinuous
Sentential postposed
CGCM (das) (16)
d-Demarcative
Phrasal CGCM
(pre- (5, 6) or
postdeterminer (18))
(das/dar +Prep)
Discontinuous
Sentential postposed
CGCM (7, 19, 20)
Connective Sentential Connective Sentential
postposed CGCM
postposed CGCM
(21, 22)
(8, 23, 24, 25, 26, 27)
Type 2 B Bound
Subordinate Clause
(circumstantial)
Type 3 Unbound
Subordinate Clause
(argumentative)
Type 4 Unbound
Subordinate Clause
(commentative)
so- CGCM
Connective Sentential
postposed CGCM
(11, 28)
Colette Corts
The complex sentence is thus not the mere concatenation of two simple sentences,
it is the result of intricate constructions which must be considered from a holistic viewpoint, because the whole sentence is more that the sum of its parts, and the meaning
added is mainly contained in the binding determinative markers and assertive markers
which build interdependent relations and set the sentence within a bigger pragmatic and
textual whole. The CGCMs are part of Fourquets categories, they are Grammatical
Category Markers allowing the speaker to control the discourse construction step by
step (i.e. phrase after phrase, and from phrase to sentence and text levels) in accordance
with the communicative intentions; determination markers project a determinative
mapping onto various levels, from the phrasal to the sentential level.
In contrast with their treatment in most German grammars, the CGCM are not
superficial and optional markers; they are fundamental (deep structure) markers of the
subordinate clause and of its relation to the complex sentence, as pointed out by historical linguists like Jean Fourquet (1970), Andr Rousseau (1984), Claude Muller (1996)
or Anne Daladier (2002). Modern CGCMs offer (lacunary) traces of a very ancient
system of correlative subordination markers, a situation shared by Romance and
Germanic languages at least, (and, according to Daladier, with most Indo-European
languages). They should be dealt with in terms of deep structure syntax and pragmatics, although they do not appear in all complex sentences, for one good reason: just as
not all noun phases are headed by a definite determiner, not all complex sentences are
headed by a determinative CGCM.
An exact understanding of CGCMs rests on the analysis of all their possible combinations with subordinate clauses, and with all possible subordinating conjunctions,
but this would take us far beyond the limits of this paper. Let us simply suggest the
three following points:
The CGCMs belong to the complex sentence and rest on a posited original deep
structure for all types of complex sentences and thus for all types of correlative uses.
The German complex sentence is marked by two types of correlative markers and
subordinating conjunctions: d- correlative markers (for factual categorisation)
and so- correlative markers (for processual categorisation). The correlative markers share the function of the grammatical category markers of modern German.
The German Correlative Grammatical Category Markers are chameleon-like markers: being pro-elements, without any lexical content, they may refer to and focalise
any kind of link established for the sake of discourse and textual construction.
References
Blhdorn, Hardarik. 1993. Deixis und Deiktika in der deutschen Gegenwartssprache. Deutsche
Sprache 1: 4462.
Blhdorn, Hardarik. 1996. Was ist Deixis? Linguistische Berichte. 156: 109142.
Corts, Colette & Rousseau, Andr. 1999. Catgories et connexions. En hommage Jean Fourquet
pour son centime anniversaire, le 23 Juin 1999. Villeneuve dAscq: Presses universitaires du
Septentrion.
Corts, Colette. 2006. Subordination et corrlation en allemand. In Coordination et subordination: Typologie et modlisation [Faits de Langues 28], Isabelle Bril & Georges Rebuschi
(eds), 107118. Paris: Ophrys.
Curme, George O. 1904. A Grammar of the German Language, rev. edn 1922, 10th printing
1970.
Daladier, Anne. 2002. Definiteness in Amwi: Grammaticalization and syntax. In Syntaxe de la
dfinitude [Recherches linguistiques de Vincennes 31], Anne Zribi-Hertz & Anne Daladier (eds), 6178. Presses universitaires de Vincennes.
Descls, Jean-Pierre. 1992. Au sujet des catgories grammaticales. In La thorie dAntoine Culioli,
Ouvertures et incidences, 203212. Paris: Ophrys.
Ducrot, Oswald. 1972. Dire et ne pas dire. Principes de smantique linguistique [Collection
Savoir]. Paris: Hermann.
Eisenberg, Peter. 1994. Grundriss der deutschen Grammatik, 3rd edn. Stuttgart: Metzler.
Fabricius-Hansen, Cathrine. 1981. Was ist nun wieder ein Korrelat? Gedanken zur Rehabilitierung eines naiven Nebensatzbegriffs. Kopenhagener Beitrge zur Germanistischen
Linguistik 18: 145.
Foley, William & Van Valin Jr, Robert D. 1984. Functional Syntax and Universal Grammar.
Cambridge: CUP.
Fourquet, Jean, 1970. Prolegomena zu einer deutschen Grammatik. Pdagogischer Verlag
Schwann Dsseldorf.
Grnig, Blanche-Nolle. 1963. tude des dmarcatifs en allemand moderne. Thse de 3e cycle.
Paris Sorbonne.
Muller, Claude. 1996. La subordination en franais. Le schme corrlatif. Paris: Armand Colin,
Masson.
Pasch, Renate, Brausse Ursula, Breindl, Eva, Wassner, Ulrich Hermann. 2003. Handbuch der
Konnektoren: Linguistische Grundlagen der Beschreibung und syntaktische Merkmale der
deutschen Satzknpfer (Konjunktionen, Satzadverbien und Partikeln). Berlin: Walter de
Gruyter.
Redder, Angelika. 1987. Wenn, so. Zur Korrelatfunktion von so. Sprache und Pragmatik, 315326.
Inger Rosengren (Hrgb). Lunder germanistische Forschungen Bd 55. Almqvist &Wiksell
International Stockholm Sweden.
Rousseau, Andr. 1984. Apparition et grammaticalisation des formes verbales priphrastiques
en germanique ancien. In Romanistique-Germanistique, 97130. Strassbourg: Presses
Universitaires de Strasbourg.
Sonnenberg, Bernhard. 1992. Korrelate im Deutschen: Beschreibung, Geschichte und Grammatiktheorie [Reihe Germanistische Linguistik 124]. Tbingen: Max Niemeyer Verlag.
Tesnire, Lucien. 1959. Elments de syntaxe structurale. Prface de Jean Fourquet. Deuxime
dition revue et corrige. Cinquime tirage. Paris: Klincksieck [1988].
Zifonun, Gisela, Hoffmann, Ludger & Strecker, Bruno. 1997. Grammatik der deutschen Sprache
[Schriften des Instituts fr deutsche Sprache]. Berlin: Walter de Gruyter.
part iv
Informational hierarchy and TAM markers

functions in clause-linkage
Focus, mood and clause linkage in Umpithamu

(Cape York Peninsula, Australia)
Fund for Scientific Research-Flanders & University of Leuven

Umpithamu is poor in markers that specifically encode clause linkage, but it relies
on markers from other domains, specifically mood and information structure.
Such markers can contribute to clause linkage in three semiotically distinct ways.
The first is compositional encoding, as with the purposive relation that is encoded
by the combination of potential mood with a schema of argument sharing. The
second is inference: with the conditional relation, potential mood provides a
feature of non-actuality, and the causal relation between the propositions is left to
inference. The third is encoded inference: with explanatory relations, use of a
focus marker invokes a presupposition, which when not found in the context
forces the inference of an explanatory relation with the preceding clause.
1. Introduction1
In this paper, I will analyse the mechanisms available for linking clauses in Umpithamu,
a Paman language of Cape York Peninsula (Australia). Like many other Australian languages, Umpithamu is interesting for the typology of clause linkage because the language
has few, if any, elements that are specialized in marking semantic relations between
1. I would like to thank the late Mrs Florrie Bassani and Ms Joan Liddy, who so patiently
taught me their language, and Bruce Rigsby, who introduced me to the Lamalama people, and
generously shared his earlier recordings of Umpithamu and his vast knowledge of Princess
Charlotte Bay languages. Fieldwork on Umpithamu was sponsored by the Fund for Scientific
Research-Flanders, which funded two research stays at the University of Melbourne, by the
Australian Department of Communication, IT and the Arts (DCITA), which funded another
fieldtrip, and by the Endangered Languages Documentation Program (MDP0133), which funded
further fieldtrips and equipment. Many thanks to Nick Evans and the Department of Linguistics
in Melbourne for hosting my stays, and to Clair Hill for organizing the DCITA grant. An earlier
version of this paper was presented at the seminar on Deixis, focus, topic and subordination, at
LACITO in Paris. I am grateful to Isabelle Bril, Alexandre Franois and Stphane Robert for very
insightful comments which helped me to improve the argument. I would also like to thank two
reviewers and the editor, for very detailed and helpful comments on an earlier draft.
clauses. Instead, the language has a number of markers from other domains, like mood
and information structure, that can be recruited to signal specific types of semantic relations. From a typological perspective, the analysis of Umpithamu shows how languages
without specialized markers for clause linkage can get by: if there is a need to be more
explicit about the interpretation of an interclausal relation than is possible by simple
inference from apposition or other general syntactic mechanisms, speakers can resort
to markers outside the domain of clause linkage. From a semiotic perspective, however,
the analysis of Umpithamu also shows that this type of mechanism is not uniform. I
will argue that markers from outside the domain of clause linkage can contribute to
the signalling of interclausal relations in at least two ways. In what I will call a compositional strategy, the marker serves to provide semantic specification in a more general
syntactic schema of clause linkage, and thus can be said to encode the interclausal relation jointly with this syntactic schema, in a compositional way. In what I will call an
inferential strategy, by contrast, the marker does not contribute to the encoding of an
interclausal relation, but merely provides the basic semantic prerequisite for it, which
must be enriched by inferential strategies to arrive at the interpretation of the relation. I
will argue that such an inferential strategy can be either purely inferential, based entirely
on world and discourse knowledge, or the result of an encoded inference (comparable to
Blakemores [1987, 2002] notion of procedural meaning), with the search for an inference (though not the inference itself) triggered by the use of the marker.
As background to the discussion on clause linkage, I will first provide some basic
information on Umpithamus structure, particularly the features that will be relevant
to clause linkage. Umpithamu is a language that belongs to an area on the east coast
of Cape York Peninsula, around Princess Charlotte Bay. Rigsby (1992) and Rigsby &
Chase (1998) provide more detailed information on the language and its speakers.
Genetically, the language is a Paman language, as defined by Hale (1964, 1966), and
can most likely be subgrouped with Middle Paman languages (see Evans 2005 for
some published evidence based on my own and Rigsbys work). Structurally, the language is of the split ergative type, like all of its close relatives, with nominals showing
ergative-absolutive alignment, and pronouns showing nominative-accusative alignment,
as illustrated in (1), which contains both an ergatively marked nominal argument, and a
pronoun complex consisting of a nominative and an accusative form.
(1) wantha-mpal warrki-ngka=iluwa-athungku
ignor-erg
follow-prs=3sg.nom-1sg.acc
Who is it that calls me (friend)?2
2. The following abbreviations are used in the glosses: abl ablative, acc accusative, appr
apprehensive, dat dative, erg ergative, exc exclusive, gen genitive, ignor ignorative, imp imperative, inc inclusive, intens intensifier, loc locative, neg negative, nom nominative, pot
potential, prs present, pst past, sub subordinator.
Focus, mood and clause linkage in Umpithamu (Cape York Peninsula, Australia)
There are two features, however, that set it apart from the standard type of split
ergative language found in the area. First, unlike its closest relatives, Umpithamu
has an emergent system of pronominal cross-reference, with a nominative-accusative
type of alignment. Nominal arguments can be cross-referenced by pronouns, most
often for subjects, as illustrated by the nominative cross-reference of the subject
wanthampal in (1), and less often for objects, as illustrated by the accusative crossreference of the object ama nhunha in (2). Pronouns are bound to each other in a
pronoun complex, which is encliticized to the inflected verb in default contexts, as
in (1) and (2), or can occur in clause-initial position in contexts of local prominence,
as in (3).
(2) ama nhunha ama kali-n=iluwa-inangku
person other person carry-pst=3sg.nom-3pl.acc
She brought other people.
(3) iluwa-athungku onongkol uwa-n
3sg.nom-1sg.acc one
give-pst
He gave me one.
Second, unlike the classic ergative language, Umpithamu does not use the ergative
marker for all (animate) transitive subjects, but only for those that are locally prominent (see further Verstraete [2009] for more details on Umpithamu, and McGregor
[2006] for a more general discussion of this type of system). This is illustrated in the
contrast between (4) and (5), both of which have a nominal as their transitive subject,
but only one of which uses the ergative marker. It should be noted that the prominence context that triggers ergative marking is the same context as the one that triggers clause-initial position for pronouns, as in (3). This will be discussed in more detail
in Section4, where I will show that prominence marking is one of the means that can
be exploited to signal interclausal relations.
(4) nhunha-mpal watyu-n=iluwa
other-erg
spear-pst=3sg.nom
Another one speared it (the crocodile).
(5) omoro watyu-n=ilu-ungku
minya minya
father spear-pst=3sg.nom-3sg.acc meat meat
Dad speared the animal (the turtle).
The verb and the pronouns together serve as the basic reference points for the rest of
the clause structure in Umpithamu. Tense and mood markers are suffixed to the verb,
with pronouns encliticized to inflected verb forms in the default context, as in (1), (2),
(4) and (5), and in clause-initial position in focus contexts, as in (3). Polarity markers
form a paradigm of elements located immediately before the verb, as illustrated with
the standard negative form alu in (6). Nominal arguments are predominantly located
before the verb, in subject-object order in rare cases when there is more than one
nominal argument. In addition, nominal arguments can also be moved to follow the
verb, in contexts of emphasis, as is the case with minya in (5) above (where the repetition of minya adds further emphasis).
(6) mayi alu atha-n=iluwa
food neg eat-pst=3sg.nom
She didnt eat food.
Taken together, these positional restrictions produce the following basic clause template
(Figure1), which is centered around the verb. Shaded linked areas represent alternative information-structural choices, both for pronouns and for nominal arguments.
Figure1. Template of Umpithamu clause structure
2. Clause-linking elements
If we look beyond the structure of the simple clause, Umpithamu is remarkably poor
in elements whose only or primary function is to mark semantic relations between
clauses. I take the basic criterion for this type of element to be a distributional restriction to clause linkage contexts, or at least a distributional preference for clause linkage
contexts over simple clauses. On the basis of this criterion, there is really only one
element that could be regarded as a candidate for a specialized clause linkage marker,
viz. the suffix -na on the verb, and even here the arguments in favour of this analysis
are not entirely convincing.
Basically, the suffix is found in two types of contexts, formally distinguished by the
presence of the ablative case on the verb. The first context, where the -na-marked verb
is followed by the ablative case form mun, is illustrated in (7), (8) and (9). In these
contexts, the structure can be called subordinate because the verb does not have any
of the properties of normal main verbs in Umpithamu. It does not take tense, it has no
cross-referencing pronouns, and it takes case marking, which is normally restricted
to nominals. In addition, the arguments of the subordinate verb can also inherit its
case form, as with the object ngoki in (9), which also receives the ablative case that is
marked on the verb form athana-.
(7) uwi-n=ilu-ungku
yoompi-na-mun / ngampu
find-pst=3sg.nom-3sg.acc stand-sub-abl / brown.snake
We found it standing (there), a brown snake.
(8) motoka ngama-n=antyampa yongki-na-mun

car
see-pst=1pl.exc.nom come-sub-abl
We saw the car coming.
(9) ngoki-mun atha-na-mun wiinka-n=ilu-ungku
water-abl drink-sub-abl scratch-pst=3sg.nom-3sg.acc
She scratched him because she drank too much (alcohol).
The type of subordination involved here is a bit harder to determine, as there seem to
be two distinct types. In both types, the main clause and the subordinate clause share
an argument, which is the subject or object of the main verb and the subject of the subordinate verb, and which is marked for its function in the main clause. Furthermore,
the case used to mark the subordinate verb is not a core case but an ablative marker,
normally used to mark spatial origin. Taken together, these features imply that the
subordinate clause cannot be functioning as an argument of the main verb, and thus
is not a complement clause. Instead, the subordinate clause seems to function in one
of two ways. It can either function as an adverbial clause, as in (9), in which case the
ablative is semantically transparent, because the causal meaning associated with the
subordinate clause can be regarded as a semantic extension of its basic spatial meaning of spatial origin. Or it can function as a modifier of one of the arguments of the
main clause, as in (7) and (8), describing a feature of the argument as it is involved
in the action described by the main clause. Unlike with adverbial interpretation, in
this structure the ablative case is not semantically transparent, because its use in NPmodifying function cannot easily be linked with its basic spatial meaning of spatial
origin. Although the use of the ablative is surprising in NP-modifying function, the
conflation of adverbial subordination and NP-modifying subordination in one single
construction type is not unknown in Australian languages, as first described by Hale
(1976) for Warlpiri (see also Nordlinger 2006 for a recent assessment). The lack of any
clear morphosyntactic distinction between adverbial uses like (9) and NP-modifying
uses like (7) and (8) suggests that Umpithamu is similar in this regard.
The second context in which -na is used is in combination with posture verbs, to
describe the activity the subject of the posture verb is engaged in, as in (10) and (11). At
first sight, this use seems to be formally subordinate just like in (7)(9): the verb does
not take tense, and does not have pronominal cross-reference. The -na-marked verb
shares its subject argument with the main verb, and this shared argument is marked
pronominally only on the main verb. In this sense, the construction is formally parallel
to the other one, and the -na-marked verb could be regarded as a secondary predicate
to the subject of the main verb. There are two basic morphosyntactic differences, however: in the structures in (10) and (11), the verb does not take case, and the main verb
does not have any arguments independent from the subordinate verb. The lack of case
marking could in theory be reconciled with a secondary predicate analysis, because in
this case the controller is an intransitive subject. The lack of separate arguments for the
main verb, however, points towards a semantic difference: semantically, this seems to
imply that the -na-marked verb is the main predicate, and that the posture verb is an
auxiliary element with an aspectual function. Newman (2002) provides a discussion of
cross-linguistic parallels.
(10) atha-na niina-n=ayu
eat-sub sit-pst=1sg.nom
I was eating.
(11) minya yutya-na niina-n=iluwa
meat cut-sub sit-pst=3sg.nom
She was cutting fish.
Because of its use in structures like (10) and (11), even the -na-suffix is not a great candidate for a specialized clause linkage marker: from a synchronic perspective, (10) and
(11) are not biclausal at all, but form an aspectual construction with a predominantly
monoclausal interpretation. In this sense, the only real candidate for a dedicated clause
linkage marker is not the suffix itself, but rather the general syntactic mechanism of
argument sharing, and the lack of pronominal cross-reference that is associated with
the use of -na.
How, then, does a language like Umpithamu manage to specify semantic relations between clauses? In the following sections, I will focus on two categories that are
distributionally outside the domain of clause linkage, but can be exploited to specify
semantic relations between clauses, in conjunction with the argument sharing schema
or general inferential mechanisms. In Section3, I will focus on mood marking, and in
Section4, I will explore case and focus marking.
3. Mood and clause linkage in Umpithamu

The basic function of mood elements is to mark the reality status of a proposition, e.g.
whether its occurrence is actual or just potential. Given this function, it is not surprising that mood elements can be recruited for purposes of clause linkage, specifically
those types of clause linkage that involve non-actual events, like purpose and condition. In Umpithamu, the basic structure of the clause has two slots where mood can be
marked: (i) in the paradigm of verbal suffixes, where basic tense and mood marking is
found, and (ii) in the paradigm of polarity markers located right before the verb, which
has some extensions into the domain of mood. These will be discussed in the following
two sections. In Section3.1, I will analyse the role of potential verb suffixes in purpose
and conditional relations, and in Section3.2, I will analyse the role of the apprehensive
marker in negative purposive relations.
3.1 Mood in verb suffixes

The verb in Umpithamu shows a basic contrast between five suffixes to mark temporal
and modal types of meaning, all in the same morphological slot: -n marks past tense,
and is also used as the citation form, -ngka marks present tense, -ku marks potential
mood, -rra marks counterfactual mood and -l marks the imperative.
The suffix that is most relevant to clause linkage is the potential suffix. This suffix
can have plain future uses in the temporal domain, but it is mainly found with modal
uses, marking events as potential rather than actual. Potentiality can be attributed to
various sources, which correspond to the major subtypes of modal meaning (see further Verstraete 2005). Potentiality can be attributed to the speakers desires (deontic
modality), as in (12)(13), where the speaker represents an event as desirable, or to the
speakers fears (apprehensive modality), as in (14), where the speaker represents it as
undesirable, or to a clause participants abilities or intentions (dynamic modality), as
in (15), where the clausal subject is unable to realize the action.
(12) iila,
oomolo
muunti-ku=aliya
younger.sister barramundi3 swim-pot=1duinc.nom
Sister, lets dive.
(13) ungkarra iya-ku=uurra
north
go-pot=2pl.nom
You have to go north.
(14) antyala ayngkini-l
neg.imp send-imp
mathi-ku=ilu
yenu / yuku-mpal yenu
climb-pot=3sg.nom up / tree-loc up
Dont let it go. It might climb up, on the tree.
(15) alu kali-ku=ayu-ungku
neg carry-pot=1sg.nom-3sg.acc
I cant take him.
This mood marker is used for two distinct types of clause linkage, associated with
different configurations of mood marking and pronominal cross-reference. The first
type is a conditional relation, illustrated in (16)(17). When two potential-marked
propositions occur in sequence, they are typically interpreted as being in a conditional relation, i.e. the speakers death as a condition for her childrens inheritance in
(16), and approaching the Story Being (known as a Dreamtime Being in other parts
of Australia) on ones own as a condition for not being attacked by it in (17). From the
3. Oomolo muunti- barramundi swim is an idiom that means to dive.
erspective of clause-internal morphosyntax, however, there is nothing in the strucp

ture of the clauses in (16) and (17) to distinguish them from independent clauses with
potential mood. Each has its own pronominal cross-reference, and there is no other
evidence of subordination as found in structures like (7)(9) in the previous section.
Semantically, therefore, these structures could just as well be interpreted as two separate potential-marked statements, e.g. I might die, the children might get it. for (16).
(16) ayuwa wuypu-ku nayngka maarra-ku=ina
1sg.nom die-pot child
get-pot=3pl.nom
If I die the children will get (it).
(17) ama-kin
iya-ku=ilu
person-priv go-pot=3sg.nom
alu watyu-ku=ilu-ungku
kayku
neg spear-pot=3sg.nom-3sg.acc head
If he goes alone, it (Eaglehawk) wont peck him on the head.
The second type is a purposive relation, illustrated in (18)(19). Unlike with conditionals, only one of the propositions has to be marked for potential mood, and it is
this proposition that receives the purposive interpretation. Thus, sitting in court is
interpreted as the purpose of going up to the community house in (18), and dancing
is interpreted as the purpose of going to the festival in (19). Apart from the restrictions on mood marking, there are two further differences between the purposive and
the conditional types. With the purposive type, the potential-marked verb shares its
subject argument with the other clause, and does not have cross-referencing pronouns.
With the conditional type, by contrast, both mood-marked verbs have their own
cross-referencing pronouns, and need not share any arguments. These formal differences allow us to distinguish purposive instances like (19) from conditional ones like
(16)(17): even if we have a sequence of potential-marked verbs in both structures,
the lack of cross-reference on ayparraku in (19) tells us we are dealing with a purpose
interpretation rather than a conditional one.
(18) S iya-ngka=iluwa uukul niina-ku
S go-prs=3sg.nom court sit-pot
S [name] is going to sit in court.
(19) wian
ayparra-ku iya-ku=ina
upper.leg play-pot go-pot=3pl.nom
They will go to dance.
Given the meaning of the potential suffix in simple clauses, the basic question is
exactly how the semantic feature it encodes contributes to the final clause linkage
interpretation in structures like (16)(19). The feature of potentiality is a necessary
condition both for purposive and for conditional interpretations, because they deal
with unrealized situations (see, for instance, Hengeveld 1998:349351), but in neither
case is it a sufficient condition. In purposive structures, the crucial difference is that
the judgement of potentiality does not belong to the perspective of the speaker, as
it would in an independent clause, but to the perspective of the subject of the main
clause (see further in Hengeveld 1998:350; Verstraete 2008). Thus, in the purposive
interpretation of a structure like (18), sitting in court is not the speakers desire (He
goes, I intend [he sit in court]), but that of the subject of the main clause (He goes,
he intends [he sit in court]). In conditional constructions, the crucial difference is
that the two potential events are not independent, but are interpreted as being causally
linked to each other. Thus, in the conditional interpretation of a structure like (17),
the two potential-marked clauses are not just statements about potential events (He
might go alone, Eaglehawk might not peck him on the head), but the interpretation
also involves a causal relation between these potential events (He might go alone, and
because of this, Eaglehawk might not peck him on the head). For both structures,
therefore, we can say that the potential suffix does not encode clause linkage the way
genuine clause linkers do, but merely contributes a necessary semantic feature to the
clause linkage interpretation.
From this perspective, the structural differences observed between the conditional
and the purposive types are interesting, because they illustrate two different ways in
which the potential suffix can be recruited to signal clause linkage. For the purposive
structure, what is required to link the meaning of the potential suffix to the final purposive interpretation is a feature that relates the potential-marked verb to the subject
of the main clause rather than to the speaker. There is, in fact, a good candidate for this
type of feature in the syntax of purposive structures. What distinguishes structures like
(18)(19) from a sequence of independent clauses is precisely the fact that the purposive clause shares its subject with the main clause and that the verb in the purposive
clause does not have any pronominal cross-reference. This is similar to the syntactic features of (7)(9) above, except for the fact that the potential-marked verb in purposive
structures does not take a nominal case form, and therefore cannot be regarded as subordinate.4 As shown in the previous section, the function of argument sharing and lack
of cross-reference is to link the subordinate clause to a specific argument in the main
clause. From this perspective, the potential suffix and the syntactic features of purpose
constructions could be said to encode the purposive interpretation compositionally.
The potential suffix contributes the semantic feature of potentiality, as it does in simple
4. Even though the potential suffix -ku is homophonous with the dative marker -ku, the
verbs in (18)(19) do not take the subordinator -na, and their arguments do not inherit the
-ku marker, which means that -ku in purposive structures cannot be analysed as a nominal
case marker in parallel with the ablative in subordinate structures like (7)(9).
clauses, and the general mechanism of argument sharing and lack of cross-reference
links this feature to the subject of the main clause rather than to the speaker.
In conditional structures, by contrast, it is not possible to derive the conditional
interpretation from the features of potentiality in a compositional way. The two clauses
are independent syntactically, which means that the basic interpretation of structures
like (16)(17) must be one of successive statements of potentiality: X might happen,
Y might happen. As already mentioned, this structure has the basic prerequisite for a
conditional interpretation, in the sense that the events described in conditional constructions are not actual but potential. What is required on top of this for a genuine
conditional interpretation is that the two events be causally linked, in the sense that the
happening of X enables the happening of Y. If we look beyond clausal morphosyntax,
prosody may actually contribute part of this link, but crucially it does not contribute
the causal feature. In the few text examples that could be checked for prosody, the first
clause shows a sharp clause-final rise in pitch, which in other contexts is associated
with a general meaning of there is more to come, for instance in multiple subsequent
renderings of an elicited item. In this sense, prosody marks the presence of a general
type of prospective link, but the meaning of this link is not specific enough to count
as a compositional contribution. Unlike in the purposive structure, therefore, there
is nothing in the conditional structure to encode the causal feature that is needed to
come to a conditional interpretation, which implies that it is inferred on the basis of
knowledge of the real world and the discourse world. Thus, the contribution of the
potential suffix in conditional structures is different from its contribution in purposive
structures. While the potential suffix helps to encode the purposive relation, in combination with the meanings encoded by other features of the construction, for the conditional relation it only provides a basic semantic prerequisite, which has to be further
enriched inferentially to arrive at the conditional interpretation.
3.2 Mood in the polarity paradigm

in addition to the verbal suffixes marking tense and mood, Umpithamu also has a paradigm of semantically related markers in the slot immediately before the verb, which
are partly relevant to mood. Basically, these elements serve to mark polarity, as with
the default negator alu in (20) for non-imperative clauses, and the specific negator
antyala in (21) for imperative clauses.
(20) alu errke-n=ilu-ungku
neg speak-pst=3sg.nom-3sg.acc
He didnt speak to him.
(21) antyala wurrpuntyi-l=inuwa
neg.imp jump-imp=2sg.nom
Dont you jump in.
Within this set of elements, there is one element that has modal functions, and that
also shows extensions to the domain of clause linkage. The marker atya, which always
occurs with a potential-marked verb, basically has an apprehensive function: it serves
to mark an event as potential but undesirable, as in (22) and (23) (see Lichtenberk
1995 on apprehensive modality in general). Pragmatically, apprehensive-marked
clauses often have the value of warnings, rather than being the subject of a general prohibition, particularly in those cases where the event described is about to occur, and
may thus function as an equivalent for negative imperatives, as in (23). This link with
polarity probably explains why the apprehensive marker belongs to the same paradigm
as genuine polarity markers.
(22) atya antyi-ku=inuwa
appr fall-pot=2sg.nom
You might fall.
(23) ngoki atya ngyampi-ku
water appr hit-pot
Dont hit the water.
As in many Australian languages, the apprehensive marker in Umpithamu also has

extensions in the domain of clause linkage, as the negative equivalent of purposive
constructions. When combined with a clause that could be interpreted as a precaution,
an apprehensive-marked clause can serve to describe the event that is to be avoided
through taking the precaution. Thus, for instance, in (24) the apprehensive marker
signals that the event marked with the apprehensive is the one to be avoided by taking
the advice in the main clause.
(24) kali-l
yinthu atya antyi-ku=iluwa
carry-imp down appr fall-pot=3sg.nom
Take him down, so he doesnt fall.
Even if the semantics of this construction links it to purposive structures, there are some
important structural differences. Unlike in purposive structures, the apprehensivemarked verb has its own pronominal cross-reference, including for the shared argument, and arguments shared between the two clauses need not have a subject role in
the main clause. Moreover, the structure is not available for describing precautionapprehension links for non-speech act participants: the precaution clause is typically
addressed to the second person, as an imperative. All of this implies that there is no
constructional feature linking the apprehensive-marked verb to an argument in the
main clause, and that structures like (24) should be regarded as consisting of two separate clauses, with the negative purposive relation inferred rather than encoded. In
this sense, a paraphrase with a separate apprehensive clause (e.g. Take him down, he
might fall) is actually more accurate than a negative purpose paraphrase with so that
not. Semantically, the contribution of the apprehensive marker to the negative purpose relation is similar to its contribution in conditionals: the marker does not encode
the negative purpose relation, on its own or jointly, but merely provides a semantic
prerequisite that is to be further enriched inferentially.
4. Information structure and clause linkage in Umpithamu

Information structure is a second domain that is relevant to clause linkage. In
Umpithamu, information structure is marked mainly in association with verbal
argument marking, through the position of pronouns for pronominally represented
arguments and the use of ergative case for nominally represented ones (as transitive
subjects). In this section, I will first provide a basic description of the system, and I will
then outline its relevance for the domain of clause linkage.
4.1 Prominent arguments

In its system of argument marking, Umpithamu allows a basic choice between an informationally neutral type of marking, and an informationally prominent one. The type of
prominence involved in Umpithamu is what I would like to call local prominence, i.e.
those cases where a specific verbal argument is prominent in relation to the immediately
preceding discourse. This does not relate to global measures of discursive prominence
like givenness or newness, but instead covers both what has traditionally been known
as focus (see, for instance, Jacobs 1984; Van Valin & LaPolla 1997) and what has been
known as contrastive topic (see, for instance, Bring 2003). The prominence associated
with focus refers to the fact that a particular argument X serves to fill out some variable
for a presupposition relating to the preceding discourse: It is X who. The prominence associated with contrastive topics refers to the fact that arguments X and Y serve
as different ways to fill out a variable within a larger set of referents Z relating to the
preceding discourse: Within the group of Z, X does A and Y does B. I am aware of
the controversies in the literature on information structure mainly based on English
about whether focus and contrastive topic should be regarded as forming one system
or not (see, for instance, Bring 2003), however, for Umpithamu at least there is no question that the two form a single unified system. The only formal difference depends on
whether the arguments are marked pronominally or nominally.
For pronominally marked arguments, the choice relates to the position of the
pronoun complex relative to the verb: the informationally neutral position is the one
encliticized to the inflected verb, while the prominent position is the clause-initial one.
Thus, for instance, when a contrast between participants from a larger set is involved, the
pronoun complex is found in clause-initial position, as in (25), which describes the distinct activities of two participants that have so far been acting together. A similar effect
is found in the elicitation of pronominal paradigms, where speakers typically construe
clauses with variations of the first elicited instance as being in contrast with it, as in (26).
(25) iluwa
iya-n=iluwa
aakurru ampanu-ku
3sg.nom go-pst=3sg.nom home own-dat
iluwa
ngaani yula-n=iluwa
3sg.nom ignor make-pst=3sg.nom
One (moon) went to his own home.
The other (wind) made something (a boomerang).
(26) watyu-ku=ayu-ungkuna
spear-pot=1sg.nom-2sg.acc
ayu-ipungku
watyu-ku
1sg.nom-2du.acc spear-pot
(Elicitation) Ill spear you. (Followed by) Ill spear you two.
For nominally marked arguments, the distinction between neutral and prominent
marking is available only for transitive subjects, through the use of the ergative marker
-mpal. As already mentioned, Umpithamu deviates from the classic ergative model
in that not all transitive subjects take the ergative marker: -mpal can be regarded as
ergative because it is distributionally restricted to transitive subjects, but its actual use
with ergative subjects is governed by factors of animacy and information structure.
This situation is known somewhat inaccurately as optional ergative marking in the
literature (see McGregor 1992, 2006 for an analysis of such systems in Australia and
beyond). For inanimate transitive subjects, ergative marking is obligatory, as shown
in (27). For animate transitive subjects, by contrast, ergative marking correlates with
information structure: the ergative marker is used only with prominent transitive subjects. Thus, we find the ergative marker in classic prominence contexts like contrast
sequences, as in (28), where the transitive subject contrasts with the expected transitive subject, and (29), where variation in elicitation is interpreted as contrasting, and
in the answers in question-answer pairs, as in (30).
(27) ngoki-mpal ungka-n=antyangana
water-erg wet-pst=1plexc.gen
The water made us wet.
(28)

[description of how someone gets a spear ready]

nhunha-mpal watyun=iluwa /
Norman
other-erg
spear-pst=3sg.nom Norman
Another one speared it, Norman.
(29) eelewen eentinti atha-n=ilu-ungku

dog
small bite-pst=3sg.nom-3sg.acc
yawul-mpal atha-n=iluwa
big-erg
bite-pst 3sg.nom
(Elicitation) The small dog bit him. (Followed by) The big dog bit him.
(30) yakaa, ngaani-mpal atha-n=athuna
aw
ignor-erg bite-pst=1sg.gen
minya-mpal atha-n=ilu-ungkuna/
koloyi
meat-erg bite-pst=3sg.nom-2sg.acc taipan
Aw, what is it that bit me? Its an animal that bit you, a taipan.
4.2 Prominence and clause linkage

Given that prominence is by definition a relational concept, its potential relevance for
clause linkage is obvious. Marking an argument as prominent in a particular clause
invokes a link with something beyond this clause, like a presupposition for focal
prominence, or a larger set of alternative referents for contrastive topic prominence.
None of these links is necessarily aimed at the immediately preceding clause, however:
they can relate to another speakers turns (e.g. in question-answer sequences), or to
slightly larger stretches of the speakers own turn (e.g. with some contrastive topics).
This means that prominence marking as such cannot be regarded as a specialized type
of clause linkage marker. Instead, it marks relations in discourse, which can, but need
not, be instantiated between individual clauses.
Nevertheless, there is at least one type of context where prominence marking does
invoke a relation with the immediately preceding clause, and where it can be regarded
as a mechanism for clause linkage rather than discourse linkage. Crucially, in these
contexts the interclausal relation invoked by prominence marking is different from
the discourse relation that is actually marked by it. In examples like (31) and (32),
for instance, the ergative marker invokes an explanatory relation to the immediately
preceding clause: in (31), the fact that the speaker is being taught by a prominent
speaker of the language serves to justify the fact that he is in the community to learn
the language, and in (32), the actions of the devils (who surreptitiously place objects
in peoples bodies) serve to explain why the interlocutor has something in his body.
(31) uuku
iminh ungku-ku=ayu
language intens hear-pot=1sg.nom
F-mpal ayngkiingkini-ngka=iluwa-athungku
F-erg teach-prs=3sg.nom-1sg.acc
I want to learn language well. F [name] is teaching me.
(32)

(English) you got something in your body

woympu-mpal ayngki-ngka=ina
devil-erg
throw-prs=3pl.nom
You have something (a bone) in your body. Its the devils who throw it.
How can these explanatory relations be reconciled with the basic definition of prominence marking, which is associated with discourse relations of contrast or presupposition? Unlike in (27)(30), in examples such as (31) and (32) there is no specific
presupposition in the preceding clauses to which the focal element can be related. As
argued by Robert (2000), however, the presupposition to which a focal element can be
related need not actually be made explicit in the preceding discourse. What matters
is that the use of a focus marker invokes such a presupposition, and that this forces
one to infer a relation with the preceding context. As shown by Robert, this schema
can cover both classic focus contexts, in which the invoked presupposition actually
belongs to the preceding context, and the instances mentioned in this section, in
which the invoked presupposition does not belong to the preceding context but is logically related to it. What is important for our analysis is that this proposal can explain
why the focus marker comes to invoke explanatory relations in instances like (31) and
(32). The presupposition is logically related to the preceding clause, and thus creates
the relation: in an example like (32), for instance, The devils [focal] throw it invokes
the presupposition that someone is responsible for the bones, which itself explains
why the interlocutor has these bones in his body. Similarly, in an example like (31), F
[focal] is teaching me invokes the presupposition that someone is actually teaching
the speaker the language, which explains the speakers presence in the community. In
this sense, the relation of the invoked presupposition to the actual preceding clause is
what creates the explanatory relation to this clause, and what indirectly associates the
prominence marker with this relation.
From the broader perspective of this paper, the mechanism for clause linkage illustrated in (31) and (32) is interesting because it illustrates a third way in which nonspecialized markers can contribute to clause linkage. This mechanism is different from
those observed for mood marking in the previous section, both in purposive and in
conditional structures. It is different from the purposive type in that the ergative marker
does not encode the argumentative relation, either by itself or compositionally with
something else. What the ergative marker encodes is the fact that the associated NP
functions as a transitive subject, and that this NP fills out a variable with respect to some
presupposition. The argumentative relation, by contrast, is a link between this presupposition and the preceding clause, and is itself a matter of inference. In this sense, it is
similar to the conditional type, which also relies on inference, but it is crucially different
on one point. Unlike in the conditional type, the inference here is actually triggered by
the use of the marker. If the presupposition invoked by the marker is not found in the
preceding clauses, interpreting the focus marker forces one to look for a link between
the presupposition and the preceding clauses. With conditionals, inferring what would
be required for a conditional interpretation is not enforced by the mere use of two successive verbs marked for potential mood, as it is for the ergative marker. Somewhat
paradoxically, therefore, in this particular case we could speak of an encoded inference:
even if the marker does not encode the explanatory relation as such, the fact that the
use of the marker forces one to look for an inference implies that it does encode the
presence of an inference. Although the notion of an encoded inference may sound like
a contradiction in terms, it is certainly not unknown in the study of interclausal relations. The distinction between conceptual meaning and procedural meaning in the
work of Blakemore (1987, 2002), for instance, tries to capture exactly this: linguistic
signs do not just encode conceptual information, but they can also encode inferential
procedures, in the form of instructions on how to constrain processes of inference.
5. Conclusions
To conclude, the system of clause linkage in Umpithamu distinguishes between four
semiotically distinct types of clause linkage. The central distinction made in this study
is between specialized markers, which are restricted to contexts of clause linkage and
can therefore be said to encode interclausal relations, and non-specialized markers
like potential mood and prominence marking, which are distributionally outside the
domain of clause linkage and therefore do not themselves encode interclausal relations. Such non-specialized markers can contribute to interclausal relations in various
ways: the basic distinction made in this study is between structures where the interclausal relation is encoded and structures where it is inferred. For encoded relations,
non-specialized markers can help to encode the relation compositionally, as in the
purposive structure, where the potential suffix contributes the feature of non-actuality,
while the general syntactic schema contributes the link to the subject of the main
clause, which together encode the purposive relation. For inferred relations, nonspecialized markers can lay the groundwork for the inference in two different ways.
One way is to provide a basic semantic prerequisite, which is then further enriched
by inference, as with conditional relations, where the potential suffix contributes the
feature of non-actuality for the two propositions, and the causal relation between them
is left to inference. Another way is to enforce the inference of an interclausal relation,
as with explanatory relations, where the use of the focus marker automatically invokes
a presupposition, and forces one to look for an explanatory link with the preceding
clause if this presupposition is not found in the preceding context (Robert 2000).
In this sense, a language like Umpithamu, which is at first sight relatively poor
in traditional means for clause linkage, nonetheless provides interesting insights
into the semantics and pragmatics of clause linkage, because it shows a wide variety of mechanisms for exploiting markers from outside the domain of clause linkage. Such mechanisms are not just interesting from a semiotic perspective, but they
also point towards potential paths of development between clause-internal functions
and clause linkage functions. If we look at other Australian languages, for instance,
the situation described in Umpithamu is not at all exceptional. Most languages have
few specialized markers of clause linkage, and have to rely heavily on elements from
other domains. The types of elements found are very similar to the ones discussed
in this study: mood markers, for instance, systematically play a role in purposive,
negative purposive and conditional constructions, compositionally or inferentially
(see Verstraete forthcoming). In this perspective, it is not surprising that there are
documented developments from mood marker to clause linkage marker and back,
and that in some instances it is impossible to decide between the two analyses (see
Evans 2007, Verstraete forthcoming). Heavy reliance on a clause-internal marker for
signalling an interclausal relation may create a path that leads from strategy to marker.
This type of situation is not unknown from work in grammaticalization, but what
a more detailed study of the semiotic mechanisms underlying the strategies could
show is that not all such paths are equally transparent and predictable from a semantic perspective. While changes based on compositional strategies can be expected to
be more or less predictable semantically, changes based on purely inferential ones
are probably far less regular, and may lead to surprising and unpredictable links, as
shown by Evanss (2007) analysis of diachronic links between conditional markers
and deontic mood markers rather than the semantically expected epistemic type.
References
Blakemore, Diane. 1987. Semantic Constraints on Relevance. Oxford: Blackwell.
Blakemore, Diane. 2002. Relevance and Linguistic Meaning. The Semantics and Pragmatics of
Discourse Markers. Cambridge: CUP.
Bring, Daniel. 2003. On D-Trees, beans, and B-accents. Linguistics & Philosophy 26: 511545.
Evans, Nicholas. 2005. Australian languages reconsidered. A review of Dixon (2002). Oceanic
Linguistics 44: 242286.
Evans, Nicholas. 2007. Insubordination and its uses. In Finiteness, Irina Nikolaeva (ed.),
Hale, Kenneth L. 1964. Classification of northern Paman languages, Cape York Peninsula,
Australia. A research report.Oceanic Linguistics 3: 248264.
Hale, Kenneth L. 1966. The Paman group of the Pama-Nyungan phylic family. Anthropological
Linguistics 8: 162197.
Hale, Kenneth L. 1976. The adjoined relative clause in Australia. In Grammatical Categories in
Australian Languages, Robert M.W. Dixon (ed.), 78105. Canberra: Australian Institute of
Aboriginal Studies.
Hengeveld, Kees. 1998. Adverbial clauses in the languages of Europe. In Adverbial Constructions
in the Languages of Europe, Johan van der Auwera (ed.), 335419. Berlin: Mouton.
Jacobs, Joachim. 1984. Funktionale Satzperspektive und Illokutionssemantik. Linguistische
Berichte 91: 2558.
Lichtenberk, Frantisek. 1995. Apprehensional epistemics. In Modality in Grammar and Discourse [Typological Studies in Language 32], Joan Bybee & Suzanne Fleischman (eds),
293327. Amsterdam: John Benjamins.
McGregor, William. 1992. The semantics of ergative marking in Gooniyandi. Linguistics 30:
275318.
McGregor, William. 2006. Focal and optional ergative marking in Warrwa (Kimberley, Western
Australia). Lingua 116: 393423.
Newman, John (ed.). 2002. The Linguistics of Sitting, Standing and Lying [Typological Studies in
Language 51], Amsterdam: John Benjamins.
Nordlinger, Rachel. 2006. Spearing the Emu drinking: subordination and the adjoined relative
clause in Wambaya. Australian Journal of Linguistics 26: 529.
Rigsby, Bruce. 1992. The languages of the Princess Charlotte Bay region. In The Language Game:
Papers in Memory of Donald C. Laycock, Thomas Edward Dutton, Malcolm Ross & Darrell
Tryon (eds), 353360. Canberra: Pacific Linguistics.
Rigsby, Bruce & Chase, Athol. 1998. The sandbeach people and dugong hunters of Eastern Cape
York Peninsula: Property in land and sea country. In Customary Marine Tenure in Australia, Nicholas Peterson & Bruce Rigsby (eds), 192218. Sydney: University of Sydney.
Focalisation dans les Langues Africaines, Bernard Caron (ed.), 229267. Leuven: Peeters.
Van Valin Jr., Robert D. & LaPolla, Randy J. 1997. Syntax. Structure, Meaning, and Function.
Cambridge: CUP.
Verstraete, Jean-Christophe. 2005. The semantics and pragmatics of composite mood marking:
The non-Pama-Nyungan languages of northern Australia. Linguistic Typology 9: 223268.
Verstraete, Jean-Christophe. 2008. The status of purpose, reason and intended endpoint in the
typology of complex sentences. Linguistics 46: 757788.
Verstraete, Jean-Christophe. 2009. Illocution and focus at the semantics/pragmatics interface in
Umpithamu (Cape York, Australia). 2009 Linguistics 47: 867884.
Verstraete, Jean-Christophe. Forthcoming. The role of mood marking in complex sentences:
A case study of Australian languages. To appear in Word.
Clause chaining and conjugations in Wolof

A typology of parataxis and its semantics
Stphane Robert
Fdration Typologie et Universaux Linguistiques-LLACAN, CNRS

Due to the specificity of its verb conjugation system, Wolof (Senegal) favors
parataxis for clause combining and provides an interesting case where
interclausal dependency can be marked by forms indicating information
hierarchy. Furthermore, the study of clause combining shows that, with some
restrictions on possible combinations, the various combinations of conjugations
(or Tense-Aspect-Modality markers) produce different but regular interclausal
meanings, such as succession, causality, opposition or consecution. Moreover,
due to the nature of the different conjugations, paratactic clause chaining in
Wolof yields different types of interclausal dependency, defining a gradient of
syntactic integration: from simple assertive juxtaposition to more integrated
syntactic dependency, through lesser known types of dependency, defined here as
situational dependency and pragmatic dependency.
1. Introduction
Wolof is spoken by approximately ten million speakers mainly in Senegal and belongs
to the Northern Atlantic branch of the Niger Congo family. It has a complex and interesting verbal system expressing, among others, (1) distinctions related to information
hierarchy (focus), and (2) what I have dubbed situational dependency. Another characteristic of this language, which is probably related to this particular verbal system, is
the tendency in Wolof to express complex sentences paratactically, i.e. by juxtaposing
clauses in a single sentence without any coordinating or subordinating morphemes.
Subordinating morphemes do exist in the language (cf. Sall 2005)1 but the specific
. A. Salls work is the only comprehensive study of subordination in Wolof. It provides a

detailed syntactic analysis of the various subordinate clauses attested in the language (complement clauses, relative clauses and various adverbial clauses), including an overview of the
semantic functions of the subordinate clauses and of the constraints on the verbal forms used
in these clauses. Most of this study is devoted to dependent clauses introduced by subordinating (or coordinating) morphemes, but it also containssome parts on paratactic structures.
Stphane Robert
properties of its conjugations naturally allow certain types of interclausal relations to

be expressed by simple parataxis and conjugation chaining.
In this paper,2 I will present an overview of the morphosyntactic and semantic patterns observed in paratactic clause chaining; I will try to account for their constraints
and regularities, and to define the nature of the specific types of dependency generated
by clause chaining in the different conjugations. We will proceed from the less to the
more integrated interclausal linkages. The study will be limited to clause combining
within the sentence, sometimes extending to questions and answers in dialogue or
clause chaining in discourse, but always only considering two adjacent clauses. While
not extending to macrostructuration into paragraphs, this level of analysis already
reveals remarkable phenomena. This overview of paratactic sentences will lead us to
grant special attention to two particular cases of clause chaining: one with focusing
conjugations and the other with Null tense (also called Aorist, Narrative or Minimal)
conjugation. Before going into the detailed study of clause chaining with conjugations,
here is an overview of the languages verbal system.
2. The Wolof verbal system

Like several other languages of the Northern Atlantic branch of the Niger-Congo
family (cf. Robert 2010), one of the characteristic features of Wolof grammar is the
synthetic expression of information structure through verbal morphology and focusing conjugations.
2.1 The conjugations

The Wolof verb constituent has two components (Robert 1991, 2000): an invariant
(unless derived) lexical stem and an inflectional marker conveying the grammatical
specifications of the verb (person, number, tense, aspect, and mood) as well as the
information structure of the sentence (focus). The inflectional marker is preposed,
postposed, or suffixed to the lexical stem, and organizes the verb system into ten paradigms or conjugations: Perfect, Presentative, Null tense (or Narrative), Verb Focus,
That is why, although A. Salls perspective is quite different from mine in this article (in particular because she has no intention of analyzing the conjugations role in the semantic value
of clause combination), I will refer to her work occasionally.
. My deepest thanks go to my two reviewers, Kevin Moore and Maarten Mous, for their
valuable comments on an earlier version of this paper.
Subject Focus, Complement Focus, Negative, Emphatic Negative, Obligative, Imperative and its negative counterpart: Prohibitive.3
In the affirmative indicative mood, there are three non-focusing conjugations
(Perfect, Presentative, and Null tense) and three focusing conjugations (traditionally
called emphatic): Verb focus, Subject Focus, and Complement Focus. The latter vary
according to the syntactic function of the focused constituent: subject, verb, or complement (in the wider sense of any constituent which is neither subject nor main verb).
Thus, for dem go, leave and foofu there, over there, one finds the inflections presented in Table 1.
Table 1. Indicative affirmative conjugations in Wolof45
1 sg
2
3
1 pl
2
3
1 sg
2
3
1 pl
2
3
Perfect
Presentative
Null tense
dem naa
dem nga
dem na
dem nanu
dem ngeen
dem nau
maa ngi4 dem

yaa ngi dem
mu ngi dem
nu ngi dem
yeena ngi dem
u ngi dem
ma dem
nga dem
mu ~ dem
nu dem
ngeen dem
u ~ dem
Verb Focus
Subject Focus
Complement Focus5
dama dem
danga dem
da(fa) dem
danu dem
dangeen dem
dau dem
maa dem
yaa dem
moo dem
noo dem
yeena dem
oo dem
foofu laa dem

foofu nga dem
foofu la dem
foofu lanu dem
foofu ngeen dem
foofu lau dem
In the absence of the imperfective suffix (-y), these conjugations or verb inflections have a present perfective value: action verbs refer to a past event, while stative
verbs refer to a present state, except for the Presentative which refers to a current
. For a presentation of the negative inflections, which will not be detailed here, see Robert
1990 or Robert 1991: 283302.
4. The Presentative bears a spatial suffix, usually the proximal (-i), which can alternate with
the distal (-a).
5. The complement-focusing paradigm consists of inflectional markers (e.g. 1sg laa) encoding
both the focusing of the (preposed) complement, the subject person marker, and the aspectual
(perfective) meaning of the following predicate.
Stphane Robert
rocess6 with all verb types (see examples in 2.2.3). More generally, these simple
p
forms, which have present perfective affirmative meaning, can bear imperfective, anterior or negative suffixes.
Table 2. The aspectual and temporal meanings of verb forms in Wolof
Simple verb forms
Suffixed verb forms
zero suffix (perfective)

zero suffix (present)
zero suffix (affirmative)
suffix -y ~ di (imperfective)
suffix -oon (past anterior)
suffixe -ul (negative)
2.2 Conjugation meanings

A careful study of the contextual uses of Wolof conjugations in previous works,7 has
led me to propose the following analyses for the simple affirmative conjugations.8
2.2.1 The focusing conjugations

The three focusing conjugations are used according to the syntactic function of the
focused constituent (subject, verb, or complement, i.e. any other constituent). The
focusing conjugations are obligatory in Wolof whenever a constituent (whatever its
syntactic function) is the rheme (informative part, rheme or commentary) in the
information structure of the sentence. Therefore their uses go beyond the case of contrastive, exclusive or replacing focus; for instance, they are also used in wh- questions9
and are obligatory in the replies to such questions (even when there is no conceivable
alternative to the focused constituent) as in (1):
. On the specific effect of the imperfective suffix on the Presentative, see Robert
(1991: 264) and (1994).
. For a full description of the Wolof verb system, see Robert (1991); for the Null tense or
Aorist in particular, see Robert (1996) and for the focusing conjugations, see Robert (2000
and 2010).
. Alongside the negative conjugations, there is also a complex affirmative conjugation with
future meaning: this form is made up of an imperfective copula (di-) suffixed with the Perfect
inflexion (e.g. dinaa dem I will go); furthermore, this conjugation can carry the imperfective
suffix (e.g. dinaa-y dem I occasionally go); for details, cf Robert, 1991: 2702.
. There is actually a double system of wh- questions in Wolof: the question markers consist
of a class consonant which can bear a suffixed -an morpheme requiring a focusing conjugation (Example 1) or the spatial suffix -u indicating the absence of localization in the deictic
space and requiring the Null tense conjugation as in (2).
(1) Na nga
tudd?
how compfoc.2sg be.named
What is your name? (lit. HOW are you named?)
Kumba laa
tudd.
Kumba compfoc.1sg be.named
My name is Kumba (lit. I am named KUMBA)
[SP]10
I reserve the term rheme (or commentary) for the semantic/pragmatic notion of
new information. I speak of focus only in reference to instances where the rheme
(1) corresponds to a syntactic constituent and (2) is morphologically marked, as is the
case with these focusing conjugations. For instance, in my terminology, the notion of
subject focus applies to a morphological form which prototypically indicates that the
syntactic subject is the rheme. In discourse however, focusing forms may have a variety
of uses in addition to this prototypical one indicating that the lexical content of the
focused constituent is the informative part of the sentence, in particular for the verb
which is a syntactic constituent having the special status of predicate.
My study of Wolof sentences in discourse has shown that the subject-focusing
form has three main uses (see Table 3) as exemplified in (2)(5) (from Robert 2000),
subject identification being just the prototypical one:
Table 3. The main uses of the subject-focusing forms
Identification or qualitative designation of the subject (2).
Definition or explanation of the current situation (3, 4).
Exclamation with an intensification of the verb (5).
(2)

Ku
la bind bataaxal bii?
who null.3sg you write letter
this
Who wrote you this letter?
Daba moo
ma ko bind.
Daba focsubj.3sg me it write
It is Daba (lit. daba wrote it to me ~ it was Daba who wrote it to me).
A person arrives and hears screaming. He asks:

(3)

Lu
xeew fi?
what null.3sg happen here
What is going on here?
Musaa moo
dor Ndey.
Musa focsubj.3sg beat Ndey
It is Musa who has thrashed Ndey.
On data references, see Appendix 1.
[SP]
Stphane Robert
People are talking about a man called Kebe and the crowd of people visiting him:
(4)

Kebe moo
am alal, mootax
nit
i di ko topp.
Kebe focsubj.3sg have wealth that.is.why human the ipfv him follow
Kebe, (it is because) he has money that the people come to him.
(lit. KEBE has money, that is why people follow him).
[SP]
One should note that on the pragmatic level, sentence (4) corresponds to sentence
focus expressing an explanation, but on the morphological level, it corresponds to subject focus; this point will be explained in Section 4.
(5) Mbuum bii, moo
gudd!
rope
this focsubj.3sg be.long
How long this rope is!
In Wolof, the intensive meaning conveyed by subject focus is only possible for verbs
expressing a measurable quality (essentially scalar stative verbs).11 Actually, these
three main uses of subject focus are not restricted to Wolof: they are equally possible in
French, even the most surprising one (predicate intensity). The complement-focusing
form, beside its uses in wh-questions and nominal predicates, serves mostly to identify
the complement, with a more or less contrastive effect12 (cf. Example 1).
Concerning the verb-focusing form, beyond its uses for focusing on the lexical
content (in parallel focus for instance), its main uses can be divided into four types,
falling into two contrasting areas, simple predication and explanation:
Table 4. The main uses of the verb-focusing form
focusing on the lexical content of the verb (6).
simple predication: qualitative definition of the subject or situation (stative verbs only) (7).
explanation (8a and b).
intensive predication (really) (9a).
or focus on the truth value or assessment of the predicate (9b).
(6) Waxuma
la sax rekk lekk, dama

ko wann
tell:neg.1sg you even only eat vbfoc.1sg opr eat.
I have not just eaten it, I have DEVOURED it.
(lit. I do not just tell you only eat, in fact I DEVOURED it)
Interestingly, its use as simple predication is found with stative verbs only: when
focused, beside their focusing use on the lexical meaning of the verb (cf. 6), action
verbs always have an explicative meaning (9), while stative verbs, or more precisely
. For details and an explanation of this type of use, see Robert (1991:135, 306307) or
Robert (2010).
. For details, see Robert 1991: 149164.
verbs expressing a quality or property, are commonly used with this focusing conjugation as mere statements serving for the predication of that property (7):
(7) Dafa
liw.
vbfoc.3sg feel.cold
It is cold.
(stative verb expressing a quality)

(unmarked statement)
By contrast, the explicative meaning is possible for all verbs, action verbs (8a) as well as
stative verbs (8b), and is actually the most common meaning for the former:
(8) a.

Dafa
dem.
13
vbfoc.3sg go
Actually, he left ~ it is because he left.
(action verb)
(confirmation or explanation)
b. Lutax ngay
ibbisi?
why null.2sg: ipfv go:back.home
Why are you coming back home?

damaa
xiif.
vbfoc.1sg:conj be.hungry
(it is because) I am hungry.
[Church, 1981: 139]
Finally, the intensive predication produced by the Verb focus form is visible in contrast
to the Perfect conjugation, as in the following example:
(9) a. Bgg naa

dem
want prf.1sg go
I want to leave
Dama
bgga
dem
vbfoc.1sg want:conj go
I firmly intend to leave
b. Momar dafa
ko gis.

Momar vbfoc.3sg opr see

Momar DID see it.
In order to explain these various uses and meanings of focusing forms, I have defined
focus (Robert 1993, 2000 and 2010) as a specific mode of identification: in a focused
sentence, assertion consists in the qualitative designation of an element whose existence
is presupposed. In other words, the focused proposition consists of a split assertion
involving a temporal presupposition14 (of the predicative relationship, e.g. I am named
somehow) and a qualitative designation (of the focused constituent: Kumba is how I
. In the absence of the imperfective suffix (-y), all Wolof conjugations have present perfective meaning: action verbs refer to a past event while stative verbs refer to a present state; cf.
above in 1.1.
. Actually, I prefer to call it a pre-constructed assertion or pre-assertion given that the
speaker explicitly indicates (by using focus markers) that the predicative relation already
holds true, independently from his present statement and commitment, and is warranted by
a prior statement.
Stphane Robert
am named). The temporal presupposition of the predicative relationship involved in

focused sentences explains why the unmarked focused conjugations (with a suffix
for tense and aspect) have perfective present meaning in Wolof. Moreover, the fundamental meaning of qualitative designation of the focused constituent explains the neutral meaning of focused quality verbs with Verb focus conjugation, contrasting with
the marked (focalized) meaning of action verbs: since verb focus is an assertion of the
qualitative properties of the verb, its common use with verbs of quality follows from
the affinity between the semantics of the verb (indicating a quality) and the semantics
of the focusing conjugation itself (expressing a qualitative identification of the predicate). This is particularly important in a language where there is no simple present
conjugation available to express a mere statement. The explanatory meaning appearing
in discourse and clause chaining will be explained in 3. In fact, the grammaticalization
of focus marking in verbal inflection is essential for the explanation of clause linkage
by conjugation chaining in Wolof.
In contrast to the focusing conjugations, the three non-focusing conjugations
indicate that the rheme (or informative part of the utterance) is not a syntactic constituent as such.
2.2.2 The Perfect

The Perfect has resultative meaning: it indicates that a process (already known to be
ongoing) has henceforth reached its expected end-point or term so that there is nothing to add, no more variation: a stable resulting state has been reached; thus its informational content is conveyed by the aspecto-temporal component of the process.
With dynamic (or action) verbs that take place over time and have a temporal
term, this conjugation indicates that any temporal variation or instability is now eliminated. It provides the classical attributes of perfect conjugations such as to have finished doing p (10) or to have already done p (11). Depending on the context, emphasis
can also be laid on the resulting state (12).
(10)

Yaa ngiy
lekk ? Ddt, lekk naa
(ba noppi).
prst.2sg:ipv eat no
eat prf.1sg (until stop)
Are you eating? No, I have finished eating (my meal).
(11) Sarax bi gg na.

alms the arrive prf.3sg
lit. The alms have (already) arrived that is I have already given (meaning:
I wont give any more).
[SP]
(12) Jaar nga
ci ngu
gr
pass prf.2sg in room:conn man.
You have gone through the mens house = you are circumcised,
i.e. you are ambitious.
[XCL]
It is worth noting that with stative verbs, which have no temporal phases, no unfolding
over time, one does not find the same aspecto-temporal meanings but, instead, modal
or subjective uses. This corresponds to what De Smet & Verstraete (2006), after Halliday (1994) and Halliday and Hasan (1976) call ideational subjectivity. Moreover,
these subjective meanings are also associated with regular argumentative effects (the
discussion is over/I disagree/I agree/I am relieved/you should do something) corresponding to what De Smet & Verstraete (2006) call interpersonal subjectivity that
deals with the positioning of the speaker with respect to [the representation of the
extralinguistic and extra-discursive world] and his or her interaction with the interlocutor. Due to the properties of these Aktionsarten, the elimination of variation or
instability corresponds here to the elimination of the epistemic variation surrounding the predicate (for more details see Robert 1991:5267 and Robert 1994). So with
stative verbs, the Perfect indicates that there is no doubt as to the assertion, it simply
conveys the speakers viewpoint, with several possible contextual meanings: expected
conformity (13), the speakers agreement (14), polemic or decisive assertion (15):
(13)

[a person who was looking for a rope of some length]

Buum bi gudd na.
rope the be.long prf.3sg
(Its all right) the rope is long (enough). [gloss: here we are! at long last]
(14) [two people looking at a boubou (cloth)]

Bubu bii, dafa
rafet.

boubou this, vbfoc.3sg be.beautiful

This boubou is beautiful.

Rafet
na
(de)!
be.beautiful prf.3sg (ptcl)
It is indeed (beautiful). ~ it is (definitely) a beautiful one.
(agreement of the speaker)
[SP]
(15) [At the end of a discussion where the speakers disagree]

Tng na!
be.hot prf.3sg
(Im telling you) it is hot!
[gloss: it is certain, there is no more discussion, there is nothing to be added, no
further comment, period].
2.2.3 The Presentative

The Presentative reports a current state of affairs by situating the predicative relationship in the speakers space-time. More specifically, this conjugation indicates that the
process is happening at the moment of the speech act, in a place close (vs. remote) to
the speakers space, as indicated by the spatial suffix: -i for proximal, -a for distal. The
rheme here consists in the specification, the localization of the predicative relation in
Stphane Robert
the discourse situation, whence its meaning of current present or a recent event which
has been updated, for example by being just witnessed by the speaker. Contrary to
the Perfect or the focusing conjugations, there is no presupposition here, no previous
expectations: the process occurs at the time of speech and serves to define the discourse
situation and the events which affect the speaker, and which happen to him suddenly
and, strictly speaking, unexpectedly. This is why the Presentative is typically used by
reporters and in the news. This point is important for explaining some modal effects of
the Presentative (such as surprise or warning) as well as the interclausal meaning of the
Presentative (cf. 5.): unexpected events tend to be perceived as detrimental.
(16)

Gisuloo
Abdu?
see: neg.2sg Abdu?
Have you not seen Abdou?
a. Mu ngi
dellu dkk bi.
prest.3sg return town the

Here he is (right here) coming back to the village. (he can be seen coming)
b. Mu nga
jm
ca dkk ba, legi laa
prest.3sg be.headed.for at town the, now compfoc.1sg

tase
ak moom.
encounter with him
He is on his way back to the village, I just ran into him.
c. Abdu? Mu ngi
mujj
ci gannaaw!
Abdu? prest.3sg be.the.last at back
Abdou? (As I am speaking to you) he is over there, way at the end
of the line!
d. Abdu? Mu nga
fa.
Abdu? prest.3sg:dist there

Abdou? He is over there.
Stative verbs rarely seem to be used in independent clauses with the Presentative, but
are common with this conjugation in paratactic structures (cf. 5.).
2.2.4 The Null tense

The Null tense (also called Narrative, Aorist or Minimal) holds a special place in the
system, defined by its being the only non-tensed conjugation. More precisely, with the
Null tense, the verb has the properties of a predicative operator, but the speaker does
not express any commitment and the event is not located in the speakers time. Aspectually, the verb is minimally specified and referred to as a pure and simple notion without
internal structuring: its aspectual meaning is therefore perfective and becomes imperfective only when suffixed with the imperfective suffix (-y). This analysis (Robert 1991
and 1996) can account for the various and, at first, apparently contradictory uses of the
Null tense: on the one hand, this conjugation is common in proverbs (17) and obligatory in tales and historical narrations (18), in which the Null tense is apparently used
in independent clauses; on the other hand, it is typically a subordinating mood since it
is obligatory with most subordinating conjunctions (19); it is also the only conjugation
used to mark clause subordination (complement clauses as in Example (20), or consecutive or purpose clauses as in Examples (21)) without any subordinating morpheme.
(17) Ku
mu,
muu.
who null.3sg be.patient, smile
The one who is patient will smile. (Patience is rewarded)
[SP]
(18) As soxna dafa amoon doom ju jigen. Bi doom ji matee sy mu maye ko. Ylla def
xale ba mb
Once upon the time there lived (Verb focus) an old woman with her daughter.
When the daughter became (Null tense) nubile, her mother married (Null tense)
her off. God willing (Null tense), the child became pregnant (Null tense) [T: 169]
As the beginning of a tale, example (18) starts with a formulaic expression using another
conjugation (the Verb Focus conjugation) and then proceeds with Null tense clauses.
(19) Bu

waan,
xale
ypp
When null.3sg come:pastiter, children all
dauy
bg.
vbfoc.3pl:ipfv be.pleased
Whenever he came, all the children were pleased.
(20) Dama
bggoon ngeen
ndal
maak
sama doom.
vbfoc.1sg like:pst null.2pl accompany me:with my child
I would like you (to) accompany my daughter for me.
(21) a.

[T: 169]
Dafa
scc, u
kaaf
ko.
vbfoc.3sg steal null.3sg imprison him
He stole (therefore) he was put in jail.
b. Jox ma ko, ma
seet.

give me it, null.1sg look

Give it to me (so I can) have a look.
[SP]
With this conjugation, the process is anchored (located) in an unspecified situation. In

other words, the verb is endowed with the properties of a predicative operator, but the
event is not located in time, nor is the speaker committed to it, therefore it is lacking
in temporal and modal specifications; the sentence is not a complete assertion and the
clause depends on some extra-clausal element to specify in which situation this event is
located and true. This element, which I call a situational locator, functions as an anchor
point and discursive landmark for the Null tense clause, and supplies it with the missing temporal and modal specifications. So the various uses of this conjugation can be
Stphane Robert
explained through the various syntactic statuses of the situational locator as it is defined in
discourse: a different sentence, another clause, another verb or zero. The various degrees
of syntactic dependency of the Null tense clause, its more or less subordinating function
and embedded status, as well as its semantic effects, depend on the degree of syntactic
integration of the situational locator and the Null tense clause, as will be explained in 5.
3. P
ermitted and prohibited clause chaining: the role of conjugations;
succession and cumulative assertion
The use of negative conjugations in clause chaining has not yet been studied; the following remarks only concern affirmative inflections.
3.1 Prohibited combinations

There are various types of constraints in paratactic clause chaining. The main constraint concerns the Null tense. It follows from the presentation in 2.2.4 that this
conjugation is the favoured conjugation for a second clause (P2) in paratactic constructions, both within a complex sentence (as a subordinating mood), and after a previous sentence in narratives, as exemplified above in Examples (1721) (for a detailed
analysis, see 5.1). However, when the Null tense appears in a first clause (P1),15 it
seems to block clause chaining (P2) with any other conjugation (cf. Sall 2005:270): for
instance, while a sequence of Null tense-clauses is possible (Example22), as well as the
chain Verb Focus followed by Null tense (23), the sequence Null tense-Verb Focus is
impossible (24). An explanation for this constraint will be proposed in 5.
(22) Mu
gg, taw bi
door.
null.3sg arrive, rain the null.3sg begin
(As soon as) he arrived, the rain started.
(23) Dafa
gg (rekk), taw bi
door.
vbfoc.3sg arrive (only), rain the null.3sg begin
(As soon as) he arrived, it started to rain.
(24) *Mu
gg, taw bi dafa
door
null.3sg arrive, rain the vbfoc.3sg begin
The combination of a first Null tense clause followed by a clause with any conjugation
other than Null tense is the only impossible combination. However, some other types
of sequences, although possible, seem to be rare or constrained. Firstly, the sequence
. Within a narrative, but the Null tense cannot appear at the beginning of a narrative, cf.
Section5.
Perfect-Null tense in juxtaposition is possible (25a), but the speaker naturally prefersto add a temporal auxiliary (25b) in the following example:
(25) a.

? gg na,
taw bi
door.
arrive prf.3sg, rain the null.3sg begin
He arrived, the rains started.
b. gg na,
taw bi
sooga
door.

arrive prf.3sg, rain the null.3sg aux(happen.after):conj begin

He arrived (then) the rains started.
It is worth noting that among their various respective uses, on the one hand, the
Perfect is the verb form used for past events in the narration of personal experiences (discours in Benvenistes terms), while, on the other hand, the Null tense is
required for past events in narratives such as tales or historical narration (rcit in
Benvenistes terms). This contrast and complementary distribution might explain the
speakers reluctance for using both of them in the same sentence without a temporal
auxiliary between the two clauses. The chaining would probably be more acceptable in a verbal context where the Null tense clause could be interpreted as a consequence of the Perfect clause: this would be a case of the Null tense being used as
asubordinator.
In addition, the chaining of two Perfect clauses (cf. 2.2), while perfectly correct
(Example26), seems to be rare:
(26) gg na,
taw bi door na.
arrive prf.3sg, rain the begin prf.3sg
He arrived, (then) it started to rain.
Nor have we found many cases of clause chaining with two Verb foci: one example,
given by Sall (2005: 269), lacks sufficient contextual indications for interpreting its
meaning otherwise than as a cumulative assertion (cf. (36) in 3.2.2.); in the second
example (41), the translation clearly indicates that the first clause (P1) is the explanation of a previous statement or situation and the second one (P2) is the explanation of
P1; it will therefore be presented in Section 4.
Lastly, a sequence of two Presentatives seems to require parallel structures where
the first one contrasts with the second one, as in (27). It is probably because the insertion of the personal pronoun you reinforces the parallel and contrast that (29) is
more acceptable than (28):
(27) Mi ngi
fot, maa ngi togg.
prest.3sg wash prest.1sg cook
He is washing the laundry, I am cooking.
(28) ? Mu ngi bokk ak yow, yaa ngi
ko koy
tere!
prest.3sg share with you prest.2sg opr opr:ipfv forbid
He has the same rights as you (and) you forbid him!
Stphane Robert
(29) Mu ngi
bokk ak yow, yow yaa ngi
ko koy
tere!
prest.3sg share with you you prest.2sg opr opr:ipfv forbid
He has the same rights as you (and) you, you forbid him!
3.2 Juxtaposition of same conjugation: Succession or cumulative assertion

Despite the constraints mentioned above (3.1.), chaining clauses with the same conjugation is possible for all conjugations. However, this type of chaining is apparently
not frequent except for a succession of Null tense clauses in narratives. Semantically,
chaining two identical conjugations corresponds to two cases: (1) a temporal succession of events (for Perfect and Null tense only) or (2) a cumulative or contrasting
assertion (all other cases).
3.2.1 Temporal succession of events

With dynamic processes (action verbs), clause chaining with the Perfect expresses a
succession of events as exemplified in (26) above. One can contrast this temporal effect
of the Perfect with that of the Presentative corresponding to two simultaneous actions
in (27) above. The comparison with the use of Verb focus also shows that, in the case of
the Perfect (30, repeated from 26), the link between the two events is purely temporal
(one event occurs after a previous one): in contrast, the same sentence with Verb Focus
in P1 and Null tense in P2 (31, repeated from 23) describes a situation in which the
first event triggers the second one:
(30) gg na,
taw bi door na.
(31) Dafa
gg (rekk), taw bi
door.
(As soon as) he arrived, the rains started His arrival was enough for
This succession effect for the Perfect is in accordance with its definition as indicating that a process has now reached its expected end-point/term: the chaining of two
Perfect-processes corresponds to the successive recording, by the speaker, of resulting
events. The Perfect is a tensed conjugation using the speech-time as reference point,
so that clause chaining with two Perfects corresponds to the successive anchoring in
speech-time of two resulting events. The (temporal) sequencing effect is the same with
the Null tense (cf. Example18 above), but produced differently. First, the Null tense
is used for expressing successive events in narratives rather than in discourse. Secondly, in this case, the temporal succession is not produced by successive anchoring
in speech-time (as with the Perfect), but by what I call situational anaphora (Robert
1996): lacking in temporal and modal specifications, the Null tense clause depends
on some extra-clausal locator (cf. 2.2.4). Thus, at the beginning of a narrative, there
must be a previous clause with a tensed conjugation (for instance the Verb Focus with
the past suffix, as in (18) which is the beginning of a tale), from which the Null tense
clause can receive its temporal specifications: the Null tense clause then refers to this
previous situation, just as a relative pronoun refers to its antecedent; this is what I call
situational anaphora. This situational anaphora goes on through the narration: all the
Null tense-events follow one another starting from this previous temporal anchoring,
as a set, inescapable chain of events, typical of historical narration (for details see
Section 5). This particular relation between the Null tense-events is visible in the following example where the first event instantly triggers the second one:
(32)

Mu
w, ma
dem.
null.3sg come null.1sg go
Ds quil est venu, je suis parti.
(As soon as) he came, I left.
[Sall 2005:267]
3.2.2 Cumulative assertion

In other cases, particularly with stative verbs which do not unfold over time, clause
chaining with Perfects corresponds to a simple juxtaposition expressing cumulative assertion, which I define as a succession of two assertions with the same pragmaticeffect or argumentative strength:
(33) Grand, rusloo nga
ma, too nga
ma rekk.
Grand, shame prf.2sg me, wrong prf.2sg me only
Grand, you shamed me, you just plain wronged me.
[XSW]
(34) Sant naa

leen lool. Jr
ngeen jf. Sama xol
thank prf.1sg you much be.worth prf.2pl deed my heart
sedd
na.
be.cold prf.3sg
Thank you very much. I say thank you. I am satisfied.
[G]
Generally, this pragmatic effect of cumulative assertion for clause chaining with the
same conjugation holds true for conjugations other than the Perfect. Here is a good
example of cumulative assertion with Subject Focus from a dialogue in a TV series:
(35) Moo
xam lu
tilim
ci sama doom, moo
xam
subjfoc.3sg know rel.pr be.dirty in my child subjfoc.3sg know
lu wara
bon
ci sama doom, moo
xam
rel must:conj be.bad in my child subjfoc.3sg know
feebaram,
moo
koy
boot
ci diggu
fever:poss.3sg, subjfoc.3sg opr:ipfv carry.on.the.back in middle:conn
ginnaawam
gi.
back:poss.3sg the
She is the one who knows what is unclean in my child, she is the one who
knows what must be bad for my child, she is the one who knows if he is sick, she
is the one who will carry him on her back.
[XSW]
Stphane Robert
The chaining of two Verb Focus conjugations, although apparently rare, also seems to
be possible, as in the following example from Sall:
(36) Gej gi dafa
aay, dauy
tere
ku fa
sangu.
see the vbfoc.3sg rage vbfoc.3pl:ipfv prohibit rel there bath
The sea is very rough, swimming is prohibited
[Sall 2005:269]
One should remember that the two main uses of Verb focus are simple qualitative
predication and explanation (cf. 2.2.1). This example lacks sufficient contextual indications for interpreting its meaning in the paratactic structure more specifically than
as a cumulative assertion. But another example (41 infra) indicates that the first clause
(P1) explains a previous statement or situation, the second one (P2) explains P1, in
accordance with the most common explanatory meaning of Verb focus in clause
chaining which will be presented in 3.
In the case of Presentatives, as in example (29) above, paratactic chaining generally expresses a contrast between two parallel clauses and situations happening at the
time of speech:
(37) Maa ngiy
gnn, yow, yaa ngiy
dugg.
prest.1sg:ipfv exit you prest.2sg:ipfv enter
I am going out (whereas) you, you are coming in.
One could probably generalize the following point concerning non-temporal clause
chaining: depending on whether the argumentative orientation of the two clauses is
convergent or divergent, the effect of the repetition of the same conjugation can either
be that of strengthening, reinforcing the preceding assertion in a cumulative fashion
(cf. 3335), or that of emphasizing contrast or discordance between the two situations
(cf. the Presentative in 29 and 37).
The remarkable argumentative effect produced here by the Presentative is in
accordance with its most common meaning of discordance in clause chaining and due
to its specific semantics, as will be explained in Section6. We can now summarize the
semantics of clause chaining with the same conjugation in Table 5.
Table 5. Semantics of clause chaining with the same conjugation
Chaining of a single conjugation
Semantics
Null tense P1, Null tense P2:

Perfect P1, Perfect P2(action verbs):
Perfect P1, Perfect P2(stative verbs) or
Any other Conjugation P1,
same Conjugation P2:
temporal succession in narratives

temporal succession in discourse
cumulative assertion:
a. reinforcing effect (all conjugations)
b. contrastive effect (Presentative)
Alongside this type of clause chaining, the Wolof system provides two particular
cases of special interest: clause chaining with focusing conjugations (Section4) and
clause chaining with Null tense (Section5). Apart from the use of negative conjugations which are not studied here, these two types of paratactic chaining seem to be
favoured since they are the most frequent. We will also mention interesting cases of
clause combining with the Presentative (Section6).
4. F
ocus in clause chaining: explanation and pragmatic
dependency (discursive landmark)
Since it entails some presupposition (cf. 2.2.1), a focused clause is a priori seldom
used alone, without clause chaining: the focused clause is usually related to a previous clause or sentence (corresponding to the presupposition) to which it adds supplementary information, contrast or correction, bearing specifically on the focused
constituent. In dialogues in particular, a focused sentence is frequently used as an
answer to a previous question or statement: for instance a sentence like JOHN went
there is frequently used after the previous question Who went there? or for correcting a preceding statement You went there which both correspond to the presupposition (somebody went there) involved in the focused clause JOHN went there.
This is generally true in Wolof. However, in this language, the focused clause may
appear in discourse where no preceding clause corresponds to the presupposition. In particular, as we saw in 2.2.1, the Verb focus form yields two paradoxical
and apparently contradictory uses: (1) it is the usual and ordinary conjugation for
stative verbs, more specifically for verbs expressing a quality or property such as
sedd to be cold, xonq to be red or xiif to be hungry (which are verbs in this
language, which has no adjectives); (2) action verbs cannot be used with a Verb
focus form outside of clause chaining, where the focused clause has explanatory
meaning. In the first case (with stative verbs), the morphologically focused clause
can appear on its own, and pragmatically corresponds to a simple statement as in
the following example:
(38) Dafa
tng.
vbfoc.3sg be.hot
It is hot/the weather is hot.
[SP]
We explained this paradoxical use in 2.2.1 through the affinity between the semantics
of the verb (expressing a quality) and the semantics of the focusing operation indicating a qualitative designation of the focused constituent (here the verb). By contrast,
while it is possible to open discourse with an action verb in the Presentative form (39a)
Stphane Robert
or with a stative verb in the Verb focus form (39c), Verb focus is inappropriate with an
action verb (39b):
(39) a.

Maa ngi sant i doxal kolok

bii ci li
prest.1sg thank rel lead conference the at rel
u
ma may ngir ma
ubbi ko.
null.3pl me offer for null.1sg open it
I thank the organizers for inviting me to open this conference.
Maa ngi sant

prest.1sg thank
I thank
b. ? Damay
sant

vbfoc.1sg:ipfv thank
c.

Dama
bgga
sant
vbfoc.1sg want:conj thank
I want to thank
(Presentative appropriate)
(Verb focus inappropriate for opening)
(Verb focus appropriate)
(39) b. would appear as the explanation of a previous sentence and would therefore
be inappropriate as an opening. So a clause with an action verb like to leave would
be incomplete if it appeared alone and could only be used in clause chaining with an
explanatory meaning as in (40):
(40)

Dafa
dem
vbfoc.3sg leave
*He has left.
its that he has left (that is why)
In fact, the common meaning of Verb focus (whatever the Aktionsart) in clause chaining is that of explanations as in Example (8b) in a dialogue, or here in a complex
sentence (41):
(41) Dangaa
narulwoona
jnd, dangeen defe ni jigen
vbfoc.2sg intend:neg:pst:conj buy vbfoc.2pl think as women
pp a
yem.
all foc be.similar
(Its that) you had no intention of buying (because) you think that all women
are the same.
[XCL]
Here we have two Verb focus clauses: as shown by the translation, the first one (P1) is the
explanation of a previous statement or situation, the second one (P2) is the explanation
of P1.16 More precisely, in clause chaining, the Verb focus clause appears as the causal
source of P2 when it is the protasis (P1), and as the explanation of P1 when it is the apodosis (P2). This causal source meaning of P2 is illustrated by the comments made on the
two contrasting examples, (42, repeated from 26) with the Perfect and (43, repeated from
23) with the Verb focus in the protasis:
(42) gg na,
taw bi door na.
(43) Dafa
gg (rekk), taw bi
door.
(As soon as) he arrived, it started to rain. His arrival was enough for
With the Perfect (42), the link between the two events is purely that of temporal succession, while (43), with the Verb focus in the protasis, indicates that the first event
triggers the second, i.e. is its causal source, as appears in one informants comment:
it could be used, for instance, to denote the supernatural power of a marabout whose
arrival would trigger the rains, would cause the rain to fall. Example (44) gives another
illustration of this causal source meaning for the Verb focus in the protasis, while (45)
and (46) illustrate its explanatory meaning in the apodosis:
(44) Dafa
ko fetal, mu
dee.
vbfoc.3sg him shoot, null.3sg die
He shot him (therefore) he is dead.
Here, the Null tense expresses a consequence. Verb focus and Null tense are therefore
complementary in this type of causal structure, the first one indicating the starting
point or causal source of a situation and the second one its resulting consequences.
(45) Sama cmmi waxal
mboog, man dama
ykkamti.
My brother17 speak:imp therefore me vbfoc.1sg be.rushed
Come my friend, make up your mind (because) I am in a rush.
(46) Moytul
paaka bi, dafa
aw
de!
avoid:imp knife the vbfoc.3sg be.sharp ptcl
Be careful with the knife (because) it is sharp!
[XCL]
[SP]
. I have found the same meaning of explanation accumulation when the two Verb focus
clauses are coordinated with te (and), but in this case, the causal link between the two clauses can
be either positive or contrastive as in the following example: [to justify the fact that he no longer
prays] Ylla dafa yex te man dama ykkamti (its that) God is slow, (whereas) me, Im in a hurry.
. brother for a sister.
Stphane Robert
I have also found a few cases where the Verb focus conjugation in the apodosis seems
to be used simply in order to characterize P1 more explicitly or to add a qualitative
description of the action expressed by P1 rather than a true explanation:
(47)

Biig,
nelawuma
benn yoon, dama
fanaane
xoole
last.night sleep:neg.1sg one way vbfoc.1sg spend.night keep.awake
ba bir-set.
till dawn
Last night, I didnt sleep once, I stayed awake until morning.
[XSW]
(48) Yow deewagoo,

dangay
door.
you die:not.yet:neg.2sg vbfoc.2sg: ipfv begin
You, you are not dead yet, you are just beginning.
[XSW]
This use is in accordance with the qualitative meaning of Verb focus (cf. 3.1.).
In both cases, as a causal source or an explanation in discourse, Verb focus in
clause chaining generally indicates the necessary and sufficient cause of a situation.
How then could we explain its specific meaning in clause chaining, and account for the
varied pragmatic meanings of focused clauses?
In fact, the range of meanings of focusing conjugations depends on interclausal
linkage and particularly on the available discourse reference points, which may
or may not coincide with the presupposition of the focused sentence. In 2.2.1, we
defined focus as a specific mode of identification: in a focused sentence, the assertion
consists in the qualitative designation of an element whose existence is presupposed.
The various pragmatic functions of the focused clause can be accounted for by a regular mechanism combining this unitary definition of focus with various discursive
linkages. The proposed definition implies (1) that the fundamental meaning of the
focused clause is that of a qualitative assertion (meaning, in the case of verb focus, that
the predicate is asserted for its lexical properties), (2) that the focused clause, by itself,
always involves a presupposition. As we said earlier, the focused proposition (JOHN
went there) consists of a split assertion involving two components having different
pragmatic statuses: a temporal presupposition of the predicative relationship (someone went there) and a qualitative designation of the focused constituent (John is the
one who went there). This means that the whole predicative relationship (R) is present
in the focused clause as a presupposed background upon which the focused element is
profiled as the salient component, constituting what is really asserted by the speaker.
This predicative background functions as an internal clause landmark.
In the prototypical case, the preceding clause corresponds to the presupposition,
and therefore to the internal landmark of the focused sentence. But this is not always
the case. The available discourse reference points (R) created by discourse chaining
may or may not coincide with this internal focused clause landmark (R): it may be
absent (), identical, or different.18 These different cases correspond to the various uses
of the focused forms as summarized in Table 3 for Subject focus and Table 4 for Verb
focus. One may describe the various discursive chainings of focused clauses asfollows:
1. If the focused clause is not connected with any preceding clause (discursive landmark = ), in conformity with its meaning of qualitative designation, it takes on
a defining qualitative predication meaning (qualifying the subject or defining the
situation in the case of sentence scope focus) as exemplified in (7) and (38);
2. If the focalized process is compared to another meaning of the same process
(R/R), e.g. in the case of a preceding question, parallel focus (6) or doubt (9a and
9b), it takes on intensive really meaning (where one predicates the truth value of
a previously predicated element);
3. If it is connected to another clause (R), it takes on causal meaning (8a and 8b).
The predicative relation (or the process) compared to which the focalized predicative relation is posited, and which constitutes the discourse landmark, may thus be
absent ( = absolute initial position), identical (R) or different (R) from the predicative relation presupposed by the focalized sentence (R). These various types of discursive chaining produce the different meanings of Verb focus clauses as summarized in
Table 6.
Table 6. Typology of Verb focus meanings in clause chaining
R/
R/R
R/R
definition
really
causal source, explanation
The fact that the fundamental meaning of the focused constituent is a designation
of quality explains the neutral meaning of focused verbs of quality with Verb focus
conjugations when there is no clause chaining (). But how is it that the connection
between the two clauses produces causal meaning? The relations between clauses can
be reduced to two basic categories: circumstance and causation (as well as absence of
relation). With predicate focalization, there is some identification of the predicative
relations core; the relations between the two clauses is thus not of a circumstantial but
rather of a causal nature. This creates a connection between two predicative relations
where one serves as the landmark for the other. One may therefore gloss the example
of the explanatory focalization above (Example 46) as follows: be careful with the
. For a more detailed account of the various meanings of focused forms, in particular on
the intensive value of Subject focus, cf. Robert 1990, 1993 and 2000.
Stphane Robert
knife, as it is sharp. In this example, one glimpses the links between the explanatory
meaning and the qualitative predication typical of stative verbs: in connection with
another sentence, indeed, mentioning the quality of knife is enough to use it to justify
the situation previously posited: it is because it is really sharp that one must be careful with the knife.
In the case of subject focus,19 the explanatory meaning rests on the same mechanism, but the speaker also chooses to omit the link of the presupposed predicative relation, which is equivalent to explaining a situation by designating the author responsible
for another situation: thus in Example3, in response to the question whats going on
here?, the speaker answers its Musa who hit Ndey. One should note that we have not
yet encountered examples where the Complement focus has causal meaning.
This causal meaning of focused forms (verb or subject) in clausal linkage relies
on the general mechanism of connections between the discursive landmark (preceding clause) and the internal landmark (presupposed assertion), and is also attested in
many languages such as French, Berber (Leguil 1987), Umpithamu (Verstraete, this
volume) or some Oceanic languages (Bril, in press). It provides an interesting case in
which interclausal dependency is marked by forms indicating information hierarchy.
This corresponds to a particular type of dependency, different from embedding, subordination or syntactic dependency, i.e. pragmatic dependency of a focused clause on
its discursive landmark. This particular type of dependency construes the semantics
of interclausal linkage.
5. Null tense, consequence and situational dependency

The Null tense also yields another particular type of interclausal dependency which
has not yet been sufficiently described and deserves special mention.
In order to understand how this works, one has to remember the constraints
in the apparently independent uses of the Null tense: in tales or narratives, a Null
tense clause never appears first, but only after the temporal specification provided by
another conjugation (cf. Example18). In discourse, it is found only in specific utterances such as proverbs (cf. 17) or stage directions, which can be described as general or
a priori statements whose truth value does not depend on the speaker and on the time
of speech (gnomic value), but which are used by the speaker in order to illustrate a
particular situation. In other cases, the Null tense clause always comes second, embedded in a previous clause with a different conjugation (Examples20 and 21) or with a
subordinating conjunction (19). Otherwise, the Null clause appears as incomplete and
requiring further specification, as in injunctions or in wh-questions where it is obligatory
. For a thorough analysis of the various uses of Subject Focus, see Robert 1993.
with the question markers of the -u series (ku who, lu what, fu where, nu who). In
other words, the Null tense clause presents a large array of dependency types.20
According to my analysis (cf. 2.2.4.), with the Null tense, the process is located with
respect to an unspecified situation: the locator-slot, defined by the speakers time and
his commitment to the utterance, is vacant; this can be symbolized as: [ ]Sit. However, as
with any utterance, the Null tense clause is expected to receive some temporal specification and the speakers endorsement in order to constitute a speech act, therefore, the
utterance has to be located via an extra-clausal locator. That is why the Null tense clause
cannot appear in the first position in a clause chain (cf. 3.1). If there is another utterance
functioning as a locator and anchor point in the preceding context, the clausal chain
provides the special characteristics of what I have called situational anaphora (cf. 3.2.1)
corresponding to both temporal anaphora and an assertive dependency at work in tales
(cf. 18) which always begin with another temporal location and for which the speaker
is not committed as warranting its truth: the process is located in a time and situation
that is specified in another clause. If the locator is not another sentence and independent
clause, but a previous clause in the same sentence, the Null tense clause is embedded in
a complex sentence with consecutive or purposive meaning. If the locator is another
verb in a previous clause, the Null tense clause is embedded in a complex sentence as a
complement clause. Finally, if there is no locator in the preceding context, the utterance
is incomplete and pragmatically not independent: it is characterized by assertive dependency (lack of speakers commitment which requires further specification of its truth
value by the hearer) as is the case with interrogation, injunction and hypothesis.
In sum, Null tense is fundamentally a dependent mode and, as shown in Table 7,
the nature of the situational locator and its integration in the Null tense clause is the
variable determining the various degrees of dependency displayed by the Null tense
clauses, which range from assertive to syntactic dependency, and extend from discourse coherence to embedding.
Table 7. Null tense-clauses and the syntactic integration of their locator
Syntactic integration
Max.
Nature of the locator
Nature of dependency
zero
different sentence
different clause
different verb
assertive dependency
situational anaphora
embedding (purposive or consecutive)
embedding (complement clause)
Now, how can we more precisely account for the meaning of the Null tense in
clause combining, namely its consecutive meaning? Being located in an unspecified
. For more details, see Robert 1991 (199234) or Robert 1996.
Stphane Robert
situation for both temporal and modal specifications, the Null tense clause is expected
to be located via an extra-clausal locator, otherwise it is incomplete. Therefore, there
is a necessary and sufficient link between the clause containing the Null tense and its
locator, to constitute a complete and valid sentence: the specification of the locator
triggers the validation of the Null clause. Depending on the temporal and epistemic
status of the main clauses event, this particular relation between the Null tense clause
and its locator corresponds to a purposive or a consecutive clause: when the locator,
i.e. the event in the main clause, is accomplished, the clause linking value of the Null
tense is that of a consecutive clause as in (21a), whereas when the first event is irrealis,
the Null tense clause takes on the meaning of a purposive clause, as in (21b). However,
in both cases, the relation between the first clause and the Null tense clause expresses
consecution: the Null tense indicates that as soon as the first clause is asserted, it triggers the validation of the second clause This specific semantic relation could account
both for the purposive and the consecutive meanings of Null tense embedded clauses
and for the specific semantics of historical narratives and tales. By contrast with the
Perfect (cf. 2.2.2. and 2.2.4.) which is used for narrating personal experiences (to
which the speaker is committed), the Null tense expresses successive events in narratives, rather than in discourse. In this case, the temporal succession of events is produced by situational anaphora and not by successive anchorings in speech-time: the
specification of a first temporal location (by another conjugation) triggers the validation of the Null tense clause which conveys a development of the situation specified in
the opening clause. After which all the Null tense events follow one after the other, out
of this previous temporal anchoring, as an inescapable chain of events, independent
from the speaker. This particular presentation of event chaining is typical of historical
narrations where the chain of events is presented a posteriori as necessary and inescapable, and which we will call a reconstructed consecution of history.
Before closing this overview of paratactic clause chaining in Wolof, I would like
to mention one last interesting case which also concerns the Null tense: the role of the
imperfective marker in clause chaining.
6. Th
e role of the imperfective in clause chaining:
Simultaneityand opposition
As mentioned in 2.2.4., the Null tense (with ) has perfective meaning. It becomes
imperfective with the -y (~ di) suffix. Apparently, when suffixed with this imperfective
marker, a Null tense clause in the apodosis tends to indicate temporal concomitance
with a connotation of contrast or opposition with regard to the protasis, rather than
consecution, as in the following examples:
(49) Maa ngi toog

ci sama biir
neg di
dgg ypp,
prest.1sg be.sitting at my inside room ipfv hear all
ngay
wax.
null.2sg:ipfv talk
I am sitting in my room hearing everything (while) you are speaking. [XSW]
(50) [A first spouse is talking to her husband about the bad behaviour of his
second spouse]
eme na
w fekk ma lay21
defaral
dare prf.3sg come find me you:ipfv prepare:ben
njar,
muy
indi istuwaar?
curdled.milk.with.water null.3sg:ipfv bring quarrel
She would dare come make trouble (while) I am making you curdled milk?
(Lit. She dares come find me preparing milk for you (and) she makes trouble?)
The effect of simultaneity is clearly produced by the specific influence of the imperfective on the dependency expressed by the Null tense. This does not hold true for the
other conjugations in clause chaining. Interestingly, I have found more or less the same
interclausal meaning (P1 though/and yet P2) when a Presentative in P1 is followed by
the perfective (51) or imperfective (52, 53) Null tense in P2:
(51)

[a child is shocked by the bad behaviour of his brother who just beat him]
Mu ngi
fekk may
nelaw, mu
dor ma!
prest.3sg find me: ipfv sleep null.3sg beat me
He finds me asleep and he hits me!
(52) Paaka bi mu ngi

aw
be, nga
koy
foye!
knife the prest.3sg be.sharp so null.2sg opr:ipfv play:ins
The knife is so sharp and (yet) you are playing with it!
(53)

Mu ngi
ko pp
ba pare,
nga
prest.3sg opr be.in.excess until be.ready, null.2sg
koy
yokk!
opr:ipfv make.bigger
Its already too big for him and (yet) you (still) make it bigger!
Stative verbs appear to be rarely used with the Presentative alone. However, they
are frequently found in the following type of structure: in the protasis of a binary
structure, where one has an action verb expressing an unexpected contradiction in the
. When there is a clitic object pronoun in the clause, it attracts the imperfective marker;
this rule applies to all conjugations and not only to the Null Tense.
Stphane Robert
apodosis; in this case the sentence takes on the meaning of he is and yet as in
(52) and (53).
This interclausal meaning is due to some specific semantic features of the Presentative which also expresses simultaneity between the event expressed by the process
and the speech act22 this time: as shown in 2.2.3., the Presentative indicates that the
present process is unforeseen (absence of presupposition or previous expectation). It
is unexpected for the speaker, and unexpected events tend to be expressed as detrimental. This point could explain some of the modal effects of the Presentative (such as
surprise or warning) as well as its interclausal meaning of discordance or opposition
(always marking surprise) when combined with a Null tense process expressing an
action unexpectedly triggered by the Presentative.
The Presentative is also used in connection with the meaning hardly has
that: the meaning is very close to the preceding one. Here too, surprise at an unexpected turn of events is expressed, the only difference being that there is more insistence on their synchronicity. The Presentative clause is often (but not necessarily)23
reinforced by rekk just:
(54) Mu ngi
takk jabaram
ba par,

byyi ko fi!
prest.3sg bind wife:poss.3sg till be.ready null.3sg leave opr here
Hardly has he married his wife that hes abandoning her!
(55) Mu ngi
tj buntam
rekk, xale
yi
tijjiwaat!
prest.3sg shut door: poss.3sg only, children the null.3pl shut:inv:iter
Hardly has he closed the door that the children are opening it again!
These uses in connection with the Presentative thus indicate that two processes perceived as discordant by the speaker coincide temporally or immediately follow each
other; this process allows the speaker to convey this fact alongside disapproval at the
turn of events.
As noted in 3.2.1. on the subject of clause chaining with the same conjugation,
depending on whether the argumentative orientation of the two clauses is convergent or divergent, the effect of the repetition of the same conjugation can be that of
strengthening, reinforcing the preceding assertion in a cumulative fashion, or empha. Of course in (51), when he complains about being beaten, the speaker is not asleep anymore,
but the event is expressed as having just happened, and the informant insists that the Presentative reports it as a current state of affairs. Its meaning is that of a current present or a recent event
with present relevance, and just witnessed by the speaker for instance (cf. 2.2.3.).
. Note also that the use of rekk to reinforce the assertion is not restricted to Presentative
clauses; it is also found with some Verb focus clauses (as in Example23), Negative clauses (6)
or Perfect clauses (as in 31) for instance. A systematic study of discourse particles in Wolof is
still to be done.
sizing a contrast or discordance between the two situations: the speakers surprise,
related to the unexpected character of the event expressed by the Presentative, seems
to result in discordance when the conjugation connects the two events.
7. Conclusion
This overview of clause combining with various conjugations in Wolof reveals a large
array of semantic interclausal meanings and constraints that can be summarized as
in Table (8):
Table 8. The semantics of parataxis in Wolof
Prohibited chains:
P1 Null tense, P2 any conjugation
other than Null tense
Chaining of the same conjugation:
P1 Null tense, P2 Null tense:
temporal succession in narratives
P1 Perfect, P2 Perfect(action verbs):
temporal succession in discourse
P1 any conjugation except Perfect

(or Perfect with stative verbs),
P2 same conjugation as P1:
cumulative assertion:
a. reinforcing effect (all conjugations)
b. contrastive effect (Presentative)
Focusing conjugations in combination:
causal source, explanation
Null tense in second position:

P1, P2 Null tense:
situational anaphora: consecution
Markers of simultaneity in combination:

P1 Presentative, P2 Null tense
(perf. or imperf.)
P1 any conjugation, P2 imperfective
Null tense
concomitance surprising discordance
Thus the current analysis reveals (1) that forms indicating information hierarchy
can be used to mark specific interclausal dependency, (2) that the general constraint
on the necessary temporal location and the speakers commitment to his utterance has
significant effects on clause chaining. Interestingly for the typology of clause linkage,
due to the nature of the different conjugations, paratactic clause chaining in Wolof
yields different types of interclausal dependency, defining an integration gradient:
from simple assertive juxtaposition to more integrated syntactic dependency (with the
embedded Null tense clauses at the endpoint of the gradient), through lesser known
Stphane Robert
types of dependency which we have defined as pragmatic dependency (for focused

sentences depending on a discursive landmark) or situational dependency (for Null
tense sentences depending on an extra-clausal locator). The various combinations of
conjugations in clause chaining produce different regular interclausal semantic effects:
succession, cumulative assertion with reinforcing or contrastive effect, causality, opposition, or consecution. Remarkably, these semantic values produced by parataxis in
Wolof echo the different discourse relations described by the Segmented Discourse
Representation Theory (SDRT), such as narration, explanation, elaboration, result,
parallel or contrast (Lascarides & Asher 2003).
Appendix 1
Corpus and data references
When not otherwise indicated, the data come from native-speaker elicitation: following the method described in Robert (2004), the elicited utterances were always contextualized, that is produced with the description of a precise situation in which they
would be used, and glossed by the informant. Otherwise the following abbreviations
are usedfor the various references:
Gancax gi, a TV play from the radio program Jamonoy Tey, broadcast by the
ORTS (Office de Radiodiffusion-Tlvision du Sngal) on July 8, 1984.
SP:
spontaneous utterances taken down by the author in Dakar in 1985 and 1986.
T:
The anthology of traditional Wolof tales and mythological narratives edited
by Kesteloot & Mbodj (1983).
XCL: A play entitled Xt cig lndm, whose manuscript was furnished in 1985 by
the company of the Daniel Sorano Theater in Dakar.
XSW: Xam sa waru gaar, a play from a TV educational program, by the ORTS
(Office de Radiodiffusion-Tlvision du Sngal) in 1986.
G:
Abbreviations
1, 2, 3
anter
aux
ben
compfoc
comp
first, second, third person

anterior suffix -ee
auxiliary
benefactive verb suffix
complement-focusing
conjugation
complementizer (introducing
complement clauses)
conn
conj
dem
dist
foc
imp
connective suffix
(-u sg, -i pl)
conjunctive verb suffix -a
demonstrative
distal spatial suffix (-a)
subject-focus particle or
suffix (-a)
imperative
imperfective suffix (-y); full

variant: di~d-: imperfective
copula
ins
instrumental verb suffix
inv
inversive verb suffix
iter
iterative verb suffix (-(w)aat)
neg
negative suffix (-ul)
neg.(3sg) Negative perfective
conjugation (3sg)
null
Null tense conjugation
opr
object pronoun
p
proposition or clause
pst
past suffix (-(w)oon)
pastiter past iterative
suffix (-(w)aan)
perf
perfective
ipfv
Perfect conjugation
3sg possessive suffix (-am)
discourse particle
plural
Presentative conjugation
(discontinuous morpheme)
pron
pronoun
prox
proximal spatial suffix (-i)
rel.(pr) relative pronoun
(class marker C+i for the
definite, +u for the indefinite)
sg
singular
subjfoc subject-focusing
conjugation
suff
verb suffix
vbfoc
Verb-focusing conjugation
prf
poss.3sg
ptcl
pl
prest
References
Bril, Isabelle. In press. Coordination, information hierarchy and subordination in some Austronesian languages. In Converbs, Medial Verbs, Clause Chaining and Related Issues, Azeb
Amha, Christian J. Rapold, Sascha Vllmin & Silvia Zaugg-Coretti (eds). Frankfurter Afrikanistische Bltter.
Church, Eric. 1981. Le systme verbal du wolof. Dakar [Documents linguistiques 27]. Dakar:
Publications du Dpartement de linguistique de lUniversit de Dakar.
De Smet, Hendrik & Verstraete, Jean-Christophe. 2006. Coming to terms with subjectivity.
Cognitive Linguistics 17(3): 365392.
Halliday, Michael A.K. 1994. An Introduction to Functional Grammar, 2nd edn. London:
Arnold.
Halliday, Michael A.K. & Hasan, Ruqaiya. 1976. Cohesion in English. London: Longman.
Kesteloot Lilyan & Mbodj, Cherif. 1983. Contes et mythes wolof. Dakar: Nouvelles Editions
Africaines.
Lascarides,Alex & Asher, Nicholas. 2003. Logics of Conversation. Cambridge: CUP.
Leguil, Alphonse. 1987. Structures prdicatives en berbre. Doctorat dtat, INALCO-Universit
de Paris 3.
Robert, Stphane. 1990. Aperu sur la ngation en wolof. Linguistique africaine 4 [Documents
de travail sur la ngation]: 167180.
Robert, Stphane. 1991. Une approche nonciative du systme verbal: le cas du wolof [collection
Sciences du langage]. Paris: Editions du CNRS.
Robert, Stphane. 1993. Structure et smantique de la focalisation. Bulletin de la Socit de
Linguistique de Paris LXXXVIII: 2547.
Stphane Robert
Robert, Stpane. 1994. Sur le rle du sujet parlant dans la construction du sens: Liens entre
temps, aspect et modalit. In Subjecthood and Subjectivity, Marina Yaguello (ed.), 209230.
Paris: Ophrys.
Robert, Stphane. 1996. Aspect zro et dpendance situationnelle: lExemple du wolof. In Dpendance et intgration syntaxique (subordination, coordination, connexion), Claude Mller
(ed.), 153161. Tbingen: Niemeyer.
Focalisation dans les Langues Africaines, Bernard Caron (ed), 229267. Leuven: Peeters.
Robert, Stphane. 2004. la recherche du sens grammatical: Contribution une mthode denqute. In Langues et cultures: Terrains dAfrique, Hommage France Cloarec-Heiss [Collection
Afrique et Langage 7], Pascal Boyeldieu & Pierre Nougayrol (eds), 6576. Louvain: Peeters.
Robert, Stphane. 2010. Focus in Atlantic languages. In The Expression of Information Structure:
A Documentation of its diversity across Africa [Typological Studies in Language 91], Ines
Fiedler & Anne Schwarz (eds), 233260. Amsterdam: John Benjamins.
Sall, Adjaratou Oumar. 2005. La subordination en wolof: Description syntaxique. Ph.D. dissertation, Universit Cheikh Anta Diop de Dakar.
Verstraete, Jean-Christophe. In this volume. Focus, mood and clause-linkage in Umpithamu
(Cape York Peninsula, Australia), 451468.
Pragmatic demotion and clause dependency

On two atypical subordinating strategies
inLoTogaand Hiw (Torres,Vanuatu)
Alexandre Franois
Langues et Civilisations Tradition Orale,

Despite the wealth of subordinators in Hiw and LoToga (Oceanic, north Vanuatu),
two of their Tense-Aspect-Mood categories the Subjunctive and the Background
Perfect can do without them, and encode clause dependency by themselves. A
pragmatic hypothesis is proposed to account for this clause-linking faculty. The
Subjunctive differs from other irrealis categories insofar as it lacks any specific
illocutionary force; the Background Perfect labels its predicate as informationally
backgrounded. In both cases, the clause lacks certain key properties (illocutionary
force; informational weight) which are normally required in pragmatically wellformed utterances. This pragmatic demotion makes the clause dependent on
external predications, which naturally results in syntactic subordination. This
case study illustrates how syntax can be reshaped by the pragmatic parameters
of discourse.
1. Two cases of subordination with no subordinator

1.1 The Torres languages
The Torres islands form a small island group located at the northwestern tip of the
Republic of Vanuatu (formerly New Hebrides), in the south Pacific (Map 1).1 Two
. The present work originates in a talk I gave in 2006 for the research group Typology of
interclausal dependencies (Fdration Typologie et Universaux Linguistiques), led by Isabelle
Bril. I am grateful to her, as well as to Alexis Michaud, Claudia Wegener and Johan van der
Auwera, for their helpful comments on earlier versions of this paper. The data presented in
this chapter were collected by the author during several field trips to the Torres islands, in
2004, 2006 and 2007. The financial support of the lacito CNRS, as well as of the French
Ministre de la Recherche (ACI Jeunes Chercheurs), is also gratefully acknowledged.
Alexandre Franois
ceanic languages are spoken there: Hiw by 150 speakers, and Lo-Toga itself consistO
ing of two very close varieties Lo and Toga by 650 speakers. They have never been the
subject of any published grammatical description.
Torres Is.
Mwotlap
Vanua
Lava
Mota
B A N K S Is .
Gaua
Santo
Hiw
Tegua
Lo
Hiw (150)
[HIW]
T O R R E S I S.
Lo-Toga (650)
[LTG]
Maewo
Santo
Ambae
Malekula
Pentecost
Ambrym
Toga
Epi
Efate
Port-Vila
Erromango
Tanna
Aneityum
Map1. The two Torres languages, at the northwestern tipofVanuatu
Hiw and Lo-Toga differ from each other in many regards, whether in their phonology, their lexicons, or details of their grammars enough to make them clearly
distinct, mutually unintelligible languages. Nevertheless, they also share parallel structures in most domains of their morphosyntax, their phraseology, and more generally the way they categorize meaning into forms. This linguistic isomorphism between
the two Torres languages is due both to their common ancestry, and to a history of
sustained social and cultural contact which their communities have long had with
each other. The linguistic phenomena to be discussed in the present chapter belong to
those many structures which are shared by the two languages: this is why I will treat
them together here, and illustrate each phenomenon with evidence taken alternatively
from Hiw and from Lo-Toga.2
While these two Torres languages also have a lot in common with the languages
of the Banks group and those of Vanuatu in general spoken further south (Map1),
they present many specific developments, which tend to give them a grammatical profile
of their own. This is especially true of the topic I will discuss here, namely the morpho
syntactic strategies for encoding clause dependency and subordination. Generally speaking, as we shall see in Section 2, the various types of dependency between clauses or
predicates (subordination, coordination) are expressed quite classically by a variety
of conjunctions and other overt morphemes that are more or less dedicated to this clauselinking function. Yet, despite the wealth of these formal devices, these two languages have
also developed certain patterns of clause dependency that lack any formal subordinator.
1.2 Parataxis or subordination?

Considered superficially, each of the following sentences consists simply of a string of
two clauses, with no formal indication whatsoever of their syntactic relationship:3
(1) Hiw Ne temt on t
yaqe me
n-w ne,

art devil sbjv go:sg appear hither like this

tekn-wa voyi.

people aor:run.away
[lit. Thedevil would appear like this, people ran away.]

(Whenever) the devil appeared, people would run away.
(2) Ltg

Ne gehuh
ve
kerkur
tle
si
mat mt.
art coconut.crab bkpf1 iter~crunch person bkpf2 cplt die
[lit. Thecoconut crab has devoured people has died.]
The coconut crab (which) had devoured people was dead.
One might propose to see in these two sentences examples of simple clause parataxis (cf.
Noonan 1985:55), or perhaps of verb serialization. In fact I will show that (1) and (2)
rather illustrate genuine patterns of syntactic subordination, in the full sense of the term.
While such instances of apparent clause parataxis are frequent in the spontaneous
speech of the two Torres languages, they are much more constrained than they at first
appear, and depend on the Tense-Aspect-Mood marking (TAM) on the verbs. Among
. When a given fact is unique to one of the two languages, this will be stated explicitly: see
for example the resultative construction, which exists only in LoToga.
. The spelling conventions adopted for the two Torres languages include the following:
g=[>]; n=[]; nw=[W]; q=[kW]; d=[z]; r=[:L]; o=[f]; =[o]; =[b]; e=[6]; =Ltg[7],
Hiw[e]; = Ltg[e], Hiw[I].
Alexandre Franois
the many TAM categories about sixteen present in each of these two languages,
only two appear to trigger seemingly paratactic structures of this sort. One belongs to
the domain of irrealis modality, and is called the Subjunctive (sbjv); this appears as on
in the Hiw sentence (1). The other belongs to the set of realis TAM markers, and more
precisely to the perfect aspect; due to its particular properties, I propose to label it the
Background Perfect (bkpf) expressed as vesi in (2).
Ultimately, these two TAM categories each one for distinct reasons and through
different mechanisms can be said to convey the status of their clause as being
syntactically subordinate to another main clause. In other words, apparently paratactic
sentences such as (1)(2), even though they may lack any formal conjunction, can still
be said to be formally marked as subordinate: this information is conveyed by the TAM
marking on the verb, instead of being coded by clause linkers. Thus, the first clause in
(1) is marked as a dependent clause by the presence of the Subjunctive; likewise, the
first predicate phrase of (2) is formally identifiable as a subordinate (relative) clause
through the use of the Background Perfect.
1.3 Formal properties, functional mechanisms

In this study, I intend not only to establish the empirical facts for these two undescribed
languages, but also to propose a functional interpretation and discussion. I will adopt
a functionalist perspective on this set of linguistic facts, and suggest that the syntactic
effect of these two TAM categories, rather than just a purely formal property, can be
shown to result from their semantic and pragmatic values.
In a nutshell, the core function of the Subjunctive in the Torres languages4 is to
represent a virtual state of affairs, with no further information on modality or illocutionary force. This pragmatic indeterminacy is fundamentally the reason why a subjunctive clause will need to attach itself to another clause, which provides it with the
modality value it lacks. Likewise, the Background Perfect can be defined as a perfect
aspect which demotes its predicate from the scope of the informational focus. Due to
this backgrounded status, the predicate then needs to attach itself to another element
under focus, in order to form a valid utterance.
The two cases thus appear to follow similar logic. Intrinsically, each of these two
TAM markers combines its purely semantic value (in terms of aspect or modality)
. Obviously, the Subjunctive category of the two Torres languages owes its name to very
similar mood categories found in other languages (Noonan 1985:91), notably Indo-European
ones. This being said, as a principle, the observations made in this article must be understood as
applying primarily to the TAM category specific to the Torres languages hence the uppercase
in its label, following the usage in Comrie (1976:10). My intention is not to make any general
claim about the properties of a universal category subjunctive supposing such a cross-linguistic
category indeed exists (see Haspelmath 2007).
with some pragmatic property. In both cases, this property corresponds to a form of
pragmatic demotion lack of a specific illocutionary force for the Subjunctive, lack
of focal status in the case of the Background Perfect and in both cases, this demotion results in a form of clause dependency. While they are ultimately grounded in
the pragmatic dimension of discourse, these two TAM-based strategies ultimately
also affect the formal syntax of the sentence, as they constitute a routinized device for
encoding clause subordination.
The following sections are organized as follows. Section 2 will provide a brief syntactic overview of the two Torres languages, and will pay special attention to overtly
marked clause-linking strategies whether subordination, coordination or verb
serialization. Section 3 will then examine in detail the functional and formal behaviour of the Subjunctive, and Section4 will be dedicated to the subordinating power of
the Background Perfect.
2. Clause linking in the Torres languages: Anoverview

I will begin this study with an overview of the syntactic structures of the two Torres
languages, with special focus on clause linking strategies.
2.1 Syntax of the simple clause

Like their Oceanic neighbours of Vanuatu, Hiw and Lo-Toga possess an accusative
alignment system, and follow a strict SVO constituent order.
2.1.1 Argument coding

Subjects take the form of noun phrases or free pronouns preceding the verb, and are
not cross-referenced on the predicate itself. Likewise, direct objects usually leave the
verb form unchanged (3a), except when they have human reference. In the latter case,
the verb form becomes marked for transitivity (3b), and sometimes bears a suffix
cross-referencing the object (3c):
(3) a. Ltg Nke na it n en-we mhe
si.

1sg prf1 see art house poss-3pl prf2

Ive seen their house(s).
b. Ltg Nke na ise
kemi si.

1sg prf1 see:tr 2pl prf2

Ive seen you[+human].
c. Ltg Nke na isihe
si.

1sg prf1 see:tr-3pl prf2

Ive seen them[+human].
Alexandre Franois
2.1.2 Tense-Aspect-Mood categories

Alongside its arguments, a well-formed verb phrase entails the presence of a marker
coding for aspect, mood and polarity. These three parameters are subsumed under
a single paradigm of portmanteau morphemes. For example, the marker labelled
Complete (a postclitic piti in Hiw, a proclitic mat in LoToga) simultaneously
encodes aspectual meaning (completed event), modal meaning (indicative), and
polarity value5 (affirmative):
(4) a. Hiw Sise motrig piti.

3pl sleep:pl cplt
b. Ltg Nihe mat
metur.

3pl cplt
sleep

Theyve already slept.
The category of tense properly speaking is not marked in these languages. Although
the paradigm of verb modifiers should thus be designated, strictly speaking, as AMP
markers (for Aspect-Mood-Polarity), throughout this chapter, I shall nevertheless
continue to use the widespread abbreviation TAM (for Tense Aspect Mood), for the
readers convenience.
The two Torres languages possess sixteen formally distinct6 TAM categories. The
realis markers (see 4.1) include the Stative, the Imperfective, the standard Perfect, the
Background Perfect, as well as the Complete, the Recent Perfect, and the Realis Negative. The irrealis categories (see 3.3) include the Future, the Prospective, the Potential, the Apprehensive, the Subjunctive, the Counterfactual, and the Irrealis Negative.
Finally, two categories labelled Aorist and Time Focus span the realis and the
irrealis domains.7
The Aorist is a particularly polysemous category, found in the Torres8 as well as
several of the Banks islands to the south (Franois, in press). It covers several meanings, both realis and irrealis, including narrative, sequential, generic, prospective,
imperative and conditional. A possible description of the Aorist would be to consider
. The morphosyntax of the negation will be mentioned in 4.2.2.2.2.
. Many of these TAM morphemes are morphologically complex, and sometimes discontinuous as in the case of the Perfect nasi in (3). See also the discussion in 4.1.2.
. See Franois (2003) for a detailed semantic analysis of a highly similar (and partly cognate)
TAM system, that of the neighbouring language Mwotlap.
. The morphology of the Aorist in the Torres languages is complex (Franois, 2009: 189).
First, it is coded by a set of preverbal markers that vary for person and number (Ltg 1sgke,
2sgwe, 3sgni); second, these preverbal markers are generally deleted in the presence of a
free personal pronoun, in which case the surface form of the Aorist is simply [see ex.(28b)].
In the present article, I will only mention the Aorist in the gloss when it is relevant to the discussion, otherwise the verb will simply be given as unmarked for TAM.
it as a zero verbal category that is underspecified with regard to tense, aspect and
mood; this would account for both its great flexibility, and its compatibility with modally bound dependent clauses (12). Interestingly, the Subjunctive [Hiwon, Ltgv(n)]
can be analysed along similar lines in terms of semantic underspecification except
that it is restricted to irrealis clauses (see 3). As we will see later, the two markers
can be synonymous in certain contexts compare (12) and (38) for modality-bound
complement clauses; or (32f) and (35b) for the hortative. Yet even though the Aorist
and the Subjunctive show a certain degree of functional overlap, the Subjunctive will
be preferred when the semantic status of the subordinate clause is explicitly irrealis or
generic.
2.1.3 Syntactic categories and their predicativeness

Another important characteristic of the Torres languages and more generally of
many of the areas languages (Franois 2005a) is the diversity of parts of speech compatible with the predicate function. A predicate head9 need not be a verb: it can be an
adjective, a noun, a numeral, etc. For example, a nominal predicate takes the form of
a simple noun phrase in a direct (zero) construction, with no copula whether it be
equational (type XistheN) or ascriptive (XisanN).
(5) Hiw Nine {rekn-ok}.

3sg mother-1sg

She (is) my mother.
When the subject is omitted, the result is a clause that consists of just a single noun
phrase:
(6) Hiw () {ne wake}.

art canoe

(Its) a canoe.
[direct noun predicate]
Several other word classes may also be directly predicative. This includes locative
phrases whether in the form of adverbs [e.g. the interrogative where in (7)] or
prepositional phrases [see y kn- in (54)] as well as certain invariant words [e.g. the
existential predicate not exist, lack in (7)].
(7) Ltg Ne hen-wvot mino {ev}? Nie {tateg}.

art knife
my where
3sg neg:exist

Where (is) my knife?
It is not here.
Direct predicativeness (Lemarchal 1989; Launey 1994) constitutes an important

property of parts of speech in the Torres languages, which will later prove crucial in
the syntactic analysis of the Background Perfect (4.2.2.2).
. In Examples (5)(7), the limits of the predicate phrase are indicated by curly brackets.
Alexandre Franois
2.2 Subordination
Hiw and Lo-Toga possess a wide array of morphological devices for encoding the syntactic relations between a subordinate and a main clause. I will successively examine
the coding of complement clauses (2.2.1); conditional clauses (2.2.2); relative clauses
(2.2.3); and adverbial time clauses (2.2.4).
2.2.1 Complement clauses

The Torres languages have a quotative particle (Hiw tom, Ltg t) for introducing direct
reported speech. It can be used as the unique predicate of the clause, or in combination
with a speech verb:
(8) Hiw Tema-ne yur-mi-e tom Ye nne? Tom Noke!

father-3sg ask-tr-3sg quot who that quot 1sg

Her father asked her [saying]: Who was that? [She said] That was me!
The same quotative particle is used to introduce indirect speech. Therefore, despite its
obvious origin as a quotative, it is better analyzed, synchronically, as a complementizer. Indeed it can combine not only with verbs of speech, but also with all sorts of
verbs governing a clause complement:10
(9) Hiw Noke tati mng, noke ttm tom ne g
kye.

1sg neg steal 1sg think comp art thing my

I didnt steal it, I thought (that) it was mine.
(10) Ltg Ne n-wi ni
holq me,

art devil aor:3sg return hither

ni
it t
nihe ve toge.
aor:3sg see comp 3pl ipfv stay
The devil came back, and saw (that) they were there.
If the complement clause is realis, its predicate is normally compatible with any realis
TAM marker (Perfect, Stative, Imperfective), with no particular restrictions. The
same applies if the clause is semantically irrealis but is modally independent from
the main clause. For example, a main verb meaning believe would allow the complement clause to take essentially the same TAM markers as in an independent clause. As
we shall see in 3.3.1, there are quite a few irrealis markers which correspond to this
definition, for example the Potential (Hiwta, Ltgsi):
. This process, whereby the quotative particle has generalised its use to cover the whole
functional array of a complementizer, is widespread in the area. The process may be compared
to the typologically common process whereby complementizers originate in a verb of speech
(see Heine & Kuteva 2002; Chappell 2008).
(11) Ltg N ige w ne, nke dem t

nke si
gn.

art fish like this 1sg think comp 1sg pot:aff eat

This sort of fish, I think I can eat.
Conversely, certain types of predicates typically, verbs of volition and manipulation

entail that the irrealis complement clause be bound to the main clause with respect to
modality. In that case, the choice of TAM marking in the complement clause becomes
essentially restricted to two possible categories: the Subjunctive [see (37)(38) below]
or the Aorist (12).
(12) Ltg Ne lie-k
na n-ih t
ke
tun d
sa n-wil.

art mind-1sg stat want comp aor:1sg buy from m chief

Id like to buy it from the chief.
Purpose clauses are also constructed along the same patterns (Comp + Aorist or
Comp+ Subjunctive): see (39)(40) and (62)(63) below. Once again, in this irrealis
context, the Subjunctive and the Aorist are essentially equivalent (cf.2.1.2).
The combination of the complementizer and Aorist markers has also grammatica
lized, in Lo-Toga (but not in Hiw), into a TAM category in its own right, called the Prospective. Its meanings encompass the desiderative (want to do), the deontic (should
do, must do), the prospective proper (be about to do)11 Although it originally
incorporates the complementizer t, this Prospective marker can appear on the main
predicate of an independent clause as in (32c) below which shows that it has lost
any connection with clause dependency. This is also proven by the possibility of combining the Prospective (here twe Prosp:2sg) with the complementizer t in the same
sentence:
(13) Ltg

Tate
pero t
nike t
we hadit.
neg:real long comp 2sg prosp 2sg be.initiated
[lit. Its not long before youre going to be initiated]
You are soon going to follow the initiation rituals.
The category of the Future is in turn a composite morpheme, which combines the
Prospective (t + Aorist) with the particle ake see (15), (26), (32a).
2.2.2 Conditional clauses

Conditional clauses may again involve the same complementizer (Hiw tom, Ltg t),
which is here translated if :
(14) Hiw Tom ike gengon n-w, ne ga tat
qisi tirtir ike.

comp 2sg aor:eat first art kava neg:irr hit:tr strong 2sg

If you eat first, the kava wont have any strong effect upon you.
. Both the morphology and the semantics of the Lo-Toga Prospective are narrowly similar
to those of the Prospective in Mwotlap (Franois 2003:218257).
Alexandre Franois
The conditional subordinator also displays longer forms which are derived from the
complementizer. One thus finds the (semantically non-compositional) combination
Hiw tom + n-w like tom-n-w or tom-n-w-tom meaning if see (49). Lo-Toga has
exactly parallel forms, either morphologically transparent (t + w like tw [t7w7]
if ) or with a slight vowel change tw tew [t6w7] ~ tew-t [t6w7t7] see (15), (48).
Several TAM categories can be found in the protasis of a conditional sentence:
Aorist; Subjunctive; Counterfactual (15):
(15) Ltg Tewt te
not ne met-ne si,

if
ctfc1 hit art eye-3sg ctfc2

nie t n ake mteqa !

3sg fut1 3sg fut2 blind obl
If they had hit his eyes, he would have become blind!
We will see below (3.5.2) that, while conditional constructions can make use of a
conjunction, they are also regularly coded by the Subjunctive alone. This TAM marker
is the only one capable of replacing a conditional conjunction.

Relative clauses are marked by a variety of morphological devices. The role of relativizer can be played, in both languages, by the (polyfunctional) form pe:
(16) Ltg Noke t
ke v k it ne gehuh

1sg prosp 1sg go 1sg see art coconut.crab

pe ve kerkur
tle
nk.
rel ipfv ipfv~crunch person there
Ill go and have a look at that coconut crab which devours people.
The relativizer function can also be played by phonologically heavier forms; these
combine several morphemes in ways that are not always semantically compositional.
Thus one finds a relativizer Hiwpetom ~ Ltgpet, etymologically the combination
{relativizer + complementizer} [also see (41) below]:
(17) Hiw Sise mi n-sa
tir n-wute petom sise

3pl with poss-3pl true place rel
3pl

toge
ie
y nwrwn.

stay:pl obl:adv loc forest
They have special places of theirs, where they dwell in the forest.
Lo-Toga also combines the relativizer pe with the comparative w like (Ltgpew),
generally with virtual or generic referents (whoever):
(18) Ltg Ni
ole ne wuhe

aor:3sg give art potion

hi hen-were pew na mo.

dat people rel stat sick

He provides medicine to who(ever) is sick.
In fact the form w alone (without pe) can also serve as a relativizer in Lo-Toga see(42).
To sum up, the forms of the relativizer in Hiw are pe or petom; those in LoToga are pe,
pet, pew or w.
Finally, despite the wealth of these relativizers, it is also common for relative
clauses to lack any formal subordinator, provided the status of the whole phrase as
a dependent clause is visible on the verbs TAM marking. This ability to constitute a
relative clause with no relativizer is attested only with two TAM categories, precisely
those which form the topic of the following sections: the Subjunctive (3.5.2), and the
Background Perfect (4.2.2.1).
2.2.4 Adverbial time clauses

Adverbial time clauses are often formed with a noun meaning time, moment: Hiw
tamern ~ (take)timern, Ltg mowe. The time clause can then be construed as a relative clause (see Thompson & Longacre 1985:179) i.e. when = literally the time in
which
(19) Hiw Ike yo-ie
ti timern

2sg see-3sg prf time
pe kimire n
rel 2du stat
pe tekn-wa te
rel people from
yumegov q, tamern
young still time
y vny ve tetayw.
loc village ipfv celebrate
You met her (at a time) when you both were still young,
as the villagers were celebrating.
But it also commonly happens that the same word appears on its own, with no overt
relativizer:
(20) Ltg Mowe
ne tarepi ke
mat th pah,

time/when art body canoe cplt carve finish

pahvn ge
rak ne hm
in.
then aor:pl make art outrigger its
Once the body of the canoe is carved, [then] one makes the outrigger.
It could be proposed to see mowe here still as a noun time followed by a relative clause
with no relativizer; however, such relative clauses, as mentioned in 2.2.3, are normally
Alexandre Franois
restricted to two TAM markers. The presence in (20) of another TAM category (mat
Complete aspect) calls for another syntactic analysis: namely, that the noun mowe has
been grammaticalized into a subordinator when.12
In addition, Lo-Toga also has a genuine time subordinator noneg when, as:
(21) Ltg Noneg nie ve vin-g
ne megole,

as
3sg ipfv climbappl art child

ni
hur ne vete sise.
aor:3sg sing art song one
As she was climbing with her baby, she began to sing a song.
We shall see other cases where time clauses lack an overt subordinator, the relation of dependency being reflected only by the TAM marking on the verb: the
Subjunctive (3.5.2).
2.3 Coordination
The Torres languages make relatively little use of coordination, and generally prefer
resorting to subordinating or serialising strategies.
Following a typologically common trend (Stassen 2000), the Torres languages
usually form the equivalent of coordination between two noun phrases by using the
comitative preposition mi with:
(22) Hiw tema-ne mi
rekn-a-ne

father-3sg with/and mother-3sg

his father with/and his mother
Quite originally, Lo-Toga has extended the use of this comitative preposition to coordination between any two phrases, including two prepositional phrases (23) or two
clauses (24):
(23) Ltg Noke na melekelake pi
megole mke, mi pi
lgie
mke.

1sg stat happy
about child your and about wedding your

Im delighted about your baby, *with/and about your wedding.
(24) Ltg Ne n-wi si dahia

ne tle,

art devil pot harm obl art person

mi nihe si kur
veri ne tle.
and 3pl pot crunch also art person
Devils can harm people, *with/and they can even devour people.
. This pattern, whereby a noun meaning time, moment grammaticalizes into a subordinator, is commonplace in the area. Mwotlap does the same with (vt)mah (Franois 2003:26),
aswell as Bislama with taem <Eng. time (Crowley 2004:188).
This functional extension of mi is unique to Lo-Toga, and is quite marginal in the

language. It would be impossible in Hiw, where mi is still used strictly as a comitative
preposition with a noun phrase. In order to coordinate two clauses, Hiw instead uses
an adverb pavn then:
(25) Hiw Timern ne, nine n
n-wotoy

time
that 3sg stat short
k,
little
pavn n uy ena n
teytoy.
then art hair her stat plaited
At that time, she was a little short, and her hair was plaited.
Other coordinate constructions include words for but (Hiw/Ltgpa), or (Hiw titom,
Ltghit), or because (Hiw [ur] npe [tom], Ltg naw).
2.4 Verb serialization

Finally, this rapid overview of clause linkage in Hiw and Lo-Toga should mention, albeit briefly, verb serialization. Serial verbs in these two languages have two
distinct forms.
The structure which is known as nuclear-layer serialization (Foley & Olson 1985;
Crowley 1987, 2002) consists in joining two verb radicals together with no intervening
element, as if through a process of lexical compounding. The resulting macro-verb
behaves in many regards as a single verbal unit, taking no more than one subject, one
object and one TAM morpheme:
(26) Ltg

T w ake vese vah noke

ne i
ige.
fut1 2sg fut2 say show 1sg obl art name fish
[lit. You will say show me of fish names]
You will teach me the names of fish.
In this pattern of nuclear-layer serialization, the second verb modifies the first verb,
both semantically and syntactically (Bril 2004; Franois 2004).
The Torres languages have also developed a pattern of core-layer serialization,
whereby two verbs follow each other in a single clause, yet each one bears its own TAM
marker (or at least the proclitic part, in the case of discontinuous markers). This TAM
marker is normally the same for the two verbs:
(27) Ltg Noke na vn na vivd si
l
n-we ror.

1sg prf1 go prf1 pray prf2 loc house holy

I went to pray in the church.
This is an example of concordant marking of tense-aspect-mood, to use the terms in

Aikhenvald (2006:42).
Alexandre Franois
The latter pattern is especially used when V1 is a verb of motion (go, run) or
of posture (sit, stand). One of the derived uses of this serial structure, involving a
posture verb in the V1slot, encodes progressive aspect:
(28) a. Ltg

Noke ve gel ve hiar ne hen-wvot mino.

1sg ipfv stay ipfv seek art knife
my
[lit. I stay I search my knife]
I am looking for my knife.
This progressive construction involves either the Imperfective ve (cf.4.1.1) as in (28b),

or the semantically neutral aspect called Aorist (2.1.2). In this case, the very special
morphology of the Aorist (fn.8 p. 504) makes the serial pattern less easy to detect:
(28) b. Ltg Noke () gel ke
hiar ne hen-wvot mino.

1sg aor stay aor:1sg seek art knife
my

I am looking for my knife.
In all these cases, the sharing of arguments and of TAM marking whether it occurs
once or is repeated clearly shows that we are dealing with serial verb constructions,13
and hence with single clauses (Durie 1997; Bril 2004). As such, these structures do
not illustrate patterns of clause linking strictly speaking, but rather linkage strategies
between predicates.
3. The Subjunctive: In search of an illocutionary force

The preceding section showed the wide array of formal devices used by the two Torres languages to encode dependency relations between clauses or predicates, whether
in the form of verb serialization, coordination, or subordination. Despite the wealth
of these clause-linking devices, two TAM categories, the Subjunctive and the Background Perfect, present atypical behaviour: these two markers, and only these two,
show a strong tendency not only to combine with subordinate clauses, but also to
directly encode clause dependency, even in the absence of any subordinating device
(see 1.2).
I shall detail these two cases successively: the Subjunctive in the present section,
and the Background Perfect in Section4.
. The Resultative constructions of Lo-Toga share certain properties with these serial verb
constructions, yet they must be analyzed as a different structure: see 3.6.
3.1 Presentation
The Subjunctive was first exemplified in sentence (1), reproduced below:
(1) Hiw Ne temt on t
yaqe me
n-w ne,


tekn-wa voyi.

people aor:run.away
[lit. Thedevil would appear like this, people ran away]

The behaviour of the Subjunctive is parallel in Hiw (form on) and in LoToga (forms
v ~ vn).14 One question arises: what exactly is the mechanism that makes this
Subjunctive marker so intimately connected with subordination? Why is it that all
other TAM categories including the various irrealis markers require the presence
of overt subordinators, whereas the Subjunctive can easily do without them? Could
one go as far as to consider this morpheme intrinsically endowed with the power of
subordination?
The position I will defend is the following. The syntactic properties of the
Torres Subjunctive, in terms of its ability to encode subordination, can be understood
as an indirect consequence of fundamentally semantic properties: this marker codes
an event as merely irrealis, with no further specification of any illocutionary force. This
modal and pragmatic indeterminacy accounts for the inability of the Subjunctive alone
to constitute well-formed utterances, and ultimately helps explain its strong tendency
to trigger syntactic dependency between clauses.
3.2 A note on irrealis sentences

An irrealis sentence involves the representation of a virtual situation which has no
reality other than that of a mental construct in the speakers discourse. Unlike realis
events, whose existence is a fact and which may therefore be recounted as such, an
irrealis situation cannot simply stand on its own: in order to form a pragmatically wellformed utterance, it needs to be embedded in some form of secondary predication, be
it a deontic predicate, an epistemic judgment, or a speech act of some sort.
. Despite the formal difference between Ltg v [7] ~ vn [7n] and Hiw on [fn], it is in fact
likely that the two forms are cognate. According to regular vowel correspondences (Franois
2005b), they could reflect a proto-form *ani, of uncertain origin. A link with Proto Oceanic
*pani give is not implausible, although it raises semantic problems. The connection between
give and subjunctives does not seem to be widely supported in other languages (see Bybee etal.
1994), and the etymology of English if (<OEif), sometimes mentioned as connected to giefan
give, is disputed.
Alexandre Franois
For example, let us consider the state of affairs {Babyget sick}. When one refers
to a realis event like (29), that state of affairs can easily be stated and provided with
various semantic properties, such as time coordinates and truth value:
(29) Eng
Baby got sick again last week.
Conversely, the same state of affairs in an irrealis context (i.e. the possibility that
Baby gets sick at some point in the future) will not be able to constitute, by itself, a
complete utterance. Even the English sentence (30), which is syntactically complete
and grammatical, appears to be an ill-formed utterance from the pragmatic point
of view:
(30) Eng
Suppose Baby got sick.
A sentence like (30) is felt to be incomplete, as if waiting for the rest of the sentence in
order to be interpretable.15
To use the terminology of Simon Diks Functional Grammar, a sentence like (30)
does little more than merely represent a possible State of Affairs i.e. the conception
of something that can be the case in some world (Dik 1989:46). In order to constitute
a well-formed utterance, such a virtual situation needs to be encapsulated within some
type of higher-level linguistic operation such as aspect and time operators that would
provide it with the status of a Possible fact; or illocutionary force and modal values
that would make it a pragmatically complete Speech act.
For example, the virtual state of affairs mentioned above could be incorporated
within various forms of speaker-centered speech acts e.g. apprehension, wish, prediction, etc.:
(31) a. Eng I fear Babymight get sick.
b. Eng I wish Babygot sick!
c. Eng [Given what I know, I hereby predict that] Babywill get sick.
It may also take the form of a question, anchoring the modal center in the addressee
(31d):
(31) d. Eng [According to you] will Babyget sick?
It may also be encapsulated within a conditional structure, either as the protasis (31ef)
or as the apodosis (31g):
. The pragmatic incompleteness of an English sentence like (30) is confirmed by historical

evidence: in English-based Melanesian Pidgins such as Bislama or Tok Pisin, the imperative form suppose has grammaticalised into a subordinator sipos/sapos meaning if (Franois
1997:22; Mhlhusler et al. 2003: 24; Crowley 2004:189).
(31) e. Eng In case Babygets sick, he will need to take this medicine.
f. Eng Every time Babygets sick, he tends to recover within two or three days.
g. Eng If he goes out in that cold weather,
[I hereby predict that] Babywill get sick.
In all of these sentences, the virtual situation which by itself has no pragmatic value
comes explicitly incorporated within a higher-level predication involving a specific
speech act or modal attitude (prediction, wish, apprehension). This is what makes
them capable of forming a valid utterance, unlike (30) above.
3.3 Two types of irrealis markers in the Torres languages

These preliminary remarks about the nature of irrealis utterances should help understand the facts in Hiw and Lo-Toga. In each of these two languages, a semantically
irrealis verb can be associated with two types of TAM markers: (a)modally specified
markers, (b)a modally underspecified marker, the Subjunctive.
3.3.1 Modally specified irrealis TAM markers

One set of irrealis TAM markers consists not only in representing a state of affairs as
virtual; they also inherently convey a specific modal value and/or speech act (such as
prediction, order, warning, etc.) within which this state of affairs is logically embedded. In a way, these modally specified morphemes could be described as semantically
composite, as they combine the [+irrealis] feature with some other modal specification. It is therefore not surprising following the reasoning in 3.2 that they should
be capable of forming pragmatically well-formed, complete utterances.
In Lo-Toga,16 this first set of irrealis markers includes the affirmative Future
tnake in (32a) and its negative counterpart tat in (32b); the Prospective tni in (32c);
the affirmative Potential si in (32d) and its negative counterpart tatho in (32e); the
Aorist used for orders in (32f); the Apprehensional mik in (32g).
(32) a. Ltg Nie
t
n
ake metur l-en-we
mino.

3sg:indep fut1 3sg:s fut2 sleep loc-house my

(I predict/promise) He will sleep at my house.
b. Ltg Nie
tat
metur l-en-we
mino.

3sg:indep neg:irr sleep loc-house my

(I predict/forbid) He wont sleep at my house.
. The forms given in this paragraph are for Lo-Toga. Hiw has corresponding markers for all
of them, except that it does not formally distinguish between the Future (32a) and the Prospective (32c).
Alexandre Franois
Nie
t
ni
metur l-en-we
mino.
3sg:indep prosp 3sg:s sleep loc-house my
(I recount somebody elses desire)
Hed like to sleep/Hes supposed to sleep at my house.
d. Ltg Nie
si
metur l-en-we
mino.

3sg:indep pot:aff sleep loc-house my

(I allow or state a factual possibility)

He can sleep at my house.
e. Ltg Nie
tat
ho
metur l-en-we
mino.
c. Ltg

3sg:indep neg:irr pot:neg sleep loc-house my

(I state a factual impossibility)
He cannot sleep at my house.
f. Ltg Nie
ni
metur l-en-we
mino!

3sg:indep aor:3sg sleep loc-house my

(I order/suggest) Let him sleep at my house!
g. Ltg Nie
mik metur l-en-we
mino!

3sg:indep appr sleep loc-house my

(I present a situation as undesirable)
(I fear) he might sleep at my house!
3.3.2 The Subjunctive, a modally underspecified TAM marker

In addition to these modally specified markers, the two Torres languages possess
another irrealis marker with slightly different properties. This proclitic, which I label
the Subjunctive, belongs to the same morphosyntactic paradigm as the TAM markers
cited in (32ag).
The reason for treating this morpheme separately is not morphological, but
semantic. In itself, the Subjunctive provides the clause with no specific modality nor
illocutionary force of any kind, and appears to convey the sole meaning [+irrealis]. To
use Diks terms, it does nothing more than to represent a purely virtual State of affairs.
It is therefore hardly surprising (following 3.2) that the Subjunctive alone is unable to
form a pragmatically valid declarative sentence:
(33) a. Hiw *N on mitir
yn-we
kye.

3sg sbjv sleep:sg loc-house my
b. Ltg *Nie vn metur l-en-we
mino.

3sg sbjv sleep loc-house my

(for him to sleep at my house)
A declarative sentence like (33ab) would be felt to be truncated or unfinished, in a

way very similar to (30) in English. This can be explained if one remembers that a virtual state of affairs can only form a complete sentence if it is embedded in a higher-level
linguistic operation. While the various irrealis morphemes cited in (32ag) incorporate
that linguistic operation intrinsically, this is not the case for the Subjunctive (33ab),
which remains modally under-specified.
This semantic property of the Torres Subjunctive entails an important corollary:
its high potential for syntactic dependency. Due to its pragmatic incompleteness, a Subjunctive clause will need to hook on to some other clause or predication operator,
in order to form a valid sentence. This essentially implies that the Subjunctive has a
strong affinity with syntactic subordination hence my choice for its name. In certain
cases, this affinity means that the Subjunctive will combine with/be required by formal
subordinators, in a way reminiscent of the subjunctives found in European languages.
But quite often and crucially for the topic of the present volume the syntactic consequence will be that the Torres Subjunctive is capable of creating a relation of dependency between two clauses, even in the absence of any specific subordinator.
These issues will form the essentials of the discussion in 3.5. But before we turn
to them, it is necessary to address the paradox of the hortative.
3.4 The special case of the hortative

The preceding paragraphs may have given the impression that the Torres languages
make it impossible for an utterance to consist of a single clause marked as Subjunctive.
Even though this may be indeed very close to the truth, there is in fact one exception
to this generalization: the case of third-person hortatives.
When the speaker orders that an action be performed by the addressee, he will
use an imperative. In the Torres languages this may be marked by the Aorist, or more
often by the verb alone:
(34) a. Hiw T
me! ~ Wt
t
me!

go:sg hither aor:2sg go:sg hither
b. Ltg Vn me! ~ We
vn me!

go hither aor:2sg go hither

Come here!
When the person in control of the desired state of affairs is distinct from the addressee,
the corresponding speech act, described typologically as a hortative (van der Auwera,
Dobrushina & Goussev 2008), may also be coded by the Aorist, as in (32f) above.
In addition, for third-person hortatives, the two Torres languages can also use their
Subjunctive:
(35) a. Hiw N on mitir yn-we
kye !

b. Ltg Nie vn metur l-en-we
mino !


(I order/suggest) Let him sleep at my house!
Alexandre Franois
This functional equivalence between the Aorist and the Subjunctive is also found with
third-person optatives:
(36) Ltg Ne ten-wte vn toge m-ke !

art peace
sbjv stay with-you

(I wish) May peace be with you!
This use of the subjunctive for hortatives or optatives is typologically common,17 as witnessed by Latin Veniat! Let him come! or Pax sit semper vobiscum May peace be always
with you (cf. Ernout & Thomas 1953:239). However it seems to be at odds with the definition I gave of the Torres Subjunctive in 3.3.2, where it was stated that this marker does
not convey any speech act value. If this is so, then where does the illocutionary force of
these hortative or optative utterances find its source? And how is it possible that sentences
such as (35ab) and (36) are perfectly well-formed, while (33ab) was ungrammatical?
The answer to this paradox does not lie with the Subjunctive itself: obviously, if
hortative/optative modality were intrinsically built into this marker, then it should
convey it in every sentence, and an utterance such as (33ab) should be correct. This
means we need to take seriously the only difference that distinguishes (33) from (35):
the prosody which is very roughly represented here by the punctuation. On the one
hand, the prosodic contour of (33ab), that of a declarative statement, results in the
pragmatic incompleteness of the sentence. On the other hand, the prosody of (35ab),
which is characteristic of orders and exclamatory sentences a high pitch plateau
ending in an instant fall makes the sentence grammatical.
In my interpretation, the particular suprasegmental profile of the sentence is the
locus where the needed illocutionary force is lodged, and must be sought. The ungram
maticality of (33ab) showed that the function of the Subjunctive, namely the mere
representation of a virtual State of affairs, did not find enough support in the declarative modality to constitute a well-formed utterance. Conversely, what (35ab) reveals is
that an intonation typical of orders and exclamations, because it is markedly anchored
in the speakers desires and emotions, is sufficient to provide that virtual State of affairs
with the modal value and illocutionary force it needs to form a correct utterance.
Semantically, this formal asymmetry indeed makes sense. Such a mental construct
as a virtual state of affairs can hardly be stated in any way; but it can still be represented
in an emotional perspective which is what exclamatory utterances tend to mimic. This
contrast accounts, respectively, for the incompatibility between the Subjunctive and the
declarative modality, and for its affinity with the intonation of orders and exclamations.18
. See Noonan (1985: 54): Main clause subjunctives tend to be used in modal, hortative,
or imperative senses.
. A similar hypothesis was proposed in Franois (1997: 66) to explain why certain languages encode their imperative with some linguistic structures (noun phrases, infinitives, subjunctive clauses) which would constitute an ill-formed declarative sentence. Despite their
In sum, (35) and (36) constitute no exception to the general principles outlined
in 3.3.2, namely that an irrealis event can constitute a sentence if, and only if, it is
involved in a modal predication of some kind. But while every other irrealis TAM
marker in the Torres languages has an inbuilt illocutionary force that makes it welldesigned for the formation of a valid utterance cf.(32ag) this is not the case with
the Subjunctive, which is under-specified in this regard. As a result, the only way for a
Subjunctive verb to form a correct sentence, is to receive its illocutionary force from
outside. Most of the time, this external source for the coding of modality will correspond to a different clause, that syntactically belongs outside the Subjunctive clause;
this point will account for the strong ties of this marker with syntactic subordination
(3.5). As for (35ab) and (36), they illustrate a more particular case, where the specific illocutionary force is lodged outside the verbal form strictly speaking, yet still
has to be found within the formal limits of the clause itself: in its prosody.
All things considered, the functions of hortative and optative which are some
times fulfilled by the Subjunctive do not contradict its earlier description as a modally
under-specified, indeterminate irrealis marker.
3.5 From modal indeterminacy to syntactic subordination

In sum, the Subjunctive is the only irrealis TAM category of the Torres languages
which does not inherently convey any modal value or illocutionary force. Unless it
receives the latter from some modally charged intonation pattern, it is therefore unable
to constitute a valid utterance on its own.19
The principal corollary of this description is the strong ties that exist between
this irrealis TAM marker and the syntax of clause dependency. I will first review the
various cases where the Subjunctive combines with a clause that is already marked
formally as subordinate: complement clauses, relative clauses, conditional sentences,
etc. In a subsequent section (3.5.2), I will show that the presence of an overt subordinator is in fact not even necessary for the Subjunctive to be able to encode syntactic
dependency between clauses.
3.5.1 The subjunctive combined with overt subordinators

Quite often, the backgrounded clause is already marked as dependent by means of a
subordinator of some sort. This is the case, for example, when a clausal complement is
morphological variety, these linguistic structures all share a similar semantic function: the
representation of a virtual State of affairs. More recently, Nick Evans has addressed similar
issues under the cover term Insubordination (Evans 2007).
. This TAM marker corresponds to what Cristofaro (1998, 2003) calls a deranked verb
form: that is, a form of which the Italian Subjunctive would be another illustration that is
structurally different from those used in independent declarative clauses (Cristofaro 2008).
Alexandre Franois
introduced by means of a complementizer (Hiw tom, Ltg t), after a verb of manipulation or expectation (see 2.2.1):
(37) Hiw Marenage sa
gatt ti
tekn-wa

chief
their say dat people

tom ne veroye on pa.

comp art war sbjv finish
[lit. The chief asked the people that the war be stopped.]
The chief asked his people to stop the war.
(38) Ltg Dege

toge sh t
ne gengn

1incl:pl stay wait comp art food

v howse pah.
sbjv cooked finish
Lets wait till the food is completely cooked.
The same formal structure {complementizer + Subjunctive} is used for purpose clauses,
either with the same subject or with one different from the main clause.
(39) Hiw Sise my ti ne tt ga n-ot

3pl pull prf art root kava one

tom sise on ni
y gemoy.
comp 3pl sbvj drink loc mens.house
Theyve pulled out a head of kava so as to drink it in the mens house.
(40) Ltg Hr t

n-wule me
vete mi hr

3du prosp du:s return hither place poss 3du

t
nie v menwe.

comp 3sg sbjv breathe

They are going back to their place for him to get some rest.
Syntactic dependency may also be marked by a relativizer (2.2.3). The Subjunctive is

required when the relative clause is semantically irrealis and/or generic (cf. Eng. whoever):
(41) Hiw Tekn-wa petom sise on tati voyi wrog,

people rel
3pl sbjv neg escape through

ne temt qur-ise.
art Ghost crunch-3pl
All those who were unable to escape, the monster would devour them.
(42) Ltg N ve w nihe v vese hivi-ke,

art thing rel 3pl sbjv say dat-2sg

nike rn-t urv.

2sg listen properly

Whatever they may tell you, you must obey them.
As we saw in 2.2.4, adverbial time clauses generally take the form of a relative clause
hooked on the noun time, moment, with or without an overt relativizer. When the time
reference of the subordinate clause is irrealis or generic, the Subjunctive is expected:
(43) Hiw Taketimer n pe ne tay

on mt,

time
rel art person sbjv die

tite
tivig n op-ne.
1incl:pl bury art body-3sg
When(ever) somebody dies, we bury their body.
(44) Ltg Mowe w si

tle
v mo, dege
leklok m.

time rel some person sbjv sick 1incl:pl help with.3sg

When(ever) somebody gets sick, we help them.
(45) Ltg Mowe
kem
v da-togin, nike vn me

time/when 1excl:pl sbjv be-ready 2sg go hither

dege
n-wule.

1incl:pl return
When were ready, you can come here so we can go back together.
(46) Ltg Noke t

ke vn ke tugtuger rem mino

1sg prosp 1sg go 1sg watch
mother my

mowe
nie v metur.
time/when 3sg sbjv sleep
I will watch my mother when shes asleep.
An irrealis clause can be embedded within another irrealis clause, in which case the
Subjunctive percolates throughout. (47) shows three instances of v(n): the first one
(vn it) is due to the semantic status of the time clause as generic (whenever); the
next two (v sw v lew) constitute a second level of subordination, being a complement clause within that time clause [see also (51) below]. Incidentally, the string /v
sw v lew/ is a serial verb construction, of the type that requires the repetition of the
TAM marker (see 2.4):
(47) Ltg {Mowe

kemm vn it [t ne ho in

time/when 1excl:pl sbjv see comp art leaf its

v sw v lew pe
si]},
sbjv grow sbjv big already prf
al kem
ge
lio.
then 1excl:pl aor:pl dig.up
When(ever) we see that [the taros] leaves have grown (and become) big,
wedig it up.
Alexandre Franois
Finally, the protasis of conditional sentences (2.2.2) constitutes another structure

where the Subjunctive often combines with the subordinator if :

(48) Ltg Tew t ne li
v n-ih, nike si hue

if
art mind:2sg sbjv want 2sg pot paddle

o rw vete qe .
out out place deep obl
If you want, you can also paddle (your canoe) out there into the deep sea.
wtom se on vn y veroye, s
(49) Hiw {Tomn

if
3pl sbjv go:pl loc war
3pl

on qtn-og ne tay
ne tay
on qt},

sbjv kill:pl art person art person sbjv die:pl
sise viye
n op-se
me
se mok er qor.
3pl take:pl art body-3pl hither 3pl put on grave
{If/ When the population went to war, and many people were killed and
died}, their bodies were then collected and deposited in stone graves.
Note that the Subjunctive never occurs in the apodosis of such conditional sentences,
because this is a section of the sentence which needs to have its own illocutionary
force as in (31g) above.
These examples (37) to (49) all illustrate the strong links of the Subjunctive with
subordinate structures. In each case, the Subjunctive verb phrase does no more than
represent a virtual state of affairs which is, in itself, deprived of any inherent modal
value. What then makes the clause interpretable, is its insertion here via overt subordination within a higher level predication, which is in turn specified for modality
and illocutionary force.
3.5.2 The subordinating effect of the Subjunctive alone

Crucially, while the Subjunctive marker often combines with a subordinator, it turns out
that it is also capable of creating an effect of syntactic dependency on its own. A clause
marked as Subjunctive will spontaneously tend to develop a relationship of syntactic
dependency with a matrix clause, even in the absence of any formal subordinator.
The most frequent case of spontaneous subordination is when the Subjunctive
alone marks the protasis of a conditional sentence. Indeed, the conditional conjunctions if illustrated in 2.2.2 and in (48)(49) become optional when the verb is
marked with the Subjunctive. In the majority of cases, the TAM morpheme is sufficient to encode the subordinate status of the clause:
(50) Ltg Nke v vese t
ne genegone v vn,

1sg sbjv say comp art war
sbjv go

ne genegone t
ni vn.
art war
prosp 3sg go
(If) I say that the war (must) go on, then the war will go on.
(51) Hiw Ik on r-n- tom s

g
on r-ak ti,

2sg sbjv hear comp some thing sbjv make prf

ike ta tw
ne wt eye.
2sg pot compose art song obl
(if) you hear that some event has happened,

you can compose your song aboutit.
(52) Hiw Ik on sr--ie

on yoqse, n ptg nne!

2sg sbjv spear-3sg sbjv miss art shame that

(If) you try to spear him and you miss, then shame on you!
Rather than hypothesizing a form of conjunction ellipsis, it is probably more accurate

to suggest that the semantic dependency is inherently encapsulated in the modal morpheme itself.20 Quite often, this leads to the impression that the Subjunctive marker
itself is in fact a sort of conjunction meaning if . Consider for example the idiomatic
phrase ifnot (Hiw ontego; Ltg v tateg):
(53) Hiw Tite
gon ne pta, on tego,
gon ne qte.

1incl:pl eat art yam sbjv neg:exist eat art taro

Well eat some yam; if not (=or else), some taro.
The similarity of the Subjunctive morpheme with a conditional conjunction is not

merely an effect of translation, but also appears to be a reality for the speakers themselves. This is clear, for example, in this sentence of Hiw:
(54) Hiw On
y kn
, sise y n
wutuye ne vti

sbjv/if loc night 3pl see only
art star

ve yay r mesaye.
ipfv shine on sky
If at night, they would just watch the stars that shine in the sky.
It is true that locative phrases including prepositional phrases like ykn at night
may be used with the syntactic function of predicate (2.1.3). However, this is always
done in the form of a direct predicate, incompatible with any TAM marker.21 Therefore, the combination of the subjunctive on with the phrase ykn, rather than being
seen as plain TAM marking which would be grammatically abnormal here would
probably be better explained by a form of specialization of on as a form of (quasi)
. A similar pattern of grammaticalisation can be found in some West Germanic languages.

Thus in English, the modal auxiliary should in sentence-initial position takes up the function of a conditional conjunction: e.g. Should you be in Paris, call me (see van der Auwera &
Plungian 1998:98).
. In other words, the part of speech locative in these languages is directly predicative,
but not TAM-sensitive (Franois 2005a: 192).
Alexandre Franois
conjunction, similar to other conditional subordinators also present in this language

(2.2.2). Incidentally, this pattern is only attested in Hiw: Lo-Toga would have to use
one of its genuine conjunctions here (tewt liqen if[itwere] at night). This last point
tends to confirm that (54) illustrates an extreme case in the evolution path of the Subjunctive, which Hiw has reached but not its close neighbour.
When a sentence-initial clause is marked by the Subjunctive alone, it can
be ambiguous between a strict conditional reading equivalent to the if-clauses of
(48)(49) and a future or generic time interpretation corresponding to the whenclauses of (43)(47) above.
(55) Hiw Ne n
wute on meyigeyige ttn , ike tn
ar

art place sbjv dark
pitch 2sg grope seek

nwutuye ne wnaye.
just
art road
(If/ When) its pitch dark, you just have to grope your way.
(56) Ltg Ne t v ele gega w nk,

art fowl sbjv crow always like this

tate pero t
ne metave ni tt.
neg long comp art morning aor chop
(Every time) the cock crows like that, (this means) day is almost breaking.
(57) Ltg Denwk noke v nwule, noke t

k
ole si
vot.

today
1sg sbjv return 1sg prosp 1sg:s take some stone

Today (when) I go back, Ill take some money with me.
The proper interpretation will be given by the context. If the situation is expected
to take place anyway e.g. short-time visitors are expected to go back to their place
sooner or later it will translate as a when clause. But if the hypothesis is uncertain,
then the topic clause will correspond to a conditional sentence proper. Obviously, the
speakers get by perfectly well with this semantic ambiguity, and do not necessarily feel
the need to disambiguate these situations, even though they do have the formal means
to do so (see 2.2.2, 2.2.4).
The irrealis value of the Subjunctive does not only cover such time references as
future and generic present. It is also found in past contexts whether real or fictitious
past, as in narratives provided the event is presented as iterative:
(58) Hiw Teknwa on nwuye me
ton ne rekove sa,

people sbjv return hither from art work their

s on vn wate me, se
vn se
motrig.
3pl sbjv go:pl until hither aor:3pl go:pl aor:3pl sleep:pl
(Every time) the group came back from their labour and reached home,
theywould go to sleep.
(59) Ltg Ne nwi v ere

nie vetek,

art Ogre sbjv hit:sg 3sg here

ni
wl vn
w nk.
aor:3sg leap thither like this
(Whenever) the Ogre tried to hit him, he would jump away like this.
This is where sentence (1) cited in 1.2 would fit:

(1) Hiw Ne temt on
t
yaqe me
nw ne,


teknwa voyi.
people aor:run.away
This use of the Subjunctive in the expression of past iterative events in time clauses,
paradoxical though it may be, finds its parallel in the Classical Latin subjunctive of
repetition (Ernout & Thomas 1953:400):
Lat Id ubi dix-isse-t,
hasta-m in fines

that when say-sbjv:pluprf-3sg spear-acc to territory

eorum emitte-ba-t.
their throw-ind:imprf-3sg
When(ever) he thus spoke, he would throw a spear into their territory.

[Livy I, 32, 13]
Irrealis relative clauses are formed along similar lines. Compare (42) above with
(60), where the subordinate status is exclusively coded by the mood marker:
(60) Ltg N ve nke vn alegr t
tat
rak,

art thing 1sg sbjv forbid comp neg:irr do

henwere pah t
ge rnt.
people all prosp pl:s listen
Whatever I may ban people from doing, they will have to comply.
The presence of the article (ne) in (61) makes it clear that mowe is a noun meaning
time (rather than a subordinator, cf. 2.2.4), and that we are dealing here with an
irrealis relative clause with no relativizer:
(61) Ltg vn wahe ne mowe nke vn tmet.

go until art time 1sg sbjv old.man

until the time (when) I get old.
Finally, while purpose clauses can include a complementizer as in (39)(40) above,

they may also do without any conjunction:
(62) Hiw Noke tegtegagy
ne megoye kye {n on toge varn}.

1sg ipfv~mislead art child my 3sg sbjv stay quiet

Im amusing my baby (so) he keeps quiet.
Alexandre Franois
(63) Ltg We
tw
ne mon, we
venk

aor:2sg aim.at art bird aor:2sg let.go
ne mesor
art arrow
{v vn v qihe nie}.
sbjv go sbjv bang 3sg
You aim at the bird, then you shoot your arrow (so) it flies and knocks it.
3.6 From clause dependency to lexical derivation

The pattern illustrated in (63), whereby a purpose clause can be coded by the Subjunctive v alone, is the source of a process of reanalysis completed by Lo-Toga, but not
Hiw. This process involves several steps leading to patterns of resultative (pseudo)
serialization, resultative compounding, and even causative derivation. I will conclude
my analysis of the Torres Subjunctive by detailing the successive steps of this reanalysis. This will confirm the powerful affinity of the Subjunctive morpheme not only with
clause dependency, but also with predicate binding, including an ultimate tendency
towards the fusion of verb roots into one word.
Lo-Toga has developed a resultative construction that is clearly derived from
the purposive subordination structure (63), yet with a tighter relationship between
the two verbal heads, in a manner reminiscent of verb serialization. When a first
dynamic event V1 (generally a verb of impact) results in a state V2, then V2 is obligatorily marked as a Subjunctive. The structure { V1 v v2 } is particularly frequent
in Lo-Toga:
(64) Ltg Ole ne gi, ge
tt v wureri,

take art kava aor:pl chop sbjv small:pl

ge
gt v men.
aor:pl chew sbjv soft
(Procedural explanations about how to process kava, a woody plant

which is ground and brewed into a narcotic drink)
Take a branch of kava, mince[Aor] it small[Sbjv], then chew[Aor] it soft[Sbjv].
A sentence like (62) above unambiguously consisted of two distinct clauses: the main verb
was immediately followed by its object (the baby), and the latter referent was repeated,
in the form of a pronoun, as the formal subject within the subordinate purpose clause.
Comparison between (62) and the two resultative constructions in (64) respectively
tt v wureri and gt v men shows similarities and differences. On the one hand, the
underlying syntactic structures are identical: the subject of V2 coincides with the object
of V1. But on the other hand, (64) shows tighter structure than (62). Its two verbs are not
separated by any noun phrase, be it the object of V1 or the subject of V2; the only morpheme
that divides V1 from V2 in each construction is the Subjunctive v. Phonologically speaking, the strings {V1vV2} are uttered under a single contour with no internal pause, as
if forming a single syntactic phrase.
The compactness of the constructions in (64) is confirmed by (65): if a noun
phrase occurs, it is preferably postposed to the whole phrase {V1vV2} rather than
inserted in between.
(65) Ltg Dr
si gt v men ne gi
ne.

1incl:du pot chew sbjv soft art kava this

We can chew this kava soft.
(66) Ltg Dege
t
ge lv v nwedl

1incl:pl prosp pl:s call sbjv short

ne i
t
Alex.
art your.name quot (name)
We shall (pronounce shortly =) shorten your name to Alex.
Functionally as well as formally, these strings {V1vv2} have a lot in common with
serial verb constructions (2.4), the only difference being that the TAM marking differs between V1 and V2. Syntactically, this sequence of verbs behaves globally like a
single, transitive macro-verb. In a way, it would even make sense to consider the whole
string a single lexical unit (gt-v-men soften by chewing; lv-v-nwedl shorten), as
through a process of lexical compounding.
Arguably, the form v in these compound forms has come to have a status of
its own:22 instead of coding the Subjunctive, it could be described here as a kind
of buffer affix linking two verb roots together, with resultative meaning. This new
analysis could result in an alternative transcription and gloss for (65):
(65) Ltg Dr
si gt-v-men
ne gi
ne.

1incl:du pot chew-result-soft art kava this

We can soft-chew this kava.
Interestingly, LoToga is the only language in north Vanuatu that has developed this
pattern of resultative structure, using a buffer morpheme like v. All its neighbours
including Hiw would simply construct their resultative macro-verbs by resorting to
a simple pattern of nuclear-layer serialization (Franois 2004, 2006). Thus, the equivalent of (65) in Mwotlap would be kuy madamdaw naga/chew soft artkava/, with
nothing between the two verb radicals.
. Note that the variant vn is never attested in these new structures. This tends to confirm
that the Subjunctive marker has adopted a new grammatical status here.
Alexandre Franois
While sentences like (64)(66) are still somewhat ambiguous and compatible with more than one interpretation, some other examples provide an even
clearer case for a compounding analysis. This is especially true when the first verb
before v is the dummy auxiliary da do (also be), which does not exist as an independent verb. The string dav thus serves as a productive prefix in LoToga for
the formation of causative (transitive) verbs out of stative (intransitive) verbs or
adjectives (Table1).
Table1. Resultative compounds of Lo-Toga, incorporating theSubjunctive/Resultative
morpheme v
Simple verb/adjective
men
soft
nwedl
mo
mmerie
luw
hemr
duwr
ror
short
sick
painful
big
laugh
false
holy
Resultative compound
gt-v-men
qihih-v-men
lv-v-nwedl
da-v-mo
da-v-mmerie
da-v-luw
da-v-hemr
da-v-duwr
da-v-ror
soften by chewing
soften by grinding
shorten (aname)
make s.o. sick, sicken
hurt (body part)
make bigger, enlarge
talk playfully, joke
pretend
consecrate, baptize
(67) Ltg Temtrn tat

ho
da-v-mo
ne tle.

healer
neg:irr pot:neg do-result-sick art person

Healers cannot make people sick.
(68) Ltg Ne ri
nwl na deda-v-mmerie

art top.of reef stat iter~do-result-painful

ne tepl tle.
art foot person
The surface of the coral reef hurts the feet.
(69) Ltg

Tate hehu da-v-ror

nihe q.
neg bathe do-result-holy 3pl still
[lit. (one) has not bathed consecrated them yet]
They havent been baptized yet.
Once again, these examples are open to two morphological analyses. It is still possible
to consider them compoundings between two lexical roots (da do + mo sick), hence
the gloss /doresultsick/. But due to the relative productiveness of the process, and
the low semantic specificity of the first auxiliary, it would be equally accurate to speak
synchronically of a process of lexical derivation that actually combines a single lexical
unit (V2) with a causative prefix dav. In the latter case, one could transcribe (67) as
davmo and gloss it /caussick/.23
The historical and/or logical processes outlined here can be described as a series
of morphosyntactic reanalyses. Starting from a clear pattern of subordination between
two clauses, each step corresponds to a tighter relationship between the verbs of each
clause, and ultimately results in a specialized pattern of causative derivation (Table2).
Table2. From biclausal purpose subordination tocausativederivation: the binding
power of the Subjunctive
Ex.
Syntactic analysis
(40)
V1 = main clause
V2 = dependent purpose clause, with subordinator
V1 = main clause
V2 = dependent purpose clause, without subordinator
V1 = first action in resultative (quasi) serialization
V2 = resulting state in resultative (quasi) serialization
V1 = first radical in resultative compound verb
V2 = second radical in resultative compound verb
V1 = (dummy verb) > causative prefix
V2 = stative verb, input of causative derivation
(62)
(64)
(65)
(67)
Roots
Verbs
Clauses
3.7 The Subjunctive: Summary

The various functions of the Subjunctive in the two Torres languages are summarized
in Table3.
Table3. The narrow ties between the Subjunctive and clause dependency: A summary
Syntax
Functional value
Examples
no subordination
hortative & optative (3sg)
(35)(36)
combines
withsubordinators
modally-bound complement clauses

purpose clauses
irrealis & generic relative clauses
irrealis adverbial time clauses
irrealis conditional protases
(37)(38)
(39)(40)
(41)(42)
(43)(47)
(48)(49)
directly encodes
subordination
irrealis conditional protases

[Hiw] reinterpreted as conjunctionif
irrealis & generic adverbial time clauses
irrealis & generic relative clauses
irrealis purpose clauses
[Ltg] resultative compounding
>causativederivation
(50)(52)
(53)(54)
(55)(59)
(60)(61)
(62)(66)
(65)(69)
. This prefix has thus, in function, replaced the Proto Oceanic causative prefix *paka,
which has essentially left no trace in the two Torres languages.
Alexandre Franois
4. The Background Perfect: In search of a focus

The TAM category I propose to label Background Perfect offers a broadly similar, yet
quite distinct illustration of the phenomenon just discussed with the Subjunctive. The
general mechanism behind the two patterns is the same: the semantic and pragmatic
identity of a TAM marker makes it particularly prone to the syntactic coding of clause
dependency. Nevertheless, the case of the perfect is sufficiently different to warrant a
section of its own.
The question addressed here is the following: how can the Background Perfect
marker (ve si) clearly form a subordinate relative clause in a sentence like (2),
and yet do without any overt subordinator? What is there in its makeup that renders
it syntactically different from other realis categories, and especially different from the
regular Perfect?
(2) Ltg

Ne gehuh
ve
kerkur
tle
si
mat mt.
[lit. Thecoconut crab has devoured people has died.]
The coconut crab (who) had devoured people was dead.
Once again, I shall argue that the syntactic power of this marker must ultimately be
understood as an outgrowth of its main functional property, namely, its ability to mark
the informational status of its predicate as presupposed. Due to this form of pragmatic
demotion, the predicate phrase thus marked needs to search for an external focus of information, which will typically result in a syntactic relation of dependency between clauses.
4.1 The two perfects and sentential focus

Among the various TAM categories that can denote a realis event (2.1.2), the two
Torres languages have a Stative, an Imperfective, and two perfects. I will briefly present
the first two of these TAM markers, before I turn to the difference between the last two
which are derived from the former.
4.1.1 Stative vs. Imperfective

The Stative [Hiwn(g), Ltgna] is followed exclusively by stative predicates, that is,
stative verbs and adjectives:
(70) Ltg Ne vavetme mi kemi na dernha.

art language poss 2pl stat difficult

Your language is difficult.
The only way for a semantically dynamic verb to be compatible with this marker is
to first be converted into a habitual (and therefore stative) predicate, by means of
reduplication:
(71) Ltg *(Nke na si.) Nke na sesi.

*1sg stat walk 1sg stat iter~walk

*(I walk.)
I usually walk, i.e. I am a (good) walker.
As for the Imperfective (Hiw/Ltg ve),24 it encompasses two aspectual values (cf.
Comrie 1976): the progressive (72) and the habitual (73):
(72) Ltg Rem m ve ker.

mother his ipfv weep

His mother is/was weeping.
(73) Ltg Nihe ve lv nie t
Temtrn.

3pl ipfv call 3sg quot Healer

People call him Healer.
The same Imperfective ve also takes part in several progressive structures based on
verb serialization {ve Posture verb V1 + veAction verb V2}: see 2.4, ex.(28a).
Verbs that are lexically stative (including adjectives) are sometimes found to combine with the Imperfective, in which case they take on a dynamic reading:
(74) a. Ltg Ne vete na medudut.

art place stat black

Its dark.
[stative reading]
b. Ltg Ne vete ve
medudut.

art place ipfv black

Its getting dark.
[dynamic reading]
However, setting aside these rare cases, it is generally true that the Stative and the
Imperfective tend to target two different sets of verbs, respectively stative and dynamic.
Obviously this makes it difficult to carry out any extensive comparison of these two
TAM markers. But as we shall now see, the situation is totally different for the two
perfects that are derived from them.
4.1.2 The two perfects

I now turn to the two perfects of the Torres languages, which will form the heart of the
following discussion: the regular Perfect (Hiw nti/Ltgnasi) and the Background
Perfect (Hiw veti/Ltgvesi).25
. Beside the widespread form ve [6], Lo-Toga also possesses a rare variant me [m6]; likewise, mesi constitutes a (rare) variant of its Background Perfect vesi. Incidentally, there is
no reason to suspect any etymological connection between the element ve [6] of the Imperfective and the Lo-Toga form of the Subjunctive v [7]: they are two unrelated morphemes.
. Unlike Lo-Toga where the contrast is systematically coded, Hiw is problematic in that
it treats the two proclitics respectively n and ve as optional (see Table4). Quite often, a
Alexandre Franois
Morphologically speaking, one may say that these two perfect markers show a
straightforward correspondence with the Stative and the Imperfective, as they simply
consist in the combination of the latter with the postclitic *ti.26 However, the clitic *ti
only occurs in combination with TAM markers, with various semantic effects, and
cannot be assigned any stable meaning unto itself. It is therefore methodologically
safer and probably more realistic from the speakers point of view to consider
each compound TAM marker as a single meaningful morpheme, albeit a discontinuous one. As a result, while the form na alone was glossed stat(ive), I shall gloss the
sequence nasi as prf1prf2, with no attempt to arrive at a compositional analysis.27
As for the semantic processes that may have led to the creation of these compound
forms, that is a matter for history, and goes beyond the limits of the present study.
Considered from a purely semantic angle, the two TAM categories under consideration are synonymous, as they both correspond to the typological definition of
the perfect aspect. They represent a realis event insofar as it is complete, and place the
cursor in the resultant state that follows that event.
(75) a. Ltg Kemm na gil o si
ne keka tekle.

1excl:pl prf1 dig out prf2 art yam some

We have dug out a few yams.
b. Ltg Ne keka tekle kemm ve
gil o si.

art yam some 1excl:pl bkpf1 dig out bkpf2

(These are) a few yams we have dug out.
Because they both point to the resultant state that follows the final boundary of a completed state of affairs, they are equally compatible with stative and with dynamic predicates. This contrasts with the Stative and the Imperfective, which tend to combine with
distinct sets of verbs stative vs. dynamic as we saw earlier (4.1.1). Thus, while the
perfect predicate will be tagged by the postclitic ti alone as in (19) or (39) blurring the
contrast between the two perfects. This is why the present section will mainly cite examples
from Lo-Toga, where the phenomenon is much more conspicuous. This being said, when
the proclitics of Hiw are overtly marked as in (76) or (79) they do conform to the same
principles as in Lo-Toga.
. To be precise, LoToga alternates between two allomorphs: an assibilated form si (<*ti),
and an elided form t [t] when preceded or followed by a vowel see (80), (87), (88). Here I
lump the two synchronic allomorphs together under the underlying (and ancestral) form *ti,
for the sake of discussion.
. I adopted similar methodological principles for the analysis of discontinuous TAM
markers in Mwotlap (Franois 2003:30sqq, 343). Incidentally, most of the compound forms
of Mwotlap involved postclitic t [t~], with which the Torres form ti/si is cognate.
dynamic verb gilo digout is incompatible with the Stative na (*nagilo), it can
perfectly take the standard Perfect which is derived from it (nagilo si).28
Yet, even though the two perfects may be said to be synonymous in terms of their
aspectual semantics, they are not functionally equivalent, and in fact occur in distinct contexts. The difference between these two TAM categories is best defined in
pragmatic terms, by contrasting the manner in which they organize the informational
hierarchy within the sentence: to use the terms of Lambrecht (1994: 52), the standard
Perfect puts its predicate under the scope of the assertion, whereas the Background
Perfect explicitly encodes its status as a pragmatic presupposition (Table4). This use of
TAM markers for coding informational hierarchy is typologically original.
Table4. Hiw and Lo-Toga have two Perfects; their difference lies in the pragmatic status
of the predicate
(Standard) Perfect
Background Perfect
Hiw
Lo-Toga
Pragmatic status ofpredicate
(n) ti
(ve) ti
na si
ve si
asserted/foregrounded
presupposed/backgrounded
4.1.3 When TAM markers encode informational hierarchy

The regular Perfect (Hiw nti, Ltgnasi) represents the predicate as a realis perfect event and it places it under the scope of the sentences informational focus. This
is the pragmatically unmarked situation, where the syntactic center of the sentence
coincides with its pragmatic center in terms of assertion as in (75a) or (76a):
(76) a. Hiw Ike ttm tom ne tir mon, pa

2sg think comp art true bird but

tego.
N
rak ti.
neg:exist prf1 make prf2
You could think its a real bird, but far from it. (Somebody) made it.
As for the Background Perfect (Hiwveti, Ltg vesi), it also construes a realis perfect predicate, but explicitly specifies its informational status as pragmatically presupposed, i.e. defocused. Crucially, a predicate phrase marked with the Background
Perfect (henceforth BkPf ), due to this backgrounded status, cannot constitute a wellformed utterance on its own:
(76) b. Hiw *Ve
rak ti.

bkpf1 make bkpf2

*{(somebody) made it}[background]
. This freedom of actionality combinations provides further support for the view explained above, that the two perfects should not be analyzed compositionally, but as (discontinuous) TAM markers in their own right, with specific properties.
Alexandre Franois
In contrast to (76a) nrakti, a sentence like (76b) would be deemed incomplete. This
is because an utterance, in order to be pragmatically valid, needs to contain at least
some new, assertional information.29 Insofar as the BkPf tags a predicate phrase as
presupposed, it makes it unable to constitute a correct utterance by itself; in order to
be interpretable, the sentence needs some other constituent with which the pragmatic
assertion can be identified.
Occasionally, the background status applies to the whole clause (i.e. the predicate
with its arguments and complements), which is then entirely marked as presupposed.
This is what happens, for example, when the speaker refers back to an event that is
already known to the addressee, as a reminder. Thus compare the regular Perfect of
(77a), where the whole clause is fully new, and the Background Perfect of (77b), where
it only serves as a reminder of an already known fact:
(77) a. Ltg Ses
na hag si!

your.sister prf1 sit prf2

Hey! {Your sister has given birth!}[focus]
b. Ltg Ses
ve
hag si:
ne tenwn

your.sister bkpf1 sit bkpf2 art male

hit ne leqvine?
or art female
{Your sister has given birth (as we know):}[background]

{is it a boy or a girl?}[focus]
(77b) could be described as a case of clause topicalization.30 The event marked as Background Perfect has no informational value in itself, that would allow it to form an
utterance on its own; rather, it is used as a reminder to help the addressee interpret the
focal part of the sentence (in this instance, the question).
4.1.4 How many clauses?

As is typically the case for topicfocus structures, the syntactic relationship between
the two clauses in (77b) is still loose. While it does illustrate a form of dependency, it
does not form subordination in the strict sense of the term. Most of the time, however,
the Background Perfect is involved in much more tightly bound structures than this.
. See Givn (1984: 241), Tomlin (1985), Lambrecht (1994: 60).

. Other strategies for clause topicalization have been observed, for example, with the
background topic clauses found in Chuave, a language of Papua New Guinea (Thurman
1979, cited by Givn 1990:870). Clause topicalization is a common phenomenon in North
Vanuatu, but in the neighbouring Banks languages, it involves the use of deictics rather than
TAM strategies (Franois, forthcoming).
As we shall see in 4.2, the presupposed predicate quite often involves genuine subordination, e.g. a relative clause:
(78) Ltg Lwie
leqvine meke {nie ve
rak si

thanks obl woman your 3sg bkpf1 make bkpf2

ne ttgal}.
art picture
Thanks to your wife {(who) drew the pictures}[background].
One ambiguous case, however, is when the sentence apparently consists of a single
predicate: this happens especially in contrastive focus sentences like (79).
(79) Hiw Noke ve
tot
ti.

1sg bkpf1 carve bkpf2

I carved it! (not you)
The predicate here (ve tot ti) is the presupposed segment of the sentence, whereas the focal
part corresponds to its grammatical subject (noke). In fact the sentences structure is quite
parallel to its English counterpart, including the contrastive focal stress that affects the subject phrase, with the same pragmatic implications. All these arguments tend to suggest that
(79), just like its English translation, consists of just one syntactic clause, with no possibility
to speak here of clause dependency. If this were true, then we would need to temper the
claim that the pragmatic mechanism of the Background Perfect almost systematically goes
along with subordination. In doing so, one would have to admit that the pragmatic properties of the BkPf sometimes trigger clause dependency as in (78), but sometimes operate on
a purely pragmatic level, with little incidence on the syntactic structures, as in (79). This
would also challenge the statement made earlier about (76b) that a main clause cannot
stand alone if it is marked with the Background Perfect.
In fact, we shall see below (4.2.2.2) that the structural similarity between LoToga and English in (79) is an optical illusion. It will appear that (79), like all contrastive focus patterns in the Torres languages, is best analyzed as consisting of not just
one, but two distinct clauses. In doing so, I will show that the Background Perfect does
not only affect the pragmatic interpretation of the sentence in terms of informational
hierarchy, but also has a syntactic impact, in creating a genuine subordination relation
between predicates.
4.2 From pragmatic presupposition to syntactic subordination

The following pages will illustrate in more detail this syntactic affinity between the
Background Perfect and clause dependency. I will first show cases where the two perfects combine with overt subordinators (4.2.1). I will then show that the BkPf alone
may in fact suffice to generate clause dependency and subordination, without requiring
Alexandre Franois
any further formal device (4.2.2). The special syntax of contrastive focus structures
will be examined in 4.2.2.2.
4.2.1 The two perfects and overtly marked subordination

The semantic principles exposed in 4.1.3 for main clauses are equally true for those
clauses which are formally marked as dependent by means of an overt subordinator.
Thus, the regular Perfect will be used whenever the subordinate clause falls under the
scope of the assertion. This is the case, in general, for complement clauses attached to
speech or thought verbs:
(80) Ltg Nke dem {t ne nwi na kur

1sg think comp art Ogre prf1 crunch

nike pe
t }.
2sg already prf2
I thought (that) the Ogre had already devoured you.
Regular Perfects are also found in the protasis of certain conditional clauses:

(81) Ltg { Tewt

if
ne temt na la
nike si},
art ghost prf1 take:tr 2sg prf2
Temtrn
Healer
t
n ake vn hr
nike Pene.
fut1 3sg fut2 go find obl 2sg Hell
If the ghosts kidnapped you, the Healer would come and find you inHell.
Conversely, if a subordinate clause refers back to an already established event,

then the Background Perfect will be required. This is especially true of restrictive relative clauses, whose function is precisely to point to a background element to help the
addressee track referents:
(82) Ltg ne revrev pe nke ve
hur si /

art song rel 1sg bkpf1 sing bkpf2

(?? pe nke na hur si)

rel 1sg prf1 sing prf2

the song {which I sang}[background]
(83) Ltg Ne lilie {pe nie ve

durlue si}

art cave rel 3sg bkpf1 drill bkpf2

ve taqe wahe me
denwk.
ipfv lie until hither today
The cave {which he broke open}[background] still exists today.
In each of these two sentences, the relative clause is unambiguously marked as subordinate by its relativiser pe (2.2.3). As for the BkPf, it arguably operates on the pragmatic
level, by providing its predicate with a background status.
The regular (assertive) Perfect is extremely rare in relative clauses. This configuration does occur however, in exceptional cases, when the informational focus is in fact
located within the relative clause. Example (84) provides an illustration of this nonstandard situation:
(84) Ltg Henwere

people
pah tat
llmern .
all neg:irr know
obl:adv
{w na huqe
wereno si}
rel prf1 initiated only
prf2
Henwere
people
nihe ve llmern .
3pl ipfv know
obl:adv
Not everybody would know (these things).

Only {those whove been initiated}[focus] know[background].
What, syntactically, forms the main clause (nihe ve llmern) of the whole sentence is
here a mere repetition of the previous sentence, with no informational weight. Conversely, the sentences assertion is located in the relative clause, which exceptionally
takes the regular Perfect rather than the Background Perfect.
A sentence such as (84) tends to show that the use conditions of the two perfects
in relative clauses do not obey a strict formal rule, whereby all relative clauses would
mechanically take the Background Perfect. Rather, the choice of TAM marker remains
a pragmatically productive device, based on the informational hierarchy chosen by the
speaker in organizing his utterance.
4.2.2 The subordinating effect of the Background Perfect on its own

In sum, it would be exaggerated to say that all relative clauses let alone all subordinate clauses require the Background Perfect: this is only the case for those clauses
which are pragmatically presupposed. Now, if we narrow down our observations to
the latter configuration, an important point remains to be made. Unlike all other
realis markers, the Background Perfect allows a subordinate clause to dispense with
any formal subordinator, as though it were sufficient per se to code for clause dependency. This, as we shall see now, is especially the case with relative clauses, and the
focus cleft constructions which are derived from them.
4.2.2.1 Relative clauses
While the BkPf is occasionally found to combine with an overt relativizer see
(82)(83) the most frequent pattern is for perfect relative clauses to dispense with a
formal subordinator, and be simply marked by the BkPf alone (see also (78) above).
(85) Ltg li megage {ve pah si}

loc month bkpf1 finish bkpf2

last month [lit. in the month {(which) has finished}[background] ]
Alexandre Franois
(86) Hiw Ike peon sawe-vog

ne temt
{tekn wa ain

2sg fut dance-appl art headdress people other

ve
rak ti}.
bkpf1 make bkpf2
You shall dance with a headdress {other people will have made}[bkg].
A superficial look at (86) could suggest a comparison with the syntax of zero-marked
relative clauses in English, which happens to be parallel here. Two differences must
however be noted.
Contrary to English, zero-marked relative clauses in Torres languages are allowed

whatever the function of the antecedent within the subordinate clause. Thus while
English allows a zero-marked clause in (86) where the relativized NP is an object,
it does not in (85), where it is a subject. The Torres languages are less constrained
than English in this regard.
Contrary to English, zero-marked relativization in the Torres languages is only
allowed in combination with certain specific TAM markers, the Background Perfect and the Subjunctive. The Torres languages are more constrained than English
in this regard.
We can now account for Example (2), which was quoted in 1.2:
(2) Ltg Ne gehuh
ve
kerkur
tle
si
mat mt.


[lit. Thecoconut crab {has devoured people}[background] {has died}[focus]]

The coconut crab (which) had devoured people had died.
On the face of it, (2) is a sequence of two clauses taking the same subject, with no formal dependency marker between the two clauses. Only the nature of the Background
Perfect, and its ability to defocus its own predicate, makes it clear here which clause
is subordinate, and which is the main clause of the sentence. It must also be noted
that setting aside the case of the Subjunctive (3.5.2) only the BkPf is capable of
encoding a relative clause in this way. Even the Imperfective, which is otherwise morphologically similar to the BkPf, makes the presence of an overt relativizer obligatory:
compare sentence (2) with its counterpart (16).
This analysis in turn helps us understand the structure of (75b), here repeated:
(75) b. Ltg Ne keka tekle kemm ve
gil o si.

art yam some 1excl:pl bkpf1 dig out bkpf2

(These are) a few yams we have dug out.
An initial approach could have proposed analysing (75b) as consisting of a single

clause, with a single predicate (vegilosi). In this case, the unusual sentence-initial
position of the object noun phrase (nekeka tekle) would probably be explained as a
form of left-dislocation. However, this analysis does not hold, for two reasons: formally, the whole sentence is uttered under a single phrase contour with no pause,
which makes it incompatible with a topic-focus pattern; and semantically, the
function of the initial NP is not that of a topic (*These yams), but of a predicate
(These are some yams). This sentence can only be properly analyzed if one remembers that the Torres languages do not make use of copula for noun predicates, i.e.
nouns and noun phrases are directly predicative [see 2.1.3, ex.(6)]. Consequently,
an appropriate syntactic analysis for (75b) would posit not one clause, but two:
first, the whole sentence consists of a zero subject followed by its NP predicate:
[These are] {a few yams we have dug out}; second, the clause we have dug out constitutes a relative clause (marked by the BkPf) that is embedded within that main
predicate phrase.
Relative clauses marked by the BkPf alone have all the syntactic properties of relative clauses in these languages. They can be embedded within a noun phrase, a prepositional phrase, etc. As mentioned above, the antecedent of the relative can play any
syntactic role both in the main clause and in the relative clause itself; and it may also
be referred to by a resumptive, anaphoric morpheme within the relative clause (e.g.
there, from it):
(87) Ltg Ne gerite ni
nwule wul vete

art octopus aor:3sg return again place

{hr v
la
}.
3du bkpf1 take:tr bkpf2 obl:adv
The octopus went back to the place {they had caught it from}.
The use of the BkPf in relative clauses is so widespread, that one often hears quite
complex sentences such as (88), which superficially consist in a string of juxtaposed
clauses, with no obvious indication of their syntactic structure.
(88) Ltg Mowe nie ve velag wahe vin, ni
vn wahe

time 3sg ipfv run until up aor:3sg go until

vn
li lilie {nihe ve
toge si
vine},
thither loc cave 3pl bkpf1 stay bkpf2 underneath
{rem m v
in si
vine},
{ne nwi
mother his bkpf1 lie bkpf2 underneath art devil
ve
li
nie t
},
nie ni
gerage.
bkpf1 replace 3sg bkpf2 obl:adv 3sg aor:3sg climb

And as he ran all the way up, he managed to reach the cave
{(where) they had been staying}, {(where) his mother had been lying},
{(and where) the devil had taken her place}, and he climbed it.
Alexandre Franois
Apart from the first clause, introduced here by the noun-conjunction mowe time,
moment (2.2.4), the five remaining clauses lack any subordinator properly speaking.
However, the status of the three medial clauses (in braces) as restrictive relative clauses is
unambiguous: this is indicated by the Background Perfect, as well as by the presence of
locative adverbials (vine underneath, there) whose function is to indicate the syntactic role of their antecedent (the noun lilie cave) within each embedded clause. Ultimately,
among the six clauses in (88), only two have the status of informatively new, syntactically
main clauses: these are the two Aorist clauses ni vn wahe he reached and nigerage he
climbed.
4.2.2.2 Focusing structures

The coding of contrastive focus, in the Torres languages as well as in other languages of
north Vanuatu, resorts to a cleft-sentence strategy which is derived from its relativization patterns.
4.2.2.2.1 Contrastive subject focus
The focal constituent, generally a noun phrase, occurs preferably to the left of the sentence whether via left-dislocation or not and is immediately followed by a relative
clause pointing to the presupposed segment of the utterance.
(89) Hiw Teknwa tameso
{pe ve
vegevage vati

people
old
rel bkpf1 talk
show

kema
ti
ie}.
1excl:pl bkpf2 obl:adv
(It is) the older generation {who taught all these stories to us}.
Clearly, the best way to analyse (89) would be to identify two distinct predicates
here, similarly to the analysis of (75b) above. The predicate phrase vegevage vati
itself a verb serialization, see (26) is marked as syntactically dependent as
much by the Background Perfect, as by the relativizer pe. It is subordinate to the
sentences main predicate that is, the nominal predicate teknwa tames (itis) the
elders.
The syntactic organization of such structures is also reflected in their prosody.
A sentence like (89) is uttered with a contrastive accent on the last stressed syllable
of the group teknwa tames. It is followed by a distinctive fall in pitch and intensity on the remainder of the sentence, which is typical of presupposed elements in
cleft constructions:
[t6kWa tam6so p6 6 6>6a>6 ati k6ma ti i6]
The analysis of (89) may also apply to a slightly different form of focusing pattern, one
that lacks any formal relativizer. Consider (90):
(90) Hiw Teknwa te Toge ve
rak ne gengon ti.

people
from Toga bkpf1 make art meal bkpf2

[lit. the toga people[focus] {made the feast}[background]]

(It was) the Toga people (who) organized the feast.
A first glance at a sentence like (90), which consists of the sequence NP+VP, might
have suggested that we are simply dealing with the syntax of a single sentence, with a
subject followed by its predicate. However, following the reasoning above for (89), this
sentence (90) can rather be shown to consist of two syntactically hierarchized clauses.
The predicate phrase veti, which is pragmatically presupposed in the context,
would thus be a relative clause with no relativizer, as in (75b) above. The phrase teknwa
te Toge, to which this relative clause attaches, is pragmatically the focus of the sentence,
and syntactically its matrix (NP) predicate. In other words, the syntactic structure of a
focusing sentence like (90) is once again parallel to the NP predicate (75b) above:
(91) noun phrase
+ verb phrase with bkpf
= {nominal equational clause1 + relative clause2 (without relativizer)}
The difference between the simple relative clause of (75b) and the focusing structure
(90) lies essentially in the prosody. Thus, (90) contrasts a stressed segment with an
unstressed one, exactly like (89) above:
[t6kWa t6 tf>6 6 :lak n6 >6n>fn ti]
4.2.2.2.2 Biclausality and the negation test

The biclausal analysis under (91) is confirmed by certain syntactic tests, such as negation. In principle, the negator is a member of the TAM paradigm (2.1.2), which means
that it normally occurs in the same slot as the corresponding affirmative TAM marker,
on the initial boundary of the negated predicate phrase. For example, a standard
Perfect like (92a) would be negated as in (92b):
(92) a. Hiw Teknwa te
Toge n rak ne gengon ti.

people from Toga prf1 make art meal prf2

The Toga people organized a feast.
b. Hiw Teknwa te
Toge tati
rak ne gengon.

people from Toga neg:real make art meal

[ordinary negation, no contrastive focus]

The Toga people didnt organize a feast.
1 clause
But the sentences overall structure turns out to be different when the negation affects a
Background Perfect sentence such as (90). Instead of combining with the verb r ak as in
Alexandre Franois
(92b), the negator then affects the initial noun phrase of the sentence, thereby proving
it has the syntactic status of a predicate:
(93) Hiw Tati
teknwa te
Toge ve
rak ne gengon ti.

neg:real people from Toga bkpf1 make art meal bkpf2
[negation of contrastive focus pattern]
{Its not the Toga people}[focus] (who) organized the feast[bkg].

2 clauses
In sum, (90) consists not just of a subject phrase with its predicate, but of two predicates: it must be analyzed as a genuine cleft construction.
Finally, exactly the same analysis could be conducted to account for example (79),
mentioned in 4.1.4 and repeated below:
(79) Hiw Noke ve
tot
ti.

1sg bkpf1 carve bkpf2

[lit. {(its) I}[focus] (who) {carved it}[background].]

I made it!
While the brevity and simplicity of (79) would spontaneously suggest we are dealing
with a monoclausal SV(O) sentence just like its English translation, it turns out that
a more accurate analysis would have to parse it into two distinct clauses: a direct NP
predicate (noke)31 followed by a relative clause with no relativizer (vetotti).32 Thus the
negation of (79) would be parallel to (93) above:
(79) Hiw Tati
noke ve
tot
ti.
Temo-k.

neg:real 1sg bkpf1 carve bkpf2 father-1sg

{ (Its) not I}[focus] { (who) carved it}[bkg]. (Its) my father.
4.2.2.2.3 Contrastive focus of non-subjects

The analysis proposed above for the contrastive focus of subject noun phrases can
be extended to other syntactic functions and other parts of speech. Indeed, we know
(from 2.1.3) that the ability to constitute a direct predicate with no copula is not
. Ex. (8) above illustrates the same pronoun noke [its] me in a direct NP predicate structure.
. Evans (2007), in his article on insubordination, cites similar instances of hidden cleft
constructions in certain Australian languages. For example, the language Ngandi (Evans
2007:414, after Heath 1985) expresses subject focus by combining an ordinary subject NP with
a verb form that is formally marked as subordinate (with ga): e.g. ni-deremu nigarudu-i,
literally [its] the man [who] wentsubord. The structural similarity with my proposed analysis
(91) is worthy of notice: in both cases, the surface form of the sentence seems to consist of a
single clause, where underlyingly there are two.
only characteristic of nouns and noun phrases, but in fact of most other parts of speech
and syntactic constituents.
It is thus possible to interpret all focus constructions as biclausal sentences, along
the lines of (91). The focus phrase forming a direct predicate may be e.g. an adverb (94)
or a predicative demonstrative (95):
(94) Hiw Ve
rak ti
nwna?

bkpf1 make bkpf2 how

[lit. {made it}[background] how[focus]?]

How was it made?
(95) Ltg

Noke ve
vn ve
tun si
Vave pe no k!
1sg bkpf1 go bkpf1 buy bkpf2 Vava foc this
[lit. {I went to buy on Vava}[background] {(its) this}[focus]]
this is what I bought on Vava island.
In those cases too, the BkPf clause can be analyzed as a relative clause followed by its
matrix predicate.
The case for this biclausal analysis is even stronger when the asserted phrase is
fronted, as commonly happens in cleft focus constructions. As mentioned in 2.1.1,
the constituent order is normally SVO. When the asserted element coincided with
the subject of the backgrounded verb, as in (90) or (79) above, the focus construction
involved no displacing of the phrase under focus; its pragmatic status was only indicated by the prosody (and of course, indirectly, by the BkPf in the rest of the sentence).
But when fronting affects an object or another complement whose normal position is
after the predicate, then the disrupted syntax of the sentence makes it clear that we are
dealing with a biclausal structure.
For example, compare the non-contrastive sentence (96a) with standard word
order and the regular Perfect and its contrastive counterpart (96b):
(96) a. Ltg Gide
na vn si
me

ne mesale pek.

1incl:pl prf1 go prf2 hither obl:prep art road this

We came through this road.
1 clause
b. Ltg

Ne mesale pek gide

ve
vn si
me
.
art road
this 1incl:pl bkpf1 go bkpf2 hither obl:adv
[lit. (it is) this road (that) we came through (it).]
this is the road we came through.
2 clauses
(96b) shows fronting of the focal element, in the form of a predicate noun phrase (ne
mesale pek [itis] this road). The remainder of the sentence, which is marked as BkPf,
has the syntactic status of a relative clause. Specifically, the antecedent mesale road
is anaphorically indexed by the locative preposition-adverb (there, through it) in
accordance with the typical syntax of relative clauses, as in (87) above. The resulting
Alexandre Franois
double-zero relative clause i.e. zero relativizer, zero anaphora on the preposition
happens to be structurally close to its English equivalent: (it is) this road { we came
through}.
We saw earlier that the surface form of subject-focusing sentences like (79) shows
some structural ambiguity, to the point that certain tests were required to determine
their underlying syntax (4.2.2.2.2). This is not necessary with other contrastive focus
cleft constructions such as (96b), because they are transparent in this regard.
In sum, a predicate marked as Background Perfect must always be understood as
forming a subordinate clause even when superficially it may seem to form the sole
verb of the utterance. The pragmatic center of assertion, as well as the syntactic center
of the sentence, are to be sought outside of its boundaries.
4.2.2.3 Wh-questions and the Background Perfect

Finally, a contrast similar to (96ab) can be found in the structure of questions. At first
sight, the different choice of aspect between (97a) and (97b) is difficult to explain:
Nike na vegevage si
mi paie?
2sg prf1 talk
prf2 with who
[lit. You were talking to whom?]
Who were you talking to?
(??ve vegevage si)
b. Ltg Paie ve
vegevage si
mke?

who bkpf1 talk
bkpf2 with-you

Who was talking to you?
(*na vegevage si)
(97) a. Ltg

[standard perfect]
[background perfect]
The rule that is empirically observed, and illustrated by (97ab), is given in (98):
(98) In content questions referring to a completed event (perfect), the verb will
normally take the regular perfect if the question word comes after the verb;
but it must be marked as background perfect if the question word precedes
the verb (whether by wh-movement or not).
The explanation for this unexpected asymmetry has to do with the placement of sentential focus, which in content questions systematically falls upon or includes the
question word. In (97a), which is unmarked for word order, the sentence-final position
of the question word paie is compatible with the interpretation of the whole predicate (including its complement) as falling under the pragmatic focus of the utterance. In
(97b) however, the sentence-initial position of paie attracts stress and sentential focus,
yielding a sentence shape that is strongly reminiscent of focalising structures such as (79)
or (96b). A consequence of this sentence-initial focus is that the rest of the sentence has
to be coded as informationally defocused, which explains the use of the Background
Perfect here. Once again, the most appropriate analysis of (97b) is to consider it as
biclausal, similarly to (91) above. In other words, what we have here is literally:
(97) b. {(it is) who}[focus] (the one that) {was talking to you}[background]?
Such a formal TAM contrast between (97ab), depending on the placement of the
question word, is unique to the Torres languages, and unknown elsewhere in the
region. Furthermore, it is even quite particular within these two languages, as it is
restricted to questions whose verbal aspect is a perfect. Uncommon though it may
be, this contrast can however be explained by the internal logics of these languages, in
terms of the handling of informational hierarchy and predicate dependencies.
4.3 The Background Perfect: Summary

The various patterns characteristic of the Background Perfect are summarized
in Table5.
Table5. The close links between the Background Perfect and clause dependency: A summary
Syntax
no subordination
combines
withsubordinators
directly encodes
subordination
Functional value
clause topicalization & backgrounding
Examples
(77b)
realis background (restrictive) relative clauses
(82)(83)
realis background clause in cleft focus patterns
(89)
realis background (restrictive) relative clauses
(85)(88)
realis background clause in cleft focus patterns
(90)(96b)
question sentences if whword is fronted
(97b)
5. Conclusion
Hiw and Lo-Toga, the two languages of the Torres islands, possess a wealth of formal
devices for encoding clause dependency, and make regular use of them with most of
their TAM markers. However, this paper has shown that two TAM categories the
Subjunctive and the Background Perfect present different behaviour when it comes
to handling interclausal relations. While they are both compatible with regular subordinators, they also show a marked tendency to do without them, and to be used alone
as a subordinating strategy in its own right.
Obviously, the two cases under study differ in many respects, if only because they
do not come under the same discourse constraints:
the Subjunctive contrasts with other irrealis markers, in lacking the necessary
information on the clauses modality status and illocutionary force.
the Background Perfect contrasts with other realis categories (especially with
the regular Perfect), in marking its target predicate as pragmatically
presupposed.
Alexandre Franois
One characteristic that is nevertheless shared by these two components is that they
both affect the pragmatic well-formedness of an utterance. A sentence, if irrealis, needs
to have some form of illocutionary force; and likewise, an utterance must include at
least some new, asserted segment. In my interpretation, the absence of either of these
two elements in a clause is precisely what makes it unable to form a sentence on its own,
and makes it dependent, both functionally and syntactically, upon external predicates
and clauses.
In sum, different as they may be, these two patterns essentially obey the same
underlying mechanism, which justifies their comparison. In both cases, the key to the
syntactic structures attested is a form of pragmatic indeterminacy, or pragmatic demotion, that is inherently conveyed by the TAM marker.
The two patterns illustrated in this paper are specific to Hiw and Lo-Toga, and
make these two languages original, even in comparison with the nearby languages
of north Vanuatu. Yet they also show a form of universal relevance. They remind us
that the existence of formal, dedicated subordinators is not the sole key to the syntax
of interclausal relations; and that patterns of clause dependency can also result, albeit
indirectly, from a clauses pragmatic properties and semantic profile. This is another
illustration of how the formal structures of languages are regularly shaped and renewed
through the pragmatic constraints that weigh upon communication.
Abbreviations
Examples are glossed according to the Leipzig rules. More specific abbreviations are
listed below.
aff
aor
appl
art
bkpf
caus
comp
cplt
ctfc
du
foc
fut
ipfv
irr
iter
hiw
loc
affirmative
Aorist
applicative
article
Background Perfect
causative
complementizer
Complete aspect
Counterfactual
dual
focus marker
Future
Imperfective
irrealis
iterative
Hiw
locative marker
ltg
m
neg:exist
obl
poc
poss
pot
prf
prosp
quot
rel
result
s
sbjv
stat
tr
Lo-Toga
masculine
Negative existential
oblique
Proto Oceanic
possessive marker
Potential
Perfect
Prospective
quotative
relativizer
resultative
subject clitic
Subjunctive
Stative
transitive verb
References
Aikhenvald, Alexandra Y. 2006. Serial verb constructions in a typological perspective. In Serial
Verb Constructions: A Cross-linguistic Typology [Explorations in Linguistic Typology],
Robert M.W. Dixon & Alexandra Y. Aikhenvald (eds), 168. Oxford: OUP.
Bril, Isabelle. 2004. Complex nuclei in Oceanic languages: Contribution to an areal typology. In
Complex Predicates in Oceanic Languages: Studies in the Dynamics of Binding and Boundness, Isabelle Bril & Franoise Ozanne-Rivierre (eds), 148. Berlin: Mouton de Gruyter.
Bybee, Joan, Perkins, Revere & Pagliuca, William. 1994. The Evolution of Grammar: Tense,
Aspect, and Modality in the Languages of the World. Chicago IL: University of Chicago
Press.
Chappell, Hilary. 2008. Variation in the grammaticalization of complementizers from verba
dicendi in Sinitic languages. Linguistic Typology 12(1): 4598.
Comrie, Bernard. 1976. Aspect. An Introduction to the Study of Verbal Aspects and Related
Problems. Cambridge: CUP.
Cristofaro, Sonia. 1998. Deranking and balancing in different subordination relations: A typological study. Sprachtypologie und Universalienforschung 51: 342.
Cristofaro, Sonia. 2003. Subordination [Studies in Typology and Linguistic Theory]. Oxford:
OUP.
Cristofaro, Sonia. 2008. Purpose clauses. In The World Atlas of Language Structures Online,
Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds), Ch. 125.
Munich: Max Planck Digital Library.
Crowley, Terry. 1987. Serial verbs in Paamese. Studies in Language 11: 3584.
Crowley, Terry. 2002. Serial Verbs in Oceanic: A Descriptive Typology. Oxford: OUP.
Crowley, Terry. 2004. Bislama Reference Grammar. Honolulu HI: University of Hawaii Press.
Dik, Simon. 1989. The Theory of Functional Grammar, Part 1: The Structure of the Clause
[Functional Grammar Series 9]. Dordrecht: Foris.
Durie, Mark. 1997. Grammatical structures in verb serialization. In Complex Predicates. Alex
Alsina, Joan Bresnan & Peter Sells (eds), 289354. Stanford: CSLI.
Ernout, Alfred & Thomas, Franois. 1953[1993]. Syntaxe latine. Paris: Klincksieck.
Evans, Nicholas. 2007. Insubordination and its uses. In Finiteness. Theoretical and Empirical
Foundations, Irina Nikolaeva (ed.) 366431. Oxford: OUP.
Foley, William A. & Olson, Mike. 1985. Clausehood and verb serialization. In Grammar inside
and outside the clause. Some approaches to theory from the field. Johanna Nichols & Anthony
C. Woodbury (eds), 1760. Cambridge: CUP.
Franois, Alexandre. 1997. La subordination sans marques segmentales: Formes de dpendance
interpropositionnelle dans le discours. Mmoire de DEA. MA thesis, Universit Paris-III
Sorbonne Nouvelle.
Franois, Alexandre. 2003. La smantique du prdicat en mwotlap (Vanuatu). Collection Linguistique de la Socit de Linguistique de Paris, 84. Louvain: Peeters.
Franois, Alexandre. 2004. Chains of freedom: Constraints and creativity in the macro-verb
strategies of Mwotlap. In Complex Predicates in Oceanic Languages: Studies in the Dynamics of Binding and Boundness. Bril & Ozanne-Rivierre (eds), 107143 Berlin: Mouton de
Gruyter.
Franois, Alexandre. 2005a. Diversit des prdicats non verbaux dans quelques langues ocaniennes. In Les constituants prdicatifs et la diversit des langues: Actes de la Journe de la
Socit de Linguistique de Paris [Mmoires de la Socit de Linguistique de Paris], Jacques
Franois & Irmtraud Behr (eds), 179197. Louvain: Peeters.
Alexandre Franois
Franois, Alexandre. 2005b. Unraveling the history of the vowels of seventeen northern Vanuatu
languages. Oceanic Linguistics 44(2): 443504.
Franois, Alexandre. 2006. Serial verb constructions in Mwotlap. In Serial Verb Constructions:
A Cross-linguistic Typology [Explorations in Linguistic Typology], Robert M.W. Dixon &
Alexandra Y. Aikhenvald (eds), 223238. Oxford: OUP.
Franois, Alexandre. 2009. Verbal aspect and personal pronouns: The history of aorist markers
in north Vanuatu. In AFestschrift for Robert Blust, AlexanderAdelaar & AndrewPawley
(eds), 103126. Canberra: Australian National University.
Franois, Alexandre. Forthcoming. From deictics to clause linkers. Discourse deixis, topicalization and clause backgrounding strategies in the languages of the Banks islands (Vanuatu).
Studies in Language.
Givn, Talmy. 1984/1990. Syntax. A Functional-typological Introduction. Amsterdam: John
Benjamins.
Haspelmath, Martin. 2007. Pre-established categories dont exist consequences for language
description and typology. Linguistic Typology 11(1): 119132.
Heath, Jeffrey. 1985. Discourse in the field: clause structure in Ngandi. In Grammar inside and
outside the Clause: Some Approaches to Theory from the Field, Johanna Nichols & Anthony
C. Woodbury (eds), 89110. Cambridge: CUP.
Heine, Bernd & Kuteva, Tania. 2002. World Lexicon of Grammaticalization. Cambridge: CUP.
Lambrecht, Knud. 1994. Information Structure and Sentence Form: Topic, Focus, and the Mental
Representation of Discourse Referents [Cambridge Studies in Linguistics 71]. Cambridge:
CUP.
Launey, Michel. 1994. Une grammaire omniprdicative: Essai sur la morphosyntaxe du nahuatl
classique. Sciences du Langage. Paris: CNRS.
Lemarchal, Alain. 1989. Les parties du discours. Syntaxe et smantique [Linguistique Nouvelle].
Paris: Presses Universitaires de France.
Mhlhusler, Peter, Dutton, Thomas E. & Romaine, Suzanne. 2003. TokPisin Texts: From the
Beginning to the Present [Varieties of English around the World T9]. Amsterdam: John
Benjamins.
Noonan, Michael. 1985. Complementation. In Language Typology and Syntactic Description,
Timothy Shopen (ed.), 42140. Cambridge: CUP.
Stassen, Leon. 2000. And-Languages and With-Languages. Linguistic Typology 4:154.
Thompson, Sandra A. & Longacre, Robert E. 1985. Adverbial clauses. In Language Typology and
Syntactic Description. Timothy Shopen (ed.), 169234. Cambridge: CUP.
Tomlin, Russell. 1985. Foreground-background information and the syntax of subordination.
Text 5: 85122.
van der Auwera, Johan & Plungian, Vladimir. 1998. Modalitys semantic map. Linguistic Typology
2: 79124.
van der Auwera, Johan, Dobrushina, Nina & Goussev, Valentin. 2008. Imperative-hortative
systems. In World Atlas of Language Structures Online. Martin Haspelmath, Matthew
Dryer, David Gil & Bernard Comrie (eds), Ch. 72. Munich: Max Planck Digital Library.
<http://wals.info/feature/description/72> 15 November, 2008.
Tense-mood concordance and clause chaining

in Mankon (a Grassfields Bantu language)
Jacqueline Leroy
Universit Paris Descartes, Laboratoire des langues et Civilisations

Tradition Orale, Fdration Typologie et Universaux Linguistiques-CNRS
In this article we explore the ways in which Mankon, a Grassfields Bantu
language, links clauses together to form certain sentences by using four distinct
verbal constructions the successive, exhortative, non-future and future
consecutive.These constructions encode the syntactic and semantic links
betweenclauses.
After a brief typological survey of the language we present the affirmative
perfective conjugation. Sections3, 4, 5 discuss sentences whose constituting
clauses are not linked by coordinating or subordinating morphemes, or
wheresuch links are optional. Even when one such morpheme is required, this
does not indicate the status of the clause it marks. Section6 deals with fixed
constructions relating to the sentences examined in the preceding sections.
Theconclusion briefly reviews research on the topic for other
Grassfields languages.
1. Introduction
Mankon is spoken in and around Bamenda, the capital of the North-West Province of
Cameroon. It belongs to the Mbam-Nkam group of the Grassfields Bantu languages.
The aim of this article is to show how choosing one of the four verbal constructions, i.e. the successive, the exhortative, the non future consecutive, or the future
consecutive, suffices to mark the syntactic and semantic links between the clauses contained in certain complex sentences.
Section2 gives a brief typological overview of Mankon; Section3 presents the affirmative perfective conjugation; Sections 4 to 6 describe the complex sentences which
do not contain a conjunction, or, if they do, where the conjunction is optional (cf. |t|
for to, in order to, so as to in 5. and |b| if in 6.) or else the obligatory presence of
the conjunction is not relevant (cf. |a| that in 4 and 5). Section7 presents several complex sentences where the verb, either in initial position or in the following clause, has
Jacqueline Leroy
become specialized in the expression of various semantic notions (comparison, possibility, etc.); these are the particular uses of the sentences presented in 4 and 5.1
2. Typological overview
2.1 Phonology
The canonical form of lexical morphemes is CV(C), that of grammatical morphemes
is (C)V(N), C or N.
2.1.1 Vowels
Nine simple vowels are found in the lexical morphemes: /i, e, , Y, 6, a, u, o, f/, three
front diphthongs /ye, y7, ya/ and seven back diphthongs /wi, we, w7, wu, w6, wa,
wf/. The grammatical morphemes contain a reduced set of vowels: /u/ (in CV(N)),
// and /a/ (archiphonemes). There are gaps in the distribution of certain vowels,
and numerous cases of neutralization are also found. Furthermore, in grammatical
morphemes, the vowels // and /a/ undergo various reduction, deletion and merging
phenomena.Thus the three nasal vowels e , 6 and f , are the result of the merging of

the sequences Vn + and Vn +a where V is an open vowel (7, a, f) and +is a morpheme or word boundary.2
2.1.2 Consonants
Morpheme initially, one finds the following consonants (C1): the stops /t, k, b, d/, the
affricates /ts, dz/, the continuants /f, s, z, >, l/ the nasals /m, n, \, /. The inventory of final
consonants (C2) is limited to /m, n, , b, r, >, /. /i/ induces palatalization and /u/ labiodentalization of certain C1 for example: k [kf ] or [pf ]; > [v]; s [] (one should
note that labio-dentalization does not depend only on /u/ but also on the C2). The voiced
continuants /z, >, l/ are articulated as stops [dz, g, d] after the nasal archiphoneme /N/3
and C2 /n/. These two nasal consonants cannot appear before C1 nasals /m, n, \, / nor
the unvoiced continuants /f, s/; they share the place features of the other C1.
2.1.3 Tones
The tones have both lexical and grammatical functions. For the lexicon, there are two
distinctive levels: high and low.
. I wish to thank Isabelle Bril and Denis Creissels for their suggestions and corrections
which were greatly helpful for constructing this article.
. Cf. Leroy, 1994 and 2007.
. When the nasal /N/ constitutes a morpheme on its own, it gives way to the archiphoneme /G/.
Tense-mood concordance and clause chaining in Mankon (a Grassfields Bantu language)
Two tones are associated to each lexical stem CV(C). Thus there are four possible
combinations: -hh, -ll, -hl and -lh (the noun prefix always carries a low tone). Verb
stems belong to two tone classes: H and L.
A frequent occurrence is that there are more tones than syllables. This can be due
to several factors: (1) the structural form itself contains tones which are not linked to
any syllable; (2) through mergings, mentioned in 1.1.1., tones become separated from
their syllable; (3) furthermore, a syllables tone may be repeated on the following syllable. These factors have the following consequences:
tonal neutralization: whereas, e.g. the structural tones of |m-bf| mushrooms

are |l-hh| and those of |m-bf| pumpkins are |l-lh|, these two nouns both
show the pattern l-lh after a L (or in citation form), but their patterns are distinct
when preceded by a H:4
|m lf>
m-bf
a|
I
suc.take 6-mushroom dm
(and) I took mushrooms.
m lf> mbff
|m lf>
m-bf
a|
I
suc.take 6-pumpkin dm
(and) I took pumpkins.
m lf> mbff
downstep and upstep: these phenomena are due to a L not being directly articulated
when surrounded by H tones. Depending on the context, a hlh(h) sequence may
either be realized hh(h) the first H is spoken on a higher pitch than the following
H tone or tones which can be said to be downstepped from the level of the preceding
super H or hh(h) the second H and following H tones are spoken on a higher
pitch than the first H tone: they are upstepped from the level of the preceding H.
Besides downstep and upstep as described above, I have been led to posit several
other types of step, namely: anticipated downstep, postponed upstep and simplified
upstep (hh simplifies to h in a non root syllable). For any given type, a step (or the
first in a succession of steps) automatically triggers the appearance of a super high
level. Moreover the downstepped or upstepped H creates a new upper limit which
affects all following H. Below are examples of downstep (hh) and upstep (hh):
|-l`
b-l6m`
a|
7-country 7 2-sorcerer dm
The country of the sorcerers.
l bl6m6
|-b` b-km a|
7-slave 7 2-notable dm
The slave of the notables.
b 6 bkm6
. Cf. Leroy, 1979 and 2007.
Jacqueline Leroy
the formation of contour tones, which are to be analyzed as the combination of

several distinct level tones on a single syllable. The language has three falling tones
HL, HL and HH, three rising tones LH, LH and HH and one rising falling
tone LHL (the falling rising tone HLH is forbidden, and there are various strategies for avoiding its formation).
2.2 Morphology
2.2.1 Nouns
Mankon is a noun class language. The nouns are divided into different groups or
genders.5 Each gender is made up of two classes with a singular/plural distinction
in the case of count nouns, or one singular or plural noun in the case of non count
nouns. A nouns class is indicated in the form of the word, through noun prefixes
(NP). A noun is thus composed of a NP followed by a noun root, and a single root
may bear different prefixes, one of which corresponds to a singular class, the other
to a plural class. There are six singular classes (1, 3, 5, 7, 9 and 19) and four plural classes (2, 6, 8 and 10). The numbers attributed to the noun classes are taken
from the Common Bantu numbering convention. In effect, the correspondences
between the class system in Mankon and in Common Bantu are clear enough so
that this convention may be used without anymodification.
This grouping of nouns into classes is accompanied by agreement phenomena,
moreover this is one of the defining features of noun class systems. Each class has
its own agreement series. The elements which agree in class with the noun are
noun modifiers, subjects and other pronouns corresponding to the nouns:
|-k` z- n` a|
7-pan 7-this dm
This pan
k z6
|-k` ts- n` a|
8- pan 8-this dm
These pans
k ts6
When a noun is in the locative (i.e. when it is governed by the locative preposition ||), it determines a specific agreement series, which does not correspond to
its inherent noun class. Moreover, it cannot be modified by a possessive, which in
such cases is replaced by an independent pronoun.6
. Cf. Leroy, 1977 and 2007.

. Cf. Leroy, 1983 and 2007.
Mankon is extremely rich in nominal derivation, consisting in deriving nouns

from verb roots: to the root, one adds a nominal prefix corresponding to one of
the noun classes as well as a derivational suffix. The derivation of verbal nouns
(the act of) is systematically carried out in class 9. Some verbs can be derived
in several classes, with a different meaning in each case:
|zf-|
rub, marry
ndzff > (9) act of rubbing, marrying
ndzff (1/2) person(s) who rub(s) (with oil)
zfff (7/8) calabash(es) containing rubbing oil
2.2.2 Pronouns
These belong to several categories:
Person (1st and 2nd persons) and class (3rd person) subject pronouns (cf.
Section3)
Person object pronouns (1st, 2nd and 3rd persons) which refer to humans only
(but cf. paragraph 2.3.1.)
Independent person pronouns (1st, 2nd and 3rd persons) referring to humans only
(but cf. paragraph 2.3.2.) and class pronouns (3rd person) referring to human as
well as non human entities. Class pronouns contain the prefix which corresponds
to the class of the entity referred to plus a root; they may serve object function.
2.2.3 Verbs
Verb stems contain a root which may be followed by a formal suffix, inseparable
from the root, or a derivational suffix. The four formal suffixes, |-n|, |-k|, |-s|
and |-t| are identical to the derivational suffixes which they undoubtedly stem
from, especially as their use is mutually exclusive. Their suffixes, whether formal
or derivational, have no inherent tones. They take the tone of the final element.
When one takes the semantic-syntactic implications of the derivational suffixes
into account, two |-n| suffixes are to be distinguished: |-n1| and |-n2| and two
|-k| suffixes: |k1| and |-k2|. Without entering into too much detail, it may be
said that |-n1|, |-k1| and causative |-s| change the verbs valency, |-n2| and |-k2|
respectively imply collective and distributive plurality of one of the arguments,
|-t| has diminutive meaning.7 These suffixes are mutually exclusive.
There is no passive voice: the use of the class 2 subject pronoun, which, besides its
anaphoric meaning has indefinite meaning, makes it possible to omit the agent (see
the C4 clause of Example12 in Section5).
. Cf. Leroy, 1982 and 2007.
Jacqueline Leroy
Taking into account the aspectual distinction perfective/imperfective and

the polar distinction between affirmative and negative, four conjugations may be
distinguished: affirmative perfective, negative perfective, affirmative imperfective and
negative imperfective (cf. Section3).
Verbs are very often preceded by one or two (very seldom more) auxiliaries. In
this case the verb and the auxiliary (or auxiliaries) form a sequence called verbal
chain here. We consider auxiliaries terms which are freely conjugated, and which
are followed by a verb which can only be conjugated in either the future or non
future consecutive mood, depending on the tense-mood concordance rules (cf.
paragraph 4.1.). If one or more auxiliaries are interposed between the first auxiliary
and the verb they also can only be conjugated in the consecutive mood. This seems
to prove that verbal chains are the result of the grammaticalization of clause chains
(cf. Section4.). Some of these auxiliaries correspond to verbs currently in use in the
language, some do not. Several auxiliaries are presented in paragraph 4.2.b. where
they are to be found in the examples. The auxiliaries may be divided into four
groups, depending on their meanings: (1) ten temporal auxiliaries, divided into
two sub groups determined by their distribution;8 (2) three frequency auxiliaries;
(3) one parallelism 9 auxiliary; (4)three manner auxiliaries.10
2.3 Syntax
2.3.1 Simple sentences
The basic word order in Mankon is SV(O)(X). The presence of a subject (S),
whether a noun or a pronoun, is obligatory, unless the verb is in the imperative or
consecutive. O is the object modifier, corresponding to the patient, and X represents the other determiners.
When the subject is focalized, the order between S and O is reversed, but syntactically
speaking S is no longer the subject since it no longer governs agreement with the verb.
In this case, the class 1 pronoun is used in subject function.
The topic is placed sentence initially. Speakers pause slightly between the topic and
the rest of the sentence, which entails a lack of segmental and/or tone sandhi between
the last word in the topic and the following word. The topic has discourse function but
. Cf. paragraph 4.2.b. for the presentation of six of them.

. Cf. paragraph 4.2.b.
. Cf. Example10 for one of them.
no syntactic function. However, it is nonetheless represented by a pronoun in the following clause, unless it is in object function, generally speaking.
Mankon uses prepositions. Many are derived from locative phrases (for example:
| -t `| t, literally: loc 7-head loc, on). It should be noted that the object
modifier is governed by the object preposition (OP) , which is homophonous
with the locative preposition , if it refers to a person or if it is in class 1. Some
pronouns, in particular object pronouns, are also governed by this preposition.
Noun modifiers follow the noun they determine, although some may precede it
to give it emphasis.
In negative clauses, the constituent order is (1) [S +Neg +predicate +O +X] or
(2) [S +Neg +O +X +V] if the negation is |s|, |k | or |t|. It would seem that
the contents of a clause whose constituent order is the second are less informative
than those in a clause whose constituents are in the first order. The negation |k|
comes clause initially. The subject must be accompanied by an object pronoun
agreeing in person with the subject. This pronoun is placed between the predicate
and the object. When the subject is not human, the 3rd person singular object
pronoun y is invariably used.
All sentence types, with the exception of yes-no questions, are marked by one of
two particles (discourse markers, dm) a or >e. Until one of them has been uttered,
the sentence is considered incomplete, and the listener waits for the rest, or else it
is a yes-no question. These particles have no inherent tones, but make it possible
for the last tone of the preceding morpheme to be realized. They are in complementary distribution. The interrogative terms in partial questions occupy the
same position as the non interrogative terms to which they correspond in assertive sentences.
2.3.2 Complex sentences

In this section, we shall only study complex sentences, or more precisely, clauses making up complex sentences which require the use of conjunctions. Complex sentences
which do not use conjunctions will be presented in Sections4 to 7.
Relative clauses
As noun modifiers, relative clauses are generally (but see below) introduced by the
relative morpheme |- | which agrees in class with the relatives antecedent. The relative morpheme may be followed by the conjunction |nb | mb that. The relative morpheme |- | has no function in the relative. The antecedent is taken up in the
relative in the form of a pronoun corresponding to its function in the relative clause,
unless it has object function (this non-referral to the object is not limited to relative
clauses). The relative morpheme |-| is not used when the relative is not necessary for
Jacqueline Leroy
the identification of its antecedent: in this case, the conjunction mb alone introduces the relative clause.
Completive clauses (or sentences)
These are introduced by the conjunction |gY

| :Y , :Y , etc. that unless it is
governed by the verb |>Y
| say, state. This conjunction is a frozen form of the verb
|>Y
| in the non-consecutive future,11 which explains why it cannot be used with this
verb. When the governing verb is a speech verb, the contents of the completive are
often reported speech. If the subject of the governing verb is in the 3rd person singular, there are three ways (styles) of reporting the speech, which are distinguished by
the use of a specific set of pronouns: direct, indirect and semi-direct styles. (1) In the
direct style, the pronouns are the same as those used in independent sentences: the
1st person (subject, object and possessive) pronouns refer to the speaker, those of the
2nd person to the hearer and those of the 3rd person to any entity absent from the discourse situation. (2) In the indirect style, only the 3rd person pronouns are used. However, if the subject of the completive corresponds to the subject of the governing verb,
it is the independent pronoun |zY | (s)he which is used, and which thereby takes
on logophoric meaning, instead of the subject pronoun of class 1 .12 (3)In the semidirect style, the 3rd person pronouns are used for the speaker and any entity absent
from the discourse situation (in subject function one uses |zY | if there is coreference
with the subject of the governing verb and for any entity absent from the discourse
situation); but it is the 2nd person pronouns which are used for the hearer.
Temporal clauses
To express the semantic relation of simultaneity or consecution, one uses a relative

clause whose antecedent, when it is expressed, is |nkf`| time (cl. 9). When it is
not expressed, the relative clause is introduced only by the conjunction mb. This
clause generally precedes the governing clause, but may also follow it.
Causal clauses
Two possibilities exist: (1) the causal clause is also a relative clause introduced by
the conjunction mb, its antecedent |-dz`| way, manner (cl. 7) rarely being
expressed. It precedes (most frequently) or follows the governing clause; (2) it is introduced by the conjunctive phrase |`mb (-k) :Y
| because a phrase where
. frozen form since it is the [F]C which is used even when the tense-mood concordance
rules demand the use of the [+F]C. Originally, this verb in this form (i.e. in the future or nonfuture consecutive) must have constituted a clause chained to the preceding clause on its own.
. This is true for all completive subordinates, even when the subject does not denote a human.
|`mb | is the grammaticalized form of the verb |b| ?13 in the non-future consecutive, |-k| is the interrogative class 7 noun what? (which is not always expressed)
and |:Y
| that, the conjunction which introduces completive clauses. The causal
clause usually follows the governing clause but may also precede it.
Finality clauses
These clauses may be introduced by the conjunction |t| in order to/that (but cf. 5.).
Hypothetical clauses (realistic domain)
There are two possibilities: (1) they may be introduced by the conjunction b or mb
if or by the conjunctive phrase b () b :Y if (it) be that (but cf. 6). In this
case they precede the main clause; (2) they follow the main clause (this order is much
less frequent than the order subordinate clause +main clause), in which case they are
obligatorily introduced by the phrase mb b that if .
Hypothetical clauses (imaginary domain)
To express a hypothetical fact belonging to the imaginary domain, the language has
recourse to two juxtaposed clauses (cf. 7.5.). The main clause follows the two clause group.
The main clause must be linked to the expression of the hypothesis by b/m then,
in that case, a morpheme which only expresses a logical link between two processes.
The conjunction introduces all clauses whose main verb is in the exhortative,
whether they are independent or not (cf. 4.1.).
3. The affirmative perfective conjugation

The Mankon verbal forms (or constructions) may be divided into four groups taking
into account the aspectual perfective/imperfective distinction and the polarity affirmative/negative distinction. Four conjugations are thus distinguished:
affirmative perfective conjugation;

negative perfective conjugation;
affirmative imperfective conjugation;
negative imperfective conjugation.
. In fact, the first term in this phrase is the prepositional element |mb ()| because of.
|b| has retained, because of its previous status of non grammaticalized verb, the government
of its object complement. Thus the object preposition || is used with class 1 nouns, person
object pronouns, etc., whereas in all other circumstances the elements are governed directly.
Jacqueline Leroy
These conjugations are not entirely symmetrical: the completed aspect of the indicative (CA) and the relative mood are absent from the negative perfective conjugation.
In the affirmative imperfective conjugation, in the indicative, there are only three constructions present, past, future whereas there are seven in the perfective; in the
relative, there are only two constructions past and present instead of the three
in the perfective. The completed aspect of the indicative and the relative mood are
absent from the negative perfective conjugation; while in the indicative there are four
constructions present, recent past, far past, and future there are only three in the
affirmative imperfective conjugation.
I will limit myself here, for lack of space, to the affirmative perfective conjugation, as
it is the one which has the most forms and is the most frequent in the texts collected.
Within this conjugation, the verb forms are distributed over seven moods. In this
article we will mostly look at the successive, exhortative and consecutive (non-future
and future) moods since, as indicated in the introduction, they are the ones found in
the non-initial clauses of certain complex sentences and thus well illustrate how, in the
absence of conjunctions, the language marks syntactic and semantic relations between
the clauses which make up these sentences:
the indicative, a personal and temporal mood. This mood shows a triple aspectual
distinction: effective/completed/virtual (i.e. which has not yet happened). Effective and Completed (CA) combine with three temporal meanings: effective P0 and
completed P0 CA are used for dynamic processes having effects which are felt in
the present or for current states; effective P1 and completed P1 CA refer to a recent
past; effective P2 and completed P2 CA refer to a far past. The virtual coincides
with the future tense;
the relative, a personal and temporal mood used in relative clauses (contrary to the
indicative, this mood has neither future forms nor forms in the completed aspect);
the conditional, personal mood used in hypothetical clauses (realistic domain),
whether the condition be temporal or logical;
the imperative, injunctive mood which has only one form, in the 2nd person
singular;
the exhortative, personal injunctive mood which can be used in independent
clauses, or in non-initial clauses in certain complex sentences (one must remember that a clause whose verb is in the exhortative is always introduced by the conjunction || that);
the successive, personal mood used in non-initial clauses in certain complex
sentences;
the consecutive, non personal mood (i.e. not having any person or class markers) is used in non-initial clauses in certain complex sentences, when there is
no change of subject in relation to the preceding clause. A distinction is made
between non-consecutive future and consecutive future. Note that consecutive

does not convey resultative but successive meaning (cf. paragraph 4.2.).
The linear organization of the verb phrase:
1
(Subject pronoun)
(Formative)
Stem
Final14
Depending on their tonal behavior, the subject pronouns or subject class markers
when the subject is a noun (cf. below) may be divided into three groups:
Person pronouns: m I, you, t us (dual),15 n you (plural) (in the

3rd person singular the class 1 pronoun is used, in the 3rd
person plural the class 2 pronoun is used)
Class pronouns:
(cl. 1), (cl. 9)
Class pronouns:
n (cl. 5), (cl. 7), f (cl. 19), b (cl. 2), m (cl. 6), ts (8 and 10).
In the presence of a subject noun, the subject marker of the nouns class is used. These
markers all of V (a or ) shape only have the same tonal characteristics as the corresponding pronouns; thus, on one hand one has the class 1 and 9 (L tone) markers
and on the other, the class 7 marker and (H tone) for all the other markers.
In the affirmative perfective conjugation, one finds three subject pronoun/subject
marker paradigms S1, S2 and S3 which show distinctions in tone (below, m represents the person pronouns, or represent the class 1 and 9 pronouns and b or b the
pronouns for all the other classes):
S1: m, , b (person and class 1 and 9pronouns: L; other pronouns: H)

S2: m, , b (person pronouns: L; class markers: H)
S3: m, , b (all pronouns are L)
The stem is made up of a verb root and an optional formal or derivational suffix
(cf. 2.2.3.).
Each construction is distinct from all others by the combined characteristics of
its subject pronoun paradigm, its formative and final. One could however consider
that the subject pronoun paradigm gives information on the mood; S1 is used in the
indicative and in the relative, S2 in the successive and the conditional, S3 in the exhortative; the absence of pronoun is characteristic of the imperative and the consecutive.
. Formative andfinal are terms borrowed from Meeussen (1967).

. we (exclusive) and we (inclusive) are expressed respectively by the independent pronouns bY>6 and bY>n6. They are followed by the subject class marker just as plural nouns
are (cf. below).
Jacqueline Leroy
The formative is what determines either the mood: successive, conditional, exhortative, consecutive or imperative (the latter being characterized by the absence of any
formatives); or the tense in the indicative and relative, the effective present of these
two moods being characterized by the absence of any formative, and the completed
present (P0 C) of the indicative by a low tone. The final mostly determines the aspect
(completed vs. effective in the indicative) or the mood (relative vs. indicative).
Table 1. The affirmative perfective conjugation
imp
p0
p1
p2
p0 r
p1 r
p2 r
`/
n`
n/n`
s1
s1
p0 ca
p1 ca
p2 ca
s1
suc
s2
cd
s2
exh
s3
[f] c
[+f] c
fut
k
k
k
k
k
k
n
`
s1 ()
m `
In Table 1, the construction labels are to the left: imp: imperative; p0: effective
indicative present; p1: effective indicative recent past; p2: effective indicative far past;
p0 r: relative present; p1 r: relative recent past; p2 r: relative far past; p0 ca: completed
indicative present; p1 ca: completed indicative recent past; p2 ca: completed indicative
far past;suc: successive; cd: conditional; exh: exhortative; [f]c: non-future consecutive; [+f]c: future consecutive; fut: indicative future.
Column 1 contains the subject paradigm, column 2 the formative paradigm, column 3 the (v) stem, and column 4 the final paradigm.
The order in the presentation of the formula for the various affirmative perfective conjugation constructions aims to highlight the formal similarities and differences
between these constructions.
4. Clause chains
Clause chains differ from the other complex sentences presented in the Sections5 and
6 in that they obey specific tense-mood concordance rules (cf. paragraph 4.1.), which
is not the case for the sentences in Sections5 and 6. Furthermore, on the semantic
level, the clauses in chains are not hierarchically ordered (cf. 4.2.).
4.1 Formal characteristics

These clauses are not linked by any conjunction, unless the verb is in the exhortative.
In fact, all clauses, even if they are independent, are introduced by the conjunction
that when their verb is in the exhortative ( b b6 They should clear!). Therefore,
the conjunction does not indicate the syntactic status of the clause it introduces.
Remark 1: THe exhortative mood always has exhortative or injunctive meaning in
independent clauses (cf. the example above). But in clause chains it may or may not
have this meaning: it all depends on the verb form in the preceding clause (cf. clause
C4 in Example1 and Section5).
In a chain of clauses, in the initial clause the verb16 is conjugated, depending on
the type or status of the clause, in the indicative, relative, imperative, conditional, etc.
moods.17 But the verb in the following clauses is in the future consecutive ([+F]C) or
in the non future consecutive ([F]C) as long as, from one clause to the next, there is
no change in subject, and in the exhortative (EXH) or successive (S) whenever there
is a change in subject.18
Furthermore, the clause chains obey strict tense-mood concordance rules which
are summarized in Table2.
. It could also be an auxiliary (cf. paragraph 2.2.3.).

. One may also speak of a relative chain when the initial clause is a relative, of an injunctive chain when the first clause is an injunctive, etc.
. There is an exception to this use of the successive (cf. Section6).
Jacqueline Leroy
Table 2. Tense-mood concordance rules

(1)
(2)
exh
[+f]c
s
[f]c
fut
[+f]c
exh
Other
[f]c
+
+
+
+
+
+
+
+
+
+
Table 2 should be read as follows: if in a clause (1) the verb is in the indicative
future (fut), then the verb in the following clause (2) will be in the exhortative (exh)
or in the future consecutive ([+f]c), but not in the successive (s) nor in the non future
consecutive ([f]c), etc. Otherrefers to any verbal construction not mentioned in
this table (cf. Table1).
Remark 2: THe cells left empty indicate that the combinations between the verbal
constructions are possible, but not in clause chains as defined above (cf. Section5).
(1) C1. m m f 6 tm6 nb t
m > wmbn //19

I fut put fruit kola loc.head child your male
C2. >7//
C3. t7>6
y n >> //

[+f]c.go
[+f]c.place.op him with distance
C4.
d6
dz z //
C5.
m tm //

that.you ehx.show way rel
you fut shoot
C6. s6 //
C7. wr //

[+f]c.split
[+f]c.cut
C8. t>a20 l7m6
mf gw

without you p0.hurt.op child the.dm
Ill put a kola nut on the head of your son, Ill place him at [a certain]
distance and youll show how you shoot and split [it] without hurting
the child.
. Mankon has numerous merges. In the example glosses, the morphemes are not all systematically explicitated.
. The conjunction t>a or tka has only been encountered twice, and has not been analyzed. The first syllable is undoubtedly the negation marker t. The second sentence in which
it is found is: m-w m wn m l k\ tka m r l Why has the pap got finished so
quickly, while I am not satisfied?
In Example1 there are two clause chains: the verb in C1 (the initial clause) in the
first chain is in the indicative future (Fut); in C2, the subject is the same as in C1,
and in C3 it is the same as in C2, therefore the verb in C2 and C3 is in the future
consecutive ([+F]C). In C4 however, the subject changes, the verb in this clause is
therefore in the exhortative (Exh). The initial clause (C5) in the second chain is a
relative: its verb is in the indicative future, in C6 the subject is the same as in C5,
and in C7 it is the same as in C6, therefore the verb in clauses C6 and C7 are in the
future consecutive ([+F]C). The last clause (C8) in the sentence is not chained to the
preceding clauses (it is a subordinate clause), therefore its verb does not obey the
tense-mood concordance rules.

k //
imp.wash yams
(2) C1. s>
C2. ndm
bw bb >

[f]c.cook against hour two
dm

Wash up the yams and cook them for two hours.
In Example2, the verb in C1 (the initial clause) is in the imperative (Imp; cf. the other
column in Table2), in C2 the subject is the same as in C1, the verb in C2 is therefore
in the non future consecutive ([F]C).
(3) C1. slm6 >
Yn6
nd
ndw w //

Silum+1 p0.aux.ca [f]c.rise.loc day
this
C2. f
ts77 t
dzY
6 zm //

[f]c.go out just without eat thing
C3. :7 //
C4. nt7
d6 nts

[f]c.go
[f]c.stay.loc place watching

ms
m //
C5. \m6 t //
C6. nd
z //
sorghum the
sun s.shine
hunger s.pain
C7. mY
> mw7r6 fY
//
C8. ze

eyes
of her s.darken
1 s.see

:ff ts7r //
C9. kff //
C10. t //
rock certain
[f]c.climb
[f]c.sit
C11. nts
wt //
C12. nt6

[f]c.stay there on top
[f]c.then. ipf

ndzfb6
:Y
[] //
C13. m:y ntff
[f]c.sing. ipf that
woman palace
wmff k6
f6
y ffn ts77
one
s.also. ipf [f]c.go out.op her field just+loc
Jacqueline Leroy
ndm m w // C14. :7 //
C15. ndz6
way
very the
[f]c.go
[f]c.hear.7
dz // C16. za mb slm6 zfbn6

l //
way
rel that
Silum.1 prt.sing.r.ipf am
C17. slm6 z7n6

m:y

Silum.1 p0.see.ca op.woman
One day, later on, after getting up, Silum went out just empty stomached,
she went and stayed keeping a watch on the sorghum, [until] the sun shone
hard, she was hungry, her eyes were darkened, she saw some rock, climbed
up and sat down on top of it, then started singing [], at that time one
chief s wife going to the field just along the same way, heard how Silum was
singing. After having seen the chief s wife, Silum
Remark 3: THe translation of Example3 deviates from the syntactic structure of the
text in that it uses temporal subordinates, thereby creating a syntactic hierarchy among
the clauses, which is absent from Mankon.
In Example3 the verb (here the temporal auxiliary |>Y
|; cf. 4.2.b) of the initial clause
(C1) is in the present of the completed indicative (P0 CA), in C2, C3, and C4, the subject is the same as in the preceding clause each time, therefore the verb in these clauses
is in the non future consecutive ([F]C). The subject of clauses C5, C6, C7, and C8 is
different in each case from the subject of the preceding clause, therefore their verbs are
in the successive (S). In C9 the subject is the same as in C8, in C10 is it the same as in
C9, in C11, it is the same as in C10 and in C12 it is the same as in C11, which explains
the use of the non future consecutive ([F]C) in these four clauses. Since the subject
in C13 is different from that of C12, the verb in C13 is in the successive (S). Lastly, the
subject in the clauses C14 and C15 being the same as in C13, the verbs in these clauses
are in the non future consecutive ([F]C). C16 is a relative clause embedded in C15:
its verb is therefore not subjected to the tense-mood concordance rules. The clause
C17, whose verb is in the indicative present, completed aspect (P0 CA), begins a new
clause chain.
There is no theoretical limit to the number of chained clauses. The chain in
Example3, with its sixteen clauses, is relatively long. It stops when reaching a clause
where the verb is in the indicative (P0 CA).
4.2 On the semantic relations between chained clauses

a. Initial clauses
It is the initial clause which, in part through the form of its verb, anchors the entire
chain in relation to the discourse situation, or the reported situation (in the case of
indirect speech), or in relation to a set of facts just presented in a narration (cf. C1 of

the Examples3, 6 and 7).
b. Subsequent clauses
The verbal constructions in the following clauses situate the process temporally as
compared to the time referred to in the initial clause. Only the initial clauses time
frame can be absolute, that of the other clauses is relative.
The relations between the clauses making up a chain are chronological. The order
of appearance of the clauses corresponds, in general, to their chronological order. Some
auxiliaries make it possible to refine the relations, and to establish a hierarchy between
the processes on the discourse level.
The succession relations can be explicitated by the auxiliary |t| then; this auxiliary also makes it possible to set apart process subsets having tighter relations with
each other than with the processes of the other subsets; furthermore, the use of
|t| often accompanies a change in aspect, especially a switch from perfective to
imperfective as illustrated by Example4 (also see C12 in Example3; and, without
a change in aspect, C3 in Example35):
(4) C1. w7r6 lf> :

Y
Y //
C2. nt7> //

she+1 s.take seeds
[f]c.put up
C3. nt6
k6bt6
[f]c.then+ipf [f]c.peel+ipf
She took the melon seeds, put them [in front of her]; then she was
cracking [the shells]
The parallelism auxiliary |k| also21 serves to express simultaneity between

processes, to establish a parallel among characters, processes(cf. C2 in Example5
below, and also C12 in Example3, C4 in Example9, C5 in Example14):
(5) C1. lY
n
y //

ant
s.drink.op her
C2. k
ndzY
6
y ts77 k6

[f]c.also [f]c.eat.op her just true.dm

the ant drank and ate a lot.
. The vowel G in this auxiliary, like that of the auxiliary t/tnG (cf. remark 6) is not found in
the lexical morphemes. This indicates, in my view, a strong degree of grammaticalization.
Jacqueline Leroy
The auxiliary |l|22 from the verb |l| come from, and especially the auxiliary |>Y
|
(in C2 below) from the verb |>Y
| do, homophonous with the verb |>Y
| say,
serves to indicate that a certain time span separates two processes (also cf. C1 of
Examples3, 7, 16 and 36, and C2 in Example6):
(6) C1. >7n6

n> gw l //

1 p0.go.ca.loc journey the am
C2. :w y w >Y

t6
mf gw

wife his the s.aux [f]c.shake.op child the
Some time after he had gone on the journey, his wife was
in labor of the child
The auxiliary |kw7n| from the verb |kw7n| come back home (cf. kwe in C5 of
Example7) serves to go back in time:
(7) C1. >

Yn6
mb //
C2.

10 p0.aux.ca [f]c.rot
1
w //
s.dig out
C3. s> // C4. :yb mb b:y

[f]c.wash
[f]c.share to
women

bntff b ts77 btsm6 [] // C5. b:y

of palace the just all
women
bntff b b kwe nd
nts
l //
of palace the 2 s.aux [f]c.ever [f]c.be am
C6. wts k
dw`6 //

certain s.neg 23 give birth.dm
After they (the calabashes) had rotted, she dug them out, washed them
and shared them out to all the chief s wives []. None of the chief s wives
had ever given birth [to a child].
The auxiliary |l| ever (cf. C5 of Example7, and C1 of the Examples28 and 36)
indicates a far past or future.
. The auxiliary |l| conjugated in the P0 of the indicative or relative serves to express that
a process took place earlier in the day. Example 30 illustrates this in its C1.
. Here it is the negation which is conjugated in the successive. In presence of the negation|k| the verb is not in the consecutive, but in a specific negative form.
5. Complex sentences with finality clauses

As is the case with clause chains, here one has a simple juxtaposition of clauses making
up complex sentences. But only the future consecutive and the exhortative may be used
in non initial clauses: the tense-mood concordance rules do not apply. It is this restriction which gives finality meaning to the non-initial clause. On the semantic level, a hierarchy is established where the finality clause is subordinate to the initial (or preceding)
clause.
(8) C1. nY

n k // C2. nt> n //

imp.put potash
[f]c.stir
C3.
f>6

that.9 exh.turn white+dm

Add potash and stir until (so that) it (the soup) turns white.
In Example8, the verb in the C1 (main) clause is in the imperative (Imp.). C2 is chained
to C1 because its verb is in the [F]C, in conformity with the tense-mood concordance
rules. But the verb in C3 is in the exhortative (Exh). C3 is therefore a (subordinate)
finality clause.
(9) C1. b sn6 nd7l // C2. fY

Ytn6
n //

2 s.tell.ipf stories
[+f]c.inform.ipf things
C3. ts mb k bY
> 7
wY
>6 //

rel que
neg we
know.op us

mb bY
>6 // C4. bt bY
> k
fY
Y
t
to
us.dm
fathers our s.also [f]c.inform
They (our mothers) would tell us stories to inform us of things that

we did not know; our fathers also informed [us]
The verb in C1 (the main clause) is in the successive (S). The verb in C2 is in the future
consecutive ([+F]C), as its subject is the same as that in C1. C2 is therefore a finality
subordinate clause. C3 is a relative clause, embedded in the finality clause; its verb is
therefore not governed by the tense-mood concordance rules. Clause C4 is chained to
C1, and as its subject is different from that of C1, its verb (here the auxiliary |k|) is in
the successive (S) as imposed by the tense-mood concordance rules.
Remark 4: In Example 9, if one had the non future consecutive in C2 instead of
the future consecutive, this clause would be chained to the preceding one and the
notion of finality would be absent. In that case the meaning would be: They (our
mothers) told us stories and informed us of things we did not know. Our fathers also
informed [us].
Jacqueline Leroy
While the tense-mood concordance rules demand in any case the use of the future
consecutive or exhortative, there is ambiguity, as the notion of finality may or may not
be implied:
(10)

( s // ) C1.
1 s.say
mnd // C2.
house
:Y zY
m ln t7
that he24 fut really25 [+f]c.stay.loc
y7
bvn //
lw77t6
[+f]c.hide.op her.loc corner
C3. z
wf

[+f]c.see.op person
In theory, this sequence of clauses could give rise to four different interpretations:
1. (He said) that he would really stay at home, hide in a corner and see the person
2. (He said) that he would really stay at home, hide in a corner to see the person
3. (He said that) he would really stay at home, to hide in a corner and see the
person
4. (He said that) he would really stay at home, to hide in a corner to see the person
The interpretation chosen by the storys narrator is the first one. Thus one has a clause
chain.
It should be noted that a clause with finality meaning can be introduced using the conjunction |t| forto, in order to, so as to. The moods used in this case are once again the
future consecutive and the exhortative. When the verb is in the exhortative the conjunction
is maintained, whence the conjunctive phrase |t | t6 in order that, so that.
Remark 5: THe conjunction |t|, which is always pronounced ta or t, most probably stems from a verb (perhaps |t| draw which still exists in the present state of the
language) conjugated in the future consecutive. All that remains of the future
consecutive formative |`| is a trace in the form of a rising or low tone on the conjunction,
whatever the preceding tonal context.
(11) C1. s6 mb bo
by //

1 s.say to
children her
C2. :6 b f //
C3. :7 // C4. nts //

that.that 2 exh.go out
[f]c.go
[f]c.stay
. One should remember that when in completive subordinates the subject is the same as
in the main clause, one uses the independent 3rd person singular pronoun |zY
| which in that
case takes on logophoric meaning.
. |lnG| really is a manner auxiliary which may stem from the verb |ln| be clear.
C4. nts //
C5. ndy7m
ntw z //

[f]c.stay
[f]c.watch maize the
C6. ta tm
m za // C7. mb

to [+f]c.shoot animal rel
that.9

yn6
C8. kfr6
l //
prs.come.r.ipf
[f]c.eat.ipf am.dm
C9. bfm b f

children the s.go out
She said to her children that they should go out, go and watch
the maize to shoot the animal which comes and eats [it]. The children
went out
In Example11, C2 is a subordinate completive clause dependant on the verb in C1: its

verb, in the exhortative, expresses injunction. C3, C4 and C5 are chained to C2. The
subject of these clauses being the same as in C2, their verbs are in the non future consecutive ([F]C) in conformity with the tense-mood concordance rules. C6 is a finality
clause dependant on C5 introduced by |t|: its subject is the same as in C5, the verb
is therefore in the future consecutive ([+F]C). C7 is a relative embedded in C6: its verb
does not follow the tense-mood concordance rules. C8 is chained to C7 and its verb is
in the [F]C since its subject is the same as that of C7 (C7 and C8 make up a relative
chain, see Note17). C9 is chained to C1: the subject in C9 being different from that in
C1, the verb in C9 is in the successive.
When the conjunction |t| is used, there is no risk of ambiguity: the clause can
only have finality meaning.
(12) C1. :6 f k6 l
tY
//

workers
pas come from.ipf.loc up
C2. >7 //
C3. bm
mkm m //

[+f]c.go
[+f]c.accept names their
C4. t6
b y7bt6
w

in order.that 2 exh.divide.op them
The workers would get up to go and answer the roll-call in order
to be divided [into groups]
In Example 12, the verb in clause C1 is in the indicative habitual past (PAS). Then
one finds a first finality clause (C2) subordinate to C1. Only the use of the future consecutive ([+F]C) indicates subordination to C2. Clause C3 is chained to C2, its verb is
therefore also in the [+F]C. Then one finds a second finality clause (C4), subordinate
to C3. It is introduced by the conjunctive phrase |t |. As the subject in C4 is different from that of C3, the verb in C4 is in the exhortative.
Jacqueline Leroy
6. Hypothetical sentences: The realistic domain

In general, verbs in hypothetical clauses are in the conditional (CD) as illustrated by
the following examples.26
When the hypothetical clause is in first position, and the main clause is simply
juxtaposed, the successive mood is used in the main clause, even if its subject is identical to the one in the hypothetical clause (also see Example15):
(13) C1. > d // C2. wa t fY
n6 //

1 cd.go place
rel night
p0.be dark.r
C3. lf> :
Y
Y z // C4. kr6

1 s.take melon the
[f]c.coil.loc

t
zY
// C5. \f // C6. ndy
head her
[f]c.lie
[f]c.sleep.dm
When she (Silum) gets to a place where it is night, she (Silum) takes the
stem of the melon, coils it round her head, lies down and sleeps.
C1 is the hypothetical clause. Its verb is in the conditional (CD). C2 is a relative embedded in C1, not subjected to the tense-mood concordance rules. C3 is the main clause
and its verb is in the successive (even though its subject is the same as in the hypothetical in C1). The clauses C4, C5 and C6 are chained to C3 and since there is no change
in subject from one clause to the next, their verbs are in the non future consecutive
([F]C).
This use of the successive in the main clause blocks its use in the clauses chained
(on the semantic level) to the hypothetical clause, when the subject of these clauses is
different from that of the hypothetical. In that case one uses the conditional (CD):
(14) C1. b :Y
ndw ln //

if
day cd.be clear
C2. t
dr
C3. >7
ndm //

9 cd.aux [+f]c.creep
[+f]c.go way
C4. wa >7n6 //
C5. k6
ndzf6 >

rel 9 prt.go.r.ipf
1 s.also.ipf follow.ipf dm
At daybreak, as soon as it (the melon) creeps, wherever it goes,
she follows [it].
. Use of the conjunction b/mb is optional. When the hypothetical clause is introduced bythe
phrase b b :Y (literally: if 1 CD.+be that; if it be that), one may use the indicative.
Remark 6: |t-| or |tni | is an immediacy temporal auxiliary. It is freely conjugated

and is followed by verbs in the consecutive, but it is always the future consecutive
([+F]C) which is used. The use of |t-| is incompatible with the use of negation.
Depending on the context, it can be translated as hardly(when), as soon as, be
on the point of , just as. (also see C2 and C4 in Example16).
C1 and C2 are hypothetical clauses. Since there is a change in subject from C1 to
C2, the verb in C2 (here the auxiliary t) is in the conditional, like the verb in C1.
The subject of C3 being the same as in C2, the verb >7 go in C3 is in the future
consecutive, as is the verb dr creep in C2.27 The successive is used in C4, which
is the main clause.
On the contrary, as illustrated by Example15, if the subject of a second hypothetical clause (C2) is the same as that of the preceding hypothetical clause (C1), the non
future consecutive ([F]C) is used, in conformity with the tense-mood concordance
rules presented in 2. The use of the successive indicates the main clause (C3).
(15) C1. m t
ky // C2. m7t //

I cd.draw water
[f]c.finish
C3. m s> k6

I s.wash pans.dm

When I have finished drawing water, I do the dishes.
In Example16 below, there is a sequence of four hypothetical clauses (C1, C2, C3 and
C4): the subject in C2 is different from that of C1, and that of C3 is different from the
one in C2, therefore the conditional is used in C2 and C3. But the subject in C4 is the
same as that in C3, therefore the non future consecutive is used (C4 is chained to C3).
C5 is the main clause: its verb is in the successive.
(16) C1. >Y
ndz // C2. d6 ky7
t

1 cd.aux hear
place water.7 cd-aux
C3. >7// C4. nt
dz6 //
k6
w // C5. zm6
y7
[+f]c.reach.ipf there
3 s.dry.op her.dm
[+f]c.murmur.ipf
1 cd.go
[f]c.aux
As soon as she hears water murmuring somewhere, even before she
(the hill partridge) gets there, it (the water) has evaporated.
. For the tense-mood concordance in verb chains, it is the last verb that counts: here therefore it is drcreep and not the auxiliary t. This is in keeping with the analysis proposed in
2.2.2. which suggests that verb chains stem from clause chains.
Jacqueline Leroy
7. Specific cases
In paragraphs 7.1. to 7.3. I will present some particular cases in clause chains. The clauses
are simply juxtaposed and the tense-mood concordance rules apply. The verb |ts| surpass has become specialized for expressing comparison, |>| give for expressing beneficiaries and |>Y
| do for factitivity. In 7.4. and 7.5. I will present particular cases
of complex sentences with finality clauses. Indeed, the clauses are once again simply
juxtaposed, but the verb in the second clause is always in the consecutive [+F]C, while
the verb in the first is usually (and even obligatorily) in an indicative construction, present or past. The verb |ts| exist, be has become specialized in the expression of ability,
possibility, and the verb |b| be in the hypothetical (imaginary) domain. The expression of the anteriority relation presented in 7.6. shares characteristics both with clause
chains and complex sentences with finality clauses. Once again, the clauses are simply
juxtaposed, the specialized verb |bf| be well/good in clause C2 agrees with the
verb in the preceding clause (C1) following the tense-mood concordance rules,28 but
the verb in the following clause (C3) is always in the future consecutive or exhortative.
7.1 Comparison
Comparison between two terms is effected by chaining two clauses:
The first (C1) can be any type of clause.

The verb in the second clause (C2) is |ts| surpass. The subject of |ts| is obligatorily
the same as that of the verb in C1, |ts| is conjugated in the consecutive, either future
([+F]C) or non future ([F]C), following the tense-mood concordance rules.
(17) C1. m m f // C2. ()tsa

fr6

I fut work
[+f]c.surpass.op Fru.dm

Ill work more than Fru.
(18) C1. m f // C2. ntsa
fr6

I p0.work
[f]c.surpass.op fru.dm

I worked more than Fru
(19) C1. nvw nm >m
bf //

fall
of rainy season p0.be good
C2. ntsa
nalm6

[f]c.surpass.op that of dry season.dm
Falling in rainy season is better than falling in dry season.
[It does not hurt as much]
. This formulation simplifies the facts which will be presented in depth in 7.6.
7.2 Expressing the beneficiary

The (semantic) role of beneficiary29 is most often expressed, on the syntactic level, by
a chain of two clauses:
The first clause (C1) can be of any type.

The verb in the second clause (C2) is |>| give. The subject of |>| is obligatorily
the same as that of C1. Therefore it is conjugated in the consecutive, either future
or non future, following the tense-mood concordance rules. The noun which
refers to the beneficiary, governed by the locative phrase |m
b`| to (literally:
to hands) or by the preposition n with, is the complement of |>|.
(20) C1. m f // C2. :

n w7r6 >

I p0.work
[f]c.give with him dm

I worked for him. (lit: I worked, I gave to him.)
(21) C1. m m f // C2. >
mb zY
6

I fut work
[+f]c.give to
him.dm

Ill work for him. (lit.: Ill work, Ill give to him.)
(22) (dzf6 ffmY
:m s n mf gw //)
giant
Fomingom 1 s.say with child the
C1. :6
fr
ndzm //

that that.1 exh.turn back his
C2. :
n zY
6

[f]c.give with him.dm

(Giant Fomingom, he said to the child) to turn his back on him.
(23) C1. s
n km n n //

[f]c.take.down basket the this
C2. :
mb m >

[f]c.give to
me dm

and take down this basket for me!
. It is a syntactic criterion which justifies the distinction made between the semantic
roles of beneficiary and recipient: in fact, the verb |>| give is never used to express the
semantic role of recipient, while the reverse is not true.
Jacqueline Leroy
7.3 Factitive constructions

Factitive constructions are also made up of a two clause chain:
The first clause only contains the verb |>Y

| make, freely conjugated, and possibly
also contains its subject (as specified earlier, when a verb is in the consecutive, its
subject is not expressed).
Since the subject of the second clause is different from that of the first, the second clauses verb is in the successive or exhortative, in accordance with the tensemood concordance rules. In Example24, the verb in C1 is in the relative perfective
present, the verb in C2 is therefore in the successive:
(24) (kb n6 zm za // ) C1. mb >

Yn //
money 1 be thing rel
that.7
p0.make.r
w
ntsm
C2. drn6
>
person everyone
1 s.be happy.ipf dm
(Money is something) that makes everyone happy.
In clause C1 in Example 25 below, the verb |>Y

| do is conjugated in the future
consecutive (its use in the [+F]C after the verb |ts| exist, be serves to express ability
(cf. 7.4.)), the verb in C2 is therefore in the exhortative. Then it is in the non future
consecutive in C5, and thus the verb in C6 is in the successive:
(25) (tsff
ko
s //) C1. :Y zY
ts//
partridge bush s.say
that she p0.be
>
Y6 //
C2. mb6 l//
[+f]c.make.that
rain exh.come
(tsff
nt s //) C3. :Y zY
ts //
partridge hill
s.say
that she p0.be
>
Y6 //
C4. lm6
t // [] ( tstn6 //
[+f]c.make.that
dry.season exh.shine
p0.be.ca
ntst //) C5. :Y
//
C6. lm6
t
[f]c.be
[f]c.make
dry season s.shine.dm
(The bush partridge says) she can make rain fall; (the hill partridge says) she
can make the dry season shine [].(After some time,) she (the hill partridge)
made the dry season shine.
7.4 Expressing ability/possibility

Ability is expressed through two clauses: a main clause C1 and a subordinate clause C2:
The first, main, clause (C1), contains only the verb |ts| exist, be, stay freely
conjugated,30 preceded by its subject.
The second clause (C2) is subordinate to C1. As its subject is the same as that of
|ts|, its verb is in the consecutive, but always future ([+F]C):
(26) (w7r6) C1. mb tsn // C2. kwbt

ky7r
he
that.1 p0.be.r
[+f]c.change appearance his
(He (the chameleon) who can change his appearance
(27) C1.
ts // C2. t7r6
lw7
ntf6

you p0.be
[+f]c.judge bitterness throat his

You can judge of his bitterness
(28) (w7r6 k //
:Y
) C1. zY
ts // l
she.1 s.swear that
she p0.be [+f]c.ever
k
ly // C2. ky7r6
ts77 ky7rn
[+f]c.neg sleep
[+f]c.stayup just stayup
(She swore that) she could not sleep anymore but just kept awake
Remark 7: This last example illustrates (1) that the verb |ts| exist, be cannot be
negated when it serves to express ability. It is ly sleep which is negated; (2) that in
this same use, it is not conjugated in the future but rather an auxiliary is used in the
following clause (here |l| ever) to situate the process in the future.
7.5 The hypothetical (imaginary domain)

To express a hypothetical fact belonging to the imaginary domain, Mankon has
recourse to a syntactic construction which is actually quite similar to how ability/possibility are expressed.
Once again, the construction is made up of two clauses:31
The first only contains the verb |b| be and its subject. The verb |b| is only
conjugated in the indicative, effective aspect. |b| may be preceded by a temporal
auxiliary, |>Y
|, |l| or |ts|; in this case it is the auxiliary which is conjugated in the
indicative, effective aspect, and |b| is conjugated in the non future consecutive
([F]C).
. There are restrictions however, it is only conjugated in the perfect and cannot be negated.
. In the following lines, we will present the expression of hypothetical facts, and not the relation between the subordinate hypothetical group (made up of two clauses) and the main clause.
Jacqueline Leroy
The subject of the second clause, which describes the hypothetical process, is
always the same as that of |b| be. Its verb is therefore in the consecutive, but
always future ([+F]C):
(29) C1. m b// C2. b

> //

I p0.be
[+f]c.be you
t7
>6 kmrn)
p0.stay.op me.loc Cameroon
(m m
then I
If I were you, (I would stay in Cameroon.)
(30) C1. m lo
mb // C2. w:t6
m

I p0.aux [f]c.be
[+f]c.kill.op child

>n // (m mk m m mmb
my then arrows the nm two
m lw77t6 >)
6 p0.carry off you.dm
If I had killed my child, (the two arrows, they would have carried you off.)
(31) C1. m b // C2. ts

l mbye nmY
6 //

I p0.be
[+f]c.be foc near mother

(m k m ndze
>6 z6 )
then neg I see.ipf.op me this.dm
If I were near my mother, (I would not go through this).
7.6 Temporal clauses: Relations of anteriority

A relation of anteriority between two processes is marked through three clauses:
The first clause (C1) belongs to any type.

The second (C2) is made up of the verb |bf| (which probably stems from the verb
|bf| be well/good), and in some cases its subject, which always refers to the
same entity as the subject of C1. The status of the second clause depends on that
of the first, as I will explain below.
The third clause (C3) is subordinate to C2: its verb is always in the future consecutive if its subject is the same as in C2 (see Examples32, 33, 35 and 36) and in the
exhortative if there is a change in subject (cf. 34); in the latter case, the conjunction
is naturally present.
(32) C1. m m z7> nd // C2. ()bf //

I fut sweep house
[+f]c.be well
C3. f6

[+f]c.go out.dm

Ill sweep the house before going out.
(33) C1. m z7>

nd // C2. mbf //
C3. f6

I p0.sweep house
[f]c.be well
[+f]c.go out.dm

I swept the house before going out.
(34) C1. \m tsmff ts y //

animals other
the s.come
C2. mbf6 //
C3. w7r6 k
:
Y6

[f]c.be well.that
he.1 exh.weave 9.do

bv6 \
nn
harm against chameleon.dm
The other animals arrived before he had spun an ill deed against the
chameleon.
Sometimes speakers use the auxiliary |t| then in C3:

(35) mfs32 ndz7 // \:Y
zY
z7>
nd
1 s.must [f]c.see that
he p0.sweep house his
C1. nt
ky [] // C2.

[f]c.draw water
mbf //
[f]c.be well
C3. ()t
>e
y7
f

[+f]c.then [+f]c.go.op him.loc field.dm
he made sure he had swept his house, drawn water [] before he then
went to the field.
In Examples3235, the clause C2 is chained to C1. Its verb |bf| is therefore in the
consecutive: future [+F]C in 32 since the C1 verb is in the indicative future, and non
future [F]C in 3335 following the tense-mood concordance rules.
But in Example36, the first clause C1 is hypothetical. C2 is therefore the main
clause governing C1 and its verb |bf| is of course in the successive, even though its
subject is identical to that of C1 (cf. 6):
(36) C1. b :Y
zY
>Y

nd
ndz7 //

if
he cd.aux [f]c.ever [f]c.see
C2. zY
bf //
C3. nf //
C4.

he s.be.well
[+f]c.lie
ly
f ly
[+f]c.sleep sleep.dm
If ever he sees them, then he will lie down and sleep. (i.e.: He must see
them before he can lie down and sleep).
. This is a borrowing from the English must.
Jacqueline Leroy
8. Conclusion
In this article I have shown how certain verb forms in Mankon, in this case the nonfuture and future consecutive, the successive and the exhortative, serve to structure
certain complex sentences on the syntactic and semantic levels.
A survey carried out in 1978 on clause chains in other Ngemba languages which are
not immediately intercomprehensible namely Bafut, Mbili and Mandankwe revealed
properties almost entirely identical to those found in Mankon: (1) a distinction between
a non future consecutive and a future consecutive having, on a structural level, the same
tonal characteristics. On the segmental level however, while Bafut does indeed have a
formative N- in the [F]C and a formative - (or perhaps a vowel that I have not been
able to make out),33 in the [+F]C, in Mbili the formative is N- in both the [F]C and the
[+F]C (N- however not being found elsewhere than before stops), and in Mandankwe
- (or once again a vowel I have not been able to make out) both in the [F]C and in the
[+F]C; (2) a second common property is a distinction between successive and exhortative, having the same tonal characteristics as in Mankon; and the presence, at least in
Bafut, of the conjunction or t, obligatory when the verb is in the exhortative.
In Dchang, a language of the Bamileke group coordinate with Ngemba within
Mbam-Nkam, L. Hyman (1980) clearly identified, in the verbal chains, two forms of
consecutive: a [F]C with, on the segmental level, the formative N-, and a [+F]C with
the formative e-. Judging by Hymans examples, the structural tones of the [F]C and
[+F]C are the same as in Mankon. Lastly, in Fefe, another Bamileke language, using
a verb in the past tense in the initial clause of a chain, L. Hyman (1971) was able to
bring to light verb forms corresponding to the [F]C (Hyman speaks of coordinate
consecutivization) with the formative N-, the other in the [+F]C with finality meaning since it is used after verbs in the past (subjunctive consecutivization) with the
formative -. Similarly, when the subject in C2 is different from that of C1, there are
two distinct constructions, but only in that in the exhortative (subjunctive consecutivization) one finds the conjunction .
In other groups genetically coordinate with Mbam-Nkam within the Grassfields
Bantu languages, similar phenomena have been noted. Languages distinguish, within
clause chains, clauses whose subject is the same or different from that of the preceding clause. In all of the cases documented, identity in subjects is manifested in the
same way, through the use of a non-personal verb form, i.e. without subject or class
marking: cf. in Babungo of the Ring group (W. Schaub, 1985: 8891 and 232233),
. The difficulty raised by the Mbam-Nkam languages is that, without profound knowledge
of their segmental morphology and morphotonology, one risks bypassing certain relevant
characteristics, if one has not determined and used adequate contexts.
in Aghem, also of the Ring group (S. Anderson, 1979:112117), in Mundani of the
Momo group which, like Mankon, distinguishes between a [F]C (with the formative
N-) and a [+F]C (with the vowel formative e-) (E. Parker 1991).
Symbols, abbreviations and conventions

| |
[ ]
//

`
^

`^

1, 2, 3, etc.
[+f]c
[f]c
am
aux
ca
cd
dm
exh
structural form
phonetic form
clause boundary
high tone
low tone
falling tone
rising tone
rising-falling tone
tonal upstep affecting all
following high tones
tonal downstep affecting
all following high (or
rather non-low) tones
class numbers designating
noun prefixes, agreement
prefixes and subject
pronouns
future consecutive
non future consecutive
anaphoric marker |l|
auxiliary
completed aspect
conditional
discourse marker |a| or |>e|
exhortative
foc
fut
h
hl, lh, etc.
imp
ipf
l
loc
neg
nm
np
op
p0
p1
p2
pas
prt
r
rel
s, suc
s1, s2, etc.
v
focalizer
future
high tone
contour tone
imperative
imperfective
low tone
locative preposition ||
or locative agreement
element
negation
Numeral morpheme
noun prefix
object preposition ||
perfective present
close perfective past
far perfective past
imperfective past
imperfective present
relative (mood)
relative morpheme
successive
pronoun or subject
marker paradigms
verb stem
References
Anderson, Stephen. 1979. Verb structure. In Aghem Grammatical Structure [Southern California Occasional Papers in Linguistics (SCOPIL) 7, Larry M. Hyman (ed.), 73136.
Hyman, Larry. M. 1971. Consecutivization in Fefe. Journal of African Languages 10: 2943.
Hyman, Larry. M. 1980. Relative time reference in the Bamileke tense system. Studies in African
Linguistics 11(2): 227237.
Jacqueline Leroy
Leroy, Jacqueline. 1977. Morphologie et classes nominales en mankon (Cameroun) [BS 6162].
Paris: SELAF.
Leroy, Jacqueline. 1979. A la recherche de tons perdus: Structure tonale du nom en ngemba.
Journal of African Languages and Linguistics 1(1),5571.
Leroy, Jacqueline. 1982. Extensions en mankon, langue bantoue des Grassfields. In Le verbe
bantou [Oralit-Documents 4], Gladys Guarisma, Gabriel Nissim & Jan Voorhoeve (eds),
125138. Paris: SELAF.
Leroy, Jacqueline. 1983. Systme locatif mankon et classes locatives proto-bantoues. The Journal
of West African Languages 13(2):91114.
Leroy, Jacqueline. 1994. La nasalit en mankon, langue bantu des Grassfields (Cameroun).
Linguistique africaine 13: 6181.
Leroy, Jacqueline. 2007. Le mankon, langue Bantou des Grassfields (Province Nord-Ouest du
Cameroun) SELAF 437 [Coll. Langues et Cultures africaines 36]. Paris: Peeters.
Meeussen, Achille Emille. 1967. Bantu grammatical reconstructions In Africana Linguistica III
[Annales du Muse Royal de lAfrique Centrale, Sciences Humaines 61], 81121. Tervuren:
Muse Royal de lAfrique Centrale.
Parker, Elizabeth. 1991. Complex sentences and subordination in Mundani. In Tense and Aspect
in Eight Languages of Cameroon, Stephen C. Anderson & Bernard Comrie (eds), 189210.
Dallas TX: SIL.
Schaub, Willi. 1985. Babungo [Croom Helm Descriptive Grammars]. London: Croom Helm.
Clause dependency relations

in East Greenlandic Inuit*
Nicole Tersis
Centre dEtudes des Langues Indignes dAmrique,

In the dialects of Inuit, two main features of clause subordination stand out: first,
the lack of subordinating conjunctions, as subordinate clauses are indicated by
verbal morphology and synthetic devices; second, there is a structural parallelism
between several verb forms found in subordinate clauses and possessive noun
phrases. Alongside verbal morphology marking subordination, several markers
found within the verb phrase also indicate subordination. We will emphasize the
frequency and role of multiple subordinate clause-chaining in East Greenlandic
Tunumiisut discourse. Our findings show that the supposed dichotomy between
the verbal markers found in dependent sentences and those found in independent
sentences might be less rigid in oral narratives. This leads us to also consider
dependency phenomena at the level of discursive paragraphs or sentence sequences.
1. Introduction
In the dialects of the Inuit continuum,1 two features2 of subordination relations in
complex sentences stand out: first of all they are characterized by the absence of subordinating elements: subordinate clauses are indicated by markers on the verb; sec*We would like to thank Isabelle Bril (LACITO-CNRS), Michael Fortescue (University of
Copenhagen) and Francisco Queixalos (CELIA-CNRS) for their helpful comments and
insightful suggestions on an earlier version of this paper.
. The Inuit continuum spans from northern Alaska to Greenland; it belongs to the
Eskaleut (Eskimo-Aleut) family and constitutes a homogeneous set of sixteen dialects with
approximately 80000 speakers. Tunumiisut is spoken in the easternmost part of the dialect
continuum by approximately 3577 people, out of an estimated total Greenland population of
56969, according to statistics compiled in 2005.
. These traits are found throughout the Inuit dialects, from northern Alaska to Greenland
(cf. Dorais 1996; Fortescue 1984; Kaplan 2000; Lowe 1991; Mennecier 1995; Sadock 2003) as
well as in the Yupik languages belonging to the same linguistic family (de Reuse 1988; Jakobson
1995; Mithun 1996; Miyaoka 1996).
Nicole Tersis
ond, there is a certain degree of structural and formal parallelism between several verb
forms found in subordinate clauses and possessive noun phrases.
Taking examples from East Greenlandic, Tunumiisut, I will begin by briefly
describing simple and complex sentence structure. Complex sentences are defined as a
sequence of two or more clauses linked together in a dependency hierarchy.3 In consequence, clause dependency is spread out along a scale. On one end, coordinated clauses
which are weakly dependent and which are linked, or not, to the main clause by a coordinator, without any modification of the verb form; on the other end, more strongly
dependent subordinate clauses usually indicated by specific verbal markers. Alongside
verbal markers indicating subordination, several morphemes within the verb group
also mark subordination, as does the presence of a 3rd person referential element which
may or may not refer to the agent of the main clause (identity or cross-reference).
The frequent sequencing of multiple subordinate clauses in narratives and their
pragmatic role at discourse level will also be underlined; I will show that the supposed
dichotomy between verbal markers in dependent sentences and those in independent
sentences seems to be less clear cut in oral discourse, where relations between clauses
go beyond sentence level to reach paragraph level.
I will then explore ways of interpreting a central issue in Inuit morphology the
similarities between possessive noun phrases and subordinate verb forms. To conclude, I will show how, on the typological level, the polysynthetic nature of this language also applies to sentence structure, through the preferential use of synthetic
processes to mark clause subordination.
2. From simple to complex sentences

Inuit is a SOV language type. In the ergative construction (7b), the nominal agent
of the two-argument verb is marked as ergative by -p, while the patient is marked as
absolutive by a zero marker. In the so-called anti-passive construction (7c), the single
nominal argument of the one-argument verb is in the absolutive and the patient is
marked by an oblique instrumental case -mi(k), which tends to change into an object
marker in less conservative dialects such as East Greenlandic. Pragmatics and definiteness are the main factors which determine the choice of one or the other structure: in
7b the topic is niqiq the meat and in 7c the topic is qimmiq the dog. The change in
word order corresponds to pragmatic variation, in particular due to focalization or
to the messages explicitation by use of an anti-topic (Tersis & Carter-Thomas 2005).
. The term dependent sentence refers to semantically or syntactically non-autonomous

sentences. Coordination and subordination describe the nature of the dependent sentence.
However one must be aware that interpretation of the sentences structural analysis,
transitivity, word order and ergativitys true nature have been the matter of some debate
among linguists over the last ten years, because of the parallelism between transitive
sentences (7b) and possessive noun phrases (Mahieu 2009; Tersis 2004); this phenomenon can be explained from a historical perspective (cf. Section4).
Simple sentences contain a single nominal or verbal predicate. In the former case,
the predicate can be a noun, a personal pronoun, a determiner noun or a noun phrase.
In this case, word order is subject-predicate.

(1) una
qimmiq.
this.one dog
This one (is) a dog.

(2) taanna
uana.
anaph.this.one mine
This one (is) mine.
(3) atiwaqpik aikkaayuk.
school
big
The school (is) big.
(4) isiqtu-p
nunata-a
tamaat nunakkaassiaq.
Isortoq-of landscape-3sg all
stone
The Isortoq landscape (is) all stone.
In the latter case, the predicate is a verb which is always followed by a morpheme
traditionally defined as a mood marker, and where the person markers are affixed verbfinally, the whole making up a complete sentence. The nominal subject is not mandatory:
(5) nii-wu-q.
eat-ind-3sg
He is eating.
(6) nii-wa-na.
eat-ind-1sg.3sg
I am eating it.
(7) a. qimmiq nii-wu-q.

dog.abs eat-ind-3sg

The dog is eating.
b. qimmi-p niqiq
nii-wa-a.

dog-erg meat.abs eat-ind-3sg.3sg

The dog is eating the meat.
c.

qimmiq niqi-mi nii-wu-q.

dog.abs meat-ins eat-ind-3sg
The dog is eating some meat.
Nicole Tersis
(8) una
nii-wu-q.
this.one eat-ind-3sg
This one is eating.
In simple independent sentences, the verb is followed either by an indicative or an

injunctive marker. The final person marker represents the single agent participant of
one-argument verbs (7c) or the fusion of the agent and patient participants, for twoargument verbs (7b). The verbal inventory is thus divided into one-argument verbs
and two-argument verbs, although a few verbs may do both. The person markers vary
depending on the verbs valence and the pre-final marker. Personal indexes referring
to fused agents and patients show many cases of neutralization and assimilation in
Tunumiisut (see Tersis 2000).
The indicative expresses general assertions, noting a past or present fact.
(9) ukiassa-kkit
tasiitaqmiit
amiqtaniit
autumn-trans Tasiilaq.inhabitant.pl many.pl
aatisaq-taq-pu-t
aatisakka-ni.
fish-habitual-ind-3pl cod-ins.pl
In autumn, most of Tasiilaqs inhabitants usually fish for cod.
(10) nutia-kai-kka
taki-ssa-wa-tit.
woman-bad-1pl see-fut-ind-2sg.3pl
You will see my bad women.
The injunctive expresses moderate injunction.

(11) itii-nia-kkit!
look-injunct-2sg.3pl
Look at them!
The sentence with the interrogative and the answer to it (12a-b) although syntactically
independent, are of a dependent nature, since they require contextual reference when
demanding an answer, or for a request.
(12) a. ani-wi-si?

go.out-inter-2pl

Are you going out?
b. kia akkiq-pa?
tikkak.

who come-inter-3sg man

Who came? (It was) a man.
Several verbal predicates may follow one another, forming an asyndetically coordinated complex sentence. We will first mention the relatively less frequent case where
the clauses of the complex sentence are on the same level without any modification
of the verbal forms. Juxtaposed verbal predicates express an enumeration, or two
simultaneous events, whether the subject is identical or not; this differs from syndetic
coordination in that there is no coordinating element between the clauses.
(13) itisaa-ita-a
aassaat
taki-qnaq-pa-a.
recognize-ind.neg-3sg.3sg right.away see-1st time-ind-3sg.3sg
She doesnt recognize him right away, it is the first time she has seen him.
(14) tikkak qitip-pu-q
aappa-a iqinaqsuq-pu-q.
man dance-ind-3sg other-3sg sing-ind-3sg
One man danced, the other man sang.
In the second case, the verbal predicates and clauses may be coordinated by a medial
coordinator, for example aamma and, also, as well, kisiat but, qaniq because, suuq
because of or clause finally by the clitic coordinators, -tu/-ti and, -tuunniit/-tiinniit or
else, even, which display vowel harmony with the preceding vowel. Aamma has insistent
and additional meaning which the clitic -tu/-ti does not have. Some coordinators may
also link nouns (15b, 16b, 17b) aamma and, as well, also, -tu/-ti and, kisiat but, while
others, such as taa, taawa so, then (18), kiisa finally, qaniq because only coordinate
clauses.
(15) a. pisiniaqpi-qaq-pu-q aamma uqatuppi-qaq-pu-q.

shop-have-ind-3sg and
church-have-ind-3sg

There is a shop and there is a church.
b. aatisaaq ataasiq aamma timittat maqtit

fish
one
and
bird.pl two

one fish and two birds
(16) a. innaq-tanaaq-sima-wu-q
kisiat itaaa-ita-q.

be.in.bed-already-pfv-ind-3sg but sleep-ind.neg-3sg

He is already in bed but he is not sleeping.
b. tikka-a-ita-q
kisiat miqsiqtiq

man-be-ind.neg-3sg but child

He is not a man but a child.
(17) a. miqsiqtit mama-a-kkaayu-u-wa-at

child.pl be.good-find-mod-be-ind-3pl.3pl

nunat
miimiitaqtaat piiwa-a-ttu-ut-tu
dandelions.pl stalk.pl
toy-have.for-conc-3pl-and
niitassa-a-qati-i-ttu-ut-tu
food-be-together-have.for-conc-3pl-and
The children find the dandelion stalks good (indicative), they use them as toys
(concomitant) and, at the same time, they use them for food(concomitant).
b. tikkak naniq-ti

man bear-and

the man and the bear
Nicole Tersis
(18) quttuatiit
suutti-i-ttu-tit
naa-taq-pu-t
leaves.of.willow.pl first-be-conc-3r.pl grow-habitual-ind-3pl
taa kiiqna-siit tuqtiiqnat
then after-trans.3sg stonecrop.leaf.pl
maiuttuqniq-asiit
takkit-taq-pu-t.
become summer.intensive.fact.of-trans.3sg come.out-habitual-ind-3pl
It is the willow leaves which grow first, then, afterwards, the stonecrop leaves
come out, in the middle of summer.
The coordinated clauses take the same verbal mood. In example (17a), the clitic coordinator -tu/-ti links the two clauses in the concomitant mood which are dependent on
the main clause in the indicative. Although it is possible to coordinate two clauses in
the indicative, the presence of the clitic tu seems to usually be accompanied by the
concomitant mood when the agent of the main clause is coreferential with that of the
coordinated clause, as also pointed out by Fortescue (1984:120) for West Greenlandic:
isirpuq ii-llu-ni-lu /enter.ind.3sg/sit-conc-3r.sg-and/ She enters and sits down.
The dependency of the coordinate clause is thus stronger and seems close to a subordinate relation, given that it has a specific verbal marker.
When the clauses of a complex sentence are in a subordinate relation, clause
dependency is essentially marked by the verb forms.
3. Dependent clauses and dependent verb forms

Subordination is characterized by the absence of any subordinator, as in the Samoyed
branch of the Uralic family, in particular in Tundra Nenets where there are no subordinating conjunctions; subordination is expressed by subordinate non-finite forms or
infinitives and participles (M. A. Mahieu p. c.). In Siberian Yupik, of the Eskaleut family, most subordinating conjunctions are borrowed from the neighboring Chukchi
language (Vaxtin, 2000:316). This situation has led to the reduction of the synthetic
processes characteristic of the Eskaleut family (Mahieu & Tersis 2009).
Of the Tunumiisut verbal system (cf. Table1) it is the attributive, concomitant, causative (or effected) and conditional (or non-effected) markers which are most commonly
found in subordinate clauses. These markers are followed by a final person and/or number marker. Only one argument is morphologically marked on verbs in the concomitant,
the agent of one-argument verbs and the patient of two-argument ones. The other mood
markers can appear with either one-argument or two-argument verbs. In the indicative
and the attributive, the one-argument form differs from the two-argument form.
The unmarked clause order is [Subordinate/Main] for dependent time and manner
clauses (cf. 23, 27, 30, 34). The order [Main/Subordinate] is found with certain perception,
thought, and declarative verbs (cf. 20) in complement clauses or when the dependant
clause expresses a logical succession of events (cause-effect), a purpose, or a cause (cf.

29). This order is pragmatically marked, and conveys new information (26a) but can
also be caused by the subordinate clauses unwieldiness. Several verbal markers form
different subordination types: the attributive introduces complement clauses, the concomitant indicates complement and adjunct clauses (temporal, purpose, cause) with
coreferential agents, the causative and the conditional correspond to adjunct subordinate clauses (temporal, manner, cause, consequence or purpose, condition, hypothesis).
Table 1. Verbal system markers
One-argument
Independent
sentences
Indicative
Injunctive
Interrogative
wu-/+pu-
Dependent
clauses
Attributive
Concomitant
-ti-ttu-
Causative (effected)
Conditional (non-effected)
Two-argument
wa-/+pa-niaqwi-/+pii-/+ki-
a/+ka-/mi/+ki /+pa-
The consonant variations w/+p and /+k are conditioned by the surrounding
vowels or the consonant (+) preceding the verbal marker. Due to amalgamation with
the non-coreferential 3rd person marker -a, the causative has a variant in m- and the
conditional has a variant in pa-. We will illustrate the use of each of these forms with
a few examples.
3.1 Attributive
The single argument attributive -ti- expresses attribution of a specific property, a state, or
the durative. Its variant -si- is determined by the preceding consonantal context (cf. 37).
This marker is used as a participial in the 3rd person, serving to modify nouns, e.g.
tikkak nii-tiq man eating (lit. man eat-attrib.sg). This form is translated by certain
authors as a relative clause. It can also appear as semantically dependent on a preceding sentence, which I define as a form of situational dependency at paragraph level (see
19). Such discursive dependency also appears with other subordination markers (see
Section3.5). This means that it cannot appear on its own in an independent sentence, contrary to what is found in other Inuit dialects but not in West Greenlandic (which is more or less like East Greenlandic). It is frequent in discourse, where
it is distinguished from the indicative which has a more general meaning and which
may appear in independent sentences. It is also found in complement clauses, after
thought, perception, and declarative verbs (cf. 20). The verb final person marker
represents the agent. In the 3rd person, there is a distinction between same- or
Nicole Tersis
cross-reference marked by the index of coreference with the agent of the main clause, -ni
or -tit pl. (cf. 19) and by the index of non-coreference -q, -t pl (cf. 20).
(19) miqsiqtit tasiita-mut nuut-taq-pu-t
tattani
child.pl Tasiilaq-dir move-habitual-ind-3pl there
atiwaqpim-mi
nayua-qaq-ti-tit.
school-loc
home-have-attrib-3pl
The children are used to going to Tasiilaq, there they stay at the school.
(20) taamani
isima-qaq-ta-atiwaq-pu-a
kuummiit
at.that.time thought-have-habitual-however-ind-1sg Kuummiit
tasiita-miaaniit ai-tii-ti-umaaq-ti-q.
Tasiilaq-sep
be.big-compar-inch-fut-attrib-sg
At that time, I often thought that Kuummiit (village) would be
bigger than Tasiilaq.
Contrary to the one-argument attributive verb, the two-argument attributive verb

i-/+ki- cannot serve as a noun modifier, but may only be found in a context of situational dependency,4 or after thought or declarative verbs, as before. The final person
marker combines both participants, the agent and the patient. The vocalic variant
a-/+ka- is found before the 3rd person -a due to vowel harmony, as in (21). In the
second Example (22), one finds both the one-argument attributive form modifying
the noun siqqut fins and the two-argument attributive in the dependent clause.
(21) isima-qaq-pu-a
kaimi-a-a.
thought-have-ind-1sg be.content-attrib-3sg.3sg
I think he is content with him.
(22) siqqut uti-sima-ti-t
ukiiq tamaat
fin.pl pluck-pfv.attrib-pl year all
aaqqissii-niqas-sinnaa-ppu-t
kisiat
prepare.intr-passif-be.able-ind-3pl but

nii-tii-ta-ni-pput
niitassat mainiq-mi.
eat-compar-hab-attrib-1pl.3pl food.pl summer-loc
Plucked fins may be prepared all year, but we are used to eating
more food insummer.
. I have borrowed the label situational dependency from Stphane Robert (1996: 154)
from her article on the meanings of the aorist in Wolof where she describes the usage of the
aorist which never appears first in independent sentences; it is necessarily integrated via a
preceding element which specifies the situational framework within which the narrative will
then unfold in the Aorist..
In this last example, it would be possible to have the indicative in the coordinated
sentence after kisiat but, to express a general fact and not a specific property.
3.2 Concomitant
The concomitant -ttu- (the more usual term is contemporative) is used with one-argument
valence in subordinate clauses expressing actions simultaneous or contemporary with
those expressed in the main clause (23), with temporal or manner specification, or in
purpose clauses where the action participates in the main action (cf. 25). The agent
is generally the same in the main and dependent clauses. The referential third person marker refers either to the agent (cf. 24) -ni (3R.sg)/-tit (3R.pl), as a reflexive
person, or to the patient (cf. 25) u (3sg)/-ut (3pl). The most common order is for
the subordinate clause to come first. This order may be reversed for semantic reasons (expressing manner), or for focalization purposes (26a), or because the sentence
becomes too unwieldy when the three dependent clauses with the concomitant come
in succession, or lastly after a declarative verb such as say, in which case the complement clause cannot be preposed (cf. 26c).
(23) miqsiqti-i-ttu-a
tattani atiwaq-ti-wa.
child-be-conc-1sg there go.to.school-attrib-1sg
When I was a child, I went to school there.
(24) suti-ttu-ni
aattaq-tip-pu-q.
work-conc-3r.sg begin-inch-ind-3sg
He is starting to work
(25) uqni-ttu-u
ani-wu-q.
meet-conc-3sg go.out-ind-3sg
He is going out to meet him.
(26) a. [mamaa-kkaayu-u-pa-qput]
[kataatti-i-ttu-ta]
mattak.

think.good-a.lot-be-ind-1pl-3sg Greenlander-be-conc-1pl narwhal.skin
We think it is very good (indicative), we who are Greenlanders,
(concomitant) narwhal skin.
b. [aatisakkat pani-qqissaq-sima-it-sit

cod.pl
be.dry-completely-pfv-neg-attrib.pl

[uu-ttu-ut]
[taawaa aammaqqaa-ttu-ut]
cook-conc-3pl then
eat.with.the.fat-conc-3pl
[iminna tiinniit misi-kaa-qtu-ut]

nii-taq-pa-qput].
or
even dip-many-conc-3pl eat-habitual-ind-1pl.3pl
We are used to eating (indicative) cod which are not quite dry (attributive,
equivalent to a relative clause) by cooking them (concomitant) and mixing
them with fat (concomitant) or even by dipping them (concomitant) several
times (in oil).
Nicole Tersis
c.

uqaq-pu-q akki-ssa-ttu-u.
say-ind-3sg come-fut-conc-3sg
He said he was going to come.
3.3 Anteriority and effected processes

The causative (to use the term traditional in Inuit studies) or the effected marker
a-/+ka/m- in subordinate clauses refers to a past action, to an action preceding either
another action or the time of speech, or a causal-explicative relation where the subordinate clause is postposed (29). The 3rd person marker may indicate reference identical to that of the main clause agent -ni (sg)/-tit (pl) or without coreference -a/-at to
the agent of the main clause. The same is true in the conditional, for both one- and
two-argument forms.
(27) nii-a-ma
suti-wu-q.
eat-caus-1sg work-ind-3sg
After I ate, he worked.
(28) itii-a-kkit
atiwaq-pu-tit.
look.at-caus-1sg.2sg read-ind-2sg
When I looked at you, you were reading.
(29) kutusu-p mittaqpi-a aqqissii-niqaq-pu-q
Kulusuk-of airport-3sg build.intr-passive-ind-3sg
nuna-ta-a
manis-sii-naaq-m-at.
ground-part-3sg be.flat-compar-time.when-caus-3sg
The Kulusuk airport was built because the ground was flatter.
3.4 Posteriority and non-effected processes

The i-/+ki/pa- markers express non-effected actions, an action posterior to another
one, cause-effect relations, and hypotheticals. The i- form has the variant -ni- when the
preceding consonant is the uvular /q/. The affix iaq- and its variants -niaq-, Vyaq- at
the time when often precedes the conditional (cf. 3.5.).
(30) tiki-ya-ni-ma
ani-ssa-wu-q.
come-time.when-cond-1sg go.out-fut-ind-3sg
When I come, he will go out.
(31) itii-ia-ni-kku
ittuu-ti-ssa-wu-q.
look.at-time.when-cond-1sg.3sg be.bothered-inch-fut-ind-3sg
When I look at him, it will bother him.
(32) a.

kataati-i-ttu-ta
niqiut-taq-pu-ut
Greenlander-be-conc-1pl hope-habit-ind-1pl
danmaqki-mi itinniaqniq-put ini-ya-ni-ttiit

Denmark-loc education-1pl finish-time.when-cond-1pl.3sg
nuna-ttinni
uti-qqi-ssa-tu-ut.
country-loc.1pl return-again- fut-attrib-1pl
As a Greenlander (concomitant), we hope (indicative) that after having

finished our studies in Denmark (conditional), we will be able to return to
our country(attributive).
b. qanni-ya-ssa-ppat
akki-nusu-ita-na

snow-time.when-fut-cond.3sg come-want-ind.neg-1sg

If it snows, I wont come.
These examples show the dichotomy noted by scholars of Inuit, i.e. the clear distinction
between the indicative, the interrogative and the injunctive which are found in independent sentences; and the attributive, the concomitant, the causative and the conditional
which appear in subordinate clauses (see Table1).
3.5 Clause chains in oral discourse

However, in the structures of narratives, this dichotomy is not always so clear-cut, as
shown by discourse analyses (Berge 1997; Fortescue 1984 for Greenlandic, Jacobson
1995; Mithun 1996, Mather, Meade & Miyaoka 2002; Woodbury 1985 for Alaskan
Yupik). Below is an example taken from a story in Tunumiisut where it seems that
several complex sentences have no indicative marker and, consequently, no main
clause, only verbal markers which are usually reserved for subordinate clauses, since
they are pragmatically linked to the preceding utterances. For example in (33), the
complex sentence has two effected verb forms (in the causative) and one concomitant, all of which depend on a clause situated much earlier in the paragraph, which
is in the indicative.
(33) qitiiqtaa-na-mii
nanni-p assinaaqta-ni
untie-caus.3sg.3sg bear-from rope.in.skin-3sg
pii-na-mii
ani-qqi-tti-ni-ti.
remove-caus-3sg.3sg go.out-anew-conc-3r.sg-and
After having untied it, he removed the rope from the bear and went out again.
Furthermore, oral narratives clearly show that dependent sentences are highly frequent. In a given story, 69% of the sentences are dependent clauses and only 31% are
independent, whence the notion of clause chains anchored in an initial clause on the
Nicole Tersis
paragraph discourse level.5 On a pragmatic and discursive level, the role of subordinate clauses with causative verb forms is essential, as it serves to retain the listeners
attention with explicative or background information up until the announcement
of the major information contained in the main clause (Kalmar 1982); subordinate
clauses also present the various phases of an event. For example in (34), at the beginning of a story, one sees the sequence of two causatives and one concomitant before
the main clause in the indicative. In the same narrative, one counts a majority of 43%
of causative verb forms in the dependent clauses (Grove and Tersis, forthcoming).
(34) [nunaqqati-qati-ii-ppatiwa-kaiit
living-together-mutual-numerous-intensive.pl
tusaa-ya-naa-tiq-m-ata]
listen-time.when-many-inch-caus-3pl
[pianiiqsa-qti-tit]
[aatta-kaa-naaq-6 tiq-m-ata]
prepare-conc-3r.pl leave-numerous-many-inch-caus-3pl
[qananisa-kaiit kiissaaq-mi itti-p
iti-wa-ni
old.man-poor.pl alone-ins house-of inside-3sg-loc
paaqsi-ti-ssa-iq-pu-t].
keep-attrib- fut-become-ind-3pl
Whereas a lot of people from the same village were all going together, lots
of them, to listen to stories (causative), while they were getting ready
(concomitant marking simultaneity between the process and the following one)
to all leave together (causative), two unfortunate old men stayed behind (indicative)
alone to guard the house.
Moreover there are several conjunctive morphemes internal to the verbal group which
also mark clause dependency. Below are some examples.
3.6 Conjunctive morphemes

Several conjunctive morphemes frequently appear in dependent sentences, either alone
or in succession, in particular: -niaq- while, whereas, -iaq- (variant -niaq-, -(V)yaq-)
at the time when, each time that, -aaq- (variant -naaq-) when, each time that:
. M. Mithun (2008) compares two narratives, one in Yupik (same family as the Inuit continuum), the other in Navajo (Athabascan family): the Yupik narrative has 2 indicative forms
for 25 subordinate clauses, whereas the Navajo narrative presents 17 indicative forms for 17
subordinate clauses.
. There are two homophonous affixes -aaq-, the verbal derivation affix a lot and the conjunctive
morpheme at the time when. It might well be that the distinct affixes aaq- et iaq- each time that
tend to formally merge in Tunumiisut (M. Fortescue p.c.). There are numerous homophones in the
affixes given the languages small phoneme inventory (three vowels i, a, u, and eleven consonants p,
t, s, k q, m, n, , n, w, y) and the numerous cases of vowel and consonant assimilations.
(35) inuu-uya-a-ma
tattani nayunaqaq-pu-ut.
be.born-time.when-caus-1sg there live-ind-1pl
When I was born, we lived there.
(36) asimiuqaqpin-ni atta-ni
aata-aya-ssa-aa-na-mit
village-ins.pl
other-ins.pl travel-time.when-fut-when-caus-3pl.coref
kutusu-miaaniit itikuptiqi-ni
timmi-taq-pu-t.
Kulusuk-sep
helicopter-ins.pl fly-habit-ind-3pl
When they are going to travel to the other villages, they take the helicopter
coming from Kulusuk.
The conjunctive morphemes -niaq- and iaq-/-(V)yaq- also appear in other contexts however, albeit in independent sentences with a different identity. The question is how to interpret these examples: is it a case of grammaticalization within
the affix inventory? Or should one consider, on the contrary, that it is a single
affix entering different combinations, and thereby having different temporal and
modal meanings?
On this point, the morpheme +niaq- is particularly instructive. In dependent sentences, it means while, whereas, at the time when; it marks the fact that the process
covers a certain time span, and can be followed by attributives, concomitants, causatives or interrogatives:
(37) tattani puta-sima-nia-qti-ni
sitaq
there enter-pfv-time.where-conc-3r.sg outside
taqqa-a-si-tiq-siq.
shadow-be-become-inch-attrib.sg
When she went in there, outside it got dark.
This same morpheme -niaq- is also found after nouns, with the meaning hunt, as
an incorporating verb which is always suffixed (38), or after a verb, as conative verbal
modality with the meaning try, seek to (39), or as a marker of moderate injunction.
The semantic denominator common to these different uses is that of intention/conation (Tersis 2008):
+niaq-1 incorporating NV(-) verb hunt
(38) miikkattaq-niaq-pu-q.
ringed.seal-hunt-ind-3sg
He hunts ringed seal.
+niaq-2 verbal modality VV seek to, try, have the intention of
(39) suutti-i-niaq-pu-q.
first-be-try-ind-3sg
He is trying to be first.
Nicole Tersis
-niaq- 3 moderate injunctive

(40) nii-niaq-ta!
eat-inj-2pl
Lets eat!
The morpheme -iaq- /-Vyaq- has the meaning when, each time that in dependent
clauses, and is most often preposed to the conditional marker (cf. 30). It is also found
in main clauses with the meaning suddenly, at the time when (41):
(41) isi-yaq-pu-q
itaqutat nii-ti-t.
enter-time.when-ind-3sg family.pl eat-attrib-3pl
When he came in the family was eating.
The etymological form of iaq-/-Vyaq- could be linked to a Proto-Inuit inchoative morpheme: *iaq- and *riaq- set about -ing (Fortescue et al. 1994:398). The morpheme
aaq- when, each time that only appears in dependent clauses with the causative
and the concomitant. It might be related to Proto-Inuit *aa- and *-ai- whenever
(Fortescue et al. 1994:398).
One of the most striking features of certain dependent verb forms, namely the
causative and the conditional, is however their structural and formal similarity to possessive noun phrases. I will first present the data and will then draw some conclusions
from them.
4. Subordination and possessive noun phrases

The entire set of person markers for causative (effected) and one-argument conditional
(non-effected) verb forms are identical in form to possessor markers in possessive
modification involving two nouns (cf. Table2).
Possessive modifiers
(42) anaana-ma ati-a

mother-1sg name-3sg
my mothers name
Subordinate clause
(43) nii-a-ma
suti-wu-q.
eat-caus-1sg work-ind-3sg
When I was eating, he was working.
(44) nii-ya-ni-ma
suti-ssa-wu-q.
eat-time.when-cond-1sg work-fut-ind-3sg
When I eat, he will work.
One must remember that in a possessive relation between two nouns, the junctor
(genitive) -p of governs the modifier (the possessor) and the 3rd person marker -a is
suffixed to what is possessed, modified:
(45) qimmi-p suuni-a
dog-of head-3sg
the dogs head
The morphophonological analysis of the verb forms in the causative (see Table 2)
shows elements shared with nouns, such as the junctor -p and the person markers:
(46) anaana-ma
|anaana-p-a|
mother-of-1sg
of my mothers
nii-a-ma
|nii-a-p-a|
eat-caus-of-1sg
when I was eating
Synchronically, this indicates convergence between the possessive noun phrase and
the verb phrase in certain subordinate clauses. The possessor and the possessee are in
a dependency relation, just like a clause with a verb in the causative or the conditional
is in a relation of dependency with the main clause.
Table 2. The causative or effected single argument; the verb nii- eat
1.
2.
3..
Singular
Plural
nii-a-ma
|nii-a-p-a|
eat-caus-of-1sg
when I was eating
nii-a-wit
|nii-a-p-tit|
eat-caus-of-2sg
when you were eating
nii-a-tta
| nii-a-p-ta|
eat-caus-of-1pl
when we were eating
nii-a-ssi
|nii-a-p-si|
eat-caus-of-2pl
when you (pl) were eating
Coreference with the agent in the main clause

nii-a-mi
|nii-a-p-ni|
eat-caus-of-3sg
when he was eating
nii-a-mit
|nii-a-p-ni-t|
eat-caus-of-3pl
when they were eating
Non-coreference with the agent in the main clause

nii-mm-at
|nii-(a)-p-at|
eat-caus-of-3sg
when he was eating
nii-mm-ata
|nii-(a)-p-ata|
eat-caus-of-3pl
when they were eating
In the morphophonology one notes the presence of the plural form -nit instead of
-tit for the 3rd coreferential person, probably by analogy with the singular ni. Some
morphophonologically irregular forms might correspond to different evolutive stages
of the language.
Nicole Tersis
These forms are parallel to those expressing possessive relations in noun phrases,
except that the singular/plural distinction is neutralized in the 3rd person. The same is
true for the one-argument conditional.

1sg

2sg

1pl

2pl

qimmi-ma
dog-of.1sg
qimmi-wit
dog-of.2sg
qimmi-tta
dog-of.1pl
qimmi-si
dog-of.2pl
suuni-a
head-3sg
suuni-a
head-3sg
suuni-a
head-3sg
suuni-a
head-3sg
my dogs head
your dogs head
our dogs head
your dogs head
3sg/pl coreferential with the sentence agent:

qimmi-mii suuni-a
dog-of.3sg head-3sg
his/their dogs head
3sg/pl non-coreferential with the sentence agent:

qimmi-ata suuni-a
dog-of.3pl head-3sg
his/their dogs head
This symmetry has given rise to the interpretation that the verbal clause was essentially
nominal, an interpretation which has been reinforced by the existence of the same
parallelism between possessive forms for nouns (47) and two-argument verb forms
in the indicative in independent clauses (48). The entire verb system could thus be
interpreted as participial nominalized forms (wu-/-wa- assertion, ti-/i durative,
ttu- concomitance, -a- effected, -i- non-effected).
(47)

qimmi-kka
qimmi-tit
qimmi-t
qimmi-a
my dogs
your dogs
your dog
his dog
(48)

taki-wa-kka
taki-wa-tit
taki-wa-t
taki-wa-a
I see them
you see them
you see him
he sees him
However, there are a certain number of discrepancies between structures belonging

to different levels (possessive phrases and subordinate clauses). Several linguists have
been led to consider that the nominal interpretation is only valid from a historical
perspective, and that synchronically, it is a case of verbal conjugation on the one hand
and of noun phrases on the other, with no possibility of their being one and the same
phenomenon:
Yes, there is indeed a parallel, with some morphological communality, but no,
the verbal and nominal constructions can not simply be equated, synchronically
or diachronically.
(Fortescue 1995:62)
The origin of two-argument verb forms is to be found in 3rd person possessed passive
participial constructions which spread throughout the verbal system (Fortescue
1995: 67). Thus the ancestor of sentence: pinaqtu-p terianniaq taku-va-a /hunterrel/fox.abs/see-3sg.3sg/ the hunter saw the fox could be construed as the hunters
seen thing (was) the fox, no longer a nominal construction used predicatively but the
blend of such a nominal construction with normal clause syntax.
In the subordinate clause in particular, the possessive structure is not wholly symmetrical, as the verbal predicate in the main clause has no specific marker, contrary
to the possessed element in the possessive relation, which is always followed by the
non-coreferential 3rd person marker -a. To use Nicholss (1986) terms, this structure
is marked twice, whereas in the complex sentence, only the subordinate clause is
marked, thus constituting a truncated possessive relation. Furthermore, the chaining
of subordinate clauses is possible and frequent, whereas there is no such chaining of
possessive noun phrases (Woodbury 1985). Moreover, word order in the possessive
phrase is set, while it is flexible in the subordinate clauses, subject only to semantic or
pragmatic criteria (Berge 1997).
One must also note that the parallelism does not affect all of the verbal systems
forms. Nominal and verbal inflections only partially overlap. In particular, reference to
the agent and the patient word-finally, is not identical for all of the two-argument verb
forms, whether in independent or dependent clauses. The 3rd person agent (50) does
not receive the same treatment as the 1st and 2nd person agent (49), according to the
morphophonological analysis which makes it possible to account for the numerous
amalgamations and deletions in Tunumiisut (Tersis 2000). Furthermore, reference to
the agent and patient are reversed for reasons of person hierarchy. Parallelism between
possessives and verb forms in independent clauses, or even dependent ones, is not
valid when the agent is a 3rd person (50): in this case, a patient is added word finally,
this has no correspondent among the possessive forms:
Marker order: patient-agent
(49)

taki-wa-kka.
|taki-wa-t-a|
see-ind-pl.1sg
qimmi-kka
|qimmi-t-a|
dog-pl-1sg
I see them.
my dogs
Nicole Tersis
Marker order: agent-patient

(50)

taki-wa-a-a.
|taki-wa-a-0-a|
see-ind-3-sg-1sg
*qimmi-a-a
He sees me.
unobserved possessive form
Historically, K. Bergsland (1989:31) posits anteriority for the causative and conditional
dependent verb forms as compared to the independent verb forms. He reconstructs a
morpheme *m to mark dependency after one-argument causative verb forms, which
is followed by a personal pronoun, and became cliticized into a person marker (see
Mahieu 2009). This morpheme *m would then have become specialized as a genitive
marker following the noun. M. Fortescue (1995) notes that, according to this hypothesis, forms with a dependency marker would have first characterized dependent clauses,
and then would have spread to the independent forms of two-argument indicatives.
The parallelism between possessive noun phrases and dependent constructions in
complex sentences, would therefore be due to historical, formal and semantic reasons,
because of the link existing between a clauses possessor and agent: this convergence
between nominal and verbal dependency could attest to a property of this linguistic
system. This flexibility would make it possible to use, after both nouns and verbs, the
same morphemes -resulting from the fusion between a person marker and a dependence morpheme. The parallelism might have weakened and grown opaque over time.
5. Conclusion
In Inuit, subordinating verbal forms have both syntactic and pragmatic functions,
since they also link clause chains which are contextually dependent on preceding
utterances. An analysis of oral texts shows the pragmatic importance of subordination
and the frequency of several successive subordinate clauses. The distinction between
verb forms appearing in dependent and independent clauses for pragmatic reasons
should thus be minimized, and dependency phenomena should be considered at a
higher level, the level of the discursive paragraph. Similarly, the notion of situational
dependency inherent to the attributive verbal marker and to the other subordinate
verbal markers should be broadened.
Subordination marked by specific verb forms shows the clearest type of dependency.
Two types of hierarchy are distinguished in subordination: complement clauses which
are postposed to the main clause and are characterized by the attributive or concomitant
verb markers, and adverbial clauses which are preposed to the main clause-with certain
possibilities however for displacement for semantic and/or pragmatic reasons- and
which are characterized by the concomitant, causative and conditional verbal markers.
Furthermore, the various devices used to mark subordinate clauses in East Greenlandic Inuit -namely specific verb forms, affixes included in the verb phrase and coreferential person markers- highlight the compactness of complex sentences and the
importance of the synthetic means used in clause dependency relations at clause or
discourse levels. These devices are in keeping, on the typological level, with the languages polysynthetic structure. The structural homology between possessive noun
phrases and adverbial clauses with the causative and conditional verb forms contributes to reducing the number of subordination markers and to reinforcing the cohesion
of complex sentences.
A study of textual structures which would also include pauses, intonation and the
role of discourse particles would make it possible to further explore the specificity of
dependent clauses within oral Inuit discourse.
Abbreviations
| | morphophonological
analysis

amalgam
abs
absolutive
anaph anaphoric
attrib attributive
caus
causative
compar comparative
conc
concomitant
cond
conditional
dir
directive
erg
ergative
fut
future
inch
inchoative
ind
indicative
injunct injunctive
ins
inter
intrans
loc
mod
neg
nom
pfv
pl
rel
3r.sg/pl
rel
sep
sg
trans
instrumental
interrogative
intransitivizer
locative
modifier
negation
nominative
perfective
plural
relative
third reflexive person singular/plural
relative
separative
singular
translative.
References
Berge, Anna. 1997. Topic and Discourse Structure in West Greenlandic, Agreement Construction.
Berkeley CA: University of California (mimeographed).
Bergsland, Knut. 1989. Comparative aspects of Aleut syntax. Aikakauskirja /Journal de la Socit
Finno-ougrienne 82: 780.
Nicole Tersis
De Reuse, Willem J. 1988. Studies in Siberian Yupik Eskimo. Morphology and Syntax. Austin TX:
University of Texas.
Dorais, Louis-Jacques. 1996. La parole inuit. Langue, culture et socit dans lArctique nordamricain. Paris: Peeters.
Fortescue, Michael. 1984. West Greenlandic [Croom Helms Descriptive Grammars]. London:
Croom Helm.
Fortescue, Michael. 1995. The historical source and typological position of ergativity in Eskimo
languages. Etudes/Inuit/Studies 19(2): 6175.
Fortescue, Michael, Jacobson, Steven A. & Kaplan, Lawrence. 1994. Comparative Eskimo Dictionary
with Aleut Cognates. Fairbanks AK: Alaska Native Language Center.
Grove, Arnaq & Tersis, Nicole. Forthcoming. Structure and Gesture in Greenlandic Oral Tradition, Tales from Eastern Greenland. Ms.
Jacobson, Steven. 1995. A Practical Grammar of the Central Alaskan Yupik Eskimo Language.
Fairbanks AK: Alaska Native Language Center.
Kalmar, Ivan. 1982. The function of Inuktitut verb modes in narratives texts. In Tense-Aspect:
Between Semantics and Pragmatics, P. Hopper (ed.), 4564. Amsterdam: John Benjamins.
Kaplan, Lawrence. 2000. Linupiaq et les contacts linguistiques en Alaska. In Les langues
eskaloutes, Sibrie, Alaska, Canada, Groenland, Nicole Tersis & Michle Therrien (eds),
91108. Paris: CNRS Editions.
Lowe, Ronald. 1991. Les trois dialectes inuit de lArctique canadien de lOuest: Analyse descriptive
et comparative [Groupe dEtudes Inuit et Circumpolaires]. Qubec: Universit Laval.
Mahieu, Marc-Antoine. 2009. Objective conjugations in Uralic and Eskaleut: Evidence from Inuit
and Mansi. In Variations on Polysynthesis: The Eskaleut Languages [Typological Studies in
Language 86], Marc-Antoine Mahieu & Nicole Tersis (eds), 115134 Amsterdam: John
Benjamins.
Mahieu, Marc-Antoine & Tersis, Nicole (eds). 2009. Variations on Polysynthesis: The Eskaleut
Languages [Typological Studies in Language 86]. Amsterdam: Johns Benjamins.
Mather, Elsie, Meade, Marie & Miyaoka, Osahito. 2002. Survey of Yupik Grammar Revised.
Endangered Languages of the Pacific Rim A2-023. Kyoto: Nakanishi.
Mennecier, Philippe. 1995. Le tunumiisut, dialecte inuit du Groenland oriental. Description et
analyse. Paris: Klincksieck.
Mithun, Marianne (ed.). 1996. Prosody, Grammar, and Discourse in Central Alaskan Yupik [Santa
Barbara Papers in Linguistics 7]. Santa Barbara CA: University of California.
Mithun, Marianne. 2008. The extension of dependency beyond the sentence. Language 84: 69119.
Miyaoka, Osahito. 1996. Sketch of Central Alaskan Yupik, an Eskimoan language. In Handbook
of North American Indians, Vol. 17. Languages, Ives Goddard (ed.), 325363. Washington
DC: Smithsonian Institution.
Nichols, Johanna.1986. Head-marking and dependent-marking grammar, Language 62: 56119.
Robert, Stphane. 1996. Aspect zro et dpendance situationnelle: lExemple du wolof. In Dpendance et intgration syntaxique, Claude Muller (ed.), 153161. Tbingen: Max Niemeyer.
Sadock, Jerrold. 2003. A grammar of Kalaallisut (West Greenlandic Inuttut) [Languages of the
World/Materials 162]. Munich: Lincom.
Tersis, Nicole. 2000. Economie structurelle et complexit syntagmatique, Groenland oriental. In
Les langues eskaloutes. Sibrie, Alaska, Groenland, Nicole Tersis & Michle Therrien (eds),
249268. Paris: CNRS Editions.
Tersis, Nicole. 2004. De lergatif laccusatif, volution des structures argumentales en inuit
du Groenland. In Actes du Colloque Ergativit, Francesc Queixalos (ed.), 4551. Villejuif:
Centre dEtudes des Langues Indignes dAmrique (CNRS, IRD).
Tersis, Nicole. 2008. Forme et sens des mots du tunumiisut, lexique inuit du Groenland oriental,
Paris: Peeters.
Tersis, Nicole & Carter-Thomas, Shirley. 2005. Integrating syntax and pragmatics: Word order and
transitivity variations in Tunumiisut. International Journal of American Linguistics, 473500.
Vaxtin, Nicolaj B. 2000. Les diffrents types de phrases polyprdicatives du yupik sibrien central. In Les langues eskaloutes, Sibrie, Alaska, Canada, Groenland, Nicole Tersis & Michle
Therrien (eds), 303333. Paris: CNRS Editions.
Woodbury, Anthony. 1985. Noun, nominal sentence and clause in Central Alaskan Yupik Eskimo.
In Grammar Inside and Outside the Clause, Johanna Nichols & Anthony C. Woodbury (eds),
6188. Cambridge: CUP.
Coordination and subordination

ma in Bulgarian dialectal Greek*
Eleni Valma
LACITO (Laboratoire des langues et Civilisations Tradition Orale),

ma su lo na miSs e eSinik, e les a ti in aft pu

miSs en to kataSavno
When I ask you to speak Greek with us you answer: but
what language are you speaking? I dont understand it
This article studies the relator ma in dialectal Greek as spoken in Bulgaria.
The relator is polysemous and marks temporality and hypotheses, it is also
polyfunctional and serves as a subordinator and a coordinator. Our goal in this
study is to isolate its syntactic and semantic properties so as to distinguish its
functions and meanings. We will also examine the notion of opposition which
ma can also convey, following contact between dialectal Greek and Bulgarian.
1. Introduction
Speakers of Greek in Bulgaria originate from two different groups: the inhabitants
of the Black Sea coastal cities (Sozopol, Nessbre and Pomori, abbreviated as MN)
*Our gratitude goes to Petya Assenova, Isabelle Bril and Victor Friedman for having
accepted to read over this article and for their precious comments. We would also like to thank
Zlatka Guentchva for having helped construct the article. It has been carried out within the
framework of the Franco-Bulgarian project RILA (integrated action programs) entitled Les
dialectes balkaniques de Bulgarie, (the Balkan dialects of Bulgaria) directed by Zl. Guentchva
(France) and P. Assenova (Bulgaria). The examples quoted are taken from an oral corpus of
free and semi-directed conversations, recorded for the needs of the project. The examples of
connectors in Standard Greek are written in the Greek alphabet, the rest of the corpus is transcribed with the IPA. Some specific abbreviations appear in the glosses: aor (aorist), imperf
(imperfect), impfv (imperfective), perf (perfective), pro (pronoun).
Eleni Valma
on one hand and the Karakatchans [abbreviated as Kar], a former nomadic population who have been sedentary for the past 5070 years, settled at the foot of the
Stara Planina and Rila mountains (in the West) on the other hand. Although they
show Bulgarian-Greek diglossia which is now giving way in favor of a strong tendency
towards Bulgarian monolingualism,1 most of the fluent speakers may currently be characterized as belonging to a culture of secondary orality as defined by Ong [1982:11]:
I style the orality of a culture totally untouched by any knowledge of writing
or print, primary orality. It is primary by contrast with the secondary orality
of present-day high-technology culture, in which a new orality is sustained by
telephone, radio, television, and other electronic devices that depend for their
existence and functioning on writing and print.
The secondary orality which is found among the Greek speakers in Bulgaria is, in our
opinion, the result of their exposure to Standard Greek via radio and television, to
which must be added short stays in Greece, generally for economic reasons.
In this article, we propose to analyze the polysemy of the conjunction ma,
which, parallel to its temporal use, has developed hypothetical uses. This subordinating conjunction can also mark opposition and, in a marginal manner, justification. Given that opposition is reserved for coordination, one may say that the uses
of ma are susceptible of covering two different syntactic functions, subordination
andcoordination.
Numerous studies attest to the difficulty of circumscribing the distinction between
these two notions, and call upon syntactic, semantic or formal criteria to do so. Dik
[1997:189] defines them mutually. Subordination and coordination thus mark a connection between two constituents: the first notion marks a hierarchical link while the
second shows a link between entities which are functionally equivalent. But although
this distinction is seductive, coordination shows great complexity and asymmetry
between clauses, as has often been noted [Bril & Rebuschi, 2007:6]. This asymmetry
would imply logical dependence between two clauses having the same syntactic status.
So as to avoid the use of terms with strong connotations such as subordinator and
coordinator, we will opt for the notion of junctor from Foley & van Valin [1984:242]:
a junctor is susceptible of joining two elements from any strata.
The ma in Bulgarian dialectal Greek qualifies as a junctor since in the case of syntactic dependence, it introduces temporality and hypotheses, and in the case of syntactic
autonomy, it is restricted to opposition. We will describe its various syntactic contexts,
the semantic criteria making it possible to distinguish temporal uses from hypothetical uses, and the links between these two notions and that of opposition reserved only
forthe dialectal use of ma. While the semantic switch from tense to hypothesis is
. For a geopolitical and linguistic presentation of these populations, cf. Assenona, 1997;
Poromanska, 2004; Tsitsilis, 1999.
f requently found cross linguistically (cf. Traugott, 1985; Kortmann, 1996: 210), the
expression of opposition via the same junctor is far from being obvious.
2. From temporality to hypotheses

We propose to apply the syntactic tests (cf. among others Bril & Rebuschi, 2007;
Maingueneau, 1994:2627, 9899, 138139) which make it possible to determine the
status of the junctor ma in dialectal Greek. Let us take the following example:
(1) vlp
pols kzm0 mazemnos / ma kodft-ase
see.prs.3sg. many people assembled
when arrive.aor.3sg.
don
a-ne/
ma l-i
crje janko
him.pro. see. aor.3pl but say.prs.3sg. mister janako
pu
-sane
where be.aor.3pl.
He sees many people assembled. When he came closer [the villagers] saw him. But
Mister Janako, they said, where did they go [the Bulgarian priests]?
[MN]
where the various syntactic manipulations of the two clauses confirm the subordinating status of ma:
permutation between the two clauses is possible: don ane ma kodftase (En.
They saw him when he approached);
one may focalize or extract the clause introduced by ma, for example by using
the adverb mno (En. only): don ane mno ma kodftase (En. They saw him
only when he approached);
one notes the presence of a pronominal cataphora which is coreferential with a
noun in the following clause: don ane ma ecnos kodftase (En. They saw him
when that one approached).
While the application of syntactic criteria demonstrates the subordinating use of

ma, the tense-aspect meanings associated with the contents of the two clauses favor
a temporal interpretation2 of the subordinate clause. This interpretation of ma in
(1) is reinforced by the use of the aorists kodftase (En. he approached) and ane
. Standard Greek has marginalized this use in favor of the hypothetical. Although the
polysemy of in Standard Greek is not unanimously recognized among Greek grammarians and syntacticians [cf. Holton et al., 1997], numerous grammarians [cf. Tzartzanos,
1928/1996; Babiniotis & Clairis, 2001; Kriaras, 1998] still agree that it has two uses (temporal
and hypothetical). Nevertheless, current linguistic observation shows a temporal use of
in familiar speech [Mackridge, 1985/1990: 418, 9.4.4.3]. Standard Greek is not used
with oppositional meaning.
Eleni Valma
(En. they saw), in the subordinate and the main clause respectively. The aorist refers
to an event and is therefore capable of entering into a structure of a logical and chronological sequence of events. The occurrence of ma makes it possible to then mark
the anteriority of an event as compared to the others. The event meaning which we
just attributed to the aorist in (1) is in distinction to the state and the process, three
notions called upon to treat the aspect-tense questions in works such as those by
Comrie [1976/1998], Lyons [1977/1996], Descls [1980, 1990a, 1990b, 1994]). We will
use the notions here such as they were defined in Descls [1994:7176]):
States express stability in the referential situation represented, i.e. absence of
movement and change. Events indicate discontinuity in a stable referential.
The validation interval for an event is a closed interval. The process expresses
a change grasped in its internal evolution. All processes necessarily express an
initial change, i.e. an initial event which indicates the beginning of the process.
In (2), ma is well suited to the same analysis: followed by the imperfect padrvodan
(Eng. they were getting married), ma is the source for the clauses temporal interpretation. In this context, the verb form in the imperfect refers to an open class of events
and the subordinate clause serves to mark iteration.
(2) ma padrvo-dan
when marry.imperf.3pl.
kaar -tan
clean be.imperf.3sg.
mazvo-dan
assemble.imperf.3pl.
ci -vlep-an
and look.imperf.3pl.
en -tan
or not be.imperf.3sg.
[Kar]
After the wedding, the members of the two families assembled to check
whether the young bride was a virgin or not.
One may apply the same syntactic tests to (3), but this time the interpretation is
hypothetical:
(3) bor-
e> pad
bor-
na
be.able.prs.1sg. I
everywhere be.able.prs.1sg. that
z-
ma -ne
o dras
mu maz mu.
live.impfv.1sg. when be.impfv.3sg. the husband my with me
[MN]
I can, me, I can live anywhere as long as my husband is with me.
The only relevant elements which make it possible to distinguish between the two uses,
temporal and hypothetical, are the tense-aspect meanings conveyed by the subordinate
clause. One must remember that Standard Greek and certain dialectal varieties (including
the Bulgarian Greek dialects) are systems which are strongly aspect-oriented. Indeed, they
are organized around two themes, the present (in the glosses: imperfective aspect), and the
aorist (perfective aspect). Thus the future, the imperative and the modal (which certain
Modern Greek grammarians classify as a subjunctive, cf., among others, Tsopanakis, 1994)
are constructed using a verb root followed by the present declensions.
When ma is followed by an aspectual form, it is open to hypothetical interpretation. This is the case in (3), where one has the imperfective ne. While in its temporal
use, ma expresses something which has already taken place, in its hypothetical use,
the contents of the subordinate (P) are posited without it being possible to make a
direct statement as to its veracity. One therefore distinguishes two cases:
a hypothetical structure where P is a condition for Q (the main clause) being carried out; through encyclopedic knowledge, one knows that P will happen in the
future (Example4)
a hypothetical structure, often with iterative aspect, where P is open to two possibilities, that of P and that of non P (Example5)
(4) X ikoj\a prp-i

na -i
tra pej /
a family must.prs.3sg. that have.impfv.3sg. three children
na mn-i
jna ma pen-un / tes-un
that stay.perf.3sg. one when die.perf.3pl. finish.perf.3pl.
i >ones na mn-un
j toma.
the parents that stay.perf.3pl. two people
[Kar]
A family must have three children, that one remains, [and] when the parents
die, that two people remain.
(5) jerdja -i
vivlirjo
e ci ma to
old
have.prs.3sg. health.book here and if
the.pro.
pjs-I
rpi aft pijn
sto farmaco
catch.perf.3sg. flu he go.prs.3sg. to pharmacy
[Kar]
Old people here, they have a health book and if they get sick,
they go to thepharmacy.
In (4), the use of ma indicates that the speaker clearly leans towards Ps
accomplishment, for reasons mostly of a pragmatic nature (everyone is mortal).
The speaker voices affirmations on the future based on his or her experience of the
way in which things happened in his or her past (observation data, according to
Toulmin, 1958/1993:153) or through encyclopedic knowledge. In (5), the accomplishment of P is not a fatality and given the pragmatic context (old people may
fall sick) it is highly probable that P will happen. Therefore we propose to consider
ma a vericonditional junctor (for the terminology used in this article, cf. Toulmin,
1958/1993 & 1983). The notion of truth is then in reference to the speakers apprehension of the subordinates content and combines the notion of time with hypotheses. Thus ma poses a fact within which the facts explicitly mentioned in the main
Eleni Valma
clause takeplace. The accomplishment of P is presented in such an obvious manner that the speaker may take refuge behind the will of a third person, as shown
by the following example:
(6) [] prksenos na me kn-i u ce rotsa-ne
matchmaker that me make deal and ask.imperf.3pl.
ton el-is
ts e> jirns-a
ts
him want.prs.2sg. and me turn.imperf.1sg. and
-le>-a
ma ton el-i
say.imperf.1sg. if
him want.impfv.3sg.
i mna mu
tson e-o
ts e>.
the mother to-me him want.prs.1sg. and me
[MN]
[One day arrived] a matchmaker to present someone to me and I was asked:

do you want him? And I turned [towards my mother] and I said: if my mother
wants him, I want him too.
On the contrary, in (7), the two an3 (En. if) introduce an eventuality:4 a fact is presented as only possible and the speaker considers both P and non P (you might come
like you might not come/I might be home like I might not be home):
(7) kaX for an r-ete
ap makr mro tip-te
one time if come.perf.2pl. from far
place ring.prs.2pl.
to kuni / an -me
e to pti ik
sas
the doorbell if be.impfv.1sg. here the house pro.poss. to-you.
[Kar]
If one day you come [to Karlovo] from afar, ring at the door,
and if I am home, the house is yours.
ma also expresses justification, even though the scope of this interpretation is very
narrow and is essentially contextual (discourse justification). In (8)
. This connector combines with the two aspectual themes.

. Also in Standard Greek, the presence of an and the type of meta-linguistic commentary
point in this direction.
, , ; If John abandoned you, I do say if, what
would you do?
, , ; If John abandoned you,
all this is hypothetical, what would you do?
(8) e> kr-o

ta ik
m ps ti
zo -zisa-n
me know.prs.1sg. the pro.poss. my how what life live.aor.3pl.
ce v\p-o
ta eg\a
m en bor-n
na
and see.prs.1sg. the grandchildren my not be.able.prs.3pl. that
ta
pan
puen. ma
en ftni.
them bring.perf.3pl. anywhere because not suffice.prs.3sg.
[Kar]
Me, I know how my children grew up, and I can see that [my children] cannot
take my grandchildren anywhere. Because it [their salary] is not enough.
Justification5 is situated on the discourse level and serves to explain what has just been
said [also see say in Ducrot, 1984]. The syntactic criteria listed above are mostly violated because permutation between the two clauses, or focalization or extraction of the
clause introduced by ma are impossible. This justification use could serve as a bridge
from the subordinating towards the coordinating use.
3. From subordination to coordination

When ma introduces an opposition, one observes a violation of all the criteria classifying
it as a subordinator. The permutation between two clauses, P extraction and the presence
of a pronominal cataphora are rendered impossible. Only post-positioning is allowed,
which is in agreement with the notion of justification which ma occasionally expresses.
However, it would be risky to assert that the justification meaning which ma
conveys at times (8) could be the link between subordination (tense and hypotheses)
and coordination (opposition).6 It would be more reasonable to posit another hypothesis according to which ma would have appeared in dialectal Greek following contact
. The dictionary asserts that in Standard Greek, is susceptible of introducing a
causal relation. This affirmation is probably dictated by the example chosen by the authors:
it brings out a sentiment verb, which probably led them to attribute a causal interpretation
[on causal constructions which call on sentiment verbs, cf. Valma, 2004:209210, 213217].
Although the justification is directly linked to causality, the term causality is inappropriate
because this junctor does not mark the link between two situations where one of them (Sit1)
is the cause and the other (Sit2) the produced effect [cf. Nedjalkov & Silnickij, 1969, quoted in
Descls & Guentchva, 1997].
. The idea of such a switch may be defended in Standard Greek when one uses the opposition coordinator (En. but), in contexts where the clause introduced by is a condition for the accomplishment of P (in the structure P opposition Q). We will not enter into
the details here. Let us further remark that our dialectal Greek corpus does not give us the
means for such an affirmation.
Eleni Valma
with Greek and Bulgarian. Indeed, am is found in Bulgarian [Feuillet, 1996:318] and
expresses the notion of opposition.7 In its interphrasal function of opposition coordinator, am is placed between the two sentences or the two elements which are opposed
[Vrinat-Paskov, 1990:106, v.1].8 Its coordinating function would be specific to the Balkan
languages and linked to Turkish influence (where one also finds the form ama). The result
is that ma is a junctor which passed from Turkish into dialectal Greek via Bulgarian.
One observes that the structure P ma Q, where P and Q are two contrary (i.e.
antonyms) or contradictory entities, presupposes that P can serve as an argument for
a certain conclusion R and that Q is an argument which annuls this conclusion. From
this perspective, P and Q are not necessarily opposites but are opposed through the
argumentation,9 thus becoming two elements contradictory in R.
The junctor punctually marks a contrast which could qualify as lexical. Thus in
(9) and (10), P and Q are opposed in a very particular context where ma necessitates
reference to a sociopolitical situation which the speaker and hearer share:
(9) vul>ara / -dan
vul>ara
Bulgaria be.imperf.3sg. Bulgaria
ma -e
elinik skola.
but have.imperf.3sg. Greek schools
[MN]
It was Bulgaria, but there were Greek schools.

(10) sti
>ala ma linas -is?
in.the France but Greek have.prs.2sg.
You live in France and your husband is Greek?
[MN]
In effect, nothing opposes, in principle, Greek and Bulgarian schools. But in (9), the
opposition marks a gap between the logical relation (in Bulgaria, the schools are Bulgarian) and the current situation (even though it is Bulgaria, one finds Greek schools).
In (10) what the speaker expresses is surprise/incomprehension when faced with the
paradox live in France and be married to a Greek which calls for an explanation.
. In dialectal Greek, the stress is on the first syllabe (ma).

. The author [1990:106] recognizes a second, expressive function for am. It would thus
be placed before its incidence base and would most often come at the beginning of a reply.
. Argumentation theory conceived of as new rhetoric covering a whole area of discourse
aiming to persuade or convince, independently of the audience addressed and the subject matter.
[] The argument is then founded on influence, interaction, speech acts, and not necessarily on
truth. [Valma, 2004:22, 23; translated by MD] Argumentation is thus the domain of disagreement,
conflict and contradiction. The latter attests, like argumentation, to a discursive polyphony.
For the Karakatchans, ma lexically opposes two terms, for example young and
oldin (11):
(11) esis kortsja -saste
pj na
prp-i
na
you girls
be.prs.2pl. more young must.prs.3sg. that
-ete javsi tis istores ma ci e pu -o pi
solo
read.prf.2pl. the histories but and I
who go.prf.1sg. school
o
tin istora tus karakatsnus
have.prs.1sg. the history the Karakatchans
[Kar]
You, girls, are very young and you must have studied history,
but I too went to school and know the history of the Karakatchans.
Furthermore, one knows [Ducrot, 1980; Cadiot, 1976] that the notion of opposition
can manifest itself on the morphological, lexical and discourse levels. Opposition can
appear e.g. between two conclusions or between two clause contents which serve as
arguments for the same conclusion [Moeschler & de Spengler, 1982: 10]. Dialectal
Greek seems to have just a single linguistic possibility for expressing both strategies
(one or two conclusions). We will therefore distinguish between these various ma
having different semantics.
Contrary to (9) and (10) where two facts are confronted, in (12) and (13) the clause
introduced by ma furnishes additional information and helps further the debate:
(12) maz
p>-ame
ma tra e> e pijn-o
together go.aor.1pl. but now me not go.prs.1sg.
jat
pras-an
because pass.aor.3pl
sta r\a.
in years
[MN]
We used to travel together but now I no longer go on vacation [with them]
because I am too old [lit. The years have passed].
(13) e> ttes
-mane
ekaept ron e
me at.the.time be.imperf.1sg. seventeen years not
-el-e
na me aravo\s-i
mikr / ma -a
want.imperf.3sg. that me betroth.perf.3s. young but have.imperf.1sg.
brba [] tse l-i
aft
-ne
kal
uncle
and say.prs.3sg. that.one be.prs.3sg. good
pSikri /
prtik-a.
young.man take.aor.1sg.
[MN]
Me, at the time, I was seventeen years old and [my mother] didnt want me
to be betrothed young. But I had an uncle who said this young man is a good
person. [Therefore] We got married.
Eleni Valma
Therefore there are two ma, one of them rectifies what was posited by P and the other
argues in favour of or against P. The data collected among the Karakatchans seems to
confirm this analysis. Example (14) shows a logical opposition with discourse based
on proof whereas in (15) the relation is given as argumentative:
(14) pr-ane
ikpea/ ma Seft pu / ta prvata
take.aor.3pl. land
but money where the sheep
e plros-an / Seft e ma -os-an.
not pay.aor.3pl. money not to.us give.aor.3pl.
[Kar]
They nationalized (lit. They took) the land. But we did not get any money.
We didnt get any for the sheep either. We were not given any money.
(15) e> e cit-o
ta pej
tra/
me here take.care.of.prs.1sg. the children now
ma skoSi i zo -n /
i daki
but difficult the life be.prs.3sg. the retirement
-ne
mikr e tn-i.
be.prs.3sg. small not suffice.prs.3sg.
[Kar]
Me, I take care of children now. But life is difficult.

My pension is not sufficient.
One must note that despite preponderance in the use of ma, other processes are also
used to express opposition. These means are different in the two groups, the Karakatchans and the inhabitants of the North Sea coast. The Karakatchans, never having
lived in Greece, use the conjunction a (from Slavic)10 to express objection, often with
a connotation of reproof and irony.11 In the case of mechanisms for linking argument
sentence chains, a is linked with non-verbal content and marks the opposition of the
speaker towards a certain type of behavior: in (16), the speaker signals to her grandson
who, intimidated, does not come into the room to greet the guests. In (17), a furnishes
further specification and in (18) it replaces a speech act:
(16) aft/ a drp-ete /
an
Sa
he
but be ashamed.prs.3sg. Thanassis come.imp.2sg.
a>r m/ aft a drp-ete
ap sas/ katlav-e.
[Kar]
boy my he but be ashamed.prs.3sg. of you understand.aor.2sg.
. According to Vrinat-Paskov [1990:112, v.1] a can rectify the contents of what precedes
it or introduce a new idea which comes to the mind of the speaker.
. On the other uses, cf. Vrinat-Paskov [1990:169, v.1]
It is him [my grandson] but he is shy [that is why he doesnt want to say
hello to you]. Come my child. It is him, but you make him shy, you understand.
(17) tsarja lo-dan
ecna fors-an
tsarouks call.imperf.3pl. those wear.imperf.3pl.
ta tsarja a e ta
-a
e
the tsarouks but not the.pro. see.aor.1sg. i
s-ti ici
mu zo e ta
-e.
to.the pro.poss. my life not the.pro. have.imperf.3sg.
[Kar]
lit. They [the shoes] were called tsarouks, they used to wear those,
the tsarouks, but I didnt see them during my life.
The used to wear tsarouks, but before my time.
(18) i me>Ss bor-s
na p-is
e
the big
be.able.prs.2sg. that say.perf.2sg. not
j-nete
ce e ftn
en -ne /
be possible.prs.3sg. and not be sufficient.prs.3sg. not be.prs.3sg.
a sto mikr pos na to p-s?
but to little how that it say.perf.2sg.
[Kar]
You can tell adults that you dont have enough money [to buy something].
But how can you admit it to a child?
On the other hand, the inhabitants of the North Sea coast alternate between ma and
aS (En. but): the first is reserved for argumentative opposition (founded on an argument) or demonstrative opposition (founded on proof), and the second is reserved
for specific lexical contrasts, as in (19), or for opposite terms which belong to a same
notional field, as in (20). aS, very common in Standard Greek, comes from Ancient
Greek [Andriotis, 1951/1995].
(19) ekat
ies jenm-aste/
tso poS/
hundred thousand become.prs.1pl. so numerous
poS/
aS -ne
mno ton v>usto poS kzmo.
numerous but be.prs.3sg. only the August many people
[MN]
We get up to 100,000 inhabitants [over the summer]. So numerous. Numerous. But there are a lot of people only in August.
(20) kataSavn-une
aS e miS-ne.
understand.prs.3pl. but not speak.prs.3pl.
[My children] understand [Greek] but they dont speak [it].
[MN]
But aS is also used when the speaker wishes to rectify information by choosing certain elements of a notion and certain specific aspects of these elements, as in (21).
Eleni Valma
(21) kras -ame

artpuSa mas -in-an
aS S>o.
meat have.imperf.1pl. coupons to.us give.imperf.3pl. but few
We had meat, we were given coupons, but not very many.
[MN]
One also observes the albeit rare presence of ma, frequently used in Standard
Greek; it stems from the Italian (ma) and is attested to in Medieval Greek epic literature [Andriotis, 1951/1995].12 It indicates a discursive chain as in (22) where it heads
the answer and does not introduce an explicit Q. Generally speaking, ma introduces
a series of counter-arguments without questioning the propositional contents of the
hearers discourse.
(22) non1: e kna
j mres r-ete.

to almost two days come.prs.3sg.
non2: vvea / tso kod -saste.

Of.course so close be.prs.2pl.
non3: ma -li
e> na majirv-o /

but want.prs.3sg. me that cook.impfv.1sg.

e>
me
na
that
ti
it
skups-o. [MN]
sweep.perf.1sg.
Speaker1: [My daughter] comes [to take care of me] every two days.
Speaker2: Of course you live close to one another.
Speaker3: But she wants me to cook for her, to sweep [her house] [].
Its discursive character is also underlined by its capacity to combine with the discursive connector af (En. since). The latter serves to justify an assertion [Valma,
2004: 140] and introduces a sentence which takes its argumentative strength from
what has already been said:
(23) me rot-san
pu
ta
ksr-is
ta elinik
me ask.imperf.3pl. where them.pro. know.prs.2sg. the Greeks
ma af -is
sptja jat r-es.
but since have.prs.2sg. houses why come.aor.2sg.
[MN]
I was asked how I had learned Greek and why I was there [in Greece]
since I had a house and a family [in Bulgaria].
One should also note that in the case of ma, the contents of the two clauses are not
necessarily in opposition, but are opposed as compared to an argumentative movement. Furthermore they can belong to the same notion, as in (24):
. Another yet unproved hypothesis is that ma is the truncated form of am.
(24) -kam-es
ta cfja su ma -is
to rma.
do.imperf.2sg. the wishes your but have.prs.2sg. the money
You did what you wanted, but you had the [financial] means to do so.
[Kar]
4. By way of conclusion

The junctor ma is polyfunctional and polysemous in introducing tense and hypotheses. It
is both subordinating and coordinating. It is a subordinator when it allows permutation
between two clauses, and coreference using an anaphoric pronoun and extraction of the
clause introduced by ma. The link between the two clauses is of a logical, chronological
and conditional nature. The distinction between temporal and hypothetical meanings
is based on the tense-aspect meanings present in the two clauses. The foreseeable
evolution of ma into a hypothetical junctor calls for some specification: it is not
hypothetical on the same level as an (En. if ) because by using ma, the speaker poses
a fact and personally guarantees its accomplishment. Whereas with an, P is simply
posed, without being endorsed by the speaker.
Parallel to this subordinating use with two semantic interpretations (temporality
and hypothesis), ma also has a coordinating function expressing opposition. To analyze ma, we therefore have the choice between two classic solutions:
either one separates the uses of ma into two homonyms disjoined in their meaning and distribution;
or one seeks a semantic invariable common to all its uses (temporal, hypothetical,
oppositional) by correlating them with the contextual/lexical constraints which
distinguish them.
The polysemy hypothesis entails that one pinpoints a common abstract property and a single meaning which includes all the contextually differentiated uses.
But our corpus does not allow us to make the connection between these three
notions, nor to propose a semantic invariable. We must therefore conclude that
they are homonyms.
In the structure P ma Q, focus is solely on whether the speaker takes responsibility for the arguments or not [also see Plantin, 1990:43]. It is interesting to compare
this use of ma with the other processes for expressing opposition; this imposes various conclusions.
Thus if the temporal and hypothetic use of ma is common both to the Karakatchans and to the inhabitants of the North Sea coast, the oppositional use which
ma has developed in the two linguistic communities is due to language contact (usage
which would have spread through the Balkans following contact with Turkish). But
Eleni Valma
these two communities differ in their use of ma. For the Karakatchans, ma coexists
with a (En. but), another borrowing from Bulgarian. By contrast, the inhabitants of
the North Sea coast have opted for aS, a borrowing from Greek. As we saw above, in
the structures P a Q and P aS Q, the clause Q belongs to a different utterance than
P. Moreover, a and aS show no unity on the content level.
Lastly, a final remark on the type of orality (primary or secondary) susceptible of
dividing the speakers into two groups:13 in the first, one finds speakers with secondary
orality, who use ma and aS to distinguish between the different types of opposition
(lexical or semantic). In the second group, one finds speakers whose orality is close to
primary orality: for them, ma still contains all sorts of oppositions, notional or discursive (argumentation and/or demonstration).
References
Andriotis, Nikolaos. 1951/1995. . :
.
Assenova, Petya. 1997. Bulgarian Greek. In Kontaktlinguistik/Contact Linguistics/Linguistique
de contact, Hans Goebl, Peter H. Nelde, Zdenek Stary & Wolfgang Wlck (eds), 15101514.
Berlin: Walter de Gruyter.
Babiniotis, Georgios. 1998. . A: .
Babiniotis, Georgios & Clairis, Christos. 1998 pour I, 1999 pour II et 2001 pour III,
, . : .
Bril, Isabelle & Rebuschi, Georges. 2007. Coordination, subordination et co-jonction: Faits linguistiques et concepts. In Faits de langues 28, Isabelle Bril & Georges Rebuschi (eds), 518.
Paris: Ophrys.
Cadiot, Pierre et al. 1979. Oui mais non mais ou il y a dialogue et dialogue. In: Langue Franaise
42: 94102. Paris: Larousse.
Comrie, Bernard. 1976/1998. Aspect. Cambridge: CUP.
Descls, Jean-Pierre. 1980. Construction formelle de la catgorie de laspect (essai). In Notion
daspect, J. David & R.Martin (eds), 198237. Paris: Klicksieck.
Descls, Jean-Pierre. 1990a. State, event, process and typology. In General Linguistics, Vol.29(3):
159200. Philadelphia PA: Pennsylvania State University Press.
Descls, Jean-Pierre. 1990b. Langages applicatifs, langues naturelles et cognition. Paris: Herms.
Descls, Jean-Pierre. 1994. Quelques concepts relatifs au temps et laspect pour lanalyse des
textes. In Studia kognitywne 1, 5788.Warszawa: SOW.
Descls, Jean-Pierre & Guentcheva, Zlatka. 1997. Causalit, causativit, transitivit smantique.
In Typology of Verbal Categories, Leonid Kulikov & Heinz Vater (eds), 727. Tbingen/
Niemeyer: Linguistische Arbeiten.
. One must be cautious when dividing speakers into groups as they tend to mix and to
borrow from each other.
Descls, Jean-Pierre & Guentcheva, Zlatka. 2000. La notion dabduction et le verbe devoir pistmique. In Cahiers Chronos 8, Patrick Dendale & Johan van der Auwera (eds.) 103122.
Amsterdam: Rodopi.
Dik Simon. 1997. The Theory of Functional Grammar. New York NY: Mouton de Gruyter.
Ducrot, Oswald. 1980. Les chelles argumentatives. Paris: Les ditions de Minuit.
Ducrot, Oswald. 1984. Le Dire et le Dit. Paris: Editions de Minuit.
Feuillet, Jack. 1996, Grammaire synchronique du bulgare. Paris: Institut dtudes slaves.
Foley William & Van Valin Jr., Robert D. 1984. Functional Syntax and Universal Grammar. Cambridge: CUP.
Holton, David, Mackridge, Peter & Philippaki-Warburton, Irini. 1997. Greek: A Comprehensive
Grammar of the Modern Langage. London: Routledge.
Kortmann, Bernd. 1996. Adverbial Subordination: A Typology and History of Adverbial Subordinators
Based on European Languages. Berlin: Mouton de Gruyter.
Kriaras, Emmanouil. 1998. . : .
Lyons John. 1977/1996. Semantics (v.2). Cambridge: CUP.
Mackridge, Peter. 1985/1990. The Modern Greek Language. Oxford: OUP.
Maingueneau, Dominique. 1994. Syntaxe du franais. Paris: Hachette.
Moeschler, Jacques & de Spengler, Nina. 1982. La concession et la rfutation interdite
approches argumentative et conversationnelle. Cahiers de linguistique franaise 4: 737.
Nedjalkov Vladimir & Silnickij, Georges. 1969. In Tipologija kauzativnyx konstrukcij,
A. A. Xolodovi (ed.), 519. Leningrad: Nauka.
Ong, Walter. 1982. Orality and Literacy The Technologizing of the Word. London: Methuen & Co.
Plantin, Chistian. 1990. Essais sur largumentation. Paris: Kim.
Poromanska, Stoyna. 2004. . Tipologija kauzativnyx
konstrukcij. In Studies in Greek Linguistics Proceedings of the 13th Annual Meeting of the
Department of Linguistics, 560567. Thessaloniki: Aristotle University of Thessaloniki, Faculty
of Philosophy.
Toulmin Stephen. 1958/1993. The Uses of Argument. Cambridge: CUP.
Toulmin Stephen.1983. Logic and criticism of argument. In The Rhetoric of Western Thought,
James L. Golden, Goodwin F. Berquist &. William E. Coleman (eds), 91401. Dubuque IA:
Kendall-Hunt.
Traugott, Elizabeth Closs. 1985. Conditional markers. In Iconicity in Syntax [Typological Studies
in Language 6], John Haiman (ed.), 289307. Amsterdam: John Benjamins.
Tsitsilis Christos. 1999. s. In Dialect
Enclaves of the Greek Language, 7375. Athens: Ministry of National Education and
religious Affairs.
Tsopanakis, Agapitos. 1994. . : .
Tzartzanos, Achilleas. 1928/1996. (v.2). : .
(dictionnaire du Greek moderne) []. 1998. :
.
Valma, Eleni. 2004. Lexpression de la causalit en franais et en Greek moderne tude contrastive. Thse non publie, Universit Paris VII Denis Diderot, Paris.
Vrinat-Paskov, Marie. 1990. Les particules expressives du bulgare moderne (v.2). Thse non publie,
INALCO, Paris.
Author index
A
Abdulaeva 105106, 124,
126, 133
Adamou, E. 5, 12, 14, 399,
401402, 404, 407, 410
Akatsuka, N. 391
Anderson, S. R. 402,
404405, 579
Andrews, A. 27
Andriotis, N. 613614
Asher, N. 496
Assenova, P. 403, 603
Authier, G. 79, 143144, 149
Ayoub, G. 393394
B
Babiniotis, G. 605
Bally, C. 405
Bandelt 84
Banys, W. 380
Basset, A. 395
Bentolila, F. 361, 378, 384, 386
Berge, A. 591, 597
Bergsland, K. 598
Berthoud, A.-C. 357
Bickel, B. 35, 7, 1012, 5156,
60, 63, 6566, 69, 7172,
75, 78, 80, 86, 98, 107,
274, 306
Bisang, Walter 107, 171, 175,
200201, 225
Blakemore, D. 219, 241,
452, 466
Blhdorn, H. 427428
Bresnan, J. 28, 30, 33
Breu, W. 402
Brewis, R. 279280
Bril, I. 12, 5, 1112, 14, 17, 106,
143, 204, 219, 269, 272,
279285, 298, 303, 342,
346, 350, 355, 392, 399, 421,
451, 490, 511512, 550, 581,
603605
Bryant 84
Bhler, K. 386
Bring, D. 462
C
Cadiot, P. 611
Carter-Thomas, S. 582
Caubet, D. 335, 337338,
342, 344
Chafe, W. 69
Chaker, S. 360
Chase, A. 193, 452
Chomsky, N. 69
Church, E. 207208, 246, 250,
283, 403, 414, 475
Clairis, C. 605
Clark 72
Cohen, D. 335, 356
Cohen, M. 335
Colin, G. S. 335
Comrie, B. 2, 106, 209, 222,
390, 392, 502, 531, 606
Corston-Oliver, S. H. 295297
Corts, C. 5, 12, 14, 421
Creissels, D. 78, 105, 111, 113,
115, 159, 550
Cristofaro, S. 2, 54, 67, 193,
233, 519
Croft, W. 3, 5455, 284
Culioli, A. 306, 393394,
400, 411
Curme, G. 425, 431
Cysouw, M. 51, 84
Cyxun, G. A. 403, 409
D
Daladier, A. 421, 446
De Reuse, W. J. 581
De Smet, H. 477
De Spengler, N. 611
Descls, J.-P. 386, 423,
606, 609
Diessels, H. 362, 364, 389, 394
Dik, S. 2, 204, 275, 306, 514,
516, 604
Dimitrova-Vulchanova, M. 403
Dixon, R.M.W. 11, 207,
253254, 283, 306
Dorais, L.-J. 581
Dress 7071, 84, 406
Drettas, G. 401, 404
Ducrot, O. 429430, 433434,
440, 609, 611
E
Eisenberg 425427
Elson, M.J. 403
Emeneau, M.B. 167, 178
Evans, N. 3, 9, 132, 144,
149, 162, 451452, 467,
519, 542
Ezard, B. 298302
F
Fabricius-Hansen, C. 221, 425
Farr, C. J. M. 44, 53, 66, 100,
304305
Fedden, S. 43
Ferguson, C. A. 420
Feuillet, J. 404, 610
Fielder, G. E. 410
Foley, W. A. 212, 15,
2729, 36, 4041, 52,
54, 60, 6566, 6970,
9192, 171, 199, 206, 221,
239240, 256, 273, 430,
511, 604
Fontinelle (de la), J. 281
Forker 58
Fortescue, M. 581, 586,
591592, 594, 597598
Fourquet, J. 422424, 426428,
445446
Franois, A. 6, 1112, 14,
17, 269, 297, 376, 451,
504507, 510514, 518,
523, 527, 532, 534
Franklin, K. 40
Friedman, V. 402, 404, 603
Author index
G
Glbov, I. 403
Galand, L. 355356,
360361, 365, 367368,
370, 377378, 384385,
388, 395
Genetti, C. 30, 53, 101
Givn, T. 206, 221, 229230,
235, 240, 271, 534
Good, J. 51, 58, 64, 70, 75,
77, 8283, 90, 121, 128,
150, 158, 160161,
183, 248, 251, 255, 289,
299301, 315, 321, 327,
330, 344, 416, 446, 459,
483, 572, 576, 611
Green, G.M. 54, 224
Greenbaum, S. 332
Greenberg, J. H. 390
Grove, A. 592
Guentcheva, Z. 386, 410, 414
H
Hagge, C. 357
Haiman, J. 2, 11, 36, 53,
65, 67, 69, 100, 204,
206, 229, 240, 275, 284,
291292, 305, 331, 359,
390391
Hale, K. 53, 452, 455
Halliday, M. 477
Harrel, R. S. 335
Hasan, R. 477
Haspelmath, M. 23, 6, 29,
54, 74, 93, 106107, 136,
145, 147148, 166, 169170,
204, 206, 212, 215216, 225,
229231, 233234, 240,
242, 249, 502
Hausser 83, 87
Heine, B. 506
Hengeveld, K. 459
Hepner, M. 43
Hoijer, H. 28
Holton, D. 605
Horn 59
Huson 84
Hyman, L. M. 578
J
Jacobs, J. 75, 462
Jacobson, S. 591
Jasperson, R. 332
Johannessen, J. B. 2, 5, 204,
219, 221, 229
Johanson 58, 63, 101, 169,
171, 181
K
Kabasanov, S. 410
Kalinina, E. 143144, 162
Kalmar, I. 592
Kanevska-Nikolova, E. 402,
404, 414
Kaplan, L. 581
Kasatkina, R. 402
Kaufman 82
Kazenin, K.I. 115
Keenan, E. L. 2, 402, 404405
Kibrik, A. 142
King, J. K. 153, 172, 188, 221,
233, 245, 247, 254, 280
Kiss, K. . 273
Kokkas, N. 402, 407, 413
Koneski 403
Knig, E. 2, 6, 29, 7778, 107,
166, 169, 220, 231, 341
Kortmann, B. 411, 605
Kossmann, M. 355,
363364, 379
Kouloughli, D. 381
Kriaras, E. 605
Kwon, N. 242
L
Lafkioui, M. 382
Lakoff 11, 54, 89, 99
Lambrecht, K. 270271,
273275, 306, 339340,
348, 533534
Landberg, C. de 335336
LaPolla, R. J. 2, 78, 206, 221,
240, 462
Larcher, P. 391
Lascarides,A. 496
Lazard, G. 54, 93
Leech, G. 332
Leguil, A. 342, 490
Lehmann, C. 55, 86, 93, 234
Leroy, J. 6, 11, 18, 550553
Levinsohn, S. H. 279280
Lichtenberk, F. 259, 303, 461
Longacre, R. E. 2, 178,
206, 216, 221, 228229,
244, 248, 256, 259,
305, 509
Lowe, R. 581
Lynch 100, 275, 278,
295, 297
Lyons, C. 403
Lyons, J. 606
M
MacDonald, L. 30, 34,
36, 41, 53, 57, 7071,
73, 101
Mackridge, P. 605
Madouni, J. 335, 337338, 342
Maechler 83
Magomedbekova, Z. M.
106, 109
Magomedova, P. 105106,
124, 126, 133
Mahieu, M.-A. 583, 586, 598
Maingueneau, D. 605
Marchello-Nizia, C. 404405
Marchese 69, 100
Margolin 87
Masica 169
Mather, E. 591
Matras, Y. 409
Matthews, P. 332
McGregor, W. 453, 463
Meade, M. 591
Meeussen, A. E. 559
Mennecier, P. 581
Mettouchi, A. 360, 382,
384, 393
Meyer 83, 8889
Mileti 403
Miller, J. & C. 280
Minova-Gurkova, L. 404
Mirev, K. 403
Mithun, M. 213, 272, 581,
591592
Miyaoka, O. 581, 591
Mladenov, M. 404
Mladenova, O. 403404,
407, 409
Moeschler, J. 611
Molochieva 51, 58, 64, 70,
74, 99
Moulton 84
Moyse-Faurie, C. 281
Muller, C. 387, 390391, 446
Munro, P. 39
N
Nat-Zerrad, K. 373, 379

Nedjalkov, V. 8, 106107,
166, 169170, 198200,
231, 235, 609
Newman, J. 456
Nichols, J. 56, 209, 597
Nikolaeva, I. 144, 233
Noonan, M. 2, 51, 61,
99, 207, 232, 254,
501502, 518
Nordlinger, R. 37, 455
O
Olson, M. 1, 28, 52, 511
Ong, W. 604
Opgenort 67, 101
OzanneRivierre, F. 281
P
Parker, E. 579
Payne, J. R. 204, 215,
219, 238, 248249
Peled, Y. 379, 390
Philippaki-Warburton, I. 617
Pilhofer 100
Plantin, C. 615
Polinsky, M. 115, 151, 242
Polotsky, H. J. 209211, 226,
228, 237
Poromanska, S. 604
Potsdam, E. 151
Prasse, K.-G. 355, 360
Q
Quirk, R. 313
R
Ramat, P. 387
Rappaport 101
Rebuschi, G. 2, 5, 204, 219,
604605
Redder 443
Reed 101
Reesink, G. 37, 53, 58, 69,
76, 101
Reilly, J. 396, 420
Rigsby, B. 451452
Author index
Robert, S. 4, 6, 1012, 1416,
51, 201, 269, 333, 339342,
376, 383384, 394395,
399400, 411, 451,
465466, 469475,
477478, 482, 489491,
496, 588
Roberts, J. 43, 52, 57, 60, 63, 65,
68, 71, 76, 98, 240, 242
Rogo, A. 402
Ross, M. 3, 69, 99, 275279,
285294, 299
Rousseau 426, 446
Rousseeuw 82
S
Sadock, J. 3, 5, 240, 581
Sall, A. O. 469470, 480481,
483484
aur 403
Schackow 59, 75
Schaub, W. 578
Schneuker, C. 46
Scott, G. 32, 47, 63, 99
Shay, E. 3, 331332
Shrestha 53
Silnickij, G. 609
Simeone-Senelle, M.-Cl. 395
Sobolev, A. N. 410
Sonnenberg 425
Stassen, L. 67, 89, 170, 213, 215,
229, 352, 510
Strimmer 83, 87
Subrahmanyam 169,
173174, 180
Sumbatova, N. 144, 162
Suter, E. 47, 53, 100
Svane, G. 403
Svartvik, J. 332
T
Tafi, M. 359
Taine-Cheikh, C. 5, 11,
1314, 355357, 363, 365,
369, 381
Takahashi 54
Ter Meulen, A. 232
Tersis, N. 4, 6, 11, 18, 581584,

586, 592593, 597
Tesnire, L. 422
Theoharidis, P. 402, 413
Thompson, S. 2, 178, 189190,
199, 248, 256, 259, 305,
331, 509
Tikkanen, B. 5253, 57, 59, 75,
99, 169, 200
Topolinjska, Z. 404
Toulmin, S. 606
Traugott, E. C. 393, 415, 605
Tsitsilis, C. 604
Tsopanakis, A. 607
Tzartzanos, A. 605
V
Vaillant, A. 404
Valma, E. 11, 19, 603,
609610, 614
van der Auwera, J. 2, 77, 107,
517, 523
Van Valin, R. D. 2, 4, 6, 2729,
4041, 5152, 54, 60,
6970, 78, 91, 206, 221,
239240, 256, 269274,
282, 430, 462, 604
Vanhove, M. 5, 11, 1314,
333335, 347, 390
Vaxtin, N. B. 586
Veleva-Bojadieva 403
Verstraete, J.-C. 6, 1013, 15,
17, 285, 338, 342, 451, 453,
457, 459, 467, 477, 490
Vrinat-Paskov, M. 610, 612
Vulchanov, V. 403
Vycichl, W. 360, 388
W
Woodbury, A. 591, 597
Y
Ylikoski, J. 107
Z
Zeileis 88
Zifonun 425, 434
Ziga, F. 107
Language index
A
Aghem 579
Aghul 143, 145
Aji 281
Akhvakh 68, 105113, 115,
133, 137139
Alaskan Yupik 591
Albanian 403
Algerian Arabic 335
Amele 43, 5255, 57, 60, 6263,
65, 68, 71, 76, 79, 82,
8586, 98
Arabic 6, 13, 159, 203, 207208,
331, 333338, 340342,
344347, 350, 352, 356,
359, 370, 373, 379, 381,
390391, 393394
Archi 143, 145
Aromanian 403
Austronesian 38, 269,
279280, 305, 392
Azerbaijani 143, 148, 159
B
Babungo 578
Badaga 7, 9, 165167, 169173,
175176, 183, 188, 190, 195,
197201
Bafut 578
Bagvalal 105, 115
Bamileke 578
Bantu 18, 52, 549, 552, 578
Bargam 43
Belhare 5255, 57, 6061,
63, 6566, 68, 7174,
7679, 82, 8586,
8990, 92, 98, 274
Berber 1314, 342, 355365,
367370, 376379,
381384, 386388,
392395, 490
Bislama 510, 514
Budugh 7, 9, 143149,
159, 162
Bulgarian 400403, 407,

409410, 414415,
603606, 610, 616
Burushaski 169, 200
C
Cmuh 281
Chamito-Semtic 355
Chantyal 61, 82, 99
Chechen 5152, 58, 6364, 70,
72, 7475, 77, 99
Chukchi language 586
Classical Arabic 381, 390391,
393394
Common Bantu 552
Coptic Egyptian 6, 10, 203,
207208, 210, 222, 225
D
Daghestanian 89, 58,
105107, 109, 115, 138,
143145, 151, 163
Dchang 578
Dialectal Greek 19, 603605,
609611
Drehu 281
Dyirbal 306
E
English 30, 34, 39, 45, 48, 54,
57, 7172, 77, 84, 89, 92,
99, 107, 109, 114, 128, 131,
134, 153, 176, 193, 204, 220,
277, 313, 320322, 342, 350,
357, 386, 388, 391, 462,
513514, 523, 535, 538, 577
Eskaleut family (Eskimo-Aleut
family) 586
F
FeFe 578
Fijian 283
Fore 3233, 36, 4748, 63,
85, 99
French 153, 285, 298, 322, 336,

339, 342, 357, 368, 387, 391,
474, 490
Fula 325, 331
Fwi 281
G
German 14, 8485, 99100,
322, 388, 421426,
428430, 433, 436, 443,
445446
Godi 84, 100
Grassfields Bantu 18, 549, 578
Greek 6, 19, 61, 100, 204,
208, 213, 248250, 258,
401402, 405, 408410,
417418, 435, 603611,
613614, 616
H
H
assniyya Arabic 359
Hausa 224, 325, 331
Hiw 1718, 499546
Hua 53, 6567, 85, 100,
284285, 291, 305
I
Iaai 281
Inuit dialects 581, 587
Iraqw 213
J
Jawe 281
K
Kabyle 373, 379, 382,
385, 393
Kadazan 280
Kalaallisut (West
Greenlandic) 586587
Kte 4647, 53, 72, 100
Kaulong 278, 307
Kewa 40, 48
Khinalug 105, 145
Language index
Kiranti 12, 53, 59, 67, 85,
274, 306
Korafe 12, 4445, 53, 66, 100,
303305
Kryz 143, 145, 148149
L
Latin 65, 72, 233, 305306, 387,
518, 525
Lenakel 100
Lezgian 107, 143, 145, 148
Lezgic 143, 145, 154, 159
Lo-Toga 17, 499546
M
Macedonian 400404, 407,
409, 415
Maghribi Arabic 337,
340342, 344
Manam 272, 303
Mandankwe 578
Mankon 18, 549, 552555,
557, 562, 564, 575,
578579
Mbam-Nkam 549, 578
Mbili 578
Meglenoromanian 403
Mianmin 43
Modern German 14, 421422,
428429, 446
Modern South Arabic 395
Momo 579
Moroccan Arabic 335, 337
Mundani 579
Mwotlap 297, 376, 504, 507,
510, 527, 532
N
Navajo 592
Nlmwa 272, 276, 281, 284,
307, 350
Nemi 281
Nepali 5355, 72, 75, 84, 100
Newar 101
Newari 30
Ngandi 542
Nyelyu 281
O
Oceanic languages 14, 17, 342,
350, 490
Old Church Slavonic 403, 414
P
Polish 320
Pomak 6, 14, 399402,
405406, 409410, 415,
417418
Proto-Inuit 594
R
Ring 578579
Romanian 403
Roviana 295, 297
Russian 51, 101, 105107,
320, 322
Rutul 143, 145
S
Samoyed 586
Semitic 208, 334, 356
Serbian 401, 403
Shilha 361, 367, 378, 388
Siberian Yupik 586
Sobei 275, 307
Spanish 154
Standard Greek 603606,
608609, 613614
Suena 4445
Swahili 52, 5455, 57, 85, 101
T
Tabassaran 143, 145
Takia 272, 276, 282, 286,
298, 307
Tamazight 378, 384, 386387
Tauya 7, 30, 3637, 4142, 47,
5355, 57, 7073, 85, 101
Tawala 272, 297299
Timugon Murut 279
Tok Pisin 514

Tombunuo 272, 280
Tonkawa 28
Touareg 365, 377, 385
Toura 101
Trans New Guinea family 36
Tsakhur 115, 143, 145
Tsez 115
Tundra Nenets 586
Tunumiisut
(East Greenlandic) 1819,
581582, 584, 586, 591592,
597, 599
Turkish 19, 58, 63, 85, 101, 392,
401402, 405, 409, 415,
418, 610, 615
U
Udi 143, 145
Umpithamu 15, 17, 338, 342,
451467, 490
Uralic family 586
Usan 7, 3739, 41, 53, 58, 69,
76, 101
W
Wambule 62, 6667, 85, 101
Wandala 12, 313314, 317
Warlpiri 455
Watam 4243
Wolof 10, 14, 16, 18, 201,
340342, 344, 376, 383,
394, 411, 469496, 588
Y
Yafi Arabic 13, 333, 341
Yemeni Arabic 335336
Yimas 31, 34, 36, 48
Yupik 101, 592
Z
Zenaga 6, 13, 18, 355361,
363365, 368370,
375382, 384385,
387395
Topic index
A
ability 106, 112113, 148,
184, 288, 395, 427, 572,
574575
action verbs 471, 474476,
482, 484485, 495
addressee 111, 219, 239, 299,
304, 335, 340341, 344345,
351, 353, 399400, 402,
405408, 410411, 418, 514,
517, 534, 536
adjunction 4, 133, 259
adverbial (clause, phrase) 2,
812, 19, 27, 3234, 38, 57,
72, 7778, 84, 105107,
135139, 165166, 168170,
176177, 180, 182185,
191195, 197200, 205, 212,
219, 225226, 230233,
240, 243, 247, 251,
256260, 273, 278, 285,
288, 295298, 300, 302,
305, 313, 318319, 326327,
331, 334, 407, 410, 415, 423,
455, 469, 509, 521, 529,
598599
Aktionsart 209, 439, 486
anaphora 5, 16, 80, 271,
291292, 294, 306, 361,
384, 482483, 491492,
495, 544
anteriority 177, 179, 188, 236,
303, 400, 411, 414415, 572,
576, 590, 606
anti-passive construction 582
anticipatory subject
marking 67
aorist 11, 1319, 269, 308,
355, 369389, 391395,
470, 472, 478, 504505,
507508, 512, 515518, 588,
603, 606
apodosis 10, 17, 251, 279, 314,
318, 327329, 379380,
390392, 443, 487488,

492, 494, 514, 522
apprehensive 452, 456457,
461462, 504
argumentative 15, 17, 430, 432,
434, 440441, 443445,
465, 477, 483484, 494,
612614
aspect 6, 912, 14, 16, 19,
31, 45, 52, 111, 138, 144,
146147, 167, 170, 179181,
185, 188189, 199200,
205207, 209211,
220, 222, 225, 227, 229,
234, 240, 246, 257, 261,
286287, 298, 316, 334335,
343, 355, 370, 383384, 394,
418, 423, 439, 469470,
499, 501502, 504505,
510512, 514, 530546, 558,
560, 564565, 575, 579,
605607, 610, 615616
assertion 45, 1213, 1518,
30, 4748, 54, 61, 199, 237,
239, 254, 269273, 275,
288289, 294, 297, 306,
335, 339340, 346, 350,
389, 394, 440, 469, 475,
479484, 488, 490492,
494496, 533544, 546,
596, 614
assertive 46, 52, 66, 111, 162,
294, 334, 337, 348, 383,
387, 391, 537, 555
asymmetry 166, 171, 216, 219,
304, 544, 604
auxiliary 48, 110111, 114,
117, 131132, 137, 169,
174, 179, 199, 211, 235,
260, 334, 364, 367368,
372, 375, 378, 390391,
413, 418, 456, 481, 523,
528, 554, 561, 564568,
571, 575, 577, 579
B
background 47, 69, 207,
285, 383, 403, 427, 432,
440441, 443, 452,
488, 592
background perfect 1718,
499, 502505, 509, 512,
530546
backward anaphora 80
balanced clause linkage 67
bound and unbound
subordinate clauses 427
C
case assignment 78
case marking 12, 107, 109, 148,
295297, 305, 454455
case stacking 37
cataphora 212, 242, 301,
387391, 389391, 395, 422,
427, 605, 609
causative 1819, 48, 108, 139,
144, 146, 149, 163, 193, 254,
260, 308, 353, 528529,
553, 586587, 590592,
594595, 598599
cause (clause) 5, 10, 13, 1617,
19, 76, 88, 99, 101, 125,
165166, 173, 175, 177,
179, 186, 189191, 195,
197200, 219220,
248, 256, 271, 273, 275,
278, 282, 285287, 292,
298, 302303, 305307,
333334, 342345, 349,
351353, 372, 410, 416417,
438439, 443444, 451,
455, 459460, 466,
487490, 495, 556557,
587, 590, 609
center-embedding 78
chaining (clause), chain 111,
1419, 27, 2930, 40,
4244, 4648, 5153,
Topic index
5879, 8285, 9293,
98103, 159, 162, 166, 171,
173, 175, 181, 196, 198, 203,
206207, 228230, 233,
235, 243248, 250253,
256, 260, 285286, 292,
305, 307, 340, 342, 370371,
384, 389, 439, 444,
469496, 549, 554, 556,
561574, 578, 591, 597, 614
clause backgrounding 519522,
530546
clause linkage 12, 58, 1516,
27, 51, 5560, 62, 6573,
75, 78, 8082, 8485, 90,
9293, 203, 205206,
222, 227, 234, 333, 335,
338, 351, 449, 451452,
454, 456459, 461462,
464467, 476, 495, 511
clause order 12, 229, 291, 306,
390, 586
cleft construction 273, 339,
540542
co-ranking 45, 206207, 228,
247, 260
co-subordination 12, 115, 147,
285, 333
comment 280, 318, 320,
325326, 328, 357,
367, 381382, 390391,
395, 434
comment clause 1213, 194,
212, 284, 313315, 317320,
322, 326327, 329331
commitment 338, 443, 475,
478479, 491492, 495
comparison 40, 108, 192, 228,
303, 352, 408, 410, 482,
550, 572
complement (clause) 2, 45,
8, 1011, 1314, 1619,
3234, 3940, 56, 59, 80,
106, 135, 151, 153154, 158,
175, 183, 207, 209, 232, 240,
253256, 273, 275, 277,
281, 284286, 288289,
291292, 295, 298302,
305307, 314, 318, 321327,
333334, 341, 344, 350351,
355356, 382, 391, 396,
426, 455, 471472, 474,
490491, 543544
complementizer 14, 138139,

223, 225, 234, 259260,
350, 352353, 376, 389,
394, 520, 525
completive (clause) 271, 280,
286287, 334, 376, 389,
395, 556557, 568569
complex sentence 15, 165, 171,
181, 229, 243, 252, 260,
277, 330331, 421422,
424, 428432, 434435,
437440, 443446, 480,
491, 549, 555, 558, 561, 567,
572, 578, 580, 582, 584, 586,
591, 597
compounding 511, 526529
concessive clause 191, 220,
390, 441
concessive converb 63, 65, 113,
133134, 139
concomitance 17, 188, 305, 492,
495, 596
conditional 9, 1314, 16, 1819,
33, 52, 56, 69, 7172, 105,
107, 113, 118, 121, 124,
130133, 139, 163, 177178,
180, 187, 189, 191, 197,
219220, 224, 250251,
260, 275, 278, 285286,
295, 298299, 302303,
305, 314, 318, 327, 329, 355,
358360, 371, 379380,
384, 390395, 400, 410,
414416, 451, 456460,
465467, 504, 506508,
514515, 522524, 529,
536, 558561, 570571,
579, 586587, 590591,
594596, 598599, 615
conditional converb 9, 113, 118,
121, 124, 130133, 139, 163,
177, 180, 191, 197
conjugation 6, 1011, 1719,
207, 210, 222, 225, 227,
234, 237, 242, 250,
340341, 343344, 372, 393,
469497, 549, 554, 557561
conjunct scope 52, 53, 56, 63,
65, 70, 82
conjunction, conjunctive 5,
17, 29, 40, 57, 6869,
76, 203204, 212213,
215216, 230231, 246,
248249, 259, 272,

287, 294, 298299, 331,
360, 368, 387, 392, 409,
429, 433, 456, 502, 508,
523525, 540, 549, 555558,
561562, 568570, 576, 578,
592593, 604, 612
connective function 424,
428, 435
consecutio temporum 65
consecutive (clause) 12, 14,
1618, 236, 243, 259,
272, 285, 287, 290, 292,
298299, 302303, 305,
342, 373, 383, 387389,
479, 491492, 549, 554,
556564, 566579
consequence 1011, 14,
139, 217, 219, 222, 237,
251, 278, 282, 290, 336,
373, 378, 394, 481, 487,
490, 587
contrast, contradiction 35,
1214, 1617, 19, 2829,
34, 36, 4142, 6061, 66,
68, 71, 7578, 108109, 112,
131, 148, 204, 211, 221, 226,
228, 236, 240, 243, 246,
249, 256, 269273, 275,
277279, 280, 282283,
287, 291, 294, 297299,
301, 304308, 320, 331,
338339, 345, 348, 356,
363, 367, 393394, 401,
407, 426, 439, 441443,
446, 452453, 457458,
460, 462464, 465466,
472, 474, 481, 484485,
487, 492493, 495496,
604, 610, 616
contrastive focus 221, 287,
320, 331, 339, 345, 535536,
540544
contrastive topic 13, 277,
279, 282, 308, 338, 393,
462, 464
control 80, 93, 116, 444,
446, 517
controlling verb 10, 153, 206,
227, 229, 234236, 240,
246, 250, 253, 257, 260
converb 810, 2930, 48,
5152, 5859, 61, 6366,

69, 75, 77, 85, 106107,
111119, 121139, 146, 159,
163, 165166, 168173,
175181, 185, 187188, 191,
194201, 222223, 225,
231233, 235
coordination 16, 8, 1011,
2730, 32, 4041, 44, 48,
5152, 57, 66, 69, 8486,
93, 107, 109, 114115, 171,
201, 203207, 212221,
226230, 232, 234235,
238243, 248250,
252253, 257, 259, 277,
281, 285287, 295, 298,
302305, 307, 342, 370371,
510512, 582, 585586,
603604, 609
copula 13, 109111, 114, 116117,
130, 136, 139, 146, 163, 222,
260, 308, 334335, 337,
340, 345, 347348, 351353,
355, 363364, 368, 381, 385,
393394, 472, 497, 505,
539, 542
core 27, 32, 36, 67, 83, 107, 118,
135, 143, 167, 182, 184, 209,
223, 240, 253, 255, 273, 295,
301, 306, 430, 511
coreferential person 595, 599
correlation 88, 106, 112, 144,
207, 225, 229, 233
correlative (markers) 1, 5, 12,
14, 249, 272, 279281, 285,
292, 294, 298, 306, 379,
387, 421422, 424431,
433437, 439, 441, 444446
cosubordination 6, 8, 2730,
5155, 84, 93, 239240, 430
counter-expectation 330
cross-clausal reflexivization 80
D
dative 108, 116, 119, 137, 139,
150, 154, 156, 163, 184186,
190, 193, 331, 401, 418, 431,
452, 459
debitive 9, 149, 154156, 163
deictic, deixis 1214, 37, 44,
222223, 226, 270271,
275277, 285, 289290,
299, 304, 306, 308, 317,
334336, 345, 347, 351353,
Topic index
355, 359366, 364365,
368, 381, 383, 385389,
393395, 399400,
402411, 413414, 417,
422, 424427, 451, 472
demarcative function 424, 438
demonstrative 13, 139, 260,
283, 293, 299302,
318320, 328, 331, 346, 391,
406407, 409, 543, 613
dependency (clause) 1, 4, 6, 9,
1719, 28, 40, 57, 82, 144,
165166, 171, 198199, 201,
221, 246, 269, 272, 308,
318, 333, 342, 352, 359, 368,
373, 379, 383, 392, 422, 430,
469, 480, 485, 491, 493,
495496, 499, 501, 503,
507, 510526, 529530,
534538, 545546,
581582, 586588, 592,
595, 598599
dependent verb forms 112113,
143, 147, 586
deranked linkage 67
derivation 146, 163, 181, 183,
205, 314, 526, 528529, 553
determiner 31, 37, 355356,
364, 381, 446
detrimental 17, 478, 494
discordance 484, 494495
discourse 4, 6, 11, 1314, 1619,
121, 162, 167, 181, 194195,
200, 203204, 211212,
219220, 222, 226, 229,
237, 242244, 247, 250,
258, 260, 269271, 273,
277, 281, 290, 306, 318,
333334, 338, 340, 342, 357,
359, 365, 367, 370, 376,
381, 383, 385387, 393394,
399400, 402, 406408,
410411, 413414, 417,
423424, 426428, 434,
438, 444446, 452, 460,
462, 464465, 473, 476,
482, 488489, 494496,
499, 503, 513, 545, 554556,
564565, 579, 581582, 587,
591592, 599, 608612, 614
discursive inference 352
disjunct scope 53, 54, 5, 5860,
65, 89, 90
distance 8284, 151, 218219,

225, 360361, 364,
404405, 407, 435
E
embedding 8, 17, 27, 32, 36, 78,
115, 147, 181182, 187, 190,
199, 223, 252253, 259, 307,
333, 429430, 490491,
521, 539540
encoded inference 11, 13, 16,
342, 451452, 466
endocentric 3132, 34
entropy 83, 87
exhortative 18, 549, 557563,
567569, 572, 574, 576,
578579
exocentric 27, 3132
expectation 330, 389,
494, 520
explanation 13, 1617, 41,
116, 122, 132, 137, 163, 271,
292, 313, 334, 340345,
348349, 351352, 387,
451, 464466, 473476,
485490, 495, 610
explanatory focusing 343, 345,
348, 351
explicative converb 107, 113,
135136, 139
extensible scope 58, 59, 82
extraction 3, 8, 19, 51, 54,
6974, 81, 86, 89, 9192,
98, 100, 609, 615
F
feature percolation 31
final clause 67, 2830, 40, 42,
4547, 63, 388, 458, 557,
567569, 572, 578
finite, finiteness 2, 46, 911,
30, 3336, 48, 5354,
6263, 6769, 81, 85,
9193, 98101, 106, 111112,
143144, 146149, 154157,
160, 162, 165, 167171,
173174, 185, 199200,
210211, 221, 224, 230,
233235, 253254, 269,
274, 305, 308, 334
flexible position 85, 93
focalization 367368, 381, 489,
582, 589, 609
Topic index
focus, focusing 2, 4, 8, 1113,
1518, 5152, 56, 69,
7172, 7475, 81, 86, 92,
93, 146, 148, 203, 205, 207,
212, 218, 221, 224227,
247248, 259260,
269275, 278279, 285,
287, 293299, 301309, 313,
320, 331, 333335, 339342,
343345, 347349, 351353,
357, 393, 395, 409, 422,
428, 431, 433, 444445,
451, 453, 456, 462,
465466, 469476,
478480, 482490,
494497, 502504,
530546, 554, 615; see also
contrastive focus
frame 14, 125, 137, 144, 247,
271, 273, 275280, 282,
284, 294, 300, 304306,
395, 411, 422, 427, 430, 433,
443444, 565
fronting 273, 293, 543
future 14, 18, 4344, 47, 49,
61, 67, 82, 122, 130131,
154, 169, 172, 174, 180, 199,
206, 210, 222223, 226,
229, 235237, 248, 254,
260, 308, 334, 338, 343,
353, 355, 375, 378380,
384, 386388, 391392,
394395, 400402,
406407, 409412,
414415, 418, 457, 472,
504, 507, 514516, 524,
549, 554, 556559,
561564, 566579,
606607
G
generic 372, 379, 504505,
520521, 524, 529
gnomic 15, 308, 490
gradual converb 107, 113,
134, 139
grammaticalization 105, 117,
124, 138, 195, 335, 352,
355, 359, 364, 376,
380381, 387388, 390,
394395, 401403, 467,
476, 554, 565
H
Hamming distance 8283
harmonic marking 89, 91, 92
hortative 47, 49, 52, 505,
517519
hypothesis 8889, 118, 125128,
132, 134, 162, 360, 387388,
390392, 443, 491, 524,
557, 587, 604, 609,
614615; see conditional
hypothetical 19, 178, 251, 283,
302, 331, 334, 359, 380,
390, 394, 415, 440443,
557558, 570572, 575577,
604608, 615
I
I features 7, 2930, 3334, 36,
3843
identification 52, 118, 135, 339,
341, 358, 385387, 394,
405, 414, 473, 475476,
488489, 556
illocutionary force 34, 68,
1011, 15, 1718, 2730,
34, 4243, 4649, 51,
56, 5960, 62, 65, 67,
70, 75, 81, 85, 89, 9193,
200, 239240, 243, 260,
273, 294, 499, 502503,
512519, 522, 545546
imperative 18, 2829, 43, 47,
49, 56, 63, 110, 113, 132, 139,
156157, 163, 167, 216, 219,
238239, 241242, 261, 323,
331, 335336, 344, 351353,
369370, 374, 376377,
379, 384, 387, 389, 418,
452, 457, 460461, 471,
496, 504, 514, 517518,
554, 558561, 563, 567,
579, 606
imperfective 11, 17, 19, 122,
126, 129131, 139, 144, 146,
148149, 152, 154, 156,
162163, 167, 224, 308,
334, 336, 338, 340, 343, 353,
355, 368370, 374377,
379380, 384, 386,
390391, 395, 471472, 475,
478, 492493, 495, 497,
504, 506, 512, 530532, 538,
546, 554, 557558, 565, 579,

603, 606607
implication 87, 91, 285, 302,
304, 383, 395
implicational universals 80,
86, 93
in situ questions 71
inconsequential 291292, 308
indirect speech 253, 255, 322,
374, 376, 389391, 395,
506, 565
inference 11, 13, 1516, 237, 342,
451452, 465466
infinitive 49, 112113, 127,
136139, 148, 156159, 163,
168169, 175, 260261, 425
information structure 11,
1516, 69, 211212, 224,
259260, 270, 296, 333,
345, 351, 356, 427428,
433435, 451452,
462463, 470, 472
informational hierarchy 12,
46, 1113, 1518, 267,
269275, 280, 282,
293294, 297, 304306,
333, 338, 342, 346, 352, 359,
449, 469, 476, 490, 495,
499, 502, 530, 533537, 545
injunction 9, 1314, 143, 145,
156, 355, 376377, 383, 387,
387, 394395, 491, 558, 561,
569, 584, 587, 591, 594, 599
insubordination 132, 144, 162,
467, 519, 542
intensive 208, 338, 340341,
348349, 474475, 489
internal landmark 341,
488, 490
interrogative 4748, 58, 386,
389, 391, 393, 395, 414, 425,
505, 584, 587, 591, 593, 599
IP 67, 15, 27, 3031, 3334,
3949, 5859, 63, 101
irrealis 7, 1718, 4244, 49,
60, 65, 169, 180, 190, 269,
279, 283284, 286, 289,
292, 294, 298299, 303,
305, 416, 492, 499, 502,
504507, 513529, 545546
iterative 67, 386, 497,
524525, 607

J
junctor 19, 595, 604605, 607,
609610, 615
juxtaposition 175, 190, 370,
383, 392, 395, 469, 481483,
495, 567
L
left-detached position 78
local scope 57, 58, 60, 62, 88
locative converb 107, 121122
locator 15, 17, 19, 356,
381382, 479480, 482,
491492, 496
logophoric pronouns 80
M
masdar 9, 112, 148154, 156,
163, 230
matrix clause 3, 9, 2728, 32,
34, 36, 38, 117, 147, 161, 199,
255, 334, 522
modality 910, 161162,
175177, 239, 292, 365, 381,
418, 457, 461, 469, 502,
505, 507, 512525, 545
mood 4, 67, 913, 1516, 18,
2729, 31, 33, 39, 4244,
63, 146147, 149, 154, 156,
163, 167170, 181, 185, 197,
199200, 205, 207, 210,
222, 225, 227, 230, 238,
251, 260, 286287, 305,
355, 370, 391, 418, 423,
438439, 451453,
456458, 460, 465467,
470471, 479480, 499,
501505, 511, 525, 549, 554,
556, 558564, 567574,
577, 579
morphology 1, 8, 18, 44, 62, 68,
80, 9192, 105, 108, 144,
147, 162163, 165, 167, 184,
205, 208210, 224, 227,
230, 259260, 270, 295,
362, 405, 425, 470, 504,
507, 512, 552, 578
multidimensional scaling 83
N
narrative 1617, 56, 7677, 111,
163, 169170, 173, 198, 222,
Topic index
237, 244, 258, 383, 439,
470, 478, 480, 482, 504
NEG transport 39, 59
negation 78, 27, 3942, 47,
5860, 62, 75, 81, 115,
139, 175176, 200, 235,
260, 273274, 279, 283,
303304, 308, 353, 363,
369, 377, 384, 387, 418, 429,
460461, 504, 541542,
555, 562, 566, 571, 579
nexus 4, 6, 2730, 3233,
3941, 4748, 207, 232,
239240, 251253, 259260
nominalization 9, 182, 234,
285286, 299, 305
non-finite 4, 6, 910, 34, 36,
40, 49, 53, 62, 68, 81, 85,
91, 98101, 106, 143144,
147149, 154, 156157,
162, 165, 167171, 174, 199,
233234, 269, 305
NP-modifying
subordination 455
null tense 16, 470472,
478485, 487, 490497
O
object focusing 341, 344
oblique 10, 27, 32, 36, 49,
108, 110, 135, 163, 173,
184185, 240, 308,
430431, 582
operators 3, 8, 29, 47, 5152,
5660, 62, 70, 75, 81,
271274, 302, 373, 514
opposition 70, 108, 144,
167, 170, 206, 211, 214,
289, 334, 368, 425, 428,
433, 440441, 443, 469,
492, 494, 496, 603605,
609616
optative 9, 14, 18, 110, 139,
156157, 160161, 163, 179,
377, 394, 518519, 529
P
parataxis 388, 469470,
495496, 501
participle 10, 61, 115, 122123,
126, 129, 149, 152154,
156158, 163, 166, 168, 170,
173174, 176178, 181185,

192, 198, 334, 353, 365
past 14, 19, 29, 40, 4345,
4849, 52, 61, 6567,
110111, 113, 122, 125, 131,
139, 146, 163, 167170,
172174, 176, 182, 190, 218,
222, 237, 258, 279, 315, 326,
331, 334, 368, 380, 392,
399400, 402, 406408,
410412, 414415, 418, 452,
457, 471472, 475, 481, 483,
497, 524525, 558, 560561,
566, 569, 572, 578579,
604, 607
perception verb 116, 119,
150151, 153, 291292,
336, 352
perfect 11, 13, 1718, 45, 111,
114, 131, 177, 179180, 188,
210, 218, 222227, 236, 258,
260261, 308, 334335,
343, 353, 418, 470472,
475478, 481484, 487,
492, 494495, 497,
499, 502506, 509, 512,
530546, 575
perfective 1719, 49, 110,
112113, 123, 126, 130131,
139, 144, 146, 148, 156157,
159, 161163, 167, 179,
188189, 224, 279, 316,
334, 340, 343, 353, 355, 358,
368370, 372, 374377,
380, 384, 387, 390393,
395, 471472, 475476,
478, 492493, 497, 549,
554, 557561, 565, 574, 579,
603, 606
peripheral operators 29
permutation test 8889
polarity 4, 7, 10, 39, 42, 49, 110,
112, 167168, 171, 176, 199,
218, 235, 249, 258259, 453,
456, 460461, 504, 557;
see negation
polyfunctionality 165166, 341,
351, 359, 415
polysynthetic structure 599
position 10, 1214, 17, 43, 48,
72, 75, 78, 82, 85, 9193,
98, 100, 108, 119, 130, 143,
Topic index
146148, 154, 162163, 173,
180, 190, 195, 207, 216, 227,
229230, 247, 250251, 253,
256, 271, 273, 281, 290295,
297, 299, 302305, 324,
356357, 373, 376, 378,
381382, 384385, 390, 395,
427428, 431432, 436,
438439, 453, 462463,
489, 491, 495, 513, 523, 539,
543544
possibility 3, 37, 44, 47, 57, 59,
65, 68, 7475, 78, 105, 107,
133, 372, 384386, 390, 414,
550, 572, 574575, 611
potential 1416, 110111, 113,
139, 199, 230, 235, 274, 308,
359, 379, 412, 443, 451452,
456461, 464, 466467,
504, 506, 514517, 546
pragmatics 1, 45, 911, 13, 15,
1718, 51, 54, 90, 118, 148,
165, 173, 199, 211, 248, 270,
272273, 299, 302303,
306, 333, 337, 342, 344,
348, 351352, 404405,
421425, 427432,
434435, 438, 440441,
444446, 467, 469,
473474, 483, 485, 488,
490, 496, 499, 502503,
513518, 530536, 543544,
546, 592, 597598, 607
pragmatic demotion 499, 503,
530, 546
pragmatic dependency 333,
469, 490, 496
predicate 4, 10, 52, 58, 77, 82,
109, 117, 145, 149, 154, 158,
167, 170171, 174175, 177,
180185, 191, 194, 197200,
217, 225, 231233, 236237,
286, 317, 334, 337341,
344345, 347349, 353, 356,
363364, 366, 378, 385, 393,
471, 473474, 476477,
488489, 499, 502507,
513, 523, 526, 530545, 555
predicate focusing 340
predicative relation 11, 18,
106, 110, 112, 167, 171,
205206, 225226, 331,
335336, 338340, 345,
348, 363364, 368, 385,

394395, 422, 475479,
488490, 505, 523, 539, 543
present 1, 14, 19, 30, 33, 4345,
49, 57, 61, 66, 107, 111112,
116, 130, 132, 137138, 144,
149, 163, 167169, 172, 174,
194, 196, 203, 222223,
226, 231, 235237, 248,
257258, 261, 270, 313314,
332334, 351, 361, 368, 385,
392395, 399, 401402,
407, 421, 427, 432, 437, 452,
457, 470472, 475476,
478, 488, 494, 524, 558,
560, 564, 572, 574575, 579,
604, 606608, 615
presentative 13, 17, 281, 331,
335337, 340, 342, 344348,
351353, 364, 368, 381,
386387, 393, 470472,
477478, 482, 484486,
493495, 497
presupposition 45, 1213,
1618, 34, 4648, 57, 92,
199, 269276, 287289,
291294, 296297, 300,
305306, 339342, 421,
424, 451, 462, 464466,
475476, 478, 485,
488490, 494, 533, 535,
537, 540541, 545
procedural meaning 452, 466
prohibition 377, 387, 395, 461
prosody 9, 16, 144, 148,
163, 270, 272, 344, 460,
518519, 540541, 543
protasis 14, 1718, 132, 251,
279, 358, 371, 379380,
390393, 395396, 443,
487, 492493, 508, 514,
522, 536
prototypical 6, 7, 27, 29, 41,
82, 93, 219, 388, 437438,
473, 488
proverbs 192, 370, 383,
479, 490
purpose (clauses) 911,
1517, 19, 52, 58, 105109,
133, 136137, 149, 154,
158, 161, 165166, 190,
193195, 198199, 226, 243,
256258, 259260, 275,
282, 284286, 298, 313,

373, 383, 434, 439, 444, 451,
456, 458461, 461462,
465467, 479, 491492,
507, 520, 525526, 529,
587, 589
purposive converb 9, 107, 113,
121, 136139
Q
qualitative predication 339,
473, 474476, 484, 485,
488490
question words 46, 6970, 72,
74, 81, 88, 90, 544
quotative 9, 1314, 165166,
170, 187188, 190, 193,
195, 197, 253, 295, 334, 350,
355, 376, 388389, 391,
394395, 506
R
raising 80, 89, 151
realis 7, 4244, 49, 60, 6566,
167, 169, 190, 284, 286,
289, 292, 301, 305, 308,
387, 416, 502, 504, 506,
513514, 530533, 537, 545
reanalysis 13, 144, 224, 334, 526
reason clause 13, 190, 260, 289,
315, 321
referential 260, 270, 278, 283,
288289, 292293, 297,
306, 350, 356357, 364,
380381, 383, 385387, 390,
395, 606
referential hierarchy 1, 5,
1112, 267, 269272,
274275, 285, 294, 297,
304306, 346
referentiality 365, 381, 392
relative (clause) 2, 8, 1011,
1314, 1819, 3233, 36,
38, 4445, 61, 7073,
8285, 8788, 91, 106, 115,
122123, 127, 152154, 157,
171, 176, 180, 183185, 199,
203, 205207, 222227,
230237, 240, 253,
259261, 275278,
281290, 292, 295,
297300, 302, 305306,
309, 326, 333334, 336337,

339, 345, 350353, 361, 363,
365367, 371, 381382,
384385, 387, 394395,
399, 408409, 415,
422, 425, 429, 434, 439,
462, 469, 483, 497, 502,
508509, 519521, 525, 530,
535546, 555556, 558561,
563567, 569570, 574, 579
relativization 115, 166, 181183,
205, 363, 538, 540
residuals 8891
resultative 13, 225, 476, 512,
526529, 559
rheme 367, 472473, 476477
right-detached position
78, 382
S
scope 23, 68, 1213, 2930,
34, 3942, 4448, 5154,
5663, 65, 6970, 72,
7475, 7982, 8586,
8893, 98, 100, 175176,
185, 200, 238, 271274,
283, 294, 299, 302, 304,
313314, 326, 335337, 340,
345346, 348349, 360,
429, 439, 445, 489, 502,
533, 536, 608
semantic role assignment
7879
semantic relations 1, 5, 1011,
15, 1718, 31, 46, 56, 74,
121, 171, 176, 166, 188, 198,
203205, 208, 232, 234,
236237, 239240, 252,
270, 272, 284, 287, 302,
334335, 344, 352, 400,
426427, 451452, 454,
456, 461, 467, 469, 476,
484485, 490, 492, 495,
533, 556, 558, 564, 611
sentence 4, 9, 1216, 1819, 29,
31, 34, 40, 4348, 55, 61,
63, 68, 7879, 91, 115116,
119120, 129, 154, 165, 167,
170174, 176, 180181,
183185, 190, 193194,
196197, 199, 204, 218,
224, 226, 228229, 243,
250, 252, 257, 259260,
270271, 274275, 277,
Topic index
281, 290, 293294, 297,
299, 302303, 305, 313314,
318, 321, 324325, 328,
330331, 337, 339341, 348,
355356, 359, 367369,
381382, 389390, 395,
405, 421424, 427440,
443446, 469470,
472475, 479482,
485486, 488492, 494,
502508, 513526,
530546, 549550,
554556, 558, 561563, 567,
570, 578, 580, 612, 614
sequential 910, 14, 18, 4445,
49, 52, 79, 143, 147148,
156157, 159160, 162163,
175176, 179, 196, 201, 236,
272, 275, 277, 280281,
284286, 298, 302306,
309, 331, 370, 383384, 392,
395, 504
sequential converb 63, 77,
162163
similarity measure 82
similative converb 107, 113,
134, 139
simple sentence 15, 313314,
331, 421, 446, 554
simultaneity 11, 1718, 125,
176, 179, 188, 235236, 411,
492495, 556, 565, 592
situational anaphora 16, 384,
482483, 491492, 495
situational dependency 6,
19, 383, 469, 490, 496,
587588, 598
speakers commitment
491, 495
specialized converb 118,
181, 198
speech act 17, 118, 145, 461,
477, 491, 494, 513518, 612;
see illocutionary force
speech-act modification 78
split assertion 339, 475, 488
split graphs 84
statement 30, 39, 4647, 238,
349, 443, 475476, 481,
484486, 518, 607
stative 208210, 236237,
257, 261, 309, 360, 471,
474475, 477478,
483486, 490, 493, 495,

504, 506, 528533
status (realis/irrealis)
marker 60, 65
subject 5, 7, 11, 16, 2934,
36, 41, 4849, 52, 55, 58,
6569, 73, 77, 81, 85, 150,
153, 158160, 170, 173175,
181, 200, 204205,
208212, 218221, 229230,
234, 238244, 246248,
250, 254258, 260, 283,
298, 301, 317, 325326, 334,
337340, 345348, 353,
356357, 363364, 368,
373, 375, 382, 385, 391, 402,
424425, 431, 435438, 453,
455461, 463, 465466,
471474, 483, 489490,
494, 496497, 503, 505,
511, 520, 526, 535, 538544,
552556, 558559, 561,
563564, 567579, 610
subject focusing 339, 473, 497,
540544; see focus
subjunctive 1718, 65, 68, 393,
401, 438, 499, 502505,
507510, 512531, 545,
578, 607
subordinate clauses 34, 912,
14, 18, 2728, 3334, 36,
3839, 48, 207, 233, 240,
252, 254, 269, 273, 284,
286, 288, 294295, 305,
334, 352, 366, 372, 383384,
389, 392, 409410,
421422, 424441,
444446, 469, 512, 537,
595599
subordination 16, 8, 10, 12,
1319, 27, 32, 34, 39, 48,
5152, 54, 57, 6566, 69,
7678, 84, 86, 93, 105107,
115, 135, 138, 147, 150,
161162, 166, 169, 171, 198,
201, 203, 207, 219, 221,
223224, 230, 233236,
239240, 252, 260, 267,
271275, 281, 285289, 293,
295298, 303307, 314,
333334, 358, 374, 376,
381, 383, 387389, 394,
399400, 410417,
Topic index
421422, 425, 430, 446,
451452, 455, 458459,
469, 479, 481, 490, 499,
501, 503, 506, 508510,
512514, 517, 519, 521522,
525526, 529530,
534540, 545, 569,
580582, 586587, 594,
598599, 603604,
609, 615
succession 10, 1617, 196, 219,
235236, 256, 371, 460,
466, 469, 480, 482484,
487, 492, 495496, 549,
558562, 565, 566567,
569571, 574, 577579
surprise 17, 111, 478,
494495, 610
switch-reference 66, 68, 71, 73,
80, 260
symmetry 4, 57, 6768, 81,
9193, 98, 100, 204, 407
syntactic dependency 171, 198,
201, 469, 480, 490491,
495, 513, 517, 519520, 522
syntax 1, 56, 11, 1718,
51, 54, 56, 74, 112, 138,
144, 147148, 165, 205,
208209, 216, 230, 232,
240241, 259, 270, 273,
405, 422, 426, 446, 499,
503, 519, 529, 536, 538,
541546, 554
synthetic processes 582, 586
T
TAM 18, 109110, 112, 115,
210211, 218, 222, 229,
355, 359, 368370, 372373,
376, 380389, 392393,
395, 409, 415, 423, 449,

499, 501546; see
tense-aspect-mood
temporal clause 184, 231,
371372, 383, 392, 484,
556, 576
temporal converb 6465
temporality 373, 603605, 615
tense 4, 611, 1316,
1819, 2729, 31, 33, 36,
3940, 4245, 4849,
54, 58, 6062, 6568, 81,
9091, 93, 110111, 122, 143,
146147, 163, 167, 169172,
180181, 185, 199200,
205207, 210211, 218, 220,
222, 225227, 229230,
233235, 237, 240, 246,
251, 257258, 261, 334, 355,
383384, 391, 394, 418, 423,
438439, 453457, 460,
469472, 476, 478485,
487, 490497, 499, 501,
504505, 511, 549, 554,
556, 558, 560564,
567574, 577578,
604606, 609, 615
tense-aspect-mood 222,
355, 499, 501, 504, 511;
see TAM
tense-mood concordance 18,
549, 554, 556, 561564,
567574, 577
time clauses 176, 184, 231,
275276, 278, 295, 298,
300, 399400, 409411,
415417, 484, 506,
509510, 521, 525, 529
time of speech 19, 286, 478,
484, 490
time of utterance 174, 411,

415, 417
topic 4, 6, 1113, 33, 3638,
49, 57, 66, 6970, 73,
7576, 7880, 8485, 92,
100101, 107, 112, 115, 144,
173175, 181, 183, 200, 207,
218, 230, 243, 247, 250, 256,
260, 269282, 284285,
293295, 297301,
303309, 313, 319320,
333335, 338, 340, 345,
348353, 357, 367, 381382,
390393, 395, 410, 444,
451, 462, 464, 524, 534,
539, 554
topic clauses 12, 69, 73, 75, 271,
274, 301, 305307, 524,
534, 545
topicalization 13, 246247,
256, 314, 333, 338, 367, 381,
390392, 395
transported scope 60, 62
typology 1, 41, 56, 65, 8384,
93, 166, 170, 198, 203, 207,
230, 355, 392, 451, 469,
489, 495
U
universals 23, 51, 80, 86,
90, 93
V
verb serialization 171, 200,
501, 503, 511512, 526,
531, 540
W
wh-questions 472, 474,
490, 544
Studies in Language Companion Series

A complete list of titles in this series can be found on the publishers website, www.benjamins.com
123 PUTNAM, Michael T. (ed.): Studies on German-Language Islands. xii,473pp.+index. Expected January
2011
122 CLANCY, Steven J.: The Chain of Being and Having in Slavic. xix,289pp.+index. Expected December 2010
121 BRIL, Isabelle (ed.): Clause Linking and Clause Hierarchy. Syntax and pragmatics. 2010. viii,632pp.
120 ROTHSTEIN, Bjrn and Rolf THIEROFF (eds.): Mood in the Languages of Europe. xiv,643pp.+index.
Expected November 2010
119 STATHI, Katerina, Elke GEHWEILER and Ekkehard KNIG (eds.): Grammaticalization. Current views
and issues. 2010. vii,379pp.
118 MHLEISEN, Susanne: Heterogeneity in Word-Formation Patterns. A corpus-based analysis of suffixation
with -ee and its productivity in English. 2010. xiii,245pp.
117 SPEVAK, Olga: Constituent Order in Classical Latin Prose. 2010. xv,318pp.
116 NORDSTRM, Jackie: Modality and Subordinators. 2010. xvii,341pp.
115 HASKO, Victoria and Renee PERELMUTTER (eds.): New Approaches to Slavic Verbs of Motion. 2010.
x,392pp.
114 ROBY, David Brian: Aspect and the Categorization of States. The case of ser and estar in Spanish. 2009.
xiii,191pp.
113 COMRIE, Bernard, Ray FABRI, Elizabeth HUME, Manwel MIFSUD, Thomas STOLZ and Martine
VANHOVE (eds.): Introducing Maltese Linguistics. Selected papers from the 1st International Conference
on Maltese Linguistics, Bremen, 1820 October, 2007. 2009. xi,422pp.
112 DUFTER, Andreas and Daniel JACOB (eds.): Focus and Background in Romance Languages. 2009.
vii,362pp.
111 POLGURE, Alain and Igor A. MELUK (eds.): Dependency in Linguistic Description. 2009.
xxii,281pp.
110 DIMMENDAAL, Gerrit J. (ed.): Coding Participant Marking. Construction types in twelve African
languages. 2009. xvi,389pp.
109 NARROG, Heiko: Modality in Japanese. The layered structure of the clause and hierarchies of functional
categories. 2009. xxii,277pp.
108 BARDAL, Jhanna and Shobhana L. CHELLIAH (eds.): The Role of Semantic, Pragmatic, and
Discourse Factors in the Development of Case. 2009. xx,432pp.
107 BUTLER, Christopher S. and Javier MARTN ARISTA (eds.): Deconstructing Constructions. 2009.
xx,306pp.
106 VANHOVE, Martine (ed.): From Polysemy to Semantic Change. Towards a typology of lexical semantic
associations. 2008. xiii,404pp.
105 VAN VALIN, JR., Robert D. (ed.): Investigations of the SyntaxSemanticsPragmatics Interface. 2008.
xxiv,484pp.
104 MUSHIN, Ilana and Brett BAKER (eds.): Discourse and Grammar in Australian Languages. 2008.
x,239pp.
103 JOSEPHSON, Folke and Ingmar SHRMAN (eds.): Interdependence of Diachronic and Synchronic
Analyses. 2008. viii,350pp.
102 GODDARD, Cliff (ed.): Cross-Linguistic Semantics. 2008. xvi,356pp.
101 STOLZ, Thomas, Sonja KETTLER, Cornelia STROH and Aina URDZE: Split Possession. An areallinguistic study of the alienability correlation and related phenomena in the languages of Europe. 2008.
x,546pp.
100 AMEKA, Felix K. and M.E. KROPP DAKUBU (eds.): Aspect and Modality in Kwa Languages. 2008.
ix,335pp.
99 HEG MLLER, Henrik and Alex KLINGE (eds.): Essays on Nominal Determination. From morphology
to discourse management. 2008. xviii,369pp.
98 FABRICIUS-HANSEN, Cathrine and Wiebke RAMM (eds.): 'Subordination' versus 'Coordination' in
Sentence and Text. A cross-linguistic perspective. 2008. vi,359pp.
97 DOLLINGER, Stefan: New-Dialect Formation in Canada. Evidence from the English modal auxiliaries.
2008. xxii,355pp.
96 ROMEO, Nicoletta: Aspect in Burmese. Meaning and function. 2008. xv,289pp.

95 OCONNOR, Loretta: Motion, Transfer and Transformation. The grammar of change in Lowland Chontal.
2007. xiv,251pp.
94 MIESTAMO, Matti, Kaius SINNEMKI and Fred KARLSSON (eds.): Language Complexity. Typology,
contact, change. 2008. xiv,356pp.
93 SCHALLEY, Andrea C. and Drew KHLENTZOS (eds.): Mental States. Volume 2: Language and cognitive
structure. 2007. x,362pp.
92 SCHALLEY, Andrea C. and Drew KHLENTZOS (eds.): Mental States. Volume 1: Evolution, function,
nature. 2007. xii,304pp.
91 FILIPOVI, Luna: Talking about Motion. A crosslinguistic investigation of lexicalization patterns. 2007.
x,182pp.
90 MUYSKEN, Pieter (ed.): From Linguistic Areas to Areal Linguistics. 2008. vii,293pp.
89 STARK, Elisabeth, Elisabeth LEISS and Werner ABRAHAM (eds.): Nominal Determination. Typology,
context constraints, and historical emergence. 2007. viii,370pp.
88 RAMAT, Paolo and Elisa ROMA (eds.): Europe and the Mediterranean as Linguistic Areas. Convergencies
from a historical and typological perspective. 2007. xxvi,364pp.
87 VERHOEVEN, Elisabeth: Experiential Constructions in Yucatec Maya. A typologically based analysis of a
functional domain in a Mayan language. 2007. xiv,380pp.
86 SCHWARZ-FRIESEL, Monika, Manfred CONSTEN and Mareile KNEES (eds.): Anaphors in Text.
Cognitive, formal and applied approaches to anaphoric reference. 2007. xvi,282pp.
85 BUTLER, Christopher S., Raquel HIDALGO DOWNING and Julia LAVID (eds.): Functional
Perspectives on Grammar and Discourse. In honour of Angela Downing. 2007. xxx,481pp.
84 WANNER, Leo (ed.): Selected Lexical and Grammatical Issues in the MeaningText Theory. In honour of
Igor Mel'uk. 2007. xviii,380pp.
83 HANNAY, Mike and Gerard J. STEEN (eds.): Structural-Functional Studies in English Grammar. In
honour of Lachlan Mackenzie. 2007. vi,393pp.
82 ZIEGELER, Debra: Interfaces with English Aspect. Diachronic and empirical studies. 2006. xvi,325pp.
81 PEETERS, Bert (ed.): Semantic Primes and Universal Grammar. Empirical evidence from the Romance
languages. 2006. xvi,374pp.
80 BIRNER, Betty J. and Gregory WARD (eds.): Drawing the Boundaries of Meaning. Neo-Gricean studies in
pragmatics and semantics in honor of Laurence R. Horn. 2006. xii,350pp.
79 LAFFUT, An: Three-Participant Constructions in English. A functional-cognitive approach to caused
relations. 2006. ix,268pp.
78 YAMAMOTO, Mutsumi: Agency and Impersonality. Their Linguistic and Cultural Manifestations. 2006.
x,152pp.
77 KULIKOV, Leonid, Andrej MALCHUKOV and Peter de SWART (eds.): Case, Valency and Transitivity.
2006. xx,503pp.
76 NEVALAINEN, Terttu, Juhani KLEMOLA and Mikko LAITINEN (eds.): Types of Variation. Diachronic,
dialectal and typological interfaces. 2006. viii,378pp.
75 HOLE, Daniel, Andr MEINUNGER and Werner ABRAHAM (eds.): Datives and Other Cases. Between
argument structure and event structure. 2006. viii,385pp.
74 PIETRANDREA, Paola: Epistemic Modality. Functional properties and the Italian system. 2005.
xii,232pp.
73 XIAO, Richard and Tony McENERY: Aspect in Mandarin Chinese. A corpus-based study. 2004. x,305pp.
72 FRAJZYNGIER, Zygmunt, Adam HODGES and David S. ROOD (eds.): Linguistic Diversity and
Language Theories. 2005. xii,432pp.
71 DAHL, sten: The Growth and Maintenance of Linguistic Complexity. 2004. x,336pp.
70 LEFEBVRE, Claire: Issues in the Study of Pidgin and Creole Languages. 2004. xvi,358pp.
69 TANAKA, Lidia: Gender, Language and Culture. A study of Japanese television interview discourse. 2004.
xvii,233pp.
68 MODER, Carol Lynn and Aida MARTINOVIC-ZIC (eds.): Discourse Across Languages and Cultures.
2004. vi,366pp.
67 LURAGHI, Silvia: On the Meaning of Prepositions and Cases. The expression of semantic roles in Ancient
Greek. 2003. xii,366pp.
66 NARIYAMA, Shigeko: Ellipsis and Reference Tracking in Japanese. 2003. xvi,400pp.
65 MATSUMOTO, Kazuko: Intonation Units in Japanese Conversation. Syntactic, informational and

functional structures. 2003. xviii,215pp.
64 BUTLER, Christopher S.: Structure and Function A Guide to Three Major Structural-Functional
Theories. Part 2: From clause to discourse and beyond. 2003. xiv,579pp.
63 BUTLER, Christopher S.: Structure and Function A Guide to Three Major Structural-Functional
Theories. Part 1: Approaches to the simplex clause. 2003. xx,573pp.
62 FIELD, Fredric: Linguistic Borrowing in Bilingual Contexts. With a foreword by Bernard Comrie. 2002.
xviii,255pp.
61 GODDARD, Cliff and Anna WIERZBICKA (eds.): Meaning and Universal Grammar. Theory and
empirical findings. Volume 2. 2002. xvi,337pp.
60 GODDARD, Cliff and Anna WIERZBICKA (eds.): Meaning and Universal Grammar. Theory and
empirical findings. Volume 1. 2002. xvi,337pp.
59 SHI, Yuzhi: The Establishment of Modern Chinese Grammar. The formation of the resultative construction
and its effects. 2002. xiv,262pp.
58 MAYLOR, B. Roger: Lexical Template Morphology. Change of state and the verbal prefixes in German.
2002. x,273pp.
57 MELUK, Igor A.: Communicative Organization in Natural Language. The semantic-communicative
structure of sentences. 2001. xii,393pp.
56 FAARLUND, Jan Terje (ed.): Grammatical Relations in Change. 2001. viii,326pp.
55 DAHL, sten and Maria KOPTJEVSKAJA-TAMM (eds.): Circum-Baltic Languages. Volume 2: Grammar
and Typology. 2001. xx,423pp.
54 DAHL, sten and Maria KOPTJEVSKAJA-TAMM (eds.): Circum-Baltic Languages. Volume 1: Past and
Present. 2001. xx,382pp.
53 FISCHER, Olga, Anette ROSENBACH and Dieter STEIN (eds.): Pathways of Change.
Grammaticalization in English. 2000. x,391pp.
52 TORRES CACOULLOS, Rena: Grammaticization, Synchronic Variation, and Language Contact. A study of
Spanish progressive -ndo constructions. 2000. xvi,255pp.
51 ZIEGELER, Debra: Hypothetical Modality. Grammaticalisation in an L2 dialect. 2000. xx,290pp.
50 ABRAHAM, Werner and Leonid KULIKOV (eds.): Tense-Aspect, Transitivity and Causativity. Essays in
honour of Vladimir Nedjalkov. 1999. xxxiv,359pp.
49 BHAT, D.N.S.: The Prominence of Tense, Aspect and Mood. 1999. xii,198pp.
48 MANNEY, Linda Joyce: Middle Voice in Modern Greek. Meaning and function of an inflectional category.
2000. xiii,262pp.
47 BRINTON, Laurel J. and Minoji AKIMOTO (eds.): Collocational and Idiomatic Aspects of Composite
Predicates in the History of English. 1999. xiv,283pp.
46 YAMAMOTO, Mutsumi: Animacy and Reference. A cognitive approach to corpus linguistics. 1999.
xviii,278pp.
45 COLLINS, Peter and David LEE (eds.): The Clause in English. In honour of Rodney Huddleston. 1999.
xv,342pp.
44 HANNAY, Mike and A. Machtelt BOLKESTEIN (eds.): Functional Grammar and Verbal Interaction. 1998.
xii,304pp.
43 OLBERTZ, Hella, Kees HENGEVELD and Jess SNCHEZ GARCA (eds.): The Structure of the
Lexicon in Functional Grammar. 1998. xii,312pp.
42 DARNELL, Michael, Edith A. MORAVCSIK, Michael NOONAN, Frederick J. NEWMEYER and
Kathleen M. WHEATLEY (eds.): Functionalism and Formalism in Linguistics. Volume II: Case studies.
1999. vi,407pp.
41 DARNELL, Michael, Edith A. MORAVCSIK, Michael NOONAN, Frederick J. NEWMEYER and
Kathleen M. WHEATLEY (eds.): Functionalism and Formalism in Linguistics. Volume I: General papers.
1999. vi,486pp.
40 BIRNER, Betty J. and Gregory WARD: Information Status and Noncanonical Word Order in English.
1998. xiv,314pp.
39 WANNER, Leo (ed.): Recent Trends in MeaningText Theory. 1997. xx,202pp.
38 HACKING, Jane F.: Coding the Hypothetical. A comparative typology of Russian and Macedonian
conditionals. 1998. vi,156pp.
37 HARVEY, Mark and Nicholas REID (eds.): Nominal Classification in Aboriginal Australia. 1997. x,296pp.
36 KAMIO, Akio (ed.): Directions in Functional Linguistics. 1997. xiii,259pp.

35 MATSUMOTO, Yoshiko: Noun-Modifying Constructions in Japanese. A frame semantic approach. 1997.
viii,204pp.
34 HATAV, Galia: The Semantics of Aspect and Modality. Evidence from English and Biblical Hebrew. 1997.
x,224pp.
33 VELZQUEZ-CASTILLO, Maura: The Grammar of Possession. Inalienability, incorporation and
possessor ascension in Guaran. 1996. xvi,274pp.
32 FRAJZYNGIER, Zygmunt: Grammaticalization of the Complex Sentence. A case study in Chadic. 1996.
xviii,501pp.
31 WANNER, Leo (ed.): Lexical Functions in Lexicography and Natural Language Processing. 1996.
xx,355pp.
30 HUFFMAN, Alan: The Categories of Grammar. French lui and le. 1997. xiv,379pp.
29 ENGBERG-PEDERSEN, Elisabeth, Michael FORTESCUE, Peter HARDER, Lars HELTOFT and Lisbeth
Falster JAKOBSEN (eds.): Content, Expression and Structure. Studies in Danish functional grammar. 1996.
xvi,510pp.
28 HERMAN, Jzsef (ed.): Linguistic Studies on Latin. Selected papers from the 6th International Colloquium
on Latin Linguistics (Budapest, 2327 March 1991). 1994. ix,421pp.
27 ABRAHAM, Werner, T. GIVN and Sandra A. THOMPSON (eds.): Discourse, Grammar and Typology.
Papers in honor of John W.M. Verhaar. 1995. xx,352pp.
26 LIMA, Susan D., Roberta CORRIGAN and Gregory K. IVERSON: The Reality of Linguistic Rules. 1994.
xxiii,480pp.
25 GODDARD, Cliff and Anna WIERZBICKA (eds.): Semantic and Lexical Universals. Theory and
empirical findings. 1994. viii,510pp.
24 BHAT, D.N.S.: The Adjectival Category. Criteria for differentiation and identification. 1994. xii,295pp.
23 COMRIE, Bernard and Maria POLINSKY (eds.): Causatives and Transitivity. 1993. x,399pp.
22 McGREGOR, William B.: A Functional Grammar of Gooniyandi. 1990. xx,618pp.
21 COLEMAN, Robert (ed.): New Studies in Latin Linguistics. Proceedings of the 4th International
Colloquium on Latin Linguistics, Cambridge, April 1987. 1990. x,480pp.
20 VERHAAR, John W.M. S.J. (ed.): Melanesian Pidgin and Tok Pisin. Proceedings of the First International
Conference on Pidgins and Creoles in Melanesia. 1990. xiv,409pp.
19 BLUST, Robert A.: Austronesian Root Theory. An essay on the limits of morphology. 1988. xi,190pp.
18 WIERZBICKA, Anna: The Semantics of Grammar. 1988. vii,581pp.
17 CALBOLI, Gualtiero (ed.): Subordination and Other Topics in Latin. Proceedings of the Third Colloquium
on Latin Linguistics, Bologna, 15 April 1985. 1989. xxix,691pp.
16 CONTE, Maria-Elisabeth, Jnos Snder PETFI and Emel SZER (eds.): Text and Discourse
Connectedness. Proceedings of the Conference on Connexity and Coherence, Urbino, July 1621, 1984.
1989. xxiv,584pp.
15 JUSTICE, David: The Semantics of Form in Arabic. In the mirror of European languages. 1987. iv,417pp.
14 BENSON, Morton, Evelyn BENSON and Robert F. ILSON: Lexicographic Description of English. 1986.
xiii,275pp.
13 REESINK, Ger P.: Structures and their Functions in Usan. 1987. xviii,369pp.
12 PINKSTER, Harm (ed.): Latin Linguistics and Linguistic Theory. Proceedings of the 1st International
Colloquium on Latin Linguistics, Amsterdam, April 1981. 1983. xviii,307pp.
11 PANHUIS, Dirk G.J.: The Communicative Perspective in the Sentence. A study of Latin word order. 1982.
viii,172pp.
10 DRESSLER, Wolfgang U., Willi MAYERTHALER, Oswald PANAGL and Wolfgang Ullrich WURZEL:
Leitmotifs in Natural Morphology. 1987. ix,168pp.
9 LANG, Ewald and John PHEBY: The Semantics of Coordination. (English transl. by John Pheby from the
German orig. ed. 'Semantik der koordinativen Verknpfung', Berlin, 1977). 1984. 300pp.
8 BARTH, E.M. and J.L. MARTENS (eds.): Argumentation: Approaches to Theory Formation. Containing
the Contributions to the Groningen Conference on the Theory of Argumentation, October 1978. 1982.
xviii,333pp.
7 PARRET, Herman, Marina SBIS and Jef VERSCHUEREN (eds.): Possibilities and Limitations of
Pragmatics. Proceedings of the Conference on Pragmatics, Urbino, July 814, 1979. 1981. x,854pp.

Isabelle Bril. Clause Linking and Clause Hierarchy

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Isabelle Bril. Clause Linking and Clause Hierarchy

Uploaded by

Copyright:

Available Formats

Clause Linking and Clause Hierarchy

Studies in Language Companion Series (SLCS)

Elly van Gelderen

Arizona State University

Max Planck Institute, Leipzig

University of New Mexico

Free University of Berlin

University of California, Santa Barbara

John Benjamins Publishing Company

The paper used in this publication meets the minimum requirements of

Library of Congress Cataloging-in-Publication Data

2010 John Benjamins B.V.

part i. Syntactic terminology and typological methods

Capturing particulars and universals in clause linkage: A multivariate analysis

part ii. Clause-chaining, converbs, masdars, absolutive constructions, etc.

Finite and non-finite: Prosodic distinctions on Budugh verb stems

Converbs and adverbial clauses in Badaga, a South-Dravidian language

Coordination, converbs and clause chaining in Coptic

part iii. Subordination, informational hierarchy and referential hierarchy

Comment clause: Crossing the boundaries between simple

Clause Linking and Clause Hierarchy

Deixis, information structure and clause linkage in Yafi Arabic (Yemen)

part iv. Informational hierarchy and TAM markers

Clause chaining and conjugations in Wolof: A typology of

Pragmatic demotion and clause dependency: On two atypical

Tense-mood concordance and clause chaining in Mankon

Clause dependency relations in East Greenlandic Inuit

Coordination and subordination: ma in Bulgarian dialectal Greek

Clause Linking and Clause Hierarchy

LACITO (Laboratoire des Langues et Civilisations Tradition orale)

Are coordination, subordination and co-subordination universal syntactic categories?

Are there mismatches between form, function and meaning of clause-linking

Part I presents some theoretical reassessment of terminologies from syntactic and

Part IV presents insights in the clause-linking and subordinating functions of some

The origin and evolution of clause-linking morphemes or strategies is a topic of

Part I. A reassessment of terminology and typological methods

Assessment of I-features: Polarity, illocutionary force, mood, tense

B. Statistical methods for typological comparability: A multi-variate approach

Part II. Converbs, masdars, clause-chaining, conjunctive conjugations

B. Converbs, masdars and prosody

based on an adjectival participle form, used with a nominal or an adverbial head in

D. Clause chaining and conjunctive conjugations

Part III. The syntax, pragmatic, semantic interface

in Part IV (Verstraete, Robert, Franois). Clause-linking and subordinating strategies

A. Informational hierarchy, referential hierarchy and clause-linkage

functions in clause-linkage. Demonstratives may have deictic, endophoric reference,

B. Deictics as conjunctive and correlative markers

Part IV. T.A.M. strategies and informational hierarchy

A. Mood, informational hierarchy and clause-linkage

B. Conjugations, informational hierarchy and clause-linkage

clauses, consecutive or purpose clauses), without any other subordinating morpheme.

C. Pragmatics, T.A.M dependency and subordination

D. T.A.M dependency and clause-chaining

E. T.A.M markers and subordinators

Index: Languages studied

Other languages surveyed in this volume

Kabyle (Berber from Algeria, Afro-asiatic)

Clause linkage and Nexus

Figure 1. Traditional Nexus Types

tekekee:k a:pa-w e:-ta ke-yae-w,

b. tekekee:k a:pa-ta ke-yae-w

Clause linkage and Nexus in Papuan languages

(S)I (S)I (S)I (S)I

Cosubordinate Nexus: (S S S S)I

Capturing particulars and universals in clause linkage: A multivariate analysis

Finite and non-finite: Prosodic distinctions on Budugh verb stems

Converbs and adverbial clauses in Badaga, a South-Dravidian language

Deixis, information structure and clause linkage in Yafi Arabic (Yemen)

part iv. Informational hierarchy and TAM markers

Clause dependency relations in East Greenlandic Inuit

Coordination and subordination: ma in Bulgarian dialectal Greek

tepau-fe-pa yate fitau-a=nae?

nen mei momune-i-nani=ra pofei-ti nen=tu-e=nae?

Maliika tyka-na=a j-ax-na ca-j-ea-r=ii?