Professional Documents
Culture Documents
Editors
Werner Abraham Elly van Gelderen
University of Vienna Arizona State University
Editorial Board
Bernard Comrie Christian Lehmann
Max Planck Institute, Leipzig University of Erfurt
and University of California, Santa Barbara
Brian MacWhinney
William Croft Carnegie-Mellon University
University of New Mexico
Marianne Mithun
Östen Dahl University of California, Santa Barbara
University of Stockholm
Heiko Narrog
Gerrit J. Dimmendaal Tohuku University
University of Cologne
Johanna L. Wood
Ekkehard König University of Aarhus
Free University of Berlin
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
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)
Deixis, information structure and clause linkage in Yafi‘ Arabic (Yemen) 333
Martine Vanhove
The role of the Berber deictic and TAM markers in dependent
clauses in Zenaga 355
Catherine Taine-Cheikh
Deixis and temporal subordinators in Pomak (Slavic, Greece) 399
Evangelia Adamou
Correlative markers as phoric “Grammaticalised Category Markers”
of subordination in German 421
Colette Cortès
1. Presentation
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 theo-
retical 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 (König & Haspelmath
(eds) 1995), adverbial constructions (van der Auwera (ed.) 1998), coordination
(Sag et al. 1985; Munn 1993; Johannessen 1998; Progovac 1998; Haspelmath 2004;
Godard & Abeillé 2005). Typological studies have also focused on clause-linking in
its various aspects (inter alia, Lehman 1988; Haiman 1988; Comrie 1989 and vari-
ous 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 orig-
inate from?
Some more general theoretical and methodological questions are also addressed:
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 uni-
versal constructions and categories, or any universal conceptual structures or for-
mal 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 con-
structions 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
volume’s contributions.
4. S
ome properties of coordination and subordination
and some distinctive tests
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 illo-
cutionary 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 sub-
ordinate 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, case-
markers 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).
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.
This study, based on two dozen languages, isolates a set of twelve analytical
variables (ranging from the scope of illocutionary force operators to extraction con-
straints) that are applied to a representative selection of clause-linkage structures. The
analysis supports Foley’s view that ‘cosubordination’ is not a distinct prototype, while
there is a cross-linguistic prototype of subordination characterised by disjunct illocu-
tionary 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 adjoin-
ing, 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.
A. Converbs
Three papers are concerned more or less centrally with converbial strategies in Daghes-
tanian (Creissels, Authier), Dravidian languages (Pilot-Raichoor) and Coptic (Reintges).
They generally follow Nedjalkov’s (1995) distinction between specialised and general-
polyfunctional converbs.
D. Creissels (Chapter 4) 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 Nedjalkov’s 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 con-
verbs may be defined as essentially marking adverbial subordination, the multipur-
pose ‘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 coordina-
tion. Another particularly strong argument in favour of subordination is that, in some
complex constructions, the general converb in Akhvakh may show external suffixal
Editor’s introduction
agreement, i.e. controlled by an NP outside the clause headed by the dependent verb
form and controlled by the main verb’s S/P agreement.
Converbal clauses usually precede the main verb, but the reverse order is also pos-
sible though not equally usual for all converbs. The order main verb–converb 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.
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); different-
subject conjunctive clauses have an illocutionary force different from that of the initial
conjunct, the different subject is contrastively focused and the spreading of the illocu-
tionary 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 coordina-
tion and complex predicates. They occur (i) in restrictive relative clauses, (ii) in predica-
tive 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 Rela-
tive tenses, also play a role in information structure and focus marking, which sets
them apart from pragmatically neutral declarative clauses.
phrase or a clause); but the form wá 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 sys-
tems 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), raΩ, 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, raΩ developed (i) a deictic function as a pre-
sentative, (ii) an assertive function as a copula, (iii) an aspectual function as a perfect-
resultative 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 reanaly-
sis 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, raΩ
marks the hierarchy between a presupposition and an assertion. When used in clause-
linkage, the clause focused by raΩ is interpreted as the cause or explanation for the
other event. Following Verstraete (this volume), Vanhove points out that the function
of raΩ 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. Cortès (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
Editor’s introduction
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.
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 poten-
tial 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 con-
tribute 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 sub-
ject 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.
The Subjunctive clause refers to some virtual state of affairs and lacks illocution-
ary 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,
I’d 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 rou-
tinised 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 sub-
ordinate clause is explicitly irrealis or generic (as in Zenaga and Wolof).
N. Tersis (Chapter 18) 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,
the concomitant, the causative/effected, and the conditional. The attributive -ti- is a
Editor’s introduction
dependent conjugation, used in relative and complement clauses (of thought, per-
ception, 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 com-
plement 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, caus-
ative 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 mark-
ers, 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.
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,
38–49.
Bril Isabelle & Rebuschi, Georges (eds). 2006. Coordination et subordination: typologie et modé-
lisation [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 coordina-
tion. Linguistic Inquiry 28(2): 195–217.
Diessel, Holger. 1999. The morphosyntax of demonstratives in synchrony and diachrony. Lin-
guistic Typology 3: 1–49.
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 cross-
linguistic typology, 1–48 [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), 115–143. Amster-
dam: John Benjamins.
Evans, Nicholas R.D. 2007. Insubordination. In Nikolaeva Irina (ed.), Finiteness, 366–431.
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):
429–492.
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 Interac-
tion [Typological Studies in Language 55]. Amsterdam: Benjamins.
Godard, Danièle & Abeillé, 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. 49–69.
Amsterdam/Philadelphia: Benjamins.
Haspelmath, Martin. 2004. Coordinating constructions: An overview. In Coordinating
Constructions [Typological Studies in Language 58], M. Haspelmath (ed.), 3–39.
Amsterdam: John Benjamins.
Haspelmath, Martin. 2007. Coordination. In Language Typology and Linguistic Description, Timothy
Shopen (ed.),1–51. Cambridge: CUP.
Haspelmath, Martin. 2007. Pre-established categories don’t exist: Consequences for language
description and typology. Linguistic Typology 11(1): 119–132.
Editor’s introduction
Haspelmath, Martin. & König, 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: 481–492.
Lehman, 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), 181–225. Amsterdam: John Benjamins.
Longacre, Robert. 1985. Sentences as combination of clauses. In Language Typology and Lin-
guistic Description, Timothy Shopen (ed.), 235–286. 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):
133–157.
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): 3–6 &
3(8): 3–9.
Rebuschi, Georges. 2003. Towards a theoretical treatment of left-dislocated subordi-
nate clauses. In P. Lafitteren sortzearen mende-mugako biltzarra, i: Gramatika gaiak
[coll. iker 14–1] Jesus Mari Makazaga & Beñat Oyharçabal (eds), 395–416. 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: 117–171.
——— 2005. La coordination et l’identité syntaxique des termes. In Godard, Danièle & Anne
Abeillé (eds.), La syntaxe de la coordination. Langages 160: 110–127.
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.), 237–300. Cam-
bridge: CUP.
van der Auwera, Johan. 1998. Adverbial Constructions in the Languages of Europe. Berlin: Mou-
ton de Gruyter.
Van Valin Robert. D. 1984. A typology of syntactic relations in clause linkage. Proceedings of the
Berkeley Linguistics Society 10, 542–558.
——— 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, Jerry M. 2002. Pseudo-subordination: A mismatch between syntax and
semantics. Journal of Linguistics 38(1): 87–111.
Isabelle Bril
Syntactic terminology
and typological methods
Clause linkage and Nexus
in Papuan languages
William A. Foley
University of Sydney
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 tra-
ditional 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 func-
tion as core arguments complements, and those which function as oblique constitu-
ents, 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 relation-
ship 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:
S S S S–S–S–S
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 con-
trast 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. tekekeôe:k šôa:pa-w ôe:-ta ke-yaše-w,
in.that.bush hide-imp and-sr 1sg.o-watch-imp
‘Hide in that bush and watch me!’
b. tekekeôe:k šôa:pa-ta ke-yaše-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
Clause linkage and Nexus in Papuan languages
clauses are linked by an explicit conjunction ôe:- “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 com-
mand 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 genera-
tive 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:
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 speci-
fied 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 nor-
mally 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 & König
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:
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 surpris-
ingly 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 coordina-
tion. 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 coordina-
tion. 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 inno-
vations, 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, adposi-
tion, etc. and the phrase types they project, NP, VP, PP, respectively, and functional
categories like I which do not typically correspond to independent lexemes, but are
Clause linkage and Nexus in Papuan languages
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] NP V
[Num: PL]
ADJ N
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 per-
son 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 be ungrammatical.
The syntactic category S is the odd man out in this framework. It is not endo-
centric like the other phrasal categories, but exocentric; in other words it lacks a pro-
jecting 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 verb’s 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 subordi-
nate 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, corre-
sponding 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 struc-
tures, as in Fore (Scott 1978):
c. a-egu-ôt-óô-ti w–a:n-ó
3sg.o-hit-np-1sg.a.sbd-all go-2sg-sq
‘Are you going to where I hit him?’
d. mi-nt-i ôkuma:ô-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.’
(6) NP
N TOP
IP [TNS: LIKELY]
S I
[TNS: LIKELY]
V
a-ka- … -ô i -ô ib -pa
3.-see … -3..
(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. Essen-
tially, 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 presum-
ably 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 ambigu-
ous 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 conjunc-
tions 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 complement’s 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 subor-
dinate 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 is typical:
(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.’
Clause linkage and Nexus in Papuan languages
IP [TNS: IRR]
S I
[TNS: IRR]
NP NP V
N CASE
S OBL
b. finite
ya-mpu-na-pay-kulanaŋ-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
S I
[TNS: PRS]
NP V
N CASE
S I [TNS: PRS]
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 non-
finite 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 possibili-
ties 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: 413–433) 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:
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:
Clause linkage and Nexus in Papuan languages
V V
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):
b. [CASE: ERG]
[CASE: TOP]
KP
KP [CASE: ERG] K
[CASE: TOP]
S K
[CASE: ERG]
XP V
The percolation of [CASE: ERG] from a lower KP node to the upper KP node is per-
missible: 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
tâp 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 didn’t 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 dif-
fusion 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:
(12) (= (11a))
[TNS: FP]
IP
S I [TNS: FP]
DP NP V
S I [TNS: FP]
NP ADV V
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 don’t think that John is the thief, in which the clause follow-
ing that is actually under the scope of negation again might suggest that that comple-
ment 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 say-
ing 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 fea-
tures like tense, mood or polarity from one clause to the other:
(14) IP
S I [TNS: NEG]
[POL: –]
DP
[TNS: FP]
[POL: +] IP D
S I [TNS: FP]
[POL: +]
NP NP V
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 I
[TNS: NEG]
S S S [POL: –]
NP NP V V V
(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 inflec-
tionally 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-analyz-
ing the former cosubordinate nexus as simply coordinate nexus. This was fore-
shadowed 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):
(17) a. (= (16a)) IP
[TNS: PRS] [TNS: PRS]
[POL: +] IP CONJ IP [POL: –]
S I [TNS: PST]
S S
NP V NP V
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 fea-
tures 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:
IP I
[IF: DECL]
[POL: +] IP [POL: –] IP
S I [POL: +] S I [POL: –]
NP V NP V
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 illocu-
tionary force, the features of the final I head. Watam is typical of this pattern; verbs in
medial clauses are marked for realis versus irrealis:
Clause linkage and Nexus in Papuan languages
The verbs of the medial clauses are marked ‑r realis when the tense of the whole sen-
tence, 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 simi-
larly. 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
I head:
NP V V
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
tense or 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:
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 is covert:
IP I
[IF: DECL]
[MOOD: R] IP IP [MOOD: IRR]
NP V NP NP V
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 lan-
guages. 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:
Clause linkage and Nexus in Papuan languages
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 spec-
ifications 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 sequen-
tial 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 imme-
diate 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 construc-
tions. 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 domi-
nating 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:
IP I [TNS: IP]
[TNS: PRS] IP IP
S I [TNS: PRS] S I
NP V NP V
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 Kâte example (Schneuker 1962):
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 illocution-
ary 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, regard-
less 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 presup-
posed, i.e. has a neutral, perhaps declarative illocutionary force, albeit not an asser-
tive 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 illocu-
tionary force of the main I head and hence the top dominating IP and the sentence
as a whole:
(27)
Clause linkage and Nexus in Papuan languages
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 Kâte (Suter 1992):
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 illocu-
tionary 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:
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
Abbreviations
References
Andrews, Avery. 2007. The major functions of the noun phrase. In Syntactic Description and
Language Typology, Timothy Shopen (ed.), 132–223. 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 lan-
guage 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.), 362–446. Cambridge: CUP.
Foley, W. & Van Valin Jr., Robert D. 1984. Functional Syntax and Universal Grammar. Cam-
bridge: CUP.
Franklin, Karl J. 1971. A grammar of Kewa, New Guinea. Pacific Linguistics C16.
Genetti, Carol. 2005. The participial construction of Dolakhā Newar: Syntactic implications of
an Asian converb. Studies in Language 29(1): 35–87.
Haiman, John. 1978. Conditionals are topics. Language 54: 564–589.
Haspelmath, Martin & König, 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: 1–32.
William A. Foley
Hoijer, Harry. 1949. Tonkawa syntactic suffixes and anaphoric particles. Southwestern Journal of
Anthropology 5: 37–55.
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), 301–318. 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: 363–401.
Schneuker, Carl L. 1962. Kâte 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 Kâte. MA dissertation, University of Zürich.
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
1. Introduction
*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 World’s Languages conference in Berlin, September 26, 2008, and as a guest lecture at
the University of Zürich, 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 (1947–2009).
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-ôe-b dana age gbo-ig-a fo? (‘chain’)
pig run.out-ds-3s man 3p hit-3p-t.pst q
‘Did the pig run out and did the men kill it?’
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 sub-
ordination, 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 Section 4, 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 addi-
tion 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
today’s past, v V gender (in Chechen), w.pst witnessed past, and y.pst yesterday’s past. Roman
numerals indicate noun classes.
. Data without a source specification are from my own fieldnotes.
Capturing particulars and universals in clause linkage
In these structures, the scope of the interrogative marker in the main clause (i), is inde-
terminate: 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 Section 3.3 on finiteness). The same pattern can also
be observed in the Indo-European (Indo-Aryan) language Nepali:
(4) Nepali
~
yahā ā-era khānā khā-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; Burúshaski:
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), Kâte (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 pos-
sibilities, 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 don’t 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 Lazard’s (2006) or Haspelmath’s (2007)
suggestions and keep issues of language-specific analysis free of comparative notions:
we could set up entities like ‘Amele ôe-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 construc-
tional entities, and it is one of the great steps of progress that typological and theo-
retical 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,
Capturing particulars and universals in clause linkage
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 alter-
native, 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 Section 2. In Section 3 I review some of the key struc-
tural properties that lead to the diversity in clause linkage noted above and in general.
Section 4 presents ways in which the diversity can be measured and discusses cross-
linguistic and possibly universal patterns emerging from this, based on a pilot database
of 69 constructions from 24 languages.
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 lan-
guages or within languages. I call such decompositions ‘multivariate analysis’, extend-
ing 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, practi-
cal 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
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 struc-
tures 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 Section 4, where the variables
are applied to a pilot database.
. 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.
Capturing particulars and universals in clause linkage
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 con-
strains 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 yuŋ-a=naa mundhupt-he i? (‘sub’)
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?’
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 –
I call 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? (‘but’)
pig run.away-3s-t.pst but man 3p hit-3p-t.pst q
‘The pig ran away but did the men kill it?’
In Tauya, local scope is associated with ‘topic’ clauses, more akin to adverbial subor-
dination in English:
(8) Tauya (MacDonald 1990)
nen mei momune-i-nani=ra pofei-ti nen=tu-e=nae? (‘topic’)
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?’
Balthasar Bickel
(9) Usan (Trans-New Guinea: Madang; Papua New Guinea, Reesink 1987)
munon iyau wârâm-a um-orei ôiyo? (‘chain.swr’)
man dog him.hit-3sDS die-3sREM.PST q
‘Did the man hit the dog and it died?’
A similar pattern has recently been noted by Molochieva (2008) for the Nakh-
Daghestanian 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:
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.
As the data in (11b–c) 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:
Burúshaski converb constructions do not constrain the scope of negation, and sen-
tences like the following can be understood with either conjunct or disjunct scope:
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 pos-
sibility, 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 don’t think that p which are regu-
larly 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 didn’t keep working until he came.’
(i.e. ‘he worked but not until he came.’)
b. yaŋ- his-sa la-ŋŋ-um-ô-ni. (‘cvb’)
around- look-cvb walk-[3nsS]neg-walk-npst-neg
‘He didn’t look around while walking.’
(i.e. ‘he walked without looking around’)
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-ŋ=lo khem-t-u-ŋ. (‘com’).
understand[sbjv]-3sP-1sA=com listen-npst-3sP-1sA
‘I’ll listen so that I understand.’
b. tupt-u-ŋ=lo pr%sta ka-lur-a! (‘com’)
understand[sbjv]-3sP-1sA=com clear 1sP-tell-sIMP
‘Talk to me clearly so that I understand!’
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 construc-
tions 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
ôe-structures illustrate conjunct tense scope (cf (1), which shows conjunct illocutionary
scope of the same construction):
(16) Amele (Roberts 1988)
a. ho busale-ôe-b dana age gbo-ig-a. (‘chain’)
pig run.out-ds-3s man 3p hit-3p-t.pst
‘The pig ran out and the men killed it.’
b. ho busale-ôe-b dana age gbo-gbag-an. (‘chain’)
pig run.out-ds-3s man 3p hit-3p-fut
‘The pig will run out (not: ran out) and the men will kill it.’
Capturing particulars and universals in clause linkage
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. (‘chain’)
house-loc 3nsA-reach[sbjv]-d-3sP seq chat-3nsS-chat[npst]-d
‘They will reach home and chat.’
or ‘When they reach home, they’ll chat.’
or ‘They reached home and now they will chat.’
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 under-
stand 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-әŋ Bya-sirә bәw-ra dBo-wa Bin. (‘chain’)
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.’
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. pollakhoû dḗ me epéskhe
often ptcl 1sACC stop.3sIMPERFECT
légo-nt-a metaxú. (‘part.coni.’)
talk-ipfv.act.ptcp-acc.s in.the.middle
‘[The oracle] has often stopped me when I was in the middle of talking.’
(Plat. Apol. 40b)
b. egṑ eréō hōs eû epistá-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 illocu-
tionary force and negation, viz. disjunct or transported scope, never seem to occur
with tense markers.
The sentence in (22a) illustrates the fact that main clause question markers have con-
junct 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! (‘chain’)
eat-imp seq go-imp
‘Eat and go!’
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 c’a-j-e’a-r=ii? (‘chain’)
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?’
b. *Maliika tyka-na=’a j-ax-na=ii c’a-j-e’a-r (‘chain’)
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’)
c. ahw toex-na-shehw j-axa-r=ii
2sg.erg hit-cvb-conc j-go-w.pst=q
Zaara tyka-na? (‘chain’)
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?’
d. *ahw toex-na-shehw=ii j-axa-ra
2sg.erg hit-cvb-conc=q j-go-w.pst
Zaara tyka-na? (‘chain’)
Z.(j).nom store-dat
Intended: ‘Did Zara go to the store, even though you hit her?’
But illocutionary marking is allowed in clauses headed by the temporal converb in -cha:
(25) Chechen (Molochieva 2008)
a. Muusa c’a v-e’a-cha, naan-na
m.(V).nom home V-come-when mother-dat
xaza xiiti-r=ii? (‘cvb.temp’)
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?’
b. Muusa c’a v-e’a-ch=ii, naan-na
m.(V).nom home V-come-when=q mother-dat
xaza xiiti-ra? (‘cvb.temp’)
beautiful seem-w.pst
‘Was (his) mother happy, when Musa came home?
(presupposing ‘mother was happy’)’
Capturing particulars and universals in clause linkage
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 (Trans-New Guinea: Kainantu-Gorokan; Haiman 1980: 421)
a. fu=mo d-mi-sa-ga-da u-gu-e. (‘ds-chain’)
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. (‘ds-chain’)
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.’
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 mark-
ing 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.’
(29) Korafe (Trans-New Guinea: Binanderean, Papua New Guinea; Farr 1999)
a. 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. (‘chain’)
give.1s-seq.realis.1sDS go.dur-f.pst.3sIND
‘I gave it and he went (two or more days ago)’
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 nomi-
nalizer was eventually reanalyzed as an illocutionary marker, with some kind of ‘assertive’
or ‘affirmative’ function.
Capturing particulars and universals in clause linkage
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
tyaŋ ‘from now on’ suggests future tense reference in the main clause.
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
In (32a), both conjuncts are finite and tense-marked, in (32b) both conjuncts are non-
finite 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 asym-
metrical 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 oth-
ers 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 construc-
tions for expressing alternating events. In this case, both conjuncts must include verb
forms from the same paradigm:
(33) Belhare
ŋka=na ten-he-ŋ=cha pheŋd-he-ŋ=cha. (‘and’)
1sNOM=top hit-pst-1sA=add drive.off-pst-1sA=add
‘I hit him and drove him off.’
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:
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 same-
subject 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 Chafe’s (1976) defini-
tion of a topic as “a spatial, temporal, or individual framework within which the main
predication holds.”
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 (MacDonald 1990)
a. 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 … (‘topic’)
who[nom] here come-3s-ass=top
Intended: ‘*if who did come here ...’
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 inde-
pendent 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 (Good 2003)
a. Maliika hu iec-na c’a j-e’a-ra? (‘chain’)
m.(j).nom what buy-cvb house j-come-w.pst
‘What did Malika buy and came home?’
b. mila c’a j-all-alc Ahwmad irs
who.nom home j-come-until a.nom happy
d-ol-ush v-a-ra? (‘cvb’)
d-be-cvb v-aux-wpst
‘Ahmed was happy until who came home?’
But the same constructions block extraction in the form of relative clauses:
(37) Chechen (Molochieva 2008)
a. Zaara koch ec-na c’a j-e’a-ra. (‘chain’)
Z.(j).nom dress.nom buy-cvb home j-come-w.pst
‘Zara bought a dress and came home.’
. 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).
Capturing particulars and universals in clause linkage
Example (38b) is a relative construction based on (38a), where the sole argument of
the dependent clause (fenaôa ‘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:
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) Kâte (Trans-New Guinea: Huon, PNG; Schneucker 1962)
ŋohe wena yu-ha-pire goŋgoŋ dâŋe-yeô. (‘chain’)
2d where be-sim-2dDS bell sound-3sn.pst
‘Where were you when the bell rang?’
(42) Belhare (Bickel 1993)
a. 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? (‘chain’)
how talk talk-2sNPST
‘Since you don’t know it either, how can you talk [about it] ?’
(43) Nepali (Clark 1963: 169)
tyo alchī-le ke garī-kana kamā-era khāncha? (‘chain’)
dem lazy.fellow-erg what do-cvb earn-cvb eat-3sNPST
‘How does that lazy fellow earn a living?’
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 orator a natura nisi
many.n.NOMp rel.n.acc orator.NOMs by nature.ABLs if.not
haberet] non multum
have.3sPST. sbjv neg much.n.ACCs
a magistro adiuvaretur] (Cic, de Or. 1, 126)
by teacher.ABLs help.3sPST.sbjv.pass
‘many [properties] such that if the orator didn’t have them by nature, he couldn’t
be much helped by a teacher.’ (Literally: ‘*many properties [whichi couldn’t be
helped by a teacher [if the orator didn’t have ti by nature]])’
Capturing particulars and universals in clause linkage
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 argu-
ments 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-ŋa,
2sPOSS-unmarried.female.agnatic-erg
doko-ep=phu chaŋ-da-he. (‘chain’)
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.’
b. u-tak-ŋa ya phekt-u=naa
3sPOSS-friend-erg call [3sA-]call-[sbjv]-3sP=top
ŋ-khatd-at-ni. (‘sub’)
neg-[3sS-]go-pst-neg
‘When his friend called him, he didn’t go.’
c. *ya m-phekt-u=naa, u-tak-ŋa,
call [3sA-]call[-sbjv]-3sP=top 3sPOSS-friend-erg
ŋ-khatdatni. (‘sub’)
neg-[3sS-]go-pst-neg
‘When they called him, his friend, that is, he didn’t go.’
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 indepen-
dent of the variable regulating extraction constraints.
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:
Capturing particulars and universals in clause linkage
While more research is needed to firmly establish the nature of =ŋa 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
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)
a. ho busale-ôe-b dana age gbo-ig-a. (‘chain’)
pig run.out-ds-3s man 3p hit-3p-t.pst
b. *dana age gbo-ig-a ho busale-ôe-b. (‘chain’)
man 3p hit-3p-t.pst pig run.out-ds-3s
‘The pig ran out and the men kill it.’
The same condition obtains for Usan clauses marked by the conjunction eng (cf. 34
above):
(52) Usan (Reesink 1987)
wau e-âb igo-iner eng unor
child cry-ss be-3suncertain.fut top mother
mâni u-t-i-b-â. (‘topic’)
yam 3sP-give-s-fut-3s
‘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 igâm-a igo-i urigerm-a end
rain night come.down be-3sDS be-cess light-3sDS because
irumban sig boru. (‘cause’)
mud very bad
‘Because it has been raining all night until daybreak, it is very muddy.’
b. irumban sig boru, ya âib dâr-a end. (‘cause’)
mud very bad rain big come.down-3sDS because
‘It is very muddy; that’s because a big rain came down.’
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-construc-
tions. 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-lleŋ
eat-inf=top [3sA-]eat-ipfv-npst-3sP ptcl where-dir
leŋ kina? (‘chain’)
[3sS-]direct seq
‘It (the cow) is eating, but after having turned towards which direction?’
Capturing particulars and universals in clause linkage
For many constructions, positional flexibility is limited, however, in that the depen-
dent clause must still be adjacent to the main clause. This is so in Belhare. By con-
trast, 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 (Good 2003)
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
c’a j-e’a-ra. (‘chain’)
home j-come-w.pst
b. Maliika tyka-na=’a j-agh-na c’a j-e’a-ra
m.nom store-dat=ss j-go-cvb home j-come-w.pst
zhejna=’a iec-na. (‘chain’)
book.nom=ss buy-cvb
c. Maliika c’a j-e’a-ra tyka-na=’a j-agh-na
m.nom home j-come-w.pst store-dat=ss j-go-cvb
zhejna=’a iec-na. (‘chain’)
book.nom=ss buy-cvb
‘Malika went to the store, bought a book, and came back home.’
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 sec-
ond clause). Such behavior is traditionally taken to be associated with ‘subordination’
and can often be observed with what translates English ‘adverbial clauses’.
(1988, 1990), KÖnig (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 con-
trast, 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 transpar-
ent 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. (‘com’)
D.[loc] [3sS-]stumble-ipfv-pstsbjv-com=foc [3sS]go-pst
‘He went to Dhankuta stumbling.’
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 comi-
tative 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 ver-
sion 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 construc-
tions, which block case assignments and therefore center-embedding:
(57) Belhare
a. u-chom pok=naa Dhankuta khaô-yu. (‘sub’)
3sPOSS-desire [3sS]rise[sbjv]=top D.[loc] [3sS]go-npst
b. *Dhankuta u-chom pok=naa khaô-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).
Capturing particulars and universals in clause linkage
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. (‘chain’)
seq [3sS]go-pst
b. [s%pp%i mai-mat-pir-he ki]
all[nom] 1s-[3sA]narrate-benef-pst seq
a-tak khar-e. (‘chain’)
1sPOSS-friend[nom] [3sS]go-pst
‘My friend told me everything and went off ’
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-ôe-b] gbo-i-ga. (‘chain’)
man 3p pig run.out-ds-3s hit-3p-t.pst
‘The men, the pig having run out, killed it.’
b. [ho busale-?e-b] [dana age gbo-ig-a]. (‘chain’)
pig run.out-ds-3s man 3p hit-3p-t.pst
‘The pig ran out and the men killed it.’
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 assign-
ment by this verb is blocked by the intervening dependent clause (ho busaleôeb ‘the pig
ran out and’), which itself has no extra role to assign. The sentence becomes grammati-
cal in (59b) where the dependent clause is not center-embedded but precedes the main
clause containing both dana age and the main verb.
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 diag-
nostic of clause linkage types), control and raising patterns (which is critical in dis-
tinguishing various complement clause types), and the presence of reference-tracking
devices such as switch-reference morphology, cross-clausal reflexivization or logo-
phoric 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 inter-
propositional 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
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 cross-
linguistic type clusters? (ii) Which variables are correlated, forming statistical impli-
cational 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.
. 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.
Capturing particulars and universals in clause linkage
discussion in Section 3, but I exclude negation scope because relevant data was not
available with sufficient analysis. I also use a simplified version of the finiteness vari-
able. 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-à-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.
4.2.1 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 dif-
ferent levels in all non-empty variables. For example, according to the pilot database
in the Appendix, the difference between the Amele ‘but’-construction and ‘chain’-
construction 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 struc-
tures may even be identical. For example, from all we know, Chantyal and Burúshaski
. Thanks to Jeff Good and Volker Gast for reminding me of this.
Capturing particulars and universals in clause linkage
0.167
B
0.334 0.084
D
D? 0.667 A 0.167 A
0.084 0.667
0.334
C 0.25
C
(a) Simple graph (b) Split graph
Figure 1. 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 Devel-
opment 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
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 Figure 2 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 Godié 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 bio-
logical 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.
Capturing particulars and universals in clause linkage
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’ con-
structions 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
Godie.topic English.fin.sub Fore.chain Newar..Dolakha..chain
Yup.ik..Central..sub Belhare.purp
Nepali.topic Belhare.cvb
Belhare.sub Nepali.cvb
English.and Lenakel.chain Chantyal.chain
German.sub Usan.chain.swr
Swahili.and Hua.ds.chain Burushaski.chain
English.although Chechen.cvb
German.and Hua.ss.chain
Russian.and Belhare.inconseq
Belhare.and Nepali.chain Kate.chain
Chechen.and Kate.serial
Amele.chain
Amele.but Turkish.chain
German.cause. Amele.or Swahili.and.INF
detached Tauya.topic Kate.TOP Swahili.chain
German.cond. Wambule.chain Godie.chain
detached Hua.alter.iter
Belhare.chain German.cvb Chechen.chain
Tauya.inconseq Greek..Ancient..part.coni.
English.go.and
Usan.cause Chechen.cvb.temp Turkish.serial
Amele.purp Hua.cond English.to.ing.detached Toura.chain
Korafe.chain Hua.inconseq German.purp
German.cvb.adV
Hua.topic Russian.cvb.detached
English.cvb.detached
Usan.topic English.cvb.adv
Belhare.com
Russian.cvb.adV
Swahili.topic
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 sur-
veyed 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 Figure 2. Thus, to the extent that there are
probabilistic clusters based on specific kinds of subordination and coordination,
as suggested by Figure 2, we can expect corresponding correlations of the proper-
ties defining these clusters.
The difference between traditional property bundles and probabilistic correlations
is the same as the difference between absolute and statistical (or ‘empirical’) implica-
tional 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. Proba-
bilistic 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 cor-
relations. One method for finding these would simply consist in performing statisti-
cal 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’
Capturing particulars and universals in clause linkage
. 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 π(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 ∈ {a1…k} 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.)
1…k
in the database, but this method (the ‘Maximum Likelihood’ method) is known to be unreli-
able 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
C → B we observe that π(A|C) < π(B|C). Then, C → A is an artifact of the stronger
implications and we set π(A|C) = 0, i.e. we effectively remove the link:
B B
A C A 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 clus-
ters, 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 et al.
(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 spe-
cific 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 residu-
als 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−ˆm
table to the χ2 sum of the table and is defined as ˆm where n is the observed count and μ̂ the
expected count.
Capturing particulars and universals in clause linkage
raint- onjunct
bann
free
Illocutionary force marking
c
Illocutionary scope
onic
sible nct
ok
Figure 3. 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 Figure 3 illustrate the method for two of the implications found, ‘ILL-
mark → extraction’ and ‘ILL-scope → WH’. Figure 3a 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 illus-
trating 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 extrac-
tion). In Figure 3a, no other residuals are significant. Figure 3b 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, constraint-
free scope is negatively associated with a ban on question formation, i.e. if there is
Balthasar Bickel
4.3.2 Results
Table 1 summarizes the findings on all correlations which have an MI-based probabil-
ity π(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 anal-
ysis, 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 Figure 3b), this is added in brackets.
Most of the findings in Table 1 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 convention-
alize 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 don’t know whether question formation is allowed or simply
not studied.
Capturing particulars and universals in clause linkage
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 finite-
ness (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 sen-
tence 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.
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
Capturing particulars and universals in clause linkage
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 study-
ing 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 ‘sub-
ordination’, which tends to associate disjunct illocutionary scope, local tense scope,
finite and symmetrical clauses, and flexible position (and, with slightly less prob-
ability, 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 con-
structions, 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 typol-
ogy closer to language-specific analysis. Instead of being separated enterprises (as
proposed by Lazard (2006) or Haspelmath (2007)), typological survey work is identi-
cal 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 the-
oretical 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 granular-
ity 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, cross-
linguistic prototypes and continua.
Balthasar Bickel
References
Bandelt, Hans-Jürgen & Andreas Dress, 1992. A canonical decomposition theory for metrics on
a finite set. Advances in Mathematics 92, 47–105.
Bearth, Thomas, 1986. L’articulation du temps et de [‘aspect dans le discours Toura. Bern: Lang.
Bickel, Balthasar, 1991. Typologische Grundlagen der Satzverkettung. Zürich: ASAS.
Bickel, Balthasar, 1993. Belhare subordination and the theory of topic. In Ebert, Karen H. (ed.)
Studies in clause linkage, 23–55. Zürich: ASAS.
Bickel, Balthasar, 1996. Aspect, mood, and time in Belhare. Zürich: ASAS.
Bickel, Balthasar, 1998. Converbs in cross-linguistic perspective [review article of Haspelmath and
König, eds., Converbs, Berlin: Mouton de Gruyter 1995]. Linguistic Typology 2, 381–397.
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
38–49.
Bickel, Balthasar, 1999b. Nominalization and focus constructions in some Kiranti languages.
In Yadava, Yogendra P. & Warren W. Glover (eds.) Topics in Nepalese linguistics, 271–296.
Kathmandu: Royal Nepal Academy.
Bickel, Balthasar, 2007. Typology in the 21st century: major current developments. Linguistic
Typology 11, 239–251.
Bickel, Balthasar, in press. Absolute and statistical universals. In Hogan, Patrick Colm (ed.)
The Cambridge Encyclopedia of the Language Sciences. Cambridge: Cambridge Univer-
sity 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, 26–27May 2002.
Nijmegen: MPI for Psycholinguistics.
Bryant, David & Vincent Moulton, 2004. Neighbor-Net: an agglomerative method for the con-
struction of phylogenetic networks. Molecular Biology and Evolution 21, 255–265.
Chafe, Wallace L., 1976. Givenness, contrastiveness, definiteness, subjects, topics, and point of
view. In Li, Charles N. (ed.) Subject and topic, 27–55. 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 typo-
logical study. Sprachtypologie und Universalienforschung 51, 3–42.
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 Wälchli (eds.) New challenges in typology: broadening the horizons and
redefining the foundations., 225–248. Berlin: Mouton de Gruyter.
Cysouw, Michael, 2008. Generalizing scales. In Malchukov, Andrej & Marc Richards (eds.)
Scales (Linguistische ArbeitsBerichte 86), 379–396. 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.
Capturing particulars and universals in clause linkage
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. Max-
Planck-Institute for Evolutionary Anthropology, Leipzig.
Genetti, Carol, 2005. The participial construction in Dolakha Newar: syntactic implications of
an Asian converb. Studies in Language 29, 35–87.
Good, Jeff, 2003. Clause combining in Chechen. Studies in Language 27, 113–170.
Green, Georgia M., 1976. Main clause phenomena in subordinate clauses. Language 52,
382–397.
Haiman, John, 1978. Conditionals are topics. Language 54, 564–589.
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, 73–95. 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, 1–55. Berlin:
Mouton de Gruyter.
Haspelmath, Martin, 2007. Pre-established categories don’t exist: consequences for language
description and typology. Linguistic Typology 11, 119–132.
Hausser, Jean & Korbinian Strimmer, 2009a. Entropy inference and the James-Stein estima-
tor, with application to nonlinear gene association networks. Journal of Machine Learning
Research 10, 1469–1484.
Hausser, Jean & Korbinian Strimmer, 2009b. entropy: entropy and mutual information esti-
mation. 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, 254–267.
Jacobs, Joachim, 1984. Funktionale Satzperspektive und Illokutionssemantik. Linguistische Berichte
91, 25–58.
Jacobs, Joachim, 1991. Focus ambiguities. Journal of Semantics 8, 1–36.
Johanson, Lars, 1995. On Turkic converb clauses. In Haspelmath, Martin & Ekkehard Konig
(eds.) Converbs in cross-linguistic perspective, 313–347. Berlin: Mouton de Gruyter.
Kaufman, Leonard & Peter J. Rousseeuw, 1990. Finding groups in data: an introduction to cluster
analysis. New York: Wiley.
König, Ekkehard, 1995. The meaning of converb constructions. In Haspelmath, Martin &
Ekkehard König (eds.) Converbs, 57–95. Berlin: Mouton de Gruyter.
König, Ekkehard & Johan van der Auwera, 1988. Clause integration in German and Dutch con-
ditionals, concessive conditionals, and concessives. In Haiman, John & Sandra A. Thompson
(eds.) Clause combining in grammar and discourse. Amsterdam: Benjamins.
König, 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, 337–355. Berlin: Mouton
de Gruyter.
Balthasar Bickel
Lakoff, George, 1984. Performative Subordinate Clauses. Proceeding of the 10th Annual Meeting
of the Berkeley Linguistics Society 472–480.
Lakoff, George, 1986. Frame-semantic control of the Coordinate Structure Constraint. Proceeding
of the 20th Annual Meeting of the Chicago Linguistics Society 152–167.
Lazard, Gilbert, 2006. La quête 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, 181–226. Amsterdam:
Benjamins.
Lynch, John, 1978. A grammar of Lenakel. Canberra: Pacific Linguistics (PL–B55).
MacDonald, Lorna, 1988. Subordination in Tauya. In Haiman, John & Sandra A. Thompson
(eds.) Clause combining in grammar and discourse, 227–246. 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, 157–164. 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, 1–48.
Meyer, David, Achim Zeileis, & Kurt Hornik, 2009a. vcd: visualizing categorical data. R pack-
age, 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 World’s Languages III conference, Berlin, September 25–28, 2008.
Noonan, Michael, 1999. Converbal constructions in Chantyal. In Yadava, Yogendra P. & Warren
G. Glover (eds.) Topics in Nepalese linguistics, 401–420. 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 Kâte-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. Yup’ik
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, 45–64.
Ross, John R., 1967. Constraints on variables in syntax. Ph.D. thesis, MIT, Cambridge, Mass.
Capturing particulars and universals in clause linkage
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. Master’s thesis, University of Zürich.
Takahashi, Hidemitsu, 2008. Imperatives in concessive clauses: compatibility between construc-
tions. 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,
487–528. Berlin: Mouton de Gruyter.
Van Valin, Robert D., Jr., 1995. Towards a functionalist account of so-called extraction con-
straints. In Devriendt, Betty, Louis Goossens, & Johan van der Auwera (eds.) Complex
structures: a functionalist perspective, 29–60. 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, 507–525.
Appendix: Pilot database on clause adjoining (NA = information not available)
Language Label Marker ILL-scope T-scope Finiteness ILL-mark T-mark Symmetry WH Extraction FOC Position Layer References and
Balthasar Bickel
examples
Amele but gba local local finite banned ok symmetrical NA NA NA fixed:pre-main ad-S Roberts (1987,
1988), (7),
(34c–d)
Amele chain -me, -ôV conjunct conjunct nonfinite banned banned asymmetrical NA NA NA fixed:pre-main ad-S Roberts (1987,
1988), (1), (16),
(20), (32b),
(51), (59)
Amele or fo ~o conjunct local finite ok ok symmetrical NA NA NA fixed:pre-main ad-S Roberts (1987,
1988), (32a)
Amele cond fi disjunct extensible any ok ok constraint-free ok NA NA flexible-relational ad-S Roberts (1987,
1988), (39)
Amele purp nu disjunct extensible any banned ok asymmetrical NA NA NA fixed:pre-main ad-S Roberts (1987,
1988)
Belhare and =cha...=cha constraint- local finite ok ok symmetrical NA banned banned flexible-adjacent ad-S Bickel (1993),
free (32)
Belhare chain ki(na)(huŋ) constraint- extensible finite harmonic ok asymmetrical ok possible ok flexible-adjacent ad-S Bickel (1993),
free (3a), (17), (23),
(42), (46a), (47),
(54), (58)
Belhare inconseq -kone local extensible nonfinite ok banned asymmetrical ok banned ok flexible-adjacent ad-S Bickel (1993)
Belhare sub -naa, -huŋ(go/ disjunct extensible finite banned ok constraint-free ok banned ok flexible-relational ad-S Bickel (1993),
do/cha), =cha (6), (40a),
(46b–c), (57)
Belhare cvb -sa constraint- conjunct nonfinite ok banned asymmetrical ok banned ok flexible-adjacent ad-V Bickel (1993),
free (3b), (14b), (40c)
Belhare purp -si constraint- conjunct nonfinite ok banned asymmetrical ok banned ok flexible-adjacent ad-V Bickel (1993)
free
Belhare com -lo(k) local conjunct finite banned harmonic asymmetrical ok banned ok flexible-relational ad-V Bickel (1993),
(14a), (26),
(40b), (56)
Bunúshaski chain n(V)-´-Σ-(i)n constraint- extensible nonfinite NA banned asymmetrical NA NA banned flexible-adjacent ad-S Tikkanen
free (1995), (13)
Chantyal chain -si, -rә NA extensible nonfinite NA banned asymmetrical NA NA NA flexible-adjacent ad-S Noonan (1999),
(18)
Chechen chain -na, -(u)sh, -i extensible conjunct nonfinite banned banned asymmetrical ok banned banned flexible-relational ad-S Molochieva
(2008), (10),
(24), (36a), (37),
(48), (55)
Chechen cvb -nach, -alc disjunct local nonfinite banned ok asymmetrical ok banned ok flexible-relational ad-S Molochieva
etc. (2008), (36b)
Chechen cvb.temp -lie, -cha disjunct conjunct nonfinite ok banned asymmetrical ok banned ok flexible-relational ad-S Molochieva
(2008), (25), (49)
Chechen and t’q’a local local finite ok ok symmetrical ok banned NA flexible-adjacent ad-S Molochieva
(2008)
English and and, but constraint- local finite ok ok symmetrical ok banned ok flexible-adjacent ad-S Ross (1967)
free
English go-and and conjunct conjunct finite harmonic ok symmetrical ok possible ok flexible-adjacent ad-S Lakoff (1986),
etc.
English fin. sub if, when etc. disjunct conjunct finite banned harmonic symmetrical banned banned ok flexible-relational ad-S
English cvb- -ing local conjunct nonfinite banned banned asymmetrical banned banned ok fixed:post-main detached
detached
English although although disjunct local nonfinite ok ok symmetrical banned banned ok flexible-relational ad-S
English to.ing. to, -ing local conjunct nonfinite banned banned asymmetrical banned banned ok fixed:pre-main detached
detached
English cvb-adv -ing disjunct conjunct nonfinite banned banned asymmetrical banned possible ok fixed:post-main ad-V
Fore chain -ki constraint- conjunct finite ok harmonic asymmetrical NA NA NA flexible-adjacent ad-S Scott (1978),
free (22)
German sub weil, wenn disjunct local finite banned ok constraint-free banned banned ok flexible-relational ad-S
German cause. weil local local finite ok ok symmetrical banned banned banned fixed:post-main detached
detached
German purp um zu, ohne zu disjunct conjunct nonfinite banned banned asymmetrical ok banned ok flexible-relational ad-V
German cvb-adv -nd disjunct conjunct nonfinite banned banned asymmetrical banned banned ok fixed:post-main ad-V
German and und constraint- local finite ok ok symmetrical ok banned ok flexible-adjacent detached
free
German cond. wenn local local finite ok ok symmetrical banned banned banned fixed:pre-main detached
detached
Capturing particulars and universals in clause linkage
Language Label Marker ILL-scope T-scope Finiteness ILL-mark T-mark Symmetry WH Extraction FOC Position Layer References and
examples
German cvb -nd disjunct conjunct nonfinite banned banned asymmetrical banned banned ok flexible-relational ad-S
Balthasar Bickel
Godié chain yi conjunct conjunct nonfinite banned banned asymmetrical NA NA NA flexible-adjacent ad-S Marchese (1977)
Godié topic n% disjunct local finite banned ok constraint-free NA NA NA flexible-relational ad-S Marchese (1977)
Greek part.coni. -men, -nt disjunct extensible nonfinite banned banned asymmetrical NA NA NA flexible-relational ad-S (19)
(Ancient)
Hua alter.iter -rohI, -rehI conjunct conjunct nonfinite banned banned symmetrical NA banned NA flexible-adjacent ad-S Haiman (1980),
(31)
Hua ds-chain -ga constraint- conjunct nonfinite banned harmonic asymmetrical ok NA NA fixed:pre-main ad-S Haiman (1980),
free (27)
Hua topic -ma local local nonfinite banned ok constraint-free banned NA NA fixed:pre-main ad-S Haiman (1980),
(28)
Hua inconseq -mana local local nonfinite banned banned asymmetrical banned NA NA fixed:pre-main ad-S Haiman (1980)
Hua cond -si local local nonfinite banned ok asymmetrical banned NA NA fixed:pre-main ad-S Haiman (1980)
Hua ss-chain Ø conjunct conjunct nonfinite banned banned asymmetrical ok NA NA fixed:pre-main ad-S Haiman (1980)
Korafe chain -o, -i, Ø constraint- local NA NA ok asymmetrical NA NA NA fixed:pre-main ad-S Farr (1999), (29)
free
Kâte chain -ra, -me, -ku, ... constraint- conjunct nonfinite banned banned asymmetrical ok NA NA fixed:pre-main ad-S Pilhofer (1933),
free Suter (1992),
(41)
Kâte serial Ø conjunct extensible nonfinite banned banned asymmetrical NA NA NA fixed:pre-main ad-S Pilhofer (1933),
Suter (1992)
Kâte topic mutsaô disjunct local finite NA ok asymmetrical ok NA NA NA ad-S Pilhofer (1933),
Suter (1992)
Lenakel chain m- conjunct local nonfinite banned ok asymmetrical NA NA NA flexible-adjacent ad-S Lynch (1978)
Nepali chain -era constraint- conjunct nonfinite banned ok asymmetrical ok NA ok flexible-adjacent ad-S fieldnotes, (4)
free
Nepali topic bhane disjunct extensible finite banned ok constraint-free ok NA ok flexible-relational ad-S fieldnotes
Nepali cvb -era constraint- conjunct nonfinite banned ok asymmetrical ok NA ok flexible-adjacent ad-V fieldnotes, (43)
free
Newar chain -en constraint- NA nonfinite NA banned asymmetrical NA NA ok flexible-adjacent ad-S Genetti (2005)
(Dolakha) free
Russian and i constraint- local finite ok ok symmetrical ok banned ok flexible-adjacent ad-S
free
Russian cvb-adv -a, -v disjunct conjunct nonfinite banned banned asymmetrical banned banned ok flexible-relational ad-V
Russian cvb. -a, -v disjunct conjunct nonfinite banned banned asymmetrical banned banned ok fixed:pre-main detached Rappaport
detached (1984)
Swahili chain ka- conjunct conjunct nonfinite banned banned asymmetrical NA NA NA fixed:post-main ad-S fleldnotes, (2)
Swahili and.inf na conjunct conjunct nonfinite banned banned asymmetrical NA NA NA fixed:post-main ad-S fleldnotes
Swahili and na conjunct local finite ok ok symmetrical NA NA NA flexible-adjacent ad-S fleldnotes
Swahili topic ki disjunct conjunct nonfinite banned banned asymmetrical NA NA ok flexible-relational ad-S fleldnotes
Tauya chain -pa, -te/-fe/ constraint- conjunct nonfinite ok harmonic asymmetrical ok possible NA flxed:pre-main ad-S MacDonald
-tefe free (1990), (5),
(37a), (40)
Tauya inconseq -na local local nonfinite ok ok asymmetrical NA banned NA fixed:pre-main ad-S MacDonald
(1990)
Tauya topic =ra local NA finite ok NA constraint-free banned banned NA fixed:pre-main ad-S MacDonald
(1990), (8),
(35b), (45)
Toura chain le conjunct conjunct nonfinite banned banned asymmetrical NA NA NA flexible-adjacent ad-S Bearth (1986)
Turkish serial Ø conjunct conjunct nonfinite banned banned asymmetrical NA NA NA fixed:pre-main ad-S fleldnotes
Turkish chain -Ip conjunct conjunct nonfinite ok banned asymmetrical NA NA ok fixed:pre-main ad-S Johanson (1995),
(11a), (21)
Usan chain.swr NA extensible conjunct nonfinite NA ok asymmetrical NA NA NA flexible-adjacent ad-S Reesink (1987),
(9)
Usan topic eng local local any banned ok constraint-free NA possible NA fixed:pre-main ad-S Reesink (1987),
(34a–b), (52)
Usan cause end disjunct NA any NA NA asymmetrical NA NA NA flexible-adjacent ad-S Reesink (1987),
(53)
Wambule chain -kho constraint- local finite NA ok asymmetrical NA NA NA flexible-adjacent ad-S Opgenort
free (2004), (30)
Yup’ik sub NA disjunct extensible nonfinite banned ok constraint-free NA NA NA flexible-relational ad-S Reed et al.
(Central) (1977)
Capturing particulars and universals in clause linkage
part ii
Denis Creissels
Université Lumière (Lyon2), Fédération Typologie et Universaux
linguistiques – CNRS
1. Introduction
Akhvakh (ašWalˉi micˉ’i, 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 š (voiceless) and ž (voiced);
palato-alveolar affricates č (voiceless) and Š̌ (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 Axaxdәrә near Zaqatala (Azerbaijan),
where I carried out fieldwork on Akhvakh. Axaxdәrә 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 Šamil’skij
Rajon of Daghestan.
The analysis of the Akhvakh specialized converbs proposed in this paper is based
on texts collected in Axaxdәrә 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 comple-
ment 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:
. Judging from Magomedbekova 1967 and Magomedova & Abdulaeva 2007, the variety
of Akhvakh spoken in Axaxdәrә 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 Axaxdәrә 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 Axaxdәrә, but in Axaxdәrә 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.
Specialized converbs and adverbial subordination in Axaxdәrә Akhvakh
. For a general discussion of the notion of converb, see Haspelmath (1995a), Nedjalkov
(1995), König (1995), Bisang (1995), van der Auwera (1998), Bickel (1998), Zúñiga (1998),
Haspelmath (1999), Ylikoski (2003).
Denis Creissels
–– ergative (‑de),
–– dative (‑La),
. The only exceptions to the semantic rule of class assignment are ãde ‘person’ and mik’e
‘child’, which in the singular trigger N agreement, whereas the corresponding plural forms ãdo
and mik’eli 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 Axaxdәrә, 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.
Specialized converbs and adverbial subordination in Axaxdәrә Akhvakh
–– genitive (Ø or -lˉi),8
–– comitative (‑k’ena),
–– 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 “NP1‑k’ena NP2”,
where ‑k’ena is the comitative case suffix (also used for comitative or instrumental
adjuncts), or “NP1‑la NP2‑la”, where ‑la is an additive particle (glossed add) also
found in contexts in which it corresponds to English ‘also’, ‘in turn’, or ‘even’. See
Section 3 for more details on this particle.
2.3 Adjectives
Like verbs (see Section 3.4.2), 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 bik’uruLa ‘be’.11
In Axaxdәrә 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 -lˉi 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 con-
figuration, 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, šaha-Sˉi-ga ‘to the town’
can optionally be pronounced šaha-Sˉ‑a).
. 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 bik’uruLa
‘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 ‑sˉu‑, F/N ‑Sˉi‑,
HPL ‑lo‑, NPL ‑le‑ ~ ‑li‑).
This list of independent synthetic verb forms calls for the following remarks:
a. The forms labeled ‘perfective₁’ and ‘perfective₂’ 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, pf₂ 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, pf₂ 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 pf₁ vs. pf₂ and ipf₁ vs. ipf₂ seem to be historically related, since
the pf₁ and ipf₁ suffixes share the ending ri, contrasting with da common to the
pf₂ and ipf₂ suffixes. Moreover, pf₂ and ipf₂ share the possibility of being used as
participles, whereas pf₁ and ipf₁ are strictly finite verb forms. However, function-
ally, the choice between ipf₁ and ipf₂ 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 ipf₂ as the marker of habitual aspect, whereas ipf₁ is used
in modal contexts where it is impossible to substitute ipf₂ for it. In AD Akhvakh,
ipf₂ is particularly common as a narrative tense (historical present), whereas ipf₁
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.
2.4.3 Participles
AD Akhvakh has four participles. Each one is characterized by a stem homonymous
with one of the independent verb forms listed above: perfective₂ ‑ada, perfective
negative ‑iLa, imperfective₂ ‑ida, and imperfective₂ negative ‑ika.
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:
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).
c. It combines with the copula or the verb bik’uruLa ‘be’ in auxiliary function to give
analytic tenses semantically similar to the English perfect, as in Example (3).
. 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.
Specialized converbs and adverbial subordination in Axaxdәrә Akhvakh
and subordination. In the literature, at least three different types of analyses can be
found for similar constructions:
In this respect, Akhvakh’s 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 con-
tradict the subordination analysis, whereas the texts I collected include many occur-
rences of the general converb in constructions in which the coreference relationships
between full NPs and pronouns are organized in a way commonly considered incom-
patible 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 Akhvakh’s
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 form’s 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
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 bik’uruLa ‘be’ in existential
function, or with one of the following three verbs: biχˉuruLa ‘remain’, hariguruLa ‘see’,
and mičunuLa ‘find’. Note that hariguruLa and mičunuLa 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. če ruša-Sˉi-ge L’a če tot’okiro b-ik’-awi
one tree-n-ess on.ess one woodpecker n-be-uw.n
[ruša-Sˉi-ge č’or-ere].
tree-n-ess knock-prog
‘On a tree there was a woodpecker knocking on the tree.’
b. huLe [lãga ũk-āno] išWada 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 mik’e [Ω-āre] m-ič-ani.
1sg-dat baby cry-prog* n-find-pf*
‘I found the baby crying.’
Specialized converbs and adverbial subordination in Axaxdәrә Akhvakh
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-sˉW-e Suda b-uqˉ’-ere godi.
dem-m-erg wood n-cut-prog cop₁.n
‘He is cutting firewood.’
b. hudu-sˉW-e Suda b-uqˉ’-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 bik’uruLa in existential predicate function. Note the following three differences
between the progressive tenses and the construction from which they originate:
I have no etymological hypothesis to put forward for this converb. Its syntactic distri-
bution could suggest a participial origin, but there does not seem to exist any concrete
evidence supporting this hypothesis.
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 lˉ’ũk’ideSˉi shows masculine singular agreement; the interpre-
tation 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 verb’s arguments.
In sentence (c), ek’WasˉWe is in the ergative case required by the main verb gWije bik’Wari
(< gūruLa ‘do’), and not in the absolute form that would point out its status as an argu-
ment of wuL’ideSˉi (< biL’uruLa ‘die’).
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).
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.
The locative converb is marked by the suffix ‑iSˉ‑ (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.
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, baker’s, 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 miša ‘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 parti-
cipial construction which constitutes its probable origin (b).
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 riSˉi ‘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 riSˉi is modified by a relative
clause with the verb in the perfective.
Denis Creissels
This analysis is supported by the fact that forms with the suffix ‑eSˉi ~ ‑adeSˉi ~ ‑ideSˉi
do not appear in Magomedova & Abdulaeva’s dictionary, which however includes
many examples of sentences in which the same meaning is carried by forms with
the suffixes ‑a(da)riSˉi and ‑idiriSˉi. These ‘suffixes’, in which riSˉi ‘at the moment’ is
still clearly recognizable, quite obviously represent a less advanced stage in the
grammaticalization process:
‑a(da)riSˉi → ‑eSˉi
‑ādariSˉi → ādeSˉi
ˉ
‑idiriS i → ‑idiSˉi
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-sˉW-a mašina b-iχˉW-iL-eSˉi] akˉ’i imaχa-ge
dem-m-dat car m-find-neg-post wife donkey-ess
d〈i〉k’-aj-e qˉ’elˉ-a j-eL-ari.
〈f〉sit-caus-cvb.f home-lat f-take.away-pf₁
‘As he didn’t find any car, he took his wife home on his donkey.’
b. [qˉ’ačali qˉ’elˉ-a b-eq’-iL-eSˉi], raši eχˉa r-olˉ-ari
dragon home-lat n-come-neg-post young.pl out npl-set.out-pf₁
ĩ-di-lˉi-da ila eqˉed-ōruLa.
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), mašina b-iχˉW-il-eSˉi is interpreted as ‘As he didn’t find any car …’,
but in a different context, mašina b‑iχˉW-il-eSˉi could equally mean ‘When one doesn’t
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 mašina b-iχˉW-iL-eSˉi ‘As I didn’t find any car …’
1sg-dat car m-find-neg-post
< di-La mašina b-iχˉW-iLa ri-Sˉi
1sg-dat car m-find-pf.neg moment-n.ess
lit. ‘At the moment when I found no car…’
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 riSˉi ‘at the moment’.
waša ušku-Sˉ-a
(28) a. w-oq’-ideSˉi
boy school-n-lat m-come-simult
‘While the boy was arriving at school …’
= waša ušku-Sˉ-a w-oq’-ida ri-Sˉi
boy school-n-lat m-come-ipf₂ moment-n.ess
b. ilˉ-e rilˉ’i qˉ’-ē̃deSˉi
1pli-erg meat eat-simult*
‘While we were eating the meat …’
= ilˉ-e rilˉ’i qˉ’-ē̃da ri-Sˉi
1pli-erg meat eat-ipf₂* moment-n.ess
Here again, this hypothesis is supported by the fact that the suffix ‑ideSˉi does not
appear in Magomedova & Abdulaeva’s dictionary, which however includes many
examples of sentences in which the same meaning is carried by verb forms with the
‘suffix’ ‑idariSˉi.
An apparently synonymous converb formed by means of a suffix ‑adada spo-
radically occurs in some of my texts – Example (29), but most of my informants do
not use it productively.
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.
The probable origin of this converb is again the contraction of a synonymous con-
struction involving an inflected form of ri ‘moment’ modified by a relative clause,
as indicated in (31).
‑ik’ena probably contains a second formative cognate with the suffix of the comitative
case ‑k’ena, 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 anal-
ysis is supported by the semantic proximity between the comitative suffix and the
additive particle.
Denis Creissels
The immediate converb of mūnuSa ‘go’, whose literal meaning is ‘just before going’,
has postposition-like uses with the meaning ‘before a particular length of time has
passed’ (‘within’).
Specialized converbs and adverbial subordination in Axaxdәrә Akhvakh
Etymologically, this is probably a complex suffix with the imperfective suffix ‑ida as
its first formative. Consequently, a possible origin for V‑idaSˉa is the contraction of a
sequence V‑ida ri‑Sˉ‑a, with the imperfective participle V‑ida modifying the allative
form of ri ‘moment’.
As for ‑idaSˉoqˉe, it probably results from the addition of the similative particle ‑oqˉe
‘like, as’ to ‑idaSˉa.
The non-posterior converb of muqˉ’unula ‘close (an eye)’ occurs in the idiom če k’eha
muqˉ’ileda 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
[hu-ji
(40) a. baza-ga m-aô-ij-ala], dene-la w-ī̃da gWida.
dem-hpl market-lat hpl-go-hpl-cond 1sg-add m-go.ipf₂* cop₂.m
‘If they go to the market, I will go too.’
b. [ha req’a beča-La ãL-ala], maSˉeq’-ide.
dem word.pl mountain-dat hear-cond be.vexed-ipf₂.n
‘If the mountain hears these words, it will be vexed.’
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.
Example (43) illustrates the contrast between conditional sentences referring to a con-
dition that may come true in the future, and a construction in which the conditional
perfect (conditional converb of bik’urula ‘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.
More generally, analytic forms of the conditional converb are used to express aspectual
shades of meaning. Note that, in the conditional, the use of mičunula ‘find’, ‘be found’,
‘be’ in auxiliary function (rather than bik’urula ‘be’) is particularly common.
(45) a. [w-ū̃xˉ-o w-ũč-ala], u-t-a w-ūnuLa.
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-ēro w-ũč-ala], w-ū̃balˉ’a!
dem-m be.afraid-prog.m* m-be-cond* m-go.opt.neg*
‘If he is (being) afraid, let him not go!’
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) heresˉi-k’ena di-La šahidiSila gWij-a, du-La ač’ada
lie-com 1sg-dat testimony do-imp 2sg-dat ten
turuši o-xˉ-uwa.
rouble n-give-pot.n
‘Give a false testimony for me, and I’ll 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 ‑ãčala, it seems reasonable to assume that it results from the
contraction of an analytic form containing mĩčala, a conditional converb of mĩčunula
‘find, be found, be’ in auxiliary function.
[jaše j-eq’-iL-ala],
(49) a. me-de čWi gW-īda gWe-da?
girl f-come-neg-cond 2sg-erg what do-ipf₂* cop₂.n
‘If the girl does not come, what will you do?’
b. [hu-be di Ωaqˉ’ilo-Sˉi-ga b-eq’-e m-ič-iL-ala],
dem-n 1sg(gen) mind-n-lat n-come-cvb.n n-be-neg-cond*
hu-sˉu Ωaqˉ’ilo-Sˉi-ga-la b-eq’-ike.
dem-m(gen) mind-n-lat-add n-come-ipf₂.neg.n
‘If it did not come to my mind, it will not come to his mind either.’
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 ču-bi, [di raqˉ’olˉi b-olˉ’-ala b-olˉ’-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?’
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 & Abdulaeva’s
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 ‑ero‑tana, 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.
This converb can be translated as ‘in the same way as …’. It is characterized by the suffix
‑eroqˉe ~ ‑ereqˉe, glossed simil.
Etymologically, the suffix of the similative converb quite obviously contains the
similative particle ‑oqˉe ‘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.
This converb, marked by the suffix ‑ũda ~ ‑ũdaSe (glossed grad), expresses the
meaning rendered in English as ‘in proportion to …’, ‘the more …, the more …’.
Morphologically, the explicative converb can be identified as the elative form of the ver-
bal 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 ‑Sˉi‑, whereas the suffix of
the explicative converb is never found with ‑Sˉi‑ 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 Section 11.2), whereas
the verbal noun only has an analytic negative form.
Denis Creissels
[me-de ači
(56) a. o-xˉ-iL-erogu] Ωazi b-eqˉ-ere godi.
2sg-erg money n-give-neg-explic complaint n-take.off-prog cop₁.n
‘He is complaining that you do not give him money.’
b. hu-Sˉi-tana recˉi gW-ēre gWeda, [imaχa-ge L’a
dem-n-purp thanks do-prog* cop₂.n donkey-ess on.ess
w-uk’-iL-erogu [b-uq-ideSˉi]].
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.
. For a general discussion of the possible historical connections between purposive and
infinitive, see Haspelmath 1989.
Specialized converbs and adverbial subordination in Axaxdәrә Akhvakh
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 syn-
chronic 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.
In such constructions, lˉ’e 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 lˉ’e can be grouped together with
other complementizer-like uses of the same converb, either in its full form elˉ’e or in its
reduced form lˉ’e – Example (61).
Consequently, in the present state of the system, there are arguments for describing
this use of lˉ’e as a particular aspect of the infinitive’s syntax. But the same facts could
equally be described by positing a second purposive converb characterized by the
complex suffix ‑urula‑lˉ’e.
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
Specialized converbs and adverbial subordination in Axaxdәrә Akhvakh
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 elˉ’urula ‘say’ is gram-
maticalizing 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
Annex
. ṽ represents the unspecified nasal vowel which constitutes one of the allomorphs of the
root of mūnula ‘go’.
Specialized converbs and adverbial subordination in Axaxdәrә Akhvakh
(39) qˉ’-ē̃leda < |qˉ’ã(b)-ileda|, w-uš-a < |w-iš-a|, w-ī̃leda < |w-ṽ-ileda|, rãcˉ’-ini <
|rãcˉ’-iri|
(40) w-ī̃da < |w-ṽ-ida|
(41) m-ič-ala < |b-ĩč-ala|, w-ũč-ala < |w-ĩč-ala|, t’-ō̃š-awa < |t’ã(b)-uš-awa|
(42) t’-ē̃da < |t’ã(b)-ida|
(43) gW-īda < |gWi(j)-ida|
(44) q-āda-sˉW-a < |qa(b)-ada-sˉu-ga|
(45) w-ū̃xˉ-o < |w-ṽ-uxˉ-o|, w-ũč-ala < |w-ĩč-ala|, w-ūnula < |w-ṽ-urula|, l-ēro <
|li(b)-ero|, w-ū̃balˉ’a < |w-ṽ-ubalˉ’a|
(46) w-uk’-ala < |w-ik’W-ala|, w-uk’-iko < |w-ik’W-ika-we|
(47) w-uk’-ãčala < |w-ik’W-ãčala|, w-uχˉ-ada < |w-iχˉW-ada|
(48) gW-īda < |gWi(j)-ida|, m-ič-il-ala < |b-ĩč-il-ala|
(49) lib-ēl-ala < |li(b)-a(j)-il-ala|
(52) w-ošqˉ-alotola < |w-ešqˉ-alotola|
(53) gW-ēreqˉe < |gWi(j)-ereqˉe|, q-āreqˉe < |qa(b)-ereqˉe|, q-āda < |qa(b)-ada|
(54) lˉ’-ū̃daSe < |lˉ’i(b)-ũdaSe|, w-ī̃da < |w-ṽ-ida|
(55) w-uk’-eregu < |w-ik’W-eregu|, qˉ’-āne < |qˉ’ã(b)-ere|, l-ēro < |li(b)-ero|, ẽSˉ-enogu
< |ẽSˉ‑erogu|
(56) w-uk’-il-eregu < |w-ik’W-il-eregu|, gW-ēre < |gWi(j)-ere|
(58) t’-ēni < |t’ã(b)-iri|, zor-ēda < |zor-a(j)-ida|
References
Gilles Authier
Institut National des Langues et Civilisations Orientales
1. Introduction
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 grat-
itude to Timur Maisak, Yura Lander, Lena Kalinina, and to Isabelle Bril for very helpful com-
ments 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 ‘intro-
flected’, 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 asso-
ciated 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
order in Budugh. The third and fourth sections present the prosody of imperfective
verb stems determined by the syntactic domains in which they are found, show-
ing the relationship between segmentally homonymous forms and the correlation
of stress distinctions with syntactic function. Sections five and six propose a clas-
sification of perfective stems and illustrate the dependent and independent uses of
these forms.
The basic frame of any Budugh lexical verb stem consists of two or three original ele-
ments: one or two preverbs, and a root. Verb stems can also display:
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 depen-
dency, they will not be addressed here, cf. Authier (2009, to appear).
Finite and non-finite
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 affect-
ing 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 agree-
ment: 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), inani-
mate (n for ‘neuter’), non-human plural (npl), and human plural (hpl). On the major-
ity 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) ma’lla’-cır lem ğùvotu-ri (2) ma’lla’-cır lemér ğùtu-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
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 üzöbülq’üla’vini can be analysed as follows:
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 speaker’s mental dictionary; then the
Tense-Mood ending, which is produced with minimal allomorphy, is identified.
Affixation in Budugh is thus of two types:
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, -ö) or a sonorant immedi-
ately 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 aspec-
tual derivation involving a change of preverbs is the verb ‘be’:
‘Be’ pf ipf
m/n/npl sa-xh-a yi-xh-ar
f sa-rxh-a yi-r-xh-ar
a soxho < *sa-v-xh-a yuxhor < *yi-v-xh-ar
hpl sabka < *sa-b-xh-a yıbkar < *yi-b-xh-ar
Finite and non-finite
‘Go’ pf ipf
m vi-xhi ça-ğar
f vi-r-xhi ça-r-ğar
a vüxhü < *vi-v-xhi çoğor < *ça-v-ğar
n/npl vidki < *vi-d-xhi ça-ğ-ar
hpl vibki < *vi-b-xhi ça-b-ğar
3. Word order, parts of speech and stress position and nature
Budugh clearly has distinct word-classes, associated with specific inflectional mor-
pheme 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. Substan-
tive 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:
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
genuine nouns.
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 mallá’-cır ğìvataci ‘It was the mullah himself who beat his daughter’
(here mallá’ 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 syn-
tactic attributes, arguments, and adjectives named respectively participles (perfective &
imperfective), verbal nouns (perfective infinitives vs. imperfective masdars)12 and con-
verbs – 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 inde-
pendent (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.
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 morphologi-
cally imperfective, while its masdar is morphologically perfective, cf. Haspelmath 1994.
Finite and non-finite
All imperfective stems have thus two accentual forms and syntactic interpretations:
‘slay’
m/n(pl)/A/hpl sorkú sòrku
‘hang’ tr. soborkú sòborku
m/n(pl) Ωanxú Ωànxu
f Ωanxú Ωànx-u
a Ωonxú Ωònxu
‘hang’ (anticausative)
m/n(pl) Ωanxán Ωànxan
f Ωanxán Ωànxan
a Ωonxón Ωònxon
‘go’
m/n/npl çağár çàğar
f çağár çàrğar
a çoğór çòğor
hpl çabğár çàbğar
‘carry’ (causative of ‘go’)
m çiğí çìği
f çirğí çìrği
a çüğǘ çǘğǘr
hpl çibğí çìbği
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.
13. For the relationship between participles and masdars in the related language Kryz, see
Authier (2007).
Gilles Authier
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 dağ-ra’ q’ala’ ğöç’ǘ başarmi sì’i-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. With ‘want’ and ‘know (how)’ the Masdar has an obligatorily coreferent subject which, if
expressed, is in the Dative case -z.
Finite and non-finite
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 çalam ‘boot’ which is Animate):
(19) u-n-uz suncu pineçi-ri áq’ul-nu çalam serví 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: cobbler’s sitting…).’
With verbs meaning ‘stop’ or ‘finish’, the matrix verb, as a rule, agrees with the Absolutive
argument of the Masdar:
(28) Ωayel-cir gita’ra’ çorot’ú Ωàvaqa-ci.
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’ araxhár laħki-vni.
1.in behind m.come.ipf.prt many-cop.pst
‘Many were courting me.’
(39) vez irqí za’ kıl-imer da-d.
5.dat see.ipf.prt 1.in arm-pl cop.neg-n
‘What you see are not my arms.’
(40) filan şeher-ce sa-r aza’n yuts’ú -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, head-
less participle constructions are used for complementation with the verb ‘see’. The
controlling verb takes gender agreement with the Absolutive participant of the sub-
ordinate clause:
(41) gada-z (ug-unda’) ada Ğuba-ca çağár ìrqa-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) padşah-cır ug-o rij-eri dogru yu’ú ìdxa-cı.
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).
16. A similar finite deliberative use of a formally non-finite form is Spanish A ver se… ‘Let’s
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.
Finite and non-finite
(49) vın gerek yıxta’ cıga’-ca çàğar, har cıga’ va-z ìrqi.
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:
A finite debitive form can also be subordinated: By means of the complementiser ki,
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 root-
consonant 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.
‘take’
m/n/npl ğèyin ğèyin-e ğéyin ğeyín ğeyin-í
f ğèrin etc. ğérin ğerín etc.
a ğèvin ğévin ğevín
hpl ğèbin ğébin ğebín
18. There is also a ‘long’ infinitive, bearing the dative marker -z.
Finite and non-finite
The Thematic perfective stems are used as (1) optatives, (2) short infinitives, (3) parti-
ciples, (4) sequential converbs.
These irregular and labialised-root (or ‘third class’/‘strong’) verb stems are used as
(1) short infinitives, (2) sequential converbs, (3) participles.
The short infinitive tends to form a single phonological word when used in very com-
mon constructions with the verb ‘be (able)’ immediately following, the subject being
expressed in the Adelative case:
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 accusa-
tive alignment, or Experiencer), is most frequently controlled by the valency of the
sequential verb form:
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:
Note there is an alternate, suffixed form with -nV which seems to be used when non-
shared arguments are inserted between the two verb forms:
The ‘Augmented’ stems must add the low vowel to their bare perfective stem in order
to express optative modality:
‘Radical’ verb stems form their optative by lowering the high vowel of the verb stem
(u > o, ü > ö, i > e):
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 dis-
criminate 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) hàzebtir ğábç’u sòğo!
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 subordination-
marking 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 depen-
dent 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, con-
version by accent pattern shift is a productive feature in the verbal paradigm, although
the language’s grammar is otherwise characterised by rich inflectional morphology.
Fairly strict word order rules, along with specific prosody associated with purely syn-
tactic 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 Daghesta-
nian languages, is very scarce. Indeed, such a ‘syntactic stress’ system may well be an
inherited feature.
Abbreviations
References
Christiane Pilot-Raichoor
LACITO (Laboratoire des langues et Civilisations à Tradition Orale),
Fédération Typologie et Universaux Linguistiques, CNRS
1. Introduction
*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 200 000 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: 53–54).
Christiane Pilot-Raichoor
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 dis-
tinction elaborated by I. and V. Nedjalkov in the 1990s (1995, 1998) and used by many
authors in typology, notably Haspelmath and König (1995).
Sections 2 to 4 describe the morphological and syntactic properties of these strat-
egies. 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 proper-
ties. Paragraph 2.2. presents the various specialized converbs based on their relevant
semantic domain. Section 3 will focus on strategies B and C, both based on an adjec-
tival 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.
Section 4 will briefly show that the uses of the quotative, derived from the verb ‘say’,
basically stem from the previous morphosyntactic strategies. Section 5 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 characteris-
tics of the verbal clause in Badaga should be briefly mentioned. First, two important
Converbs and adverbial clauses in Badaga, a South-Dravidian language
roperties should be highlighted. Whatever the actual categorial status (finite, non-
p
finite, nominalized) of the verb heading a clause:
Secondly, a brief account of the basic verb forms involved in the construction of com-
plex 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 α for class 1 verbs, the most
numerous and regular, and type β for class 2-3-4 verbs. The following table with the
verb mād.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 ref-
erence 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. Table 1 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 in-
dicate 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(“Non-
Past”) + 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.).
Table 1. Main forms of the verb mād.u “to do”
Christiane Pilot-Raichoor
Stems S2 (“Past”) mād.-i(d)- S1 (“Non-Past”) mād.-uv- Negative mād.-ād- Bare stem mād.-
Polarity Positive Negative Unspecified
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 infini-
tive stands apart as it is nearly exclusively used as a component [verb root+a_Aux] of
modal auxiliary forms.
The notion of ‘converb’ as “a nonfinite verb form whose main function is to mark
adverbial subordination” (1995: 3) was firmly established in Haspelmath & König’s
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 con-
verbs in other language families, notably Turkic languages (Johanson) and Burushaski
(Tikkanen), as described in Haspelmath & König 1995. In Indian linguistics, converbs
were identified as an areal feature by Masica (1976: 108–140), and Subrahmanyam
(2006) gives a more recent account of their functions.
In Badaga, as in other Dravidian languages, some constraints on and charac-
teristic features of the construction of complex sentences perfectly fit the typo-
logical set up favoring the development of converbal strategies: (i) only one main
Christiane Pilot-Raichoor
verbal predicate per sentence and (ii) a general absence of coordinative conjunc-
tions for clauses.
More precisely, in Badaga, it follows from (i) that, except for the main, morpho-
logically ‘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 (partici-
ples) 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.
compound verbs.8 Alongside these predicative uses, the Pc form is a common deriva-
tive 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, concern-
ing the various levels of their uses.11
Moreover, as stressed by Bisang (1995) for Tamil converbs, the Badaga Pc con-
verb, 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.
8. Ex. ottu bā (carry.Pc_come) ‘to bring; kan.d.(u)-id.i (see.Pc catch) ‘to find’; bītu nōd.u (keep.
Pc see) ‘to compare’.
9. Ex. tirigi ‘again’ < Pc of tirigu ‘to turn’. The Pc of ‘to become’ -āgi is the most common
derivative for adverbs: sandōsa ‘happiness’ > sandōsavāgi ‘happily’, neja ‘true’ > nejavāgi ‘really’.
10. The Pc of āgu ‘to become’ is added to the dative case suffix -ga and grammaticalized into
a postposition -gāgi 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 (aspec-
tual) auxiliaries carry the inflectional markers.
Christiane Pilot-Raichoor
The Pc converb can be used in any temporal context,12 past (1), present (2) or
future (3):
(1) Jōgi 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) Bēla Māda avasara hōgi hat.t.iga kad.edu banna.
Bela Mada haste go-pc village-dat leave-pc come:T1:3m
‘Bēla Māda hurries up and leaves for the village.’
(3) [nanga hosa rāja ettuvo]. dēsa pūra.ga tandura ūdu
we new king takeT3.1pi country whole.dat drum beat-pc
kaida […] ākuvo.
notice put.T3.1pi
‘[We will select a new king.] Throughout the country we’ll beat the drums
and put up notices.’
The Pc converb can be linked both to modal verbs (4) and to nominal predicates (5):
(4) hāv-ē, nī 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)
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 sen-
tence’ (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 com-
mon 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. ā jōli.ya kētu Sevana āciriya pat.t.a.
that news.obl hear.pc Sevana surprise feel.T2.3
b. Sevana ā jōli.ya kētu āciriya pat.t.a.
Sevana that news.obl hear.pc surprise feel.T2.3
‘Hearing that news Sevana felt surprised.’
(9) ā jōli.ya kētu Sevana.ga āciriya ātu.
that news.obl hear.pc Sevana.dat surprise become.T2.3n
/*Sevanaga ā jōliya kētu…
‘Hearing that news, Sevana became surprised.’
On the whole, subject – or maybe more exactly topic – continuity across clauses is pre-
ferred, 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
Christiane Pilot-Raichoor
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.
In some cases, a nominal clause may even follow a converbal agentive clause as in (11):
In sentence (12), the change of subjects between ava ‘she, the demoness’ in the first
clause, ava and Kariabetta in the second clause (sande ād.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.
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 alterna-
tion 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).
Christiane Pilot-Raichoor
The sentence “sōpa āki nīr at.t.ōdu 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 (āki) 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) kūsa kattara.vā ille.yā 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!’
~ ~
(17) ēnagū ama hā embadu ille hū embadu ille.
See below, Examples (28) and (29) for other instances of disjunction in dependent
clauses.
However, as shown in Table 1, 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 inte-
grative 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 barāde ibbad.enda…
he come.PNEG be.N1z.abl. [= Neg. Perfect]
‘Because of the fact that he did not come…’
of meanings (see Thompson, Longacre & Shin Ja Hwang 2007: 255–262 for a typologi-
cal 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 āsaga mād.ile gana.
he so.much do.Hyp enough
‘If he does that much, it is enough.’
(27) ama bandale nanga hat.t.iga hōgāku.
The same marker is used to express alternate possibilities. It should be noted here that
the clitic14 -ū, which usually cannot coordinate clauses, is grammatical in this condi-
tional disjunctive use:
(28) idu ninaga sēdaleyū enaga sēdaleyū ēna?
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: -āde) followed
by the independent form of the conditional participle (āle).
(29) avakaga hid.atāleyū hidiyāde āleyū
. nā enna
they.dat like.pc.Hyp-foc like.Pneg Hyp-foc I me.obl
gelasava gīdane.
work.obl do.T1.1
‘Whether they like it or not, I will do my work.’
14. The clitic -ū, and its equivalents in other Dravidian languages, is a very polysemous
marker. Emeneau (1980: 199–200) identified five distinct meanings: ‘also’, ‘and’ (when repeated),
‘even’, ‘all’, ‘any’ (any element of the whole) which can be grouped under additive [add], total-
izing and indefinite meanings [any], all with a shade of emphasis dominant in its focalizing
meaning [foc]: ‘also’ and ‘even’.
Converbs and adverbial clauses in Badaga, a South-Dravidian language
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.
In Example (36), the combination of the Perfect aspect with the conditional converbal
form conveys irrealis meaning.
(34) modalu nanga Todanād.u ayya Tūnēriya ān.d.iddu, …
first our Todanad grand-father Tuneri.obl rule.Ppf
‘Formerly, our Todanad ancestor had been ruling Tuneri and…’
(35) āga had.udun.d.ibbane, innu jāmanōge iva ēna 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) nī ninne bandidale amana nōd.irāku.
you yesterday come.pc.Pf.Hyp him.obl see.pc.Pf.Pot
‘If you had come yesterday, you would have seen him.’
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. Subrahmanyam’s remark (paragraph 2.1.1.2) that, contrary to a Pc form which is
Converbs and adverbial clauses in Badaga, a South-Dravidian language
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: 330–331) 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.
Christiane Pilot-Raichoor
predicate. In both cases, the verbal forms are derived from the basic morphological
subsystem of “Past”, “Non-Past” and Negative stems as displayed in Table 1.
16. Similar nominalized forms of the verb can be made with 3rd pers. pronominal deriva-
tives: -ma ‘3m’ bapp.a-ma ‘the man who comes’, barād.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’, barād.a-du the fact of not coming’, band.a-du ‘the fact of having come’.
Converbs and adverbial clauses in Badaga, a South-Dravidian language
In (40), the narrator expresses the idea that the traditional Hindu way of honoring
the gods is not followed by the Badagas:
we 1 1
coconut break.N z-dub other ritual do.N z-dub. flower
ākōd-ō ondu.n-ū 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 straight-
forward. Here the affirmative substantivized forms of the verbs are apposed to ondu.n-ū
(one-any), the subject of the negative existence verb (ille). They are chained with the
use of the clitic -ō (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’:
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 under-
stand 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 adjec-
tives 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
Christiane Pilot-Raichoor
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 adjec-
tives accept a wider range of lexical heads. Notably, they can be syntactically depen-
dent 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 māke ‘as, like’:
(44) nā tappa porul.a hottōgi, nā ēgida māke mād.ire-yā?
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 mēle ‘after happiness’ [happiness.obl on/after]
samudara kere.gu mēle 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 predi-
cate is a Padj, see for instance, ex. (100): appanella satta mēle ‘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’; ī keppu… ‘this red one…’ vs. ī dod.d.adu… ‘this big one…’) and idu1
āpat.t.i2 keppu3 ille4 ‘this one1 is not4 so2 red3’ vs. idu1 āpat.t.i2 dod.d.adu3 ille4 ‘this one1 is not4
so2 big3’.
18. The lexeme mēle, 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: mēle
iru ! ‘Stay upstairs! , noun modifier: mēle kade ‘the above side’, sentence localizing adverb: āga1,
mēle2, obba3 …enda4’ ‘then1, upstairs2, someone3 said4 …’, which can also be case marked in a
localizing adverbial phrase: mēlanda ‘from above (cf. 46) or postposition: pustaga1 pet.t.i2 mēle3
hadade4 ‘The book1 is4 on3 the table2’ (see also 58).
.
Converbs and adverbial clauses in Badaga, a South-Dravidian language
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):
bāgu1 āda2 mēle ‘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 āpadu.gu2 mundād.u3 ‘Before3 darkness1 has come2…’ [become.N1z.dat]
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): ēgidēnge ‘as I said’ [tell.P2adj.like],
ex. (91): ēgōnēnge [tell.P1adv.like], as well as the use of a nominalized form with a case
suffix ex. (90): aradōdunōge [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 hōi [go.Pc] inside the relative clause, whose regular predicate is an adjec-
tival participle (sēda) and the head an adverb (indād.u ‘behind, after’):
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).
Christiane Pilot-Raichoor
Various postpositions can also mark the dependent clause, expressing for instance a
spatial or temporal localisation, mēle ‘upon/after’ or mundād.u ‘in front of/before’:
The quotative markers derived from the verb ennu ‘to say’ offer another powerful
device to construct complex sentences. They appear with various morphological spec-
ifications: 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 ēgu ‘to say’, nenacu ‘to think’, kē ‘to ask’, ari ‘to
know’, but not with the verb ennu itself (cf. ex. (32)).
As shown above, embedding with the quotative is a substitute for the conjunctive use
of adverbs such as ‘how’ (ette… endu… in 56), ‘why’ (ēka… 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].
Christiane Pilot-Raichoor
In many occurrences, the literal meaning of the quotative ‘say > think’ is retained, with
a frequent shade of intentionality ‘think so that.’
(59) embattāru rāja makka, ‘ondu rājana kōt.ega rāja
86 king sons one king.obl palace.dat king
ille’ embaduga ittitte tandru hūdudāre.
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 āva kobbadu endu itte dēvaru
Nellitore be.P1adj people.obl kill1.Oblig quot so god
bandudane endu ava ondu.nū 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.’
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.
5.1.1 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):
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.
5.2.1 Cause/time
In many cases, both temporal and causal readings are possible:
(64) ava itte ibbane appara siggāda.
she thus be.P1adv much shy.become.T2.3
‘While/Because sitting like this she was feeling embarrassed.’
(65) nīnū bandadenda mane bīsalāgi tōridega…
you-foc come.N2z.abl house bright.adv seem…
‘From your coming/Because you have come, the house seems bright…’
Christiane Pilot-Raichoor
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) nā āsaga ā mundādun ibbaduga siggu.
I so much people in front.obl be.N1z.dat shy (cp. with 64)
‘I feel embarrassed to stand before so many people.’
Finally, we should mention that the simple juxtaposition of two independent sentences,
here in reverse order, may also express a cause:
(68) avve, satakana bā, 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 bēku ‘be wanted’, requires the
experiencer to be case marked as dative. In the first sentence (66a), kūsu, the subject agreeing
with the verb (pat.t.itu), is omitted in the embedded clause “kāru bēku”, while in (66b), the
experiencer kūsu.ga, marked in the dative and placed in initial position, is shared by both
clauses, but the second clause has its own subject kōpa bandutu: “anger came to the boy…”
Converbs and adverbial clauses in Badaga, a South-Dravidian language
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.
A restrictive meaning ‘only if, unless’ can be obtained by adding the emphatic particule
-tā to the conditional converb:
(73) kāreman.iya tandale-tā […] nā 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 won’t bring the news to Kariabetta
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 suf-
fixed to an adjectival or an adverbial (79) participle as well as to a substantivized form
(82) of the verb. Two elements: māke as in (44) and enge, are used for simple manner
or comparison clauses.
It should be noted however that a distinct adverbial vid.a ‘in comparison with, instead
of, (rather) than’ is used for evaluative comparison:
20. Note that this is a marginal use. There are other devices, morphological as well as peri-
phrastic, to express direct and indirect causativity.
Christiane Pilot-Raichoor
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 sug-
gested 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 struc-
tural 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 clause-
linking 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 (66a–b) for cause function and
(84)–(85) for purpose function.
Christiane Pilot-Raichoor
To balance the analytic approach adopted in the preceding sections, we will now pres-
ent a few excerpts from narratives to show how the various devices are used and com-
bine to construct complex sentences.
As said at the beginning, converbs are frequently used to express simple sequen-
tial 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 hōgi 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 bā ‘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 hōgi, jōsi 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 ujjōne which, in this case, delimits the only domain relevant for the main
clause ‘their teeth were chattering’.
(99) orakadu, [ella eddu, mōga tōgi, < hallu ujjōne],
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.
Converbs and adverbial clauses in Badaga, a South-Dravidian language
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
Christiane Pilot-Raichoor
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 morpho-
syntactic 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 spe-
cialized 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 par-
ticiple) or may depend on a nominal head (typically a noun of the spatial, temporal:
jāma ‘time’ or manner domains) or may be directly dependent on the postposition/
adverbial marker which express the semantic relation to the main clause. The spe-
cialized 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 con-
textual converbs (Pc) which have some implied adverbial meaning (cause, conces-
sion, 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 Nedjalkov’s 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.
Structural, pragmatic and prosodic factors play an important role in the interpre-
tation 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 differ-
ently 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 49–50), the construction
NVz.ga is used for purpose clauses as well as for the Potential mood.
From a syntactic point of view, Nedjalkov’s 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 pre-
asserted 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 syn-
tactically 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 sim-
ply 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 simultane-
ous or anterior to the verb/predicate to which it is linked. It signals both syntacti-
cally and semantically that the sentence/sequence/narration is not finished and that
Christiane Pilot-Raichoor
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 inter-
esting, 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 con-
struction 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
References
Chris H. Reintges
CNRS/LLF & University Paris 7 – Denis Diderot, Fédération Typologie
et Universaux linguistiques – CNRS
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 volume’s 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․ī, itself a corruption of the
Greek adjective (ai)gypt(ios) ‘Egyptian’. Ancient speakers of the language called themselves
nә-rәm-әn-kε˜ me ‘the people (rәm) of Egypt’ (kε˜ me lit. the black country) and used the abstract
noun tә-mәnt-rәm-әn- kε˜ me ‘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 inven-
m
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 syntac-
tically 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 coordina-
tive construction (see Dik 1968: 52–54; Schachter 1977: 90; Haiman 1983b: 86–87;
Johannessen 1998: 39; Huddleston, Payne & Peterson 2002: 1275–1276 paragraph 1.1.,
1299–1300 paragraph 2.2.3.; Haspelmath 2004: 3–4, 34, 2007: 5–6, 46; Bril & Rebuschi
2007: 10–13, 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 identi-
cal 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 coor-
dination 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 Coptic-
Greek alphabet є and н and о and ϖ 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. kϖT /kft/ ‘to build’ vs. kϖTє /kf˜te/ ‘to turn’;
Coordination, converbs and clause chaining in Coptic Egyptian typology
b. The identical subject in the first clause is the 1st sing. personal pronoun ti
(Budge, Miscellaneous Coptic Texts 266, 13–14)
ti=әr khria әm-pei-saε˜ әn-∫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.’
Unstressed open syllables, on the other hand, have short vowels, e.g. the /ke/ in kєĂϖĂ /kelfl/
‘pitcher, jar’. There is some evidence that the digraph Nг represents a word-final velar nasal,
e.g. Nг /aŋ/ ‘I’. Greek loan words are given in their original orthography. A more comprehensive
discussion on Coptic phonology is offered in Reintges (2004a: Chapter 1).
Chris H. Reintges
The conjunctive plays a pivotal role in the grammar of clause combining and inter-
clausal 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, conjunc-
tive 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, 13–14)
se=na-mere pә-ßios әn-t-anakhf˜rε˜sis
3pl=fut-love.nom def.m.sg-life link-def.f.sg-seclusion
et-waaß әn=se-apotasse әm-pә-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: 238–239, 263–283; Givón 1990: 864–874; Haspelmath 1995: 20–27; Van Valin &
LaPolla 1997: 448–454).
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
Coordination, converbs and clause chaining in Coptic Egyptian typology
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 descrip-
tive 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.
Sahidic as the official church language and had become the sole representative of Cop-
tic 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 com-
prises two macro-stages with distinct typological features. On the one hand, there is
Earlier Egyptian, which includes Old Egyptian (2750–2400 BCE) and Middle Egyptian
(2150–1750 BCE). On the other hand, there is Later Egyptian, which consists of Late
Egyptian (1350–950 BCE), Demotic (950 BCE–200 CE) and Coptic (300–1300 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 non-
Egyptian 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.
Table 1. The four grades of transitive-active verbs (√kt ‘to build’)
Grade Form Semantics Valence
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, desig-
nates states resulting from prior events or states irrespective of their origin.
Aside from marking a core distinction in the aspectual domain, the grade sys-
tem 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 comple-
ment, must be adjacent to each other.
(4) The head-marking pattern of the nominal state grade (Mark 14, 58)
awf˜ ti=na-ket ke-wa әn-at-mu^ әn-kјit∫
and 1sg=fut-build.nom other-one as-neg.pfx-build with-hand
hәn-∫omәnt әn-hou˜
in-three link-day
‘And I will build another (i.e. temple) not as a hand-made one in three days.’
The stative functions as a detransitivizing voice, which eliminates the agent and
aligns the patient argument with the subject function (Polotsky 1987/1990: 203–205
. 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 /ә/, 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: 201–206 paragraph 6.3.1,
218–219 paragraph 6.3.1.1) for a more detailed analysis of Coptic verb morphology.
Chris H. Reintges
paragraphs 2–4; Layton 2000: 129 paragraph 168a; Reintges 2004a: 228 paragraph 6.3.3).
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.
(6) The detransitivizing stative grade (Luke 6, 48)
ne=f-k7t
t∫e kalf˜s
comp pret=3m.sg-build.stat well
‘because it (the house) was built well’
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, 197–196 paragraph 40; Layton 2000: 125
paragraph 160). Indeed, absolute and nominal state stems are readily available for nomi-
nalizations, e.g. pә-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.
Dryer (1992: 99, paragraph 3.5) distinguishes auxiliary verbs and tense/aspect parti-
cles 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 morphologi-
cal innovation of the Sahidic dialect, which is not shared by all neighbouring dia-
lects. 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 pat-
terns display any kind of agreement with the immediately following subject.
is placed in the left periphery and connected to the associated comment clause via a
resumptive pronoun.
(11) Left-dislocation (Mena, Martyrdom 4b, 6–9)
p-aggelos de әm-pә-t∫oeis
def. m.sg-angel pcl link-def.m.sg-lord
a=f-wonhә=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.’
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 constitu-
ents have in the discourse.
According to the Word Atlas of Language Structures, 125 languages out of 301 distin-
guish 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 con-
junction of both noun phrases and clauses (Haspelmath 2005: Chapter 64). 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.
Coordination, converbs and clause chaining in Coptic Egyptian typology
. 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.) hif˜f˜= < hi + f˜f˜= ‘back, dorsum’ (Reintges 2004a: 101 paragraph 3.2.1.1). Cru-
cially, the pronominal state hif˜f˜= is excluded from the domain of nominal coordination,
since coordinated pronouns are marked by the comitative preposition mәn.
. 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 dif-
ferent 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 coordi-
nation, the Iraqw coordinator nee displays a much broader syntactic distribution. See Mous
(2004) for a detailed discussion.
Chris H. Reintges
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 category-
sensitivity 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 mәn C mәn D (Mark 6, 3)
әm-pai an pe pә-ham∫e pә-∫ε˜re әm-Maria
neg-dem.m.sg not cop.m.sg def.m.sg-carpenter def.m.sg-son link-Mary
awf˜ pә-son әn-Iakoßos men-If˜sε˜ men-Iu˜da men-Simf˜n?
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 coordi-
nators 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 coor-
dinand (Haspelmath 2004: 4–5). There are only a few attested examples of bisyndetic
coordination, in which the coordinands are bound together by the two coordinators
awf˜ and mәn (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 mәn changes semantically to become a conjunction marker, it does not adopt the
coding properties of a coordinative construction, with the nominal complex NP1 mәn-NP2
triggering third person plural agreement. Examples like (i) are marginal.
. The coordinating conjunction awf˜ ‘and’ has been grammaticalized out of a petrified
imperative j‑wôħ or r-wôħ ‘Add!’ < wôħ ‘to place, add’. The coordinating function of r-wôħ is
first attested in Late Egyptian and Demotic (see Müller 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.
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 coor-
dinator and linkage markers implies a tighter unity between the juxtaposed clauses
(Layton 2000: 260 paragraph 335b).
(18) Asyndetic clause coordination (Hilaria 5, 1–4)
pә-diakonos de a=f-eire әm-pe=s-wf∫
def.m.sg-god pcl perf=3m.sg-make.abs prep-def.m.sg=poss3f.sg-wish
a=f–t∫i әm-toot=s әm-pә-nomisma
perf=3m.sg-take.abs from-hand=poss.3f.sg prep-def.m.sg-solidus
a=u˜-wfm mәn-ne=u˜-erεu
perf=3pl-eat.abs with-def.pl=poss.3pl-recip
әm-pe-hou˜ et-әmmau
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 tempo-
ral 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.
. 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, 1–3)
hen-tfm әn-hεt hen-mәnt-sokj
indef.pl-hardness link-heart indef.pl-nominal-foolish
hen-mәnt-at-sowәn hen-pornia hen-mi˜∫e
indef.pl-nominal-neg-know.abs indef.pl-fornication indef.pl-strife
hen-nokj әm-mәnt-asε˜ßεs a pә-nu˜te
indef.pl-great link-nominal-impious perf def.m.sg-god
әnt=e eßol әn-hεt=u˜
bring.pron=2f.sg pcl from=3pl
‘(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 әmpә-, whose past tense
reference and negative polarity extends to the entire VP coordination.
(19) VP coordination with awf˜ (Shenoute, Wess. 9, 87a, 30–33)
mәn wa әn-wot t∫e әmpә=f-f˜mәs awf˜
be.not one link-single comp neg.perf=3m.sg-drown.abs and
ff˜te eßol
wipe.abs pcl
‘There is no one such that he did not sink and get wiped out.’
The coordinated clauses need not have entirely parallel syntactic structures. The pro-
logue of the Gospel of John is particularly instructive, since the Sahidic translation pre-
serves the chiastic structure of the Greek original. In the first conjunct, pә-∫at∫e ‘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˜.
Coordination, converbs and clause chaining in Coptic Egyptian typology
conjoined clause awf˜ f-na-nahmә=k ‘and he will save you’ has the semantic interpreta-
tion 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 I’m leaving (Culicover & Jackendoff 1997).
(24) Conditional reading of awf˜-coordination (Teaching of Antonius 37–38)
politeue mәn-pe-Khristos awf˜ f-na-nahmә=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!’
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 Féry & Hartmann 2005 and the references cited therein).
. The tripartite tense system of Coptic Egyptian allows for more fine-grained distinctions
in the past domain. In Reintges (2004a: 271–75 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-sf˜tәm ‘he heard, was hearing’ (present-in-the-past), the
preterit past (pluperfect) ne-a=f-sf˜tәm ‘he had heard’ (past-in-the-past) and the preterit
future ne=f-na-sf˜tәm ‘he was going to hear’ (future-in-the-past).
Coordination, converbs and clause chaining in Coptic Egyptian typology
Table 2. The inflectional paradigms of Coptic deictic tenses (sf˜tәm ‘to hear’)
Present
Basic Relative Converb
sg 1 ti=sf˜tәm e=i-sf˜tәm e=i-sf˜tәm
2m k=sf˜tәm e=k-sf˜tәm e=k-sf˜tәm
2f te=sf˜tәm ere-sf˜tәm ere-sf˜tәm
3m f=sf˜tәm e=f-sf˜tәm e=f-sf˜tәm
3f s=sf˜tәm e=s-sf˜tәm e=s-sf˜tәm
pl 1 tәn=sf˜tәm e=n-sf˜tәm e=n-sf˜tәm
2 tetәn=sf˜tәm e=tetәn-sf˜tәm e=tetәn-sf˜tәm
3 se=sf˜tәm e=u˜-sf˜tәm e=u˜-sf˜tәm
Future
Basic Relative Converb
sg 1 ti=na-sf˜tәm e=i-na-sf˜tәm e=i-na-sf˜tәm
2m k=na-sf˜tәm e=k-na-sf˜tәm e=k-na-sf˜tәm
2f te=na-sf˜tәm ere-na-sf˜tәm ere-na-sf˜tәm
3m f=na-sf˜tәm e=f-na-sf˜tәm e=f-na-sf˜tәm
3f s=na-sf˜tәm e=s-na-sf˜tәm e=s-na-sf˜tәm
pl 1 tәn=na-sf˜tәm e=n-na-sf˜tәm e=n-na-sf˜tәm
2 tetәn=na-sf˜tәm e=tetәn-na-sf˜tәm e=tetәn-na-sf˜tәm
3 se=na-sf˜tәm e=u˜-na-sf˜tәm e=u˜-na-sf˜tәm
Perfect
Basic Relative Converb
sg 1 a=i-sf˜tәm әnt-a=i-sf˜tәm e-a=i-sf˜tәm
2m a=k-ssf˜tәm әnt-a=k-sf˜tәm e-a=k-sf˜tәm
2f are=sf˜tәm әnt-are-sf˜tәm e-are=sf˜tәm
3m a=f-sf˜tәm әnt-a=f-sf˜tәm e-a=f-sf˜tәm
3f a=s-sf˜tәm әnt-a=sf˜tәm e-a=s-sf˜tәm
pl 1 a=n-sf˜tәm әnt-a=n-sf˜tәm e-a=n-sf˜tәm
2 a=tetәn-sf˜tәm әnt-a=tetәn-sf˜tәm e-a=tetәn-sf˜tәm
3 a=u˜-sf˜tәm әnt-a=u˜-sf˜tәm e-a=u˜-sf˜tәm
(27) Relative clauses with the relative perfect (Testament of Isaac 230, 10–11)
t-irε˜nε˜ әm-pa-t∫oeis
def.f.sg-peace link-def.sg.m.poss.1sg-lord
[RC әnt-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, rela-
tive tenses are not simply subordinate verb forms, but display a broad syntactic dis-
tribution 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
(Apophth. Patrum, Chaîne n0 139, 31, 7)
әnt-a u ßfk e-pe=k-hεt?
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, 19–21)
әmmon, alla әnt-a=u˜-sәßtotә=f
no but rel-per=3pl-prepare.pron=3m.sg
әm-pә-diaßolos mәn-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 Footnotes 4 and 5) recently proposed to replace the conventional term ‘relative’
by the semantic cover term ‘focus’, which provides an appropriate semantic characteriza-
tion 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: 185–186 for preliminary observations).
Coordination, converbs and clause chaining in Coptic Egyptian typology
. According to Bisang (1995), there is no direct correlation between canonical morpho-
logical 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 produc-
tive use of various types of converbs.
. To account for the presence of relative tenses in predicative adjuncts, one may capi-
talize on the focus role of depictive predication. Depictives make a predication that is par-
tially 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: Chapter 6). Alternatively, one might argue that
the depictive predicate functions in much the same way as a restrictive relative clause, nar-
rowing down the predication conveyed by the main verb, to the exclusion of other possibilities
Chris H. Reintges
paragraph 138; Steindorff 1951: 171 paragraph 366; Mallon 1953: 119 paragraph 249;
Layton 2000: 276 paragraph 351; Reintges 2004a: 296–297 paragraph 8.1.1.4.1).
. The inflectional paradigm of the Coptic conjunctive represents a language-internal in-
novation. The first attestation of the base morpheme nә- 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 nә- (London/Leiden Magical Papyrus V33, 3)
xr=w Ħr imi n-i-ir=k-wnm
say=3pl Horus come.imp conj-aug-aux=2m.sg-eat.inf
‘They said Horus, come and eat!
Coordination, converbs and clause chaining in Coptic Egyptian typology
clause chaining. Series of conjunctive clauses display the hallmarks of chaining struc-
tures. 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
(Testament of Isaac 231, 14–16)
mәnnsf=s on f=na-ale et∫әn-u-∫e әn-stauros
after=3f.sg pcl 3m.sg=fut-ascend.abs on-indef.sg-wood link-cross
nә=f-mu˜ ha-pә-tεrә=f
conj=3m.sg-die.abs for-def.m.sg-all=poss.3m.sg
nә=f-tf˜wәn әm-pe=f-meh-∫omәnt әn-how
conj=3m.sg-raise.abs on-def.m.sg=poss.3m.sg-nominal-three link-day
nә=f-∫fl әn-amәnte nә=f-fei
conj=3m.sg-destroy.abs prep-Hell conj=3m.sg-carry.abs
әn-tә-mәnt-rf˜me әn-tootә=f әm-pә-t∫at∫e
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; Givón
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
Givón (1990: 889) contends that the position of the clause-linking morpheme pro-
vides another feature of interclausal grammar that correlates with word order typol-
ogy. 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 language’s analytic morphological type. The con-
junctive base morphemes nә- and әnte- are functional particles rather than morphologi-
cal 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 for-
mative 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.
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 (Mena, Miracles 27b, 19–20)
kә=nau ero=i e=i-∫f˜ne
(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 counter-
parts 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 (König 1995: 75–81). 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, 12–14)
en hoson de e=n-∫at∫e mәn-ne=n-erεu
in so.far pcl rel(-pres)=1pl-talk.abs with-def.pl=poss.1pl-recip
a=f-ei ehun әnkјi pә-hε˜gemf˜ n
perf=3m.sg-come.abs pcl foc def.m.sg-governor
әn-te-khf˜ra et-әmmau
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: 23–26)
pә-ran de әn-te-heneete et-әmmau pe
def.m.sg-name pcl link-def.f.sg-monastery compREL-there cop.m.sg
Erε˜te e=n–o әn-u-hεt әn-wft
Erête rel(-pres)=1pl-do.stat in-indef.sg-heart link-single
mәn-ne=n-erεu hi-u-sop
with-def.pl=poss.1pl-recip in-indef.sg-time
ere t-irinε˜ ∫oop hәn-tә=n-mεεte
rel(-pres) def.f.sg-peace happen.stat in-def.f.sg=poss.1pl-midst
e=n-∫oop hәn-u-esykhia mәn-erεu
rel(-pres)=1pl-happen.stat in-indef.sg-ease with-recip
e=n-ti eou˜ әm-pә-nu˜te
rel(-pres)=1pl-give.nom praise to-def.m.sg-god
‘The name of that monastery (was) Erête 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: Chapter 2). Since converbs form a complex predicate with the aspectual verb,
they function syntactically as complement clauses rather than clausal modifiers, which
are adverbial.
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.
There are other instances of conjunctive clauses that behave like subordinate struc-
tures. A case in point is conjunctive complement clauses to manipulative verbs.
Coordination, converbs and clause chaining in Coptic Egyptian typology
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 desententializa-
tion. Absolute finiteness and non-finiteness, on their part, are connected to the pres-
ence 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 expo-
nent of asymmetric clause linkage appears in the structural slot of the tense/aspect
particle. Conjunctive verbs exhibit temporal restrictions in the sense that a coordi-
native 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:
The key differences between conjunctive verbs and infinitives can conveniently be
illustrated with examples like (39), in which the two forms are coordinated (Shisha-
Halevy 1986: 203–204 paragraph 7.2.5.1).
nә=f-∫f˜pe ha-t-eksu˜sia
conj=3m.sg-become.abs under-def.sg.f-power
әn-u-pneuma әn-t∫a˜t∫e
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 tәm ‘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, 89–90)
nim pe pә-rf˜me e=f-na-kјf˜∫t
who cop.m.sg def.m.sg-man rel=3m.sg-fut-watch.abs
әnsa-pә=f-∫ε˜re e=f-ßεk әn-emtf˜
after-def.m.sg=poss.3m.sg-son rel(-pres)=3m.sg-go.stat to-depth
nә=f-tәm-ßoε˜thei 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 Givón’s (1990: 875) typological generalization that “in the clause-
chaining type, sequentially-conjoined clauses are the least finite of all clause types” [his
emphasis, CHR].
The stative grade is generally excluded from the event-oriented conjunctive. Conjunc-
tive verbs may, however, appear in the inchoative construction, which is formed with
the existential verb ∫f˜pe ‘to happen, become’. In this context, they combine periphras-
tically 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∫ εrp әn-ßrre e-askos әn-ßrre
but rel-hab=3pl-put.nom wine link-new to-wineskin link-new
әn=se-∫f˜pe e=u˜-wot∫ әm-pә-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.
. The traditional nomenclature of the inferred evidential as a conjunctivus futuri ‘future con-
junctive’ (Stern 1880: 284 paragraph 450) reflects the alleged semantic affinity of the two asym-
metrically coordinative patterns. Stern’s terminology has been revived in Layton’s (2000: 283–284
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 Till’s grammar (1966: 157–156 para-
graph 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 state-
ment (or prediction) introduced by a conjunctive and inferential verb form, respectively.
Within such mixed coordinations, the coordinating device has wider scope than the illo-
cutionary component (Huddleston, Payne & Peterson 2002: 1332–1333 paragraph 3.5).
(48) Mixed coordination with initial imperative and different-subject conjunctive clause
(Eudoxia 68, 10–11)
tenu kјe tf˜wәn maro=n әnta-t∫it=e
interj pcl raise.imp opt-1pl conj.1sg-take.pron=2f.sg
et∫әm-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 Christ’s) tomb!’
(49) Mixed coordination with initial imperative and different-subject inferential clause
(Ephesians 5, 14)
tf˜wәn=g p-et-әnkotәk ŋ=lo
raise.imp=2m.sg def.m.sg-compREL-sleep.abs conj.2m.sg-cease.abs
u˜te-n-et-mowt tare pe-Khristos әr
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: 157–56, 219–233 for a more detailed discussion
on first-person effects in evidentials).
(i) First person singular inferred evidential (Testament of Isaac 230, 9–10)
awf˜ fi pә-rou∫ tar=i-ei әnsf=k
and take.imp def.m.sg-care infer=1sg-go.abs behind=2m.sg
hәn-u-ra˜∫e e-men f˜sәk
with-indef.sg-joy rel-not.be hesitate.inf
‘And take care and I shall fetch you with joy without delay.’
Coordination, converbs and clause chaining in Coptic Egyptian typology
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
(Testament of Isaac 235, 5–6)
fei әn-ne=k-ßal ehrai
lift.imp prep-def.pl=poss.2m.sg-eye pcl
ŋ-nau e-nә-kolasis
conj.2m.sg-see.abs prep-def.pl-punishment
‘Lift your eyes and look at the punishments!’
(51) Mixed coordination with initial imperative and same-subject inferential clause
(Sirach 6, 18)
pa-∫ε˜re t∫in-te=k-mәnt-koui
def.m.sg.poss.1sg-child since-def.f.sg=poss.2m.sg-nominal-young
sf˜tәm әn-te-sßo 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-mәnt-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: 324–327
paragraph 8.2.4 for further discussion).
Foley & Van Valin (1984: Chapter 6) acknowledge three types of interclausal nexus
relations: coordination, subordination, and cosubordination. The three types of inter-
clausal nexus are defined in terms of two syntactic parameters [± dependent] and
[± embedded]; see Table 4 for further illustration.
Chris H. Reintges
. In Van Valin & LaPolla’s (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).
Coordination, converbs and clause chaining in Coptic Egyptian typology
fused into a single complex event, as in (52a), or events that are contrasted with one
another, as in (52b) (see Blakemore & Carston 2005: 570–571, 580–581).
(52) Binary coordination with same-subject conjunctive clause
a. Coordinated clauses describe a complex event (1 Corinthian 15, 32)
marә=n-wfm nә=tәn-sf˜ t∫e tәn=na-mu˜
opt=1pl-eat.abs conj=1pl-drink.abs comp 1pl=fut-die.abs
gar әn-raste
pcl in-tomorrow
‘Let us eat and drink, since we will die tomorrow!’
b. Coordinated clauses describe contrasted events (Psalm 36, 21)
∫are pә-ref-әr-noße t∫i et∫o=f
hab def.sg.m-nominal-do.nom-sin take.abs for=3sg.m
nә=f-tәm-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
individual conjunct.
(53) Binary coordination with same-subject conjunctive clause and repetition
of the identical subject (Shenoute, Leipoldt III 40, 28–29)
ere hen-psykhε˜ na-ti hεu awf˜ әnte hen-psykhε˜
rel indef.pl-soul fut-give.nom profit and conj indef.pl-soul
ßfk e-rat=f әm-pә-nu˜te hәn-tßßo nim
come.abs to-foot=3m.sg prep-def.m.sg-god in-purity each
e-rat=f әm-pә-nu˜te hәn-tßßo 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 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 essen-
tially coordinative nature of the conjunctive verb conjugation (Roberts 1988: 56–57;
Haspelmath 1995: 25; Kwon & Polinsky 2008: 91–93).
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)
pә-rime na-∫f˜pe e-ru˜he
def.m.sg-weeping fut-become.abs at-sunset
әnte pә-telεl ∫f˜pe e-htf˜we
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 chain-
initial clause to the conjunctive clause is blocked. Consequently, the resulting con-
struction is not interpreted as a coordinate structure, but instead as a complex sentence
with an adverbial clausal modifier (Schachter 1977: 99–100). 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 (Besa’s Life of Shenoute, [Pap. Brit.Mus. 10820, folio 2 rct.])
e=n-na-kjәn rf˜me tfn әn-te=k-he
rel=1pl-fut-find.nom man where link-def.f.sg=2m.sg-kind
nә=f-ti sßo na=n nә=f-trypha
conj=3m.sg-give.nom teaching to=1pl conj=3m.sg-nourish.abs
әmmo=n әn-ne-graphε˜ et-waaß ?
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?’
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.
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.
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 ten-
dency originates from the prevalence of same-subject conjunctives with anaphorically
linked subject pronouns.
(64) Bipartite coordination with awf˜ ‘and’ conjunctive verbs (Psalm 36, 21)
∫are pә-dikaios ∫әn htε=f
hab def.m.sg-righteous be.pitiful.nom heart=poss.3m.sg
awf˜ nә=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 prag-
matic scale of the speaker’s disapproval or outrage at the actions reported therein.
(65) Multiple coordination with awf˜ and conjunctive verbs (John 10, 10)
pә-ref-t∫i˜we me=f-ei etße-laau
def.m.sg-nominal-steal neg.hab=3m.sg-come.abs for-anything
eimε˜ti t∫ekaas e=f-e-hfft awf˜ nә=f-∫fft
unless comp rel-3m.sg-prep-steal.abs and conj=3m.sg-slaughter.abs
awf˜ nә=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.’
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: 1293–1297 paragraph 2.2.1).
(66) Open disjunction with the Greek conjunction ε˜ and conjunctive verbs
(Budge, Coptic Martyrdoms 211, 31–32)
әr-∫an u-thlipsis taho=u˜
rel-cond indef.sg-affliction rise.pron=3pl
7˜ әnte u-kindynos tf˜wәn et∫f=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 (…).’
(67) Open disjunction with correlative ε˜… ε˜ and conjunctive verbs (Matthew 6, 24)
mәn kјom
әn-laau e-r hәmhal әn-t∫oeis
snau
not.be power link-someone to-make.nom servant link-lord two
7˜ gar f-na-meste wa nә=f-mere wa
or pcl 3m.sg-fut-hate.nom one conj=3m.sg-love.abs one
7˜ nә=f-kјolt∫ә=f әn-wa nә=f-kataphronei
or conj=3m.sg-entangle.pron=3m.sg to-one conj=3m.sg-despise.abs
әm-pә-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.’
(68) Adversative coordination with the coordinator alla and a conjunctive verb
(2 Timothy 2, 24)
u-hәmhal әnte-pә-t∫oeis nә-∫∫e an
indef.sg-servant link-def.m.sg-lord neg(-pres)-is.proper not
ero=f e-mi∫e alla nә=f-∫f˜pe
for=3m.sg to-fight.inf but conj=3m.sg-become.abs
e=f-thßßiεu˜ әnnahrәn-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’
(69) The topic-shifting particle de in conjunctive clause chains (1 Corinthians 14, 23)
e∫f˜pe kјe er∫an t-ekklε˜sia tε˜rә=s ei
if pcl cond def.f.sg-church entire=poss.3f.sg come.abs
e-u-ma әn-wft nә=se-∫at∫e tε˜r=u˜
to-indef.sg-place link-one conj=3pl-speak.abs entire=poss.3pl
hәn-u-aspe nә=se-ei de ehun әnkјi
in-indef.sg-tongue conj=3pl-come.abs pcl pcl foc
hen-idif˜tε˜s ε˜ hen-apistos n–e=u˜-na-t∫oo=s
indef.pl-laymen or indef.pl-impious neg-rel=3pl-fut-say.pron=3f.sg
an t∫e e=tetәn-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»?’
between subordinate conjunctive clauses that concern the level of embedding: core
versus periphery.
What interests us here are clausal chains in which the controlling verb and the con-
junctive verb are separated by a clause boundary. In a syntactic variant of the comple-
ment 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 con-
stituents. The absence of the finite subordinating complementizer t∫e ‘that’ is a charac-
teristic feature of this construction.
(75) Indirect speech complement formed with a conjunctive verb (Till, KHML I 8, 6–7)
ta-sf˜ne etße-u әmper-t∫oo=s
def.f.sg-sister for-what neg.perf.2f.sg-say.pron=3f.sg
әn-ne=s-eiote nә=se-ei әnmma=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 comple-
mentation 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).
With verbs of volition, intent or desire, conjunctive verbs may alternate with infini-
tives. Members of this class are characterized by an experiencer subject and a com-
plement proposition that is asserted to be realizable in the nearby future (Noonan
1985: 121–125; Dixon 2006: 31). With first person singular subjects, volitional and
Coordination, converbs and clause chaining in Coptic Egyptian typology
desiderative verbs designate a promissive speech-act where the speaker commits him-
self to a particular action.
(77) Conjunctive direct object clauses to desiderative verbs (Mena, Miracles 10a, 10–15)
anok gar e=i-wf∫ әnta-pft
i pcl rel(-pres)=1sg-want.abs conj.1sg-depart.abs
6nta-pf˜hәt eßol әm-pa-snof
conj.1sg-shed.abs pcl prep-def.sg.m.poss.1sg-blood
et∫әm-pә-ran әm-pa-t∫oeis Jε˜sus 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 acqui-
sition of knowledge, which convey a semifactive sense. As a rule, these verbs presup-
pose the truth of the complement clause (see Kiparsky & Kiparsky 1970 and much
related research).
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
. 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, 1–4)
awf˜ e=s-∫an-∫f˜pe anok ta-әr rf˜me
and rel=3f.sg-cond-happen.abs i conj.1sg-do.nom man
ta-mu˜ ta-tf˜wәn eßol hәm-n-et-mowәt
conj.1sg-die.abs conj.1sg-stand.up.abs pcl from-def.pl-comp.rel-die.stat
әm-pә-meh-∫omәnt әn-hou˜ ti-na-tre won nim
on-ordinal-three link-day 1sg-fut-caus.inf one each
er pә-meewe әm-pe=tәn-ran
do.nom def.m.sg-remembrance link-def.m.sg=poss.2pl-name
әn=se-epikalei әmmf˜=tәn na=u˜ әn-eif˜t
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.’
Coordination, converbs and clause chaining in Coptic Egyptian typology
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 (Sahidic Vita of Pachomius 90, 3–6)
u˜de әmpe=i-ei e=i-na-ßfk e-u-ma eneh
and.not neg.perf=1sg-go.abs rel=1sg-fut-go.abs to-indef.sg-place ever
ε˜ u-sowhәs ta-t∫oo=s
or indef.sg-congregation conj.1sg-say.pron=3f.sg
hf˜s e-wәnta=i t-eksu˜sia t∫e ma
comp rel(-pres)-have=1sg def.f.sg-authority comp give.imp
na=i әn-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».’
(84) Conjunctive clauses with a temporal shift from historical perfect to present
tense (Shenoute, pap. Brit.Mus. 198, 82b: 22–25)
a= u˜-net∫ pә-Satanas eßol hәn-tә-pe
perf=3pl-throw.nom def.m.sg-Satan pcl from-def.f.sg-heaven
t∫e a=f-kf˜te әnsa-t-he әn-әr
comp perf=3m.sg-seek.abs after-def.f.sg-manner link-do.nom
ne=f-ßote ŋ-aa=u˜ de
def.pl=poss.3m.sg-scandal conj.2m.sg-do.pron=3pl pcl
hff=k awf˜ ŋ-t∫ok=u˜
reflex=poss.2m.sg and conj.2m.sg-complete.pron=3pl
hәn-nә-paraphysis nә=se-t∫itә=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)?’
(85) Conjunctive clauses introduced by hf˜s ‘such that’ and (affirmative) polarity
agreement (Budge, Coptic Martyrdoms 216, 2–3)
a=f-pf˜ne hf˜s nә=f-әr
perf=3m.sg-change.abs comp conj-3sg.m-make.nom
tә-he әn-ke-rf˜me
def.sg.f-manner link-other-man
‘He changed such that he became like another man.’
Coordination, converbs and clause chaining in Coptic Egyptian typology
The negative counterparts mә˜pf˜s and mә˜pote 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 t∫e may be stacked on top of the negative complementizer mә˜pote.
(86) Conjunctive clauses introduced by the stacked complementizers t∫e and mε˜pote
‘that’ and (negative) polarity agreement (Sahidic Vita of Pachomius 138, 32)
a=i–t∫oo=s t∫e mε˜pote әn=se-kјf˜
perf=1sg-say.pron=3f.sg comp comp conj=3pl-remain.abs
e=u˜-ti woi hәn-tә-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 subordina-
tive 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 coor-
dinate 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 coordi-
nation, 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 sym-
metric 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 con-
junctions 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 inter-
est, 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
Abbreviations
References
Kahle, Paul E. 1954. Bala’izah – Coptic Texts from Deir el-Bala’izah, 2 Vols. London: OUP.
Kiparsky, Paul & Kiparsky, Carol. 1970. Fact. In Progress in Linguistics: A Collection of Papers,
Manfred Bierwisch & Karl Erich Heidolph (eds.), 143–173. The Hague: Mouton.
König, 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 König
(eds.), 57–95. 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 Communica-
tion Research 11], Barbara Lewandowska-Tomaszczyk (ed.), 87–102. 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.), 181–225. Amsterdam: John Benjamins.
Lichtenberk, Frantisek. 1995. Apprehensional epistemic. In Modality in Grammar and Discourse
[Typological Studies in Language 32], Joan Bybee & Suzanne Fleischmann (eds.), 293–327.
Amsterdam: John Benjamins.
Longacre, Robert E. 1979. The paragraph as a grammatical unit. In Syntax and Semantics 12 –
Discourse and Syntax, Talmy Givón (ed.), 115–134. 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.), 235–286. 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), 331–359. 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.), 109–122. Amsterdam: John Benjamins.
Müller, Max. 1888. Zur Etymologie des koptischen awf˜: woh. Zeitschrift für Ägyptische Sprache
und Altertumskunde 26: 94–95.
Muysken, Pieter 2008. Functional categories [Cambridge Studies in Linguistics 117].
Cambridge: CUP.
Nedjalkov, Vladimir P. 1995. Some typological parameters of converbs. In Converbs in Cross-
linguistic Perspective: Structure and Meaning of Adverbial Verb Forms – Adverbial Parti-
ciples, Gerunds – [Empirical Approaches to Language Typology 13], Martin Haspelmath &
Ekkehard König (eds.), 97–136. Berlin: Mouton de Gruyter.
Nichols, Johanna. 1986. Head-marking and dependent-marking grammar. Language 61: 56–119.
Nikoleva, Irena. 2007. Introduction. In Finiteness: Theoretical and Empirical Foundations, Irena
Nikolaeva (ed.), 1–19. Oxford: OUP.
Noonan, Micheal. 1985. Complementation. In Language Typology and Syntactic Description,
Vol. II: Complex Constructions, Timothy Shopen (ed.), 42–140. 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.), 3–41.
Cambridge: CUP.
Chris H. Reintges
Polotsky, Hans Jakob. 1960 . The Coptic conjugation system. Orientalia 27: 392–422.
Polotsky, Hans Jacob. 1944. Études de syntaxe copte [Publications de la société d’archéologie
copte]. Cairo.
Polotsky, Hans Jacob. 1987/1990. Grundlagen des koptischen Satzbaus [American Studies in
Papyrology 28–29]. 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), 51–81. Philadelphia PA:
Fortress Press.
Reintges, Chris H. 1994. Egyptian root-and-pattern morphology. Lingua Aegyptia 4: 213–244.
Reintges, Chris H. 2001. Code-mixing strategies in Coptic Egyptian. Lingua Aegyptia 9: 193–237.
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.),
363–408. Amsterdam: John Benjamins.
Reintges, Chris H. 2004a. Coptic Egyptian (Sahidic Dialect): A Learner’s Grammar [Afrikawis-
senschaftliche Lehrbücher 15]. Cologne: Rüdiger Köppe.
Reintges, Chris H. 2004b. Coptic Egyptian as a bilingual language variety. In Lenguas en con-
tacto: El testimonio escrito [Manuales y Anejos de ‘Emerita’ XLVI], Pedro Bádenas de
la Peña, Sofía Torallas Tovar, Eugenio R. Luján & María Ángeles Gallego (eds.), 69–86.
Madrid: Consejo Superior de Investigaciones Científicas.
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.), 249–287. 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.), 185–220. Berlin: Mouton de Gruyter.
Reintges, Chris H. & Green, Melanie. 2004. Coptic second tenses and Hausa relative aspects:
A comparative view. Lingua Aegyptia 12: 157–177.
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.), 165–194. Cambridge MA: The MIT
Press.
Roberts, John R. 1988. Amele switch-reference and the theory of grammar. Linguistic Inquiry
19: 45–63.
Schachter, Paul. 1977. Constraints on coordination. Language 53: 86–103.
Schultze-Berndt, Eva & Himmelmann, Nickolaus P. 2004. Depictive secondary predicates in
crosslinguistic perspective. Linguistic Typology 8: 59–131.
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: VWGÖ.
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: 1–54.
Coordination, converbs and clause chaining in Coptic Egyptian typology
Stassen, Leon. 2005. Noun phrase coordination. In The World Atlas of Language Structures,
Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.), 258–261.
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.), 171–234.
Cambridge: CUP.
Till, Walter C. 1942. Koptische Kleinliteratur. Zeitschrift für Ägyptische Sprache und Altertum-
skunde 77: 101–111.
Till, Walter C. 1961. Koptische Dialektgrammatik: Mit Lesestücken und Wörterbuch. München:
C.H. Beck.
Till, Walter C. 1966. Koptische Grammatik (Saïdischer Dialekt) mit Bibliographie, Lesestücken und
Wörterverzeichnissen, Leipzig: Enzyklopädie 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 Handwörterbuch. Heidelberg: Carl Winter.
Winkler, Susanne. 1997. Focus and Secondary Predication [Studies in Generative Grammar 43].
Berlin: Mouton de Gruyter.
Yuasa, Etsuyo & Sadock, Jerry M. 2002. Pseudo-subordination: A mismatch between syntax and
semantics. Journal of Linguistics 38: 87–111.
Zanuttini, Raffaella. 2008. Encoding the addressee in the syntax: Evidence from English imperative
subject. Natural Language and Linguistic Theory 26: 185–218.
part iii
Isabelle Bril
LACITO, (Laboratoire des Langues et Civilisations à Tradition Orale),
Fédération Typologie et Universaux Linguistiques, CNRS
1. Introduction
*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
2. Th
e syntax of information and referential hierarchy in clause-linking:
Some definitions
1. “Presupposed” is defined as “not part of the same assertion” as the main clause (Givón,
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 argu-
ments or modifiers of another clause (see for instance Sections 5.4, 7.4).
Another function of demonstratives and definite markers is found in clause-chains,
as in Takia (Section 5.5.2) where they encode sequential and consecutive dependency
relations, or in correlative strategies.
Topic, focus markers and demonstratives do not exhaust the types of informa-
tional hierarchy markers; position, clause ordering, T.A.M. markers are also widely
distributed, but they will not be the core of this study.
Domain and scope are two essential notions to account for the varying functions of
morphemes. According to their scope and syntactic domain, demonstratives and deic-
tics, 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 Sections 4; 5.3; 8.2).
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).
Informational and referential hierarchy
(ii) factual presupposition as in he is sorry that she didn’t come (in which the subordi-
nate clause contains the presupposition: she didn’t come). By contrast with categori-
cal propositions, thetic propositions such as it’s 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, explica-
tive 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 hierar-
chy operate as markers of syntactic hierarchy in clause-linking.
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: 169–171).
(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) (Sterner & Ross 2002: 175)
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?’ (Sterner & Ross 2002: 175)
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
‘Don’t 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 determin-
ers and definite markers are more tightly integrated and embedded. Similar facts will
be shown for Nêlêmwa (Section 4.3.2) and Takia (Section 5.4.1).
Without any time noun, clauses are structured as topic (time frame)-comment
clauses (3b):
Informational and referential hierarchy
The complex sentence in (4) illustrates the interaction of several markers: (i) sequen-
tial 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 weren’t there.’
(Sterner & Ross 2002: 184)
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
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).
Informational and referential hierarchy
‘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.’ (Throop & Ross 2002: 393)
In Austronesian languages, some coordinators often develop into topic or focus mark-
ers (see Bril in press); this occurs via topic maintainance (with ‘and’ coordinators), via
topic switch (with contrastive ‘but’ coordinators) or via correlative functions.
9. Note also constrastive function of the topic marker men on the NPs ‘men’ and ‘women’.
Isabelle Bril
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).
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 forward-
looking po, it marks the first part of the construction as the topic and points forward to the
comment which follows (Miller 1991: 126–127).
Informational and referential hierarchy
11. Loyalty Islands languages use similar constructions: in Drehu ame … tre ‘as for…then’ or
ame … ke ‘as for … so’ (tre ‘then’ or ke ‘so, as’) (Moyse-Faurie 1983: 197, 201); in Iaai, haba…
me ‘as for …and’ (Ozanne‑Rivierre 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) Nêlêmwa
Na i shumwiny mwa Pwâ-Hivic: “xe yo axe-ve xe yo axe
and 3sg do.thus seq Pwâ-Hivic conj 2sg see-dir tpc 2sg see
o da?, fo idaama-m?
instr what? there.is eye-poss.2sg
‘And Pwâ-Hivic then says: “so, (what) you saw (tpc), what did you see it with?
Do you have eyes?” (Bril fieldnotes 1995)
12. In van Valin’s model, the fact that they cannot be topicalised is a sign that they are co-
subordinate.
Informational and referential hierarchy
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) Nêlêmwa
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’ (Bril fieldnotes)
d. Kio wa tuuli foliik o cêê 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)
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: 251–257).
Isabelle Bril
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 Nêlêmwa’ to ‘I met the woman who is learning
Nêlêmwa’). 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 comple-
ment clauses have different constructions:
14. Croft (2001: 351–353) points out that complement clauses often originate from coordi-
nate or purposive subordinate clauses.
Informational and referential hierarchy
“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 fre-
quent cross-linguistic use as correlative markers in time, conditional and consecu-
tive 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 Takia’s16 clause-linking strategies, which epitomise some types
often found in Western Oceanic languages. Three markers will be analysed: (i) the coor-
dinator 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). Takia’s clause linking types include (paratactic)
coordination, co-subordination, clause-chaining, and subordination for relative, com-
plement 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 foreground-
background 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
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: 58–59).
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 dichot-
omy; it appears in constructions that are equivalent to both, including relative clauses
(Ross 2002: 235).
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 con-
strasts with the enclitic definite marker =n (see Section 5.4).
the two clauses are asserted and the reported speech clause is more loosely concate-
nated; 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-loŋ 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
ŋai man [nek du-fun-ag ŋa-mat ane-n] ŋa-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)
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)
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).
Informational and referential hierarchy
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 plac-
ing 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, fol-
lows this verb). There are thus two constructions of perception verbs in Hua; the first
treats the perception as the verb’s 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’ kutta’na ripa’ na-mo] kgoe.
money theft take(2sg.rel) thing-(nmz) I.saw.you
‘I saw you stealing the money.’ (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’ ’a’baivao.
look (3du.incons) in.there really they.are.not
‘The two of them looked (and saw) that they really weren’t in there.’
(Haiman 1988: 61)
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.
(38) Takia
a. ŋai22 pein ŋu-le la i=k e,| oŋ 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)
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 it’s very long.’ ‘Don’t worry about that.’
(Ross 2002: 245)
22. Independent pronouns signal a change of topic. When there is no topic switch, the refer-
ence 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 depen-
dent morpheme =p) is marked as a topic clause by man:
(40) Takia
You nam gireŋ mi-gane ta=p man | aŋar 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).
5.8 Discussion
Coordinate clauses with equal assertive/illocutionary force thus stand in contrast with
clause-linking strategies based on asymmetric assertive force, using either (i) informa-
tional 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 analy-
sed in Roviana.
Informational and referential hierarchy
6. C
ase-marking and focus strategies in clause-linking: The case of
Roviana (Oceanic, Solomon Islands)
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 speci-
ficities which set them apart from other subordinate clauses: (i) they occur in clause-
initial 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 argu-
ment 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 erga-
tive system, while adverbial subordinate clauses have a neutral case system. In (44a–b),
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
“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: 26–27)
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.’
(Corston-Oliver 2002: 495–497)
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.’
(Corston-Oliver 2002: 497)
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)
Informational and referential hierarchy
7. R
eferential hierarchy: Demonstratives and deictics as markers
of subordination and clause hierarchy in Tawala
French: il lui a menti, /ceci ~ cela ~ ce qui/ n’arrangea 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 func-
tions (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.
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 (Section 7.4.1), are generally pre-
posed to the main clause and are simply juxtaposed or correlated to the following main
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: 143–144).
(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 speaker’s or hearer’s sight). They are reflexes of
Proto-Oceanic *a/*na ‘near addressee’; *o/*no ‘distant from both speaker and addressee’ (Ross
1988: 100).
Isabelle Bril
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.
28. The bracketing of clauses (64a-b) is Ezard’s, 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.
Informational and referential hierarchy
In (53b), the topic time clause is marked by a prosodic rise, thus houga-na hi-gele-
geleta 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)
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 Ezard’s, 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.’ (Ezard 1997: 224)
c. Hi-wiwogatala [po apo hi-na-bagibagi].
3pl-plan conj fut 3pl-pot-work
‘They planned to work.’ (Ezard 1997: 248)
7.5 Discussion
When used as conjunctive markers, demonstratives may express sequencing and con-
secutive functions (as in Takia), but they mostly function as subordinators (in relative,
complement or adverbial clauses). As adverbial clause markers, they appear in time, con-
ditional, 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.
Clause ordering is another pragmatic and syntactic indicator. Sequential (in line) con-
structions 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
Informational and referential hierarchy
32. The postposition bóôana ‘like’ has various subordinating functions (causal, manner,
similative-comparison, conditional, and counterfactual irrealis conditional (Lichtenberk
1983: 372–75, 528–529, 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 strength-
ened by informational hierarchy (topic or focus) devices or by referential hierarchy
devices (demonstratives), as in Korafe.
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, ã 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)
Informational and referential hierarchy
Conjunct participles and absolute constructions actually fulfil the same ad-
sentential, 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 infor-
mational structure markers often originate from a pool of connectors or coordinators:
additive or sequential connectors express clause sequencing, logical and topic continu-
ity, 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 presupposi-
tion, 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 Culioli’s notions of pointage and fléchage 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: “Prag-
matically structured propositions are (…) paired with appropriate lexicogrammatical
structure.” (1994: 334).
Table 1. Summary
Tawala Manam Takia Sobei Kaulong Nêlêmwa
(PNG) (PNG) (PNG) (Irian Jaya) (New Britain) (New Caledonia)
Abbreviations
References
Bickel, Balthasar. 1993. Belhare subordination and the theory of topic. In Studies in Clause
Linkage, Karen H. Ebert (ed.), 23–55. Zürich: Arbeiten des Seminars für Allgemeine
Sprachwissenschaft.
Bickel, Balthasar. 1999. From ergativus absolutus to topic marking in Kiranti: A typological per-
spective. Proceedings of the 25th Annual Meeting of the Berkeley Linguistics Society, 38–49.
Bickel, Balthasar. 1999. Nominalization and focus constructions in some Kiranti languages. In
Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren W. Glover (eds), 271–296.
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.), 29–43. Canberra: Australian National University.
Bril, Isabelle. 2001. Postmodification and the structure of relative clauses in Nêlêmwa 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), 261–284. Canberra: Australian National
University.
Bril, Isabelle. 2002. Le Nêlêmwa, analyse syntaxique et sémantique [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.), 499–534. 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 Völlmin & Silvia Zaugg-Coretti (eds). Frankfurt:
Frankfurter Afrikanistische Blätter 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), 467–497. Richmond: Curzon Press.
Croft, William. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspec-
tive. Oxford: OUP.
Culioli, Antoine. 1990. Pour une linguistique de l’énonciation, Tome 1: Opérations et représenta-
tions. 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 Lan-
guage 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 Lan-
guage Typology and Syntactic Description, Vol. I: Clause Structure, Timothy Shopen (ed.),
282–364. Cambridge: CUP.
Foley, William A. 2007. A typology of information packaging in the clause. In Language Typol-
ogy and Syntactic Description, Vol. I: Clause Structure, 2nd edn), T. Shopen (ed.), 362–446.
Cambridge: CUP.
Fontinelle (de la), Jacqueline. 1976. La langue de Houaïlou (Nouvelle-Calédonie): Description
phonologique et description syntaxique [TO 17]. Paris: SELAF.
François, Alexandre. 2000. Contraintes de structures et liberté dans l’organisation du dis-
cours. Une description du mwotlap, langue océanienne du Vanuatu. Thèse de Doctorat en
Linguistique, Université Paris-IV Sorbonne.
Givón, Talmy. 1980. The binding hierarchy and the typology of complements. Studies in
Language 4(3): 333–378.
Haiman, John. 1978. Conditionals are topics. Language 54(3): 564–589.
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 Clause-
combining in Grammar and Discourse [Typological Studies in Language 18], John Haiman
& Sandra A. Thompson (eds), 49–69. 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.), 75–92. Canberra: Australian National University.
Kiss, Katalin É. (ed.). 2001. Discourse configurationality. In Language Typology and Language
Universals, Vol. 2, Martin Haspelmath, Ekkehard König, Wulf Österreicher & Wolfgang
Raible (eds), 1442–1455. 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 Kada-
zan narratives. In Thematic Continuity and Development in Languages of Sabah [Pacific
Linguistics C-118], Stephen H. Levinsohn (ed.), 105–135. Canberra: Australian National
University.
Informational and referential hierarchy
Zygmunt Frajzyngier
University of Colorado
1. Introduction
. 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 distinc-
tion 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 wá,
whose structure and derivation is described below:
Temporal apodosis:
(2) mádá dùmá màlŠárwá
má dá-d má màVár wá
hyp go-go hyp now com
‘And when he [the groom] goes,’
ә́lvà tàr ŋánnà bà1ә̀mmè á dә́ zә̀ dàdà
speech 3pl def all pred go hl father
‘everything that they have to say is with the father.’
Conditional apodosis:
(3) Vàmá má cìnà-kùr-áncín jùbóonә̀ náwá
Vàmá má cìnà-kùr-án-cìn jùbà únà ná wá
assembly hyp hear-2pl-3sg-hear matter def dem com
‘Listeners, if you understand this matter,’
'ákà-r gìná
'ákà ár gì ná
neg.ex over to dem
‘there is no greater [truth/goodness] than that.’
Comment clause
An afterthought clause:
(7) ә́lvà ŋánnà yè sә̀ ndà-n-ée nà kwá [error] wá
story def 1sg come tell-3sg-1sg dem com
‘The story I came to tell…’
á kàtá lvà ndzà
á kàtà ә́lvà ndzà
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).
palatal segments, high round in the environment of round segments, and high central
in other environments. The epenthetic vowel is e before a pause:
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 determin-
ers; 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:
Here is an illustration of the root form of the verb (with epenthetic high-front vowel)
before the nominal object:
In one type of the head-modifier constructions, the head has the root form:
The root + a form characterizes morphemes when they are followed by the
nominal subject:
(13) ә̀bVyàsә́bVyà dùksá
ә̀bVy-à-sә́-bVy-à dùksá
fall-go-s-fall-pb thing:q
‘Something fell out?’
The root + á is used in a variety of structures which involve connecting two elements
that do not constitute an expected, natural sequence. The vowel á 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) VóVà lvá hә̀r1á [má added in analysis]
Và-ú-V-à lv-á hә̀r1 [má
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
+ á ending:
(15) tà dә̀ hàl-tá zárvá tàr ŋánnà tà púmá d-ә́m bùhá
tà dә̀ hàl-t-á zárv-á-tàr ŋánnà tà pú-m á dә́-m bùhá
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 hàlà ‘gather’ is intransitive)
The comment clause marker wá always ends in the high tone vowel á. The form wá 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. ábáŋánè kòndáŋgù kòbúunà
á bá ŋànè kò ndá-n gә̀ wè kò bwá nà
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. á, kàndángùgdzárwà màkí9yé
á, kà ndá-n g w gdz-á-rwà mà kí9yé
ah, 3sg say-3sg to what child-gen-1sg 1incl three
“Why did you say [so] my son? ‘There are three of us.”’
In the following example, the last word of the topic phrase kùlà ends in the vowel a:
(19) mábà kә̀Và kùlà wá ájìyù kә̀gìyà
má bà kә̀Và kùlà wá à jì-y-ú k gә̀ ìyà
hyp foc count calculus com 3sg surpass-1sg-vent neg to 1sg
‘If it concerns counting, nobody surpasses me.’
(20) ŋánwá dùksә̀ŋánnә̀wá séí mábànì ákSyànà bàddámyà
ŋán wá dùksә̀ ŋánnә̀ wá séí má bà nì
def com thing def com only hyp foc intens
á kSyà-nà bà ddá-myà
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) (kSyà-nà ‘cool’ + 3sg obj)
The comment marker may form a unit with the demonstrative ná, where the demon-
strative 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 wá anticipates
the comment clause. In the following example, the adverbial phrase is the topic:
(21) ánó vàcí stwà náwá yágánìkà
á nó vàcí s-t-wà ná-wá yá gá nìkà
eh pres day dem-dem-dem dem-com 1sg contract marriage
‘On such and such day I am going to get married.’
(22) áwáyà náwá tàdá kràwà
á wáyà ná wá tà dá kràwà
pred yesterday dem com 3pl go:go Krawa
‘As for yesterday, they went to Krawa …’
(23) a. yó ә́gdzà dàwàlná à dә̀ žàràn ùrárà
yó ә́gdzà dàwàl-ná à dә̀ žàrà-n ùr-á-rà
well young man-dem 3 go see-[pause] friend-gen-3sg
‘Well, the boy went to see his friend,’
Zygmunt Frajzyngier
fátàrŋánnә̀ kínì wá
f-á-tàr ŋánnә̀ kínì wá
field-gen-3pl def c.foc com
‘And now, their father, who was buried in the field …’
àbàŋánè kòndàŋgù kòbúunà
à bà ŋánè kà ndà-à-n gә̀ wè kò búu nà
3sg say 3sg 2pl say-go-3sg to what 2pl two dem
‘He says, “Why do you say you two?”
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 indi-
rect 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
The marker wá can precede a clause in the imperative, a function shared by comple-
mentizers in IE languages:
(38) yò 'ә́llàhá ŋ ŋánnà Vàbá dàlyè ŋáccìnà
yò 'ә́l làh-á-ŋ ŋánnà Vàbè á dàlyè wá ŋá ccìnà
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) é, àbàSә̀ksә́wá pàšó pàšә́vgә̀ŋánnà
é, à bà Sә̀ksә́ wá pàšà-w-pàš ә́vgә̀ ŋánnà
eh, 3sg say Sultan com dig-pl-dig grave def
‘And the Sultan said, “Dig up this grave.”’
(pàšà-w-pàš is a reduplicated imperative form)
The marker wá can precede the imperative verb of the complement clause. The de dicto
complementizers of IE language cannot be used in a similar environment:
(40) àbàSә̀ksùwá dómbàré nóyà'lámdè
à bà Sә̀ksè wá dú-wà mbàré
3pl say Sultan com go-pl.imp together
nó yá 'lá mdè á dә́ zә̀ kùr ŋánnà
pres 1sg send people pred go hl 2pl def
‘The sultan said, “Go, I am sending people to your place.”’
Zygmunt Frajzyngier
The marker wá 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 wá in the following example:
(41) à dә́ndàvà ázà dә́ ŋàrwà gyálә̀ ná
à dә́ ndàvà á zà dә́ ŋà-r wá gyálә̀-ná
3sg go ask pred hl father poss-3 com girl-dem
hàyèn hàyá àmá dә́ ŋàrwá à kàtá yénvyè
hàyé-n-hàyà àmá dә́ ŋà-r wá à kàtà á
love-3sg-love but father poss-3sg com 3sg want 3sg
yé-n-v-yè
refuse-3sg-appl-refuse
‘He went to ask her father, saying that the girl loves him.
But her father wants to refuse him.’
(42) tàsә́ndàvìwә́lvà kàVànáa rúgyà wá
tà sә́ ndàv-ì-w ә́lvà kàVàn-áa-rú g-yà
3pl come ask-1sg-vent matter goods-gen-1sg to-1sg
‘They came to ask me about the matter of the goods’
wá 'ákà ә́lvá kàVánà
com neg.ex matter goods
(I said) ‘There is no matter of goods.’
The only example of the comment-clause marker wá occurring after a verb of percep-
tion is in an elicited sentence, hence its validity is somewhat doubtful:
(43) yà ná wá tà gyá 1áfà
1sg see com 3pl prepare food
‘I saw that they prepared food.’
The comment-clause marker wá is not an obligatory component of the complements
of verbs of saying. Thus, it does not have to precede direct speech:
(44) à bә̀ pàllè ddá Sìmà
à bà pàllè dd-á Sìmà
3sg say one man-gen ear
‘The one who hears all said:’
šíilyá zàrvá mì pàllè mbә̀1á mbә̀1á dә́m yáwnà
šíilì-á zàrv-á myà pàllè mbә̀1á-m-mbә̀1 á dә́m yáw nà
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 mbә̀1à 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 vVyà [bVyà] ‘fall’ should have
been used instead. In the subsequent description of the same event the speaker does indeed
use the verb bVyà ‘fall’.
Comment clause
The comment marker wá 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 wá marks the clause as a comment on the preceding material
within the same sentence:
Most examples in which the comment-clause marker wá does occur have some
material after the verb of the first clause and before the comment-clause marker wá. It
is usually the subject of the clause. If the complement clause immediately follows the
verb of saying, the comment-clause marker wá most often does not occur:
The comment-clause marker wá does not occur if the complement clause uses
the deontic complementizer séi ‘then, therefore’ borrowed from Hausa or Fula.
Zygmunt Frajzyngier
The reason the form wá does not occur is that its function is subsumed by the
complementizer séi:
(49) máVàrá béetàrè séetá dá šúlà
máVàrà á bà ítàrè séi tá d-á šúlà
now (f) 3sg say 3pl then 3pl go-go travel
‘Now they say that they have to travel.’
(50) àbàŋàn kàyà sèyìdә̀ sánsә̀ sábìnà
à bà ŋàn kàyà sèi yà dә̀ sá-n-sә̀-sè á b-ìnà
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 wá 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 wá is not used if the complement clause precedes the verb of
saying. If the complement clause has the deontic complementizer séi the comment-
clause marker wá is not used. This points out that the role of the comment-clause
marker wá is thus to mark the clause as being a comment on the preceding material.
3. Synthesis
The syntactic environments listed above do not have much in common in terms of the
categories that precede the form wá. That indicates that the presence of the comment-
clause marker wá is not triggered by the properties of the phrases or clauses that
precede it. In all cases, the marker wá 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 wá. The fact that the presence of the marker
wá 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 complemen-
tizers in other languages (Frajzyngier 1996).
The form wá 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 wá indicates that
something else follows in the utterance and that it constitutes a comment on the
immediately preceding material.
4. Implications
The widespread use of the comment clause puts into the question the funda-
mental 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
References
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
Développement in Maroua, which has also provided me with much-needed institu-
tional and logistic support. I am most grateful to its current director, Dr. Noé Woin,
for support during the years 2004–2008.
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, Fédération 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 Robert’s (1993,
2000) findings on focusing strategies, these links are discussed in detail for two
polyfunctional particles, raΩ, 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 raΩ
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 clause-
linking 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 sub-
ordinator liôanna ‘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 mountain-
ous district in central Yemen, belongs to the so-called ‘-k dialect’ group and is located
in the latter’s 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 Perfec-
tive 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 hypo-
thetical 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 qa˜l ‘say’, only direct reported discourse
can follow.
The verbal system is based on a binary aspectual opposition between the Perfec-
tive (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 raΩ), Progressive (b or y) and Future (ba˜, ôa˜, ∫(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 ka˜n ‘be’.
Nominal (i.e. verbless) sentences (with Subject-Predicate word order) are very fre-
quent: 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 ka˜n ‘be’ may be inserted between S and P.
Yafi‘ Arabic also has a wide range of topic and focus particles in assertive, inter-
rogative 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.
Deixis, information structure and clause linkage in Yafi‘ Arabic (Yemen)
The origin of the particle raΩ goes back to the imperative form of a verb ra˜Ω (cognate
form of classical Arabic raôa) meaning ‘look, see’, which, unlike other Arabic varieties,
is still in use in the language as such, as in (1):
Very frequently, raΩ 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 raΩ
still retains traces of its verbal origin and cannot be considered as fully grammatical-
ized yet, in the sense that (a) its form most often varies with the number and gender
of the addressee: m.sg. raΩ, f.sg. raΩ-i˜, m.pl. raΩ-u˜, f.pl. raΩ‑e˜n, and (b) object pro-
nouns can be 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 raΩ 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 vari-
eties, 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 grammati-
calization 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 raΩ imply that some potentiality
for such developments exists. Harrel (1962: 215) for instance mentions the following
on Moroccan Arabic:
The use of․rā is especially common when a sort of logical connexion or consequent
result is implied, e.g. xud ․taksi u ․rāh iw6․s․s․l 6k ‘take a taxi and he’ll get you (s.g.)
there’, d․․r6 b l-xadiža tilifun, ․raha f6 -d․-d․ar
․ ‘give Khadija a telephone call, she’s
at home’.
. 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 d’être’.4
The following sections will describe the different uses of the particle raΩ.
2.1 Presentative
One function of the particle raΩ 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, raΩ has the deictic function of a presentative with
scope over either an entity or a whole predicative relation, quite similarly to voici or voilà
in French (whose origin is also the imperative form of the verb meaning ‘see’ voir).
When raΩ simply has scope on a pronoun, and is followed by an independent ver-
bal clause in the Imperfective, or by a coordinated clause, the sequence raΩ + pronoun
constitutes a nominal clause of its own. In this particular case, raΩ often highlights the
speaker himself, expressed as an object suffix pronoun:
Example (3) is one of the rare instances when the particle remains invariable in the
3m.sg (there are several addressees, not just one):
When raΩ has scope over noun phrases or prepositional phrases (ex. 4, 5, 6), point-
ing out/referring to animate as well as inanimate, definite or indefinite entities, the
sequence raΩ (+ pronoun) + NP or PP constitutes a nominal clause. Unlike most
. “Quant à la signification de toutes ces particules, il est à observer qu’elles ne désignent pas
seulement l’apparition subite et inattendue d’une chose, ou seulement l’antithèse, mais aussi la
conséquence, la situation continuée ou la raison d’être.” (Landberg commented upon several
other discursive particles, hence the plural form (‘these particles’).
Deixis, information structure and clause linkage in Yafi‘ Arabic (Yemen)
Maghribi Arabic varieties, in this case, pronouns cannot be suffixed to the particle,
except with indefinite nouns (4):
But when raΩ has scope over a relative clause (ex. 7), the object pronoun is obligatorily
suffixed to raΩ, and functions as the antecedent of the relative marker:
2.2 Copula
In Yafi‘ Arabic, nouns, prepositional phrases, participles and adjectives can also func-
tion as predicates of existence, property or location, without the need for a verb mean-
ing ‘be’ or a copula.
The particle raΩ may be used with the assertive function of a copula for both
nominal properties and locative utterances (not only for locative utterances ones as
in Moroccan Arabic, Caubet 1993: 35). In my data, this occurs only when the subject
of the nominal utterance is a pronoun, expressed by the set of suffix object pronouns,
as when raΩ is used as a presentative. The word order is thus Copula – Subject – Nomi-
nal Predicate.
The use of raΩ 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 one-
nominal constituent utterance: ru∫a˜∫ (lit. ‘rain’) ‘it’s raining’.
Martine Vanhove
. The non intensive nominal utterance would use the independent subject pronoun fol-
lowed by the nominal predicate: ôani bi ħa˜l-ah (1f.sg with state-f) ‘I am in a strange state’;
hu˜ mulxo˜f (3m.sg mad.ptcp) ‘he is crazy’.
Deixis, information structure and clause linkage in Yafi‘ Arabic (Yemen)
Such a definition corresponds, all things being equal, to Lambrecht’s 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
. Just like the subject focus conjugation of Wolof (for an explanation of the dissociation
between the marked element and its scope, see Robert, 1993).
Deixis, information structure and clause linkage in Yafi‘ Arabic (Yemen)
Example (15) below presents another syntactic construction where the direct object
argument of the predicate is attached to the focusing particle (=oh in raΩ=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 ħetta ‘until; even’ (see König
1991). In this particular example, selling something was a totally unexpected event.
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).
. Note that the Yafi‘ particle raΩ 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 Robert’s 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 raΩ is based on a mecha-
nism of “encoded inference”, inherent to the definition of the process of focus promo-
tion, 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 pro-
vides 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 raΩ in Maghribi Arabic is also very
frequent (it is called ‘reassertive value’ by Caubet 1992, and Madouni 1993), and exam-
ples 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 raΩ is often found in dialogues, as
an answer to ‘why’ questions, linking the focused clause to a question asked by
another speaker; raΩ is even compatible with an overt coordination marker as in
(17) below:10
. Des exemples de réassertion avec recentrage (de type “c’est que/c’est parce que”…), où
l’énonciateur reprend ce qu’il vient de dire, parce qu’il estime avoir été mal compris, ou qu’on
a pu se méprendre sur ses intentions. (My English translation).
. Note that the structure of Example (17) is identical to that of the presentative use of
raΩ in (3). The difference is that in (4), the raΩ clause is joined to the previous one by the
consecutive particle baΩe˜n ‘then’, which marks a temporal, not a causal, relation between the
two clauses.
Deixis, information structure and clause linkage in Yafi‘ Arabic (Yemen)
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 raΩ brings to the Perfective
conjugation, the particle also makes the causal relation between the two clauses explicit:
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:
The same explanatory relation is found with nominal predicates, when they are juxta-
posed 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
raΩ the prosody blocks these interpretations which are no longer salient, and the causal
interpretation predominates:11
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
Section 2.3.4 above).12
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 raΩ 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 raΩ 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 it’s 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.
Deixis, information structure and clause linkage in Yafi‘ Arabic (Yemen)
2.4 Summary
The above description has shown a certain number of semantic and syntactic constraints
on and restrictions in the use of raΩ, depending on the syntactic scope of the particle. In
the deictic domain raΩ has a wide range of uses as a presentative with NPs, PPs and rela-
tive clauses, but with pronouns its use is limited to 1st persons. As a copula, raΩ 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 raΩ
are summarized in Table 1.
In Yafi‘ Arabic, the use of the particle ta as a clause-linking device is less frequent than
that of raΩ. Its lexical source being different from the latter, it is nevertheless interest-
ing to discuss it in the light of similar developments in various languages worldwide.
. maΩ ‘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‘ Ara-
bic 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):
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 nomi-
nal or verbal; in nominal utterances, ta is limited to locative utterances (28, 29). Contrary
to raΩ, no object suffix pronoun can be cliticized to ta (a trace of its verbal origin).
3.3 Copula
Ta is also used as a copula in nominal property utterances.15 The word order, different
from the one with raΩ, is Subject – Copula – Predicate. The predicate can be an indefinite
noun (31) or a clause (32) and, unlike raΩ, there is no constraint on subject category:
. Example (35) shows that another presentative ôan 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
Ta often combines with other focusing markers, which have different pragmatic or
assertive connotations. In the parenthetical clause below (ta qa hu˜ aθ-θaΩlab makka˜r),
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 raΩ, has no intensity value as a copula
and marks a simple thetic nominal utterance, but examples are too scarce to be sure.
Deixis, information structure and clause linkage in Yafi‘ Arabic (Yemen)
relation and is the explanatory focus marker, while qa is an intensity focusing particle
whose syntactic scope is limited to the verbal predicate itself:
In (37) the speaker’s statement explains why it would be difficult for her to avoid killing
her son, and ma˜∫i˜ is the intensive negative focusing marker:
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 wasΩa ‘ask’.17 Note that it cannot be used in reported speech after the quotative
verb qa˜l ‘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:
. The other request verbs can be followed by the complementizer ôinna, by no comple-
mentizer at all, or more rarely by one of the other relative pronouns.
Deixis, information structure and clause linkage in Yafi‘ Arabic (Yemen)
3.9 Summary
The polyfunctionality of ta is somewhat different from that of raΩ. As a presentative,
ta is only used with full utterances (not with NPs or PPs), or as a copula with prop-
erty utterances, but both with pronominal and nominal subjects. Its role as a marker of
information structure is narrower than that of raΩ: 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 conjunc-
tive 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.
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 raΩ can be used as an intensity-
focusing particle, and only ta is a demonstrative and a marker of relative and com-
plement 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 raΩ, independent utterances for ta. As a copula, raΩ is more
limited than ta as it can only be used with pronominal subjects, not with nominal
ones, while raΩ 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 raΩ
can focus the addressee (cf. ex. 24 where the object pronoun suffixed to raΩ refers to
the addressee). In addition, only raΩ 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 clause-
linking device for causal clauses.
Martine Vanhove
*imp dem pres prop prop loc top s,o,p int expl caus rel comp
pr n pr add
raΩ + – + + – + + + + + + – –
ta – + + + + – + – – + + + +
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 raΩ:
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.
Deixis, information structure and clause linkage in Yafi‘ Arabic (Yemen)
Abbreviations
References
Bril, Isabelle. 2000. Postmodification and the structure of relative clauses in Nêlêmwa and
other New Caledonian languages. In SICOL. Proceedings of the Second International Con-
ference on Oceanic Linguistics, Vol. 2: Historical and Descriptive Studies [Pacific Linguis-
tics 505], Bill Palmer & Paul Geraghty (eds), 261–284. Canberra: Australian National
University.
Bril, Isabelle. 2002. Le Nêlêmwa, analyse syntaxique et sémantique. Paris: Peeters.
Bril, Isabelle. In press. Coordination, information hierarchy and subordination in some Austro-
nesian languages. In Converbs, Medial Verbs, Clause Chaining and Related Issues [Frank-
furter Afrikanistische Blätter], Azeb Amha, Christian J. Rapold, Sascha Völlmin & Silvia
Zaugg-Coretti (eds).
Caubet, Dominique. 1992. Deixis, aspect et modalité: Les particules hā- et ․rā- 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
Catherine Taine-Cheikh
LACITO (Laboratoire des langues et Civilisations à Tradition Orale)
Fédération de Typologie et Universaux Linguistiques-CNRS,
Paris III and Paris IV-Sorbonne Universities
1. Introduction
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 account-
able for the imperfections to be found in this final version.
2. Abbreviations: a Aorist, ann annexation state, cas case marker, do direct object, io indi-
rect object, lit. literally, nipfv negative Imperfective, npfv negative Perfective, np noun phrase,
Catherine Taine-Cheikh
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-әrmäš ‘he took’ to
t-әrmäš ‘she took’ and t-әrmäšš-äm ‘you (pl) took’). When the person marker is explic-
itated by a lexeme, the two are coreferential and the lexeme’s 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-әrmäš iDy t-ägz̄ә-tt=әn=š
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)
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 mīn ‘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 Sections 1 and 2, pp. 1–150). 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]: 287–307),
followed in this by numerous scholars of Berber, dubbed this lexeme which bears the “an-
nexation 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 Galand’s, 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 ele-
ment 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 (“rhème” 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 organiza-
tion proposed in Hagège 1982) the discourse-hierarchic level, a level closely related to
the semantic-referential level, without however coinciding with it.
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 focal-
ized (ex. Un père, ça se respecte) (‘fathers, they are to be respected’), but this entails the noun’s
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 conso-
nant d (so that d > đ). But this lax tendency in Zenaga has also spread to the voiced fricatives
(so that z > θ) 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 [đ] 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
Catherine Taine-Cheikh
–– The root, made up of ordered root consonants (variable in number, but usually
two or three), is reserved for a given word family:
–– RMŠ in yәrmäš ‘he took, pleased’, trämS < trämšt ‘act of taking, pleasing
(once)’, änärmiš ‘s.o. who habitually takes, pleases’, äššärmiš ‘pond; handle’
and yäTyärmäš ‘he has been taken, pleased’;
–– RM in yәräm ‘he opened (his mouth, a trunk, etc.)’, trämt ‘act of opening (one’s
mouth, etc.) (once)’ and yäTyäräm ‘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 yәc(c)äc of the 3rd m.sg Perfective of yәrmäš
and yәräm;
–– morpho-lexical derivations, such as the pattern änäcc(c)ic for m.sg habitual
nouns with the affix -n- in änärmiš and the pattern äššäc(c)ic for m.sg nouns
with the affix -šš- in äššärmiš;
–– morpho-syntactic derivations such as tc(c)äct, the f.sg verbal noun in trämS
< trämšt and trämt, as well as yäTyäc(c)äc, the passive verb pattern with the
affix -Ty- of the 3rd m.sg Perfective of yäTyärmäš and yäTyäräm.
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 naô, with or without -đ) 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 әđ).
–– 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 äyš. Here, we will
concentrate on the first, as it is not uniquely used as a subordinating element.
analysis) but, to facilitate the identification of these morphemes, some assimilations (indicated
by ˆ) are not noted. Apocope is also indicated by ˆ, but the deleted phonemes (generally a vowel,
at times followed by the laryngeal consonant h) are given in parentheses.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
by the hypothetical system – namely potential vs. irreal), but they often contain the
nasal element m, thus mara, imer, xem, Meš, mur, lMr, mr, etc. (see Taïfi 1993).
One also finds a particle m (äm / häm / h.äm) in Zenaga, i.e. Mauritanian Berber,
but only in particular counter-factual hypothetical uses:
(3) (h)äm=đä t-әšš-äđ t-näzzä-T t-az.r.-ađ=iôh
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 didn’t come).
The particle ad (usually pronounced äđ) is used in the case of ‘standard’ conditional
clauses. Thus, in the following proverb:8
(4) äđ=ki y-äTyәf äm y
. už- äT әf=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, Zenaga’s peripheral situation, where it has been in
contact with Arabic for centuries, also favors specific developments. Using conver-
gence 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.
This section is concerned with Zenaga’s demonstratives and with the various gram-
maticalized 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 don’t 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.assāniyya).
Catherine Taine-Cheikh
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 world’s 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.
The three series have several points in common, but the mid series is characterized
by the presence of the laryngeal consonant ô, while the farthest series is distinguished
by a final -n (instead of đ), alongside the presence of the long vowel -ā-, found only in
the proximity adverb đāđ.
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 dăġ ‘in’. This analysis, however, is difficult to apply
to Zenaga as its preposition ‘in’ (đäg ~ đägg ~ đäôg ~ đäô) is never reduced to d.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
away from the speaker (op2) (see Bentolila 1969). In Zenaga, the ops take on the
following forms:
op1 -đ / -d and its allomorphs -ôđ / -đä(h) / -dä(h) / -ddä(h)
op2 -n and its allomorphs -ôn / -nä(h) / -nnä(h).
Zenaga (examples with obligatory op)
(5) y-әššä=ddäh y-әDyi=nän=däh
3m.sg-go.pfv=op1 3m.sg-leave.pfv=pr.do.3pl=op1
‘He came (here) [and] he left them (around) here.’
(6) y-әššä=nnäh y-әDyi=nän=näh
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 lan-
guage 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 den-
tal 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:
–– äräbīy=iôđ ‘that child there’, as opposed to äräbīy=äđ ‘this child (here)’ and
äräbīy=ān ‘that child there (far away), that child over there’,
–– but also äräbīy=iôđ ‘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 Aït 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 Aït Seghrushen’s 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.
Catherine Taine-Cheikh
/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 ‑äđ, which has a
specific plural form (-iđ), i.e.:
(7) äräbīy=äđ / äräbīy=iôđ / äräbīy=ān m. älliy
‘This child / that child / that child over there [is] white.’
(8) äräbān=iđ / äräbān=iôđ / äräbān=ān m
. älliy-iđ
‘These children / those children / those children over there [are] white.’
(9) tarbađ.=äđ / tarbađ.=iôđ / tarbađ.=ān m. älliy-äđ
‘This girl / that girl / that girl over there [is] white.’
(10) träbīn=iđ / träbīn=iôđ / träbīn=ān m
. älliy-iđ
‘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) äđ / iôđ / ān maz.z.ūg
‘This one (m) / that one (m) / that one (m) over there [is] small.’
(12) әđniđ / әđniôđ / әđnān maz.z.ūg-iđ
‘These ones (m) / those ones (m) / those ones (m) over there [are] small.’
(13) täđ / tiôđ / tān maz.z.ūg-äđ
‘This one (f) / that one (f) / that one (f) over there [is] small.’
(14) tәđniđ / tәđniôđ / tәđnān maz.z.ūg-iđ
‘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. taôK äđ ?
what this.one?
‘Who is he? Who is this one?’
b. taôK äyđ ?
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”.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
is distinct from the demonstrative’s (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 con-
taining 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 world’s languages.17
2.3 Presentatives
In clauses containing the copula ad, the subject, as we saw above, can be a demonstra-
tive 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) äyđ / äôđ / ān äđ [cop] burdällih / äđäm
. i / äytäb=әn=k
‘This / that / that (over there) is a centipede / a baobab / your book.’
(21) äđ / iôđ / ān äđ [cop] ah.mäd / aġma=n=š / äyiôm m
. älliy-än
‘This one / that one / that one (over there) [is] Ahmed / his brother /
a white camel.’
(22) täđ / tiôđ / tān äđ [cop] kumbä / tyađma=n=š / täyiômt m
. älliy-än
‘This one / that one / that one (over there) [is] Kumba / her sister /
a white she-camel.’
(23) әđniđ / әđniôđ / әđnān äđ [cop] äyđmän=š / iôymän m
. älliy-nin
‘These ones / those ones / those ones (over there) [are] his brothers /
white camels.’
17. For a discussion on the grammaticalization processes leading to the use of the non verbal
copula, see Diessels (ibid: 143 sq.).
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
–– amis wa nayeġ Ø ‘camel [namely:] the one I-saw Ø’ = ‘the camel that I saw’ or wa
nayeġ Ø ‘the one I-saw Ø’ = ‘the one that I saw’,
–– amis wa-reġ ‘camel this one’ = ‘this camel’ or wa-reġ ‘this one’
–– amis wa n amġar ‘camel of the chief ’ = ‘the chief ’s camel’ or wa n amġar ‘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 depen-
dence is marked by the simple presence of the participle suffix -än:
Zenaga
(27) mīn y-әrmәšš-än äđäġđi đäg ävuôš
someone 3m.sg-take.pfv-ptcp finger in hand
y-әrm䚈ti koL
3m.sg-take.pfv=pr.do.3m.sg all
‘Whoever takes a finger, takes the whole hand.’
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 äđ as a satellite prop.18
Whence the presence of äđ in (31b) – when the op is added – but not in (31a):
(31) a. taôK äyđ y-uđ.ur.-an
what this 3m.sg-fall.pfv-ptcp
b. taôK äyđ äd=däh y-uđ.ur.-an
what this rel=op1 3m.sg-fall.pfv-ptcp
‘What fell (around here)?’
18. In (30), äyđˆnäh could be the contraction of äyđ + äđ + näh. This simplification appears only
with po2 (with po1, one would have äyđ + äđ + däh > äyđ äddäh, cf. (31b). Other cases have been
noted, however, with äyđ 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 iôđ would not be used.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
This specific use of äđ 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 äđ to grammaticalize in this specific context.
2.5 Focalization
Focalization belongs to the domain of the discourse-hierarchic perspective. Intona-
tion, which always plays a role in the organization of a message, can indicate focaliza-
tion 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: 331–355). 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, it’s 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.
Catherine Taine-Cheikh
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 äyđ-äđ is used in “standard” rhematization, in associa-
tion with the deictic iôđ ( ... 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.bađ.-iôđ ‘the girl (there)’ and träbīn-iôđ
‘the girls (there)’ are not introduced, respectively, by the demonstratives f.sg täđ / tiôđ
/ tān + äđ and f.pl tәđniđ / tәđniôđ / tәđnān + äđ (as in täđ-äđ tar.bađ. ‘it’s a girl’), but
instead by the presentative äyđ-äđ in (35):
Zenaga
(35) a. äyđ-äđ t-ar.ba-đ.=iôđ ār kәnt šāwäy-äg
‘It is the girl (there) with whom (past) I spoke.’
b. äyđ-äđ t-räbī-n=iôđ ār kәnt šāwäy-äg
‘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 predica-
tion particle, in presentatives) are quite removed from its uses as subject demonstra-
tive 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.
21. An example without ayđ-äđ has been noted however, linked perhaps to the presence of
the relater äyš (formed by the determination prop ay + the relation particle š):
However, this same sentence (albeit incomplete) with different intonation could mean ‘Those
girls with whom you may speak (...)’.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
The forms used in negative contexts (after the negation wär ‘not’) often have sev-
eral 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 ä and ә.22
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 perfec-
tive 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.
22. And even more so between a and i or a and u. For more details, see Cohen and
Taine-Cheikh 2000.
Catherine Taine-Cheikh
Zenaga
(36) äbđīh naô ällәg
leave.imp or stay.imp
‘Leave or stay!’
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 impera-
tives 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 hare’s cow.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
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) āggūn ә<Ty>uz.̄uz.z.ug-än unnugur-än=ti
bard.pl <pass>beat.impfv-3pl hide.a-3pl=pr.do.3m.sg
‘The bards, one hits them [and] they hide it.’
It can also be the protasis of a conditional clause (see Examples (3) and (4)) or a speci-
fying relative (43):
(43) mīn y-әttättәr-än ädˆy-äykīh äššäyx=әn=š
someone 3m.sg-want.ipfv-ptcp ad=3m.sg-disappoint.a marabout=of=pr.3
.
y-әssәnkәr=ti oôf äy[đ] wärˆy-īr(i)ˆ 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 doesn’t want.’23
23. This proverb contains another one of the uses of ad + Aorist (after yәttättәr ‘he wants’)
which we shall examine in §3.3.1.
Catherine Taine-Cheikh
In the preceding examples, the subordinate clauses refer to some generic, usual, or
possible situation. If the subordinate clause corresponds to a temporally specified situ-
ation, the main clause verb is not in the Aorist. Thus, after a temporal subordinate
clause introduced by äll-ār (lit. ‘place where’), the verb in the main clause is in the
perfective form:
(44) ällār=đˆ әšš-äg aôr nwakšōT.
when=op1 come.pfv-1sg to Nouakchott
uz.̄r.-ag adya·ôÒn
find.pfv-1sg family.pl·pr.1sgÒ
‘When (as soon as) I arrived in Nouakchott, I found my family.’
The same holds true for the causal subordinate clause introduced by äđ-ār-äyš ‘since’
(lit. ‘this where that’):
(45) äđāräyš=täđ t-uz.z.aôr.-äđ t-әssän-äđ aôll=әn=š
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 verb’s tam generally depends on the aspectual or modal mean-
ing of the auxiliary. It is always in the Aorist after two auxiliaries with modal meaning.
yahađ. expresses a possibility (‘he can, was able’). The invariable impersonal äDyīri
denotes a necessity (‘it is necessary that’).24
(46) nәkni äššäbbäš=äđ äDyīri n-äyәg đäwr-әn äffälläh
us year=this must 1pl-ascend.a in.direction-of up
‘Us, this year, we must go to the countryside.’
24. The etymology could be: ädˆyīri ‘that he want’, from the verb yārä (A yīri) ‘want’.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
–– 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 oôf .
(52) ilāh äđ äž-b-äg
this.way ad drink.a-1sg
‘This way, that I drink!’
(53) ilāh ädˆtäđ ž-išb-äg
this.way ad=pr.do.3f.sg make.drink.a-1sg
‘This way, that I make her drink!’
25. ilāh is an adverb (borrowed from dialectal Arabic) not a verb, but it also expresses a process.
Catherine Taine-Cheikh
This construction with ad and the Aorist is sometimes equivalent to the construction
with the subordinator hānәš ‘so that’ (which probably contains the element š), but this
particle (always followed by the perfective or imperfective) is used to express the goal
specifically, contrary to ad.
. .
(56) akf (i)ˆiô=đ az.̄әr.f i hānәš aôS-äg әtšiyän
wait.imp=pr.do.1sg=op1 money so.that buy.i-1sg shoe.pl
‘Give me (here) money so that I (may) buy shoes.’
ad + Aorist is used after certain verbs such as yәnnäh ‘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):
When several orders are given, the particle ad is repeated before each Aorist, as in the
following example taken from a story:
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
26. Such formal similarities are found in many languages having a so-called Aorist form (for
example in Mwotlap and Wolof, see François 2003, Robert 1991, etc.). This cannot be simple
coincidence. In Berber, it does indeed correspond to a semantic proximity.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
With prohibitives, i.e. negative orders, the verbal negation particle wär 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. ätyši b. äđ wär=t-ätyšiô-đ
eat.imp ad neg=2-eat.a-sg
‘Eat!’ ‘Don’t eat!’
In the following proverb, the same verb is used successively as an order and as a
prohibition:
(69) ägtawäšni ägm
. ug āwäy=әn=š
wise.man follow.imp words[sg]=of=pr.3
äđ wär=t-ägm
. ug-äđ äôđ y-әssäkkär
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) äđ=ki wärˆy-uz.z.uôr. allah
ad=pr.do.2sg neg=3m.sg-see.a Allah
đäg äll=ār wärˆy-ūkiy
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-…[-đ] 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 wär, 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 Aïr Tuareg,
analyzed by Galand (2002 [1984]: 126): Ialla1, a2 di3 t4әqqәn5a4! ‘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 awedәn2 i3kkәl4et5 teyint6 -net7! ‘that5
each1 man2 he3-take4 the-pot6-of-him7’.
Catherine Taine-Cheikh
(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) mīn y-әttättәr-än ädˆy-äykīh
someone 3m.sg-want.ipfv-ptcp ad=3m.sg-disappoint.a
äššäyx=әn=š ädˆti y-әssәnkәr
marabout=of=pr.3 ad=pr.do.3m.sg 3m.sg-invoke.a
.
oôf =äy[đ] wärˆy-īr(i)ˆ 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!’
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”.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
–– 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.
–– 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 domi-
nant, see (78b). The following example – the morale to a story which illustrates the
need for brothers to stick together – illustrates its generic meaning.
The choice of the tam in the apodosis determines the meaning: general condition in
(78a) and goal in (78b):
..
(78) a. äđ wärˆy-äšbi y-uf f uđ
ad neg=3m.sg-drink.a 3m.sg-be.thirsty.a
‘Although he isn’t drinking, he is thirsty.’
b. äđ wärˆy-äšbi / y-әšbi y-änhäyä y-әtfuđ
ad neg=3m.sg-drink.a/npfv 3m.sg-future.pfv 3m.sg-be.thirsty.ipfv
‘If he doesn’t 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 fol-
lowed by yu(u)gä (> äDyu(u)gä), 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 explica-
tive (Banys 1996: 222), all forms of the perfective (pfv and npfv) and the imper-
fective (ipfv and nipfv) seem to be allowed, in both the protasis and the apodosis,
with various changes in meaning.
..
(79) a. äDyu(u)gä y-әšbä wärˆyuf f uđ
adˆyu(u)gä 3m.sg-drink.pfv neg=3m.sg-be.thirsty.npfv
‘If he drank, he is not thirsty.’
..
b. äDyu(u)gä y-әšbä kān y-äf f uđ
adˆyu(u)gä 3m.sg-drink.pfv past.3 3m.sg-be.thirsty.pfv
‘If he drank, [then] he was thirsty.’
..
(80) äDyu(u)gä y-әz̄ässä y-äf f uđ
adˆyu(u)gä 3m.sg-drink.ipfv 3m.sg-be.thirsty.pfv
‘If he drinks, [then] he is thirsty.’
..
(81) äDyu(u)gä wärˆy-uf f uđ wärˆy-әz̄issi
adˆyu(u)gä neg=3m.sg-be.thirsty.npfv neg=3m.sg-drink.nipfv
‘If he isn’t thirsty, he doesn’t drink.’
In Example (82), the lexeme “explicitating” the person marker (here bābaô-n-š ‘his father’)
is placed between äDyūgä and the verb y6=žäy. This is the sign of strong solidarity between
ad and yu(u)gä, and could be additional proof of the grammaticalization of äDyūgä 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) äDyūgä bābaô=n=š y-6ž-äy y-6ttiž-iy
adˆyu(u)gä 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.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
Numerous are the uses of ad in Berber. The deictic origin of the morpheme undoubt-
edly explains most of these uses (copula, presentative, “relative” pronoun, etc.), as com-
parable 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.
. 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 gram-
maticalizing the numeral one).
Catherine Taine-Cheikh
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: 118–122), between
the post-verbal subject (or “explicative complement”) and the right dislocated subject
(dislocation being indicated by an intonational break)34 should not similarly distin-
guish 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 mīn ‘someone’,
which can only be fronted if it is determined by a relative clause. Thus one may make
a distinction between the fronted mīn (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 wär-än yuz․‒․r i and its expan-
sion äyđ inäbbä:
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 cer-
tain 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: 266–70). 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 role of the Berber deictic and TAM markers in dependent clauses in Zenaga
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 indi-
cating 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 trans-
mitted 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 pre-
cise 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.
. 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.
Catherine Taine-Cheikh
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 under-
stand the causes of these phenomena, starting with the use of ad as tam particle. Cer-
tain peculiarities in its uses deserve more attention.
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 neces-
sarily 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
. 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.
Catherine Taine-Cheikh
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 Bühler 1934: 121–140).4041
. On the different types of referentials, see Desclés and Guentchéva 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.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
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 oblitera-
tion 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
. 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).
Catherine Taine-Cheikh
for clauses governed by modal verbs (can, must) or verbs expressing a request, a wish,
or an order.
The term “governing” follows Vycichl’s 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 depen-
dence 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 con-
sider 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
. The one which one finds e.g. after ira ‘want’, in the construction which gave rise, after gram-
maticalization of the modal verb, to the ‘true’ Tachelhit future: iraˆad 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.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
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 sub-
sequent 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-әnn(äh)ˆāš äzzәn wäđәn äđ
3m.sg-say.pfv=pr.io.3 say.imp someone ad
taôK=täđ đäg t-räbī-n y-әttättär
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-әnn(äh)ˆāš taôK ayaôxty
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 sub-
ordinate clauses in the same sentence, without the second introducing verb (y6nnäh in
the sense of ‘ask’) being repeated:
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 y6nnäh ‘say’, the clause expresses the object of a request,
a wish, just as after other verbs expressing a wish, expectation or intention.
Catherine Taine-Cheikh
–– If the verb is in the Perfective or Imperfective, the contents of the clause are given
independent referential anchorage. Following y6nnäh ‘say’, it is a case of indirect
speech.
–– If a verbal auxiliary, yu(u)gä or yum․ rä, precedes the verb in the Perfective or Imper-
fective (§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 pos-
sibility 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.
. This succinct semantic interpretation of indirect interrogation (inspired by Muller 2001)
is compatible with the meanings of the auxiliaries yu(u)gä and yum ․ rä outside of this par-
ticular context.
. We have found a proverb where the clause introduced by the counterfactual marker äm
is in second position, but the change in order goes along with a semantic change, where äm
‘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 role of the Berber deictic and TAM markers in dependent clauses in Zenaga
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)gä).
–– 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 adˆyu(u)gä 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)gä, it could be a case of topicalization (without any
particular formal marker) of the subordinate clause governed by a request verb
or the verb y6nnäh ‘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 it’ll definitely rain, (as X says), then I’ll 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 (thème/rhème
in French).50
. If the future auxiliary yänhäyä 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: 201–2.
Catherine Taine-Cheikh
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.
The difference between the two constructions lies in the way the framework is con-
strued. In the time clause introduced by the conjunction oôgđ-ār, 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.
Zenaga, are not, specifically, those of a topic marker, but the behavior of a demonstra-
tive 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 ôin and
ôinna, the latter being a particle with presentative meaning, serving to introduce topics
(definite or not), especially discourse initially.53
Classical Arabic
(90) ôinna rağul=a=n ğāôa
ôinna man=cas.acc=indf come.prf.3m.sg
‘There is a man (who) came’.
However, the discourse particle ôinna is related to the conditional marker ôin which
is very similar to Zenaga ad. Indeed, ôin, 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. ôin can however
also be followed by the suffixal conjugation (“Perfective”), like ôiđā.54
Classical Arabic (after Ayoub 2003)
(91) a. ôin t-ağiô=nī ô-ukrim=ka
if 2-come.sbjv=pr.do.1sg 1sg.honor.sbjv=pr.do.2m.sg
b. ôin ğiô-ta=nī ôakram-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 com-
pletely neutral: the Perfective (faΩala) and the Apocopate (yafΩal) 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 world’s languages: (i) modalities
(especially epistemic modalities), (ii) copula constructions, (iii) interrogatives (especially in-
terrogative 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) tem-
poral notions. The Zenaga ad could belong to (ii) and/or (iv).
. However, the topic is not focalized after ôinna, whereas the contrastive topic introduced
by ôammā ‘as for’ is focalized (the operation is marked in particular by a prosodic break).
. ôiđā, the other Classical Arabic conditional marker, commutates with ôin in (91b), not
in (91a).
Catherine Taine-Cheikh
request, an implicit ‘that P be’ on the part of the speaker: who asks the hearer to tempo-
rarily 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 Ayoub’s “donné admis” (‘allowed fact’)) so
that the process is constructed in the world of possibles: “One construes a fictitious ref-
erence 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 crosslinguis-
tically (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, conse-
quence 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 “iden-
tification 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.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
ialects. As they are also rarer among cases of grammaticalization in languages in gen-
d
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 adˆyu(u)gä, 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 adˆyu(u)gä, 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 virtu-
ally instantiated form: followed by the verbal system’s 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 sequen-
tial 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 polygrammati-
calization 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 iso-
lated 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 intona-
tion, 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 đ- and its multiple functions (Simeone-Senelle 2003).
Catherine Taine-Cheikh
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), 333–351.
Cambridge: CUP.
Ayoub, Georgine. 2003. Corrélation et rupture modales. Formes verbales et particules énoncia-
tives dans les hypothétiques en arabe littéraire. In Mélanges David Cohen, Jérome Lentin &
Antoine Lonnet (eds), 29–45. Paris: Maisonneuve et Larose.
Banys, Wieslaw. 1996. Propositions “conditionnelles”: Coordination, subordination, connexion.
In Dépendance et intégration syntaxique. Subordination, coordination, connexion, Claude
Muller (ed.), 221–226. Tübingen: Niemeyer.
Basset, André. 1952. La langue berbère. London: Dawsons.
Bentolila, Fernand. 1969. Les modalités d’orientation du procès en berbère. La linguistique 1969
(1): 85–96 & 1969 (2): 91–111.
Bentolila, Fernand. 1981. Grammaire fonctionnelle d’un parler berbère: Aït Seghrouchen d’Oum
Jeniba (Maroc). Paris: SELAF.
Berthoud, Anne-Claude. 1994. Indéfinis et thématisation. Faits de langues 4: 161–8.
Bril, Isabelle. In press. Coordination, information hierarchy and subordination in some Aus-
tronesian languages. In Converbs, Medial Verbs, Clause Chaining and Related Issues Azeb
Amha, Christian J. Rapold, Sascha Völlmin & Sylvia Zaugg-Coretti (eds). Frankfurter
Afrikanistische Blätter.
Bühler, Karl. 1934. Sprachtheorie, Jena: Fischer.
Chaker, Salem. 1997. Quelques faits de grammaticalisation dans le système verbal berbère. In
Grammaticalisation et reconstruction, 103–121. Paris: Klincksieck.
Cohen, David. 1984. La phrase nominale et l’évolution du système verbal en sémitique. Études de
syntaxe historique. Paris: Société de Linguistique de Paris.
Cohen, David & Taine-Cheikh, Catherine. 2000. À propos du zénaga. Vocalisme et morphologie
verbale en berbère. Bulletin de la Société Linguistique de Paris 95(1): 269–322.
Comrie, Bernard. 1986. Conditionals: A typology. In On Conditionals, Elizabeth Closs Traugott,
Alice ter Meulen, Judith Snitzer Reilly & Charles A. Ferguson (eds), 77–99. Cambridge:
CUP.
Culioli, A. 1990. Pour une linguistique de l’énonciation. Opérations et représentations 1. Paris:
Ophrys.
Desclés, Jean-Pierre & Guentchéva, Zlatka. 2006. Référentiels aspecto-temporels dans les textes.
Études cognitives 7: 7–34.
Diessel, Holger. 1999. Demonstratives: Form, Function, and Grammaticalization [Typological
Studies in Languge 42]. Amsterdam: John Benjamins.
François, Alexandre. 2003. La sémantique du prédicat en mwotlap (Vanuatu), Leuven: Peeters.
Galand, Lionel. 1957. Un cas particulier de phrase non verbale: “l’anticipation renforcée” et
l’interrogation en berbère. In Mémorial André Basset (1895–1956), 27–37. Paris: Adrien
Maisonneuve.
Galand, Lionel. 1977. Continuité et renouvellement d’un système verbal: Le cas du berbère.
Bulletin de la Société Linguistique de Paris 72: 275–303.
The role of the Berber deictic and TAM markers in dependent clauses in Zenaga
Galand, Lionel. 1988. Le berbère. In Les langues dans le monde ancien et moderne, III: Les lan-
gues chamito-sémitiques, Jean Perrot & David Cohen, 207–242. Paris: CNRS.
Galand, Lionel. 2002. Études de linguistique berbère. Présentation d’un parler: Le touareg de
l’Aïr [1974], 117–143; Typologie des propositions relatives: La place du berbère [1984],
219–240; “Subordination résultant de la relation”: À propos de la relative berbère [1987a],
241–256; Les emplois de l’aoriste sans particule en berbère [1987b], 259–271; L’énoncé ver-
bal en berbère. Étude de fonctions [1964], 287–308; Propositions relatives, rhématisation
et thématisation. L’exemple du berbère, 331–355. Leuven: Peeters.
Galand, Lionel. 2003. L’aoriste berbère, l’aspect et les valeurs modales. In Mélanges David Cohen,
Jérome Lentin & Antione Lonnet (eds), 235–46. 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.), 73–113.
Cambridge MA: The MIT Press.
Hagège, Claude. 1982. La structure des langues. Paris: PUF.
Haiman, John. 1978. Conditionals are topics. Language 54(3): 564–589.
Kossmann, Maarten G. 1997. Grammaire du parler berbère de Figuig (Maroc oriental). Paris:
Peeters.
Kossmann, Maarten G. 2000. Esquisse grammaticale du rifain oriental. Paris: Peeters.
Kouloughli, Djamel E. 1994. Indéfini et structure thématique en arabe. Faits de langues 4:
169–75.
Lafkioui, Mena. 2002. L’intonation et ses fonctions syntaxiques en rifain. In Articles de linguistique
berbère. Mémorial Werner Vycichl, Kamal Naït-Zerrad (ed.), 253–281. Paris: L’Harmattan.
Larcher, Pierre. 2000. Subordination vs coordination “sémantiques”: l’exemple des systèmes
hypothétiques de l’arabe classique. Annales Islamologiques 34: 193–207.
Mettouchi, Amina. 2002. La forme ad + aoriste en berbère (kabyle). In Articles de linguistique
berbère. Mémorial Werner Vycichl, Kamal Naït-Zerrad (ed.), 335–347. Paris: L’Harmattan.
Mettouchi, Amina. 2006. “Sujet” postverbal et état d’annexion en kabyle (berbère). Faits de Lan-
gues 27 [numéro spécial Les langues chamito-sémitiques (afro-asiatiques) Vol. 2]: 113–129.
Muller, Claude. 1996. La conjonction que: Rection vs. dépendance immédiate et concurrence
avec que pronominal. In Dépendance et intégration syntaxique. Subordination, coordina-
tion, connexion, Claude Muller (ed.), 97–111. Tübingen: Niemeyer.
Muller, Claude. 2001. Sémantique de la subordination: l’interrogation indirecte. In La séman-
tique des relations, André Rousseau (ed.), 163–177. Villeneuve d’Ascq: Université Charles-
de-Gaulle/Lille 3.
Naït-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 touarègue (tahăggart), VIII–IX, Syntaxe. Schwülper:
Cargo.
Ramat, Paolo. 1985. Typologie des langues. Paris: PUF.
Robert, Stéphane. 1991. Approche énonciative du système verbal (le cas du Wolof). Paris: CNRS.
Robert, Stéphane. 1996. Aspect zéro et dépendance situationnelle: l’exemple du wolof. In Dépen-
dance et intégration syntaxique. Subordination, coordination, connexion, Claude Muller
(ed.), 153–161. Tübingen: Niemeyer.
Robert, Stéphane. 2003. Polygrammaticalisation, grammaire fractale et propriétés d’échelle. In
Perspectives synchroniques sur la grammaticalisation. Polysémie, transcatégorialité et échelles
syntaxiques, Stéphane Robert (ed.), 85–120. Louvain: Peeters.
Catherine Taine-Cheikh
Robert, Stéphane. 2006. Deictic space in Wolof. Discourse, syntax and the importance of
absence. Typological Studies in Language 66: 155–174.
Simeone-Senelle, Marie-Claude. 2003. De quelques fonctions de đ- dans les langues sudarabi-
ques modernes. In Perspectives synchroniques sur la grammaticalisation. Polysémie, transca-
tégorialité et échelles syntaxiques, Stéphane Robert (ed.), 239–252. Louvain: Peeters.
Taïfi, Miloud. 1993. L’ expression de l’hypothèse en berbère. In À la croisée des études libyco-
berbères. Mélanges offerts à Paulette Galand-Pernet et Lionel Galand, Jeannine Drouin &
Arlette Roth (eds), 215–28. Paris: Geuthner.
Taine-Cheikh, Catherine. 1999a. Le zénaga de Mauritanie à la lumière du berbère commun. In
Afroasiatica Tergestina, Marcello Lamberti & Livia Tonelli (eds), 299–324. Padova: Unipress.
Taine-Cheikh, Catherine. 1999b. Topicalisation, thématisation et anaphore en arabe. In La thé-
matisation dans les langues, Claude Guimier (ed.), 247–261. Bern: Peter Lang.
Taine-Cheikh, Catherine. 2003a. La corrélation de gémination consonantique en zénaga (ber-
bère de Mauritanie). Comptes rendus du Groupe Linguistique d’études chamito-sémitiques
34 (1998–2002): 5–66.
Taine-Cheikh, Catherine. 2003b. L’adjectif et la conjugaison suffixale en berbère zénaga. In
Mélanges David Cohen, Jérome Lentin & Antoine Lonnet (eds), 661–674. Paris: Maison-
neuve & Larose.
Taine-Cheikh, Catherine. 2005a. Le rôle des phénomènes d’agglutination dans la morphogenèse
du chamito-sémitique. Exemples de l’arabe et du berbère. In Linguistique typologique,
Gilbert Lazard & Claire Moyse-Faurie (eds), 288–315. Villeneuve d’Ascq: Presses Univer-
sitaires du Septentrion.
Taine-Cheikh, Catherine. 2005b. Du rôle de la quantité vocalique en morphogénie. Réflexions
à partir de l’arabe et du berbère de Mauritanie. Faits de Langues 26 [numéro spécial Les
langues chamito-sémitiques (afro-asiatiques) Vol. 1]: 41–63.
Taine-Cheikh, Catherine. 2007. Les propositions relatives en zénaga et la question des relateurs
en berbère. In XII Incontro Italiano di Linguistici Camito-semitica (Afroasiatica). Atti, M.
Moriggi (a cura di), 301–310. Rubbettino: Medioevo Romanzo e Orientale.
Traugott, Elizabeth Closs. 1985. Conditional markers. In Iconicity in Syntax [Typological Studies
in Language 6], John Haiman (ed.), 289–307. Amsterdam: John Benjamins.
Vanhove, Martine. 2002. Conditionnelles et concessives en Arabe de YafiΩ (Yemen). In “Sprich
doch mit deinen Knechten Aramäisch, wir verstehen es!”. 60 Beiträge zur Semitistik. Fest-
schrift für Otto Jastrow zum 60. Geburtstag, Werner Arnold & Hartmut Bobzin (eds),
755–777. Wiesbaden: Harrassowitz.
Vycichl, Werner. 1992. Der Ursprung der Partikel ad- zur bildung des Konjunktivs, des Optativs
und des Futurum im Berberischen. Frankfurter Afrikanistische Blätter 4: 77–80.
Deixis and temporal subordinators
in Pomak (Slavic, Greece)*
Evangelia Adamou
LACITO (Laboratoire des Langues et Civilisations à Tradition Orale),
Fédération Typologie et Universaux Linguistiques, CNRS
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 speaker’s sphere (-s-), the addressee’s 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 addressee’s
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 anchor-
ing 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 Stéphane Robert; their work and comments were funda-
mental for understanding the Pomak temporal subordinators. I also thank Zlatka Guentchéva
and the two anonymous readers for their useful remarks.
1. Similar spatial-temporal uses are observed cross-linguistically, see for example the case of
Nêlêmwa 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).
Ag´ato/kug´ato ‘when (past)’ is used for a past moment in relation to the discourse
situation. Ag´ano/kug´ano ‘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; ag´a ‘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 kug´ato ‘as soon as (past)’ and li kug´ano ‘as soon as (future)’.
Pom´atsko ‘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 gen-
eration, men and women, have learnt Greek and Turkish at school, within the “Minori-
tariste” 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, subjunc-
tive, 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 con-
text of shift to Turkish, the Pomak language is subject to political and ideological con-
flicts 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).
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.
These three suffixes form the definite articles and the demonstratives presented
below (see §3.1.2 and §3.2.2).
4. Old Church Slavonic [OCS] refers to the eldest available Slavic documents, mostly eccle-
siastical. 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 dif-
ferences; 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 sys-
tem (Kanevska-Nikolova 2006, and fieldwork notes Adamou 2005–2006 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 & Keenan’s (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):
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 impor-
tant to specify.
ated in Bulgaria, ‘Macedonian dialects’ for those in FYROMacedonia - even though for Mac-
edonian 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.
Deixis and temporal subordinators in Pomak (Slavic, Greece)
1. For here and now situations the system depends on whether the entity is consid-
ered as being part of the speaker’s sphere, the addressee’s 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 crite-
ria as well. The objective distance can thus be manipulated following pragmatic crite-
ria, but such uses will not be explored in detail here.
-s-: ‘speaker’s sphere’
(1) ſe mi p´ane kinit´o-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-: ‘addressee’s sphere’
(2) bul´uz-at ti je h´ubaf
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 didn’t have the chance to record. The speak-
er’s 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 2005–2007.
Evangelia Adamou
The suffix -n- is used for objects that are not included in the participants’ spheres.
(3) j´ela nah kap´uje-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 addressee’s -t- article no longer concerns the addressee’s 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 sf´adba-na ſe n´adena-m tſerv´en-en fust´an
at wedding-def.dist mod wear-1sg red-def.dist dress
‘At the wedding I’ll wear the red dress.’ (Q, F2, 30, Pmk, Ell, Tur)
-t-: ‘past moment in relation to the moment of utterance’
(5) na sf´adba-ta beh sas tſerv´en-et fust´an
at wedding-def.loc.pas was.1sg with red-def.loc.pas dress
‘At the wedding I wore the red dress.’ (Q, F2, 30, Pmk, Ell, Tur)
-s-: ‘speaker’s sphere’
(6) na sf´adba-sa sam / ´ima mn´ogo ins´an
at wedding-def.s am.1sg / is.3sg a.lot people
‘I’m at the wedding. There are a lot of people.’ (Q, F2, 30, Pmk, Ell, Tur)
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
Deixis and temporal subordinators in Pomak (Slavic, Greece)
‘speaker’s sphere’
(7) ais´os je d´æd-va-sa k´oſta11
dem.s is.3sg grandfather-poss-def.s house
‘This (my sphere) is grandfather’s house.’ (Q, F2, 30, Pmk, Ell, Tur)
‘addressee’s sphere’
(8) ts´alo aitv´a ſe jad´e-ſ
all dem.a mod eat-2sg
‘You’ll eat all of this (your sphere).’ (S, F8, 21, Pmk, Ell, Tur)
10. izí (m.), isáy (f.), izí, isáy (n.) and iséy (pl.) for close entities, inazí (m.), ináy (f.), inazí, ináy
(n.) and inéy (pl.) for distant entities.
11. Note that the vowels /o/, /e/, /a/ when stressed are diphthongized [u´o], [j´e], [j´a], and,
when not stressed, are reduced.
Evangelia Adamou
‘distal’
(9) ain´os l´elka hi z´œla je b´eſ bin evr´o
dem.dist auntie poss.3sg.f took.prf aux.3sg five.thousand euro
‘That auntie of hers, she has got 5 000 euros.’ (S, F5, 15, Pmk, Ell, Tur)
‘addressee’s sphere’
(11) gjuŠl´utſi-te Š´œ-te n´osi-ſ par´atiki sa
glasses-def.a which-def.a wear-2sg ugly are
‘The glasses that you’re wearing are ugly.’ (Q, F5, 15, Pmk, Ell, Tur)
‘speaker’s sphere’
(12) gjuŠl´utſi-se Š´œ-se n´ose-m ais´a j´atse sa h´ubavi
glasses-def.s which-def.s wear-1sg now very are nice
‘The glasses that I’m wearing now are very nice.’ (Q, F5, 15, Pmk, Ell, Tur)
12. Translated literary texts of a religious and didactic nature, in what is referred to as
Modern and, for some of them, Middle Bulgarian (17th–18th century).
Evangelia Adamou
1. Presence of a deictic suffix indicates the anchoring of the utterance in the discourse
situation: ag´a-to, g´ato or kug´ato “when (past)”, ag´a-no g´ano or kug´ano14 “when
(future), whenever”. Absence of deictics indicates absence of anchoring in the dis-
course situation (ag´a). Thus, the temporal marker with no deictic suffix is used for
folk tales, jokes and other narratives, whether fictional or not. Ag´a seems to be tak-
ing 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 ‘addressee’s’ -t- suffix, in ag´a-to, is used for a ‘past’ event in relation to the dis-
course situation. The ‘distal’ -n- suffix, in ag´a-no, refers to generic or habitual time-
frames and future reference.
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; Guentchéva 1995).
13. This is not a unified analysis for all the Rhodope varieties using deixis in the tem-
poral 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 2005–2006). Moreover, a quick look at data from
Siroka Laka (Bulgaria) in Sobolev (2001), doesn’t 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, Kabasanov’s description of Tixomir
(1963), a Rhodope variety in Bulgaria, spoken by Muslims very close to the Greek border.
14. Variation between kug´a-no/to (based on the interrogative kug´a) and ag´a-no/to (based on
the temporal subordinator ag´a) depends on the speakers’ age and possibly education; older
speakers, or younger ones with little education, seem to prefer ag´a-, although I have not con-
ducted any precise sociolinguistic study on this topic. In any case, both forms are accepted by
the speakers when asked.
Deixis and temporal subordinators in Pomak (Slavic, Greece)
Note that the three temporal markers presented above are not specialized in a spe-
cific type of temporal interclausal relation: for example ag´ato is used for anteriority,
posteriority, simultaneity, terminus a quo, while all three markers can express anterior-
ity (for this terminology see Kortmann 1997).
To describe the temporal subordinator system of Pomak1, I will use the terminol-
ogy developed by Culioli (1971, 1978, and 1990) on the notion of situational anchor-
ing, 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 (ω).
Pomak1 temporal conjunctions express these relations as follows:
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 unmar-
ried 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.
Deixis and temporal subordinators in Pomak (Slavic, Greece)
25 the speaker gives two possible versions, changing the TMA markers while main-
taining the temporal subordinator ag´ano.
(24) ag´a-no ´ima sf´adba zb´ira sa s´elo
when-loc.hab.fut have.3sg marriage gather.3sg refl village
‘Whenever there is a marriage, the whole of the village gathers.’
(On local traditions) (Q, F2, 30, Pmk, Ell, Tur)
Also see Examples 15 and 16 illustrating the fact that the subordinator’s choice is not
determined by the TMA.
The constant use of ag´a in tales is also confirmed by other corpuses of the area, such as
the tales published in Theoharidis (1995) and Kokkas (2004b).
Evangelia Adamou
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 stud-
ies on the meaning and placement of li in Slavic languages). Example in Pomak1:
h´ubeve li ste? ‘Are you OK?’
j´ala li si? ‘Have you eaten?’
t am (from Turkish) ‘just as, as soon as’16, 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.
5. Polyfunctionality
5.1 Conditionals
ag´a: from ‘temporal’ to ‘hypothetical’
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
ak´u ‘conditional’
is is the most frequent conditional subordinator in Pomak1, used for both realis
Th
and irrealis clauses.
5.2 Causal
Ga: from ‘temporal’ to ‘causal’
´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 sub-
ordinate clause.
6. Conclusion
The study of the Pomak variety spoken in the Xanthi area (Greece) reveals an interest-
ing 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 Bulgaria
Macedonia Pomak
Albania
Greece Turkey
Abbreviations
References
Adamou, Evangelia. 2006. Le nashta. Description d’un parler slave de Grèce en voie de disparition.
Munich: Lincom.
Adamou, Evangelia. 2008. Sur les traces d’une dégrammaticalisation: Le médiatif en pomaque
(Grèce). Revue des Études Slaves: Proceedings of the French delegation at the XIV Interna-
tional Congress of Slavists 78(1–2): 177–189.
Adamou, Evangelia & Drettas, Georges. 2008. Slave. In Le patrimoine plurilingue de la Grèce,
Evangelia Adamou (ed.), 107–132. Leuven: Peeters.
Anderson, Stephen R. & Keenan, Edward L. 1985. Deixis. In Language, Typology and Syntactic
Description, Shopen Timothy (ed.), 259–308. Cambridge: CUP.
Assenova, Petya. 2002 [1st edition 1989]. Balkansko ezikoznanie. Veliko Tŭrnovo: Faber.
Bally, Charles. 1926. L’expression des idées de la sphère personnelle et de solidarité dans les
langues indo-européennes. In Festschrift Louis Gauchat, Franz Fankhauser & Jakob Jakob
(eds), 68–78. Aarau: Verlag Sauerlander.
Breu, Walter. 1994. Der Faktor Sprachkontakt in einer dynamischen Typologie des Slavischen.
In Slavistische Linguistik 1993, Hans Robert Mehlig (ed.), 41–64. Muenchen: Verlag Otto
Sagner.
Bril, Isabelle. 2002. Le nêlêmwa (Nouvelle-Calédonie): Analyse syntaxique et sémantique. Louvain:
Peeters.
Culioli, Antoine. 1971. À propos d’opérations intervenant dans le traitement formel des langues
naturelles. Mathématiques et Sciences humaines 34: 7–15.
Culioli, Antoine.1978 [1983]. Valeurs aspectuelles et opérations énonciatives: La notion
d’aoristique. In Enonciation: Aspect et détermination, Sophie Fisher & Jean-Jacques Franckel
(eds), 99–114. Paris: EHESS.
Culioli, Antoine. 1990. Pour une linguistique de l´énonciation. Opérations et représentations.
Paris: Ophrys.
Cyxun, Gennadij A. 1981. Tipologičeskie problemy balkanoslavjanskogo jazykovogo areala.
Minsk: Nauka i texnika.
Dimitrova-Vulchanova, Mila & Vulchanov, Valentin In press. An Article Evolving. In DIGS Vol-
ume, Stephen R. Anderson & Dianne Jonas (eds). Oxford: OUP.
Drettas, Georges. 1990. Le dialecte bulgaro-macédonien de Xr. (Edhessa, Grèce). Questions de
typologie. Bulletin de la Société de Linguistique de Paris 85(1): 227–265.
Elson, Mark J. 1976. The definite article in Bulgarian and Macedonian. Slavic and East European
Journal 20: 273–279.
Feuillet, Jack. 1999. Grammaire historique du bulgare. Paris: Institut d’études slaves.
Fielder, Grace E. 1985. Aspect and Modality in Bulgarian subordinate clauses. In The scope of
Slavic aspect, Michael S. Flier & Alan Timberlake (eds), 181–193. 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 Tomić (ed.), 101–134.
Amsterdam: John Benjamins.
Guentchéva, Zlatka. 1995. L’imparfait perfectif en bulgare. Modèles linguistiques 16 (2): 73–94.
Gălăbov, Ivan. 1950. Za člena v bălgarskija ezik. Izvestija na narodnija muzej v Burgas 1: 171–227.
Kabasanov, Stajko. 1963. Edin starinen bălgarski govor. Tixomirskijat govor. Sofia: Izd. na
Bălgarskata akademija na naukite.
Evangelia Adamou
Colette Cortès
Université Paris 7-Denis Diderot, Fédération 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 Cortès
–– The first level is sub-phrasal, comprising a “base” (i.e. a basic lexical node impos-
ing 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 Tesnière, with whom Jean Fourquet had long discussions at the
University of Strasbourg. Following Tesnière’s terminology, a “connexion” is a
dependency relation, involving a valency driven “actantial” relation or a “circum-
stantial” 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
Correlative markers as phoric “Grammaticalised Category Markers”
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 große Haus
‘the old big house’ is represented by a different connexion scheme from das große
alte Haus ‘the big old house’ in Figure (1).
–– On the second level, the syntactic sub-phrase which comprises the intercon-
nected “base” and its member(s) must be “overarched” by specific category mark-
ers (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. Fourquet’s
“categories” anchor the phrase in the discourse situation, involving an assertive
mood selected by the speaker, in accordance with its semantic and pragmatic rel-
evance within the sentence and the communication context. Fourquet’s “catego-
ries” (marked by a dotted line in Figure (1)) insert the phrase into the speaker’s
discursive and pragmatic process. These “categories” are the grammatical tools
developed by natural languages to express the metalinguistic and cognitive pro-
cesses 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 com-
parative markers. To avoid possible misinterpretations of the highly polysemic
term “category”, henceforth I will refer to Fourquet’s “categories” as “Grammati-
cal Category Markers” (abbreviated as GCMs).
Figure 1. A sample analysis of a noun phrase: das große alte Haus (the big old house)
–– 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 speak-
er’s 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 func-
tion and are so frequently found as subordinating or coordinating conjunctions
in German.
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 Example 1B
below, Wer das sagt, der lügt ‘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 usu-
ally not omitted. Wer einmal lügt, 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 syn-
tactic 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 conside-
ration: “Als Korrelate werden in der Literatur Ausdrücke unterschiedlicher Kategorie
bezeichnet, die in phorischer Funktion mit Nebensätzen sowie Infinitiv- und Partizi-
pialgruppen auftreten. Die wichtigsten Korrelate sind es, Formen des Demonstrati-
vums (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 “correla-
tive markers” belong to different morphological classes (pronouns, pronominalized
prepositional phrases and “adverbs”), which have a common phoric function analysed
under Section 2.2.
Colette Cortès
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 demonstra-
tive 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 anständig ‘well-behaved’ or am 9. November 1989), 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 mor-
phemes referring to the context or the co-text, and which may be used as phrasal substi-
tutes (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 gram-
maticalisation 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 Fourquet’s 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 reac-
tion 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 cor-
relative 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 doctor’s.)
Correlative markers as phoric “Grammaticalised Category Markers”
(2) b. Wie benahm er sich? Er benahm sich so, wie zu erwarten war.
(How did he behave? He behaved as expected.)
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 refer-
ring to the subordinate clause or to its relation to the complex sentence. This confirms
that it can be analysed as a Grammatical Category Marker (Fourquet’s “Catégorie”).
Like the other “Catégories” and according to the context, a correlative marker
may fulfilll a demarcative and/or a connective function. As expected, these func-
tions depend on the syntactic, semantic and pragmatic role of the subordinate clause.
Part 2 will provide further evidence for the fundamental observation that the correla-
tive 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 Anknüpfung sich, wenn man das Kognitionsgeschehen mitberücksichtigt, als
Spezialfall von situationsbezogener Anknüpfung erweist. Die besprochenen Sachverhalte (die
Nachricht) und das materielle Zeichen (der Text) sind nämlich nichts anderes als Teilnehmer
der Kommunikationssituation neben dem Sender und dem Empfänger, die ebenso wie letzterer
in der Kognition der Kommunikationspartner repräsentiert werden müssen, um als Substrat
für Anknüpfungen verfügbar zu sein. Mit anderen Worten : Bei der phorischen Anknüpfung
wird vorausgesetzt, dass im Laufwissen des Empfängers eine (eventuell noch unvollständige)
Repräsentation derjenigen Sachverhalte und Sachverhaltsbestandteile vorhanden ist, zu deren
Repräsentation die im Zuge des aktuell stattfindenden Kommunikationsereignisses bereits
übermittelten Zeichenbestandteile anleiten sollten”. Blühdorn (1996: 132)
Correlative markers as phoric “Grammaticalised Category Markers”
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 sub-
ordinate conjunction and ending with a conjugated verb-form) will be dealt with,
excluding relative clauses whose complex relations between the antecedent noun-
phrase 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.
Table 1.
(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 nämlich ‘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 [P´ [P, Q]] [P´ [P, Q]] [[P´ P] Q] interpolation
into P´ (ex. P´ = er glaubt, dass…
(P’ = ‘he believes that…’))
Colette Cortès
As Table 1 shows, Ducrot’s 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 subordi-
nate clauses and matrix sentences, and therefore called “bound” subordinate clauses
(“subordonnées liées”) by Ducrot, and two other classes (type 3 and 4) of “unbound”
subordinate clauses (“subordonnées non liées”) which all test negative, thus evidenc-
ing 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.
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 Valin’s model
(1984). Our “bound subordinate clauses” correspond to what the authors call “subordination”
([+ dependent], [+ embedded]) with its two subclasses, one being characterised as “core em-
bedding” (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).
Correlative markers as phoric “Grammaticalised Category Markers”
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 func-
tion of the subordinate clause, depending on the governing verb’s 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 two-
fold 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 clause’s 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
Realität bei hunderten von Firmen nicht mehr in 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 handlungsfähige 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” func-
tion, 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 Cortès
(8) (Type 3) Sowohl in der Politik wie in den Schriften zur Philosophie wird
Aristoteles nicht müde, den Wert aller theoriegeleiteten, also allein auf Anschauung
gestützten Disziplinen für die menschliche Sozialisation zu bekräftigen: “Obwohl
das hier erworbene Wissen doch theoretisch ist, so vollbringen wir doch unzählige
Handlungen nach seinem Muster, indem wir nach seiner Maßgabe das eine
Correlative markers as phoric “Grammaticalised Category Markers”
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 über seine rummelige
Altbauwohnung am Prenzlauer Berg, wo er doch die noble Dienstvilla in
Zehlendorf hätte beziehen können, die er Bundespräsident Johannes Rau überließ.
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
President Johannes Rau.)
–– Last but not least, in type 4 unbound subordinate clauses, which fulfilll a meta-
linguistic (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 Part 4 below).
(10) (Type 4) Wenn er brav ist, wofür alles spricht, wird er, einundsechzigjährig, 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 Identität und die unterstellten Motive angeht, so sind sie für 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, Ducrot’s 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 Cortès
For type 3 subordinate clauses, Ducrot’s tests produce evidence for a looser syn-
tactic 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 dis-
course 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 prag-
matically 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 Part 4 for the functions of d- correlatives
and in Part 5 for the functions of the so-correlatives. Before that, let us briefly inquire
into the possible status of the CGCMs.
5. Elemente der Kategorie Korrelatverbindung sind Paare, bestehend aus einem
phorischen/ deiktischen Ausdruck und einer Infinitivkonstruktion oder einem Subjunk-
torsatz, die in der Kette diskontinuierlich und gegebenenfalls akzentuiert auftreten können.
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 für die Verwendung als Korrelat überdeckt.
– die Infinitivkonstruktion / der Satz bleibt über die Regel als valenzgesteuertes Element
erkennbar. (Zifonun et alii 1997: 1487–1488).
Correlative markers as phoric “Grammaticalised Category Markers”
two interdependent parts of the same sentence constituent, but their interdependency
does not always have the same pragmatic function. Three possible solutions can
be imagined:
Part 4 (d- correlatives) and Part 5 (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 postdeter-
miner, 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, daß Länder wie Irland und Spanien, die noch
höhere 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 Enttäuschung ist groß”, 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 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 lan-
guage 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, depend-
ing 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)).
(14) Herzog: Dass es Staaten gibt, die hier sehr genau aufpassen, darüber gibt es
überhaupt keinen Zweifel. Spiegel 2000, n0 25, p. 445 (That some states are very
prudent on that matter, that is beyond any doubt).
(15) Daran, dass das Parken in der Stadt gebührenpflichtig ist, hat man sich längst
gewöhnt Spiegel 2000, n0 37, p. 255 (That we are charged for parking in a town,
that is something we have long been used to).
Correlative markers as phoric “Grammaticalised Category Markers”
(16) SPIEGEL: Warum dann noch die Präsentation im Fernsehen? Caballé: Das ist
wie ein Vorspielen. Wenn Intendanten von Konzerthäusern die Sendung sehen,
kann das der Weg zum Ruhm sein. Spiegel 2000, n0 1, p. 96 (Why still present
shows on TV? Mr. Caballé: 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 sub-
groups, 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 pro-
totypical 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 subordi-
nate 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 d-
correlative delimiting type 1A subordinate clauses is the last “overarching” gram-
maticalised 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 (Figure 2):
TAM Daran, dass das Parken in der Stadt gebuehrenpflichtig ist, man laengst sich gewoehn
Figure 2.
Colette Cortès
(17) Es wäre schön, wenn es ein besseres Beispiel gäbe. 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 prepo-
sitional 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 demar-
cative 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 “Deter-
minative, 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.
(19) Die Romane und Erzählungen kreisen immer wieder um die Katastrophe des
Gedächtnisverlustes. Weil Literatur für ihn ein einzigartiger Gedächtnisspeicher
ist, deshalb ist der Autor Lenz bis heute ein besessener Leser geblieben, wie seine
umfangreichen Essaybände 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 excep-
tional 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 Deutschland aufzuspüren, hat Winfried Kräling, Mitbe-
gründer der Anti-Licht-Kampagne, deshalb am Computer Satellitenfotos ausge-
wertet. Spiegel 1997, n0 5, p. 157 (This is to list ill-lit places in Germany, (that’s
why) Winfried Kräling, 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 subordi-
nate 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 differ-
ence 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 con-
tent 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 determi-
native CGCM (like dann in ((6) Dann, wenn dem Schriftsteller das Vollkommene
gelingt,) is a horematic, demarcative marker delimiting the entire prepositional cir-
cumstantial 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 “discontinu-
ous Sentential postposed CGCM”.
Colette Cortès
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”.
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 Table 2.
Table 2. d- CGCM
Type 1 A Bound Subordinate Clause d-Demarcative Phrasal CGCM
(actantial) (pre- (4, 15) or postdeterminer (13, 14))
(das/dar + Prep)
Type 1 B Bound Subordinate Clause Discontinuous Sentential postposed CGCM
(actantial) (das) (16)
Type 2 A Bound Subordinate Clause d-Demarcative Phrasal CGCM
(“circumstantial”) (pre- (5, 6) or postdeterminer (18))
(das/dar + Prep)
Type 2 B Bound Subordinate Clause Discontinuous Sentential postposed CGCM
(“circumstantial”) (7, 19, 20)
Type 3 Unbound Subordinate Clause Connective Sentential postposed CGCM
(argumentative) (21, 22)
In Part 5, we shall compare the function of the CGCMs dann and so in type 3
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 Cortès
(23) Obwohl er ein begnadeter Lyriker war und blieb, so gibt es doch im Spätwerk
Ausrutscher, die der junge Brecht, mit seinem untrüglichen Gespür für das
mot juste und den richtigen Vers, spielend vermieden hätte. 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 töricht wäre zu glauben, daß Demokratien keinen Krieg
miteinander führten, so steht doch fest, daß sie sich im allgemeinen sehr viel lang-
samer 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 hat’s 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 Großbritannien vor allem England paßt 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) Gegenüber Scharons Vorgänger Barak mag er sich gedacht haben, dass eine Show
des Widerstands nicht schaden könnte (…). Nur: Falls dies das Kalkül war, so war
es eine Fehlkalkulation. Denn die Situation entwickelte eine eigene Dynamik und
wurde von den Israelis für ihre Zwecke ausgenutzt. (Z02/204.02337 Die Zeit
(Online-Ausgabe), 25.04.2002, Nr. 18, Ressort: Politik; “Frieden ist noch immer
möglich”, S. 6) (Unlike Sharon’s 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.)
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 propo-
sition 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 “fac-
tual” and the “processual” relation to the “protasis”:
The contrast between dann and so as CGCMs in a hypothetical context can be under-
stood 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: 323–324) writes: “Die Nicht-Linearität der Ausführungen durch so
(steht) in Opposition zur Linearität der Ausführungen durch dann.” With the CGCM
6. Conclusion
The correlative markers in German are “grammatical category markers”, i.e. hore-
matic “categories” in Fourquet’s model, “overarching” the subordinate clause or
focusing its relation to the interpretation of the whole sentence, and enabling its cog-
nitive 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.
The semantic and pragmatic functions of the CGCMs vary with the type of subordi-
nate clause and the type of complex sentence, as shown in Table 3:
Table 3.
es CGCM d- CGCM so- CGCM
Type 1 A Bound Discontinuous d-Demarcative
Subordinate Clause Sentential CGCM Phrasal CGCM
(actantial) (3, 12) (pre- (4, 15) or
postdeterminer (13,
14)) (das/dar + Prep)
Type 1 B Bound Discontinuous Discontinuous
Subordinate Clause Sentential CGCM Sentential postposed
(actantial) (17) CGCM (das) (16)
Type 2 A Bound d-Demarcative
Subordinate Clause Phrasal CGCM
(“circumstantial”) (pre- (5, 6) or
postdeterminer (18))
(das/dar + Prep)
Type 2 B Bound Discontinuous
Subordinate Clause Sentential postposed
(“circumstantial”) CGCM (7, 19, 20)
Type 3 Unbound Connective Sentential Connective Sentential
Subordinate Clause postposed CGCM postposed CGCM
(argumentative) (21, 22) (8, 23, 24, 25, 26, 27)
Type 4 Unbound Connective Sentential
Subordinate Clause postposed CGCM
(commentative) (11, 28)
Colette Cortès
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 view-
point, 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 Fourquet’s “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 histor-
ical linguists like Jean Fourquet (1970), André 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 pragmat-
ics, 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 com-
binations 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 mark-
ers share the function of the grammatical category markers of modern German.
–– The German “Correlative Grammatical Category Markers” are chameleon-like mark-
ers: 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
Blühdorn, Hardarik. 1993. Deixis und Deiktika in der deutschen Gegenwartssprache. Deutsche
Sprache 1: 44–62.
Correlative markers as phoric “Grammaticalised Category Markers”
Blühdorn, Hardarik. 1996. Was ist Deixis? Linguistische Berichte. 156: 109–142.
Cortès, Colette & Rousseau, André. 1999. Catégories et connexions. En hommage à Jean Fourquet
pour son centième anniversaire, le 23 Juin 1999. Villeneuve d’Ascq: Presses universitaires du
Septentrion.
Cortès, Colette. 2006. Subordination et corrélation en allemand. In Coordination et subordi-
nation: Typologie et modélisation [Faits de Langues 28], Isabelle Bril & Georges Rebuschi
(eds), 107–118. 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
définitude [Recherches linguistiques de Vincennes 31], Anne Zribi-Hertz & Anne Dala-
dier (eds), 61–78. Presses universitaires de Vincennes.
Desclés, Jean-Pierre. 1992. Au sujet des catégories grammaticales. In La théorie d’Antoine Culioli,
Ouvertures et incidences, 203–212. Paris: Ophrys.
Ducrot, Oswald. 1972. Dire et ne pas dire. Principes de sémantique 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 Rehabi-
litierung eines naiven Nebensatzbegriffs. Kopenhagener Beiträge zur Germanistischen
Linguistik 18 : 1–45.
Foley, William & Van Valin Jr, Robert D. 1984. Functional Syntax and Universal Grammar.
Cambridge: CUP.
Fourquet, Jean, 1970. Prolegomena zu einer deutschen Grammatik. Pädagogischer Verlag
Schwann Düsseldorf.
Grünig, Blanche-Noëlle. 1963. Étude des démarcatifs en allemand moderne. Thèse de 3e cycle.
Paris Sorbonne.
Muller, Claude. 1996. La subordination en français. Le schème corrélatif. 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 Satzknüpfer (Konjunktionen, Satzadverbien und Partikeln). Berlin: Walter de
Gruyter.
Redder, Angelika. 1987. Wenn…, so. Zur Korrelatfunktion von so. Sprache und Pragmatik, 315–326.
Inger Rosengren (Hrgb). Lunder germanistische Forschungen Bd 55. Almqvist &Wiksell
International Stockholm – Sweden.
Rousseau, André. 1984. Apparition et grammaticalisation des formes verbales périphrastiques
en germanique ancien. In Romanistique-Germanistique, 97–130. Strassbourg: Presses
Universitaires de Strasbourg.
Sonnenberg, Bernhard. 1992. Korrelate im Deutschen: Beschreibung, Geschichte und Grammatik-
theorie [Reihe Germanistische Linguistik 124]. Tübingen: Max Niemeyer Verlag.
Tesnière, Lucien. 1959. Eléments de syntaxe structurale. Préface de Jean Fourquet. Deuxième
édition revue et corrigée. Cinquième tirage. Paris: Klincksieck [1988].
Zifonun, Gisela, Hoffmann, Ludger & Strecker, Bruno. 1997. Grammatik der deutschen Sprache
[Schriften des Instituts für deutsche Sprache]. Berlin: Walter de Gruyter.
part iv
Jean-Christophe Verstraete
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 lan-
guages, 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 François and Stéphane 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.
Jean-Christophe Verstraete
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 rela-
tions. 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 ‘compo-
sitional’ 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 rela-
tion 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
Blakemore’s [1987, 2002] notion of “procedural meaning”), with the search for an infer-
ence (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 Umpithamu’s 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 Rigsby’s work). Structurally, the lan-
guage 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.
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 im-
perative, 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 cross-
reference 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.’
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 promi-
nent (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 promi-
nence context that triggers ergative marking is the same context as the one that trig-
gers clause-initial position for pronouns, as in (3). This will be discussed in more detail
in Section 4, 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).’
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
Jean-Christophe Verstraete
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 repeti-
tion of minya adds further emphasis).
(6) mayi alu atha-n=iluwa
food neg eat-pst=3sg.nom
‘She didn’t eat food.’
Taken together, these positional restrictions produce the following basic clause template
(Figure 1), which is centered around the verb. Shaded linked areas represent alterna-
tive information-structural choices, both for pronouns and for nominal arguments.
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 restric-
tion 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.’
Focus, mood and clause linkage in Umpithamu (Cape York Peninsula, Australia)
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 sub-
ordinate 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 mean-
ing 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 NP-
modifying 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, how-
ever: 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
Jean-Christophe Verstraete
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 can-
didate 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 rela-
tions 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 Section 3, I will focus on mood marking, and in
Section 4, I will explore case and focus marking.
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 surpris-
ing 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 condi-
tion. 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 Section 3.1, I will analyse the role of potential verb suffixes in purpose
and conditional relations, and in Section 3.2, I will analyse the role of the apprehensive
marker in negative purposive relations.
Focus, mood and clause linkage in Umpithamu (Cape York Peninsula, Australia)
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 condi-
tional relation, i.e. the speaker’s death as a condition for her children’s inheritance in
(16), and approaching the Story Being (known as a Dreamtime Being in other parts
of Australia) on one’s 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”.
Jean-Christophe Verstraete
The second type is a purposive relation, illustrated in (18)–(19). Unlike with condi-
tionals, 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 restric-
tions 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 differ-
ences 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) wi’an 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
Focus, mood and clause linkage in Umpithamu (Cape York Peninsula, Australia)
with unrealized situations (see, for instance, Hengeveld 1998: 349–351), 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 speaker’s 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 pur-
posive 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 purpo-
sive 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 fea-
tures 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 sub-
ordinate.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).
Jean-Christophe Verstraete
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 con-
structions 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 combi-
nation with the meanings encoded by other features of the construction, for the con-
ditional relation it only provides a basic semantic prerequisite, which has to be further
enriched inferentially to arrive at the conditional interpretation.
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 pro-
hibition, 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.’
Even if the semantics of this construction links it to purposive structures, there are some
important structural differences. Unlike in purposive structures, the apprehensive-
marked verb has its own pronominal cross-reference, including for the shared argu-
ment, 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 precaution-
apprehension 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 sep-
arate 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
Jean-Christophe Verstraete
not”. Semantically, the contribution of the apprehensive marker to the negative pur-
pose 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.
pronoun complex is found in clause-initial position, as in (25), which describes the dis-
tinct 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) I’ll spear you. (Followed by) I’ll 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 sub-
jects. Thus, we find the ergative marker in classic prominence contexts like contrast
sequences, as in (28), where the transitive subject contrasts with the expected transi-
tive 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.’
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 suc-
cessive 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 rela-
tions. 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
into the semantics and pragmatics of clause linkage, because it shows a wide vari-
ety of mechanisms for exploiting markers from outside the domain of clause link-
age. 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 seman-
tic 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 Evans’s (2007) analysis of diachronic links between conditional markers
and deontic mood markers rather than the semantically expected epistemic type.
References
Hengeveld, Kees. 1998. Adverbial clauses in the languages of Europe. In Adverbial Constructions
in the Languages of Europe, Johan van der Auwera (ed.), 335–419. Berlin: Mouton.
Jacobs, Joachim. 1984. Funktionale Satzperspektive und Illokutionssemantik. Linguistische
Berichte 91: 25–58.
Lichtenberk, Frantisek. 1995. Apprehensional epistemics. In Modality in Grammar and Dis-
course [Typological Studies in Language 32], Joan Bybee & Suzanne Fleischman (eds),
293–327. Amsterdam: John Benjamins.
McGregor, William. 1992. The semantics of ergative marking in Gooniyandi. Linguistics 30:
275–318.
McGregor, William. 2006. Focal and optional ergative marking in Warrwa (Kimberley, Western
Australia). Lingua 116: 393–423.
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: 5–29.
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), 353–360. 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 Austra-
lia, Nicholas Peterson & Bruce Rigsby (eds), 192–218. Sydney: University of Sydney.
Robert, Stéphane. 2000. Le verbe wolof ou la grammaticalisation du focus. In Topicalisation et
Focalisation dans les Langues Africaines, Bernard Caron (ed.), 229–267. 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: 223–268.
Verstraete, Jean-Christophe. 2008. The status of purpose, reason and intended endpoint in the
typology of complex sentences. Linguistics 46: 757–788.
Verstraete, Jean-Christophe. 2009. Illocution and focus at the semantics/pragmatics interface in
Umpithamu (Cape York, Australia). 2009 Linguistics 47: 867–884.
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
Stéphane Robert
Fédération 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 inter-
esting verbal system expressing, among others, (1) distinctions related to information
hierarchy (focus), and (2) what I have dubbed “situational dependency”. Another char-
acteristic 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. Sall’s 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 (comple-
ment 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 subordi-
nating (or coordinating) morphemes, but it also contains some parts on paratactic structures.
Stéphane Robert
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 focus-
ing conjugations.
That is why, although A. Sall’s perspective is quite different from mine in this article (in par-
ticular 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.
Clause chaining and conjugations in Wolof
In the absence of the imperfective suffix (-y), these conjugations or verb inflec-
tions 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: 283–302.
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.
Stéphane 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, ante-
rior or negative suffixes.
Table 2. The aspectual and temporal meanings of verb forms in Wolof
Simple verb forms Suffixed verb forms
. 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: 270–2.
. 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 conjuga-
tion (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).
Clause chaining and conjugations in Wolof
People are talking about a man called Kebe and the crowd of people visiting him:
(4) Kebe moo am alal, mootax nit ñi 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 sub-
ject 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:
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, 306–307) or
Robert (2010).
. For details, see Robert 1991: 149–164.
Clause chaining and conjugations in Wolof
verbs expressing a quality or property, are commonly used with this focusing conjuga-
tion as mere statements serving for the predication of that property (7):
(7) Dafa liw. (stative verb expressing a quality)
vbfoc.3sg feel.cold
‘It is cold’. (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. (action verb)
13
vbfoc.3sg go
‘Actually, he left ~ it is because he left.’ (confirmation or explanation)
b. – Lutax ngay ñibbisi ?
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. Bëgg naa dem Dama bëgga dem
want prf.1sg go vbfoc.1sg want:conj go
‘I want to leave’ ‘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 perfec-
tive 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.
Stéphane Robert
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 Hal-
liday (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…’) cor-
responding 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 inter-
locutor”. Due to the properties of these Aktionsarten, the elimination of variation or
instability corresponds here to the elimination of the epistemic variation surround-
ing the predicate (for more details see Robert 1991: 52–67 and Robert 1994). So with
stative verbs, the Perfect indicates that there is no doubt as to the assertion, it simply
conveys the speaker’s viewpoint, with several possible contextual meanings: expected
conformity (13), the speaker’s 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
‘(It’s 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]
Tàng na!
be.hot prf.3sg
‘(I’m telling you) it is hot!’
[gloss: it is certain, there is no more discussion, there is nothing to be added, no
further comment, period].
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 dëkk bi.
prest.3sg return town the
‘Here he is (right here) coming back to the village.’ (he can be seen coming)
b. – Mu nga jëm ca dëkk ba, léegi 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.).
Null tense: on the one hand, this conjugation is common in proverbs (17) and obliga-
tory 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 consec-
utive or purpose clauses as in Examples (21)) without any subordinating morpheme.
(17) Ku Ø muñ, muuñ.
who null.3sg be.patient, smile
‘The one who is patient will smile.’ (Patience is rewarded) [SP]
(18) As soxna dafa amoon doom ju jigéen. Bi doom ji matee sëy mu maye ko. Yàlla def
xale ba ëmb…
‘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 Ø ñówaan, xale yépp
When null.3sg come:pastiter, children all
dañuy bég.
vbfoc.3pl:ipfv be.pleased
‘Whenever he came, all the children were pleased.’
(20) Dama bëggoon ngeen àndal maak sama doom.
vbfoc.1sg like:pst null.2pl accompany me:with my child
‘I would like you (to) accompany my daughter for me.’ [T: 169]
(21) a. Dafa sàcc, ñu 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]
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 fol-
lowing remarks only concern affirmative inflections.
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.
Section 5.
Clause chaining and conjugations in Wolof
Perfect-Null tense in juxtaposition is possible (25a), but the speaker naturally pre-
fers to add a temporal auxiliary (25b) in the following example:
(25) a. ? Ágg na, taw bi Ø door.
arrive prf.3sg, rain the null.3sg begin
‘He arrived, the rains started.’
b. Ágg na, taw bi Ø 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 experi-
ences (“discours” in Benveniste’s terms), while, on the other hand, the Null tense is
required for past events in narratives such as tales or historical narration (“récit” in
Benveniste’s 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 accept-
able in a verbal context where the Null tense clause could be interpreted as a con-
sequence of the Perfect clause: this would be a case of the Null tense being used as
a subordinator.
In addition, the chaining of two Perfect clauses (cf. 2.2), while perfectly correct
(Example 26), seems to be rare:
(26) Ágg 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 explana-
tion 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 inser-
tion of the personal pronoun “you” reinforces the parallel and contrast that (29) is
more acceptable than (28):
This succession effect for the Perfect is in accordance with its definition as indicat-
ing 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. Example 18 above), but produced differently. First, the Null tense
is used for expressing successive events in narratives rather than in discourse. Sec-
ondly, 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
Clause chaining and conjugations in Wolof
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 fol-
lowing example where the first event instantly triggers the second one:
(32) Mu ñëw, ma dem.
null.3sg come null.1sg go
‘Dès qu′il est venu, je suis parti.’ [Sall 2005: 267]
‘(As soon as) he came, I left.’
The chaining of two Verb Focus conjugations, although apparently rare, also seems to
be possible, as in the following example from Sall:
(36) Géej gi dafa aay, dañuy 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 indi-
cations 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 gener-
ally expresses a contrast between two parallel clauses and situations happening at the
time of speech:
(37) Maa ngiy génn, 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. 33–35), 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 Section 6. We can now summarize the
semantics of clause chaining with the same conjugation in Table 5.
Alongside this type of clause chaining, the Wolof system provides two particular
cases of special interest: clause chaining with focusing conjugations (Section 4) and
clause chaining with Null tense (Section 5). Apart from the use of negative conjuga-
tions 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 (Section 6).
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 previ-
ous clause or sentence (corresponding to the presupposition) to which it adds sup-
plementary 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 correct-
ing a preceding statement You went there which both correspond to the presup-
position (‘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 presupposi-
tion. 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:
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 indicat-
ing 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)
Stéphane Robert
or with a stative verb in the Verb focus form (39c), Verb focus is inappropriate with an
action verb (39b):
(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):
In fact, the common meaning of Verb focus (whatever the Aktionsart) in clause chain-
ing is that of explanations as in Example (8b) in a dialogue, or here in a complex
sentence (41):
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
Clause chaining and conjugations in Wolof
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 apo-
dosis (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) Ágg na, taw bi door na.
arrive prf.3sg, rain the begin prf.3sg
‘He arrived, (then) it started to rain.’
(43) Dafa àgg (rekk), taw bi Ø door.
vbfoc.3sg arrive (only), rain the null.3sg begin
‘(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 suc-
cession, 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 informant’s 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 càmmiñ waxal mboog, man dama yàkkamti.
My brother17 speak:imp therefore me vbfoc.1sg be.rushed
‘Come my friend, make up your mind (because) I am in a rush.’ [XCL]
(46) Moytul paaka bi, dafa ñaw de!
avoid:imp knife the vbfoc.3sg be.sharp ptcl
‘Be careful with the knife (because) it is sharp!’ [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] Yàlla dafa yéex te man dama yàkkamti ‘(it’s that) God is slow, (whereas) me, I’m in a hurry’.
. brother for a sister.
Stéphane 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 didn’t 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 reg-
ular 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 (‘some-
one 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
Clause chaining and conjugations in Wolof
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 as follows:
1. If the focused clause is not connected with any preceding clause (discursive land-
mark = ∅), 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 (Rʹ), it takes on causal meaning (8a and 8b).
The predicative relation (or the process) compared to which the focalized predica-
tive relation is posited, and which constitutes the discourse landmark, may thus be
absent (∅ = absolute initial position), identical (R) or different (Rʹ) from the predica-
tive relation presupposed by the focalized sentence (R). These various types of discur-
sive chaining produce the different meanings of Verb focus clauses as summarized in
Table 6.
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
relation’s 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.
Stéphane 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 care-
ful with the knife.
In the case of subject focus,19 the explanatory meaning rests on the same mecha-
nism, but the speaker also chooses to omit the link of the presupposed predicative rela-
tion, which is equivalent to explaining a situation by designating the author responsible
for another situation: thus in Example 3, in response to the question ‘what’s going on
here?’, the speaker answers ‘it’s 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 (preced-
ing 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, sub-
ordination 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.
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. Example 18). In discourse, it is found only in specific utter-
ances 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, embed-
ded in a previous clause with a different conjugation (Examples 20 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.
Clause chaining and conjugations in Wolof
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 speaker’s 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 specifica-
tion and the speaker’s 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 depen-
dency (lack of speaker’s 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 dis-
course coherence to embedding.
Table 7. Null tense-clauses and the syntactic integration of their locator
Syntactic integration Nature of the locator Nature of dependency
Ø
zero assertive dependency
different sentence situational anaphora
different clause embedding (purposive or consecutive)
Max. different verb 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 (199–234) or Robert 1996.
Stéphane 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 clause’s 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 trig-
gers 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 narra-
tives, rather than in discourse. In this case, the temporal succession of events is pro-
duced by “situational anaphora” and not by successive anchorings in speech-time: the
specification of a first temporal location (by another conjugation) triggers the valida-
tion 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 inescap-
able, 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:
Simultaneity and opposition
As mentioned in §2.2.4., the Null tense (with Ø) 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:
Clause chaining and conjugations in Wolof
(50) [A first spouse is talking to her husband about the bad behaviour of his
second spouse]
Ñeme na ñëw 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 imperfec-
tive 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 dóor ma!
prest.3sg find me: ipfv sleep null.3sg beat me
‘He finds me asleep and he hits me!’
(52) Paaka bi mu ngi ñaw 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 ëpp ba pare, nga
prest.3sg opr be.in.excess until be.ready, null.2sg
koy yokk!
opr:ipfv make.bigger
‘It’s 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.
Stéphane 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 Presen-
tative 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 detri-
mental. 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 unex-
pected turn of events is expressed, the only difference being that there is more insis-
tence on their synchronicity. The Presentative clause is often (but not necessarily)23
reinforced by rekk “just”:
(54) Mu ngi takk jabaram ba paré, Ø bàyyi ko fi!
prest.3sg bind wife:poss.3sg till be.ready null.3sg leave opr here
‘Hardly has he married his wife that he’s abandoning her!’
(55) Mu ngi tëj buntam rekk, xale yi Ø 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 per-
ceived 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 conver-
gent 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 Presenta-
tive 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 Example 23), Negative clauses (6)
or Perfect clauses (as in 31) for instance. A systematic study of discourse particles in Wolof is
still to be done.
Clause chaining and conjugations in Wolof
sizing a contrast or discordance between the two situations: the speaker’s 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):
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 speaker’s 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
Stéphane Robert
Appendix 1
G : Gancax gi, a TV play from the radio program Jamonoy Tey, broadcast by the
ORTS (Office de Radiodiffusion-Télévision du Sénégal) 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 Xët cig lëndëm, 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-Télévision du Sénégal) in 1986.
Abbreviations
References
Bril, Isabelle. In press. Coordination, information hierarchy and subordination in some Aus-
tronesian languages. In Converbs, Medial Verbs, Clause Chaining and Related Issues, Azeb
Amha, Christian J. Rapold, Sascha Völlmin & Silvia Zaugg-Coretti (eds). Frankfurter Afri-
kanistische Blätter.
Church, Eric. 1981. Le système verbal du wolof. Dakar [Documents linguistiques 27]. Dakar:
Publications du Département de linguistique de l’Université de Dakar.
De Smet, Hendrik & Verstraete, Jean-Christophe. 2006. Coming to terms with subjectivity.
Cognitive Linguistics 17(3): 365–392.
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 prédicatives en berbère. Doctorat d’état, INALCO-Université
de Paris 3.
Robert, Stéphane. 1990. Aperçu sur la négation en wolof. Linguistique africaine 4 [Documents
de travail sur la négation]: 167–180.
Robert, Stéphane. 1991. Une approche énonciative du système verbal: le cas du wolof [collection
Sciences du langage]. Paris : Editions du CNRS.
Robert, Stéphane. 1993. Structure et sémantique de la focalisation. Bulletin de la Société de
Linguistique de Paris LXXXVIII: 25–47.
Stéphane Robert
Robert, Stépane. 1994. Sur le rôle du sujet parlant dans la construction du sens: Liens entre
temps, aspect et modalité. In Subjecthood and Subjectivity, Marina Yaguello (ed.), 209–230.
Paris : Ophrys.
Robert, Stéphane. 1996. Aspect zéro et dépendance situationnelle: l’Exemple du wolof. In Dépen-
dance et intégration syntaxique (subordination, coordination, connexion), Claude Müller
(ed.), 153–161. Tübingen: Niemeyer.
Robert, Stéphane. 2000. Le verbe wolof ou la grammaticalisation du focus. In Topicalisation et
Focalisation dans les Langues Africaines, Bernard Caron (ed), 229–267. Leuven: Peeters.
Robert, Stéphane. 2004. À la recherche du sens grammatical: Contribution à une méthode d’en-
quête. In Langues et cultures: Terrains d’Afrique, Hommage à France Cloarec-Heiss [Collection
Afrique et Langage 7], Pascal Boyeldieu & Pierre Nougayrol (eds), 65–76. Louvain: Peeters.
Robert, Stéphane. 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), 233–260. Amsterdam: John Benjamins.
Sall, Adjaratou Oumar. 2005. La subordination en wolof: Description syntaxique. Ph.D. disser-
tation, Université Cheikh Anta Diop de Dakar.
Verstraete, Jean-Christophe. In this volume. Focus, mood and clause-linkage in Umpithamu
(Cape York Peninsula, Australia), 451–468.
Pragmatic demotion and clause dependency
On two atypical subordinating strategies
in Lo‑Toga and Hiw (Torres, Vanuatu)
Alexandre François
Langues et Civilisations à Tradition Orale,
Fédération Typologie et Universaux Linguistiques, CNRS
Despite the wealth of subordinators in Hiw and Lo‑Toga (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 well-
formed 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.
. The present work originates in a talk I gave in 2006 for the research group Typology of
interclausal dependencies (Fédération 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
Ministère de la Recherche (ACI “Jeunes Chercheurs”), is also gratefully acknowledged.
Alexandre François
ceanic languages are spoken there: Hiw by 150 speakers, and Lo-Toga – itself consist-
O
ing of two very close varieties Lo and Toga – by 650 speakers. They have never been the
subject of any published grammatical description.
Vanua B A N K S Is .
Mota
Lava
Gaua
Santo
Hiw Hiw (150) Maewo
[HIW] Santo
Tegua T O R R E S I S. Ambae Pentecost
Lo-Toga (650) Malekula
Lo [LTG] Ambrym
Toga
Epi
Efate Port-Vila
Erromango
Tanna
Aneityum
Hiw and Lo-Toga differ from each other in many regards, whether in their pho-
nology, their lexicons, or details of their grammars – enough to make them clearly
distinct, mutually unintelligible languages. Nevertheless, they also share parallel struc-
tures in most domains of their morphosyntax, their phraseology, and more gener-
ally 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
Pragmatic demotion and clause dependency
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 (Map 1),
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 speak-
ing, 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 clause-
linking 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.
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 Lo‑Toga.
. The spelling conventions adopted for the two Torres languages include the following:
g = [>]; n– = [ŋ]; n–w = [ŋW]; q = [kW]; d = [z]; r– = [:L]; o = [f]; ō = [o]; ö = [b]; e = [6]; ë = Ltg [7],
Hiw [e]; ē = Ltg [e], Hiw [I].
Alexandre François
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 ve… si 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.
. 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).
Pragmatic demotion and clause dependency
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 demo-
tion 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 syn-
tactic 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 behav-
iour of the Subjunctive, and Section 4 will be dedicated to the subordinating power of
the Background Perfect.
I will begin this study with an overview of the syntactic structures of the two Torres
languages, with special focus on clause linking strategies.
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 A‑M‑P
markers (for “Aspect-Mood-Polarity”), throughout this chapter, I shall nevertheless
continue to use the widespread abbreviation TAM (for “Tense Aspect Mood”), for the
reader’s 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 Nega-
tive. The irrealis categories (see §3.3) include the Future, the Prospective, the Poten-
tial, 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 (François, in press). It covers several mean-
ings, both realis and irrealis, including narrative, sequential, generic, prospective,
imperative and conditional. A possible description of the Aorist would be to consider
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 mod-
ally bound dependent clauses (12). Interestingly, the Subjunctive [Hiw on, Ltg vë(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.
When the subject is omitted, the result is a clause that consists of just a single noun
phrase:
(6) Hiw (Ø) { ne wake }.
art canoe
‘(It’s) 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 yö kön- in (54)] – as well as certain invariant words [e.g. the
existential predicate ‘not exist, lack’ in (7)].
(7) Ltg Ne hen-wëvot mino { evë }? – Nie { tategë }.
art knife my where 3sg neg:exist
‘Where (is) my knife? – It is not here.’
. In Examples (5)–(7), the limits of the predicate phrase are indicated by curly brackets.
Alexandre François
2.2 Subordination
Hiw and Lo-Toga possess a wide array of morphological devices for encoding the syn-
tactic 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).
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 complemen-
tizer. Indeed it can combine not only with verbs of speech, but also with all sorts of
verbs governing a clause complement:10
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 comple-
ment 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 (Hiw ta, Ltg si):
. 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).
Pragmatic demotion and clause dependency
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 Pro-
spective. 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 të, 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 com-
bining the Prospective (here të we ‘Prosp:2sg’) with the complementizer të in the same
sentence:
‥
(13) Ltg Tate pero të nike t we hadit.
neg:real long comp 2sg prosp 2sg be.initiated
[lit. It’s not long before you’re 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 (të + Aorist) with the particle ake – see (15), (26), (32a).
. Both the morphology and the semantics of the Lo-Toga Prospective are narrowly similar
to those of the Prospective in Mwotlap (François 2003: 218–257).
Alexandre François
The conditional subordinator also displays longer forms which are derived from the
complementizer. One thus finds the (semantically non-compositional) combination
Hiw tom + n-wë ‘like’ → tom-n-wë or tom-n-wë-tom meaning ‘if ’ – see (49). Lo-Toga has
exactly parallel forms, either morphologically transparent (të + wë ‘like’ → tëwë [t7w7]
‘if ’) or with a slight vowel change tëwë → tewë [t6w7] ~ tewë-të [t6w7t7] – see (15), (48).
Several TAM categories can be found in the protasis of a conditional sentence:
Aorist; Subjunctive; Counterfactual (15):
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.
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 Hiw petom ~ Ltg petë, etymologically the combination
{relativizer + complementizer} [also see (41) below]:
Lo-Toga also combines the relativizer pe with the comparative wë ‘like’ (→ Ltg pewë),
generally with virtual or generic referents (whoever…):
Pragmatic demotion and clause dependency
In fact the form wë 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 Lo‑Toga are pe,
petë, pewë or wë.
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 verb’s 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).
But it also commonly happens that the same word appears on its own, with no overt
relativizer:
(20) Ltg Mowe ne tarepi ēke mat tëh pah,
time/when art body canoe cplt carve finish
pahvēn ge rak ne hēm’ 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 François
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 nonegë ‘when, as’:
(21) Ltg Nonegë nie ve vin-gë ne megole,
as 3sg ipfv climb‑appl 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 rela-
tion 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 r̄ekn-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 coor-
dination between any two phrases, including two prepositional phrases (23) or two
clauses (24):
(23) Ltg Noke na melekelake pi megole mēke, mi pi lëgie mēke.
1sg stat happy about child your and about wedding your
‘I’m delighted about your baby, *with/and about your wedding.’
(24) Ltg Ne n-wië si dahia ē ne tēle,
art devil pot harm obl art person
mi nihe si kur verië ne tēle.
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 subordi-
nator, is commonplace in the area. Mwotlap does the same with (vēt)mahē (François 2003: 26),
as well as Bislama with taem < Eng. time (Crowley 2004: 188).
Pragmatic demotion and clause dependency
Other coordinate constructions include words for ‘but’ (Hiw/Ltg pa), ‘or’ (Hiw titom,
Ltg hitë), or ‘because’ (Hiw [ur–] nëpe [tom], Ltg nawë).
In this pattern of nuclear-layer serialization, the second verb modifies the first verb,
both semantically and syntactically (Bril 2004; François 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:
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 V1‑slot, encodes progressive aspect:
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.
The preceding section showed the wide array of formal devices used by the two Tor-
res 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 Back-
ground 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 Section 4.
. 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.
Pragmatic demotion and clause dependency
3.1 Presentation
The Subjunctive was first exemplified in sentence (1), reproduced below:
(1) Hiw Ne temët on tō yaqe me n-wë ne,
art devil sbjv go:sg appear hither like this
tekn-wa voyi.
people aor:run.away
[lit. The devil would appear like this, people ran away]
‘(Whenever) the devil appeared, people would run away.’
The behaviour of the Subjunctive is parallel in Hiw (form on) and in Lo‑Toga (forms
vë ~ vën).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.
. Despite the formal difference between Ltg vë [β7] ~ vën [β7n] and Hiw on [fn], it is in fact
likely that the two forms are cognate. According to regular vowel correspondences (François
2005b), they could reflect a proto-form *Äβani, 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 et al.
1994), and the etymology of English if (< OE ġif), sometimes mentioned as connected to giefan
‘give’, is disputed.
Alexandre François
For example, let us consider the state of affairs {Baby get 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 Dik’s 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, pre-
diction, etc.:
(31) a. Eng I fear Baby might get sick.
b. Eng I wish Baby got sick!
c. Eng [Given what I know, I hereby predict that] Baby will 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 Baby get sick?
It may also be encapsulated within a conditional structure, either as the protasis (31e–f)
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 impera-
tive form suppose has grammaticalised into a subordinator sipos/sapos meaning ‘if ’ (François
1997: 22; Mühlhäusler et al. 2003: 24; Crowley 2004: 189).
Pragmatic demotion and clause dependency
(31) e. Eng In case Baby gets sick, he will need to take this medicine.
f. Eng Every time Baby gets sick, he tends to recover within two or three days.
g. Eng If he goes out in that cold weather,
[I hereby predict that] Baby will 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.
. 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 Prospec-
tive (32c).
Alexandre François
that linguistic operation intrinsically, this is not the case for the Subjunctive (33a–b),
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 Sub-
junctive 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 con-
sequence will be that the Torres Subjunctive is capable of creating a relation of depen-
dency 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.
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:
This functional equivalence between the Aorist and the Subjunctive is also found with
third-person optatives:
(36) Ltg Ne ten-wēte vën toge mē-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 wit-
nessed 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 defi-
nition 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 (35a–b) and (36) are perfectly well-formed, while (33a–b) 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 (33a–b) 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 (33a–b), that of a declarative statement, results in the
pragmatic incompleteness of the sentence. On the other hand, the prosody of (35a–b),
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 (33a–b) showed that the function of the Subjunctive, namely the mere
representation of a virtual State of affairs, did not find enough support in the declara-
tive modality to constitute a well-formed utterance. Conversely, what (35a–b) reveals is
that an intonation typical of orders and exclamations, because it is markedly anchored
in the speaker’s 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 François (1997: 66) to explain why certain lan-
guages encode their imperative with some linguistic structures (noun phrases, infinitives, sub-
junctive clauses…) which would constitute an ill-formed declarative sentence. Despite their
Pragmatic demotion and clause dependency
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 well-
designed for the formation of a valid utterance – cf. (32a–g) – 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 corre-
spond 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 (35a–b) and (36), they illustrate a more particular case, where the spe-
cific 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.
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 François
introduced by means of a complementizer (Hiw tom, Ltg të), after a verb of manipula-
tion or expectation (see §2.2.1):
(37) Hiw Mar̄enage sa gatēt ti tekn-wa
chief their say dat people
tom ne ver̄oye 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 sëh t ne gengën
1incl:pl stay wait comp art food
vë howse pah.
sbjv cooked finish
‘Let’s 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.
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 tayö on mët,
time rel art person sbjv die
tite tivig n’ opë-ne.
1incl:pl bury art body-3sg
‘When(ever) somebody dies, we bury their body.’
‥
(44) Ltg Mowe w si tēle vë mōo, dege leklok mē.
time rel some person sbjv sick 1incl:pl help with.3sg
‘When(ever) somebody gets sick, we help them.’
An irrealis clause can be embedded within another irrealis clause, in which case the
Subjunctive percolates throughout. (47) shows three instances of vë(n): the first one
(vën itë) is due to the semantic status of the time clause as generic (‘whenever’); the
next two (vë sōw vë lewō) constitute a second level of subordination, being a comple-
ment clause within that time clause [see also (51) below]. Incidentally, the string /vë
sōw vë lewō/ is a serial verb construction, of the type that requires the repetition of the
TAM marker (see §2.4):
‥
(47) Ltg {Mowe kemëm vën itë [t ne ho in
time/when 1excl:pl sbjv see comp art leaf its
vë sōw vë lewō pe si ] },
sbjv grow sbjv big already prf
alē kemë ge lio.
then 1excl:pl aor:pl dig.up
‘When(ever) we see that [the taro’s] leaves have grown (and become) big,
we dig it up.’
Alexandre François
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 sub-
ordination – within a higher level predication, which is in turn specified for modality
and illocutionary force.
It is true that locative phrases – including prepositional phrases like yö kön̄ ‘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 There-
fore, the combination of the subjunctive on with the phrase yö kön̄, 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 func-
tion 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” (François 2005a: 192).
Alexandre François
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 Tekn̄wa on n̄wuye me ton ne r̄ekove sa,
people sbjv return hither from art work their
s’ on vën wate me, se vën se motr̄ig.
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,
they would go to sleep.’
Pragmatic demotion and clause dependency
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
tōt vë wureri and gët vë menō – 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
Pragmatic demotion and clause dependency
that divides V1 from V2 in each construction is the Subjunctive vë. Phonologically speak-
ing, the strings { V1 vë V2 } 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 { V1 vë V2 } rather than
inserted in between.
(65) Ltg Dōr si gët vë menō ne gi ne.
1incl:du pot chew sbjv soft art kava this
‘We can chew this kava soft.’
(66) Ltg Dege të ge lōv vë n̄wedōl
1incl:pl prosp pl:s call sbjv short
ne iē të ‘Alex’.
art your.name quot (name)
‘We shall (pronounce shortly =) shorten your name to Alex.’
Functionally as well as formally, these strings { V1 vë v2 } have a lot in common with
serial verb constructions (§2.4), the only difference being that the TAM marking dif-
fers 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 (gët-vë-menō ‘soften by chewing’; lōv-vë-n̄wedōl ‘shorten’), as
through a process of lexical compounding.
Arguably, the form vë 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 Dōr si gët-vë-menō ne gi ne.
1incl:du pot chew-result-soft art kava this
‘We can “soft-chew” this kava.’
Interestingly, Lo‑Toga is the only language in north Vanuatu that has developed this
pattern of resultative structure, using a buffer morpheme like vë. All its neighbours –
including Hiw – would simply construct their resultative macro-verbs by resorting to
a simple pattern of nuclear-layer serialization (François 2004, 2006). Thus, the equiva-
lent of (65) in Mwotlap would be kuy madamdaw na‑ga/chew soft art‑kava/, with
nothing between the two verb radicals.
. Note that the variant vën is never attested in these new structures. This tends to confirm
that the Subjunctive marker has adopted a new grammatical status here.
Alexandre François
While sentences like (64)–(66) are still somewhat ambiguous and compat-
ible 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 vë is the dummy auxiliary da ‘do’ (also ‘be’), which does not exist as an inde-
pendent verb. The string da‑vë‑ thus serves as a productive prefix in Lo‑Toga for
the formation of causative (transitive) verbs out of stative (intransitive) verbs or
adjectives (Table 1).
Once again, these examples are open to two morphological analyses. It is still possible
to consider them compoundings between two lexical roots (da ‘do’ + mōo ‘sick’), hence
the gloss /do‑result‑sick/. 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
Pragmatic demotion and clause dependency
unit (V2) with a causative prefix davë‑. In the latter case, one could transcribe (67) as
davë‑mōo and gloss it /caus‑sick/.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 (Table 2).
Table 3. The narrow ties between the Subjunctive and clause dependency: A summary
Syntax Functional value Examples
no subordination hortative & optative (3sg) (35)–(36)
modally-bound complement clauses (37)–(38)
purpose clauses (39)–(40)
combines irrealis & generic relative clauses (41)–(42)
with subordinators irrealis adverbial time clauses (43)–(47)
irrealis conditional protases (48)–(49)
irrealis conditional protases (50)–(52)
[Hiw] reinterpreted as conjunction if (53)–(54)
irrealis & generic adverbial time clauses (55)–(59)
directly encodes irrealis & generic relative clauses (60)–(61)
subordination irrealis purpose clauses (62)–(66)
[Ltg] resultative compounding (65)–(69)
> causative derivation
. 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 François
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 tēle si mat mēt.
art coconut.crab bkpf1 iter~crunch person bkpf2 cplt die
[lit. The coconut 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 infor-
mation, which will typically result in a syntactic relation of dependency between clauses.
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:
Pragmatic demotion and clause dependency
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 Remë mē ve kerë.
mother his ipfv weep
‘His mother is/was weeping.’
(73) Ltg Nihe ve lōv nie të “Temētrōn̄”.
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 + ve Action verb V2 }: see §2.4, ex. (28a).
Verbs that are lexically stative (including adjectives) are sometimes found to com-
bine with the Imperfective, in which case they take on a dynamic reading:
(74) a. Ltg Ne vete na medudut.
art place stat black
‘It’s dark.’ [stative reading]
b. Ltg Ne vete ve medudut.
art place ipfv black
‘It’s 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.
. Beside the widespread form ve [β6], Lo-Toga also possesses a rare variant me [m6]; like-
wise, me…si constitutes a (rare) variant of its Background Perfect ve…si. Incidentally, there is
no reason to suspect any etymological connection between the element ve [β6] of the Imper-
fective and the Lo-Toga form of the Subjunctive vë [β7]: 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 në and ve – as optional (see Table 4). Quite often, a
Alexandre François
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 speaker’s point of view – to consider
each compound TAM marker as a single meaningful morpheme, albeit a discontinu-
ous one. As a result, while the form na alone was glossed stat(ive), I shall gloss the
sequence na…si as prf1…prf2, 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 con-
sideration 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.
Because they both point to the resultant state that follows the final boundary of a com-
pleted state of affairs, they are equally compatible with stative and with dynamic predi-
cates. 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, Lo‑Toga 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 (François 2003: 30 sqq, 343). Incidentally, most of the compound forms
of Mwotlap involved postclitic tō [t~], with which the Torres form ti/si is cognate.
Pragmatic demotion and clause dependency
dynamic verb gil o ‘dig out’ is incompatible with the Stative na (→ *na gil o), it can
perfectly take the standard Perfect which is derived from it (→ na gil o 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 dis-
tinct 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 (Table 4). This use of
TAM markers for coding informational hierarchy is typologically original.
Table 4. Hiw and Lo-Toga have two Perfects; their difference lies in the pragmatic status
of the predicate
Hiw Lo-Toga Pragmatic status of predicate
As for the Background Perfect (Hiw ve…ti, Ltg ve…si), it also construes a realis per-
fect predicate, but explicitly specifies its informational status as pragmatically pre-
supposed, i.e. defocused. Crucially, a predicate phrase marked with the Background
Perfect (henceforth “BkPf ”), due to this backgrounded status, cannot constitute a well-
formed utterance on its own:
(76) b. Hiw *Ve r̄ak ti.
bkpf1 make bkpf2
*{ (somebody) made it… }[background]
. This freedom of actionality combinations provides further support for the view ex-
plained above, that the two perfects should not be analyzed compositionally, but as (discon-
tinuous) TAM markers in their own right, with specific properties.
Alexandre François
In contrast to (76a) në r̄ak ti, 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:
(77b) could be described as a case of clause topicalization.30 The event marked as Back-
ground 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).
. See Givón (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 Givón 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 (François, forthcoming).
Pragmatic demotion and clause dependency
As we shall see in §4.2, the presupposed predicate quite often involves genuine subor-
dination, e.g. a relative clause:
(78) Ltg Lōwie ē leqëvine meke { nie ve rak si
thanks obl woman your 3sg bkpf1 make bkpf2
ne tōtōgalē }.
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 sentence’s structure is quite
parallel to its English counterpart, including the contrastive focal stress that affects the sub-
ject 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 proper-
ties 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 Lo-
Toga and English in (79) is an optical illusion. It will appear that (79), like all contras-
tive 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.
any further formal device (§4.2.2). The special syntax of contrastive focus structures
will be examined in §4.2.2.2.
In each of these two sentences, the relative clause is unambiguously marked as subordi-
nate 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.
Pragmatic demotion and clause dependency
The regular (assertive) Perfect is extremely rare in relative clauses. This configura-
tion 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 non-
standard situation:
(84) Ltg Hen̄were pah tat lōlmerën ē.
people all neg:irr know obl:adv
‥
Hen̄were { w na huqe weren̄o si }
people rel prf1 initiated only prf2
nihe ve lōlmerën ē.
3pl ipfv know obl:adv
‘Not everybody would know (these things).
Only { those who’ve been initiated }[focus] know[background].’
What, syntactically, forms the main clause (nihe ve lōlmerën) of the whole sentence is
here a mere repetition of the previous sentence, with no informational weight. Con-
versely, the sentence’s 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.
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.
We can now account for Example (2), which was quoted in §1.2:
(2) Ltg Ne gehuh ve kerkur tēle si mat mēt.
art coconut.crab bkpf1 iter~crunch person bkpf2 cplt die
[lit. The coconut 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 for-
mal 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 mor-
phologically 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:
position of the object noun phrase (ne keka tekële) would probably be explained as a
form of left-dislocation. However, this analysis does not hold, for two reasons: for-
mally, 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 remem-
bers 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 con-
stitutes 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 rela-
tive clauses in these languages. They can be embedded within a noun phrase, a prepo-
sitional 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. ē
‘there, from it’):
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.
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 (viēne ‘underneath’, ē ‘there’) whose function is to indicate the syntac-
tic 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 vēn wahe ‘he reached’ and ni gerage ‘he
climbed’.
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
sentence’s main predicate – that is, the nominal predicate tekn̄wa tamesō ‘(it is) 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 tekn̄wa tamesō. It is followed by a distinctive fall in pitch and inten-
sity on the remainder of the sentence, which is typical of presupposed elements in
cleft constructions:
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 Tekn̄wa te Toge ve r̄ak 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 ve… ti, which is pragmatically presupposed in the context,
would thus be a relative clause with no relativizer, as in (75b) above. The phrase tekn̄wa
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:
[t6kÃŋWa t6 Ätf>6 ↓β6 :lak n6 >6nÄ>fn ti]
(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 tekn̄wa te Toge ve r̄ak ne gengon ti.
neg:real people from Toga bkpf1 make art meal bkpf2
[negation of contrastive focus pattern]
‘{ It’s 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 predi-
cates: 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. ‘{ (it’s) 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 (ve tot ti).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
‘{ (It’s) not I }[focus] { (who) carved it}[bkg]. (It’s) my father.’
. Ex. (8) above illustrates the same pronoun noke ‘[it’s] 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. n·i-d·eremu n·i‑ga‑r·ud·u-ŋi,
literally ‘[it’s] 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.
Pragmatic demotion and clause dependency
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):
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 indi-
cated 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):
(96b) shows fronting of the focal element, in the form of a predicate noun phrase (ne
mesale pek ‘[it is] 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 ē (‘there, through it’) – in
accordance with the typical syntax of relative clauses, as in (87) above. The resulting
Alexandre François
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 {Ø we came
through Ø}.
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.
Such a formal TAM contrast between (97a–b), 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.
Table 5. The close links between the Background Perfect and clause dependency: A summary
Syntax Functional value Examples
no subordination clause topicalization & backgrounding (77b)
realis background (restrictive) relative clauses (82)–(83)
combines
with subordinators realis background clause in cleft focus patterns (89)
realis background (restrictive) relative clauses (85)–(88)
directly encodes
realis background clause in cleft focus patterns (90)–(96b)
subordination
question sentences if wh‑word 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 subor-
dinators, 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 clause’s 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 François
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 demo-
tion, 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 clause’s 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 affirmative ltg Lo-Toga
aor Aorist m masculine
appl applicative neg:exist Negative existential
art article obl oblique
bkpf Background Perfect poc Proto Oceanic
caus causative poss possessive marker
comp complementizer pot Potential
cplt Complete aspect prf Perfect
ctfc Counterfactual prosp Prospective
du dual quot quotative
foc focus marker rel relativizer
fut Future result resultative
ipfv Imperfective s subject clitic
irr irrealis sbjv Subjunctive
iter iterative stat Stative
hiw Hiw tr transitive verb
loc locative marker
Pragmatic demotion and clause dependency
References
François, Alexandre. 2005b. Unraveling the history of the vowels of seventeen northern Vanuatu
languages. Oceanic Linguistics 44(2): 443–504.
François, 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), 223–238. Oxford: OUP.
François, Alexandre. 2009. Verbal aspect and personal pronouns: The history of aorist markers
in north Vanuatu. In A Festschrift for Robert Blust, Alexander Adelaar & Andrew Pawley
(eds), 103–126. Canberra: Australian National University.
François, Alexandre. Forthcoming. From deictics to clause linkers. Discourse deixis, topicaliza-
tion and clause backgrounding strategies in the languages of the Banks islands (Vanuatu).
Studies in Language.
Givón, Talmy. 1984/1990. Syntax. A Functional-typological Introduction. Amsterdam: John
Benjamins.
Haspelmath, Martin. 2007. Pre-established categories don’t exist – consequences for language
description and typology. Linguistic Typology 11(1): 119–132.
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), 89–110. 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 omniprédicative: Essai sur la morphosyntaxe du nahuatl
classique. Sciences du Langage. Paris: CNRS.
Lemaréchal, Alain. 1989. Les parties du discours. Syntaxe et sémantique [Linguistique Nouvelle].
Paris: Presses Universitaires de France.
Mühlhäusler, Peter, Dutton, Thomas E. & Romaine, Suzanne. 2003. Tok Pisin 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.), 42–140. Cambridge: CUP.
Stassen, Leon. 2000. And-Languages and With-Languages. Linguistic Typology 4: 1–54.
Thompson, Sandra A. & Longacre, Robert E. 1985. Adverbial clauses. In Language Typology and
Syntactic Description. Timothy Shopen (ed.), 169–234. Cambridge: CUP.
Tomlin, Russell. 1985. Foreground-background information and the syntax of subordination.
Text 5: 85–122.
van der Auwera, Johan & Plungian, Vladimir. 1998. Modality’s semantic map. Linguistic Typology
2: 79–124.
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
Université Paris Descartes, Laboratoire des langues et Civilisations à
Tradition Orale, Fédération Typologie et Universaux Linguistiques-CNRS
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 construc-
tions, 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 con-
tained in certain complex sentences.
Section 2 gives a brief typological overview of Mankon; Section 3 presents the affir-
mative 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. |tàŋ¢|
“for… to, in order to, so as to” in 5. and |bé| “if ” in 6.) or else the obligatory presence of
the conjunction is not relevant (cf. |á| “that” in 4 and 5). Section 7 presents several com-
plex sentences where the verb, either in initial position or in the following clause, has
Jacqueline Leroy
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, ε, 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)),
/і‒/ 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 /і‒/ 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 + і‒ and Vn + a where V is an open vowel (7, a, f) and + is a mor-
pheme 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, \, ŋ/. The inventory of final
consonants (C2) is limited to /m, n, ŋ, b, r, >, ô/. /i/ induces palatalization and /u/ labio-
dentalization 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, \, ŋ/ 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 syllable’s tone may be repeated on the following syl-
lable. These factors have the following consequences:
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 (h↑h 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 (↑h↓h) and upstep (h↑h):
|à-láô` á bì‒-l¡6m` a| àl↑áô↓á bí‒l¡6m¡6
7-country 7 2-sorcerer dm
‘The country of the sorcerers.’
|à-bùô` á bì‒-kúm΄ a| àbùô ¡6 bí‒k↑úm΄6
7-slave 7 2-notable dm
‘The slave of the notables.’
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 noun’s 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 plu-
ral 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 any modification.
–– 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:
–– When a noun is in the locative (i.e. when it is governed by the locative preposi-
tion |á|), 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
2.2.2 Pronouns
These belong to several categories:
–– Person (1st and 2nd persons) and class (3rd person) subject pronouns (cf.
Section 3)
–– 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, |-nі‒|, |-kі‒|, |-sі‒|
and |-tі‒| 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 |-nі‒| suffixes are to be distinguished: |-nі‒1| and |-nі‒2| and two
|-kі‒| suffixes: |‑kі‒1| and |-kі‒2|. Without entering into too much detail, it may be
said that |-nі‒1|, |-kі‒1| and causative |-sі‒| change the verb’s valency, |-nі‒2| and |-kі‒2|
respectively imply collective and distributive plurality of one of the arguments,
|-tі‒| 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 Example 12 in Section 5).
–– 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. Section 4.). 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 repre-
sents 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 à 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
–– Mankon uses prepositions. Many are derived from locative phrases (for example:
|á à-tú ‒í `| át↑ú↓, literally: loc 7-head loc, “on”). It should be noted that the object
modifier is governed by the object preposition (OP) á, which is homophonous
with the locative preposition á, 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 |sèÜ|, |kèô Ü| or |tú|. 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 |ká|
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 yí 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 comple-
mentary distribution. The interrogative terms in partial questions occupy the
same position as the non interrogative terms to which they correspond in asser-
tive sentences.
–– Relative clauses
As noun modifiers, relative clauses are generally (but see below) introduced by the
relative morpheme |-à Ü| which agrees in class with the relative’s antecedent. The rela-
tive morpheme may be followed by the conjunction |n¡bàô Ü| mb↑áô↓á “that”. The rela-
tive morpheme |-à Ü| 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 |-àÜ| is not used when the relative is not necessary for
Jacqueline Leroy
the identification of its antecedent: in this case, the conjunction mb↑áô↓á alone intro-
duces the relative clause.
These are introduced by the conjunction |ŋ¡gY ¡ Ü| ŋ:↑YÜ ↓, ŋ:Y̌, 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 singu-
lar, 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 dis-
course situation. (2) In the indirect style, only the 3rd person pronouns are used. How-
ever, if the subject of the completive corresponds to the subject of the governing verb,
it is the independent pronoun |zYÜ Ü| “(s)he” which is used, and which thereby takes
on logophoric meaning, instead of the subject pronoun of class 1 à.12 (3) In the semi-
direct style, the 3rd person pronouns are used for the speaker and any entity absent
from the discourse situation (in subject function one uses |zYÜ Ü| if there is coreference
with the subject of the governing verb and à for any entity absent from the discourse
situation); but it is the 2nd person pronouns which are used for the hearer.
–– Temporal clauses
–– Causal clauses
Two possibilities exist: (1) the causal clause is also a relative clause introduced by
the conjunction mb↑áô↓á, its antecedent |à-dzàŋ`| “way, manner” (cl. 7) rarely being
expressed. It precedes (most frequently) or follows the governing clause; (2) it is intro-
duced by the conjunctive phrase |`mÜbùô‒í (à-kéÜ) ŋ:Y ¡ Ü| “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 non-
future 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.
Tense-mood concordance and clause chaining in Mankon (a Grassfields Bantu language)
|`mÜbùô‒í | is the grammaticalized form of the verb |bùô| “?”13 in the non-future consec-
utive, | à-ké Ü| is the interrogative class 7 noun “what ?” (which is not always expressed)
and |ŋ:Y ¡ Ü| “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 | tàŋÜ| “in order to/that” (but cf. 5.).
There are two possibilities: (1) they may be introduced by the conjunction bé or mbé
“if ” or by the conjunctive phrase bé (á) bé ŋ:↑YÜ ↓ “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↑áô↓á bé “that if ”.
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 búŋ/múŋ “then”,
“in that case”, a morpheme which only expresses a logical link between two processes.
–– The conjunction á introduces all clauses whose main verb is in the exhortative,
whether they are independent or not (cf. 4.1.).
The Mankon verbal forms (or constructions) may be divided into four groups taking
into account the aspectual perfective/imperfective distinction and the polarity affir-
mative/negative distinction. Four conjugations are thus distinguished:
. In fact, the first term in this phrase is the prepositional element |m¡Übùôí (á)| “because of”.
|bùô| has retained, because of its previous status of non grammaticalized verb, the government
of its object complement. Thus the object preposition |á| 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 indica-
tive (CA) and the relative mood are absent from the negative perfective conjugation.
In the affirmative imperfective conjugation, in the indicative, there are only three con-
structions – 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). Effec-
tive 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 remem-
ber that a clause whose verb is in the exhortative is always introduced by the con-
junction |á| “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 mark-
ers) 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
Tense-mood concordance and clause chaining in Mankon (a Grassfields Bantu language)
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: mà “I”, ò “you”, tì‒ “us (dual)”,15 nì‒ “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: à (cl. 1), ‒ì (cl. 9)
–– Class pronouns: ní‒ (cl. 5), á (cl. 7), fí‒ (cl. 19), bí‒ (cl. 2), mí‒ (cl. 6), tsí‒ (8 and 10).
In the presence of a subject noun, the subject marker of the noun’s class is used. These
markers – all of V (a or ‒і ) shape only – have the same tonal characteristics as the cor-
responding pronouns; thus, on one hand one has the class 1 à and 9 ‒ì (L tone) markers
and on the other, the class 7 marker á and ‒í (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, mà repre-
sents the person pronouns, à or á represent the class 1 and 9 pronouns and bí‒ or bì‒ the
pronouns for all the other classes):
–– S1: mà, à, bí‒ (person and class 1 and 9 pronouns: L; other pronouns: H)
–– S2: mà, á, bí‒ (person pronouns: L; class markers: H)
–– S3: mà, à, bì‒ (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 exhor-
tative; the absence of pronoun is characteristic of the imperative and the consecutive.
The formative is what determines either the mood: successive, conditional, exhorta-
tive, 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).
imp – – v ‒í `/í‒
p0 –
p1 kì‒
s1 v ‒í `
p2 kì‒Ü
¡Ü
p0 r –
p1 r kì‒
p2 r s1 v ní‒`Ü
kì‒Ü
¡Ü
p0 ca ¡
p1 ca kì‒
p2 ca s1 v Ünà/Ünà`Ü
kì‒Ü
¡Ü
suc s2 v ‒ì
cd s2 Ü
exh s3
v ‒í
[–f] c n¡Ü
[+f] c – í‒`
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
Tense-mood concordance and clause chaining in Mankon (a Grassfields Bantu language)
far past; suc: successive; cd: conditional; exh: exhortative; [–f]c: non-future consecu-
tive; [+f]c: future consecutive; fut: indicative future.
Column 1 contains the subject paradigm, column 2 the formative paradigm, col-
umn 3 the (v) stem, and column 4 the final paradigm.
The order in the presentation of the formula for the various affirmative perfec-
tive conjugation constructions aims to highlight the formal similarities and differences
between these constructions.
Clause chains differ from the other complex sentences presented in the Sections 5 and
6 in that they obey specific tense-mood concordance rules (cf. paragraph 4.1.), which
is not the case for the sentences in Sections 5 and 6. Furthermore, on the semantic
level, the clauses in chains are not hierarchically ordered (cf. 4.2.).
exh + + +
(2) [+f]c + + –
s – – – + + +
[–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. “Other” refers to any verbal construction not mentioned in
this table (cf. Table 1).
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. Section 5).
(1) C1. mà m↑‒í ŋ¡f ŋ΄6 t↑ám΄6 ní‒bì át↑ú m↓ú >ò wûmbâŋní‒ //19
I fut put fruit kola loc.head child your male
C2. ‒ì >7¦// C3. t↑7΄>6΄ yí nâ >á>‒ì //
[+f]c.go [+f]c.place.op him with distance
C4. ô d↑íô6΄ dzàŋ zá // C5. ò m↑‒í t↓úm //
that.you ehx.show way rel you fut shoot
C6. ‒í s↓΄6 // C7. wàrí‒ //
˜
[+f]c.split [+f]c.cut
C8. tí‒>ǎ20 ò l¡7m↑΄6 m↓f΄ŋ g↓wâ
without you p0.hurt.op child the.dm
‘I’ll put a kola nut on the head of your son, I’ll place him at [a certain]
distance and you’ll show how you shoot and split [it] without hurting
the child.’
. Mankon has numerous merges. In the example glosses, the morphemes are not all sys-
tematically explicitated.
. The conjunction tí>a¦ or tíka¦ has only been encountered twice, and has not been ana-
lyzed. The first syllable is undoubtedly the negation marker tí. The second sentence in which
it is found is: mì-wàŋ mê w↑áŋn↓í mì lâ kè\ tíka¦ mà Šùrú l↑é “Why has the pap got finished so
quickly, while I am not satisfied?”
Tense-mood concordance and clause chaining in Mankon (a Grassfields Bantu language)
In Example 1 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.
In Example 2, the verb in C1 (the initial clause) is in the imperative (Imp; cf. the “other”
column in Table 2), in C2 the subject is the same as in C1, the verb in C2 is therefore
in the non future consecutive ([–F]C).
Remark 3: THe translation of Example 3 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 Example 3 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 sub-
ject 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
Example 3, with its sixteen clauses, is relatively long. It stops when reaching a clause
where the verb is in the indicative (P0 CA).
–– The succession relations can be explicitated by the auxiliary |tíô| “then”; this auxil-
iary 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
|tíô| often accompanies a change in aspect, especially a switch from perfective to
imperfective as illustrated by Example 4 (also see C12 in Example 3; and, without
a change in aspect, C3 in Example 35):
. The vowel G in this auxiliary, like that of the auxiliary tí/tínG (cf. remark 6) is not found in
the lexical morphemes. This indicates, in my view, a strong degree of grammaticalization.
Jacqueline Leroy
–– The auxiliary |lò|22 from the verb |lò| “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
Examples 3, 7, 16 and 36, and C2 in Example 6):
–– The auxiliary |kw7΄n| from the verb |kw7΄n| “come back home” (cf. kwê in C5 of
˜
Example 7) serves to go back in time:
–– The auxiliary |láô| “ever” (cf. C5 of Example 7, and C1 of the Examples 28 and 36)
indicates a far past or future.
. The auxiliary |lò| 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 nega-
tion |kèôÜ| the verb is not in the consecutive, but in a specific negative form.
Tense-mood concordance and clause chaining in Mankon (a Grassfields Bantu language)
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 restric-
tion which gives finality meaning to the non-initial clause. On the semantic level, a hier-
archy is established where the finality clause is subordinate to the initial (or preceding)
clause.
In Example 8, 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.
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 |kí‒|) 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:
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 story’s 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 con-
junction |tàŋÜ| “for…to, 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
á is maintained, whence the conjunctive phrase |tàŋÜ á| tàŋ6Ü “in order that, so that”.
Remark 5: THe conjunction |tàŋÜ|, which is always pronounced tǎŋ or tàŋ, most prob-
ably stems from a verb (perhaps |tàŋ| “draw” which still exists in the present state of the
language) conjugated in the future consecutive. All that remains of the future
consecutive formative |í‒`| is a trace in the form of a rising or low tone on the conjunction,
whatever the preceding tonal context.
(11) C1. á súŋ¡6 mb↑ô bô byí //
˜
1 s.say to children her
C2. ŋ:↑΄6↓΄ bì‒ fú // C3. ὴ:↑7΄ // C4. nts↓í //
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
΄ Ü| which in that
in the main clause, one uses the independent 3rd person singular pronoun |zY
case takes on logophoric meaning.
. |lánG| “really” is a manner auxiliary which may stem from the verb |lán| “be clear”.
Tense-mood concordance and clause chaining in Mankon (a Grassfields Bantu language)
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 con-
secutive ([+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 |tàŋÜ á|. As the subject in C4 is differ-
ent from that of C3, the verb in C4 is in the exhortative.
Jacqueline Leroy
C1 is the hypothetical clause. Its verb is in the conditional (CD). C2 is a relative embed-
ded 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 hypotheti-
cal 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):
. Use of the conjunction bé/mbé is optional. When the hypothetical clause is introduced by the
phrase bé á bé ŋ:↑YÜ↓Ü (literally: “if 1 CD.+be that”; “if it be that”), one may use the indicative.
Tense-mood concordance and clause chaining in Mankon (a Grassfields Bantu language)
. For the tense-mood concordance in verb chains, it is the last verb that counts: here there-
fore it is dàrí “creep” and not the auxiliary tí. This is in keeping with the analysis proposed in
2.2.2. which suggests that verb chains stem from clause chains.
Jacqueline Leroy
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 |tsà| “sur-
pass” has become specialized for expressing comparison, |>á| “give” for expressing ben-
eficiaries 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, pres-
ent or past. The verb |tsí| “exist, be” has become specialized in the expression of ability,
possibility, and the verb |bé| “be” in the hypothetical (imaginary) domain. The expres-
sion 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 |b¡fŋ| “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:
. This formulation simplifies the facts which will be presented in depth in 7.6.
Tense-mood concordance and clause chaining in Mankon (a Grassfields Bantu language)
. It is a syntactic criterion which justifies the distinction made between the semantic
roles of “beneficiary” and “recipient”: in fact, the verb |>á| “give” is never used to express the
semantic role of “recipient”, while the reverse is not true.
Jacqueline Leroy
–– 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 sec-
ond clause’s verb is in the successive or exhortative, in accordance with the tense-
mood concordance rules. In Example 24, the verb in C1 is in the relative perfective
“present”, the verb in C2 is therefore in the successive:
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 |tsí| “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:
Y΄6 //
>¡ C4. lòm¡6 tá // […] (à tsìtì‒n6̂ //
[+f]c.make.that dry.season exh.shine p0.be.ca
–– The first, main, clause (C1), contains only the verb |tsí| “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
|tsí|, its verb is in the consecutive, but always future ([+F]C):
Remark 7: This last example illustrates (1) that the verb |tsí| “exist, be” cannot be
negated when it serves to express ability. It is lyè “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 |láô| “ever”) to situate the process in the future.
–– The first only contains the verb |bé| “be” and its subject. The verb |bé| is only
conjugated in the indicative, effective aspect. |bé| may be preceded by a temporal
auxiliary, |>Y
¡ |, |lò| or |tsí|; in this case it is the auxiliary which is conjugated in the
indicative, effective aspect, and |bé| 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 rela-
tion 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 |bé| “be”. Its verb is therefore in the consecutive, but
always future ([+F]C):
(29) C1. mà bè↑΄// C2. b↓é >ô //
I p0.be [+f]c.be you
(múŋ mà t∫7¦ >↑΄6 ↓΄ kàmì‒rún)
˜
then I p0.stay.op me.loc Cameroon
‘If I were you, (I would stay in Cameroon.)’
(30) C1. mà lo¦ mb↑é // C2. Šw↓í:t΄6 m↓ú
I p0.aux [–f]c.be [+f]c.kill.op child
>àn // (múŋ mì‒ŋkàôà m↑á m↓í mí‒mbá↓΄
my then arrows the nm two
mí‒ lw↑7΄ô↓7΄t΄6 >ô)
6 p0.carry off you.dm
‘If I had killed my child, (the two arrows, they would have carried you off.)’
(31) C1. mà bè↑΄ // C2. ts↓í ΄ ŋ¡6 //
lá mbye¦ nì‒mY
˜
I p0.be [+f]c.be foc near mother
(múŋ k↑á mà ndzê >6̂ zíŋ¡6 )
˜ ˜
then neg I see.ipf.op me this.dm
‘If I were near my mother, (I would not go through this).’
In Examples 32–35, the clause C2 is chained to C1. Its verb |b¡fŋ| 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 33–35 – following the tense-mood concordance rules.
But in Example 36, the first clause C1 is hypothetical. C2 is therefore the main
clause governing C1 and its verb |b¡fŋ| is of course in the successive, even though its
subject is identical to that of C1 (cf. 6):
8. Conclusion
In this article I have shown how certain verb forms in Mankon, in this case the non-
future 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 Ø- (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
Ø- (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 exhorta-
tive, having the same tonal characteristics as in Mankon; and the presence, at least in
Bafut, of the conjunction á or tá, 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 Hyman’s examples, the structural tones of the [–F]C and
[+F]C are the same as in Mankon. Lastly, in Feôfeô, 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 mean-
ing since it is used after verbs in the past (“subjunctive consecutivization”) with the
formative Ø-. 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 consecu-
tivization”) one finds the conjunction á.
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 preced-
ing 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: 88–91 and 232–233),
. 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.
Tense-mood concordance and clause chaining in Mankon (a Grassfields Bantu language)
in Aghem, also of the Ring group (S. Anderson, 1979: 112–117), 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).
References
Anderson, Stephen. 1979. Verb structure. In Aghem Grammatical Structure [Southern Califor-
nia Occasional Papers in Linguistics (SCOPIL) 7, Larry M. Hyman (ed.), 73–136.
Hyman, Larry. M. 1971. Consecutivization in Feôfeô. Journal of African Languages 10: 29–43.
Hyman, Larry. M. 1980. Relative time reference in the Bamileke tense system. Studies in African
Linguistics 11(2) : 227–237.
Jacqueline Leroy
Leroy, Jacqueline. 1977. Morphologie et classes nominales en mankon (Cameroun) [BS 61–62].
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), 55–71.
Leroy, Jacqueline. 1982. Extensions en mankon, langue bantoue des Grassfields. In Le verbe
bantou [Oralité-Documents 4], Gladys Guarisma, Gabriel Nissim & Jan Voorhoeve (eds),
125–138. Paris: SELAF.
Leroy, Jacqueline. 1983. Système locatif mankon et classes locatives proto-bantoues. The Journal
of West African Languages 13(2): 91–114.
Leroy, Jacqueline. 1994. La nasalité en mankon, langue bantu des Grassfields (Cameroun).
Linguistique africaine 13: 61–81.
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 Musée Royal de l’Afrique Centrale, Sciences Humaines 61], 81–121. Tervuren:
Musée Royal de l’Afrique 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), 189–210.
Dallas TX: SIL.
Schaub, Willi. 1985. Babungo [Croom Helm Descriptive Grammars]. London: Croom Helm.
Clause dependency relations
in East Greenlandic Inuit*
Nicole Tersis
Centre d’Etudes des Langues Indigènes d’Amérique,
Fédération Typologie et Universaux Linguistiques, CNRS
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
*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 80 000 speakers. Tunumiisut is spoken in the easternmost part of the dialect
continuum by approximately 3577 people, out of an estimated total Greenland population of
56 969, 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 conse-
quence, 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 coor-
dinator, 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 con-
clude, I will show how, on the typological level, the polysynthetic nature of this lan-
guage also applies to sentence structure, through the preferential use of synthetic
processes to mark clause subordination.
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 Ø 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 definite-
ness 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 message’s explicitation by use of an ‘anti-topic’ (Tersis & Carter-Thomas 2005).
However one must be aware that interpretation of the sentence’s structural analysis,
transitivity, word order and ergativity’s 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 phenom-
enon can be explained from a historical perspective (cf. Section 4).
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 aŋikkaayuk.
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 verb-
finally, 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
(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 coordi-
nated 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
Clause dependency relations in East Greenlandic Inuit
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-ŋŋita-a aassaat taki-qnaq-pa-a.
recognize-ind.neg-3sg.3sg right.away see-1st time-ind-3sg.3sg
‘She doesn’t recognize him right away, it is the first time she has seen him.’
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. aatisaŋaq ataasiq aamma timittat maqtit
fish one and bird.pl two
‘one fish and two birds’
The coordinated clauses take the same verbal mood. In example (17a), the clitic coor-
dinator -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 iŋi-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 subor-
dinate 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.
The consonant variations w/+p and ŋ/+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 preced-
ing 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
Section 3.5). This means that it cannot appear on its own in an independent sen-
tence, contrary to what is found in other Inuit dialects – but not in West Greenlan-
dic (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 nayuŋa-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-ŋatiwaq-pu-a kuummiit
at.that.time thought-have-habitual-however-ind-1sg Kuummiit
tasiita-miŋŋaaniit aŋi-tii-ti-ŋumaaq-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.’
. I have borrowed the label ‘situational dependency’ from Stéphane 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.”.
Clause dependency relations in East Greenlandic Inuit
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 per-
son marker refers either to the agent (cf. 24) -ni (3R.sg)/-tit (3R.pl), as a reflexive
person, or to the patient (cf. 25) ‑ŋu (3sg)/-ŋut (3pl). The most common order is for
the subordinate clause to come first. This order may be reversed for semantic rea-
sons (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 comple-
ment 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-ŋu 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-ŋŋit-sit
cod.pl be.dry-completely-pfv-neg-attrib.pl
[uu-ttu-ŋut] [taawaa aammaqqaa-ttu-ŋut]
cook-conc-3pl then eat.with.the.fat-conc-3pl
[iminna tiinniit misi-kaa-qtu-ŋut] 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-ŋu.
say-ind-3sg come-fut-conc-3sg
‘He said he was going to 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 indepen-
dent sentences; and the attributive, the concomitant, the causative and the conditional
which appear in subordinate clauses (see Table 1).
Furthermore, oral narratives clearly show that dependent sentences are highly fre-
quent. 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 subordi-
nate 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 begin-
ning 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-ŋii-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-ŋŋiq-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.
. M. Mithun (2008) compares two narratives, one in Yup’ik (same family as the Inuit con-
tinuum), the other in Navajo (Athabascan family): the Yup’ik 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 -ŋaaq-, the verbal derivation affix “a lot” and the conjunctive
morpheme “at the time when”. It might well be that the distinct affixes ŋaaq- et ‑ŋiaq- ‘each time that’
tend to formally merge in Tunumiisut (M. Fortescue p.c.). There are numerous homophones in the
affixes given the language’s small phoneme inventory (three vowels i, a, u, and eleven consonants p,
t, s, k q, m, n, ŋ, n, w, y) and the numerous cases of vowel and consonant assimilations.
Clause dependency relations in East Greenlandic Inuit
The conjunctive morphemes -niaq- and ŋiaq-/-(V)yaq- also appear in other con-
texts however, albeit in independent sentences with a different identity. The ques-
tion 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 sen-
tences, 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, caus-
atives or interrogatives:
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/cona-
tion (Tersis 2008):
+niaq-1 incorporating N–V(-) verb ‘hunt’
(38) miikkattaq-niaq-pu-q.
ringed.seal-hunt-ind-3sg
‘He hunts ringed seal.’
+niaq-2 verbal modality V–V ‘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
The morpheme -ŋiaq- /-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):
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. Table 2).
–– Possessive modifiers
–– Subordinate clause
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 dog’s 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 nii-ŋa-ma
|anaana-p-ŋa| |nii-ŋa-p-ŋa|
mother-of-1sg eat-caus-of-1sg
‘of my mother’s’ ‘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”
Singular Plural
1. nii-ŋa-ma nii-ŋa-tta
|nii-ŋa-p-ŋa| | nii-ŋa-p-ta|
eat-caus-of-1sg eat-caus-of-1pl
‘when I was eating’ ‘when we were eating’
2. nii-ŋa-wit nii-ŋa-ssi
|nii-ŋa-p-tit| |nii-ŋa-p-si|
eat-caus-of-2sg eat-caus-of-2pl
‘when you were eating’ ‘when you (pl) were eating’
3.. Coreference with the agent in the main clause
nii-ŋa-mi nii-ŋa-mit
|nii-ŋa-p-ni| |nii-ŋa-p-ni-t|
eat-caus-of-3sg eat-caus-of-3pl
‘when he was eating’ ‘when they were eating’
Non-coreference with the agent in the main clause
nii-mm-at nii-mm-ata
|nii-ŋ(a)-p-at| |nii-ŋ(a)-p-ata|
eat-caus-of-3sg eat-caus-of-3pl
‘when he was eating’ ‘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 qimmi-ma suuni-a ‘my dog’s head’
dog-of.1sg head-3sg
2sg qimmi-wit suuni-a ‘your dog’s head’
dog-of.2sg head-3sg
1pl qimmi-tta suuni-a ‘our dog’s head’
dog-of.1pl head-3sg
2pl qimmi-si suuni-a ‘your dog’s head’
dog-of.2pl head-3sg
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-/‑ŋi‑ durative,
‑ttu- concomitance, -ŋa- effected, -ŋi- non-effected).
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 /hunter-
rel/fox.abs/see-3sg.3sg/ “the hunter saw the fox” could be construed as “the hunter’s
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 sym-
metrical, 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 Nichols’s (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 system’s
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. ‘I see them.’
|taki-wa-t-ŋa|
see-ind-pl.1sg
qimmi-kka ‘my dogs’
|qimmi-t-ŋa|
dog-pl-1sg
Nicole Tersis
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 hypothe-
sis, 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 clause’s 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 depen-
dence 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.
Clause dependency relations in East Greenlandic Inuit
Furthermore, the various devices used to mark subordinate clauses in East Green-
landic Inuit -namely specific verb forms, affixes included in the verb phrase and core-
ferential 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 lan-
guage’s polysynthetic structure. The structural homology between possessive noun
phrases and adverbial clauses with the causative and conditional verb forms contrib-
utes 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
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 Société
Finno-ougrienne 82: 7–80.
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 société dans l’Arctique nord-
américain. 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): 61–75.
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 Tradi-
tion, Tales from Eastern Greenland. Ms.
Jacobson, Steven. 1995. A Practical Grammar of the Central Alaskan Yup’ik 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.), 45–64. Amsterdam: John Benjamins.
Kaplan, Lawrence. 2000. L’inupiaq et les contacts linguistiques en Alaska. In Les langues
eskaléoutes, Sibérie, Alaska, Canada, Groenland, Nicole Tersis & Michèle Therrien (eds),
91–108. Paris: CNRS Editions.
Lowe, Ronald. 1991. Les trois dialectes inuit de l’Arctique canadien de l’Ouest: Analyse descriptive
et comparative [Groupe d’Etudes Inuit et Circumpolaires]. Québec: Université 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), 115–134 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 Yup’ik 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 Yup’ik [Santa
Barbara Papers in Linguistics 7]. Santa Barbara CA: University of California.
Mithun, Marianne. 2008. The extension of dependency beyond the sentence. Language 84: 69–119.
Miyaoka, Osahito. 1996. Sketch of Central Alaskan Yupik, an Eskimoan language. In Handbook
of North American Indians, Vol. 17. Languages, Ives Goddard (ed.), 325–363. Washington
DC: Smithsonian Institution.
Nichols, Johanna.1986. Head-marking and dependent-marking grammar, Language 62: 56–119.
Robert, Stéphane. 1996. Aspect zéro et dépendance situationnelle: l’Exemple du wolof. In Dépen-
dance et intégration syntaxique, Claude Muller (ed.), 153–161. Tübingen: 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 complexité syntagmatique, Groenland oriental. In
Les langues eskaléoutes. Sibérie, Alaska, Groenland, Nicole Tersis & Michèle Therrien (eds),
249–268. Paris: CNRS Editions.
Tersis, Nicole. 2004. De l’ergatif à l’accusatif, évolution des structures argumentales en inuit
du Groenland. In Actes du Colloque Ergativité, Francesc Queixalos (ed.), 45–51. Villejuif:
Centre d’Etudes des Langues Indigènes d’Amérique (CNRS, IRD).
Clause dependency relations in East Greenlandic Inuit
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, 473–500.
Vaxtin, Nicolaj B. 2000. Les différents types de phrases polyprédicatives du yupik sibérien cen-
tral. In Les langues eskaléoutes, Sibérie, Alaska, Canada, Groenland, Nicole Tersis & Michèle
Therrien (eds), 303–333. 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),
61–88. Cambridge: CUP.
Coordination and subordination
Áma in Bulgarian dialectal Greek*
Eleni Valma
LACITO (Laboratoire des langues et Civilisations à Tradition Orale),
Fédération Typologie et Universaux Linguistiques, CNRS
This article studies the relator áma 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
áma 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, Nessèbre and Pomorié, 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 Guentchéva 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. Guentchéva
(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 tran-
scribed 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 popu-
lation who have been sedentary for the past 50–70 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 char-
acterized 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 áma,
which, parallel to its temporal use, has developed hypothetical uses. This subordi-
nating conjunction can also mark opposition and, in a marginal manner, justifica-
tion. Given that opposition is reserved for coordination, one may say that the uses
of áma are susceptible of covering two different syntactic functions, subordination
and coordination.
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 con-
nection 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 áma in Bulgarian dialectal Greek qualifies as a junctor since in the case of syn-
tactic 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 hypotheti-
cal uses, and the links between these two notions and that of opposition reserved only
for the dialectal use of áma. 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.
Coordination and subordination
f requently found cross linguistically (cf. Traugott, 1985; Kortmann, 1996: 210), the
expression of opposition via the same junctor is far from being obvious.
We propose to apply the syntactic tests (cf. among others Bril & Rebuschi, 2007;
Maingueneau, 1994: 26–27, 98–99, 138–139) which make it possible to determine the
status of the junctor áma in dialectal Greek. Let us take the following example:
(1) vlép polís kózm0 mazeménos / áma kodóft-ase
see.prs.3sg. many people assembled when arrive.aor.3sg.
don íða-ne/ ma lé-i círje janáko
him.pro. see. aor.3pl but say.prs.3sg. mister janako
pu í-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 subordinat-
ing status of áma:
–– permutation between the two clauses is possible: don íðane áma kodóftase (En.
They saw him when he approached);
–– one may focalize or extract the clause introduced by áma, for example by using
the adverb mòno (En. only): don íðane móno áma kodóftase (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 íðane áma ecínos kodóftase (En. They saw him
when that one approached).
. Standard Greek has marginalized this use in favor of the hypothetical. Although the
polysemy of άμα in Standard Greek is not unanimously recognized among Greek gram-
marians 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 άμα
in familiar speech [Mackridge, 1985/1990: 418, §9.4.4.3]. Standard Greek άμα 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 chron-
ological sequence of events. The occurrence of áma 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], Desclés [1980, 1990a, 1990b, 1994]). We will
use the notions here such as they were defined in Desclés [1994: 71–76]):
In (2), áma is well suited to the same analysis: followed by the imperfect padrévodan
(Eng. they were getting married), áma is the source for the clause’s temporal interpre-
tation. In this context, the verb form in the imperfect refers to an open class of events
and the subordinate clause serves to mark iteration.
One may apply the same syntactic tests to (3), but this time the interpretation is
hypothetical:
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
Coordination and subordination
Modern Greek grammarians classify as a subjunctive, cf., among others, Tsopanakis, 1994)
are constructed using a verb root followed by the present declensions.
When áma is followed by an aspectual form, it is open to hypothetical interpreta-
tion. This is the case in (3), where one has the imperfective íne. While in its temporal
use, áma 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 car-
ried out; through encyclopedic knowledge, one knows that P will happen in the
future (Example 4)
–– a hypothetical structure, often with iterative aspect, where P is open to two pos-
sibilities, that of P and that of non P (Example 5)
In (4), the use of áma indicates that the speaker clearly leans towards P’s
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 accom-
plishment 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
áma 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 speaker’s appre-
hension of the subordinate’s content and combines the notion of time with hypoth-
eses. Thus áma poses a fact within which the facts explicitly mentioned in the main
Eleni Valma
On the contrary, in (7), the two an3 (En. if) introduce an eventuality:4 a fact is pre-
sented 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):
áma also expresses justification, even though the scope of this interpretation is very
narrow and is essentially contextual (discourse justification). In (8)
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 vio-
lated because permutation between the two clauses, or focalization or extraction of the
clause introduced by áma are impossible. This justification use could serve as a bridge
from the subordinating towards the coordinating use.
When áma 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 áma occasionally expresses.
However, it would be risky to assert that the justification meaning which áma
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 hypoth-
esis according to which áma would have appeared in dialectal Greek following contact
. The ΛΚΝ dictionary asserts that in Standard Greek, άμα 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: 209–210, 213–217].
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 & Sil’nickij, 1969, quoted in
Desclés & Guentchéva, 1997].
. The idea of such a switch may be defended in Standard Greek when one uses the opposi-
tion coordinator αλλά (En. but), in contexts where the clause introduced by αλλά is a condi-
tion 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, amá is found in Bulgarian [Feuillet, 1996: 318] and
expresses the notion of opposition.7 In its interphrasal function of opposition coordina-
tor, amá 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 áma is a junctor which passed from Turkish into dialectal Greek via Bulgarian.
One observes that the structure “P áma 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 áma necessitates
reference to a sociopolitical situation which the speaker and hearer share:
(9) vul>aría / í-dan vul>aría
Bulgaria be.imperf.3sg. Bulgaria
áma íç-e eliniká skolía. [MN]
but have.imperf.3sg. Greek schools
‘It was Bulgaria, but there were Greek schools.’
(10) sti >alía áma élinas éç-is ? [MN]
in.the France but Greek have.prs.2sg.
‘You live in France and your husband is Greek?’
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 Bul-
garian) 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.
For the Karakatchans, áma lexically opposes two terms, for example “young” and
“old” in (11):
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 áma
having different semantics.
Contrary to (9) and (10) where two facts are confronted, in (12) and (13) the clause
introduced by áma furnishes additional information and helps further the debate:
Therefore there are two áma, 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:
One must note that despite preponderance in the use of áma, other processes are also
used to express opposition. These means are different in the two groups, the Kara-
katchans 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:
. 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]
Coordination and subordination
‘It is him [my grandson] but he is shy [that is why he doesn’t want to say
hello to you]. Come my child. It is him, but you make him shy, you under-
stand.’
(17) tsarúçja léγo-dan ecίna forús-an
tsarouks call.imperf.3pl. those wear.imperf.3pl.
ta tsarúçja a δe ta ίδ-a eγό
the tsarouks but not the.pro. see.aor.1sg. i
s-ti δiciá mu zoί δe ta ί-çe. [Kar]
to.the pro.poss. my life not the.pro. have.imperf.3sg.
lit. They [the shoes] were called tsarouks, they used to wear those,
the tsarouks, but I didn’t see them during my life.
‘The used to wear tsarouks, but before my time.’
On the other hand, the inhabitants of the North Sea coast alternate between áma and
aSá (En. but): the first is reserved for argumentative opposition (founded on an argu-
ment) 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). aSá, very common in Standard Greek, comes from Ancient
Greek [Andriotis, 1951/1995].
But aSá is also used when the speaker wishes to rectify information by choosing cer-
tain elements of a notion and certain specific aspects of these elements, as in (21).
Eleni Valma
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 litera-
ture [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
hearer’s discourse.
(22) énon1: ∫e kána şjó méres ér-çete.
to almost two days come.prs.3sg.
énon2: vévea / tóso kodá í-saste.
Of.course so close be.prs.2pl.
énon3: ma θé-li e>ó na majirév-o /
but want.prs.3sg. me that cook.impfv.1sg.
e>ó na ti skupís-o. [MN]
me that it 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 dis-
cursive connector afú (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-úsan pu ta ksér-is ta eliniká
me ask.imperf.3pl. where them.pro. know.prs.2sg. the Greeks
ma afú é-çis spítja jatí írθ-es. [MN]
but since have.prs.2sg. houses why come.aor.2sg.
‘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 move-
ment. Furthermore they can belong to the same notion, as in (24):
. Another yet unproved hypothesis is that ma is the truncated form of amá.
Coordination and subordination
The junctor áma 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 áma. 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 áma into a hypothetical junctor calls for some specification: it is not
hypothetical on the same level as an (En. ‘if ’) because by using áma, 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), áma also has a coordinating function expressing opposition. To ana-
lyze áma, we therefore have the choice between two classic solutions:
–– either one separates the uses of áma into two homonyms disjoined in their mean-
ing 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 prop-
erty 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 áma Q”, focus is solely on whether the speaker takes responsi-
bility for the arguments or not [also see Plantin, 1990: 43]. It is interesting to compare
this use of áma with the other processes for expressing opposition; this imposes vari-
ous conclusions.
Thus if the temporal and hypothetic use of áma is common both to the Kara-
katchans and to the inhabitants of the North Sea coast, the oppositional use which
áma 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 áma. For the Karakatchans, áma coexists
with a (En. ‘but’), another borrowing from Bulgarian. By contrast, the inhabitants of
the North Sea coast have opted for aSá, a borrowing from Greek. As we saw above, in
the structures “P a Q” and “P aSá Q”, the clause Q belongs to a different utterance than
P. Moreover, a and aSá 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 áma and aSá 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, áma still contains all sorts of oppositions, notional or dis-
cursive (argumentation and/or demonstration).
References
. One must be cautious when dividing speakers into groups as they tend to mix and to
borrow from each other.
Coordination and subordination
Desclés, Jean-Pierre & Guentcheva, Zlatka. 2000. La notion d’abduction et le verbe devoir épis-
témique. In Cahiers Chronos 8, Patrick Dendale & Johan van der Auwera (eds.) 103–122.
Amsterdam: Rodopi.
Dik Simon. 1997. The Theory of Functional Grammar. New York NY: Mouton de Gruyter.
Ducrot, Oswald. 1980. Les échelles argumentatives. Paris: Les éditions de Minuit.
Ducrot, Oswald. 1984. Le Dire et le Dit. Paris: Editions de Minuit.
Feuillet, Jack. 1996, Grammaire synchronique du bulgare. Paris: Institut d’études slaves.
Foley William & Van Valin Jr., Robert D. 1984. Functional Syntax and Universal Grammar. Cam-
bridge: 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. Λεξικό της Κοινής Νεοελληνικής. Θεσσαλονίκη: ΙΝΣ – ΑΠΘ.
Lyons John. 1977/1996. Semantics (v.2). Cambridge: CUP.
Mackridge, Peter. 1985/1990. The Modern Greek Language. Oxford: OUP.
Maingueneau, Dominique. 1994. Syntaxe du français. Paris: Hachette.
Moeschler, Jacques & de Spengler, Nina. 1982. La concession et la réfutation interdite –
approches argumentative et conversationnelle. Cahiers de linguistique française 4: 7–37.
Nedjalkov Vladimir & Sil’nickij, Georges. 1969. In Tipologija kauzativnyx konstrukcij,
A. A. Xolodovič (ed.), 5–19. Leningrad: Nauka.
Ong, Walter. 1982. Orality and Literacy – The Technologizing of the Word. London: Methuen & Co.
Plantin, Chistian. 1990. Essais sur l’argumentation. Paris: Kimé.
Poromanska, Stoyna. 2004. Οι νεοελληνικές διάλεκτοι στη Βουλγαρία. Tipologija kauzativnyx
konstrukcij. In Studies in Greek Linguistics – Proceedings of the 13th Annual Meeting of the
Department of Linguistics, 560–567. 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), 91–401. Dubuque IA:
Kendall-Hunt.
Traugott, Elizabeth Closs. 1985. Conditional markers. In Iconicity in Syntax [Typological Studies
in Language 6], John Haiman (ed.), 289–307. Amsterdam: John Benjamins.
Tsitsilis Christos. 1999. Το γλωσσικό ιδίωμα των Σαρακατσάνων της Βουλγαρίαs. In Dialect
Enclaves of the Greek Language, 73–75. Athens: Ministry of National Education and
religious Affairs.
Tsopanakis, Agapitos. 1994. Νεοελληνική γραμματική. Θεσσαλονίκη: Κυριακίδης.
Tzartzanos, Achilleas. 1928/1996. Νεοελληνική σύνταξις (v.2). Θεσσαλονίκη: Κυριακίδης.
Λεξικό της κοινής νεοελληνικής (dictionnaire du Greek moderne) [ΛΚΝ]. 1998. Θεσσαλονίκη:
ΑΠΘ – Ινστιτούτο Νεοελληνικών Σπουδών.
Valma, Eleni. 2004. L’ expression de la causalité en français et en Greek moderne – étude con-
trastive. Thèse non publiée, Université Paris VII – Denis Diderot, Paris.
Vrinat-Paskov, Marie. 1990. Les particules expressives du bulgare moderne (v.2). Thèse non publiée,
INALCO, Paris.
Author index
Nedjalkov, V. 8, 106–107, Robert, S. 4, 6, 10–12, 14–16, Tersis, N. 4, 6, 11, 18, 581–584,
166, 169–170, 198–200, 51, 201, 269, 333, 339–342, 586, 592–593, 597
231, 235, 609 376, 383–384, 394–395, Tesnière, L. 422
Newman, J. 456 399–400, 411, 451, Theoharidis, P. 402, 413
Nichols, J. 56, 209, 597 465–466, 469–475, Thompson, S. 2, 178, 189–190,
Nikolaeva, I. 144, 233 477–478, 482, 489–491, 199, 248, 256, 259, 305,
Noonan, M. 2, 51, 61, 496, 588 331, 509
99, 207, 232, 254, Roberts, J. 43, 52, 57, 60, 63, 65, Tikkanen, B. 52–53, 57, 59, 75,
501–502, 518 68, 71, 76, 98, 240, 242 99, 169, 200
Nordlinger, R. 37, 455 Rogo, A. 402 Topolinjska, Z. 404
Ross, M. 3, 69, 99, 275–279, Toulmin, S. 606
O 285–294, 299 Traugott, E. C. 393, 415, 605
Olson, M. 1, 28, 52, 511 Rousseau 426, 446 Tsitsilis, C. 604
Ong, W. 604 Rousseeuw 82 Tsopanakis, A. 607
Opgenort 67, 101 Tzartzanos, A. 605
Ozanne‑Rivierre, F. 281 S
Sadock, J. 3, 5, 240, 581 V
Sall, A. O. 469–470, 480–481, Vaillant, A. 404
P Valma, E. 11, 19, 603,
483–484
Parker, E. 579 609–610, 614
Šaur 403
Payne, J. R. 204, 215, van der Auwera, J. 2, 77, 107,
Schackow 59, 75
219, 238, 248–249 517, 523
Schaub, W. 578
Peled, Y. 379, 390 Van Valin, R. D. 2, 4, 6, 27–29,
Schneuker, C. 46
Philippaki-Warburton, I. 617 40–41, 51–52, 54, 60,
Scott, G. 32, 47, 63, 99
Pilhofer 100 69–70, 78, 91, 206, 221,
Shay, E. 3, 331–332
Plantin, C. 615 239–240, 256, 269–274,
Shrestha 53
Polinsky, M. 115, 151, 242 282, 430, 462, 604
Sil’nickij, G. 609
Polotsky, H. J. 209–211, 226, Vanhove, M. 5, 11, 13–14,
Simeone-Senelle, M.-Cl. 395
228, 237 333–335, 347, 390
Sobolev, A. N. 410
Poromanska, S. 604 Vaxtin, N. B. 586
Sonnenberg 425
Potsdam, E. 151 Velčeva-Bojadžieva 403
Stassen, L. 67, 89, 170, 213, 215,
Prasse, K.-G. 355, 360 Verstraete, J.-C. 6, 10–13, 15,
229, 352, 510
Strimmer 83, 87 17, 285, 338, 342, 451, 453,
Q Subrahmanyam 169, 457, 459, 467, 477, 490
Quirk, R. 313 173–174, 180 Vrinat-Paskov, M. 610, 612
Sumbatova, N. 144, 162 Vulchanov, V. 403
R Suter, E. 47, 53, 100 Vycichl, W. 360, 388
Ramat, P. 387 Svane, G. 403
W
Rappaport 101 Svartvik, J. 332
Woodbury, A. 591, 597
Rebuschi, G. 2, 5, 204, 219,
604–605 T Y
Redder 443 Taïfi, M. 359 Ylikoski, J. 107
Reed 101 Taine-Cheikh, C. 5, 11,
Reesink, G. 37, 53, 58, 69, 13–14, 355–357, 363, 365, Z
76, 101 369, 381 Zeileis 88
Reilly, J. 396, 420 Takahashi 54 Zifonun 425, 434
Rigsby, B. 451–452 Ter Meulen, A. 232 Zúñiga, F. 107
Language index
Kiranti 12, 53, 59, 67, 85, Ngandi 542 Tok Pisin 514
274, 306 Nyelâyu 281 Tombunuo 272, 280
Korafe 12, 44–45, 53, 66, 100, Tonkawa 28
303–305 O Touareg 365, 377, 385
Kryz 143, 145, 148–149 Oceanic languages 14, 17, 342, Toura 101
350, 490 Trans New Guinea family 36
L Old Church Slavonic 403, 414 Tsakhur 115, 143, 145
Latin 65, 72, 233, 305–306, 387, Tsez 115
518, 525 P Tundra Nenets 586
Lenakel 100 Polish 320 Tunumiisut
Lezgian 107, 143, 145, 148 Pomak 6, 14, 399–402, (East Greenlandic) 18–19,
Lezgic 143, 145, 154, 159 405–406, 409–410, 415, 581–582, 584, 586, 591–592,
Lo-Toga 17, 499–546 417–418 597, 599
Proto-Inuit 594 Turkish 19, 58, 63, 85, 101, 392,
M 401–402, 405, 409, 415,
Macedonian 400–404, 407, R 418, 610, 615
409, 415 Ring 578–579
Maghribi Arabic 337, Romanian 403 U
340–342, 344 Roviana 295, 297 Udi 143, 145
Manam 272, 303 Russian 51, 101, 105–107, Umpithamu 15, 17, 338, 342,
Mandankwe 578 320, 322 451–467, 490
Mankon 18, 549, 552–555, Rutul 143, 145 Uralic family 586
557, 562, 564, 575, Usan 7, 37–39, 41, 53, 58, 69,
578–579 S 76, 101
Mbam-Nkam 549, 578 Samoyed 586
Mbili 578 Semitic 208, 334, 356 W
Meglenoromanian 403 Serbian 401, 403 Wambule 62, 66–67, 85, 101
Mianmin 43 Shilha 361, 367, 378, 388 Wandala 12, 313–314, 317
Modern German 14, 421–422, Siberian Yupik 586 Warlpiri 455
428–429, 446 Sobei 275, 307 Watam 42–43
Modern South Arabic 395 Spanish 154 Wolof 10, 14, 16, 18, 201,
Momo 579 Standard Greek 603–606, 340–342, 344, 376, 383,
Moroccan Arabic 335, 337 608–609, 613–614 394, 411, 469–496, 588
Mundani 579 Suena 44–45
Mwotlap 297, 376, 504, 507, Swahili 52, 54–55, 57, 85, 101 Y
510, 527, 532 Yafi‘ Arabic 13, 333, 341
T Yemeni Arabic 335–336
N Tabassaran 143, 145 Yimas 31, 34, 36, 48
Navajo 592 Takia 272, 276, 282, 286, Yup’ik 101, 592
Nêlêmwa 272, 276, 281, 284, 298, 307
307, 350 Tamazight 378, 384, 386–387 Z
Nemi 281 Tauya 7, 30, 36–37, 41–42, 47, Zenaga 6, 13, 18, 355–361,
Nepali 53–55, 72, 75, 84, 100 53–55, 57, 70–73, 85, 101 363–365, 368–370,
Newar 101 Tawala 272, 297–299 375–382, 384–385,
Newari 30 Timugon Murut 279 387–395
Topic index
69, 75, 77, 85, 106–107, 355, 359–366, 364–365, distance 82–84, 151, 218–219,
111–119, 121–139, 146, 159, 368, 381, 383, 385–389, 225, 360–361, 364,
163, 165–166, 168–173, 393–395, 399–400, 404–405, 407, 435
175–181, 185, 187–188, 191, 402–411, 413–414, 417,
194–201, 222–223, 225, 422, 424–427, 451, 472 E
231–233, 235 demarcative function 424, 438 embedding 8, 17, 27, 32, 36, 78,
coordination 1–6, 8, 10–11, demonstrative 13, 139, 260, 115, 147, 181–182, 187, 190,
27–30, 32, 40–41, 44, 48, 283, 293, 299–302, 199, 223, 252–253, 259, 307,
51–52, 57, 66, 69, 84–86, 318–320, 328, 331, 346, 391, 333, 429–430, 490–491,
93, 107, 109, 114–115, 171, 406–407, 409, 543, 613 521, 539–540
201, 203–207, 212–221, dependency (clause) 1, 4, 6, 9, encoded inference 11, 13, 16,
226–230, 232, 234–235, 17–19, 28, 40, 57, 82, 144, 342, 451–452, 466
238–243, 248–250, 165–166, 171, 198–199, 201, endocentric 31–32, 34
252–253, 257, 259, 277, 221, 246, 269, 272, 308, entropy 83, 87
281, 285–287, 295, 298, 318, 333, 342, 352, 359, 368, exhortative 18, 549, 557–563,
302–305, 307, 342, 370–371, 373, 379, 383, 392, 422, 430, 567–569, 572, 574, 576,
510–512, 582, 585–586, 469, 480, 485, 491, 493, 578–579
603–604, 609 495–496, 499, 501, 503, exocentric 27, 31–32
copula 13, 109–111, 114, 116–117, 507, 510–526, 529–530, expectation 330, 389,
130, 136, 139, 146, 163, 222, 534–538, 545–546, 494, 520
260, 308, 334–335, 337, 581–582, 586–588, 592, explanation 13, 16–17, 41,
340, 345, 347–348, 351–353, 595, 598–599 116, 122, 132, 137, 163, 271,
355, 363–364, 368, 381, 385, dependent verb forms 112–113, 292, 313, 334, 340–345,
393–394, 472, 497, 505, 143, 147, 586 348–349, 351–352, 387,
539, 542 deranked linkage 67 451, 464–466, 473–476,
core 27, 32, 36, 67, 83, 107, 118, derivation 146, 163, 181, 183, 485–490, 495, 610
135, 143, 167, 182, 184, 209, 205, 314, 526, 528–529, 553 explanatory focusing 343, 345,
223, 240, 253, 255, 273, 295, determiner 31, 37, 355–356, 348, 351
301, 306, 430, 511 364, 381, 446 explicative converb 107, 113,
coreferential person 595, 599 detrimental 17, 478, 494 135–136, 139
correlation 88, 106, 112, 144, discordance 484, 494–495 extensible scope 58, 59, 82
207, 225, 229, 233 discourse 4, 6, 11, 13–14, 16–19, extraction 3, 8, 19, 51, 54,
correlative (markers) 1, 5, 12, 121, 162, 167, 181, 194–195, 69–74, 81, 86, 89, 91–92,
14, 249, 272, 279–281, 285, 200, 203–204, 211–212, 98, 100, 609, 615
292, 294, 298, 306, 379, 219–220, 222, 226, 229,
387, 421–422, 424–431, 237, 242–244, 247, 250, F
433–437, 439, 441, 444–446 258, 260, 269–271, 273, feature percolation 31
cosubordination 6, 8, 27–30, 277, 281, 290, 306, 318, final clause 6–7, 28–30, 40, 42,
51–55, 84, 93, 239–240, 430 333–334, 338, 340, 342, 357, 45–47, 63, 388, 458, 557,
counter-expectation 330 359, 365, 367, 370, 376, 567–569, 572, 578
cross-clausal reflexivization 80 381, 383, 385–387, 393–394, finite, finiteness 2, 4–6, 9–11,
399–400, 402, 406–408, 30, 33–36, 48, 53–54,
D 410–411, 413–414, 417, 62–63, 67–69, 81, 85,
dative 108, 116, 119, 137, 139, 423–424, 426–428, 434, 91–93, 98–101, 106, 111–112,
150, 154, 156, 163, 184–186, 438, 444–446, 452, 460, 143–144, 146–149, 154–157,
190, 193, 331, 401, 418, 431, 462, 464–465, 473, 476, 160, 162, 165, 167–171,
452, 459 482, 488–489, 494–496, 173–174, 185, 199–200,
debitive 9, 149, 154–156, 163 499, 503, 513, 545, 554–556, 210–211, 221, 224, 230,
deictic, deixis 12–14, 37, 44, 564–565, 579, 581–582, 587, 233–235, 253–254, 269,
222–223, 226, 270–271, 591–592, 599, 608–612, 614 274, 305, 308, 334
275–277, 285, 289–290, discursive inference 352 flexible position 85, 93
299, 304, 306, 308, 317, disjunct scope 53, 54, 5, 58–60, focalization 367–368, 381, 489,
334–336, 345, 347, 351–353, 65, 89, 90 582, 589, 609
Topic index
146–148, 154, 162–163, 173, 348, 363–364, 368, 385, 282, 284–286, 298, 313,
180, 190, 195, 207, 216, 227, 394–395, 422, 475–479, 373, 383, 434, 439, 444, 451,
229–230, 247, 250–251, 253, 488–490, 505, 523, 539, 543 456, 458–461, 461–462,
256, 271, 273, 281, 290–295, present 1, 14, 19, 30, 33, 43–45, 465–467, 479, 491–492,
297, 299, 302–305, 324, 49, 57, 61, 66, 107, 111–112, 507, 520, 525–526, 529,
356–357, 373, 376, 378, 116, 130, 132, 137–138, 144, 587, 589
381–382, 384–385, 390, 395, 149, 163, 167–169, 172, 174, purposive converb 9, 107, 113,
427–428, 431–432, 436, 194, 196, 203, 222–223, 121, 136–139
438–439, 453, 462–463, 226, 231, 235–237, 248,
489, 491, 495, 513, 523, 539, 257–258, 261, 270, 313–314, Q
543–544 332–334, 351, 361, 368, 385, qualitative predication 339,
possibility 3, 37, 44, 47, 57, 59, 392–395, 399, 401–402, 473, 474–476, 484, 485,
65, 68, 74–75, 78, 105, 107, 407, 421, 427, 432, 437, 452, 488–490
133, 372, 384–386, 390, 414, 457, 470–472, 475–476, question words 46, 69–70, 72,
550, 572, 574–575, 611 478, 488, 494, 524, 558, 74, 81, 88, 90, 544
potential 14–16, 110–111, 113, 560, 564, 572, 574–575, 579, quotative 9, 13–14, 165–166,
139, 199, 230, 235, 274, 308, 604, 606–608, 615 170, 187–188, 190, 193,
359, 379, 412, 443, 451–452, presentative 13, 17, 281, 331, 195, 197, 253, 295, 334, 350,
456–461, 464, 466–467, 335–337, 340, 342, 344–348, 355, 376, 388–389, 391,
504, 506, 514–517, 546 351–353, 364, 368, 381, 394–395, 506
pragmatics 1, 4–5, 9–11, 13, 15, 386–387, 393, 470–472,
17–18, 51, 54, 90, 118, 148, 477–478, 482, 484–486, R
165, 173, 199, 211, 248, 270, 493–495, 497 raising 80, 89, 151
272–273, 299, 302–303, presupposition 4–5, 12–13, realis 7, 42–44, 49, 60, 65–66,
306, 333, 337, 342, 344, 16–18, 34, 46–48, 57, 92, 167, 169, 190, 284, 286,
348, 351–352, 404–405, 199, 269–276, 287–289, 289, 292, 301, 305, 308,
421–425, 427–432, 291–294, 296–297, 300, 387, 416, 502, 504, 506,
434–435, 438, 440–441, 305–306, 339–342, 421, 513–514, 530–533, 537, 545
444–446, 467, 469, 424, 451, 462, 464–466, reanalysis 13, 144, 224, 334, 526
473–474, 483, 485, 488, 475–476, 478, 485, reason clause 13, 190, 260, 289,
490, 496, 499, 502–503, 488–490, 494, 533, 535, 315, 321
513–518, 530–536, 543–544, 537, 540–541, 545 referential 260, 270, 278, 283,
546, 592, 597–598, 607 procedural meaning 452, 466 288–289, 292–293, 297,
pragmatic demotion 499, 503, prohibition 377, 387, 395, 461 306, 350, 356–357, 364,
530, 546 prosody 9, 16, 144, 148, 380–381, 383, 385–387, 390,
pragmatic dependency 333, 163, 270, 272, 344, 460, 395, 606
469, 490, 496 518–519, 540–541, 543 referential hierarchy 1, 5,
predicate 4, 10, 52, 58, 77, 82, protasis 14, 17–18, 132, 251, 11–12, 267, 269–272,
109, 117, 145, 149, 154, 158, 279, 358, 371, 379–380, 274–275, 285, 294, 297,
167, 170–171, 174–175, 177, 390–393, 395–396, 443, 304–306, 346
180–185, 191, 194, 197–200, 487, 492–493, 508, 514, referentiality 365, 381, 392
217, 225, 231–233, 236–237, 522, 536 relative (clause) 2, 8, 10–11,
286, 317, 334, 337–341, prototypical 6, 7, 27, 29, 41, 13–14, 18–19, 32–33, 36,
344–345, 347–349, 353, 356, 82, 93, 219, 388, 437–438, 38, 44–45, 61, 70–73,
363–364, 366, 378, 385, 393, 473, 488 82–85, 87–88, 91, 106, 115,
471, 473–474, 476–477, proverbs 192, 370, 383, 122–123, 127, 152–154, 157,
488–489, 499, 502–507, 479, 490 171, 176, 180, 183–185, 199,
513, 523, 526, 530–545, 555 purpose (clauses) 9–11, 203, 205–207, 222–227,
predicate focusing 340 15–17, 19, 52, 58, 105–109, 230–237, 240, 253,
predicative relation 11, 18, 133, 136–137, 149, 154, 259–261, 275–278,
106, 110, 112, 167, 171, 158, 161, 165–166, 190, 281–290, 292, 295,
205–206, 225–226, 331, 193–195, 198–199, 226, 243, 297–300, 302, 305–306,
335–336, 338–340, 345, 256–258, 259–260, 275, 309, 326, 333–334, 336–337,
Topic index
339, 345, 350–353, 361, 363, 281, 290, 293–294, 297, 483–486, 490, 493, 495,
365–367, 371, 381–382, 299, 302–303, 305, 313–314, 504, 506, 528–533
384–385, 387, 394–395, 318, 321, 324–325, 328, status (realis/irrealis)
399, 408–409, 415, 330–331, 337, 339–341, 348, marker 60, 65
422, 425, 429, 434, 439, 355–356, 359, 367–369, subject 5, 7, 11, 16, 29–34,
462, 469, 483, 497, 502, 381–382, 389–390, 395, 36, 41, 48–49, 52, 55, 58,
508–509, 519–521, 525, 530, 405, 421–424, 427–440, 65–69, 73, 77, 81, 85, 150,
535–546, 555–556, 558–561, 443–446, 469–470, 153, 158–160, 170, 173–175,
563–567, 569–570, 574, 579 472–475, 479–482, 181, 200, 204–205,
relativization 115, 166, 181–183, 485–486, 488–492, 494, 208–212, 218–221, 229–230,
205, 363, 538, 540 502–508, 513–526, 234, 238–244, 246–248,
residuals 88–91 530–546, 549–550, 250, 254–258, 260, 283,
resultative 13, 225, 476, 512, 554–556, 558, 561–563, 567, 298, 301, 317, 325–326, 334,
526–529, 559 570, 578, 580, 612, 614 337–340, 345–348, 353,
rheme 367, 472–473, 476–477 sequential 9–10, 14, 18, 44–45, 356–357, 363–364, 368,
right-detached position 49, 52, 79, 143, 147–148, 373, 375, 382, 385, 391, 402,
78, 382 156–157, 159–160, 162–163, 424–425, 431, 435–438, 453,
175–176, 179, 196, 201, 236, 455–461, 463, 465–466,
S 272, 275, 277, 280–281, 471–474, 483, 489–490,
scope 2–3, 6–8, 12–13, 29–30, 284–286, 298, 302–306, 494, 496–497, 503, 505,
34, 39–42, 44–48, 51–54, 309, 331, 370, 383–384, 392, 511, 520, 526, 535, 538–544,
56–63, 65, 69–70, 72, 395, 504 552–556, 558–559, 561,
74–75, 79–82, 85–86, sequential converb 63, 77, 563–564, 567–579, 610
88–93, 98, 100, 175–176, 162–163 subject focusing 339, 473, 497,
185, 200, 238, 271–274, similarity measure 82 540–544; see focus
283, 294, 299, 302, 304, similative converb 107, 113, subjunctive 17–18, 65, 68, 393,
313–314, 326, 335–337, 340, 134, 139 401, 438, 499, 502–505,
345–346, 348–349, 360, simple sentence 15, 313–314, 507–510, 512–531, 545,
429, 439, 445, 489, 502, 331, 421, 446, 554 578, 607
533, 536, 608 simultaneity 11, 17–18, 125, subordinate clauses 3–4, 9–12,
semantic role assignment 176, 179, 188, 235–236, 411, 14, 18, 27–28, 33–34, 36,
78–79 492–495, 556, 565, 592 38–39, 48, 207, 233, 240,
semantic relations 1, 5, 10–11, situational anaphora 16, 384, 252, 254, 269, 273, 284,
15, 17–18, 31, 46, 56, 74, 482–483, 491–492, 495 286, 288, 294–295, 305,
121, 171, 176, 166, 188, 198, situational dependency 6, 334, 352, 366, 372, 383–384,
203–205, 208, 232, 234, 19, 383, 469, 490, 496, 389, 392, 409–410,
236–237, 239–240, 252, 587–588, 598 421–422, 424–441,
270, 272, 284, 287, 302, speaker’s commitment 444–446, 469, 512, 537,
334–335, 344, 352, 400, 491, 495 595–599
426–427, 451–452, 454, specialized converb 118, subordination 1–6, 8, 10, 12,
456, 461, 467, 469, 476, 181, 198 13–19, 27, 32, 34, 39, 48,
484–485, 490, 492, 495, speech act 17, 118, 145, 461, 51–52, 54, 57, 65–66, 69,
533, 556, 558, 564, 611 477, 491, 494, 513–518, 612; 76–78, 84, 86, 93, 105–107,
sentence 4, 9, 12–16, 18–19, 29, see illocutionary force 115, 135, 138, 147, 150,
31, 34, 40, 43–48, 55, 61, speech-act modification 78 161–162, 166, 169, 171, 198,
63, 68, 78–79, 91, 115–116, split assertion 339, 475, 488 201, 203, 207, 219, 221,
119–120, 129, 154, 165, 167, split graphs 84 223–224, 230, 233–236,
170–174, 176, 180–181, statement 30, 39, 46–47, 238, 239–240, 252, 260, 267,
183–185, 190, 193–194, 349, 443, 475–476, 481, 271–275, 281, 285–289, 293,
196–197, 199, 204, 218, 484–486, 518, 607 295–298, 303–307, 314,
224, 226, 228–229, 243, stative 208–210, 236–237, 333–334, 358, 374, 376,
250, 252, 257, 259–260, 257, 261, 309, 360, 471, 381, 383, 387–389, 394,
270–271, 274–275, 277, 474–475, 477–478, 399–400, 410–417,
Topic index
421–422, 425, 430, 446, 395, 409, 415, 423, 449, time of utterance 174, 411,
451–452, 455, 458–459, 499, 501–546; see 415, 417
469, 479, 481, 490, 499, tense-aspect-mood topic 4, 6, 11–13, 33, 36–38,
501, 503, 506, 508–510, temporal clause 184, 231, 49, 57, 66, 69–70, 73,
512–514, 517, 519, 521–522, 371–372, 383, 392, 484, 75–76, 78–80, 84–85, 92,
525–526, 529–530, 556, 576 100–101, 107, 112, 115, 144,
534–540, 545, 569, temporal converb 64–65 173–175, 181, 183, 200, 207,
580–582, 586–587, 594, temporality 373, 603–605, 615 218, 230, 243, 247, 250, 256,
598–599, 603–604, tense 4, 6–11, 13–16, 260, 269–282, 284–285,
609, 615 18–19, 27–29, 31, 33, 36, 293–295, 297–301,
succession 10, 16–17, 196, 219, 39–40, 42–45, 48–49, 303–309, 313, 319–320,
235–236, 256, 371, 460, 54, 58, 60–62, 65–68, 81, 333–335, 338, 340, 345,
466, 469, 480, 482–484, 90–91, 93, 110–111, 122, 143, 348–353, 357, 367, 381–382,
487, 492, 495–496, 549, 146–147, 163, 167, 169–172, 390–393, 395, 410, 444,
558–562, 565, 566–567, 180–181, 185, 199–200, 451, 462, 464, 524, 534,
569–571, 574, 577–579 205–207, 210–211, 218, 220, 539, 554
surprise 17, 111, 478, 222, 225–227, 229–230, topic clauses 12, 69, 73, 75, 271,
494–495, 610 233–235, 237, 240, 246, 274, 301, 305–307, 524,
switch-reference 66, 68, 71, 73, 251, 257–258, 261, 334, 355, 534, 545
80, 260 383–384, 391, 394, 418, 423, topicalization 13, 246–247,
symmetry 4, 57, 67–68, 81, 438–439, 453–457, 460, 256, 314, 333, 338, 367, 381,
91–93, 98, 100, 204, 407 469–472, 476, 478–485, 390–392, 395
syntactic dependency 171, 198, 487, 490–497, 499, 501, transported scope 60, 62
201, 469, 480, 490–491, 504–505, 511, 549, 554, typology 1, 41, 56, 65, 83–84,
495, 513, 517, 519–520, 522 556, 558, 560–564, 93, 166, 170, 198, 203, 207,
syntax 1, 5–6, 11, 17–18, 567–574, 577–578, 230, 355, 392, 451, 469,
51, 54, 56, 74, 112, 138, 604–606, 609, 615 489, 495
144, 147–148, 165, 205, tense-aspect-mood 222,
208–209, 216, 230, 232, 355, 499, 501, 504, 511; U
240–241, 259, 270, 273, see TAM universals 2–3, 51, 80, 86,
405, 422, 426, 446, 499, tense-mood concordance 18, 90, 93
503, 519, 529, 536, 538, 549, 554, 556, 561–564,
541–546, 554 567–574, 577 V
synthetic processes 582, 586 time clauses 176, 184, 231, verb serialization 171, 200,
275–276, 278, 295, 298, 501, 503, 511–512, 526,
T 300, 399–400, 409–411, 531, 540
TAM 18, 109–110, 112, 115, 415–417, 484, 506,
210–211, 218, 222, 229, 509–510, 521, 525, 529 W
355, 359, 368–370, 372–373, time of speech 19, 286, 478, wh-questions 472, 474,
376, 380–389, 392–393, 484, 490 490, 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, 473 pp. + index. Expected January
2011
122 CLANCY, Steven J.: The Chain of Being and Having in Slavic. xix, 289 pp. + index. Expected December 2010
121 BRIL, Isabelle (ed.): Clause Linking and Clause Hierarchy. Syntax and pragmatics. 2010. viii, 632 pp.
120 ROTHSTEIN, Björn and Rolf THIEROFF (eds.): Mood in the Languages of Europe. xiv, 643 pp. + index.
Expected November 2010
119 STATHI, Katerina, Elke GEHWEILER and Ekkehard KÖNIG (eds.): Grammaticalization. Current views
and issues. 2010. vii, 379 pp.
118 MÜHLEISEN, Susanne: Heterogeneity in Word-Formation Patterns. A corpus-based analysis of suffixation
with -ee and its productivity in English. 2010. xiii, 245 pp.
117 SPEVAK, Olga: Constituent Order in Classical Latin Prose. 2010. xv, 318 pp.
116 NORDSTRÖM, Jackie: Modality and Subordinators. 2010. xvii, 341 pp.
115 HASKO, Victoria and Renee PERELMUTTER (eds.): New Approaches to Slavic Verbs of Motion. 2010.
x, 392 pp.
114 ROBY, David Brian: Aspect and the Categorization of States. The case of ser and estar in Spanish. 2009.
xiii, 191 pp.
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, 18–20 October, 2007. 2009. xi, 422 pp.
112 DUFTER, Andreas and Daniel JACOB (eds.): Focus and Background in Romance Languages. 2009.
vii, 362 pp.
111 POLGUÈRE, Alain and Igor A. MEL’ČUK (eds.): Dependency in Linguistic Description. 2009.
xxii, 281 pp.
110 DIMMENDAAL, Gerrit J. (ed.): Coding Participant Marking. Construction types in twelve African
languages. 2009. xvi, 389 pp.
109 NARROG, Heiko: Modality in Japanese. The layered structure of the clause and hierarchies of functional
categories. 2009. xxii, 277 pp.
108 BARÐDAL, Jóhanna and Shobhana L. CHELLIAH (eds.): The Role of Semantic, Pragmatic, and
Discourse Factors in the Development of Case. 2009. xx, 432 pp.
107 BUTLER, Christopher S. and Javier MARTÍN ARISTA (eds.): Deconstructing Constructions. 2009.
xx, 306 pp.
106 VANHOVE, Martine (ed.): From Polysemy to Semantic Change. Towards a typology of lexical semantic
associations. 2008. xiii, 404 pp.
105 VAN VALIN, JR., Robert D. (ed.): Investigations of the Syntax–Semantics–Pragmatics Interface. 2008.
xxiv, 484 pp.
104 MUSHIN, Ilana and Brett BAKER (eds.): Discourse and Grammar in Australian Languages. 2008.
x, 239 pp.
103 JOSEPHSON, Folke and Ingmar SÖHRMAN (eds.): Interdependence of Diachronic and Synchronic
Analyses. 2008. viii, 350 pp.
102 GODDARD, Cliff (ed.): Cross-Linguistic Semantics. 2008. xvi, 356 pp.
101 STOLZ, Thomas, Sonja KETTLER, Cornelia STROH and Aina URDZE: Split Possession. An areal-
linguistic study of the alienability correlation and related phenomena in the languages of Europe. 2008.
x, 546 pp.
100 AMEKA, Felix K. and M.E. KROPP DAKUBU (eds.): Aspect and Modality in Kwa Languages. 2008.
ix, 335 pp.
99 HØEG MÜLLER, Henrik and Alex KLINGE (eds.): Essays on Nominal Determination. From morphology
to discourse management. 2008. xviii, 369 pp.
98 FABRICIUS-HANSEN, Cathrine and Wiebke RAMM (eds.): 'Subordination' versus 'Coordination' in
Sentence and Text. A cross-linguistic perspective. 2008. vi, 359 pp.
97 DOLLINGER, Stefan: New-Dialect Formation in Canada. Evidence from the English modal auxiliaries.
2008. xxii, 355 pp.
96 ROMEO, Nicoletta: Aspect in Burmese. Meaning and function. 2008. xv, 289 pp.
95 O’CONNOR, Loretta: Motion, Transfer and Transformation. The grammar of change in Lowland Chontal.
2007. xiv, 251 pp.
94 MIESTAMO, Matti, Kaius SINNEMÄKI and Fred KARLSSON (eds.): Language Complexity. Typology,
contact, change. 2008. xiv, 356 pp.
93 SCHALLEY, Andrea C. and Drew KHLENTZOS (eds.): Mental States. Volume 2: Language and cognitive
structure. 2007. x, 362 pp.
92 SCHALLEY, Andrea C. and Drew KHLENTZOS (eds.): Mental States. Volume 1: Evolution, function,
nature. 2007. xii, 304 pp.
91 FILIPOVIĆ, Luna: Talking about Motion. A crosslinguistic investigation of lexicalization patterns. 2007.
x, 182 pp.
90 MUYSKEN, Pieter (ed.): From Linguistic Areas to Areal Linguistics. 2008. vii, 293 pp.
89 STARK, Elisabeth, Elisabeth LEISS and Werner ABRAHAM (eds.): Nominal Determination. Typology,
context constraints, and historical emergence. 2007. viii, 370 pp.
88 RAMAT, Paolo and Elisa ROMA (eds.): Europe and the Mediterranean as Linguistic Areas. Convergencies
from a historical and typological perspective. 2007. xxvi, 364 pp.
87 VERHOEVEN, Elisabeth: Experiential Constructions in Yucatec Maya. A typologically based analysis of a
functional domain in a Mayan language. 2007. xiv, 380 pp.
86 SCHWARZ-FRIESEL, Monika, Manfred CONSTEN and Mareile KNEES (eds.): Anaphors in Text.
Cognitive, formal and applied approaches to anaphoric reference. 2007. xvi, 282 pp.
85 BUTLER, Christopher S., Raquel HIDALGO DOWNING and Julia LAVID (eds.): Functional
Perspectives on Grammar and Discourse. In honour of Angela Downing. 2007. xxx, 481 pp.
84 WANNER, Leo (ed.): Selected Lexical and Grammatical Issues in the Meaning–Text Theory. In honour of
Igor Mel'čuk. 2007. xviii, 380 pp.
83 HANNAY, Mike and Gerard J. STEEN (eds.): Structural-Functional Studies in English Grammar. In
honour of Lachlan Mackenzie. 2007. vi, 393 pp.
82 ZIEGELER, Debra: Interfaces with English Aspect. Diachronic and empirical studies. 2006. xvi, 325 pp.
81 PEETERS, Bert (ed.): Semantic Primes and Universal Grammar. Empirical evidence from the Romance
languages. 2006. xvi, 374 pp.
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, 350 pp.
79 LAFFUT, An: Three-Participant Constructions in English. A functional-cognitive approach to caused
relations. 2006. ix, 268 pp.
78 YAMAMOTO, Mutsumi: Agency and Impersonality. Their Linguistic and Cultural Manifestations. 2006.
x, 152 pp.
77 KULIKOV, Leonid, Andrej MALCHUKOV and Peter de SWART (eds.): Case, Valency and Transitivity.
2006. xx, 503 pp.
76 NEVALAINEN, Terttu, Juhani KLEMOLA and Mikko LAITINEN (eds.): Types of Variation. Diachronic,
dialectal and typological interfaces. 2006. viii, 378 pp.
75 HOLE, Daniel, André MEINUNGER and Werner ABRAHAM (eds.): Datives and Other Cases. Between
argument structure and event structure. 2006. viii, 385 pp.
74 PIETRANDREA, Paola: Epistemic Modality. Functional properties and the Italian system. 2005.
xii, 232 pp.
73 XIAO, Richard and Tony McENERY: Aspect in Mandarin Chinese. A corpus-based study. 2004. x, 305 pp.
72 FRAJZYNGIER, Zygmunt, Adam HODGES and David S. ROOD (eds.): Linguistic Diversity and
Language Theories. 2005. xii, 432 pp.
71 DAHL, Östen: The Growth and Maintenance of Linguistic Complexity. 2004. x, 336 pp.
70 LEFEBVRE, Claire: Issues in the Study of Pidgin and Creole Languages. 2004. xvi, 358 pp.
69 TANAKA, Lidia: Gender, Language and Culture. A study of Japanese television interview discourse. 2004.
xvii, 233 pp.
68 MODER, Carol Lynn and Aida MARTINOVIC-ZIC (eds.): Discourse Across Languages and Cultures.
2004. vi, 366 pp.
67 LURAGHI, Silvia: On the Meaning of Prepositions and Cases. The expression of semantic roles in Ancient
Greek. 2003. xii, 366 pp.
66 NARIYAMA, Shigeko: Ellipsis and Reference Tracking in Japanese. 2003. xvi, 400 pp.
65 MATSUMOTO, Kazuko: Intonation Units in Japanese Conversation. Syntactic, informational and
functional structures. 2003. xviii, 215 pp.
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, 579 pp.
63 BUTLER, Christopher S.: Structure and Function – A Guide to Three Major Structural-Functional
Theories. Part 1: Approaches to the simplex clause. 2003. xx, 573 pp.
62 FIELD, Fredric: Linguistic Borrowing in Bilingual Contexts. With a foreword by Bernard Comrie. 2002.
xviii, 255 pp.
61 GODDARD, Cliff and Anna WIERZBICKA (eds.): Meaning and Universal Grammar. Theory and
empirical findings. Volume 2. 2002. xvi, 337 pp.
60 GODDARD, Cliff and Anna WIERZBICKA (eds.): Meaning and Universal Grammar. Theory and
empirical findings. Volume 1. 2002. xvi, 337 pp.
59 SHI, Yuzhi: The Establishment of Modern Chinese Grammar. The formation of the resultative construction
and its effects. 2002. xiv, 262 pp.
58 MAYLOR, B. Roger: Lexical Template Morphology. Change of state and the verbal prefixes in German.
2002. x, 273 pp.
57 MEL’ČUK, Igor A.: Communicative Organization in Natural Language. The semantic-communicative
structure of sentences. 2001. xii, 393 pp.
56 FAARLUND, Jan Terje (ed.): Grammatical Relations in Change. 2001. viii, 326 pp.
55 DAHL, Östen and Maria KOPTJEVSKAJA-TAMM (eds.): Circum-Baltic Languages. Volume 2: Grammar
and Typology. 2001. xx, 423 pp.
54 DAHL, Östen and Maria KOPTJEVSKAJA-TAMM (eds.): Circum-Baltic Languages. Volume 1: Past and
Present. 2001. xx, 382 pp.
53 FISCHER, Olga, Anette ROSENBACH and Dieter STEIN (eds.): Pathways of Change.
Grammaticalization in English. 2000. x, 391 pp.
52 TORRES CACOULLOS, Rena: Grammaticization, Synchronic Variation, and Language Contact. A study of
Spanish progressive -ndo constructions. 2000. xvi, 255 pp.
51 ZIEGELER, Debra: Hypothetical Modality. Grammaticalisation in an L2 dialect. 2000. xx, 290 pp.
50 ABRAHAM, Werner and Leonid KULIKOV (eds.): Tense-Aspect, Transitivity and Causativity. Essays in
honour of Vladimir Nedjalkov. 1999. xxxiv, 359 pp.
49 BHAT, D.N.S.: The Prominence of Tense, Aspect and Mood. 1999. xii, 198 pp.
48 MANNEY, Linda Joyce: Middle Voice in Modern Greek. Meaning and function of an inflectional category.
2000. xiii, 262 pp.
47 BRINTON, Laurel J. and Minoji AKIMOTO (eds.): Collocational and Idiomatic Aspects of Composite
Predicates in the History of English. 1999. xiv, 283 pp.
46 YAMAMOTO, Mutsumi: Animacy and Reference. A cognitive approach to corpus linguistics. 1999.
xviii, 278 pp.
45 COLLINS, Peter and David LEE (eds.): The Clause in English. In honour of Rodney Huddleston. 1999.
xv, 342 pp.
44 HANNAY, Mike and A. Machtelt BOLKESTEIN (eds.): Functional Grammar and Verbal Interaction. 1998.
xii, 304 pp.
43 OLBERTZ, Hella, Kees HENGEVELD and Jesús SÁNCHEZ GARCÍA (eds.): The Structure of the
Lexicon in Functional Grammar. 1998. xii, 312 pp.
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, 407 pp.
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, 486 pp.
40 BIRNER, Betty J. and Gregory WARD: Information Status and Noncanonical Word Order in English.
1998. xiv, 314 pp.
39 WANNER, Leo (ed.): Recent Trends in Meaning–Text Theory. 1997. xx, 202 pp.
38 HACKING, Jane F.: Coding the Hypothetical. A comparative typology of Russian and Macedonian
conditionals. 1998. vi, 156 pp.
37 HARVEY, Mark and Nicholas REID (eds.): Nominal Classification in Aboriginal Australia. 1997. x, 296 pp.
36 KAMIO, Akio (ed.): Directions in Functional Linguistics. 1997. xiii, 259 pp.
35 MATSUMOTO, Yoshiko: Noun-Modifying Constructions in Japanese. A frame semantic approach. 1997.
viii, 204 pp.
34 HATAV, Galia: The Semantics of Aspect and Modality. Evidence from English and Biblical Hebrew. 1997.
x, 224 pp.
33 VELÁZQUEZ-CASTILLO, Maura: The Grammar of Possession. Inalienability, incorporation and
possessor ascension in Guaraní. 1996. xvi, 274 pp.
32 FRAJZYNGIER, Zygmunt: Grammaticalization of the Complex Sentence. A case study in Chadic. 1996.
xviii, 501 pp.
31 WANNER, Leo (ed.): Lexical Functions in Lexicography and Natural Language Processing. 1996.
xx, 355 pp.
30 HUFFMAN, Alan: The Categories of Grammar. French lui and le. 1997. xiv, 379 pp.
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, 510 pp.
28 HERMAN, József (ed.): Linguistic Studies on Latin. Selected papers from the 6th International Colloquium
on Latin Linguistics (Budapest, 23–27 March 1991). 1994. ix, 421 pp.
27 ABRAHAM, Werner, T. GIVÓN and Sandra A. THOMPSON (eds.): Discourse, Grammar and Typology.
Papers in honor of John W.M. Verhaar. 1995. xx, 352 pp.
26 LIMA, Susan D., Roberta CORRIGAN and Gregory K. IVERSON: The Reality of Linguistic Rules. 1994.
xxiii, 480 pp.
25 GODDARD, Cliff and Anna WIERZBICKA (eds.): Semantic and Lexical Universals. Theory and
empirical findings. 1994. viii, 510 pp.
24 BHAT, D.N.S.: The Adjectival Category. Criteria for differentiation and identification. 1994. xii, 295 pp.
23 COMRIE, Bernard and Maria POLINSKY (eds.): Causatives and Transitivity. 1993. x, 399 pp.
22 McGREGOR, William B.: A Functional Grammar of Gooniyandi. 1990. xx, 618 pp.
21 COLEMAN, Robert (ed.): New Studies in Latin Linguistics. Proceedings of the 4th International
Colloquium on Latin Linguistics, Cambridge, April 1987. 1990. x, 480 pp.
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, 409 pp.
19 BLUST, Robert A.: Austronesian Root Theory. An essay on the limits of morphology. 1988. xi, 190 pp.
18 WIERZBICKA, Anna: The Semantics of Grammar. 1988. vii, 581 pp.
17 CALBOLI, Gualtiero (ed.): Subordination and Other Topics in Latin. Proceedings of the Third Colloquium
on Latin Linguistics, Bologna, 1–5 April 1985. 1989. xxix, 691 pp.
16 CONTE, Maria-Elisabeth, János Sánder PETÖFI and Emel SÖZER (eds.): Text and Discourse
Connectedness. Proceedings of the Conference on Connexity and Coherence, Urbino, July 16–21, 1984.
1989. xxiv, 584 pp.
15 JUSTICE, David: The Semantics of Form in Arabic. In the mirror of European languages. 1987. iv, 417 pp.
14 BENSON, Morton, Evelyn BENSON and Robert F. ILSON: Lexicographic Description of English. 1986.
xiii, 275 pp.
13 REESINK, Ger P.: Structures and their Functions in Usan. 1987. xviii, 369 pp.
12 PINKSTER, Harm (ed.): Latin Linguistics and Linguistic Theory. Proceedings of the 1st International
Colloquium on Latin Linguistics, Amsterdam, April 1981. 1983. xviii, 307 pp.
11 PANHUIS, Dirk G.J.: The Communicative Perspective in the Sentence. A study of Latin word order. 1982.
viii, 172 pp.
10 DRESSLER, Wolfgang U., Willi MAYERTHALER, Oswald PANAGL and Wolfgang Ullrich WURZEL:
Leitmotifs in Natural Morphology. 1987. ix, 168 pp.
9 LANG, Ewald and John PHEBY: The Semantics of Coordination. (English transl. by John Pheby from the
German orig. ed. 'Semantik der koordinativen Verknüpfung', Berlin, 1977). 1984. 300 pp.
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, 333 pp.
7 PARRET, Herman, Marina SBISÀ and Jef VERSCHUEREN (eds.): Possibilities and Limitations of
Pragmatics. Proceedings of the Conference on Pragmatics, Urbino, July 8–14, 1979. 1981. x, 854 pp.