Coordinating Constructions (Typological Studies in Language) - PDF Free Download (2024)

Coordinating Constructions

Typological Studies in Language (TSL) A companion series to the journal Studies in Language

General Editor

Michael Noonan

Assistant Editors

Spike Gildea, Suzanne Kemmer

Editorial Board Wallace Chafe (Santa Barbara) Bernard Comrie (Leipzig) R. M. W. Dixon (Melbourne) Matthew Dryer (Buffalo) John Haiman (St Paul) Bernd Heine (Köln) Paul Hopper (Pittsburgh) Andrej Kibrik (Moscow) Ronald Langacker (San Diego)

Charles Li (Santa Barbara) Edith Moravcsik (Milwaukee) Andrew Pawley (Canberra) Doris Payne (Eugene, OR) Frans Plank (Konstanz) Jerrold Sadock (Chicago) Dan Slobin (Berkeley) Sandra Thompson (Santa Barbara)

Volumes in this series will be functionally and typologically oriented, covering specific topics in language by collecting together data from a wide variety of languages and language typologies. The orientation of the volumes will be substantive rather than formal, with the aim of investigating universals of human language via as broadly defined a data base as possible, leaning toward crosslinguistic, diachronic, developmental and live-discourse data.

Volume 58 Coordinating Constructions Edited by Martin Haspelmath

Coordinating Constructions Edited by

Martin Haspelmath Max Planck Institute for Evolutionary Anthropology, Leipzig

John Benjamins Publishing Company Amsterdam/Philadelphia



The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Coordinating Constructions / edited by Martin Haspelmath. p. cm. (Typological Studies in Language, issn 0167–7373 ; v. 58) Includes bibliographical references and indexes. 1. Grammar, Comparative and general--Coordinate constructions. 2. Contrastive linguistics. 3. Typology (Linguistics) I. Haspelmath, Martin, 1963II. Series. P299.C6C66 2004 415--dc22 isbn 90 272 2966 X (Eur.) / 1 58811 479 1 (US) (Hb; alk. paper)


© 2004 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

To the memory of Helma van den Berg

Table of contents

Map showing the languages discussed in this volume General list of abbreviations of grammatical categories Acknowledgments

ix x xiii

General articles 1. Coordinating constructions: An overview Martin Haspelmath


2. Coordination in Mentalese Toshio Ohori


3. Coordination: An adaptationist view Jeffrey Heath


4. Conjunction and personal pronouns D. N. S. Bhat


Africa 5. The grammar of conjunctive and disjunctive coordination in Iraqw Maarten Mous 6. Coordinating constructions in Fongbe with reference to Haitian Creole Claire Lefebvre 7. Comitative, coordinating, and inclusory constructions in Hausa Mahamane L. Abdoulaye




Caucasus 8. Coordinating constructions in Daghestanian languages Helma van den Berg 9. Where coordination meets subordination: Converb constructions in Tsakhur (Daghestanian) Konstantin I.Kazenin and Yakov G.Testelets



viii Table of contents

10. Coordination in Chechen 241 Liane Jeschull Middle East 11. Coordination in three Western Iranian languages: Vafsi, Persian and Gilaki Donald Stilo


Southeast Asia 12. Coordination in Hakha Lai (Tibeto-Burman) David A. Peterson and Kenneth VanBik 13. Conjunction and concatenation in Sgaw Karen: Familiarity, frequency, and conceptual unity Carol Lord and Louisa Benson Craig 14. Riau Indonesian sama: Explorations in macrofunctionality David Gil




Pacific 15. Coordination in Lavukaleve Angela Terrill


16. Coordination in Oceanic languages and Proto Oceanic Claire Moyse-Faurie and John Lynch


17. Coordination strategies and inclusory constructions in New Caledonian and other Oceanic languages Isabelle Bril


Americas 18. Coordination in Upper Kuskokwim Athabaskan Andrej A. Kibrik


Language index


Name index


Subject index


4 2


5 6

Arrernte 14 Chechen 7 Daghestanian languages 7 Dargi 7 Drehu 18 English 5 Fongbe 2


7 8 9 10


German 6 Haitian Creole 22 Hausa 3 Iraqw 1 Japanese 11 Karen 10 Koyraboro Senni 4

13 14


15 17 16 18




Lai 9 Lavukaleve 15 Lenakel 16 Maori 19 Nêlêmwa 17 New Caledonian languages 17 Nunggubuyu 13


Persian 8 Riau Indonesian 12 Samoan 20 Tsakhur 7 Upper Kuskokwim Athabaskan 21 Vafsi 8 Xârâcùu 17

Map 1.Languages described in this volume.


General list of abbreviations of grammatical categories

General list of abbreviations of grammatical categories

abl abs acc adj adv agr aor AP appl art asp aux ben caus comp cond conv dat def dem det dist distr du dur erg excl f foc fut gen imp impf incl indef

ablative absolutive accusative adjective adverb(ial) agreement aorist adjective phrase applicative article aspect auxiliary benefactive/beneficiary causative complementizer conditional converb dative definite demonstrative determiner distal distributive dual durative ergative exclusive feminine focus future genitive imperative imperfect(ive) inclusive indefinite

iness inf instr intr loc m n neg nmlz nom NP obj obl pass past perf pfv pl poss PP pred pres pret prog prox pst ptcp purp refl rel sg subj TA(M) tns top

inessive infinitive instrumental intransitive locative/localizer masculine neuter negation, negative nominalizer/nominalization nominative noun phrase object oblique passive past tense perfect perfective plural possessive adpositional phrase predicate present preterite progressive proximal/proximate past participle purposive reflexive relative singular subject tense-aspect(-mood) tense topic

General list of abbreviations of grammatical categories

tr VP

transitive verb phrase

1 2 3

first person second person third person



The majority of the papers published in this volume were first presented at a seminar on coordination at the Max Planck Institute for Evolutionary Anthropology in 2001. I am grateful to all those who took part in the discussions of that seminar. I was helped in the editing of this volume by Claudia Schmidt, and especially by Orin Gensler, who did a great job of copy-editing, proofreading and preparing the indexes. Finally, I thank Michael Noonan for accepting the volume in the series, and Kees Vaes and the other Benjamins staff for their assistance and their patience.

Leipzig, June 2004 Martin Haspelmath

General articles

Chapter 1

Coordinating constructions An overview Martin Haspelmath Max-Planck-Institut für evolutionäre Anthropologie, Leipzig

1. Coordination: Basic concepts and terms 1.1 Coordination 1.2 Basic patterns 1.3 Semantic types of coordination 2. The position of the coordinator(s) 3. Category-sensitivity of coordinating constructions 4. Semantic distinctions in conjunction 5. From comitative to conjunctive marker 6. Semantic maps 7. Inclusory constructions 8. Disjunction 9. Extraction from coordinating constructions 10. Ellipsis in coordination 11. Coordination and subordination/dependency


Coordination: Basic concepts and terms

1.1 Coordination This volume contains seventeen papers on coordinating constructions in languages from different families and different continents (see p. vii for a world map showing the most important languages treated in this book). The definition of the term coordination will be discussed in some detail in § 11 below. For the moment we take it for granted that coordinating constructions can be identified on the basis of their symmetry: A construction [A B] is considered coordinate if the two parts A and B have the same status (in some sense that needs to be specified further), whereas it is not coordinate if it is asymmetrical and one of the parts is clearly more salient or important, while the other part is in some sense subordinate. In practice, we


Martin Haspelmath

typically suspect that a construction will be coordinate if it is systematically used to render English constructions with the coordinating particles and, or and but. The contributions to this volume often use the terminological conventions proposed by Haspelmath (to appear a), so they should be recapitulated here briefly. Illustrations in this chapter mostly come from the papers in this volume. 1.2 Basic patterns A coordinating construction consists of two or more coordinands, i.e. coordinated phrases. Their coordinate status may be indicated by coordinators, i.e. particles like and, or and but, or affixes like Chechen -ii (e.g. shyyr-ii dik-ii ‘thick and good’; Jeschull°, ex. 15).1 If one or more coordinators occur in a coordinating construction, it is called syndetic. Asyndetic coordination consists of simple juxtaposition of the coordinands. An example is (1). (1) Lavukaleve (Terrill°, ex. 9) nga-bakala nga-uia tula 1sg.poss-paddle(m) 1sg.poss-knife(f) ‘my paddle and my small knife’ Two types of syndesis can be distinguished: monosyndetic coordination, which involves only a single coordinator (when not more than two coordinands are present), and bisyndetic coordination, which involves two coordinators. Examples are given in (2) and (3). (Here and elsewhere in this volume, the coordinators are boldfaced in the examples.) (2) monosyndetic: Iraqw (Mous°, ex. 29) kwa/angw nee du’uma hare and leopard ‘the hare and the leopard’ (3) bisyndetic: Upper Kuskokwim Athabaskan (Kibrik°, ex. 1) dineje ‘ił midzish ‘ił moose with caribou with ‘moose and caribou’ In addition to binary coordinating constructions with two coordinands, languages also allow multiple coordinands, i.e. more than two. In such constructions, “bisyndetic” coordination in fact means one coordinator per coordinand (cf. ex. 5), while monosyndetic coordination has one coordinator fewer (cf. ex. 4).

1.A small circle after an author’s name means that the author’s contribution to this volume is referred to. Thus, “Jeschull°” stands for “Jeschull (this volume)”.

Coordinating constructions: An overview

(4) monosyndetic: Iraqw (Mous°, ex. 16) Kwermuhl, nee Tlawi, nee Dongobesh, nee Haydom nee Daudi Kwermuhl and Tlawi and Dongobesh and Haydom and Daudi ‘Kwermuhl, Tlawi, Dongobesh, Haydom, and Daudi [place names]’ (5) bisyndetic: Upper Kuskokwim Athabaskan (Kibrik°, ex. 3) maladija ‘ił jamena ‘ił denk’a ‘ił łeka mama‘ ‘ił tent with stove with gun with dog food with ‘a tent, a stove, a gun, and dog food’ In many languages, all but the last coordinator can be omitted in monosyndetic constructions, as in the English translation of example (5) (e.g. in Iraqw: Mous°, ex. 15–18; in Hausa: Abdoulaye° § 3.1; in Hakha Lai NP conjunction: Peterson & VanBik°, ex. 9; Lavukaleve: Terrill°, ex. 14). However, there are also quite a few cases where coordinator omission has been reported to be impossible (e.g. in Hakha Lai clauses; Peterson & VanBik°, ex. 8). It seems that in bisyndetic constructions, coordinator omission is generally not possible with multiple coordinands. The coordinator usually combines with one of the coordinands, so that the construction is not entirely symmetric. If it precedes this coordinand, it is called prepositive; if it follows this coordinand, it is called postpositive. The evidence for prepositive status is clearest when the coordinator is a prefix or a procl*tic (e.g. 6), and for postpositive status when the coordinator is a suffix or encl*tic (e.g. 7). (6) prepositive: Lenakel (Moyse-Faurie & Lynch°, ex. 28a) I-em-va m-6m-ang6n. 1sg-past-come and-past-eat ‘I came and ate.’ (7) postpositive: Japanese (Ohori°, ex. 16) hon-to zasshi book-and magazine ‘a book and a magazine’ 1.3 Semantic types of coordination Three different semantic types of coordination are usually distinguished: conjunction (= conjunctive coordination, ‘and’ coordination, e.g. 8), disjunction (=disjunctive coordination, ‘or’ coordination, e.g. 9), and adversative coordination (‘but’ coordination, e.g. 10). (8) conjunction: Persian (Stilo°, ex. 25) tir=o kæman arrow=and bow ‘bow and arrow’



Martin Haspelmath

(9) disjunction: Iraqw (Mous°, ex. 17) tsíiyáhh laqáa tám laqáa tsár four or three or two ‘four or three or two’ (10) adversative coordination: Lavukaleve (Terrill°, ex. 40) Oina sou fale-re kini a-e-ve-meon taman rise stand-nonfin act 3sg.m.O-sbd-go-surp but lake ga e-nua-ri-re… road(n) 3sg.n.O-be.amiss-caus-nonfin ‘He stood up, but he took the wrong path…’ Sometimes an additional type “causal coordination” is distinguished (e.g. English constructions involving for, or German constructions involving denn ‘for, because’), but causal constructions much more often make use of subordinate clauses. In nonEuropean languages, causal constructions are rarely described as involving coordination (see, however, van den Berg° § 2.2.4 for causal coordination in Dargi). The coordinands of a conjunction are also called conjuncts. In the older literature, the term conjunction is often used as a cover-term for coordinators and subordinators, but this usage is avoided in this volume to minimize confusion.


The position of the coordinator(s)

In monosyndetic coordination, there are four logically possible types, which are listed in (11) in descending order of cross-linguistic frequency. In the formulas in (11), A and B stand for two coordinands, and co stands for the coordinator. (11) a.

[A] [co B]

e.g. Hausa

b. [A co] [B]

e.g. Lai


e.g. Latin

[A] [B co]

Abdù dà Feemì ‘Abdu and Femi’ (Abdoulaye°, ex. 14a) vòmpii=leé phè]tee ‘a bear and a rabbit’ (Peterson & VanBik°, ex. 4) senatus populus-que romanus ‘the senate and the Roman people’

d. [co A] [B] The fourth type seems to be unattested, and the third type is very rare — it does not occur in the languages discussed in this volume. Distinguishing between the first and the second type is often not straightforward, and there is the additional logical possibility of a symmetrical tripartite structure [A] [co] [B]. However, there is a broad consensus that English and other European languages have the bracketing [A] [and B], and it seems to be generally assumed that all languages have an asymmetry one way or the other (especially in the recent

Coordinating constructions: An overview

generative literature, cf. Progovac 2003). The following kinds of criteria have been mentioned for determining the constituency of coordinating constructions: (i) Clisis: In some languages, the coordinator is clearly phonologically attached to one of the coordinands, either as a procl*tic or as an encl*tic (or even as a prefix/ suffix — the difference between clisis and affixation is not relevant in the present context). For instance, in Hunzib (and in some other languages of the Tsezic branch of Nakh-Daghestanian), the shape of the coordinator is -no after a consonant (e.g. kid-no ‘girl and’) and -n after a vowel (e.g. ože-n ‘boy and’) (van den Berg°, ex. 49, and van den Berg 1995: 51).2 (ii) Intonational phrasing: When the coordinators are short, a coordinating construction A co B is pronounced as a single intonational phrase, but when they are longer (e.g. two full clauses), there is usually an intonation break between them (cf. Stilo° § 1.4.2), and the coordinator is then either attached at the beginning of the second phrase (as in 12), or at the end of the first phrase (as in 14b below). The intonation break is indicated by a comma. (12) Chechen (Jeschull°, ex. 80) So hwan gospodin vu, tq’a hwo san jalxoo vu. I you.gen master be.pres and you I.gen servant be.pres ‘I am your master, and you are my servant.’ For Hausa, Abdoulaye° (§ 3.1) observes that constructions with multiple coordinands can have separate intonational phrases for each coordinand (e.g. Abdù, dà Bàlki, dà Muusaa, dà Mo˜rù ‘Abdu and Balki and Musa and Moru’). (iii) Extraposition: Many languages allow extraposition of coordinands to the end of the clause, so that the construction is no longer continuous. In English, such extrapositions seem to occur mostly in afterthought constructions, but in German, they are perfectly natural even in carefully planned utterances because of the rigid object-verb order in certain constructions (cf. 13a). (13bc) show examples from Iraqw and Hausa (and Ohori°, ex. 42, cites one from Old English). (13) a.

German Schröder hat mit Fischer telefoniert und mit dem Agenten, der das Waffengeschäft aufgedeckt hat. ‘Schröder spoke to Fischer on the phone and to the agent who uncovered the weapons deal.’

2.However, Hunzib conjunction is bisyndetic (ože-n kid-no ‘a boy and a girl’), so that the question of deciding between [A co] [B] and [A][co B] does not arise for this language.



Martin Haspelmath

b. Iraqw (Mous°, ex. 35) nee masoomo bir-ta doog-iyé’ laqáa dasi… with youth cond-rec:perf meet-3pl:past or girl ‘If he meets a youth or a girl, …’ c. Hausa (Abdoulaye°, § 3.1) Abdù nee dà Muusaa sukà tàfi. (= Abdù dà Muusaa nèe sukà tàfi.) Abdu cop and Musa 3pl.pfv go ‘It is Abdu and Musa who went.’ This criterion for constituency would be even more convincing if we had examples of extraposition with postpositive (i.e. [A co][B]) constructions in which the postposed coordinator stays behind together with the first coordinand, but I know of no such cases (though van den Berg°, ex. 56, cites an example of extraposition with a bisyndetic construction, resulting in a pattern “… [A co] verb [B co]”). These three criteria often yield an unambiguous constituent structure, but occasionally they do not. Some languages have coordinators which may go either way in intonational phrasing. Kibrik° (§ 5) reports for Upper Kuskokwim Athabaskan that the adversative coordinator ‘edinh ‘but’ can belong either to the first or to the second coordinand: (14) Upper Kuskokwim Athabaskan (Kibrik°, ex. 33–34) a. hiyoko tsiłdilghwsr, ‘edinh mikwl for.her they.are.sobbing but ‘The are bemoaning her but she is gone.’ b. sileka ch’ildon’ nich’i toghedak ‘edinh, my.dogs part too but ch’ildon’ chu‘da tinh k’its’ ‘ohighet’a ts’e‘… part still ice on they.are.there and ‘While some of my dogs fell into the water, the others still stayed on the ice, and…’ Similarly, Peterson & VanBik° note that Hakha Lai =‘ii ‘and’ can occur clausefinally or clause-initially (ex. 1–3), and Terrill° (§ 4.1) finds for Lavukaleve that “pauses before coordinators are roughly equal in frequency to pauses after the coordinator”. In these cases, a symmetrical bracketing (i.e. [A] [co] [B]) may well be justified. Even more problematic for the criteria in (i)–(iii) above are cases of mismatches between them. One such case is cited by Stilo° from colloquial Persian, where the coordinator =ò ‘and’ is encl*tic to the preceding element (e.g. tir=o kæman ‘arrow and bow’, ex. 25). However, when a conjunct is extraposed, =ò is extraposed along with the conjunct and encl*ticizes onto whatever happens to precede it:

Coordinating constructions: An overview

(15) colloquial Persian (Stilo°, ex. 10) Xoda ye (dune) bæradær dad beh=éš=o ye xahær. God one (clf brother gave to=3s.obl=and one sister ‘God gave him a brother and a sister.’ Stilo argues that this shows that the syntactic bracketing is [A][co B], and that the encl*ticization is purely phonological. Allowing syntax-phonology mismatches amounts to discarding the first two criteria for determining syntactic constituent structure, leaving us only with the extraposition criterion. It thus appears that the constituent structure of coordinating constructions is much more problematic than has been generally thought. In the great majority of cases (including all examples we have seen so far), the coordinator is in a peripheral position with respect to the coordinand that it links. However, when the coordinands are long, and especially when they are clauses, the coordinator may occasionally stand in an internal position. For example, Dargi has the conjunctive suffix -ra which occurs bisyndetically in NP conjunction (A-ra B-ra). When this suffix is used to conjoin clauses, it occurs just once, following the first NP of the last conjunct clause: (16) Dargi (van den Berg°, ex. 15) Il nu-ni abit’=aq-un-ra idzala-ra Gaybik-ib. this I-erg remove=caus-aor-1 disease(abs)-and stop-aor(3) ‘I had it (viz. the tooth) removed and the pain stopped.’ In the distantly related language Chechen, the cl*tic =’a also occurs bisyndetically in NP conjunction (A=’a B=’a), but when it links finite clauses, it occurs twice as well (contrasting with Dargi). Its position is immediately before the finite verb, and it cl*ticizes to the word preceding it: (17) Chechen (Jeschull°, ex. 48) as sialxana wovdalalla ’a lieliira, hwuuna xala ’a I.erg yesterday foolishly and behave.wp you.dat difficult and xietiitira. let.seem.wp ‘Yesterday I behaved like a fool and offended you.’ When a conjoined clause has an intransitive verb and thus no argument that =’a could cl*ticize onto, a “copy” of the verb is produced for =’a to cl*ticize onto: (18) Chechen (Good 2003: 134; see also Jeschull°, ex. 51–52) Maalik viela=’a viilara vialxa=’a vilxara. Malik laugh=and laugh.wp cry=and cry.wp ‘Malik laughed and cried.’



Martin Haspelmath

A surprisingly similar construction is found in Hakha Lai, where zó] ‘also’ is used bisyndetically for emphatic conjunction (A zó] B zó] ‘both A and B’). In clausal emphatic conjunction, zó] follows the object NP (see 19a), and when the verb is intransitive, a verb copy is produced (see 19b). (19) Hakha Lai (Peterson & VanBik°, ex. 46–47) a. Làwthlawpaa=ni‘ ‘aàr zó] ] ‘a-tsook vok zò] ] ‘a-zuár. farmer=erg chicken also 3sg.subj-buy2 pig also 3sg.subj=sell2 ‘The farmer both bought a chicken and sold a pig.’ b. Làwthlawpaa ‘a-tluuk zó] ] ‘a-tluú, ‘a-thi‘ zo] ] farmer 3sg.subj-fall2 also 3sg.subj-fall1 3sg.subj-die2 also ‘a-thiì. 3sg.subj-die1 ‘The farmer both fell and died.’


Category-sensitivity of coordinating constructions

In English and other European languages, the coordinators ‘and’ and ‘or’ can link a diverse range of categories: noun phrases, verb phrases, clauses, adjective phrases, prepositional phrases, and others. The coordinator ‘but’ is mostly confined to clauses, but this seems to be for semantic reasons. But many languages have category-sensitive coordinating constructions (see also the discussion in Ohori° § 2.2). In particular, about half of the world’s languages show different conjunctive constructions for nominal and verbal/clausal conjunction (see Haspelmath to appear b). For example, in Upper Kuskokwim Athabaskan, noun phrases are conjoined by means of bisyndetic postpositive ‘ił (cf. 3), while clauses are conjoined by means of the particle ts’e‘ (cf. 20). (20) Upper Kuskokwim Athabaskan (Kibrik°, ex. 19) “hondenh ghwla‘ sidadza‘” yinezinh ts’e‘‘ hwts’its’ay’nełghwt “where unknown my.sister he.thought and ‘He wondered where his sister was and took off with a sled.’ Identity of nominal and verbal conjunction is found throughout Europe and southwestern Asia, most of Southeast Asia, and Mesoamerica, while differentiation (as in Upper Kuskokwim) is found throughout Africa, in eastern Asia, and many areas of the Pacific and North and South America (see the map in Haspelmath to appear b). In this volume, identity is represented by Iraqw, Chechen, Dargi (and other Daghestanian languages), Western Iranian, Sgaw Karen, and Riau Indonesian, while differentiation is represented by Koyraboro Senni (Heath°), Fongbe, Hausa, Lai, and Lavukaleve.

Coordinating constructions: An overview

In Haspelmath’s (to appear b) cross-linguistic survey, the notion “verbal conjunction” lumps together clauses and verb phrases, because in many cases they cannot be easily distinguished (for instance, in (20) it is quite unclear what criteria one would use to argue that we are dealing with clausal or verb-phrase conjunction). But in some languages they can be distinguished, and then we sometimes find that VPs are conjoined like NPs, not like clauses. This seems to be particularly common in the Oceanic languages (cf. Moyse-Faurie & Lynch° § 3.1). An example comes from Xârâcùù (a language of New Caledonia). (21) Xârâcùù (Moyse-Faurie & Lynch°, ex. 5, 19, 22) a. NP conjunction gu mê gè 2sg and 1sg ‘you and I’ b. VP conjunction Ru cha mê mara. 3du clear.bush and ‘They cleared the bush and worked in the fields.’ c. clausal conjunction È nä fädë nä è nä bare tèpe. 3sg impf walk and 3sg impf also talk ‘He speaks as he is walking.’ Some languages even show three different conjunction strategies for NPs, VPs and clauses (e.g. Somali, cf. Haspelmath to appear a). However, there do not seem to be any languages with the same strategy for NP and clause conjunction and a different strategy for VP conjunction. Thus, we can set up an implicational sequence “NP – VP – clause”, such that each conjunction strategy covers a contiguous segment. The four different language-particular distributions are shown in (22). (22)








Upper Kuskokwim Athabaskan









The implicational sequence can be enlarged by bringing adjective phrases (APs) into the picture. In some languages, these are conjoined like NPs (e.g. Mandarin Chinese, Ohori° ex. 12–14; Chechen, Jeschull° ex. 15), while in other languages they are conjoined like VPs (e.g. Japanese, Ohori° ex. 16–18). In Hausa, some adjectives are conjoined like NPs, while others are conjoined like VPs (Abdoulaye° § 3.1).



Martin Haspelmath

Thus, adjective phrases are intermediate between NPs and VPs, and we get the implicational sequence NP – AP – VP. (23) again shows various language-particular distributions: (23)




English, Sgaw Karen








Chinese, Chechen






Payne (1985: 5) proposes an even longer implicational sequence also involving adpositional phrases (NP – PP – AP – VP – clause), but does not provide much evidence for it. Some counterexamples have been noted, but they mostly concern languages in which one of the intermediate categories cannot be conjoined at all. Thus, Koyraboro Senni does not permit the conjunction of PPs (Heath° § 3), Lavukaleve does not permit the conjunction of APs (Terrill° § 6), and Tîrî does not permit the conjunction of VPs (Moyse-Faurie & Lynch° § 3.1.1, ex. 24–26).

4. Semantic distinctions in conjunction In English and other European languages, there is a single conjunctive coordinator ‘and’ whose use is independent of the meaning of the conjuncts or any semantic nuances of conjunction that might be conveyed. But many languages have different conjunctive constructions depending on semantic factors (see also Ohori° § 3.2). One factor is the animacy of the conjuncts. In Takia (an Oceanic language of Papua New Guinea), noun phrases with human referents are conjoined by means of the comitative postposition da (e.g. o] ]ai da [2sg 1sg com]) ‘you and I’, Meit Kabun da ‘Meit and Kabun’), whereas non-human NPs are conjoined by juxtaposition (e.g. mau dabel fud ‘taro, yam and banana’; Ross 2002: 228). Animacy is also relevant for the distinction between me and ma in Nêlêmwa (Bril° § 2.1–2), and for the distinction between men, o and ma in Nemi (Moyse-Faurie & Lynch°, ex. 12). Another factor is the distinction between proper names and common nouns. In Asmat (West Papua), the coordinator en˘erím ‘and’ is only used to link proper names (e.g. Pisím en˘erím Wasí ‘Pisím and Wasí’), whereas other NPs are conjoined by bisyndetic A-am B-am (e.g. onów am, ós am ‘thatch and wood’; Voorhoeve 1965: 171–2). This distinction is also relevant, for instance, in Tamabo (Oceanic; northern Vanuatu; Jauncey 2002: 614), and for eastern Polynesian languages such as Maori (Moyse-Faurie & Lynch°, ex. 15). In some languages, different conjunctive constructions are used when the conjuncts form a conceptual unit and when they are thought of as separate entities.

Coordinating constructions: An overview

This is how Jeschull° (§ 2.1.2–3) describes the difference between Chechen A-ii B-ii and A ’a B ’a (e.g. shish-ii stak-ii ‘a bottle and a glass’ (ex. 14); waerzha mazh ’a, q’eegash shi bwaerg ’a ‘a black beard and two shining eyes’ (ex. 38)). This distinction seems to be related to the distinction between tight and loose coordination that is made by Moyse-Faurie & Lynch° (especially § 2.1) for several Oceanic languages. Tight coordinators are used with items that can be thought of as couples or pairs which are closely associated in the real world. Loose coordinators are used with items which are less closely associated. An example is Lenakel tight m (as in n6mataag m nihin ‘wind and rain’) vs. loose m6ne (as in kuri m6ne pukas ‘a dog and a pig’). Wälchli (2003) uses the terms natural coordination vs. accidental coordination for the semantic distinction, and loose vs. tight coordination for the corresponding formal distinction. Wälchli observes that tight coordination often implies less explicit formal means, typically simple juxtaposition, and two juxtaposed conjuncts are often treated as coordinative compounds. A language showing such an asyndetic/syndetic contrast is Sgaw Karen (Lord & Craig° § 3): for instance, j6-mo j6-pa [my-mother my-father] ‘my parents’ requires no overt coordinator, but ‘my mother and my brother’ does (j6-mo df‘ j6-w7, *j6-mo j6-w7). In Dargi, juxtaposed combinations such as neš-dudeš [mother-father] ‘parents’, berAi-dugi [day-night] ‘day and night’, and šalbar-Aewa [trousers-shirt] ‘clothes’ are described as “compounds” (van den Berg° § 2.1.1). In Iraqw (Mous° § 2) and in Hausa (Abdoulaye° § 3.1), natural coordination is not formally different from accidental coordination, but Abdoulaye observes that conjunctions such as uwaa dà ùbaa [mother and father] are more integrated and would not be used in two different intonational phrases, for instance. Such pairs of naturally coordinated items often develop idiomatic meanings (cf. Sgaw Karen ha‘ df‘ f‘ lokw7 ‘go and play’ vs. ha‘ lokw7 [go play] ‘go for a stroll’ (Lord & Craig°, ex. 62–63); Hausa bàakii dà hancìi [mouth and nose] ‘very close’ (Abdoulaye° § 3.1)). Clause coordinators are often translated as ‘and’, ‘and then’, ‘then’ (e.g. Lavukaleve aka and hano, Terrill° § 7.1). It is difficult to judge to what extent this sequential meaning is part of the coordinators’ meaning and to what extent it simply derives from contexts in which sequences of events are reported. Lefebvre° (§ 1) notes that Fongbe b`f and bó are translated as ‘and then’ when they occur in perfective clauses, and as ‘and’ when they occur in imperfective clauses. She concludes that the sequential sense is a contextual, not an inherent, property of these coordinators. A semantic distinction that is sometimes conveyed by clause coordination is switch-reference, i.e. the distinction between same-subject and different-subject clause combinations. The term switch-reference is typically used for constructions in which one clause is dependent on another clause (see Stirling 1993: 6), and for some reason independent coordinating particles that signal a same-subject/differentsubject contrast have not been prominent in the literature. However, such a



Martin Haspelmath

contrast is described in detail by Lefebvre° for Fongbe bó (same-subject) and b`f (different-subject): (24) Fongbe (Lefebvre°, ex. 5–6) a. Ùn wá bó yì. 1sg arrive coord leave ‘I arrived and then left.’ b. K`fkú wá b`f Àsíbá yì. Koku arrive coord Asiba leave ‘Koku arrived and then Asiba left.’ Similarly, in Nêlêmwa the clausal coordinator xa requires subject identity in both clauses (cf. 25a), whereas me also allows different subjects (cf. 25b). (25) Nêlêmwa (Bril°, ex. 71–72) a. I oda Teâ Pwayili shi Teâ Ovaac xa (i) khabwe 3sg go.up Teâ Pwayili side Teâ Ovaac also (3sg) say ushi-n … ben-poss.3sg ‘Teâ Pwayili goes up to Teâ Ovaac and tells him…’ b. I oda Teâ Pwayili shi Teâ Ovaac me i khabwe … 3sg go.up Teâ Pwayili side Teâ Ovaac and 3sg say ushi-n a Teâ Ovaac… ben-poss.3sg agt Teâ Ovaac ‘Teâ Pwayili goes up to Teâ Ovaac and Teâ Ovaac tells him…’ Also, van Klinken (2000) reports that Tetun (Timor; Central Malayo-Polynesian) has a clause coordinator -odi that requires subject identity between the clauses.


From comitative to conjunctive marker

As was noted by Stassen (2000), many of the world’s languages use the same marker for expressing conjunctive (‘A and B’) and comitative (‘A with B’) relations (he calls these languages “with-languages”). Quite a few of the languages discussed in this volume are with-languages in Stassen’s sense: Iraqw, Fongbe, Haitian Creole, Hausa, Vafsi, Sgaw Karen, Riau Indonesian, Nêlêmwa and most other Oceanic languages, as well as Japanese (discussed by Ohori°) and Koyraboro Senni (discussed by Heath°). Examples from Fongbe and Riau Indonesian are given in (26)–(27). (26) Fongbe (Lefebvre°, ex. 53a, 49) (note: kpó2ó…kpó is a circumposition) a. Àsíbá yì àxì m`7 [kpó2 2ó K`fkú kpó]. Asiba go market in [with Koku with ‘Asiba went to the market with Koku.’

Coordinating constructions: An overview

b. Àsíbá [kpó2 2ó K`fkú kpó] yì àxì m`7. Asiba [with Koku with go market in ‘Asiba and Koku went to the market.’ (27) Riau Indonesian (Gil°, ex. 9, 3) a. Damsir beli celana sama si Man sudah bulu-bulu. Damsir buy trousers sama pers Mansudir pfct distr~feather ‘The trousers that Damsir bought with Mansudir are already all frayed.’ b. Doni sama Amat mau di-tumbuk dia. Doni sama Amat want pat-hit 3 ‘He wants to hit Doni and Amat.’ There are at least two different ways in which this formal identity can be understood. On the one hand, one can argue that the comitative/conjunctive markers in with-languages have just one single function, which happens to be rendered in two different ways in and-languages like English that must differentiate between ‘and’ and ‘with’.3 This is what Lefebvre and Gil claim for their languages. On the other hand, one could argue that the comitative marker and the conjunctive marker are different synchronically, both semantically and syntactically, and that the identity of their shape is due to a very common semantic-syntactic change from comitative marker to conjunctive coordinator. Of course, it is quite possible (and actually very likely) that some with-languages are of the former type, while others are of the latter type. Both Gil and Lefebvre argue that the single-function (or monosemy) description should be the default, to be replaced by a multiple-function (or polysemy/ ambiguity) description only if relevant differences in the object language are discovered (differences in the translations to other languages obviously do not qualify as arguments against a monosemic description). In the following, a list of types of semantic and syntactic differences is provided that could be used to argue that the comitative and conjunctive markers/constructions are different in a given language. (i) Semantics: Abdoulaye° (§ 4.2) describes the difference between and and with in English: A and B act suggests that both A and B are equally in control of the action, but not necessarily simultaneously or in the same place, whereas A acts with B entails that A and B are in the same place and their involvement is simultaneous, but it does not suggest that they are equally in control (A could be in full control,

3.Lefebvre actually translates (26b) (her ex. 49) as ‘Asiba with Koku went to the market’, in order to underline her claim that (26a–b) do not differ semantically, but the translation given here sounds much more natural in English.



Martin Haspelmath

with B as a co-actor, or vice versa). Thus, the following English sentences are odd for semantic reasons: (28) a. #Pedro watched the world cup final with Yumiko, but Pedro was in Cordoba, and Yumiko was in Kumamoto. b. #Clelia watched The Tin Drum with Niklas, but Clelia watched it in 1986, and Niklas in 2004. (29) #In the 2000 election campaign, Gore ran for U. S. president with Bush. Sentence (29) is fine only if (contrary to fact) Bush is Gore’s running mate (i.e. candidate for vice president) and thus has less control, not if Gore and Bush are both candidates for the presidency. (ii) Topicality: Abdoulaye° (§ 4.1) also notes that in Hausa, subject NPs are topics and must be referred to as pronouns in the following clause: (30) Hausa (Abdoulaye°, ex. 21) Abdù yaa tàfi maka˜rantaa dà Bàlki an bâa *Abdù/Bàlki Abdu 3sg.m.pfv go school with Balki imps.pfv give *Abdu/Balki àlloo. board ‘Abdu went to school with Balki, and a [writing] board was given to *Abdu/Balki.’ Here the full NP Abdù is not possible in the second clause, and it would have to be replaced by an anaphoric pronoun. By contrast, Bàlki is possible because it is not a topic in the first clause. When Abdù and Bàlki are conjoined with dà, they are both jointly the topic of the sentence, and each can be repeated individually in the next clause: (31) Hausa (Abdoulaye°, ex. 22) Abdù dà Bàlki sun tàfi maka˜rantaa an bâa Abdù/Bàlki Abdu and Balki 3pl.pfv go school imps.pfv give Abdu/Balki àlloo. board ‘Abdu and Balki went to school and a board was given to Abdu/Balki.’ (iii) Word order. In many SVO languages with relatively rigid word order, such as Hausa and English, the verb generally follows the subject immediately and any adpositional-phrase adjuncts come after the verb. This makes it fairly easy to distinguish comitative dà from conjunctive dà in Hausa sentences like (30)–(31). In SOV languages like Iraqw, however, this criterion does not help:

Coordinating constructions: An overview

(32) Iraqw (Mous°, ex. 11) Muu-dá’ nee dama-r-ín ta-ri waráahh. people-dem4 and calf-f-3pl.poss imps-nar pass:past ‘Those people and their calf passed.’ Or: ‘Those people passed with their calf.’ In addition to functioning as clausal adjuncts, comitative phrases may also occur adnominally in many languages, cf. (33): (33) Hausa (Abdoulaye°, § 3.3) Wata màcè dà jàariirìn-tà ta-nàa zàune wàje. one woman with baby-3sg.poss 3sg.f-cont sit outside ‘A woman with her baby is sitting outside.’ Thus, word order only differentiates conjuncts from clausal comitative phrases.4 (However, additional semantic criteria often easily exclude the adnominal comitative reading, e.g. with proper nouns as in (26b), (27b), and (31); proper nouns are extremely unlikely to have an adnominal modifier.) It is an interesting question which comitative construction is the diachronic source of the conjunctive construction in those with-languages that clearly have a separate conjunctive construction. Stassen (2000: 26) seems to presuppose that it is the clausal comitative construction (he appeals to “movement” of the comitative phrase from its canonical position in the sentence), but it may well be that the adnominal construction (as in 33) is the source construction in most languages. (iv) Bisyndetic conjunction. In several languages, comitative-derived conjunction markers are used bisyndetically, thus showing that they occur in a new construction. For instance, the comitative postposition ‘ił in Upper Kuskokwim Athabaskan occurs on both conjuncts when it is used as a coordinator (see ex. 3 above; Kibrik° §2.1–2). Hausa dà may also be used bisyndetically with an emphatic sense (dà Abdù dà Bàlki ‘both Abdu and Balki’, Abdoulaye° §3.1). In Hakha Lai, the comitative encl*tic =heè is only used bisyndetically in emphatic conjunction (e.g. ‘aàrpii=heé ‘a-faà=leé=heè [hen=com 3sg.poss-children=coll=com] ‘both the hen and her children’, Peterson & VanBik°, ex. 43); it is not used for non-emphatic conjunction. (v) Multiple conjuncts. When a marker occurs in a construction with multiple noun phrases, this shows that the construction does not involve clausal adjuncts, because there can be only one comitative clausal adjunct. A construction such as

4.A clausal comitative interpretation is also excluded when the expression is the complement of an adposition, e.g. Koyraboro Senni X nda Y še [X with/and Y for] ‘for X and Y’ (Heath° § 3). This kind of structure could conceivably be an adnominal comitative phrase (if the language allows prenominal PP modifiers), but it could not be a clausal comitative phrase.



Martin Haspelmath

“A co-B co-C co-D” could in principle represent nested adnominal comitative phrases (as in a father with a child with a doll with a red dress), but this occurs only in extremely specialized circ*mstances. Normally such a construction represents conjunction. Comitative-derived conjunction markers that are used with multiple conjuncts are found in Hausa, Iraqw (ex. 4 above), Upper Kuskokwim Athabaskan (ex. 5 above), and many Oceanic languages (e.g. West Uvean, Moyse-Faurie & Lynch°, ex. 74). (vi) Coordinator omission. Many languages allow non-final coordinators to be dropped when there are three or more coordinands. When these coordinators have the same shape as a comitative marker, the possibility of coordinator omission is a sure indication that we are dealing with a different construction. Coordinator omission for comitative-derived coordinators is reported, for instance, for Iraqw (Mous°, ex. 16) and Hausa (Abdoulaye° § 3.1). (vii) Use of independent pronouns. It is a general property of coordinating constructions that personal pronoun coordinands appear as independent pronouns, not as cl*tic or affixal pronouns. In those languages that use cl*tic/affixal pronouns with their comitative marker, the formally identical conjunctive marker typically requires the independent pronoun. This can be illustrated by Hausa and Upper Kuskokwim Athabaskan (see also Moyse-Faurie & Lynch°, note 16, on Polynesian languages). (34) Hausa (Abdoulaye°, ex. 12b, § 2.2) a. Naa ganee =shì dà àbikkiyà-˜r-shì. 1sg.pfv see =3sg.m.obj with friend-of-3sg.m ‘I saw him with his friend.’ b. Naa ga shii dà àbikkiyà-˜r-shì. 1sg.pfv see he and friend-of-3sg.m ‘I saw him and his friend.’ (35) Upper Kuskokwim Athabaskan (Kibrik°, ex. 8, 2) a. nut si-‘‘ił ton dalts’enh-na here me-with town they.stay-those.people ‘the people who live in town with me’ b. Timothy ‘ił se ‘ił kayih ts’ideghilts’e‘. Timothy with me with house we.stayed ‘Timothy and I stayed at home.’ (viii) Number agreement. Conjunctive constructions often contrast with comitative constructions in requiring nonsingular agreement on the verb, e.g. (36) East Uvean (Moyse-Faurie & Lynch°, ex. 77) Ne‘e momoe te kiu mo te fo‘i ‘uga i Vaitupu. past def egret and def clf hermit.crab obl Vaitupu ‘The egret and the hermit-crab slept at Vaitupu.’

Coordinating constructions: An overview

This criterion is also discussed for Iraqw by Mous° (§ 4), but he concludes that it is not decisive, because Iraqw also shows plural agreement when the meaning is clearly comitative. (See also the unexpected nonsingular agreement in Hausa as in Abdoulaye°’s example 30, and in Tolai as in Bril°’s example 99.) (ix) Use with non-NP categories. Many conjunction markers that are formally identical to comitative markers can conjoin non-NP categories such as adjective phrases and clauses, where the comitative meaning/construction cannot be involved. This is the case in Iraqw, Sgaw Karen, and many of the Oceanic languages, for instance. (x) Extraction and focusing. Clausal comitative modifiers can be extracted and focused, but individual conjuncts cannot in general be extracted and focused (see § 9 below). Thus, in the East Uvean sentence (37a), only te tama ‘the boy’ can be focused (as shown in 37b), not te tama mo koe ‘the boy with you’, because this does not form a constituent.5 (37) East Uvean (Moyse-Faurie & Lynch°, ex. (78a–b)) a. ‘E ‘alu te tama mo koe. ns def boy with you ‘The boy went off with you.’ b. Ko te tama ‘e ‘alu mo koe. pred def boy ns with you ‘It’s the boy who went with you.’ Extraction and focusing is also discussed for Hausa by Abdoulaye° (§ 4.1). Thus, there are a fair number of properties that can distinguish between comitative and conjunctive constructions even when the marker is the same in both constructions. It appears that when a comitative marker changes semantically to become a conjunction marker, there is strong pressure for it to adopt the formal properties that are associated with conjunctive constructions. What the source of this pressure might be is an intriguing question that I will not try to address here. Before leaving this section, we should briefly consider the question whether the change from comitative to conjunctive is unidirectional, or whether the reverse change is also possible and might account for some of the synchronic cases of comitative-conjunctive marking identity. An interesting case in point is the unusual construction described for Dargi by van den Berg° (§ 5, “conjunctive-comitative construction”). Dargi has bisyndetic postpositive conjunction (A-ra B-ra), and one way of expressing the comitative role is by conjoining the comitative NP with the reflexive pronoun say/sari/sabi:

5.It could be a constituent if the language allows adnominal comitative phrases, but this would be a pragmatically implausible expression.



Martin Haspelmath

(38) Dargi (van den Berg°, ex. 86a) ˆ Xunul-ra say-ra udzi šadi-w arq’-uli say. wife-and self.m-and brother(abs) walk-m leave-ger be.m ‘My brother went for a walk with his wife.’ Literally, this seems to mean ‘My brother went for a walk, his wife and himself ’, but it cannot be an apposition construction synchronically, because the part ‘his wife and himself ’ does not agree in case with the subject (this can be seen in sentences like van den Berg°’s example 87a, where the subject is in the ergative case). Thus, this is a special construction that comes close to being an example of a change from conjunctive construction to comitative construction.

6. Semantic maps Coordinators often have other meanings/functions besides the function of marking coordinating constructions. In the last section we discussed this for the two functions ‘comitative’ and ‘conjunctive’, and in this section we will take an even broader view. When one looks at the patterns of polyfunctionality (or macrofunctionality, to use Gil’s° term) across languages, one notices that there are many differences, but also recurrent patterns. Some examples are given in (39). (39) a. b. c. d. e. f. g. h. i.

Karen df‘ (Lord & Craig°): conjunctive, comitative, instrumental Iraqw nee (Mous°): conjunctive, comitative, instrumental, agent Hausa dà (Abdoulaye°): conjunctive, comitative, instrumental, existence Fongbe kpó2ó … kpó (Lefebvre°): comitative, instrumental, manner Upper Kuskokwim Athabaskan ‘ił (Kibrik°): conjunctive, comitative, ‘instrumental’, ‘also’ Dargi -ra (van den Berg°): conjunctive, ‘also’, ‘even’ Riau Indonesian sama (Gil°): conjunctive, comitative, instrumental English with: comitative, instrumental Russian -om/-oj/-ju (=instrumental case): instrumental, agent

Coordinating constructions: An overview

None of these particles or affixes has exactly the same range of functions, but some uniform pattern can be detected within the diversity: Certain combinations of meanings/functions of a polysemous/polyfunctional element do not occur. For instance, there are no particles that express ‘also’ and ‘comitative’, but not conjunctive, and no markers that express verbal conjunction and comitative, but not nominal conjunction. An elegant way of expressing these regularities is by drawing a semantic map that consists of meanings linked by connecting lines (see Croft 2001, Haspelmath 2003 for the concept of “semantic map”). A very tentative semantic map for conjunction and related meanings is given in Figure 1.


N -conjunction









Figure 1.A semantic map for conjunction and related notions

Semantic maps express universals of polysemy because they are associated with the connectivity hypothesis: Every language-particular element or category occupies a connected region on the semantic map. This is illustrated in Figures 2–10, which show the regions occupied by each of the elements in (39a–i).


N -conjunction









Figure 2.Karen df‘

None of these particles/markers is exactly like any of the others, but the striking similarities between them are expressed in a salient way by the semantic-map notation. Of course, Figure 1 and Figures 2–10 are only partial representations: Many of the particles/markers have additional functions not shown here, and the main map in Figure 1 could be enlarged by adding further items. For instance, one could easily add the implicational sequences that we saw in (22)–(23) above. It is only for the sake of expository convenience that the two sets of facts are described separately here.



Martin Haspelmath


N -conjunction









Figure 3.Iraqw nee


N -conjunction









Figure 4.Hausa dà


N -conjunction









Figure 5.Fongbe kpó2ó … kpó


N -conjunction









Figure 6.Upper Kuskokwim Athabaskan ‘ił

The only contribution to this volume that makes use of semantic maps is Gil°, who argues that the various uses of Riau Indonesian sama are not different meanings, but just different ways of translating the unitary meaning of sama (‘togetherness’) in different contexts. Thus, he argues for monosemy rather than polysemy of markers that seem to have a broad range of different uses from the point of view of another language. The semantic-map notation is neutral between these two ways of looking at polyfunctionality. It is compatible both with a polysemic view that regards the

Coordinating constructions: An overview


N -conjunction









Figure 7.Dargi -ra


N -conjunction









Figure 8.Riau Indonesian sama


N -conjunction









Figure 9.English with


N -conjunction









Figure 10.Russian instrumental case

nodes in the network of a semantic map as distinct meanings, and with a monosemic view claiming that each of the connected regions in Figures 2–10 has just a single, unified meaning. The challenge for the polysemic approach is to show that speakers really make all the distinctions that the lingusts make, because, as both Gil° (§ 2) and Lefebvre° (§ 1) point out, the default hypothesis in the description of an individual language should be that of monosemy. The challenge for the monosemic approach is to show how the exact boundaries of each element can be derived from



Martin Haspelmath

the single unified meaning. Single unified meanings (or “Gesamtbedeutungen”) are generally very abstract (like Gil°’s ‘togetherness’), and it is often difficult to see how one gets from the abstract meaning to the various concrete uses. Semantic maps express universals of polyfunctionality, but like many other universals, they may have exceptions. A possible exception to the map in Figure 1 comes from Welsh (cf. Stolz 1998), where the conjunction marker is a [a] (ac [ag] before vowels), and the instrumental preposition is â [a] (ag [ag] before vowels). These two are different orthographically, but not in pronunciation, and they are etymologically identical as well. Given the semantic map in Figure 1, we expect that a(c)/â(g) also occurs in the comitative function, but the modern comitative preposition is gyda(g) in most cases. As Stolz (1998) shows, older Welsh had a single preposed element for conjunctive, comitative and instrumental functions, but at a later stage this element was replaced by gyda(g) (originally yn nghyd â ‘in unison with’), so that now we have a situation where conjunction and instrumental are expressed by the same marker (pronounced [a(g)]), while comitative is expressed by a different marker. This constitutes a synchronic exception to the connectivity claim of Figure 1. However, this Welsh case still shows connectivity at the diachronic level: The marker [a(g)] presumably first expressed comitative and was then extended to instrumental and finally to conjunctive function. It was later replaced by a renewed comitative marker, but in its original development it did not jump over the comitative function. Semantic maps can also be looked at as sets of diachronic pathways, and the synchronic polyfunctionality of a marker is then simply the result of its diachronic extension along the permitted paths. Some of these pathways are unidirectional, i.e. changes are possible only in one direction. Such unidirectional pathways can be indicated by an arrow connecting two functions on a semantic map. A first hypothesis of the diachronic version of the map in Figure 1 is shown in Figure 11.


N -conjunction 'also'




instrumental agent

Figure 11.Diachronic links between conjunction and related functions

Coordinating constructions: An overview


Inclusory constructions

Many languages have constructions that are notionally like conjunction and are rendered by ‘and’-conjunction in English, but that are crucially different from ordinary conjunction in that one of the constituents has the same reference as the entire construction. This is best explained with some concrete examples: (40) a.

Koyraboro Senni (some dialects; Heath°, § 10) ir nda ni we with ‘you and I’ b. Hausa (Abdoulaye°, ex. 2b) muu dà shii we with he ‘he and me’ (or: ‘he and us’) c. Nêlêmwa (Bril°, ex. 15) hla ma Kaavo 3pl with Kaavo ‘they (two) and Kaavo’ d. Lavukaleve (Terrill°, ex. 26) el Mima 1du.excl Mima ‘Mima and I’ e. Mparntwe Arrernte (Ohori°, ex. 31. from David Wilkins) kake ilerne 1du ‘elder brother and I’ f. Old English (Bhat°, ex. 6, from Edgerton 1910: 112) wit Scilling 1du Scilling ‘Scilling and I’

In the first example ir nda ni ‘you(sg) and I’, literally ‘we and you(sg)’, the reference of the first constituent (ir ‘we’) includes the reference of the second constituent (ni ‘you(sg)’). This construction type is called inclusory construction in this volume (following Lichtenberk 2000). The first constituent is almost always a nonsingular pronoun, typically first or second person (called the inclusory pronoun), and the second constituent (called the included NP) can be juxtaposed (as in ex. (40d–f)), or it can be linked by a particle, often a particle that also means ‘and’, ‘with’ (as in (40a–c)). The inclusory pronoun most often precedes the included NP, but it can also follow it (as in (40e)).



Martin Haspelmath

In those languages where the linking particle is identical to the ‘and’ coordinator, these constructions may be ambiguous. Thus, in Hausa, muu dà shii (lit. ‘we and he’) can have both an inclusory meaning (‘he and me’) and an additive meaning (‘he and us’). And since Hausa does not make a distinction between dual and plural, there are actually two translations of the inclusory reading: if muu ‘we’ refers to two people, muu dà shii is translated as ‘he and me’, but if it refers to more than two, it is translated as ‘he and us’. In this latter case, the inclusory and the additive meanings are not differentiated in the English translation. In a language with a dual-plural contrast such as Nêlêmwa (cf. Bril°, ex. 14), it is clear that an inclusory construction like yaman ma axaleny [1du.excl and] must be translated as ‘me and this man’, not ‘us and this man’. In the examples in (40), the inclusory pronoun and the included NP occur as a contiguous constituent and form a phrasal inclusory construction. In these cases, the inclusory pronoun is always an independent pronoun. But the inclusory pronoun may also be a bound pronoun, and then the included NP does not form a constituent with it. This construction is called split inclusory construction. Two examples are given in (41). (41) a.

Hausa (Abdoulaye°, ex. 2a) Mun jee kàasuwaa dà Abdù. 1pl.pfv go market with Abdu ‘Abdu and I went to the market.’ b. Polish (Bhat°, ex. 4, from Schwartz 1988: 52) Posz-li-s´my z matkø do kina. go-past-1pl with mother to movies ‘Mother and I went to the movies.’

In these examples, the bound pronoun is a verbal argument. Split inclusory constructions may also occur with bound possessive pronouns on nouns, as in (42). (42) Nêlêmwa (Bril°, ex. 17) mwa-wa ma kââma-m ma axomoo-m house-poss.2pl and father-poss.2sg and mother-poss.2sg ‘your(sg) house and your father’s and your mother’s’ (It appears that the included NP in this example is itself a coordinated phrase; i.e. the first ma serves as inclusory particle, and the second ma serves as ordinary additive coordinator.) Inclusory constructions are discussed from a general perspective by Bhat°, and detailed discussion of inclusory constructions in particular languages is found in Abdoulaye°’s and Bril°’s contributions.

Coordinating constructions: An overview

8. Disjunction Disjunctive (‘or’) coordination is much less prominent in this volume than conjunctive coordination, and this is not surprising because it is also less prominent in language use. Ohori° (§ 3.2) observes that ‘and’ words are much more frequent in discourse than ‘or’ words. Thus, we also expect that ‘or’ words are typically longer (and rarely shorter) than ‘and’ words. A few examples illustrating this trend are given in (43). (43) German Russian Hausa Iraqw Persian Lavukaleve Dargi

‘and’ und i dà nee =ò o …-ra …-ra

‘or’ oder ili koo laqáa ya… ya… ve ya(-ra)… ya(-ra)…

Ohori° (§3.2) also notes that disjunctive coordinators tend to be more autonomous (i.e. they are more often free words rather than cl*tics or affixes), and that disjunction less often differentiates between NP coordinands and clause coordinands. All this can probably be explained as due to the lower frequency of disjunction markers. In some languages, there does not seem to be any grammaticalized way of expressing disjunction at all. Ohori° (§ 3.1) describes cases of neutralization between ‘and’ and ‘or’. According to Kibrik° (§ 4), who describes coordination in Upper Kuskokwim Athabaskan, “there does not seem to exist any native way to express disjunction” in this language, only the English borrowing o is sometimes used. In general, disjunction markers are fairly easily borrowed (see Matras 1998 for the borrowing hierarchy ‘but’ > ‘or’ > ‘and’); another example is Dargi ya… ya…, borrowed ultimately from Persian. In Lai, there is a native conventional way to express disjunction, but this is rather complicated and seems to be quite young. An example is (44). (44) Hakha Lai (Peterson & VanBik°, ex. 12a) làwthlawpaa=ni‘ vok ‘a-sií-làw=leè ‘aàr ‘a-tsook farmer=erg pig 3sg.subj-be2-neg=cond chicken 3sg.subj-buy2 ‘The farmer bought a pig or a chicken.’ Originally, the word for ‘or’ (‘asiílàwleè) was analyzable as ‘a-sií-làw=leè [3sg.subj-be2-neg=cond] and the literal translation of (44) is ‘The farmer, if it wasn’t a pig, bought a chicken.’ This is not a possible synchronic analysis because =leè is not used as a conditional marker anymore, but it is clear that the disjunctive marking strategy must be of very recent origin.



Martin Haspelmath

Like conjunction markers, disjunction markers are often polyfunctional, and it should be possible to draw a semantic map of the various uses that ‘or’ words can have (see Haspelmath 1997: 164–69 for some relevant observations). The papers in this volume give too little information on disjunctive markers to start drawing a semantic map, but some other meanings of ‘or’ words are noted, e.g. doubt or possibility in Iraqw (Mous°, ex. 13–14), and question tag in Nêlêmwa (Bril° § 1.2).

9. Extraction from coordinating constructions Since Ross 1967 [1986], constraints on extraction from coordinate structures have been discussed in the literature. Ross (1967: 89) formulated the following constraint (cited here after Schachter 1977: 94; see also Kazenin & Testelets° § 2.2). (45) The Coordinate Structure Constraint In a coordinate structure, no conjunct may be moved, nor may any element contained in a conjunct be moved out of that conjunct. The first part of this constraint explains the ungrammaticality in English of sentences like (46a–b). In (46a), the second conjunct has been fronted, and in (46b), the first conjunct has been fronted. (In this and the following examples, the extraction site is signaled by _.) (46) a. *What sofa will he put the chair between some table and _? (Schachter 1977: 94) b. *What records did you buy _ and books on civil engineering? Since English allows (and often prefers) preposition stranding, one might expect (46a) with “coordinator stranding” to be possible, but it is not. Coordinator piedpiping (i.e. movement of the coordinator along with the questioned phrase) is not possible either (*And what sofa will he put the chair between some table _?). The Coordinate Structure Constraint is sometimes used as a test for coordinatehood where there might be some doubt over the coordinate status of a construction. Thus, Abdoulaye° (§ 4.1) invokes it to argue that in Hausa, the element dà occurs in a coordinate construction when it is translated as ‘and’, and in a dependent prepositional phrase when it is translated as ‘with’. (47a) shows dà in its ‘with’ meaning, and (48a) shows it in its ‘and’ meaning. The (b) sentences show fronting of the dà-phrase, and the ungrammaticality of (48b) is due to the Coordinate Structure Constraint, showing that the two different dàs behave differently syntactically. (47) a.

Yaa zoo nannìyà (tàare) dà yâara-n-shì. 3sg.m.pfv come here (together with children-of-3sg.m ‘He came here with his children.’

Coordinating constructions: An overview

b. (Tàare) dà yâara-n-shì fa, yaa zoo nannìyà _. (together with children-of-3sg.m indeed 3sg.m.pfv come here ‘With his children (indeed), he came here.’ (48) a.

Abdù yaa kashè kàree dà mussàa. Abdu 3sg.m.pfv kill dog and cat ‘Abdu killed the dog and the cat.’ b. *Dà mussàa kàm, Abdù yaa kashè kàree _. and cat indeed Abdu 3sg.m.pfv kill dog ‘And the cat, Abdu killed the dog _.‘

The second part of the Coordinate Structure Constraint is responsible for the impossibility of extraction from clausal conjuncts, as illustrated in (49). (49) a. *What did Maria sell _ and Robert got angry? b. *Who did Robert sleep too long and Maria phoned _? The Coordinate Structure Constraint as formulated in (45) only bans extraction out of coordinate structures, but questions (and other focusing constructions) targeting an element of a coordination are often disallowed even if the focused phrase is not displaced. Thus, in Tsakhur, sentence (50) is impossible, even though the question word occurs in situ: (50) Tsakhur (Kazenin & Testelets°, ex. 24b) *Rasul-e¯ hiˇŠo¯n alja‘-u, mašin ališ¯-u? Rasul-erg what(4cl) build-coord(4cl) car(4cl) buy-pfv(4cl) (lit.) ‘Rasul built what and bought the car?’ Again, these phenomena have often been cited in arguing for the coordinate status of clause combining constructions. The ungrammaticality of (49a) can be taken as evidence that the second constituent is not a subordinate clause, because extraction from a main clause that is followed by a subordinate clause is possible (cf. What did Maria sell _ so that Robert got angry?). However, (49a) says nothing about the status of the first constituent, because in general it is not possible to extract from a finite subordinate clause either (cf. *What did Robert get angry because Maria sold _?). Sentence (49b) is even less useful in distinguishing between subordination and coordination, because extraction would be impossible anyway even if one or the other constituent were a subordinate clause (Since Robert slept too long, Maria phoned X; Robert slept too long so that Maria phoned X; in both these sentences, X cannot be replaced by who and fronted). Thus, the impossibility of extraction from a clause in a complex sentence is not an automatic argument for coordinate status — first it has to be demonstrated that extraction from a corresponding subordinate structure would be possible. Baker (1996: 459–60) claims that (51a) is a coordinate structure, despite its English translation.



Martin Haspelmath

(51) Mohawk (Iroquoian; Baker 1996: 460) a. S-hon-aht´%ty-u s-hon-ather-unyá-hn-u. ‘They have gone home to make baskets.’ b. *Nahót% s-hon-aht´%ty-u s-a-hun-unyá-hn-e’? what ‘What have they gone home to make?’ Baker’s main argument for the coordinate status of (51a) is the impossibility of (51b), which contrasts with the grammaticality of the English translation (involving a nonfinite subordinate structure). But this argument is convincing only if we know that extraction from corresponding subordinate structures is possible in Mohawk, as it is in English (note that even a closely related language like German does not allow such extraction: *Was sind sie nach Hause gegangen, um zu machen? ‘What have they gone home to make?’). Conversely, when extraction is possible in a construction whose English counterpart is coordinate and disallows extraction, this is usually taken as proof that we are not dealing with coordination, but with subordination. Thus, in Lai extraction is possible from the second constituent clause, as shown in (52) (which is similar to 49b): (52) Hakha Lai (Peterson & VanBik°, ex. 28) Zày=da‘ làwthlawpaa=ni‘ tihaà] ‘a-dín-di‘-naa‘‘in what=interr farmer=erg soup 3sg.subj-drink2-comp-concess sayaàpaa=ni‘ ‘a-‘ay-zhiamzhiam? teacher=erg 3sg.subj-eat2-still ‘What did the farmer drink all the soup but the teacher still ate _?’/ ‘What, although the farmer drank all the soup, did the teacher still eat?’ Peterson & VanBik take this as showing that the Lai construction is not coordinate, but shows subordination of the first constituent (cf. the second translation). However, it is not clear whether the Coordinate Structure Constraint is a true universal, because extraction from coordination has not been studied systematically for very many languages. Perhaps there are languages with constructions that look like coordinations by all other criteria, but that still allow extraction freely. A widespread view is that the Coordinate Structure Constraint can be subsumed under the general requirement that coordinated categories should be of the same syntactic type (Schachter 1977, Gazdar 1981). However, this view of the Coordinate Structure Constraint as a syntactic constraint has recently been challenged by Culicover & Jackendoff 1997 and Yuasa & Sadock 2002. These authors argue that the Coordinate Structure Constraint is a semantic constraint. Some observations made in this volume (Peterson & VanBik°, § 2.5.4, and Kazenin & Testelets°, § 4.1.2, § 4.2) seem to confirm this view because they show that the acceptability of extraction sometimes depends on the interpretation of the complex sentence.

Coordinating constructions: An overview

10. Ellipsis in coordination Whenever two elements are coordinated that are smaller than complete clauses, one can suspect that we are dealing with special constructions allowing ellipsis (or “deletion”, or “reduction”) in coordinate structures. Consider the examples in (53), where possible ellipsis sites are indicated by “Ø”. (53) a. (English) b. (German) c.


Robert is at home but Ø has no time. Maria möchte einen roten Ballon und einen weißen Ø. ‘Maria wants a red balloon and a white one.’ rìkìrkitaccen dookìi dà Ø jàakii horse and donkey ‘confused horse and confused donkey’ (Abdoulaye°, ex. 16a)

However, in many cases of apparent ellipsis, a description in different terms is clearly preferable. Thus, for (53a) one would say that we are not dealing with two coordinated clauses (one of which has an ellipted subject), but with two coordinate verb phrases: [NP Robert][VP[VP1 is at home] but [VP2 has no time]]. Terrill° discusses a similar case in Lavukaleve (her ex. 16): (54) solo-al o kofitaol vo-na mountain-pl and 3pl-in ‘over mountains and valleys’ Here the structure could conceivably be [PP [PP soloal Ø] o [PP kofitaol vona]], with an ellipted postposition in the first PP, but the evidence of number agreement shows that it must be [PP [NP [NP1 soloal] o [NP2 kofitaol]] vona]: When both conjuncts are singular, the postposition still shows plural agreement, so it must combine with a conjoined NP (e.g. ovulita o ki’kile vo-na [his.shield and axe 3pl-in] ‘on his shield and axe’). Another alternative to coordinate ellipsis is simple discourse ellipsis (or “anaphoric ellipsis”). Many languages allow discourse ellipsis more freely than the European languages, and this makes it more difficult to find true cases of coordinate ellipsis. As (53b) shows, there are even cases where German is different from English in allowing discourse ellipsis: Headless NPs with adjectives may be used in balloon anaphoric contexts (e.g. Welchen Ballon möchtest du? Den roten. ‘Which balloon would you like? The red one’). Thus, the ellipsis construction in (53b) has nothing to do with coordination.6

6.Abdoulaye° cites a similar example from Hausa (his ex. 17), which should probably be analyzed like the German example as well (cf. Newman 2000: 32, § 3.2.4).



Martin Haspelmath

But there are of course genuine examples of ellipsis that occur only in coordinating constructions. An example of this kind is (53c) from Hausa. This cannot be an example of coordination of a smaller constituent (say, [NP rìkìrkitaccenA [N dookìiN dà jàakiiN]]) because the adjective rìkìrkitaccen shows singular agreement, and a conjoined agreement controller [dookìi dà jàakii] would require plural agreement. It cannot be simple anaphoric ellipsis either because there is no construction of anaphoric ellipsis of adjectives. Thus, it must be an example of forward coordinate ellipsis (or analipsis). An example of backward coordinate ellipsis (or catalipsis) in NPs is shown in (55). (55) German Eine Stimme für Ø und drei Stimmen gegen den Kanzler waren ungültig. ‘One vote for Ø and three votes against the chancellor were invalid.’ This cannot be a coordination of a smaller constituent because eine Stimme für is not a constituent, and it cannot be cataphoric ellipsis because prepositional complements cannot be anaphorically omitted (let alone cataphorically). Most of the discussion of coordinate ellipsis in the literature has concerned constituents of clauses rather than constituents of noun phrases. In clausal coordination, it seems that we most often find analipsis of a constituent in the second coordinand. If the constituent is in a clause-medial position (thus leaving a gap), this type of analipsis is called gapping, as in (56b). (56) a.

Basque (Saltarelli 1988: 90) Gu-re herri-ko mutil-ak trakets-ak d-i-ra we-gen village-rel boy-pl.abs clumsy-pl.abs 3abs-pres-be neska-k ordea iaio-ak Ø. girl-pl.abs however agile-pl.abs ‘They boys in our village are clumsy, the girls, however, Ø agile.’ b. Gulf Arabic (Holes 1990: 68) ’ali ’a»ta ‘aAmad galam u Ø Maryam kitaab. Ali gave Ahmad pen and Maryam book ‘Ali gave Ahmad a pen and Ø Maryam a book.’

Catalipsis is largely restricted to the ellipsis of a constituent on the right periphery of the first constituent. In verb-final structures, the verb may be omitted in this way: (57) German (subordinate clause) …dass Robert Saft Ø und Maria Bier trinkt. ‘…that Robert drinks juice and Maria beer.’ A similar construction is cited from Lai by Peterson & VanBik° (they also report an alternative construction analogous to 56a).

Coordinating constructions: An overview

(58) Hakha Lai (Peterson & VanBik°, ex. 68b) Làwthlawpaa Falaám sayaàpaa Tidím ‘àn-kal. farmer Falam teacher Tedim 3pl.subj-go1 ‘The teacher went to Tedim and the farmer to Falam.’ What is unexpected here is the plural agreement on the verb: Ordinarily we expect plural subject agreement only if the subject is a conjoined NP. However, here only an analysis in terms of catalipsis seems to be possible (i.e. [S Làwthlawpaa Falaám Ø][S sayaàpaa Tidím ‘àn-kal]), because the subject and the locational phrase are not a constituent (thus ruling out the analysis [S [X Làwthlawpaa Falaám] [X sayaàpaa Tidím] ‘àn-kal]). Much too little is currently known about coordinate ellipsis in the world’s languages. The phenomenon has been studied extensively for the major European languages (see Schwabe & Zhang 2000 for a recent collection of articles), but for non-European languages we know very little about it. Unfortunately, ellipsis is not widely discussed in the contributions to this volume either (but see Peterson & VanBik° § 4, Abdoulaye° § 3.2, Terrill° § 4.3).

11. Coordination and subordination/dependency In this section, we return to the question of how coordination should be defined in such a way that the notion can be applied cross-linguistically. In particular, it is often not immediately obvious whether a construction consisting of two constituents involves coordination or subordination/dependency.7 One possibility is to define coordination and dependency in a purely formal sense. Thus, one could say that a coordinating construction is one in which all of the constituents are of the same syntactic category and this is also the category of the whole construction (as shown in 59a). By contrast, a dependency construction would be one in which the category of the whole construction is determined only by one of the constituents (the head), while the other constituents (the dependents) play no role in this respect (as shown in 59b).8 (The figures in (59a–b) were taken from Yuasa & Sadock 2002: 90.)

7.The term subordination is generally restricted to clauses in the current literature (cf., e.g., Cristofaro 2003, who says nothing about other constituent types), while the term dependency is used more widely. 8.The structure in (59a) is close to the one traditionally assumed in generative grammar, but more recently, a rather different structure (looking much more similar to dependency) has been widely adopted. See, e.g., Johannessen (1998) and Progovac (2003) for discussion, and Borsley (1994) for a skeptical voice.



Martin Haspelmath


(59) a. X1 X2


b. ... Xn





This definition of coordination is adopted by Gil (1991) and Gil° (§ 5.3), who claims that in Riau Indonesian, a simple sentence such as makan ayam [eat chicken] ‘chicken is eating, etc.’ is an instance of coordination. (This is so because in Gil’s analysis of Riau Indonesian, words like makan ‘eat’ and ayam ‘chicken’ belong to the same syntactic category (“S”) as sentences; see Gil 2000.) Appositive constructions like Ms. Bannerjee, the teacher (or the teacher, Ms. Bannerjee) would presumably also fall under this definition of coordination. However, it is not immediately clear that constructions like Robert and Maria would be classified as coordination by this definition, because they are generally thought to have the structure [[Robert] [and Maria]], i.e. they are not perfectly symmetrical (see the discussion above in § 2). For these reasons, it seems that a semantic definition better captures the actual current usage among linguists. The following definition is given in Haspelmath (to appear a): The term coordination refers to syntactic constructions in which two or more units of the same type are combined into a larger unit and still have the same semantic relations with other surrounding elements.

This definition entails that constructions with a comitative-derived coordination marker like Russian Saša s Mašej ‘Sasha and Masha’ are also considered as instances of coordination (see the discussion in § 5 above), although they are syntactically asymmetrical in that only the first conjunct shows the case assigned to the entire NP, while the second conjunct shows a case assigned by the (preposition-derived) coordinator s ‘with; and’. Similarly, complex sentences with two clausal constituents like (60) can be considered coordinations: (60) Japanese (Yuasa & Sadock 2002: 92) Ojii.san-ga yama-de hatarai-te, obaasan-ga mise-no mountain-at work-coord old.woman-nom store-gen ban-o shi-ta. sitting-acc do-past ‘The old man worked at the mountain, and the old woman tended the store.’ Yuasa & Sadock (2002) mention five criteria that show that we are dealing with coordination here, and not with a sentence consisting of a subordinate and a main clause:

Coordinating constructions: An overview


Reversibility: Changing the order of the conjuncts does not affect the truth conditions. ii. Application of the Coordinate Structure Constraint: The constituents of one clause cannot be questioned separately (e.g. ‘The old man worked/working at the mountain, (and) who tended the store?’; such sentences are impossible). iii. No backward anaphora: A pronoun in the first clause cannot corefer with a full NP in the second clause (e.g. ‘Hisi wife worked/working at the mountain, (and) the old mani tended the store’; such sentences are impossible). iv. Multiple conjuncts are possible. v. All the conjuncts are equally asserted. However, the Japanese -te-construction differs from the English and-construction in that the tense information on the verb is omitted from the first verb. In this respect, the construction is not symmetrical. Yuasa & Sadock (2002) propose that this state of affairs can be understood as a mismatch between the semantic and the syntactic properties of the construction. While (60) is a coordinate structure semantically, it is a subordinate structure syntactically. They make a similar argument with respect to the Russian comitative conjunction construction Saša s Mašej (or rather its very similar analog in Yiddish). At first glance, it appears that this proposal is a perfect compromise between the two positions that we contrasted earlier in this section: Rather than asking ourselves whether coordination should be defined formally/syntactically (as in Gil 1991, Gil°) or semantically (as in Haspelmath, to appear a), we could say that coordination can be defined at both levels, and that some constructions show coordination both semantically and syntactically, while other constructions are semantically coordinate and syntactically subordinate (“pseudo-subordinate”).9 Unfortunately, the properties of constructions do not always line up so nicely. First of all, Yuasa & Sadock have to make the surprising claim that not only constraints on anaphora (cf. (iii) above), but also the Coordinate Structure Constraint is sensitive to the semantic structure rather than the syntactic structure. Second, they have to claim that agreement is semantically determined as well, because comitative-conjunctive constructions (which they claim are syntactically subordinate) typically show number and gender agreement with both conjuncts rather than just the purported syntactic head.10

9.Constructions of the opposite type, showing semantic subordination and syntactic coordination, are also possible according to Yuasa & Sadock. They cite the example described by Culicover & Jackendoff (1997) as a case in point. 10.I give Polish examples rather than Russian examples here because Polish has gender agreement also in the plural.



Martin Haspelmath

(61) Polish (Tomasz Bak, p.c.) Jelena z Aniø by-ł-y cały dzien´ na wycieczce. Jelena with Ania all day on excursion ‘Jelena and Ania were all day on an excursion.’ Third, if restrictions on anaphora are sensitive to the semantic structure rather than the syntactic structure, one would expect that Polish i-conjunction (‘and’) and z-conjunction (‘with’) should behave alike. However, they do not: i-conjunction requires a non-reflexive possessive pronoun in the second conjunct, while z-conjunction requires a reflexive pronoun, like dependency constructions: (62) a.

Ania i jej nauczycielka były na wycieczce. Ania and her teacher were on excursion ‘Ania and her teacher were on an excursion.’ b. Ania ze swojø nauczycielka były na wycieczce. Ania with her.refl teacher were on excursion ‘id.’

Fourth, it is not so clear what exactly it means for a constituent to be categorydetermining — recall that this is the main criterion for establishing head status, and hence dependency structures. According to Yuasa & Sadock, in the Japanese example (60), we have syntactic subordination because only the second clause contributes the tense information of the sentence, i.e. the second clause is the syntactic head, and the first clause is the syntactic dependent. However, the omission of identical elements in one of the conjuncts is of course very common. For example, we have the following gapping construction in German, where the auxiliary verb is omitted in the second clause (the literal English translation seems to be ungrammatical): (63) Die Frau hatte gearbeitet und der Mann Ø den Kindern das Essen gegeben. ‘The woman had worked and the man (had) given the children the food.’ As in the Japanese example, the conjunct that contains the coordination marker lacks the tense element. Presumably, Yuasa & Sadock would not want to say that this sentence is syntactically subordinate, or else they would have to regard all cases of coordinate ellipsis as subordinate, contrary to almost everything that has been said about these constructions in the past. Fifth, Yuasa & Sadock’s discussion implies that differences in case-marking automatically lead to an analysis in terms of syntactic subordination/dependency. However, cases like the following are actually very common:

Coordinating constructions: An overview

(64) Mari (Uralic; Wälchli 2001: 46) Tide acˇa-m da ava-m-lan pölek. this father-1sg.poss and mother-1sg.poss-dat present ‘This is a present for my father and my mother.’ Wälchli (2001) (see also Wälchli 2003: § 2.4) discusses the question whether there are reasons to regard these constructions, where only one of the conjuncts has the marking that applies to the whole coordinate construction, as “asymmetrical” (i.e. “syntactically subordinate” in Yuasa & Sadock’s terms). He concludes that this is not the case if one allows the possibility of affixes combining with phrases, so that the dative suffix -lan can be said to have scope over the entire coordinate phrase acˇam da avam. Similarly, one could also say that the Japanese tense marker -ta in (60) has scope over the entire coordinate phrase [X-te Y], and the construction would then look formally symmetrical, i.e. coordinate syntactically as well as semantically. Thus, it appears that the simple idea of semantics/syntax mismatches in coordination will not solve all the problems. It remains difficult to operationalize the basic undisputed intuition that coordination involves symmetry, while subordination involves asymmetry. There are many constructions showing mixtures of both, and we are only at the beginning of understanding what constraints there might be on such mixtures. It is hoped that the data and discussions of this volume will ultimately contribute to a deeper understanding of this and other puzzles about coordination.

Abbreviations act action particle agt agent cl class (= gender) clf classifier coll collective com comitative concess concessive cont continuous coord coordinator cop copula fact factual ger gerund imps impersonal

inter iter nar nonfin O pat pers punc rec sbd stat surp wp

interrogative iterative narrative past nonfinite object/patient-like argument patient personal article punctual reciprocal subordinate stative surprise witnessed past



Martin Haspelmath

References Baker, Mark C. 1996. The polysynthesis parameter. New York & Oxford: Oxford University Press. Borsley, Robert. 1994. “In defense of coordinate structures.” Linguistic Analysis 24: 218–46. Cristofaro, Sonia. 2003. Subordination. Oxford: Oxford University Press. Croft, William. 2001. Radical Construction Grammar. Oxford: Oxford University Press. Culicover, Peter W. & Ray Jackendoff. 1997. “Semantic subordination despite syntactic coordination.” Linguistic Inquiry 28: 195–217. Edgerton, Franklin. 1910. “Origin and development of the elliptic dual and of dvandva compounds.” Zeitschrift für vergleichende Sprachforschung auf dem Gebiete des Deutschen, Griechischen und Lateinischen 43: 110–20. Gazdar, Gerald. 1981. “Unbounded dependencies and coordinate structure.” Linguistic Inquiry 12: 509–29. Gil, David. 1991. “Aristotle goes to Arizona and finds a language without ‘and’.” In: Dietmar Zaefferer (ed.), Semantic universals and universal semantics. Berlin: Foris, 96–130. Gil, David. 2000. “Syntactic categories, cross-linguistic variation and universal grammar”. In: Petra M. Vogel & Bernard Comrie (eds.), Approaches to the typology of word classes (Empirical Approaches to Language Typology, 23). Berlin: Mouton de Gruyter, 173–216. Good, Jeff. 2003. “Clause combining in Chechen.” Studies in Language 27.1: 113–70. Haspelmath, Martin. 1997. Indefinite pronouns (Oxford Studies in Typology and Linguistic Theory). Oxford: Clarendon Press. Haspelmath, Martin. 2003. “The geometry of grammatical meaning: Semantic maps and crosslinguistic comparison.” In: Michael Tomasello (ed.), The new psychology of language, vol. 2. Mahwah, NJ: Lawrence Erlbaum, 211–42. Haspelmath, Martin. To appear a. “Coordination.” In: Timothy Shopen (ed.), Language typology and syntactic description. 2nd ed. Cambridge: Cambridge University Press. Haspelmath, Martin. To appear b. “Nominal and verbal conjunction.” In: Matthew S. Dryer, Martin Haspelmath, David Gil & Bernard Comrie (eds.), World atlas of language structures. Oxford: Oxford University Press. Holes, Clive. 1990. Gulf Arabic. London: Routledge. Jauncey, Dorothy. 2002. “Tamabo.” In: John Lynch, Malcolm Ross & Terry Crowley (eds.), The Oceanic languages. Richmond: Curzon, 608–25. Johannessen, Janne Bondi. 1998. Coordination. New York: Oxford University Press. Lichtenberk, Frantisek. 2000. “Inclusory pronominals.” Oceanic Linguistics 39: 1–32. Matras, Yaron. 1998. “Utterance modifiers and universals of grammatical borrowing.” Linguistics 36: 281–331. Newman, Paul. 2000. The Hausa language: An encyclopedic reference grammar. New Haven: Yale University Press. Payne, John R. 1985. “Complex phrases and complex sentences.” In: Timothy Shopen (ed.), Language typology and syntactic description, vol. 2: Complex constructions. Cambridge: Cambridge University Press, 3–41. Progovac, Ljiljana. 2003. “Structure for coordination.” In: Lisa Cheng & Rint Sybesma (eds.), The second Glot International state-of-the-article book. Berlin: Mouton de Gruyter, 241–87. Ross, John R. 1986. Infinite syntax. Norwood, NJ: Ablex. (Original: “Constraints on variables in syntax”, 1967 Ph.D. dissertation, MIT.) Ross, Malcolm. 2002. “Takia.” In: John Lynch, Malcolm Ross & Terry Crowley (eds.), The Oceanic languages. Richmond: Curzon, 216–48. Saltarelli, Mario. 1988. Basque (Descriptive Grammar Series). London: Routledge.

Coordinating constructions: An overview

Schachter, Paul. 1977. “Constraints on coordination.” Language 53: 86–103. Schwabe, Kerstin & Ning Zhang (eds.). 2000. Ellipsis in conjunction. Tübingen: Niemeyer. Schwartz, Linda. 1988. “Conditions on verb-coded coordinations.” In: Michael Hammond, Edith Moravcsik & Jessica Wirth (eds.), Studies in syntactic typology. Amsterdam: Benjamins, 53–73. Stassen, Leon. 2000. “AND-languages and WITH-languages.” Linguistic Typology 4.1: 1–54. Stirling, Lesley. 1993. Switch-reference and discourse representation. Cambridge: Cambridge University Press. Stolz, Thomas. 1998. “und, mit und/oder und/mit? — Koordination, Instrumental und Komitativ — kymrisch, typologisch und universell.” Sprachtypologie und Universalienforschung 51.2: 107–30. van den Berg, Helma. 1995. A grammar of Hunzib. Munich: Lincom Europa. van Klinken, Catharina. 2000. “From verb to coordinator in Tetun.” Oceanic Linguistics 39.2: 350–63. Voorhoeve, C. L. 1965. The Flamingo Bay dialect of the Asmat language. ’s-Gravenhage: Martinus Nijhoff. Wälchli, Bernhard. 2001. “Ist Koordination in syntaktischer Hinsicht symmetrisch oder asymmetrisch?” In: Bernhard Wälchli & Fernando Zúñiga (eds.), Sprachbeschreibung & Typologie (Arbeitspapiere, 38). Bern: Universität Bern, Institut für Sprachwissenschaft, 45–64. Wälchli, Bernhard. 2003. Co-compounds and natural coordination. Ph.D. dissertation, University of Stockholm. Yuasa, Etsuyo & Jerry M. Sadock. 2002. Pseudo-subordination: a mismatch between syntax and semantics. Journal of Linguistics 38.1: 87–111.


Chapter 2

Coordination in Mentalese* Toshio Ohori University of Tokyo at Komaba

1. Introduction: Mental logic meets linguistic typology 2. Against the universality of coordination 2.1 The universalist claim formulated 2.2 Cross-categorial uniformity 2.3 Category-internal uniformity 3. The relationship between and and or 3.1 Underdifferentiation 3.2 Some asymmetries between and and or 3.3 Implications 4. Mentalese from the bottom up 5. Conclusion


Introduction: Mental logic meets linguistic typology

Mentalese, or the language of thought, has been a focus of discussion among cognitive scientists, especially psychologists and linguists. One common assumption among them is that the human mind is equipped with a system of thought which shares certain properties with formal logic. The aim of this paper is to examine this idea critically and to propose, in a preliminary fashion, some alternative schemas with more empirical validity. As far as I am aware, there is no reasonably comprehensive grammar and lexicon of mentalese. So in what follows, I will begin by specifically focusing on the following claim by leading advocates of mental logic theory: “[T]he well-known languages of Europe and Asia appear to have connectives that are at least approximately equivalent to and, or, and if.” (Braine and O’Brien 1998: 51)

*The present study was in part supported by a fellowship from the Alexander von Humboldt Foundation for the year 2000, which is here gratefully acknowledged. Also, I wish to thank Martin Haspelmath and Brendan Wilson for very helpful comments. All remaining faults are my own.


Toshio Ohori

“[T]he existence of similar early-developing connectives in unrelated languages, taken with the great frequency of such connectives across languages, is certainly consistent with the notion of a universal mental propositional logic and may be hard to explain otherwise.” (ibid.: 52)

Let us take an example. When I speak English and say and, I mean and (i.e. logical conjunction). Conversely, when I mean and, I say and in English. The assumption here is that the relationship between linguistic forms and logical connectives is transparent.1 But can this idea remain tenable when confronted with the diversity of the world’s languages? What variations are found, and what tendencies? These are the questions I wish to address in this paper. Examples will be drawn from a variety of languages, though no extensive and systematic sampling is claimed here. Repeated reference will be made to Japanese, and in this sense this paper can also be used as reference material on this language. The organization of this paper is as follows. In § 2, universalist claims about coordination will be examined. In § 3, it will be pointed out that there is an asymmetry between strategies for marking and and or cross-linguistically. In § 4, some speculations will be made on possible schemas for coordination. Finally, in § 5, a summary and conclusion will be provided.


Against the universality of coordination

2.1 The universalist claim formulated The basic claim of mental logic has already been quoted. The following is a straightforward reformulation of it. (1) Universalist claim (i) All natural languages have forms that are equivalents of logical connectives. But if we take this to mean that all languages have translation equivalents of logical connectives with different degrees of naturalness, it sounds like a truism, offering no

1.In this paper, I will use use the term logical connectives to refer to operators in propositional calculus and adopt capital letters to represent them, e.g. and, or, if. The term coordination is used to refer to constructions in natural language that are structurally, though not always semantically/pragmatically, symmetrical. Examples whose sources are not indicated are drawn either from my intuition or from my own materials. I am indebted to Kim Hoang for the Vietnamese data (1988–1989) and to Ruetaivan Kessakul for the Thai data (1999). Japanese conversational data are from my archive, created with assistance from Satoko Toyosaki (1995).

Coordination in Mentalese

meaningful constraints on the way natural language is structured. What is assumed by mental logic theorists is a much stronger claim, which I interpret in the following way. (2) Universalist claim (ii) All natural languages grammaticize logical connectives. It is this strong version of the universalist claim which has graver theoretical significance. Indeed, there have been earlier studies that addressed this issue, e.g. Gil (1991), and this paper is an addition to the not-too-long list along this line. To be fair, mental logic theory does not claim that formal logic exists in our mind unmodified. But it is worth demonstrating that the linguistic argument for the existence of universal formal logic is far less strong than has been naively assumed. The examination of the universalist claim (2) proceeds in two steps. First, I will examine cross-categorial uniformity of marking, i.e. whether one and the same strategy is used to connect units of all syntactic categories in every language. Second, I will also examine category-internal uniformity of marking, i.e. whether there is a privileged marker for a given syntactic category in every language. 2.2 Cross-categorial uniformity Support for the universalist claim of coordination will be obtained if the same marking strategy is used for all syntactic categories. Certainly, this is true for English and many Indo-European languages. In what follows, discussion is limited to and. English (3) books and articles


(4) young and promising scholar


(5) Your argument lacks evidence and it won’t persuade anyone.


And it is not difficult to come up with analogous cases from languages that are neither genetically nor areally related. Here, for want of space, only two sets of examples are given. More examples would spring to a Eurocentric mind in a flash. Vietnamese (6) Con chó và thˇa`ng nhoÑ clf dog and clf kid ‘the dog and the kid’


(7) môt » ngþüi ` cô khôn và 0ep » one clf girl smart and pretty ‘a smart and pretty girl’




Toshio Ohori

(8) Con cú bay ra và thˇa`ng nhoÑ té xuâong dâoc. (Clause+Clause) clf owl fly go.out and clf kid stumble down slope ‘The owl flies out and the kid stumbles down the slope.’ Tagalog (Ramos and Cena 1990: 78, 112) (9) aso at pusa dog and cat ‘(a) dog and (a) cat’ (10) mayaman tanyag, at mabait rich famous and good ‘rich, famous, and good’ (11) Nag-sigarilyo si Pepe at nag-tabako si Daniel. do-cigar top Pepe and do-tobacco top Daniel ‘Pepe smokes cigars and Daniel smokes tobacco.’




Thus, the reasoning goes, coordination can be considered primitive and basic to human language. Being primitive, it can also be innate, and hence one may jump to the conclusion that logical connectives directly map onto linguistic forms. This thread of supposition does not hold, however. Counter-examples abound. Mandarin Chinese (12) Zho¯ngguó hé Meˇiguó China and America ‘China and America’


(13) yánsù hé xı¯ngfèn de xı¯nqíng awestruck and excited attr feeling ‘(an) awestruck and excited feeling’


(14) Jı¯ngtia¯n heˇn hˇao, wˇo xiˇang qù jia¯oyóu. today very fine 1sg would go picnic ‘It’s fine today, and I wish to go on a picnic.’


Here, hé is used to connect NPs in (12) and APs in (13), but the zero-marking strategy is used to connect clauses in (14). This is not, however, the whole story. Zero-marking (or more precisely marking by intonation) is common for all categories. For NPs, besides hé which connects entities of equal salience, the comitative coverb ge¯n ‘follow>with’ is also used. Hé for connecting APs has a rather limited distribution (mainly in written style), and there are other possibilities with different pragmatics. For example, yòu Adj1 yòu Adj2 is a common strategy for expressing the additive sense ‘also’. For coordination of clauses, besides zero-marking as in (14), weak interlacing (a term introduced by Lehmann 1988) by adverbial particles is also observed (from Chao 1968: 106).

Coordination in Mentalese

(15) N˘ı yeˇ bú rèn de wˇo, wˇo yeˇ bú rèn de n˘ı. 2sg also neg know res 1sg 1sg also neg know res 2sg ‘You don’t know me, and I don’t know you.’ This situation in Mandarin Chinese is significant. Even though it is typologically similar to Vietnamese (or even “identical” to naive eyes), the two languages show interesting differences with regard to cross-categorial uniformity of coordination strategy. There are also typologically different languages that lack cross-categorial uniformity of coordination marking. Consider the following examples from Japanese, which show a common pattern across verb-final agglutinating languages. Japanese (16) hon-to zasshi(-to) book-and magazine (and) ‘(a) book and (a) magazine’


(17) takakute yoi seki expensive.te good seat ‘(an) expensive and good seat’


(18) Hon-o sutete machi-ni deyoo. (Clause+Clause) book-acc throw.away.te town-dat go.out.vol ‘Let’s throw away the books and walk out onto the street.’ In (16), NPs are coordinated by to, used also as a comitative marker. To can be optionally attached to the last NP. In (17) and (18), adjectives and verbs respectively take the participial form, known as the te-form, in order to form a coordination construction (we shall return later to the question whether this is genuine coordination).2 Possible alternatives for (17)–(18) are the corresponding forms without te as in the following examples. (17¢) takaku yoi seki expensive good seat (same as (17)) (18¢) Hon-o sute machi-ni deyoo. book-acc throw.away town-dat go.out.vol (same as (18))



2.In Japanese, adjectives belong to the verbal category, i.e. they can form a predication without being supported by the copula. There is another adjective-like category whose internal structure is property noun+copula, but this category is not discussed here.



Toshio Ohori

This type of form is variously called the gerundive, infinitive, or stem form. When the verb stem ends with a consonant, as in kak- ‘to write’ (citation form kaku), the coordinating form becomes kaki, so it is often called the i-form as opposed to the te-form. There are other strategies as well. Here it is only pointed out that a relic of the negative form in classical Japanese shows up in the coordination of VPs. Below, (19) is a more standard negation of the predicate while (20) is a relic of the classical negative form. (19) Gakkoo-ni ikanaide tomodachi-ni au. school-dat go.neg.te friend-dat meet ‘(X) doesn’t go to school and see (her/his) friend.’ (20) Gakkoo-ni ikazu tomodachi-ni au. school-dat go.neg friend-dat meet ‘(X) doesn’t go to school and see (her/his) friend.’ In (19), ikanai is analyzed as ik- (stem) plus anai, an allomorph of nai (negative auxiliary in modern Japanese). Coordination is marked by de, which is an allomorph of te. In (20), ikazu shares the stem ik-, and azu is an allomorph of zu (negative auxiliary in classical Japanese).3 Formerly, (a)zu occurred both sentencefinally and in clause chains, but today it is used only in clause chains. See the following examples. (19¢) Gakkoo-ni ikanai. school-dat go.neg ‘(X) doesn’t go to school.’ (20¢) ??Gakkoo-ni ikazu. school-dat go.neg ‘(X) doesn’t go to school.’ Admittedly, (20) is less colloquial than (19), but (20¢) is extremely deviant even in written style. In fact, since zu does not inflect today, it could be labeled “negative conjunction” by an imaginary field worker who had no access to historical data. The situation found in Japanese is fairly common among verb-final agglutinative languages. They tend to grammaticize clause chains and hence clause-chaining morphology may suffice for the purpose of clause coordination. For example, many Papuan languages show this feature:

3.To be more precise, azu originally comes from the combination of a (the vocalic element inserted between the consonant-final verb stem and the auxiliary having an irrealis meaning) and zu.

Coordination in Mentalese

Tauya (Papua New Guinea; MacDonald 1990: 137, 222) (21) pename-sou wa‘a nono ne-pi-sou Pename-com female child 3sg-gen-com ‘Pename and her daughter’


(22) na bramani yate-a-te ten awi etiti fofe-ene-‘a (Clause+Clause) 2sg Brahman go-2sg-ds 2pl two return come-1/2pl-ind ‘You (sg) went to Brahman and you two came back.’ Below are two more sets of examples from Eurasia. Kolyma Yukaghir (isolate, eastern Siberia; Maslova 2003: 313, 370) (23) alandin tude aduø-n’e Alandin his son-com ‘Alandin and his son’


(24) [cˇaj o:že-t [modo-l-u-ke met ejme-n ada:-n [tea drink-ss:ipfv] [sit-12-0-ds] I across-prol here-prol pugeže-s’ (Clause+Clause) dart.out-pfv:intr:3sg ‘I was sitting and drinking tea, (and) it darted out right there, at the other side.’ Shizunai Ainu (isolate, northern Japan; Refsing 1986: 162, 240) (25) mun ka nonno shrubs and flower ‘shrubs and flowers’ (26) ponno rurihi ci kar wa ci e a.little (fish)broth we make and we eat ‘We prepared a little fish broth and ate it.’



Indeed, it is highly problematic to consider the clause-chaining construction as coordination. As has been argued (e.g. Olson 1981, Foley and Van Valin 1984, Van Valin and LaPolla 1997), the chained clauses crucially depend on the final clause for certain grammatical features (for a discussion specifically aimed at Papuan languages, cf. Roberts 1988). For example, the te-marked clause in Japanese is tenseless, while the Tauya example shows suppression of mood in the first clause. This structure, called “cosubordination”, is widespread across languages of the world (cf. Haspelmath and König 1995 under the rubric of “converbs”). It is neither embedding (=subordination) nor juxtaposition (=coordination) (for details, see for example Van Valin and LaPolla 1997: 448ff). From these considerations, it is now clear that cross-categorial uniformity of coordination is not a universal feature of natural languages. Even if logical



Toshio Ohori

connectives like and may be universal to the human mind, the coordination construction is not. 2.3 Category-internal uniformity Next, if it is the case that many languages do not grammaticize any uniform strategy for coordinating units across different syntactic categories, do we always find a strategy that is used uniformly or at least dominantly within each category? 2.3.1 NP coordination Let us start with NP coordination. The following examples are taken from Wilkins (1989, 1994ms). Mparntwe Arrernte (Pama-Nyungan; Wilkins 1994ms) (27) arne-le, pwerte-le stick-inst rock-inst ‘with sticks and with stones’ (28) arlkwerte, amirre, irrtyarte, ante alye-kerte shield womera spear and boomerang-purposive ‘with a shield, a spear-thrower, spears, and a boomerang’ (29) ayle uthene irrtyarte uthene boomerang binary.and spear binary.and ‘a boomerang and a spear’ (30) Bob therre Debbie therre Bob two Debbie two ‘Bob and Debbie (together)’ (31) kake ilerne 1du ‘elder brother and I (=we-two)’ (32) nye-nhenge father-dyadic ‘father and child(ren)’ Example (27) is coordination by juxtaposition. The coordinands (in this case arne ‘sticks’ and pwerte ‘stones’) need not exhaust all relevant entities in the universe of discourse.4 There could be other NPs to be mentioned if required. In contrast, (28) is a case of exhaustive listing. Unlike in (27), where the instrumental suffix le is

4.The term coordinand is used here after Haspelmath (to appear).

Coordination in Mentalese

attached to each coordinated NP, the purposive suffix kerte occurs only next to the last coordinand, i.e. it is attached to the whole complex NP. The strategy in (29) is not only exhaustive but is restricted to NPs that form a frame-semantically connected pair. It cannot be used to express, for example, ‘a man and a spear’. (30) also expresses a strong conceptual connection. The difference between (29) and (30) is that while the strategy of the former allows a distributive interpretation, that of the latter does not. Thus when the coordinate NP (30) is combined with the predicate meaning ‘own a house’, the only possible interpretation is that Bob and Debbie jointly own a house. If the strategy of (29) were used instead, it would be possible to have the interpretation that Bob has a house and Debbie has her own too. In (31), an inclusory construction, the interpretation proceeds in the following way: First, kake ‘elder brother’ is mentioned, and then the pronominal ilerne ‘we-two’ is given. The latter form specifies the set which the former is part of. Consequently, the partwhole relation between the two elements gives rise to the interpretation ‘elder brother and I’. Finally, in (32) the existence of one of the coordinands is only entailed. The suffix nhenge indicates a dyadic relation. When attached to nye ‘father’, one plausible interpretation is father plus some concept that makes a pair, i.e. ‘child’ in the present case. Admittedly, not all the above examples qualify as prototypical coordination. But they show a wide range of grammatical strategies that are used to express a pair (or a longer list) of entities put together and functioning as a structural unit. At the extreme, one can think of nominal compounds as the most conventionalized form of coordination. Examples include, but are not restricted to, so-called “irreversible binomials” such as bow and arrow (??arrow and bow), bread and butter (??butter and bread), and heaven and hell (??hell and heaven) (Malkiel 1959, Lambrecht 1984), as well as coordinative compounds in languages like Chinese with much stronger lexical creativity. Do, for example, tia¯n-dì ‘heaven-earth’ or tia¯n-shàng-tia¯n-xià ‘heaven-up/top-heaven-down/bottom (=in the universe)’ qualify as coordination? Though not as dramatically as Mparntwe Arrernte, there are other languages which show multiple possibilities for NP coordination. Let us consider Japanese for illustrations. We have already seen in (16) that to is used to connect NPs. In addition, ya is also possible. Unlike to, this form has no comitative function. Also, it does not attach to the last NP. Japanese (33) hon ya zasshi book and magazine ‘books and magazines’ Partly because it has no comitative sense, coordination by ya seems to express a looser connection, suggesting the existence of wider variety implicit in the mentioned entities; hence the plural interpretation in (33). Ya has a related form yara



Toshio Ohori

(the form ra is hom*ophonous with the non-productive plural suffix ra, as in kare-ra ‘he-pl’, but it is not certain whether this knowledge is significant for modern speakers). The existence of variety or some feeling of abundance is stronger with yara. For whatever reason, yara can be attached to the last NP. (34) hon yara zasshi (yara) book and magazine (and) (same as (33), with the implicature ‘books here and magazines there’) Another strategy is the use of mo ‘also’ (which is used in the sense of ‘even’ in other contexts). Unlike the others, it has to be attached to the last NP as well. (35) hon mo zasshi mo book also magazine also (same as (33), with the implicature ‘not only books but also magazines’) Also, the dative ni can be used to enumerate NPs with some additive meaning. It cannot be attached to the last NP. (36) hon ni zasshi book and magazine (same as (33), with the implicature ‘books plus magazines’) Ni is otherwise used to mark location and goal. Its use in a coordination construction may come from its meaning ‘in addition to’ or ‘on top of ’. In colloquial speech, the form toka is widely used. To is a comitative marker and ka is a particle used to mark questions (sentence finally), indefinite pronouns (suffixed to wh-words), and logical disjunction (to which we shall return in § 3.1). Since this form has not been discussed in any depth in the literature, let us look at it in some detail. An example from natural discourse is given in (37).5 Literal and idiomatic translations are given separately when necessary. (37) A: de, juusu ireru-no-ga ne, nanka, koo, then juice pour-nz-nom prt somehow like.this kyuu-tte osh*te sa, jaa-tto sa, onom-quot push.te prt onom-quot prt ano o-cha ireru-no-toka aru-desho that pre-tea pour-nz-and be-mod (lit.) ‘then, that juice-pouring thing, I mean, like this, (you) push (it) like kyuu (sound of pushing on a lever or button), (then) like jaa

5.As can be seen, spoken Japanese drastically diverges from the regularized and well-known variety, and seems almost impenetrable at first sight.

Coordination in Mentalese

(sound of liquid being poured), (so) there is that tea-pouring thing and things like that, you know’ (id.) ‘and there is this server for drinks such as juice and tea, whose lever you push to serve, you know’ B: hun hun hun hmm hmm hmm (lit. & id.) ‘hmm’ A: are-de nanka koora-toka orenji-toka supuraito-toka that-inst somehow cola-and orange-and Sprite-and (lit. & id.) ‘with that, well, (you serve) cola, orange juice, and Sprite’ Since ka generally conveys uncertainty and hence the possibility of choice, the composite morpheme toka means more than ‘and’. The relevant lines in the above passage are intended to be enumeration, but they are by no means exhaustive. A more detailed semantic description behind the use of toka would be something like ‘(after mentioning X) there could be something else which equally fits the subject matter of the ongoing discourse’. Interestingly, there are extended uses of toka in natural discourse, which in fact are far more frequent in contemporary spoken Japanese. The following examples all occur in the same discourse as (37). Given their deviance from standard coordination, the instances of toka in (38)–(41) are not given English-based glosses. (38) A: nanka ne, nanka uuroncha-toka ne somehow prt somehow oolong.tea-toka prt (lit. & id.) ‘well, well, oolong-tea and things like that’ B: un yeah (lit. & id.) ‘yeah’ A: nanka koo chuumon-ga ooi-no-wa somehow like.this order-nom be.many-nz-top motomoto tsuidoku no ne beforehand serve.prep prt prt (lit.) ‘well, like this, as for things with many orders, (we) prepare to serve beforehand’ (id.) ‘well, we get many cupfuls ready to go for those kinds of drinks which many people order (such as oolong tea)’ (39) B: de, coffee-no nioi-toka, moo, tonjattete, then, coffee-gen aroma-toka already go.away.perfect.stat.te nanka ne, puuru-no mizu-ni suru somehow prt swimming.pool-gen water-dat make yoo-na ne like-cop prt



Toshio Ohori

(lit.) ‘then, coffee’s aroma or anything like that was already gone, and somehow (it’s) like something that (you) use for swimming pool’s water’ (id.) ‘then, the coffee had no aroma, and smelled/tasted like the sort of water you find in a swimming pool (i.e. smelled/tasted awful)’ (40) A: tenchoo-toka sasugani chotto manager-toka a.bit waratte-shimau-tteiu kanji datta laugh.te-perfect-quot feeling cop.past (lit.) ‘the feeling/atmosphere was that the manager, (while) living up to expectation, had to laugh a bit’ (id.) ‘the story was that even the manager (of the restaurant) couldn’t help giggling (encountering that funny situation)’ (41) B: ja nebaroo-to omotte sa, “ja, moo then stay.longer.vol-quot think.te prt “then yet ippai”-toka itte tanonda no yo one.clf-toka say.te ask.past prt prt (lit. & id.) ‘then (I) thought (I) would stay longer, and asked (for another coffee) saying something like “(I’d like) another cup”’ In (38), toka occurs next to a NP uuroncha ‘oolong tea’, but is not followed by any other NP. In this sense the toka construction is reminiscent of (32) from Arrernte, but actually the difference is greater than the similarity. In the dyadic construction in (32), the missing half of the coordination is (or ought to be) identified with ease. But toka goes the opposite way, only suggesting the existence of a like entity which can be enumerated if required. It is a marker to evoke some extra candidate in (38), hence the translation ‘oolong tea and things like that’ (cf. also the first occurrence of toka in (37)). In a way, toka functions to express what may be called optionalization. The speaker may not have any other specific kind of drink in mind besides oolong tea. She may just have a vague idea about its superset, chuumon-ga ooi-no ‘things with many orders’. In (39), toka is attached to a NP again, but here it is less easy to think of anything which jointly forms some hom*ogeneous set with the aroma of coffee. In the present case, the speaker is specifically talking about the aroma of coffee, but he is intending to leave room for some extra candidate by implicating the existence of a set from which the subject matter (i.e. the aroma of coffee) has been chosen, whatever that set may be. This effect is much stronger in (40). Here it does not seem to make sense to interpret the phrase tenchoo-toka as ‘the manager and people like him’. A more reasonable interpretation would be ‘I am talking about the manager (of the restaurant), as he is the one who most deserves mention’. Interactively, the use of toka makes it easier for the interlocutor to co-construct discourse by expanding the range of subject matter. At the same time, it seems to

Coordination in Mentalese

have a function of pragmatic highlighting, by implicating that an entity is specifically chosen out of some set. The last example (41) is different from the others in that toka is attached to a quoted utterance, not a NP. This quote is an instance of reenactment, and toka functions as a hedge to indicate that the speech may be not reproduced verbatim. These facts about Japanese toka are quite intriguing in that a marker for NP coordination/enumeration is developing into a cl*tic-like form attached to a single NP, acquiring its own pragmatics of optionalization. Finally, I shall cite one example of the comitative-coordination continuum from an unexpected language, English. Very often examples are drawn from languages outside Europe, but it is interesting to note that if one extends the perspective a little bit to include diachronic data, one finds that even English and used to have a comitative-like property. Old English (Anglo-Saxon Chronicle, year 755) (42) Her Cynewulf benam Sigebryht his rices ond Westseaxna here Cynewulf deprived Sigebryht his kingdom and West.Saxon wiotan for unryhtum dædum, buton Hamtunscire. elders for unrighteous deeds but Hampshire ‘In this year, Cynewulf and the West Saxon elders deprived Sigebryht of his kingdom for unrighteous deeds, except for Hampshire.’ If analyzed with modern eyes, this example must be ungrammatical, since it violates the coordinate structure constraint, the subject NP being Cynewulf ond Westseaxna wiotan. A more reasonable analysis would be that Old English ond was not fully grammaticized as a coordination marker. This fact suggests that balanced coordination in modern English, directly reflecting logical and, was not necessarily the norm a millennium ago. 2.3.2 Clause coordination As already seen, many languages have cosubordinate/converb constructions, whose closest translation equivalents in better known European languages are coordination. Structurally, however, they are not. Symmetrical and balanced constructions are rather marked in such languages. This point is illustrated by (43)–(45). Kewa (Papua New Guinea; Franklin 1983: 43) (43) no-a pira-wa eat-seq sit-1sg.past ‘I ate and sat down.’ (44) no pira-wa eat(sim.unit) sit-1sg.past ‘I sat eating.’



Toshio Ohori

(45) na-ri pira-wa eat-sim.split sit-1sg.past ‘While eating, I sat down.’ These are all instances of same-subject switch-reference constructions. (43) is marked by sequential -a, (44) by “unitary” symmetrical zero, and (45) by “split” symmetrical ri (here the term “symmetrical” is roughly equal to simultaneity). When it is necessary to express conceptual symmetry explicitly, the following strategy is possible. (46) saa eta no agaa lo pi-pa we=two food eat talk say sit-1du.perfect ‘We have been eating and talking together.’ (47) ne-me sapi gusu nipu-mi ipa iru pi-pa I-ag kaukau cook he-ag water boil sit-1du.perfect ‘He and I have been respectively boiling water and cooking kaukau.’ In each of these examples, the second activity (agaa lo ‘talk say’ in (46) and iru ‘boil’ in (47)), which could be marked by a final, finite ending with person and aspect marking as in (43)–(45), is deprived of such features and is on an equal footing with the first activity. Note that in (47), even though the two clauses have different subjects (lit. ‘I have been cooking kaukau and he has been boiling water’), a switchreference marker for different subjects is absent. The coordinated activities, both marked by unitary symmetrical zero, jointly depend on the syntactically main and semantically bleached final verb pi ‘sit’. In this way, Kewa requires a special balanced construction to express conceptual symmetry. Japanese has a similar construction using tari (originally the perfective auxiliary in classical Japanese), as in the following example. Japanese (48) Min’nade okashi-o tabe-tari terebi-o mi-tari sh*ta. together snack-acc eat-tari TV-acc watch-tari do.past ‘We all ate snacks and watched TV.’ (=We spent time eating snacks and watching TV) Here, the suffix tari is playing a function equivalent to the unitary symmetrical marking in Kewa. The predicates in the two clauses are both in non-finite form, and are jointly governed by the final verb sh*ta ‘did’, whose occurrence is obligatory.6

6.Interestingly, tari has taken on a function comparable to that of toka. That is, in colloquial speech, tari (sh*te) (sh*te = te-form of suru ‘do’) can be attached to a single clause, thereby becoming like an utterance-final particle, with roughly the meaning ‘and things like that’. Again, the speaker is able to suggest alternative possibilities by this strategy.

Coordination in Mentalese

This is also analogous to the specialized symmetrical coordination in Kewa. Since it is widely acknowledged that clause-chaining languages tend to grammaticize a variety of clause linkage forms, I need not spend much space to show the lack of category-internal uniformity of clause coordination in such languages. For Japanese, we have seen the te-form, the i-form (realized as zero with vowel-final verb stems), beside the tari-form. Besides stylistic differences (te is ubiquitous while i is mainly limited to the written style at the clause level, while both of them are used in the formation of complex predicates), the te-form allows a concessive interpretation, while the i-form does not. In this sense the former is more inferenceintensive compared with the latter. Another possibility is to use toka, which is attached to the finite (i.e. tensed) form of the verb, and has to be supported by the matrix verb suru as in (48). See the following example.7 (49) Kao-o arau-toka ha-o migaku-toka shinasai. face-acc wash-toka tooth-acc brush-toka do.imp ‘Wash your face and brush your teeth!’ Toka can occur after tari, though it sounds most natural to use it only for the last coordinand. (50) Min’nade okashi-o tabe-tari(-toka) terebi-o together snack-acc eat-tari-toka TV-acc mi-tari-toka sh*ta. watch-tari-toka do.past ‘We all ate snacks and watched TV.’ (=We spent time eating snacks and watching TV) The reverse order, toka-tari, is ill-formed. This fact reflects the relative strength of each form’s bond to the verbal root, i.e. toka attaches to a finite (=tensed) form while tari occurs directly after the i-form (=untensed). The last construction to be taken up here is shi-coordination. It is attached to a tensed form of the verb and marks a weaker connection between the two events. Causal and sequential connotations are weak compared to i- and te-forms, especially the latter. Even in subject-sharing constructions, the scope of the volitional auxiliary is different. (51) Gaikoku-ni itte shinken’ni kenkyuu-sh*tai. go.te seriously research-do.vol ‘(I) would like to go abroad and do my research seriously.’

7.This is probably the only strategy applicable to both NPs and clauses besides intonationally linked juxtaposition, and is very widespread in Japanese.



Toshio Ohori

(52) Gaikoku-ni iki shinken’ni kenkyuu-sh*tai. go seriously research-do.vol (same as (52)) (53) Gaikoku-ni iku-shi shinken’ni kenkyuu-sh*tai. go-shi seriously research-do.vol ‘(I) go abroad and (so) would like to do my research seriously.’ While (53) allows a causal interpretation because of its semantics, more symmetrical relations can also be expressed by shi. (54) Ani-wa sakkyokuka-da-shi otooto-wa gaka-da. composer-cop-shi painter-cop ‘The elder brother is a composer and the younger brother is a painter.’ From all these examples, it is clear that category-internal uniformity of coordination strategy does not hold universally, either for NPs or for clauses. The universalist claim formulated in (2) does not seem defensible.


The relationship between and and or

While the above discussion may be persuasive enough to discredit the idea that there is a single, privileged marking strategy for each logical connective such as and, one might still argue that this shows only that natural languages tend to have finer distinctions than logic. Thus, one might suppose that coordination in each language (and for each syntactic category) is a set of constructions having a radial-category structure (in the sense of Lakoff 1987). This concession itself is problematic for a straightforward mental logic theory which would assume some kind of transparency principle in language-logic mapping. In this section, we shall see some further problems with this assumption. 3.1 Underdifferentiation Besides the problem of overdifferentiation, for which there is ample evidence, what if it is the case that natural languages also underdifferentiate logical relations? This is the case in the following Maricopa examples. Maricopa (Yuman; Gil 1991: 99) (55) John-š Bill-š v‘aawuum John-nom Bill-nom ‘John and Bill will come.’

Coordination in Mentalese

(56) John-š Bill-š v‘aawuumšaa John-nom Bill-nom ‘John or Bill will come.’ Here, what matters is the epistemic status of the state of affairs. In (55), the verb is marked as future only and the state of affairs being described is securely believed by the speaker with higher certainty. But in (56), the verb has an inferential suffix, and the speaker is not entirely certain about what is going to happen. The and-or distinction is thus dependent on the choice of a modal-like element on the verb. If certain, the resulting interpretation is and. If uncertain, it is or. This kind of situation often involves very subtle pragmatic factors, and it is not easy to gather relevant information from grammars. But here are a few more examples. In Upriver Halkomelem (Salish), coordination is marked by q6 for both and and or (and ‘but’ as well). The interpretation “depends on semantic environment” (Galloway 1993: 363). See the following examples (ibid.: 416). (57) L6 l6m´6lst6xw6s t6 Bill t6 sq’´6m´6l xw6l´7m t6 Jim q66 Bob. 3 throw.3 dem Bill dem paddle to dem Jim and Bob ‘Bill threw the paddle to Jim and Bob.’ (58) Lí l´7m k’w6 Bill q66 Bob? q go dem Bill or Bob ‘Did Bill or Bob go?’ While I was not able to reconstruct the precise mechanism of interpretation from the grammar, the declarative construction seems to allow a conjunctive reading in (57) and the interrogative construction a disjunctive reading in (58). In Hua (Papua New Guinea), the morpheme ve can be used to coordinate NPs with the meaning of and, but this morpheme is also used as an interrogative suffix. Consequently, its meaning is “identified as that of disjunction: the reading ‘and/or’, that of inclusive disjunction, is a consequence of the fact that -ve, unlike -Ki, marks non-exhaustive enumeration” (Haiman 1980: 251). In the following Japanese examples, the interpretation is dependent on framesemantic knowledge. NP coordination by juxtaposition is used here. (59) Doko-ni ikitai no? where-dat go.vol prt ‘Where do you wish to go?’ – Kyoto, Nara, Kobe da naa. Kyoto Nara Kobe cop prt ‘Kyoto, Nara, and Kobe, I suppose.’



Toshio Ohori

(60) Doko-ni sumitai no? where-dat live.vol prt ‘Where do you wish to live?’ – Kyoto, Nara, Kobe da naa. Kyoto Nara Kobe cop prt ‘Kyoto, Nara, or Kobe, I suppose.’ It is possible to visit three cities in one trip, if one wishes to, but it is not normally possible to live in three cities at one time. Hence the most probable interpretation for (59) is conjunctive and that for (60) is disjunctive. Instead of juxtaposition, the use of to and ka makes explicit the meanings of and and or respectively, which in fact sound more natural. (59¢) Doko-ni ikitai no? where-dat go.vol prt ‘Where do you wish to go?’ – Kyoto-to Nara-to Kobe da naa. Kyoto-and Nara-and Kobe cop prt ‘Kyoto, Nara, and Kobe, I suppose.’ (60¢) Doko-ni sumitai no? where-dat live.vol prt ‘Where do you wish to live?’ – Kyoto-ka Nara-ka Kobe da naa. Kyoto-or Nara-or Kobe cop prt ‘Kyoto, Nara, or Kobe, I suppose.’ To make the matter more complicated, the form toka can be used instead of zero in (59)–(60), with the same kind of ambiguity. This is understandable given its etymology, which is a composite of conjunctive to and disjunctive ka. Another complication comes from the use of a comitative marker in the disjunctive situation. Here are examples from Thai. Thai (61) Cffn kàp m77rîi paj duu nˇa]. John with/and Mary go see movie ‘John and Mary went to see a movie.’ (62) Th66 cà‘ lyâk tham ‘araj rá‘wàa] paj duu nˇa] kàp paj you will choose do what between go see movie and/or go sýy khˇff]? buy thing ‘Which would you like to do, go to see a movie or go shopping?’

Coordination in Mentalese

Example (61) illustrates the use of comitative kàp for NP coordination. As a comitative phrase, kàp m77rîi can be moved to the end of the sentence, as in cffn paj duu nˇa] kàp m77rîi. Example (62) is also a case of NP coordination, but here the comitative marker must be interpreted as or, not as and. A similar situation is found in Japanese too. Japanese (63) Eega-to kaimono(-to) docchi-ga ii? movie-and/or shopping-and/or which-nom be.good ‘Which is good (for you), movie or shopping?’ A subtle difference from the Thai example is that in Japanese the coordinated NPs, eega-to kaimono(-to), are in an appositive relation to docchi ‘which’, as in the English translation. Note that the case marker ga is attached to docchi ‘which’, not to the coordinated NPs. Hence the following example is ungrammatical. (64) *Eega-to kaimono(-to)-ga docchi ii? movie-and/or shopping-and/or-nom which be.good (intended: same as (63)) So a possible interpretation process is ‘given the set consisting of movie and shopping, which do you select?’. Unlike in English, however, the disjunctive ka is unnatural when used in this construction. (65) ??Eega-ka kaimono(-ka) docchi-ga ii? movie-and/or shopping-and/or which-nom be.good (intended: same as (63)) The disjunctive relation is not a part of the meaning of to, but of the whole construction. The implication of the foregoing array of facts is this: and and or, the two basic logical connectives in formal logic, can sometimes be underdifferentiated in natural languages. An appropriate interpretation is obtained only by considering the epistemic status of the proposition as in Maricopa or by considering the relevant universe of discourse and frame-semantic knowledge behind an utterance as in Japanese and Thai. Of particular importance for the language-logic connection is the case of Maricopa. In modal logic, the epistemic status, or modus, of a proposition is defined with reference to a theory of quantification, which in turn is based on primitive propositional calculus. But in natural language, there are cases in which epistemic status is the crucial determinant of sentence interpretation at the level of propositional calculus. This subversiveness raises the problem of what is really basic to the structure of natural language and what belongs to “higher order” thinking.



Toshio Ohori

3.2 Some asymmetries between and and or Having seen that logical connectives themselves may not transparently map onto natural language forms, let us examine another aspect of and and or in natural languages, namely the asymmetries between them. Again, this is important in considering the basicness of logical connectives, for in formal logic there is no specification as to which is more basic to the other, as the following formulae indicate. (66) a. p and q = not(not p or not q) b. p or q = not(not p and not q) But this symmetry is not observed in some natural languages. We shall look at some manifestations of this. First, there can be a difference in the coherence of marking strategies, as the following Maori (Austronesian) examples show. Maori (Bauer 1993: 126–127, 130) (67) Kei te oma te maahita, me ngaa tamariki t/a run the teacher and/with the(pl) children i runga i te one. at top at the beach ‘The teacher and the children are running along the beach.’ (68) Ka kite raatou i te ngaawhaa, i te whare t/a see do the boiling.pools do the house whakairo, i te aha, i te aha. carved do the what do the what ‘They saw the boiling pools, the carved houses and other things.’ (69) Ka aawhina a Mere, a Marama raanei i a Pani. t/a help pers Mary pers Marama or do pers Pani ‘Either Mary or Marama will help Pani.’ (70) Me turituri koe, me haere koe ki too ruuma, oblig quiet 2sg oblig move 2sg to sg.gen.2sg room me haere atu koe ki waho raanei. oblig move away 2sg to outside or ‘Either be quiet, (or) go to your room, or go outside.’ Maori has various strategies for expressing the and-relation. For subject NPs, zero marking is possible, but the use of comitative me is said to be preferred as in (67). For NPs with prepositions, there is an obligatory copying of preposition (in the above case, i), as in (68). For VPs and clauses, juxtaposition is common, and aa and hoki are used as secondary strategies with sequential and additive meanings respectively. But this diversity is not found with the disjunctive relation. The

Coordination in Mentalese

morpheme raanei is used throughout, as (69)–(70) show. The fact that or has a uniform marking strategy compared to and suggests that it is conceptually more specialized, with a more articulated meaning. This kind of situation is not uncommon among languages that lack cross-categorial uniformity for and. For example, Japanese uses participial morphology for clausal coordination and a comitative particle (besides some other particles) for NP coordination if the semantic relation to be expressed is and. But when expressing the or-relation, the particle ka is uniformly employed. Second, when we look at the morphological type, strategies for the and-relation appear to be less autonomous than those for the or-relation. Here we look at Yuchi (isolate, presumably related to Siouan; Wagner 1934). (71) is a list of cl*tics for connecting constituents. Yuchi (71) -lah˛c’ -ya’h˛c -d7 -l7’nd7 -la

‘therefore, and so’ ‘therefore, and so (after negation)’ ‘and, also’ ‘only and then’ ‘but’

“The coordinative and subordinative relation between two clauses and a word and a clause is expressed by a number of verbal encl*tics. However, … only a few conjunctive ideas are expressed by the formal device of enclisis while in the majority the conjunction is an independent particle” (Wagner 1934: 362). The or-relation, unlike those in (71), is expressed by an independent word na’˛7ya’nd7, glossed ‘or else’. To be sure, there are other independent words meaning ‘but’, ‘and then/also/ so’, ‘further’, but the asymmetry between and and or is clear enough. Third, let us briefly look at frequency of occurrence in discourse. This can only be examined with languages that grammaticize both and and or, preferably with cross-categorial uniformity, so here I take up English and German. I used formal conversation for English and storytelling for German.8 The results are as follows. English (72) ca. 350,000 words 8786 occurrences of and 1414 occurrences of or [and-or ratio: 6.2 to 1]

8.The data used were three texts from the Corpus of Spoken Professional American English compiled by Michael Barlow for English, and ten storytelling texts from the Motion Events and Discourse project conducted by the present author for German.



Toshio Ohori

German (73) ca. 9,000 words 546 occurrences of und 32 occurrences of oder [and-or ratio: 17.1 to 1] The much lower rate of occurrence of oder in German compared to that of or in English is attributed to the fact that the survey of the former was based on storytelling. In natural discourse, especially in storytelling, und is used for advancing the narrative, while oder, mainly used for provision of options, is far less useful for simple storytelling. In English, on the other hand, the data were formal conversation (committee meetings and press conferences), so or occurs more frequently to provide options in exposition, as well as in questions from the interviewer. Even so, the frequency asymmetry is obvious. 3.3 Implications The above survey indicates: (i) The or-relation tends to be more transparently mapped onto natural language and may be more well-defined in its conceptual content. (ii) The or-relation tends to be marked with greater morphological bulk and/or autonomy than the and-relation. (iii) The or-relation tends to be far more restricted in its occurrence in discourse than the and-relation. Taken together, the asymmetry between and and or is that of markedness, i.e. or is a marked concept.9 This idea can be shown in the form of a matrix as follows. Table 1.

and or





+ +

+ −

− +

− −

For any of the properties (i)–(iii), the situation (b) does not hold in any language. Thus in English, (a) holds for consistency of coding strategy, (d) for relative bulk and/or autonomy of morpheme, and (c) for restrictedness of frequency. In Japanese, (c) holds for consistency of coding strategy, (d) (or (a) depending on the

9.In passing, it is revealing that some sophisticated varieties of formal logic adopt the convention that juxtaposition of symbols is to be interpreted as having the and-relation while explicit notation is employed to represent the or-relation. This fact seems to be an unwitting manifestation of the asymmetry between and and or that is felt among logicians oriented to natural language. I owe this point to Shuichi Yatabe.

Coordination in Mentalese

form) for relative bulk and/or autonomy of morpheme, and (c) for restrictedness of frequency. A universal that may be proposed now is as follows. (74) There is no language in which and is more marked than or in terms of consistency of coding strategy, morphological bulk/autonomy, or restrictedness of frequency in discourse. Let us now think about the consequences of the facts surveyed so far for the theory of the language of thought. 4. Mentalese from the bottom up We have seen in this paper that coordination in natural languages is so diverse that it is far from easy to postulate anything like Platonic coordination. At the same time, if all humans, at least adults, share some ability to perform logical inferences, we face a new and very important problem. This problem, called the “mapping problem”, concerns the question how it is possible that children develop logical thinking given such diverse linguistic inputs (cf. Slobin 1997, Bowerman and Choi 2001 on “typological bootstrapping”). The position I am taking is an interactionist one, with a commitment to the view that there is a dynamic interaction between the pre-wired brain state and the environmental input. Given this orientation, the issue then concerns the processes whereby logical operators such as and and or are abstracted from linguistic expressions, though exact developmental paths may differ across languages. It seems that logical operators are higher-order abstractions, and there can be a system of representations which mediates between language and logic. If there is to be a sensible discussion of mentalese, it is in terms of this level of representation. Here I only proceed in a conjectural way, listing conceptual schemas that can be gathered from the data surveyed in this paper. There will be redundancies, apparently. But this cannot be avoided if one aims to be realistic, for each individual language draws its own inventory of conceptual schemas from the pool of schemas potentially available to the human mind. The following list is unstructured for now, but when enough knowledge is accumulated, it can be turned into a conceptual map with stronger explanatory power (for the use of conceptual maps, cf. Croft 2001). The first schema to be proposed is symmetry–asymmetry. Ultimately this schema is related to the figure/ground distinction in perception (Talmy 1983, 2000). When projected onto the temporal domain, we find the schema of temporal sequence. In many cases this schema is the default interpretation of clausal coordination, at the same time activating various sorts of frame-semantic knowledge. Next, especially for NP coordination, there is a schema of accompaniment. It entails the co-presence of individuals and hence maps onto and rather straightforwardly.



Toshio Ohori

Another important schema is incremental activation. In its pure form, it is manifested as an act of enumeration. This is one of the essential human acts across cultures, regardless of the numeral system being used. It is also one of the earliest developing linguistic faculties of children, prior to the acquisition of complex constructions and a numeral system. Probably this schema develops in conjunction with the concept of quantification (all and some). Related to the above schemas is binary set. It is, for example, closely connected with accompaniment. Many concepts are acquired as pairs, and this fact gives motivation to the formation of binary set, whose exhaustive enumeration (=coactivation of two entities or states of affairs) leads to an overt recognition of the pair. Also, the realis-irrealis distinction of mode is a crucial factor in forming the concepts of and and or. As we have seen, a list of entities can be interpreted conjunctively (=and) in some languages when the predication is in the realis mode. On the other hand, when the predication is in the irrealis mode, this means the acknowledgement of alternative possibilities (more generally the construction of mental spaces). The disjunctive (=or) interpretation is a result of this process. In this connection, it is intriguing to think of the possible order of acquisition of linguistic forms associated with or and those associated with irrealis mode (and, if possible, in relation to the stage where children understand the concept of displacement). Admittedly, the above listing is no more than a first step toward a realistic characterization of mentalese. The discovery of the exact nature of the mutual relations among schemas and the formulation of inference rules are tasks for future research.



In this paper, we started with the claim of the advocates of mental logic theory, focusing on the linguistic argument for its universality. We have seen that many languages lack cross-categorial uniformity of coordination marking, and many also lack category-internal uniformity. Further, besides overdifferentiation of marking strategies, we examined cases of underdifferentiation. Thus, while our mind may be equipped with the ability to perform logical inference, the sheer diversity of natural languages makes it difficult to believe in transparent mappings from linguistic forms to logical connectives. Finally, a set of conceptual schemas have been proposed as mediators between the two levels. Working out precise mechanisms for the typological bootstrapping of the acquisition of logical notions will require the joint efforts of typologically-minded linguists and cognitive scientists.

Coordination in Mentalese

Abbreviations 0 acc ag attr clf com cop dat dem do ds du fut gen imp ind infer inst intr ipfv mod neg nom nz oblig

sub-morphemic unit accusative agent attributive classifier comitative copula dative demonstrative direct object different subject dual future genitive imperative indicative inferential instrumental intransitive imperfective modality negative nominative nominalizer obligation

onom pers pfv pl pre prep prol prt q quot res seq sg shi sim ss stat t/a tari te toka top unit vol

onomatopoeia person perfective plural prefix preparatory prolative particle question quotative resultative sequential singular suffix shi simultaneous same subject stative tense/aspect suffix tari suffix te suffix toka topic unitary volitional

References Bauer, Winifred. 1993. Maori. London: Routledge. Bowerman, Melissa and Soonja Choi. 2001. “Shaping meanings for language: universal and language-specific in the acquisition of spatial semantic categories”. In Melissa Bowerman and Stephen C. Levinson (eds.), Language Acquisition and Conceptual Development. Cambridge: Cambridge University Press, 475–511. Braine, Martin D. S. and David P. O’Brien. 1998. “How to investigate mental logic and the syntax of thought”. In Martin D. S. Braine and David P. O’Brien (eds.), Mental Logic. Mahwah, NJ: Lawrence Erlbaum Associates, 45–61. Chao, Yuenren. 1968. A Grammar of Spoken Chinese. Berkeley and Los Angeles: University of California Press. Croft, William. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: Oxford University Press. Foley, William and Robert D. Van Valin, Jr. 1984. Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press.



Toshio Ohori

Franklin, Karl J. 1983. “Some features of interclausal reference in Kewa”. In John Haiman and Pamela Munro (eds.), Switch-Reference and Universal Grammar. Amsterdam: John Benjamins, 39–49. Galloway, Brent D. 1993. A Grammar of Upriver Halkomelem. Berkeley and Los Angeles: University of California Press. Gil, David. 1991. “Aristotle goes to Arizona and finds a language without ‘and’”. In Dietmar Zaefferer (ed.), Semantic Universals and Universal Semantics. Berlin: Foris, 96–130. Haiman, John. 1980. Hua: A Papuan Language of the Eastern Highlands of New Guinea. Amsterdam: John Benjamins. Haspelmath, Martin. To appear. “Coordination”. In Timothy Shopen (ed.), Language Typology and Syntactic Description, 2nd. ed. Cambridge: Cambridge University Press. Haspelmath, Martin and Ekkehard König (eds.) 1995. Converbs in Cross-linguistic Perspective. Berlin: Mouton de Gruyter. Lakoff, George. 1987. Women, Fire, and Dangerous Things: What Categories Reveal about the Mind. Chicago: University of Chicago Press. Lambrecht, Knud. 1984. “Formulaicity, frame semantics and pragmatics in German binomial expressions”. Language 60: 753–796. Lehmann, Christian. 1988. “Towards a typology of clause linkage”. In John Haiman and Sandra A. Thompson (eds.), Clause Combining in Grammar and Discourse. Amsterdam: John Benjamins, 181–225. MacDonald, Lorna. 1990. Tauya. Berlin: Mouton de Gruyter. Malkiel, Yakov. 1959. “Studies in irreversible binominals.” Lingua 8: 113–160. Maslova, Elena. 2003. A grammar of Kolyma Yukaghir. Berlin & New York: Mouton de Gruyter. Olson, Michael L. 1981. Barai Clause Junctures: Toward a Functional Theory of Interclausal Relations. Ph.D. Dissertation, Australian National University. Ramos, Teresita V. and Resty M. Cena. 1990. Modern Tagalog: Grammatical Explanations and Exercises for Non-native Speakers. Honolulu: University of Hawaii Press. Refsing, Kirsten. 1986. The Ainu Language: The Morphology and Syntax of the Shizunai Dialect. Aarhus: Aarhus University Press. Roberts, John R. 1988. “Amele switch-reference and the theory of grammar”. Linguistic Inquiry 19: 45–63. Slobin, Dan I. 1997. “The origins of grammaticizable notions: beyond the individual mind”. In Dan I. Slobin (ed.), The Crosslinguistic Study of Language Acquisition. Mahwah, NJ: Lawrence Erlbaum Associates, 265–323. Talmy, Leonard. 1983. “How language structures space”. In Herbert L. Pick, Jr. and Linda P. Acredolo (eds.), Spatial Orientation: Theory, Research, and Application. New York: Plenum Press, 225–282. Talmy, Leonard. 2000. Toward a Cognitive Semantics. Cambridge, MA: MIT Press. Van Valin, Robert D., Jr. and Randy J. LaPolla. 1997. Syntax: Structure, Meaning, and Function. Cambridge: Cambridge University Press. Wagner, Günter. 1934. Yuchi. Extract from Franz Boas (ed.), Handbook of American Indian Languages 3. New York: Columbia University Press. Wilkins, David. 1989. A Grammar of Mparntwe Arrernte (Arranda). Ph.D. Dissertation, Australian National University. Wilkins, David. 1994ms. “Semantic structuring of syntax or syntactic reflections of semantics: a semantics prior approach to noun phrase ‘coordination’ in Mparntwe Arrente (Arranda: Central Australia)”. SUNY Buffalo.

Chapter 3

Coordination An adaptationist view Jeffrey Heath University of Michigan

1. Introduction 2. Languages 3. Scope delimitation: English and Koyraboro Senni 4. “Coordinating” particles and “scope” delimitation in Nunggubuyu 5. Lexical alternatives to NP coordination: associative plurals 6. Lexical alternatives to NP coordination: kinship collectives and dyadics 7. Progressive pronominal incorporation 8. Pronominal coordination versus inclusive/exclusive 9. Pronominal hierarchies in coordinand sequencing 10. Inclusory left coordinands 11. Ordering asymmetries among nouns 12. Summary



The issues in the analysis of coordination have been defined differently by typologists and formalists, neither formulation being compatible with adaptationism. For the formalists, the central issue is the arboreal representation of coordinated phrases and clauses. Is a coordinated NP base-generated, or produced from a twoclause combination by conjunction reduction? Is the coordinating particle the “head” of an “&P” (or “CoP”) with one coordinand as specifier and the other as complement (e.g. Johannessen 1998)? Or is the first coordinand the head of a phrase (e.g. NP), with an adjoined Boolean phrase containing the particle and the second coordinand (Munn 1999)? An adjoined structure is most obviously present in languages where a comitative adposition functions as ‘and’ particle, as in the Russian inclusory construction (McNally 1993). For typologists, the prevailing Greenbergian survey methodology directs attention to cross-linguistic asymmetries among coordinating particles, particularly


Jeffrey Heath

‘and’ and ‘or’ (conjunction versus disjunction), with particular reference to combinability: NPs only in “with languages”, NPs along with VPs and clauses in “and languages” (Stassen 2000). The rise of grammaticalization theory as the semiofficial diachronic model for typologists has also borne fruit (Mithun 1988). To an adaptationist, neither of these approaches will do. While the typological approach is vaguely functionalist, its survey methodology effectively confines it to the study of those local asymmetries within the coordination system that can be expressed as implicational hierarchies. It therefore fails to pursue connections with other grammatical subsystems, which to an adaptationist represent the main general interest of coordination. Grammaticalization theory is almost totally antistructural and anti-functional, and I disregard it here. Grammars are highly complex adaptive systems, consisting of numerous interlocking subsystems that have co-evolved over long periods of time under continuous pressures of several (partially conflicting) types. Grammars take semantic representations that are hierarchically (rather than linearly) organized, and encode them as linear strings of morphemes. There are several fundamental engineering problems that must be addressed by any grammar. One challenge is how to indicate how each NP fits into the argument structure of a predicator; solutions include fixed linear order, case-marking of NPs, and agreement morphology. Another is how to indicate coindexation of NPs within and across clauses; this is a particularly difficult coding challenge that is partially solved in various languages by switch-reference, infinitives (versus finite clauses), fine-grained noun classes, special anaphoric pronominals, and other mechanisms (Heath 1975). A third, lower-level problem is how to mark the left and right edges of embedded clauses and other extended phrases (scope bounding). And so forth. Grammatical solutions to these challenges are additionally constrained by online cognitive considerations. On the one hand, there is a preference for a rhythmical alternation of foregrounded and backgrounded material. On the other hand, the speaker can rely on listeners to use their powers of inference to flesh out incomplete semantic representations, so long as a critical mass of information is directly expressed. We may think of grammars as having several “functional systems” (for example, referential tracking) that do not correspond one-to-one to the traditional formal systems that we recognirze in reference grammars (e.g. coordination). The lexicon, which co-evolves with the grammar, can also play a role in some functional systems in a way that relieves the grammar of some of the load. Sociolinguistic factors (particularly regarding pronominals) are also relevant. A remarkable range of grammars and lexical systems have evolved under these conditions. Their components are so tightly interwoven that diachronic change (in the absence of language contact) is slow, often involving the formal renewal of pre-existing constructions using new morphemic material.

Coordination: An adaptationist view

In this paper I try to show how coordination (a formal system) relates to these functional systems, drawing selected data from Nunggubuyu, Koyraboro Senni, and English.



Nunggubuyu is a radically nonconfigurational language of southeastern Arnhem Land in northern Australia, which I studied in the field between 1972 and 1976 with some interruptions (Heath 1978, 1982, 1984). A capsule description of the morphosyntax and word typology is Heath (1986). In a radically nonconfigurational language there is no hierarchical phrase structure beyond word level, although words (particularly verbs) may have considerable internal morphological structure. There is no multi-word “NP” or “VP”, and “S” is poorly defined. There is no government by verbs of NP complements (hence no “accusative” or “ergative” case on nouns), no adjectival modification (though loose apposition is possible), no clausal embedding, no indirect speech, no agentives, no imperative category, and no NP quantification. I use the adverb “radically” since the concept of nonconfigurationality has been debased by being applied to languages like Basque and Hungarian where there is some question of the validity of “VP”. Even Warlpiri, supposedly an extreme of nonconfigurationality, has considerably more configurational structure than Nunggubuyu. Koyraboro Senni (=Songhay of Gao) is a Songhay language spoken along the Niger River in northern Mali, West Africa. It has a rigid syntactic structure and weakly developed morphology. Nominal morphology is essentially limited to marking Definite and Plural on the final word in the core NP (which may be followed by a demonstrative, a discourse-functional morpheme like moo ‘too’ or binde ‘Topic’, and/or the quantifier kul ‘all’). Verbal morphology is limited to valency-changing derivation (Causative, two kinds of passives, and UnspecifiedObject), and there is no pronominal agreement. Basic constituent order for canonical transitives like ‘cut’ and ‘eat’ is S-MAN-O-V-X (where X is “everything else”). A smaller set of low-impact transitives like ‘see’, ‘obtain’, and ‘remember’ have a sharply different S-MAN-V-X ordering (O is now contained in X). MAN is a set of mood, aspect, and negation elements. If there is no other nonzero MAN morpheme, a special Transitive morpheme na appears in this slot in a canonical transitive, separating S from O NPs. Koyraboro Senni pronouns occur in the same slots as do full NPs. However, in basic argument positions (subject, object, possessor), several pronouns take reduced cl*tic-like forms that differ from their full form used in independent position. There are several postpositions (dative, locative, etc.). Any NP (including “adverbial”) or PP may be focalized. A focalized subject is followed by a “Subject Focus” morpheme;



Jeffrey Heath

an object NP or a PP is fronted to clause-initial position and followed by a more general “Focus” morpheme. When an object is focalized, its original position is filled by a resumptive pronominal in the appropriate person and number. There is also a way of focalizing the VP using special MAN morphemes. Interclause relations are expressed using either serialization (an “infinitival VP”, without an overt subject or MAN marking, is added to a main clause), or embedded finite clauses (indicative and subjunctive subtypes). For data see Heath (1998a, b; 1999).


Scope delimitation: English and Koyraboro Senni

An and particle, hereafter “&”, makes no denotative sense in the abstract. Even titles such as Bonny and Clyde or Le rouge et le noir are understood as logically embedded: this film/book is about [X&Y]. One of the standard typological concerns regarding coordination is whether &’s are used to link clauses as well as NPs. We should begin by asking whether an overt & is equally useful or necessary in the two cases. In the case of NPs, an overt & is useful since NPs are always embedded at least implicitly in a larger structure, for example as subjects or complements of a verb, as complements of an adposition, or as possessors of another noun: ‘[X&Y] came to dinner’, ‘I saw [X&Y] at the theater’, ‘[X&Y]’ s ship has come in’. In many languages, in the absence of an overt &, the sequence [XY] of two NPs is likely to be construed not as coordination, but rather as some other logical or grammatical relationship (apposition, possessor-possessed, compound, or distinct arguments that happen to be adjacent in the clause). True, some languages reportedly express NP coordination by simple juxtaposition [XY]. I strongly suspect that these languages distinguish [XY] in the sense ‘X&Y’ from non-coordinative relationships. For example, if X and Y are NPs with the same nonpossessive case-marker C, a phrase [X-C Y-C] without appositional intonation might be uniquely interpretable as coordinative. Such a coordinating strategy would not work in a language without case-marking. Koyraboro Senni lacks case marking for subject and object NPs (which are distinguished from each other by linear position), and expresses possession and (loose as well as tight) compounding by NP juxtaposition: [XY] can mean ‘the Y of X’ or can be a compound (even if X and Y are separately marked for definiteness and number). This language is therefore a poor candidate for NP coordination by juxtaposition without an overt & morpheme. NP coordination is in fact expressed in Koyraboro Senni as [X [nda Y]] ‘X and Y’. Like most West African languages, Koyraboro Senni is a “with” language. The preposition nda is also the regular instrumental morpheme (‘by means of ’). A schematic representation of an instrumental construction is [1sg Transitive 3pl hit [nda stick] = ‘I hit them with a stick’. nda can also be comitative, though with motion verbs the tendency is to denote a conveyed entity (human or nonhuman):

Coordination: An adaptationist view

[1sg come [nda X] = ‘I brought X’. In the comitative-like cases, nda frequently fuses with the verb as a suffix, the effect being to transform the verb from intransitive to a low-impact transitive (with postverbal O). While [X [nda Y]] ‘X and Y’ looks like an instance of grammaticalization of an older construction meaning *‘X with Y’ to yield the modern sense ‘X and Y’, there is no evidence that a change of this type has occurred since Proto-Songhay. Both the western and eastern Songhay languages use their reflex of *nda in the same set of constructions (instrumental, comitative, &) as Koyraboro Senni. Postpositional phrases (PP’s) cannot be coordinated with nda. A putative dative ‘for X and for Y’ with constant postposition and different NP’s is expressed as [[X [nda Y]] še], where Dative še takes the conjoined NP in its scope. There is no pressing need to express e.g. ‘[in X] and [for X]’, or ‘[in X] and [for Y]’, with different postpositions, in a single coordinated PP. Directional combinations like ‘[into X] and [out of X]’ or ‘[to X] and [from X]’ have no translation equivalents, since directionality is expressed in Koyraboro Senni by verbs (there are no adpositions meaning ‘into’, ‘out of ’, ‘from’, or allative ‘to’). ‘[Over X] and [under X]’ is expressible within a clause, but not by a coordinated PP; it comes out as, schematically, [[X Locative] above & below], where ‘above’ and ‘below’ are nouns in adverbial function. When we turn from NP (and PP) to S, we again must ask whether an overt & morpheme is useful. One possible function is to specify temporal sequence: [T&U] = ‘T, then U’, where T and U are clauses (here, indicatives denoting bounded eventualities). This could be of some use in languages that have no other routine way of indicating or implying sequencing (switch-reference including relative-tense specification, serialization). However, the & would not really function in this case as a logical coordinator since the coordinated structure is not embedded. Moreover, sequencing can be indicated or implied in other ways, by iconic linearization of clauses, or by use of more specific conjunctions such as ‘then’ or ‘suddenly’. In English and other Standard Average European languages, the full functionality of & in unembedded contexts is brought out in “conjunction reduction” cases, especially where the subject (and some inflection) is omitted in the second clause: I came in [and Ø sang a song]. Here the zeroing of the subject, made possible by use of the coordinating particle, provides the clue necessary for referential tracking. The sequential nuance is secondary, and is inapplicable in progressive counterparts: I was coming in [and Ø singing a song]. English but and or occur in similar constructions. Together, these particles play an important role in cross-clause reference-tracking. In Koyraboro Senni, with its productive serialization system, an & has no role to play in clausal linkage. The serialization system is used not only when the first clause has a specialized control verb (‘be able to’, ‘ought to’, ‘used to’, etc.), but also for stringing narrative clauses together. The second clause, which has the same logical subject as the first clause, is reduced to an infinitival VP beginning with



Jeffrey Heath

Infinitival morpheme ka (allomorph ha optionally before a velar stop), with no other inflection (1). (1) ay ga huru [nee ra], ka do˜ni don 1sg impf enter [here loc], infin song sing ‘I will come in here, and (or: in order to) sing a song.’ Since this construction requires sameness of subjects, the next question is whether an & might be useful in narrative sequencing of indicative clauses with different subjects. It turns out that there is no such compelling need. The relevant clauses can simply be juxtaposed, their linear order implying sequence. Alternatively, the speaker can use gar ‘find, encounter’ as the infinitival verb in a serial construction, since this verb can itself take a fully formed indicative clause as complement. So the subject switch is syntactically independent of the serialization process, and one can get along without an &, as in (2). (2) ay huru, ka gar [a ga don] 1sg enter, infin find [3sg impf sing] ‘I came in, to find that (s)he was singing.’ So much for strings of narrative clauses. However, there remains the issue of whether an & could be useful when the two clauses, T and U, are embedded together in some larger structure. This is where coordination in the core logical sense can be relevant at clause-level in the same way it is at NP-level. Consider, for example, the schematic surface strings in (3). (3) a. want TUV b. if TUV The coding problem is to ensure that the listener understands how far the scope of the governing verb or the logical operator extends in the sequence of clauses TUV. I will use «…» to indicate scope boundaries. Given that the first clause T is always in this scope, we can formulate the problem as how to distinguish between narrow scope «T» and broad scope «TU». Whatever mechanism makes this distinction can presumably be iterated to specify still broader scope «TUV», so I confine the discussion to «T» versus «TU» and take V as outside the scope. In this context, the functionality of English and at clause level becomes clear. In I want you to go and them to stay, the coordinating particle and the reductions it triggers make it clear that want governs both following clauses (with go and stay). In if he comes and I fight with him, I’ll be in trouble, the particle makes it clear that both the come and fight clauses are part of the conditional antecedent; compare if he comes, I’ll fight with him and I’ll be in trouble, which has a different logic (the fight clause is now part of the consequent). Again, but and or have similar scopedemarcating functions in these constructions.

Coordination: An adaptationist view

In cases like (3a), many languages distinguish «T» from «TU» scope without an &, because each clause embedded as complement to a verb like ‘want’ is in subordinated form (infinitive, subjunctive, irrealis, etc.). For example, with Subju[nctive] marking, a surface string [… want T-Subju UV] is decoded with «T» scope, while [… want T-Subju U-Subju V] is decoded with «TU» scope. To be sure, in some combinations there is a possible ambiguity between coordinative (single-level) and stacked (double) clause embeddings, but delimiting the right edge of the scope of ‘want’ is a far more pressing engineering problem. In cases like (3b), the scope issue is more serious, since most languages phrase both antecedent and consequent clauses as ordinary indicative clauses. I focus on hypothetical (rather than counterfactual) conditionals, and assume for present purposes that the antecedent clause precedes the consequent. There is normally an ‘if ’ morpheme at the beginning of the antecedent, so the left edge of the antecedent is marked. The key coding issue is how to mark the antecedent’s right edge (or, equivalently, the left edge of the consequent). One important clue to the antecedent/consequent boundary is aspect, since many (perhaps most) languages have strong tendencies to use different aspects in the antecedent and consequent of hypothetical conditionals. English uses the simple (nonprogressive) present in the antecedent, while many other languages use a perfective. English uses a future in the consequent, while many other languages use an all-purpose imperfective. However, aspect switches are rarely reliable as boundary markers in conditionals. If the logical relationship between antecedent and consequent is one of forensic inference rather than causation, the consequent may be in the ordinary form for past-time events: if there’s blood on his clothing, then he killed them. Likewise, if the antecedent clause denotes gnomic eventualities, it may appear in an imperfective form even in a language that normally has perfective antecedents. Moreover, aspectual shifts characteristic of hypothetical conditionals may be irrelevant to counterfactuals. An overt & can therefore be quite useful in marking scope boundaries in conditionals. Still, there are many “with” languages that express NP coordination with an ‘and, with’ morpheme that is not used in clausal coordination, and Koyraboro Senni is one of these. I sketched above how Koyraboro Senni takes care of unembedded clause sequences, for same and different subjects. Scope demarcation in the context (3a) is unproblematic. If a verb like ‘want’ that takes subjunctive complements has broad scope over two clauses T and U, either U is attached as an infinitival VP to T in the serial construction (this assumes that T and U have the same subject), or both T and U appear in subjunctive form (if T and U have different subjects). In the latter case, if T also happens to contain a subjunctive trigger there might be a choice between stacked and parallel interpretations, i.e. between ‘[…want [TU]]’ and ‘[…want[T[U]]]’, but in practice this is not a major problem.



Jeffrey Heath

The more serious coding issue for languages without a clause-level & is how the scope of an ‘if ’ operator is specified. Ironically, in Koyraboro Senni the ‘if ’ particle is none other than nda, identical to the & particle and instrumental-comitative preposition (compare and in the sense ‘if ’ in Early Modern English, e.g. in Shakespeare). At the beginning of a clause, ‘if ’ is the only sense nda can have. Indeed, avoidance of clause-initial nda in other functions seems to be a powerful factor in the syntax. In Koyraboro Senni, adpositional phrases are among the constituents eligible for focalization, in which case they are extracted from their normal postverbal location and positioned clause-initially before the subject, with optional Focus morpheme no. All relevant adpositions, except nda, are postpositions. Consider unfocalized (4a) and focalized (4b). (4) a.

ay n-a tee [har-oo še] 1sg trans-3sg do [ dat] ‘I did it for the man.’ b. [har-oo še] (no) ay n-a tee Ø [ dat] (foc) 1sg trans-3sg do Ø ‘It was for the man [focus] that I did it.’

Since postpositional phrases are extracted as wholes, one would expect a postverbal instrumental phrase like nda bundu ‘by means of a stick’ to be extractable in the same manner. But when this is done, the normally prepositional nda flip-flops with its complement NP and becomes a postposition (5b). This conveniently prevents processing conflicts between focalized Instrumental nda, and clause-initial nda ‘if ’ in conditional function. (5) a.

ay n-a kar [nda bundu] 1sg trans-3sg hit [with stick] ‘I hit it/him/her with a stick.’ b. [bundu nda] ay n-a kar Ø [stick with] 1sg trans-3sg hit Ø ‘It is with a stick that I hit it.’

In hypothetical conditionals, the unmarked aspects are perfective for the antecedent and imperfective for the consequent, but neither correlation is rigorous for reasons given earlier. The left edge of the antecedent is reliably marked by nda. Interestingly, the right edge is often (though not obligatorily) marked by kul, which occurs elsewhere as the universal quantifier ‘all’ (6). (6) nda a kaa kul, ay ga diy-aa if 3sg come all, 1sg impf see-3sg ‘If he comes, I will see him.’

Coordination: An adaptationist view

Right-edge marking kul may be uttered at the end of the last intonational group of the antecedent, and this is normal when the entire sequence is delivered fluently. However, it may also be delayed, especially when there is a noticeable pause between antecedent and consequent, and in this case it merges intonationally with the onset of the consequent. In effect, speakers may pause to reflect whether a further elaboration of the antecedent is needed, or if the consequent can begin. The antecedent may consist of more than one clause, strung together with no coordinating particle (or repetition of ‘if ’). Since the left edge of the antecedent is marked by nda ‘if ’ and the right edge can be marked by kul, speakers have the resources to avoid ambiguity as to the break point between antecedent and consequent. Moreover, if the antecedent consists of two eventualities that can be verbalized as clauses with the same subject NP, the second can be expressed as a subjectless infinitival VP (a form also used in true serial-verb constructions). An infinitival VP following an ‘if ’ clause, as in (7), can only be interpreted as an elaboration of the antecedent, since the consequent must begin with a complete finite clause. (7) nda a koy [ha goro] … if 3sg go [infin sit] ‘if she goes and sits …’ Given this machinery (right-edge marker, infinitivals), Koyraboro Senni has no need for an overt clause-level &, even in multi-clause conditional antecedents where & is vitally needed in English and many other languages. In formal analyses of coordination, the fact that an & can be used with both NPs and clauses in some languages, but with NPs only in other languages, is likely to be interpreted in essentialistic terms. The & particles (it is thought) may differ from language to language in some way (e.g. ‘and’ versus ‘with’) that is brought out by their differential combinatory potential. Or (it is thought) the particles are the same, but languages differ as to whether S and NP are comparable objects. Studying how coordinating particles combine with NPs and S’s is, in this view, a kind of chemical experiment, whose results allow us to infer something about the hidden (“deep”) structure of the atomic entities. In the adaptationist approach advocated here, we study how the grammar handles universal engineering problems such as scope demarcation, where the need for coordinating particles depends critically on what other resources can be deployed. By observing the tight functional integration of particles with other formal features, we begin to appreciate how grammatical “types” take shape.

4. “Coordinating” particles and “scope” delimitation in Nunggubuyu Since Nunggubuyu lacks hierarchical phrase structure beyond word level, it has no [X&Y] construction, indeed no multi-word syntactic phrase of any kind lower than



Jeffrey Heath

S (=clause). Even S is an ill-defined concept, since Nunggubuyu utterances that translate as English sentences may have one, two, or zero words carrying tenseaspect inflection. Nevertheless, translatability from Nunggubuyu to English is reasonably good, since the basic content of encoded messages is the same, and since Nunggubuyu has ways of expressing the messages we encode using possessives, relative clauses, indirect speech, conditionals, universal quantification, and other familiar constructions. The morphology of Nunggubuyu is extremely rich. Basic word classes are noun (marked for noun class, human number, and spatial cases) and verb (marked for subject and object pronominal agreement, in addition to tense and other inflectional categories). “Adjectives” are a subset of nouns that, in addition to normal nominal function, can also be used predicatively with a scaled-down verblike inflection. There are also many uninflectable particles, functioning as discourse particles and the like, glossable roughly as ‘only’, ‘again’, and the like. One particle, mari, is frequently translatable as ‘and’. However, given the lack of phrase structure in Nunggubuyu, any gloss that evokes an English syntactic frame is misleading. There is, for example, no constraint on what type of word mari can co-occur with. One could argue that a more revealing gloss is ‘moreover’ (in the discourse-referential sense, i.e., ‘I hereby add …’). The particle is usually uttered in an intonational group with a following element F, hence [mari F] ‘and (moreover) F’. When two elements E and F are “conjoined”, the sequence is something like [E, … mari F], with a pause (and perhaps other words) intervening between E and the particle. Occasionally mari is postposed, giving a sense of emphatic finality, hence [F mari]. With referential nouns, the sense of mari is close to ‘and (moreover)’ in the additive sense, though there is no sharp boundary between ‘and’ and ‘with’. Usually when both referents are expressed as nouns, they are separated intonationally from adjoining material, as a topical expression or an afterthought clarification. Here is the beginning of a story about two mythical birds: “now they-FeDu were sitting, at (the place) wara\i\ Ÿ at the head of a river, Emu-FeSg mari Brolga-FeSg”. Here the “coordinated NP” is a delayed clarification of the 3FeDu pronominal on the verb ‘sit’. Such [E, mari F] combinations involving two nouns are infrequent in the texts. In (8), we have an idiomatic sequence that would call for an and in English, but is expressed in Nunggubuyu by simple juxtaposition: (8) niri-ya˜rijgi˜-ni, a-maram-ba˜ Ÿ a-mu˜n-ba˜ 1expl>nc-transport-infl nc-hand-dual nc-foot-dual ‘we lugged it (=lumber), with both hands and both feet’ Nunggubuyu speakers can also express coordination indirectly, by specifying one component referent as an independent noun, in the proximity of an inclusory (or summative) pronominal, such as the exclusive 3+ plural subject prefix on the verb

Coordination: An adaptationist view

in (9). In examples like this, the difference between conjunctive and comitative function is blurry. (9) nu˜-ya-]gi, wara-mij-bura˜yu] we(expl)-go-infl, pl-pl-children ‘we (=men) went, along with (our) children.’ mari has even lower text frequency “conjoining” verbs. It is perfectly normal to string clauses together in narrative without conjunctions, and mari is merely one among several discourse particles that can be used to foreground or otherwise emphasize an action verb. (The most common such particle is adaba Ÿ or aba ‘now, then’; others include araga ‘suddenly’ and ]ija] ‘again, furthermore’.) Consider (10) and (11). (10) aba an-uwa-ga˜-la ni-garaji-\ ni-ga now from.there 3masg-jump-infl he mana˜-Ÿl]a-wala, mari gudbij! Ÿ ]u-biÛl arga-\ Ÿ nc-mud-abl, and seize! 3masg>3fesg-seize-infl ‘Then he jumped up, him, out of the mud, and he seized her (=Jabiru).’ In (10), mari is used before a “root form”, an uninflectable interjection with verblike sense, often used (as here) in conjunction with an inflectable verb of the same meaning. The narrative punch can be conveyed (roughly) by modifying the gloss to ‘he jumped up …, and pow! he punched her’, though Nunggubuyu root forms are not onomatopoeic. The man had been hiding in ambush, and ‘he seized her’ is the dramatic high point. (11) wini-ja]gal-]ambi-\ mari ]a wa]i˜-mbiŸldÛ i-\ 3madu-leg-be.immersed-infl and and 3fesg>3pl-strike-infl ‘They (=boys) had their legs underwater (due to an abrupt flood), and (suddenly) she (=Python) struck them.’ In (11), ‘she struck them’ is again a culminating dramatic event. Here mari is combined, as often in such contexts, with another particle ]a that can also be glossed ‘and’ (when they are juxtaposed the order is usually mari ]a but occasionally ]a mari). While the core sense of mari is essentially additive, ]a brings out the continuity between two events, as when a backgrounded verb in continuous aspect is followed by a punctual verb denoting a foregrounded event. This is a very common feature of Nunggubuyu narrative, the continuous verb often ending in a near falsetto with protracted final syllable (12). (12) niwa˜-buŸlu-Ûl arma-]i-wugi˜˜˜j ]a ba-gu ]u-mi-\ 3masg>nc-mud-follow-infl-still and there 3masg>3fesg-get-infl ‘He kept tracking [continuous] the muddy water, and then he caught [punctual] her (=Python).’



Jeffrey Heath

]a can also “conjoin” nouns, or more commonly deictics. However, the sequence [E ]a F] is not additive ‘E and F’, rather connective ‘(the distance) from E to F’. The speaker typically points to objects or positions in the here-and-now space to indicate the distance between entities in the narrative (13). (13) ma-Ûl a]a\-]al]ala-]i ]u\ju y-a˜-ni ]a o˜-wa˜-ni nc-guts-gleam-infl like and ‘Its (=Snake’s) entrails were gleaming, (extending) like (the distance) from here to over there.’ The discussion so far has not considered scope demarcation issues. Indeed, “scope” is a meaningless syntactic context in a radically nonconfigurational language. But Nunggubuyu does have functional equivalents to conditionals and other familiar constructions, and therefore must tackle the same engineering problems described earlier. Indeed, the two greatest hurdles for a radically nonconfigurational language are negation and conditionals. For everything else, it is relatively easy to find a nonsyntactic alternative to an English syntactic phrase. For example, possession can be expressed by an appositional construction with the “possessor” marked by a Relative suffix (which is also added to verbs to form relative “clauses”). Scopebearing quantifiers are absent, but one can express e.g. ‘they went in all directions’ by a complex paraphrase ending with a demonstrative word roughly glossable as ‘etcetera’: ‘they went this way [pointing], they went that way [pointing], they went that way [pointing], etcetera’. But negation is a real problem. yagi and ari Ÿ are the basic negative elements. yagi is an uninflectable particle, expressing negation in non-actual modal categories (future negative including prohibitive, or past irrealis negative). When used by itself, with interjection intonation, yagi can be glossed ‘don’t!’ (or ‘may it not happen!’). ariŸ is really a predicative existential or locational “adjective” that requires a pronominal subject, as in ni˜-ri Ÿ ‘he is absent, he is not (here, there)’. However, its default form wa˜-ri Ÿ ‘it is absent’ (with nonhuman subject marker) can function abstractly as a “clausal” negative. The coding problem is how to indicate where, in an ensuing sequence of nouns and verbs not organized into well-defined clauses, the “scope” of the negation ends and the discourse reverts to positive polarity. (The negative word normally precedes everything in its “scope”, but when a negative sequence is repeated we can get stylistic inversion whereby the negative word comes last.) To resolve “scope” ambiguity, Nunggubuyu morphology has evolved a unique tagging system that has the effect of indexing those nouns (including “adverbs”) and verbs that occur in the scope of a negative word — or, more generally, in an irrealis modal space. Specifically, an irrealis space requires certain verbal inflection forms, and a particular series of noun-class prefixes on nouns. Some of these forms are not strictly limited to irrealis spaces, but in practice the tagging system does

Coordination: An adaptationist view

indicate where the irrealis space ends. As a result, even a post-negation sequence containing two (or more) verbs can be indexed as forming a single irrealis space created by the negation (14). (14) yagi an-u-guni wu˜-rumi-Ø Ÿ neg there 3pl-go-infl ana˜-rijbiŸla wu˜-wurÛ Ÿ di-Ø nc-[place name] 3pl-be.buried-infl ‘They cannot go there to Arijbirla and be buried (there).’ There is no basis for assigning the place name ana˜-rijbiŸla exclusively to the “clause” on the left or to the “clause” on the right, but this raises no truth-conditional problem. In the positive counterpart of (14), both the demonstrative adverb ‘there’ (yu˜-guni) and the place name arijbiŸla would normally occur without prefixes. The use of ana- (a nonhuman noun-class prefix) with adverbs and place names is one of the tagging devices indexing an irrealis space. The two verbs in (14) are likewise tagged; compare e.g. positive wu˜-ruma-] Ÿ ‘they will (=can) go’. Because of this tagging, we know that the irrealis space begun by yagi continues through wu˜-wurÛ Ÿ di-Ø, even without repeating yagi or adding an explicit right-edge marker. At the first sign of an untagged noun, adverb, or verb, the listener deduces that the irrealis space has ended. The same tagging applies to conditional antecedents (positive as well as negative), which involve a cl*tic -maji˜ ‘if ’ that attaches to a verb (or other word). Again, at the point where the first untagged noun, adverb, or verb occurs, the listener knows that the speaker has finished the antecedent and has begun the consequent. Nunggubuyu therefore has a remarkable tagging system that compensates for the absence of other indicators of clausal boundaries. In spite of the dramatic grammatical differences between Koyraboro Senni and Nunggubuyu, each has evolved a scope-demarcating strategy (e.g. for conditional antecedents) that does not rely on &. This distinguishes these two languages jointly from others like English, where & plays a key role in scope demarcation, as seen again in the free gloss of (14).


Lexical alternatives to NP coordination: Associative plurals

Overt coordinated NPs compete with single-word, referentially plural or collective nouns. This is trivial in the case of wide-ranging collectives (English people, the French, cattle), where enumeration of individuals is not feasible. However, when dealing with pairs of individuals and other small sets a speaker clearly has choices, e.g. between my Dad and Mom and my parents, and there are some differences among languages in this respect.



Jeffrey Heath

Both Nunggubuyu and Koyraboro Senni allow a plural suffix to be added to a referentially singular noun, such as a personal name, to denote a group centered on that individual. In Nunggubuyu, it is not the usual plural marking, rather a special plural suffix, as in na-ma˜di-wa] ‘Maadi and his bunch (family or clan)’ (with MaSg prefix na-, hence logically bracketed [[na-ma˜di]-wa]]). The only other use of plural -wa] is to pluralize kin terms with first person propositus (‘my/our …-s’). In Koyraboro Senni, the relevant suffix is the normal indefinite plural suffix -yan (or -ya]). In Koyraboro Senni, a bare noun is interpreted as indefinite singular, and forms an indefinite plural by adding plural -yan. Example: bo]-koyni ‘(a) chief ’, bo]-koyni-yan ‘(some) chiefs’. Contrast bo]-koyn-oo ‘the chief ’ (DefSg suffix -oo) and bo]-koyn-ey ‘the chiefs’ (DefPl suffix -ey). Since bo]-koyn-oo ‘the chief ’, in ordinary contexts, denotes a unique individual (e.g. the chief of the local village), one can add -yan to it to produce the associative plural (=individualcentered group) reading: [bo]-koyn-oo]-yan ‘the chief and his bunch’ (note the bracketing). Since this is distinct from bo]-koyn-ey ‘the chiefs’, and since DefSg -oo cannot otherwise be followed by -yan, no ambiguity is possible.

6. Lexical alternatives to NP coordination:

Kinship collectives and dyadics The functions of & particles are also affected by language-specific lexical patterns. In particular, languages differ in whether pairs and other sets of kin types are expressed as coordinated NPs or as monolexemic plurals and collectives. In English, plurals like folks/parents, grandparents, kids/children (instead of …sons and daughters), grandchildren, and in-laws can be used instead of lists, but they constitute a limited subsystem (no monolexemic ‘brother and sister’, ‘aunt and uncle’, or ‘nephew and niece’ term, excluding the somewhat literary siblings). Plurals like fathers are generally used only for multiple propositus (egos), as in our (respective) fathers. English is also very weak in specialized dyadic kin terms (which indicate the relationship between the referents). Only couple in the sense ‘married pair’ is common, though cousins can be used dyadically since cousin is self-reciprocal. As a result, English has many high-frequency coordinated NPs such as brothers and sisters and (dyadic) mother and child, often with fixed order (Cooper & Ross 1975). These routinized (lexicalized) coordinated expressions are very typical of English. By contrast, in Nunggubuyu and many other Australian languages, any kin term can be pluralized (‘fathers’, ‘mothers’, etc.) even with a fixed propositus, since each singular term is applicable to a range of near and distant kintypes. (Everyone is “related” to everyone else through one of the kin categories.) Moreover, plurals may commingle male and female types, so e.g. na-mi-ni-\ara ‘your fathers’

Coordination: An adaptationist view

subsumes the addressee’s (real) father, father’s brothers, father’s sisters, and many distant classificatory “fathers”. Australian languages also have well-developed sets of dyadic terms, whereby any reciprocal set such as ‘father and child’ or ‘mother’s brother and sister’s child’ can be expressed as a single noun. There is much irregularity and suppletion, but the basic pattern is to add a Dyadic suffix to the term denoting the more senior member of the pair, e.g. awan-\ij Ÿ (father-Dyadic) ‘father and child pair’. Dyadic terms normally appear at the first introduction of the two referents into the discourse, obviating the need for a coordinating particle. The development of monolexemic kinship expressions denoting pairs or sets is expectable in Nunggubuyu, which has no hierarchical phrase structure and therefore cannot lexicalize high-frequency coordinated NPs. It illustrates the mutual adaptation of grammatical and lexical systems.


Progressive pronominal incorporation

Examples (8–9), above, give a taste of what Nunggubuyu speakers can do with their rich pronominal system, which distinguishes first exclusive (1Ex), first inclusive (1In), second (2), and third (3) persons. Each person further distinguishes Sg, Du, and (3+) Pl numbers (bump the number values up one for 1In), though 3Pl is generally limited to human referents. Gender (Ma, Fe) is distinguished in all structurally Du forms, and in the (human) 3Sg. For nonhumans, Pl is not usually expressed, but there are several nonhuman noun classes. These pronominal categories are expressed not only in independent pronouns (which occur chiefly in topical, emphatic, and focusing functions) but also in pronominal agreement prefixes on the verb (subject, object), though in transitive prefix combinations there is some syncretism and reduction. Of particular relevance to “coordination” is the rich set of nonsingular forms. Example (9) was a fragment of a larger text segment given more fully as (15). (15) nu˜-ya-]gi, wara-mij-bura˜yu], 1expl-go-infl pl-pl-children F ni-ni Ÿ mari G nu-ru, nuru-gu-gubaÛda-]i F we( and G we(, ‘We (=men) went, along with (our) children, F and G and me, we went walkabout.’ As usual, the sequence glossed ‘F and G and me’ can be construed either as a delayed specification of (the adult portion of) the subject of the preceding verb ‘went’, or as an anticipatory (“topical”) specification of the subject of the following verb (‘go walkabout’). Since the three men are in fact included in the ‘we’ of ‘we went’, and are the subject of ‘we went walkabout’, no truth-conditional issue is in play.



Jeffrey Heath

The gloss ‘F and G and me’, denoting the adult men (F and G are personal names), corresponds to a Nunggubuyu sequence ‘F 1Ex.Ma.Du (=the two of us), mari G 1Ex.Pl (=the 3+ plurality of us)’. Instead of a running list of the type ‘F&G&me’, the Nunggubuyu sequence is built up piece by piece, the pronominal sum being recomputed at each point (‘F we-Du, moreover G we-Pl’).

8. Pronominal coordination versus inclusive/exclusive The pattern of high-frequency coordinated NPs in English extends to pronominal combinations like you and me and me and him. A language with such phrases has little need for an inclusive/exclusive distinction in the 1st person nonsingular pronouns. A possible problem for a ‘you and me’ language as opposed to an inclusive/ exclusive language is that an overt coordination would be clumsy in hortatives — the most common context for inclusives; a sequence ‘may you and I/me go’ is awkwardly lengthy in such a high-frequency utterance. But as long as hortative mood is marked, a simple ‘we’ pronoun can be understood as inclusive (let’s go!), since exclusive hortatives are rare. So a language with a single 1Pl pronoun (‘we/us’), a high-frequency ‘you and me’ coordination (used in non-hortative utterances), and an explicit hortative construction (let’s), is adaptively viable. Koyraboro Senni basically follows English: 1Pl pronoun iri (or ir), fixed pronominal coordinations including agey nda ni ‘I and you(sg)’, and a subjunctive mood that is interpreted as hortative with 1Pl subject and no external embedding frame (ir ma koy ‘let’s go!’ with Subjunctive ma and verb koy). Another adaptively viable type is one with an inclusive/exclusive opposition and either a dedicated hortative or a more general irrealis or future inflection. In Nunggubuyu, there are no routinized coordinations of any kind (NP or pronominal). (15) above shows how third-person referents are consolidated into nonsingular pronouns including the speaker and/or addressee. Since there is a set of inclusive pronominal categories (1In.Du, optional 1In.Ma.Tr[ial] and 1In.Fe.Tr, and 1In.Pl for 3+ referents), there is never a need for a ‘you and me’ coordination. Indeed, Nunggubuyu tolerates partial phonological merger of exclusive and second persons (nuru-ya-]gi ‘we-Ex.Pl went’ or ‘you(pl) went’), but consistently distinguishes the inclusive from the exclusive in verbal agreement (]uru-ya-]gi ‘we-In.Pl went’). The nonpast irrealis (translatable elsewhere as future, capacitative, or imperative) is normally interpreted as hortative in combination with an inclusive pronominal subject: na\-ja˜-ri˜ ‘let’s-In.Du go!’.

Coordination: An adaptationist view

9. Pronominal hierarchies in coordinand sequencing The routinization of ‘you and me’ and similar combinations can lead to rigidification of the internal ordering of coordinands. Modern (American) English is actually not a very good example of this, since the whole system of pronominal coordinations has been destabilized by the cumulative efforts of thousands of schoolmarms correcting their wards’ wayward case-marking (from e.g. you and me are going to you and I…). In Koyraboro Senni, the ordering of pronominal coordinands is not completely rigid, but there are strong hierarchical tendencies that can be summarized as 1st>2nd>3rd (I leave logophorics aside). Thus agey nda ni ‘I and you-sg’ is more common than ni nda agey ‘you-sg and I’, though the latter is possible. The rest of the hierarchy can be seen in the predominance of ni nd-a˜ ‘you-sg and him/her’ over ]ga nda ni ‘he/she and you-sg’, and of agey nd-a˜ ‘I and he/she’ over ]ga nda agey ‘he/she and I’. A hierarchy 1st>2nd>3rd is similar to the 1st>2nd>3rd, 2nd>1st>3rd, and {1st, 2nd}>3rd linearization hierarchies that we observe in direct-inverse systems. Such systems occur in various languages as an organizational strategy for transitive pronominal agreement morphology (cl*tics or verbal affixes). Nunggubuyu pronominal agreement is such a direct-inverse system, with a primary hierarchy {1st, 2nd} > human 3pl > human 3sg > nonhuman, and less systematic ordering rules (along with portmanteaus and other suppletion) for subject–object pairs within a single hierarchical set (1sg on 2sg, 3pl on 3pl, etc.). Thus ]a-wu- for 1sg subject acting on nonhuman ANA class object, but /]a-N-wu-/ > ]a-]-gu- for ANA subject acting on 1sg object. Examples are ]a-wu-na-\ ‘I saw it’ and ]a-]-gu-na-\ ‘it saw me’ (-na-\ ‘see-past’). The 1sg morpheme ]a- always precedes the ANA class morpheme -wu- (pronominal hierarchy), but the presence or absence of Inverse -N- (derived from a Proto-Australian Accusative suffix) distinguishes the two combinations. Roughly similar direct-inverse systems occur in Algonquian and some Siberian languages (e.g. Chukchi). The preference for placing a {1st, 2nd} pronominal to the left of a non-participant pronominal, seen both in Koyraboro Senni coordination and Nunggubuyu agreement, is therefore not isolated. Because these examples involve linear placement, they probably have a somewhat different dynamic than the superficially similar {1st, 2nd} versus 3rd person opposition observed in splitergative case marking (Silverstein 1976), where markedness relations among basic grammatical relations are skewed by the frequency of 1st/2nd person subjects and of 3rd person objects. Algonquian is especially interesting since its 2nd>1st>3rd rule applies to the single prefix slot on verbs, while the remaining pronominal material and the Inverse morpheme (a reflex of *-ekw-) occur in the suffix complex. Clearly in Algonquian, but probably also in Nunggubuyu agreement and in



Jeffrey Heath

Koyraboro Senni coordination, the 1st/2nd person morpheme in leftmost position functions as a kind of appetizer, letting the listener know that a speech-act participant is involved, the relational details coming later. Calling the leftmost pronominal a “grammaticalized topic”, as in some literature, therefore has a grain of truth. However, the constructions mentioned are distinct from true synchronic topicalization processes, and none of them originated by compression of reconstructible topicalized phrases into morphology.

10. Inclusory left coordinands In some dialects of Koyraboro Senni, a pronominal left coordinand is inclusory and therefore always plural. In these dialects, instead of agey nda ni ‘I and you-sg’ we get ir nda ni (or variant), literally ‘we and you(sg)’, even when only two individuals are involved. Likewise war nd-a˜ lit. ‘you(pl) and him/her’ and ir nd-a˜ ‘we and him/her’, which can denote a set of two (as well as three) individuals. I know of no other instance of this inclusory pattern in left coordinand in other Songhay languages. I therefore take it to be a dialectal innovation in Koyraboro Senni. I believe that the innovation was not instigated by random typological drift, but rather that it was catalysed by language-specific phonological mergers. The full set of pronominals that can appear as left coordinands in the non-inclusory dialects are those in (16). These are “full” (=independent), as opposed to procl*tic, forms of these pronouns, as required of left coordinands. (16) person 1st 2nd 3rd

sg agey ni ]ga

pl iri (or ir) war ]gey (or ]gii)

The probable locus of the innovation was the 3rd person. The & morpheme nda begins with a cluster, so the combination ]gey nda has an ey diphthong in a superheavy syllable. In Koyraboro Senni, the short diphthong transcribed ey does not contrast with ay. Instead, there is a single diphthong that is pronounced [aj] phrase-initially (as in 1sg procl*tic ay) and after h (hay ‘give birth’), elsewhere as phonetic [7j] (halfway between [ej] and [aj]), as in 3Pl independent pronoun ]gey []g7j]. Some orthographies for Koyraboro Senni transcribe “ay” in all these positions, e.g. 3pl “ngay”. In a superheavy syllable, ey may lose its semivowel and monophthongize. The 3Pl combination ]gey nda can then appear as []g7 nda], which is perilously close phonetically to 3sg ]ga nda. I believe that 3sg and 3pl fell together at least partially in this combination. It is true that there is a dialectal 3pl variant ]gii that could remain clearly distinct from 3sg ]ga in left coordinand position. But this ]gii is

Coordination: An adaptationist view

another dialectal Koyraboro Senni innovation that probably originated in left coordinand position as an alternative monophthongization of ]gey. At one stage, ]gey nda fluctuated between []g7 nda] (not easily distinguishable from 3sg ]ga nda) and ]gi nda. The next stage, in some dialects, was a merger between 3sg and 3pl, the variant ]gi nda tending to generalize. Once the singular-plural merger was completed in the 3rd person, it began spreading to the first and second persons as left coordinands. The data in this and the preceding section show how coordination can intersect with strategies of “front-loading” topical or inclusory pronominals in leftmost position.

11. Ordering asymmetries among nouns Other left-right asymmetries in coordination are the subject of Cooper & Ross (1975), whose real title is “World order” although it is often mis-cited as “Word order” (as it was in the table of contents of the volume in which it appeared!). This is a seminal work in several contexts and is a must read. It seeks to explain why high-frequency coordinated NPs often have a fixed order: up and down, brothers and sisters, dribs and drabs. Some of these “freezes” are best explained in semantic terms — not standard quantificational semantics, rather the primordial conceptual semantics of gender, spatial orientation, and other domains. As the authors show, an explanatory model must delve into these domains, exploring the relationships among each set of polar categories, and motivating an asymmetry between them that is appropriately expressed in a surface linear ordering relationship. Are the asymmetries universal, or culture-specific? Are they unconscious, or are they subject to pragmatic manipulation? Topics of considerable moment, but perhaps too hot to handle in the current climate, and even cursory citations of this paper are largely absent from the typological literature. Cooper & Ross also devote considerable attention to nonsense freezes like dribs and drabs, pingpong, riffraff, and so forth, which include frozen coordinated phrases as well as compounds. Connecting them with cases like bread and butter, where a cultural semantic basis for the asymmetry seems unlikely, the authors argue that the ordering in such sequences is controlled by a phonological heaviness metric involving prosody (stress, syllabic structure) and a hierarchy of vocalic features that applies in a pure form when the prosody is neutral. They suggest that [i] is at one end of the hierarchy, favored in the left coordinand (or compound initial), with [u] at the opposite end, favored in the rightmost element. This analysis connects with another long-forgotten research tradition, the work on vocalic sound symbolism by Otto Jespersen, Edward Sapir, Stanley Newman, and many other scholars, mainly in the first half of the 20th century. Sapir and Newman ran experiments with



Jeffrey Heath

English and other (Chinese, American Indian) subjects concerning appropriate names for imaginary new animals, ranging in size from humongous [hjum%]g6s] to teensy-weensy [ti˜nsiwi˜nsi], where sound-symbolic and physical heaviness intersect. Basically, they found that [i] was strongly and universally associated with small size, and that the opposite pole, while less precisely defined, was somewhere in the area of [o] or [f], with other vowels such as [æ] intermediate. Cooper & Ross suggested [u] as the polar opposite of [i], but the data point rather to a pole around [f]. We may add that the lightness of [i] is enhanced by lengthening and repetition, and by juxtaposing [w] to [i] (teensy-weensy, and parallel forms in many other languages, e.g. Nunggubuyu wirig ‘small’ and more emphatically diminutive wi\ig). There is no enhancement of the less well-defined [f] pole that has comparable power, but brevity, stress, and a following voiceless stop seem to have some effect. Indeed, the import of vocalic sound-size symbolism extends into Labovian sociolinguistics, most of which concerns correlations between vocalic articulations and social categories (sex, age, class), lamely explained as driven by “social influence” of some unexplained type. A light-heavy opposition has obvious connections to sex and age; consider, for example, the social significance of pitch differences (which interact both physically and subjectively with vowel-quality differences). An interpretation of Labovian vocalic cycles, emphasizing the lead role of female speakers in the recurrent [f > " > æ > 7 > e > i] progression, is given in Gordon & Heath (1998). The cross-linguistic study of these phenomena initiated within the Cooper & Ross paper itself (they argued that their principles worked for several foreign languages, but not for Hindi!) is well worth reviving and extending. Koyraboro Senni and Nunggubuyu happen to have too few recorded instances (Koyraboro Senni verb \uumi\aama ‘swarm’ is one example) for a comparative discussion, but there are other languages with abundant freeze-like formations. Like many topics I have raised, this one falls through the cracks among standard research topics, since it involves both coordinated NPs (dribs and drabs) and single-word compounds (riffraff), and it quickly leads us into currently unfashionable topics like cultural semantics and sound-size symbolism. I suggest, then, that an important and interesting aspect of coordination needs to be approached from a perspective centered in sound-size symbolism, following the trail that leads from there to (among other things) coordination typology.

12. Summary Coordination is one of many topics that has a section devoted to it in just about any reference grammar, so it lends itself to the standard typological survey methodology. This can produce an inventory of variable features, and some implicational

Coordination: An adaptationist view

results internal to the domain. To me, the resulting publications are of slender interest, except for students getting an initial exposure to “what’s out there”. In other empirical human sciences, such as social anthropology, the age of correlationist cross-cultural surveys has long since given way to a more penetrating, holistic study of social dynamics, with an increasingly historical scope and a broadly adaptationist rather than structural interpretive framework. Linguistics, “functionalist” as well as formalist, lags far behind in this respect. To appreciate how coordination functions, we need to visualize it as located in a space where several semantic and functional subsystems intersect, the details being often language-specific. These include two subsystems about which I have said nothing: (a) the logical properties of particles (‘and’, ‘or’) in the context of a modeltheoretic or similar semantics; and (b) a flexible production system that makes allowance for afterthought additions and open-ended lists. They also include several topics we have considered: (c) the presence or absence of hierarchical phrase structure (configurationality); (d) backgrounding and foregrounding of clauses in narrative; (e) scope demarcation for external operators such as ‘if ’ and negation (a particularly significant connection); (f) mechanisms by which clauses or VPs are strung together or embedded; (g) front-loading of speech-act participant pronominals in leftmost position; (h) prosodic and vocalic “heaviness”; (i) the system of pronominal categories; (j) the modal system with particular reference to hortatives; (k) dyadic and collective lexical items; and (l) cultural symbolism (with a basis in universal human experience) as a source of asymmetries between polar lexical items (like ‘up’ and ‘down’). And to understand how the grammar of coordination evolves over time, we need to consider not only this complex “force field”, but also the fortuitous, language-specific phonetic mergers and similarities that may catalyze change.

Abbreviations ex infin man

exclusive infinitive mood-aspect-negation

nc Subju X>Y

noun class subjunctive X acts on Y

References Cooper, William E. and Ross, John Robert. 1975. “World order.” In: Papers from the Parasession on Functionalism, 47–62. Chicago: Chicago Linguistic Society. Gordon, Matthew and Heath, Jeffrey. 1998. “Sex, sound symbolism, and sociolinguistics.” Current Anthropology 39: 421–49. Heath, Jeffrey. 1975. Some functional relationships in grammar. Language 51: 89–104.



Jeffrey Heath

Heath, Jeffrey. 1978. Nunggubuyu Myths and Ethnographic Texts. Canberra: Australian Institute of Aboriginal Studies. Heath, Jeffrey. 1982. Nunggubuyu Dictionary. Canberra: Australian Institute of Aboriginal Studies. Heath, Jeffrey. 1984. Functional Grammar of Nunggubuyu. Canberra: Australian Institute of Aboriginal Studies. Heath, Jeffrey. 1986. “Syntactic and lexical aspects of nonconfigurationality in Nunggubuyu (Australia).” Natural Language and Linguistic Theory 4: 375–408. Heath, Jeffrey. 1998a. Dictionnaire songhay – anglais – français. Vol 3: Koroboro Senni. Paris: l’Harmattan. Heath, Jeffrey. 1998b. Texts in Koroboro Senni. Cologne: Köppe Verlag. Heath, Jeffrey. 1999. A Gammar of Koyraboro (Koroboro) Senni. Cologne: Köppe Verlag. Johannessen, Janne Bondi. 1998. Coordination. New York: Oxford Univ. Press. McNally, Louise. 1993. “Comitative coordination: a case study in group formation.” Natural Language and Linguistic Theory 11: 347–379. Mithun, Marianne. 1988. “The grammaticization of coordination.” In: Clause combining in grammar and discourse, J. Haiman and S. Thompson (eds.), 331–59. Amsterdam: Benjamins. Munn, Alan. 1999. “First conjunct agreement: against a clausal analysis.” Linguistic Inquiry 30: 643–68. Silverstein, Michael. 1976. “Hierarchy of features and ergativity.” In Grammatical Categories in Australian Languages, R. M. W. Dixon (ed.), 112–71. Canberra: Australian Institute of Aboriginal Studies. Stassen, Leon. 2000. “AND-languages and WITH-languages.” Linguistic Typology 4 (1): 1–54.

Chapter 4

Conjunction and personal pronouns D. N. S. Bhat Central Institute of Indian Languages, Mysore

1. 2. 3. 4. 5. 6. 7.


Introduction Inclusory conjunction A problem for analysis Plurality and conjunction Conjunction instead of plurality in personal pronouns Other varieties of conjunction Hierarchy of number marking


Personal pronouns are rather different from other pronouns and nouns in their association with the category of number. One of the interesting constructions in which this difference finds expression is the so-called “inclusory conjunction”. There are several languages in which personal pronouns show dual or plural forms when occurring in coordinated noun phrases, even though they denote only a single individual (§ 2–3). In order to account for this puzzling use of non-singular pronouns, I suggest that in the case of personal pronouns, the dual or plural forms are generally used for denoting conjunction rather than duality or plurality (§ 4–5). There are also certain additional types of constructions in which a number marker or a non-singular personal pronoun appears to indicate conjunction rather than plurality (§ 6). This association of conjunction rather than plurality with personal pronouns is also helpful in resolving some of the problems that are faced by Corbett (2000) in establishing a cross-linguistic hierarchy of number marking (§ 7).


Inclusory conjunction

There is a construction called “inclusory conjunction”, occurring in several languages, that involves either (i) the conjunction of a personal pronoun with another nominal (in the form of a noun phrase) or (ii) the use of a personal


D. N. S. Bhat

pronoun along with a comitative argument, with the two occurring in different phrases. In the case of most languages, an action performed jointly by either first or second person with a third person (or by first person with a second person) can be expressed in either of two ways. (i) One may use a coordinated noun phrase in which the two expressions are conjoined, with the noun phrase functioning as a single argument as in (1a), or (ii) one may use one of the participants as a distinct comitative argument as in (1b). It is also possible, of course, to use the plural (or dual) form of the pronoun as in (1c). (1) a. I and John went to the market. b. I went to the market with John. c. We went to the market. On the other hand, there are several languages that require the personal pronouns occurring in constructions of the type (1a) and (1b) to be used in their “dual” or “plural” form, in spite of the fact that they do not by themselves express a dual or plural meaning. That is, the languages appear to indicate the number of the whole phrase through those pronouns rather than the number of their own referent(s). The sentence (2) of Toqabaqita, an Oceanic language, shows this characteristic in a coordinate construction (Lichtenberk 2000: 2). (2) Kamareqa doqora-ku meki lae mai qusungadi. 1.du.excl brother-1sg 1.du.excl.fut go at tomorrow ‘I and my brother will come tomorrow.’ Notice that in (2) only two individuals are involved in the action of ‘coming’, but the pronoun kamareqa ‘we (du.excl)’ occurring as one of the constituents of the coordinated noun phrase is in the dual and not in the singular as in the English gloss. The pronoun appears to indicate the number of the whole coordinated noun phrase, which is dual, rather than the number of its own referent, which is only singular (one in number). Constructions of this type have been reported to occur throughout the Austronesian language family and also in several other languages that belong to the Australian, Indo-European, Polynesian and Mayan families (Haspelmath, forthcoming). Lichtenberk notes two main parameters along which these constructions can differ. (i) The first parameter concerns the relation between the inclusory pronoun and the noun phrase whose referent gets included in it. They may occur in the same phrase or in different phrases; the latter are called “split” constructions. Sentence (3), also from Toqabaqita (Lichtenberk 2000: 22), exemplifies a split inclusory construction. (3) Kera thaitoqoma-mareqa wela nau. 3pl know-1du.excl child 1sg ‘They know me and my child.’

Conjunction and personal pronouns

Notice that in (3) the inclusory pronoun occurs as part of the verb phrase, whereas the included noun phrase occurs as a distinct constituent. Schwartz (1988) refers to these split constructions as “verb coded coordinations”, as the argument that is referred to by a pronoun occurs with the verb rather than as an independent argument. (ii) The second parameter concerns the occurrence of a marker for conjunction. An inclusory construction may or may not have a marker for denoting the relationship between the inclusory pronoun and the included noun phrase. If it contains a marker, it is considered to be “explicit”, and if it does not, it is considered to be “implicit”. There is no marker for denoting the relationship in (2) and (3), whereas there is one in (4), which is from Polish (Schwartz 1988: 52). (4) Poszlis´my z matkø do kina. went.1pl.aux with mother to cinema ‘Mother and I went to the cinema.’ Corbett (2000: 232) refers to the possibility of this construction occurring in the case of pronominal possessive adjectives, as pointed out by Isacˇenko (1962: 481). The phrase is from Russian. (5) naš-i s toboj vospominanija with 2sg.instr ‘our memories (yours and mine)’ Corbett (2000: 233) also refers to the occurrence of what is called the “sylleptic dual”, which involves an “inclusory construction”. He gives an Old English example for this structure from Edgerton (1910: 112). (6) Wit Scilling song aho¯fn. we (two) Scilling song raised ‘Scilling and I raised a song.’ According to Corbett, the construction is not restricted to the dual, but it tends to be frequent in languages with a dual because conjoining most often involves just two conjuncts, and these more usually denote singulars rather than other values.


A problem for analysis

The occurrence of a dual or plural personal pronoun in these constructions, apparently for denoting a single individual, has led to a dispute regarding its analysis among linguists, especially because other languages use only a singular pronoun in such constructions. Some consider the inclusory pronoun to be basically a singular pronoun, whereas others argue that it is in no way different from dual and plural pronouns occurring in other contexts.



D. N. S. Bhat

For example, Schwartz (1988: 243) argues that the plurality of inclusory pronouns is internal to the construction, meaning, apparently, that it is not the same as the plurality that is expressed by agreement markers occurring with the verb (see below § 5). Schwartz (1988: 243) suggests that an explanation for this constraint may have to do with the heterogeneity of first and second person nonsingular pronouns. That is, these non-singular pronouns represent heterogeneous groups that consist of a speech act participant and one or more persons who are not of the same status, whereas non-singular forms of other nouns represent a hom*ogeneous grouping. On the other hand, Lichtenberk (2000: 8) argues that the inclusory pronoun encodes a set of which the noun phrase that occurs with it encodes a proper subset. That is, the inclusory pronoun, according to him, is the head of the construction and the included noun phrase is its modifier, an adjunct. The former indicates the plurality of the construction that is also expressed by the agreement marker occurring in the verb. The solution suggested by Schwartz appears to be superior in that it can account for some of the constraints that affect the occurrence of inclusory constructions. For example, there is a preference for first and second person pronouns to occur as inclusory pronouns, with some languages like Tagalog restricting the construction to only these two pronouns (Schwartz 1988: 241). It is extended to third person pronouns in some languages, but the use of ordinary nouns as inclusory plurals is extremely rare. Haspelmath (forthcoming) mentions one such case, namely that of Margi, a Chadic language (Hoffmann 1963: 57). (7) kàmb´6ràwázhá-’yàr àgá màlà g´6ndà with wife of.him ‘Kamburawazha and his wife’ Notice that this exceptional non-pronominal inclusory construction also involves a heterogeneous combination. There are certain other types of inclusory constructions that could be problematic for an analysis that regards inclusory pronouns as in no way different from plural pronouns. For example, Tinrin (Melanesian; Osumi 1995: 41) uses an inclusory pronoun along with a comitative noun phrase, but in addition to this, the sentence also contains a non-inclusory (singular) subject marker (see also Bril, this volume, § 4.3.2). (8) a.

U fi komu nrî pwere numea. 1sg go 1du.excl 3sg to Noumea ‘I go to Noumea with him.’

The construction containing the inclusory pronoun and the comitative noun phrase can be shifted to the sentence-initial position or to the final position.

Conjunction and personal pronouns

(8) b. Komu nrî u fi pwere numea. 1du.excl 3sg 1sg go to Noumea c. U fi pwere numea komu nrî. 1sg go to Noumea 1du.excl 3sg The notion of inclusory pronouns denoting an “internal” heterogeneity appears to be more suitable for describing the occurrence of komu ‘1du.excl’ in these sentences. A similar use of non-singular pronouns can be seen in Mundani, a BenueCongo language (Parker 1986: 136). The three personal pronouns of Mundani, used for denoting the subject, have three suppletive plural forms as shown in (9). These pronouns do not make any distinction between dual and plural. (9) 1 2 3

Singular mØa a ta

Plural bØa bØ-ı bØf

In the case of an inclusory construction, Mundani uses a complex form for denoting the included argument. When only two persons are involved, this form consists of one of the plural forms of pronouns given above, followed by nè ‘with you (sg)’ or tò ‘with him’. Parker (1986: 135) suggests that the form nè involves the combination of ne ‘with’ and the second person singular pronoun. The form tò is identical with the third person singular object form. (10) a.

Máa¯ ghˇa bá nè. 1sg go 1pl with.2sg ‘I shall go with you.’ b. Áa¯ ghˇa b-í tò. 2sg go 2pl with.3sg ‘You (sg) will go with him.’ c. Táa¯ ghˇa b´f tò. 3sg go 3pl with.3sg ‘He will go with him.’

The form tò can denote the association with first person as in (10d) when the plural pronoun occurring before it denotes first person. (10) d. Táa¯ ghˇa bá tò. 3sg go 1pl with.3 ‘He will go with me.’ When more than two persons are involved, a more complex construction is used, in which the plural forms of pronouns are followed by nG ‘with 2pl’ (made up of ne



D. N. S. Bhat

‘with’ followed by the second person plural pronoun, which has the object form w-í) or to ‘with 3’ and also a plural marker bá as shown in (11). The form to can indicate the association of first person when it is preceded by a first person plural form, as shown in (11c). (11) a.

Báa¯ ghˇa bá nG bá. 1pl go 1pl with.2pl pl ‘We shall go with you (pl).’ b. B-ía¯ ghˇa b-í to bá. 2pl go 2pl with.3 pl ‘You (pl) will go with him/them.’ c. B´ff¯ ghˇa bá to bá. 3pl go 1pl with.3 pl ‘They will go with me/us.’

It appears to me that in these inclusory constructions of Mundani also, the plural forms of pronouns can be considered to have the function of denoting the internal heterogeneity of the group.

4. Plurality and conjunction In order to understand the logic behind the use of dual or plural pronouns in these inclusory constructions, it is necessary, I think, to examine the relationship between plurality and conjunctive coordination. These are two different devices that most languages utilize for indicating the involvement of two (or more) persons or things in a particular semantic role. The crucial difference between these two devices is that conjunction emphasizes the difference between the two persons or things that are conjoined by it, whereas plurality emphasizes the similarity between the persons or objects that are grouped together by it. In fact, the differences occurring among the grouped individuals are to be disregarded when a plural marker is used. (12) a. John, Bill and Tom have gone to the movies. b. The boys have gone to the movies. Notice that the use of a conjunction in (12a) depends upon the specification of John, Bill and Tom as three distinct individuals, whereas the use of plurality in (12b) depends upon their non-specification as distinct individuals, and their specification as just boys. One cannot use conjunction if no difference is indicated, either explicitly or implicitly, whereas one cannot use plurality if differences cannot be disregarded. We cannot have *he and he as a conjunction because no difference between the two conjoined entities can be perceived, whereas John and John is

Conjunction and personal pronouns

possible only if the context indicates that the two personal names denote different individuals. When we come to the plurals of personal pronouns, however, we find this basic requirement of plurality being completely disregarded. For example, the first person plural pronoun, we, does not indicate or imply any similarity among the referents that are referred to by it. In fact, there cannot be any similarity between those referents, namely ‘I and he/they’ or ‘I and you’, at least not one that can be indicated or implied by the first person pronoun. It is as if the plural marker has been forced to do the duty of a conjunction marker. The problem is even more acute in the case of languages in which there is a distinction between exclusive and inclusive plurals. For example, there is no distinction of “number” between first person exclusive dual and first person inclusive dual, but still, grammarians describe them as representing two different “dual” forms. What we have here is only a distinction in the type of individuals that get included or excluded, and the proper device for expressing this distinction is conjunction and not plurality. I argue in Bhat (2004) that first and second person pronouns differ from other pronouns and nouns in not identifying their referents. They only identify the speech roles of ‘being the speaker’ and ‘being the addressee’ respectively and relate those roles to semantic roles with the help of case markers. They generally remain unresponsive to any changes that occur among the actual individuals who perform those speech roles. This characteristic of first and second person pronouns derives from their being “shifters”, i.e. words that remain unchanged as different persons assume the speech roles that they represent. In view of this unique characteristic of personal pronouns, any association of these pronouns with the category of number would primarily represent combinations of speech roles rather than of the individuals who perform those roles. This is another reason why the association of first and second person pronouns with the category of number results in the conjunction (of speech roles) and not in the plurality of speech act participants. Several linguists like Jespersen (1924: 192), Benveniste (1971: 201–2) and J. Lyons (1968: 277) refer to this interesting characteristic of first and second person pronouns. Cysouw (2003: 69) suggests using the term “group” rather than plurality for referring to the association of personal pronouns with the category of number, especially when it is used for denoting combinations like 1+2, 1+3, 1+2+3 and 2+3. However, the notion of “group” has apparently misled him, because he includes 3+3 also in his category of groups, and further, he sets aside pronouns that specify a number distinction, like dual or trial, from this category of “group”. I think the more important point here is the distinction between conjunction and plurality, which rests upon the notion of difference (or individuality) as against similarity among the referents concerned. The notion of “group” actually belongs to plurality rather than to conjunction, and is therefore inappropriate in characterizing the association of the number distinction with personal pronouns. Further,



D. N. S. Bhat

the specification of number does not conflict with the notion of conjunction as it does with the notion of a “group”. In fact, in the case of personal pronouns, the most prototypical conjunctive form is a dual involving 1+2 in which number gets very clearly specified. Most languages differentiate between first and second person pronouns on the one hand, and the rest of the nominals on the other, in their marking of number distinctions. There are very few languages like Trumai (Isolate; Guirardello 1999: 27) that use the same marker for denoting the number distinction in personal pronouns and nouns. The general tendency is to have suppletive forms in the case of personal pronouns as shown, for example, by English, which has the form we for denoting the “plural” of I (and the form us for denoting the “plural” of me). Several languages use distinct markers for personal pronouns in contrast to other pronouns and nouns. For example, Kannada uses -vu in the case of first and second person pronouns and -ru (human) and -ga»lu (non-human) in the case of nouns. In spite of this differentiation in the number markers used, and in the meaning that the markers represent, grammarians have tended to describe the markers occurring with personal pronouns (and suppletive forms of pronouns) as allomorphically related to other number markers. There is clearly a need to re-examine the assumption that underlies this general procedure, especially in the case of languages in which number marking is restricted to first and second person pronouns. This is because in the case of such languages, non-singular forms are used only for denoting conjunction, and hence from the language-internal point of view, there cannot be any basis for regarding them as denoting “plurality” with the help of those forms.


Conjunction instead of plurality in personal pronouns

Consider, for example, the case of Chalcatongo Mixtec (Macaulay 1996: 81), which has two distinct personal pronouns, namely rù’ù ‘first person’ and ro’o ‘second person’. Third person pronouns are general nouns such as ˇcàà ‘man’ (for 3rd masculine) and ñã’ã ‘woman’ (for 3rd feminine). There is only one non-singular form for the two pronouns, namely žó’ó ‘first and second person’ (“first person dual exclusive”). Plurality is expressed nowhere else among nouns in the language. In order to indicate the plurality of the subject, the language uses the prefix ká- that is attached to verbs. If we apply the traditional analysis to this language, we would have to regard the language as having a dual category that is restricted to first person. That is, we perceive a defective paradigm. On the other hand, if we consider the language as expressing only conjunction and not duality or plurality among its pronouns, the paradigm of pronouns will not appear to be defective. There can only be a single conjunctive form in a system that consists of just two

Conjunction and personal pronouns

persons, namely 1+2, and the language correctly possesses a single conjunctive form to represent it. There are languages like Kwakiutl (Wakashan: Boas 1911: 527) that also appear to restrict number marking to first person in the traditional analysis. Kwakiutl has singular forms for all three persons, but two plural forms for the first, namely an inclusive and an exclusive, as shown in (13). A similar system is reported to occur in Acehnese (Durie 1985), which also has a first person pronoun restricted to singular use, and two additional forms used for denoting exclusive and inclusive plurals. Second and third person pronouns do not show any number distinction. (13)

Kwakiutl: 1 2 3

Singular en es (e¯)

Inclusive eu7xu »

Exclusive ens

If we regard these forms as representing the possible conjunctions of persons, rather than as “plural” forms of the first person pronoun, the paradigm would appear to be less defective. Notice that the three persons can involve three possible conjunctions, namely 1+2, 1+3 and 2+3. The language has representations for two of them. In fact, Boas suggests that the two forms “are not conceived as plurals” but as inclusive and exclusive forms. Several languages show this kind of disparity between the number of singular forms on the one hand and the number of corresponding non-singular forms on the other. Most of these languages appear as possessing defective paradigms if we apply the traditional analysis to them, whereas if we regard them as representing conjunction with the help of different types of number markers or suppletive forms, most of the paradigms turn out to be perfectly regular and logical. For example, the “gap” occurring in the system of English personal pronouns, namely the absence of second person plural, can have a similar explanation if we regard the first person “plural” (the suppletive form) as basically a conjunction of first and second persons. In fact, the third person does not really belong to the system of personal pronouns in English. There are several languages in which a distinct suppletive form occurs for denoting the 1+2 combination. In the case of several such languages, like those of Arnhem Land (Australia) and some of the Philippines, this suppletive form behaves very much like a singular pronoun in taking dual and plural markers. In Rembarrnga (McKay 1978, as referred to by Corbett 2000: 166), for example, the traditional representation of personal pronouns provides a disjointed occurrence of the ‘dualtrial’ affix -bbarrah as can be seen in (14).



D. N. S. Bhat

(14) 1excl 1incl 2 3m 3f

Singular ng}n} – k} nav} mf ngad}

Dual yarr-bbarrah y}kk} nakor-bbarrah barr-bbarrah

Trial – ngakorr-bbarrah – –

Plural yarr} ngakorr} nakorr} barr}

McKay proposes to regard the inclusive dual form y}kk} ‘1+2’ as “minimal” along with the singular pronouns 1, 2 and 3 and to consider the remaining dual forms along with the trial form of y}kk}, namely ngakorr-bbarrah, as “unit augmented” forms and the plural forms as “augmented” forms. This results in a better arrangement of the forms into paradigms (especially from the point of view of the suffix -bbarrah) as shown in (15). (15)

Minimal 1excl ng}n} 1incl y}kk} 2 k} 3m nav} mf 3f ngad}

Unit augmented yarr-bbarrah ngakorr-bbarrah nakor-bbarrah barr-bbarrah

Augmented yarr} ngakorr} nakorr} barr}

Similar suggestions to shift the 1+2 form to the column that contains singular pronouns have been made in the case of languages like Tarma Quechua (Adelaar 1977: 218), Maranao (McKaughan 1972), Popolocan languages (Veerman-Leichsenring 2000: 322) and some additional American Indian languages (Mithun 1999: 218). However, the analysis appears to be rather ad hoc because one cannot escape from the fact that the 1+2 form is unlike singular forms in that it represents non-singular referents. If, on the other hand, we regard the inclusive form as representing a conjunction, its separation from dual and plural forms would be less ad hoc.

6. Other varieties of conjunction Languages use plural markers for indicating conjunction in certain other contexts as well. For example, Sanskrit can have a noun phrase conjunction marked either (i) by the coordinator ca attached to both the nouns or (ii) by juxtaposition. In the latter case, the noun phrase must obligatorily be marked as dual or plural. (16) a.

ra:mas´-ca lak»smanas » ´-ca Rama-and Lakshmana-and ‘Rama and Lakshmana’

Conjunction and personal pronouns

b. ra:ma-lak»sman-au » Rama-Lakshmana-du ‘Rama and Lakshmana’ Sanskrit also allows one of the constituents to be deleted in the latter case, but this appears to be possible only in the case of “natural pairs” like ma:ta:-pitar-au (mother-father-dual), which is equivalent to pitar-au (father-dual) ‘mother and father’. Some of the Dravidian languages like Kannada also make use of these two different ways of indicating conjunction. The use of the plural marker in the latter case is obligatory: an» na » mattu tamma and ‘elder brother and younger brother’ b. an» na-tamma-ndi-ru » ‘elder brother and younger brother’

(17) a.

The latter construction, however, is used only in contexts in which the contrast between the two coordinands is to be emphasized. It is especially used for denoting pairs of terms such as ‘father-mother’, ‘elder-brother/younger-brother’, ‘mother-inlaw/daughter-in-law’, etc. We can regard these appositional constructions as representing another context in which the plural marker has to function as a conjunction marker. There is an interesting contrast in Babungo (Grassfields Bantu; Schaub 1985: 197) between appositive constructions of the above type on the one hand, and constructions in which the noun phrase that occurs after a personal pronoun indicates the identity of its referent on the other. The language makes a distinction between these two different types of pronoun-noun phrase constructions, which Schaub calls “reference” and “selective” constructions. In the former case, the reference of the noun phrase includes the referent(s) of the pronoun that precedes it, whereas in the latter case it does not. In both these cases, however, the pronoun includes the referent(s) of the noun phrase. Yìa v-íG ndâa g´6 ntó’. we (excl) people smithy go palace ‘We, the blacksmiths, go to the palace.’ b. Yía v-íG ndâa g´6 ntó’. we (Selective) people smithy go palace ‘I and the blacksmiths go to the palace.’

(18) a.


100 D. N. S. Bhat

(18a) is actually the so-called appositive construction and (18b) is an inclusory construction. Babungo differentiates between these two types of constructions through the use of a floating high tone. Notice that the initial pronoun in (18b) is in high tone. This example indicates further that Lichtenberk’s (2000) claim that there is no difference between plural pronouns and pronouns that occur in inclusory constructions is not true of all languages. Another interesting case is that of Pero (Chadic; Frajzyngier 1989: 159), in which the conjunction marker is attached to the second person pronoun in order to indicate that it involves “conjunction”. That is, the conjunction marker indicates that the pronoun includes first person. It does not, however, involve any plurality as such, and hence we cannot claim that it constitutes an inclusory construction. Notice that the construction resembles the Sanskrit dual form given earlier. We can perhaps regard this special use of the conjunction marker as deriving from the unique relation that exists between personal pronouns and conjunction. (19) a.

Kán-mà tà-wáatò gombé. conj-2pl fut-go Gombe ‘We are going to Gombe.’ b. Kán-ci kopo. go ‘Let us go!’

The “inclusory” characteristic of first and second person pronouns has given rise to certain other puzzles as well. For example, C. Lyons (1999: 313) points out that the possessive construction of English allows an anaphoric pronoun to be either in agreement (20a) or non-agreement (20b) with it. (20) a. Some of us like our beer chilled. b. Some of us like their beer chilled. Similarly, Spanish allows a third person subject to have an agreeing verb in third person plural (21a) or a non-agreeing verb in first person plural (21b). In the latter case, the subject is considered to include the speaker. (21) a.

Los estudiantes trabajan mucho. the students work(3pl) much ‘The students work hard.’ b. Los estudiantes trabajamos mucho. the students work(1pl) much ‘We students work hard.’

All these puzzling uses of first and second person pronouns derive from their unique characteristic of being “shifters” — expressions that identify speech-roles rather than the actual individuals who perform those roles. The pronouns need to

Conjunction and personal pronouns

be distanced from the identity of their referents in order to function as shifters and hence when they are associated with the category of number, they tend to denote, primarily, the conjunction of speech roles rather than the plurality of speech act participants.


Hierarchy of number marking

Corbett (2000: 56) discusses in detail an Animacy Hierarchy for number-marking, proposed earlier by Smith-Stark (1974), in which the three personal pronouns are considered to occupy the top position. (22) 1 > 2 > 3 > kin > human > animate > inanimate The placement of the first person pronoun on top of this hierarchy is based upon the existence of two languages, namely Kwakiutl (Boas 1911) and Acehnese (Durie 1985) mentioned above (§ 5), in which number marking is reported to be restricted to first person. On the other hand, the placement of the second person pronoun in the second position is based upon the existence of two other languages, namely Asmat (Papuan; Voorhoeve 1965: 142) and Guaraní (Tupi; Gregores & Suárez 1967: 140), in which number marking is restricted to first and second persons. The pronouns of Asmat are given in (23). (23) 1 2 3

Singular no o a

Plural na ca –

This positioning of first and second person pronouns on top of this number hierarchy appears to be rather odd because we generally find the most prototypical members of a group occupying the top position of a hierarchy into which the group is organised. For example, in an accusative-ergative case-marking hierarchy, first and second person pronouns rightly belong at the top because their occurrence with an accusative system of case marking is the most prototypical one (Dixon 1994). However, in the case of the number hierarchy, personal pronouns do not prototypically belong at the top, especially because the notion of plurality has an extended use among them — a use that actually conflicts with its primary characteristic of disregarding the differences that occur among individuals. Corbett (2000) has some difficulty in justifying this animacy hierarchy. For example, he finds a variety of number-marking, called associative plural, to occur only with personal names, kin terms and certain other human nouns but not with personal pronouns. That is, it occupies the middle portion of the hierarchy and therefore forms an exception. Corbett refers to the case of Hungarian (based on


102 D. N. S. Bhat

Edith Moravcsik, personal communication), in which personal names, kin terms and nouns denoting professions can occur either with a plural suffix -ok or an associative suffix -ék, whereas personal pronouns have only plural forms. (24) jános apa tanító


jános-ok jános-ék ‘father’ apa-ék ‘teacher’ tanító-ék

‘Johns’ (more than one person called John) ‘John and his associates’ ‘father and his group’ ‘teacher and his group’

Following a suggestion made by Moravcsik, Corbett considers the possibility of regarding personal pronouns also as representing an associative plural rather than ordinary plural. However, this would make the occurrence of the ordinary plural exceptional for the hierarchy because according to this reinterpretation, personal pronouns, though occurring at the top of the hierarchy, would not be showing the ordinary plural. In view of this difficulty, Corbett (2000: 107) proposes to regard “associative plural” as not forming part of the category of number. He considers such a proposal to be supported by the occurrence of a post-base (derivational marker) in Central Alaskan Yupik (Jacobson 1995), namely -nku ‘and family’. This post-base can be attached only to proper names, but it can be followed either by the dual suffix -k or the plural suffix -t (cuna-nku-k ‘Cuna and his friend’, cuna-nku-t ‘Cuna and his family/friends’). The dual and plural markers apparently indicate the total number of persons that are denoted by the expression. Two other types of distinctions that are associated with the notion of number also appear to be exceptions to the animacy hierarchy. They are distributives and collectives, which occur with nouns but not with personal pronouns. Corbett proposes to regard them also as not belonging to the number category. One can actually extend this argument to non-singular forms of personal pronouns that involve either suppletion or the use of special number markers for denoting different combinations of speech act participants as well. In several languages, such forms can take dual and plural markers just like other singular pronouns. We have already seen this situation occurring in the case of languages that have “augmented” forms rather than plural forms (§ 5). Another type of situation occurs in Nivkh (Paleo-Siberian; Gruzdeva 1998: 26). This language has a 1+2 dual form, inclusive and exclusive first person plural forms, and second person and third person plural forms. The formation of these non-singular forms involves either suppletion or irregular affixation. All these forms can occur with the plural suffix -gu. Similarly, Imonda (Papuan; Seiler 1985: 44) has a single suppletive form for denoting inclusive meaning, and all its personal pronouns, along with this inclusive one, can take -id ‘men’ to denote plural meaning. Exceptions to the animacy hierarchy of number marking also result from the fact that there are some languages in which dual occurs with some nouns but not

Conjunction and personal pronouns 103

with personal pronouns. According to Plank (1989: 297) several languages of the Semitic family, and also Irish and possibly Polish, manifest such an occurrence of the dual. He points out, however, that these are diachronically unstable. On the other hand, Hopi appears to have newly developed a dual that is restricted to some nouns. Plank refers to an interesting distinction proposed by Humboldt (1830) between languages that restrict the dual (i) to pronominals, (ii) to nominals (in order to represent natural pairs) and those that (iii) allow them to occur with nominals as well as pronominals. In the latter case, the category would be of an allpervasive type. Plank (1996) is a more detailed study of these exceptional duals. He finds 26 languages (in a sample of 205) in which the dual occurs only with some nouns, and in twelve of these the dual does not occur with personal pronouns. Among the remaining languages in which the dual is restricted to personal pronouns and some nouns, the latter include primarily animate nouns. Thus, the animacy hierarchy, as represented in (22), is not only defective in placing non-prototypical members on top, but also in having several exceptional cases to deal with. In view of these problems, it is perhaps better to regard the hierarchy as involving two different criteria rather than a single one. That is, it may be split into two different levels. We can regard the notion of conjunction as moving down from the top of the hierarchy and plurality as moving down from third person pronouns. The dual will have to be regarded as representing (i) conjunction in the case of personal pronouns (especially when it is represented by distinct number markers or by suppletion) and (ii) the category of number in other cases. The restricted use of duality to denote “animate pairs” can perhaps be viewed as an extension of its use for denoting conjunction among personal pronouns. (25) a.

1>2> Conjunction Æ 3 > kin > human > animate > inanimate Plurality Æ

It is now possible to regard different types of plurals like associative, collective and distributive, which do not extend up to personal pronouns, as occurring in the second level of the hierarchy starting from the third person pronoun or kin terms. They are connected with the notion of plurality in the sense that they do not differentiate between individual members that are regarded as associates or collections, or entities that are distributed over a particular location. On the other hand, the notion of inclusory conjunction can be added to the first level of the hierarchy as it forms part of the notion of conjunction. Its preferred occurrence is with first and second persons and may be extended to third and also to nouns involving differentiation.

104 D. N. S. Bhat

(25) b. 1 > 2 > Conjunction (animate dual) Æ (Inclusory construction) Æ 3 > kin > human > animate > inanimate Plurality Æ Associative Æ Collective Æ Distributive Æ

Abbreviations ass conj

associative conjunction

References Adelaar, W. F. H. 1977. Tarma Quechua grammar, texts, dictionary. Lisse: The Peter Ridder Press. Benveniste, Emile. 1971. Problems in general linguistics. Translated by M. E.Meek. Coral Gables: University of Miami Press. Bhat, D. N. S. 2004. Pronouns. Oxford: Oxford University Press. Boas, Franz. 1911. “Kwakiutl.” In Boas, Franz (ed.), Handbook of American Indian Languages. Washington: Government Printing Press, 423–557. Bril, Isabelle. This volume. “Coordination strategies and inclusory constructioins in NEw Caledonian and other Oceanic languages.” Corbett, Greville. 2000. Number. Cambridge: Cambridge University Press. Cysouw, Michael. 2003. The paradigmatic structure of personal marking. Oxford: Oxford University Press. Dixon, Robert M. W. 1994. Ergativity. Cambridge: Cambridge University Press. Durie, Mark. 1985. A grammar of Acehnese: on the basis of a dialect of North Aceh. Dordrecht: Foris. Edgerton, Franklin. 1910. “Origin and development of the elliptic dual and of dvandva compounds.” Zeitschrift für vergleichende Sprachforschung auf dem Gebiete des Deutschen, Griechischen und Lateinischen 43: 110–20. Frajzyngier, Zygmunt. 1989. A grammar of Pero. Berlin: Dietrich Reimer Verlag. Gregores, Emma and Suárez, Jorge A. 1967. A description of colloquial Guaraní. The Hague: Mouton. Gruzdeva, Ekaterina. 1998. Nivkh. Munich: Lincom Europa. Guirardello, Raquel. 1999. A reference grammar of Trumai. Ph.D. Dissertation Rice University. Haspelmath, Martin. Forthcoming. “Coordination.” In Shopen, Timothy (ed.), Language typology and syntactic description, 2nd edition. Cambridge: Cambridge University Press. Hoffmann, Carl. 1963. A grammar of the Margi language. London: Oxford University Press. Humboldt, Wilhelm von. 1830. “Uber den Dualis.” Reproduced in Königlich Preussische Akademie der Wissenschaften (ed. by), 1907, Wilhelm von Humboldts gesammelte Schriften, Volume 6. Berlin: B. Behr, 4–30.

Conjunction and personal pronouns 105

Isacˇenko, Aleksandr V. 1962. Die russische Sprache der Gegenwart, I: Formenlehre. Halle: Max Niemeyer. Jacobson, S. A. 1995. A practical grammar of the Central Alaskan Yup’ik Eskimo language. Fairbanks: Alaska Native Language Center. Jespersen, Otto. 1924. The philosophy of grammar. London: George Allen & Unwin. Lichtenberk, Frantisek. 2000. “Inclusory pronominals.” Oceanic Linguistics 39: 1–32. Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press. Lyons, John. 1968. Introduction to theoretical linguistics. Cambridge: Cambridge University Press. Macaulay, Monica. 1996. A grammar of Chalcatongo Mixtec. Berkeley: University of California Press. McKaughan, Howard P. 1972. “Semantic components of pronoun systems: Maranao.” Word 15: 101–2. McKay, Graham R. 1978. “Pronominal person and number categories in Rembarrnga and Djeebbana.” Oceanic Linguistics 17: 27–37. Mithun, Marianne. 1999. The languages of Native North America. Cambridge: Cambridge University Press. Osumi, Midori. 1995. Tinrin grammar. Honolulu: University of Hawaii Press. Parker, Elizabeth. 1986. “Mundani pronouns.” In Wiesemann, Ursula (ed.), Pronominal systems. Tübingen: Gunter Narr, 131–165. Plank, Frans. 1989. “On Humboldt on the dual.” In Corrigan, R., Eckman, F., and Noonan, M. (eds.), Linguistic categorization. Amsterdam: John Benjamins, 293–333. Plank, Frans. 1996. “Domains of the dual, in Maltese and in general.” Vista di Linguistica 8.1: 123–140. Schaub, Willi. 1985. Babungo. London: Croom Helm. Schwartz, Linda. 1988. “Conditions for verb-coded coordinations.” In Hammond, Michael, Moravcsik, Edith, and Wirth, Jessica (eds.), Studies in syntactic typology. Amsterdam: John Benjamins, 53–73. Seiler, Walter. 1985. Imonda, A Papuan language. Canberra: The Australian National University. Smith-Stark, T. Cedric. 1974. “The plurality split.” Chicago Linguistic Society 10: 657–61. Veerman-Leichsenring, Annette. 2000. ”Popolocan independent personal pronouns: comparison and reconstruction.” International Journal of American Linguistics 66: 318–59. Voorhoeve, C-L. 1965. The Flamingo Bay dialect of the Asmat language. ‘s-Gravenhage: Martinus Nijhoff.


Chapter 5

The grammar of conjunctive and disjunctive coordination in Iraqw* Maarten Mous University of Leiden

1. Introduction 2. Basic properties of nee ‘and’ and laqáa ‘or’ 3. Coordination in noun phrases 3.1 Coordinated noun phrase as a unit 3.2 Coordinated modifiers as a unit 4. Agreement 5. Word category and origin



The purpose of this article is to present the grammatical properties of nee ‘and, with’ and laqáa ‘or’, which are the elements that are used in coordination in Iraqw: nee for conjunctive coordination and laqáa for disjunctive coordination. In addition to their coordinating functions, they have a number of preposition-like functions as well. In § 2 I discuss the basic properties of these two coordinators; in § 3 I show that modifiers can be coordinated but that the coordinated noun phrase does not so easily form a unit under modification. In § 4 I discuss the agreement behavior of coordinated noun phrases. In § 5 it is argued that both coordinators belong to a limited word class of prepositions, while there is no evidence for a word class of conjunctions. I also suggest that nee is of Bantu origin. This paper was inspired by the overview on coordination presented in Haspelmath (to appear). Examples come partly from my grammar (Mous 1993) and from the extensive text collection in Berger and Kiessling (1998).

*This article was written during a research period at the Max Planck Institute for Evolutionary Anthropology in Leipzig. I would like to thank my colleagues there for stimulating discussions.


Maarten Mous

Iraqw is a Southern Cush*tic language spoken by roughly half a million people in northern Tanzania. It is a verb-final language. Modifiers follow the head. Modifiers agree in gender with the head noun, which is in construct state when modified. The subject is always marked on the verb and an NP or pronoun subject need not be expressed when retrievable from the context. Every sentence contains an inflectional word termed “selector” in Iraqw studies or “indicator particle” in Somali studies. The selector precedes the verb and constitutes the beginning of a syntactic unit, the “verbal piece”. The selector, iri in (1), contains subject agreement plus tense/aspect/ mood. Objects, fa/a ‘food’ in (1), and verbal adverbs, tsuwa ‘for sure’ in (1), occur inside the verbal piece, i.e. between the selector and the verb. When objects are preposed, as is the case for gitladá’ in (2), they need an object pronoun as selector, u in ’una in (2). For more details on Iraqw structure, see Mous (1993). (1) /ameeni-r-dá’ i-ri tsuwa fa/ár /agagiin. woman-f-that 3-nar for.sure food:f:con eat:impf:3sg.f ‘And that woman was surely eating.’ (2) (’aníng) gitla-dá’ ’u-na aahhiit. I man-that obj.m-past hate:1sg ‘I hate that man.’ The letter / represents the voiced pharyngeal fricative, hh the voiceless pharyngeal fricative; hl is the lateral fricative, ts and tl are ejective affricates; for details, see Mous (1993).


Basic properties of nee ‘and’ and laqáa ‘or’

Both coordinators, nee and laqáa, are prepositive and monosyndetic, i.e the coordinator is used only once for two coordinands and its position is in front of the second coordinand (A co-B), e.g. /ayto’o nee loosí ‘maize and beans’, lahhóo laqáa tsiiyahh ‘six or four’. The coordinators are considered to be prepositive rather than postpositive on the basis of the fact that coordinator plus coordinand can occur without a preceding coordinand, as we will see later (cf. 35), while the reverse is impossible. The bisyndetic disjunctive construction, laqáa X laqáa Y, is used in particular in order to express that the choice between the two coordinands is irrelevant, as in (3). It does not express that a choice is compulsory and in this respect it differs from either X or Y.

The grammar of conjunctive and disjunctive coordination in Iraqw

(3) baráa hlaqwaraa-r-o hee bur-kw-aa ’óh in fighting-f-bgnd man cond-obj3:imps:obj.m-perf catch:3sg.m:past laqáa Tarmo laqáa ’Irqutu umú-qo hee-w-o kuu or Datoga or Iraqw every-emph man-m-bgnd obj3:imps:obj.m gaas-i-ká. kill:3sg.m-inf-neg ‘If someone was caught during the fight, whether a Datoga or an Iraqw, no one would be killed.’ For conjunctive coordination, we have no examples of bisyndetic constructions (*nee X nee Y). In situations in which one wants to stress that both coordinands are involved (‘both…and’ construction) the monosyndetic, and not a bisyndetic, construction is used. This is, for example, the case in the fixed last lines of a poetic prayer: tlaawu nee daanda ‘blanket and back’, ya/ati nee kolo ‘sandal and heel’, dawa nee afa ‘hand and mouth’, where the comparable coordinands are conceptually inherently linked, creating an association of universality. Extra emphasis can be expressed by the addition of hleemee ‘also’ after the second coordinand. The ‘also’ coordinator is postpositive and monosyndetic (A B-co), as shown in (4). (4) faarór tlakw tám nee ko’án, fáanqw hleemee née gwaléel number:f:con bad:f three and five seven also and nine 1 ka tlaakw. obj3:imps:obj.f bad ‘Uneven numbers are three and five; seven and nine are also uneven.’ In addition to NP coordination, both the conjunctive (5) and the disjunctive (6) coordinator are also used for clause coordination. The first nee in (5) coordinates two nouns while the second nee coordinates two sentences. In (6), laqáa links two sentences, each expressing a possibility of how to create the conditions for a game. (5) [daaqay nee baabá ta laaw bar qoomá-r dóohla] [boys and father imps go:cultivate:pres cond period:con-f cultivation nee [daxa qooma-r-í ham ta-wa bará fiqit…] and [now period-f-dem1 now in:con harvesting ‘… the boys and father go to cultivate if it is the period of cultivation, and now this period is harvest time…’

1.The object pronoun a obj.f in k-a refers back to the preposed feminine object faaro (f) ‘number’.



Maarten Mous

(6) [peehhír niina ’a-n tleehh-áan] laqáa [yaamu [board:f:con small:f obj.f-prog prepare-1pl or [ground i-n fool-áan boho’ee gwaléel hháw wak-e.] obj.p-prog dig-1pl holes nine line:m:con one-bgnd ‘We prepare a small wooden board or we dig nine holes in the ground, in one line.’ The examples (3) through (6) show clear-cut cases of conjunctive and disjunctive coordination. This distinction is not always obvious. In (7) the concept ‘even number’ is explained by presenting instances of various even numbers. Both laqáa and nee are used for coordinating these instances; there is no conjunctive–disjunctive opposition here. (7) faarór hhoo’ ’a tsiiyáhh laqáa tsár nee lahhóo’ née dakáat. number:f:con good:f cop four or two and six and eight ‘An even number is four, two, six or eight.’ In addition to a conjunctive function, nee has comitative (cf. 8), agentive (cf. 9) and comparative meaning (cf. 10, and 18 below). (8) tlaw-áng nee hikwa-wo. leave-imp.hith with cattle-bgnd ‘Leave with the cattle!’ (9) geesó duqa, kúu tu gaas geesó duqa 2sg.m imps:obj2sg.m kill:pres nee múk dóo taatú hare-r-ók. by people:m1:con house:m:con grandfather:m:con wife-f-2sg.poss ‘Geso Duqa, you will be killed by the people of your wife.’ (10) faaro ’ar kángw ’iraqw a dimbé’ nee faaro ’ar counting instr matter:con Iraqw cop different with counting instr kiswahili. Swahili ‘Counting in Iraqw is different from counting in Swahili.’ The distinction between comitative and conjunctive is not always easy to make; compare (11) and (12). In (12), only a comitative interpretation is admissible, while in (11) both a comitative and a conjunctive reading are possible. (11) muu-dá’ nee dama-r-ín ta-ri waráahh. people-dem4 and calf-f-3pl.poss imps-nar pass:past ‘Those people and their calf passed.’ Or: ‘Those people passed with their calf.’

The grammar of conjunctive and disjunctive coordination in Iraqw

(12) ’ino’ín ta qaró waráahh nee hikwa-ín. 3pl imps already pass:past with cattle-3pl.poss ‘The others had already passed with their cattle.’ Laqáa not only expresses disjunctive coordination, but is used in non-coordinating functions as well. In (13), it expresses possibility or probability, and in (14), it expresses doubt. (13) ’i-n hluq-ká laqáa ká-n tsuu/ nee subj.3-prog gain:booty:3f-neg or obj3:imps:obj.f-prog by hooma-dáa-dá’. enemies-dem4-dem4 ‘They certainly won’t gain any booty, and probably they will be killed by those enemies.’ (14) ’i-ri ’óo’ laqáa hhab-a-ká. subj3-nar say:3m:past or be:weak:2-inf-neg ‘He said: “So you are not weak, huh?”’ In multiple conjunction the last coordinator is obligatory, and earlier coordinators are optional: A (co-)B (co-)C co-D; both for nee, (15) and (16), and for laqáa (17) and (18). (15) faarór hhoo’ a tsiiyáhh laqáa tsár nee lahhóo’ née dakáat. number:f:con good:f cop four or two and six and eight ‘An even number is four or two and six and eight.’ (16) dí-r ni keemú hláa’ a Kwermuhl, (nee) Tlawi, place:con-f hith going:con want:1sg cop Kwermuhl (and Tlawi, (nee) Dongobesh, (nee) Haydom nee Daudi. (and Dongobesh and Haydom and Daudi ‘I want to go to Kwermuhl, Tlawi, Dongobesh, Haydom, and Daudi.’ (17) faaroo-r-ós ’a tsíiyáhh laqáa tám laqáa tsár. number-f-3sg.poss cop four or three or two ‘[When they are ready, the boys that will play this game,] their number is four, three or two.’ (18) mugú’ gár ka-r tlehhahhiti ’a-qo game thing:f:con obj3:imps:obj.f-instr prepare:3m:sbjv cop-emph [’iláa /ayto’o] [tlaa/áa tsitsihhár niinakw] [grains:p:con maize [stones:p:con pebbles:f:con small:pl:f] laqáa [gár hlaqás nee ’iláa /ayto’o]. or [thing:f:con similar to grains:p:con maize ‘The Mugú’-game is played with grains of maize, with small pebbles or with anything that is like maize grains.’



Maarten Mous

Idiomatic conjunctions (or coordinative compounds) such as ga/ale nee mahhangw ‘bow and arrow’ are not different from non-idiomatic conjunctions.


Coordination in noun phrases

3.1 Coordinated noun phrase as a unit A property of coordinated noun phrases is that they form a unit that can fall under the scope of a single preposition, adverb, or of a single modifier. In Iraqw too it seems to be the case that a coordinated noun phrase can fall under the scope of a single preposition: ar ‘with’ in (19), or of a single locational noun dír ‘place-of, to’ in (21), and of the agentive nee ‘by’ in (20). (19) ’i-na ’a’ii káy ’ar [ya’a nee gari]. subj3-past journey:dir go:3sg.m instr [leg and car ‘He travelled on foot and by car.’ (20) ta-ri maanimóo hláy nee [hhiiyaa-w-ós laqáa imps-nar Bantu:m:con find:3m by [brother-m-3sg.poss or mulqumoo-w-ós]. friend-m-3sg.poss ‘Then some Bantu will be summoned by his brother or his friend.’ (21) ’i [dír [hhawate nee /ameena-r]]=i ’axwees subj3 [place:f:con [men and women-f=dir talk:3m:pres ‘He is talking to men and women.’ A single “case cl*tic” =i dir appears to refer to a coordinated noun phrase in (21). These so-called case cl*tics are structurally linked to the verb, and not a property of the noun (phrase) as such because they have a fixed syntactic position before the verb and they indicate that there is an NP somewhere in the sentence with that particular role, but not necessarily the NP that it gets cl*ticized to. Thus, in (22) the directional dir =i has scope over atén ‘us’ but it gets cl*ticized to ‘thing’ which happens to precede it. As a consequence, the example in (21) does not so much show the scope properties of the cl*tic that is attached to the NP; rather, it shows that the conjoined NP can be conceived as a single referent to which a role is assigned. (22) atén ’iringée-r-í tí alaa gáa-r=í geexéera. 1pl crime-f-dem1 obj1pl but thing-f=dir:con leave:3f:inf:int ‘Will this sin leave anything to us?’ Coordination creates a single referent for the postposed adjunct hleemeero in (23). The conditional adverb bar in (24) has the three disjunctively coordinated options in its scope; in some other examples the conditional is repeated.

The grammar of conjunctive and disjunctive coordination in Iraqw

(23) [hiikwa nee ’aara] hleemeero awa kijiji [cattle.p and goats.p all of.p village daaqay gi bará qawo-r-i tlees. boys obj3:obj.p in:con meadow-f-dir bring:3m ‘The boys bring all the cows and goats of the village to the meadows.’ (24) bar [[hée ’a duuxun] laqáa [hée cond [[man:con subj3:perf marry:dur:3m:subj or [man:con ’a duuxuniká] laqáa [qawri]] hlaqsi’ii-wós ’a subj3:perf marry:3m:inf:neg or [baby examples-3sg.poss cop diimbadiimbé. various ‘[If a wife dies among the Iraqw,] or if [the one who died] is a married man or an unmarried man, or if it is a baby, [in all cases] the burial customs are different.’ Modifiers, however, do not easily take a coordinated noun phrase in their scope. Adjectives which follow a sequence of nouns only modify the last noun. If it is to refer to both nouns, the adjective must be repeated, as in (25). Similarly, the conjoined NP cannot fall under the scope of one possessive or demonstrative suffix; these modifying suffixes have to be repeated, like the possessive suffix in (26). Within such coordinated noun constructions, a demonstrative or possessive suffix or pronoun refers only to the noun to which it is suffixed, as is the case with the demonstrative suffix in (26, 27). If a genitive phrase is used after a coordinated noun structure, the ‘of ’-pronoun refers only to the last noun, and agrees only with the last noun, as shown in (27). (25) kitángw ’úr nee kabatí-r ’ur gi tleehhiit. chair:con big and cupboard:con-f big obj3:obj.p make:3m ‘He is making a big chair and a big cupboard.’ (26) [dasi-r-ók nee [/ameni-r-ók ti-dá’]] [girl-f-2sg.poss and [woman-f-2sg.poss indep.f-dem4 ki saaw-en. obj.3:imps:obj.p far-pl ‘Your daughter and that wife of yours are far.’ (27) [mulqumo-w-í] nee [hhiya-’ée ku-dá’ ’oo ’ayá [friend-m-dem1 and [brother-1sg.poss indep.m-dem4 of.m land:con Tumati] naa hardah-iyé’. Tumati hith:perf arrive-3pl:past ‘This friend and my brother from Tumati have arrived.’ On the other hand, the scope of a modifier that consists of a noun is the coordinated construction, despite the fact that only the last noun is marked as being modified



Maarten Mous

by the construct case, as in (28). Thus the first NP in (28) (‘this boy and this girl’) is a unit which is considered plural for the semantic number agreement on the adjective and which is modified by the name Buura. (28) [dasi-r-í nee garma-w]ú Buura ki hlahla/ar-en. [girl-f-dem1 and boy-dem1:m:con Buura obj3:imps:obj.p ugly-pl ‘The girl and the boy of Buura are ugly.’ In conclusion, coordinated nouns, whether conjunctively or disjunctively coordinated, do appear as a unit when under the scope of operators (such as prepositions) that precede them, but less clearly for modifiers that follow them (see Marten (2000) for a discussion of left-to-right processing as an explanatory factor in the properties of coordinating constructions in Swahili). 3.2 Coordinated modifiers as a unit Having discussed coordinated heads in the previous section, we now turn to the properties of coordinated modifiers. Within the noun phrase, modifiers can be coordinated to form a single modifying entity. For example, in (29) a coordinated noun structure modifies the preceding noun ‘story’, and in (30) numeral modifiers are coordinated with laqáa ‘or’. (29) ti’itá-r [kwa/angw nee du’uma] story:con-f [hare and leopard ‘the story of the hare and the leopard’ (30) ’ay deelo [tsár laqáa tám] towards day [two or three ‘up to two or three days’ In (31) there are two examples of coordinated color adjectives referring to the head noun ‘colors’. Each adjective has plural agreement referring to the plural head noun ‘colors’. The two constructions are coordinated with an adversative conjunction alaa and the head noun of this matrix construction, ’inqwari ‘sheet’, is represented and repeated by a construct case pronoun (glossed with ‘of ’), which is obligatory here as it always is when the head noun is already modified (in this case by a possessive suffix). Example (31) could also be interpreted as one NP in which the adjectives are used attributively. The first conjunctive NP falls under the scope of one negative suffix in (31). The scope of the negation suffix is marked by a background suffix -ee- added to the last element of the larger phrase containing the first two coordinated colors.

The grammar of conjunctive and disjunctive coordination in Iraqw

(31) ’inqwari-r-’ée’ [’ar kal’á [da/at-en nee qansar-en]]-ee-ká sheet-f-1sg.poss [of.f colors:con [red-pl and green-pl-bgnd-neg alaa [’ar kal’á [bahhay-én nee tsee’at-én]]. but [of.f colors:con [grey-pl and yellow-pl ‘My sheet (is) not red and green but grey and yellow.’ In other cases the head noun is repeated, in full, or in the form of a construct case pronoun, ‘of ’ in (32). (32) hée [ya’ee tíiq] laqáa hée [tiqaaqíim] laqáa man-con [foot ill:m or man:con [being ill:hab:dur:m or /ameenír gwaagwa’ati laqáa ’ar [sihhimamiit] woman:f:con stillbirth or of.f [staying without pregnancy:3f ’adór ’i laq ’a-tí… kind:f:con subj3 do:3f:sbjv cop-indep.f:dem1 ‘Someone who has a bad foot, or someone who is ill all the time, or a woman who has had a stillbirth, or one who constantly stays without pregnancy, what she will do is this…’ Despite these cases in which the head noun is repeated, it is possible for modifiers of all sorts to be coordinated conjunctively, disjunctively or adversatively to form a single modifying unit.

4. Agreement A possible criterion leading us to consider nee as a coordinator rather than a comitative preposition is plural agreement. Indeed, if the coordinated noun phrase is the subject, it has plural agreement on the verb, as in (33). By contrast, the agreement is singular if the nee phrase has comitative meaning, or if the nee-phrase is post-verbal, as in (34). (33) kitaangw nee mesa i gwaranggwarimiit-iyá’ asmá kunseli. chair and table subj3 shake-3pl because earthquake ‘The chair and the table shake because of the earthquake.’ (34) ’ag’alé ’aayi ’i-n amó-r-d=i always mother subj3-prog place:con-f-dem4=dir hara-kí/ nee hárgarma-wos-ee asma towards-return:3sg.f:pres with daughter:in:law-3sg.poss-bgnd because tseehhee qasa-r-wa alé. manure placing-f-abl respro ‘Mother always returns home with the daughter-in-law to put the manure inside.’



Maarten Mous

However, the characteristics of clauses with the verb doog ‘meet’ in (35) suggest that plural marking on the verb does not exclude comitative meaning. The meeting is conceived from the perspective of the unexpressed subject and thus the person to meet with (a boy or a girl) is comitative but the verb ‘meet’ requires a plural subject and a reciprocal pronoun. In this case the plural agreement on the verb indicates that the unexpressed subject and the comitative noun can be conceptualized as plural. (35) nee masoomo bir-ta doog-iyé’ laqáa dasi… with youth cond-rec:perf meet-3pl:past or girl ‘If he meets a youth or a girl, …’ Thus, the test of plural agreement does not seem to be able to distinguish between a coordinated and a comitative reading. Our initial impression that plural agreement can distinguish between nee as a coordinator and as a comitative preposition is therefore incorrect. Instead, we have to look into the nature of subject agreement on the verb. Subject marking in the verbal piece has semantic aspects. A noun phrase in which people are coordinated can be referred to in two ways in the verbal piece: it can be referred to with 3pl marking on the verb and i third person marking on the selector, or with a default third person masculine subject marking on the verb and “impersonal” ta as subject marking in the selector, as in (36). This second option is used for a noun phrase that is conceived as a collective agent in which the individuals are not distinguished. Ta can only refer to human subjects, either a collective agent or an unspecific agent (impersonal subject). The first coordinand can be understood from context, as in (37). Thus, the subject is conceptualized as either singular (collective) or as plural, and this is reflected in the subject marking in the verbal piece. (36) ’aayi nee dasu-w-ós ta doo’=i meet. mother and girls-m-3sg.poss imps house=dir remain:pres ‘Mother and her girls stay at home.’ (37) bar hár garma ’i déer, nee hárgarma cond wife:con boy subj3 be:present:3sg.f with daughter:in:law ta-n dí-r doo’=i meet. imps-prog place:con-f house=dir remain:pres ‘If there is a daughter-in-law, the mother stays at home with the daughter-in-law.’ Preposed objects require an object pronoun agreeing in gender with the head noun of the object. In the case of coordinated nouns, this object pronoun has plural gender, regardless of the gender of the individual nouns.2

2.Plural is both an exponent of the category number and of the category gender in Iraqw (and in other Cush*tic languages). To avoid confusion I used the term Neuter for Plural as

The grammar of conjunctive and disjunctive coordination in Iraqw

(38) loosí nee kasiis ’i-na /ay-áan. beans.f and potatoes.f obj.p-past eat-1pl ‘We have eaten beans and potatoes.’ Subject agreement on the verb with a noun phrase containing laqáa is with the first noun, suggesting that the laqáa phrase is subordinated to the first head noun, as in (39). Object agreement in a noun phrase with laqáa is with the last noun, mahaangw ‘arrow.m’ in (40), suggesting a proximity effect for the resolution of the gender agreement conflict. (39) baabúu-w-ós laqáa aayo-r-ós ’i-n daqáy. father-m-poss.3sg or subj3-prog go:3sg.m ‘Its father or its mother will be going.’ (40) kwahlaahli laqáa mahaangw g-u-n haniis. bead.f or arrow.m obj.3-obj.m-prog give:3sg.m:pres ‘He will give him a bead or an arrow.’ To conclude, a phrase with nee can induce plural agreement on the verb, which shows, at least, that conjunction is part of the meaning of nee. However, the nature of subject agreement is such that it cannot distinguish between conjunctive and comitative nee.


Word category and origin

The language-internal definition of some of the word classes in Iraqw is straightforward. Nouns can be defined by their property of triggering gender agreement, verbs by their tense and person marking, adjectives by their special plural marking, verbal adverbs by their position immediately before the verb. What remains is a number of words for which it is impossible to show that they are nouns. Words like nee ‘and’, laqáa ‘or’ and aláa ‘but’ fall in this group. There is a clear, albeit small, class of prepositions: ar ‘with’, as ‘because’, ay ‘to’. All three are etymologically related to the case cl*tics, =r instrumental, =sa reason, =i directional respectively. As prepositions they precede the noun with which they form a phrase. Since they do so without receiving the construct case, they cannot be nouns themselves. These two properties define them as a separate word category. Following this definition, nee is also a preposition. Given their high final tone, laqáa ‘or’ and aláa ‘but’ could be defunct nouns in construct case. However, if laqáa were a noun in construct case, it should trigger gender agreement in sentences such as (6) where the agreement is

an exponent of Gender in Mous (1993). Words with this gender trigger plural subject agreement on the verb.


120 Maarten Mous

with yaamu ‘land’ and not with laqáa yaamu ‘or land’. Thus, laqáa cannot be a noun in construct case synchronically. The same can be argued for aláa ‘but’. So nee and laqáa have the same basic properties as the case-derived prepositions and complement the word class of prepositions. They are different from the casederived prepositions in one important respect. The case cl*tics indicate that somewhere in the sentence there is an NP with that particular role, not necessarily the NP to which the cl*tic is attached, and the same holds for the prepositions (see the case-derived instrumental preposition ’ar in (41)), but this does not hold for nee, nor for laqáa, which always form a syntactic and a semantic unit with what follows. (41) nee gu tsaxáar ’ar dasi. and obj.3:obj.m instr girl ‘And he beats the girl with it (the ball.m).’ Now that we have established a word class of prepositions and shown that both nee and laqáa are (deviant) members of this class, the question arises whether it is possible to define a word class of conjunctions in Iraqw. If such a word class exists, nee and laqáa are candidate members on the basis of the fact that they can link both phrases and clauses. However, the fact that nee and laqáa can occur clause initially does not distinguish them from the other prepositions. The instrumental preposition ’ar and the reason preposition ’as can be used as complementizers, as illustrated in (42) and (43). (42) ’i aldakuut ’ar do’ ’ahla ngw-a subj3 instr house fire obj.3:dep:obj.m-perf /ák. eat:3sg:f ‘He wonders whether the house has burnt down.’ (43) ’aníng ’a tseewa tláw ’as ni-wa hardah 1sg subj1/2 early rise:1sg reas dep.subj1sg-bgnd arrive:1sg:sbjv ’afíqooma-dá ni hlaa’. until-dem4 dep.subj1sg want:1sg:sbjv ‘I rise early in order to arrive when I want.’ The conditional bar can also appear clause-initially and seems to link clauses, see (24) above. But more often it appears in second or third position, immediately before the selector with which it forms a phonological unit, as in (44) and (3). (44) hare-r-ók bar-a tsahh-a-ká, ’a-qo cond-obj.f cop-emph naagáy. for:nothing ‘If you don’t recognize your wife, it is bad luck for you.’

The grammar of conjunctive and disjunctive coordination in Iraqw

Although nee and laqáa have functions similar to those of conjunctions, there is no language-internal evidence to warrant a separate word category of conjunctions. Other prepositions also occur clause-initially and other words that occur clauseinitially do so at other adjunct positions as well. Thus nee and laqáa are in the same word category as the prepositions ‘to’, ‘with’ and ‘because’ and all these prepositions have conjunction-like functions. The case-cl*tic-derived prepositions are different from the “coordinating” prepositions in several respects. The case-cl*ticderived prepositions do not link modifiers, for example, nor can they have a verb as complement, a property which nee and laqáa share: laq as a short form of laqáa, followed by the verb form hlaa’i ‘that he wants’ (without the usually obligatory inflectional element), is used in leading questions, as in (45); see also nee in (46). (45) ’aari ’i kôond-a, laqhlaa’i. goats obj.p have:2sg:int-inf isn’ ‘You have goats, isn’t that so?’ (46) ’aama nee huuringw. mother and cooking ‘And the mother was cooking.’ Summing up, Iraqw is one of those languages that use the same word to express the comitative (and agentive and comparative) role of an NP and for conjoining NPs (and clauses). [NP nee NP] may act as one unit, for example to be modified or as a modifier, but as such it is not necessarily a unit. Agreement phenomena, too, show the hybrid nature of this function word. In terms of word class, the question whether nee is a preposition or a conjunction turns out to be futile since the distinction in word class cannot be made on language-internal grounds. In this respect Iraqw is not very dissimilar from Bantu languages. In fact it is very likely that nee was borrowed from or at least influenced by Bantu na. In Gorwaa, Iraqw’s closest relative, there is variation between nee and naa or na. Alagwa, next in kin linguistically, has both naa and haa in comitative, agentive, and conjunctive functions, and also functioning as a copula, while this combination of functions is also found in Bantu. I have no comparative evidence on laqáa ‘or’.

Abbreviations 1/2 bgnd con cop dem(1–4)

first or second person background construct case copula demonstrative (distance 1–4)

dep dir emph hab hith imps

dependent clause directional emphasis habitual hither impersonal


122 Maarten Mous

indep int nar p past reas rec respro

independent pronoun interrogative narrative plural (as exponent of Gender) past reason reciprocal/reflexive resumptive pronoun

sbjv :


subjunctive between glosses of morphemes that are not separated in the Iraqw line between abbreviations within a gloss for one morpheme

References Berger, Paul and Roland Kiessling (ed.). 1998. Iraqw texts. Köln: Rüdiger Köppe. Haspelmath, Martin (to appear). “Coordination.” In Timothy Shopen (ed.), Language typology and syntactic description, second edition. Cambridge: Cambridge University Press. Marten, Lutz. 2000. “Agreement with conjoined noun phrases in Swahili.” AAP 64/Swahili Forum 7: 75–96. Mous, Maarten. 1993. A grammar of Iraqw. (Cush*tic Language Studies 9.) Hamburg: Helmut Buske.

Chapter 6

Coordinating constructions in Fongbe with reference to Haitian Creole* Claire Lefebvre Université du Québec à Montréal

1. Introduction 2. The conjunction b`f 2.1 b`f as a coordinating conjunction 2.2 b`f as a complementiser 2.3 A unified analysis of b`f 3. The conjunction bó 3.1 bó as a coordinating conjunction 3.2 bó as a complementiser 3.3 A unified account of bó 3.4 Does bó constitute a separate lexical entry from b`f? 4. The theoretical relevance of the properties of b`f and bó 4.1 On the marked character of and-then conjunctions 4.2 The disjoint/coreferential subjects distinction 4.3 From connective adverb to complementiser through conjunction of coordination 5. Can NPs be conjoined? 5.1 The kpó2ó…kpó ‘with…with’ circumposition 5.2 Fongbe and other with-type languages 6. Coordinating constructions in Haitian Creole 6.1 The clausal conjunction epi 6.2 Can NPs be conjoined in Haitian Creole?

*The content of this paper builds on earlier research on Fongbe (see Lefebvre and Brousseau 2002, and the references therein). The research underlying this paper has been funded by SSHRCC, FIR-UQAM and the Max Planck Institute for Evolutionary Anthropology, Leipzig. I would like to thank Marijo Denis and Virginie Loranger for their assistance in documenting the issues discussed in this paper, and Andrée Bélanger for formatting the manuscript. I am grateful to Bernard Comrie, Martin Haspelmath, the participants in the MPI seminar on coordinating constructions and to the participants in the McGill-UQAM research group on syntactic categories for their helpful comments and questions on an earlier version of this paper.

124 Claire Lefebvre

How did the properties of the Haitian lexical items get to be the way they are? 7. Concluding remarks: The typological features of Fongbe 6.3



This paper is concerned with coordinating constructions in Fongbe, a Kwa language spoken mainly in Benin, and in Haitian Creole, a Caribbean creole spoken mainly in Haiti. These two languages are historically related in a way that will be specified below. Akoha (1980: 210) identifies the Fongbe conjunctions in (1).1 b`f bó kpó/kpó2ó … kpó b. àm´f àlo¯ kàbı¯ àdì lobó lob`f hú có lo-có lo¯-´f

(1) a.

‘and’ ‘and then’ ‘and’ ‘but’ ‘or else’ ‘or else’ ‘or’ ‘and then’ ‘and then’ ‘then’ ‘but’ ‘nonetheless’ ‘while, but’ (from Akoha 1980: 210)

This paper is concerned only with the lexical entries in (1a): the clausal conjunctions b`f and bó, and the so-called NP conjunction made up of two lexical items, kpó2ó…kpó ‘with…with’. Anonymous (1983: IX, 1) glosses both b`f and bó as ‘and’. The author specifies that when b`f and bó occur in combination with lo yielding lob`f and lobó, respectively, the interpretation ‘and then’ obtains. Akoha (1980: 108 and 210, respectively) glosses b`f as ‘and, then’2 and bó as ‘and then’. He glosses both lob`f and lobó as ‘and then’ (p. 210). Hounkpatin (1985: 160 and 233, respectively) glosses b`f as ‘then’ and bó as ‘and’. As can be seen from this brief review of the literature, there is variation among authors as to the meaning of b`f and bó.

1.The orthographic conventions used in this paper correspond to the official orthographic conventions of Benin (for details, see Lefebvre and Brousseau 2002: 29–37). 2.Note, however, that in (1) b`f is glossed as ‘and’ by the same author.

Coordinating constructions in Fongbe with reference to Haitian Creole

According to my informants,3 when b`f and bó coordinate clauses that are in the perfective aspect, the coordinate clauses are generally interpreted as denoting related events occurring sequentially, and b`f and bó are both glossed as ‘and then’. This is illustrated in (2). K`fkú wá b`f Àsíbá yì. Koku arrive conj Asiba leave ‘Koku arrived and-then Asiba left.’ b. K`fkú 2ù nú bó nù sín. Koku eat thing conj drink water ‘Koku ate and-then drank water.’

(2) a.

In contexts such as those in (3), however, even though the two clauses coordinated by b`f and bó occur in the perfective aspect, they are interpreted as denoting two independent events (in terms of both sequentiality and causality) and the two conjunctions are glossed as ‘and’. This is illustrated in (3). M´7 wé 2ù-2é-jí b`f m`7 wé xó-kpò. person two win and person two fail ‘Two persons won and two persons failed.’ b. K`fkú 2ù-2é-jí 2ò wèzù m`7 bó xó-kpò 2ò. Koku win run in and-he lose kàn-línl´fn m`7 jump in ‘Koku won at the race and lost at the jump.’

(3) a.

When b`f and bó coordinate clauses that occur in the imperfective aspect, the coordinate clauses are always interpreted as denoting events that may be unrelated and that may occur simultaneously. This is illustrated in (4). K`fkú 2ò wíwá w`7 b`f Àsíbá 2ò yìyì w`7. Koku arriving post conj Asiba leaving post ‘Koku is arriving and Asiba is leaving.’ b. K`fkú 2ò nú 2ù w`7 bó 2ò sín nù w`7. Koku thing eat post conj water drink post ‘Koku is eating and drinking water (at the same time).’

(4) a.

It thus appears that the variation between authors as regards the meaning of the conjunctions b`f and bó finds an explanation when the aspectual properties of the clauses they conjoin are taken into account.

3.Several informants provided the original data discussed in this paper. They are named in the Preface to Lefebvre and Brousseau (2002). Marcellin Gangbe provided me with subtle judgements on data that are crucial for the analyses presented in this paper.


126 Claire Lefebvre

The difference between b`f and bó lies in the fact that, while the former basically coordinates clauses with referentially disjoint subjects, hence clauses involving switch-reference, the latter is restricted to coordinating clauses with coreferential subjects. This is illustrated in (5) and (6), respectively. (5) K`fkú wá b`f Àsíbá yì. Koku arrive conj Asiba leave ‘Koku arrived and-then Asiba left.’ (=(1) in Lefebvre and Brousseau 2002: 113) (6) Ùni wá bói yì. 1sg arrive conj leave ‘I arrived and-then left.’ (=(3a) in Lefebvre and Brousseau 2002: 113) The difference in the referential properties of the subjects of clauses coordinating by b`f and bó has been noted on several occasions (e.g. Akoha 1980: 210; 1990: 229–234; Anonymous 1983: IX, 1). For some speakers reported on in Lefebvre and Brousseau (2002: 113), b`f can also coordinate clauses whose subjects are coreferential, as is illustrated in (7). (7) K`fkú wá b`f é l´7k`f yì. Koku arrive conj 3sg again leave ‘Koku arrived and-then he left again.’ Another difference between clauses coordinated by b`f or by bó lies in the fact that, when the two clauses are coordinated by b`f, the subject of the second conjunct clause must be overt, as in (5) and (7), whereas when the two clauses are coordinated by bó, the subject of the second conjunct must be covert, as in (6). This discrepancy raises the question of the syntactic status of b`f and bó. What features do they have in common, and what features distinguish them? Are they both clausal coordinators, as is generally assumed in the literature cited above, or could they be distinguished on the basis of the type of constituents that they are coordinating, e.g. clauses versus verb phrases? It will be argued that b`f and bó are both clausal coordinators, and that furthermore, they can only coordinate clauses. Another set of facts concerning these two lexical items is that, as will be seen below, in addition to being used as coordinating conjunctions, both can serve as complementisers in specific contexts. This raises the question of whether there are two different b`fs and two different bós, or alternatively, whether it is possible to account in a unified way for the properties of b`f and for those of bó. My theoretical standpoint on this issue is that monosemy is to be preferred over polysemy wherever possible (see also Bouchard 1995; Cowper 1989, 1995; Ghomeshi and Massam 1994; Johns 1992; Lefebvre 1999; Nida 1948; Ruhl 1989; etc.). I assume the One Form/One Meaning Principle as formalised in Johns (1992: 84): “Where morphemes are identical or similar in phonological properties, in the unmarked

Coordinating constructions in Fongbe with reference to Haitian Creole 127

case, they are identical or similar in all lexical properties”. Assuming this general principle, one should avoid proposing several lexical entries with the same phonological form provided that the meanings corresponding to these forms are semantically related. I will argue that it is possible to provide a unified analysis for b`f and a unified analysis for bó. Another property of b`f and bó is that they cannot be used to coordinate noun phrases. Since bó occurs exclusively in contexts involving two coreferential subjects, it is not expected to occur in the context of NPs. However, b`f, being freer in this respect, might be expected to occur in the context of NPs. Nevertheless, b`f cannot coordinate two NPs, as is shown by the ungrammaticality of phrases like (8). (8) *K`fkú b`f Àsíbá Koku conj Asiba The use of distinct coordinators for NPs and sentences is very widespread crosslinguistically. But why can’t b`f conjoin noun phrases? To my knowledge, a sound explanation of this fact has not been provided as yet. An account of this distribution will be proposed in § 2. The properties of b`f and bó enumerated so far show that these two lexical items are quite similar. Both share the core meaning ‘and/and then’, both can serve as a coordinator of clauses and as a complementiser. Neither can coordinate NPs. The difference between them is that, while bó is restricted to coordinating clauses with coreferential subjects, b`f coordinates clauses with either referentially disjoint or with coreferential subjects. This situation raises the question of whether b`f and bó could be analysed as contextually determined allomorphs. Although this may be an appealing way of looking at the data at first glance, it will be argued that this cannot be the correct analysis. Although b`f and bó appear to have a rather similar distribution in the linguistic contexts focused on in this paper, bó has a wider distribution than b`f, and therefore, b`f and bó cannot be analysed as contextually determined allomorphs. The equivalent of coordination of NPs is achieved by means of a circumposition involving adpositions meaning ‘with’, a typologically common strategy. This is illustrated in (9). (9) K`fkú kpó(2ó) Àsíbá kpó/kpán Koku with Asiba with/with ‘Koku and Asiba’ Some authors consider the circumposition in (9) as a NP conjunction and gloss it as ‘and’ (see e.g. Akoha 1980: 210; Anonymous 1983: VII, 1). It will be argued that the circumposition occurring in (9) also occurs in comitative, instrumental and manner constructions, and that in all of its occurrences, the phrase containing the circumposition kpó2ó…kpó/kpán is a syntactic adjunct. There thus appears to be no true NP conjunction in Fongbe.

128 Claire Lefebvre

In the course of the last twenty years, Fongbe has come to be known as an important substratum language of some Caribbean creoles (see e.g. Lefebvre 1986, 1998, and the references therein; Lefebvre and Kaye 1986; Singler 1996). In Lefebvre (1998), it is argued that the properties of a significant portion of the West African lexicons have been reproduced in Haitian Creole through the process of relexification. The question arises as to whether the particular properties of the Fongbe lexical items involved in clausal and NP coordination were in fact carried over into Haitian Creole through the process of relexification. This issue will be taken up in § 6. It will be shown that, to a large extent, the properties of the Fongbe lexical items involved have been reproduced in the creole. The paper is organised as follows. Section 2 discusses the properties of b`f and proposes a unified account of these properties. Section 3 does the same for bó. It ends with a subsection addressing the question of whether b`f and bó really constitute separate lexical entries. Section 4 addresses theoretical issues raised by the properties of b`f and bó, including the sources of the functions of b`f and bó. Section 5 discusses the facts concerning coordination of NPs or lack thereof. Section 6 compares the Haitian Creole data with the Fongbe data within the framework of the relexification account of creole genesis as formulated in Lefebvre (1998 and the references therein). Section 7 concludes the paper with remarks on the typological features of the Fongbe lexical items discussed in this paper. The data discussed in this paper are drawn either from the literature, in which case they are identified as such, or from my field notes, in which case no source is mentioned. Variation in the data between authors or between informants will be pointed out throughout.


The conjunction b`f

This section bears on the properties of b`f. Its properties as a coordinating conjunction and as a complementiser are discussed in turn in the first two subsections. A unified analysis of b`f is proposed in § 2.3. 2.1 b` f as a coordinating conjunction In (5), b`f coordinates clauses having referentially disjoint subjects. In (7), b`f coordinates clauses having coreferential subjects. The conjunct clause introduced by b`f has to have an overt subject (see (5) and (7)). In (10), the subject of the second conjunct is not overt and the sentence is not grammatical (compare (10) with (5) and (7)).

Coordinating constructions in Fongbe with reference to Haitian Creole 129

(10) *K`fkú wá b`f yì Koku arrive conj leave [Lit.: ‘Koku arrived and-then (s)he left.’] Since the second clause coordinated by b`f has to have an overt subject, b`f is excluded from contexts where there is no overt subject. The various sets of data presented below document this distributional property. First, the infinitival complement of verbs of the ‘want’ class has no overt subject when the subject of the main clause and that of the embedded clause are coreferential. This is shown in (11). (11) K`fkú jló ná nù sìn. Koku want def.fut drink water ‘Koku wants to drink water.’ b`f cannot coordinate two complement clauses of a verb of the ‘want’ class, as is shown by the ungrammaticality of (12). (12) *K`fkú jló ná nù sìn b`f 2ù nú Koku want def.fut drink water conj eat thing [Lit.: ‘Koku wants to drink water and to eat.’] Second, the complement of the modal verb sìxú ‘may’ is an infinitival complement lacking an overt subject, as is shown in (13). (13) K`fkú sìxú wá. Koku may come ‘Koku may come.’ b`f cannot conjoin two complements of sìxú, as is shown by the ungrammaticality of (14). (14) *K`fkú sìxú wá b`f yì Koku may come conj go [Lit.: ‘Koku may come and go.’] Finally, some contexts requiring deverbal nominalisations do not allow for an overt subject. The complement of the aspectual verb meaning ‘to begin’ constitutes such a context. This verb selects a complement headed by the postposition jí ‘on’, which, in turn, selects a nominalised VP. This nominalised phrase contains no overt subject, as is shown in (15). As is extensively discussed in Lefebvre and Brousseau (2002: 195–215), in nominalisation contexts, the object precedes the deverbal noun.

130 Claire Lefebvre

Hence, in (15), the object nú ‘thing’ precedes the nominalised verb 2ù ‘eating’.4 (15) Ùn j`7 [[nú 2ù] jí]. 1sg fall [[thing eating on ‘I began eating.’ (=(136) in Lefebvre and Brousseau 2002: 287) The lexical item b`f cannot conjoin complements so formed, as is shown by the ungrammaticality of (16). (16) *ùn j`7 nú 2ù b`f sìn nù jí 1sg fall thing eating conj water drinking on [Lit.: ‘I began eating and drinking water.’] The sentence in (16) can be rescued as (17), where b`f conjoins two full clauses with overt subjects. (17) Ùn j`7 nú 2ù jí b`f ùn j`7 sìn nù jí. 1sg fall thing eating on conj 1sg fall water drinking on ‘I began eating and-then I began drinking water.’ The ungrammatical data in (10), (12), (14) and (16) all show that b`f cannot conjoin clauses lacking an overt subject. The ungrammaticality of the sentences in (14) and (16) further shows that b`f cannot conjoin VPs, regardless of whether they are nominalised (as in (16)), or not (as in (14)). The fact that b`f is excluded from contexts lacking an overt subject (that is, infinitival clauses of the type in (12) and (14), and nominalisations of the type in (16)) suggests that b`f cannot conjoin nonfinite clauses.5,6

4.In deverbal nominalisations, the verb appears in its reduplicated form unless it has an overt object, or some other particle preceding it (see Fabb 1992a, 1992b). When the nominalised verb has an overt object, the nominalised verb appears in its basic form and it is preceded by its object. Hence: wíwá ‘arrival’ < wá ‘to arrive’, and nú 2ù ‘eating’ from 2ù nú ‘to eat’, where nú ‘thing’ is the generic inherent object of the verb 2ù. 5.In Fongbe, there is no tense morphology. The temporal interpretation of a clause is computed from the various components of the clause that participate in establishing its aspectual properties (see Lefebvre and Brousseau 2002: 85–113, and the references therein). Non-infinitival clauses are here referred to as finite clauses. Finite clauses must have an overt subject. This subject occurs in the nominative case; this case is visible when the subject is a pronominal cl*tic: [+nominative] pronominal cl*tics bear a high tone, as opposed to [−nominative] pronominal cl*tics that bear a low tone. Finite clauses also contrast with nonfinite clauses in allowing markers that give the speaker’s point of view on the proposition, including the negative marker (see e.g. (36)). 6.This conclusion would gain support if it could be shown that b`f is also excluded from infinitival clauses containing an overt subject. In addition to the infinitival structure in (11), in which the subject is covert, Fongbe exhibits another infinitival structure, in which the

Coordinating constructions in Fongbe with reference to Haitian Creole

Summarising: b`f coordinates clauses (not VPs). Clauses coordinated by b`f must have overt subjects. Consequently, b`f is excluded from infinitival clauses lacking an overt subject and from deverbal nominalisations. Finally, b`f is excluded from nonfinite clauses. The latter claim will be shown to gain support from the distribution of b`f occurring as a complementiser. 2.2 b` f as a complementiser The lexical item b`f may also be used to introduce the clausal complement of the prepositions káká ‘until’ and có ‘before’. In (18), b`f introduces the clausal complement of the preposition káká ‘until’. In this context, b`f is optional (a fact that is represented by the parentheses in the examples below). The example in (18a) shows that the subjects of the two clauses related by b`f may be referentially disjoint. The example in (18b) shows that (for some speakers) the subjects of the two clauses related by b`f may be coreferential. In either case, the subject of the second conjunct has to be overt.

subject is overt. The latter structure is exemplified in (i). As is the case in the corresponding structure in English, the subject of the main clause and that of the infinitival clause have to be referentially disjoint. Furthermore, as is again the case in the corresponding structure in English, the subject of the infinitival clause bears accusative case. In English, accusative case is visible in the suppletive form him of the third person pronoun. In Fongbe, this case is manifested by the low tone on the third person cl*tic. (For an extensive discussion of this structure, see Lefebvre and Brousseau 2002: 280–281.) (i) É jló è yì. 3sg want 3sg go ‘He wants him/her to go.’ (=(116) in Lefebvre and Brousseau 2002: 281) If b`f is not allowed to conjoin two infinitival complements of the type seen in (i), there would be additional evidence supporting the claimed relationship between b`f and finiteness. Hence, (ii) is predicted to be ungrammatical. (ii) *é jló wè yì b`f è wá 3sg want 2sg leave conj 3sg come [Lit.: ‘He wants you to leave and him/her to come.’] It should be possible to rescue (ii) as (iii). (iii) É jló wè yì [b`f é] jló è wá. 3sg want 2sg leave conj 3sg want 3sg come ‘He wants you to leave and he wants him/her to come.’ Unfortunately, I do not have this piece of data in my notes, and the sole informant who is available to me at the time I am writing this paper does not have the infinitival structure of the type in (i) in his grammar. The result of this test will thus have to await future research. On the basis of the data of the type seen in (11) to (17), however, I will assume that b`f is restricted to conjoining finite clauses.



Claire Lefebvre

K`fkú y´fl´f è káká (b`f) é wá. Koku call 3sg until conj 3sg come ‘Koku called her/him until (s)he came.’ b. K`fkú 2ù nú káká (b`f) é j`7 àz`fn. Koku eat thing until conj 3sg fall sick ‘Koku ate until he fell sick.’ (=(19a) in Lefebvre and Brousseau 2002: 118)

(18) a.

The lexical item b`f also introduces the clausal complement of có, which, in one of its uses, may be glossed as ‘before’, as is illustrated in (19a). In (19b), the temporal clause has been topicalised. In Fongbe, topicalised constituents are headed by the definite determiner f´. Note that, when the subordinate clause follows the matrix, the definite future marker is optional, as in (19a), whereas it is obligatory when the subordinate clause precedes the matrix, as in (19b). K`fkú kò yì có b`f à (ná) wá. Koku ant leave before conj 2sg def.fut arrive ‘Koku had left before you arrived.’ b. Có b`f à ná wá f´, K`fkú kò yì. before conj 2sg def.fut arrive def Koku ant leave ‘Before you arrived, Koku had left.’ (=(122) in Lefebvre and Brousseau 2002: 172)

(19) a.

The clausal complements of káká and có (see (18) and (19), respectively) are obligatorily finite; that is, there is no infinitival alternative available. So, as a conjunction of subordination, b`f introduces finite clauses. As a conjunction of subordination, b`f has the properties of complementisers. First, in (18) and (19), b`f occurs at the begining of the complement clause, before the subject. This is the position where we find clause-initial complementisers (e.g. that in English). Second, in (18) and (19), b`f introduces only finite clauses. Complementisers may be specified for whether they introduce finite or non-finite clauses (e.g. that [+finite] versus for [−finite] in English). Third, b`f is selected by the prepositions káká and có. To my knowledge, b`f is selected by no other preposition. It is a property of complementisers to be selected by specific lexical items or by classes of lexical items (e.g. in English, some verbs select that as a complementiser, others select to). Fourth, as can be seen in (18a) and (18b), the realisation of b`f is optional in the context of káká. Complementisers are optionally realised in specific contexts (e.g. English He thinks (that) he will come). Fifth, b`f occurs in the same position as the complementiser nú does. (The complementiser status of nú is discussed in Lefebvre and Brousseau (2002: 116–117).) Compare (20a) and (21a), and (20b) and (21b). Note that in (20b) and (21b), the temporal clause has been topicalised.

Coordinating constructions in Fongbe with reference to Haitian Creole

K`fkú kò yì có b`f à wá. Koku ant leave before conj 2sg arrive ‘Koku had left before you arrived.’ b. Có b`f à ná wá f´, K`fkú kò yì. before conj 2sg def.fut arrive def Koku ant leave ‘Before you arrived, Koku had left.’ (=(122) in Lefebvre and Brousseau 2002: 172)

(20) a.

D´f àml`fn có nú à ná wá. sleep sleep before comp 2sg def.fut come ‘Sleep before you come.’ b. Có nú à ná wá f´, d´f àml`fn. before comp 2sg def.fut come def sleep sleep ‘Before you come, sleep.’ (=(123) in Lefebvre and Brousseau 2002: 172)

(21) a.

In the context of (20) and (21), nú and b`f are interchangeable, as is illustrated in (22a) and (22b). According to my informants, the choice of either one of the two forms entails no difference in meaning. (22) a.


àml`fn có

{ nú } à ná wá. { b`f } sleep sleep before comp 2sg def.fut come ‘Sleep before you come.’ b. Có { nú } à ná wá f´, d´f àml`fn. { b`f } before comp 2sg def.fut come def sleep sleep ‘Before you come, sleep.’

Since nú is a complementiser (see (21)), and since b`f can occur in complementary distribution with it (see (22)), the analysis that b`f is a complementiser in (18), (19) and (20) is a likely one. There are thus five arguments supporting the claim that, in the context of káká and of có, b`f serves as a complementiser: position, finiteness, selectional properties, optionality and complementary distribution with the complementiser nú. 2.3 A unified analysis of b` f In § 2.1, we saw that b`f serves as a conjunction of coordination, and in § 2.2, we saw that it serves as a conjunction of subordination, and more precisely, as a complementiser. The double function of b`f raises the question of whether it is necessary to postulate two lexical entries for b`f. Alternatively, is it possible to provide a unified account for this lexical item? In the introduction to this paper, I took the theoretical


134 Claire Lefebvre

standpoint that monosemy is to be preferred over polysemy wherever possible. Is a monosemic analysis of b`f supported by the properties of this lexical item across the environments in which it occurs? I argue below that the data support a monosemic analysis of b`f. First, in both coordination and subordination contexts, b`f relates two clauses with subjects that are either disjoint (see (5) and (18a)) or coreferential (see (7) and (18b)). In both cases, the subject of the second conjunct or of the subordinate clause must be overt. This appears to be a consequence of the fact that, in both environments, b`f only occurs in the context of finite clauses. It thus appears that the properties of b`f conjoining two clauses are the same as those of b`f introducing the sentential complement of káká ‘until’ and of có ‘before’. The difference between the two contexts is that, in one case, b`f serves as a conjunction of coordination, whereas in the other one, it serves as a complementiser. Interestingly enough, there are contexts of occurrence of b`f where its semantics seems intermediate between that of a coordinating and that of a subordinating conjunction. For example, in the context of the temporal adverbial clause in (23), b`f is intermediate between being interpreted as a coordinating conjunction (e.g. ‘Koku arrived and-then Asiba left’), or as a subordinating one (e.g. ‘It is as soon as Koku arrived that Asiba left’). (23) Wá K`fkú wá (tlóló) b`f Àsíbá yì. arrive Koku arrive ( conj Asiba leave ‘As soon as Koku arrived, Asiba left.’ (=(120) in Lefebvre and Brousseau 2002: 172) Similar examples are provided by Anonymous (1983:VI, 7). One of them is reproduced as (24); it can be paraphrased either as ‘Something happened (what is it) and then you came’, or as ‘What happened that you came/that caused you to come?’. (24) N`7 (w`7) ká gb`fn b`f à wá. what ( adv happen conj 2sg come ‘What happened and-then/that you came.’ (Anonymous 1983: VI, 7) The type of fuzziness reported above has been noted in the literature (see e.g. Payne’s 1985 discussion of similar cases on the basis of Fijian data).7 The ambiguity relative to the interpretation of b`f in (23) and (24) is possibly

7.Haspelmath (1995, to appear) notes that in most cases, subordination structures may be distinguished from coordination structures on the basis of syntactic tests. I believe that this is correct. Of the four tests he formulates, however, none apply to the structure in (24). This suggests that tests distinguishing between subordination and coordination structures are, to a large extent, language specific. Tests distinguishing between these structures in Fongbe remain to be designed.

Coordinating constructions in Fongbe with reference to Haitian Creole

related to the fact that, regardless of its grammatical function as a marker of coordination or as a marker of subordination, b`f generally relates two events that are sequentially ordered. Hence, the ordering of events denoted by (5) can be paraphrased as (25a), that by (18b) as (25b), that by (19) as (25c), that by (21) as (25d), that by (23) as (25e), and that by (24) as (25f). (25) a. b. c. d. e. f.

‘Koku arrived and-then Asiba left’ ‘Koku ate to the point that then he fell sick.’ ‘Koku left and-then you arrived.’ ‘Sleep and-then come.’ ‘Koku arrived and-then Asiba left.’ ‘Something happened and-then you came.’

Note that the surface order of the conjuncts does not need to reflect the sequential order of events in (25). For example, in the (a) version of (19), (20) and (21), the temporal clause follows the matrix clause. In the (b) version of the same sentences, the temporal clause has been topicalised and hence, it precedes the matrix.8 Given that b`f generally relates events that are sequentially ordered, it is not surprising to find that the second conjunct of two clauses related by b`f may be assigned a purposive interpretation, as is illustrated in (26). Note that the clause interpreted as purposive must contain the definite future marker ná. (26) Ùn x`f wémâ 2ókpó b`f à ná sìxú xà. 1sg buy book one conj 2sg def.fut may read ‘I bought a book so that you may read it.’ On the basis of the above discussion, I conclude that it is possible to describe the properties of b`f in a unified way, and that these can be recorded within a single lexical entry. This lexical entry would minimally contain the information informally identified in (27). (27) b`f: coordinator and complementiser [+finite] The fact that the subject of the clause introduced by b`f has to be overt follows from the [+finite] character of b`f, and hence, of the clauses that it relates. This information does not need to be specified in the lexical entry because this is what is expected: finite clauses do have overt subjects. The fact that b`f can relate two clauses whose subjects are either referentially disjoint or not does not need to be specified either, for it is also the unmarked case. The reason why b`f does not occur with NPs

8.For a discussion of the fact that some languages impose a surface constraint on the order of temporally ordered propositions, see Longacre (1985).


136 Claire Lefebvre

follows from the feature [+finite] associated with it. Typically, NPs are not identified for finiteness. Consequently, they cannot be related by b`f.


The conjunction bó

Like b`f, bó serves both as a coordinating conjunction and as a complementiser. The properties of bó in each of these two functions are discussed in § 3.1 and § 3.2, respectively. Section 3.3 proposes a unified analysis of bó. Given the fact that bó and b`f share a number of properties, the question arises as to whether they constitute two separate lexical entries. This question is addressed in § 3.4, where it is argued that bó and b`f do indeed constitute separate lexical entries. 3.1 bó as a coordinating conjunction The lexical item bó conjoins clauses whose subjects are coreferential. As is shown in (28a–e), this applies throughout the person paradigm. Recall from (6) that, in this case, the subject of the conjunct clause is not, and cannot be, overt. (28) a.

Ùn wá bó yì. 1sg arrive conj leave ‘I arrived and-then I left.’ b. À wá bó yì. 2sg arrive conj leave ‘You arrived and-then you left.’ c. É wá bó yì. 3sg arrive conj leave ‘(S)he arrived and-then (s)he left.’ d. Mí9 wá bó yì. 1/2pl arrive conj leave ‘We/you arrived and-then we/you left.’ e. Yé wá bó yì. 3pl arrive conj leave ‘They arrived and-then they left.’

9.As is discussed at length in Lefebvre and Brousseau (2002: 61–63), first and second person plural personal pronouns are rendered by the same form, and likewise first and second person plural cl*tics. In other words, Fongbe does not distinguish between first and second person plural.

Coordinating constructions in Fongbe with reference to Haitian Creole

In order to account for the fact that the subject of the second conjunct cannot be overt in the environment of bó, I will assume that bó binds the subject position of the second conjunct. Adjacency is required for bó to bind this position.10 In order to account for the fact that the subjects of the clauses conjoined by bó must be coreferential in (28), I will assume that the subject of the first conjunct and bó are coindexed. In this view, the two subject positions are related through bó. All three positions form a chain, schematically represented in (29), where [e] stands for ‘empty position’. (29) DPi…………… bói[e]i …………… According to the representation in (29), bó coordinates two clauses. The data in (30) and (31) may, however, lead one to the conclusion that bó may coordinate phrases that are smaller than a clause. In (30), bó conjoins two finite clauses with jló ‘to want’ occurring as the main verb in each clause. The verb jló may be omitted (that is, not pronounced) from the second conjunct, a fact that is represented by parentheses in the example. (30) K`fkú jló ná nù sìn bó (jló) ná 2ù nú. Koku want def.fut drink water conj (want def.fut eat thing ‘Koku wants to drink water and (he wants) to eat.’ In (31), the verb sìxú occurring in the second conjunct may be left unpronounced. (31) K`fkú sìxú wá bó (sìxú) yì. Koku may come conj (may leave ‘Koku may come and (he may) leave.’ The ellipses in (30) and (31) may be analysed as stylistic (that is, they would have the function of avoiding repetitions) rather than as syntactic. The fact that bó is otherwise not allowed to conjoin VPs or non-finite complements supports this claim. For example, bó cannot conjoin two complements of the verb ‘to begin’, as is shown by the ungrammaticality of the sentence in (32).

10.The morpheme bó may combine with nú to form the complex expression bó-nú ‘in order that’. In this case, bó and the following subject position are no longer adjacent, and bó cannot bind the subject position of the conjunct clause anymore. If so, the subjects of the two clauses related by bó cannot be interpreted as being coreferential; in fact, in this case, a disjoint reference is obligatorily induced. This is depicted in (i). (i) Ùn jì hàn (bó)-nú à ní kò-nú. 1sg produce song conj for 2sg irr smile ‘I sing in order to make you smile.’ (=(130) in Lefebvre and Brousseau 2002: 174) Similar examples are provided in Akoha (1980: 211, 1990: 266–273, 278, 290–293) and Anonymous (1983: IX, 6).



Claire Lefebvre

(32) *ùn j`7 nú 2ù bó sìn nù jí 1sg fall thing eat conj water drink on [Lit.: ‘I began eating and drinking water.’] The sentence in (32) can be rescued as (33), where bó coordinates two full finite clauses. (33) Ùn j`7 nú 2ù jí bó j`7 sìn nù jí. 1sg fall thing eat on conj fall water drink on ‘I began eating and I began drinking water.’ Moreover, the ungrammaticality of the sentence in (34) shows that bó cannot conjoin verbs. (34) *K`fkú n`f dó bó sá kwèkwè Koku hab cultivate conj sell banana [Lit.: ‘Koku cultivates and sells bananas.’] Again, the sentence in (34) can be rescued as (35), where bó conjoins two finite clauses. (35) K`fkú n`f dó kwèkwè bó n`f sá (è). Koku hab cultivate banana conj hab sell (3 ‘Koku cultivates bananas and sells them.’ The data in (32) and (33), (34) and (35) show that bó cannot conjoin VPs or Vs. On empirical grounds, it is not possible to exclude the possibility that bó could conjoin phrases that are smaller than a clause but larger than a VP. I leave further discussion of this possibility open for future research. Since bó entails that the conjuncts it relates have coreferential subjects, it is not expected to occur as a NP conjunction. This prediction is borne out, as bó is excluded from NPs. As is pointed out in Anonymous (1983: IX, 3), even in the case of NPs of the type ‘He does not eat salt or pepper’ the coordination is rendered by two finite clauses related by bó. This is illustrated in (36). Note that the presence of the negative marker aˇ in (36) argues for the finite character of the clause, for, as is shown in Lefebvre and Brousseau (2002: 128–130), the negative marker aˇ only occurs in finite clauses.11 (36) É n`f 2ù j`7 bó n`f 2ù tàkín aˇ . 3sg hab eat salt conj hab eat pepper neg ‘He does not eat salt or pepper.’ (from Anonymous 1983: IX, 3)

11.The negative marker in (36) (to be distinguished from the negation marker mà) is part of the paradigm of markers that give the speaker’s point of view on the proposition. As per the analysis in Lefebvre 1998, these markers have scope over the proposition that they are part of.

Coordinating constructions in Fongbe with reference to Haitian Creole 139

Summarising: the data in (28), (33) and (35) suggest that bó coordinates clauses. The data in (32) show that bó cannot coordinate nominalised VPs. The data in (34) show that bó cannot coordinate Vs or truncated VPs. Since bó does not coordinate NPs either, I conclude that bó can only coordinate clauses. As we saw above, clauses coordinated by bó have to be finite (see (36); also (33) and (35)). The distribution of bó in contexts of subordination (discussed in § 3.2 below) further argues that bó only occurs in finite clauses. The fact that, on the one hand, bó does not coordinate VPs, and the fact that, on the other hand, it is restricted to finite clauses, further support the suggestion that the ellipses in (30) and (31) are stylistic rather than syntactic.12 3.2 bó as a complementiser As is the case of b`f, bó can be used as a complementiser. As such, it introduces the clausal complement of the preposition káká ‘until’, as is shown in (37). K`fkú 2ù nú káká bó j`7 àz`fn. Koku eat thing until conj fall sick ‘Koku ate until he got sick.’ (=(19a) in Lefebvre and Brousseau 2002: 118) b. K`fkú kán-wèzùn káká bó wá. Koku run until conj arrive ‘Koku ran until he arrived.’

(37) a.

The form bó also introduces the clausal complement of có in its use meaning ‘before’, as is illustrated in (38a) and (38b). In (38b) the temporal clause has been topicalised. K`fkú kò 2ù nú có bó yì. Koku ant eat thing before conj leave ‘Koku had eaten before he left.’ b. Có bó ná yì f´, K`fkú kò ù nú. before conj def.fut leave def Koku ant eat thing ‘Before he left, Koku had eaten.’

(38) a.

12.An account of possible contexts of ellipses in Fongbe is far beyond the scope of this paper. To my knowledge, the contexts in (30) and (31) are among the rare ones which allow for ellipses in the language (but see Lefebvre and Brousseau (2002: 67–70) for other cases of ellipses). The question of why ellipsis is permitted in these two contexts is a topic for future research. For discussions of the motivation for ellipsis, see Haspelmath (to appear: 34 and following, and the references therein).

140 Claire Lefebvre

In the above examples, bó has all the characteristics of a complementiser. The arguments supporting this analysis are of the same type as those used in the discussion of b`f. First, as is the case of b`f, bó occurs at the beginning of the complement clause of káká and có (see (37) and (38)). This is the position where we expect complementisers to occur. Second, bó is [+finite] since it introduces only finite clauses (see e.g. (37), (38)). Recall from § 2.2 that káká and có select only [+finite] clausal complements. Complementisers are either finite or non-finite. Third, bó is selected by the prepositions káká and có. It is a property of complementisers to be selected. Fourth, bó is obligatory in the context of káká and có. This follows from the analysis that bó binds its adjacent subject position. Complementisers that bind their adjacent subject position are obligatory (see e.g. French qui).13 Fifth, the syntactic position occupied by bó introducing a subordinate clause is the same as that occupied by the complementiser nú. This is shown in (39) and (40). Note, however, that, since the complementiser nú does not have the property of binding its adjacent subject position, the subject position following nú is obligatorily spelled out in (40), in contrast to that following bó in (39). (39) D´f àml`fn có bó (ná) wá. sleep sleep before conj (def.fut come ‘Sleep before you come.’ (40) D´f àml`fn có nú à ná wá. sleep sleep before comp 2sg def.fut come ‘Sleep before you come.’ According to my informants, there is no difference in meaning between (39) and (40). The fact that bó occurs in complementary distribution with the complementiser nú supports the analysis according to which bó in (39) is a complementiser. Bó also occurs in purposive clauses. In this case, it is obligatorily followed by the definite future marker ná.14 Examples of this structure are given in (41) and (42). (41) Ùn ná yì bó ná wà àz`f. 1sg def.fut go conj def.fut do work ‘I will leave in order to work.’ (=(127) in Lefebvre and Brousseau 2002: 174)

13.All accounts of the distribution of the French complementiser qui hold some version of an analysis according to which qui binds the subject position that it is adjacent to, see e.g. Kayne (1981). 14.Note that bó and ná can be contracted as [bá].

Coordinating constructions in Fongbe with reference to Haitian Creole

(42) Àsíbá x`f lìnfín bó ná 2à w´f. Asiba buy flour conj def.fut prepare dough ‘Asiba bought flour in order to prepare dough.’ (=(128) in Lefebvre and Brousseau 2002: 174) The sentence in (43a) shows that a purposive clause introduced by bó can be topicalised. The sentence in (43b) shows that it can be clefted; in this case, the purposive clause must contain the word wútú ‘cause’.15 Bó ná 2à w´f f´, Àsíbá x`f lìnfín. conj def.fut prepare dough def Asiba buy flour ‘In order to prepare the dough, Asiba bought flour.’ (=(129a) in Lefebvre and Brousseau 2002: 174) b. Bó ná 2à w´f wútú w`7, Àsíbá x`f lìnfín. conj def.fut prepare dough cause Asiba buy flour ‘It is in order to prepare dough that Asiba bought flour.’ (=(129b) in Lefebvre and Brousseau 2002: 174)

(43) a.

In the two sentences in (43), bó cannot be analysed as a conjunction of coordination. It is best analysed as a complementiser. There are thus five arguments supporting the claim that, in addition to fulfilling the function of coordinator, bó also fulfills the function of complementiser when occurring in the context of káká, có and purposive clauses: position, finiteness, selectional properties, obligatoriness (due to the fact that bó binds its adjacent subject position), and complementary distribution with the complementiser nú. 3.3 A unified account of bó In § 3.1, we saw that bó is a coordinator, and in § 3.2 we saw that it may also serve as a complementiser. As in the case of b`f, the double function of bó raises the question of whether two lexical entries are needed for bó or whether it is possible to provide a unified account for this lexical item. As in the case of b`f, I believe that it is possible to provide a unified account of the properties of bó discussed so far, on the following grounds. In both environments, bó relates clauses that have the same subject. In both environments, bó coordinates finite clauses. It thus appears that the properties of bó conjoining two clauses are the same as those of bó introducing the sentential complement of káká and of có, or of bó introducing purposive clauses. As in the case of b`f, the difference between the two contexts in which bó occurs is that, in the one

15.Similar data on purposive structures may be found in Akoha (1980: 210–211, 1990: 290–293) and in Anonymous (1983: IX, 3–7).


142 Claire Lefebvre

case, bó serves as a coordinator, whereas in the other, it serves as a complementiser. As in the case of b`f, there are contexts where bó is semantically ambiguous. For example, in the context of (44), the meaning of bó is intermediate between that of a coordinator (e.g. ‘Koku arrived and-then he left’) and that of a complementiser (e.g. ‘It is as soon as Koku arrived that he left’).16 (44) K`fkú wá tlóló bó yì.17 Koku arrive conj leave ‘As soon as Koku arrived, he left.’ The semantic ambiguity observed in (44) may be related to the fact that bó generally relates clauses denoting events that are sequentially ordered with one another. The sequences of events that are related by bó are of the same type as those related by b`f in (23). Finally, purposive clauses in which bó occurs (see (41), (42)) also involve sequences of events. I thus conclude that it is possible to describe the properties of bó discussed so far in a unified way, and that these can be recorded within a single lexical entry. This lexical entry would minimally contain the information identified in (45): Bó is a coordinator and a complementiser, and it is [+finite]. The feature [+F] represents the property that bó has of binding the subject position that is adjacent to it. (45) bó: coordinator, complementiser [+finite] [+F] under adjacency18 On this analysis, the reason why bó does not occur with NPs follows from its being marked for both [+F] and [+finite]. On the above proposal, bó in (45) differs from b`f in (27) only by its feature [+F], the feature that identifies bó as binding the subject position that is adjacent to it. This raises the question of whether b`f and bó could be analysed as contextually conditioned allomorphs. This issue is the topic of the next section.

16.As is the case with b`f, none of the tests proposed by Haspelmath (1995, to appear) to disambiguate between coordinating and subordinating structures apply in the case of clauses related by bó (see note 7). 17.For my informants, the verb doubling construction corresponding to (23) is not available in this case. This type of verb doubling construction requires disjoint reference of subjects. Bó conjoins clauses that have coreferential subjects. Hence, (i) is not grammatical. (i) *wá K`fkú wá tlóló bó yì arrive Koku arrive conj leave 18.As has been pointed out to me by Martin Haspelmath, the feature [+F] could alternatively be represented as [+bind subject position].

Coordinating constructions in Fongbe with reference to Haitian Creole 143

3.4 Does bó constitute a separate lexical entry from b`f? The lexical properties of b`f (discussed in § 2) and those of bó (discussed in § 3) are summarised in (46). (46) b`f

– conjunction of coordination of subordination ambiguous cases

– conjunction of coordination of subordination ambiguous cases

– [+finite] conjoins finite clauses does not conjoin NPs

– [+finite] conjoins finite clauses does not conjoin NPs

– The subject of the second conjunct is overt.

– The subject of the second conjunct is covert. The empty position is bound by bó.

– The subjects of the two – The subjects of the two conjuncts are referentially free. conjuncts are coreferential. On the basis of the properties of b`f and bó in (46), it could be hypothesised that these two forms are contextually conditioned allomorphs of a single morpheme. On this view, b`f could be said to occur in the context of an adjacent overt subject and bó could be said to occur elsewhere. This would account in a simple way for the distribution of b`f and bó. This analysis could be a plausible one if the list of properties in (46) exhausted the distributional properties of both forms. Such is not the case, however. While b`f does not appear in other environments than those discussed so far, bó does. For example, bó also occurs in contexts such as in (47). Various authors (e.g. Akoha 1990: 161; Anonymous 1983: V, 1) assign it the French gloss ‘donc’ in this context. In this case, bó obligatorily links the content of the clause it appears in to something that has been said earlier in discourse. ‘Then’ thus appears to be an adequate translation for bó occurring in this context. Bó n`f fí. then stay here ‘Then stay here.’ (from Akoha 1990: 161) b. Bó yì. then go ‘Then go.’ (from Anonymous 1983: V, 1)

(47) a.

144 Claire Lefebvre

Bó may also occur between the subject and the verb, as is illustrated in (48). In this case, bó is sometimes referred to as a ‘permissive’ marker (e.g. Akoha 1980: 176; Hounkpatin 1985: 114). In this context as well, bó cannot be used unless it relates the clause it is part of to something that has been said earlier in discourse. I will thus gloss bó occurring in this position as ‘then’ as well. (48) É bó wá. 3sg then come ‘Then, he should come.’ (from Akoha 1980: 176; Hounkpatin 1985: 114) These additional uses of bó distinguish bó from b`f. They constitute a serious drawback for an allomorphy analysis of the two forms. I thus conclude that b`f and bó constitute two distinct lexical entries. The new facts concerning bó, introduced in (47) and (48), raise yet another question: do these new facts constitute counter-examples to a unified analysis of bó? In (45), bó has been identified as a coordinating or subordinating conjunction. However, bó occurring in the context of (47) and (48) has been identified as a connective adverb (see Avolonto 1992: 43). In spite of these differences, I believe that it is possible to maintain a unified analysis of bó. For example, when bó is used in contexts such as (47) and (48), it only occurs in finite clauses. It also relates two events that are sequentially ordered; in this case, however, bó relates the event of the clause it is part of to an event that was mentioned earlier in discourse. So, the properties of bó in contexts such as (47) and (48) do not differ from those of bó summarised in (45), in a way that would force an analysis according to which bó would signal two distinct lexical entries.

4. The theoretical relevance of the properties of b`f and bó The content of this section is dedicated to discussion of the properties of b`f and bó that bear on theoretical issues. The following points will be discussed in turn: the marked character of and-then conjunctions, the disjoint/coreferential subjects distinction and finally, the historical development of the functions of b`f and bó. 4.1 On the marked character of and-then conjunctions As we saw in the previous sections, while b`f and bó may conjoin clauses denoting unrelated events occurring simultaneously in the context of the imperfective aspect (see (4)), and in some cases involving the perfective aspect (see (3)), they otherwise relate events that are sequentially ordered with one another. This sequential interpretation obtains in the context of clauses in the perfective aspect (see § 2.1 and § 3.1). Consequently, both lexical items are interpreted as either ‘and-then’ or ‘and’

Coordinating constructions in Fongbe with reference to Haitian Creole

depending on the context in which they occur. Only the sequential interpretation is available, however, in cases where b`f and bó introduce a subordinate clause (see § 2.2 and § 3.2), and furthermore, with this interpretation bó can also relate an event described by a simple clause to an event referred to in discourse (see (47), (48)). Some languages that exhibit clausal and-then coordination are discussed in e.g. Longacre (1985), Payne (1985) and in the references therein. A point of interest for the present discussion is that, on Payne’s (1985) typology of conjunctions, andthen-type conjunctions are analysed as marked as opposed to and-type ones. This fact will be shown to be relevant for the discussion of the Haitian data in § 6. 4.2 The disjoint/coreferential subjects distinction As we saw in the preceding sections, while b`f can conjoin clauses that have either disjoint subjects (see (5)) or coreferential ones (see (7)), bó is restricted to conjoining clauses that have coreferential subjects (see (6)), provided that it is adjacent to the subject position of the second conjunct (see (28) and note 10). The disjoint versus coreferential subjects distinction associated with conjunctions (or with conjunctive affixes) is also found in languages of various genetically unrelated language families. For example, Ibaloi, a language spoken in the Philippines, has a conjunction meaning ‘and then’ that conjoins clauses having coreferential subjects (Longacre 1985). (There is no mention of another conjunction that would coordinate clauses having disjoint subjects.) In Wojukeso, a language spoken in Papua New Guinea, conjunctive suffixes that indicate temporal relations also indicate same versus different subject(s) in reference to the conjunct clause (Longacre 1985). Another example is Paez, a language spoken in Colombia that has two conjunctive morphemes: one used to coordinate clauses with same subjects and one used to coordinate clauses with different subjects (Longacre 1985). 4.3 From connective adverb to complementiser through conjunction of

coordination We saw that b`f and bó serve as conjunctions of coordination. Both lexical items also serve as complementisers, in contexts where the event described by the subordinate clause is temporally ordered with respect to that described by the matrix clause. It was argued that the multifunctional properties of each lexical item can be accommodated within single lexical entries. It was shown, however, that b`f and bó constitute two separate lexical entries. This section presents a hypothesis concerning a likely historical development of the multifunctional character of b`f and bó. I begin with bó, which has a wider range of functions than b`f, as per the discussion in § 3.4. Recall from § 3.4 that bó can occur in simple clauses as a connective adverb linking the content of the clause it is part of to an event that has been referred to


146 Claire Lefebvre

earlier in discourse. This connective adverb — which I suggested translating as ‘then’ — may very well be the source of the coordinating function of bó occurring between two matrix clauses. The similarity of the properties of bó in these two contexts, as discussed in § 3.4, supports such a hypothesis. The historical relationship between connective adverbs and conjunctions of coordination has already been noted. For example, Mithun (1988: 345) reports that, typically, in languages with no overt coordinators, particles with meanings like ‘also’, ‘then’, ‘and so’, ‘and now’, etc. appear in separate sentences. According to the author, the primary function of these particles “is to provide a semantic or pragmatic link to previous discourse, not to specify a syntactic one”. Mithun (1988: 346) adds that “the fluidity of the boundary between discourse adverbials and syntactic conjunctions is significant. The adverbial particles appear to be the source of most clausal coordinating conjunctions”. Given this situation, it is not unlikely that the connective adverbial bó may have been the source of the coordinating function of bó. Now, bó also serves as a complementiser (see § 3.2). In this function, bó has properties that are similar to those it has as a conjunction of coordination (see § 3.3). For example, as a complementiser, bó is restricted to contexts where the event described by the clause it is part of is sequentially ordered with respect to that described by the matrix clause. As a conjunction of coordination, bó conjoins clauses describing events that are generally interpreted as being sequentially ordered with one another in the context of the perfective aspect. Given this situation, it is logical to hypothesise that the subordinating function of bó is the result of the expansion of its function as a coordinator of clauses. Reported cases of such reanalysis are extremely rare. Complementisers have been shown to have evolved from various sources. For example, the pronoun that gave rise to the complementiser that in English (see e.g. Hopper and Traugott 1993; Langacker 1977; Lockwood 1968; Noonan 1985; etc.); the preposition to was the source of the complementiser to (see e.g. Noonan 1985; etc.); verbs meaning ‘to say’ gave rise to that-type complementisers in West African languages (see e.g. Lord 1976). (For extensive discussions of the source of complementisers, see e.g. König 1985; Lord 1973; Ransom 1988; Traugott 1985; etc.) To my knowledge, the closest case to the Fongbe one discussed here has been reported by Pepicello (1982). On the basis of Sanskrit, Greek and Latin, the author shows that markers of connectivity in discourse may develop in several ways; inter alia they may develop as coordinating and subordinating markers that may come to introduce subordinate clauses of purpose, cause or result. In the Fongbe case reported on here, temporal and purposive clauses are involved (see § 2.2 and § 3.2), but not cause or result clauses. A few other similar cases of linguistic change have been pointed out to me. Bernard Comrie (p.c.) notes that the coordinating conjunction than in English is being reanalysed as a subordinating conjunction at the same time as it is becoming a preposition. Martin Haspelmath (p.c.) points out that, in spoken Norwegian, the word og ‘and’ has

Coordinating constructions in Fongbe with reference to Haitian Creole 147

come to be used as an infinitival complementiser. In Haspelmath (to appear), he further reports (based on Culicover and Jackendoff 1997) on a class of English clause-combining constructions that show mixed subordinate-coordinate behavior. So, even if the change hypothesised for the Fongbe data above is not entirely unheard of, it is sufficiently unusual to be worth emphasising. The Fongbe data discussed in this paper thus appear to constitute an original contribution to our current knowledge of the possible sources for complementisers. Have the functions of b`f followed the same developmental path as bó? According to available data, b`f shows no evidence of being or having been an adverbial connector in simple clauses (see § 3.4). So, as far as we know, the history of b`f starts with its function as a conjunction of coordination (see § 2.1). Recall from § 2.2 and § 2.3 that, as a complementiser, b`f has properties that are similar to those it has as a conjunction of coordination. Therefore, I see no reason why the complementiser function of b`f would not have developed in a way similar to that of bó. Consequently, I assume that, as is the case of bó, the subordinating function of b`f is the result of an expansion of its coordinating function. The hypothesised developmental path of b`f and bó can be summarised as follows. connective adverb


> conjunction of coordination ‘and then’ in the context of perfective aspect

> conjunction of subordination (complementiser) after temporal adverbs

‘and’ in the context of imperfective aspect and in some other particular cases

The hypothesised historical development of the functions of b`f/bó is compatible with the fact that these lexical items may not conjoin NPs.


Can NPs be conjoined?

This section addresses the question of whether NPs can be conjoined in Fongbe. In § 5.1, it is shown that the equivalent of coordination of NPs is achieved by means of a circumposition made up of two lexical items meaning ‘with’. It is argued that these lexical items do not have the properties of conjunctions and that therefore

148 Claire Lefebvre

there is no true and-conjunction of NPs in this language. In § 5.2, the Fongbe data are discussed in light of the properties of other with-type languages. 5.1 The kpó2 2ó…kpó ‘with…with’ circumposition The equivalent of coordination of NPs is achieved by means of a circumposition made up of a preposition kpó2ó ‘with’ (lit.: ‘’), and of the postpositions kpó or kpán which both mean ‘with’. This is examplified in (49).19 (49) Àsíbá [kpó2ó K`fkú kpó/kpán] yì àxì m`7. Asiba [with Koku with/with go market in ‘Asiba with Koku went to the market.’ As we saw in the introduction, some authors consider this circumposition a conjunction (see e.g. Akoha 1980: 210; Anonymous 1983: VII, 1) and gloss it as ‘and’. It is argued below that kpó2ó is best analysed as a preposition and kpó as a postposition, that the phrase headed by kpó2ó is a prepositional phrase and that, from a syntactic point of view, this phrase is a syntactic adjunct. Note that kpó2ó may always reduce to kpó; no difference in meaning or in syntactic properties is involved in the selection of either one of the two forms. In (49), kpó2ó introduces a comitative phrase. In (50) it introduces an instrumental phrase. (50) K`fkú xò Àsíbá kpó2ó àtín kpó/kpán. Koku hit Asiba with stick with ‘Koku hit Asiba with a stick.’ While it is possible to assign a conjunctive interpretation to the phrase headed by kpó2ó in (49), it is not possible to do so in the case of (50), nor is it possible to do so in the case of (51), where kpó2ó introduces a manner phrase. (51) K`fkú gbá xwé f´ kpó2ó àyì kpó. Koku build house def with heart with ‘Koku built the house with care.’ Finally, kpó2ó may also introduce phrases of the type in (52). In this case also, a conjunctive interpretation is impossible. (52) K`fkú g´f hùn f´ kpó2ó gbàdé kpó. Koku fill truck def with corn with ‘Koku filled the truck with corn.’ (=(38a) in Brousseau 1998: 102)

19.For an extensive discussion of the Fongbe circumpositions, see Lefebvre and Brousseau 2002: 299–346.

Coordinating constructions in Fongbe with reference to Haitian Creole 149

The distribution of kpó2ó is thus not compatible with that of conjunctions of coordination. This strongly suggests that kpó2ó is not a conjunction. This conclusion is further supported by other properties of this lexical item. From a categorial point of view, kpó2ó is a preposition. In Lefebvre and Brousseau (2002: 303–312), it is extensively argued that kpó2ó shares no properties with verbs. Rather, it is shown that it shares its syntactic properties with the prepositions of the language. The conclusion is thus that kpó2ó is of the syntactic category P, defined by the features [−N, −V]. From a syntactic point of view, there are several arguments attesting to the adjunct status of the phrase introduced by kpó2ó. These constitute further arguments against a conjunction analysis of kpó2ó. First, the phrase headed by kpó2ó can always be extraposed, as is illustrated in (53). Àsíbá yì àxì m`7 [kpó2ó K`fkú kpó]. Asiba go market in [with Koku with ‘Asiba went to the market with Koku.’ b. K`fkú zé m`flìkún f´ 2ó m´ftò f´ m`7 [kpó2ó súklè f´ kpó]. Koku take rice def put car def in [with sugar def with ‘Koku put the rice in the car with the sugar.’

(53) a.

Phrases conjoined by ‘and’ cannot be extraposed. Second, the phrase headed by kpó2ó is optional, as is shown in (54), where optionality is signalled by parentheses. (54) K`fkú yì àxì m`7 (kpó2ó Àsíbá kpó/kpán). Koku go market in (with Asiba with ‘Koku went to the market (with Asiba).’ (=(51b) in Lefebvre and Brousseau 2002: 314) While arguments are obligatory, adjuncts are optional (see e.g. Baker 1996; Pinker 1989; Randall 1987). Third, the phrase headed by kpó2ó may occur outside of nominalised VPs. The imperfective construction provides an appropriate context to illustrate this fact. The imperfective construction makes use of 2ò ‘to be at’, which selects a phrase headed by w`7, which in turn selects a nominalised VP. In this construction, the arguments of the verb all occur within the phrase headed by w`7. This is exemplified in (55) for a serial verb construction involving the verbs s´f ‘to take’ and yì ‘to go’. (55) K`fkú 2ò [[às´fn s´f yì àxì] w`7]. Koku [[crab taking going market post ‘Koku is bringing crab to the market.’ (=(52) in Lefebvre and Brousseau 2002: 315) The arguments of the verbs in (55) cannot occur to the right of w`7. In contrast, all PPs including a phrase headed by kpó2ó may occur to the right of w`7 (that is, outside of the phrase headed by w`7). This is shown in (56).

150 Claire Lefebvre

(56) K`fkú 2ò [àxì yì w`7] kpó2ò Àsíbá kpó. Koku [market going post with Asiba with ‘Koku is going to the market with Asiba.’ (=(53b) in Lefebvre and Brousseau 2002: 315) The fact that the kpó2ó phrase can occur outside of the nominalised VP in (56) follows from its adjunct status. Fourth, like other PPs, the kpó2ó phrase may be left behind in VP fronting. For example, the nominalised VP of (55) can be clefted, as in (57). In this case, the whole nominalised VP is fronted, including all the internal arguments. (57) [Às´fn s´f yì àxì] w`7, K`fkú 2è. [crab taking going market Koku ‘It is bringing crab to the market that Koku is doing.’ (=(54) in Lefebvre and Brousseau 2002: 316) What happens when the nominalised VP containing a PP is clefted? In this case, the PP may be left behind, as is illustrated in (58). (58) [Àxì yì] w`7, K`fkú 2è kpó2ó Àsíbá kpó. [market going Koku with Asiba with ‘It is going to the market that Koku is doing with Asiba.’ (=(55b) in Lefebvre and Brousseau 2002: 316) These extraction facts follow directly if the prepositional phrase is adjoined to VP. The syntactic tests illustrated in (53) to (58) thus all point to the conclusion that kpó2ó is not a conjunction. Typically, conjunctions cannot be separated from one of their conjuncts. The phrase headed by kpó2ó can be separated from one of its potential conjuncts in various ways (see (53), (56), (58)). These facts rather strongly argue for an adjunct analysis of the phrase headed by kpó2ó. Summarising: the lexical item kpó2ó introducing the so-called conjunction of NPs does not have the properties of conjunctions. Rather, it has distributional properties that manifest its status as a major category lexical item, namely as a preposition. Finally, several arguments demonstrate that the phrase headed by kpó2ó is a syntactic adjunct. This conclusion holds even in the context of the sentence in (59). (59) Ùn wà àz´f 2ò kùtónú, kpó2ó àgbómé kpó. 1sg do work at Cotonou with Abomey with ‘I worked in Cotonou, and in Abomey.’ I now turn to a brief discussion of the properties of the synonymous postpositions kpó/kpán ‘with’. Out of some twenty postpositions in the language, kpó and kpán are the only postpositions that do not have a nominal counterpart. In Lefebvre and Brousseau (2002:327–329) it is argued that the properties of the Fongbe postpositions,

Coordinating constructions in Fongbe with reference to Haitian Creole

including those of kpó and kpán, differ from those of case markers, and that therefore, postpositions are not case markers. Rather, they have the status of major category lexical items. It is further argued that the properties of postpositions contrast in a systematic way with those of nouns (Lefebvre and Brousseau 2002: 330–334) and with those of verbs (Lefebvre and Brousseau 2002: 334–337), thus defining postpositions as being of the category [−N, −V] in this language.20 In all the examples above, the postpositions are obligatory even though, from a semantic point of view, they are redundant with respect to kpó2ó. So, the equivalent of coordination of NPs is achieved in Fongbe by means of the circumposition kpó(2ó)… kpó/kpán. To my knowledge, the sequence *NP kpó NP, where kpó could be perceived as a conjunction of NPs, is not possible. None of the informants consulted accept it and I found no example of this structure in the available literature. I thus conclude that there is no and-conjunction of NPs in Fongbe. 5.2 Fongbe and other with-type languages According to Stassen (2000: 41), with-type languages are found in Asia, in the Americas and in Africa. “With the possible exception of Khoisan, all the languages of Africa in and below the Sahara exhibit some degree of with-encoding” (Stassen 2000: 41). With respect to coordination of NPs, Fongbe is thus of the same type as the languages of its areal group. Stassen (2000:44) remarks that the distinction between with-type and and-type languages correlates with two parameters: case and tense. On the basis of a large sample of languages, he observes that tensed and cased languages tend to be and-type languages, whereas [−tensed] and [−cased] languages tend to be with-type languages. The Fongbe data support this correlation. On the one hand, Fongbe expresses aspect rather than tense (see Lefebvre and Brousseau 2002: Chapter 5). On the other hand, although it exhibits case markers in nominal structures (see Lefebvre and Brousseau 2002: 47–48), the language is generally not a cased language. Finally, it has been noted in several instances that with-type languages have the tendancy to drift towards and-status by the reanalysis of the comitative marker as a conjunction (see e.g. Haspelmath to appear: 26–30; Mithun 1988; Stassen 2000: 1, and the references cited therein). Such a change has been proposed to have occurred in some West African languages. For example, Lord (1973) proposes that in Yoruba, Gã and Ewe, a comitative verb has been reanalysed as a comitative

20.The issue of whether prepositions and postpositions constitute a uniform syntactic class, aside from their directionality properties, is discussed in Lefebvre and Brousseau (2002: 340–342).



Claire Lefebvre

preposition which, in turn, has been reanalysed as a conjunction of NPs. Note that in all these cases the sequence NP ‘with’ NP is possible, thus supporting a reanalysis account. As was mentioned in § 5.1, however, the sequence *NP kpó NP is not licit in Fongbe. It thus appears that Fongbe is more conservative than some neighbouring languages as far as the properties of this particular lexical item are concerned. In the course of my field work, however, I have come across data involving the postposition kpó which depart from the pattern described so far. These are reproduced in (60). M´7-kpó Àsíbá-kpó w´7 yì àxì-m`7? who with Asiba with this go market in ‘Who did Asiba go to the market with?’ b. Nú.kíkó kpó àwà.jíj´7 kpó m`7 w`7, é n`f n`f t´7gbé. thing-smiling with joy-falling with in 3sg hab stay always [Lit.: ‘It is with smiling with enjoying that he always is.’] ‘He lives in happiness.’ c. [À kó.nú] kpó], [à yà.ví] kpó] f´, … [2sg laugh with [2sg cry with def ‘Whether you laugh or whether you cry, …’

(60) a.

The uses of the postposition kpó in the three sentences above are quite unusual as compared with those discussed in § 5.1. Do they signal an incipient change whereby the postposition kpó would be becoming a case marker? This would explain the fact that its meaning in (60) appears to be removed from the original one ‘with’. Could it also explain its clausal complement in (60c)? I leave further investigation of this possible incipient change for future research. 6. Coordinating constructions in Haitian Creole Haitian Creole has a lexical item (e)pi used to conjoin clauses. The equivalent of NP coordination is achieved by means of the lexical item (kòl)ak. In this section, it is shown that, to a large extent, the properties of (e)pi correspond to those of Fongbe b`f, and that those of (kòl)ak correspond to those of Fongbe kpó(2ó). To my knowledge, there is no Haitian lexical item corresponding to Fongbe bó. The section ends with a short discussion of how the properties of b`f and those of kpó2ó are hypothesised to have been transferred into the creole, and why there is no lexical item corresponding to Fongbe bó in the Haitian lexicon. The data discussed in this section are based on the literature and on my own field notes gathered from speakers who speak a rather conservative variety of Haitian Creole.21

21.I am particularly indebted to Joseph Sauveur Joseph for his contribution to this topic.

Coordinating constructions in Fongbe with reference to Haitian Creole

6.1 The clausal conjunction epi Haitian has a conjunction (e)pi used to conjoin clauses, as is shown in (61). (61) Jan pati (e)pi Mari rive. John leave conj Mary arrive ‘John left and-then Mary arrived.’ (=(70) in Lefebvre 1993)


In Valdman et al. (1981), Haitian epi is glossed as ‘and, then, and then’. This conjunction derives its phonological representation from the French sequence of words et puis [lit.: ‘and then’], pronounced [(e)pi] in popular French. In this variety of French, (e)pi is used in complementary distribution with et ‘and’ to conjoin clauses and noun phrases, as is shown in (62). Jean est parti et/(e)pi22 Marie est arrivée. John aux leave conj Mary aux arrive ‘John left and Mary arrived.’ (=(73) in Lefebvre 1993) b. Jean et/(e)pi Marie John conj Mary ‘John and Mary’ (=(74) in Lefebvre 1993)

(62) a.



While Haitian (e)pi derives its phonological representation from the French sequence of words identified above, it does not have the same distributional properties as this French sequence of words. For example, in contrast to French (e)pi, Haitian (e)pi cannot be used to conjoin NPs, as is shown by the ungrammaticality of (63). Compare (63) with (62b).23 (63) *Jan (e)pi Mari John conj Mary


In fact, Haitian epi has properties that are quite similar to those of Fongbe b`f. According to my informants, when epi relates clauses that are in the perfective aspect, the conjoined clauses are interpreted as denoting related events occurring sequentially. This is illustrated in (64). (64) Jan rive epi Mari pati. John arrive conj Mary leave ‘John arrived and-then Mary left.’


22.The spelling of (e)pi reflects its pronunciation in popular French. There are no orthographic conventions for the spelling of this lexical item. 23.According to Valdman’s et al. (1981) dictionary epi may conjoin NPs as well as clauses. My understanding of the situation is that epi will be found as a conjunction of NPs in the grammar of those speakers who have had more exposure to French than my informants, who reject this use of epi.



Claire Lefebvre

In contexts such as those in (65), even though the two clauses coordinated by epi occur in the perfective aspect, they are interpreted as denoting two independent events (in terms of both sequentiality and causality) and the conjunction is glossed as ‘and’. (65) a.

De moun genyen epi de moun pedi. two person win conj two person fail ‘Two persons won and two persons failed.’ b. Jan genyen kous la epi li pedi nan sote a. John win race def conj 3sg lose in jump def ‘John won at the race and he lost at the jump.’



When epi relates clauses that are in the imperfective aspect, the conjoined clauses may be interpreted as denoting events that are unrelated and that may occur simultaneously. This is shown in (66). (66) Jan ap rive epi Mari ap pati. John impf arrive conj Mary impf leave ‘John is arriving and Mary is leaving.’


Compare the Haitian data in (64), (65) and (66) with the corresponding Fongbe data in (2), (3) and (4). Like Fongbe b`f, Haitian (e)pi may conjoin clauses that have different or coreferential subjects, as is shown in (67a) and (67b). Compare the Haitian data in (67) with the Fongbe data in (5) and (7). (67) a.

Jan rive epi Mari pati. John arrive conj Mary leave ‘John arrived and-then Mary left.’ b. Jan rive epi li pati. John arrive conj 3sg leave ‘John arrived and-then he left.’



As is the case of Fongbe b`f, the subject of the second conjunct introduced by epi has to be overt. The sentence in (68) is not grammatical because the second conjunct has no overt subject. Compare Haitian (68) with Fongbe (10). (68) *Jan rive epi — pati John arrive conj leave


As is the case of b`f, epi is excluded from subjectless clauses. As is shown in (69), epi cannot conjoin two infinitival complements of the verb meaning ‘to want’. Compare Haitian (69) with Fongbe (12). (69) *Jan vle bwè dlo epi manje pen John want drink water conj eat bread


Coordinating constructions in Fongbe with reference to Haitian Creole

The sentence in (69) can be rescued as (70) where epi conjoins two complete clauses with two overt subjects. (70) Jan vle bwè dlo epi li vle manje pen. John want drink water conj 3sg want eat bread ‘John wants to drink water and-then he wants to eat bread.’


Epi cannot be used to conjoin two infinitival complements of the modal verb kap ‘may’ as is shown by the ungrammaticality of (71). Compare Haitian (71) with Fongbe (14). (71) *Jan kap vini epi pati John may come conj go


The sentence in (71) can be rescued as (72) where epi relates two full clauses with two overt subjects. (72) Jan kap vini epi li kap pati. John may come conj 3sg may go ‘John may come and-then he may go.’


Haitian epi cannot relate two complements of the verb meaning ‘to begin’. This is shown by the ungrammaticality of (73). (73) *Jan kòmanse manje pèn epi bwè dlo John begin eat bread conj drink water


The sentence in (73) can be rescued as (74) where epi relates two full clauses with two overt subjects. (74) Jan kòmanse manje pen epi li kòmanse bwè dlo. John begin eat bread conj 3sg begin drink water ‘John began to eat bread and-then he began to drink water.’


Compare the Haitian sentences in (73) and (74) with the Fongbe ones in (16) and (17). The Haitian data in (69)–(74) show that epi is excluded from subjectless clauses. This suggests that epi cannot conjoin non-finite clauses. This conclusion would gain support if it could be shown that epi is also excluded from infinitival clauses containing an overt subject. In addition to the infinitival structure in (70), in which the subject of the infinitival complement of vle ‘to want’ is covert, Haitian exhibits another infinitival structure in which the subject of the infinitival complement of vle is overt. This structure is exemplified in (75). In (75), the subject of the main clause and that of the embedded clause are obligatorily disjoint, and, as per the analysis in Sterlin (1988, 1989), the subject of the infinitival clause bears


156 Claire Lefebvre

accusative case, assigned to it by the verb vle under Exceptional Case Marking.24 (75) Jan vle Mari/li bwè dlo. John want Mary/3sg drink water ‘John wants Mary/him/her to drink water.’


Epi cannot relate two infinitival complements of the type in (75). This is attested by the ungrammaticality of (76). (76) *Jan vle Mari bwè dlo epi li manje pen John want Mary drink water conj she eat bread


The sentence in (76) can be rescued as (77) where epi conjoins two complete finite clauses. (77) Jan vle Mari bwè dlo epi li vle li manje pen. Haitian John want Mary drink water conj 3sg want 3sg eat bread ‘John wants Mary to drink water and-then he wants her to eat bread.’ The contrast in grammaticality between (76) and (77) shows that epi is indeed excluded from infinitival clauses. It thus appears that, like Fongbe b`f, Haitian epi only conjoins finite clauses (for Fongbe see also note 10). This is an interesting conclusion in view of the fact that French (e)pi may conjoin infinitival clauses. For example, the grammaticality of the French sentence in (78a) contrasts with the ungrammaticality of the corresponding Haitian sentence in (69), that of (78b) with that of (71), and that of (78c) with that of (73). (78) a.

Jean veut boire de l’eau (e)pi manger du pain. ‘John wants to drink water and eat bread.’ b. Jean peut venir (e)pi partir. ‘John may come and go.’ c. Jean a commencé à boire de l’eau (e)pi à manger du pain. ‘John started drinking water and eating bread.’

French French French

Unlike Fongbe b`f (see (18), (19), (26)), Haitian epi does not introduce the sentential complements of adverbs meaning ‘until’ or ‘before’, nor does it participate in purposive clauses. Haitian epi does, however, occur with the verb doubling construction involved in the expression of temporal clauses. Consider the structure in (79). (79) Rive Jan rive epi Mari pati. arrive John arrive conj Mary leave ‘As soon as John arrived, Mary left.’ (=(19) in Lefebvre 1994)


24.The theory of Case adopted by Sterlin for her analysis is that in Chomsky (1981).

Coordinating constructions in Fongbe with reference to Haitian Creole

The Haitian data in (79) parallel in a striking way the Fongbe data in (23). Both languages contrast in a similar way with French in presenting the structure in (79) involving verb doubling phenomena. French presents none of the verb doubling phenomena observed in both Haitian and Fongbe. (For a thorough discussion of these facts, see Lefebvre 1998: 363–374.) The properties of Haitian (e)pi presented in this section replicate in a remarkable way those of Fongbe b`f presented in § 2 and § 3, instead of those of the French form from which it is phonologically derived. How did this situation obtain? This question will be taken up in § 6.3. 6.2 Can NPs be conjoined in Haitian Creole? The equivalent of coordination of NPs is achieved by means of the preposition (kòl)ak ‘with’, as is illustrated in (80). Kòl-ak is a complex word made up of kòle ‘close’ and ak ‘with’ (see Gilles 1988). Haitian ak is a reduced form of Haitian avèk ‘with’, phonologically derived from French avec ‘with’. (80) Jan (kòl)ak Mari John with Mary ‘John with Mary’


Haitian (kòl)ak has the properties of Fongbe kpó(2ó). Like kpó2ó (see (49)), it occurs as a comitative preposition, as in (81). (81) Mari ak Jan ale nan mache. Mary with John go in market ‘Mary with John went to the market.’


Like kpó2ó (see (50)), it occurs as an instrumental preposition, as in (82). (82) Jan frape Mari ak yon baton. John hit Mary with a stick ‘John hit Mary with a stick.’


Like kpó2ó (see (51)), it occurs in manner phrases, as in (83). (83) Jan bati kay la ak swen. John build house def with care ‘John built the house with care.’


Like kpó2ó (see (52)), it occurs in the context of (84). (84) Jan plèn kamyon an ak mayi. John load truck def with corn ‘John loaded the truck with corn.’




Claire Lefebvre

Note that while the distribution of Haitian ak is systematically parallel to that of its Fongbe counterpart, it is not systematically parallel to that of the French lexical item avec ‘with’ from which it is phonologically derived. In French, avec ‘with’ cannot relate two NPs, hence, *Marie avec Jean is not grammatical as compared to the corresponding grammatical Haitian structure in (81). Likewise, the use of Haitian ak in (84) does not correspond to that of French avec. The French sentence *Jean a rempli le camion avec du maïs [lit.: ‘John filled the truck with corn.’] is not grammatical. The preposition de has to be used in this case instead of avec, yielding Jean a rempli le camion de maïs. As is the case of the Fongbe phrase headed by kpó2ó (see (53)), the Haitian phrase headed by ak can be extraposed, as is shown in (85). Furthermore, like the Fongbe phrases headed by kpó2ó in (53), the Haitian phrase headed by ak in (85) is optional. Optionality is signalled by parentheses. (85) a.

Mari ale nan mache (ak Jan). Mary go in market (with John ‘Mary went to the market with John.’ b. Jan mete diri a nan kamyon an (ak sik la). John put rice def in truck def (with sugar def ‘John put the rice in the truck with the sugar.’



The fact that the phrase headed by ak can be extraposed, and the fact that it is optional, argue for the adjunct status of this phrase. This conclusion holds even in the context of the sentence in (86). (86) M travay potoprens, ak jakmèl. 1sg work Port-au-Prince with Jacmel ‘I worked in Port-au-Prince and Jacmel.’


The Haitian data in (86) parallel the Fongbe data in (59). Haitian (kòl)ak thus has the semantic and distributional properties of Fongbe kpó(2ó). As will be seen below, Haitian (kòl)ak also has the morphological structure of corresponding lexical items in West African languages. 6.3 How did the properties of the Haitian lexical items get to be the way

they are? How did the properties of the Haitian lexical items (e)pi and (kòl)ak get to be the way they are? In Lefebvre (1998), it is argued that the process of relexification has played a major role in the formation of the lexicons of creole languages. In this process, a given lexical entry is relabelled on the basis of a phonetic string found in a contact language. The resulting lexical entry thus has the properties of the original lexical entry with a phonological representation taken from another language (for

Coordinating constructions in Fongbe with reference to Haitian Creole 159

various representations of the process, see Lefebvre 1998; Lefebvre and Lumsden 1994a, 1994b; Mous 1995, 2001; Muysken 1981). On this view, the Fongbe lexical entry b`f would have been relabelled on the basis of the French phonetic string [epi], yielding Haitian /epi/ with the semantic and distributional properties of Fongbe b`f. As was mentioned in § 6.1, however, in contrast to b`f, epi does not introduce the complements of the prepositions meaning ‘until’ and ‘before’. The relexification account of creole genesis predicts that epi would have occurred in these contexts as well in the incipient creole and that it ceased to be used in these contexts as the creole developed. Another possibility is that relexification took place prior to the time when Fongbe b`f had acquired the function of complementiser. In this case, the Haitian lexical entry would reflect the properties of Fongbe b`f prior to the hypothesised change. In conclusion, the remarkable similarity between epi and b`f supports the relexification account of the history of the Haitian lexical entry epi ‘and then, and’. As was mentioned earlier, there is no Haitian lexical entry corresponding to Fongbe bó, discussed in § 3. As unexpected as it may be, this fact also follows from the relexification account of creole genesis. As is shown in Lefebvre (1998), the relabelling of a given lexical entry is only possible if the superstratum language of an incipient creole offers a phonetic string available to relabel an original lexical entry. A suitable phonetic string must share some semantics with the original lexical entry for relabelling to take place (see Muysken 1981). Did French offer an appropriate phonetic string to relabel Fongbe bó? There does not appear to be any French lexical material that could have been used to relabel bó. The original lexical entry could thus not be relabelled. The Haitian lexical entry (kòl)ak was also derived by the process of relexification. The form (kòl)ak is made up of two morphemes kòle.ak ‘close.with’. The forms of these morphemes are derived from French collé ‘close’ and avec ‘with’. The resulting Haitian compound word is built on the model of the West African compound prepositions meaning ‘with’. For example, Fongbe kpó2ó is made up of kpó ‘with’ and of 2ò ‘’. Lord (1973) documents the fact that corresponding words in other West African languages also involve a verb meaning ‘to come in contact’, ‘to collide’, ‘to bring together’, ‘to assemble’, etc. combined with a form meaning ‘with’. The claim that the complex preposition meaning ‘with’ in West African languages has been relabelled on the basis of French words compounded to yield the Haitian lexical entry (kòl)ak on the model of corresponding West African languages is thus borne out. Furthermore, as we saw in § 6.2, the distributional properties of the Haitian complex form are modelled on those of the corresponding substratum lexical entry rather than on those of French avec ‘with’. Concluding: the Haitian lexical entries involved in clausal and nominal coordination reproduce the details of the corresponding substratum lexical entries rather than those of the superstratum form from which the Haitian forms are

160 Claire Lefebvre

phonologically derived. This follows from the relexification account of creole genesis. This provides a straightforward explanation for the fact that Haitian is typologically like its West African substratum languages in having an andthen-type of clausal conjunction, a marked form as per the discussion in § 4.1, and a with-type of ‘so-called’ NP conjunction.


Concluding remarks: The typological features of Fongbe

This section concludes the paper with remarks on the properties of the lexical items discussed in this paper considered from the point of view of language typology. It has long been noted that, in African languages, coordination of NPs and coordination of clauses are achieved by means of different lexical items (e.g. Welmers 1973: 305). Fongbe is no exception: while b`f and bó, ‘and then, and’, are used to conjoin clauses, the circumposition kpó2ó…kpó ‘with…with’ is used to achieve the equivalent of NP coordination. It has been proposed that the reason why b`f and bó are excluded from NPs is for the same reason that they are excluded from non-finite clauses: they bear the feature [+finite]. This also explains why Fongbe b`f/bó do not conjoin verbs or VPs.25 As is noted by Welmers (1973: 365), the coordinating constructions indicate a following or simultaneous action. “Simultaneous constructions do not appear to be widespread in Niger-Congo languages, but consecutive constructions are frequently found; […].” Fongbe falls into this general pattern. Both b`f and bó introduce consecutive constructions. A simultaneous interpretation is, however, triggered in the context of clauses occurring in the imperfective aspect and in specific cases involving the perfective. Fongbe exhibits the difference found in some languages between disjoint and coreferential subjects. While this distinction is found in other West African languages, it is also found in unrelated language families (see § 4.2). In modern Fongbe, both b`f and bó have the double function of coordinating and subordinating conjunction. It has been argued that, in the latter function, b`f and bó have the properties of complementisers. This is an interesting point, for there are only a few cases of coordinating conjunctions that have been reported to have been reanalysed as complementisers (see § 3.3). As has been pointed out by Mithun (1988: 351), “a surprising number of coordinating constructions do share one characteristic […]: their youth.” According to her, there are two possible paths for the development of conjunctions; these are schematically represented in (87).

25.Conjunction of verbs appears to be a rare phenomenon in African languages, as is pointed out by Welmers (1973: 365).

Coordinating constructions in Fongbe with reference to Haitian Creole

(87) a. connective adverb > clause conjunctor > phrase conjunctor b. comitative marker > phrase conjunctor > clause conjunctor The first path is illustrated by Nguna, which has an adverbial connector go that links new sentences to previous discourse. As is observed by Mithun (1988: 348), this connector can also conjoin full clauses, which may represent sequential events or generic ones, and it can conjoin noun phrases. Fongbe bó is partially similar to Nguna go. Recall from § 3 that, in one of its uses, bó is an adverbial connector that links new sentences to previous discourse. Bó is also used to conjoin full clauses that generally represent sequential events. Clauses conjoined by bó may be interpreted as representing non-sequential events only in the context of the imperfective aspect. However, unlike Nguna go, Fongbe bó is not a phrase conjunctor; recall from § 2 that bó only conjoins clauses. b`f is like bó except that no connective adverbial function is associated with it. So, on the one hand, Fongbe may be considered a conservative language in that b`f/bó have not become phrasal conjunctions. On the other hand, Fongbe may be considered innovative in that both b`f and bó appear to also fulfill the function of complementiser in contexts involving temporal subordination. The second path in Mithun’s diagram in (87b) is illustrated by some West African languages discussed in Lord (1973), where comitative markers appear to have been reanalysed as phrasal conjunctors but not (yet) as clausal conjunctors. On this point also, Fongbe appears to be more conservative than some neighboring languages for, according to the data presented in § 5, the phrase headed by the preposition kpó2ó ‘with’ in Fongbe is still a syntactic adjunct. So on this path, Fongbe is still at the first of the three stages hypothesised by Mithun. Haitian Creole was shown to be like Fongbe in manifesting an and-then-type clausal coordinator and a with-type nominal coordinator. As regards these constructions, then, Haitian Creole is typologically similar to Fongbe. This should not come as a surprise, for similar results obtain when a wide range of lexical items and constructions are considered (see Lefebvre 1998, and the references therein, 1999, 2001). This follows from the relexification account of creole genesis. Interestingly enough, even marked aspects of lexical entries get transferred into a creole through relexification. Indeed, as per the discussion in § 4.1, and-then conjunctions are marked as opposed to and ones. Both Fongbe and Haitian exhibit the first type. This counterexemplifies Bickerton’s (1981) claim, according to which creoles manifest only unmarked features. (For further discussion of this point, see Lefebvre 1998, 2001.)

Abbreviations ant conj

marker of anteriority conjunction

def.fut definite future marker dp determiner phrase


162 Claire Lefebvre

hab irr post

habitual marker irrealis mood marker postposition

res sub

resumptive pronoun subjunctive marker

References Akoha, A. B. 1980. Quelques éléments d’une grammaire du fongbe : Nominal et syntagme nominal. Unpublished Ph.D. dissertation, Université Paris III (Sorbonne Nouvelle), Paris. Akoha, A. B. 1990. Syntaxe et lexicologie du fon-gbe (République du Bénin). 2 vols. Unpublished thesis for ‘doctorat d’état-ès-lettres’, Université Paris III (Sorbonne Nouvelle), Paris. Anonymous. 1983. Éléments de recherche sur la langue fon. Cotonou. [No indication of publisher.] Avolonto, A. 1992. De l’étude sémantico-syntaxique des marqueurs préverbaux à la structure de la phrase en f`fngbè. Unpublished MA thesis, Université du Québec à Montréal, Montreal. Baker, M. C. 1996. The Polysynthesis Parameter. New York, NY: Oxford University Press. Bickerton, D. 1981. Roots of Language. Ann Arbor, MI: Karoma. Bouchard, D. 1995. The Semantics of Syntax. Chicago, IL: University of Chicago Press. Brousseau, A.-M. 1998. Réalisations argumentales et classes de verbes en ffngbè. [Langues et cultures africaines 22]. Paris: Peeters. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris Publications. Cowper, E. A. 1989. Thematic underspecification: The case of Have. Paper presented at the Canadian Linguistic Association annual meeting. Université Laval, Quebec. Cowper, E. A. 1995. “English participle construction”. Canadian Journal of Linguistics 40: 1–38. Culicover, P. W. and Jackendoff, R. 1997. “Semantic subordination despite syntactic coordination”. Linguistic Inquiry 28: 195–217. Fabb, N. 1992a. “The licensing of Fon verbs”. Topics on the Syntax and Semantics of Fongbe 22(1): 27–34. Fabb, N. 1992b. “Reduplication and object movement in Ewe and Fon”. Journal of African Languages and Linguistics 13(1): 1–41. Ghomeshi, J. and Massam, D. 1994. “Lexical/syntactic relations without projection”. Linguistic Analysis 24: 175–217. Gilles, R. 1988. Réalisation du Cas en créole haïtien dans trois environnements: les prépositions, les marqueurs de Cas et les noms. Unpublished Master’s Thesis, Université du Québec à Montréal, Montreal. Haspelmath, M. 1995. “The converb as a cross-linguistically valid category”. In M. Haspelmath and E. König (eds), Converbs in Cross-Linguistic Perspective. Berlin: Mouton de Gruyter, 1–55. Haspelmath, M. To appear. “Coordination”. In T. Shopen (ed.), Language Typology and Syntactic Description, 2nd edition. Cambridge: Cambridge Univeristy Press. Hopper, P. J. and Traugott, E. C. 1993. Grammaticalization. Cambridge: Cambridge University Press. Hounkpatin, B. B. 1985. Le verbal et le syntagme verbal du fon-gbe parlé à Massè. Unpublished Ph.D. dissertation, Université Paris III (Sorbonne Nouvelle), Paris. Johns, A. 1992. “Deriving ergativity”. Linguistic Inquiry 17: 587–622. Kayne, R. S. 1981. “ECP extensions”. Linguistic Inquiry 12: 93–133. König, E. 1985. “Where do concessives come from? On the development of concessive connectives”. In J. Fisiak (ed.), Historical Semantics, Historical Word Formation [Trends in Linguistics, Studies and Monographs 29]. Berlin: Mouton de Gruyter, 263–282.

Coordinating constructions in Fongbe with reference to Haitian Creole 163

Langacker, R. W. 1977. “Syntactic reanalysis”. In C. N. Li (ed.), Mechanisms of Syntactic Change. Austin and London: University of Texas Press, 57–140. Lefebvre, C. 1986. “Relexification in creole genesis revisited: The case of Haitian Creole”. In P. C. Muysken and N. Smith (eds.), Substrata Versus Universals in Creole Genesis [Creole Language Library 1]. Amsterdam and Philadelphia: John Benjamins, 279–301. Lefebvre, C. 1993. “The role of relexification in creole genesis: The case of functional categories”. Travaux de recherche sur le créole haïtien 14: 23–52 [UQAM]. Lefebvre, C. 1994. “On spelling out E”. Travaux de recherche sur le créole haïtien 23: 1–33 [UQAM]. Lefebvre, C. 1998. Creole Genesis and the Acquisition of Grammar: The Case of Haitian Creole [Cambridge Studies in Linguistics 88]. Cambridge: Cambridge University Press. Lefebvre, C. 1999. “Substratum semantics in the verbal lexicon of Haitian Creole”. Studies in Language 23: 61–103. Lefebvre, C. 2001. “What you see is not always what you get: Apparent simplicity and hidden complexity in creole languages”. Linguistic Typology 5(2/3): 186–213. Lefebvre, C. and Brousseau, A.-M. 2002. A Grammar of Fongbe [Mouton Grammar Library 25]. Berlin: Mouton DeGruyter. Lefebvre, C. and Kaye, J. 1986. Études syntaxiques, morphologiques et phonologiques. Research report prepared for FCAR and PAFAC on the project Fon-Créole Haïtien, Université du Québec à Montréal, Montreal. Lefebvre, C. and Lumsden, J. S. 1994a. “Le rôle central de la relexification dans la genèse des langues créoles”. Plurilinguismes 8: 47–93. Lefebvre, C. and Lumsden, J. S. 1994b. “Relexification in creole genesis”. In C. Lefebvre and J. S. Lumsden (eds.), The Central Role of Relexification in Creole Genesis: The Case of Haitian Creole. Research report prepared for SSHRCC on the project La genèse du créole haïtien: un cas particulier d’investigation sur la forme de la grammaire universelle, 28 pages. Université du Québec à Montréal, Montreal Lockwood, W. B. 1968. Historical German Syntax. Oxford: Oxford University Press. Longacre, R. E. 1985. “Sentences as combination of clauses”. In T. Shopen (ed.), Language Typology and Syntactic Description, vol. 2. Cambridge: Cambridge University Press, 235–286. Lord, C. 1973. “Serial verbs in transition”. Studies in African Linguistics 4(3): 269–295. Lord, C. 1976. “Evidence for syntactic reanalysis: From verb to complementiser in Kwa”. In S. B. Steever, C. A. Walker and S. S. Mufwene (eds), Papers from the Parasession on Diachronic Syntax. Chicago: Chicago Linguistic Society, 179–191. Mithun, M. 1988. “The grammaticization of coordination”. In J. Haiman and S. Thompson (eds.), Clause Combining in Grammar and Discourse. Amsterdam: John Benjamins, 331–359. Mous, M. 1995. Language intertwining. Paper read at the Amsterdam Creole Workshop: Creole Genesis and Language Contact, Amsterdam. Mous, M. 2001. “Ma’a as an ethno-register of Mbugu”. Sprache und Geschichte in Afrika 16/17: 293–320. (Historical Language Contact in Africa, ed. Derek Nurse.) Cologne: Rüdiger Köppe. Muysken, P. C. 1981. “Half-way between Quechua and Spanish: The case for relexification”. In A. R. Highfield and A. Valdman (eds.), Historicity and Variation in Creole Studies. Ann Arbor, MI: Karoma, 52–79. Nida, E. 1948. “The identification of morphemes”. Language 24: 414–441. Noonan, M. 1985. “Complementation”. In T. Shopen (ed.), Language Typology and Syntactic Description, vol. 2. Cambridge: Cambridge University Press, 42–140. Payne, J. R. 1985. “Complex phrases and complex sentences”. In T. Shopen (ed.), Language Typology and Syntactic Description [Vol. II]. Cambridge: Cambridge University Press, 3–41.

164 Claire Lefebvre

Pepicello, W. J. 1982. “On the sources of Indo-European conjunctions of purpose, cause, and result”. In A. Ahlqvist (ed.), Papers from the 5th International Conference on Historical Linguistics. Amsterdam: John Benjamins, 256–264. Pinker, S. 1989. Learnability and Cognition: The Acquisition of Argument Structure. Cambridge, MA: The MIT Press. Randall, H.J. 1987. Indirect positive evidence: Overtuning overgeneralisations in language acquisition. Unpublished Ms., Indiana University Linguistics Club, Bloomington. Ransom, E. N. 1988. “The grammaticalization of complementizers”. In S. Axmaker, A. Jaisser and H. Singmaster (eds.), Proceedings of the Fourteenth Annual Meeting. General Session and Parasession on Grammaticalization. Berkeley: Berkeley Linguistics Society, 364–373. Ruhl, C. 1989. On Monosemy: A Study in Linguistic Semantics. Albany, NY: State University of New York Press. Singler, J. V. 1996. “Theories of creole genesis, sociohistorical considerations, and the evaluation of evidence: The case of Haitian creole and the relexification hypothesis”. Journal of Pidgin and Creole Languages 11: 185–231. Stassen, L. 2000. “AND-languages and WITH-languages”. Linguistic Typology 4: 1–54. Sterlin, M.-D. 1988. “Les différentes caractéristiques de pou en créole haïtien”. Travaux de recherche sur le créole haïtien 3: 1–34 [UQAM]. Sterlin, M.-D. 1989. “Les caractéristiques de pou: Un modal en position de complémenteur”. Revue québécoise de linguistique 18(2): 131–147. Traugott, E. C. 1985. “Conditional markers”. In J. Haiman (ed.), Iconicity in Syntax: Proceedings of a Symposium on Iconicity in Syntax, Stanford [Typological Studies in Language 6]. Amsterdam: John Benjamins, 289–307. Valdman, A. et al. 1981. Haitian Creole-English-French Dictionary [2 volumes]. Bloomington: Indiana University, Creole Institute. Welmers, W. E. 1973. African Language Structures. Berkeley and Los Angeles: University of California Press.

Chapter 7

Comitative, coordinating, and inclusory constructions in Hausa* Mahamane L. Abdoulaye Abdou Moumouni University, Niamey (Niger)

1. Introduction 2. The comitative preposition dà ‘with’ 2.1 The development of the comitative dà from the existential dà 2.2 General features of the comitative construction 3. The NP coordinator dà ‘and’ 3.1 General features of NP coordination 3.2 Possibilities of ellipsis in coordination 3.3 Grammaticalization from dà ‘with’ to dà ‘and’ 4. Properties of comitative and coordinating constructions 4.1 Syntactic properties of comitative and coordinating constructions 4.2 Semantic and pragmatic properties of comitative and coordinating constructions 5. Inclusory comitative structures and inclusory coordinations 5.1 The inclusory comitative constructions 5.2 The inclusory coordination constructions 6. Conclusion



The particle dà in Hausa is known to have a surprising range of meaning and usages (cf. Abraham 1962: 153–157, Kraft 1970, Newman 2000: 467, etc.). For example, in Newman and Newman (1977: 21–22) dà is translated as: ‘with, have, and, by, by means of, at, during, regarding, in relation to, with respect to, than, that, who, whom, which, while, when, on, since, from the time of, there is/are’. The particle

*I thank Martin Haspelmath for his invitation to write this chapter and for his very helpful comments on earlier versions of the chapter. My thanks also go to the various informants for their grammaticality judgments, in particular Lawali Adjinaku and Sani Sale.

166 Mahamane L. Abdoulaye

also functions as a case-marker for NPs following certain cognition verbs, and it participates in the formation of efferential and causative constructions (see Abdoulaye 1996). This chapter will be concerned with only two of the most frequent usages of dà: dà as the comitative preposition ‘with’, and dà as the NP coordinator ‘and’. These usages are exemplified in (1) and (2) (cf. also Newman and Newman 1977: 21): (1) a.

Maalàm yaa zoo (tàare) dà wani bàaMoo. teacher 3sg.m.pfv come (together with some guest ‘The teacher came together with a guest.’ b. Jiyà ya 1àuki bàkaa dà kwàrii dà kibau. yesterday 3sg.m.sp take bow and quiver and arrows ‘Yesterday he took a bow, a quiver, and some arrows.’

(2) a.

Mun jee kàasuwaa dà Abdù. 1pl.pfv go market with Abdu ‘We went to the market, Abdu and me.’ Or: ‘We went to the market with Abdu.’ b. Muu dà shii mun jee kàasuwaa. 1pl and 3sg.m 1pl.pfv go market ‘He and me, we went to the market.’ Or: ‘He and us, we went to the market.’

In these examples, the dà-phrase appears in different syntactic environments. In sentence (1a), for example, the dà-phrase appears after the verb zoo ‘come’ (optionally with tàare ‘together’) and the meaning is clearly that of the comitative ‘with’. In sentence (1b) two dà particles appear, marking the last two NPs as conjuncts in a three-term ‘and’-coordination. The sentences in (2) illustrate what will be referred to in this chapter as inclusory constructions (also called “asymmetric coordinations” or “plural pronoun constructions”, cf. Schwartz 1989b, Newman 2000: 135–142, etc.), where a plural pronoun anticipates the dà-NP with the possibility of a dual reading, as indicated. In this chapter, I will sketch the grammaticalization of an originally existential predicate dà into the comitative preposition dà ‘with’ and the coordinator dà ‘and’. The chapter also describes the basic features of the comitative and coordinating constructions, and examines some of their semantic, morphological, and syntactic properties. Using these properties, I will show that the Hausa inclusory constructions illustrated in (2a) and (2b) are, respectively, varieties of the basic comitative and coordinating constructions. The chapter is organized as follows. § 2 presents the development of the comitative construction and its basic features. § 3 is devoted to the coordinating construction, its basic features, and the possible elliptical constructions. § 4 looks at some key distinguishing properties of the comitative and coordinating constructions.

Comitative, coordinating, and inclusory constructions in Hausa 167

In § 5 I turn to the inclusory constructions and show that they can be assimilated either to the comitative construction or to the coordinating construction.


The comitative preposition dà ‘with’

This section first focuses on the origin of the comitative marker dà and proposes that it derives from the existential predicate dà ‘there is/are’ through a grammaticalization process. Then some details will be given about the general distribution and the forms of the comitative construction. 2.1 The development of the comitative dà from the existential dà As indicated in the introduction, among the translations of dà one finds the existential meaning ‘there is/are’. This usage seems to be the most basic where dà (or its negative counterpart bâa/baabù ‘there is/are not’) appears in isolation minimally with its object NP. Both dà and its negative bâa are illustrated in the following: (3) a.

Dà sanyii. exist cold ‘It is cold.’ b. Dà ruwaa cikin wannàn kwaanò-n. exist water inside this bowl-def ‘There is water in this bowl.’

(4) a.

Bâa sanyii. not.exist cold ‘It is not cold.’ b. Bâa ruwaa cikin wannàn kwaanò-n. not.exist water inside this bowl-def ‘There is no water in this bowl.’

In (3a) dà predicates the existence of sanyii ‘cold’ in the context of the utterance. In (3b) dà predicates the existence of ruwaa ‘water’ with a locative complement. In both cases, it is possible to replace dà with the synonymous existential predicate àkwai ‘there is/are’ (which does not have a comitative function; cf. àkwai ruwaa ‘there is water’). In its existential use, dà cannot be directly negated; instead, it must be replaced by the negative existential bâa/baabù ‘there is/are not’, as seen in (4a–b). In this respect, the existential dà differs from the comitative dà, which is negated directly with the negator bàa…ba, as bàa dà X ba ‘not with X, without X’ (cf. naa jee kàasuwaa dà Abdù ‘I went to the market with Abdu’, negated as: bàa dà Abdù ba na

168 Mahamane L. Abdoulaye

jee kàasuwaa ‘it is not with Abdu that I went to the market’, cf. discussion of (6) below). All of the sentences in (3)–(4) with dà or bâa are complete propositions with clause-like properties (cf. note 1). In this chapter I will assume that the comitative dà developed from the existential dà through a grammaticalization process. The facts are certainly very complex and there will be no attempt to account for every little shade of meaning or every context of occurrence. Nonetheless, I will propose a semantic development path between the two constructions, and suggest some contexts where the comitative sense will be favored. It is indeed very likely that the change from the existential dà to the comitative dà ‘with’ went through at least two semantic stages, where the dà has the meaning ‘be included among’ or ‘be present’, instead of simply ‘exist’. Two contexts might be distinguished depending on whether the dà-phrase appears initially or sentenceinternally. The first case is illustrated next in (5)–(6): (5) Dà Àali cikin màasu tàfiyàa Lòme. exist Ali inside owner.of going Lome ‘Ali is among those who are going to Lome.’ (6) a.

Bâa Àali cikin màasu tàfiyàa Lòme. not.exist Ali inside owner.of going Lome ‘Ali is not among those who are going to Lome.’ b. Ban-dà Àali cikin màasu tàfiyàa Lòme. neg-with Ali inside owner.of going Lome ‘Ali should not be among those going to Lome.’ / ‘Ali is not among those who are going to Lome.’ c. Bàa dà Àali ba cikin màasu tàfiyàa Lòme. neg with Ali neg inside owner.of going Lome ‘Ali should not be among those going to Lome.’ / ‘Ali is not among those who are going to Lome.’

The affirmative sentence (5) is most naturally translated with ‘be among/be included’, as indicated. It is parallel in structure to the existential sentence in (3b), only here, the existence of Àali is predicated of the group going to Lome. Just as in (3b), the dà can be replaced by the existential àkwai (cf. àkwai Àali cikin màasu tàfiyàa Lòme ‘Ali is among those who are going to Lome’). It may be assumed that the ‘be included’ sense takes on more prominence because the predication universe is a group of individuals, and not a location, as in (3b). But sentence (5) also implies a comitative sense, since Ali and the group are in the same place, i.e., there is an accompaniment relation (see § 4.2 below). To all these meanings correspond indeed three possible negations of (5). First, the negative existential bâa can be used, as seen in (6a), in which case the speaker reports the fact that Ali is not among the travellers. As seen in (6b), sentence (5) can also be negated with bandà, which

Comitative, coordinating, and inclusory constructions in Hausa 169

specifically marks the negation of an inclusion and basically means ‘except’. For example, only bandà can negate emphatic inclusion, as in: koowaa yaa fìta, ha˜r (dà) Abdù ‘everybody got out, even/including Abdu’, which is negated by koowaa yaa fìta, bandà Abdù ‘everybody got out, except Abdu’, and not by *koowaa yaa fìta, bâa Abdù or *koowaa yaa fìta, bàa dà Abdù ba). In this use, bandà is equivalent to sai ‘until, only, except, etc.’ (cf. koowaa yaa fìta, sai Abdù ‘everybody got out, except Abdu’, see Newman 2000: 133 on sai). Finally, (5) can be negated with the comitative negation where the dà-phrase is combined with the general negator bà(a)…ba (cf. bàa Abdù ba ‘not Abdu’; zaa sù fìta ‘they will get out’, and bàa zaa sù fìta ba ‘they will not get out’). In both (6b–c), as the translation indicates, the speaker orders Ali not to be included, or he predicts that he will not be included. The forms can also be used to report the fact that Ali is not among the travellers. It may be concluded then that in compatible contexts, the basic existential sense of dà can shift to marking the ‘be included’ sense or the comitative sense. The dà-phrase can also appear as a complement in sentence-final position. In this case, the exact meaning depends on the nature of the main verb and the type of NP object of dà. Some cases are illustrated in the following: (7) a.

Sun azà dà Àali cikin màasu tàfiyàa Lòme. 3pl.pfv think exist Ali inside owner.of going Lome ‘They thought Ali was among those going to Lome.’ b. Sun azà bâa/ban-dà Àali cikin màasu tàfiyàa Lòme. 3pl.pfv think not.exist/neg-with Ali inside owner.of going Lome ‘They thought Ali was not among those going to Lome.’

(8) a.

Sarkii ya-nàa zàune dà fàadàawa-n-sà tsàye emir 3sg.m-cont sitting exist attendants-of-3sg.m standing à hagu. on left ‘The emir (he was) sitting, with his attendants standing on the left.’ b. Sarkii ya-nàa zàune bâa fàadàawaa. emir 3sg.m-cont sitting not.exist attendants ‘The emir (he was) sitting, without attendants.’

(9) a.

Mun yi sallàa dà lìimaamìi. 1pl.pfv do prayer with imam ‘We prayed with an imam [who was leading the prayer].’ b. Mun yi sallàa bâa lìimaamìi. 1pl.pfv do prayer not.exist imam ‘We prayed without an imam.’

(10) a.

Mun yi sallàa dà Abdù. 1pl.pfv do prayer with Abdu ‘We prayed with Abdu.’

170 Mahamane L. Abdoulaye

b. Mun yi sallàa bàa dà Abdù ba. 1pl.pfv do prayer neg with Abdu neg ‘We prayed without Abdu.’ In (7a), the same sentence given in (5) appears as complement to a cognition verb. The dà-phrase essentially retains the same shades of meaning it has in isolation. Just as in (5), dà in this context can be replaced with the mainly existential predicate àkwai ‘there is/are’ (cf. sun azà àkwai Àali cikin màasu tàfiyàa Lòme ‘they thought Ali was among those who are going to Lome’). The sentence is better negated with the negative existential bâa or the inclusion negation bandà, as indicated in (7b), but the comitative negation is also possible (sun azà bàa dà Àali ba cikin màasu tàfiyàa Lòme ‘they thought Ali was not among those who are going to Lome’). In (8a), the dà particle is not replaceable by àkwai (cf. *sarkii yanàa zàune àkwai fàadàawansà… ‘the emir was sitting with his attendants…’). Dà here probably predicates more the presence of the attendants, so that it can felicitously be replaced by gàa ‘here is/are’ (cf. sarkii yanàa zàune gàa fàadàawansà tsàye ‘the emir was sitting, with his attendants standing by’). Nonetheless, the sentence can be negated with the negative existential bâa (as illustrated in (8b)), or with the comitative negation (sarkii yanàa zàune bàa dà fàadàawaa ba ‘the emir was sitting, without attendants’), which shows that the accompaniment sense is also present. The sentences in (9)–(10) contrast a generic and a specific noun. In both cases, dà cannot be replaced with the existential àkwai (*mun yi sallàa àkwai lìimaamìi/Abdù ‘we prayed with an imam/Abdu’), which shows that the two constructions are not existential. With the generic noun in (9), the sentence can be negated with the existential bâa (as given in (9b)) or with the comitative negation (cf. mun yi sallàa bàa dà lìimaamìi ba ‘we prayed without an imam’). With the specific noun in (10), only the comitative negation is possible, as given in (10b), the negative existential bâa being quite marginal (cf. ??mun yi sallàa bâa Abdù ‘we prayed without Abdu’; the meaning ‘in the absence of Abdu’ must be rendered as: mun yi sallàa bâa Abdù nan ‘we prayed while Abdu was not around’). It seems then that certain contexts (with cognition or speech predicates for example) do not alter the possible meanings of a dà-phrase inside a clause, whereas other contexts favor the ‘be present’ or the comitative senses. There are doubtless many factors, pragmatic, semantic, syntactic, etc., which can influence the readings. For example, the comitative meaning in (8) will be favored if fàadàawaa ‘attendants’ is replaced with gwamnà ‘governor’ (cf. sarkii yanàa zàune dà gwamnà ‘the emir was sitting with the governor’). Specification of some complements may also favor one meaning or the other. For example, Feemì yaa yi kasòo dà màataa can mean ‘Femi did prison with a wife’ or ‘Femi did prison together with his wife’. But in Feemì yaa yi kasòo dà màataa à gidaa the locative à gidaa ‘at home’ forces a ‘be present’ sense, i.e., ‘Femi did prison, with a wife (being present) at home’.

Comitative, coordinating, and inclusory constructions in Hausa

In conclusion, the data taken together allow one to assume an original existential dà predicate, which may appear as a proposition with its object in isolation, as in (5), or inside a clause, as in (7). This dà-phrase may inferentially have a ‘be included, be present’ meaning or a comitative meaning. Some contexts may favor one of these secondary meanings only, with the consequence that dà can no longer be replaced there with the existential àkwai ‘there is/are’. In the comitative constructions, such as (9–10), one may consider the dà-phrase as a simple prepositional phrase ‘with NP’ inside a simple predication.1 2.2 General features of the comitative construction As shown in (1a) there are clear contexts where dà is the comitative preposition ‘with’. As may be expected, a major syntactic difference between comitative dà and existential dà is the fact that the former typically appears inside regular clauses, usually following some verb, whether transitive or intransitive, as seen next: (11) a.

Naa gàmu dà Boolà bàaki-n kàasuwaa. 1sg.pfv meet with Bola border-of market ‘I met with Bola near the market.’

1.Although there is no explicit discussion of the relation between existential dà and comitative dà, most students of Hausa assume the centrality of the comitative preposition dà in relation to all other usages (cf. Wolff 1993: 454, Newman 2000: 178ff, 222ff). There are however many reasons to take the existential dà as basic. First, existential predications behave like regular propositions in taking clausal conjunctions and connectors such as the conditional ìdan ‘if ’ or the clausal coordinator kuma ‘and (also)’ (cf. ìdan dà laimàa ‘if there is humidity’; dà hatsii kuma dà masà˜raa ‘there is millet and there is corn’, cf. discussion of (15) below). Secondly, if the existential dà is like a verb, it should normally not derive from an adposition, since in grammaticalization theory, verbs change into adpositions and not the reverse (cf. for example Heine and Reh 1984, etc.). A third reason for taking the existential usage as basic is the fact that dà alternates with the predicate àkwai ‘there is/are’ and the negative bâa/baabù, all of which have no clear origin. It should be noted that even if àkwai still has not grammaticalized to a comitative, it can be used with the ‘be included’ sense, as in: àkwai mù cikin tsaarìn naakà? ‘are we included in your plans?’. It happens that predicates meaning ‘be included in, be with’ are thought to be the source of the comitative preposition ‘with’ in Yoruba and Ewe (cf. Heine and Reh 1984: 58, 62; also see Stassen 2000: 17 with a similar analysis for Choctaw, Korean and Classical Mongolian). I will assume that dà is becoming a bleached existential, which must be reinforced with or replaced by the (new?) existential predicate àkwai (cf. dà sanyii = dà àkwai sanyii = dàkwai sanyii = àkwai sanyii ‘there is cold’). A very similar situation obtains in Kanuri, where a bleached existential adposition with a comitative use (-a) contrasts with an existential predicate (mbeji/mbezai) and their shared negative existential (bâa/baawò; cf. Hutchison 1981: 168–169).


172 Mahamane L. Abdoulaye

b. Zaa kì tàfi (tàare) dà Muusaa nèe? Aa’àa, (tàare) fut 2sg.f go (together with Musa cop no (together dà Bàlki. with Balki ‘Are you going with Musa?’ ‘No, with Balki.’ c. Mun îskee Kòofîi ya-nàa zàune (tàare) dà bàaMii. 1pl.pfv find Kofi 3sg.m-cont sitting (together with guests ‘We found Kofi sitting with guests.’ d. Maalàm yaa noomè goonaa (tàare) dà àlmàajì˜ra-n-shì. teacher 3sg.m.pfv hoe farm (together with students-of-3sg.m ‘The teacher hoed the farm with his students.’ e. Sarkii yaa koomàa cikin gidaa (tàare) dà emir 3sg.m.pfv return inside house (together with fàadàawa-n-shì. attendants-of-3sg.m ‘The emir returned inside with his attendants.’ f. Duk wan-dà kèe yîn hakà ya-nàa tàare dà wàhalàa. all who-that rc do so 3sg.m-cont together with trouble ‘Whoever does this is bound to live with trouble.’ The comitative dà-NP appears most naturally following an intransitive association/ dissociation verb, as seen in (11a). In this case one cannot use the adverb tàare ‘together’. In all the other sentences of (11), however, the adverb tàare (derived from the verb taarà ‘gather’) can be used before the dà-NP, especially with a motion verb as in (11b), another derived adverbial form as in (11c), a transitive verb and its direct object as in (11d), a locative phrase as in (11e), etc. The adverb tàare ‘together’ is however obligatory with the locative copula nàa/kèe, as seen in (11f), unless some other element, such as a locative adverbial, appears in between, as in: tanàa can (tàare) dà bàaMii ‘she is there (together) with the guests’. Other derived adverbs (called “statives” in Hausa literature) that frequently occur with comitative dà are bìye ‘followed’, rìMe ‘holding’, and ràke ‘accompanied’. The comitative dà-phrase can appear in isolation, but this happens only in contexts such as an answer to a question, as shown in (11b), or in formulaic expressions (cf. the beginning formula in letters: dà faatan kunàa laafiyàa ‘[I am writing] with the wish that you are in good health’). Comitative constructions involving nominal or pronominal direct objects may be subject to confusion with a coordination structure, as seen in the following: (12) a.

Naa ga Abdù (tàare) dà àbikkiyà-˜r-shì. 1sg.pfv see Abdu (together with friend-of-3sg.m ‘I saw Abdu with his friend.’

Comitative, coordinating, and inclusory constructions in Hausa

b. Naa ganee shì dà àbikkiyà-˜r-shì. 1sg.pfv see 3sg.m with friend-of-3sg.m ‘I saw him with his friend.’ In sentence (12a), the noun Abdù is the sole direct object, while the dà-NP is a comitative adjunct. However, only the insertion of tàare ‘together’ can distinguish (12a) from a sentence with coordinated objects, such as: naa ga Abdù dà àbikkiyà˜rshì ‘I saw Abdu and his friend’. In (12b) the verb assumes a special prepronominal form ganee ‘see’ before direct object pronouns. Also, the pronoun itself assumes a special (reduced, weak) object form, different from the (full, strong) independent form. For this reason, the comitative sentence (12b) is different on the surface from a sentence with coordinated direct objects, as in: naa ga shii dà àbikkiyà˜rshì ‘I saw him and his friend’, where the verb does not have the prepronominal form, and the conjunct pronoun takes the independent form. The comitative dà can also have an instrumental or associative/possessive meaning, especially in NP syntax, as illustrated in (13): (13) a.

Yaa yankà naamàa dà wuMaa. 3sg.m.pfv cut meat with knife ‘He cut the meat with a knife.’ b. Aikìi dà kwàashee yaa fi aikìi dà hannuu. work with hoe 3sg.m.pfv surpass work with hand ‘Working with a hoe is better than working with bare hands.’ c. Naa ga yaaròo dà kàree. 1sg.pfv see boy with dog ‘I saw a child with a dog.’

In sentence (13a) the phrase dà wuMaa ‘with a knife’ has an instrumental reading. In (13b) the dà-phrases also have an instrumental reading, but they appear linked to the action noun aikìi ‘work’ in an NP construction. In (13c) the dà-phrase occurs with a concrete noun yaaròo ‘boy’, and in such cases, the sense can be comitative/possessive (or comitative/associative if the dà-NP expresses a quality, cf. naa ga wata goonaa dà faa1ii ‘I saw such a large farm’; for the ‘and’-coordination sense of (13c) see discussion of (18a) in § 3.3 below). Next we turn to the coordinating conjunction.


The NP coordinator dà ‘and’

This section explores some aspects of conjunctive coordination using the marker dà, a particle clearly related to the dà particles presented in the previous sections. I will first review some general features of coordination and then focus on the elements which can be subject to ellipsis. Finally I will look at the development of dà ‘and’ from the preposition dà ‘with’.


174 Mahamane L. Abdoulaye

3.1 General features of NP coordination As seen in (1b), the particle dà can function as the coordinator ‘and’, and it can relate NPs (including head nouns, pronouns, and nominalized clauses) as well as adjectives, adverbs, and prepositional phrases, but not regular tensed clauses. This is illustrated in (14): (14) a.

Abdù dà Feemì sun ga juunaa kàasuwaa. Abdu and Femi 3pl.pfv see each.other market ‘Abdu and Femi met one another in the market.’ b. An zàa'i nii dà kai. / Nii dà kai akà zàa'aa. imps.pfv choose 1sg and 2sg.m 1sg and 2sg.m imps.rp choose ‘One chose us, you and me. / It is you and I that were chosen.’ c. Hànkaakàa nee tsuntsuu baMii dà farii. crow cop bird black and white ‘The crow is the white and black bird.’ d. Ruwâ-n sun màlaalà cikin 1aakìi dà (kuma) cikin water-def 3pl.pfv flow inside room and (also inside fiilin gidaa. yard ‘The water filled the room and the yard (as well).’

In (14a) two nouns are coordinated with dà as subjects. Sentence (14b) shows two conjoined pronouns, and both must be of the independent series, no matter what their syntactic function (cf. also Heath (1999: 149) for similar facts in Koroboro Senni, and see also the discussion of (12b) above for the corresponding comitative structures). In (14c), a rather rare construction, two adjectives are conjoined to qualify a noun. This simultaneous qualification is possible, as far as I can tell, only for colors, to express different but related qualities. One cannot express unrelated qualities in this way. So, one cannot say: *mùtum tsoohoo dà doogoo for ‘a tall and old man’, and a simple juxtaposition or the relator kuma must be used, as in: mùtum tsoohoo (kuma) doogoo ‘a tall old man’.2 Finally in (14d) two prepositional phrases are conjoined with an optional kuma ‘and (also)’. In some cases, dà may conjoin terms with a propositional content, as seen in the following: (15) a.

À goonaa sun faarà sàssàben itàacee dà shuukà˜r daawàa. in farm 3pl.pfv start clearing bush and sowing millet ‘In the farm they started clearing the bush and sowing the millet.’

2.Newman (2000: 132) does give: wata yaarinyàa faraa dà (kuma) dooguwaa ‘a light-skinned and (also) tall girl’, but on p. 135 he says such constructions are not possible.

Comitative, coordinating, and inclusory constructions in Hausa

b. *Sun sassàbè itàacee dà sun shuukà daawàa. 3pl.pfv clear bush and 3pl.pfv sow millet ‘They cleared the bush and sowed the millet.’ c. Ha˜r yànzu su-nàa haMa˜r tàare dà saya˜r dà lu’ùlu’ù. till now 3pl-cont extract together with selling diamond ‘Even now they [RUF (Revolutionary United Front) rebels] extract and sell diamonds.’ In (15a) two nominalized clauses are conjoined. The particle dà cannot be used to conjoin sentences, as seen in (15b), and the clausal conjunction kuma must be used instead (thus: sun sassàbè itàacee kuma sun shuukà daawàa ‘they cleared the bush and sowed the millet’). Sentence (15c) presents a mixed construction very frequent in journalistic style, where a finite clause is conjoined with a nominalized clause. In this type of sentence, the adverb tàare ‘together’ is obligatory, and yet the semantics is that of conjunction. If the direct object is identical for both verbs, it is ellipsed for the first verb. In ordinary usage such sentences are expressed with clausal coordination using the coordinator kuma ‘and (also)’, as in: ha˜r yànzu sunàa haMa˜r lu’ùlu’ù kuma sunàa saya˜r dà shii ‘even now they extract diamonds and sell them’. Following the typology given in Haspelmath (to appear), the constructions in (14)–(15) above illustrate monosyndetic prepositive coordination. It is (mono)syndetic since an overt, but single conjunction dà is specified for two coordinates. It is prepositive because the conjunction marks the second term of the coordination, in a structure ‘A dà-B’. Indeed, the second coordinand is easily shown to form a constituent with dà. For example, in a coordination with multiple coordinands, as in a careful listing of the items by the speaker, the pauses will intervene before the dà’s (cf. kin ga, Abdù, dà Bàlki, dà Muusaa, dà Mo˜rù, duk bàa mutàanen ki˜rkìi ba nèe ‘you see, Abdu, Balki, Musa and Moru, they are all unrecommendable people’). Secondly, the multiple terms can be left unmarked except for the last one (so the previous example can also be: kin ga, Abdù, Bàlki, Muusaa, dà Mo˜rù, duk bàa mutàanen ki˜rkìi ba nèe). In emphatic coordination on the other hand, both terms are marked by a preposed dà (cf. dà Abdù dà Bàlki, duk sun tàfi ‘both Abdu and Balki have gone’; cf. Newman 2000:135 for more on these constructions). Also, the movable focus marker nee/cee can be inserted before, but not after, dà in a conjunction (cf. Abdù dà Muusaa nèe sukà tàfi = Abdù nee dà Muusaa sukà tàfi ‘it is Abdu and Musa who went’, and not *Abdù dà nee Muusaa sukà tàfi). This test relies on the property of the focus marker nee/cee to stand after or inside a complex focused structure, such as a coordination or a conditional or temporal subordinate clause. For example, in [dà Abdù yanàa gyaaran mootàa] nee bàaMii sukà zoo ‘it is [when Abdu was repairing the car] that the guest arrived’, the focused temporal clause in brackets is followed by the focus marker nee. However, the sentence has the same sense as [dà Abdù nee yanàa gyaaran mootàa] bàaMii sukà zoo, where the whole temporal clause is still under focus, despite the fact that the focus marker nee has moved inside it.


176 Mahamane L. Abdoulaye

Beside these (and doubtless other indications), the structure ‘A dà-B’ conforms with the Hausa SVO word order and it is compatible with the likely source of the conjunction, i.e., the comitative preposition ‘with’, as discussed in § 3.3 below. Nonetheless in some constructions, characterizable as “natural coordinations”, the two terms may be more integrated and it is not clear whether one has an ‘A dà-B’ or an ‘A-dà-B’ structure. Such sentences are for example: màcè dà namijì, bàbba dà yaaròo, dukà tàfiyàa akài wurin fàrautàa ‘men and women, old and young, all are going to the hunt party’ (adapted from Caron 1991: 93f), or: à sana˜r dà mutàanee, mazaa dà maataa, yaaròo dà bàbbaa, bàaMoo dà 1an gàrii, … ‘that one let it be known to everybody, men and women, young and old, strangers and citizens, …’ (adapted from NNPC 1968: 192). In these constructions, one has coordination chunks, involving naturally occurring terms only, and dà appears only in the middle of the chunks, though a supplementary dà can appear in between the chunks in a kind of “superordinate” coordination of coordinations (cf. mazaa dà maataa dà yaaròo dà bàbbaa ‘[men and women] and [young and old]’, etc.). Another natural coordination is uwaa dà ùbaa ‘parents’, lit. ‘mother and father’ (= iyàayee), which appears in generic context, cf. baa àa Mîn màganà˜r uwaa dà ùbaa ‘one does not reject parental advice’, vs. Abdù yaa Mi màganà˜r uwa˜rsà dà ùbansà/*uwaa dà ùbaa ‘Abdu refused the advice of his parents’. Other such expressions are: bàakii dà hancìi ‘very close’ (lit. ‘mouth and nose’); gàbaa dà gàbaa ‘face to face’; mijìi dà màataa ‘couple’ (with no plural modification possible), wâa dà Manèe ‘tandem of an elder and a younger brother’ (despite the fact that in Katsinanci magàajii is the normal term for ‘elder brother’); àbookii dà àbookii ‘a tandem of friends’; Hasàn dà Hùsainì ‘a tandem of twins’. While in regular comitative constructions the dà can be emphasized with kuma ‘also’, this is not possible in these fixed expressions (cf. mijìi dà kuma màataa ‘the husband and also the wife’, which does not mean ‘couple’). 3.2 Possibilities of ellipsis in coordination As the examples of this section show, dà-coordination conforms to the general definition of coordination as given in Haspelmath (to appear) since it builds a syntactic structure in which “two or more units of the same type are combined into a larger unit and still have the same semantic relations with other surrounding elements.” Nonetheless, in Hausa as in other languages, there are cases where the conjuncts are not structurally identical, and such structures with dissimilar terms are usually said to involve the ellipsis of repeated elements. For example, Newman (2000: 31f, 136) reports many such cases involving ellipsis of modifiers or heads, as seen in the following (example (16a) is from Newman 2000: 31): (16) a.

rìkìrkìtaccen dookìi dà Ø jàakii horse and donkey ‘confused horse and confused donkey’

Comitative, coordinating, and inclusory constructions in Hausa 177

b. rìiga-˜r Boolà dà taakalmàa Ø gown-of Bola and shoes ‘Bola’s gown and shoes’ In example (16a) the preposed (i.e. prenominal) singular adjective can modify both nouns, as given in the translation (the meaning ‘a confused horse and a donkey’ is also possible). In (16b) the modifier is a possessor noun, and Boolà is understood to be the owner of both the gown and the shoes. Example (16a) involves ellipsis, and not an adjective modifying two coordinated nouns, since in this case the adjective would be plural, as in mafà1àatan mijìi dà màataa ‘quarrelsome husband and wife’, where mafà1àataa ‘quarrelsome’ is plural (cf. Newman 2000: 32). As for relative clauses, they may not be deleted, but one relative clause can modify two coordinated nouns (cf. 'aunaa dà giiwa˜r dà mukà gàmu dà suu ‘the buffalo and the elephant that we met (with them)’, where the relative clause contains a plural pronoun referring to both nouns). According to Newman (2000: 32), for examples such as (16a) to be possible, the nouns must match in gender. But, for Niger speakers at least, the two nouns can differ in gender, as in: farin ràagoo dà Ø tumkìyaa ‘white ram and white ewe’ (with a masculine adjective for ràagoo ‘ram’ despite the feminine gender of tumkìyaa ‘ewe’), though this may be due to the natural character of the set. It appears, however, that only a preposed adjective can allow the ellipsis, so that one cannot have: *dookìi rìkìrkìtacce dà jàakii Ø ‘confused horse and confused donkey’, with a postposed adjective. A plural adjective modifying two coordinated nouns can be preposed or postposed (so that ‘quarrelsome husband and wife’ can also be mijìi dà màataa mafà1àataa). Besides the modifiers, the heads, too, can be deleted, but this is restricted to adjective modification, as seen in the following (see also Newman 2000: 135–136): (17) fara˜r mootàa dà baMaa Ø (= faraa Ø dà baMa˜r mootàa) white car and black white and black car ‘a white car and a black one’ In (17) the head mootàa ‘car’ is modified by a preposed adjective and ellipsis is possible. Catalipsis (anticipatory ellipsis) of the head is also possible, as seen in the parentheses. Catalipsis is even better with emphatic coordination where each term is marked with dà: dà faraa dà baMa˜r mootàa ‘a white car and a black one indeed’. The head can also be deleted with a postposed adjective, as seen in: mootàa faraa dà Ø baMaa, but apparently catalipsis is no longer possible, even with emphasis (cf. *(dà) Ø faraa dà mootàa baMaa). Non-color adjectives essentially function in the same way: bàbba˜r mootàa dà Màramaa Ø ‘a big car and a small one’ = (dà) bàbba Ø dà Màrama˜r mootàa; or with postposed adjectives: mootàa bàbba dà Ø Màramaa, but *(dà) Ø bàbba dà mootàa Màramaa. The type of modifiers can be mixed, as long as they are semantically close, as seen in: kàaza˜r gidaa dà Ø bàaMuwaa ‘in-house hen

178 Mahamane L. Abdoulaye

and a stranger one’ (i.e. a hen belonging to the household and one which is a stranger), where kàaza˜r gidaa is lit. ‘hen-of house’ and bàaMuwaa a simple adjective. Ellipsis in numeral phrases will not be discussed here, as different speakers — including myself — have difficulties in judging the constructions. With possessive constructions and relative clauses the obligatory use of a resumptive pronoun does not allow one to speak of head ellipsis. So for example in dooyà˜r Muusaa da ta Hàbii ‘the yam of Muusaa and that of Habi,’ the head dooyàa ‘yam’ is referred to by the pronoun ta. Omission of the pronoun yields a head modified by two coordinated possessors (dooyà˜r Muusaa dà Hàbii ‘the yam of Musa and Habi’, which cannot mean ‘the yam of Musa and the yam of Habi’). However, non-possessive genitive constructions may allow omission of the pronoun and with head ellipsis, as in: mun sawoo bùhun daawàa dà Ø masà˜raa ‘we bought a sack of millet and (a sack of) corn’, or: bugùunaa dà Ø kee wàndà sukà yi ‘the beating-of me and (with?) you which they did’). Similarly, in the sentence giiwaa waddà akà kaamàa dà waddà akà kashèe ‘the elephant that one caught and the one that one killed’, giiwaa ‘elephant’ is referred to by a relative pronoun (waddà) in both relative clauses. 3.3 Grammaticalization from dà ‘with’ to dà ‘and’ A grammaticalization process from comitative to a conjunction ‘and’ has been proposed for many languages all over the world where the two notions are expressed with the same morpheme. Languages where the comitative preposition and the ‘and’-coordinator are related include Kanuri, Ewe, Yoruba, Logbara, Songhay, among others (cf. Hutchison 1980, Heine and Reh 1984:255, Crazzolara 1960, Heath 1998:60, 112). According to Liu (1991:86) most Chadic languages and many other languages, including some Bantu languages, would also have the same morpheme for ‘with’ and ‘and’ (cf. Schuh 1998: 276 for Miya). Other languages that display the same pattern are cited in Haspelmath (to appear) (see also the discussion in Payne (1997: 339)). Two general discussions of the grammaticalization of ‘with’ prepositions to ‘and’ conjunctions are Mithun (1988: 339–340) and Stassen (2000: 25–34). As suggested in Haspelmath (to appear), the principal ground for a shift from a comitative marker to a conjunction ‘and’ is semantic. Indeed, a sentence such as John left with his father is essentially the same as John and his father left. Nonetheless the two constructions are not the same in their semantic and pragmatic implications, as will be seen later in § 4.2. However, in a process similar to the one underlying the change from existential dà to comitative dà, one can assume that in certain contexts, the secondary ‘and’ meaning of some comitative constructions is favored over the ‘with’ sense, and this realizes the grammaticalization shift. Below I present some illustrative cases. In § 2.2 we have seen that comitative phrases appear most typically after a verb, and this verb can be transitive or intransitive. When the verb is transitive, then the

Comitative, coordinating, and inclusory constructions in Hausa 179

comitative phrase follows the direct object, and the nouns may be read as coordinands. When the verb is intransitive, for example a motion verb, then only multiple comitative phrases may be interpreted as coordinations. Both cases are illustrated next: (18) a.

Maalàm bâi sô-n yaaròo dà kàree. teacher neg.cont.3sg.m liking-of child da dog ‘The teacher does not like to see a child with a dog.’ Or: ‘The teacher fears children and dogs.’ b. Nà’oomì taa zoo tàare dà Saanii dà Bàlaa. Naomi 3sg.f.pfv come together with Sani da Bala ‘Naomi came with Sani and with Bala.’ / ‘Naomi came with Sani and Bala.’

In (18a), the sequence yaaròo dà kàree can be interpreted as comitative where only yaaròo is direct object, i.e. the teacher dislikes only those cases where children are walking with dogs. It can also have a conjunctive meaning, in which case the teacher fears children and dogs separately. In the case of (18b) things are more intertwined and in practice the two constructions may be indistinguishable, as seen in the translations. This also emphasizes the semantic relations as the most decisive factor in the grammaticalization shift. This account is at variance with that given in Stassen (2000: 25–34), who claims that an SVO language may acquire the coordinate structure by a movement of the comitative phrase to pre-verbal position, next to the subject, i.e. “NP1 V with NP2” Æ “NP1 with NP2 V”. I assume here, rather, that an NP may take an adnominal comitative phrase, as in (18a), and the whole construction may then shift to a coordinate structure (cf. for the subject position: wata màcè dà jàariirìntà tanàa zàune wàje ‘a woman with her baby is sitting outside’ and wata màcè dà 1ìya˜rtà sunàa zàune wàje ‘a woman and her child are sitting outside’). In conclusion to this section, there is a distinct coordinator dà ‘and’ which conjoins various non-clausal constituents. Although historically it derives from the comitative dà ‘with’, it has its own basic semantic and morphosyntactic properties, as we will see in the next section.

4. Properties of comitative and coordinating constructions In this section I explore a few of the syntactic and semantic/pragmatic properties of the comitative and coordinating constructions. This will allow — later in § 5 — the classification of the Hausa inclusory constructions into two types. I will successively look at preposing in topicalization and focalization, the patterns of topicality and coreference with pronouns, and finally the degree of involvement of the participants in the verb’s action.

180 Mahamane L. Abdoulaye

4.1 Syntactic properties of comitative and coordinating constructions In § 2 and § 3 we saw some basic environments for comitative or conjoined dà-phrases. In the comitative construction, it is possible for the dà-phrase to appear in sentence-initial position in topicalization or focalization. Such preposing, however, is not possible in the case of coordination. The two constructions are illustrated in the following: (19) a.

(Tàare) dà yâara-n-shì fa, yaa zoo nannìyà. (together with children-of-3sg.m indeed 3sg.m.pfv come here ‘With his children indeed, he came here.’ b. (Tàare) dà Bàlaa nèe na jee kàasuwaa. (together with Bala cop 1sg.rp go market ‘It is with Bala that I went to the market.’

(20) a. *Dà mussàa kàm, Abdù yaa kashè kàree Ø. with cat indeed Abdu 3sg.m.pfv kill dog ‘As for the cat, Abdu killed the dog and (it).‘ b. *Dà mussàa nee Abdù ya kashè kàree Ø. with cat cop Abdu 3sg.m.rp kill dog ‘It is the cat that Abdu killed the dog and (it).‘ In (19a–b), a comitative phrase is topicalized and focused, respectively, with the optional adverb tàare ‘together’. In contrast, the sentences in (20) are ungrammatical with a preposed conjunct NP. The same contrast is also observed with other focus constructions such as wh-questions and relative clauses.3 Hausa would then be like most languages in that extraction is blocked out of coordinated structures (cf. for example, the discussion of Russian in Haspelmath (to appear) and the vast generative literature on this subject). Another syntactic property of the comitative construction is the fact that the subject NP alone is the topic of the sentence; it cannot be repeated in a following clause, but must be referred to with a pronoun. Repetition of the dà-NP on the other hand seems acceptable. This is illustrated next:

3.Schwartz (1989b: 38n7) gives two purported examples of extraction out of a coordination, but one sentence is clearly ungrammatical, and in any case it does not illustrate focalization (da Binta ce, da Audu sun tafi ‘as for Binta, she and Audu went’ — a sentence with a nonfocus tense/aspect marker). In the other sentence, both conjuncts are under focus (da Musa ne da Audu za: sù zo, which she translate as: ‘MUSA and Audu will come’, but which means ‘It is both Musa and Audu who are going to come’, and is equivalent to: dà Muusaa dà Abdù nee zaa sù zoo, with a movable focus marker, cf. the discussion and the examples in § 3.1 above).

Comitative, coordinating, and inclusory constructions in Hausa

(21) Abdù yaa tàfi maka˜rantaa dà Bàlki an bâa *Abdù/Bàlki Abdu 3sg.m.pfv go school with Balki imps.pfv give Abdu/Balki àlloo. board ‘Abdu went to school with Balki, and a [writing] board was given to *Abdu/Balki.’ This sentence indeed shows that the subject Abdù is the sole topic of the sentence. The dà-NP Bàlki does not have the same topicality as the subject and can consequently be repeated, as indicated. Conjoined NPs on the other hand globally constitute the topic of the sentence, and if taken separately they can each be repeated. This is illustrated in the following: (22) Abdù dà Bàlki sun tàfi maka˜rantaa an bâa Abdù/Bàlki Abdu and Balki 3pl.pfv go school imps.pfv give Abdu/Balki àlloo. board ‘Abdu and Balki went to school and a board was given to Abdu/Balki.’ This sentence shows that neither conjunct NP alone is the topic of the sentence, otherwise they would not be repeatable in the second clause. That the topic of the sentence is the whole conjunction is shown by the fact that the two NPs can be referred to by a plural pronoun, as in: Abdù dà Bàlki sun tàfi maka˜rantaa an baa sù àlloo ‘Abdu and Balki went to school and they were given a board’. Comitative and coordinating constructions involving direct objects also contrast in their coreference patterns with following pronouns. Comitative direct objects can control following pronouns together or separately, while coordinated direct objects must be jointly referred to by a plural pronoun. This is illustrated in the following: (23) a.

Naa ganee shì (tàare) dà àbikkiyà-˜r-shì ya-nàa / 1sg.pfv see 3sg.m (together with friend-of-3sg.m 3sg.m-cont/ ta-nàa /su-nàa sàye-n goo˜rò. 3sg.f-cont/3pl-cont buying-of kolanuts ‘I saw him with his friend, he was/she was/they were buying kolanuts’. b. Naa ga shii dà àbikkiyà-˜r-shì *ya-nàa /*ta-nàa / 1sg.pfv see 3sg.m and friend-of-3sg.m 3sg.m-cont/3sg.f-cont/ su-nàa sàye-n goo˜rò. 3pl-cont buying-of kolanuts ‘I saw him and his friend buying kolanuts.’


182 Mahamane L. Abdoulaye

In sentence (23a) the accusative object pronoun shì ‘3sg.m’ and the optional tàare ‘together’ signal a comitative construction. As is indicated, the two NPs can have separate control of the pronoun in the following clause. In (23b) the independent pronoun shii ‘3sg.m’ signals a coordinated structure, and the two NPs must jointly control the pronoun in the following clause. If one of the NPs must be referred to separately, then it will be repeated. (For the distinction between the accusative and the independent pronouns, see the discussion of (12b, 14b) in § 2 and § 3.) It can be concluded that comitative and coordinating constructions have different syntactic properties, which is not surprising given the fact that a coordinate structure forms one constituent, while an NP and a comitative dà-phrase are separate constituents. 4.2 Semantic and pragmatic properties of comitative and coordinating

constructions As seen in § 3.3, in some appropriate contexts, the comitative and the coordinating constructions are equivalent, or they may even be indistinguishable (cf. discussion of (18) and note 4). Nonetheless, it seems that in most cases, comitative and coordinating constructions can still be distinguished. First, let us consider the following data: (24) a. John went to the party with Mary. = John and Mary went to the party. b. John served the soup with the entrée. = John served the soup and the entrée. In an initial study of these examples, Foley and Van Valin (1984: 84) proposed that in (24a), John and Mary are co-actors in both sentences, and in (24b) the entrée and the soup are patients in both sentences. They then proposed that for certain English verbs, comitative and coordinating constructions can be taken to be equivalent. Jolly (1993: 299) however reports that native speakers are evenly split in deciding, of John or Mary, who is in control in the first sentence of (24a). The judgement is clear in the case of the second sentence, where both John and Mary are in control of the verb’s action. Similarly, the soup and the entrée may have different salience in the first sentence of (24b), but they have equal salience in the second sentence (also see Stassen 2000: 6 on the differences and similarities between ‘with’ and ‘and’). So it seems, as suggested by Jim Miller in 1985 (cited in Jolly 1993: 297), that the comitative construction in English only specifies that one object is in the same place as another object. This is clearly seen in examples such as the following: (25) a.

Feemìi yaa tàfi Ìkko dà ùba-n-shìi. Femi 3sg.m.pfv go Lagos with father-of-3sg.m ‘Femii went to Lagos with hisi father.’

Comitative, coordinating, and inclusory constructions in Hausa

b. Feemìi dà ùba-n-shìi sun tàfi Ìkko. Femi and father-of-3sg.m 3pl.pfv go Lagos ‘Femi and his father went to Lagos.’ In sentence (25a), Feemì is the subject/topic of the sentence and is referred to by the possessive pronoun in ùbanshì ‘his father’. Nonetheless speakers understand that Femi is a child and has no control of the situation. To show control of the situation by Femi, an explicit verb must be used, as in: Feemì yaa kai ùbanshì Ìkko ‘Femi took his father to Lagos’. Since Hausa does not allow backward reference in this context, the possessive (and volitional) ùbanshì cannot appear as the subject of the sentence. In sentence (25b) on the other hand, control of the situation is equally vested in Femi and his father, and they may go to Lagos together or separately. Comparable semantic and pragmatic contrasts also obtain with constructions involving object NPs, as seen in the following: (26) a.

Abdù yaa saidà gida-n-shì tàare dà Abdu 3sg.m.pfv sell house-of-3sg.m together with garka-˜r-shì, (*gidân bàara, garkâ˜r kuma bana). garden-of-3sg.m (*house last.year garden and this.year ‘Abdu sold his house with his garden, (*the house last year, and the garden this year).’ b. Abdù yaa saidà gida-n-shì dà garka-˜r-shì, Abdu 3sg.m.pfv sell house-of-3sg.m and garden-of-3sg.m (gidân bàara, garkâ˜r kuma bana). (house last.year garden and this.year ‘Abdu sold his house and his garden, (the house last year, and the garden this year).’

As seen in (26a), the comitative construction does not allow a reading where the house and the garden are sold separately. There is also no implication as to whether the garden was negotiated or simply given away as a bonus to the house buyer. In (26b), however, the house and the garden are both negotiated, but there is no implication as to whether they were sold together or separately. This shows that in general, comitative constructions signify that the two referents are in the same place (the basic accompaniment sense), or they are considered at the same time with respect to the verb’s action. Conjunction on the other hand presupposes an equal involvement of the two participants, but this involvement can be in different circ*mstances.4

4.There are some reciprocal association/dissociation verbs or other verbs and expressions where the action equally concerns the two participants in a similar way. With these verbs and expressions, the syntactic context, i.e. the comitative and coordinating constructions, or the


184 Mahamane L. Abdoulaye

This section has shown that the comitative and the coordinating constructions contrast in their syntactic properties (topicalization, focalization, topicality, and control of following pronouns) and in their semantic properties (degree of involvement in the verb’s action and spatial or temporal contiguity). In the following section, we will see that the inclusory constructions can be divided into those that are fundamentally comitative structures and those that are fundamentally coordination structures.


Inclusory comitative structures and inclusory coordinations

As seen in the introduction, Hausa, apparently like many other languages (cf. Liu 1991, Haspelmath, to appear), can use a plural pronoun in the verbal complex with a comitative NP later in the sentence which is normally understood as a co-actor in the verb’s action. Or else, a simple coordination may involve a plural pronoun and a coordinate NP, and yet the whole is understood as involving only two referents.

inclusory constructions, is not important. Some examples are given next (all sentences adapted from Imam 1980: 19, 148, 331): (i) a. Don mùtûm, Allàh yà kaawoo raana˜r dà zaa mù gàmu dà shii. for man Allah 3sg.m.subv bring day that fut 1pl meet with 3sg.m ‘As for man, may God bring the day when I meet him.’ (The sentence, taken from a tale, is spoken by a young lion.) b. Tankòo kùwa dà mu-kèe wàasaa dà shii! Tanko indeed that 1pl-rc play with 3sg.m ‘Well, that Tanko, the one with whom I play!’ c. Ai bâa ni dà daamaa tun dà mukà yi àlMawà˜rii dà ita. well neg.cont 1sg have way since 1pl-rp do agreement with 3sg.f ‘Well, I will not have that possibility, since I have made an agreement with her.’ d. Dâa su-nàa cîi tàare dà wani maMwàbci-n-shì. before 3pl-cont eat together with some neighbor-of-3sg.m ‘Previously, he was eating with some neighbor of his.’ e. Sun cùuci juunaa dà Bàlki. 3pl.pfv cheat each.other with Balki ‘They cheated one another, Balki and him.’ In sentence (i-a), the verb gàmu ‘meet’ equally involves both participants in the same type of activity, regardless of whether a comitative or coordinating construction is used (zân gàmu dà shii ‘I will meet with him’, zaa mù gàmu, nii dà shii ‘we will meet, he and me’). In (i-b), with playing events that involve two people, one can normally assume an equal level of involvement from the participants. In (i-c), the Hausa àlMawà˜rii primarily means ‘promise’ and should involve different types of activities for the two participants. However, here it has the meaning of ‘agreement’ and is hence reciprocal (for more on association and dissociation verbs see Abdoulaye 1996: 127–128). In (i-d), too, with a reciprocal construction, the context implies an equal involvement of the participants in the same type of activity.

Comitative, coordinating, and inclusory constructions in Hausa

The two constructions are illustrated again in the following (cf. Schwartz 1989b: 30, 52, repeated from (2) above): (27) a.

Mun jee kàasuwaa dà Abdù. 1pl.pfv go market with Abdu ‘We went to the market, Abdu and me.’ Or: ‘We went to the market with Abdu.’ b. Muu dà shii mun jee kàasuwaa. 1pl and 3sg.m 1pl.pfv go market ‘He and me, we went to the market.’ Or: ‘He and us, we went to the market.’

In (27a) the verb jee ‘go’ is preceded by an obligatory subject pronoun carrying person, gender, number and tense/aspect information. The pronoun here is 1st person plural, and yet, as the translations show, a dual reading of the participants is possible. Similarly, in (27b), a 1st person plural pronoun is coordinated with a third person masculine singular pronoun, and again a dual interpretation is possible. Though the constructions involving a 1st person pronoun are more frequent, constructions with a second or third person pronoun can appear as well (cf. examples (28a) and (28d) below). Data like those in (27) have been extensively discussed, in particular in a series of papers by Schwartz (Schwartz 1988a, 1988b, 1989a, 1989b, 1991, etc.) and in Newman (2000: 136). The constructions have been variously labelled in Hausa literature as “plural pronoun constructions”, “asymmetric coordination”, or “thematic coordination” (see Schwartz 1989b, for example). In this section, as the presentation of the examples already suggests, I will consider data such as (27a) to be a type of comitative construction but where the subject pronoun anticipates the comitative NP, while in (27b) we have a type of conjunction, where the first pronoun also anticipates the second conjunct pronoun. To refer to the general phenomenon, I will adopt the term “inclusory constructions”, thus following Haspelmath (to appear), among others. First I will discuss the comitative structures and then the coordinations. It should be noted that personal and dialectal factors may intervene in speaker judgements of the examples (cf. Schwartz 1989b: 53). The section will be based on the already published data from Standard Hausa (mainly from Schwartz 1989b and Newman 2000) and my own collected data (western/Niger dialects). 5.1 The inclusory comitative constructions As explained in Schwartz (1989b) and Newman (2000: 136), the inclusory comitative construction may concern a subject pronoun, as illustrated in (27) above, a direct object pronoun, an applied object, and a possessive pronoun (see the examples in (29) below). This section will try to show that all these structures have


186 Mahamane L. Abdoulaye

the syntactic, semantic, and pragmatic properties of regular comitative constructions. Some of the data discussed in the literature are clearly comitative in nature judging only from syntactic evidence. This is illustrated in the following (giving normally only the dual sense; cf. also Schwartz 1989b: 37ff): (28) a.

Sun mutù tàare dà àbooki-n-sà. 3pl.pfv die together with friend-of-3sg.m ‘They died, he and his friend.’ b. Dà Bàlaa kàm, mun jee kàasuwaa. with Bala indeed 1pl.pfv go market ‘As for Bala, we did indeed go to the market, he and me.’ c. Dà Bàlaa nèe mu-kà jee kàasuwaa. with Bala cop 1pl-pfv go market ‘It is with Bala that we two went to the market.’ d. Dà wàanee nèe Làamii ta baa kù àbinci? with who cop Lami 3sg.f.rp give 2pl food ‘With whom did Lami served you two food?’

In sentence (28a) the comitative nature of the sentence is clearly marked by the adverb tàare ‘together’. In (28b–d) the dà-NP is topicalized, focused and questioned, respectively. As seen in § 4.1, these properties also characterize the regular comitative construction, but not the coordination construction. One also finds examples of inclusory structures involving a weak direct-object or possessive pronoun which may be immediately followed by a dà-NP. These structures, too, are not conjunctions, but are comitative constructions, parallel to those discussed in (12b) above. Some such sentences are given next (also see Schwartz 1989b: 35, Newman 2000: 136, Kraft 1970: 101): (29) a.

Abdù yaa gan mù (jiyà) dà Bintà. Abdu 3sg.m.pfv see 1pl (yesterday with Binta ‘Abdu saw us (yesterday), Binta and me.’ b. Zaman nàn naa-mù (tàare) dà kee bâa shi living this that.of-1pl (together with 2sg.f neg.cont 3sg.m dà daa1ii. have pleasure ‘This living of ours, i.e., of me (together) with you, is unpleasant.’ c. Bàlaa ya kaawoo ma-nà àbinci dà Muusaa. Bala 3sg.m.rp bring to-1pl food with Musa ‘Bala brought us food, Musa and me’

In sentences (29a–b) the weak direct object or possessive pronouns mù ‘1pl’ can be contiguous or separated from the following dà-NP, which shows that the sentences are comitative. In sentence (29c) the dative pronoun -nà ‘1pl’ is separated from the

Comitative, coordinating, and inclusory constructions in Hausa 187

dà-NP, and this sentence too is comitative. The dative pronoun can be immediately followed by the dà-phrase only if the patient àbinci ‘food’ is moved, as in whquestions for example: mìi Bàlaa ya kaawoo manà Ø dà Muusaa? ‘What did Bala bring us, Musa and me?’. In spoken Hausa, speakers can most naturally start a sentence with a singular topic NP but continue with a plural subject pronoun referring to the topic NP and a dà-NP later in the sentence, or — less spontaneously — an understood participant. Typical contexts for such sentences are the replies to queries about someone, as in: ìnaa Abdù? ‘where is Abdu?’; Abdù sun tàfi maka˜rantaa dà Bintà ‘Abdu (they) are gone to school with Binta’. In these inclusory constructions, the topic NP must be referred to with a pronoun in a following clause, and it cannot be repeated. The NP object of dà can be repeated. This is illustrated in (30):5 (30) Abdù sun tàfi maka˜rantaa dà Bàlki an bâa *Abdù/Bàlkî Abdu 3pl.pfv go school with Balki imps.pfv give Abdu/Balki àlloo. board ‘Abdu went to school with Balki, and a board was given to *Abdu/Balki.’ This sentence shows that Abdù is the sole topic and cannot satisfactorily be repeated, in contrast to the dà-NP Bàlki. We have seen in § 4.1, in the discussion of (21), that these same patterns of topicality characterize the regular comitative constructions. The inclusory comitative constructions involving direct objects allow a separate control of a pronoun in a following clause, as seen in (31): (31) Abdù, naa ganee sù (tàare) dà àbikkiyà-˜r-shì ya-nàa / Abdu 1sg.pfv see 3pl (together with friend-of-3sg.m 3sg.m-cont/ ta-nàa /su-nàa sàye-n goo˜rò. 3sg.f-cont/3pl-cont buying-of kolanuts ‘As for Abdu, I saw him with his friend, he was/she was/they were buying kolanuts’. In this sentence the plural direct object pronoun refers to the topic NP Abdù and

5.The constructions exemplified in (30, 32), which apparently violate understood rules of agreement, are particularly useful for reporting someone’s being engaged in typical group activities. For example, if Speaker A says ìnaa Abdù? ‘where is Abdu?’, then Abdu is the interest of Speaker A and should be the topic of Speaker B’s answer. But for some typical group activities (say, children’s hunting or children’s swimming in the river), the plurality of the participants tends to be expressed. The solution for Speaker B is then to retain Abdu as the topic NP, but use a plural pronoun to signify group activity. Sometimes, a pause intervenes between the topic NP and the rest of the clause, especially when the co-actors are not expressed, as in: ìnaa Abdù? ‘Where is Abdu?’ Abdù, sun tàfi rwàafii wankaa ‘ah Abdu! they have gone to the river for swimming’.

188 Mahamane L. Abdoulaye

the following dà-NP. As indicated, the construction allows separate or joint control of the following pronoun. This same pattern of coreference also characterizes the regular comitative constructions (see the discussion of (23) in § 4.1 above). Semantically the inclusory comitative constructions also resemble the regular comitative constructions in requiring contiguity between the two participants and in not implying the same level of involvement for the two participants. This is illustrated next: (32) a.

Feemì sun tàfi Ìkko dà Bàlki. Femi 3pl.pfv go Lagos with Bàlki ‘Femi went to Lagos with Bàlki.’ b. Feemìi sun tàfi Ìkko dà ùba-n-shìi. Femi 3pl.pfv go Lagos with father-of-3sg.m ‘Femii went to Lagos with hisi father.’

In sentence (32a), Femi and Balki are understood to necessarily travel together. However, only the context can tell who is in control of the situation. In sentence (32b), too, Femi and his father are understood to travel together to Lagos. However, because of the slanted context, the father is understood to be the controller of the action, in spite of the favorable syntactic position of Feemì as topic. In conclusion, I consider the inclusory sentences reviewed in this section to be a subtype of comitative constructions, judging from the morphosyntactic and semantic evidence. 5.2 The inclusory coordination constructions The constructions discussed in this subsection, and which I label inclusory coordinations, are maybe less prominent in Hausa than the inclusory comitative constructions. For example, Schwartz (1989b: 52) gives only one illustration, reproduced in the following (other potential meanings are left out): (33) Muu dà shii mun jee kàasuwaa. 1pl and 3sg.m 1pl.pfv go market ‘He and I went to the market.’ The relevant part in this example is the sequence muu dà shii meaning ‘he and me’, where the first pronoun is plural in anticipation of the second pronoun. The sequence can be analyzed as a coordination with two conjuncts, in a typical structure for coordinations, i.e., “NP + dà + NP”. Following the conjoined pronouns, we find a subject pronoun in the clause proper, which is naturally plural, referring to the two pronouns. Other possible inclusory coordinations are: muu dà Muusaa ‘Musa and I’ (cf. Newman 2000: 136), kuu dà Saanii ‘you (sg) and Sani’, etc. The inclusory coordination can appear at the beginning of the sentence, in a

Comitative, coordinating, and inclusory constructions in Hausa 189

topicalization or subject position, as seen in (33). It can appear in direct object position, in indirect object position, and at the end of the sentence, in a rightdislocation position. It can also appear after psychological or speech-act main predicates. All these contexts are illustrated in the following: (34) a.

Wannàn ha˜rakàa taa shàafi kuu dà ita. this matter 3sg.f.pfv concern 2pl and 3sg.f ‘This matter concerns you and her’ b. An aikoo mà muu dà shii goo˜rò. imps.pfv send to 1pl and 3sg.m kolanuts ‘One sent us kolanuts, he and me.’ c. An aikoo ma-mù goo˜rò, muu dà shii. imps.pfv send to-1pl kolanuts 1pl and 3sg.m ‘One sent us kolanuts, he and me.’ d. Màigidaa yaa cêe muu dà kai mu-nàa tàfiyàa Lòme. boss 3sg.m.pfv say 1pl and 2sg.m 1pl-cont going Lome ‘The boss said that you and I are going to Lome.’

In (34a), both coordinands are direct objects and independent pronouns, and for this reason, the pre-nominal form of the verb is used (i.e, shàafi ‘concern’ instead of the pre-pronominal form shàafee). Similarly, in (34b), the independent pronouns appear in the applied object position. Here, too, the pre-nominal form of the applicative marker mà (with a low tone) is used (cf. the pre-pronominal form ma-, as in (34c): an aikoo mamù… ‘one sent us…’). In (34c) the coordination appears in a right-dislocation construction, following the applied object ((ma-)mù) and the direct object (goo˜rò ‘kolanuts’). Finally, in (34d) it appears as subject/topic of a clause which is complement to a speech verb. It should be noted that the grammaticality judgements about the inclusory coordinations are less consistent than those for inclusory comitative constructions. There is considerable evidence that the inclusory coordinations in (33–34) are indeed coordinations. First, in all of the sentences, it is not possible to insert the adverb tàare ‘together’. Doing so would either result in complete ungrammaticality (cf. for (34b) *An aikoo mà muu (tàare) dà shii goo˜rò) or else, the dual meaning would be destroyed (cf. for (34a) ?wannàn ha˜rakàa taa shàafi kuu (tàare) dà ita ‘this matter concerns you-pl and her’). Syntactically the inclusory coordinations behave like ordinary coordinations in not allowing the preposing of a single conjunct, the topicalization of a single conjunct, or the separate control of a pronoun in a following clause. This is illustrated next: (35) a. *Dà kai fa, muu Ø zâa mu Lòme. and 2sg.m indeed 1pl going 1pl Lome ‘*And you indeed we (= I) will go to Lome.’

190 Mahamane L. Abdoulaye

b. *Dà ita nèe wannàn ha˜rakàa ta shàafi kuu Ø. and 3sg.f cop this matter 3sg.f.rp concern 2pl ‘*It is her that this matter concerns you and (her)’ (36) a.

Abdù, suu dà Bàlki sun tàfi maka˜rantaa Abdu 3pl and Balki 3pl.pfv go school an bâa Abdù/Bàlki àlloo. imps.pfv give Abdu/Balki board ‘As for Abdu, he and Balki went to school and a board was given to Abdu/Balki.’ b. Naa ga kuu dà Nà’oomì *ka-nàa/*ta-nàa/ku-nàa 1sg.pfv see 2pl and Naomi 2sg.m-cont/3sg.f-cont/2pl-cont sàye-n goo˜rò. buying-of kolanuts ‘I saw you and Naomi, *you-sg were/*she was/you-pl were buying kolanuts.’

In sentences (35a–b) one conjunct cannot be preposed in topicalization or focalization, respectively. In (36a), despite the fact that the NP Abdù figures prominently in front of the sentence, it can be repeated in a following sentence, just like the dà-NP Bàlki. In this respect, the construction behaves like a normal coordination in not allowing one NP to behave as the sole sentence topic (cf. discussion of (22) in §4.1). In (36b) the conjoined direct object NPs cannot separately control a following pronoun, again as in regular coordination (see discussion following the example (23b)). Finally, at the semantic level the inclusory coordinations imply the same degree of involvement for the two conjuncts in the verb’s action, and they do not imply temporal or spatial contiguity for the two conjuncts. This is illustrated in the following: (37) Feemì, suu dà ùbanshì duk sun tàfi kàasuwaa, Femi 3pl and father-of-3sg.m all 3pl.pfv go market shii Feemì-n maa tun dà saafe. 3sg.m Femi-def indeed since on morning ‘Femi and his father have all gone to the market, Femi in fact left very early.’ This sentence has an NP topic, Feemì, followed by an inclusory coordination. Despite the topicality of one NP, both are understood to be co-actors. And as in ordinary coordination, the two participants need not be involved in the verb’s action together or at the same time, as indicated in the example. Considering the properties selected (preposing of the dà-phrase, topicalization, control of following pronouns, equal involvement and contiguity), the inclusory coordination constructions are indeed a variety of basic NP conjunction coordinations. There are of course many other tests one can devise to show the relations

Comitative, coordinating, and inclusory constructions in Hausa

between the ordinary and the inclusory comitative or coordinating constructions. One potential area can be scope phenomena. For example, the focus particle sai ‘only’ has scope over the subject, and not the comitative NP, as in: sai Feemì ya yi wàasaa dà Abdù, which means: ‘only Femi played with Abdu’, and not: ‘only Femi and Abdu played’. But sai ‘only’ has scope over both conjunct NPs in a basic coordination (cf. sai Feemì dà Abdù sukà yi wàasaa ‘only Femi and Abdu played’), or in an inclusory coordination (cf. kinàa zuwàa Çan Gàawoo? ‘do you go to Dan Gawo?’, sai muu dà kai! ‘only if you and me!’). Thus, in view of the facts adduced in this section, it is not necessary to set up a special thematic/asymmetric coordination structure for Hausa, which would stand in contrast to the comitative and the regular coordinating constructions. Instead, we basically have a comitative and a coordinating construction, both of which can have a dual reading with an anticipatory plural pronoun. The mechanism by which this interpretation happens may well be the unification of two sets where the elements common to the two are counted only once in the final set (cf. Haspelmath, to appear). It must be assumed that this general process maps onto both the comitative and the coordinating constructions, which remain fundamentally the same. Nonetheless, properties particular to the inclusory constructions must not be excluded. For example, inclusory constructions involving non-human referents are bad in my judgment (ìnaa teebù˜r? ‘where is the table?’; *suu dà kujèeraa an saidàa su ‘they (it) and the chair were sold’). The inclusory constructions may also be somewhat peripheral in Hausa syntax. For example, negation seems to destroy the inclusory interpretation, as seen in: munàa zuwàa dà kai Ma˜raa1i ‘we are going to Maradi, we and you/you and me’, bâa mu zuwàa dà kai Ma˜raa1i, which only means ‘we are going to Maradi, but not with you’; an baa mù goo˜rò dà kai ‘one gave us kolanuts, we and you/you and me’, bà à baa mù goo˜rò dà kai ba ‘one did not give us kolanuts with you’, or: ‘one gave us kolanuts, but not with you’, etc. Nonetheless, the inclusory constructions are frequent in speech and I have heard them seamlessly used by a 4.8-year-old child. According to Schuh (1998: 277) most, if not all, Chadic languages use the inclusory constructions and Miya speakers actually prefer them to the regular comitative or coordinating constructions (Schuh 1998: 305). In Hausa, too, this is likely, but it remains to be established.

6. Conclusion This chapter has offered an overview of two prominent usages of the Hausa particle dà, as the comitative preposition and as the NP coordinator. I have suggested some ways in which an existential predicate dà ‘there is/are’ developed into the comitative preposition dà, which in turn gave rise to the coordinator. Despite their historical relationships and some situations where they are equivalent, the basic comitative


192 Mahamane L. Abdoulaye

and conjunction constructions have different semantic and morphosyntactic properties. Hausa, like many other languages, in particular African languages, has inclusory constructions where a plural pronoun with a dual meaning anticipates the existence of an NP object of dà. These constructions have been conceived of as a special brand of thematic or asymmetric coordination in previous Hausa studies. We have seen that these dual constructions can in fact be divided into inclusory comitative and inclusory coordinating constructions, which can respectively be assimilated to the regular comitative and the regular coordinating constructions.

Abbreviations Primary data in this chapter are from the Katsinanci dialect and Standard Hausa. The transcription follows the Hausa standard orthography with some changes. Long vowels are represented as double letters, low tone as grave accent, and falling tone as circumflex accent. High tone is unmarked. The symbol ‘˜r’ represents an alveolar trill distinct from a flap ‘r’. Written ‘f ’ is pronounced [h] (or [hw] before [a]) in Katsinanci and other western dialects. cont cop imps rc

continuative copula impersonal relative continuative

rp sp subv

relative perfective simple past subjunctive

References Abdoulaye, Mahamane L. 1996. “Efferential ‘verb + da’ constructions in Hausa.” Journal of African Languages and Linguistics 17: 113–151. Abraham, Roy Clive. 1962. Dictionary of the Hausa language. [2nd edition]. London: London University Press. Caron, Bernard B. 1991. Le haoussa de l’Ader. Berlin: Dietrich Reimer Verlag. Crazzolara, J. P. 1960. A study of the Logbara (Ma’di) language: Grammar and vocabulary. London: Oxford University Press. Foley, William A. and Robert D. Van Valin, Jr. 1984. Functional syntax and universal grammar. Cambridge: Cambridge University Press. Haspelmath, Martin. (to appear). “Coordination”. In Timothy Shopen (ed.), Language typology and syntactic description. [2nd edition]. Cambridge: Cambridge University Press. Heath, Jeffrey. 1998. A grammar of Koyra Chiini: The Songhay of Timbuktu [Mouton grammar library 19]. Berlin/New York: Mouton de Gruyter. Heath, Jeffrey. 1999. A grammar of Koyraboro (Koroboro) Senni: The Songhay of Gao, Mali [Westafrikanische Studien Series 19]. Köln: Rüdiger Köppe. Heine, Bernd and Mechthild Reh. 1984. Grammaticalization and reanalysis in African languages. Hamburg: Helmut Buske Verlag. Hutchison, John P. 1980. “The Kanuri associative postposition: A case for subordination.” Studies in African Linguistics 11(3): 321–351.

Comitative, coordinating, and inclusory constructions in Hausa 193

Hutchison, John P. 1981. The Kanuri language: A reference grammar. Madison: African Studies Program, University of Wisconsin. Imam, Abubakar. 1980. Magana jari ce [Stories are valuable]. Zaria: Northern Nigerian Publishing Company (First edition: 1939). Jolly, Julia A. 1993. “Preposition assignment in English.” In Robert D. Van Valin, Jr. (ed.), Advances in Role and Reference Grammar. Amsterdam: John Benjamins, 275–310. Kraft, Charles. 1970. “Hausa sai and dà — a couple of overworked particles.” Journal of African Languages 9: 92–109. Liu, Mei-Chun. 1991. “Nominal conjunction and associative marker in Chadic languages.” In Herrmann Jungraithmayr and Henri Tourneux (eds.), Etudes tchadiques: La phrase complexe. Paris: Geuthner, 85–102. Mithun, Marianne. 1988. “The grammaticization of coordination.” In John Haiman and Sandra A. Thompson (eds.), Clause combining in grammar and discourse. Amsterdam: John Benjamins, 331–359. Newman, Paul. 2000. The Hausa language: An encyclopedic reference grammar. New Haven: Yale University Press. Newman, Paul and Roxana M. Newman. 1977. Modern Hausa-English dictionary. Ibadan: University Press. NNPC. 1968. Labaru na da da na yanzu [Ancient and modern stories]. Zaria: Northern Nigeria Publishing Company. Payne, Thomas E. 1997. Describing morphosyntax: A guide for field linguists. Cambridge: Cambridge University Press. Schuh, Russell G. 1998. A grammar of Miya [University of California Publications in Linguistics 130]. Berkeley: University of California Press. Schwartz, Linda. 1988a. “Asymmetric feature distribution in pronominal ‘coordination’.” In Michael Barlow and Charles Ferguson (eds.), Agreement in natural language. Stanford: CSLI, 237–250. Schwartz, Linda. 1988b. “Conditions on verb-coded coordinations.” In Michael Hammond et al. (eds.), Studies in syntactic typology. Amsterdam: John Benjamins, 53–73. Schwartz, Linda. 1989a. “Asymmetrical syntax and symmetrical morphology in African languages.” In Paul Newman and Robert Botne (eds.), Current Approaches to African Linguistics, Volume 5. Dordrecht: Foris Publications, 21–33. Schwartz, Linda. 1989b. “Thematic linking in Hausa asymmetric coordination.” Studies in African Linguistics 20(1): 29–62. Schwartz, Linda. 1991. “Category asymmetries in Hausa asymmetric coordination.” BLS 17S: 222–230. Stassen, Leon. 2000. “AND-languages and WITH-languages.” Linguistic Typology 4: 1–54. Wolff, H. Ekkehard. 1993. Referenzgrammatik des Hausa. Münster & Hamburg: LIT Verlag.


Chapter 8

Coordinating constructions in Daghestanian languages* Helma van den Berg Max-Planck-Institut für evolutionäre Anthropologie, Leipzig

1. Introduction 2. Dargi 2.1 Conjunctive coordination 2.2 Disjunctive coordination 3. Conjunctive coordination in other Daghestanian languages 3.1 Bisyndetic conjunctive coordination 3.2 Monosyndetic conjunctive coordination 3.3 Languages with both monosyndetic and bisyndetic conjunction 3.4 Other uses of the coordinator 4. Overview and discussion of conjunctive coordination 4.1 Variation of coordinators 4.2 Grammaticalization 4.3 Diffusion of wa 5. Dargi revisited: ‘and’ and ‘with’



This article aims at giving an overview of the unity and diversity in coordinating constructions in Daghestanian languages. The Daghestanian languages belong to the East Caucasian (or Nakh-Daghestanian) language family, which can be subgrouped as follows (see Nikolayev & Starostin 1994, Schulze 1997: 11):

*Most of the examples in this article come from the literature. However, some additional elicitation was carried out with Hadizhat Omarova (Derbent) and Magomed-Shapi Isaev (Makhachkala) for Dargi, and Magomed Magomedov (Makhachkala) for Avar. I am grateful for their helpfulness in this matter. Furthermore, discussions with Sveta Makhmudova, Uzlipat Gasanova, Bukar Talibov (Makhachkala), Wolfgang Schulze and Lars Johanson are gratefully acknowledged. Full responsibility for the analyses and hypotheses proposed remains with the author.

198 Helma van den Berg

– – – – – –

Nakh (Chechen, Ingush, Tsova-Tush) Avar-Andic (Avar and Andic, e.g. Godoberi, Bagvalal) Tsezic (e.g. Hunzib, Bezhta) Dargi Lak Lezgic (e.g. Lezgian, Tabasaran, Agul, Tsakhur, Rutul, Archi)

Typical features of Daghestanian languages are SOV word order both in the clause and the noun phrase, i.e. genitive-noun, adjective-noun, postpositions. Most Daghestanian languages, with the exception of Lezgian, Agul and Udi, have a gender system with 2 to 5 genders for nouns. The usual targets of gender agreement are verbs, adjectives, and demonstratives, less commonly adverbs and numerals; agreement is indicated mainly by affixes, sometimes also by apophony. Daghestanian languages have rich suffixation on nouns and verbs, a well-developed case inventory, and ergative/absolutive case marking. For an overview of the East Caucasian language family, see van den Berg (forthc.). Eight Daghestanian languages are written languages, viz. Avar, Dargi, Lak, Lezgian, Tabasaran, Agul, Tsakhur, and Rutul. As the information on coordination in the literature on Daghestanian languages is usually rather limited, the article concentrates on languages from which a sufficient amount of data is available. § 2 gives a survey of coordinating constructions in Dargi. Conjunctive coordination in other Daghestanian languages is treated in § 3, while § 4 discusses the data from a general perspective, pointing out implications for the historical development of Daghestanian coordinating constructions. Beside comitative case marking, Dargi, and probably also other Daghestanian languages, has a coordinating construction involving a reflexive pronoun which denotes a comitative situation. This conjunctivecomitative construction has not been described before and will be discussed in § 5.



Dargi is one of the written languages of Daghestan. It has wide dialect variation: at least 13 dialect groups can be distinguished. The mutual intelligibility between dialects is low, to the extent that some dialects are regarded as distinct languages, e.g. Kubachi, Chirag, Megeb. The written standard is based on the dialect of Akusha, with considerable influence from the Urakhi dialect. The data in this article come from Akusha Dargi.1 It has coordinating constructions of four types: conjunction

1.Most Dargi examples in this article come from the texts in van den Berg (2001): text and line numbers are provided with each example. The examples without text and line numbers were elicited.

Coordinating constructions in Daghestanian languages 199

with the cl*tic -ra and the borrowed coordinator wa, disjunction with the coordinators ya, ya-ra and aAi, adversative coordination with -gu and borrowed amma, and causal coordination with -nu. 2.1 Conjunctive coordination 2.1.1 Noun phrase conjunction The regular way of forming Dargi noun phrase conjunction is bisyndetic coordination with the cl*tics -ra…-ra. The coordinators follow the case markers.2 (1) murul-ra xˆunul-ra quli ˇcar·bÒi‘-ubli sa·bÒi husband(abs)-and wife(abs)-and home return:hpl-ger be:hpl ‘The man and his wife returned home.’


(2) il.a-la buruš-ra yurgˇan-ra ’änala-ra kas-ili this-gen mattress(abs)-and blanket(abs)-and pillow(abs)-and take-ger sa·rÒi (30:5) be:pl ‘(They) took his mattress, blanket and pillow.’ (3) dudeš.li-ra neš.li-ra emAe b-abg-ili father(erg)-and mother(erg)-and donkey(abs) n-harness-ger sa·bÒi be:hpl ‘Father and mother harnessed the donkey.’


(4) han·bÒirk-ur-u, … Ae-la-ra di-la-ra džal? (18:23) remember:n-fut.3-int you-gen-and me-gen-and quarrel(abs) ‘Do you remember your and my argument?’ The particle -ra also has the additive focussing meaning ‘also, too’ or the scalar focussing meaning ‘even’, e.g.: (5) qum·maÒrt-id b-arx yagˇlaw-ra kas-es forget:proh-fut.2 n-with frying.pan(abs)-and take-inf ‘Don’t forget to take the frying pan with you as well.’


2.Glosses are given for the essential parts of the examples. In most cases, glosses are mine, except with examples from Haspelmath (1993), Kibrik (ed.) (1996), (1999) and (2001). Their transcription has been preserved as well. The oblique case endings are usually added to a stem extension following the noun. Besides the hyphen for morpheme boundaries, = is used for a derivational morpheme boundary, + for a compound, · Ò for infixes; a dot in the morpheme analysis indicates the stem extension with nouns and zero infixes in verbs.

200 Helma van den Berg

(6) xˆunul+adam-t.a-ni duxu-ti-ra meAur+b-ar-es b-ir-ar (18:2) woman+person-pl-erg clever-pl-and mad+hpl-do-inf hpl-can-fut.3 ‘Women can even drive clever people mad.’ There are a few instances in the text corpus of noun phrase conjunction with the borrowed coordinator wa, which is more common for clause conjunction, see §2.1.2. (7) eger nu-ni duxu-ti wa ’äq’lucˇe-b-ti adam-ti if me-erg clever-pl and wise-hpl-pl person-pl(abs) b-i‘-ni ka·bÒiz=aq-asli, … hpl-be-masd(abs) pose:hpl=caus-cond.1 ‘If I prove that these people are clever and wise…’


(8) urš dudeš.li-zi han·bÒuš-ili sa.y… mu‘ˇila wa sune-la boy(erg) father-ill remind:n-ger be:m dream-cont and self-gen qä.li-cˇila (17:21) vow-cont ‘The boy reminded his father of the dream and his vow.’ wa sen Au ila kat-ursi-ri-w? (9) ˇ who(erg) and why you(abs) here put:m-paprt-2-int ‘Who put you into this bag and why?’


(10) nuša ka·dÒik-ni-la wa ˇculaq·dÒi‘-ni-la ’äyib we(abs) fall:pl-masd-gen and be.cripple:pl-masd-gen guilt(abs) šayt’ay-cˇi-b sa·bÒi (14:9) devil-sup-n be:n ‘It is certainly the devil’s fault that we fell down and hurt ourselves.’ Dargi and other Daghestanian languages use zero-marked NP-coordination only with a restricted function to denote conceptual units (cf. Mithun 1988: 332). In Dargi, these constructions are usually described as compounds, e.g.: neš+dudeš ‘parents (mother+father)’, šalbar+Aewa ‘clothes (trousers+shirt)’, berAi+dugi ‘day and night’, dam+däd ‘music (drum+flute)’, ˇcarx+bek’ ‘complexion (body+head)’, Aäl+t’abi’ät ‘behaviour (situation+ behaviour)’. 2.1.2 Clause conjunction The main type of Dargi clause conjunction involves the monosyndetic medial use of the particle wa ‘and’. It was borrowed from Arabic, and also occurs in the Turkic languages of the area (Kumyk and Azerbaijani). The diffusion of wa in Daghestanian languages is discussed in § 4.3. (11) … tupang ih-ubli sa.y wa yuldaš ka·wÒš-ili sa.y rifle(abs) hit-ger be:m and friend(abs) kill:m-ger be:m ‘… he shot and killed his friend.’


Coordinating constructions in Daghestanian languages 201

(12) šahar-la adam-t.a-ni il-di meAur-t.a-zi b-ey’-un wa town-gen person-pl-erg this-pl mad-pl-ill hpl-consider-aor(3) and xu-ri kus·dÒar-ib (27:22) dog-pl(abs) provoke:pl-aor(3) ‘The people of the town considered them fools and set the dogs on them.’ The particle wa also conjoins subordinate clauses, like infinitives, masdars, and gerunds, e.g.: (13) muruy xˆ unuy-cˇi qar·bÒar-ili sa.y džawab husband(erg) woman-sup order:n-ger be:m [answer(abs) b-ed=aq-es wa quli t’alab+w-ar=aq-es (6:7) n-give=caus-inf] and [home invite+m-do=caus-inf] ‘The husband told his wife to answer and to invite (the man) home.’ (14) eger [di-la k’äl’ä.li-zi-w ca abdal le-w-ni if [me-gen castle-ill-m one stupid present-m-masd(abs)] wa [il-ra Au w-i‘-ni nu-ni and [this-and you(abs) m-be-masd(abs)] me-erg ka·bÒiz=aq-asli, … pose:n=caus-cond.1 ‘If I prove there is one fool in my castle and that is you, …’


At the same time there are a few cases in the corpus in which a single coordinator -ra coordinates two finite clauses; in that case it follows the first NP of the last conjunct clause, e.g.: (15) di-la cula idz-es b-äA-ib-Aeli il nu-ni me-gen tooth(abs) hurt-inf n-begin-aor-when this me-erg a·bÒit’=aq-un-ra idzala-ra Gay·bÒik-ib (31:15) remove:n=caus-aor-1 disease(abs)-and stop:n-aor(3) ‘When my tooth started to ache, I had it removed and the pain stopped.’ (16) Malla.Nasradin huni b-uc-ili hay.i‘-ubli sa.y Malla.Nasradin(abs) road(abs) n-catch-ger be:m žah-ti-ra … hay·bÒi‘-ubli sa·bÒi (22:17–18) young-pl-and be:hpl ‘Malla Nasradin went on his way; and the youngsters continued their way as well, …’ The particle -ra also frequently occurs after the absolutive NP of a subordinate clause. Its meaning here is not fully understood; the use and meaning of similar constructions in other Daghestanian languages are treated in § 3.4. E.g.:

202 Helma van den Berg

(17) il [as-ili Gaz-ra talqay-s ar·bÒux-ili this [take-ger goose(abs)-and] [chief-dat take.away:n-ger] ar.äq’-i sa.y leave:m-ger be:m ‘He took a goose and left to bring it to the chief.’


(18) [Aed Guruš-la baha-ra b-ed-ili adam [you:dat rouble-gen price(abs)-and n-give-ger] person(abs) w-arˆx-i-ra-gu sen urcˇi ˇcar·AeÒ·bÒar-ibsi-w? (2:9) m-send-aor-1-but why horse(abs) return:neg:n-paprt-int ‘I sent a man, having given (him) the price of a rouble for you, but why didn’t you return the horse?’ There are also clauses in which the first or the last NP is followed by -ra, which seems to structure a larger episode of a text, as in (19)–(20) below. This is especially frequent with words like axir ‘end’, ’ur ‘then’, ilAeli ‘then’, hacˇam ‘once’, as in (21)–(23). (19) Au-ni-ra b-al-ad adam Aäri.irˆx-uAeli Aäˇi you-erg-and n-know-fut.2 person(abs) bury:m-when grave-sup sadaq’a b-ix-usi ’ädat le·bÒsi sa·bÒi (5:2) alms(abs) n-bring-paprs custom(abs) present:n be:n ‘And you also know that when we bury someone, there is the custom of bringing alms to the grave.’ (20) u‘-ili sa.y ca ’äq’lucˇe-w-si Gurban b-ik’-usi m:be-ger be:m one wise-m-adj Kurban(abs) hpl-say-paprs adam il.i-cˇi ’äq’lu b-ur=aq-es person(abs) this-sup advice(abs) n-say=caus-inf b-aš-uti-ri šan-ti-ra (19:1–2) hpl-come-paprs-pret villager-pl(abs)-and ‘There was a wise man called Kurban. And the villagers came to him to ask for advice.’ (21) hacˇam-ra il.a-la qu-cˇi-b dawlacˇe-w-si-la urcˇi once-and this-gen field-sup-n rich-m-adj-gen horse(abs) dak’u·bÒi‘-ub appear:n-aor(3) ‘And once a horse of a rich man appeared in his field.’


(22) ’ur-ra ’ämal b-u‘-i ca w-ik’-uli sa.y then-and patience(abs) n-disappear-ger one m-say-ger be:m yuldaš-un.a-zi (8:6) friend-pl-ill ‘And then they lost patience and one of them said to his friends:…’

Coordinating constructions in Daghestanian languages 203

(23) axir-ra iš-d.a-ni q’as·bÒar-ili sa·bÒi q’adi-zi end(abs)-and this-pl-erg decide:n-ger be:hpl cadi-ill xˆ ar·bÒa‘-es ˇ b-alx-es ’ä’nisi-l emAe (25:7) ask:n-inf who(erg) n-feed-inf necessary-indir donkey(abs) ‘And eventually, they decided to ask the cadi who should feed the donkey.’ 2.2 Disjunctive coordination 2.2.1 Noun phrase disjunction Dargi employs three strategies with regard to NP-disjunction. In declarative sentences there is medial monosyndesis with the particle ya ‘or’, borrowed from Persian, followed by the conjunctive coordinator -ra. Clause disjunction, as in example (28), is preferred, though. (24) b-išt’a-ti Aäžta·bÒirq’-uli sa·bÒi anq’-la-b ya-ra quli-b hpl-small-pl play:hpl-ger be:hpl garden-loc-hpl or-and home-hpl ‘The children play in the garden or at home.’ In negated sentences the particle ya is used bisyndetically, e.g.: (25) nu-ni umˆxu sune-laˇi-b b-arg-i-ra, amma me-erg key(abs) self-gen place-sup-n n-find-aor-1 but ya pulaw, ya ’är’ä Ae-d-arg-i-ra (22:46) or pilaf(abs) or hen(abs) neg-pl-find-aor-1 ‘I found the key at its place, but neither the pilaf nor the chicken was there.’ Disjunctive noun phrases in interrogative sentences are problematic for Dargi speakers; clause disjunction is preferred, see example (26) and §2.2.2. In that case, a disjunctive coordinator is not obligatory, but aAi ‘or’3 or ya-ra may optionally be used. (aA Ai) nergˇ-u? (26) ?pilaw-u pilaf(abs)-int (or soup(abs)-int {‘What shall we make for lunch?’} ‘(Shall we make) pilaf or soup?’ (26¢) pilaw b-ir-eAe-w, (aA Ai/ya-ra) nergˇ b-ir-eAe-w? pilaf(abs) n-do-fut.1p-int (or/or-and) soup(abs) n-do-fut.1p-int {‘What shall we make for lunch?’} ‘Shall we make pilaf or (shall we make) soup?’ 2.2.2 Clause disjunction The same three strategies used with NP-disjunction are used in clause disjunction. Coordinated declarative sentences use ya-ra…ya-ra, e.g.:

3.aAi is most probably a petrified gerund form of the negative auxiliary root aA ‘not be’.

204 Helma van den Berg

(27) ya-ra pilaw b-ir-eAe, ya-ra nergˇ b-ir-eAe or-and pilaf(abs) n-do-fut.1p or-and soup(abs) n-do-fut.1p {‘What shall we make for lunch?’ } ‘We’ll make pilaf or soup.’ (28) Au džä’äl ya-raˇi arq’-ad, ya-ra Aäncˇi-la you(abs) tomorrow or-and doctor-sup leave-fut.2 or-and work-loc dura.ulq-ad go.out:m-fut.2 ‘Tomorrow you’ll either go to the doctor or you’ll go to work.’ In negated sentences the bisyndetic particle ya is used, e.g.: (29) ya il nab-cˇi Ae-k’-ib, ya nu sune-cˇi or this me-sup neg-m:come-aor(3) or me(abs) self-sup Aä-wq’-un-ra neg-m:go-aor-1 ‘Neither did he come to me, nor did I go to him.’ Interrogative clauses are linked with ya-ra or aAi ‘or’, e.g. (30) nab hanna han·bÒirk-uli aAen, segˇuna-ri-l me:dat now remember:n-ger be.not what.kind-pret-indir nergˇ, ’är’ä-la-ri-w, (ya-ra/aA Ai) qara-la-ri-w? soup(abs) chicken-gen-pret-int (or-and/or) bean-gen-pret-int ‘I don’t recall what kind of soup it was, (was it) chicken or (was it) bean soup.’ 2.2.3 Adversative coordination Dargi has an adversative coordination cl*tic -gu ‘but’, and the coordinator amma, borrowed from Arabic. Whereas amma is clause-initial, -gu is usually clause-final. The proportion of -gu : amma in the text corpus is 24 : 19. The use of -gu is illustrated in example (18) above, and in (31)–(32); the use of amma in example (25) above, and in (33)–(34). There seems to be a slight semantic difference between the two adversative strategies, in that -gu is used predominantly with unfulfilled expectations. The nuance of expectation may be absent from the clauses coordinated with amma, but in a number of cases the particles occur in similar contexts (compare examples (32) and (34)). (31) [sen džäwli taman+d-i‘-ubti-w? [gˇal barxa-cad [how early end+pl-be-paprt-int] [twenty measure(abs)-eq rizq’i-la dag d-elq’-un-ra-gu (6:25) cereals-gen yesterday pl-grind-aor-1-but] ‘Did the cereals come to an end that quickly? But yesterday I ground you some twenty measures!’

Coordinating constructions in Daghestanian languages 205

(32) xan [yurt b-ic-a ‘-ili r-uqna-cˇi w-äq’-i khan(abs) [house(abs) n-sell-imp] say-ger f-old-sup m-come-ger sa.y-gu r-uqna q’abul·rÒik-ili aAen (1:3) be:m-but f-old agree:f-ger be.not ‘The khan went to the old lady and said “Sell me the house”, but the old lady did not agree.’ (33) darman-k’un nuša-ni as-i-ra amma il-d.a-ni medicine(abs)-foc we-erg buy-aor-1 but this-pl-erg wac-ni d-ubk’=aq-uli aAen mouse-pl(abs) pl-die=caus-ger be.not ‘We bought the poison, but it is not killing mice.’


(34) il-Aeli xan [xalasi baha wec’nu.šura Guruš this-when khan(abs) [big price(abs) fifteen rouble(abs) lug-as-nu b-ic-a w-ik’-uli sa.y, amma give-fut.1-therefore n-sell-imp] m-say-ger be:m but r-uqna ’ur-ra q’abul·rÒik-ili aAen (1:4–5) f-old then-and agree:f-ger be.not ‘Then the khan said: “I will give you a good price, fifteen roubles, so sell it”. But again the old lady would not agree.’ 2.2.4 Causal coordination Dargi has a causal coordinator -nu. It basically denotes the cause, reason or background of a situation in the following or — less frequently — preceding clause. See example (34) above, and (35)–(38). (35) ila-w ’äyarq’äna ’äsi·wÒi‘-ubli sa.y-nu … here-m hunter(abs) be.angry:m-ger be:m-therefore Malla.Nasradin w-it-ili sa.y (29:17) Malla.Nasradin(abs) m-hit-ger be:m ‘The hunter got very angry and (therefore) beat Malla Nasradin fiercely.’ (36) Au-ni nab Aä b-äq-i-ri-nu w-aš-i you-erg me:dat joke-dat n-hit-aor-2-therefore m-come-imp q’adi-cˇiˇi cadi-sup jurisdiction-sup ‘You hit me for fun, (so) come to the cadi for justice.’


(37) Ae-b-alx-as il džäwlil gaš b-ebk’-ili-nu (25:18) neg-n-feed-fut.1 this already hunger(erg) n-die-ger-therefore ‘I won’t (have to) feed it, for it died long ago of hunger.’

206 Helma van den Berg

(38) dura.wq-en quli-w-ad Au-ni way-ti xabur-ti leave:m-imp home-m-ela you-erg bad-pl story-pl(abs) d-ur-i-ri-nu pl-say-aor-2-therefore ‘Leave my house, for you have only told me bad stories.’


The particle -nu also indicates the order of activities, as in (39), and denotes alternatives, as in (40), e.g.: (39) d-ucˇ-a-nu le·rÒilra Ae-la d-äcˇ-unti t’alAäna pl-gather-imp-therefore all:pl you-gen pl-break-paprt pottery(abs) d-elq’-en (20:11) pl-grind-imp ‘Gather all your broken pottery and grind it.’ (40) nu-ni Ae-b-alx-as-nu Au-ni b-alx-en me-erg neg-n-feed-fut.1-therefore you-erg n-feed-imp ‘I won’t feed it, you feed it.’



Conjunctive coordination in other Daghestanian languages

Most descriptions of Daghestanian languages lack an elaborate treatment of coordination. What little is said about coordination concentrates on conjunctive coordination. Some languages predominantly have bisyndetic coordination (§ 3.1), others monosyndetic coordination (§ 3.2), and a third group use both strategies, with different coordinators (§ 3.3). 3.1 Bisyndetic conjunctive coordination Bisyndetic conjunction is found in Avar, Andic, Tsezic, Lak and Archi. Avar uses the coordinators -gi…-gi for NP-coordination; alternatively we find monosyndetic NP-coordination with wa, e.g.: (41) wac-gi jac-gi emen-gi ebel-gi brother(abs)-and sister(abs)-and father(abs)-and mother(abs)-and ‘-ana xuri-r-e (Uslar 1889:241) go-pst field:in-pl-ill ‘Brother, sister, father and mother went to the field.’

Coordinating constructions in Daghestanian languages 207

(42) hes c’aq’ t’adcˇ’u-n ‡azar-ul-e-l r-uk’-un r-ugo this very jump-ger learn-pres-part-pl pl-be-ger pl-cop xalq’.a‡-ul kic-abi, q’isa-bi, bicank’-abi, people-gen proverb-pl(abs) story-pl(abs) saying-pl(abs) marha-bi wa kucˇ’dul (Bokarev 1949:220) tale-pl(abs) and song:pl(abs) ‘He started frantically studying popular proverbs, stories, sayings, tales and songs.’ A perfect native Avar variant of this would be: (42¢) …xalq’.a‡-ul kic-abi-gi q’isa-bi-gi …people-gen proverb-pl(abs)-and story-pl(abs)-and bicank’-abi-gi marh-abi-gi kucˇ’dul-gi saying-pl(abs)-and tale-pl(abs)-and song:pl(abs)-and ‘… popular proverbs, stories, sayings, tales and songs.’ With clause coordination -gi…-gi can be added to the absolutive NP of the clause. Alternatively, the coordinator wa is used. (43) was hecˇ’ogo kawu-gi xut’utegi, yas boy(abs) not.being gate(abs)-and stay:opt.neg daughter(abs) hecˇ’ogo ebel-gi xut’utegi (Alekseev & Ataev 1998:64) not.being mother(abs)-and stay:opt.neg ‘Let a house not be without a son, let a mother not be without a daughter (i.e. may a house have a son, may a mother have a daughter).’ (44) maAmud ha-w-una wa ’-una qaAab.roso Makhmud(abs) be.born-m-pst and grow.up-pst Qahab.Roso ab-ura-b hit’ina·bÒgo ros.o-‡ (Bokarev 1949:220) say-paprt-n small:n village-in ‘Makhmud was born and grew up in a small village, called Qahab-Roso.’ Likewise, other Daghestanian languages have bisyndetic NP-coordination, e.g. Lak with the coordinators -gu…-gu, Bagvalal and Godoberi with -la…-la, Hunzib with -n(o)… -n(o), Bezhta with -na…-na, and Archi with -(w)u…-(w)u. Lak (45) bu¯ta-l q¯ urba¯n-un lu-gu surat-gu father-erg Kurban-dat book(abs)-and picture(abs)-and š¯eqqi-gu q¯ alantbicˇu-gu la·rÒs-una (Žirkov 1955:132) ink(abs)-and buy:pl-pst ‘Father bought Kurban a book, a picture, ink and a pencil-box.’

208 Helma van den Berg

Godoberi (46) zini-la, hanixa-la, unsa-la išqa r-a‘a cow(abs)-and calf(abs)-and ox(abs)-and home pl-come.pst ‘The cow, the calf and the ox came home.’ (Saidova 1973:147) Bagvalal (47) Šˇ¯e-b-o ek’wa mažit-la mimaro-la do-n-conv be mosque(abs)-and minaret(abs)-and ‘A mosque and minaret were built.’ (Kibrik (ed.) 2001:179) (48) o-b mak’a b-uk’a XindiX L’or-la ek’wa-b this-n place(abs) n-be [around rock(abs)-and be:part-n] [hini s’u˜j-la b-ižu-r-o¯-b mak’a [inside reed(abs)-and n-grow-ipf-part-n] place(abs) ‘This was a place surrounded by rocks and overgrown with reeds.’ (Kibrik (ed.) 2001: 558) Hunzib4 (49) sGd iyu-s zuq’u-n lo q’anu q’6ra, ože-n one mother-gen be-ger cop:hpl two child(abs) boy(abs)-and kid-no girl(abs)-and ‘A mother had two children, a boy and a girl.’ (50) Ilti-s-no Žagind.a-s-no hare b6cdul-en li Ilti-gen-and Zhagindo-gen eye(abs) be.blind-ger cop:pl ‘Ilti’s and Zhagindo’s eyes are blind.’



Bezhta (51) c’uddo-na c’odolo-na b-ox-ijo do q’alam red-and black-and III-buy-pres me pencil(abs) ‘I bought red and black pencils.’ (Madieva 1965: 143) (52) edo-na giso-na j-acˇ’cˇ’o gej inside-and outside-and IV-cold cop ‘It is cold inside and outside.’ (Madieva 1965: 143) Archi (53) X¯ wak lo-bur-u Aeleko-wu edi-li near child-pl(abs)-and hen(abs)-and be-infer ‘There were hen and chicks nearby.’ (Kibrik et al. 1977a: 322)

4.Hunzib examples are from the texts in van den Berg (1995): they are supplied with text and line numbers.

Coordinating constructions in Daghestanian languages 209

3.2 Monosyndetic conjunctive coordination In the Lezgic branch of Nakh-Daghestanian some languages have medial monosyndetic conjunction, viz. Lezgian, Tsakhur, Budukh, Kryz, Udi and Khinalug. The borrowed coordinator wa ‘and’ occurs additionally in most Lezgic languages. The native Lezgian marker for NP-conjunction is -ni, which is suffixed to all coordinands but the last one, e.g.: (54) Isa.di-ni Ali.di sada-sada-w gˇil-er wuga-na Isa(erg)-and Ali(erg) one(erg)-one-ad hand-pl(abs) give-aor ‘Isa and Ali shook hands.’ (Haspelmath 1993: 327) With special emphasis all coordinands are suffixed with -ni (see example (55)). In that case, the last conjunct can be extraposed to the end of the sentence after the finite verb, as in (56). (55) zi buba-ni, buba.di-n buba-ni I:gen father(abs)-and father-gen father(abs)-and ˇcuban-ar xˆ a-ji-bur ja shepherd-pl(abs) cop ‘Both my father and my father’s father were shepherds.’ (Haspelmath 1993: 327) (56) dax bürq’ü-ni xˆ a-nwa biši-ni dad(abs) blind-and become-prf deaf-and ‘Dad has become both blind and deaf.’ (Haspelmath 1993: 328) NP-coordination may alternatively take place by using the borrowed coordinator wa ‘and’, e.g.: (57) abur Isaq’.a-z wa ata-nwa-j mükü itim-r.i-z kilig-na they Isaq-dat and come-prf-part other man-pl-dat look-aor ‘They looked at Isaq and the other people who had arrived.’ (Haspelmath 1993: 330) In clause coordination the same two coordinators are used; -ni is suffixed to the first constituent of the last conjunct clause, e.g.: (58) [abur ˇca-l-aj wik’eh ja pacˇah.di-k-aj-ni kicˇ’e tuš [they we-sup-ela brave cop] [czar-sub-ela-and afraid cop:neg] ‘They are braver than we and they are not afraid of the czar.’ (Haspelmath 1993: 336)

210 Helma van den Berg

(59) [wun zi gaf-ar.a-l qhüre-na-j wa za-z … [you:(abs) me:gen word-pl-sup laugh-aor-pst] and [me:dat žawab ga-na-j answer(abs) give-aor-pst] ‘You laughed about my words and gave me the answer, …’ (Haspelmath 1993: 336) Tsakhur has a monosyndetic strategy for NP-coordination with -ij (Ibragimov 1990) or -jı¯ (Kibrik (ed.) 1999), e.g.: (60) dak-ij jedj Mask’aw-qa habk’Gn father(abs)-and mother(abs) Moscow-all go:pst ‘Father and mother went to Moscow.’ (Ibragimov 1990: 143) ¯ (61) hiˇŠo¯-je¯ Ru jed.i-s-jı¯ daki-s ha‘-u what.IV-q you:erg mother-dat-and father-dat hama-n-Gd wa-s awajk-es that.IV-attr-coh you-dat IV.find-pot ‘What you do to your father and mother, will happen to you as well.’ (Kibrik (ed.) 1999: 186) Schulze (1997: 66) mentions the additional use of bisyndesis, e.g.: dak-ij jedj-ij ‘father-and mother-and’. Examples of wa in both noun phrase and clause coordination are, e.g.: (62) man-biš-e¯ iš jugba wa 6k’ba ha¯‘a this-pl-erg work(abs) well.III and quickly.III ‘They do their work well and quickly.’ (Schulze 1997: 66) (63) rasul qarG wa haj-na iš haw‘-u rasul and this-attr work(abs) ‘Rasul came and did the work.’ (Kibrik (ed.) 1999: 460) 3.3 Languages with both monosyndetic and bisyndetic conjunction Both Tabasaran and Agul have three alternative strategies for conjunctive NPcoordination, with monosyndetic -na, with bisyndetic -ra…-ra and with the borrowed coordinator wa; however, no full examples of the use of wa are given in the literature. E.g.: Tabasaran (Magometov 1965: 331) (64) a.

gagaj-na dadaj-na baw šahur.di-s uš-nu father(abs)-and mother(abs)-and grandmother town-dat go-pst ‘Father, mother and grandmother went to the city.’

Coordinating constructions in Daghestanian languages

b. gagaj-ra dadaj-ra baw-ra šahur.di-s father(abs)-and mother(abs)-and grandmother(abs)-and uš-nu town-dat go-pst ‘Father, mother and grandmother went to the city.’ Agul (Magometov 1970: 173) ¯ Aajwan.di-s-na Aani-s jak’war keq horse-dat-and cow-dat hay(abs) give:imp ‘Give the horse and the cow hay.’ ¯ b. Aajwan.di-s-ra Aani-s-ra jak’war keq horse-dat-and cow-dat-and hay(abs) give:imp ‘Give the horse and the cow hay.’

(65) a.

In Rutul, the picture is a bit more complicated.5 Rutul seems to have two conjunctive coordination strategies, with monosyndetic -na/-ne (no distribution given) and bisyndetic -ki…-ki (Maxmudova 2001: 206). According to Maxmudova, the default coordinator is -na/-ne, whereas -ki…-ki is used with enumeration. Ibragimov (1978: 113), however, mentions only the monosyndetic strategy with -na/-ne, and the use of wa ‘and’, while regarding -ki as a mere additive focus particle.6 (66) dagˇGstandi y-irq’-Gd mihman.a-s hürmät x’an-na to.Daghestan I-come-part guest-dat refreshment(abs) beer(abs)-and sGv wij oatmeal(abs) cop ‘For a guest coming to Daghestan, refreshments are beer and oatmeal.’ (Ibragimov 1978: 112) (67) a.

day-ne yiwan-na iessi bala q’’Gcˇi d-iširi foal(abs)-and horse(abs)-and master(abs) much be.tired pl-get:pst ‘The foal, the horse and master got tired.’ (Maxmudova 2001: 206)

5.The Rutul examples given in Maxmudova and Ibragimov are from the Mukhad dialect; the corresponding elements in the Luchek dialect are -nG and -ˆxG…-ˆxG (Alekseev 1994b: 248). 6.Linking the Rutul focus particle -ki to the bisyndetic construction is not unproblematic, though. For the Mukhad dialect -ki…-ki is described as emphatic coordination ‘both…and’ (Maxmudova 2001: 206), but no such reference is made for the related coordinators -ˆxi…-ˆxi in Luchek (Alekseev 1994b: 248). The focus particle -ki occurs in both Tsakhur and Rutul, and is regarded as a loan from Azerbaijani by Ibragimov (1990: 143). Given the dialectal variant -ˆxi…-ˆxi, it would have to be an early loan.


212 Helma van den Berg

b. day-ki yiwan-ki iessi-ki bala q’’Gcˇi foal(abs)-and horse(abs)-and master(abs)-and much be.tired d-iširi pl-get:pst ‘Both the foal, and the horse and the master got tired.’ (Maxmudova 2001: 206) (68) wä pGlGw-a u¯ sipil kašniš wa maddG you pilaf-loc onion(abs) coriander(abs) and other uq’-bGr sG‘G-r ulärä herb-pl(abs) throw-ger eat.prf ‘Having put onion, coriander and other herbs on the pilaf, you ate it.’ (Ibragimov 1978: 113) 3.4 Other uses of the coordinator The coordinators treated so far in general also occur as additive focus markers, e.g.:7 Avar (69) ebel.a‡ di-da ma‡-ul-e-b b-ugo, ha-b nux.a‡ mother(erg) me-sup teach-pres-part-n n-cop this-n time(erg) ˇcu-gi b-ugo di-da ma‡-ul-e-b horse(abs)-and n-cop me-sup teach-pres-part-n ‘My mother teaches me a lesson, this time my horse teaches me a lesson as well.’ (Charachidzé 1981: 175) Bagvalal (70) sangut-abi partal-la b-uk’a ˇc’ihi chest-pl(abs) things(abs)-and n-be ‘There were chests and (other) things as well on top (of the truck).’ (Kibrik (ed.) 2001: 706) Hunzib (71) 6g ‡e‘-o‡ b"d-ra-lo-n nGs6-n ilu-s-no that.I laugh-when other-pl-erg-and say-ger we-gen-and interes·bÒ"qu-n lo (19:74) be.interesting:IV-ger cop.IV ‘When he laughed, the others, too, said: “It is interesting for us as well.”’

7.An elaborate description of the Bagvalal focus marker -la is given in Kibrik (ed.) (2001: 706–713).

Coordinating constructions in Daghestanian languages

Lezgian (72) insan q’e-ji-la ada-n gˇed-ni awat-zawa person(abs) die-aop-temp s/he-gen star(abs)-and fall-impf ‘When a person dies, his or her star also falls down.’ (Haspelmath 1993: 328) Rutul (73) zG-ki y-Gq’a-si me(abs)-and I-come-fut ‘I will come too.’ (Maxmudova 2001: 206) Archi (74) zon šutto-wu w-irxwni-qi me(abs) tomorrow-and I-work-fut ‘I will work tomorrow as well.’ (Kibrik et al. 1977a: 321) In a number of languages the focus marker is regularly found after the absolutive NP of the subordinate clause, see Dargi examples (17)–(18), and, e.g.: Hunzib (75) [o‡u-s quy nGd-o‡, [ogu q’6ra-n [that-gen noise(abs) hear-when] [that.II child(abs)-and b-ecˇamiye-n gišoke-n lo (1:6) hpl-leave.behind-ger] go.outside-ger cop.II ‘When she heard his noise, she went out, leaving her children behind.’ (76) [oq’en boco-n bu†ii-n ecˇe-n, bed b6l [four month(abs)-and home-and I.stay-ger] then this.I ože… Aalt’i.l"-" ˜e†’e-n lo boy(abs) work-dat I.go-ger cop.I ‘Having stayed at home for four months, the boy went to work.’


Godoberi (Kibrik (ed.) 1996: 198, cited from Saidova 1973: 156) ¯ (77) [haL’.e¯-Li L’eL’il-la b-ik-u, [hilak¯wa [armful-inter saddle(abs)-and n-put.pst-conv] [ ¯ burtina-la b-ik-u, [b-eXut’u X¯ wani-la coat(abs)-and n-put.pst-conv] [n-behind horse(abs)-and ¯ gyann-u, cellada rekin-o-la w-un-a, pull.pst-conv] another cavity-in-and m-go.pst-conv ¯ w-oraX-u-da m-lie.down.pst-conv-cop ‘(He) put the saddle under his arm, put the coat on (his shoulders), pulled the horse, passed to another cavity and lay down.’


214 Helma van den Berg

Archi nen (78) jemim-u ak’u-li, q’a these-and chase-conv go.prf we(abs) ‘Having chased them away, we went.’ (Kibrik et al. 1977a: 324) Bagvalal (79) [o-b-la ˇc’al’a-m-o [han-ab.a-r q’ot’i Šˇ¯e [this-n-and bother-n-conv] [village-pl-erg agreement(abs) do [sani b-uh-a¯, ce-b han-la [together hpl-gather-pot.inf] [one-n village(abs)-and Šˇ-a¯la do-pot.inf]] ‘Getting tired of it, the villages decided to unite and to make one village.’ (Kibrik (ed.) 2001: 555) Kibrik (ed.) (2001: 554–558) gives a detailed account of this phenomenon in Bagvalal. The explanation put forward there is that the Bagvalal particle seems to be obligatory when the subordinate clause and the main clause are semantically relatively independent. Example (79) could also be translated as ‘the villages got tired of it and decided to unite and make one village’. According to Kibrik, the use of the particle in subordinate clauses is clearly linked to its use as a coordinator. The analysis proposed by Kibrik may hold for the other Daghestanian languages as well, but more study is needed to confirm this explanation.

4. Overview and discussion of conjunctive coordination The following table gives an overview of all the attested conjunctive coordinators in Daghestanian languages, not just of those that were exemplified in the preceding section.8

8.The literature on Akhvakh, Tindi, Botlikh, and Khvarshi mentions just a single coordinator. It can however be assumed that bisyndetic coordination is indeed the rule, because the early literature on Chamalal, Godoberi and Bagvalal also recorded just a single coordinator, whereas later, more elaborate, studies of these languages described bisyndetic coordination only. Probably the bisyndetic character of the Andic and Tsezic coordination, being rather common in Daghestanian languages, was taken for granted by the authors, and therefore not explicitly mentioned.

Coordinating constructions in Daghestanian languages



Lak Dargi Lezgic


coordinating construction borrowed


-gi…-gi -lo…-lo -la(…-la) -la…-la -la(…-la) -la(…-la) -la…-la

Chamalal Karata

Bokarev 1949: 220 Cercvadze 1965: 292 Magomedbekova 1967: 347 Kibrik (ed.) 2001: 179 Gudava 1967: 379 Gudava 1962: 153 Saidova 1973: 147; Kibrik (ed.) 1996: 90 Magomedova 1999: 433 Magomedbekova 1971: 177


Bokarev 1967a: 415 Lomtadze 1963: 179 Bokarev 1967b: 431 Madieva 1965: 143 van den Berg 1995: 51 Žirkov 1955: 132 van den Berg 2001: 69 Haspelmath 1993: 327 Magometov 1965: 331 Magometov 1970: 173 Ibragimov 1990: 143; Kibrik (ed.) 1999: 186 Ibragimov 1978: 112; Maxmudova 2001: 206; Alekseev 1994b: 248 Mejlanova 1984: 198; Alekseev 1994a: 290 Saadiev 1994: 441 Kibrik et al. 1977a: 322 Pancˇvidze & Džejranišvili (1967: 685); Schulze 1984: 201 Dešeriev 1959: 152

Avar Andi Akhvakh Bagvalal Tindi Botlikh Godoberi

-la…-la -(e)l…-(e)l (Tokita dialect -lV…-lV) Tsez -n(o)… (-n(o)) Hinukh -n(o)… -n(o) Khvarshi -n(o)… (-n(o)) Bezhta -na…-na Hunzib -n(o)… -n(o) -gu…-gu -ra…-ra Lezgian -ni Tabasaran -na; -ra…-ra Agul -na; -ra…-ra Tsakhur -ij -jı¯ Rutul -na/-ne (Mukhad dialect) -na/-ne; -ki…-ki (Mukhad) -nG; -ˆxG…-ˆxG (Luchek dialect) Budukh -na, -Gn -n6 Kryz -nä Archi -(w)u…-(w)u Udi -q’a -q’a(n)9 yä


wa wa wa wa wa wa wa


wa wä/w’a

9.According to Wolfgang Schulze (p.c.), the final n of Udi -q’a(n) is historically connected to the Proto-Lezgic coordinator *-nV.

216 Helma van den Berg

Map 1 gives an overview of the relative position of the major Daghestanian and Turkic languages of the area discussed in this section.



Avar Tsezic

Caspian Sea

Dargi Lak


Tabasaran Agul Lezgian Rutul/ Tsakhur Azerbaijani

Map 1.Relative position of the major Daghestanian and Turkic languages of the area

An interesting correlation is observed between monosyndetic coordination in the Lezgic group versus bisyndetic coordination in the remaining Daghestanian languages. In the Lezgic group, however, bisyndesis is also not uncommon. Lezgian, for example, uses the same coordinator -ni in a bisyndetic strategy to denote emphatic conjunction. We may assume that this is found in other Lezgic languages as well. Three other Lezgic languages, Tabasaran, Agul and Rutul, also have bisyndetic coordination. Finally, Archi is the single Lezgic language with a bisyndetic strategy only. 4.1 Variation of coordinators All Daghestanian languages are genetically related and their subgrouping is clear. In this light, the variation in bisyndetic coordinators is rather remarkable. Whereas the conjunctive coordination construction itself is in principle similar for all the languages, the coordinator is different for all the subgroups: Avar -gi…-gi, Dargi -ra…-ra, Lak -gu…-gu, Andic -l(V)… -l(V), Tsezic -n(V)… -n(V), Agul and

Coordinating constructions in Daghestanian languages 217

Tabasaran -ra…-ra, Rutul -ki…-ki, Archi -w(u)… -w(u). The reconstruction of a Proto(-Nakh)-Daghestanian coordinator seems hardly feasible.10 4.2 Grammaticalization In most Daghestanian languages, the coordinator also occurs as an additive focus particle ‘also, too’ or with a scalar meaning ‘even’: this has been attested for Avar, Hunzib, Bagvalal, Godoberi, Dargi, Lezgian, Rutul, and Archi. We can safely assume that the conjunctive coordinators have been grammaticalized from this source. This is in line with the overall correspondence between additive focus particles and coordination, discussed by König (1991: 1).11 Given the division between monosyndetic and bisyndetic coordinating constructions observed, the development of this division needs to be addressed: was (Proto-)Nakh-Daghestanian coordination bisyndetic and did Lezgic, with a few exceptions, develop a monosyndetic construction? In that case we would need to explain the retention of the Lezgic bisyndetic cases. The reverse development seems somewhat more plausible: monosyndesis is the original situation, which was retained in Lezgic. The other Daghestanian languages developed bisyndetic coordination, probably on the basis of emphatic conjunction, as exemplified in the Lezgian examples (55)–(56). As bisyndetic conjunction is less common than monosyndetic conjunction cross-linguistically, we would have to explain why it developed in the majority of the Daghestanian languages. No resolution of this question is offered here. A possible explanation for the Lezgic bisyndetic cases is language contact. Archi, which is a Lezgic language, is detached from the Lezgic area and surrounded by Avar and Lak; most Archi are multilingual in Archi, Avar and Lak (and Kumyk and Russian). Furthermore, Archi speakers are politically regarded as Avars, and Avar has been taught as their ‘native’ language in school since the 1930s. In that case Archi would have been influenced by Avar and Lak through indirect diffusion, i.e.

10.Trubetzkoy (1930: 78), however, linked the Lak, Dargi and Avar coordinators to West Caucasian coordinators: Abkhaz -gi, Ubykh -gä, Adyghe k’6 (Kjach also g6) ‘and’ ~ Avar -gi, Lak -gu ‘and’, perhaps also Dargi -gwa ‘though (but)’; Adyghe (Kjach) -re ‘and, both…and’ ~ Dargi, Agul, Tabasaran -ra ‘and’. Trubetzkoy used the term Adyghe to indicate Circassian as a whole (including both West Circassian/Kjach, currently called Adyghe, as well as East Circassian, currently known as Kabardian). The Dargi adversative coordinator -gwa referred to by Trubetzkoy is an Urakhi form, whereas Standard Dargi -gu (see §2.3) is based on the Akusha form. A reconstruction of the coordinators for Nakh-Daghestanian or even North-Caucasian would only hold if it is part of a further pattern of regular sound correspondences. 11.Other interesting correspondences of additive focus particles mentioned by König, e.g. with interrogative pronouns, conditionals, also seem to hold for a great number of Daghestanian languages.

218 Helma van den Berg

it developed bisyndetic coordination without taking over the Avar or Lak coordinators. As for Tabasaran and Agul, the bisyndetic coordination with -ra…-ra, alongside Lezgic -na, is identical to the coordination in neighbouring Dargi. Tabasaran borders the Kaytag dialect of Dargi; Agul is in contact with the Chirag and Sirkha dialects of Dargi. Historically, the Kaytag and Tabasaran belonged to the same political administrative unit, and so did the Agul and Chirag. There has always been a regular exchange of goods between the ethnic groups concerned. Both in the case of Archi and of Tabasaran and Agul we also need to take into account the potential influence from Kumyk or Azerbaijani. Further research on language contact in Daghestan is needed to confirm this hypothesis. 4.3 Diffusion of wa Beside the various native means for conjunction, medial monosyndetic coordination with the borrowed coordinator wa is mentioned in a number of sources. This is mainly the case with publications on the written languages Avar, Dargi, Lak, Lezgian, Tabasaran, Agul, Tsakhur and Rutul; examples of its use are usually provided. The most elaborate description of the use of wa is found in Haspelmath (1993: 336). Coordination with wa is described here as a non-native syntactic construction for Lezgian, which was borrowed along with the coordinator itself. The use of wa is now quite common, especially in translations from Russian; it is restricted to the written language only, and rejected by speakers not literate in Lezgian. A number of descriptions of unwritten languages also mention the use of wa, though without adducing examples, e.g.: Godoberi (Saidova 1973: 147, but not Kibrik (ed.) 1996), Hunzib (Bokarev 1959: 62, but not van den Berg 1995), Bezhta (Xalilov 1995: 408, but not Madieva 1965), Hinukh (Lomtadze 1963: 179), Archi (Mikailov 1967: 143, but not Kibrik et al. 1977a,b). It seems that wa is considered by these authors to be a ‘proper’ coordinator, which is even attributed to languages in which it is hardly used. With regard to the written languages, it is interesting to note that wa seems to have been introduced only in the 20th century. None of Uslar’s grammars of Avar (1889), Lak (1890), Dargi (1892), Lezgian (1896) and Tabasaran (1979) mentions this coordinator. Education in Uslar’s time was primarily concerned with Arabic language, science and culture and was conducted in Arabic. Even people who were attending Russian schools would at the same time have had an additional education in Arabic. Furthermore, Avar, Dargi, and Lak were written already in the 17th and 18th centuries using the Arabic script. There is no indication, however, that the coordinator wa was used earlier than the 20th century. Therefore we can conclude that the introduction of wa was closely related to the wide-spread development of literacy in Daghestan in the 1920s. The languages used in writing in Daghestan at the time, Arabic and Russian, both have a medial

Coordinating constructions in Daghestanian languages 219

monosyndetic coordinator. The recent emergence of wa fits nicely into the argument put forward in Mithun (1988:351–357), that conjunctive coordinators emerge under the influence of contact with languages with a literary tradition. The choice of an Arabic coordinator might have been influenced by the close cultural affinity. Daghestanian languages have a tradition of borrowing ‘intellectual’ vocabulary (e.g. terms for religious, spiritual and cultural phenomena) from Arabic, and this tradition seems to have been respected in the choice of the coordinator as well. Furthermore, the written use of Turkic languages for interethnic communication may have supported this development. Azerbaijani was used for this purpose in the southern part of Daghestan, Kumyk in central and northern Daghestan. The use of the particle wa in Azerbaijani at the time was restricted to the written language, and wa was not used in colloquial speech. Azerbaijani was used in writing and in education in southern Daghestan between 1923 and 1928. The language had and continues to have a considerable impact on the Lezgic languages. Kumyk acquired its literacy at the same time as the East Caucasian languages. Its own development in this respect was influenced by other written Turkic languages, such as Tatar, and, again, Azerbaijani. Even today, the use of wa has remained characteristic of written texts. Furthermore, there seems to be a largely complementary distribution between the native Daghestanian (monosyndetic or bisyndetic) NP-coordinators, and the borrowed clause coordinator wa.


Dargi revisited: ‘and’ and ‘with’

Dargi and the other Daghestanian languages seem to be typical and-languages in Stassen’s (2000: 21–25) terminology: they distinguish between conjunctive coordination and a comitative construction. Daghestanian languages usually have a comitative case, which historically is often derivable from a local case. The Dargi comitative case is -cˇil, e.g.: (80) Nasradinˇil wac’ urcuy-cˇi w-äq’-i sa.y (32:1) Nasradin(abs) donkey-com forest-ill timber-sup m-go-ger be:m ‘Nasradin went with his donkey into the forest for timber.’ (81) arc-la gawlag-un.a-cˇil š ˇcar.uq-unsi Nasradin (27:13) … silver-gen bag-pl-com village-ill return:m-paprt Nasradin(abs) ‘Nasradin, who had returned to the village with bags filled with silver,’ (82) nu razideš.li-cˇil q’abul.irk-asi me(abs) happiness-com agree:m-impf.1 ‘I would agree with great gladness.’


220 Helma van den Berg

However, Dargi also employs another, more or less synonymous, comitative strategy, which involves the conjunction of a noun and the reflexive pronoun sa·BÒi.12 This construction is called the conjunctive-comitative construction here. It renders the expression ‘X with Y’ as [Y-and pronoun-and X]. In the examples given below the antecedent of the reflexive pronoun is underlined. (83) Malla.Nasradin t’ent’-ra sa.y-ra q’adi-cˇi Malla.Nasradin(abs) fly(abs)-and self:m(abs)-and cadi-sup hay.i‘-ubli sa.y (22:42) be:m ‘Malla Nasradin set off with the fly (lit. Malla Nasradin set off, the fly and self).’ (84) š uAna.uq-i sa.y emAe-laˇi-r ’inc-bi-ra village-ill m:enter-ger be:m donkey-gen back-sup-pl apple-pl-and (19:6) sa.y-ra ca c’udqran self:m-and one Tsudakharian ‘A man from Tsudakhar entered the village with apples on the back of his donkey (lit. a man from Tsudakhar entered the village, apples on the back of his donkey and self).’ (85) il-d.a-ni ˇce·bÒa‘-ili sa·bÒi b-ic’-ili k’abat’ this-pl-erg see:n-ger be:hpl pilaf(erg) n-fill-ger tray(abs) ˇcedi-b b-erc’-ibsi ’är’ä-ra sa·bÒi-ra (22)–(21) above-n n-fry-paprt hen(abs)-and self:n(abs)-and ‘They saw a tray filled with pilaf, with a fried hen on top of it (lit. they saw a tray filled with pilaf, a fried hen on top and self).’ The conjunctive-comitative construction occurs in both written and oral texts and is possible both with equally and non-equally agentive participants. (86) a.

xˆunul-ra sa.y-ra udzi šadi-w arq’-uli sa.y wife(abs)-and self:m(abs)-and brother(abs) walk-m leave-ger be:m ‘(My) brother went for a walk with (his) wife.’ b. ’äsa-ra sa.y-ra udzi šadi-w arq’-uli sa.y stick(abs)-and self:m(abs)-and brother(abs) walk-m leave-ger be:m ‘(My) brother went for a walk with (his) stick.’

12.The reflexive pronoun is identical to the auxiliary sa·BÒi, and both inflect for gender agreement: m sa.y, f sa·rÒi, n sa·bÒi, hpl sa·bÒi, pl sa·rÒi. The reflexive pronoun is used with third person antecedents only. The auxiliary is not only used as the copula, but is also part of the analytic tense forms of the Dargi verb, like the Perfective and the Present Progressive.

Coordinating constructions in Daghestanian languages 221

(87) a.

halmagˇ-ra sa·rÒi-ra kino friend(abs)-and self:f(abs)-and Salimat(erg) film(abs) ˇce·bÒa‘-ib see:n-aor(3) ‘Salimat saw the film with a friend.’ b. konfet-uni-ra sa·rÒi-ra kino sweet-pl(abs)-and self:f(abs)-and Salimat(erg) film(abs) ˇce·bÒa‘-ib see:n-aor(3) ‘Salimat saw the film with (a box of) sweets.’

It is also clearly distinguished from a coordinate construction, where the gender agreement on the verb is plural instead of singular. (86) a¢. xˆunul-ra udzi-ra šadi-b arq’-uli sa·bÒi wife(abs)-and brother(abs)-and walk-hpl leave-ger be:hpl ‘(My) wife and (her) brother went for a walk.’ or ‘(My) brother and (his) wife went for a walk.’ A striking feature of the conjunctive-comitative construction is that the first coordinated NP is often preceded by a gerund. This is the case in more than half of these constructions in the Dargi corpus. E.g.: (88) ca bulqran yäˇi ˇce·dÒiˆx-ili t’alAäna-ra one Balkharian(abs) horse-sup put.on:pl-ger pottery(abs)-and sa.y-raˇi… hay.i‘-ubli sa.y (20:1) self:m(abs)-and market-sup be:m ‘A man from Balkhar went to the market with his horse loaded with pottery.’ (89) usta-ra sa·bÒi-ra il-di ila lead:m-ger master(abs)-and self:hpl(abs)-and this-pl(abs) here ar·bÒäq’-i sa·bÒi (9:10) leave:hpl-ger be:hpl ‘Taking the master with them, they left.’ Alternatives to a conjunctive-comitative construction are a subordinate construction, as in example (89¢a), or a comitative noun phrase, as in (89¢b), e.g.: (89¢) a.

[usta-ra il-di ila ar·bÒäq’-i sa·bÒi [master(abs)-and lead:m-ger] this-pl here leave:hpl-ger be:hpl ‘Taking the blacksmith, they left.’ b.ˇil il-di ila ar·bÒäq’-i sa·bÒi master-com this-pl here leave:hpl-ger be:hpl ‘They left with the blacksmith.’

222 Helma van den Berg

The antecedent of sa·BÒi is either the intransitive subject, examples (88)–(89), the patient, examples (85), (90) — both in the absolutive —, or even the agent in the ergative, example (91). E.g.: (90) d-äq-ili dam+däd-ra sa·bÒi-ra xalasi pl-hit-ger drum+flute(abs)-and self:hpl(abs)-and big š adam-ti b-arg-ili sa.y (29:7) party-ill-hpl-pl person-pl(abs) hpl-find-ger be:m ‘He found people at a big party, with drums and flute being played.’ (91) hacˇam b-erc’-ili ’är’ä-ra sa.y-ra once Malla.Nasradin(erg) n-fry-ger hen(abs)-and self:m(abs)-and pulaw b-ar-ili sa.y (22:1) pilaf(abs) n-do-ger be:m ‘Once Malla Nasradin prepared some pilaf with a fried hen.’ There is one other Daghestanian language where an analogous conjunctive-comitative construction is possible, viz. Hunzib. Hunzib was analyzed in van den Berg (1995:52) as having a comitative case -gˇur and a comitative suffix -žun in addition, e.g.: (92) a.

ože ok’""k’ žin-do halmagˇ.li-gˇur boy(abs) wander.pres self-genobl friend-com ‘The boy wanders with his friend.’ b. ože ok’""k’ žini-s halmagˇ-no-žun boy(abs) wander.pres self-gen friend(abs)-and-with ‘The boy wanders with his friend.’

According to that analysis, the comitative suffix -žun in example (92b) follows the noun plus coordinator -n(o).13 Hunzib however also has a reflexive pronoun žu. In the light of the Dargi data above, and Hunzib coordination described in § 3.1, it seems to me that we should reanalyze the Hunzib construction in (92b) as a conjunctive-comitative construction, e.g.:14 (92) b¢. ože ok’""k’ žini-s halmagˇ-no žu-n boy(abs) wander.pres self-gen friend(abs)-and self(abs)-and ‘The boy wanders with his friend (lit. the boy wanders, and his friend and self).’

13.The construction could not be clarified further while working on Hunzib in the beginning of the 1990s; it seemed somewhat unusual for a suffix to have scope over the coordination construction, but a better solution could not be found at the time. Moreover, the informants did not point out a relationship between the comitative suffix -žun and the reflexive pronoun žu either. 14.Note that Hunzib žu, unlike Dargi sa·BÒi, does not inflect for gender.

Coordinating constructions in Daghestanian languages 223

Examples from the Hunzib corpus are, e.g.: (93) bercinab aqe-n žu-n maduhan-no beautiful wife(abs)-and self(abs)-and neighbour(abs)-and zuq’u-n lo o‡u-s be-ger be.I that-gen ‘He also had a neighbour with a beautiful wife.’


(94) gišoke-n y-ãq’-o‡, bedd"" giho ruquru leave-ger II-come-when over.there downwards long.grown mucˇ’-no žu-n ’adatab ca+boco=‡eru hair(abs)-and self(abs)-and beautiful star+moon=like kid lo †e (21:36) girl(abs) be.II qt ‘When she came out (i.e. of a tree), she turned out to be a beautiful girl with braids down to there, resembling the stars and the moon.’ (95) ãq’e-r †e h˜6s suk’u laAc’6dil s6ryo-n I.come-pret qt one man(abs) horse(abs)-and žu-n, … ãq’e-r †e h˜6s suk’u b6dil self(abs)-and I.come-pret qt one man(abs) like.this r-Gx6l macˇ’a-n žu-n (22:11) V-long sword(abs)-and self(abs)-and ‘There came a man on a very black horse, … there came a man with such a long sword.’ The development of a conjunctive construction into a comitative marker would be highly unusual in the light of earlier publications, which explicitly mention a grammaticalization process in the opposite direction (Stassen 2000: 26). As conjunctive-comitative coordination is not the only strategy for expressing a comitative relation, the construction is not (yet) fully grammaticalized. The proportion of comitative case vs. conjunctive-comitative coordination in the Dargi corpus is 34 : 21, in the Hunzib corpus 11 : 10. The presence of the conjunctive-comitative construction in two distantly related Daghestanian languages makes one curious about the presence of this construction elsewhere in the East Caucasian language family.

Abbreviations 1 1p 2

1st pers. sg (person agreement) 1st pers. plural (person agreement) 2nd pers. (person agreement)

3 3 pers. (person agreement) I, II, III, IV, Vgender markers (for languages with more than 3 genders)

224 Helma van den Berg

ad adj aff all aop aor appr attr caus cnt coh com cond cont conv cop dat ela eq erg f foc fut gen genobl ger hpl ill imp impf impf in in

adessive adjective affective allative aorist participle aorist approximative attributive causative contactive coherence marker comitative conditional contentive converb copula dative elative equative ergative feminine focus future genitive genitive with oblique nouns gerund 3rd pers. human plural (gender agreement) illative imperative imperfect imperfective (Lezgian) inessive localization ‘in’ (Godoberi)

indirect infinitive inferential interrogative localization ‘inter’ imperfective aspect locative masculine masdar neuter negation optative present participle preterite participle participle perfective plural (suffix with nouns) 1st, 2nd pers. hum., 3rd pers. non-hum. plural (gender agr.) pot potentialis pres present pret preterite prf perfect proh prohibitive pst past ptc particle ‘da’ q question marker qt quotation particle sbst substantivizer sub subessive sup superessive (in Avar, Lezgian) sup superlative temp temporal converb trans translative indir inf infer int inter ipf loc m masd n neg opt paprs paprt part pf pl pl

References Alekseev, Mikhail E., (1994a) Budukh. In: Smeets, Rieks (ed.), The indigenous languages of the Caucasus, vol. 4. Delmar: Caravan, 259–296. Alekseev, Mikhail E., (1994b) Rutul. In: Smeets, Rieks (ed.), The indigenous languages of the Caucasus, vol. 4. Delmar: Caravan, 213–258. Alekseev, Mixail E., Ataev, Boris M., (1998) Avarskij jazyk. Moskva: Academia. Bokarev, Anatolij A., (1949) Sintaksis avarskogo jazyka. Moskva-Leningrad: AN SSSR. Bokarev, Evgenij A., (1959) Cezskie (didojskie) jazyki Dagestana. Moskva: AN SSSR.

Coordinating constructions in Daghestanian languages 225

Bokarev, Evgenij A., (1967a) Tsezskij jazyk. In: Jazyki narodov SSSR, IV: Iberijsko-kavkazskie jazyki. Moskva: Nauka, 404–420. Bokarev, Evgenij A., (1967b) Xvaršinskij jazyk. In: Jazyki narodov SSSR, IV: Iberijsko-kavkazskie jazyki. Moskva: Nauka, 421–435. Cercvadze, Il’ja I., (1965) Andiuri ena: (gramatikuli analizi tekstebit). Tbilisi: Sakartvelos SSR Mecnierebata Akademia. Enatmecnierebis Instituti. Charachidzé, Georges, (1981) Grammaire de la langue avar. Paris: Jean-Favard. Dešeriev, Junius D., (1959) Grammatika xinalugskogo jazyka. Moskva: AN SSSR. Gudava, Togo E., (1962) Botlixuri ena: grammatikuli analizi, tekstebi, leksikoni. Tbilisi: Sakartvelos SSR Mecnierebata Akademia. Ernatmecnierebis Instituti. Gudava, Togo E., (1967) Tindinskij jazyk. In: Jazyki narodov SSSR, IV: Iberijsko-kavkazskie jazyki. Moskva: Nauka, 368–383. Haspelmath, Martin, (1993) A grammar of Lezgian. Berlin: Mouton de Gruyter. Ibragimov, Garun X., (1978) Rutul’skij jazyk. Moskva: Nauka. Ibragimov, Garun X., (1990) Caxurskij jazyk. Moskva: Nauka. Kibrik, Aleksandr E. et al., (1977a) Opyt strukturnogo opisanija arcˇinskogo jazyka. II: Taksonomicˇeskaja grammatika. Moskva: MGU. Kibrik, Aleksandr E. et al., (1977b) Arcˇinskij jazyk. Teksty i slovari. Moskva: MGU. Kibrik, Aleksandr E. (ed.), (1996) Godoberi. München: Lincom Europa. Kibrik, Aleksandr E. (ed.), (1999) Èlementy caxurskogo jazyka v tipologicˇeskom osvešcˇenii. Moskva: Nasledie. Kibrik, Aleksandr E. (ed.), (2001) Bagvalinskij jazyk. Grammatika, teksty, slovar’. Moskva: Nasledie. König, Ekkehard, (1991) The meaning of focus particles. A comparative perspective. London: Routledge. Lomtadze, Èližbar A., (1963) Ginuxskij dialekt didojskogo jazyka. Tbilisi: AN Gruzinskoj SSR. Madieva, Gjul’žagan I., (1965) Grammaticˇeskij ocˇerk bežtinskogo jazyka. Maxacˇkala: DGU. Magomedbekova, Zagidat M., (1967) Axvaxskij jazyk: grammaticˇeskij analiz, teksty, slovar’. Tbilisi: Mecniereba. Magomedbekova, Zagidat M., (1971) Karatinskij jazyk: grammaticˇeskij analiz, teksty, slovar’. Tbilisi: Mecniereba. Magomedova, Patimat T., (1999) Chamalinsko-russkij slovar’. Maxacˇkala: DNC RAN. Magometov, Aleksandr A., (1965) Tabasaranskij jazyk (issledovanie i teksty). Tbilisi: Mecniereba. Magometov, Aleksandr A., (1970) Agul’skij jazyk (issledovanie i teksty). Tbilisi: Mecniereba. Maxmudova, Sveta M., (2001) Morfologija rutul’skogo jazyka. Moskva: Institut Jazykoznanija RAN. Mejlanova, Unejzat A., (1984) Buduxsko-russkij slovar’. Moskva: Nauka. Mikailov, Kazbek Š., (1967) Arcˇinskij jazyk. Maxacˇkala. Mithun, Marianne, (1988) The grammaticization of coordination. In: Haiman, J. & Thompson, S. A. (eds.), Clause combining in grammar and discourse. Amsterdam: Benjamins, 331–359. Nikolayev, Sergey L., Starostin, Sergey A., (1994) A North Caucasian etymological dictionary. Moscow: Asterisk. Pancˇvidze, V. N., Džejranišvili, E. F., (1967) Udinskij jazyk. In: Jazyki narodov SSSR, IV: Iberijskokavkazskie jazyki. Moskva: Nauka, 676–688. Saadiev, Shamsaddin M., (1994) Kryts. In: Smeets, Rieks (ed.), The indigenous languages of the Caucasus, vol. 4. Delmar: Caravan, 407–446. Saidova, Patimat A., (1973) Godoberinskij jazyk. Maxacˇkala: Dagfilial AN SSSR. Schulze, Wolfgang, (1984) Die Sprache der Uden in Nord-Azerbajdžan. Wiesbaden: Harrassowitz. Schulze, Wolfgang, (1997) Tsakhur. München: Lincom Europa (LM/W 133). Stassen, Leon (2000) AND-languages and WITH-languages, Linguistic Typology, vol. 4, 1–54. Trubetzkoy, Nikolay S., (1930) Nordkaukasische Wortgleichungen, Wiener Zeitschrift für die Kunde des Morgenlandes 38, 76–92.

226 Helma van den Berg

Uslar, Petr K., (1889) Ètnografija Kavkaza: Jazykoznanie. Avarskij jazyk. Tiflis: Upravlenie Kavkazkago Ucˇebnago Okruga. Uslar, Petr K., (1890) Ètnografija Kavkaza: Jazykoznanie. Lakskij jazyk. Tiflis: Upravlenie Kavkazkago Ucˇebnago Okruga. Uslar, Petr K., (1892) Ètnografija Kavkaza: Jazykoznanie. Xjurkilinskij jazyk. Tiflis: Upravlenie Kavkazkago Ucˇebnago Okruga. Uslar, Petr K., (1896) Ètnografija Kavkaza: Jazykoznanie. Kjurinskij jazyk. Tiflis: Upravlenie Kavkazkago Ucˇebnago Okruga. Uslar, Petr K., (1979) Ètnografija Kavkaza: Jazykoznanie. Tabasaranskij jazyk. Tbilisi: Mecniereba. van den Berg, Helma E., (1995) A grammar of Hunzib (with texts and lexicon). München: Lincom Europa. van den Berg, Helma E., (2001) Dargi folktales. Oral stories from the Caucasus with an introduction to Dargi grammar. Leiden: CNWS. van den Berg, Helma E., (forthc.) The East Caucasian language family. To appear in: special issue of Lingua. Xalilov, Majid Š., (1995) Bežtinsko-russkij slovar’. Maxacˇkala: DNC RAN. Žirkov, Lev I., (1955) Lakskij jazyk. Fonetika i morfologija. Moskva-Leningrad: AN SSSR.

Chapter 9

Where coordination meets subordination Converb constructions in Tsakhur (Daghestanian) Konstantin I. Kazenin and Yakov G. Testelets Moscow State University / Russian State University for the Humanities

1. Introduction 2. Tests for distinguishing between coordination and subordination 2.1 The morphosyntactic locus test 2.2 The Coordinate Structure Constraint 2.3 Center embedding 3. Some basics of Tsakhur grammar 4. Coordination and subordination in constructions with General Converbs 4.1 The alternation of coordination and subordination properties 4.2 The semantic conditions on coordination and subordination in constructions with converbs 5. Conclusion



The opposition between coordination and subordination is treated as one of the most basic ones by almost all theories of syntax. The syntactic representation of these structures has primarily occupied formal syntacticians, but functionalists acknowledge this fundamental distinction as well. Some modern studies have shown, however, that on a number of occasions one and the same construction exhibits either properties of coordination or properties of subordination, depending upon the semantics of its elements, the context, etc. For instance, it has been known since Ross (1986[1967]) that some English sentences with and can syntactically behave as constructions with subordination. In (1a), for example, the two clauses are linked by and, so that one would treat this sentence as coordinate. The problem is, however, that in this particular sentence, unlike in other sentences coordinated with and, relativization out of the first conjunct clause is allowed, as shown by (1b): (1) a. She read this book for two hours and did not fall asleep. b. Show me the book which Jane read for two hours and did not fall asleep.

228 Konstantin I.Kazenin and Yakov G.Testelets

However, as has also been known since Ross (1986[1967]), relativization of coordinands or constituents of coordinands is prohibited. Thus, examples like (1b) undermine the standard assumption that coordination and subordination are always distinguished by the category of the linking element: At least some subordination properties may be observed in constructions with a coordination marker. Recently cases similar to (1b) were analyzed in detail by Johannessen (1998) (see also Culicover and Jackendoff (1997), who consider other similar problems in English syntax). Analogous facts of Korean are discussed by Rudnitskaya (1998). A typological survey of such phenomena is given by Yuasa and Sadock (2002). Stassen (2000) explores a similar phenomenon observed in noun phrases. He shows that in a number of languages the comitative construction ‘X with Y’ exhibits certain coordination properties in syntax, although formally it looks like dependency or subordination.1 Once we admit that such ‘mismatches’ between syntax and morphology are possible, the question immediately arises under what conditions they can appear. The present paper attempts to make a certain step towards answering it, at least for one language. This paper focuses on converb constructions in Tsakhur, a Nakh-Daghestanian (or East Caucasian) language spoken by ca. 30,000 people in Daghestan (northeastern Caucasus, Russia) and in northern Azerbaijan (see van den Berg, this volume, for detailed discussion of coordination in Daghestanian languages). Tsakhur has verbal forms which are traditionally called converbs (Russian deepricˇastie), but can actually head both adjunct clauses and independent sentences. Therefore, if a clause with a converb head combines with some other clause, two possibilities are available: The converb clause is either an adjunct of the clause it combines with, or coordinated with it. As we will see below, some instances of Tsakhur converb constructions give evidence for the first possibility, but others point to the second one. Furthermore, it turns out that this alternation is not arbitrary, but depends upon certain semantic properties of the construction. We argue below that the construction is subordinate if (i) the converb clause and the matrix clause have the same subject, and/or (ii) the converb clause has a causal interpretation. If neither of these conditions is met, the construction exhibits syntactic properties of coordination. The paper is organised as follows. In § 2 we briefly outline syntactic tests distinguishing between coordination and subordination. In § 3 we introduce some basics of Tsakhur morphology and syntax that are necessary for understanding the data. § 4 deals with converb constructions in Tsakhur. In § 4.1 we show the alternation of

1.The term subordination is often restricted to the relation between clauses, but here we use it in a wider sense, to comprise dependent constituents more generally. We say, for instance, that in the noun phrase the sparrow on the roof, the PP on the roof is subordinate to the head sparrow.

Where coordination meets subordination: Converbs in Tsakhur (Daghestanian) 229

coordination and subordination properties in converb constructions. Finally in § 4.2 we deal with semantic conditions which determine this alternation.


Tests for distinguishing between coordination and subordination

Various diagnostics for coordination have been mentioned in the literature. Here we give only those that are applicable to Tsakhur. 2.1 The morphosyntactic locus test In a subordinate phrase, the relation between this phrase and some element outside it has to be marked once; in a coordinate phrase, it is generally marked on both coordinands. Notice the double occurrence of ’s in (2a) and the lack thereof in (2b): (2) a. the king’s and the queen’s palace b. the king(*’s) of England’s palace In Zwicky (1985), the element of a phrase to which the relation marker is attached is called the morphosyntactic locus of that phrase. The contrast between (2a) and (2b), therefore, shows that a coordinate phrase does not have a single morphosyntactic locus: there are as many loci as there are coordinands. In the case of subordination, however, the proliferation of morphosyntactic loci is not expected (normally, as shown by Zwicky, the morphosyntactic locus of a phrase including subordination is its head, although (2b) demonstrates that this is not always the case). In English, if a sentence containing a subordinate clause is embedded within a matrix clause that requires the subjunctive, the subjunctive is used in the embedded main clause, but not in its (doubly) embedded subordinate clause (for this test, see Zaliznjak and Paducˇeva 1975): (3) a. I demand that John answer the questions that Bill asks (*ask). b. I demand that John ask (*asks) questions and Bill answer (*answers) them. 2.2 The Coordinate Structure Constraint This is one of the so-called island constraints discovered by Ross (1986[1967]). Island constraints prohibit syntactic extraction, focusing and relativization of elements which belong to certain constructions. Specifically, the Coordinate

230 Konstantin I.Kazenin and Yakov G.Testelets

Structure Constraint prevents these processes from targeting a coordinand or an element belonging to a coordinand:2 (4) a. I met Mary and Peter. b. *Whom did you meet and Peter? (cf. Whom did you meet?) (5) a. I saw a photo of John and a picture of Mary. b. *Whom did you see a photo of John and a picture of? (cf. Whom did you see a picture of?) Therefore, if a certain structure blocks extraction of any of its elements, this is, ceteris paribus, a reason to treat it as an instance of coordination. 2.3 Center embedding This test, especially in languages with great freedom of word order (which Tsakhur belongs to, as we will see below), is perhaps the weakest one. It is based on the assumption that coordinands cannot be center-embedded into one another, whereas center embedding of a subordinate clause into the superordinate one is possible, cf.: (6) Last year, when she entered the University, Mary was very happy. (7) a. Last year Mary entered the University and she was very happy. b. ??Last year, and she was very happy, Mary entered the University. The problem with this test is that, at least in some languages, the order seen in (7b) is permitted with a certain intonation pattern, which marks the center-embedded conjunct as a “parenthetical” constituent (some speakers of English permit (7b) in this case as grammatical, too). Below we will see, however, that in Tsakhur the possibility of center embedding of a converb clause correlates with its “subordinate” behaviour according to (at least one of) the other diagnostics listed above.


Some basics of Tsakhur grammar

Tsakhur is an ergative language with preferred SOV word order and a highly

2.A well-known exception, discussed in detail by Williams (1978), concerns extraction of an element common to all coordinands of a coordinate phrase (so-called ‘Across-the-Board’ extraction), which is grammatical: (i) Which linguistic theory does [John like Ø] and [Johanna hate Ø]?

Where coordination meets subordination: Converbs in Tsakhur (Daghestanian)

developed system of noun-class (= gender) agreement.3 The Tsakhur verb possesses three stems: the imperfective stem, the perfective stem, and the potential stem. Class agreement on the verb is controlled by its absolutive argument. Imperfective and perfective converbs can combine with several kinds of auxiliaries, building up a developed system of analytic verb forms: eminat-e¯ ma-n Šˇuwab iwho wo-d. Aminat-erg this-atr.4cl word(4cl) say.pfc aux-4cl ‘Aminat said this word.’ b. eminat-e¯ ma-n Šˇuwab iwho ix-es. Aminat-erg this-atr.4cl word(4cl) say.pfc become.4cl-pot ‘Aminat will have said this word.’

(8) a.

SOV is the basic order, but it is in no way obligatory. As shown by (8a¢), all order permutations of S, O and V are available: (8) a¢. eminat-e¯ iwho wo-d ma-n Šˇuwab ma-n Šˇuwab eminat-e¯ iwho wo-d ma-n Šˇuwab iwho wo-d eminat-e¯ iwho wo-d eminat-e¯ ma-n Šˇuwab iwho wo-d ma-n Šˇuwab eminat-e¯ Apart from verbal forms with auxiliaries, independent sentences in Tsakhur can be headed by attributive verbal forms, used otherwise as participles. Heading an independent sentence, an attributive verbal form agrees in class with the absolutive argument: (9) eminat-e¯ ma-n Šˇuwab iwho-jn. Aminat-erg this-atr.4cl word(4cl) say.pfc-atr.4cl ‘Aminat said this word.’ Finally, independent clauses can be headed by converbs lacking an auxiliary. Thus, (10) is possible on a par with (8a): (10) eminat-e¯ ma-n Šˇuwab iwho. Aminat-erg this-atr.4cl word(4cl) say.pfc ‘Aminat said this word.’ The possibility of (10) shows that converbs have a rather non-typical distribution, as they can not only head adjunct clauses or function as a part of an analytical form, but also be used independently. In (11), a use more expected for converbs is illustrated: the clause headed by the converb is combined with a finite clause:

3.For a detailed account of Tsakhur morphology and syntax, see Kibrik (ed.) (1999).


232 Konstantin I.Kazenin and Yakov G.Testelets

(11) [eminat-e¯ ma-n Šˇuwab iwho, rasul [Aminat-erg this-atr.4cl word.4 say.pfc] Rasul.1cl arkÁGn-na. leave.pfc.1cl-atr.1cl ‘Aminat having said this word, Rasul left.’ The problem is, however, that, as it stands, (11) is structurally ambiguous. The converb clause may be analyzed as an adjunct, but this is only one of the a priori given possibilities. The other possibility is that the converb clause in (11) is an independent clause, coordinated with the following clause. One cannot reject this analysis out of hand, because Tsakhur allows clausal conjunction lacking any special marking of the conjuncts. The question which we address in the next section is which of these two analyses of constructions like (11) is correct. (Note that the structural ambiguity found in (11) does not depend upon the form of the predicate in the clause combined with the converb. In (11) it is attributive, but nothing changes if it is a form with an auxiliary, or another converb.) In order to highlight the special nature of the converbs illustrated above, we will call them General Converbs. Below they will be contrasted with Adverbial Converbs.

4. Coordination and subordination in constructions with

General Converbs 4.1 The alternation of coordination and subordination properties Using the tests discussed in § 2, we will now demonstrate that General Converb constructions may have either coordination or subordination properties. 4.1.1 The “morphosyntactic locus” test Consider the sentences in (12). Here the General Converb construction is subordinate to the verb ac’a ‘know’. The complement of this verb may be marked by the subordinator morpheme -wG. The subordinator may appear either only in the rightmost clause of the General Converb construction (12a), or in both clauses (12b): (12) a.

za-k’le ac’a-n [[rasul qarG, I-aff know.4cl-atr.4cl [[Rasul(1cl) come.pfc.1cl] [ma|hammad4 ark’Gn-wG. [Mohammed(1cl) leave.pfc.1cl-subord]] ‘I know that Rasul has come, but Mohammed has left.’

4.The vertical stroke in ma|hammad and elsewhere indicates pharyngealization.

Where coordination meets subordination: Converbs in Tsakhur (Daghestanian) 233

b. za-k’le ac’a-n [[rasul qarG-wG, I-aff know.4cl-atr.4cl [[Rasul(1cl) come.pfc.1cl-subord] [ma|hammad ark’Gn-wG. [Mohammed(1cl) leave.pfc.1cl-subord]] ‘I know that Rasul has come, but Mohammed has left.’ In (12a), therefore, the General Converb of the second clause behaves as the morphosyntactic locus of the whole sentence. This is expected if the sentence displays subordination. However, (12b) shows another option: both converbs behave as morphosyntactic loci of the embedded sentence. As stated in § 2, this is expected for coordination. Note that the option illustrated by (12b) is not available if the first converb is replaced by some other verbal form which can only head subordinate clauses. Tsakhur possesses a large number of such “subordinate” forms, each expressing its own meaning typical for adjuncts (‘time’, ‘place’, ‘purpose’, etc.). Below such forms will be called Adverbial Converbs. If an Adverbial Converb with temporal semantics occurs in place of the General Converb, the morpheme -wG cannot be attached to it: (13) za-k’le ac’a-n [[rasul qarı¯-nGa| (*-wGG), I-aff know.4cl-atr.4cl [[Rasul.1cl come.pfc.1cl-temp (*-subord)] [ma|hammad ark’Gn-wGG. [Mohammed(1cl) leave.pfc.1cl-subord]] ‘I know that after Rasul had come, Mohammed left.’ The comparison between (12) and (13) shows that variability with respect to morphosyntactic locus is available to complex sentences with General Converbs, but not to other types of complex sentences. 4.1.2 The Coordinate Structure Constraint As shown by Ljutikova (1999), Tsakhur relativization obeys island constraints. Let us now consider relativization in constructions with General Converbs. In the converb construction in (14a), relativization is blocked for NPs of both clauses, as shown by (14b–c): (14) a.

eminat-e¯ njak abG zG gGnej alja¯t’-u. Aminat-erg milk(3cl) carry.pfc.3cl I.erg bread(4cl) buy.4cl-pfc ‘Aminat brought milk, and I bought bread.’ b. *[[[eminat-e¯ njak abG zG [[[Aminat-erg milk(3cl) carry.pfc.3cl] I.erg gGnej jug-un-o-d. alja¯t® ’-in buy.pfc.4cl-atr.4cl] bread(4cl)] good-atr.4cl-aux-4cl ‘The bread which Aminat brought milk and I bought is good.’

234 Konstantin I.Kazenin and Yakov G.Testelets

c. *[[[eminat-e¯ abG zG gGnej [[[Aminat-erg carry.pfc.3cl] I.erg bread(4cl) alja¯t® ’-in njak jug-un-o-d. buy.pfc.4cl-atr.4cl] milk.3] good-atr.4cl-aux-4cl ‘The milk which Aminat brought and I bought bread is good.’ The ungrammaticality of (14c) does not actually help us to distinguish between coordination and subordination in the construction with a General Converb. Indeed, if the construction with a General Converb were subordinate, relativization of the NP ‘milk’ would not be possible either. Being subordinate, the clause headed by the converb would have the role of an adjunct rather than of a complement. However, it is known that adjunct clauses are also islands for relativization in Tsakhur. This is illustrated by (15), where the General Converb is replaced by an Adverbial Converb: (15) *[[[eminat-e¯ a-b-ı¯-nGa| zG gGnej [[[Aminat-erg carry.pfc.3cl-temp] I.erg bread(4cl) alja¯t® ’-in njak jug-un-o-d. buy.pfc.4cl-atr.4cl] milk(4cl)] good-atr.4cl-aux-4cl ‘The milk which I bought bread after Aminat brought is good.’ In contrast with (14c), the ungrammaticality of (14b) gives crucial evidence in favor of the coordinate nature of the General Converb construction. If the converb clause were subordinate, and the clause following it superordinate, we would expect that relativization out of the latter should be possible, contrary to the fact. If, however, that clause is a coordinand, we expect that neither of its elements can be relativized — exactly what we observe in (14b). However, in some other constructions with General Converbs, relativization out of the clause combined with the General Converb is possible, cf.: (16) a.

rasul, pGl [alja¯t’-u, Rasul(1cl) [money(4cl) take.4cl-pfc] šad-xa-na. glad-become.pfc.1cl-atr.1cl ‘Having got the money, Rasul became happy.’ b. hama-na [[pGl alja¯t’-u this-atr.1cl [[money(4cl) take.4cl-pfc] šad-xa-na rasul wo-r-na. glad-become.pfc.1cl-atr.1cl] Rasul(1cl) aux-1cl-atr.1cl ‘This is Rasul, who became happy, having got the money.’

Here again we see that constructions with General Converbs can have either coordination or subordination properties.

Where coordination meets subordination: Converbs in Tsakhur (Daghestanian) 235

4.1.3 Center embedding Center embedding of the General Converb clause into the clause it combines with is possible, as demonstrated by (17): (17) rasul, [eminat-e¯ ma-n Šˇuwab iwho, Rasul(1cl) [Aminat-erg this-atr.4cl word(4cl) say.pfc] ark’Gn-na. leave.pfc.1cl-atr.1cl ‘Aminat having said this word, Rasul left.’ In a number of occasions, however, the General Converb clause resists center embedding; (18) is an example: (18) *zG, [še-na solulqa ark’Gn, Šˇiga-j-lj I(1cl) [he-atr.1cl to.the.left leave.pfc.1cl] place-obl-sup aX-u. stay.1cl-pfc ‘He having gone to the left, I stayed.’ We conclude, therefore, that the Tsakhur construction with General Converbs demonstrates variability of coordination and subordination properties on all the three tests which we have included in our study. In § 4.2 we shall see that this variability is determined by the same semantic conditions. 4.2 The semantic conditions on coordination and subordination in

constructions with converbs On a number of occasions, one and the same construction with a General Converb can behave either as coordination or as subordination, depending on its interpretation. Consider the following pair of examples: ¯¯ diX Xa¯-ncˇe ark’Gn, dak-e kaRGz son(1cl) house-elat leave.pfc.1cl father-erg letter(4cl) ot’k’un. write.pfc.4cl ‘The son having left home, the father wrote a letter.’ ¯ ¯, b. dak-e [diX Xa¯-ncˇe ark’Gn, kaRGz father-erg [son(1cl) house-elat leave.pfc.1cl] letter(4cl) ot’k’un. write.pfc.4cl ‘The son having left home, the father wrote a letter’ (probably, with the request for the son to return).

(19) a.

236 Konstantin I.Kazenin and Yakov G.Testelets

In (19a) the General Converb construction is understood as reporting two independent events: The son left, and the father wrote a letter to somebody after that. In this case center embedding is impossible. However, as shown by (19b), center embedding becomes possible when a causal relation exists between the two events: the son left, therefore the father wrote a letter to him, presumably, asking him to return. In some other cases center embedding is not possible, as shown by the comparison between (18) and (20): (20) [še-na solulqa ark’Gn, zG Šˇiga-j-lj [that-atr.1cl to.the.left leave.pfc.1cl] I(1cl) place-obl-sup aX-u. stay.1cl-pfc ‘He having gone to the left, I stayed.’ Note that in (20) a causal relation between the two events is hardly imaginable. This allows us to suggest that a causal relation between two clauses of a converb construction implies the possibility of center embedding. However, center embedding is also possible without a causal relation, but only when the two clauses have the same subject, as in (21): (21) rasul, [ma-n Šˇuwab iwho, ark’Gn-na. Rasul(1cl) [this-atr.4cl word(4cl) say.pfc] leave.pfc.1cl-atr.1cl ‘Having said this word, Rasul left.’ Note that we have to treat (21) as an example of center embedding because of the nominative case of the name rasul. The verb iwho ‘said’, which heads the converb clause, requires the ergative case of its subject, as seen in (22). (22) rasul-e¯ ma-n Šˇuwab iwho. Rasul-erg this-atr.4cl word(4cl) say.pfc ‘Rasul said this word.’ Therefore, if rasul belonged to the converb clause in (21), it would be in the ergative case. Its occurrence in the absolutive shows that the NP belongs to the finite clause and thus that the whole construction is an example of center embedding. But, unlike in (19b), center embedding here does not signal that the General Converb clause expresses the reason of the action expressed by the superordinate clause: (21) can merely denote a sequence of two events having the same subject, following one another. Consider now relativization. Its possibility in General Converb constructions depends upon the same conditions as the possibility of embedding does. Consider first (14) and (16) discussed above. In (14) no causal relation between the two events, buying bread and bringing milk, is implied. In contrast, (16) presupposes a causal relation: Rasul took the money, therefore he became glad. In this way, the

Where coordination meets subordination: Converbs in Tsakhur (Daghestanian) 237

existence of a causal relation is sufficient to allow relativization in General Converb constructions. Consider now (23). Here no causal relation is observed, but the subjects of the two clauses coincide. This makes relativization possible, just as it made center embedding possible: (23) a.

rasul, [pGl alja¯t’-u, ark’Gn-na. Rasul(1cl) [money(4cl) take.4cl-pfc] leave.1cl.pfc-atr.1cl ‘Having got the money, Rasul left.’ b. hama-na [pGl alja¯t’-u ark’Gn-na this-atr.1cl [money(4cl) take.4cl-pfc] leave.1cl.pfc-atr.1cl rasul wo-r-na. Rasul(1cl) ‘This is Rasul who, having got the money, went away.’

In addition to relativization, another effect of the Coordinate Structure Constraint in Tsakhur is that wh-questions can occur within a converb clause only if there is a causal relation between the clauses, as is the case in (24): (24) a.

rasul-e¯ hiˇŠo¯n mas¯a hiwo, mašin ališ¯-u? Rasul-erg what(4cl) sell give.pfc(4cl) car(4cl) buy-pfc(4cl) lit.: ‘What did Rasul sell and bought the car’? (I.e. What did he sell so that he could buy the car?)

In (24a), it is implied that Rasul sold something in order to get the money necessary for him to purchase a car. If the situations denoted by the two clauses are logically independent, wh-questions are prohibited in the converb construction (24b): b. *rasul-e¯ hiˇŠo¯n alja‘-u, mašin ališ¯-u? Rasul-erg what(4cl) build-pfc(4cl) car(4cl) buy-pfc(4cl) lit.: ‘What did Rasul build and bought the car’? Our informants note, however, that (24b) can become acceptable if Rasul purchased the car for the money that he got as payment for having built something for somebody, i. e. if the situation denoted by the first clause has somehow caused the situation denoted by the second one. Finally, let us turn to the morphosyntactic locus test. In (12) we saw that if a construction with a General Converb occurs in a complement clause, the subordinator attaches either only to the finite verb (as expected for subordination), or both to the finite verb and to the converb (as expected for coordination). Here, however, we were not able to discover any semantic factors which govern the choice between these options. For the speakers whom we consulted, the use of the subordinator on the converb or lack thereof did not correspond to any more or less clear semantic characteristic of the construction. It must be mentioned, however, that the distribution

238 Konstantin I.Kazenin and Yakov G.Testelets

of the subordinator in Tsakhur is generally not quite clear: the subordinator, strictly speaking, is optional for the complement of any verb, and specific conditions on its use need further study.



We have seen that Tsakhur constructions with General Converbs are variable with respect to at least three properties in which coordination and subordination differ. For two of these properties, we have seen that the alternation depends on certain semantic conditions, which, remarkably, are the same for both properties. What conclusion can be drawn from these data? One possibility would be to conclude that constructions varying in their coordination/subordination properties allow us to discover semantic prototypes of both syntactic patterns. If we adopt this view, we will have to admit that General Converb constructions in Tsakhur actually exhibit structural ambiguity between coordination and subordination, that is, they may have different constituent structures with one and the same morphological marking. Alternatively, however, one could suggest that the properties on which converb constructions may differ (the possibility of center embedding, of relativization/focusing) are not, against the traditional view, to be explained on the basis of major syntactic oppositions such as the opposition between coordination and subordination. They could rather be derived directly from the semantic linkage between clauses, irrespectively of morphosyntactic marking. If we pursue this line of explanation, we will not need to acknowledge syntactic ambiguity of General Converb constructions, as the alternation of the above-mentioned properties will not depend upon the syntactic structure. At the present point, we do not see strong reasons for or against taking either of these two views. Such reasons could appear, however, after further study of General Converb constructions in Tsakhur and comparison between constructions with similar coordination/subordination variability in other languages.

Abbreviations 1cl 3cl 4cl aff atr

1st agreement class (male human) 3rd agreement class (animals and some inanimate nouns) 4th agreement class (inanimate) affective case attributive

elat pfc pot subord sup temp

elative case perfective stem = perfective General Converb potential stem subordinator superessive case temporal converb

Where coordination meets subordination: Converbs in Tsakhur (Daghestanian) 239

References Culicover, Peter W., and Ray Jackendoff. 1997. “Semantic subordination despite syntactic coordination.” Linguistic Inquiry 28: 195–217. Johannessen, Janne B. 1998. Coordination. New York: Oxford University Press. Kibrik, Aleksandr E. (ed.). 1999. Èlementy grammatiki caxurskogo jazyka v tipologicˇeskom osvešcˇenii (= Elements of Tsakhur grammar in a typological perspective). Moskva: Nasledie. Ljutikova, Ekaterina A. 1999. “Otnositel’noe predloženie” (= The relative clause). In Kibrik (ed.) 1999, 461–480. Ross, John R. 1986[1967]. Infinite syntax! Norwood: Ablex. [Published version of Ross, John R. 1967. Constraints on variables in syntax. Ph.D. Dissertation, MIT.] Rudnitskaya, Elena L. 1998. “Syntactic properties of the Altaic coordination construction in Korean.” Sprachtypologie und Universalienforschung 51: 179–198. Stassen, Leon. 2000. “AND-languages and WITH-languages.” Linguistic Typology 4: 1–54. van den Berg, Helma. This volume. “Coordinating constructions in Daghestanian languages.” Williams, Edwin. 1978. “Across-the-Board rule application.” Linguistic Inquiry 9: 31–43. Yuasa, Etsuyo, and Jerry M. Sadock. 2002. “Pseudo-subordination: a mismatch between syntax and semantics.” Journal of Linguistics 38: 87–111. Zaliznjak, Andrej A., and Elena V. Paducˇeva. 1975. “K tipologii otnositel’nogo predloženija” (= Toward a typology of the relative clause). Semiotika i Informatika 6: 51–101. Zwicky, Arnold. 1985. “Heads.” Journal of Linguistics 21: 109–132.

Chapter 10

Coordination in Chechen* Liane Jeschull University of Leipzig

1. Introduction 2. Conjunction 2.1 Conjunction of NPs and non-finite clauses 2.2 Conjunction of finite clauses 2.3 Other uses of the cl*tic ’a 3. Disjunction 4. Negative coordination 5. Adversative coordination 6. Discussion



Grammars of Chechen unfortunately provide unsatisfyingly little information on coordination beyond a few conjunctions and scarce examples of their use. However, as early as in 1940, Jakovlev set up a hypothesis about the development of coordination and subordination in Chechen similar to what Mithun (1988) showed is a common phenomenon in coordination throughout the world’s languages. From a diachronic perspective, Jakovlev (1940: 176ff.) suggested that coordination and subordination without a conjunction are older than with a conjunction. According to Mithun, coordination can be expressed by two major strategies: by intonation alone or by the use of overt coordinators. Both authors claim that overtly grammaticized coordination has emerged in line with the rise of literacy and exposure to written languages. Thus in writing, and according to Jakovlev also in epic and singing traditions, conjunctions take the place of intonation in oral speech. Jakovlev furthermore concluded that Chechen takes an intermediate position in the development from

*I am grateful to Zamir Yushaev, Fatima Almursaeva, Zarina Molochieva, Ramazi Petirova, Madina Salazhieva and Elina Sambieva for providing me with valuable insights into their native language and some of the data in this paper and to Martin Haspelmath and Johanna Nichols for their helpful comments.

242 Liane Jeschull

intonational to fully grammaticized coordination and subordination. He assumed that subordinating conjunctions have developed rather recently, paralleling those in Russian. Consequently Chechen makes use of both subordinating verb forms, on the one hand, and conjunctions, on the other hand, in order to express subordination. Coordination in Chechen still deserves a more thorough analysis, and this will be the aim of the present paper. The study will be based on literary Chechen; it will illustrate the syntactic structures and lexical means available to express coordination, and it will trace their origins. Chechen belongs to the Nakh group of the Northeast Caucasian or NakhDaghestanian language family. It is mainly spoken in Chechnya and in a part of Daghestan (Russian Federation). Literary Chechen is based on the variety spoken in central Chechnya, around the capital Grozny, but there is considerable dialectal variation in oral speech. Chechen has been a written language with various alphabets since the early 1920s. The current alphabet, which was introduced in the 1930s, is based on Cyrillic. Recently, Russian has had a major impact on Chechen as a source of borrowings, as education, administration and communication with members of other nationalities have mostly been in the Russian language. Speakers of the Aeqqi dialect live in Daghestan and are illiterate in their native language because of the lack of instruction in school. Therefore they use Russian exclusively in writing and partially in oral communication. Chechen morphosyntax is characterized by a system of nominal gender classes marked on agreeing verbs, adjectives, numerals and adverbs,1 a rich verbal tenseaspect-mood system and agglutinative case-marking including eight basic and several derived nominal cases. The alignment of verb arguments in Chechen is ergative both in case marking and agreement. The basic word order in a simple sentence is SOV. Modifiers such as adjectives and relative clauses precede their head nouns. Subordinate relationships are chiefly expressed by the masdar (or “verbal noun” in Nichols’ (1994) terms), infinitives, participles, converbs and subordinate moods. For a more detailed introduction to Chechen, see Nichols (1994), Maciev (1961), Dešeriev (1967), Dešerieva (1999) and Aliroev (1999). Coordination in Chechen, as mentioned above, is expressed by intonation alone or by the use of overt conjunctions or coordinators following Haspelmath’s (to appear) terminology.2 Conjunction is asyndetic or, if coordinators are present, bisyndetic. Bisyndetic conjunction prevails in written contexts. Conjunctive coordinators are always postpositive. Two different coordinators are available for

1.The four classes are called B, D, J, V, using the principal exponents of gender agreement marking. 2.Following Haspelmath (to appear), coordinating conjunctions will be called coordinators. The term conjunction names one semantic type of coordination.

Coordination in Chechen 243

conjoining NPs and non-finite clauses, but only one of them is also used for conjoining finite clauses. Conjunction will be discussed in § 2: conjunction of NPs and non-finite clauses in § 2.1, and conjunction of finite clauses in § 2.2. The various other functions of one of the conjunctive coordinators will be illustrated in § 2.3. Disjunction in Chechen can be monosyndetic or bisyndetic. Bisyndetic disjunction has contrastive meaning. The single disjunctive coordinator, unlike the conjunctive ones, is prepositive. Disjunction will be discussed in § 3. Negative coordination shows properties of both conjunction and disjunction and will be illustrated in § 4. Adversative coordination, however, is always monosyndetic and prepositive. It differentiates two basic types and coordinators. Adversative coordination will be treated in § 5. The possible origin of coordinating structures and the coordinators found in Chechen will be discussed in § 6. Table 1 gives a summary of the coordination types and coordinators available in Chechen. Table 1.Overview of coordinators in Chechen Coordination type Conjunction

Coordinators NPs and non-finite clauses

A B (juxtaposition) A-(i)i B-(i)i A ’a B ’a

Finite clauses

A B (juxtaposition) A ’a B ’a


A ja B ja A ja B

Negative coordination

A ’a B ’a A ’a ja B ’a ja A ’a ja B ’a

Adversative coordination

A amma B A tq’a B

Except where elicited, examples are drawn from El’mursaev (1964), Noxchiin fol’klor, a collection of Chechen folklore and fairy tales (indicated as NF), and the Chechen text of an originally Russian novel, Pushkin (1961), Kapitanan jow (indicated as KJ). The transcription from the Cyrillic alphabet of Chechen follows that developed by Johanna Nichols, see

244 Liane Jeschull



2.1 Conjunction of NPs and non-finite clauses 2.1.1 Juxtaposition NPs in Chechen can be conjoined with or without coordinator. Asyndetic conjunction of nouns is potentially ambiguous as it could also be N-N compounding. Examples of this type combine two semantically related entities to form a conceptual and intonational unit as described by Mithun (1988: 332f.). In Chechen examples like daa-naana (‘mother-father’ = parents) in (1) and de-byysa (‘day-night’ = 24 hours) in (2), there is no intonation break between the two constituents and, in line with the general stress patterns of the language, stress is placed on the first syllable of the first constituent: (1) daa-s-naana-s ca v-yysh v-itina hara shi’, father-erg-mother-erg neg V-kill.scv V-leave.perf this two dov ’a d-ina. (NF p.18) scolding cl D-make.acv ‘The parents didn’t kill these two, but scolded them.’ (2) ocu buussiehw so Simbirsk-ie qeechira, […] that.obl at.night I Simbirsk-all arrive.wp cwa de-byysa txaesh d-aaqqa d-iezash one day-night ourselves.excl D-spend.inf D-need.scv j-olchu. (KJ p. 10) J-be.prespart.obl ‘That night I arrived in Simbirsk, where we had to spend one day and night.’ In these cases the coordinands behave like true compounds in that they form a unit that is inseparable not only intonationally, but also semantically, as in dog-oila (‘heart-thought’ = intention, wish, opinion), dov-dosh (‘quarrel-word’ = argument), ghan-naabarsh (‘’ = doze) and others. On the other hand, the individual constituents of examples like daa-naana (‘father-mother’ = parents) in (1) and de-byysa (‘day-night’ = 24 hours) in (2) can also be conjoined syntactically either by mere juxtaposition or by an overt coordinator. In oral speech, juxtaposition employs Mithun’s “comma intonation”, which is realized by an intonation break (1988: 332f.). In Chechen this results in stress on the first syllable of each of the coordinands. Juxtaposition thus names the constituents as separate entities, e.g. ‘mother and father’ rather than ‘parents’ in (3). (3) daa-s, naana-s niaq’ dika b-ira san. father-erg mother-erg way good B-make.wp I.gen ‘Father and mother wished me a pleasant journey.’

(KJ p.10)

Coordination in Chechen 245

Conjunction by “comma intonation” occurs with NPs that denote separate entities and properties regardless of whether they are conceptually related or not. (4) gives an example where two proper names are conjoined: (4) […] shi vasha xilla, Aahwmad, Hwaasan c’erash two brother be.perf Aahwmad Hwaasan ’a j-olush, […]. (NF p.40) cl J-be.scv ‘[…] there were two brothers, whose names were Aahwmad and Hwaasan […].’ Although common in oral communication, juxtaposition of NPs is rather hard to find in written texts. In writing, it seems to be more typical of conjoined adjectives (cf. (5) and (6)), converbs (cf. (7)) and participles (cf. (8)): (5) […] sirla, d-aqqii bwaerg-ash chexka lielash bright D-big eye-pl fast move.scv d-ara cynan. D-be.impf he.gen ‘[…] his bright, big eyes were moving quickly.’

(KJ p.24)

(6) kiet’ahw laettash shira chyynan j-oqqa top gira outside stand.scv old cast.iron J-big rifle see.wp suuna. me.dat ‘I saw an old, cast-iron cannon standing outside.’

(KJ p. 31)

(7) […] iza lovzuosh, hwoestush xilla daa-s, naana-s. (NF p.16) he play.scv caress.scv be.perf father-erg mother-erg ‘Father and mother were playing with him and caressing him.’ (8) i ginchu, hilla d-olchu he see.past.part.obl cleverness D-be.prespart.obl aaxarxuochuo, diexar d-ina. (NF p.106) farmer.erg request D-make.perf ‘The farmer, who had seen him and was clever, made a request.’ 2.1.2 The coordinator -(i)i The difference in coordination between conjunction of conceptually related constituents and spontaneous conjunction is mirrored in the distribution of the two conjoining coordinators available in Chechen: the suffix -i(i) and the cl*tic ’a. The first is restricted to the coordination of NPs and non-finite clauses and occurs only where the conjuncts are semantically related in some particular sense. The latter coordinator is used for enumerating entities which may be semantically unrelated. Both these coordinators follow their coordinands and attach to each of the conjuncts.

246 Liane Jeschull

The coordinator -(i)i usually conjoins not more than two coordinands, which often form a conceptual unit. It attaches to inflected NPs and non-finite verb forms. Conjunction by use of the coordinator -(i)i thus behaves just like the compoundlike type of conjunction described above in § 2.1.1 in that it has the same semantics and selects the same types of coordinands. Some coordinands are even used alternately with and without the coordinator, compare daa-naana in (1) above with daa-i-naan-ii, as in (9) and (10) below, both of which mean ‘parents’: (9) […] san daa-i-naan-ii mila d-u […]? I.gen father-and-mother-and who D-be.pres ‘[She wanted to know] who my parents are […].’

(KJ p.37)

(10) Zho*ra-Baaba-s hwiexa-r d-ina deen-ii, Zho*ra-Baaba-erg teach-masd D-make.perf father.dat-and naann-ii […]. (NF p. 18) mother.dat-and ‘[The witch] Zho*ra-Baaba gave the parents an instruction […].’ Conjunction by the coordinator -(i)i is not limited to nouns. In (11) two antonymous adverbs are conjoined: (11) occul d-olu jalta xaerzhina, dwaa-i-swaa-i that.much D-be.prespart grain choose.acv there-and-here-and q’aasto-r qoochush-d-aalur d-oocush xala divide-masd reach.scv-D-can.make.fut D-be.not.scv difficult ghullaq d-u. (NF p. 58) matter D-be.pres ‘Having chosen that much grain, it is a difficult and unattainable matter to divide them from one another [lit. here and there].’ Nor is the number of coordinands strictly limited to two, cf. (12): (12) so guobaeqqina dwaa-swaa hwoezhush v-ara, I turn.around.acv V-be.impf byrsa-chu ghopa-n baastionashsh-ii, bwovv-ii, vaall-ii threatening-obl fortress-gen tower-and rampart-and gan dagahw; […]. (KJ p.31) see.inf in.mind ‘Having turned around, I was looking around expecting to see the bastions of a threatening fortress, a tower and a rampart […].’ Nor do the coordinands necessarily have to form a conceptual unit. More precisely, this kind of conjunction indicates that the coordinands participate jointly in the same event, e.g. the sight of the white fields and the clear sky together denotes the theme of the seeing event in (13), and the bottle and the glass together are taken out

Coordination in Chechen 247

of the cupboard in (14): (13) suuna humma.’a ca go k’ain aarienashsh-ii me.dat nothing neg see.pres white j-eqna stigl-ii bien. J-clear sky-and except ‘I don’t see anything but white fields and clear sky.’

(KJ p.17)

(14) xozjain-a phweegha-sh chohw j-olchu shkaaf-a host-erg china-pl inside J-be.prespart.obl cupboard-gen chyra shish-ii, stak-ii swa-iicira, […]. (KJ p.24) from.inside bottle-and glass-and here-take.wp ‘The host took a bottle and glass out of the china cupboard […].’ The suffix -(i)i conjoins not only participants of the same event, but also properties as denoted by adjectives, cf. (15): (15) […] kiexat shyyr-ii dik-ii xila-r-na, […]. paper thick-and good-and be-masd-dat ‘[…] because the paper was thick and good, […].’

(KJ p.6)

Moreover, it combines clauses headed by nominalized verb forms like infinitives (cf. (16)) and masdars (cf. (17), (18) and (19)). Since these describe simultaneous parts of one event, the verb forms that -(i)i attaches to are sometimes even identical, as with the infinitives in (16) and the masdars in (17): (16) iza dogovor-ca suuna francuzski ’a, nemecki ’a he contract-instr me.dat French and German and mott waamuo-i, d-errigie.’a wilmanash waamuo-i language teach.inf-and D-all teach.inf-and d-ieqar d-olush v-ollushiehw, […]. (KJ p.5) D-duty D-have.scv V-be.pastcond ‘Although by contract it was his duty to teach me French and German and to teach me all the sciences, […]’ (17) qiera-d-ella xila-r-uo-i, hoqu ghullaq-ax ca frighten-D-can.acv be-masd-erg-and this.obl matter-lat neg qietash xila-r-uo-i dwaleecira so. (KJ p.22) understand.scv be-masd-erg-and seize.wp I ‘Fear and lack of understanding of this matter seized me.’ (18) zhima xila-r-uo-i waalam-uo-i so too-v-a-r young be-masd-erg-and nature-erg-and I healthy-V-become-masd sixdira. (KJ p. 55) speed.up.wp ‘Youth and nature sped up the process of my recovering.’

248 Liane Jeschull

(19) hwaqarchii j-aazhuo v-uoxyytur v-u as hwo, J-tend.inf V-send.fut V-be.pres I.erg you baq’-d-olu ghullaq ahw lachq’uo-r-n-ii, true-D-be.prespart matter you.erg hide-masd-dat-and zhimchu stag-ie vuon ghullaq d-aita-r-n-ii. (KJ p.59) young.obl man-all bad matter D-let.make-masd-dat-and ‘I’ll send you to keep watch over the pigs if you hide the truth and allow a young man to do bad things.’ In all these cases, the coordinands relate to one event. Therefore, the suffix -(i)i never conjoins finite clauses. Nominal coordinands conjoined by -(i)i display some interesting characteristics: they behave like compounds in two respects. First, they are often of D gender, regardless of the gender-class membership of their constituents. This is typical of compounds in Chechen. The gender class is shown by the agreeing verbs d-iexash (‘D-living’) in (20) and (21) and d-u (‘D-are’) in (21): (20) vorx k’ant ’a v-olush d-iexash wash xilla seven son cl V-be.scv D-live.scv remain.scv be.perf cwa daa-i, naan-ii. one father-and mother-and ‘There were a father and mother who had seven sons.’

(NF p.54)

(21) […] miarza-chu biezam-ca d-iexash-wash d-u sweet-obl love-instr D-live.scv-remain.scv D-be.pres boxu Chinghazz-ii, cynan mair-ii. (NF p.34) speak.pres Chinghaz-and she.gen husband-and ‘[…] Chinghaz and her husband, people say, are living in sweet love.’ Elicitation of the present tense form of the verb ‘be’ with coordinands of this type reveals that speakers differ in their choice of gender marker. Across speakers and examples, the overall trend is to the D gender class, cf. (22) through (24):3 (22) a.

de-i byys-ii d-u. day-and night-and D-be.pres b. de-i byys-ii j-u. day-and night-and J-be.pres

(23) a.

tyx-ii burch-ii d-u. salt-and pepper-and D-be.pres

3.In the default case of coordination, verbs will agree with the NP closest to them if the coordinands differ in their gender-class membership. Thus the verb agrees with byysa (‘night’) in (22b), with burch (‘pepper’) in (23b) and with maaxa (‘needle’) in (24b).

Coordination in Chechen 249

b. tyx-ii burch-ii j-u. salt-and pepper-and J-be.pres (24) a.

tai maax-ii d-u. thread needle-and D-be.pres b. tai maax-ii b-u. thread needle-and B-be.pres

The same tendency can be observed with asyndetic coordinands and lexicalized compounds, cf. mott-ghaiba (‘bed-pillow’ = bedding) in (25) and jyhw-sibat (‘faceappearance’ = outer appearance) in (26): (25) a.

mott-ghaiba d-u. bed-pillow D-be.pres


(26) a.

jyhw-sibat d-u. b. face-appearance D-be.pres

mott-ghaiba b-u. bed-pillow B-be.pres jyhw-sibat j-u. face-appearance J-be.pres

Other examples, like zhizhig-galnash (‘meat-pasta’ = Chechen national dish) in (27) and koch-xecha (‘shirt-pants’ = clothes) in (28), indisputably demand D gender, regardless of the gender class of the coordinands: (27) zhizhig-galnash d-u. meat-pasta D-be.pres (28) koch-xecha d-u. shirt-pants D-be.pres A second way in which -(i)i constructions are reminiscent of compounds is the possibility for the suffix -(i)i to occur monosyndetically and to attach to the righthand constituent only, while keeping the same meaning as in bisyndesis, cf. (29): (29) hwan daa-naan-ii d-olchu ’a you.gen father-mother-and D-be.prespart.obl cl ghuoi, […]. go-prespart ‘Let’s go to your parents’ place […]’

(KJ p. 61)

The monosyndetic option of the conjoining suffix -(i)i was also reported by Jakovlev (1940:164ff.). The following elicited examples reveal that some speakers prefer monosyndesis, but others prefer bisyndesis with exactly the same coordinands: (30) a.

daa-naan-ii father-mother-and


daa-i-naan-ii father-and-mother-and

(31) a.

maira-zud-ii husband-wife-and


mair-ii-zud-ii husband-and-wife-and

250 Liane Jeschull

(32) a.

de-byys-ii day-night-and


de-i-byys-ii day-and-night-and

(33) a.

urs-mawar-ii knife-fork-and


urs-ii-mawar-ii knife-and-fork-and

Similarly, there are other morphological environments where the rightmost constituent alone may be affected by morphological processes as in the derivation in (34) and inflection in (35b). In (34), the adverb derived from de-byysa (‘daynight’) shows the adverbial form of only the second constituent, but not the first one. In example (35b) it is again only the second constituent that is affected by plural formation. (34) guttara.’a cwana hwool-ie-hw v-u, hara pxoe’algha still one.obl situation-all-in V-be.pres this fifth de-busa d-u hara mott secna v-olu. (KJ p. 53) day-night.adv D-be.pres this speech stop.acv V-be.prespart ‘He is still in the same situation, this is the fifth day and night without speech.’ (35) a.

de-byysa day-night ‘day and night’


de-byysanash ‘days and nights’

In sum, coordinands that form a conceptual unit can be combined asyndetically without an intonation break, by the bisyndetic coordinator … -(i)i … -(i)i and by monosyndetic … -(i)i. The three structures may coexist with precisely the same conjuncts. The original way of conjoining constituents by use of the suffix -(i)i is bisyndetic. Yet the tendencies toward monosyndesis of the coordinator and toward the D gender class suggest that coordinands of this type are coming to behave rather like compounds. As a result, there is an ambiguity between coordination and compounding in Chechen. 2.1.3 The coordinator ’a In contrast to the suffix -(i)i, the cl*tic ’a conjoins NPs that denote separate entities. Examples (36) and (37) illustrate what the difference is, when the same coordinands are conjoined. (36) talks about the grain as a whole, but (37) about the separate sorts of grain: (36) […] vovshax-’iina, duqqa.’a k’a-i, muqq-ii, borcc-ii each.other-mix.acv great.many wheat-and barley-and millet-and xilla. (NF p. 58) be.perf ‘[…] mixed with each other, there was a lot of wheat, barley and millet.’

Coordination in Chechen

(37) qaana wyyranna maalx cwaqietalie k’a ’a, muq ’a borc tomorrow in.morning sun rise.until wheat and barley and millet ’a hoora shaa-shaa q’aastosh gaaliash chu ahwa and every.obl self-self divide.scv in you.erg d-oxkahw san jow Maelxa-Aesin hwan j-u. (NF p. 58) D-put.prescond I.gen daughter Maelxa-Aesin you.gen J-be.pres ‘If, dividing the wheat, the barley and the millet each individually, you put them into sacks by the time the sun rises tomorrow morning, my daughter Maelxa-Aesin will be yours.’ In more general terms, the coordinator ’a conjoins entities and properties of separate parts of events, which take place independently of each other, whether simultaneously or not. The cl*tic can attach to absolutive objects, as in (38); oblique objects, as in (39); adjectives and adverbs, as in (40) and (41); as well as converbs and masdars, as in (42) and (43). When NPs are conjoined, an agreeing verb shows the gender class of the NP closest to it. (38) laqa hwala-hwaezhna-chu suuna waerzha mazh ’a, upward up-look.pastpart-obl I.dat black beard and q’eegash shi bwaerg ’a gira. shine.scv two eye and see.wp ‘Looking up above, I saw a black beard and two shining eyes.’

(KJ p.23)

(39) diesha-r-ie ’a, q’amial-ie ’a san b-olu read-masd-all and conversation-all and I.gen B-be.prespart biezam dwa-b-eelira. (KJ p.63) love away-B-go.wp ‘My love for reading and conversation departed.’ (40) uuram-ash gotta ’a, sittina ’a d-ara. street-pl narrow and bend.pastpart and D-be.impf ‘The streets were narrow and bent.’

(KJ p.31)

(41) peetarsh […] geenna cwana aaghoonga-hwa ’a, aarahw inn further one.obl side.all-to and outside ’a massuo.’a jyrtana geennahw ’a j-ara, […]. (KJ p.25) and all village.dat far and J-be.impf ‘The inn […] was further off [the road], on the outside and far from any village […].’ (42) hara moxk suuna dika b-evzash b-u, […] hwaala ’a, this country I.dat good B-know.scv B-be.pres up and ohwa ’a chiaq-vaalla-lc so t’exula lialla ’a, down and through-go-until I across go.acv and


252 Liane Jeschull

chiaq-v-aella ’a hara b-olush. (KJ p.20) through-V-go.acv and this B-be.scv ‘I know this country very well […] because I went across and through until I had gone through it up and down.’ (43) toe’ur d-u cynan mexkarii b-olchu suffice.fut D-be.pres he.gen B-be.prespart.obl ida-r ’a, qoqiin baannash chu hwiezha-r ’a […]. (KJ p.7) run-masd and in look-masd and ‘He has long enough gone after the girls and looked into pigeons’ nests […].’ With respect to its placement, the coordinator ’a thus shows the same distribution as the coordinator -(i)i, when combining NPs and non-finite clauses. Yet the two differ in semantic terms. The constituents conjoined by the suffix -(i)i are jointly involved in the same event, whereas the constituents conjoined by the cl*tic ’a are seen as involved in an event separately. As ’a can, moreover, combine separate events themselves, it is also used for conjunction of finite clauses (cf. §2.2 below). Consequently, coordination by ’a is equivalent to asyndetic coordination by “comma intonation”, which was discussed in §2.1.1 above. 2.2 Conjunction of finite clauses 2.2.1 Juxtaposition Finite clauses can also be conjoined with and without a coordinator. Asyndetic coordination of finite clauses juxtaposes different events, cf. (44)–(46): (44) mox c’iiza b-yylira; darc hwovziira. wind howl.inf B-start.wp blizzard turn.around.wp ‘The wind started to howl, and the blizzard turned around.’

(KJ p.18)

(45) cwana aaghoorhwa laettash masiax c’a d-ara, one.obl sidewards stand.scv some house D-be.impf uuram-ash-ka-xula qiarstash lielash masiax kootam j-ara. (KJ p.34) street-pl-all-across stroll.scv go.scv some hen J-be.impf ‘To one side, there were some houses, and some hens were strolling across the streets.’ (46) hweeshii hwoega hwoezhush wiesh b-u, you.all wait.scv remain.scv B-be.pres borsh shiel-lur j-u. soup cold-get.fut J-be.pres ‘The guests keep waiting for you, and the soup will get cold.’

(KJ p.36)

Coordination in Chechen 253

2.2.2 The coordinator ’a Conjunction of finite clauses with an overt coordinator uses the cl*tic ’a again, as it combines different events. It does not, however, make use of the coordinator -(i)i. When conjoining finite clauses, ’a is bisyndetic and postpositive as well. It cl*ticizes onto the element immediately to the left of the lexical verb when independent clauses are conjoined; to the subject when the finite clauses are contrasted in some sense; and to the verb itself when finite subordinate clauses are conjoined. In most cases of conjunction of finite clauses, ’a attaches to the complements of the verbs. Hence the coordinator cl*ticizes onto the direct objects of the transitive verbs in (47) and to the adjectival and adverbial complements in (48): (47) […] peetar-ie juxa-v-axaniehw chai ’a mer inn-all back-V-GO.pastcond tea and drink.fut d-ara, byysa ’a j-oaqq-ur j-ara, […]. (KJ p.19) D-be.impf night and J-spend-fut J-be.impf ‘[…] if we had returned to the inn, we could have drunk tea and spent the night […].’ (48) as sialxana wovdalalla ’a lieliira, hwuuna xala I.erg yesterday foolishly and behave.wp you.dat difficult ’a xietiitira. and let.seem.wp ‘Yesterday I behaved like a fool and offended you.’

(KJ p.16)

When complex verbs are conjoined, the coordinator attaches to the bound element preceding the verb root, rather than the complement of the entire complex verb. Thus in prefix verbs like t’eqaacha (‘reach’) in (49), it attaches to the prefix. Even in frozen complex verbs such as taavala (‘recover’), appearing in (50) in the form too v-elira, the coordinator intervenes between the bound preverbal morpheme and the verb root: (49) so hwuuna t’e ’a qeechi, hwo gan ’a gina suuna. I you.dat on and reach.perf you see and see.perf I-dat ‘I reached you, and I saw you.’

(NF p.46)

(50) kestta so too ’a v-elira, sain soon I healthy and V-become.wp myself.gen husam-ie dwa-v-axa nicq’ ’a qeechira. (KJ p.56) lodging-all away-V-go.inf power and receive.wp ‘Soon I recovered and got strong enough to return to my lodging.’ In the case of intransitive verbs and verbs with oblique objects, the verb stem reduplicates in order to host the cl*tic, cf. (49) above, (51) and (52) below:

254 Liane Jeschull

(51) hoor wyyranna as j-ieshaa ’a j-oeshura, perevod-ash every morning I.erg J-read and J-read.impf translation-pl j-an ’a weemara so, tq’a cwacca J-make.inf and study.impf I but/and one.each j-olchu zaaman chowh stixotvoreniash J-be.prespart.obl time in jaz-j-a ’a weemara so. (KJ p.40) write-J-do.inf and study.impf I ‘Every morning I read and I practiced translating, and in every remaining moment I practiced writing poems.’ (52) […] deshnasha bwaerg-ash ’a b-illira san, eye-pl and B-open.wp I.gen duqqa.’a ghullaq-iix qietaa ’a qiitira so. (KJ p.51) great.many understand and understand.wp I ‘[…] the words opened my eyes and I understood a great many things.’ The cl*tic ’a attaches to subjects when it conjoins finite clauses and they describe parallel situations, cf. (53) and (54): (53) massuo.’a aaghoor hwoettina booda ’a all sideward stand.up.pastpart darkness and b-ara, mexan darc ’a d-ara. B-be.impf wind.gen blizzard and D-impf ‘All around, there was darkness and there was a blizzard.’

(KJ p.18)

(54) govr-ash ’a neexan j-u, ghuomat-ash ’a hwan horse-pl and foreign J-be.pres harness-pl and you.gen j-aac, […] (KJ p.20) J-be.not.pres ‘The horses belong to a stranger, and the harness is not yours […].’ Only when finite subordinate clauses are conjoined does the coordinator cl*ticize onto the finite verb itself. It attaches to the subjunctive verb forms in (55), to the present conditional in (56) and to the indicatives of the conjoined subordinate clauses in (57): (55) hun eer d-u caara, shain k’ant-a what say.fut D-be.pres they.erg themselves.gen boy-erg moliila ’a, i lovzush v-yyla ’a xi’cha. (KJ p.17) drink.subj and he play.scv V-be.subj and find.out.when ‘What will they say if they find out that their son drinks and he plays?’

Coordination in Chechen 255

(56) cynan, v-oqqa-stag, hwuuna baala b-aac, as iza that.gen V-old-man you.dat worry B-be.not.pres I.erg it mala-r-x j-oxkahw ’a, ca j-oxkahw ’a. (KJ p.26) drink-masd-lat J-sell.prescond and neg J-sell.prescond and ‘That, old man, is no business of yours whether I booze it away or I don’t.’ (57) […] iza hinca.’a qaacha ca veellera aerruu aaghoo he still reach.inf neg finish.impf left side mylxa j-u ’a aettuu aaghoo mylxa j-u ’a which J-be.pres and right side which J-be.pres and caarna xa’a-r-ie. (KJ p.40) they.dat know-masd-all ‘[…] he hasn’t yet succeeded in teaching them where left is and where right is.’ 2.2.3 Other conjunctive coordinators For conjoining finite clauses, Chechen has still more coordinators. Most notable are those that involve a bisyndetic structure. In contrast to the obligatorily bisyndetic coordinator ’a, these convey contrastive meaning, for example cq’a … t’aaqqa … (‘first … then …’) as in (58) and cq’a … cq’a … (‘now … now …’) as in (59): (58) shiak xila-r-ca cq’a xozjain-ie, t’aaqqa vozhati-ga suspicion be-masd-instr first host-all then guide-all hwoezhura iza. (KJ p.25) look.impf he ‘He was looking suspiciously first at the host, then at the guide.’ (59) […] ma ch’oogha satiisira-qa as hwo sema how strong long.wp-cl I.erg you awake ga-r-ie, cq’a j-istxilla-lc, cq’a dagara ’a see-masd-all once J-start.speak-until once from.heart cl d-iicca-lc, […]. (NF p. 53) D-speak-until ‘How badly I longed to see you in reality, so that you now start speaking, now talk from the heart […].’ The prepositive placement of the coordinators and the contrastive meaning of bisyndesis in these types of conjunction mirror precisely the corresponding structures in Russian, cf. snacˇala … potom … (‘first … then …’) and to … to … (‘now … now …’), and strongly suggest that they were calqued from Russian.

256 Liane Jeschull

2.3 Other uses of the cl*tic ’a The cl*tic ’a fulfills a number of other functions besides conjunction. In the context of coordination, it is also an obligatory part of any structure in negative coordination and can co-occur with the disjunctive coordinator (cf. §4 below for treatment). Moreover, the cl*tic is also used as a focus particle meaning ‘too, also’, cf. (60) and (61): (60) kestta deriggie.’a c’iinuo naabaran xur-t’ur soon all house.erg sleep.gen snore d-ira, v-elcha sanna, suuna ’a naab qiitira. (KJ p.25) D-make.wp V-dead like I.dat and sleep encounter.wp ‘Soon the entire house snored in sleep, and I too fell asleep as if dead.’ (61) so ohwa-v-yyzhira, mott ’a secira san. I down-V-fall.wp speech and stop.wp I.gen ‘I fell down and also lost consciousness.’

(KJ p.52)

Furthermore, ’a occurs in subordinate structures. In (62) it appears in a finite subordinate clause, where it attaches to the finite verb form; in (63) in a non-finite subordinate clause headed by a converb, where it attaches to the direct object of the verb: (62) eeluo shien jalxoo v-axiitina, i shi’ mila prince.erg himself.gen servant V-let.go.perf this two who v-u ’a hwazha […]. V-be.pres cl see.inf ‘The prince sent his servant to see who these two are […].’

(NF p.43)

(63) c’a ’a d-axana, Zho*ra-Baaba-s shaina home cl D-go.acv Zho*ra-Baaba-erg themselves.dat d-ina hwiexa-r ma.darra shienan k’ant-ie D-make.pastpart teach-masd frankly themselves.gen son-all dwa-aella qaara. (NF p.18) out-say.perf these.obl ‘After going home, they frankly spoke out to their son about the instructions [the witch] Zho*ra-Baaba had given them.’ Further, the cl*tic ’a occurs in a number of lexicalized forms, cf. shii ’a (‘both’) vs. shi’ (‘two’), humma ’a (‘nothing’) vs. huma (‘thing’), duqqa ’a (‘great many’) vs. duqa (‘many, much’), guttara ’a (‘still’) vs. guttar (‘always’), hinca ’a (‘still’) vs. hinca (‘now’), uggarie ’a (‘most of all’) vs. uggar (‘most’) and others. Peterson (2001) gives an account of the placement of the counterpart to the cl*tic ’a in Ingush, the language to which Chechen is most closely related, although he does not analyze coordination proper. This reveals that the Ingush cl*tic shares the distributional properties of Chechen ’a in the structures discussed in this

Coordination in Chechen 257

section, so the same might be predicted for coordination too. Interestingly, van den Berg (this volume) makes similar observations with regard to Daghestanian languages, according to which conjunctive coordinators also occur as emphatic particles denoting ‘also’, ‘too’, ‘even’ and in subordinate clauses. Finally, conjunctive coordinators can also combine with disjunctive coordinators in a number of Daghestanian languages. This is similar to the observation that the conjunctive coordinator ’a can appear together with the disjunctive coordinator in negative coordination in Chechen.



There is one common coordinator that expresses both disjunction of NPs and disjunction of finite and non-finite clauses: the particle ja. Unlike the conjunctive coordinators, it is always prepositive and can be monosyndetic or bisyndetic. For the disjunction of NPs, the bisyndetic structure prevails in written texts and is strongly preferred in elicitation, cf. (64) and (65): (64) ja borz, ja stag xila mega iza. or wolf or man be.inf may it ‘It may be a wolf or human being.’

(KJ p.19)

(65) so massuo.’a aaghoor guonax hwoezhush v-ara, I all sideward all.around look.scv V-be.impf ja naax b-iexash mettig, ja niaq’ karuo-r-ie or people B-live.scv place or way find-masd-all satyysush […]. (KJ p. 19) hope.scv ‘I was looking all around, hoping to find a lived-in place or a path […].’ When non-finite or finite clauses are coordinated by disjunction, both the monosyndetic and the bisyndetic structures are possible. In the case of monosyndesis, ja appears in front of the last coordinand, as in the disjunction of non-finite clauses in (66) and (67) and finite clauses in (68). (69) gives an example of bisyndetic disjunction of clauses. (66) so qoerush v-ara, so hweeq’al-ax tila-r-na, I fear.scv V-be.impf I mind-lat lose-masd-dat ja talxa-r-an niaq’-a t’e v-aala-r-na. (KJ p.63f.) or spoil-masd-gen way-gen on V-go-masd-dat ‘I was afraid that I could lose my mind or get on the wrong path.’

258 Liane Jeschull

(67) dwa-d-oelxush ja juxa-d-ooghush, cq’a bien ocu away-D-go.scv or back-D-come.scv once except this.obl c’a chu d-axa megar d-aac shuna. (NF p.55) house in D-go.inf may D-be.not.pres ‘On the way there or on the way back, you must not go into this house but once.’ (68) to’iitaa aeshpash, swadaa aaxcha ja as stop.imp hand.over.imp money or I.erg kaach-belsh-ash laecna aara-qussu hwo, - eelira collar-shoulder-pl catch.acv out-throw.pres you say.wp as ch’oogha. (KJ p.15) I.erg loud ‘Stop the lies and hand over the money or I will collar you and throw you out, said I loudly.’ (69) […] ja so v-ala v-ieza, ja asa i Zhovhar or I V-die.inf V-must or I.erg this Zhovhar 89-chu busu saina swa-j-aaluo j-ieza, […]. (NF p. 50) 89-in at.night myself.dat here-J-bring.inf J-must ‘[…] either I must die, or I must take this Zhovhar here to me [i.e. I must marry her] in the 89th night.’ Jakovlev (1940: 175) concludes that Chechen ja corresponds exactly to the Russian disjunctive coordinator ili (‘or’) in its meaning and use. He assumes that monosyndesis conveys the ordinary disjunctive meaning, while bisyndetic ja … ja …, unlike the obligatorily bisyndetic conjunctive coordinators, has contrastive meaning, just like Russian ili … ili … and English either … or … If this is the case, it means that the structures of conjunctive and disjunctive coordination in Chechen have very different characteristics. 4. Negative coordination Negative coordination can be expressed in two ways. First, a negated verb can occur together with the conjunctive coordinator ’a. Second, the disjunctive coordinator ja may co-occur with ’a and a negated verb. The first structure is illustrated by examples (70) and (71), where (70) gives an affirmative sentence with the conjunctive coordinator ’a and (71) a corresponding negative sentence with the same coordinator and negation of the verb:

Coordination in Chechen 259

(70) t’e-xa’a govr ’a, t’e-j-ucha bedar ’a, on-sit.down.inf horse and on-J-dress.inf garment and juq’-ax dwa-d-iexka giarz ’a as lur waist-lat over-D-tie.inf weapon and I.erg give.fut d-u hwuuna, […]. (NF p.11) D-be.pres you.dat ‘I will give you a horse to sit down on, a garment to put on and a weapon to tie around the waist […].’ (71) t’e-j-ucha bedar ’a j-aac, t’e-xa’a on-J-dress.inf garment and J-be.not.pres on-sit.down.inf govr ’a j-aac, juq’-ax dwa-d-iexka giarz horse and J-be.not.pres waist-lat over-D-tie.inf weapon ’a d-aac san. (NF p.11) and D-be.not.pres I.gen ‘I don’t have a garment to put on, a horse to sit down on or a weapon to tie around the waist.’ The other structure of negative coordination is illustrated in examples (72) and (73). (72) again gives an affirmative sentence with ’a and (73) a corresponding negative sentence that uses both the prepositive disjunctive coordinator ja and the postpositive conjunctive coordinator ’a: (72) dinara ohwa ’a v-oessina, […] korta ’a, […] from.steed down and V-descend.acv head and koch-xecha ’a Xozhina t’iera swa-d-aeqqina shirt-pants and Xozha.dat from.on here-D-take.perf Zho*ra-Baabin Zhonsarq’-as. (NF p.15) Zho*ra-Baaba.gen Zhonsarq’-erg ‘Having gotten off his steed, Zho*ra Baaba’s [i.e. the witch’s] Zhonsarq’ cut off the […] head and took away […] the clothes Xozha had on.’ (73) so v-aewhaar v-aac, as b-oaqqur b-aac I V-dare.fut V-be.not.pres I.erg b-take.fut B-be.not.pres cynan korta ’a, ja cunna t’iera koch-xecha ’a, he.gen head and or he.dat from.on shirt-pants and v-aewhar v-aac so, […]. (NF p.14) V-dare.fut V-be.not.pres I ‘I won’t dare, I won’t take his head nor the clothes he has on, I won’t dare […].’ Again, ja can be monosyndetic or bisyndetic, just as in disjunction, while ’a is always bisyndetic, just as in conjunction. Thus ja can either precede just the last coordinand, as in (74) and (75), or each coordinand, as in (76) and (77). This holds

260 Liane Jeschull

equally for the coordination of non-finite clauses (cf. (74) and (76)) and finite clauses (cf. (75) and (77)): (74) […] Hwaasan-na biaxkam-ash b-iash ’a, ja Hwaasan-dat prohibition-pl B-make.scv and or hwiexa-r-sh d-iash ’a v-aacara i, […]. (NF p.40) teach-masd-pl D-make.scv and V-be.not.impf he ‘[…] he wasn’t giving Hwaasan prohibitions nor giving him instructions […].’ (75) ahw cigahw d-an humma ’a d-aac, you.erg there D-do.inf thing and D-be.not.pres ja hwan cigahw ghullaq ’a d-aac, […]. (NF p.36) or you.gen there matter and D-be.not.pres ‘There isn’t anything for you to do there, nor do you have business there […].’ (76) […] ghoopa chohw ja hoezhush ’a, ja waamuosh ’a, fortress in or look.scv and or learn.scv and ja lar-d-iash ’a huma d-aac-ara. (KJ p.40) or guard-D-do.scv and thing D-be.not-impf ‘In the […] fortress there was neither anything to look at, nor anything to learn, nor anything to guard.’ (77) ja caarna ghullaq d-an ’a ca weema, or they.dat service D-do.inf and neg learn.pres ja hwuuna caerga i ghullaq muuxa d-aita or you.dat they.all this service how D-let.make.inf d-ieza ’a ca xae’a. (KJ p.37) D-must and neg know.pres ‘Neither do they learn to do the service, nor do you know how you must let them do this service.’ Bisyndetic or polysyndetic structures in negative coordination, in particular those involving bisyndetic ja, seem to have contrastive meaning, parallel to disjunction. This is shown in the latter two examples.


Adversative coordination

Adversative coordination is always monosyndetic and prepositive. It differentiates two types with two different coordinators. The first one makes use of amma, cf. (78) and (79):

Coordination in Chechen 261

(78) gotta-chu, amma toe’arg c’ena-chu c’a chu v-igira narrow-obl but enough clean-obl room in V-lead.wp so cuo. I he.erg ‘He led me into a narrow, but clean enough room.’

(KJ p.23)

(79) shien k’entan k’antana ketar j-ina herself.gen son.gen son.dat fur.coat J-make.perf baaba-s, amma phwyyshash-n-ii, ketaran koochan-ii old.woman-erg but fur.gen collar.dat-and c’uoka ca toe’na. (NF p.100) pelt neg be.enough.perf ‘The old woman made her grandson a fur coat, but for the sleeves and the fur collar the pelt was not sufficient.’ The other type contrasts two parallel situations. It is expressed by use of the coordinator tq’a, which means something between ‘and’ and ‘but’, cf. (80) and (81): (80) so hwan gospodin v-u, tq’a hwo san I you.gen master V-be.pres and/but you I.gen jalxoo v-u. servant V-be.pres ‘I am your master, and/but you are my servant.’

(KJ p.15)

(81) Jieshap-an hwynax cwa c’a ’a d-ina d-iexash Jieshap-gen one house and D-make.acv D-live.scv xilla, tq’a Aeltamar sheera-chu arie-n t’iahw be.perf but/and Aeltamar flat-obl plain-gen on v-iexash xilla. (KJ p.137) V-live.scv be.perf ‘[The witch] Jieshap was living in a house she built in the forest, and/but Aeltamar was living on the plain.’ Significantly, these two types correlate with the two types of adversative coordination available in Russian. The first, amma, parallels the Russian no (‘but’), whereas the second, tq’a, mirrors the Russian a (‘and, but’). Despite differing lexical means, their distribution is exactly the same in both languages.

6. Discussion As seen in the data above, Chechen primarily makes use of asyndesis and bisyndesis in coordination. Only negative coordination and disjunction can be monosyndetic, as is adversative coordination in general. Although asyndetic coordination can be

262 Liane Jeschull

found in oral communication for combining NPs, non-finite clauses and finite clauses alike, writing shows a strong preference for use of coordinators and seems to limit coordinator-less conjunction to compound-like NPs and adjectival and participial modifiers of NPs, cf. examples (5) through (8) above. The latter property of coordination happens to be common in Russian and also in other European languages (cf. Haspelmath, to appear) and might therefore be due to Russian influence. Moreover speakers of the Aeqqi dialect of Chechen, most of whom are illiterate in their native language, partially reject the coordinators that literary Chechen employs. Instead they rely on intonation and basically distinguish between coordination with and without intonation break, which has the characteristics outlined in § 2.1.1 above. In Aeqqi, intonation seems to be the most natural strategy of coordination, whereas in the literary variety, bisyndetic coordinators seem to replace the two major intonational structures suggested by Mithun (1988: 331ff.). This supports Mithun’s thesis that intonation alone is the primary means of coordination and that the rise of overt coordinators throughout the world’s languages correlates with exposure to literacy. In literary Chechen, bisyndesis, rather than monosyndesis, appears to be the original structure in coordination. The bisyndetic structure is obligatory in conjunction and also occurs with other coordinators. The observation that the conjunctive suffix -(i)i may occur monosyndetically can be explained by independent reasons. Bisyndesis in disjunction and negative coordination probably conveys contrastive meaning, but only because monosyndesis is available in these cases too, at least with non-finite and finite clauses. In disjunction of NPs, the bisyndetic structure is strongly preferred or perhaps even the only one possible. The most striking evidence for bisyndesis as the original structure comes from outside Chechen. Conjunction and disjunction in Chechen show common features with other Northeast Caucasian languages (see van den Berg, this volume, for an overview of coordination in Daghestanian languages). In most Daghestanian languages, except the Lezgic group, the bisyndetic structure of conjunction and disjunction appears to be the native one. Just as in Chechen, the native conjunctive coordinators also show up in focussing contexts and in subordinate clauses. The lexemes, however, vary across languages, just as the two basic conjunctive coordinators -(i)i and ’a are specific to Chechen. Disjunctive ja and adversative amma are the same in most Northeast Caucasian languages. These findings suggest a common Northeast Caucasian basis for the bisyndetic syntactic structure of coordination, and a common Northeast Caucasian source for early borrowing of disjunctive and adversative coordinators themselves. The distinction between the two conjoining coordinators and the present development of the conjoining suffix -(i)i (cf. § 2.1.2) seem to be specific to Chechen, but do not contradict the present analysis. Semantically, -(i)i conjoins coordinands that jointly participate in the same event, whereas ’a conjoins coordinands that denote separate participants of an event and different events themselves.

Coordination in Chechen 263

In syntax therefore conjunction by -(i)i is restricted to the NP domain and nonfinite clauses, whereas ’a can also conjoin finite clauses. While its origin is not quite clear, the conjunctive suffix -(i)i behaves like a morphological marker rather than a syntactic coordinator, perhaps something like a dual marker. The constituents conjoined by it exhibit characteristics of compounds rather than coordinands. Thus -(i)i usually combines no more than two constituents, which are often semantically related in some way or even form a conceptual unit. Although the suffix is predominantly bisyndetic, its tendency is toward monosyndesis, so that it attaches to the rightmost constituent only. Similarly, morphological processes such as derivation or inflection may apply to the rightmost constituent only. Just like true N-N compounds, then, nominal coordinands of this type often belong to the D-gender class, regardless of the class membership of their constituents. All these facts point in one direction: the suffix -(i)i treats the constituents it attaches to as compounds rather than coordinands, so that they are more likely to be products of morphology than syntax. Adversative coordination in particular (cf. § 5), but also the contrastive structures in disjunction (cf. § 3) and negative coordination (cf. § 4) and some specific conjunctive coordinators (cf. § 2.2.3), are highly reminiscent of the respective structures in Russian. As opposed to the native Chechen structures, they are all prepositive, just as in Russian. Bisyndesis in these cases conveys some kind of contrast, just as it does in Russian. The bisyndetic coordinators in disjunctive ja …, ja … (‘either … or …’) and negative ja … ’a, ja … ’a (‘neither … nor …’) are identical for each coordinand, just as in Russian ili …, ili … (‘either … or …’) and ni …, ni … (‘neither … nor …’). The use of the two adversative coordinators amma and tq’a mirrors exactly that of Russian no (‘but’) and a (‘and, but’) respectively. The further conjunctive coordinators illustrated in examples (58) and (59) in § 2.2.3 show strong parallels with Russian not only in syntactic structure, but also in lexical means. Thus the lexical material is often precisely equivalent to that in Russian. Therefore the more recent borrowings must have been sensitive to both the lexicon and the syntax of the source language Russian.4 To conclude, coordinating structures as found in contemporary Chechen must be attributed to different sources. The role of asyndesis is not quite clear. In written language, however, it plays only a minor role. Instead, the literary variety heavily employs bisyndesis. Significantly, the bisyndetic structure coincides with that of most Daghestanian languages, except the Lezgic group. This strongly suggests that the bisyndetic coordinating structures of Chechen and the majority of Daghestanian

4.One might wonder whether the equivalence of structures in Russian and Chechen in my examples drawn from a translated novel (Pushkin 1961) might be due to the translation itself. However, the same structures and coordinands can be found in the Chechen folklore (El’mursaev 1964).

264 Liane Jeschull

languages originate from the same source, either from a common native Northeast Caucasian basis or from the same line of borrowing. At any rate, it is very likely that in Chechen bisyndesis emerged earlier than monosyndesis. Virtually all monosyndetic coordinators can be attributed to borrowing the relevant structures from Russian and employing lexical equivalents in Chechen. This is not surprising, as Russian has been the major source of influence ever since Russian became the language of education and official business during the Soviet period. Therefore the facts analyzed here support Jakovlev’s (1940) and Mithun’s (1988) conclusions about the origin of overt coordinators. Indeed it must be the case that literacy in Chechen itself and exposure to other written languages, i.e. Russian most recently, have caused or at least contributed to the rise of coordinators in Chechen.

Abbreviations acv all cl impf lat masd pastcond prescond

anterior converb allative cl*tic imperfect lative masdar past conditional present conditional

pastpart prespart scv subj wp B, D, J, V

past participle present participle simultaneous converb subjunctive witnessed past gender-class markers

References Aliroev, I.Ju. 1999. Cˇecˇenskij jazyk. Moskva: Academia. Dešeriev, Ju.D. 1967. “Cˇecˇenskij jazyk”. In Jazyki narodov SSSR: Iberijsko-kavkazskie jazyki, V. V. Vinogradov (ed.), 190–209. Moskva: Nauka. Dešerieva, T. I. 1999. “Cˇecˇenskij jazyk”. In Jazyki mira: Kavkazskie jazyki, M. E. Alekseev et al. (eds.), 173–186. Moskva: Academia. El’mursaev, S. (ed.) 1964. Noxchiin fol’klor. Vol. 2. Grozny: Noxch-ghalghain knizhni izdatel’stvo. Haspelmath, Martin (to appear). “Coordination”. In Language Typology and Syntactic Description, T. Shopen (ed.) [2nd edition]. Cambridge: Cambridge University Press. Jakovlev, N. F. 1940. Sintaksis ˇcecˇenskogo literaturnogo jazyka. Moskva & Leningrad: Izdatel’stvo akademii nauk SSSR. Maciev, A. G. 1961. “Kratkij grammaticˇeskij ocˇerk cˇecˇenskogo jazyka”. In Cˇecˇensko-russkij slovar’, 571–625. Moskva: Gosudarstvennoe izdatel’stvo inostrannyx i nacional’nyx slovarej. Mithun, Marianne. 1988. “The grammaticization of coordination”. In Clause Combining in Grammar and Discourse, J. Haiman & S. Thompson (eds.), 331–359. Amsterdam: Benjamins.

Coordination in Chechen 265

Nichols, Johanna. 1994. “Chechen”. In North East Caucasian Languages, R. Smeets (ed.) [The Indigenous Languages of the Caucasus, J. A. C. Greppin (ed.), Vol. 4, Part 2], 2–77. Delmar & New York: Caravan. Peterson, David A. 2001. “Ingush ’a: The elusive type 5 cl*tic?”. Language 77 (1): 144–155. Pushkin, A. Je. 1961. Kapitanan jow. Trans. X. Arsanukaev. Grozny: Noxch-ghalghain knizhni izdatel’stvo. van den Berg, Helma. This volume. “Coordinating constructions in Daghestanian languages.”

Middle East

Chapter 11

Coordination in three Western Iranian languages Vafsi, Persian and Gilaki Donald Stilo Max-Planck-Institut für evolutionäre Anthropologie, Leipzig

1. Introduction 1.1 Vafsi, Persian and Gilaki 1.2 The sets of coordinate conjunctions 1.3 Historical origins and derivations 1.4 Stress and intonation 1.5 Background grammatical information: Vafsi 2. Conjunctive coordination 2.1 Main points 2.2 Noun phrase conjunctive coordinators 2.3 Verb phrases and clauses 2.4 Subordinator ke as coordinator 3. Adversative coordinative conjunctions 3.1 Main points 3.2 The adversative coordinators 3.3 Initial occurrence in intonational unit 3.4 Adversative conjunctions deleted 3.5 bælke ´ : Contrastive after negative 4. Disjunctive coordinators 4.1 Phrase level 4.2 VP and clause level 4.3 The coordinator yá-ìnke 5. Bisyndetic coordinative conjunctions 5.1 Main points 5.2 The bisyndetic coordinators 6. hæm ‘also’ and other uses 6.1 The uses of hæm 6.2 Further discussion of hæm 7. Conclusion

270 Donald Stilo



1.1 Vafsi, Persian and Gilaki The following is an overview of coordination in three western Iranian languages: Vafsi, Gilaki, and two very different registers of Modern Persian: colloquial Tehrani Persian (CP) and Formal Written Persian (FWP). Most attention, however, will be given first to Vafsi due to a quite extensive audio taped corpus in Vafsi from the village of Gurchan (Stilo: forthcoming). Over five years of additional fieldwork on Vafsi (village of Vafs) and almost two years (in addition to secondary sources) on Gilaki, both with various speakers — with no special focus on coordination — also went into formulating the present description. The second emphasis of the article is devoted to spoken Persian, for which I have used my own examples of the language.1 Vafsi (4 villages, perhaps 20,000 speakers) belongs to the Tatic subgroup (Stilo 1981) of Northwest Iranian and the material below appears in two very close subdialects (villages of Vafs and Gurchan). Since Vafsi is surrounded by Azerbaijani and since all education and contact beyond the local level is through Persian, almost all speakers are trilingual, but with varying degrees of mastery. Gilaki, also Northwest Iranian, belongs to the Caspian subgroup and consists of two major dialect clusters, Western Gilaki (center: Rasht) and Eastern Gilaki (center: Lahijan) (Stilo 2001). Most Gilaki speakers (2–3,000,000) are (generally non-coordinately) bilingual in Gilaki and Persian. Persian is Southwestern Iranian. FWP, the written norm of Persian in Iran (Farsi), Afghanistan (Dari) and Tajikistan (Tajiki), represents a historical stage of some 500 or more years ago and is not used as a spoken language. The difference between CP and FWP is not parallel to the gap between modern colloquial and written English but more analogous to some level between spoken modern English and written Biblical English.2 At any rate, the gap in the two styles of Persian is clearly not as wide as the one between written Modern Standard Arabic and modern colloquial Arabic dialects.3

1.Sentences that seemed questionable were tested with native speakers of Persian. 2.If the differences were just a matter of rigid vs. not-so-rigid verb-finality or a question of formal style (“going to”) vs. informal style (“gonna”), the parallels to spoken-written English would be more accurate. The analogy of an older stage of written English to FWP and their modern spoken counterparts, however, applies to all levels: morphology (verb endings, object marking, etc.); outmoded syntactic features; pronouns archaic to the modern language; lexical choices, etc. Since the written language has become second nature to literate Iranians, people untrained in linguistics are usually not aware of how wide the rift between the two registers is. 3.Since there is also a wide range of variation in styles in FWP, I have chosen FWP examples from both a rather informal, narrative style, Al-e Ahmad [1973], and a somewhat formal

Coordination in three Western Iranian languages 271

A normalized form of Colloquial Tehrani Persian, the sociolinguistic register used for most polite spoken communication, as for example among educated people who do not know each other well, has become the general spoken standard of the modern language for all of Iran. Local variants of Persian still exist in the major cities, however, even Tehran, where the non-normalized version, called ‘xodemuni’ (or ‘our own’, i.e., ‘just amongst ourselves’), should be considered a separate sociolinguistic register for use with the family and intimate friends (Stilo, Clinton, Talattof: forthcoming). Examples of CP given in this article are in normalized CP rather than in ‘xodemuni’ (e.g., normalized gorosnæmun=e ´ vs. ‘xodemuni’ gošnæmun=e ´ ‘we are hungry’, goft/ræft vs. góft=eš/ræft=eš ´ ‘s/he said, s/he went’ or the retention of final =nd and =st rather than ‘xodemuni’ =n and =s(s)).4 There is a probable time depth of ±2800 years between Northwest and Southwest Iranian, with a multiplicity of minor, unwritten (and usually poorly documented) languages in each group. A brief examination of simple sentences with fairly ordinary lexical items, such as sentences (8), (10) and (20) below for example, will suffice to show the average degree of linguistic divergence in Vafsi, Persian, and Gilaki. While the emphasis of the article goes to Vafsi, it should also be noted, however, that many comments pertain equally to all three languages (but usually bypassing FWP, due to its formal, non-spoken nature). Vafsi, Persian, and Gilaki mostly share the same forms and rules for coordinators (but see also footnote 5 below for an important caveat). Important differences will be indicated where necessary. When statements are made concerning all three languages together, they will simply be abbreviated VPG. Abbreviations on the left side of examples are: v, Vafsi; cp, standardized colloquial Persian; p, general Persian (CP or FWP, possibly with slight differences); g, western Gilaki (Rashti); lj, eastern Gilaki (Lahijani); c, common to VPG (with slight phonological differences, e.g., Persian o, e, æ corresponding to Gilaki u, i, 6, etc.). 1.2 The sets of coordinate conjunctions VPG have three types of monosyndetic and bisyndetic coordinate conjunctions: conjunctive, adversative, and disjunctive:

style, Zarrin-kub [1996]. The latter tends toward longer sentences with many more embeddings and is comparatively lower in coordination. 4.xodemuni as I use it with a sociolinguistic point of view in mind is certainly narrower than the way it is used among the general Iranian populace, for whom it usually means simply CP.

272 Donald Stilo

Simple Coordinate Conjunctions

Bisyndetic Coordinate Conjunctions

A) Conjunctive

A) Conjunctive

c =ò ~ =vò c væ v hemra v =ra, hemra v =iz ~ =zi p, g, v =(h)æm c ke


c hæm … (=ò) hæm … ‘and’ c næ … næ … ‘with (comit.); and’ v næ tena … ‘with (comit., instr.); v bælke ´ … (=iz) and’ ‘also, and’ p, g næ tænha … ‘also, and’ p, g bælke ´ … (=hæm) subordinator*, ‘and’

‘both … and …’ ‘neither … nor …’ ‘not only … but … (also)’ ‘not only … but … (also)’

*ke is almost always a subordinating particle used in all types of subordinate clauses. It is used extremely rarely as a coordinator, and only on the clause level.

Simple Coordinate Conjunctions B) Adversative c væli ´ c vælikæn ´ p æmma; ´ v æmma ´ ~ p ámma; g ámma c fæqæt c bælke* ´

‘but’ ‘however’ ‘but’ ‘only’ ‘but’

*contrastive after negative, cf., Sp. sino, Ger. sondern

Simple Coordinate Conjunctions

Bisyndetic Coordinate Conjunctions

C) Disjunctive

C) Disjunctive

c ya cp yá-ìnke

‘or’ ‘or’

c ya… ya… c ce… ce… fwp xah …xah … fwp gah …gah …

‘either… or…’ ‘whether … or …’ ‘whether … or …’ ‘sometimes … sometimes …’

Coordination in three Western Iranian languages 273

1.3 Historical origins and derivations =ò (c) unstressed, encl*tic in the modern languages; Old Persian: uta¯ > ud > uδ > Middle Persian: u > =o, ‘and’ (Kent 1953: 175; Horn 1893: 240), cf. Avestan: uta, Sanskrit: utá. (In written Persian, =ò falls together with Arabicderived væ, both written identically with a single , the Arabic waw.) væ (c) mostly FWP but also in more formal registers of spoken VPG; Arabic loanword; appears before the noun or clause and may, albeit rarely, be stressed. =iz~=zi (v)< Avestan =ci»t, cf. Old Persian =ciy ‘particle emphasizing or generalizing’ (Kent 1953: 184; Horn 1893: 100) < *Proto-Indo-European *kwi=d. hæm (p, g, v)< Old Persian ham- ‘together, with’, hama- ‘one and the same’ < *Proto-Iranian *ham ~ hama- (Kent 1953: 213) < *Proto-Indo-European sem- ~ *somo- (cf. Sanskrit sam-, samá-, Gothic sama, English same, Old Church Slavonic sam-, Greek homós, etc.). =ra (v) < Middle Persian -ra¯¯ı < Old Persian postposition ra¯diy ‘on account of ’, (Horn 1893: 134, Kent: 205) (cf. Old Church Slavonic radi). The equivalent in Modern Persian and Gilaki marks specific direct objects (and indirect objects in Gilaki), a usage lacking in most Tatic languages, where it generally means ‘for.’ Vafsi is unique in using =ra for the instrumental and the comitative. hemra (p, v)‘(along) with’, i.e., hæm ‘same’ (as above) + rah ‘road’ (cf. Sanskrit rathya¯ ‘(carriage) road’ < rátha, Old Persian raθa ‘chariot’, Latin rota, OHG rad ‘wheel’ < *Proto-Indo-European *ret(h)-, ‘walk’). While hæmra exists in Persian and Gilaki, its use as a coordinator has not yet been identified there. The Vafsi variants are hemra~henra. ke (v, cp, g)< Middle Persian ka < Old Persian ka-, ci-, Avestan ka-, ko¯, cf. Sanskrit ká, kas, kim ‘who, which, what’ (Kent 1953: 195) < *Proto-Indo-European *kwo-, see also ce below. næ… (v, cp, g)< Middle Persian na < Avestan na < *Proto-Indo-European *ne‘no, not’ (Horn 1893: 236; Kent 1953: 192). næ tænha … bælke ´ (v, cp, g)Persian tænha and Vafsi tena both mean ‘alone, lonely, sole.’ bælke ´ (v, cp, g)< ba˘l Arabic loanword, ‘nay, rather, yet, but’ + Iranian ke ‘subordinating particle.’ væli, ´ æmma ´ (c)both Arabic borrowings. The vowel in the Vafsi alternate ámma is probably due to Azerbaijani influence. In Gilaki, 6, the expected reflex for æ, i.e., *a˘, never occurs initially but is realized as a, a different phoneme. ya (c) cf. Early Judaeo-Persian: aya¯b, ya¯w ~ yaw ~ yaβ < Middle Persian: aya¯b ‘or’ (Paul: unpubl.). Further derivation is unknown. The analogous Old Persian form was an encl*tic, =va¯.

274 Donald Stilo

yá-ìnke (cp)A combination of ya ‘or’ plus the subordinator ínke (usually used to form a subordinate conjunction from a preposition). (See Section 4.3 below.) ce (c) Middle Persian cih < Old Persian ciy, ci- < Avestan ciš, ci»t, cf. Sanskrit =cit < *Proto-Indo-European *kwi-, *kwo-, Interrogative and relative pronoun base: ‘who, what’ (Horn 1893: 100; Kent 1953: 195) (cf. Latin qui=s, quo=d, English wha=t, etc.); means ‘what?’ in FWP. xah (fwp, c?)Present stem of the Persian verb xastæn ‘to want.’ Generally very formal written style. gah (fwp, c?)< Old Persian: ga¯θu ‘place, throne’, cf. Avestan ga¯tu, ga¯θw, Sanskrit ga¯tu < *Proto-Indo-European *gwem- ‘come’ (Kent 1953: 195); Modern Persian: ‘place, time; sometimes.’ Additional sources: Pokorny 1951–59 (Indo-European), Wehr 1979 (Arabic). 1.4 Stress and intonation Intonation plays a critical role in coordination in VPG. These languages clearly agree with the cross-linguistic tendency indicated in Mithun (1988: 332), in which coordinate constituents are typically separated by ‘comma intonation’, called Sustained here, characterized as a non-final pitch contour. Not only is the ‘comma intonation’ sustained (i.e., flat and high) in VPG, it is also extended and, as discussed below in 1.4.2 just after the five intonational contours are presented, there is typically also a very brief but clearly perceptible break in the intonation. Before presenting the role of intonation in coordination, a brief summary of the intonational patterns of these languages is in order. These intonational patterns are areal features that go through most of western Iran, including not only Persian, but also other Northwest Iranian languages such as Gilaki, Vafsi, etc.5 There is only a

5.I do not mean to imply that all statements made here hold automatically for all languages and dialects of western Iran. Sometimes even stress and intonation are drastically different. An outstanding case in point are the mutually intelligible forms of colloquial Persian spoken in Tehran and Esfahan. Stress and intonation in the latter are so markedly different from the former that I could not even form a judgment at this point about the relationship of intonation to coordination in colloquial Esfahani, to wit, I would venture to say that the system of pitches in this dialect is highly unusual cross-linguistically in that it puts a low pitch on a stressed syllable and a high pitch on an unstressed syllable. A possible similar system exists in the English dialects of Tyneside (aka Geordie) where the first syllable is stressed but has a low pitch which then rises to a high pitch on the following unstressed syllable and the pitch of syllables preceding the stressed syllable are typically mid to high (Bernard Comrie — personal communication). Thus when it comes to the discussion of how stress, pitch and intonation are intertwined with issues of coordination, broad generalizations across western Iran, while they do hold true for many areas, would not hold for every locale. Issues of suprasegmental phenomena of Persian dialects, especially subdialects of Persian (e.g., Esfahani, Yazdi) where stress and

Coordination in three Western Iranian languages 275

very slight difference in the Question intonation between Vafsi and the other languages, which does not concern the present discussion of coordination. As the following paragraphs show, intonational contours are in turn intimately bound to the patterns of clause-level stress, and secondarily to word-level stress. 1.4.1 Word-level and clause-level stress The general rule for VPG is that there is only one clause-level stress per clause and, except for the Subordinate contour, it generally falls on that syllable of the stressed word expected according to the rules of word stress. Other word stresses are then markedly diminished in force. The choice of which word in the clause should receive the clause-level stress depends for the most part on focus, emphasis and other matters of pragmatics and discourse. In unmarked situations, however, stress generally falls on the word just before the verb. 1.4.2 Clause-level intonational contours VPG appear to have five different levels of pitch in various contrasting intonational contours. These intonational patterns, not all of which are shown below, are integrally bound up with the syntax of these languages in general, and three, possibly four, of these contours are relevant to a discussion of coordination. The following are the various contrasting intonational patterns mentioned in this article and the five different pitches (levels 1–5) of these languages in summary: (5)







(5) (3)


Mid Low


(1) ‘Declarative 1’

(2) ‘Declarative 2’

(3) ‘Question’

(5) (3)




Mid High


(4) ‘Sustained’

(5) ‘Subordinate’

The contours of the first four patterns above are identical up to and including the clause-level stressed syllable and diverge only after this syllable. All syllables after the stressed syllable remain fairly flat, i.e., without any noteworthy rise or fall in pitch, and are extended in the Sustained pattern and considerably more so in the Question pattern. The first four patterns each have a different contrasting level of pitch, in addition to a fifth pitch (the highest) which falls on the single stressed syllable. The fifth intonational pattern consists of a rising pitch, and usual sentence

pitch sound totally different from other forms of Persian, are fertile ground for investigation.

276 Donald Stilo

stresses are generally neutralized and automatically placed on the first syllable of the clause. For cases of emphasis, however, sentence stress may be reinstated on the emphasized word. While subordination is not the topic of this paper, there are a few examples of coordinated sentences in this article, however, in which subordinate clauses also appear. Thus it is relevant to include Contour 5 (Rising) here as well. Declarative 2 is also shown, but it is relatively uncommon and seems to be restricted mostly to narratives, and even in these contexts it is not used as often as Declarative 1, with which it is interchangeable. Point (iv) under 4.1.2, however, does discuss the possibility that Declarative 2 intonation, or a contour very similar to it, may in fact play a distinctive role in certain types of disjunctive coordination. (Recorded examples of all these intonational patterns can be found on the compact disk accompanying Vafsi Folk Tales (Stilo: forthcoming), where each one of these contours has been marked with different punctuation throughout the texts.) While the first three contours are complete and independent, contours four and five are non-final and imply a continuation to a following contour. In the latter cases, there is a clear break before the next contour rather than a glide from the higher pitch to the initial level 3 pitch of the next segment. Examples of these intonations are from CP: 1. Declarative 1 is the intonation most commonly used for statements, as in (1), and wh-questions, as in (3). Pitch levels of contour: 3–5–1. While clause-level stress usually occurs on the word before the verb, it may shift to any word in discourse. In these cases, only the placement of the clause stress changes to the appropriate syllable, but the intonation contour still remains exactly the same. Even when the segment after the clause-level stress is considerably longer, as in (2), the following contour is still flat. (1) una míxand NUN bóxorænd.

‘They want to eat (some) bread.’ (2) uNA míxand nun bóxorænd.

‘They want to eat (some) bread.’ (3) una koJA míxand nun bóxorænd?

‘Where do they want to eat bread?’ 2. Question intonation for non-wh-questions: there is no other syntactic device to form non-wh-questions except in FWP where the initial question particle áya

Coordination in three Western Iranian languages 277

optionally occurs (but the question intonation remains). Pitch levels of contour: 3–5–3. The Vafsi variant is only slightly different. (4) una míxand NUN bóxorænd?

‘Do they want to eat (some) bread?’ 3. Sustained intonation is used to connect two longer coordinated phrases and often (but not always) any two coordinated nouns or other words. Pitch levels of contour: 3–5–4. (5) una míxand NUN bóxorænd=o...

‘They want to eat (some) bread and…’ Coordinated nouns in a list often each have their own stress (an exception to the rule of one stress per clause). (6) is an additional example of a list of coordinated nouns: (6) nún=o, pænír=o, kæré=vo, caí=o, zeytún=o, moræbbá,... ‘bread=and, cheese=and, butter=and, tea=and, olives=and jam…’ (an Iranian breakfast) 4. Subordinate intonation is used for most subordinate clauses. There are variants of this intonation, but these are beyond the scope of the present article. Pitch levels of contour: 5–3-rising to 5/5+. (7) ´ una mixand nun boxorænd,... Æge

‘If they want to eat (some) bread, …’ Intonation is an integral part of coordination. It is common to omit coordinating conjunctions in speech leaving juxtaposed phrases/clauses, and the only indicator of the presence or lack of coordination in otherwise asyndetic constructions is then the intonation. Note the following contrasts between coordinated clauses (Sustained + Declarative) and juxtaposed sentences (Declarative + Declarative): coordinated clauses (Sustained + Declarative) (8) v: æli hottæ, aqàte næ-r-kær-e. ´ lj: ali xóttI, hærf ` næ-zæn-e. ´ cp: æli xabide, hærf ` né-mi-zæn-e. Ali asleep speech neg-t/a-hit-3s ‘Ali is asleep and is not talking’

juxtaposed sentences (Declarative + Declarative) æli hottæ. aqàte mæ-kæ. ´ ali xóttI. hærf ` næ-zæn. ´ æli xabide. hærf ` næ-zæn. ´ Ali asleep. speech neg-hit. ‘Ali is asleep. Don’t talk!’

278 Donald Stilo

1.4.3 Punctuation The intonations that are relevant to the present discussion of coordination in VPG are Declarative 1 (Low = level 1), Question (Mid = level 3), and Sustained (Mid High = level 4), the latter being the most important for this topic. The following punctuation is used in the examples given below to represent the various intonational contours: period/full stop (.) for Declarative 1; Question mark (?) for Question intonation; comma (,) for Sustained intonation (with clauses or individual words); and double comma (,,) for Subordinate intonation. Hence the punctuation used in some examples will look quite different from that of a typical English sentence, e.g., question marks in the middle of a sentence (cp: emruz mir-i? ya færda. ‘Are you going today or tomorrow?’ where we find a Question intonation in the middle of a clause that finishes with a Declarative intonation). The stresses of words and clauses are shown as /´/ for primary stress (omission of stress marks assumes final stress), /`/ for secondary stress in compounds (which would be primary if the compounding word stood alone), and full caps for the one syllable where clause stress occurs, when relevant. (9) ` cp: mæn mí-xa-m ŠAM doróst kon-æm. I t/a-want-1s dinner right do-1s ‘I want to make dinner.’

The symbol ^ is used below to show the omission of a coordinate conjunction. Since this type of deletion is usually accompanied by Sustained intonation, this symbol will then mostly cooccur with a preceding comma that shows this intonation. 1.5 Background grammatical information: Vafsi Since Vafsi in general has a more complex grammatical system than either Persian or Gilaki, especially in its tense-based split marking of Agent and Patient with the Nominative-Accusative (Direct-Oblique) in the present system, and both the Ergative-Absolutive (Oblique-Direct) (alternate 1) and the Ergative-Accusative (Oblique-Oblique) (alternate 2) in the past system, I feel it would make the examples — and the accompanying interlinear glossing — throughout this paper somewhat clearer to the reader to point out some central features of Vafsi syntax: 1. There are two cases in the nominal system (two masculine singular affixes, two feminine singular affixes for each of two feminine noun classes, and two common plural affixes). The two cases, as well as two corresponding sets of pronominal agreement markers, are called simply ‘Direct’ and ‘Oblique.’ Their forms and functions are presented in brief in the table below. An asterisk indicates parallel uses between the nominal cases and the two corresponding sets of agreement markers in

Coordination in three Western Iranian languages 279

the verb (as well as possessives on nouns). While there are other uses for the Direct and Oblique cases, I present only their main uses here: Direct Case – Subject of intransitive verbs (all tenses)* – Agent in present system* – Patient in past system* – Patients that are not both specific and animate nouns (all tenses) – Adjuncts of Adpositions (usually inanimate)

1s: 2s: 3s: 1p: 2p: 3p:

Direct Markers

Oblique Case

Oblique Encl*tics

-om~-im -i -e~Ø -am -a -end(e)

– Agent in past system* -(o)m~-im – Patient in past system -i (specific, animate nouns) -(e)s~-is – Definite, Animate Patients -oan in present system* -ian – Possessive* -(e)san~-isan – Indirect Object* – Adjuncts of Adpositions (usually animate)*

2. While the direct agreement markers always follow the verb root and the oblique cl*tics always precede it, the latter are somewhat more mobile in their placement. In a process I call Fronting, the oblique encl*tics of past transitive verbs optionally (and with constraints) may move off the verb and cl*ticize to a preceding word, often the direct object, as seen in the following segment from sentence (10): v: bera =s hæ-da tini brother-3s pvb-gave him ‘he gave him a brother...’

6d, xuš->am6t, c6har-šan6. p: jæván-i bud bolænd-qæd, xoš-qamæt, cæhar-šane. (CP: car-šune) youth-indef was tall-stature well-stature four-shoulder ‘He was a tall, straight-standing, square-shouldered young man.’ (vii) In addition to connecting two coordinate clauses, a conjunctive coordinator may often begin a sentence/intonational unit, but since =ò may not occur initially, the variant væ will appear in these situations. More common than væ in this situation, however, especially in CP, would be the use of the variants of ‘also’ as a coordinator (see discussion of hæm, point ii in § 6.1): (19) cp: væ æz un bædtær, xæsis=e! and from that worse stingy-is alternate: æz ún-æm bædtær, xæsis=e! from that-also worse stingy-is ‘And what’s worse, he’s stingy!’ (viii) In VPG a coordinator almost always occurs obligatorily within the NP, whether between two or each of the various elements of a longer series, but may occasionally be deleted (see point vi above). CP and FWP behave quite differently from each other in this regard. In FWP, the various initial elements of a list of NPs may or may not be linked by a coordinator, but the last two elements of the list have a virtually obligatory coordinator. See § ‘Lists of Nouns’, especially (32), (33), (35), and (36) for FWP. (ix) Vafsi seems to be the only language here to have comitative-type conjunctive coordination. There are two comitative adpositions in Vafsi, =ra (postposition) and henra ~ hemra (postposition, occasionally a preposition). The former also functions as the instrumental and the benefactive. Both of these postpositions also function as conjunctive coordinators, meaning ‘and’, used only when conjoining nouns. =ra in this function is more common than henra ~ hemra. Even though Vafsi has a comitative pattern for coordination, it still qualifies as an and-language as expounded in Stassen (2000: 21–40) because of the clear predominance of coordinate encoding over a minor pattern of comitative encoding, as well as alignment with and-languages according to his Parameters of Casedness (Stassen 2000: 44) and Tensedness (Stassen 2000: 45–46). The comitative is clearly a coordination strategy in Vafsi as it has the support of the morphosyntactic evidence of plural agreement in the verb triggered by comitative-conjoined NPs (see (Haspelmath, to appear) and (Stassen 2000: 26)). Vafsi also fits with Haspelmath’s observation that languages (e.g., Russian) may show the comitative-origin of WITH-coordination by restricting this structure ‘to animate conjuncts, and the two conjuncts are typically thought of as participating in the situation together’ (Haspelmath, to appear). I have found no examples of comitative coordination with

284 Donald Stilo

inanimates in any of my Vafsi data. This coordinate function of the comitative adposition also occurs in other minor Iranian languages and clearly needs further investigation outside of VPG.9 (x) As sentences (50) and (55) show, Vafsi also has rare cases of double marking of coordination with unusual typology (see § 2.2.4): (50) A B-co-co, that is, A B-cowith-coand (55) A co-B-co, that is, A coand-B-cowith 2.2 Noun phrase conjunctive coordinators The following are examples of coordination in various types of NPs: 2.2.1 Coordinator: =ò (the non-comitative coordinator) two elements (not naturally associated). (20) v: kelle-mæzæn-e=o lazæ-viljæ=s daughter-big-f.dir=and son-small=3s.obl cp: doxtær-bozorge=vo pesær-kucikæ=š ´ daughter-big=and son-small=3s.obl g: ani pilla-duxt6r=u kuce-p6s6r his big-daughter=and small-son ‘his/her oldest daughter and youngest son’ (21) c:

nesbæt be zar-in=o karegær-an relation to farmer-s=and worker-pl ‘in relationship to the farmers and workers’

9.Comitative coordination is encountered elsewhere in Northwest Iranian languages (Tatic, Central Plateau Group, etc.) as a minor coordination pattern, but until now not much attention has been paid to this phenomenon. Two representative examples come from Chali (Yarshater 2001: 252) and Gazi (Eilers & Schapka 1979: 152) Chali: zani-o¯ ye vaya wife-and with bride ‘the wife and the bride’ Gazi:

hæsirbaf xow refi>-až æšnoft-ežun-e mat.weaver with friend=3s.obl heard=3p.obl-t/a ‘the mat weaver and his friend were hearing’

Coordination in three Western Iranian languages 285

Adjectives (22) c:

mærdom-án-e xoda-šenas=o motædæyyen people-pl-ez god-knowing=and religious ‘God-fearing and religious people.’ three elements (not naturally associated). (23) c:

æli=o æhmæd=o mohæmmæd ‘Ali and Ahmad and Mohammad’ v: viz-dar=o vìæ-dár=o hær-ci walnut-tree=and willow-tree=and every-what ‘walnut trees, willow trees and everything’

Occasionally the coordinator =ò is accompanied by =iz/=hæm ‘also’: (24) p: pustin=o æsb=o pul=(h)æm v: pusduney=o æsb=o pul=iz pustin=and horse=and money-also ‘a pustin (inverted sheepskin coat) and a horse and money, too’ Two or more elements (usually naturally associated). (25) c:

cæp=o rast ‘left and right’

siah=o sefid ‘black and white’

tir=o kæman ‘arrow and bow’ (reverse of English)

(26) v: sáz=o dordoræ=vo dayræ p: saz=o dohol=o dayere g: saz=u duhul=u day6r6 (three musical instruments that are usually played together) Note that some nouns or adjectives that usually occur together have been lexicalized as compounds, some with a coordinator and some without, although the latter is less common: (27) With =ò: p: gærd=o xak v: gærd=o xa dust=and dirt ‘dust’ p: tær=o taze g: t6r=o taz6 wet=and fresh ‘fresh’

286 Donald Stilo

p: tæk=o tænha g: t6k=u t6nha single=and alone ‘all alone’ p: kæm=o ´ biš less=and more ‘more or less’ (28) Without =ò: p: kot-šælvar coat-pants ‘suit’ g: arus-damad bride-bridegroom ‘wedding couple’ Coordinators and numbers. A coordinator is obligatory with fractions and compound numbers above 20 (but not multiples such as VPG se-hezar ‘3,000’): (29) c:

bist=o car 20=and 4 ‘24’

se=o se-rob 3=and 3-quarter ‘3¾’

se hezar=o noh-sæd=o do 3 1000=and 9-100=and 2 ‘3,902’

No coordinator or Sustained intonation is used with ranges of numbers, e.g., English ‘three or four’, which are rather treated phonologically as compounds: (30) v: se-car dane 3-4 clf ‘3 or 4; a few’

hæft-hæšt-dæh dane 7-8-10 clf ‘several’ (p, g use ta instead of dane) Lists of nouns. With long lists of nouns, =ò and Sustained intonation (see sentence (6) above) are almost always obligatory on each coordinand. While less common in FWP, coordinators after each NP in a list also can be found there as well. Examples: (31) c:

noh mah=o noh ru=o noh saæt=o noh dæqiqe 9 month=and 9 day=and 9 hour=and 9 minute ‘Nine months and nine days and nine hours and nine minutes’ (v: ru ‘day’ = p, g: ruz)

Coordination in three Western Iranian languages 287

(32) fwp:

dær órf=o siasæt=o ´ færhæng=o ´ mætbu’-át-e in tradition=and politics=and culture=and printed-pl-ez moasér contemporary ‘in the common tradition, politics, culture, and contemporary printed media’ (Al-e Ahmad 1973: 5)

(33) fwp:

ru-ye-hæm-ræfte bolænd-qamæt-ænd=o por-kar=o on-ez-hÊm-gone tall-stature-are=and full-work=and aftab-suxte væ zud be pir-i nešæste. sun-burned and early to old-ness seated ‘All in all they are tall-statured, hard-working, sunburned, and prematurely aged.’ (Al-e Ahmad 1973: 12) (last three elements extraposed post-verbally)

(34) v: [viz=o sía=w ælúa=w ælucæ-súr=o ` šilána=w váva=w béva=w ængúra=w túta=w qeysía=w moxtæsær gilása=w] dir-énde. ‘They-have [walnuts=and apples=and plums=and sour.cherries=and peaches=and almonds=and quinces=and grapes=and mulberries=and apricots=and, in.short, black.cherries=and.]’ (Details: The Vafsi feminine unstressed -æ` converts to -a before =w: síæ ‘apple’ fi sía=w ‘apple and.’ Only viz and ælucæ´ are masculine here.) In more formal styles of FWP, however, lists of nouns, e.g., names as seen here, are given with only one coordinator appearing before the last coordinand in the series: (35) fwp:

mirzade-ye ešqi, yæhya dowlætabadi, dehxoda, væ tæqí-e ræf ’æt pn-ez pn pn pn pn and pn-ez pn dær jostoju-ye … in search-ez ‘Mirzade Eshghi, Yahya Dowlatabadi, Dehkhoda and Taghi Raf ’at in searching for…’ (Zarrin-kub 1996: 543)

but even there, rigid adherence to the deletion of coordinators is not always the case: (36) fwp:

hæmanænd-e ´ aref=o ešqi=o kæmalí-e esfœhaní=o færroxí-e similar-ez pn=and pn=and pn-ez pn=and pn-ez yæzdi hæm… pn also ‘…similar to Aref, Eshghi, Kamali-Esfahani, and Farrokhi-Yazdi also…’ (Zarrin-kub 1996: 543)

Sentence (34) shows a common phenomenon in long lists in VPG (except FWP). Since =ò is encl*tic and cannot appear initially, a speaker giving a long open list of items often does not know where s/he will end the list. For this reason, =ò is

288 Donald Stilo

automatically attached to all nouns even if it ends up being the last item in the list. If it is not included, each time the speaker wants to add another word, s/he is obliged either to repeat the previous word with =ò attached in order to continue or to use væ, which is usually avoided in speech. On the other hand, if a list happens to end with =ò without a final noun in the series there is no need to correct this by adding another element. =ò is simply left dangling in final position and the sentence continues. Another example: (37) v: bez-é fætirey=o kærey=o ron=o qeymaq=o xawatu=vo goat-f.obl fatire=and butter=and oil=and cream=and cream=and b=ís=sattæ=o b=ís=bærdæ yey osda-y soraq. t/a=3s.obl=made=and t/a=3s.obl=took 1 master-m.obl to ‘The goat made fatire, butter, ghee, thick cream, and heavy cream and took (them) to an artisan.’ This final =ò even occurs in shorter lists when possible additions are expected: (38) v: bí=m=resda bærˇí=o koleng=o b=ísan=ard t/a=1s.obl=sent spade=and pick.axe=and t/a=3p.obl=brought ‘I sent for a spade and a pick-axe.’ This process occurs quite commonly in spoken VPG, but sounds awkward and unnatural when singled out and causes native speakers to balk if elicited out of context. 2.2.2 væ ‘and’ As stated above, the use of væ is more common in FWP but, even there, it is considered better style when reading a text aloud to use the encl*tic =ò except in initial position. væ occurs in VPG, especially initially or after a break in speech, but sounds somewhat formal and does not occur regularly. (39) v: lotf-e dæwlæt-i ´ væ šahinša favor-ez government-m.obl and shah ‘the favor of the government and the shah’ (40) cp: r6>s v6 avaz b6 miyan b-amo dance and song to middle t/a-came ‘The dancing and singing began.’ (Rastorgueva et al. 1971: 234)10

10.The Gilaki texts in Rastorgueva et al. [1971] have a much higher occurrence of væ than I would expect. Rather than showing a true difference in Gilaki vs. other languages of Iran, however, this is probably due more to the fact that the texts seem to be composed colloquial narrations rather than transcriptions of spontaneous oral texts.

Coordination in three Western Iranian languages 289

(41) g:

duxt6r-an v6 p6s6r-an r6>s bu-kúd-idi. girl-pl and boy-pl dance t/a-made-3p ‘The girls and the boys danced.’ (Rastorgueva et al. 1971: 234)

While (42) shows the variant væ in Vafsi, it is also another instance of the extraposition of a coordinand to postverbal position discussed in § 2.1, point iii and in sentences (10)–(12) that also appear there: (42) v: fæqæt yey qædr-i ´ mærhæmmæt is=ær=goæ væ tæ:mir. only one bit-indef consideration 3s.obl=t/a=want and repair ‘It only needs a bit of consideration and repairs.’ 2.2.3 Comitative coordinators: =ra ‘with, for, and’; hemra ~ henra ‘with, and’ =ra and hemra ~ henra are not uncommon in Vafsi as coordinators but are definitely less common than =ò. Their typology as coordinators is generally monosyndetic: A B-co. With comitative conjunction the verb usually still shows plural agreement (cf. Russian and other languages with this construction, and passim in this volume). This construction occurs in both known subdialects of Vafsi. The following are examples of these two adpositions in non-coordinating functions: Benefactive (=ra): (43) v: tæmen=ra yey danæ sænduq bæ-saz. ´ I.obl=for 1 clf trunk t/a-make ‘Make a trunk for me.’ Instrumental (=ra): (44) v: soan-e=ra bez-e šax=es tíz=a kærdæ. ` file-f.obl=with goat-f.obl horn=3s.obl sharp-change made ‘He sharpened the goat’s horns with a file.’ Purely Comitative (with no conjunctive sense) (=ra): (45) v: æz soan=ra ma:melæ` næ-r-kær-om. ´ I you=with transaction neg-t/a-do-1s ‘I won’t make deals with you.’

290 Donald Stilo

hemra ~ henra (as comitative): As a preposition: (46) v: xær-i=s ´ vær bæ-veynda=o ´ hemra gændem donkey-m.obl=3s.obl ahead t/a-drove=and with wheat ór=es=girætt=o bæ-ss-e. ´ pvb=3s.obl=took=and t/a-went-3s ‘He drove the donkey on and took it with the wheat and left.’ As a postposition: (47) v: pateša in kelj-i hemra -t-a-Ø. king this girl-f.obl with t/a-come-3s ‘The king comes with this girl.’ Conjunctive examples (all examples are in the Gurchani dialect of Vafsi): (48) v: in kecæl-i=ra ´ ráxd=es dægíš ær-kær-ènd. he Baldie-m.obl=with clothes=3s.obl change t/a-make-3p ‘He and Baldie exchange clothes.’ (49) v: in zelle ræfíq-i=s=ra or-éysd-ende this woman friend-m.obl=3s.obl=with pvb-arose-3p šu=s sær=esan há-biri. husband=3s.obl head=3p.obl pvb-cut ‘The woman and her boyfriend got up and cut off her husband’s head.’ (50) shows that comitative coordinators are not necessarily restricted to two coordinands (the double marking of coordination on the third coordinand here is also discussed in § 2.2.4 below): (50) v: in kærg-é=ra xorús-i=ra=o hay éna-éna he chicken-f.obl=with rooster-m.obl=with=and intens thus-thus xošdæbi -r-kær-ènde. gamboling t/a-make-3p ‘He and the rooster and the chicken go on gamboling about this way and that.’ As with any subject in Vafsi, comitative coordinated subjects can also occasionally be extraposed to post-verbal position for purposes of focus: (51) v: bæ:d-æz kará-i ke hár=esan=kærdæ, in after-from things-indef sub pvb=3p.obl=did she ræféq-i=s=ra,, ór-eysd-ændi=o… friend-m.obl=3s.obl=with pvb-arose-3p=and… ‘After the things they did, she and her boyfriend, they got up and …’

Coordination in three Western Iranian languages 291

Occasionally the postposition =ra in Vafsi is also found in a split inclusory construction as discussed by Haspelmath in the Introduction to this volume (as well as in Haspelmath (to appear)). In this case only the second participant is overtly expressed and is marked with the comitative postposition but the first participant is implied because the verb still shows plural agreement: (52) v: b-úri nær.taxdey káwa kær-am ` esdæ=ra. t/a-come backgammon play do-1p you.obl=with ‘Let’s play backgammon, you and I.’ (53) v: b-úri esdæ=ra nær.taxdey káwa kær-àm. t/a-come you.obl=with backgammon play do-1p ‘Come, let’s you and I play backgammon.’ Occasionally hemra~henra is used in the inclusory construction: (54) v: in bæ-ss-e ´ tine zen-i=s hemra b-ótt-ænde. he t/a-went-3s his wife-f.obl=3s.obl with t/a-slept-3p ‘He1 went and slept with his2 wife.’ (double marking of pronominal possessive) Note that the subject and first verb in (54) are singular but, even though there is no other overt subject, the second verb is plural after the sentence picks up a comitative. Comitative coordination is not only used for subjects as we see in sentence (55) in the next section where a coordinator is linking two genitives. 2.2.4 Double coordination Vafsi very occasionally has double coordination with comitative coordinators, i.e., both a comitative and an ‘and’ coordinator as in (50) above. Since it occurs so rarely (only twice, at least so far, in my corpora of Vafsi — and of course either one or both occurrences could even be a matter of a slip of the tongue) it is difficult to make any wider generalizations about the typology or function of double coordination, other than to state simply what the typological structures of these two examples are. As mentioned above under point (x) in § 2.1, the typology of (50) is A B-co-co, that is, A B-cowith-coand. (55), which also has double encoding of coordination, however, is slightly different. It exhibits A co-B-co, that is, A coand-B-cowith typology. Note, however, that the second coordinand is extraposed and both coordinators jointly frame the second coordinand: (55) v: haji dæsd=es dæ-bæsdæ=o ´ zell-i henra. Haji hands=3s.obl pvb-bound=and woman-f.obl with ‘He tied the Haji’s and the woman’s hands.’ (i.e., separately from each other) (subdialect of Vafs)

292 Donald Stilo

2.3 Verb phrases and clauses VPs and clauses are also A co-B. As with NPs, the main conjunctive coordinator is =ò, with væ as a much less common variant in speech. Forms of ‘also’ are also quite commonly used for coordinating VPs and clauses (see § 6.1, point ii). Clauses may be coordinated more often with the bisyndetic hæm… hæm… than NPs are. Conjoined verbs and clauses, however, as opposed to nouns, are very often asyndetic. Deletion of =ò is not obligatory but when it is deleted, the Sustained intonation is still retained, thus becoming the significant factor in indicating coordination.11 2.3.1 NPs vs. VPs/clauses Since the languages presented in this article are all pro-drop languages, it is really not possible to make a distinction between conjoined verb phrases and conjoined clauses. I have tried in the examples below, however, to separate verbs with same subjects from verbs with different subjects. Contrary to the tendency Mithun (1988: 337) describes whereby coordinators are more commonly deleted from NPs than clauses, VPG coordinators are more commonly deleted between VPs and/or clauses than between NPs — even lexicalized NP compounds, e.g., pedær=o ´ madær ‘parents’ (see also (25)–(28) above). The VPG pattern is more similar to the situation Mithun (1988: 350) quotes for African languages. Note the deletion of the coordinator (but not Sustained intonation) in: ´ kærd ` mæsjed. (56) v: dua=s bæ-værdæ, ´ ^ pært=es prayer=3s.obl t/a-took throw=3s.obl made mosque ‘He took the prayer (and) tossed it into the mosque.’ The tendency to total asyndesis (including lack of Sustained intonation), however, is clearly more common with conceptually unitary events. ‘[Clauses] conjoined with no intonation break typically describe subparts of what is conceived as a single event’ (Mithun 1988: 335), and since I cannot see any true distinction in VPG between VPs and clauses, the statement applies equally well to VPs:

11.Since the Sustained intonation is still an indicator of coordination when an overt coordinator is deleted, we cannot claim this to be true asyndesis. It should also be kept in mind, however, that Sustained intonation is not used exclusively with coordination, but has other functions in other grammatical domains.

Coordination in three Western Iranian languages 293

(57) v: yéki=æm qazán-æ -t-ar-i, in šey sabún-i=ra hæy one-also pot-f.dir t/a-bring-2s this shirt soap-m.obl=with intens ær-mal-i isvi ær-bu-Ø, ær-mal-i dæ pišd-e qazan-é. t/a-rub-2s white t/a-become-3s t/a-rub-2s to back-ez pot-f.obl ‘Then you bring a pot, too, and you keep rubbing this shirt with soap (and) it gets real white (and then) you rub it on the pot (to get it dirty again).’ We cannot assume in (57) that the acts of bringing the pot and washing the shirt logically follow from one another. They are thus conjoined with Sustained intonation since ‘…clauses separated by comma intonation typically represent conceptually distinct aspects of an action, event, or scene’ (Mithun 1988: 335). The two actions of washing the shirt and its becoming clean are clearly seen as a unitary action, thus conjoined completely asyndetically. But the two actions, again not necessarily expected ones, of washing the shirt and then rubbing it on the (dirty) pot are conjoined with Sustained intonation. Examples of Verb Phrases and Clauses are introduced in the next two subsections. Coordinator not deleted: (usually with Sustained intonation retained). Not deleted, same subjects: Actions usually associated, =ò, Sustained intonation (58) g:

xurd-i-m=u, xuft-i-m. eat-t/a-1s=and slept-t/a-1s p: mí-xord-æm=o, mí-xabid-æm. t/a-eat-1s=and t/a-slept-1s v: im=æd-ord=o, æd-ote-ym. 1s.obl=t/a=eat=and t/a-slept-1s ‘I used to (just) eat and sleep.’

b6-k´6ft-6=vu, b6-šk´6st-6. t/a-fell-3s=and t/a-broke-3s oftad=o, šekæst. fell=and broke dæ-kætt=o, ´ bæ-škia. ´ pvb-fell=and t/a-broke ‘It fell and broke.’

Actions usually associated, =ò, no Sustained intonation Verb combinations that are usually associated are occasionally heard with Sustained intonation as above, but usually lack the Sustained intonation because they are serial verbs (and pronounced more quickly?). (59) v: tæ æsb-i ´ ú-gir=o b-úri. you horse-m.obl pvb-take=and t/a-come! ‘You bring the horse here.’ (lit: take the horse and come) (58) and (59) show, however, that closely related actions or even unitary actions are not necessarily asyndetically conjoined. The examples in (58) have both an overt coordinator and Sustained intonation and the verbs in (59) are only conjoined by a coordinator but no special intonation. Thus even verbs that are syntactically and semantically closely connected can be conjoined by coordinators.

294 Donald Stilo

As mentioned above, CP and most other languages of western Iran have serial-like verbs in which the second verb is ‘come’ or ‘go’ and is used to indicate the direction of the first verb towards and away from the speaker respectively. Even these verbs, which are thus syntactic units, may occur either (occasionally) with a coordinator (as in (59)) or (usually) without overt coordinators (as in (60) and (61)): (60) cp: bé-riz bé-r-e. t/a-pour t/a-go-3s v: bæ-riz ´ bæ-sso-Ø. ´ t/a-pour t/a-go-3s ‘Pour it out.’

bé-ndaz bí-a-d. t/a-drop t/a-come-3s há-veyn b-a-Ø. pvb-drop t/a-come-3s ‘Drop it this way; drop it down towards me.’

(61) v: há-xiliz-en b-a-Ø. pvb-slide-caus t/a-come-3s ‘Slide it this way.’

há-xiliz-en bæ-sso-Ø. ´ pvb-slide-caus t/a-go-3s ‘Slide it that way.’

Not Deleted, different subjects: (62) v: mu-sær-i=s ´ pak isbi ve=o næzdik-bin be. hair-head-m.obl=3s.obl all white was=and near-sighted was ‘The hair on his head was all white and he was nearsighted.’ (63) v: sændæli=san há-nnia-vo in há-nešesd. chair=3p.obl pvb-put=and he pvb-sat ‘They put down a chair and he sat down.’ Coordinator deleted: (usually with retained Sustained intonation). Conjunction Deleted, same subjects As mentioned above, coordinators are often deleted in VPs and clauses, but Sustained intonation usually remains: (64) g: da ruz murax6si dar-6m, ^ Ø-xay-6m a ruz-án=a cp: dæ ruz moraxæsi dar-æm, ^ mí-xa-m un ruz-á=ro 10 days leave have-1s t/a-want-1s that day-pl=do r´6št-6 mian bú-guz6ran-6m. tu ræšt bé-gzærun-æm. in R. t/a-pass-1s ‘I have ten (or) fifteen days vacation (and) I want to spend those days in Rasht.’ (65) v: kæræ-æsæl dær=om=mala ´ dæ dív-e nán-i, butter-honey pvb=1s.obl=spread on face-ez bread-m.obl b=ím=ordæ t/a=1s.obl=ate ‘I spread butter and honey on the bread (and) ate (it).’


Coordination in three Western Iranian languages 295

Occasionally both Sustained intonation and the coordinator are omitted in VPG, yielding truly asyndetic sentences as in (66) (see also (73), (79) and others): (66) v: kærey=s bæ-værd ´ molla=y ^ hár=es=da butter=3s.obl t/a-took pvb=3s.obl gave mulla-m.obl ‘He took the butter (and) gave it to a mulla.’ (comma for Sustained intonation is lacking) Conjunction Deleted, different subjects: (67) v: gázæ=s b-árdey, ´ dænnán=es bæ-keša. ´ ^ værg-i pincers=3s.obl t/a-brought wolf-m.obl teeth=3s.obl t/a-pulled ‘She brought the pincers (and) he pulled the wolf ’s teeth out.’ (68) p: púst=eš=o kænd, ^ xóšk šod. bark=3s.obl=do dry became ‘It (the flood) pulled off the bark (and) it (the tree) dried up.’ (69) g:

šime xan6-bija i=ta m6>az6 va-bóst6, ^ úya šoon-d6r-6m. your house-near 1-clf store open-became there go-prog-am cp: næzdík-e xune-ye šoma ye-dune mæqaze baz šode, ^ unja near-ez house-ez you 1-clf store open-became there dàr-æm mí-r-æm. prog-1s t/a-go-1s ‘A store has opened up near your house (and) I am going there.’

Thus in a sequence of two clauses with no overt coordinator, the key to distinguishing coordinated clauses from separate sentences is the sequence of Sustained + Declarative intonations in the former vs. two Declarative intonations in the latter as in the next example, each with two conjoined verbs (see also (8)): (70) v: luas b-áw-e ´ in.tinan payæ=san ^ b=ís=bærdæ. fox t/a-came-3s t/a=3s.obl=took they.obl stick=3p.obl ú-girætt, ´ in luás-i dómbæ. ^ dæ-kætt-ænd pvb-picked.up pvb-fell-3p this fox-m.obl after ‘A fox came (and) took (a chicken). They got a pole (and) chased the fox.’ In many cases, the alternatives of two clauses connected by a Sustained intonation vs. two separate Declarative intonations indicating separate sentences are equal options: Sustained intonation connecting two clauses (71) cp: un ráh=o pedær=æm ´ ræft, bæradær=æm ´ ræft. g: u rá=y6 mi per bu-šo, mi berar bu-šo. that road=do my father t/a-went my brother t/a-went ‘My father went that route (and) my brother went (it).’

296 Donald Stilo

Two separate Declarative intonations (72) cp: un ráh=o pedær=æm ´ ræft. bæradær=æm ´ ræft. g: u rá=y6 mi per bu-šo. mi berar bu-šo. that road=do my father t/a-went my brother t/a-went ‘My father went that route. My brother went (it).’ Sentence (73) is a good example of how =ò is used more for the phrase level than the clause level, i.e., noun=and noun, verb=and verb, clause ^ clause, etc. A longer and more detailed statistical analysis may show whether coordinators are deleted from the verb phrase as often as from clauses or whether verb phrases behave somewhere between noun phrases and clauses. (73) v: læbas=o yey sæd toæn=esan há-da tini ^ clothes=and 1 100 tuman=3p.obl pvb-gave him ú-gira=o b-áwe. pvb-took=and t/a-came ‘They gave him clothes and 100 tumans (and) he took (them) and left.’ In (74), we find two clauses (different subjects), no coordinator and an ellipted verb: (74) v: æsb molla-y [b-æwæ], ´ væzir-i [ ]. ^ xær horse mullah-m.obl [t/a-be donkey vazir-m.obl ‘The horse should go [lit: be] to the mulla and the donkey to the Vazir.’ 2.3.2 Lists of verbs As opposed to a list of nouns, where each member almost always takes a coordinator, a long list of verbs may or may not be connected with overt coordinators (each verb, however, receives stress and usually Sustained intonation). While there is a preference in some styles to omit the coordination until the end of the list (see (77)), it is not uncommon to find all (or most) members of a list or sequence conjoined by coordination as in the next two examples: (75) v: áwæ=s bæ-paša, ´ kærd=o, ^ jaru=s water=3s.obl t/a-sprinkled broom=3s.obl did=and dær=es=rua qæšeng=o, qeylán=es pvb=3s.obl=swept beautiful=and water.pipe=3s.obl b=ís=keša=vo, jens-a suræt=es t/a=3s.obl=smoked=and good-pl inventory=3s.obl ú-girættæ=o, hær-ci=s bæ-nivisdæ=o, ´ hæft sal in pvb-took=and every-what=3s.obl t/a-wrote=and 7 year this

Coordination in three Western Iranian languages 297

dokan=dæ há-nešesdæ=vo, kár=es ær-kærd. shop-in pvb-sat=and work=3s.obl t/a-did ‘He sprinkled water, swept and swept nicely and he smoked his water pipe and took inventory of the goods and wrote everything down and sat in this shop for 7 years and worked (there).’ (76) fwp:

zæmin=ešan=ra ba xiš šóxm mì-zæn-ænd=o land=3p.obl=do with ploughshare plough t/a-hit-3p=and bær sær-e ´ tæqsím-e áb=eš hæmiše dæ’va upon on-ez division-ez water=3s.obl always fight bær-pa=st=o mærdóm=eš be nodræt hæmmám upon-foot-is=and people=3s.obl to rarity hæmmam dàr-ænd=o cai=šan=ra ba xorma mí-xor-ænd. have-3p=and tea=3p.obl=do with date t/a-eat-3p ‘They plough their land with ploughshares and there is always a fight afoot over the division of water (of the village) and the people there rarely have a hammam (public bath) and they drink their tea with dates.’ (Al-e Ahmad 1973: 6)

(77) v: arˇu árd=i -t-eysen-om, ^ nán=i -r-besd-om, ^ today flour-your t/a-knead-1s bread=2s.obl t/a-bake-1s æd-d-om hazún=i=o, -t-š-om. t/a-give-1s oblpr=2s.obl=and t/a-go-1s ‘Today I will make your dough, bake your bread, give it to you and go.’ =ò is also often deleted from lists of verbs even in cases of different subjects: (78) v: kolaxúd=om hár=om=nia sær=om, ´ ´ ^ xænjær=om helmet=1s.obl pvb=1s.obl=put head=1s.obl dagger=1s.obl bæ-zæ ´ kæmær=om, ´ ´ bær-k ´ ærd, ` ^ zín=es ^ æsb=es t/a-hit belt=1s.obl horse=3s.obl out-made saddle=3s.obl æsb-i ´ bæ-v-im=o ´ bæ-ss-im ´ kærd, ´ ^ sævár-e made mounted-ez horse-m.obl t/a-got-1s=and t/a-went-1s táq nià. ^ bær-dærvaze-y=m door-gate-m.obl=1s.obl open put ‘I put my helmet on my head, stuck the dagger in my belt, she brought out the horse, saddled it up, I got on the horse, and went (and) opened the gate.’ =ò can be sporadic and unpredictable in a list of verbs. Sustained intonation is often optional, depending on the rate of speed of speech, in which case, as in the next example, the sentence has asyndetic elements:

298 Donald Stilo

(79) v: in hær ru æt-š-e ælæf=es ^ æ-cærray=o, she every day t/a-went-3s t/a-grazed=and grass=3s.obl æt-arda ær-kærdey=o, in bezi-an ^ šir=es t/a-brought milk=3s.obl t/a-made=and this goat-pl.obl æd-ordæ. t/a-ate ‘Every day (the nanny-goat) would go (and) graze and bring grass (and) make milk and the kids would eat.’ In analyzing (77) it would be tempting to make a generalization that only the final verb in a series is commonly connected by =ò, but this pattern is not a regular occurrence, as seen in (78). Sometimes two unrelated clauses are coordinated but represent a break in the narrative and have no relationship to each other: kær-ènde meydan-dey=o, in xæbær æz kigi (80) v: dar bær-pá gallows upon-foot make-3p square-in=and this news from whom æ-d-om? ´ pvb-give-1s ‘They set up the gallows in the square, and who shall I tell you about (now)?’ 2.4 Subordinator ke as coordinator It must be stressed from the outset of this discussion that the use of the subordinating particle ke as a coordinator is rare. It clearly occurs in this latter function in all of these (spoken) languages, but it must be observed in spontaneous speech since elicitation may cause confused reactions or outright rejection from many native speakers. The following examples are some of the occurrences of this use of ke taken from four hours of recorded Vafsi folk tales. The construction would definitely arise in similar amounts of recorded spoken Persian or other languages of Iran. Examples: (81) v: ær-vaz-e ‘hár=i=kærdæ?’ ke -r-vaz-e ‘la.’ t/a-say-3s pvb=2s.obl=did sub t/a-say-3s ‘yes’ ‘She asks, “Did you do it?” and he says, “Yes.”’ (82) v: tofæng bæ-zæ ´ ke tofæng=es ´ næ-r-vešaz-uæ-Ø. ´ gun t/a-hit sub gun=3s.obl neg-t/a-open-inch-3s ‘He shot, but his gun would not work.’

Coordination in three Western Iranian languages 299

(83) v: píš-gerdà-nde tine suraq ke tiní=zi yey bár=es back-turned-3p his to sub he.obl=too 1 load=3s.obl ú-girættæ bæ-ss-e. ´ pvb-lifted t/a-went-3s ‘They went back looking for him, but he, too, had picked his load up and left.’ (84) v: æz æt-š-om miš-e -r-vaz-om ke æt-ta esdæ I t/a-go-1s ewe-f.obl t/a-say-1s sub t/a-comes you.obl æ-cærr-e a! t/a-graze-3s warn ‘Watch out, I’ll tell the sheep [lit: ewe] and she will come and graze on you.’ (85) v: ægærce ´ xeyr ke bidar=e,, an tæklíf=es mæ:lú ær-kær-òm.’ if no sub awake-is that task=3s.obl clear t/a-make-1s ‘If not, and she is asleep, I will tell you what to do.’ The use of ke as a coordinator vs. its predominant usage as a subordinator are contrasted in (86) (originally collected in Gilaki but the Vafsi and CP equivalents are equally natural): a. Coordinator b. Subordinator næ-šti-m ´ ke fæ-æn-dæ-m. ´ næ-šti-m ´ ke fæ-d-6m. ´ neg-had-1s sub pvb-neg-gave-1s neg-had-1s sub pvb-gave-1s cp: næ-dašt-æm ´ ke næ-dad-æm. ´ næ-dašt-æm ´ ke bé-d-æm. neg-had-1s sub neg-gave-1s neg-had-1s sub t/a-give-1s v: n=ím=dærdæ ke n=ím=da. n=ím=dærdæ ke há-d-om. neg=1s.obl=had sub neg=1s.obl=gave neg=1s.obl=had sub pvb-give-1s ‘I didn’t have (it) and I didn’t give (it to anyone).’ ‘I didn’t have (it) to give.’

(86) g:

In the intonational pattern in (86a), the two coordinands are independent clauses and take two separate declarative intonation contours, each with clause-level stress, whereas (86b) takes one declarative intonation contour with one clause-level stress (although secondary word stress does appear on the dependent verb). The two clause-level stresses in (86a) are crucial because this pattern contrasts with another completely different intonational pattern with one clause-level stress that can appear on exactly the same elements as in (86a) and convey a completely different meaning from either of those represented here.

300 Donald Stilo


Adversative coordinative conjunctions

3.1 Main points The following points can be made about the adversative coordinate conjunctions: (i) væli ´ and æmma ´ may be used interchangeably in VPG as well as FWP. An erudite form, vælikæn, ´ similar to English ‘however’, is also used, mostly in more formal speech. (ii) The use of adversatives occurs most commonly on the clause level: (87) v: par áwæ kem be væli ´ ænsal ferawan=e. cp: parsal ab kæm bud væli ´ emsal færavun=e. g: parsal ab k6m bu væli ´ imsal f6ravan=6. last.year water little was but this.year plentiful=is ‘There was little water last year, but this year it is plentiful.’ (88) g:

m6dr6s6 vá=b-e ámma næ-n-im ´ key. school open=become-3s but neg-know-1p when cp: mædrese báz mì-š-e æmma ´ né-mi-dun-im key. school open t/a-become-3s but neg-t/a-know-1p when v: mædræsæ ær-vešaz-úæ-Ø æmma ´ næ-j-jan-am ´ key. school t/a-open-inch-3s but neg-t/a-know-1p when ‘The school will open, but we don’t know when.’

(iii) væli ´ and æmma ´ connect two equal verbs or two clauses, but are not used in NPs. (89) v: mánd=a -r-bu-æ væli ´ bi kár ær-kær-è. tired-change t/a-get-3s but again work t/a-do-es ‘He gets tired but works anyway.’ (iv) bælke ´ is used after a negative to contrast any two equal NPs, VPs or clauses: (90) cp: pæræstar nist bælke ´ doktor=e. nurse isn’t but doctor-is ‘She’s not a nurse, but a doctor.’ In actuality, however, this coordinator — as well as the bisyndetic næ tænha… bælke… ´ (=hæm), which also contains bælke ´ — is not very common in speech. The usual strategy for this type of statement is simply two separate sentences: (91) cp: pæræstar nist. doktor=e. nurse isn’t doctor-is ‘She’s not a nurse. She’s a doctor.’

Coordination in three Western Iranian languages 301

(v) Ellipsis: The use of an adversative with an elliptical clause — ‘He speaks French but very little’ — has the same possibilities as in English, but with somewhat different structures in Iranian languages. Since VPG are all SOV, the term ‘gapping’ does not apply in the same way it does to English and other SVO languages but rather ellipsis is almost exclusively analiptic, i.e., forward elliptic (Haspelmath to appear): (92) cp: færanse [hærf ´ mì-zæn-e] væli ´ xéyli kæm [ ]. French [speech t/a-hit-3s but very little ‘He [speaks] French but [ ] very little.’ (vi) The use of ‘but’ in English connecting two adjectives or two adverbs — ‘he looks tired but happy’ — occasionally occurs in VPG, but sounds rather awkward and unnatural in normal speech. This construction may be expressed via extraposition of the second coordinand with ellipsis of the verb as in the previous point, but even then it is not a common pattern in these languages: (93) cp: ?xæste væli ´ xošhal be næzær ´ mì-a-d. (?)xæste be næzær ´ mì-a-d væli ´ xošhal. tired to eye t/a-come-3s but happy ‘He looks tired but happy.’ The question mark in the first example is not so much that its grammaticality is in question, but indicates rather a certain awkwardness of expression and is generally avoided. The second form is somewhat less questionable, but still not quite usual. This construction does occur more commonly, however, with the copula, but with the second verb usually included, i.e., ellipsis is not very common with the copula: (94) cp: xæstæ=st væli ´ xošhal=e. tired-is but happy-is ‘He is tired but (is) happy.’ Though not unheard in speech, this pattern belongs more to the written domain: (95) fwp:tæht-e ´ tæ’sír-e še’ár-e puya æmma arám-e xiš under-ez influence-ez slogan-ez searching but calm-ez own qærár=dàd… placing=gave ‘…placed (the younger talents) under the influence of his own inquiring, but calm, maxims…’ (Zarrin-kub 1996: 543) (vii) As with =ò, adversative conjunctions may be deleted but the Sustained intonation is retained, thus still necessitating an analysis of syndesis:

302 Donald Stilo

(96) v: boxaru=m ó-vešena, ^ bi særd=om ´ be. heater=1s.obl pvb-lit again cold=1s.obl was ‘I lit the heater, (but) I was still cold.’ (97) cp: bæra=š qæza ovórd-æm, ^ næ-xord. ´ for=3s.obl food brought-1s neg-ate ‘I brought him some food, (but) he didn’t eat (it).’ ´ (98) lj: mu diruz h6s´6n-7 b´6-dy-7m, ^ u mæ næ-de. I yesterday H.=do t/a-saw-1s he me neg-saw ‘I saw H. yesterday, (but) he didn’t see me.’ (viii) Statistics of occurrence: There are about 50 cases in the Gurchani folk tales (4 hours of recorded speech) where I have interpreted adversative syndesis in the absence of a coordinate conjunction. By contrast, there are only 8 cases where an overt conjunction (væli ´ and æmma ´ ~ ámma, four times each) is used. This shows a very high ratio of deleted to overt adversative conjunctions (±50 to 8) for four hours of spoken corpus (folk narrative). In addition, three pages of transcription in the Vafsi subdialect of Vafs (one conversation, one personal narrative) show no examples of any overt adversative conjunction. The low functional yield of overt adversatives in Vafsi is reminiscent of the situation in Lezgian — a possible candidate for an areal feature. In Lezgian there seems to be no native way of expressing adversative coordination and the forms in the literary language are borrowed from Arabic (as are the forms in VPG). In fact speakers of Lezgian who are not literate in Lezgian may not even be familiar with the Arabic terms when asked (Martin Haspelmath, personal communication). (ix) væli, ´ æmma ´ in initial position: In addition to connecting two coordinate clauses directly, adversative conjunctions often begin an intonational unit (see also point vii of § 2.1). In many cases the clause beginning with ‘but’ is either somewhat removed from the clause it is connecting or actually begins an independent utterance in a conversation (as in the English translations as well): (99) cp: Speaker A:

Speaker B:

mæn færda ba šoma mí-a-m. I tomorrow with you t/a-come-1s ‘I’m coming with you tomorrow.’ væli ´ dìr næ-kon-i=a! ´ but late neg-make-2s-warn ‘But don’t be late (or else)!’

Adversative conjunctions, especially initially, may be replaced by fæqæt ‘only’ or tænha ‘alone, sole, only’ (cf. English but ~ only, French mais ~ seulement, Russian no ~ tol’ko, Georgian magram ~ mxolod, Azeri ámma ~ anjax, etc.), e.g., an alternate to (99B): fæqæt dìr nækoni=a! ´ ‘Only don’t be late (or else)!’

Coordination in three Western Iranian languages 303

(x) væli, ´ æmma ´ in final position: the conjunctions væli, ´ æmma ´ may occasionally occur as tags at the end of a clause, but only in the case of intonational units standing apart as mentioned in the previous point. They can never occur in clausefinal position if they are directly coordinating two consecutive clauses that are related intonational units. Clause-final coordinator tags are most typically found in conversational discourse where the independent utterances of this type are more commonly found. There are no occurrences of clause-final væli, ´ æmma ´ in any of the Vafsi folk tales since this pattern is not usually found in longer narratives, especially folk tales. They would conceivably be found in more personal narratives that are told in a more relaxed, almost conversational style. This pattern is common in CP, but in all these cases it is also a matter of individual style and idiolect. Example: (100) cp: Speaker A:

Speaker B:

in qæza ceqæd gerún=e! this food how.much expensive-is ‘How expensive this food is!’ xošmæzæ=st væli ´ alternate: xošmæzæ=st æmma ´ tasty-is but ‘It’s tasty, though.’

3.2 The adversative coordinators Conjunction not deleted: 3.2.1 væli ´ ‘but’ (101) cp: do dæf ’e pul dád-ænd bæráye tæ:mir væli ´ tæ:mír=eš 2 time money gave-3p for repair but repair=3s.obl na-tæmum=e. un-finished-is ‘Twice they paid for the repairs, but the repairs are incomplete.’ (102) lj: >eblen gušt xórd-Ø-6m, v´6li ælney de nó-xo-n-6m. before meat ate-t/a-1s but now more neg-eat-t/a-1s ‘I used to eat meat before, but now I don’t (eat it) any more.’ (103) v: in dærviš=e væli ´ dærviš-e gæda ne. this dervish-is but dervish-ez beggar isn’t ‘He is a dervish but not a begging dervish.’ 3.2.2 æmma ´ ‘but’ (104) v: sæd toæn=e, æmma ´ do-ezar kæm=e. 100 toman-is but 2 crown little-is ‘It is 100 tomans, but it is two crowns short.’

304 Donald Stilo

(105) v: zarrú-e=s pak teran=dær=ende ámma in ke child-pl.dir=3s.obl all Tehran-in-are but this house ór=es=vušuttæ. pvb=3s.obl=ruined ‘His children are all in Tehran but it (i.e., the flood) destroyed the house.’ (106) cp: telefón kærd-æm ` æmma ´ xune næ-bud. ´ telephone did-1s but house neg-was ‘I called but he wasn’t home.’ (107) g:

u duxt6r æni ´ zey=I ámma un p6s´6r=æ n´6-š6nas-6m. that girl his child-is but that boy=do neg-know-1s ‘That girl is his/her daughter, but I don’t know that boy.’

3.2.3 vælikæn ´ ‘but, however’ (108) v: ærzéš=es pansæd toæn=e, vælikæn ´ in zævin xéyli æz in value=3s.obl 500 toman-is however this land very from this behtær=e. better-is ‘It is worth 500 tomans, however this (plot of) land is much better than that.’ 3.3 Initial occurrence in intonational unit Sequential but separate intonational units (109) v: æz hær mivéi ke ru-ye doniay=dæ he,, væws=dæ behtær=e. from every fruit sub on-ez world-in exists Vafs-in better-is væli ´ ær-go ke be suræt-e ´ æmæli obær ar-i. but t/a-must sub to fashion-ez practical out bring-2s ‘Of all the fruits that exist on the earth, there are better ones in Vafs. But you have to develop them in a practicable way.’ (110) cp: ra=š xéyli dur mí-oft-e. æmma ´ dær-suræti ´ ke in pol v: ra=s xéyli dur ær-gen-e. æmma ´ dær-suræti ´ ke in pol road=3s.obl very far t/a-fall-3s but in-case sub this bridge saxte-be-š-e,, un væxt-a jadde xub mí-š-e bæ-saz-uæ-Ø,, ´ an vaxt-an jæddæ næc ær-bu-Ø t/a-build-inch-3s that time-pl.obl highway good t/a-get-3s ‘It’s a long way. But if this bridge gets built, then the highway will be good.’

Coordination in three Western Iranian languages 305

(111) g:

m6n n´6-t6rs-6m. ámma xéyli vásti m6vaz´6b bu-òn. I neg-fear-1s but very must careful be-inf ‘I am not afraid. But you have to be very careful.’

3.4 Adversative conjunctions deleted (112) cp: ye cíz-i porsíd-æm, ^ né-mi-dunest. 1 thing-ind asked-1s neg-t/a-knew v: yey ci=m há-pærsa, ^ n=ís=æj-jana. 1 what=1s.obl pvb-asked neg=3s.obl=t/a-knew ‘I asked (him/her) something, (but) s/he didn’t know (the answer).’ næ-šod. ´ (113) cp: mí-xast beh=éš bé-g-e, ^ ru=š t/a-wanted to=3s.obl t/a-say-3s face=3s.obl neg-became v: is=ær-goa dæ tani bá-waz-e, ^ dìv=es næ-væ. ´ his=t/a=wanted to him t/a-say-3s face=3s.obl neg-became ‘He wanted to tell him something, (but) he didn’t have the nerve.’ næ-šnæ. (114) v: tæmen juw´6w hà-da, ^ esdæ I.obl answer pvb-gave you.obl neg-heard ‘I answered (but) you didn’t hear.’ ´ (115) v: áorˇæ=san pérr=om=a kærd, ^ næ-d-or-ond. trough=3p.obl fill=1s.obl=change made neg-t/a-eat-3p ‘I filled their trough (but) they won’t eat.’ (116) g:

bá>=æ b6-g6rd´6st-6m, ^ peyda nú-kud-6m. garden=do t/a-searched-1s find neg-make-1s ‘I searched the garden, (but) I didn’t find (it).’

´ : Contrastive after negative (Spanish sino, German sondern) 3.5 bælke As mentioned in § 3.1, point iv, the use of bælke ´ is not very common, but it does occasionally occur: (117) v: an gæsd adui næ-ve ´ bælke ´ hævi-án=es komæk ´ ær-kærdæ. ` he bad person neg-was but all-pl.obl=3s.obl help t/a-did ‘He was not a bad person, but rather he (always) helped everybody.’

306 Donald Stilo

4. Disjunctive coordinators 4.1 Phrase level 4.1.1 Usual cases of disjunctive coordinators on the phrase level There are various possibilities for fulfilling the function of disjunctive coordination on the phrase level when connecting two nouns, adverbs, etc. They differ somewhat depending on whether they appear in a statement or a question. The following general points can be made about the disjunctive coordinate conjunctions: i. Both monosyndetic and bisyndetic conjunctions are usual (see also §5.1, point vi). ii. Extraposition of the second coordinand is slightly favored. iii. Disjunctive coordinators may often be deleted, on which see § 4.1.2 below. Disjunctive coordination in statements. It is my impression that while both monosyndetic and bisyndetic disjunctive conjunctions are commonly used, there is a slight tendency to favor bisyndetic coordinators in statements, i.e., ya… ya … ‘either … or …’. Whether monosyndetic or bisyndetic, however, the second coordinand is often shifted to post-verbal position. The following are examples of all four options: Monosyndetic, not shifted: (118) cp: gúšt-e gav ya gúšt-e bærre=æm bé-gir. meat-ez cow or meat-ez lamb=also t/a-take ‘Get some beef or lamb, too.’ Bisyndetic, not shifted: (119) v: ya yey quti-e kirfít-i ya yey danæ qælæm ya yey qéyr-e or 1 box-ez match-m.obl or 1 clf pen or 1 other-ez ciz b=is=nia an kelj-i jib. thing t/a=3s.obl=put that girl-f.obl pocket ‘He would put a box of matches or a pen or something else in that girl’s pocket.’ (120) cp: hær ruz ya bæradær=eš=o ´ ya xód=eš=o mí-did-æm. every day or brother=3s.obl=do or self=3s.obl=do t/a-saw-1s ‘I would see either his brother or (him) himself every day.’ Monosyndetic, shifted: (121) cp: ye kæmi ´ šir mí-ovord, ya ye zærre æsæl. a little milk t/a-brought or a bit honey v: yey zærræ šir is=æt-ardæ ya ye zæ æsæl. a little milk he=t/a-brought or a bit honey g: i picæ šir av6rd-i-Ø ya i picæ as6l a little milk brought-t/a-3s or a bit honey ‘He would bring a little milk or a little honey.’

Coordination in three Western Iranian languages 307

Bisyndetic, shifted (probably the slightly favored option for statements): (122) v: yaqin in ya púl-e kisé-y=s hesáw kærdè, ya certain he or money-ez purse-m.obl=3s.obl account made or pul-e qaséd-i. money-ez messenger-m.obl ‘He surely (either) accounted for the cost of the purse or the delivery boy.’ (123) cp: mæn bayæd hætmæn sær-e ´ qæza ya nušabe bó-xor-æm, ya áb-e I must surely on-ez food or drink t/a-eat-1s or water-ez mive. fruit ‘I absolutely have to drink either a soft drink or fruit juice with my meal.’ Bisyndetic coordinators are discussed more fully in § 5 below. Disjunctive coordination in questions. In questions, monosyndetic ya is generally used but the second coordinand is usually shifted to post-verbal position. An unshifted variant is also possible, but is even less common than the statement form in (118). While some unshifted questions, such as (142) below, sound more acceptable (object position?), many sound awkward (subject position?) and are therefore avoided (but probably not totally ungrammatical as such): (124) cp: (?) æhmæd ya æli mí-re? fi æhmæd mí-r-e? ya æli. pn t/a-go-3s or Ali ‘Is Ahmad going or Ali?’

(unshifted, awkward) (shifted, more acceptable)

(125) cp: (?) vida pesær ya doxtær dar-e? (unshifted, awkward) fi vida pesær dar-e? ya doxtær. (shifted, more acceptable) pn son have-3s or daughter ‘Does Vida have a son or a daughter?’ Unshifted NPs in questions are even more unusual and awkward with the copula: (126) cp: ?bæcæ=šun ´ pesær ya doxtæ ´ bæcæ=šun ´ pesær=e? ya doxtær=e. ´ child=3p.obl son or girl-is child=3p.obl son-is or girl-is ‘Is their child a boy or a girl?’ 4.1.2 Deletion of disjunctive coordinators In both statements and questions, the disjunctive conjunction ya between NPs (and the like) may occasionally be deleted. Deletion of the conjunction occurs on both the phrase level and the clause level as with conjunctive coordinators. (See also § 4.1.3 below for special cases of deleted ya.)

308 Donald Stilo

(i) A coordinator between two NPs (including noun-derived adverbs of place, time, etc.) may occasionally be deleted, but the two words are still considered conjoined, because the stress pattern changes. The word stress of the first word is deleted and is taken over by that of the second word, i.e., essentially sounding like a compound word: (127) cp: emruz-færda mí-r-e. today-tomorrow t/a-go-3s ‘S/he’s going today or tomorrow.’ (ii) When the coordinator between two NPs is deleted, it is not always clear whether the deleted coordinator is conjunctive (see § 2.1, point vi above) or disjunctive; the intonational and word stress patterns are the same in both cases. (128) cp: mésl-e xahær-bæradÊr ba-hæm hærf ´ mì-zæd-ænd. like-ez sister-brother with-hÊm speech t/a-hit-3p ‘They would talk to each other like (a) brother and sister.’ (cannot be interpreted as ‘brother or sister’) (129) cp: mæn medad-xodkar be=t mí-d-æm. I pencil-pen to=2s.obl t/a-give-1s ‘I’ll give you a pencil or a pen.’ (implication: I’ll give you something to write with — one or the other but probably not both, although possibly) (130) cp: pepsi-koka be-færma-id. (= a polite offer) Pepsi co*ke t/a-command-2p ‘Have a Pepsi or a co*ke.’ (‘Have both’ would be an unlikely interpretation) In some cases of deleted coordinators, a conjunctive or disjunctive meaning is irrelevant: (131) cp: xahær-bæradÊr ` dár-in? sister-brother have-2p ‘Do you have brothersÆ or sisters≠?’ ‘Do you have a brother Æ or sister≠?’ (English stress on last word in each, not on the coordinator) (see § 4.1.3 for the use of the arrows) (Note: overt plurals in VPG are only used in combination with specificity.) Note that the choice between the first two English sentences is also irrelevant, as long as the conjunctive or disjunctive conjunction is not stressed, in which case the meaning changes. That is, ‘Do you have brothers and sisters?’ has a different meaning from the two sentences given. The interpretation ‘*Do you have a brother≠ or a sisterØ?’ is not possible for (131) (see point iii below).

Coordination in three Western Iranian languages 309

(iii) A more in-depth study needs to be conducted on coordination deletion in general and specifically in the case of disjunctive coordinators. The deletion seems to have something to do with definiteness or possibly some other grammatical factor. Note that while it is possible to say the various forms in (132) and (133) below, the starred examples in (134) and (135) sound ungrammatical or questionable: (132) cp: pepsi-koka geréft-i? ‘Did you get PepsiÆ or co*ke≠?’ (133) cp: emšæb-færda miad. ‘S/he’ll arrive today or tomorrow.’ (134) cp: *?æli-æhmæd miand. *mæn-to bérim? ‘*Is AliÆ or Ahmad≠ coming?’ ‘*Should youÆ or I≠ go?’ (135) cp: *in palto-un paltó=ro bépuš. but: in vær-un vær ræftæm. ´ ‘*Put this coat or that coat on.’ ‘I went this way and that way.’ (iv) Other intonational patterns, however, may indeed play a role in questions with deleted coordinating conjunctions. I will make the following formulation with the proviso that it is only tentative and will need further investigation: in some cases in CP, probably also VPG, certain disjunctive questions can receive their own intonational pattern. That is, the intonational pattern heard on the first coordinand in (136) does not seem to be the usual Sustained intonation (pitch level 4), but, to my ear at least, sounds lower, possibly pitch level 3 (Question) or pitch level 2 (Declarative 2). (Or is there possibly a second type of Sustained intonation?) I will mark this intonation with the symbol (,-) for the present. This intonation is optional and not all questions with deleted conjunctions — possibly the majority, but which ones? under what circ*mstances? — will have the (,-) intonation. Example: (136) cp: kár-i,færmayéš-i dár-in? job-indef command-indef have-2p ‘Is there a job or a command?, i.e., Can I do anything for you?’ My attention was drawn to the intonation on (136) only in the course of working on the final draft of this article and I will reserve any further judgment on this issue until more intensive collection and careful analysis can be conducted (preferably taken from recorded speech, but possibly also elicitation) than fits within the framework of this article. 4.1.3 Alternative option questions As opposed to the above type of Yes-No Question, which expresses a choice between two items where it is not clear if anything at all is desired — English: ‘Do you want teaÆ or cóffee≠?’ or maybe nothing — and which are joined together like

310 Donald Stilo

compounds with one stress in VPG, a disjunctive coordinator is used for an Alternative Option Question where a choice is presented (English: ‘Do you want téa≠ or cóffeeØ?’), and the second coordinand is almost always extraposed: Anything vs. Nothing

Choice between two alternatives

(137) cp: cai-qæ:ve mí-xor-i? tea-coffee t/a-drink-2s ‘Do you want teaÆ or cóffee≠?’

cai mí-xor-i? ya qæ:ve. tea t/a-drink-2s or coffee ‘Do you want téa≠ or cóffeeØ?’

(138) cp: xahær-bærad ` Êr ' dár-in? sister brother have-2p ‘Do you have brothersÆ or sisters≠?’ ‘Do you have a brotherÆ or (a) sister≠?’ ‘Do you have brothersÆ and sisters≠?’

xahÊr dár-in? ya bæradÊr. sister have-2p or brother ‘Do you have brothers≠ or sistersØ?’ ‘Do you have a brother≠ or a sisterØ?’

Related to the above two types of choices affecting the intonation, there is also a similar issue of ‘one-or-the other but I don’t care’, vs. ‘one-or-the other and I do care, or I want to know which choice between two alternatives’: (139) cp: tond-yæva¶ kár mì-kon-e? fast-slow work t/a-make-3s

tond-yæva¶, færq ` fast-slow difference né-mi-kon-e. neg-t/a-make-3s ‘Does it work fastÆ or slow≠?’ ‘Fast or slow, it doesn’t matter.’ (implication: as opposed to no variation in speed)

(140) cp:*tond-yævaš kár mì-kon-e? fi tond kár mì-kon-e? ya yæva¶. fast-slow work t/a-make-3s fast work make-3s or slow ‘Does it work fast≠ or slowØ?’ (implication: I want the information) Then when the coordinator is used it is more commonly shifted out as in the second alternative in (137) or in: (141) cp: esfæhan-širaz bé-r-im? Esfahan Shiraz t/a-go-1p ‘Shall we go to EsfahanÆ or Shiraz≠?’ (or somewhere like that, or e.g., stay home)


esfæhan bé-r-im? ya širaz. Esfahan t/a-go-1p or Shiraz ‘Shall we go to Esfahan≠ or ShirazØ?’ (i.e., which one?)

The stress and intonation patterns also contribute to marking the two types of option questions introduced above. In the unshifted version (142, left column), if

Coordination in three Western Iranian languages

there is only a clause stress (either on the last noun or on the verb), the interpretation is the first type introduced above, i.e., the compound-like type (i.e., pepsi-koka, but with the coordinator retained) where the answer sought by the speaker of the question is one of anything versus nothing. In the second unshifted subtype, the alternative option question (142, right column), where the speaker is clearly seeking a choice out of the options presented, clause stress falls on each of the options (and my impression is, strangely enough, that question intonation is converted to Declarative 1, even though it is a question). The distinction is made in the shifted questions by both intonation and stress. In the anything-versus-nothing question (143, left column), there is only one intonational unit (clause) with one clause stress — either on the first noun or on the verb — and the question intonation is extended through the extraposed coordinand to the end of the sentence. In the alternative option question (143, right column), however, there are two intonational contours, QuestionDeclarative, and two clause stresses (on the two alternatives, not on the verb): Anything vs. Nothing Unshifted (142) cp: gúšt-e gav ya gúšt-e gúšt-e gav ya gúšt-e meat-ez cow or meat-ez bærre geréft-i? bærre gere'ft-i? lamb took-2s Shifted (143) cp: gúšt-e gav geréft-i ya meat-ez cow took-2s or gúšt-e bærre? meat-ez lamb gúšt-e gav gere'ft-i ya meat-ez cow took-2s or gúšt-e bærre? meat-ez lamb ‘Did you get beefÆ or lamb≠?’

Alternative Option ?gúšt-e

gav ya gúšt-e

meat-ez cow or meat-ez bærre geréft-i. lamb took-2s gúšt-e gav geréft-i? ya meat-ez cow took-2s or gúšt-e bærre. meat-ez lamb

‘Did you get beef≠ or lambØ?’

As already mentioned in §, I would like to reiterate that there is a clear preference for the shifted question option in both cases. (This solution also eliminates the confusion over the loss of question intonation in the unshifted alternative option question, although future investigation should include a study of this construction and a solid description of its intonational pattern.)



Donald Stilo

4.2 VP and clause level 4.2.1 Disjunctive coordination of VPs and full clauses As mentioned earlier, VPs and clauses cannot really be distinguished from each other for coordination, but there are both similarities to and differences from NPs in the realm of disjunctive coordination. (i) As with NPs, monosyndetic and bisyndetic disjunctive coordinators are commonly used in VPs/clauses but with a slight tendency to favor bisyndetic coordinators in statements and monosyndetic coordinators in questions. (See also § 5, point vi.) A single ya seems to be more common when conjoining verbs, especially commands: (144) cp: bé-gu ya mí-koš-æm=et t/a-tell or t/a-kill-1s=2s.obl v: bá-waz ya i=r-koš-ome t/a-tell or 2s.obl=t/a-kill-1s ‘Tell (me) or I will kill you.’ Bisyndetic disjunctives are probably equally as common in most VP coordination: (145) cp: ya ræft-i, ´ ya næ-ræft-i. ´ kodúm=eš=e. or went-2s or neg-went-2s. which=3s.obl-is ‘Either you went or you didn’t go. Which is it?’ (ii) Extraposition of the VP is generally irrelevant because the verb is usually already in final position, but see § 4.2.3 for some occasions for shifting. (146) cp: mí-r-i? mí-a-y? t/a-go-2s t/a-come-2s ‘Are you going? Coming?’

mí-r-i(?) ya mí-a-y. t/a-go-2s or t/a-come-2s ‘Are you going or coming?’

(iii) Deletion of disjunctive coordinators is not a strategy as it is with NPs. *mí-r-i mí-a-y? t/a-go-2s t/a-come-2s ‘*(as disjunctive)’ (but can mean ‘Are you going and coming?’) (iv) The intonational patterns for VPs/clauses with ya… (ya…) in statements (147) and questions (148) are also the same as for NPs (recall also §1.4.3 on punctuation): (147) cp: færda ya xód=æm mí-a-m dombál=et, ya æhmæd=o ´ tomorrow or self=1s.obl t/a-come-1s after=2s.obl or pn=do mí-ferest-æm. t/a-send-1s ‘Tomorrow either I will come get you or I will send Ahmad.’

Coordination in three Western Iranian languages

(148) v: æ-cu ešden bæ-vær-i? ´ ya komæk=i ´ kær-òm. t/a-can self t/a-take-2s or help=2s.obl make-1s g: Ø-tan-i xúd=6t b´6-v6r-i? ya t´6=ra kum´6k kun-`6m. t/a-can-2s self=2s.obl t/a-take-2s or you-to help make-1s cp: mí-tun-i xód=et bé-bær-i? ya komæk=et ´ kon-æm. ` t/a-can-2s self=2s.obl t/a-take-2s or help=2s.obl make-1s ‘Can you carry it yourself or should I help you?’ Further information on bisyndetic coordinators is provided in § 5 below. (v) The Question-Declarative intonational pattern of the previous point may, of course, contrast with a sequence of two questions — with separate question intonations (also as in the first alternative in (146) above): Separate questions (149) cp: pul dar-i? bé-r-im money have-2s t/a-go-1p bé-gir-im? t/a-get-1p ‘Do you have money? Shall we go get (some)?’

Coordinated questions pul dar-i? ya bé-r-im money have-2s or t/a-go-1p bé-gir-im. t/a-get-1p ‘Do you have money or shall we go get (some)?’

4.2.2 Elliptical clauses An alternate analysis to extraposition/shifting of a second coordinand would be to state that each NP coordinand occurs in a separate clause and that the verb in the second clause is ellipted. Further support for the latter analysis over extraposition is that in the case of the copula, the second verb is almost never deleted in VPG. Again, it is slightly more common for statements to have a double ya and questions to have ya only before the second clause: Statement (150) cp: ya narahæt=e ´ ya æsæbani=e. or upset-is or angry-is ‘He’s (either) upset or angry.’

Question narahæt=e? ´ ya æsæbani=e. upset-is or angry-is ‘Is he upset or angry?’

The only time the copula is generally deleted is in the case of elliptical clauses that delete several elements, at least the adjective along with the copula: (151) v: kafær-e ´ ba-morovvæti ´ [næc=e]? ´ ya mosælman-e bi-morovvæti ´ [ ]. infidel-ez with-mercy [good-is or Moslem-ez without-mercy ‘Is a merciful infidel better than a merciless Moslem?’ In conversation, all other elements may be elided, leaving only the coordinands (especially in questions):


314 Donald Stilo

(152) cp: ki. mæn ya to. v: ke. æz ya tæ. g: ki. m6n ya tu. ‘Who? Me or you?’ (all declarative intonations) (153) v: yeylaq=dæ ya an zer=dæ? or that down-in ‘Up in the summer camp or down there?’

mí-bin-i ya næ? t/a-see-2s or no ‘Do you see (it) or not?’

4.2.3 Extraposition of the verb For VPs in questions (see § 4.2.1. (ii) above), if the verb is already in final position, shifting becomes irrelevant and ya simply coordinates the two verbs. However, while FWP is more rigidly verb-final, there are many grammatical categories in CP, Gilaki and Vafsi that are commonly shifted to post-verbal position in speech, in which case the second coordinated verb may then be extraposed to sentence-final position: Verb-Final (154) cp: æhmæd mí-r-e (?) ya mí-mun-e. A. t/a-go-3s or t/a-stay-3s ‘Is Ahmad going or staying?’ Non-Verb-Final with Extraposition æhmæd mí-r-e tehran (?) ya mí-mun-e. A. t/a-go-3s T. or t/a-stay-3s ‘Is Ahmad going to Tehran or staying?’ 4.2.4 Additional points (i) In addition to connecting two coordinate clauses, disjunctives (like conjunctives and adversatives) often begin an intonational unit. Often the piece beginning with ya is either somewhat removed from the first coordinand, or it can actually begin an independent turn-taking utterance, as in the following conversation: (155) cp: Sp. 1) mæn æz reza mí-pors-æm. I from R. t/a-ask-1s ‘I’ll ask Reza.’ Sp. 2) ya æz æli bé-pors. hætmæn mí-dun-e. or from A. t/a-ask surely t/a-know-3s ‘Or ask Ali. He’ll know for sure.’ (ii) In FWP (and occasionally in more formal speech), ‘or’ may be expressed with the compound conjunction væ-ya, literally ‘and-or’ (not equivalent to English ‘and/or’).

Coordination in three Western Iranian languages

4.3 The coordinator yá-ìnke ya may take the form yá-ìnke in two coordinated non-elliptic clauses (usually questions) in CP. The interesting point here is that ínke is usually used to convert adpositions into subordinate conjunctions in VPG,12 yet another connection between coordination and subordination (see § 2.4). While I have no examples of yá-ìnke in Vafsi or Gilaki, I expect it to occur regularly there also. Examples: In a statement: (156) cp: mí-tun-i ba taksi bé-r-i, yá-ìnke, æge ´ bé-xa-y,, t/a-can-2s by taxi t/a-go-2s or-sbc if t/a-want-2s piadÊ=m mí-š-e. on.foot-also t/a-become-3s ‘You can go by cab, or, if you like, walking is also possible.’ In a question: (157) cp: namæ=ro ´ xo'd=et póst mí-kon-i? ya-ínke mÊn bæra=t letter=do self=2s.obl post t/a-do-2s or-sbc I for=2s.obl póst kon-æm? ` post do-1s ‘Will you mail the letter yourself, or should I mail it for you?’


Bisyndetic coordinative conjunctions

Bisyndetic coordinators will only be discussed briefly here, as most of the material to be presented on them has already been stated in one form or another throughout this article. A list of their general characteristics and a few examples of each type should suffice. 5.1 Main points The following points, except where noted, apply to all bisyndetic coordinators (statements relating to specific coordinators are given in their respective sections): (i) They are generally parallel in Iranian languages, i.e., the two are the same —

12.Specifically ínke breaks down into two component parts: (1) in meaning “this”, a dummy NP (devoid of deictic qualities) filling the role of the required object of the adposition, and (2) the subordinator ke, which is used — among many other types of subordination — to convert a preposition to a subordinate conjunction. Thus in-ke is exactly parallel to the French ce + que in jusqu’à ce que, par ce que, etc. or Russian po-tomu ˇcto, and others.


316 Donald Stilo

hæm … hæm…, ya… ya…, næ… næ…, ce… ce…, xah… xah…, gah… gah… — with the exception of næ tænha … bælke ´ … (see § 5.2.2), and of an occasional expansion of the second member in the disjunctive, ya fi yá-ìnke. (158) cp: hæm sæhih=e hæm beja=st ‘It is both corrrect and à propos.’ ce sæhih baš-e ce beja (baš-e) ‘…whether it is correct or à propos’ ya sæhih=e ya beja=st ‘It is either correct or à propos.’ næ sæhih=e næ beja=st ‘It is neither correct nor à propos.’ not correct=is not à.propos=is (ii) On the phrase level, the second coordinator and its coordinand are more often than not shifted to post-verbal position. (iii) They may be used in both statements and questions. (159) cp: Statement: hæm caí-e mæn=o ´ xórd-i(=ò), hæm mál-e xód=et=o. Question: hæm caí-e mæn=o ´ xórd-i(=ò)? hæm mál-e xód=et=o? also tea-ez I=do ate-2s(=and) also poss-ez self=2s.obl=do ‘You drank/Did you drink both my tea and your own (tea)?’ (iv) When the second coordinand in a noun phrase is shifted to the position after the verb, the intonation patterns on the two segments are the same as those of two clauses: a.

for statements the intonational patterns on the two coordinands is SustainedDeclarative; b. for questions the intonational patterns on the two coordinands is QuestionQuestion, as opposed to the typical Question-Declarative for extraposed monosyndetic coordinands (see (159) above). (v) Bisyndetic coordinators are either conjunctive or disjunctive. The bisyndetic disjunctive is a parallel repetition of the monosyndetic form ya, whereas the bisyndetic conjunctive is not a repetition of =ò, but rather of hæm ‘also.’ (vi) An unsubstantiated impression on my part: Bisyndetic coordinators are much more common in VPG than they are in a language such as English. I cannot at present back this statement up with hard facts culled from hours of recorded natural speech, but based on my own intuition for Persian — as well as many years, though less intensive, of doing fieldwork and collecting data on the non-Persian languages of Iran — it does seem intuitively to be the case. I think the following additions to this statement, however, may offer some support to the analysis that bisyndetic coordinators may be more common in Iranian than is observed cross-linguistically: a. As we have seen thus far, these languages have a strong tendency toward extraposing the second coordinand — or, alternatively, forming an elliptic clause from it — in which case parallel bisyndetic coordinators, while not necessary, seem to fit easily into this process.

Coordination in three Western Iranian languages

b. Bisyndetic coordinators, in being more common, are not necessarily particularly emphatic or contrastive in VPG (unless specifically stressed) and do not stand out as much as in other languages. By this I mean that (162) below places less contrast or emphasis on the two coordinands than the English equivalent ‘I drink both coffee and tea.’ Haspelmath (to appear: 14) states that there are ‘languages where bisyndetic coordination is the normal, non-contrastive construction’, and VPG seem to fit neatly with this typology. Conjunctives show a difference between VPG and Daghestanian languages in that they are postposed in the latter, but disjunctives are not, cf. Dargi, Lak, Avar, etc. in van den Berg (this volume). Note, however, that I am not claiming that bisyndesis is the normal construction in VPG, but rather that it is simply more common there than, for example, in European languages. c. When emphasis is needed on the coordinator, however — particularly conjunctive coordinators, since the most common ones, =ò and encl*tic (h)æm, phonologically cannot be stressed — hæm… hæm… would be the easiest way to resolve the problem since both of its elements can be stressed. d. A tentative additional argument in support comes from the possible areal tendency to use bisyndetic coordinators more commonly than, say, European languages, cf. Daghestanian coordinators discussed in point (b) here. Stassen (2000: 12) also lists the Caucasus as one of the areas where bisyndetic coordination is the norm. A northern Iran-to-Caucasus link would not be implausible for an areal feature considering that the distance between non-Lezgic Caucasian languages and the Caspian/Tatic-Talyshi area is not that far (Lezgic is predominately monosyndetic). Areal phenomena, as well as a detailed analysis of Azerbaijani, however, are topics for a separate investigation. Such a possibility is not as far-fetched as it might seem initially given the borrowing of coordinators from Iranian in Daghestanian languages, cf. the borrowed disjunctive ya in Dargi, Avar, Andic, Tsezic, etc. as well as the first element in Dargi ya-ra (and its parallel equivalent in Avar, Lezgian, etc., cf. the Persian væ-ya introduced in § 4.2.4, point ii). Note also that WITH-coordination also occurs in Tatic and Daghestanian languages (van den Berg: this volume) (vii) The term ‘bisyndetic’ should not imply that the number of coordinands is restricted to two. There may be two or more coordinands connected by bisyndetic coordinators expressed in a sequence, although in practice the number of coordinators rarely exceeds three. See also (171), (179), and (183). (viii) It is very common to find ellipsis of repeated elements (or extraposition of the coordinand) with bisyndetic coordinators: (160) cp: næ ín=o [mí-xa-m], næ ún=o [ ]. not this=do [t/a-want-1s not that=do ‘I want neither this one nor that one.’ (more usual English: ‘I don’t want this or that one.’)



Donald Stilo

In the context of discourse, a new sentence (intonational-unit) may be made with more than one element ellipted: (161) cp: hic kodum=ešun=o [né-mi-xa-m]. næ ín=o [ ], næ none which=3p.obl=do [neg-t/a-want-1s not this=do not ún=o [ ]. that=do ‘I don’t want either one of them. Neither this nor that one.’ (ix) Except for ya… ya…, most bisyndetic coordinators — certainly hæm… hæm, næ… næ …, ce… ce…, gah… gah… — can optionally take the usual monosyndetic coordinator =ò or =(h)æm/=ìz ‘also.’ Example (see also (159) above and (182), (183) and (186) below): (162) cp: hæm qæhve [mí-xor-æm]=o hæm cai [ ]. also coffee [t/a-eat-1s=and also tea ‘I drink both coffee and tea.’ 5.2 The bisyndetic coordinators Examples and rules pertaining only to specific coordinators: 5.2.1 Conjunctive bisyndetic coordinators (hæm… (=ò) hæm…) Examples of hæm… (=ò) hæm… NP not shifted: (163) v: hæm mæ:mud-xan hæn zarrú-e=s pak teran=dær=ende. also M.-Kh. also child-pl.dir=3s.obl all T.-in-3p ‘Both Mahmud-Khan and his children are in Tehran.’ (164) cp: hæm ruzname hæm mæjælle mí-xun-æm. also newpaper also magazine t/a-read-1s g: ham ruznam6 ham m6j6ll6 Ø-xan-6m. also newpaper also magazine t/a-read-1s v: hæm ruznamæ hæm mæjællæ ær-xan-om. also newpaper also magazine t/a-read-1s ‘I read both newspapers and magazines.’ It is slightly more common to extrapose the second coordinand in the examples in (164).

Coordination in three Western Iranian languages 319

VP: (165) g:

b-iš-im s6fr6-s6r, ham bú-xor-im ham g´6b b`6-z6n-im. t/a-go-1p table-on also t/a-eat-1p also words t/a-hit-1p ‘Let’s go to the table both to eat and to talk.’ (Rastorgueva et al. 1971: 239)

(166) cp: hæm mí-xa-m(=ò), hæm né-mi-xa-m. also t/a-want-1s=and also neg-t/a-want-1s ‘I both want (to) and I don’t want (to).’ (167) fwp:

hæm zud-tær be-dæst ´ mì-ay-æd, hæm ba særma bištær also early-er to-hand t/a-come-3s, also with cold more oxt=æst. friendly-is ‘It (millet) both ripens sooner and is more compatible with the cold (weather).’

Shifted (copula and other material ellipted): (168) v: b=ís=di hæm ríš=es=dæ [aš=e], hæm t/a=3s.obl=saw also beard=3s.obl=in [soup-is also yaxæ=s=dæ [ ]. collar=3s.obl=in ‘He saw (that) there was soup both on his beard and on his collar.’ Shifted (copula not ellipted): (169) v: hæm zelle pir be, hæm merdæ pir be. also woman old was also man old was ‘Both the man and the woman were old.’ Shifted (main verb ellipted): (170) cp: hæm ín=o [mí-xa-m], hæm ún=o [ ] also this=do [t/a-want-1s also that=do ‘I want both this and that.’ Three coordinands: (171) cp: in qæza hæm peste, hæm badum=o hæm zerešk dar-e. this food also pistachio also almond=and also barberry have-3s ‘This dish has pistachios, almonds and barberries (in it).’ 5.2.2 Conjunctive bisyndetic coordinators (næ tænha… bælke… ´ (=hæm)) As mentioned above under 3.1, point iv, this coordinator is slightly less common in speech than in FWP. næ tænha… bælke… ´ (=hæm) is slightly more common than bælke ´ alone. Examples:

320 Donald Stilo

(172) fwp:

næ tænha jomæl-át-e mæ’mulí-e roman-há-ye pæst-e ´ not alone sentence-pl-ez usual-ez novel-pl-ez low-ez ešq-alud=ra tekrár mì-kærd, bælke ´ bazigær-e ´ anha love-tainted=do repetition t/a-made but actor-ez they šode-bud. become-was ‘Not only did he repeat the common sentences from vulgar love stories, but he (had) acted them out.’ (Lazard, 1992: 215)

(173) fwp:

næ tænha æz dæbestan=o behdari dær an xæbær-i ´ not alone from in it news-indef nist, bælke ´ æhali hænuz ba sæng-e ´ cæxmaq copoq-ha-ye isn’t but inhabitants still with stone-ez flint pipe-pl-ez xód=ešan=ra atéš mì-kon-ænd. self=3p.obl=do fire t/a-make-3p ‘Not only is there no sign of an elementary school or a medical clinic, but the inhabitants also still light their pipes with flintstone.’ (Al-e Ahmad: 8)

5.2.3 Disjunctive bisyndetic coordinators Affirmative: ya … ya … ‘either … or …’. NP not shifted: (174) cp: ya bærdær=eš ´ ya xód=eš be mæn goft. or brother=3s.obl or self=3s.obl to I said v: ya bera=s ya ešdén=es dæ tæmen va-Ø. or brother=3s.obl or self=3s.obl to I.obl said-3s g: ya æni berar ya xúd=eš be m6n bu-gúft. or his brother or self=3s.obl to I t/a-said ‘Either his brother or he himself told me.’ NP shifted, verb ellipted: ya ti midád=a m´6=ra [fæ-d6n] ´ ya ti >6l´6m=a [ ]. or your pencil=do I-to [pvb-give or your pen=do v: ya medád=i hazun=om [há-dæ] ya qælæm=i ´ [ ]. or pencil=2s.obl oblpr=1s.obl [pvb-give or pen=2s.obl cp: ya medád=et=o be mæn [bé-deh] ya qælæm=et=o ´ [ ]. or pencil-your=do to I [t/a-give or pen=2s.obl=do ‘Give me either your pencil or your pen.’

(175) g:

Coordination in three Western Iranian languages

(176) v: ya púl-e kisé-y=s [hesáw kærdè], ya púl-e or money-ez purse-m.obl=3s.obl [account did or money-ez qaséd-i [ ]. messenger-m.obl ‘He either accounted for the cost of the purse or of the delivery boy.’ (177) v: ya tæ [bæ-mærde-y], ´ ya æz [ ]. or you [t/a-die-2s or I ‘(Either) you will die or I (will).’ VP: (178) v: ya b=is=bærdi ´ ya qærri=es bæ-košdé. ´ or t/a=3s.obl=took or witch=3s.obl t/a-killed ‘He either took her away or he killed the witch.’ Shifted clause, series of three (copula not ellipted): (179) v: ya bera=s be ya ræfíq=es be ya medrés=es be. or brother=3s.obl was or friend=3s.obl was or lover=3s.obl was ‘He was either her brother or her boyfriend or her lover.’ Negative: næ … næ … ‘neither … nor …’. Individual rule: It should be noted that although VPG require double negatives, i.e., a negative verb in addition to whatever negative pronouns or adverbs are used, negative bisyndetic coordinators do not require negative verbs. VP (copula not ellipted): (180) v: adui næ jævan be, næ pir be. person not young was not old was ‘The man was neither young nor old.’ VP (verb ellipted — with extra =ò): (181) cp: næ næhar bæra=t [mí-mun-e](=vo), næ šam [ ]. not lunch for=2s.obl [t/a-stay-3s(=and) not dinner ‘There won’t be any lunch or supper left for you.’ (182) cp: næ to [mí-dun-i](=o), næ mæn [ ]. not you [t/a-know-2s(=and) not I ‘Neither you nor I know.’ (‘You don’t know and neither do I.’)


322 Donald Stilo

Clause (sequence of 3 næ): (183) v: næ ci há-riz-ia, næ yey cay=s bæ-satt=o ´ not fodder pvb-spread-inch not 1 tea=3s.obl t/a-made=and næ ci=o.* not what=and ‘Neither did the hay get spread, nor did she make a tea or anything’ (*Note: ci ‘fodder’ and ci ‘what, anything’ are hom*ophones) (184) fwp:

næ tond-tær æz eqteza-ye zæman hærækæt ´ mì-kærd-ænd, not fast-er from demand-ez time movement t/a-do-3p næ ælaqé-i be héfz-e sonnæt-ha-ye mætlub=o not tie-indef to retaining-ez tradition-pl-ez sought=and qabél-e defá’… nešán mì-dad-ænd. -able-ez defense show t/a-give-3p ‘… they neither progressed faster than the demands of the time nor showed any attachment to retaining desired and defensible traditions…’ (Zarrin-kub 1996: 543)

5.2.4 ce … ce … ‘whether … or …’ Individual rule: It should be noted that as in various other languages, ce… ce… cannot occur without an accompanying independent clause. Both shifted, verb ellipted: (185) v: komæk ´ ær-kær-ènde: ce æz læház-e maddi, ce æz help t/a-make-3p what from regard-ez material what from læház-e mæ:nævi. regard-ez spiritual ‘They help, whether materially or spiritually (= moral support).’ Unshifted, additional =ò: (186) fwp:

tævæjjóh-e anha ce dær omúr-e maddi=o attn-ez they what in affairs-ez material=and eqtesadi=o ce dær mæsayél-e mærbut be zæban=o economic=and what in issues-ez related to language=and færhæng be mazænderan=æst. culture to pn-is ‘Their attention, whether in material and economic affairs or in issues related to language and culture, is (directed) to Mazanderan.’ (Al-e Ahmad 1973: 10)

Coordination in three Western Iranian languages 323

6. hæm ‘also’ and other uses 6.1 The uses of hæm The word hæm has many functions beyond the narrow meaning of its usual gloss, ‘also’, including various types of coordination (cf. David Gil’s statement, this volume, on the equivalent word sama in Riau Indonesian). The points relating to coordination will be developed slightly more than the others in this list. The following are some of the uses of hæm in VPG. The Vafsi form =iz is only used for points i through iv below but hæm also exists in Vafsi — probably a borrowing from Persian (the native Vafsi equivalent hev < hæm has a much more limited usage) — and works for most of the other six functions listed here: (i) hæm (and equivalents) is an inclusive focus particle meaning ‘also’ with affirmatives and ‘(n)either’ with negatives; this is by far the most common usage in VPG. It is always encl*ticized to the word to be included rather than being fixed in one standard place in the sentence (cf., as the usual position, right after the copula or in clause-final position in English): Affirmative (187) g:

m6n j6vad=6m bi-dé-m. I pn-also t/a-saw-1s cp: mæn jævad=æm díd-æm. I pn-also saw-1s v: tæmen jævad-i=zi I.obl pn-m.obl=also b=im=diæ. t/a=1s.obl=saw ‘I saw Javad, too.’

Negative m6n j6vad=6m ní-de-m. I pn-also neg-saw-1s mæn jævad=æm næ-did-æm. ´ I pn-also neg-saw-1s tæmen jævad-i=zi I.obl pn-m.obl=also n=ím=diæ. neg=1s.obl=saw ‘I didn’t see Javad, either.’

(ii) As a conjunctive coordinator, hæm simply means ‘and’, losing the focusing sense. (188) g:

æni ´ ism h6s6n=I. u z6ney=6m æni ´ zen=I. his name H.-is that woman-also his wife-is ‘His name is Hassan. And that woman is his wife.’

(189) v: do næfær se næfær tajér-i ke=san æ-cáppa. 2 person 3 person merchant-m.obl house=3p.obl t/a-ransack hær-ci=æm ær-gerdá-nde,, dózd=esan every-what=also t/a-searched-3p thief=3p.obl n=ísan=ær-venda. neg=3p.obl=t/a-find ‘Two or three merchants’ houses were ransacked. And no matter how much they looked, they did not find the thief.’

324 Donald Stilo

Note that in the above examples ‘also’ is not a possible interpretation. If we say ‘That woman is also his wife’ in (188), the implication is that Hassan has more than one wife, which does not convey the meaning of this sentence as there was no reference to another wife. (See also (19).) Of course, the use of an initial ‘also’ in English introducing the whole sentence would have the meaning of ‘and’ or ‘additionally’, as in ‘His name is Hassan. Also, that woman is his wife.’ (iii) hæm also acquires a sense of ‘even’: (190) cp: dær amrika xærcæng=æm mí-xor-ænd. in A. crab-also t/a-eat-3p ‘In America they even eat crabs.’ This emphatic sense also includes certain rhetorical effects, such as sarcasm, irony, etc.: (191) cp: ærær-e ´ xær=æm ´ qæbúl mì-kon-i? braying-ez donkey=also accept t/a-make-2s ‘And so you accept the braying of a donkey?’ (i.e., over my word) (iv) As an adversative coordinator usually glossed as ‘and, but’: (192) cp: fæqæt mí-xast ke pedær=eš bær-gærd-è. ´ pedære hæm g: fæqæt xast-i-Ø ke æni per vá-g6rd-e. per hæm only wanted-t/a-3s sub his father pvb-turn-3s father-also cp: fékr-e bær-gæšt-æn næ-dašt. ´ g: fékr-e va-g6rd6st-6n n-ášti. thought-ez pvb-turn-inf neg-had ‘He only wanted his father to return. But the father had no thought of returning.’ Lazard (1989: 281) interprets the use of =(h)æm as in (192) as a way to mark a new theme. This insight also pertains to the discussion of sentence (188). (v) When doubled, hæm means ‘both … and …’. See § 5.2.1 above. Beginning with point (vi), we find that hæm functions very differently from the above points, where it is a grammatical encl*tic. As will be shown below, it is also a derivational morpheme and has reciprocal meanings with pronominal functions. The reciprocal and word-building senses of hæm will be glossed as hÊm in the interlinear glosses, since their meanings and functions are very varied and quite different from the focus particle uses above: (vi) ‘co-, -mates’ (most of these words are also Persian borrowings in Vafsi and Gilaki): (193) p: hæm-vætæn hÊm-homeland ‘compatriot’

hæm-kelas(-i) hÊm-class-‘ite’ ‘classmate’

hæm-saye hÊm-shade ‘neighbor’

hæm-kar hÊm-work ‘coworker’

Coordination in three Western Iranian languages 325

(194) p: hæm-zæman hæm-dærd-i hæm-zist-i hÊm-time hÊm-pain-abst hÊm-exist-abst ‘contempor- ‘sympathy’ ‘coexistence’ aneous’

hæm-kar-i hÊm-work-abst ‘cooperation’

(vii) Reciprocal pronouns, and the like. In its various reciprocal uses, hæm is interchangeable with — and probably an abbreviated form of — Persian hæm-digær ‘each other’, literally ‘hÊm-other’, which has also been borrowed into Gilaki. This usage of hæm is found only with adpositions; as a direct object or possessive, only the full form hæm-digær (or its alternate yek-digær, literally ‘one-other’) is permissible. – ‘each other’ (hæm alone or the alternate: P: hæm-digær < hæm + digær ‘other’; cp: hæm-dige) (The usual Vafsi reciprocal as a direct object or possessive is from a different source, but Vafsi hev13 is occasionally used, again only with adpositions, probably all borrowed Persian prepositions.) (195) v: æz hev jodá -r-búæ-nd. cp: æz hæm jodá mi-š-ænd. from hÊm separate t/a-become-3p ‘They separate from each other.’ –

ba-hæm ‘together, with each other’ (ba ‘with’)

(196) cp: ba-hæm dar-ænd mí-ræqs-ænd. with-hÊm prog-3p t/a-dance-3p ‘They are dancing with each other.’ –

ba-hæm qæzá mì-xor-ænd. with-hÊm food t/a-eat-3p ‘They eat together.’

be-hæm ‘together, to each other’ (be = ‘to’)

(197) v: bæ-hev bæ-resa-nde. ´ to-hÊm t/a-reach-3p cp: be-hæm resíd-ænd. to-hÊm reached-3p ‘They got to each other.’

be-hæm sælám kærd-ænd. ` to-hÊm hello make-3p ‘They greeted each other.’

– Persian be-hæm and Vafsi dæ-hev ~ o-hev are also used to form a few compound verbs, e.g., P: be-hæm zædæn ‘to disrupt, cancel, break up; stir, whip up (eggs, etc.)’, (zædæn ‘to hit, strike, beat’); V: ríæ o-hev kætt ‘get a cramp (lit: vein upon-hÊm fall)’ and others.

13.Vafsi hev and the alternates found in the Vafsi emphatic deictics in point (viii) here have undergone a typical Vafsi sound change (shared with Kurdish) of medial and final *m > v/w: hev < hæm, Persian emphatics hæmin, hæman = Vafsi hævin, hawan.

326 Donald Stilo

(viii) The emphatic deictic forms hæmin ‘this very (one), the same (one)’, hæman ‘that very (one), the same (one)’ < hæm + in ‘this’, an ‘that.’ hæmin, hæman are also used to form other emphatic deictics (Vafsi and Gilaki use other formants): a.

ja ‘place’ hæman-ja ‘right there’, væqt ‘time’ hæman-væqt ‘right then’, æl’an ‘now’ hæmin-æl’an ‘right now’, etc. b. p, g: hæman (cp: hæmun) ´ is also used to mean ‘Exactly!; That’s it!’ (ix) p: cenin ‘this way’, cenan ‘that way’ > hæmcenin ‘likewise’ (x) Other lexicalized derived forms: p: hæme, v: hævíe ‘all’; c: hæmiše ‘always’, etc. 6.2 Further discussion of hæm

The first four uses of hæm listed above seem to be areal features that include at least Western Iranian languages, Azerbaijani, Turkish, Armenian, and Georgian. In all these languages, as well as Daghestanian languages, the form of hæm is always encl*tic. Note that hæm in its bisyndetic uses is a borrowing in actual form as well as in usage in Vafsi, Armenian, Azeri, Turkish, and others. The feature of interest for the present discussion is the use of hæm in its various forms in Iranian and non-Iranian languages of the area in points i through iv above. The focusing (point i) and the adversative (point iv) are especially of import here (see also Haspelmath (to appear)). Examples are given in Persian and Turkish here but work the same in the other languages listed above: (198) Persian: jorj mahi míxore. bil=æm míxore. Turkish: jorj balık yiyer. bil=de yiyer. ‘George eats fish, Bill does, too.’’

(Affirmative Focusing)

(199) Persian: jorj mahi némixore. bil=æm némixore. Turkish: jorj balık yemez. bil=de yemez. ‘George doesn’t eat fish, Bill doesn’t, either.’

(Negative Focusing)

Adversative: (200) Persian: Turkish:

jorj mahi míxore, bil=æm gušt. jorj balik yiyer, bil=de et. G. fish eats B.-also meat ‘George eats fish, but Bill meat.’

To say that the hæm word is simply focusing and means ‘also’ or is simply adversative and means ‘but, and (on the other hand)’ misses a larger generalization about this word: since it has both functions — somewhat opposite in meaning, at that — one does not know which of these functions is intended until the sentence is completed:

Coordination in three Western Iranian languages 327



Persian: jorj mahi míxore, bil=œm Turkish: jorj balik yiyer, bil=de ‘George eats fish,

{ míxore. yiyer. ‘and Bill does, too.’ ‘eat.3s

{ guet št.

‘but Bill [eats] meat.’


The adversative function can also apply to verbs: (202) Persian: Gilaki: Vafsi:


æli mahi mí-xær-e, hæsæn=æm mí-foruš-e. æli mahi Ø-hin-I, hæsæn=6m Ø-furuš-I. æli mahi æ-rˇin-e, hæsæn=iz æ-ruš-e. pn fish t/a-buy-3s pn=also t/a-sell-3s ‘Ali buys fish, but Hassan sells (them).’


This discussion of coordination in Iranian languages was not intended to be exhaustive and is only preliminary in two ways. First, other languages, both major and minor languages and dialects (e.g., Esfahani), will have to be included to make the survey of western Iranian languages more complete. Second, even in those languages discussed here, more work must be done to develop a more complete picture of coordination. This is particularly true of Colloquial Persian. There are two main reasons why CP should be developed further. First, among the various languages discussed it is the most accessible to linguistic investigation because there are so many native speakers of Persian available and because of the rich and multifaceted research already being conducted by Iranian linguists in Iran and abroad. Secondly, Colloquial Persian is still a gold mine for linguistic research since it is really the natively spoken language of Iran today, whereas FWP is only a very formalized written form of the language, quite unnatural in spoken contexts even among educated native speakers in all spheres of society including educational milieus, the media, political arenas, as well as every-day life, even taking into account that FWP has different registers as with any other language. Since Colloquial Persian does not enjoy the prestige of FWP, however, it has by and large been ignored in much linguistic research conducted both in Iran and abroad. With so many accessible native speakers and Iranian linguists, however, this is a perfect opportunity for in-depth work on corpora of natural speech. We might add to this suggestion that the issues of sociolinguistics with respect to deletion in coordination, as well as the concomitant intonational patterns in CP, are also a very interesting and fertile area of study for future research endeavors.

328 Donald Stilo

An additional area of future research on coordination should also include a more careful search for WITH-conjunction in Iranian languages, their forms, usage and frequency. The areal distribution of this construction is also a fruitful area for research.

Abbreviations Languages: VPG Vafsi, Persian, Gilaki (when the same comment refers to all three languages) v Vafsi fwp formal written Persian cp standardized colloquial Persian p common to both cp and fwp (possibly with slight differences) g western Gilaki, specifically the dialect of Rasht lj eastern Gilaki, specifically the dialect of Lahijan c common to VPG (used in examples) Interlinear Glosses: 1s, 2s, 3s 1st, 2nd, 3rd person singular (Set 1) 1p, 2p, 3p 1st, 2nd, 3rd person plural (Set 1) 1s.obl, 2s.obl, 3s.obl1st, 2nd, 3rd person singular oblique 1p.obl, 2p.obl, 3p.obl1st, 2nd, 3rd person plural oblique abst abstract noun formant change change of state, a morpheme that accompanies the adjectival element of both causative and inchoative light verb forms clf classifier do direct object ez Ezafe (see 2.1, point v) f.dir feminine direct case f.obl feminine oblique case hÊm the reciprocal and word-building senses of hæm, ‘also’, etc. inch inchoative indef indefinite marker intens intensifier m.obl masculine oblique oblpr oblique pronoun base hazun-, to which person markers are added pl.dir plural direct case pl.obl plural oblique case pn proper noun prog progressive pvb preverb recip reciprocal sbc (ínke) (creates subordinate conjunctions)

Coordination in three Western Iranian languages 329

sbj sub t/a warn

subjunctive subordinating particle tense/aspect marker indicates warning, admonition, threat

References Al-e Ahmad, J. (1352/1973) Owraza¯n (4th printing). Tehran: Mazeyar.

6zizb6jov, X. 6. (1965). Az6rbajcanca-ruscˇa lügˇ6t. Bakı: Az6rbajcan dövl6t n6šrijjati. Gil, David (this volume). ‘Riau Indonesian sama.’ Eilers, W. and U. Schapka (1979). Westiranische Mundarten aus der Sammlung Wilhelm Eilers, Band II: Die Mundart von Gäz, 2 vols. Wiesbaden: Franz Steiner Verlag. Haspelmath, M. (1993). A Grammar of Lezgian. Berlin: Mouton de Gruyter. Haspelmath, M. (to appear). ‘Coordination.’ In T. Shopen (ed.), Language Typology and Syntactic Description, 2nd ed. Cambridge: Cambridge University Press. Horn, P. (1893). Grundriss der neupersischen Etymologie. Strassburg: Karl J. Trübner. Kent, R. G. (1953). Old Persian. New Haven: American Oriental Society. Kerimova, A. A., Mamedzade, A. K., Rastorgueva, V. S. (1980). Gilyansko-Russkij Slovar’. Moscow: Nauka. Lazard, G. (1957). Grammaire du persan contemporain. Paris: Klincksieck. Lazard, G. (1989). ‘Le persan.’ In Compendium Linguarum Iranicarum, R. Schmitt (ed.), 263–293. Wiesbaden: Reichert. Lazard, G. (1992). A Grammar of Contemporary Persian. Costa Mesa, California: Mazda. Mithun, M. (1988). ‘The grammaticalization of coordination.’ In Haiman, J. & Thompson, S.A. (eds.), Clause Combining in Grammar and Discourse, 331–359. Amsterdam: Benjamins. Paul, Ludwig (unpublished Habilitation thesis, submitted in January, 2002, Göttingen). (Tentative title of forthcoming publication:) Grammatical and Philological Studies on the Early Judaeo-Persian Texts in the Cairo Geniza. Pokorny, J. (1951–59). Indogermanisches etymologisches Wörterbuch. Bern and Munich: Francke. Rastorgueva, V. S. et al. (1971). Giljanskij jazyk. Moscow: Nauka. Stassen, L. (2000). AND-Languages and WITH-Languages. Linguistic Typology, vol. 4: 1–54. Steingass, F. (1892, repr. 1975). A Comprehensive Persian-English Dictionary. Beirut: Librairie du Liban. Stilo, D. L. (1981). ‘The Tati Language Group in the Sociolinguistic Context of Northwestern Iran and Transcaucasia’. Iranian Studies, Volume XIV: 137–187. Boston: Society for Iranian Studies. Stilo, D. L. (2001). ‘Gila¯n, ix. Languages’. In Encyclopædia Iranica, Vol. 10, Fasc. 6: Germany VI.-Gindaros, pp. 660–668. New York: Bibliotheca Persica Press. Stilo, D. L. (forthcoming). Vafsi Folk Tales. Wiesbaden: Reichert. Stilo, D. L., Clinton, J. W., Talattof, K. (forthcoming). Modern Persian: Spoken and Written. New Haven: Yale University Press. van den Berg, H. (this volume). ‘Coordinating constructions in Daghestanian languages.’ Wehr, H. (1979). A Dictionary of Modern Written Arabic. Wiesbaden: Otto Harrassowitz.

330 Donald Stilo

Yarshater, E. (2001). ‘“A Peasant Marriage”, A Poem in Cha¯li by Mohammad-Ba » ¯ qer ‘A¯meli’. Studia Iranica, 30: 245–289. Paris: Association pour l’avancement des études iraniennes. Zarrin-kub, A. H. (1375/1996). Az Gozašte-ye Adabi-e Ira¯n. Tehran: Alahadi.

Southeast Asia

Chapter 12

Coordination in Hakha Lai (Tibeto-Burman) David A. Peterson and Kenneth VanBik Dartmouth College / University of California, Berkeley

1. Introduction 2. Basic elements of coordination 2.1 Conjunction 2.2 Disjunction 2.3 Adversative coordination 2.4 Causal coordination 2.5 Extraction 3. Special types of coordination 3.1 Emphatic coordination 3.2 Representative and augmentative conjunction 3.3 Alternative vs. disjunctive questions 3.4 Comitative, inclusory, and summary conjunction 4. Ellipsis in coordination constructions 5. Conclusion



Hakha Lai is a Tibeto-Burman language (Kuki-Chin branch) spoken by about 300,000 people as a first language in and around the city of Hakha in Chin State (western Burma). A considerable number of other people speak Lai as a second language in Chin State. In this paper, we provide a descriptive account of coordination constructions in Lai, using the conceptual distinctions and terminological conventions which Haspelmath (to appear) has recently proposed for cross-linguistically adequate description of constructions of this sort. Where possible, we have drawn examples from a corpus of about ten lengthy narrative texts; examples constructed by the second author, who is a Lai native speaker, and usually verified with other Lai speakers, illustrate points about the system which our text material does not illuminate. In gross typological terms, Lai exhibits relatively rigid verb-final word order characteristics and a highly agglutinative word structure. Verbs show accusatively

334 David A. Peterson and Kenneth VanBik

aligned agreement markers for up to two participants. There is also a limited set of cl*tic case postpositions which pattern ergatively. Verb roots have a complicated system of ablaut, which is characteristic of Kuki-Chin languages. One or the other ablaut grade may be required by particular syntactic constructions. For instance, one ablaut grade, dubbed Form 1 and indicated in examples by a subscript 1 in the gloss, is required by negation and interrogative constructions, and a second ablaut grade, Form 2 (indicated by subscript 2), is required in many subordinate clauses; a small number of verb stems, which have no subscripts in our interlinear glosses, are invariant. Some constructions allow either ablaut form to appear, in which case one or another participant is accessible or inaccessible in the construction (e.g. in relativization). A synopsis of these and other characteristics of the language may be found in Peterson 2003; a discussion of the tonal system in Lai is provided by Hyman and VanBik 2002a and b. The basic facts of coordination are relatively simple, and we introduce them in the immediately following section. More specialized aspects of coordination are treated in § 3. In the last section of the paper we discuss ellipsis in Lai coordination constructions. We conclude with some brief remarks concerning the historical development of Lai’s coordinators.


Basic elements of coordination

The essentials of coordination in Lai are straightforward. In this section, we focus on the main characteristics that coordination has for the four semantic types of coordination identified by Haspelmath: conjunctive, disjunctive, adversative, and causal coordination. 2.1 Conjunction Sentential or clausal conjunction involves monosyndetic coordination marked by a postcl*tic element =‘ií at the end of each non-final coordinated clause, as in (1), with two coordinands, and (2), with three coordinands. (1) ma‘khán min tsuú ‘a-mìn=‘‘ií ma‘tsuu then avalanche deic 3sg.subj-avalanche1=and that min mín kàr=‘a‘ tsùn tlaà] kua avalanche avalanche2 interval=loc deic mountain nine ‘a-liàm=‘eé ‘àn-tií 3sg.subj-cross.over=emot 3pl.subj-say ‘Then, there was an avalanche, and during that avalanche, she crossed over nine mountains, they say.’

Coordination in Hakha Lai (Tibeto-Burman) 335

(2) ma‘ hnuu khán tlà]vaal laay mii tlà]vaal tsuú that back deic bachelor Lai person bachelor deic khuàkip=‘ìn mii tsuú ‘àn-tshuak=‘‘ií every=adv person deic 3pl.subj-come.out1=and ‘àn-kaáy-hno‘-leèn=‘‘ií ‘aháwhma‘n=ni‘ 3pl.subj-climb2-appl-repeatedly=and nobody=erg ‘àn-kaáy-hno‘-khàw-láw 3pl.subj-climb2-appl-pot-neg ‘Then, the bachelors, the Lai bachelors, people from every village, came out, and they tried many times to climb up to her, but nobody was able to climb up to her.’ The cl*tic element which marks these constructions is intonationally incorporated into the preceding phonological material, and often is followed by a pause. However, it, or a clearly related independent element, may also occur as a sentence-initial independent conjunctive particle, especially following major pauses in narrative (which are often coincident with paragraph shifts), as in the following example: (3) ‘ií phè]tee tsuú haày-ku] kaáy tsuù thiàm-hla‘-koò and rabbit deic mango-tree climb2 deic know2-neg-since vòm-pii=ni‘ khàn kha‘ koòy phe] ‘a-ná] bear-aug=erg deic interj friend rabbit emph-2sg ‘a-ta]=‘a‘ rak-‘ùm-loó káy ‘a-tsu]=‘a‘ 3sg.poss-under=loc imp-exist1-pol.imp 1sg 3sg.poss-top=loc ka-kaày-laay=‘ií kayma‘=ni‘ ka-va-lo‘-laày 1sg.subj-climb1-irr=and 1sg.pron=erg 1sg.subj-direc-pick2-irr tia‘ khàn ‘a-tií ‘àn-tií quot deic 3sg.subj-say 3pl.subj-say ‘And, since the rabbit didn’t know how to climb the mango tree, the bear said, “Friend rabbit, you stay below, and I’ll climb up on top and pick (fruit),” they say.’ In §2.5 below we will present evidence from extraction which suggests that the clauses =‘ií joins in these examples involve true coordination rather than subordination. NP conjunction involves a postcl*tic element =leé attached to each non-final NP, as shown in (4), with two coordinands, and (5), with three coordinands. (4) ‘a-hlaán-liàw-pii=‘a‘ ‘àn-tií vòm-pii=leé 3sg.poss-before-during-aug=loc 3pl.subj-say bear-aug=and phè]tee ‘àn-rak-‘ùm=‘eé ‘àn-tií rabbit 3pl.subj-past-exist1=emot 3pl.subj-say ‘Once upon a time, they say, there was a bear and a rabbit, they say.’

336 David A. Peterson and Kenneth VanBik

(5) hii tluk=‘ií hu-hàm na-]a‘y-mii=leé deic same=obl power 2sg.subj-have2-relzr=and na-zhoonnaak=leé na-zhatnaak hií khoy-kaa=da‘ 2sg.poss-strength=and 2sg.poss-goodness deic where-place=interr na-laak ka-tshím tua‘ ‘a-tií ‘àn-tií. 2sg.subj-take2 1sg.obj-tell1 do 3sg.subj-sayi 3pl.subj-sayi ‘“Tell me where you got such power that you have, your strength and your goodness,” she said, they say.’ Instances in which this coordinator is found on all three conjuncts have been recorded, as in (6). (6) ‘a-nuù=leé ‘a-paà=leé ‘a-‘uù=leé 3sg.poss-mother=and 3sg.poss-father=and 3sg.poss-elder.sibling=and ‘àn-khorua‘ ‘a-haàr-tuk 3pl.poss-thinking 3sg.subj-difficult1-very ‘Her mother and father and sisters were very surprised.’ Probably what is involved in this case, however, is actually a second use of the marker =leé as a collective marker on the final conjunct; sentences like (7), in which three conjoined NPs are all marked by the coordinator, are uniformly judged to be unacceptable. (7) *làwthlawpaa=ni‘ bànhlaà=leé bu‘=leè saa=leè ‘a-‘áy farmer=erg banana=and rice=and meat=and 3sg.subj-eat2 ‘The farmer ate a banana and rice and meat.’ The clausal coordinator =‘ií is not omissible, such that sentences like (8b–c) are not acceptable as variants of (8a): làwthlawpaa doòr=‘a‘ ‘a-kàl=‘‘ií ‘aàr farmer market=loc 3sg.subj-go1=and chicken ‘a-tsook=‘‘ií ‘ìn=‘a‘ ‘a-ziìn 3sg.subj-buy2=and house=loc 3sg.subj-return1 b. *làwthlawpaa doòr=‘a‘ ‘a-kàl, ‘aàr ‘a-tsook=‘‘ií ‘ìn=‘a‘ ‘a-ziìn c. *làwthlawpaa doòr=‘a‘ ‘a-kàl=‘‘ií, ‘aàr ‘a-tsook, ‘ìn=‘a‘ ‘a-ziìn ‘The farmer went to the market, bought a chicken, and returned home.’

(8) a.

The NP coordinator =leé, on the other hand, is omissible, though there are no clear examples of its omission in our text corpus. Thus, instead of (9a), the second variant (9b), in which the coordinator is missing from the first conjunct, and in its place there is a clear intonational break following the first conjunct, is also regarded as acceptable.

Coordination in Hakha Lai (Tibeto-Burman) 337

làwthlawpaa=ni‘ bànhlaà=leé bu‘=leè saa ‘a-‘áy farmer=erg banana=and rice=and meat 3sg.subj-eat2 ‘The farmer ate a banana and rice and meat.’ b. làwthlawpaa=ni‘ bànhlaà, bu‘=leè saa ‘a-‘áy

(9) a.

Usually there must be at least one instance of =leé in a conjunction construction, such that sentences like (10a–b) do not occur. (10) a. *làwthlawpaa=ni‘ bu‘, saa ‘a-‘áy farmer=erg rice meat 3sg.subj-eat2 ‘The farmer ate rice and meat.’ b. *làwthlawpaa=ni‘ bu‘, tsoo-saa, ‘aàr-saa ‘a-‘áy farmer=erg rice cow-meat chicken-meat 3sg.subj-eat2 ‘The farmer ate rice, beef, and chicken.’ There is at least one context in which all coordinators may be omitted from the construction, though. If a set of conjoined NPs is functioning as subject or predicate nominal in an equational copular sentence, like the NPs in (11), use of =leé is optional. làwthlawpaa=ni‘ ‘a-‘ay-mii tsuú bànhlaà, bu‘, farmer=erg 3sg.subj-eat2-relzr deic banana rice saa ‘àn-sií meat 3pl.subj-be2 ‘What the farmer ate was a banana, rice, and meat.’ b. làwthlawpaa, sayaàpaa, tsewmaá] ka-hoòy ‘àn-sií farmer teacher name 1sg.poss-friend 3pl.subj-be2 ‘The farmer, the teacher, and Tsewmang are my friends.’

(11) a.

2.2 Disjunction Other types of coordination which have structures formally similar to coordination structures in other languages do not typically have corresponding constructions in Lai. The functional end that NP disjunction in English serves, for instance, is accomplished by a multi-clausal construction like the ones shown in (12). (12) a.

làwthlawpaa=ni‘ vok ‘a-sií-làw=leè ‘aàr farmer=erg pig 3sg.subj-be2-neg=cond chicken ‘a-tsook 3sg.subj-buy2 ‘The farmer bought a pig or a chicken.’

338 David A. Peterson and Kenneth VanBik

b. làwthlawpaa ‘a-sií-làw=leè sayaàpaa doòr=‘a‘ farmer 3sg.subj-be2-neg=cond teacher market=loc ‘àn-kal 3pl.subj-go1 ‘The farmer or the teacher went to the market.’ There is evidence that the sequence ‘a-sií-làw=leè should not simply be analyzed as a monolithic coordinator synchronically. In particular, it can show agreement with non-third person singular subjects, so it still behaves like a fully verbal predicate; clauses ending with it also may be displaced from the position seen in (12a), so that the clause it occurs at the end of is a clearly separate constituent from what follows it. The structure involved in these examples apparently arose historically from a type of conditional clause structure, such that the =leé marked clause in each one might be translated as such (i.e. ‘The farmer, if it wasn’t a pig, bought a chicken’, ‘If it wasn’t the farmer, the teacher went to the market’). This is not the normal means Lai has at its disposal for marking conditional clauses. Nonetheless, support for this view of the construction includes both language-internal and comparative evidence. A common conjunctive adverb which occurs only sentence initially in texts is the form ‘a-sií-neè-leé ‘if that’s the case/that being the case’; this item appears to contain a fossilized use of this conditional clause marker. In addition, if one looks at related languages, it is easy to find conditional clauses marked by morphology transparently cognate with this element, cf. the Tedim and Thadou sentences given in (13) and (14). (13) bui=f: pu: da‘pa:=in tu: a-ho]-pai-kei=le‘, bamboo.rat=voc resp name=erg here 3sg-direc-go-neg=cond ‘a-kua=a‘ tui sa: su]-di]=hi:] 3sg.poss-hole=loc water boiling pour-fut=?assertive ci: hi: say conclusive ‘Bamboo rat, Lord Dahpa says that if you do not come he is going to pour boiling water down your hole.’ (Henderson 1965: 134) (14) a-hù] a-hí-lé 3sg.subj-come 3sg.subj-be-cond ‘If he comes’ (Krishan 1980: 80) A similar construction is also attested in closely related Sizang (Stern 1963: 255). Clausal disjunction involves a similar strategy. However, the verbal complex that the =leé marker appears on is that of the first clause rather than an invariant sii ‘be’, as in phrasal disjunction. Simple examples of this are seen in (15).

Coordination in Hakha Lai (Tibeto-Burman) 339

làwthlawpaa falaám ‘a-kál-làw=leè haàkhaà=‘a‘ farmer Falam 3sg.subj-go2-neg=cond Hakha=loc ‘a-‘ùm 3sg.subj-exist1 ‘The farmer goes to Falam or he stays in Hakha.’ b. làwthlawpaa=ni‘ ‘aàr ‘a-tsook-láw=leè vok farmer=erg chicken 3sg.subj-buy2-neg=cond pig ‘a-zuár 3sg.subj-sell2 ‘The farmer either bought a chicken, or he sold a pig.’

(15) a.

In cases involving disjunction of multiple clauses, like (16), one clause is combined with a second, using the strategy just outlined; this disjunction structure is then treated as a zero-derived nominal, which is the subject of ‘be’ in the first clause of another disjunction structure: (16) làwthlawpaa falaám ‘a-kál-làw=leè, farmer Falam 3sg.subj-go2-neg=cond haàkhaà=‘a‘ ‘a-‘ùm-laay ‘a-sií-làw=leè, Hakha=loc 3sg.subj-exist1-irr 3sg.subj-be2-neg=cond tidím=‘a‘ ‘a-raà-laay Tedim=loc 3sg.subj-come1-irr ‘The farmer will go to Falam, or stay in Hakha, or come to Tedim.’ Sentences like (16) then involve two variations on the general disjunction theme: disjunction of the first two clauses involves the normal strategy for clausal disjunction, and the addition of these two clauses to the last involves the normal strategy for phrasal disjunction. The distinction between exclusive and inclusive disjunction is not something we are fully prepared to comment on, but it does not appear to be associated with distinct disjunctive coordinator types. Instead, while a sentence like (17) can have only an exclusive reading, in which only one of the options must be true, (17) tsewmaá] tsuù làwthlawpaa ‘a-sií-làw-leè sayaàpaa name deic farmer 3sg.subj-be2-neg-cond teacher ‘a-siì-laay 3sg.subj-be1-irr ‘Tsewmang is (would be) a farmer or a teacher.’ (18) has only an inclusive reading:

340 David A. Peterson and Kenneth VanBik

(18) tsewmaá] tsuù làwthlawpaa ‘a-sií-làw-leè sayaàpaa name deic farmer 3sg.subj-be2-neg-cond teacher ‘a-siì 3sg.subj-be1 ‘Tsewmang is a farmer or a teacher.’ (18) is only true if Tsewmang is sometimes a farmer and sometimes a teacher, but it cannot be true if he has only one of these professions. The difference between these sentences is clearly the presence of the irrealis marker -laay in the first one, and in light of this, it is probable that there is a subtle interaction between verbal aspect and the ability to get an inclusive or exclusive reading. This is clearly something that will have to be addressed by painstaking investigation in future work with Lai. 2.3 Adversative coordination As Haspelmath notes, many languages lack a coordination construction which expresses adversative semantics (‘but’). Instead, this meaning is expressed by concessive subordinate clauses. This is the case in Lai. The normal, and seemingly only, way to express the notion of adversative coordination is by means of a concessive converbal construction marked by -naa‘in, as seen in (19). (19) ]uùnnuu=leé tsuu ruul-pii khuà]thlarliit ‘àn-tii-mii tsuú name=and deic snake-aug name 3pl.subj-say-relzr deic ‘àn-rak-‘ii-du‘=‘ií tsutsuú 3pl.subj-past-rec-love=and this ‘a-naàw-nuú=ni‘ 3sg.subj-younger.sibling-woman=erg ‘a-h]ál-kàw-naa‘‘in ‘a-paà sin=‘a‘ 3sg.subj-know2-affirm-concess 3sg.poss-father near=loc ‘a-phuà]-]am-láw 3sg.subj-tell.on1-dare-neg ‘]uùnnuu and that big snake who they called khua]thlarliit were in love, and although her little sister knew, she didn’t dare tell on her to her father.’ 2.4 Causal coordination The semantics of causal coordination also typically involves a subordination construction utilizing a zero-nominalized reason clause in combination with an obliquely marked relational noun tsaà ‘reason, sake’, illustrated in (20).

Coordination in Hakha Lai (Tibeto-Burman) 341

(20) ma‘khán suy lá]‘aak-piì=ni‘ tsùn vày ‘à ma‘-tií then gold raven-aug=erg deic interj interj pron-say ‘a-sií tsùn vày tsuù-kàw na-tií ‘a-sií=‘a‘ 3sg.subj-be2 deic interj peck1-affirm 2sg.subj-say 3sg.subj-be2=loc tsùn ná] zò]=ni‘ tsuù-kàw na-ka-tií-veè deic 2sg also=erg peck1-affirm 2sg.subj-1sg.obj-say2-also tsaa=‘‘a‘‘ káy zò] ka-‘ìn=‘a‘ roòl ‘áy reason=loc 1sg also 1sg.poss-house=loc meal eat2 ka-n´-ko‘-veè-laày thaayzi]=‘a‘ roòl ‘áy 1sg.subj-2sg.obj-invite2-also-irr tomorrow=loc meal eat2 rak-ka-hroo]-te‘ tia‘ khàn là]‘aak-piì=ni‘ tsùn direc-1sg.obj-go.eat2-req quot deic raven-aug=erg deic ‘a-vón-tiì ‘àn-tií 3sg.subj-direc-say 3pl.subj-say ‘Then the raven said to her, “Well, ah, if that’s the case, if you say peck on, since you said peck on to me, I too will invite you to eat at my house. Come to eat at my place tomorrow!” they say.’ The word ruà], which has approximately the same meaning as tsaà, may be substituted in all of these cases, yielding a construction with identical syntactic characteristics, exemplified in (21). (21) ma‘ tsaa=‘a‘ tsùn vòmpii ‘a-hlaán=‘a‘ tsùn deic reason=erg deic bear 3sg.poss-before-loc deic ‘a-raà] mii ‘a-rak-sií-kàw-naan phè]tee=ni‘ 3sg.subj-white1 person 3sg.subj-past-be2-affirm-concess rabbit=erg ‘a-kha‘] rua] ]=‘‘a‘‘ ‘a-nak=‘ií 3sg.subj-burn2 reason=loc 3sg.subj-black1=and ‘a-hmaay-lay=‘ií ‘a-raà]-mii-tee khií 3sg.poss-front-towards=loc 3sg.subj-white1-relzr-dim deic ‘a-paáypeèr-hrii=ni‘ ‘a-say-hrii=ni‘ 3sg.poss-basket-rope=erg 3sg.poss-basket-rope=erg ‘a-rak-khu‘-naak ‘a-siì=‘eé ‘àn-tií 3sg.subj-past-cover2-loc.relzr 3sg.subj-be1=emot 3pl.subj-say ‘For this reason, although the bear was once white, because the rabbit burned him, he turned black, and the little part of him on his front which is white is where his basket rope covered, they say.’ In addition to these subordination constructions, the marker of clausal conjunction, =‘ií, also marks a construction in which there is a causal relationship between the two clauses. However, extraction facts indicate a subtle difference between the causal and conjoining uses of this element, as we will see in the next section.

342 David A. Peterson and Kenneth VanBik

2.5 Extraction A consideration of extraction possibilities for the clausal constructions considered in § 2.1–4 suggests that, with a couple of exceptions, most notably the basic conjunctive use of =‘ií, all of the constructions involve syntactic subordination rather than coordination. 2.5.1 Disjunctives One of the most widely acknowledged indicators that a construction involves coordination is the inability to extract from its parts (Ross’s Coordinate Structure Constraint); subordination constructions, on the other hand, usually allow extraction. In Lai, content questions are either in situ questions, or they may involve extraction of a question word to a pre-clausal focus position. Extraction facts do not clearly indicate that disjunctive clauses marked by ‘a-siílàw=leè are subordination structures rather than ones which involves coordination. For a sentence like (22), (22) làwthlawpaa=ni‘ tihaà] ‘a-dín-làw=leè farmer=erg soup 3sg.subj-drink2-neg=cond sayaàpaa=ni‘ bu‘ ‘a-‘áy teacher=erg rice 3sg.subj-eat2 ‘The farmer drank the soup or the teacher ate the rice.’ questions involving extraction, like (23), may be formed. (23) zày=da‘ làwthlawpaa=ni‘ ‘a-dín-làw=leè what=interr farmer=erg 3sg.subj-drink2-neg=cond sayaàpaa=ni‘ bu‘ ‘a-‘áy teacher=erg rice 3sg.subj-eat2 ‘What didn’t the farmer drink and then the teacher ate the rice?’ However, the semantics in all cases of extraction from this construction does not involve disjunction but rather simple coordination. The reason for this may be that historically it was possible to coordinate clauses using =leé (as it still is in some closely related dialects), and this semantics is retained in this small subset of cases. Thus, extraction does not really yield a clear picture as to the coordinate or subordinate status of this structure. 2.5.2 Adversatives On the other hand, extraction facts indicate that adversative clauses marked by -naa‘in are subordination structures rather than ones which involve coordination. Consider sentence (24).

Coordination in Hakha Lai (Tibeto-Burman) 343

(24) làwthlawpaa=ni‘ tihaà] ‘a-dín-di‘-naa‘‘in farmer=erg soup 3sg.subj-drink2-comp-concess sayaàpaa=ni‘ bu‘ ‘a-‘ay-zhiamzhiam teacher=erg rice 3sg.subj-eat2-still ‘Although the farmer drank all the soup, the teacher still ate the rice.’ (=‘The farmer drank all the soup, but the teacher still ate the rice.’) The context for this situation is that the teacher would have preferred to have soup to eat along with the rice. However, the farmer drank all the soup. Nonetheless, the teacher still ate the rice. Questions like (25)–(28) may be formed relating to this situation. (25) questions the A participant of the first clause, (26) questions the O participant of the first clause, (27) questions the A participant of the second clause, and (28) questions the O participant of the second clause. (25) ‘aháw=ni‘=da‘ tihaà] ‘a-dín-di‘-naa‘‘in who=erg=interr soup 3sg.subj-drink2-comp=concess sayaàpaa=ni‘ bu‘ ‘a-‘ay-zhiamzhiam teacher=erg rice 3sg.subj-eat2-still ‘Who drank all the soup but the teacher still ate the rice?’ (26) zày=da‘ làwthlawpaa=ni‘ ‘a-dín-di‘-naa‘‘in what=interr farmer=erg 3sg.subj-drink2-comp-concess sayaàpaa=ni‘ bu‘ ‘a-‘ay-zhiamzhiam teacher=erg rice 3sg.subj-eat2-still ‘What did the farmer drink all of but the teacher still ate the rice?’ (27) ‘aháw=ni‘=da‘ làwthlawpaa=ni‘ tihaà] who=erg=interr farmer=erg soup ‘a-dín-di‘-naa‘‘in bu‘ ‘a-‘ay-zhiamzhiam 3sg.subj-drink2-comp-concess rice 3sg.subj-eat2-still ‘Who, although the farmer drank all the soup, still ate the rice?’ (28) zày=da‘ làwthlawpaa=ni‘ tihaà] ‘a-dín-di‘-naa‘‘in what=interr farmer=erg soup 3sg.subj-drink2-comp-concess sayaàpaa=ni‘ ‘a-‘ay-zhiamzhiam teacher=erg 3sg.subj-eat2-still ‘What, although the farmer drank all the soup, did the teacher still eat?’ In fact, for (27) and (28), sentences in which the question word is left in situ in the second clause (29a–b), or in which the question word is extracted only as far as the beginning of the second clause (30), are easier to process, and in that sense preferable; sentences (27) and (28) are nevertheless perfectly acceptable.

344 David A. Peterson and Kenneth VanBik

làwthlawpaa=ni‘ tihaà] ‘a-dín-di‘-naa‘‘in farmer=erg soup 3sg.subj-drink2-comp-concess ‘aháw=ni‘=da‘ bu‘ ‘a-‘ay-zhiamzhiam who=erg=interr rice 3sg.subj-eat2-still ‘Who, although the farmer drank all the soup, still ate the rice?’ b. làwthlawpaa=ni‘ tihaà] ‘a-dín-di‘-naa‘‘in farmer=erg soup 3sg.subj-drink2-comp-concess sayaàpaa=ni‘ zày=da‘ ‘a-‘ay-zhiamzhiam teacher=erg what=interr 3sg.subj-eat2-still ‘What, although the farmer drank all the soup, did the teacher still eat?’

(29) a.

(30) làwthlawpaa=ni‘ tihaà] ‘a-dín-di‘-naa‘‘in famer=erg soup 3sg.subj-drink2-comp-concess zày=da‘ sayaàpaa=ni‘ ‘a-‘ay-zhiamzhiam what=interr teacher=erg 3sg.subj-eat2-still ‘What, although the farmer drank all the soup, did the teacher still eat?’ A consideration of these extraction facts should make it clear that sentences involving the concessive construction do not operate in accordance with Ross’s Coordinate Structure Constraint, and thus do not appear to involve coordination. 2.5.3 Causals As in the case of the -naa‘in concessive converb construction, extraction shows that causal constructions with tsaà=‘a‘ or ruà]=‘a‘ involve subordination, as (32)–(35), based on sentence (31), indicate. (31) làwthlawpaa=ni‘ tihaà] ‘a-dín tsaà=‘‘a‘‘ farmer=erg soup 3sg.subj-drink2 reason=loc sayaàpaa=ni‘ bu‘ ‘a-‘áy teacher=erg rice 3sg.subj-eat2 ‘Because the farmer drank the soup, the teacher ate the rice.’ (32) ‘ahaw=ni‘=da‘ tihaà] ‘a-dín tsaà=‘‘a‘‘ who=erg=interr soup 3sg.subj-drink2 reason=loc sayaàpaa=ni‘ bu‘ ‘a-‘áy teacher=erg rice 3sg.subj-eat2 ‘Because who drank the soup did the teacher eat the rice?’ (33) zày=da‘ làwthlawpaa=ni‘ ‘a-dín tsaà=‘‘a‘‘ what=interr farmer=erg 3sg.subj-drink2 reason=loc sayaàpaa=ni‘ bu‘ ‘a-‘áy teacher=erg rice 3sg.subj-eat2 ‘Because the farmer drank what did the teacher eat the rice?’

Coordination in Hakha Lai (Tibeto-Burman) 345

(34) ‘aháw=ni‘=da‘ làwthlawpaa=ni‘ tihaà] ‘a-dín tsaà=‘‘a‘‘ who=erg-interr farmer=erg soup 3sg.subj-drink2 reason=loc bu‘ ‘a-‘áy rice 3sg.subj-eat2 ‘Who ate the rice because the farmer drank the soup?’ (35) zày=da‘ làwthlawpaa=ni‘ tihaà] ‘a-dín tsaà=‘‘a‘‘ what=interr farmer=erg soup 3sg.subj-drink2 reason=loc sayaàpaa=ni‘ ‘a-‘áy teacher=erg 3sg.subj-eat2 ‘What did the teacher eat because the farmer drank the soup?’ Versions of these questions in which the wh-element is left in situ are of course possible, and in the case of (35), highly preferred. If the first clause is intransitive, however, placing less of a burden on processing, sentences in which a wh-element is extracted from the second clause sound perfectly fine, as in (36). (36) zày=da‘ làwthlawpaa ‘a-tlún tsaà=‘‘a‘‘ what=interr farmer 3sg.subj-return2 reason=loc sayaàpaa=ni‘ ‘a-tshuán teacher=erg 3sg.subj-cook2 ‘What did the teacher cook because the farmer returned?’ 2.5.4 Causal =‘‘ií and conjoining =‘‘ií Lastly, a consideration of extraction facts suggests that while the causal use of =‘ií involves subordination, the neutral conjunction use of =‘ií is more likely to be an instance of coordination. So, the sentence in (37) has two possible interpretations: one which is a more or less neutral conjunction of the two events, without even a necessary temporal order between them, and another in which the first clause is understood to be the cause of the second clause. (37) làwthlawpaa=ni‘ vok-saa ‘a-‘áy=‘‘ií, farmer=erg pig-meat 3sg.subj-eat2=and sayaàpaa=ni‘ zuù ‘a-dín teacher=erg rice.alcohol 3sg.subj-drink2 ‘The farmer ate the pork and the teacher drank the rice alcohol.’ or ‘Because the farmer ate the pork, the teacher drank the rice alcohol.’ Extractions like the ones we have seen above are judged to be best when there is a causal relationship between the two events, to such an extent that it is sometimes impossible to get the neutral conjunction interpretation. In (38) and (39), in which the extracted interrogative elements come from the first clause, the preferred interpretation is that there is a causal relationship between the two events, though

346 David A. Peterson and Kenneth VanBik

the neutral conjunction interpretation is also possible. The same holds if the interrogative word is left in situ. (38) ‘aháw=ni‘=da‘ vok-saa ‘a-‘áy=‘‘ií who=erg=interr pig-meat 3sg.subj-eat2=and sayaàpaa=ni‘ zuù ‘a-dín teacher=erg rice.alcohol 3sg.subj-drink2 ‘Who ate the pork such that the teacher drank the rice alcohol?’ or ‘Who ate the pork and the teacher drank the rice alcohol?’ (39) zày=da‘ làwthlawpaa=ni‘ ‘a-‘áy=‘‘ií, what=interr farmer=erg 3sg.subj-eat2=and sayaàpaa=ni‘ zuù ‘a-dín teacher=erg rice.alcohol 3sg.subj-drink2 ‘What did the farmer eat such that the teacher drank the rice alcohol?’ or ‘What did the farmer eat and the teacher drank the rice alcohol?’ As we have seen for earlier cases of extraction, when the extracted element is from the second clause, the in situ question type, or a question in which the interrogative word is extracted only to the beginning of the second clause, is preferable. In such cases, while the causal interpretation is still more likely, the neutral interpretation is not judged as entirely impossible. Extraction of the interrogative word to the beginning of the entire sentence is also possible; in these cases, though, only the causal interpretation is possible ((40) and (41)). (40) ‘aháw=ni‘=da‘ làwthlawpaa=ni‘ vok-saa ‘a-‘áy=‘‘ií who=erg=interr farmer=erg pig-meat 3sg.subj-eat2=and zuù ‘a-dín rice.alcohol 3sg.subj-drink2 ‘Who drank the rice alcohol because the farmer ate the pork?’ (*‘Who drank the rice alcohol and the farmer ate the pork?’) (41) zày=da‘ làwthlawpaa=ni‘ vok-saa ‘a-‘áy=‘‘ií what=interr farmer=erg pig-meat 3sg.subj-eat2=and sayaàpaa=ni‘ ‘a-dín teacher=erg 3sg.subj-drink2 ‘What did the teacher drink because the farmer ate the pork?’ (*‘What did the teacher drink and the farmer ate the pork?’) These extraction characteristics are thus largely what we would expect if =‘ií in its neutral conjunction sense is a coordinator rather than a subordinator. However, the fact that sentences involving extraction may sometimes have a secondary interpretation in which the sense is that of neutral conjunction is curious. One possible interpretation is that the neutral conjunction construction is actually intermediate

Coordination in Hakha Lai (Tibeto-Burman) 347

between true coordination and true subordination. Another possibility is that the ban on extraction from coordination constructions that is generally assumed is not as inviolable as we might like to believe, or that it may actually have more to do with straightforward processing issues rather than with actual universal restrictions on possible linguistic structures. With regard to the latter, note in particular that when the extracted element comes from the first clause, the neutral conjunction reading is possible, even if dispreferred ((38) and (39)). If extraction is from the second clause, it is not possible to get the neutral reading, as in (40) and (41). On the face of it, this indicates that the degree of extraction, whether it involves extraction over a relatively long rather than a short distance, may sometimes be a crucial factor in determining acceptability.


Special types of coordination

3.1 Emphatic coordination Phrasal emphatic conjunction makes use of bisyndetic coordinators. However, the normal monosyndetic coordinator is not the marker of such constructions, as seen in (42). (42) *làwthlawpaa=leé sayaàpaa=leé ‘àn-kal farmer=and teacher-and 3pl.subj-go1 ‘Both the farmer and the teacher went.’ Rather, the bisyndetic coordinator consists either of repeated instances of the comitative case cl*tic, =heè, or of the particle zó] ‘also’, as in the following sentences: (43) …‘ado‘]naak‘a‘ tsùn sah]àr=ni‘ khàn ‘aàrpii tsuú …finally deic wildcat=erg deic hen deic ‘a-táy=‘ií ‘aàrpii=heé ‘a-faà=leé=heè khàn 3sg.subj-defeat2=and hen=com 3sg.poss-children=coll-com deic ‘a-‘áy-di‘-hnaà 3sg.subj-eat2-comp-pl.obj ‘Finally, the wildcat defeated the hen, and he completely ate both the hen and her children.’ (44) làwthlawpaa=ni‘ ‘aàr-saa zó] ] vok-saa zó] ] ‘a-zuár farmer=erg chicken-meat also pig-meat also 3sg.subj-sell2 ‘The farmer sold both chicken and pork.’ The comitative case marker in such instances replaces any normal case marking that the relevant participant would normally require with the given predicator, as seen in (45).

348 David A. Peterson and Kenneth VanBik

(45) sah]àr=ni‘ tsùn nizaán=‘a‘ khàn tshiàr-tshak=‘a‘ wildcat=erg deic last.night=loc deic coop-uphill=loc kàn-riak-laay ‘àn-tii=‘‘ií tsiàr-thlà]=‘a‘ 1pl.subj-overnight1-irr 3pl.subj-say=and coop-downhill=loc ‘àn-riak tuuzaán tsuù tshiàr-tshak=heè 3pl.subj-overnight1 tonight deic coop-uphill=com tshiàr-thlà]=heé ka-bo‘-hnaa-laay tia‘ coop-downhill=com quot ‘a-tii=‘‘ií ‘apahni‘=‘ìn ‘a-bo‘-hnaa 3sg.subj-say1=and both=obl ‘The wildcat said, “Last night they said, ‘We’re going to sleep in the uphill coop’, and they slept in the downhill coop. Tonight I’ll lie in wait for them both at the uphill coop and the downhill coop,” and he lay in wait for them at both of them.’ Typically, the predicator bo‘ ‘to lie in wait’ requires a location marked by the oblique case marker =‘ìn, which is absent in the first instance of the use of the predicator in this sentence because of the presence of the comitative marker. The oblique marker is seen in its normal use with bo‘ in the final clause of the sentence. Emphatic conjunction at the clause level is somewhat more complicated. This requires the use of zó] rather than the comitative marker, as shown in (46a–b). làwthlawpaa=ni‘ ‘aàr zó] ] ‘a-tsook farmer=erg chicken also 3sg.subj-buy2 vok zò] ] ‘a-zuár pig also 3sg.subj=sell2 ‘The farmer both bought a chicken and sold a pig.’ b. *làwthlawpaa=ni‘ ‘aàr=heé ‘a-tsook farmer=erg chicken=com 3sg.subj-buy2 vok=heè ‘a-zuár pig=com 3sg.subj-sell2

(46) a.

When the coordinated clauses do not have an object for zó] to be associated with as they do in (46), i.e. when the verb is intransitive, the construction involves a further complication. See example (47) (47) làwthlawpaa ‘a-tluuk zó] ] ‘a-tluú, ‘a-thi‘ zo] ] farmer 3sg.subj-fall2 also 3sg.subj-fall1 3sg.subj-die2 also ‘a-thiì 3sg.subj-die1 ‘The farmer both fell and died.’

Coordination in Hakha Lai (Tibeto-Burman) 349

In (47), where both of the coordinated clauses are intransitive, zó] is associated not with an object, but with a Form 2 copy of the verb in the predicate. In a number of respects Form 2 is generally more nominal in character than Form 1 (e.g., it can be used alone without overt nominalization to yield an action nominalization), so it makes sense for Form 2 to occur here. The overall structure in sentences like this, however, is still highly unexpected. Negative emphatic coordination likewise differs from simple neutral cases of negated sentences containing coordinated phrases and from cases of coordinated negated clauses in that it uses bisyndetic coordinators rather than monosyndetic ones. (48), for instance, which is negated and contains an instance of phrasal coordination, has only a single coordinator. (49), on the contrary, gives examples in which the phrasal coordination is emphatic; here, both coordinands are marked by one of three possible coordinators: si-se‘ (compositionally made up of a reduced form of the equational copula sii and the exhortative marker, i.e., literally ‘be it’), =heè (the comitative case marker), or zó] ‘also’. (48) làwthlawpaa=leé sayaàpaa ‘àn-kal-láw farmer=and teacher 3pl.subj-go1-neg ‘The farmer and the teacher didn’t go.’ (49) a. làwthlawpaa si-se‘‘ sayaàpaa si-se‘‘ ‘àn-kal-láw b. làwthlawpaa=heé sayaàpaa=heé ‘àn-kal-láw c. làwthlawpaa zó] ] sayaàpaa zó] ] ‘àn-kal-láw ‘Neither the farmer nor the teacher went.’ d. làwthlawpaa=ni‘ vok-saa=heé/zó] ] ‘aàr-saa=heé/zó] ] farmer=erg pig-meat=com/also chicken-meat=com/also ‘a-tsoo-láw 3sg.subj-buy1-neg ‘The farmer bought neither pork nor chicken.’ In fact, all but the comitative marker may occur simultaneously, as seen in (50). (50) làwthlawpaa zó] ] si-se‘‘ farmer also be-exhort sayaàpaa zó] ] si-se‘‘ ‘àn-kal-láw teacher also be-exhort 3pl.subj-go1-neg ‘Neither the farmer nor the teacher went.’ The comitative marker only cooccurs with the coordinators in its true case sense, in sentences like (51).

350 David A. Peterson and Kenneth VanBik

(51) làwthlawpaa zó] ]=heè si-se‘‘ farmer also=com be-exhort sayaàpaa zó] ]=heè si-se‘‘ ‘àn-kal-láw teacher also=com be-exhort 3pl.subj-go1-neg ‘They went neither with the farmer nor with the teacher.’ Negative emphatic coordination of clauses involves only the zó] coordinator (the comitative case marker does not work). Just as in the cases of affirmative emphatic conjunction we saw above, in phrasal coordination, the zó] coordinator occurs in association with the part of the clauses which is emphasized, rather than in association with the predicates themselves. (52) làwthlawpaa=ni‘ ‘aàr zó] ] ‘a-tsoo-láw farmer=erg chicken also 3sg.subj-buy1-neg vok zò] ] ‘a-zuàr-láw pig also 3sg.subj-sell1-neg ‘The farmer neither bought a chicken nor sold a pig.’ (53) làwthlawpaa zó] ]=ni‘ ‘aàr ‘a-tsoo-láw farmer also=erg chicken 3sg.subj-buy1-neg sayaàpaa zó] ]=ni‘ vok ‘a-zuàr-láw teacher also=erg pig 3sg.subj-sell1-neg ‘Neither did the farmer buy a chicken, nor did the teacher sell a pig.’ Both (52) and (53) involve objects, however. When there is no object (that is, when the verb in the coordinated clauses is intransitive), as we saw for affirmative sentences earlier, there is an interesting doubling of the verb. See the example in (54). (54) làwtlawpaa ‘a-tluuk zó] ] ‘a-tluú-làw, farmer 3sg.subj-fall2 also 3sg.subj-fall1-neg ‘a-thi‘ zó] ] ‘a-thiì-làw 3sg.subj-die2 also 3sg.subj-die1-neg ‘The farmer neither fell nor died.’ Again here, the coordinated intransitive clauses involve a Form 2 verbal element for zó] to be associated with. We have identified no specialized means of expressing emphatic disjunction which would distinguish it from normal disjunction. 3.2 Representative and augmentative conjunction Of the other special conjunction types that Haspelmath describes in his survey, Lai has two: representative and augmentative conjunction. The sketch of Lai grammar by Peterson mentioned earlier describes the

Coordination in Hakha Lai (Tibeto-Burman)

postnominal elements -teè and -poòl as marking collectives, though their exact function is not entirely clear due to their elusive semantics and low text-frequency. Both of these, and the more clearly plural marker -hnaà, are used in a construction which has the semantics that Haspelmath describes for representative conjunction, though not all speakers agree on these interpretations. Each of the options in (55) is glossed as ‘The farmer bought pigs and such (e.g. other domesticated animals).’ (55) làwthlawpaa=ni‘ vok-teè-pool ‘a-tsook farmer=erg vok-teè-hnaa 3sg.subj-buy2 vok-teè-hnaa-pool vok-teè-pool-hnaa pigThe sentence in (56), where multiple elements are all marked by -teè, has approximately the same meaning: (56) làwthlawpaa=ni‘ vok-teè ‘aàr-tee tsoo-tee farmer=erg pig-coll chicken-coll cow-coll ‘a-tsook-hnaa 3sg.subj-buy2-pl.obj As is commonly the case, what Haspelmath dubs augmentative conjunction requires repetition of a predicate, shown in (57) and (58), which both involve a particular high-pitched intonation, especially on the first occurrence of the repeated item. (57) làwthlawpaa ‘a-tlii ‘a-tlii ‘a-tlii farmer 3sg.subj-run1 3sg.subj-run1 3sg.subj-run1 ‘The farmer ran and ran.’ (58) làwthlawpaa falaám=‘a‘ niì tampii ‘a-‘ùm ‘a-‘ùm=‘ií… farmer Falam=loc day many 3sg.subj-exist1 3sg.subj-exist1=and ‘The farmer stayed and stayed in Falam for many days, and…’ In these examples, no coordinator would be appropriate. 3.3 Alternative vs. disjunctive questions Alternative questions are also distinguished from disjunctive questions quite clearly in Lai. A disjunctive question simply involves a normal disjunction construction in combination with the polar question particle, =maá/=moó, which occurs sentence finally, as seen in (59).


352 David A. Peterson and Kenneth VanBik

(59) làwthlawpaa=ni‘ vok-saa ‘a-sií-làw-leè ‘aàr-saa farmer=erg pig-meat 3sg.subj-be2-neg-cond chicken-meat ‘a-‘ày=maá 3sg.subj-eat1-quest ‘Did the farmer eat pork or chicken?’ (Yes/No answer expected.) Alternative questions, on the other hand, involve a much more complicated structure, containing both the =maá question particle and the =da‘ interrogative particle seen in content questions earlier. All non-final alternatives are marked with =maá, and the last element is typically marked with =da‘, as seen in (60). (60) ‘aa ‘a-siì-maá ‘a-sií=‘a‘ tsùn suy hlayhlaak interj 3sg.subj-be1=quest 3sg.subj-be2=loc deic gold ladder thlaak=maá na-du‘ ]uùn hlayhlaak thlaak=maá drop2=quest 2sg.subj-want silver ladder drop2=quest na-du‘ thiàr hlayhlaak thlaak=da‘‘ tia‘ khàn 2sg.subj-want iron ladder drop2=interr quot deic là]‘aak-piì=ni‘ tsùn ‘a-vón-hàl ‘àn-tií raven-aug=erg deic 3sg.subj-direc-ask2 3pl.subj-say ‘“Ah, is that so? If that’s the case, do you want me to drop a gold ladder, a silver ladder, or an iron ladder?” the great raven asked her, they say.’ However, for some speakers it is also possible to simply mark all of the alternative elements with =maá, as in (61), in fact from the same text as (60). (61) ma‘khán là]‘aak-piì ‘ùm-naak=‘a‘ tsùn ‘a-khaán tshù]=‘a‘ then raven-aug exist1-nmlz=loc deic 3sg.subj-room inside=loc tsùn ‘a-vón-lu‘-piì ‘àn-tií kha‘ ma‘ deic 3sg.subj-direc-enter2-com.appl 3pl.subj-say interj deic thi]kua] hií ka-n-peek-laay suy thi]kua]=maá box deic 1sg.subj-2sg.obj-give2-irr gold box-quest na-du‘ ]uùn thi]khua]=maá na-du‘ thiàr 2sg.subj-want silver box=quest 2sg.subj-want iron thi]kua]=maá na-du‘ tia‘ khàn ‘a-tií ‘àn-tií box=quest 2sg.subj-want quot deic 3sg.subj-say 3pl.subj-say ‘Then the great raven took her into his room with him, they say, “Here, I’m going to give you a box. Do you want a gold one, or a silver one, or an iron one?” he said, they say.’ 3.4 Comitative, inclusory, and summary conjunction Other specialized types of coordination noted by Haspelmath are apparently absent in Lai.

Coordination in Hakha Lai (Tibeto-Burman) 353

In particular, though the comitative marker is involved in emphatic coordination (see § 3.1 above), comitatively marked items are not coordinated with others, as is the case in comitative coordination. Evidence against a comitative coordination analysis of comitatively marked participants includes verb agreement, which is always singular if the non-comitative S/A participant is singular. See (62). làwthlawpaa sayaàpaa=heé falaám ‘a-kàl/*‘àn-kal farmer teacher=com Falam 3sg.subj-go1/*3pl.subj-go1 ‘The farmer went to Falam with the teacher.’ b. làwthlawpaa=ni‘ sayaàpaa=heé ‘aàr-saa farmer=erg teacher=com chicken-meat ‘a-‘áy/*‘àn-‘áy 3sg.subj-eat2/*3pl.subj-eat2 ‘The farmer ate the chicken with the teacher.’

(62) a.

The comitative participant can also be extracted, as seen in (63), suggesting that it is not in a coordinate relationship with the associated NP. Thus, ‘aháw=heè=da‘ làwthlawpaa=ni‘ ‘aàr-saa ‘a-‘áy who=com=interr farmer=erg chicken-meat 3sg.subj-eat2 ‘Who did the farmer eat the chicken with?’ b. ‘aháw=heè=da‘ làwthlawpaa falaám ‘a-kal who=com=interr farmer Falam 3sg.subj-go1 ‘Who did the farmer go to Falam with?’

(63) a.

Versions of these questions with in situ question words are also possible. Note that this is in contrast to the facts for coordinands in the phrasal conjunction construction discussed earlier: làwthlawpaa=leé ‘ahaw=da‘ falaám ‘àn-kal farmer=and who=interr Falam 3pl.subj-go1 ‘The farmer and who went to Falam?’ b. *‘ahaw=da‘ làwthlawpaa=leé falaám ‘àn-kal

(64) a.

Thus, in (64a), where the second coordinand is left in situ, the question is fine, but if we attempt to extract it, as in (64b), the sentence is impossible (cf. the examples in (63), where extraction is possible). It does not appear possible to form a question in which the first of the two conjuncts is questioned, even if no clear extraction is involved: (65) a. *‘aháw=leé=da‘ sayaàpaa falaám ‘àn-kal b. *‘ahaw=da‘=leè sayàpaa falaám ‘àn-kal Modifiers also do not necessarily have scope over comitatively marked participants, as in (66), although they may.

354 David A. Peterson and Kenneth VanBik

(66) ‘a-‘it-mii làwthlawpaá-poòl sayaàpaa=heé ‘àn-thii 3sg.subj-sleep1-relzr farmer-coll teacher=com 3pl.subj-die1 ‘The sleeping farmers died together with the teacher.’

4. Ellipsis in coordination constructions In texts, there are no clear instances of ellipsis in coordination constructions. The notion of core participant ellipsis in Lai is problematic, since core participants are actually represented in each clause either by explicit or implicit verbal pronominal morphology. So, in (67), (67) làwthlawpaa=ni‘ ‘aàr ‘a-dooy=‘‘ií farmer=erg chicken 3sg.subj-chase2=and sayaàpaa=ni‘ ‘a-thla‘y teacher=erg 3sg.subj-catch2 ‘The farmer chased the chicken and the teacher caught it.’ there is no NP expression of the chicken in the second clause, but it is nonetheless represented in the verbal complex by the absence of an object agreement marker (the normal indicator of a third person singular object with verbs like thla‘y). It is not really appropriate to count this sort of anaphoric reference as an instance of ellipsis. While there are no instances we have found in texts, it does turn out to be possible to subject the predicator in either clause to ellipsis. (68) gives examples involving catalipsis of the first clause verb. Note that in these cases the verb must bear plural agreement, which is actually different from what would have been the case for two coordinate clauses without ellipsis. làwthlawpaa=ni‘ vok-saa sayaàpaa=ni‘ ‘aàr-saa ‘àn-‘áy farmer=erg pig-meat teacher=erg chicken-meat 3pl.subj-eat2 ‘The teacher ate chicken and the farmer pork.’ b. làwthlawpaa falaám sayaàpaa tidím ‘àn-kal farmer Falam teacher Tedim 3pl.subj-go1 ‘The teacher went to Tedim and the farmer to Falam.’

(68) a.

(69) contains instances of analipsis of the second clause verb: làwthlawpaa=ni‘ vok-saa ‘a-‘áy sayaàpaa=ni‘ ‘aàr-saa farmer=erg pig-meat 3sg.subj-eat2 teacher=erg chicken-meat ‘The farmer ate pork and the teacher chicken.’ b. làwthlawpaa falaám ‘a-kàl sayaàpaa tidím farmer Falam 3sg.subj-go1 teacher Tedim ‘The farmer went to Falam and the teacher to Tedim.’

(69) a.

Coordination in Hakha Lai (Tibeto-Burman)

In these cases of analipsis, the adverb beltee ‘on the other hand’ might be added following the A and S participants of the second clause. These findings regarding ellipsis are largely consonant with what Haspelmath identifies as typical patterns of ellipsis in SOV languages.



We will conclude with a few brief remarks of a diachronic nature. The clausal coordinator =‘ií is not clearly attested in Kuki-Chin with a use as a coordinator except in the most closely related Central Chin languages. We will note, however, that there is an identical element in Lai, the generalized oblique case cl*tic which occurs in subordinate clauses, as seen in (70) and (71), where we would otherwise expect the ergative and oblique case markers, respectively. Possibly the source for the clausal coordinator is to be found in the same element, although we cannot provide a reasonable developmental scenario at this point. This element in turn may be related ultimately to the Khumi genitive case cl*tic, =ie. (70) làwthlawpaa=‘ii ‘a-tsook-mii ‘aar khaá ka-hmu‘ farmer=obl 3sg.subj-buy2-relzr chicken deic 1sg.subj-see2 ‘I saw the chicken the farmer bought.’ (71) naàm=‘ií ‘a-tha‘-mii ‘aar khaá ka-hmu‘ knife=obl 3sg.subj-kill2-relzr chicken deic 1sg.subj-see2 ‘I saw the chicken killed with the knife.’ Elements which are likely to be cognate with =leé occur in Thadou, Rangkhol, Chiru and others as a comitative case cl*tic, in Hallam as an instrumental case cl*tic, and in Bawm, Pangkhua, and Hallam, and possibly in other languages in marking event coordination (Grierson 1903–4); in addition, we have noted that dialects closely related to the Hakha variety of Lai (e.g. Thantlang Lai) use =leé as a neutral clause conjoiner. In its phrase level incarnation, probably this element was originally a marker of a comitative conjunction construction which has been reanalyzed in Lai as a simple conjunction construction. Perhaps the reanalysis of this construction obscured an earlier construction based on the comitative case cl*tic, =heè, which is still seen in the bisyndetic use of that marker in emphatic conjunction. Thus, both of these constructions would have their origins in comitative conjunction constructions, even though there is technically no comitative conjunction operative at the present stage of the language’s evolution.


356 David A. Peterson and Kenneth VanBik

Abbreviations affirm affirmative aug augmentative coll collective com comitative concess concessive deic deictic dim diminutive direc directional emot emotive emph emphatic exhort exhortative interj interjection

interr irr pol pot pron quest quot rec req resp voc

interrogative irrealis polite potential pronominal question quotative reciprocal request respect vocative

References Grierson, G. A., ed. 1903–1904. Linguistic survey of India, Vol. III: Tibeto-Burman family (3 parts). Calcutta: Office of the Superintendent, Government Printing. Haspelmath, Martin. To appear. “Coordination”. In Timothy Shopen (ed). Language typology and syntactic description, 2nd edition. Cambridge: Cambridge University Press. Henderson, E. J. A. 1965. Tiddim Chin: A descriptive analysis of two texts. Oxford: Oxford University Press. Hyman, Larry and Kenneth VanBik. 2002a. “Tone and Stem 2 formation in Hakha Lai”. Linguistics of the Tibeto-Burman Area 25 (1): 113–120. Hyman, Larry and Kenneth VanBik. 2002b. “Tone and syllable structure in the Hakha Lai noun”. To appear in BLS 28. Krishan, Shree. 1980. Thadou: A grammatical sketch. Calcutta: Anthropological Survey of India. Peterson, David A. 2003. “Hakha Lai”. In Graham Thurgood and Randy LaPolla (eds). The SinoTibetan languages. London: Routledge, 409–426. Stern, Theodore. 1963. “A provisional sketch of Sizang (Siyin) Chin”. Asia Major 10 (2): 222–278.

Chapter 13

Conjunction and concatenation in Sgaw Karen Familiarity, frequency, and conceptual unity* Carol Lord California State University Long Beach

and Louisa Benson Craig

1. 2. 3. 4. 5. 6. 7. 8.

Introduction Functions of a comitative conjunction Noun coordination Modifier coordination Verb coordination Purposive concatenations Idiomatic concatenations Conclusions

Speakers of Sgaw Karen use an all-purpose conjunction for a variety of functions.1 Coordinate noun phrases, modifiers, and verb phrases occur with this conjunction, but they also occur without a conjunction in concatenative structures. The choice

*We appreciate helpful comments from Martin Haspelmath and Sandra Thompson on an earlier draft. 1.Karen is a Sino-Tibetan language spoken in Burma (Myanmar) and Thailand. A writing system, based on Burmese script, was developed in 1832 by the Rev. Dr. Jonathan Wade. The language has predominantly analytic structure; morphemes tend to be CV monosyllables with tone. Word order is primarily VO, and some verbs have grammaticalized to some extent and have taken on aspectual functions; some are described by Kato (1993). The complex tonal distinctions in Karen are primarily lexical. They are not crucial for the focus of this paper, so we have not marked them here; see Jones (1961) and Chappell (1992) for discussion. The data used here are from recorded conversations and elicitation sessions with two native speakers of the Sgaw dialect, and from written narratives and dialogs collected by Robert Jones (Jones 1961). The Sgaw dialect is spoken along the southern Burma-Thailand border, in the Tenasserim district to the south, and in the Irrawaddy delta area. Data and judgments are primarily from the second author, who was born in Insein, a suburb of Rangoon, and lived there for 23 years. The other speaker, Genievre Sandoke Tyler, lived in Rangoon; her family is from the Bassein/Myaungmya area in the Irrawaddy delta.

358 Carol Lord and Louisa Benson Craig

between these two structures appears to reflect the speakers’ view of the closeness of the relationship between the conjoined elements and their frequency of cooccurrence in the experience of the speakers. The data show the influence of relative frequency and familiarity on grammatical form. We begin with a brief overview of coordinate structure typology, iconicity, and the role of conceptual status and frequency in grammar. We describe the functions of the Sgaw Karen conjunction morpheme and consider evidence for its historical source. We then illustrate the forms of coordination for noun phrases, modifiers, and verb phrases in Sgaw Karen. We consider the contexts in which speakers choose concatenation and conjunction, and we suggest an explanation for the speaker’s choice of construction based on his conceptualization of his experience.



Reviewing data on coordinate structures from a survey of spontaneous spoken texts in a variety of languages, Mithun (1988: 331) concludes that conjoined noun phrases with no overt conjunction and no intonation break typically designate a single conceptual unit; in contrast, conceptually distinct components are separated by comma intonation (as in Gurung, a Sino-Tibetan language of Nepal). Similarly, a single intonation unit without conjunction is used for conjoined predicates constituting a conceptually unitary event, while separate intonation units are used for conceptually distinct components. In a survey of the typology of NP coordination, Stassen (2000) finds that juxtaposition or zero-marking (referred to as asyndetic coordination) is a minor strategy, typically used in list enumerations or for NP pairs which habitually go together and can be said to form a conventionalized whole. The term natural conjunction is used for things that habitually go together, distinguishing it from accidental conjunction (Haspelmath, to appear, citing Wälchli 2003). Haspelmath suggests economy as an explanation for zero-marking; for frequent conjuncts the relation between them is predictable, so overt marking is redundant. The contrast between concatenation (juxtaposition) and overt conjunction provides an example of iconicity in language (Haiman 1983, 1985): a formal structure with two components juxtaposed with no intervening grammatical morpheme is used to express a close semantic/pragmatic connection. For looser semantic/pragmatic relationships, the components are separated by a conjunction morpheme. This contrast can be viewed as an illustration of the claim that grammar is a set of recurrent regularities that arise from doing the communicative work humans do, with structure being shaped by discourse (Hopper 1998). If the code is maximally isomorphic to the experience, the experience is easier to store, retrieve and communicate (Givón 1995).

Conjunction and concatenation in Sgaw Karen 359

It was observed by Zipf (1935: 272) that the complexity of a speech configuration bears an inverse relationship to its relative frequency and its familiarity in common usage. More recent studies have also found correlations between discourse frequency and grammatical status (Bybee and Hopper 2001). Thus, the speaker’s experience of familiar routines and frequent combinations would be consistent with the speaker’s use of a less complex speech configuration — e.g., simple concatenation rather than a conjunction morpheme — to encode them.


Functions of a comitative conjunction

Karen speakers use an all-purpose conjunction df‘ in coordinate structures for noun phrases, verb phrases, and clauses: (1) j6-mo df‘ f‘ j6-pa my-mother and my-father ‘my mother and my father’ (2) h7 df‘ f‘ ja df‘ f‘ p6 k6 l7 shu thiklo pu df‘ f‘ lY thi come with me and we will go toward river in and bathe water ‘Come with me and we will go to the river and bathe.’ In (2), this morpheme also functions as a comitative marker. In this respect, Karen is similar to the many languages with comitative conjunctions cited in Stassen (2000). In addition, df‘ marks instrument noun phrases. (3) j6 lu-θi me‘u df‘ f‘ thi I pour-die fire with water ‘I put out the fire with water.’ (4) ‘6w7 khw7 hi df‘ f‘ nfkhw7θo 3sg sweep house with broom ‘She swept the house with a broom.’ (5) kwa‘ pw7 df‘ f‘ thi cup full df‘ f‘ water ‘The cup is full of water.’ The morpheme also occurs in lexicalized form as part of a nominal which suggests a literal translation of ‘thing-with (rice)’. (6) ta-df‘ f‘ thing-df‘ f‘ ‘green vegetables that people put on rice as topping or accompaniment’

360 Carol Lord and Louisa Benson Craig

Many languages utilize a single morpheme to mark comitative/instrument/manner noun phrases (Stassen 2000), and often such morphemes can be traced to a verb source diachronically, often ‘join, accompany’ (Lord 1993). In Karen, df‘ occurs in verb contexts with meaning roughly translatable as ‘get/gather/collect’: (7) p6 df‘ f‘ thi l6 θ6re-pu we get water loc pot-inside ‘We collected water in a pot.’ (8) p6 df‘ f‘ nya l6 p>a‘ we get fish loc net ‘We caught fish in a net.’ (9) p6 df‘ f‘ thf‘-mi we get pig-wild ‘We trapped wild pigs.’ (10) j6 ‘7-df‘ f‘ me t6-kwa‘ I like-get rice one-plate ‘I want a plate of rice.’ (11) t7 k6-df‘ f‘ ‘6θf t6blf say fut-df‘ new one.time ‘Say it again.’ (12) p6 mi t6-ne k6-df‘ f‘ ba we sleep neg-get fut-df‘ f‘ neg ‘Again we could not sleep.’ Verbs with high token frequency are more likely to occur in lexicalized (“dispersed”) predicates with idiosyncratic meaning and to undergo semantic depletion, and get is a high-frequency word in English conversation that illustrates this tendency (Thompson and Hopper 2001: 49–50). Although we do not have word frequency data for Karen conversation, development along these lines is a reasonable hypothesis for Karen. The pattern of pronoun use with the conjunction is consistent with a possible earlier identity as a verb. The 3sg subject pronoun is ‘6w7, and the 3sg object pronoun is ‘f. (13) ‘6w7 thi mf 3sg.subj see mother ‘He saw Mother.’ (14) mo thi ‘f mother see 3sg.obj ‘Mother saw him.’

Conjunction and concatenation in Sgaw Karen 361

However, when the 3sg object pronoun is the first member of a coordinate structure in object position, it is the subject pronoun form that is strongly preferred: (15) mo thi ‘6w7 df‘ f‘ ja mother see 3sg.subj df‘ f‘ 1sg.obj ‘Mother saw him and me.’ Languages sometimes retain morphosyntactic agreement rules from earlier stages of the language, and the Karen pronoun data can be viewed as supporting a verb as a possible historical source for the conjunction.2


Noun coordination

In Karen, nouns in coordinate structures occur with and without the overt conjunction. In (16), the conjunction df‘ is required; (17), without the conjunction, is unacceptable. (16) ‘6w7 kw7 j6-mo df‘ f‘ j6-w7 3sg invite my-mother and my-older.sibling ‘He invited my mother and my brother.’ (17) *‘6w7 kw7 j6-mo j6-w7 3sg invite my-mother my-older.sibling ‘He invited my mother and my brother.’ A similar sentence, (18), is acceptable with the conjunction, but so is (19), without the conjunction. (18) ‘6w7 kw7 j6-mo df‘ f‘ j6-pa 3sg invite my-mother and my-father ‘He invited my mother and my father.’

2.Areal similarities suggest contact as a possible influence. Matisoff (1982) cites Tibetan dor ‘pair of draught cattle’ and Jirel jor ‘pair’ as likely borrowings from Indo-Aryan; cf. Nepali jor ‘pair’, Bengali gora ‘to join’, Kashmiri dula ‘pair’. Phonologically, the Karen conjunction is atypical in containing the slightly retroflex voiced stop, which is otherwise quite rare in Karen; in a wordlist of 859 lexical items, the d appeared in only a little over three percent of the words. This could be interpreted as supporting borrowing as a source. However, as Matisoff points out (p.c.), it is not unusual for grammatical markers to contain relatively infrequent phonemes (cf. voiced interdental fricatives in English). Phonologically similar conjunctions are found in Vietnamese and Cantonese, and the Tibetan conjunction and instrument marker is the phonologically similar da]. Stassen (2000) suggests that in Central America and Siberia zero marking for NP coordination has recently been replaced by overt markers borrowed from other languages.

362 Carol Lord and Louisa Benson Craig

(19) ‘6w7 kw7 j6-mo j6-pa 3sg invite my-mother my-father ‘He invited my parents.’ In (19), ‘my mother’ and ‘my father’ are a familiar combination, most often experienced as a single conceptual unit, and as such they occur without the conjunction in Karen; the best English translation is the single word, ‘parents.’ Consider (20) and (21): ‘shoes and hat’ are not a conceptual unit, so the conjunction is necessary, and (21) is not acceptable. (20) ‘6w7 do‘-tf khfphi‘ df‘ f‘ khoθ6lY‘ 3sg ride-up shoe and hat ‘He put on shoes and a hat.’ (21) *‘6w7 do‘-tf khfphi‘ khoθ6lY‘ 3sg ride-up shoe hat ‘He put on shoes and a hat.’ In contrast, ‘shoes and socks’ are a familiar combination, a conceptual unit, and they can occur without the conjunction; in (23), an appropriate translation might be ‘footwear.’ (22) ‘6w7 do‘-tf khfphi‘ df‘ f‘ khfplf 3sg ride-up shoe and sock ‘He put on shoes and socks.’ (23) ‘6w7 do‘-tf khfphi‘ khfplf 3sg ride-up shoe sock ‘He put on shoes and socks (footwear).’ Similarly, in (24) and (25), ‘onion and ginger’ require the overt conjunction, but in (26) and (27), ‘onion and garlic’ is permissible with or without the conjunction. (24) ‘6w7 to p6θ7θa>f df‘ f‘ t6‘e 3sg pound onion and ginger ‘She pounded onion and ginger.’ (25) *‘6w7 to p6θ7θa>f t6‘e 3sg pound onion ginger ‘She pounded onion and ginger.’ (26) ‘6w7 to p6θ7θa>f df‘ f‘ p6θ7θawa 3sg pound onion and garlic ‘She pounded onion and garlic.’ (27) ‘6w7 to p6θ7θa>f p6θ7θawa 3sg pound onion garlic ‘She pounded onion and garlic.’

Conjunction and concatenation in Sgaw Karen 363

In Karen cooking, onion and garlic are used together frequently in sauces. Onion and ginger may also be used together, but this combination is used less frequently. In all these instances (16)–(27) of noun coordination, Karen speakers use simple juxtaposition, with no overt conjunction, for combinations that are common in their experience. In each case, the single conceptual unit is encoded with a more compact grammatical construction with no intervening conjunction. As (24)–(27) illustrate, these grammatical choices are culture-specific, reflecting, for example, familiar food experiences within a particular cultural tradition.

4. Modifier coordination In Karen, the use of modifiers in coordinate constructions shows a similar pattern. As illustrated in the examples below, for Karen speakers, thinness and tallness are a familiar combination; they frequently “go together”. Accordingly, both (28) and (29) occur in Karen. However, thinness and prettiness do not; someone may be thin and pretty, but one may also be plump and pretty. Thus, ‘thin tall’ occurs attributively without the conjunction, as in (28), but ‘thin pretty’ does not, as (30) shows, and the conjunction is required, as in (31). (28) ta pf‘mY x7x7 thfthf ne me y6 pY pf‘mY thing woman thin tall prt is my younger.sibling woman ‘The thin tall woman is my younger sister.’ (29) ta pf‘mY x7x7 df‘ f‘ thfthf ne me y6 pY pf‘mY thing woman thin and tall prt is my younger.sibling woman ‘The thin and tall woman is my younger sister.’ (30) *ta pf‘mY x7x7 >e>e ne me y6 pY pf‘mY thing woman thin pretty prt is my younger.sibling woman ‘The thin pretty woman is my younger sister.’ (31) ta pf‘mY x7x7 df‘ f‘ >e>e ne me y6 pY pf‘mY thing woman thin and pretty prt is my younger.sibling woman ‘The thin and pretty woman is my younger sister.’ Similarly, work may be done quickly, or well, or correctly, as in: (32) ‘6w7 mata khlekhle 3sg work quickly ‘He worked quickly.’ (33) ‘6w7 mata >e>e 3sg work well ‘He worked well.’

364 Carol Lord and Louisa Benson Craig

(34) ‘6w7 mata baba 3sg work correctly ‘He worked correctly.’ If work is done quickly, it may or may not be done well. Accordingly, ‘quickly’ and ‘well’ do not constitute a conceptual unit here, and when they occur together, as in (35), they require the conjunction; the two adverbial modifiers may not be juxtaposed, and (36) is unacceptable. (35) ‘6w7 mata khlekhle df‘ f‘ >e>e 3sg work quickly and well ‘He worked quickly and well.’ (36) *‘6w7 mata khlekhle >e>e 3sg work quickly well ‘He worked quickly and well.’ On the other hand, work that is done well is ordinarily done correctly, and so the two modifiers can be juxtaposed without the conjunction, as in (37); here a coordination with the conjunction, as in (38), is considered very awkward. (37) ‘6w7 mata >e>e baba 3sg work well correctly ‘He worked well and correctly.’ (‘He is a good and dependable worker.’) (38)(?)‘6w7 mata >e>e df‘ f‘ baba 3sg work well and correctly ‘He worked well and correctly.’


Verb coordination

As (2) illustrates, one of the functions of df‘ is to conjoin verb phrases. However, compare (39) and (40): (39) ‘6w7 thu‘ m7 df‘ f‘ pla m7‘ 3sg brush teeth and wash face ‘He brushed his teeth and washed his face.’ (40) ‘6w7 thu‘ m7 pla m7‘ 3sg brush teeth wash face ‘He brushed his teeth and washed his face (performed his morning ablutions).’ In (39) there is a coordinate structure with two verb phrases and the conjunction df‘. The structure is VP df‘ VP. But in (40) the conjunction does not appear, and

Conjunction and concatenation in Sgaw Karen 365

there is the concatenation VP VP. The concatenated structure (40) encourages a single-event construal. The juxtaposition of the two verb phrases is iconic; it corresponds to the closer relationship (intended and understood) between the event components named by each verb phrase. Concatenation is not an option for just any two sequential actions or events; compare (41) and (42): (41) ‘6w7 pla m7‘ df‘ f‘ l7 mi-lf 3sg wash face and go sleep-down ‘He washed his face and went to bed.’ (42) *‘6w7 pla m7‘ l7 mi-lf 3sg wash face go sleep-down Concatenation, as in (40) and (42), is appropriate only for familiar routines, frequent or customary patterns of experience that the speaker construes as an event. Although face-washing and going to sleep may be actions carried out in a sequence, they lack the frequent contiguity in experience that is present for the tooth-brushing-and-face-washing routine. Similarly, the two verbs in a coordinate structure in (43) name separate events which are not accepted in a concatenated sequence, as (44) shows. (43) ‘6w7 lfxi df‘ f‘ hf 3sg fall.down and cry ‘He fell down and cried.’ (44) *‘6w7 lfxi hf 3sg fall.down cry According to speakers, (44) sounds “very awkward”, and is “not a natural combination”. If we reverse the order of the verbs in (43) and (44), both coordinate and concatenated forms are felicitous — but they differ in meaning. (45) ‘6w7 hf df‘ f‘ lfxi 3sg cry and fall.down ‘He cried and he fell down.’ (“Maybe he wasn’t watching his step, and he tripped.”) (46) ‘6w7 hf lfxi 3sg cry fall.down ‘He collapsed from grief.’ The concatenation carries meaning that is not fully predictable; the meaning is not a compositional sequence of the meanings of the individual verb phrases that it comprises. The meaning difference between (45) and (46) follows in part from the

366 Carol Lord and Louisa Benson Craig

hearer’s expectation of, and creation of, a coherent single-event interpretation for the concatenated structure. But there is an additional element of meaning attached to this concatenation. For concatenated VPs, there appears to be a continuum from an analytic structure to a familiar routinized expression to an idiom whose meaning may not be fully predictable from the meanings of the component verb phrases. The ‘cry’ + ‘fall down’ combination appears to have moved away from the analytic end of this continuum. If speakers construct a single-event interpretation for the concatenation in (46), why not for (44)? If a child falls down and scrapes his knee and starts to cry, isn’t the scene a familiar one, and don’t the events cohere? From one Karen speaker’s perspective, the events in (44) did not cohere. She recalled her childhood experience and observed that, when one of the children would misbehave, the mother would punish him but would say, ‘Now don’t you cry!’ It is well known that speakers can be highly creative in concocting explanations for linguistic phenomena. However, if we take the speaker’s observation at its face value, the inference would be that the perception of eventhood depends on one’s experience, including personal experience of culture as well as linguistic input; the perception of eventhood is culture-specific. The verb sequence in (47) is acceptable as coordination but not as concatenation: (47) ‘6w7 phY‘ df‘ f‘ lfxi 3sg startle and fall.down ‘He startled and fell.’ ‘He became startled and fell.’ (48) *‘6w7 phY‘ lfxi 3sg startle fall.down The sequence of startling and falling down would seem to be pragmatically plausible as a candidate for a single-event interpretation. However, plausibility is apparently not sufficient; there must also be familiarity and frequency. The importance of repetition of familiar talk in chunking and in constituent definition has been noted by Krug (1998) and Bybee and Scheibman (1999), among others. For some verb phrase combinations, both coordination and concatenation may be possible, but with different inferences. The structures will be used in different contexts, for different communicative purposes. The coordination form, as in (49), is more likely to be used for reporting separate events, or for giving an explanation in detail. For a single-event inference, the concatenation structure, as in (50), is preferred. (49) kwa‘ lft7 df‘ f‘ θepa cup fall and break ‘The cup fell down, and it broke.’

Conjunction and concatenation in Sgaw Karen 367

(50) kwa‘ lft7 θepa cup fall break ‘The cup fell and broke.’ The coordinate structure, as in (49), has the effect of downplaying the relationship between the two subparts of the event and might be chosen for social interaction or politeness considerations. For example, someone may know that it was your favorite cup, and may not want to break the news directly; the coordinate form might be seen as conveying a kinder message. (But this is a subtle nuance at most.)

6. Purposive concatenations When two verbs are understood as having the same object, the concatenated structure typically carries a purpose inference, as in (52). The object occurs only once, and the two verbs are juxtaposed. Similar VV forms in other languages have been described as resultative compounds. (51) mahak6‘ j6 θu bYkhe df‘ f‘ ‘f bYkhe yesterday I plant corn and eat corn ‘Yesterday I planted corn and ate corn.’ (inference: not the same corn) (52) mahak6‘ j6 θu ‘f bYkhe yesterday I plant eat corn ‘Yesterday I planted corn.’ (for the purpose of eating it later) The salience of the second verb is downplayed in these constructions, as in (52) and (53). (53) j6 pwe ‘f ko I buy eat bread ‘I bought bread to eat.’ (I may or may not have eaten it yet) In these structures, the whole event is questioned or negated; a negative cannot apply just to the second verb: (54) ‘6w7 phf ‘o nya 3sg cook eat fish ‘She is cooking fish.’ (to eat) (55) ‘6w7 phf ‘o nya haa 3sg cook eat fish q ‘Is she cooking fish?’ (to eat) (56) n6 phf ‘o m6nu l7 you cook eat what q ‘What are you cooking?’ (the question is not ‘What are you eating?’)

368 Carol Lord and Louisa Benson Craig

(57) ‘6w7 t6 phf ‘o nya ba 3sg neg cook eat fish neg ‘She is not cooking fish.’ (to eat) (58) *‘6w7 phf t6 ‘o ]ya ba 3sg cook neg eat fish neg The same purpose or expected-result inference for concatenations is present when the understood object of the first verb is the understood subject of the second verb, as in: (59) ‘6w7 shf ja 3sg push me ‘She pushed me.’

j6 lfxi I fall.down ‘I fell down.’

(60) ‘6w7 shf lfxi ja 3sg push fall.down me ‘He pushed me so that I would fall.’ (his intent was for me to fall, but I may or may not have actually fallen) (61) ‘6w7 shf lfxi ja basha j6 t6 lfxi ba 3sg push fall.down me but I neg fall.down neg ‘He shoved me, but I didn’t fall down.’ In Karen, the purposive concatenative construction carries the implication that the second verb is the expected result. In some West African languages, a parallel construction would carry a similar expected-result implication, but there would be the additional inference that the expected result was in fact realized. The differences between the Sino-Tibetan and West African readings, where they exist, are due to differences between the language-specific inferences inherent in the purposive concatenative construction itself.


Idiomatic concatenations

As we saw in (46), some concatenations have taken on unpredictable idiomatic meaning and can be seen as moving along a compositional-to-idiomatic continuum. The coordinate structure, however, retains the literal meaning. Note the difference in meaning between the concatenated and coordinate alternatives: (62) ‘6w7 ha‘ df‘ f‘ lokw7 3sg go and play ‘He is going and playing.’ (63) ‘6w7 ha‘ lokw7 3sg go play ‘He is going walking for pleasure.’ (going for a stroll)

Conjunction and concatenation in Sgaw Karen 369

(64) ‘6w7 sf‘ phoθa 3sg carry child ‘He carried the child.’

phoθa thf θe child ascend tree ‘The child climbed a tree.’

(65) ‘6w7 sf‘ thf phoθa 3sg carry ascend child ‘He picked up the child.’ As (64) and (65) illustrate, ‘carry’ + ‘ascend’ results in ‘pick up’. Another concatenation, ‘walk’ + thf ‘ascend’, means ‘leave’. The verb ‘ascend’ occurs frequently as the second verb in a concatenation, e.g., as in ‘call up’, ‘take up’, ‘open up’, ‘ride up’, ‘change up’, ‘leave up’, ‘complete up’. There is a similar pattern in concatenations with lf ‘descend’, for example, ‘come down’, ‘look down’, ‘arrive down’, ‘sleep down’. Another pattern has verbs with kwi‘ ‘throw away, finish, be done with’, as in ‘give away’, ‘cut die away’, ‘walk up away’. In these instances, the generalization of the meaning and the productivity of the pattern suggest a lexical compound template. Once the meanings of these commonplace verbs become semantically depleted, generalized or partly grammaticalized, they can be used as affixes with other verbs and become an even more frequent and familiar pattern.

8. Conclusions The Karen comitative conjunction df‘ is used for a range of functions. Its use in coordinate constructions contrasts with its absence in concatenated structures. The speaker’s grammatical choices depend on her construal of the natural and cultural environment and the extent to which entities, attributes, activities and events are experienced as conceptual units. When circ*mstances or characteristics are frequently or routinely experienced together, an overt conjunction is not used; in fact, its use in many such contexts is ungrammatical. The structures suggest a continuum from syntactic to lexical, from analytic constructions to routinized expressions to idiomatic chunks. The form-meaning relationship of coordinate and concatenated structures is iconic; however, the inferences made by speakers of different languages show that constructions (such as purposive concatenations) can carry language-specific instructions for their interpretation.

Abbreviations prt particle q question

370 Carol Lord and Louisa Benson Craig

References Bybee, Joan L. and Hopper, Paul (eds.) 2001. Frequency and the Emergence of Linguistic Structure. Amsterdam: Benjamins. Bybee, Joan L. and Scheibman, Joanne. 1999. “The effect of usage on degrees of constituency: the reduction of don’t in English”. Linguistics 37 (4), 575–596. Chappell, Hilary. 1992. “The benefactive construction in Moulmein Sgaw Karen”. Linguistics of the Tibeto-Burman Area 15 (1), 11–30. Givón, T. 1995. Functionalism and Grammar. Amsterdam: Benjamins. Haiman, John. 1983. “Iconic and economic motivation.” Language 59.4, 781–819. Haiman, John (ed.) 1985. Iconicity in Syntax. Amsterdam: Benjamins. Haspelmath, Martin. To appear. “Coordination.” In Language Typology and Syntactic Description, 2nd ed., T. Shopen (ed). Cambridge: Cambridge University Press. Hopper, Paul J. 1998. “Emergent grammar.” In M. Tomasello (ed.), The New Psychology of Language, 155–175. Mahwah, NJ: Lawrence Erlbaum. Jones, R.K. 1961. Karen Linguistic Studies [University of California Publications in Linguistics 25]. Berkeley: University of California Press. Kato, A. 1993. “Verb serialization in Sgaw Karen.” Journal of Asian and African Studies 45. Krug, Manfred. 1998. “String frequency”. Journal of English Linguistics 26 (4), 286–320. Lord, Carol. 1993. Historical Change in Serial Verb Constructions. Amsterdam: Benjamins. Matisoff, James A. 1982/1987. “Conjugal bliss: an Indo-Aryan word-family PAIR/YOKE/JOIN in Tibeto-Burman”. South Asian Review 6 (3), 42–50. Reprinted 1987 in Wang Li Memorial Volumes (English Volume), 309–320. The Chinese Language Society of Hong Kong. Mithun, Marianne. 1988. “The grammaticization of coordination.” In J. Haiman and S. Thompson (eds.), Clause Combining in Grammar and Discourse, 331–359. Amsterdam: Benjamins. Stassen, Leon. 2000. “AND-languages and WITH-languages.” Linguistic Typology 4, 1–54. Thompson, Sandra A. and Hopper, Paul. 2001. “Transitivity, clause structure, and argument structure: Evidence from conversation.” In J. Bybee and P. Hopper (eds.), Frequency and the Emergence of Linguistic Structure, 27–60. Amsterdam: Benjamins. Wälchli, Bernhard. 2003. Co-compounds and Natural Coordination. Ph.D. dissertation, University of Stockholm. Zipf, George K. 1935. The Psycho-biology of Language: An Introduction to Dynamic Philology. Houghton Mifflin (Republished 1965 by MIT Press.)

Chapter 14

Riau Indonesian sama Explorations in macrofunctionality* David Gil Max Planck Institute for Evolutionary Anthropology, Leipzig

Introduction Macrofunctionality Riau Indonesian Sama: Description 4.1 The functions of sama 4.2 Sama plus reduplication 4.3 Sama and other forms 5. Sama: Analysis 5.1 Cross-linguistic replicability 5.2 Fuzziness of boundaries 5.3 Unified syntactic and semantic representations 6. Monosemy and polysemy

1. 2. 3. 4.



In the Riau dialect of Indonesian, there is a single word, sama, whose translations into English make use of various forms such as and, with, to, each other, also, same and others. Thus, from an English point of view, the Riau Indonesian word sama is associated with a wide range of meanings which may or may not be connected to each other in various ways; in other words, sama appears to be polyfunctional. However, such an Anglocentric perspective tells us more about English than

*This paper would not have been possible without the speakers of Riau Indonesian who provided all the data: Aidil, Aliawar, Anton, Arip, Damsir, Dedi, Elly Yanto, Ijal, Indra, Kairil, Nano, Nasar, Pai, Ricky, Rudi Candra, Sap, Zainudin, Zulfikar, and assorted anonymous others. This paper has also benefited from discussions with Peter Cole, Gabriella Hermon, Michael Israel, Aleksandr E. Kibrik, Martin Haspelmath and Uri Tadmor, as well as by comments from participants of the Coordination Seminar at the Max Planck Institute for Evolutionary Anthropology, Department of Linguistics, in the course of 2001.

372 David Gil

about the language under investigation, Riau Indonesian. Just because English and, with, to, and so forth have distinct functions does not mean that Riau Indonesian sama must necessarily be carved up isomorphically, into several distinct items, a sama1 ‘and’, a sama2 ‘with’, a sama3 ‘to’, and so forth. Rather, Riau Indonesian sama should be described on its own terms, in a way that reflects the realities of the language it is part of, and the grammatical competence of the speakers that use it. Moreover, for such a description, the point of departure and default hypothesis should be the characterization of sama not as polyfunctional but rather as associated with a single function, or in other words as macrofunctional. This paper, then, tells the story of the Riau Indonesian word sama as it emerges from within the language itself; in doing so it provides a case study of how to deal with the phenomenon of macrofunctionality. The organization of the paper is as follows. § 2 contains a general discussion of macrofunctionality. § 3 provides some background information about Riau Indonesian. § 4 presents a detailed description of sama. § 5 proposes a unified analysis of sama accounting for its macrofunctionality. And § 6 concludes with some remaining open questions concerning the description and analysis of sama.



The term macrofunctionality implies a “large” function, where, admittedly, the attribution of size may reflect a bias stemming from comparison with other languages. However, unlike polyfunctionality, which presupposes a plurality of functions, macrofunctionality assumes a single function, which may or may not be viably decomposable into a set of constituent subfunctions. In cases of macrofunctionality, distinct analytical strategies are available. At one end of the spectrum, macrofunctionality may be construed as involving a single holistic function with no meaningful internal divisions into distinct subfunctions; in such a case one might characterize the form in question as monosemous, and hence as vague with respect to any partitioning of its single unified meaning into submeanings. Alternatively, macrofunctionality may be seen as comprising a plurality of functions which are distinct in the sense that they bear different consequences with regard to grammatical structure and semantic interpretation, but at the same time related in that they involve a core meaning, or perhaps a network of core meanings; in such a situation one might characterize the form in question as being polysemous. Finally, at the other extreme, macrofunctionality may be viewed as lumping together totally unrelated functions; in such an instance one might characterize the form in question as being ambiguous. Thus, macrofunctionality subsumes a continuum of possibilities ranging from monosemy through various degrees of polysemy all the way to ambiguity.

Riau Indonesian sama 373

How might one go about choosing between such alternative analyses? In general, the default hypothesis in all cases of macrofunctionality should be that of monosemy, or a single unified meaning. One of the dominant design features governing the structure of language is the principle of one-form-one-meaning, which says that each form in a language has a unique meaning different from that of each other form. An overwhelming body of empirical evidence can be cited in support of this principle; see, for example, Tobin (1990). Nevertheless, the principle of one-form-one-meaning is far from exceptionless: polysemy and ambiguity are widespread throughout language. Thus, for each case of macrofunctionality that is encountered, objective criteria should be invoked in order to choose between the various analytical strategies. Following are three such criteria: (1) A single form is associated with a single meaning to the extent that: a. in a variety of genealogically, geographically and typologically unrelated languages, there exists a single form associated with a similar range of meanings; b. the boundaries between the putative distinct alternative meanings are ill-defined; c. the meaning in question can be defined in a unified manner, without recourse to disjunctions. The first criterion involves cross-linguistic replicability. If a particular broad meaning is associated with a single form in a variety of genealogically, geographically and typologically diverse languages, then within each language the form in question is monosemous or perhaps polysemous; on the other hand, if an apparent instance of macrofunctionality is found in just one language but does not recur cross-linguistically, then the form in question is ambiguous. The second criterion pertains to the degree to which the boundaries between would-be distinct meanings are well defined, and points towards the Speaker’s Intention Test: Given a form associated with putative distinct meanings, the form is vague between the meanings if it is possible to imagine a normal context in which a speaker uttering the form is indifferent with regard to the distinctions between the various meanings; otherwise it is ambiguous. For example, in English, it is easy to imagine a situation in which one might use the word spring without caring whether the spiral object in question was made of metal or plastic; this would count as vagueness. However, one would be hard put to use the word spring in a naturally-occurring context without caring whether it denoted a spiral object or a source of water; this would constitute an instance of ambiguity. Finally, the third criterion is the obvious one, namely, if it is possible to define a single common and coherent meaning without recourse to an ad hoc and unstructured listing of submeanings, then the form in question is monosemous or polysemous, whereas if no such common meaning can be defined then it is ambiguous.

374 David Gil

Application of the above criteria to a given form may yield results that are clear cut, or alternatively it may fall short of such an ideal. Ambivalent results may occur because each of the above criteria is itself fuzzy rather than categorical, and because the different criteria may sometimes conflict with one another. The functionality of linguistic forms, ranging from monosemy through polysemy to ambiguity, may be thought of as analogous to any number of other complex structures which display varying amounts of hom*ogeneity or heterogeneity, and may accordingly be characterized to differing degrees as either singular or plural. For example, one might think of monosemy as a football team so cohesive that the individual members no longer have any distinct identity. Polysemy, then, would be a loosely organized team of moody individualistic players whose individual identities compete with that of the team as a whole. And ambiguity would be a random assortment of players from different teams who just happen to find themselves one evening drinking in the same pub. In this paper, it is argued that the Riau Indonesian form sama is monosemous, associated with a single unified meaning embodying the notion of “togetherness”. However, in the final section, the possibility is acknowledged that the single function of sama might be decomposed into a few interlocking and overlapping subfunctions, in which case it might also be characterized as exhibiting a limited amount of polysemy.


Riau Indonesian

Riau Indonesian is the variety of Malay/Indonesian spoken in informal situations by the inhabitants of Riau province in east-central Sumatra.1 The population of Riau province is linguistically and ethnically heterogeneous. Although the indigenous population is mostly Malay, a majority of the present-day inhabitants are migrants from other provinces, speaking a variety of other languages. Riau Indonesian is acquired as a native language by most or all children growing up in Riau province, whatever their ethnicity. It is the language most commonly used as a lingua franca for inter-ethnic communication, and in addition, like other colloquial varieties of Indonesian, it is gradually replacing other languages and dialects as a vehicle for intra-ethnic communication. Riau Indonesian is quite different from Standard Indonesian, familiar to many general linguists from a substantial descriptive and theoretical literature. Riau

1.At the time of writing, it appears likely that Riau province will split into two distinct provinces, one comprising the mainland parts of Sumatra, the other containing the smaller offshore islands. The dialect described in this paper encompasses both regions.

Riau Indonesian sama 375

Indonesian is also distinct from a set of dialects generally referred to as Riau Malay, also used in Riau province, by ethnic Malays, primarily for intra-ethnic communication. In addition, Riau Indonesian is distinguished from another set of Malayic dialects spoken by various indigenous peoples in Riau province, known as Orang Sakai, Orang Akit, Orang Hutan and Orang Laut. Finally, Riau Indonesian is also different from the variety of Malay/Indonesian used by the ethnic Chinese residents of Riau province when speaking to non-Chinese, and by the non-Chinese when speaking to Chinese, which is sometimes referred to as “Bazaar Malay”. The Riau Indonesian data presented in this paper are the product of several years of field work in Riau province, reported on in Gil (1994c, 1999, 2000a,b, 2001a,b,c, 2002a,b,c, to appear b, 2003). From a general typological perspective, Riau Indonesian is a strongly isolating language, with no inflectional morphology, and relatively little derivational morphology or compounding. It is also a language with very flexible word order. Perhaps the most unusual feature of Riau Indonesian is the pervasiveness of underspecification: the absence of obligatory overt grammatical expression for a wide variety of categories, including number, definiteness, tense, aspect, thematic role, and ontological type. Consider, for example, a simple two-word expression such as the following: (2) Makan ayam eat chicken ‘an association of eating and chicken’ In the above sentence, makan ‘eat’ is unspecified for tense and aspect, while ayam ‘chicken’ is unmarked for number and definiteness. In addition, the semantic relation between the two words is indeterminate: the chicken could bear any thematic role whatsoever, such as agent, patient, and so forth. Finally, the expression as a whole may be associated with any ontological category: it can denote an activity, for example ‘The chicken is eating’; a thing, for example ‘the chicken that is eating’; a time, for example ‘when the chicken is eating’; a place, for example ‘where the chicken is eating’, and so forth. As suggested by the given translation, sentence (2) has a single underspecified meaning, involving eating and chicken, associated with each other in an arbitrary manner. Only in actual discourse is the range of possible interpretations substantially reduced, by the context in which the sentence is uttered. Finally, it is worth noting that, since word order is flexible, the two words may be interchanged, and the resulting construction, Ayam makan, retains the same range of possible interpretations. In Gil (1994c) and subsequent works, it is suggested that such facts reflect a grammatical organization that does without many of the staple categories of grammatical theory. In particular, it is argued that in Riau Indonesian there is no distinction between syntactic categories of noun, adjective, verb and sentence, nor

376 David Gil

between lexical categories and their phrasal projections. Instead, almost all words belong to a single open syntactic category, corresponding roughly to the traditional notion of Sentence. A theory of syntactic categories within which such claims are made more rigorous is provided in Gil (2000b). Thus, for example, in (2) above, both makan ‘eat’ and ayam ‘chicken’ belong to the category S(entence). This reflects the fact that, like most other words in the language, they can stand alone as nonelliptical sentences, and can combine freely, in any order, with other words belonging to the category S. A corollary of this analysis is that Riau Indonesian is also lacking in traditional grammatical relations such as subject and direct object. Indeed, the reader may note that everything that needs to be said about Riau Indonesian in this paper can be couched within a very limited technical vocabulary, without recourse to distinct syntactic categories or grammatical relations. Methodologically, the study of Riau Indonesian faces a number of practical difficulties. Since it is a basilectal language variety, it is often difficult or impossible to elicit reliable judgements from native speakers. What happens all too often is that the moment speakers realize that they are being questioned in a “learned” context, they switch from colloquial Riau Indonesian into the standard language, or rather their sometimes imperfect variants thereof. And when speakers do provide judgements for ordinary or everyday language, they frequently make claims that are in gross conflict with their actual linguistic behaviour, for example by characterizing as ungrammatical forms or constructions that they use all the time. Faced with such obstacles, the study of Riau Indonesian reported on here makes use of an alternative method of data collection, based on the gathering of naturalistic corpora: actual utterances produced by native speakers in real live situations, either jotted down right away into a notebook, or else recorded and subsequently transcribed. All of the data presented below is of such a character.

4. Sama: Description Although the concern of this paper is exclusively synchronic, a few introductory words about the history of sama are in order, not the least because of a little mystery that, to the best of my knowledge, has not yet been resolved. The received etymology of sama is that it is a loanword from Sanskrit; see, for example, Jones (1984: 10), Adelaar (1992: 137), and Casparis (1997: 31). This would entail that sama is cognate with, among others, Persian hæm discussed in Stilo (this volume) and briefly in § 5.1 below, and also English same. However, perusal of Zorc’s (1995) reconstructions suggests the tantalizing alternative of a pure Austronesian lineage, from a combination of Proto-Austronesian *sa- ‘one’ plus *ma ‘and, with’, or perhaps alternatively *mai ‘come’. Two of these three items have reflexes in Adelaar’s (1992) reconstructions for Proto-Malayic, *sA- ‘one’ (p. 116) and *mari(‘) ‘come’, ‘hither’

Riau Indonesian sama 377

(p. 204), thereby increasing the plausibility of an Austronesian origin. Further support for this scenario derives from the observation that in many varieties of contemporary Malay/Indonesian, the procl*tic form of the numeral ‘one’, se- (from Proto-Malayic *sA- above), is also used to express sameness, in constructions such as, from Riau Indonesian, sebesar ‘one-big’ or ‘the same size as’, and sekampung ‘one-village’ or ‘the same village as’. Of course, it could be the case that both of the alternatives are right; that is to say, that sama is a Sanskrit borrowing that landed on the fertile soil provided by the fortuitous presence, in the borrowing language, of one or more indigenous collocations with similar sound and meaning. As suggested by Zuckermann (2000), such coincidences are actually quite common in the history of words.2 Returning, now, to the present, as suggested in the previous section, almost all words in Riau Indonesian belong to the single open syntactic category S. And indeed, sama is no exception; it, too, is a member of S. What this means is that in terms of its syntactic behaviour, it may stand alone as a complete, non-elliptical sentence. When occurring as a complete sentence, Sama can have a variety of meanings, such as ‘It’s the same’, ‘They’re together’, and others. Alternatively, as a member of S, sama can combine freely with other S expressions; most of the examples considered below are of that kind. 4.1 The functions of sama A convenient mode for the representation of macrofunctionality is that of semantic maps, as developed by Kemmer (1993), Haspelmath (1997, 2003) and others. A semantic map for Riau Indonesian sama is presented in Figure 1. The semantic map makes reference to 21 would-be functions associated with sama. These 21 functions are introduced for purely expository purposes, and have no theoretical status whatsoever: one could imagine alternative and equally adequate descriptions of sama making reference to 7 functions or to 63. Indeed, the extent to which these or other functions are distinct from each other is precisely what is at issue. Of the 21 functions, 16 are represented in boldface: these are

2.In their discussion of colloquial Malay varieties, Adelaar and Prentice (1996) suggest that the presence of sama or some other marker with a similarly wide range of functions is one of eight diagnostic features for what they refer to as “Pidgin Malay Derived” isolects. As pointed out in Gil (2001a), sama is in fact the only one of these eight diagnostic features exhibited by Riau Indonesian, thereby suggesting that, to the extent that their criteria are valid, Riau Indonesian is not one of their “Pidgin Malay Derived” varieties. Indeed, the presence of a macrofunctional marker sama is characteristic of a wide range of colloquial Malay and Indonesian dialects, although, unsurprisingly, its actual range of usages varies from dialect to dialect.

378 David Gil

universal quantification


additive focus




maledictive theme

instrumental reciprocal


symmetric participant

animate locative

experiencer of excessivity

animate source

standard of comparison

animate goal


theme of cognition

transitive agent

trigger of emotion


Figure 1.A semantic map for Riau Indonesian sama

functions for which sama is a primary mode of expression, either the sole one or a common alternative. The remaining 5 functions are ones for which sama is a relatively minor mode of expression, in comparison with other more common alternatives. Some functions are connected to neighbouring functions with lines; these lines represent proximity, as motivated by the following criteria: (a) connected functions are conceptually closer; and (b) connected functions are more likely to be expressed by an identical form in Riau Indonesian or in other languages. Of the 21 functions, 16 are enclosed within a round-cornered box. This box also represents proximity, as reflected in the above two criteria. (Using lines to connect closely related functions within the box would have yielded a visually uninterpretable jumble.) However, as is suggested in § 6 below, the box also demarcates what might constitute a distinct and coherent subfunction of sama. Finally, the conjunction function is in upper case, since it constitutes the theme of this volume, and the point of departure for this paper. In the remainder of § 4.1, examples of each of the above 21 functions of sama are presented in turn.

Riau Indonesian sama 379

The first set of examples illustrate usages of sama expressing conjunction:3 (3) Doni sama Amat mau ditumbuk dia Doni sama Amat want pat-hit 3 [About a quarrel among a group of friends] ‘He wants to hit Doni and Amat’ (4) Teh o sama teh o beng, ya? tea black sama tea black ice yes [Man at coffee stall returns to table to double-check order] ‘A black tea and a black iced tea, right?’ (5) Ambil topi sama hidupkan tivi take hat sama live-ep TV [Playing card game in which loser has to do what other players bid; speaker throws his cap across the floor and then orders his friend] ‘Bring the hat and turn the TV on’ (6) Makan sama ojek, itu aja eat sama dem-dem:dist conj.op [Speaker telling how he spent 5000 Rupiah in one day] ‘On eating and motorcycle taxis, that’s all’ (7) Kerja sama sekolah, gitu work sama school like-dem-dem:dist [Discussing life] ‘Working and school, that’s how it is’ In the above examples, sama occurs between two conjoined items, the first of which is double-underlined while the second is single-underlined; in these examples, sama is rendered into English as ‘and’. In example (3), the conjoined items are names of people, in (4) they are drinks, and in (5) activities with an associated participant. In examples (6) and (7), the first item denotes an activity while the second item denotes an object; nevertheless, these constructions are still understood as conjunctions of equally ranked items. As suggested in the previous section, Riau Indonesian does not distinguish between major open syntactic categories such as NP and VP. Examples (3)–(7) above highlight one of the consequences of the absence of such a distinction. In languages that distinguish between NPs and VPs, this distinction typically interacts with conjunction in one of two ways. In many languages, different forms are used

3.In all of the examples of naturalistic data cited in this paper, the context in which the utterance occurred is indicated in square brackets. The translation given beneath it is one which is appropriate to the context in question, and does not reflect the usually much wider range of meanings the sentence could have had in other hypothetical contexts.

380 David Gil

for conjoining NPs and VPs; see, for example, Haspelmath (to appear a,b). For example, in Dagbani, mini is used to conjoin NPs while ka is used to conjoin VPs; see Olawsky (1999). Alternatively, many languages make use of the same form but impose a constraint of hom*ogeneity: both of the conjuncts must belong to the same category. Thus, in English, one can say John and Bill, sang and danced, but not *John and danced. However, as evident in (3)–(7) above, neither of these effects is present in Riau Indonesian. Examples (3)–(5) show that the same form sama is used for the equivalents of both NP conjunction and VP conjunction. And examples (6) and (7) demonstrate that sama may be used for the equivalent of conjoining a VP and an NP. In order to render the latter examples into English, the translation of the VP must be nominalized (or alternatively, the translation of the NP verbalized); however, there is no trace of such a category-changing process in the original Riau Indonesian. Thus, examples (6) and (7) provide a vivid instantiation of the generalization to the effect that in Riau Indonesian, expressions denoting objects and activities enjoy the same distributional privileges, and therefore need not be assigned to distinct syntactic categories such as NP and VP. The next example illustrates a somewhat different use of sama to express conjunction, in what is sometimes referred to as an inclusory construction. (8) Kami main sama Kairil 1 play sama Kairil [Boy taking leave from adults] ‘Me and Kairil are going to play’ In many languages, there are constraints on the occurrence of pronouns within conjunctions. In such languages, instead of constructions of the form pronoun1 coordinator X (for example, English Me and John), one finds constructions of the form pronoun2 (…) marker (…) X, where marker is typically some kind of adposition, and pronoun2 is a pronoun whose reference includes both pronoun1 and the other conjunct X (hence the term inclusory). Thus, in Riau Indonesian, for ‘Me and Kairil’ it is possible to say Saya sama Kairil (where saya is the first-person singular pronoun ‘I’); however, the preferred construction, indicated in (8), involves the pronoun kami, whose core meaning is that of first-person plural exclusive, which, in the above example, refers to both the speaker and the third-person Kairil.4 In a wide range of examples, encompassing (9)–(38) below, the function of sama resembles that of an adposition or case-marker in other languages. In these

4.As noted by Lichtenberk (2000a), inclusory constructions come in two varieties: phrasal, where the three parts of the construction (pronoun2, marker, and X) form a single constituent; and split, where they do not. Riau Indonesian has both kinds of inclusory construction; example (8) illustrates the latter, split kind, in that Kami is separated from sama Kairil by the word main ‘play’.

Riau Indonesian sama

examples, sama occurs in front of an expression, which is single-underlined, characterizing that expression as standing in a particular semantic relation, or thematic role, to some other expression, which is double-underlined. In the semantic map in Figure 1, these usages of sama comprise the functions (other than conjunction and inclusory) enclosed within the round-cornered box. As suggested by the box, these functions are all very close to each other, and the boundaries between them are often ill-defined. The first class of cases are those in which sama expresses a comitative relation, translatable into English with the preposition ‘with’. (9) Damsir beli celana sama si Man sudah bulu-bulu Damsir buy trousers sama pers Mansudir|fam pfct distr~feather [About a pair of trousers] ‘The trousers that Damsir bought with Mansudir are already all frayed’ (10) Kita makan sama bang Ai itu? 1.2 eat sama fam| Aidil|fam dem-dem:dist [Speaker asks interlocutor where he bought the rice; interlocutor says by the harbour; speaker asks] ‘The place we ate with elder brother Aidil?’ The comitative use of sama is one of its core functions, and one that occurs with high frequency. From a cross-linguistic perspective, the formal identity of conjunctions and comitatives is widespread and well-documented; see, for example Stassen (2000, to appear a) and many of the articles in this volume. In addition, the formal identity of inclusory markers and comitatives is also well-attested, see Haspelmath (to appear a), Moravcsik (to appear) and others. An example of this is the Russian preposition s, as described by Urtz (1994). A somewhat more limited function of sama is to express an instrumental relation, also translatable into English with the preposition ‘with’. (11) Berangkat sama apa? depart sama what [At harbour, waiting for boats] ‘Which boat are you leaving on’ (12) Abang sama Mikonata balik? sama Mikonata return [At harbour, asking which boat I am planning to take back to Singapore] ‘Are you returning on the Mikonata?’ The cross-linguistically widespread formal identity of comitatives and instrumentals has been described by Stolz (1996a,b), Stolz, Stroh and Urdze (to appear) and others; an obvious example of this is provided by English with. In Riau Indonesian, however,


382 David Gil

most instrumentals are more commonly expressed by means of a different form, namely pakai ‘use’; the use of sama to mark instrumentals is largely if not exclusively limited to instrumentals of vehicular motion, as in the above two examples. Another function of sama is expressing a relation that might be referred to as animate locative: (13) Saya simpan sama David mana? 1sg deposit sama David which [About some money that the speaker had left with me for safekeeping but now wants back] ‘Where’s what I deposited with you?’ (14) Tak ada duit kecil sama saya neg exist money small sama 1sg [Wanting to make purchase] ‘I don’t have any small change on me’ In general, in Riau Indonesian, ordinary locatives involving places are not marked with sama, but rather with the locative marker di. However, one particular type of locative-like relation, involving animates, can be marked with sama, as shown above. As suggested by (13) and (14), both of which happen to involve money, this semantic relationship might alternatively be characterized as temporary possessor. Indeed, as noted by Heine (1997), Stassen (to appear b) and many others, possessors are often expressed by means of locative markers, for example Russian u ‘at’ and Hebrew l- ‘to’. Another related function of sama is the expression of a relation that could be termed animate source: (15) Minta uang sama dia request money sama 3 [One beggar, catching sight of me, says to another] ‘Ask him for money’ (16) Aku beli sama David 1sg buy sama David [Offering to buy my camera off me] ‘I’ll buy it from you’ For the most part, in Riau Indonesian, ordinary sources, involving places, are marked with the source marker dari ‘from’. However, as suggested by the above examples, when the source is animate, sama is used instead. Yet another related function of sama, and a much more common one, is the expression of the opposite relation animate goal:

Riau Indonesian sama 383

(17) Kenapa David tak kasi ikan sama dia? why David neg give fish sama 3 [After fishing; usually I would give the fish that we caught to the cleaning lady; this time I hadn’t, and speaker asks why] ‘Why didn’t you give her the fish?’ (18) Tanya sama orang Cina ask sama person China [From a tale about a group of tribesmen coming to town; their radio needs repairing, so they take it to a workshop owned by a Chinaman] ‘Then they asked the Chinaman’ (19) Orang bahasa Inggeris sama David? person language English sama David [Discussing my trip to the Philippines] ‘Did people speak English to you?’ Once more, an animacy split is in evidence. In general, for inanimate goals, such as places towards which motion is directed, the procl*tic form ke ‘to’ is used; however, for animate goals, as in the above examples, sama is used instead. The formal marking of goal marking and the marking of other oblique expressions is of course so widespread cross-linguistically that it hardly warrants further discussion. A number of closely related functions of sama, exemplified in (20)–(26) below, involve relations associated with mental states. One less common function is the expression of a relation which might be denoted theme of cognition: (20) Bapak saya sama David sudah lupanya father 1sg sama David pfct forget-assoc [After long absence, I encounter speaker’s father, who does not recognize me] ‘My father has already forgotten you’ However, in such contexts, sama is relatively infrequent, while zero-marking is the more natural option. A more common function of sama is the expression of a relation which might be referred to as trigger of emotion: (21) Mister takut sama tadi? white.person fear sama pst:prox [Walking through town, we are followed by group of men who then disappear; speaker asks] ‘Were you afraid of those guys just before?’

384 David Gil

(22) Sama dua orang ini aja yang marah dia sama two person dem-dem:prox conj.op rel angry 3 [Group of people watching a trained monkey climb a coconut palm and pick coconuts; when the monkey comes back down he becomes very agitated at two children; his owner reassures me] ‘It’s only these two persons that he’s angry with’ (23) Malu lah sama pengawal ashamed contr sama official [From a tale about a king and queen; the queen is crying out loudly to have sex, but the king is embarrassed, and afraid the officials will overhear] ‘I’ll be embarrassed in front of the officials’ In each of the above examples, an expression marked with sama that is singleunderlined refers to a participant that is the trigger of an emotion, described in the expression that is double-underlined. A very common function of sama involves the expression of the experiencer relation: (24) Putih ’kan nampak sama ikan white q ag-see sama fish [Watching TV program about fishing lures] ‘The white ones, the fish can see’ (25) “Tak apa lah, nanti saya kasi tahu sama, a, bapak saya, “neg what contr fut:prox 1sg give know sama fill father 1sg kamu tidur di tempat saya, di istana”, ha 2 sleep loc place 1sg loc palace deic [From a tale of a jackfruit vendor and a princess; the princess invites the jackfruit vendor to spend the night with her; the jackfruit vendor is reluctant, and the princess coaxes him] ‘She said “It doesn’t matter, I’ll let my father know that you’re sleeping at my place, in the palace”, there’ (26) Semua enak sama David all nice sama David [Complaining that I like too many different kinds of food] ‘You like everything’ In each of the above examples, an expression preceded by sama that is singleunderlined refers to a participant that is the experiencer of a mental activity, described in the expression that is double-underlined. The above examples illustrate three different subtypes of mental activities: perception in (24), cognition in (25),

Riau Indonesian sama 385

and emotion in (26). The formal identity of experiencer marking on the one hand, and datives and obliques on the other, is of course well-documented cross-linguistically; see, for example, the articles in Verma and Mohanan (eds.) (1991). A major function of sama is to express the relation of transitive agent: (27) Minum sama abang drink sama [Speaker wants me to finish my drink quickly] ‘Drink up’ (28) Sama David pegang sama David hold [My wallet goes missing; speaker and I are looking for it] ‘You were holding it’ (29) Nampar sama komandan ag-slap sama commander [About army life, and what happens to new recruits who are caught smoking] ‘They get slapped by the commander’ (30) Talinya digigit-gigit sama ikan string-assoc pat-distr~bite sama fish [While fishing, after the speaker’s line snapped] ‘The line was bitten by a fish’ (31) Kita kasikan sama dia lagi? 1.2 give-ep sama 3 conj.op [Discussing what would happen if a passport were to be lost] ‘Would they give it back to one again?’ In each of the above examples, an expression marked with sama that is singleunderlined refers to a participant that is the agent of an activity, described in the expression that is double-underlined. Moreover, in each case, the activity is one whose semantic frame contains another participant with the role of theme or patient. For activities whose semantic frame does not contain such a second participant, for example lari ‘run’, tidur ‘sleep’, and senyum ‘smile’, sama cannot be used to mark the agent expression. Thus, the relation expressed by sama in such examples is that of transitive agent. Examples (27)–(31) above provide some indication of the range of constructions in which sama may mark transitive agents. Whereas in (27) and (28) the activity expression occurs in bare form, in (29)–(31) it bears one of a set of affixes which denote particular thematic roles and, in doing so, assign them greater salience; elsewhere, these affixes are characterized as generalized voice markers (Gil 2002c). Thus, in (29) tampar ‘slap’ is prefixed with the actor-oriented marker N-, in (30) gigit ‘bite’ is prefixed with the patient-

386 David Gil

oriented marker di-, and in (31) kasi ‘give’ is suffixed with the end-point-oriented marker -kan. In unison, then, examples (27)–(31) show that the transitive-agentmarking function of sama is independent of generalized voice and the affixes that may occur on the expression denoting the relevant activity.5 A similar function of sama involves the expression of the cause relation: (32) Jatuh sama siapa? fall sama pers-what [About a damaged camera] ‘Who dropped it?’ (33) Sama saya banyak sekali masuk itu sama 1sg much one-time enter dem-dem:dist [Watching game in which people are throwing bottletops into bucket] ‘I would get lots in’ (34) Ini rambut enak sama dia dem-dem:prox hair nice sama 3 [About some hairspray] ‘This makes one’s hair nice’ In the above examples, an expression marked with sama that is single-underlined refers to a participant that is the cause of an activity, described in the expression that is double-underlined. Some additional functions of sama, of a rather heterogeneous nature, may be characterized as involving the expression of various “minor” thematic roles. One such role is standard of comparison: (35) Vid, kalau Manado sama Jakarta mana jauh Vid? fam|David top Manado sama Jakarta which far fam|David [Discussing a friend’s planned move to Manado] ‘David, which is further, Manado or Jakarta?’ In the above example, Jakarta, marked with sama, is the standard of a comparison of which the theme of comparison is Manado. Another minor role expressed by sama is one which might be referred to as experiencer of excessivity:

5.With reference to other colloquial varieties of Indonesian, such as that of Jakarta, it is occasionally suggested that constructions corresponding to that in (30) are passive, and that sama is the colloquial equivalent of the standard form oleh, which marks the agent phrase in passive constructions. In Gil (2002c) it is argued that constructions containing the prefix diare not passive constructions in either Jakarta or Riau Indonesian. But whatever the merits of that argument, the data in (27)–(31) show clearly that sama plays no role in any putative distinction between active and passive sentences in Riau Indonesian.

Riau Indonesian sama 387

(36) Ini kecil sama David dem-dem:prox small sama David [About trousers in shop that I was about to try on] ‘They’re too small for you’ In the above example, David, marked with sama, is the experiencer of excessivity with respect to the scalar property kecil ‘small’. Yet another minor role expressed by sama is one which could be termed maledictive theme: (37) Ini jatuh sama kau dem-dem:prox fall sama 2 [Playing Mario, having just knocked off a bad mushroom] ‘Away with you’ In general, neither patients nor themes of motion can be marked with sama; in particular, jatuh ‘fall’ is an activity whose single participant cannot ordinarily be marked with sama. However, in (37), illustrating a rather uncommon construction type of high expressivity, the theme of jatuh, namely kau, is in fact marked with sama, the presence of sama contributing maledictive force to the utterance. It is interesting to note that this function also seems to be expressible in English with with, as in the translation to (37) above, and also in the Queen of Hearts’ Off with her head (from Lewis Carroll’s Alice in Wonderland). The final function of sama involving thematic roles to be discussed here is the expression of the symmetric participant relation: (38) Saya tabrak sama dia 1sg collide sama 3 [Playing Nintendo racing-car game, speaker’s car collides with other one] ‘I collided with him’ In the above example, the third person pronoun dia bears the same thematic role as the first person singular pronoun saya with respect to tabrak ‘collide’. Accordingly, example (38) could have been paraphrased symmetrically as a conjunction, bringing us full circle back to the first function of sama discussed in the beginning of this section. (Note, however, that given the constraints on conjunction pointed out above, such a paraphrase would probably have assumed the form of an inclusory construction similar to (8), for example Kami sama dia tabrak ‘Me and him collided’.) However, as is the case in (38), the same symmetric event may also be cast in an asymmetric syntactic mould reflecting an alternative asymmetric, and in this particular case speaker-oriented, perspective. As evidenced by the translation above, English has a very similar construction, again involving the use of with. All of the usages of sama discussed so far involve the expression of functions which, in the semantic map in Figure 1, are enclosed within the round-cornered box. In particular, all of the usages exemplified in (9)–(38) involve the expression of

388 David Gil

functions in which sama occurs in front of a particular expression, and marks it as standing in a certain semantic relation, or thematic role, to some other expression. As evidenced by the diversity of the examples, the semantic relation in question may vary quite a bit. However, as was observed in the course of the preceding discussion, such variation is not unconstrained. Some of the core semantic relations not expressible by sama include patients and intransitive agents. Moreover, outside the core relations, there are several more peripheral semantic relations that are not expressible by sama, including temporals, many locatives and many instrumentals. Cross-linguistically, the range of thematic roles expressible by sama overlaps considerably with that of some comitative markers which, in certain languages, have expanded their functions to a considerable degree. An example of this is provided by Seychelles Creole ek, discussed in Michaelis and Rosalie (2000) and also Haspelmath (2003). In addition, the range of thematic roles expressible by sama bears a certain resemblance to that associated with ergative case markers in languages where ergative case generalizes to cover a variety of so-called oblique functions. Some examples of case markers with a range of functions similar to sama include the “oblique” in Kabardian as described in Colarusso (1989, 1992), and in Adyghe as discussed in Kibrik (2003) based on Jakovlev and Ashxamaf (1940). We now turn to consider the remaining functions of sama, those lying outside the round-cornered box in the semantic map in Figure 1. These functions differ from the preceding ones in that they do not involve thematic roles; that is to say, they do not assert a particular semantic relation between the phrase containing sama and some other phrase. As a result, in the examples that follow, there is no counterpart to the double-underlined expressions in the preceding examples. As in the preceding examples, the expression occurring in immediate construction with sama is single-underlined. However, whereas in the preceding examples sama invariably precedes the relevant expression, in the examples below sama may either precede or follow the expression in question. The first function to be considered in this group is that of the reciprocal:6 (39) Ha, bagus, nembak sama deic good ag-shoot sama [Viewing a drawing of two people pointing guns at each other] ‘There, good, they’re shooting each other’

6.Example (39), as well as examples (50) and (54) below, also expressing reciprocals, were obtained though a method of “directed spontaneous speech production”, whereby subjects are brought into a situation conducive to a certain type of grammatical construction in the hope that they will spontaneously produce an utterance exemplifying that construction. In the case at hand, the speakers happened to observe me working on a laptop computer displaying a picture of two people pointing guns at one another, and, without being asked to, spontaneously produced the given reciprocal constructions.

Riau Indonesian sama 389

In the above example, sama marks nembak ‘shoot’ as reciprocal. Constructions such as the above are relatively infrequent; more commonly, reciprocals are expressed by reduplication, or by two more specific constructions involving a combination of sama plus reduplication; these latter constructions are discussed in § 4.2 below. The formal identity of reciprocal and other functions of sama has been described in a variety of other languages; see for example Maslova (2000) on reciprocals and comitatives in Yukagir and Bantu languages. A second function is that of additive focus, of the type commonly expressed in English by means of ‘too’, ‘also’ and ‘as well’: (40) As sama, ya? ace sama yes [Adding up score at end of card game; since an ace counts for extra points, the speaker reminds the person adding up that he had one and that he should include it] ‘An ace too, right?’ In the above example, sama marks as ‘ace’ as being within the scope of a contextually provided activity, the adding up of the score, in addition to some other cards which are presupposed to be in the scope of the same activity. Again, constructions such as the above are relatively infrequent; more typical would have been the use of one or more of a set of conjunctive operators such as juga, pun and lagi, each of which has a wide range of usages including, among others, those corresponding to English too, also and as well. Cross-linguistically, the formal identity of conjunction and additive focus is extremely widespread; see for example König (1991) and Gil (1994a,b, 1995a). A third and closely related function is that of universal quantification: (41) Sama bawa lah sama bring contr [Speaker and friend had an accident on a motorcycle; interlocutor is angry at speaker for driving carelessly; speaker responds] ‘We were both driving’ In the above example, sama marks bawa ‘bring’ or ‘drive’ as having a participant undergoing universal quantification, reflected in the English translation with ‘both’. Once again, the use of sama to express universal quantification is relatively uncommon; other more frequently occurring strategies for the expression of universal quantification include the dedicated lexical items semua ‘all’ and tiap ‘every’, reduplication of numerals as in dua-dua, ‘distr~two’ or ‘both’, and the combination of an interrogative pronoun with a conjunctive operator as in mana juga ‘which too’ or ‘any one’ in example (65) below. Cross-linguistically, it is quite common for universal quantifiers to be formally related to conjunctions and/or

390 David Gil

additive focus markers; see for example Emeneau (1980), König (1991), Gil (1994a,b, 1995a,b, to appear a), Haspelmath (1995, 1997), and others. For example, in Hungarian, minden ‘every’ is formed from mind, an emphatic additive focus marker meaning ‘also’. Indeed, in many languages of South, Southeast and East Asia, universal quantifiers are formed, like mana juga above, from the collocation of an interrogative pronoun and a conjunctive operator. One of the most basic functions of sama is the expression of togetherness; this function is sometimes referred to as the sociative: (42) Sama pergi kita, Vid sama go 1.2 fam|David [Speaker and I planning to go out of house to different places; speaker suggests that we leave together] ‘Let’s go out together’ In the above example, sama applies to pergi ‘go’ and qualifies the manner of going as involving the togetherness of the two agents referred to by the first-person inclusive pronoun kita. The above example bears a close semantic affinity to comitative constructions as in (9) and (10), but differs from them in an important way. Whereas in comitative constructions the entities that are semantically together are formally distributed over two expressions, Damsir and si Man in (9), kita and bang Ai in (10), here the entities are united within a single referring expression, the pronoun kita. We are nearing the end of this extensive description of the variegated functions of the single macrofunctional word sama. What remains, however, is another of its most central functions, namely the expression of sameness: (43) Hari yang sama Jakarta Jogja day rel sama Jakarta Jogja [Garuda airline office clerk typing booking into computer] ‘The same day, Jakarta-Jogja’ (44) Sama dengan David tasnya sama with David bag-assoc [Watching movie, man appears with shoulder bag] ‘His bag is the same as yours’ (45) Sama Arip bagus sama Arip good [Speaker wants to play game on laptop computer; I ask him which one; he refers to a game that his friend was playing just before] ‘The same one that Arip was playing is good’

Riau Indonesian sama 391

(46) Sama juga seperti ke Sungai Apit sama conj.op as dir Sungai Apit [At village called Lalang; interlocutor asks how far is it from here to Buton; speaker answers with reference to a small town called Sungai Apit whose location he assumes the interlocutor must know] ‘Same as from here to Sungai Apit’ (47) Sama sama aku sama sama 1sg [Speaker and I both leaving Batam, speaker for Padang, I for Kuala Lumpur; I tell speaker I’ll only arrive the next day; speaker says] ‘Same as me’ (48) Sama panjang, panjang ini lagi sama long long dem-dem:prox conj.op [Going through a collection of foreign currency notes and comparing their sizes] ‘Just as long .. this one’s longer’ (49) Besar sama? big sama [In graphics program on laptop computer, trying to select a line width that is identical to that used previously] ‘Is it as thick?’ The word sama constitutes the primary means for attributing or predicating the notion of sameness of any kind, including sameness of reference and sameness of properties. Semantically, the notion of sameness is trivalent, involving the theme, or that which is the same, the standard of comparison, or that which the theme is the same as, and the criterial property, or that with respect to which the theme is the same as the standard of comparison. In the above examples, these three participants are marked, where present, with single underline, sans serif font, and sans serif bold font respectively. In (43), the theme alone is present. In (44)–(47) the standard of comparison alone is present; these examples illustrate the variety of ways in which the standard of comparison may be marked. By far the most common way is that exemplified in (44), involving the form dengan ‘with’ discussed in more detail in § 4.3 below. Other, much less frequent strategies include zeromarking as in (45), use of the form seperti ‘as’ as in (46), and, interestingly, the use of sama as in (47), in its function as marker of standard of comparison, as previously exemplified in (35). Finally, in (48) and (49), the criterial property alone is present; in these examples, the English translation makes use not of ‘same’ but rather the preposition ‘as’. The formal identity of sameness and several of the other functions of sama considered above has been reported in a variety of other unrelated languages. For

392 David Gil

example, Stilo (this volume) describes Persian hæm and derived forms as expressing, among others, conjunction, reciprocal, additive focus, universal quantification, sociative and sameness, while McGregor (1990) characterizes Gooniyandi -nyali as expressing, inter alia, additive focus, universal quantification, sociative and sameness. The conceptual basis for such formal identity is quite clear. To say that two entities are “the same” is to say that they “go with each other” or are located “together” in some relevant conceptual space. For example, in (43), the two flights might be entered together into the same box in a calendar, reflecting their occurrence on the same day. Or, in (44), the two bags might appear together in a “map of bag types”, such as, for example, airlines present to customers whose check-in luggage has been lost, in order to assist them in the identification of their bags. And so forth. In the preceding pages, we have described, in some detail, the range of functions exhibited by sama, following the semantic map in Figure 1. However, in order to complete the picture, it is necessary to dwell briefly on the interaction between sama and another important morphosyntactic device, reduplication. 4.2 Sama plus reduplication In Riau Indonesian, reduplication is a very frequently-occurring morphosyntactic device. Formally, reduplication is most often complete, though occasionally it may be partial; when partial, it is most frequently the initial (C)V sequence that is copied, though occasionally other initial sequences undergo copying. Most commonly, reduplication results in a single additional copy, though less frequently a larger number of copies are created. Semantically, reduplication is, not surprisingly, macrofunctional; while its basic meaning is distributivity, it has a number of other core meanings, including iterativity and negative polarity, plus a variety of other less basic meanings, including plurality, large amount, extent or number, universal quantification, concessivity, atelicity and playfulness. There exists no correlation between the forms and the meanings of reduplication: any of the possible forms can bear any of the possible meanings. Reduplication and sama may interact in either of two ways. First, the form sama may itself be reduplicated, resulting in forms such as sama-sama and other less common variants. Alternatively, sama may occur between two copies of another expression X, resulting in a construction of the form X sama X. The interpretation of sama-sama and its variants is, broadly speaking, compositional, combining the meanings of sama and of reduplication. However, many of the possible combinations are semantically incoherent and hence unattested. The meaning that reduplication most commonly confers on sama is its most basic meaning, namely distributivity. Since distributivity presupposes a plural entity, the distributive key, in the sense defined in Gil (1995b), the effect of reduplication is

Riau Indonesian sama 393

thus to reinforce the four functions of sama which presuppose a plurality: reciprocal, universal quantification, sociative and sameness. Three of these functions are illustrated below, while the remaining one, universal quantification, for which there are no unambivalent examples in the corpus, is exemplified in § 5.2, in the discussion of the ill-definedness of boundaries between functions. The reciprocal function is illustrated by the following example: (50) Sama-sama musuh distr~sama enemy [Viewing a drawing of two people pointing guns at each other] ‘They’re each other’s enemies’ For the expression of reciprocity, reduplicated sama, as in (50), is much more common than simple sama, shown in (39). In examples such as the above, reduplication expresses the distributivity that is inherent in the notion of reciprocity. The formal identity of reciprocal and distributive markings constitutes a recurrent cross-linguistic pattern; see, for example, Lichtenberk (2000b) for several Oceanic languages. An obvious example of such formal identity is provided by the English translation of (50) and the each in each other’s. The sociative function is instantiated by the following example: (51) Sama-sama pergi distr~sama go [Speaker and interlocutor about to leave hotel, but in different directions] ‘Let’s leave together’ In the case of sociative, both reduplicated sama, as in (51), and simple sama, as in (42), are commonly used. At first blush, sociative appears to involve a notion of collectivity contradicting the distributive meaning of reduplication. But this conflict is only apparent, since the collectivity and the distributivity are of different scope, that is to say, they pertain to different concepts. Thus, in the above example, togetherness applies collectively to the speaker and the interlocutor, as it makes no sense to say that each of the two, individually, is together. However, the activity of leaving distributes over the speaker and the interlocutor, since it is in fact the case that each of them, individually, is going to leave. The sameness function can be observed in the following examples: (52) Sama-sama distr~sama [Watching professional wrestling on TV, speaker commenting that the two players are evenly matched] ‘They’re the same’

394 David Gil

(53) Sa-sama warna hilang dia distr~sama colour disappear 3 [Establishing reference to computer game in which the object is to shoot coloured disks at other coloured disks: when three disks of the same colour come into contact with each other, they disappear] ‘The one which when the colours are the same they disappear’ In (52) and (53), reduplication of sama adds emphasis, by underscoring that each and every one of the relevant items is the same; it is, however, less frequent than the ordinary expression of sameness by means of simple sama, as in (43)–(49). Again, an apparent conflict between the collectivity inherent in sameness and the distributivity expressed by reduplication resolves itself under closer scrutiny: whereas the relevant set of entities are collectively the same, each entity distributively participates in the sameness relation. The second way in which reduplication interacts with sama is in the X sama X construction. In this construction X is typically a single word, though less commonly it may also consist of a closely-knit two-word expression. The X sama X construction is an instance of a more general construction type which might be referred to as intercalated reduplication, which takes the form X m X, where m is some typically grammatical marker. Intercalated reduplication occurs quite frequently across languages, though I am not familiar with any general discussion of the construction in the linguistic literature. Examples of intercalated reduplication in English include distributive numerals such as two by two, distributive manner adverbials such as suitcase after suitcase, and, interestingly, reciprocals such as man to man.7 In Riau Indonesian, the X sama X construction is the only productive example of intercalated reduplication. Thus, unlike for ordinary reduplication, there is no independent source of evidence with respect to the function of intercalated reduplication in Riau Indonesian. Broadly speaking, the range of functions of the X sama X construction is similar to that of sama-sama discussed above, encompassing reciprocal, universal quantification, sociative and sameness. However, the relative weight of these four functions varies from one construction to another. By far the most common function of the X sama X construction is the expression of reciprocity. Thus, it would seem reasonable to suggest that whereas ordinary reduplication

7.In a variant form of intercalated reduplication, the marker m attaches to the second reduplicand, represented schematically as X m-X. In Riau Indonesian there is a single lexicalized example of this, the form adik beradik ‘brothers and sisters’, derived from adik ‘younger sibling’ and the middle-voice prefix ber-. In Standard Indonesian there is a more productive construction involving the agent-oriented prefix meN-, yielding forms such as surat menyurat ‘write to each other’, from surat ‘write’. Again, it is worthy of note that both examples involve a reciprocal meaning.

Riau Indonesian sama 395

bears a basic meaning of distributivity, intercalated reduplication is inherently associated with a reciprocal interpretation. The reciprocal function of the X sama X construction is evident in the following examples: (54) Tembak sama tembak shoot sama shoot [Viewing a drawing of two people pointing guns at each other] ‘They’re shooting each other’ (55) KO sama KO, Vid knockout sama knockout fam|David [Playing a video game; both fighters fall to the ground] ‘They knocked each other out’ (56) Ayo, main kuat sama main kuat exhrt play strong sama play strong [During horseplay, interlocutor complains that speaker is playing hard while he’s playing easy; speaker responds] ‘Come on then, let’s both play hard’ (57) Kawan sama kawan mana boleh kaya gitu, ’kan? friend sama friend which can similar like-dem-dem:dist q [About somebody who has just walked out, slamming the door] ‘Amongst friends you shouldn’t do that, should you’ (58) Kalau henpon sama henpon banyak bayar top sama much pay [About the prices of telephone conversations] ‘One mobile phone to another costs a lot’ In examples (54)–(56), the reduplicated expression, underlined, denotes an activity, which the X sama X construction marks as being reciprocal. In contrast, in examples (57) and (58), the reduplicated expression denotes an object; here the effect of the X sama X construction is to characterize the object as being engaged in some contextually determined reciprocal activity. Thus, in (57) the kawan sama kawan construction means ‘friends doing things in each other’s company’, while in (58) henpon sama henpon means ‘mobile phones being used to call each other’. As noted above, the X sama X construction may also express additional functions. However, such examples are of a considerably more ambivalent nature, and are accordingly discussed in § 5.2 below. 4.3 Sama and other forms Having concluded the description of sama, it is worth acknowledging, however briefly, the existence of other forms in Riau Indonesian whose functions intersect

396 David Gil

to various degrees with those of sama. Figure 2, based on the semantic map for sama provided in Figure 1, represents the range of functions of five other forms: dengan, juga, and the three generalized voice affixes N-, di- and -kan. Since the focus of this paper is on sama, Figure 2 only indicates those functions of these five forms which overlap with those of sama; additional functions, which some of these forms have, are not represented in Figure 2. juga universal quantification


additive focus

sociative dengan CONJUNCTION

maledictive theme

comitative instrumental



experiencer of excessivity

animate locative -kan

standard of comparison


animate source

symmetric participant

animate goal


theme of cognition

transitive agent

trigger of emotion



Figure 2.Sama and selected other forms in Riau Indonesian

Of the forms represented in Figure 2, the one with the widest range of functions is dengan. In fact, the range of functions of dengan is identical to those enclosed by the round-cornered box within Figure 1, at the beginning of § 4.1. For all of the functions enclosed within the box, dengan may freely replace sama, without any change in grammaticality or meaning. The only exception to this generalization, a marginal one, is that alluded to in the discussion of examples (44) and (47): in the context of sama meaning ‘same’, when marking the standard of comparison, dengan as in (44) is strongly preferred over sama as in (47). Presumably, the motivation

Riau Indonesian sama 397

behind this constraint is to avoid the repetition of sama in two different functions. Other than this, however, it is hard to identify any language-internal principles governing the choice between the two forms, sama and dengan. Rather, the choice between the two would seem, at least in part, to reflect sociolinguistic factors: the use of dengan is more closely associated with speakers of Malay ethnicity, and/or speakers aiming for a somewhat less colloquial register. As pointed out in § 4.1, the functions enclosed within the box for dengan share certain substantive properties. Two of them involve some kind of conjunction, while all of the remaining ones are commonly associated with adpositions or casemarkers. In fact, the functions within the box are precisely those for which sama or dengan must be followed by an obligatory expression to which it assigns a certain semantic role, either conjunct, or one of the thematic roles. Moreover, since, for dengan, these are the only available functions, this entails that dengan must always be followed by an obligatory expression to which it assigns a semantic role. This brings to the fore a crucial difference between sama and dengan pertaining to syntactic category membership. As suggested at the beginning of § 4, sama belongs to the open syntactic category S: this reflects the fact that it can occur by itself as a complete non-elliptical sentence, or in construction with other S expressions with flexible word order. In contrast, dengan cannot stand on its own as a complete non-elliptical sentence; instead it must always occur in front of another expression belonging to the category S. As suggested in Gil (2000b), this property of dengan groups it together with a small set of other words in Riau Indonesian, which are characterized as belonging to the closed syntactic category S/S, that is to say, expressions which combine with an S to yield another S. Thus, sama and dengan provide a striking example of the autonomy of syntactic category membership vis-à-vis semantic function. Although exhibiting a great degree of semantic overlap with respect to the range of expressible functions, these two forms differ with respect to their distributional privileges, and accordingly belong to distinct syntactic categories. Unlike dengan, the remaining forms represented in Figure 2 are never interchangeable with sama; although they overlap with sama with regard to their meanings, their grammatical behaviour is quite different. The conjunctive operator juga shares with sama the functions of conjunction, additive focus and universal quantification. Semantically, its basic meaning is ‘too’, ‘also’, or ‘as well’. In addition, it is used to express emphatic conjunction, in which case it occurs after each of the coordinated elements. And in combination with interrogative pronouns, it forms universal quantifiers, such as mana juga ‘where too’ or ‘any one’, as in example (65) below. Syntactically, however, juga is a bird of a quite different feather. Unlike sama, but like dengan, it belongs to the category S/S; however, unlike dengan it invariably follows the element to which it attaches. The remaining three forms, N-, di- and -kan are generalized voice markers, whose function is to denote a particular thematic role and by doing so assign it

398 David Gil

greater salience; see Gil (1999, 2002b,c) for more details. Each of the three thematic roles denoted by these markers overlaps in part with the thematic roles expressed by sama. The marker N- picks out the thematic role of actor, which, as indicated in Figure 2, includes experiencer, transitive agent and cause; however, it also includes intransitive agents, which cannot be marked with sama. An illustration of the overlap between N- and sama is provided by example (29) above, Nampar sama komandan ‘They get slapped by the commander’; in this construction, the thematic role of komandan may be said to be doubly marked, once by sama and once by the actor-oriented voice prefix N- on tampar ‘slap’. Similarly, the marker di- denotes the thematic role of generalized patient, which includes symmetric participant and animate goal, but also prototypical patients, which cannot be marked with sama. And the marker -kan refers to the thematic role of end point, which includes animate source, animate goal (again) and theme of cognition, but also roles such as patient of causation, which cannot be marked with sama. From a formal perspective, the marker N- is a prefix, while the markers di- and -kan, although typically written joined on to their host words, are actually cl*tics; see Gil (2002a) for some evidence with respect to -kan.


Sama: Analysis

The preceding section provided a detailed description of the usages of sama, in terms of 21 functions, represented in the semantic map in Figure 1. This was an exercise that was appropriate for expository reasons. However, it is now time to argue that the preceding description involved a fundamental misrepresentation of reality, and that in actual fact, sama is associated not with 21 functions but rather with one single function, expressed in various ways by all of the examples in (3)–(58). This single function is a generalization of the sociative function indicated in Figure 1, centered around the notion of togetherness. In order to argue for a single unified meaning underlying the macrofunctionality of sama, the three diagnostic criteria set forward in (1) will be invoked in turn: similarly macrofunctional forms will be adduced from other languages, the boundaries between the 21 would-be functions will be shown to be fuzzy, and the unified meaning of sama will be defined in a more rigorous manner. 5.1 Cross-linguistic replicability In accordance with (1a), sama may be associated with a single meaning to the extent that, in a variety of genealogically, geographically and typologically unrelated languages, there exists a single form associated with a similar range of meanings. In the course of the discussion in § 4, note was repeatedly made of how patterns of

Riau Indonesian sama 399

Persian hæm universal quantification

Gooniyandi -nyali

sameness English with

additive focus



maledictive theme

comitative instrumental

inclusory reciprocal

symmetric participant

animate locative

experiencer of excessivity

animate source

standard of comparison

animate goal


theme of cognition

transitive agent

trigger of emotion

cause Kabardian -m

Figure 3.Sama and selected other forms in other languages

formal identity exhibited by sama have common, well-known and well-documented counterparts in other languages. These cross-linguistic patterns accordingly provide support for a unified analysis of Riau Indonesian sama. Four of the specific forms alluded to in § 4 are represented, with reference to the same semantic map as for sama, in Figure 3. Figure 3 represents the range of functions of English with, Persian hæm and its derivatives (following Stilo this volume, personal communication), Gooniyandi -nyali (as described by McGregor 1990), and the Kabardian oblique case-marker -m (in accordance with Colarusso 1989, 1992).8 Note that for each of these four forms, Figure 3 represents just the

8.It should be acknowledged that Colarusso explicitly argues that the conjunction -m is distinct from the case-marker -m. However, his examples show merely that -m can be used to conjoin expressions that are in non-oblique cases; they do not impinge on the question

400 David Gil

subset of their functions that overlap with those of Riau Indonesian sama; each of these forms has additional functions not shared with sama and hence not represented in Figure 3. As is evident from Figure 3, these four forms differ greatly with regard to their range of usages; indeed, they were chosen in part because of their diversity. Nevertheless, each of these four forms covers a substantial number of functions associated with Riau Indonesian sama. In doing so, these forms, and many others like them, which considerations of space alone preclude from discussing here, suggest that the patterns of macrofunctionality exhibited by Riau Indonesian sama are cross-linguistically widespread, attested in a wide range of genealogically, geographically and typologically diverse languages. Accordingly, they support the claim that the macrofunctionality of Riau Indonesian sama is not accidental but rather the result of general principles underlying the organization of language. 5.2 Fuzziness of boundaries In accordance with (1b), sama may be associated with a single meaning to the extent that the boundaries between the putative distinct alternative meanings are fuzzy and ill-defined. This is most certainly the case. Indeed, the 21 putative functions of sama described in the previous section were very much a Procrustean bed into which naturalistically occurring utterances containing sama had to be forced, whether or not they actually fit. In most of the cases discussed above they did manage to fit, more or less, but now is the time to examine a few of the many cases that did not. In examples (59)–(66), an utterance containing sama may be characterized as expressing more than one of the 21 supposed functions of sama posited in the preceding section. In each example, alternative English translations are provided for each of the would-be functions. Examples (59)–(61) contain the simple form sama.9 (59) Sama makan sama kopi sama eat sama coffee [At coffee shop, customer asks how much the bill is; woman in charge says 19,000, and then explains] ‘It’s for the food and the coffee’ conjunction ‘It’s for the food and the coffee too’ additive focus

whether the two functions are special cases of a more general function, as is suggested to be the case here. 9.In example (60), sama is pronounced as samo; this is due to interference from the local variety of Malay, in which most instances of word-final -a are realized as -o.

Riau Indonesian sama 401

(60) Panjangnyo, samo ini-ni, samo ini-ni, long-assoc sama dem-distr~dem:prox sama dem-distr~dem:prox samo ini-ni sama dem-distr~dem:prox [Speaker teasing interlocutor, points to his nose and says how long it is, then points to his nipple, then to his other nipple, and then to his belly button] ‘It’s so long, and this, and this, and this’ conjunction ‘It’s so long, this too, this too, this too’ additive focus ‘It’s so long, same as this, same as this, same as this’ sameness (61) Abang pergi sama? go sama [Group of people is about to leave; one asks the other] ‘Are you coming too?’ ‘Are you coming with?’ ‘Are you coming, the same as us?’

additive focus sociative sameness

Examples (62) and (63) contain reduplicated forms of sama.10 (62) Kurang ajar, sama-sama meledak itu deficient learn distr~sama ag-explode dem-dem:dist [Playing car-racing Nintendo game, speaker’s car crashes into other one and both explode] ‘Stupid, they exploded each other’ reciprocal ‘Stupid, they both exploded’ universal quantification ‘Stupid, they exploded together’ sociative (63) Sa-sama-sama ikut aku distr~sama follow 1sg [Interlocutor preparing to go out, speaker wants to come with] ‘I’m also coming’ additive focus ‘I’m coming with’ sociative ‘I’m coming the same as you’ sameness And examples (64)–(66) contain the X sama X construction.

10.Another similar example is provided by the use of reduplicated sama-sama as a conventionalized response to the expression terima kasih ‘thanks’, corresponding to the English ‘you’re welcome’. This use could also be considered as vague between various responses such as ‘the feeling is mutual’ expressing the reciprocal, ‘you too’ expressing additive focus, and ‘the same to you’ expressing sameness.

402 David Gil

(64) Masuk sama masuk syaitan semua enter sama enter devil all [Watching ghost story about western family living in Japan; two western men are engaged in fistfight, when one Japanese ghost enters one of the men and he starts using martial arts; then a second ghost enters the second man and he starts using martial arts as well] ‘They’re fighting each other with devils’ reciprocal ‘They’ve both been entered by devils’ universal quantification ‘They’ve been entered by devils the same way’ sameness (65) Hebat sama hebat mana juga mau masuk dia great sama great which conj.op want enter 3 [Watching TV sports program with digest of over one hundred of the best goals scored in world cup competition: speaker observes that whoever has possession of the ball seems to be able to get it in to the goal] ‘They’re all so good any one can score’ universal quantifier ‘They’re just as good, any one can score’ sameness (66) Bernang sama bernang swim sama swim [In swimming pool, speaker playing with friend; after uttering the above, they both submerge themselves, facing each other] ‘Let’s swim facing each other’ reciprocal ‘Let’s swim together’ sociative ‘Let’s swim the same way’ sameness Crucially, for each of the above examples, the Speaker’s Intention Test, introduced in § 2, suggests that these sentences are not ambiguous with respect to the different putative functions and associated English translations. Rather, each of these sentences was uttered in a normal every-day context in which the speaker was clearly indifferent with respect to the distinctions implied by the different would-be functions and corresponding English translations. Indeed, from the perspective of the Riau Indonesian speakers, the different functions have no reality whatsoever; only when attempting to translate the constructions into English is one forced to make arbitrary decisions, and to select one function at the expense of the others. Thus, the Speaker’s Intention Test suggests that these sentences are vague, rather than ambiguous, with respect to the different putative functions. Accordingly, the ill-definedness of boundaries between such would-be functions provides further support for the existence of a single unified meaning underlying the macrofunctionality of Riau Indonesian sama.

Riau Indonesian sama 403

5.3 Unified syntactic and semantic representations But the proof of the pudding is in the eating. In accordance with (1c), sama may be associated with a single meaning to the extent that the meaning in question can be defined in a unified manner, without recourse to disjunctions. As suggested at the beginning of § 5, the single unified meaning of sama is that of “togetherness”. However, in order to see how this single meaning may underlie all of the usages of sama considered so far, we need to step back and look at sama within the context of the general syntactic and semantic patterns of Riau Indonesian. Let us return to sentence (2) Makan ayam, provided with the very general translation ‘An association of eating and chicken’. As suggested in § 3, makan ‘eat’ and ayam ‘chicken’ are, like almost all other words in the language, members of the category S. Thus, the syntactic structure of sentence (2) is that given below: (67) Syntactic Structure of (2): S S makan

S ayam

As represented in (67) above, the syntactic structure of sentence (2) is in fact a coordination of sentences, though without an overt coordinator. The characterization of sentence (2) as a coordination reflects that fact that the two constituent words are equally ranked: they belong to the same syntactic category, and there is no evidence for any structural asymmetry between them, such as might be induced, in other languages, by agreement, government, and other similar grammatical processes. Indeed, almost all sentences in Riau Indonesian are, like (2) above, built up from the coordination of sentences.11 But what about the meaning of (2)? As suggested by the translation provided, the meaning of (2) is in fact that of a conjunction. It may be represented as follows: (68) Semantic Structure of (2): A (eat, chicken) eat


The semantics of Riau Indonesian centers around the association operator, represented above with the letter A. In its monadic, or one-place guise, the association operator

11.The term coordination is being used here to denote a formal construction type, in contrast to conjunction, which pertains to semantic properties. This usage thus differs from that in Haspelmath (to appear a).

404 David Gil

provides a semantic representation for markers of association, possession, and genitive case in many languages. For example, in English, in an expression such as John’s, the possessive ’s is interpreted as the association operator A, applying to the denotation John, yielding the formula A(John), which can be read as ‘entity associated with John’, where the detailed nature of the association is left unspecified by the grammar and is instead determined by context. However, in a typical Riau Indonesian sentence, the association operator applies polyadically, to a sequence of items, and without any overt morphosyntactic realization. For example, in (68) above, it applies dyadically to the two meaning components of the sentence, yielding the formula A(eat, chicken), which may be read as ‘an association of eating and chicken’, where the precise nature of the association is left unspecified, to be determined by context. Accordingly, sentence (2), Makan ayam, is endowed with a single unitary semantic representation which is indeterminate with respect to categories such as tense, aspect, number, definiteness, thematic roles and ontological categories. One of the ways in which the range of possibilities may be reduced is by the imposition of headedness. In general, in a coordination of the form represented in (67) above, one of the daughter Ss may optionally assume the role of head, the other daughter S thereby being characterized as its modifier. Headedness bears semantic consequences: if a daughter S is head, some of its semantic properties project up to the mother S node, thereby becoming properties of the entire construction. For example, if in sentence (2) makan ‘eat’ assumes the role of head, then the resulting interpretation is one that might be represented as follows: (69) Semantic Structure of (2), with makan as head: eat | A (chicken) eat


In the above representation, boldface indicates that eat is the head of the construction. As suggested by the double line, the head projects its identity up to the mother node of the construction, which accordingly assumes the denotation eat. In other words, when makan ‘eat’ is head, the meaning of Makan ayam is narrowed down in such a way that it must denote an activity of eating. However, it remains indeterminate with respect to other categories such as tense, aspect, number, definiteness, and thematic roles. Thus, when makan is ‘head’, what Makan ayam means is ‘eating associated with chicken’, where the nature of the association is unspecified, to be determined by context. This meaning is represented in the formula eat | A(chicken), where A is the association operator in its monadic guise, and the vertical line is to be read as “such that”, as is customary in set-theoretical notation. Alternatively, if in sentence (2) ayam ‘chicken’ assumes the status of head, then the resulting interpretation is represented as in the following:

Riau Indonesian sama 405

(70) Semantic Structure of (2), with ayam as head: chicken | A (eat) eat


In the above representation, it is now chicken that projects up to the mother node, yielding the interpretation chicken | A(eat). Thus, when ayam ‘chicken’ is head, the meaning of Makan ayam is narrowed down in such a way that it must denote chicken: ‘chicken associated with eating’, where, as before, the nature of the association is indeterminate, left open to context. Thus, headedness contributes towards the reduction of indeterminacy in Riau Indonesian. However, such reduction is of limited scope, and besides, the occurrence of headedness is in any case far from universal, since, in many naturally occurring constructions, there is no evidence for its presence. In addition to its semantic consequences, headedness also has an important effect on the syntax, and, in particular, the linear order of constituents. In order to appreciate this, it must first be noted that the syntactic and semantic tree diagrams provided in (67)–(70) above, and all of those proposed for sama below, are meant to be understood as unordered representations. For example, representation (67) should be read as saying that sentence (2), Makan ayam, is an S consisting of two daughter Ss, Makan and ayam; however, it should not be understood as encoding the relative order of the two daughter Ss, Makan and ayam. Indeed, representation (67) would be equally appropriate also for the sentence resulting from the interchange of the two words in (2), namely, Ayam makan. In other words, the fact that the nodes of tree diagrams appear in left-to-right order should be viewed as a mere artefact of the physical contingencies of the three-dimensional space we inhabit, and the resulting two-dimensionality of the paper on which we represent our thoughts. A number of scholars working within different theoretical frameworks have provided arguments in support of unordered syntactic and semantic tree structures and the representational separation of hierarchic structure and linear order; see, for example Sanders (1975), Keenan (1978), Keenan and Faltz (1986), Abels (2001) and Bury (2003). Although, as pointed out in § 3, Riau Indonesian has very flexible word order, there are still a variety of constraints on linearization running the gamut from weak to categorical. Three of the most important principles governing word order in Riau Indonesian are summarized in (71) below: (71) Principles Governing Word Order in Riau Indonesian a. Left-Headedness Heads precede modifiers; b. Iconicity The linear order of expressions mirrors a conceptualized linear order of their referents;

406 David Gil


Information Flow The linear order of expressions reflects the communicative strategies of speakers.

The above principles are formulated in extremely general terms. Not all of them apply in all circ*mstances; moreover, when two or more apply to the same domain, they may either converge in effecting the same order or else diverge in pointing towards conflicting orders. Thus, the actual word order of any given sentence is the product of a complex interplay of the above principles as well as others. These principles should accordingly be thought of as “preference rules” in the spirit of Lerdahl and Jackendoff (1983), or as constraints in the currently fashionable Optimality Theory. The first principle, Left-Headedness, places Riau Indonesian in the company of perhaps one-half of the world’s languages for which a similar principle is operative, while distinguishing it from the remaining languages for which a mirror-image Right-Headedness principle is appropriate. However, within Riau Indonesian, Left-Headedness does not apply to the many constructions that are headless; moreover, even in cases where it does apply, it occasionally conflicts with and is overridden by one or both of the other principles, resulting in rightheaded constructions. The second principle, Iconicity, is probably universal; see, for example, Haiman (1985b) and the papers in Haiman (ed.) (1985a). However, given the greater flexibility of word order in Riau Indonesian, its effect is perhaps more salient in Riau Indonesian than in many other languages. For example, in a sentence such as Adi pukul Hendrik, which can be understood either as ‘Adi hit Hendrik’ or as ‘Hendrik hit Adi’, iconicity is responsible for the former interpretation being more readily available than the latter one, since the former interpretation is one in which the linear order of referents, from agent through activity to patient, is echoed by the linear order of expressions. The third principle, Information Flow, is a cover for what are probably a variety of distinct though related principles governing the organization of information, such as old-precedes-new, the placement of certain expressions in focus position, and so on. Again, although such principles are probably universal, their application to Riau Indonesian may differ in some of the details from that in other languages. The preceding couple of pages provided a brief and rather dense sketch of the major syntactic and semantic patterns of Riau Indonesian, with numerous details left to be filled in. However, it is sufficient for the task at hand, which is to provide a characterization of the single unified meaning of the Riau Indonesian form sama. As suggested at the beginning of §4, sama, too, belongs to the category S. Thus, in terms of its distributional privileges, sama is completely interchangeable with both makan ‘eat’ and ayam ‘chicken’. Accordingly, when sama occurs in construction with an expression X, the resulting syntactic structure is identical to that in (67):

Riau Indonesian sama 407

(72) Syntactic structure of constructions containing sama: S S sama


Like (67), then, the above construction is a coordination of sentences, though without an overt coordinator. A corollary of the above is that, in terms of its formal properties, sama has no specific idiosyncratic characteristics that would support its characterization, within the language itself, as a “grammatical item” as opposed to a “content word”. We come now to the unified semantic analysis. Again, in complete analogy to (68), the semantics of constructions containing sama may be assigned the following unified representation: (73) Semantic structure of constructions containing sama: A (together, M) M


In the above representation, together is the fundamental meaning of sama, while M is the meaning of the expression X. What the above says, then, is that the meaning of the construction sama X is A(together, M), which may be read ‘an association of togetherness and M’, where, as with the eating and the chicken above, the precise nature of the association is unspecified, left open to the context. All of the usages of sama in (3)–(66) are accounted for as particular instances of the above representation. In the remainder of this section, we shall examine a few exemplary cases. Syntactic and semantic representations for the expression Doni sama Amat ‘Doni and Amat’ in example (3) are provided in (74) and (75) below: (74) Syntactic Structure of (3) S S

S Doni S sama

S Amat

408 David Gil

(75) Semantic Structure of (3) doni | A (together | A (amat)) doni

together | A (amat) together


As suggested in (74), the syntactic structure of Doni sama Amat is right-branching. Examination of (74) reveals some rather piquant differences between Riau Indonesian and languages such as English. While in English, Doni and Amat is a coordination of Doni and Amat, with and playing the role of coordinator, in Riau Indonesian, Doni sama Amat is not a coordination of Doni and Amat, and sama is not a coordinator. However, Doni sama Amat is a coordination of Doni and sama Amat, without an overt coordinator; moreover, as already suggested previously, sama Amat is itself a coordination of sama and Amat, again without an overt coordinator. This is of course no more than a striking instance of a well-known fact, namely that formal construction types such as coordination are not necessarily preserved under translation from language to language.12 Moving on to the semantics, as suggested in (75), the structure of Doni sama Amat is left-headed, following the principle of linearization in (71a). Thus, in accordance with (75) the interpretation of sama Amat is together | A(amat), to be read as ‘togetherness associated with Amat’, while the interpretation of Doni sama Amat is doni | A(together | A(amat)), or ‘Doni associated with togetherness associated with Amat’. This interpretation is still considerably broader than that of the English translation ‘Doni and Amat’, allowing for a variety of ways in which Doni might be associated with togetherness, and togetherness associated with Amat. But this is precisely how it should be, given that in different syntactic environments and real-world contexts, constructions such as Doni sama Amat may indeed have different kinds of meanings. What this means, then, is that in Riau Indonesian, Doni sama Amat does not express conjunction, even in those contexts where its most appropriate translation into English is ‘Doni and Amat’. Thus, when put alongside similar accounts for other constructions, for example Doni dengan Amat, which also translate into English conjunctions, the above analysis leads towards the conclusion that Riau Indonesian is a language without a dedicated means for the expression of conjunction. In this respect, the present paper may be viewed as a

12.It should, however, be acknowledged that there is some evidence for right-branching structure also in coordinations such as the English Doni and Amat; see, for example Haspelmath (to appear a). So perhaps the differences between Riau Indonesian and English on this score are not as great as suggested above.

Riau Indonesian sama 409

sequel to Gil (1991), which reaches a similar conclusion, concerning the absence of ‘and’, for a language of a very different syntactic profile, namely Maricopa. In (76) and (77) below, syntactic and semantic representations are provided for example (27) Minum sama abang ‘elder brother drink’. (76) Syntactic Structure of (27) S S

S Minum S sama

S abang

(77) Semantic Structure of (27) drink | A (together) | A (elder brother)) drink

together | A (elder brother) together

elder brother

As is clearly evident, the representations in (76) and (77) are completely parallel to those in (74) and (75). What this means is that in terms of syntactic and semantic structure, the expression Minum sama abang ‘elder brother drink’ in (27) is indistinguishable from the expression Doni sama Amat ‘Doni and Amat’. Both are right branching; both are left-headed; and in both, sama, bearing an abstract meaning of togetherness, provides a conceptual glue connecting the items on either side. What sets them apart, other than the obviously different words they contain, is entirely within the bounds of the vagueness inherent within the meaning of the association operator. Let us explore the bounds of this vagueness in somewhat more detail. Given a leftheaded expression of the form sama X, with a left-headed interpretation, the meaning of the construction is together | A(M), to be read as ‘togetherness associated with M’. How this corresponds, in different contexts, to translations such as ‘and M’ or ‘with M’ should be obvious. But what about some of the other thematic roles within the roundcornered box in Figure 1, such as instrumental, experiencer, or, as in (27) and (77), transitive agent? A moment’s reflection should reveal that here, too, one can conceive of the relevant activity as being conceptually together with an instrumental, experiencer or transitive-agent participant. Thus, for example, in (27) and (77), a drinking is characterized as being together with an elder brother: not together in the comitative sense of somebody else’s drinking being in the company of the elder brother, but rather together in the more abstract sense of drinking and elder brother appearing together on the stage provided by the event. With the precise nature of the elder brother’s involvement in the drinking left open to context.

410 David Gil

But what, now, of the other thematic roles that sama cannot express: how does the above analysis rule out the use of sama in these cases? Here there are two different answers, for core and peripheral roles respectively. With respect to the former, recall that the two main core roles not expressible with sama are patient and intransitive agent. Observe now that these are the two core roles that are most closely connected with the activity; this shared property provides the semantic motivation underlying the existence of absolutive case markers. What this suggests, then, is that the notion of togetherness conveyed by sama presupposes a certain degree of conceptual distance which the attribution of togetherness subsequently partially mitigates; in the case of patients and intransitive agents, this distance is lacking, and hence the precondition for the usage of sama is not met. Turning now to the peripheral roles, here the explanation is of a different and somewhat more ad hoc nature. Essentially, sama cannot be used in those cases where its function would appear to be usurped by a more semantically specific form. Thus, as noted in § 4.1, in the case of locatives, the form di is most commonly used, while in the case of instrumentals, the word pakai ‘use’ is available. However, this account cannot be the whole story, since in other cases, mentioned in § 4.3, alternative forms coexist side-by-side, without one form replacing the other; one such example is provided by dengan ‘with’. Ultimately, it would seem unrealistic to seek a completely principled explanation for why, when words encroach on each other’s semantic space, sometimes one forces the other one out, while in other cases the two manage to coexist. One just has to recognize competition from other words as a complicating factor in any account of macrofunctionality. The third and final example whose analysis we shall consider here is the construction pergi sama, part of example (61) Abang pergi sama. Recall that this example was provided three alternative English translations corresponding to three distinct putative functions: ‘Are you coming too?’ expressing additive focus, ‘Are you coming with?’ expressing the sociative, and ‘Are you coming the same as us?’ expressing sameness. However, it was argued that these three translations and associated functions are the artefact of an Anglocentric perspective, and that in reality, the construction has but a single general meaning that is vague with respect to the above distinctions. We are now in a position to define that single unified meaning more explicitly: (78) Syntactic Structure of (61) S S Abang

S S pergi

S sama

Riau Indonesian sama

(79) Semantic Structure of (61) (together | A (go)) | A (elder brother) together | A (go)

elder brother go


As in the preceding examples analyzed in this subsection, sama heads the construction it occurs in; however, unlike the preceding constructions, the construction is right-headed, and sama follows its sister constituent pergi ‘go’. Similarly, at the higher level, pergi sama heads the construction it occurs in, and this construction is right-headed, with pergi sama following Abang. As shown in (79) the expression pergi sama is interpreted as together | A(go), or ‘togetherness associated with going’, while the expression Abang pergi sama is interpreted as (together | A(go)) | A(elder brother), or ‘togetherness associated with going associated with elder brother’. In the given context, that of a group of people about to leave, the relationship of togetherness is understood as holding between two distinct agents of the going activity: speaker plus his companions, and the overtly expressed abang, or elder-brother addressee. Alas, however, it is a fact about English that there is no easy and unawkward way to say all of the above. English forces the would-be translator of sentence (61) to choose between one of a number of translations with different and more specific shades of meaning, expressing different putative functions. Choosing ‘too’ adopts the perspective of an additive focus marker: ‘Are you coming too (in addition to us)?’. Choosing ‘with’ expresses the sociative, while underscoring the close affinity between sociative and comitative functions: ‘Are you coming with (us)?’ (One could imagine a generative grammarian positing a null pronoun following sama and characterizing the construction as “comitative-drop”.) And choosing ‘same’ sets up a same-or-different comparison between speaker plus companions and addressee with regard to their respective activities: ‘Is your going the same (as ours)?’. So much the worse for our would-be English translator; but that is his or her problem, most definitely not ours. As argued above, the Riau Indonesian sentence (61) Abang pergi sama is actually vague with respect to the distinctions forced upon the unfortunate translator; it has a single unified meaning, that represented in (79) above.13

13.The reader may be wondering why the semantic representation given in (79) makes no reference to the understood ‘us’, together with whom the addressee is invited to go. This is because (79) is a representation of the sentence-level semantics of example (61), without reference to context, or to other general cognitive factors influencing the understanding of the utterance. Obviously, a deeper and more general representation of the meaning of the utterance would involve reference to the speaker and his friends, as well as to other relevant entities.


412 David Gil

Limitations of space preclude the presentation of further analyses of constructions containing sama; however, by now the pattern should be clear, since, in effect, all of the usages of sama considered in this paper are analyzed in the same way, as shown in (72) and (73). These syntactic and semantic representations constitute the unified analysis of sama, as specified in criterion (1c). This, then, is what sama really means. Unfortunately, experience suggests that some readers may feel uncomfortable with the discussion of the last few pages, which might seem like so much loose verbiage, involving the fast and easy weaving of nebulous metaphors. True, most of the English paraphrases offered for the various unified meanings are extremely awkward and ungainly; but they are provided only as a crutch for those readers who are averse to trees, brackets, and other formal symbols. Indeed, it is these formal symbols that constitute the actual representations for the syntax and semantics of sama. However, with regard to the metaphorical nature of the discussion, no apology is necessary. As Lakoff and Johnson (1980), Lakoff (1987) and many others have told us, all meaning is metaphoring. And the challenge facing any linguist dealing with other languages is that different languages construct their metaphors in different ways, and then conventionalize these metaphors into their grammar and lexicon, at which point they cease to be metaphors. To close our eyes to such processes and their outcomes is to ignore one of the most important aspects of linguistic diversity, and to impose our own language’s world view and modes of expression on languages that have different world views and different ways of saying things. So when analyzing another language, there is no alternative but to try to break free from the straitjacket of one’s native language, exercising the imagination in order to gain access into the target language’s metaphors, and by doing this to come closer to the language’s true spirit.

6. Monosemy and polysemy As shown in the preceding section, all three criteria proposed in (1), cross-linguistic replicability, fuzziness of boundaries, and the availability of a single unified meaning, lead inexorably towards a monosemic analysis of sama, underlying its entire range of usages as exemplified in (3)–(66). Accordingly, the analysis proposed in the preceding section characterizes sama as a member of the syntactic category S with a single broad meaning of ‘togetherness’. However, as suggested in § 2, monosemy is the end point of a continuum which stretches though polysemy all the way to ambiguity. Thus, the unified analysis proposed above does not preclude the possibility that the single general function associated with sama be endowed with internal structure involving the existence of two or more distinct subfunctions, thereby entailing a certain degree of polysemy. At present, though, I am aware of no clear evidence in support of the existence of

Riau Indonesian sama 413

such internal structure within the range of usages of sama. In particular, I have not yet found any reason to believe that the semantic map in Figure 1 with its 21 putative functions says anything real about sama and how it should be represented in the grammar of Riau Indonesian. Nevertheless, it should be acknowledged that the present account of sama does leave a number of important issues unanswered, and it is not beyond the realm of possibility that some of these answers might involve the positing of a certain amount of polysemy. In conclusion, we shall briefly address what is perhaps the most important outstanding question: How does the unified analysis of sama proposed above account for the numerous minute details of its semantic range, in order to specify which usages are included and which others excluded? The cogency of this question is most evident when viewed from a crosslinguistic perspective. As suggested previously, many languages possess macrofunctional forms whose ranges of usages overlap partially but not completely with that of Riau Indonesian sama. Presumably, for many such cases, arguments analogous to those presented here could be developed in support of a similar unified analysis in terms of a single general meaning involving togetherness. But then how is it possible to explain why togetherness results in one range of usages for Riau Indonesian sama but in other ranges of usages for other forms in other languages? This question may be sharpened through the consideration of a specific example in more detail. The form sama occurs also in the Jakarta dialect of Indonesian, where it is associated with a nearly but not quite identical range of usages.14 Two of the differences that I have had occasion to observe between sama in the Jakarta and Riau dialects are the following: (a) for expressing the standard of comparison following sama meaning ‘same’, sama is rarely used in Riau Indonesian, as in (47), but is frequently used in Jakarta Indonesian; and (b) for marking the inclusory construction, sama is used in Riau Indonesian, as in (8), whereas Jakarta Indonesian lacks an inclusory construction, and uses ordinary conjunction instead. These differences are minor, and indeed, I could have chosen to have written an almost identical paper, with the same conclusions, about sama in Jakarta rather than Riau Indonesian. However, if sama has the same meaning in Jakarta and Riau Indonesian, how then can the differences between the two forms be accounted for?

14.Jakarta Indonesian is the general colloquial language used in Jakarta, the capital and largest city of Indonesia, in most everyday contexts, for inter-ethnic and increasingly also intra-ethnic communication; in addition, it is gaining in currency as an informal lingua franca throughout Indonesia. Jakarta Indonesian is distinct from Betawi Malay, the native dialect of the indigenous ethnic community of Jakarta, now a small minority of the total population of the city. Some recent descriptions of Jakarta Indonesian include Wouk (1989, 1999), Cumming (2002) and Sneddon (2002a,b).

414 David Gil

Three possible modes of explanation suggest themselves: (80) Modes of explanation for variable patterns of macrofunctionality a. Polysemy Macrofunctional forms with different ranges of usages may or may not have identical unified meanings; however, they are associated with different sets of submeanings; b. Monosemy sharpened Macrofunctional forms with different ranges of usages have different unified meanings, though the differences may be subtle; c. Monosemy contextualized Macrofunctional forms with different ranges of usages have identical unified meanings; however, interaction of these identical meanings with the different grammatical systems of the respective languages results in different ranges of usages. In general, there is no reason to believe that one of the above modes of explanation should be appropriate to the exclusion of the others; rather, all three modes of explanation might be relevant with respect to a certain subset of the observable facts. In accordance with a polysemy explanation, as in (80a), the grammars of individual languages would make reference to various designated submeanings or subfunctions, with respect to which the range of usages of individual forms would be delimited. Thus, for example, while sama may share the same unified meaning in Riau and Jakarta Indonesian, additional stipulations would specify that Riau Indonesian sama may mark the inclusory construction while Jakarta Indonesian sama may not. One specific application of this method, proposed by Croft (2003) and perhaps also implied by Haspelmath (2003), involves the use of semantic maps. In accordance with this approach, the grammars of individual languages would contain semantic maps like those in Figures 1–3, with respect to which the range of functions of individual forms would be stated. Thus, for example, while sama may share the same unified meaning in Riau and Jakarta Indonesian, each dialect would, in addition, contain a specification of the range of functions expressed by sama — in effect a replication of the semantic map in Figure 1 and its counterpart for Jakarta Indonesian. This, then, would provide the locus for representing the different range of functions of sama in the two dialects. However, the method of semantic maps faces some difficult, perhaps even intractable challenges. Foremost among these is the problem of deciding which functions to include in the map. Haspelmath suggests a cross-linguistic criterion: two functions will be considered distinct if there is a form in some language that distinguishes between them, by expressing one but not the other. However, this criterion is clearly problematical. First, it presupposes that there is a single set of

Riau Indonesian sama

functions that is common to all languages, essentially universal, with respect to which all of the relevant forms in all of the world’s languages are mapped. For example, one such universal function might be the inclusory. Whereas Riau Indonesian sama would be marked as expressing the inclusory, Jakarta Indonesian sama would be marked as not expressing it, as would many other forms, including English and and with. However, neither Riau nor Jakarta Indonesian has a dedicated marker for the inclusory; nor for that matter does English. One wonders, then, how speakers of Riau and Jakarta Indonesian, and speakers of English, gain knowledge of the existence of the inclusory function, in order to mark the relevant forms in their languages as either expressing it or not expressing it, whatever the case may be. The usual solution to such quandaries is to say that the knowledge was there all the time, that is to say, it is part of an innate Universal Grammar. Although such a state of affairs cannot be ruled out on a priori grounds, it should be kept as an analytical last resort, for when all other, better motivated accounts fail. An even more severe problem with the cross-linguistic criterion for the construction of semantic maps is that by looking at more and more forms in more and more languages, one is rapidly overwhelmed with an arbitrarily large number of arbitrarily specific “small” functions. Begin, for example, with conjunction. Even a cursory perusal of Haspelmath’s (to appear a) cross-linguistic survey reveals languages that distinguish between natural and accidental conjunction, for example Erzya Mordvin (Wälchli 1988); normal and representative conjunction, for example Koasati (Kimball 1991); normal and oppositive conjunction, for example Ponapean (Rehg 1981); and the list is open-ended, containing any number of additional distinctions made by at least one of the world’s five or six thousand languages. And, by assumption, all of the above distinctions, in fact all possible combinations of the above distinctions, would end up being part, in fact just a small part, of the universal semantic map accessed by, among others, English and and Riau Indonesian sama. But this is clearly wrong; it is obviously both impractical and implausible to attribute to a single form tens, hundreds or perhaps even thousands of distinct microfunctions. However, there is a third and more principled problem with the use of semantic maps to represent the meanings of individual forms in individual languages. By definition, semantic maps make reference to lattices of disjoint functions of an atomic nature. This is fine for the purpose for which semantic maps were originally introduced, namely, providing a means for representing cross-linguistic implicational universals. However, for the representation of the meaning of individual forms in individual languages, the atomic nature of the individual functions renders such maps inadequate. Whichever way one looks at it, the architecture of meaning is inherently hierarchical. All current theories of semantics reflect this hierarchical organization, with more specific meanings contained within more general ones in a recursive manner. For example, in model-theoretical semantic theories based on


416 David Gil

Ajdukiewicz and Montague, the hierarchical nature of meanings is captured in the set-theoretical relation of inclusion defined on Boolean algebras; see, for example, Keenan and Faltz (1986). Similarly, in computational theories of lexical semantics, hierarchical structure is manifest in nonmonotonic inheritance networks, as in, for example, Evans and Gazdar (1996). Even in the realm of clause structure and thematic roles, recent work has suggested that thematic roles are not atomic but rather form tree structures containing “hyperroles” of increasing generality; see, for example, Foley and Van Valin (1984), Dowty (1991) and Kibrik (1997). Thus, according to Van Valin (1999), a hearer is a kind of perceiver, a perceiver a kind of experiencer, an experiencer a kind of actor. Within such a hierarchical organization, there is no one privileged level which is a priori the right level at which to draw semantic maps. This is why, at the beginning of § 4.1, it was stated that the semantic map for sama could just as easily have contained 7 functions or 63, rather than the arbitrarily chosen figure of 21. Returning now to sama, if indeed it turns out to be polysemous, it would be nothing more than a coincidence if one of its submeanings turned out to be coextensive with one of the 21 functions of the semantic map in Figure 1. Moreover, it is extremely unlikely that one of its submeanings would be coextensive with one of the multitude of microfunctions that one would arrive at through the method of cross-linguistic comparisons. Indeed, a potential subfunction of sama could be of any size whatsoever, ranging from very small to very large. And if several such subfunctions were shown to exist, they would not have to be disjoint; they could easily stand in more complex relations involving overlapping and inclusion. What this means, then, is that the problems posed by macrofunctionality are recursive in nature. If a certain form is polysemous, its constituent meanings will themselves present the same kinds of questions, involving choices between monosemic and polysemic analyses. It’s turtles all the way down, which is why atomic semantic maps are inadequate for the task at hand, and independent motivation must be sought for the positing of polysemy. Such motivation is most appropriately obtained from the syntactic and semantic patterns of the language itself. This might involve either specific grammatical properties of the form in question that distinguish between different subfunctions, or else independently motivated patterns that make reference to such distinctions. In the case of sama there are, perhaps, some suggestive leads which may point towards a partitioning of sama into two distinct subfunctions, one enclosed by the round-cornered box in Figure 1, the other comprising its complement set. One source of evidence pertains to the relative order of sama and its sister constituent. Whereas for some of the functions in the semantic map of Figure 1 sama may occur either before or after its sister constituent, in others, in fact, for those enclosed within the box in Figure 1, sama invariably precedes its sister constituent. This suggests a division of the usages of sama into two subfunctions,

Riau Indonesian sama 417

one allowing flexible word order, the other forcing sama to precede its sister constituent.15 A second source of evidence involves the interaction of sama and reduplication as discussed in § 4.2. Both reduplicated sama and the X sama X construction cover a range of functions that is coextensive with those that lie outside the box in Figure 1. A third and final source of evidence is provided by the form dengan discussed in § 4.3. As observed there, dengan may substitute for sama in precisely those usages where sama obligatorily precedes its sister constituent; thus, the map of dengan in Figure 2 yields a box that is coextensive with the box in Figure 1. Accordingly, a monosemic analysis of dengan might posit a single unified meaning that would then constitute one of two submeanings in a polysemic analysis of sama. So there would seem to be some prima facie evidence for a modest amount of polysemy, involving the existence of two distinct subfunctions of sama; however, the evidence is hardly overwhelming. But polysemy does not provide the only mode of explanation for variable patterns of macrofunctionality. As suggested in (80b), monosemy sharpened, a second mode of explanation allows for the maintaining of purely monosemic analyses, while pointing towards the fine-tuning of the unified meanings, in ways that will account for the different usages of different forms. Obviously, both within and across languages there exist different forms with different unitary meanings underlying different patterns of macrofunctionality. However, in cases involving forms with similar ranges of usages, this mode of explanation becomes implausible. For example, it might just be possible to propose a unified meaning for dengan that is wholly contained within the unified meaning of sama proposed in § 4.3, and which accounts in a principled way for the more limited range of usages of dengan. However, it is hard to imagine how one might tweak the unified meaning of Riau Indonesian sama proposed above in order to come up with an almost but not quite identical unified meaning for its Jakarta Indonesian counterpart, one that would account for the similar but not identical range of usages of the latter form. Clearly, in such cases, a sharpened monosemic analysis will not do, and instead a different mode of explanation is called for. This brings us to the third and final mode of explanation for variable patterns of macrofunctionality proposed in (80c), monosemy contextualized. In accordance with this approach, two forms in different languages could be associated with the same unified meaning; however, their range of usages would differ for other reasons, pertaining to the overall grammatical systems of the respective languages. One could imagine any number of such reasons. For example, the two forms could

15.However, under an alternative approach, such a partitioning would not be necessary. Given the unified analysis of sama proposed in § 5.3, the chain of causality could be construed as pointing in the opposite direction: of the variety of syntactic configurations that are available, only a subset would give rise to the relevant interpretations.

418 David Gil

have different formal properties in the respective languages: one could be an independent word and the other an affix, or both could be independent words but belonging to different syntactic categories. Or, alternatively, the two forms might find themselves competing with a different array of functionally adjacent forms in their respective languages. As emphasized by Saussure, forms do not exist in isolation, but only in the context of all of the other forms that are present in the language. Returning one last time to the minor differences in usage evinced by the form sama in Riau and in Jakarta Indonesian, it is not hard to see how a monosemic analysis might be contextualized in plausible ways, in order to account for these patterns of variation. Consider the first difference, namely the use of sama to mark the standard of comparison following sama meaning ‘same’, which is rare in Riau but common in Jakarta. As noted in the discussion of examples (43)–(49), in Riau Indonesian dengan is usually used in such contexts, as exemplified in (44). One might speculate that sama is dispreferred because it creates an awkward sequence of two consecutive occurrences of the same word, as is the case in the rare (47). So why are things different in Jakarta Indonesian? Two reasons. First, in Jakarta Indonesian the form dengan is very infrequently used; as a result it is not a readily available alternative to sama in such contexts. Secondly, in Jakarta Indonesian sama occurs in two distinct variants, sama and ama.16 And in fact, in the context under consideration here, the second of the two occurrences of sama will often assume the form ama, resulting in constructions such as Sama ama aku, corresponding to (47), and thereby mitigating the undesirable occurrence of two identical forms in sequence. Thus, the differences between Riau and Jakarta Indonesian with respect to sama in the construction in question can be straightforwardly accounted for in terms of their interaction with other grammatical features: the competition of sama with dengan in Riau but not in Jakarta, and the applicability of s-deletion to sama in Jakarta but not in Riau. The second difference between the two forms involves the availability of sama for the inclusory construction in Riau but not in Jakarta Indonesian. In this case, an obvious difference between the two dialects leaps to the eye: whereas Riau Indonesian has a dedicated exclusive first-person plural pronoun kami, as evident in the inclusory construction in (8), Jakarta Indonesian has no such pronoun, and instead uses the first-person plural pronoun kita with either inclusive or exclusive meaning. But the story is not that simple. In many languages which, like Jakarta Indonesian,

16.In both Riau and Jakarta Indonesian, there is an optional rule of initial s-deletion which applies to a small set of high frequency words; however, only in Jakarta does this set of words include sama. Whereas in many cases forms with and without initial s are in free variation, work in progress suggests that in Jakarta Indonesian, the distribution of sama and ama might be different. If this turns out to be the case, that it is possible that this will provide additional support for a polysemic analysis of sama in the Jakarta dialect.

Riau Indonesian sama 419

lack an inclusive/exclusive distinction, the unmarked first-person plural pronoun does participate in an inclusory construction; one example of such a language is Russian. Interestingly, however, Riau Indonesian also has the form kita, which, as in Jakarta, can have either inclusive or exclusive meaning. But crucially, in Riau, as in Jakarta, kita cannot enter into an inclusory construction. So what is going on here? The key to the puzzle is provided by various cross-linguistic studies of the inclusory construction, such as Schwartz (1985, 1988) and Moravcsik (to appear), which demonstrate the existence of a “focal referent hierarchy” of pronouns with respect to the inclusory construction (as well as other constructions). At the top of the hierarchy are first-person plural pronouns: if only one pronoun can enter into an inclusory construction it is a first-person plural pronoun. Next in the hierarchy are second-person plural pronouns, then finally third-person plural pronouns. Note, however, that whereas exclusive first-person plural pronouns such as Riau Indonesian kami are purely first person, inclusive or unmarked first-person plural pronouns such as Riau and Jakarta Indonesian kita may be construed as combining first- and second-person features. Thus, the focal referent hierarchy may be expanded to read as follows: exclusive first-person plural pronouns > inclusive or unmarked first-person plural pronouns > second-person plural pronouns > thirdperson plural pronouns. The relevant facts may now be accounted for by stating that for both dialects of Indonesian, the cut-off point on the focal reference hierarchy is between exclusive first-person plural pronouns and all the other pronouns. Indeed, neither dialect of Indonesian has an inclusory construction for the general firstperson plural pronoun kita, or for any second or third person plural pronouns. And now we are in a position to conclude that the reason only Riau Indonesian has an inclusory construction with sama is that only Riau Indonesian has an exclusive firstperson plural pronoun. Here too, then, the differences between Riau and Jakarta Indonesian with respect to the usage of sama can be accounted for in terms of their interaction with other aspects of the grammar, in this case the presence or absence of an exclusive first-person plural pronoun licensing the inclusory construction.17 Thus, it would seem that in many cases, the contextualization of monosemy offers a more promising mode of explanation for variable patterns of macrofunctionality. Nevertheless, it also comes replete with its own set of challenges. Emerging out of the dictum that a language is a system “où tout se tient”, this approach entails that

17.Of course, this account just pushes the puzzle one step further back, since it does not explain why, for both Riau and Jakarta Indonesian, the cut-off point on the focal reference hierarchy is above inclusive or unmarked first-person plural pronouns: it could just as easily have been below inclusive or unmarked first-person plural pronouns, as it is in Russian, or, for that matter, right at the bottom of the hierarchy, as in fact is the case in the Lampung dialect of Indonesian, which forms inclusory constructions even with the third person plural pronoun dia orang. But this is another story for another paper.

420 David Gil

in order to construct the most appropriate analysis of a single form in a language, it is necessary to have a clear picture of the entire language within which the form is embedded. Although the practical consequences of such a methodological moral are far-reaching and possibly discomforting, this is perhaps a conclusion worth bearing in mind.

Abbreviations ag assoc conj.op contr deic dir ep exhrt

agent orientation associative conjunctive operator contrastive deictic direction end-point orientation exhortative

fam fill loc pat pers pfct q

familiar filler locative patient orientation personal perfect question

References Abels, Klaus (2001) Move? Unpublished Paper, University of Connecticut, Storrs. Adelaar, K. Alexander (1992) Proto Malayic: The Reconstruction of Its Phonology and Parts of Its Lexicon and Morphology, Pacific Linguistics Series C — 119, The Australian National University, Canberra. Adelaar, K. Alexander and D. J. Prentice (1996) “Malay: Its History, Role and Spread”, in S. A. Wurm, Peter Mühlhäusler and D. T. Tryon eds., Atlas of Languages of Intercultural Communication in the Pacific, Asia, and the Americas, Mouton de Gruyter, Berlin, 673–693. Bury, Dirk (2003) “Preverbal Particles in Verb-Initial Languages”, Paper presented at the Workshop on the Syntax of Verb Initial Languages, University of Arizona, Tucson, AZ, USA, 21 February 2003. Casparis, J.G. de (1997) “Sanskrit Loan-Words in Indonesian: An Annotated Check-List of Words from Sanskrit in Indonesian and Traditional Malay”, NUSA, Linguistic Studies of Indonesian and Other Languages in Indonesia, Universitas Katolik Indonesia Atma Jaya, Jakarta. Colarusso, John (1989) “East Circassian (Kabardian Dialect)”, in B. G. Hewitt ed., The Indigenous Languages of the Caucasus, Volume 2, The North West Caucasian Languages, Caravan Books, Delmar, 261–355. Colarusso, John (1992) A Grammar of the Kabardian Language, University of Calgary Press, Calgary. Croft, William (2003) Typology and Universals, Second Edition, Cambridge University Press, Cambridge. Cumming, Susanna (2002) “On -in”, Paper presented at the 9th International Conference on Austronesian Linguistics, The Australian National University, Canberra, Australia, 10 January 2002. Dowty, David (1991) “Thematic Proto-Roles and Argument Selection”, Language 67: 547–619.

Riau Indonesian sama 421

Emeneau, Murray B. (1980) “The Indian Linguistic Area Revisited”, in M. B. Emeneau, Language and Linguistic Area, Stanford University Press, Stanford, 197–249. Evans, Roger and Gerald Gazdar (1996) “DATR: A Language for Lexical Knowledge Representation”, Computational Linguistics 22: 167–216. Foley, William A. and Robert D. Van Valin, Jr. (1984) Functional Syntax and Universal Grammar, Cambridge University Press, Cambridge. Gil, David (1991) “Aristotle Goes to Arizona, And Finds a Language without And”, in D. Zaefferer ed., Semantic Universals and Universal Semantics, Foris Press, Berlin, 96–130. Gil, David (1994a) “Conjunctive Operators: A Unified Semantic Analysis”, in P. Bosch and R. van der Sandt eds., Focus and Natural Language Processing, Volume 2, Semantics, Proceedings of a Conference in Celebration of the 10th Anniversary of the Journal of Semantics, 12th–15th June, 1994, Hotel Schloß Wolfsbrunnen, Meinhard-Schwebda, Germany, Working Papers of the Institute for Logic and Linguistics, Working Paper 7, ISSN 0946–7521, IBM TR-80.94–007, 311–322. Gil, David (1994b) “Conjunctive Operators in South-Asian Languages”, in A. Davison and F. M. Smith eds., Papers from the Fifteenth South Asian Language Analysis Roundtable Conference, South Asian Studies Program, University of Iowa, Iowa City, 82–105. Gil, David (1994c) “The Structure of Riau Indonesian”, Nordic Journal of Linguistics 17: 179–200. Gil, David (1995a) “Conjunctive Operators: A Cross-Linguistic Study”, IATL — Proceedings of the Tenth Annual Conference of The Israel Association for Theoretical Linguistics, 72–90. Gil, David (1995b) “Universal Quantifiers and Distributivity”, in E. Bach, E. Jelinek, A. Kratzer, and B. H. Partee eds., Quantification in Natural Languages, Kluwer Academic Publishers, Dordrecht, 321–362. Gil, David (1999) “Riau Indonesian as a Pivotless Language”, in E. V. Raxilina and Y. G. Testelets eds., Tipologija i Teorija Jazyka, Ot Opisanija k Objasneniju, K 60-Letiju Aleksandra Evgen’evicha Kibrika (Typology and Linguistic Theory, From Description to Explanation, For the 60th Birthday of Aleksandr E. Kibrik), Jazyki Russkoj Kul’tury, Moscow, 187–211. Gil, David (2000a) “Riau Indonesian: A VO Language with Internally-Headed Relative Clauses”, Snippets 1. Gil, David (2000b) “Syntactic Categories, Cross-Linguistic Variation and Universal Grammar”, in P. M. Vogel and B. Comrie eds., Approaches to the Typology of Word Classes, Empirical Approaches to Language Typology, Mouton, Berlin and New York, 173–216. Gil, David (2001a) “Creoles, Complexity and Riau Indonesian”, Linguistic Typology 5: 325–371. Gil, David (2001b) “Escaping Eurocentrism: Fieldwork as a Process of Unlearning”, in P. Newman and M. Ratliff eds., Linguistic Fieldwork, Cambridge University Press, Cambridge, 102–132. Gil, David (2001c) “Reflexive Anaphor or Conjunctive Operator? Riau Indonesian sendiri”, in P. Cole, C.-T. J. Huang and G. Hermon eds., Long Distance Reflexives, Syntax and Semantics 33, Academic Press, New York, 83–117. Gil, David (2002a) “Ludlings in Malayic Languages: An Introduction”, in Bambang Kaswanti Purwo ed., PELBBA 15, Pertemuan Linguistik Pusat Kajian Bahasa dan Budaya Atma Jaya: Kelima Belas, Unika Atma Jaya, Jakarta, 125–180. Gil, David (2002b) “Riau Indonesian -kan in Synchrony and Diachrony”, Paper presented at the Workshop on Benefactives, 9th International Conference on Austronesian Linguistics, The Australian National University, Canberra, Australia, 10 January 2002. Gil, David (2002c) “The Prefixes di- and N- in Malay/Indonesian Dialects”, in F. Wouk and M. Ross eds., The History and Typology of Western Austronesian Voice Systems, Pacific Linguistics, Canberra, 241–283.

422 David Gil

Gil, David (2003) “Intonation Does Not Differentiate Thematic Roles in Riau Indonesian”, in A. Riehl and T. Savella eds., Proceedings of the Ninth Annual Meeting of the Austronesian Formal Linguistics Association (AFLA9), Cornell Working Papers in Linguistics 19: 64–78. Gil, David (to appear a) “Conjunctions and Universal Quantifiers”, in M. Dryer, M. Haspelmath, D. Gil and B. Comrie eds., World Atlas of Language Structures, Oxford University Press, Oxford. Gil, David (to appear b) “Intonation and Thematic Roles in Riau Indonesian”, in M. Gordon, D. Büring and C. M. Lee eds., Topic, Focus and Intonation [tentative title], Kluwer, Deventer. Haiman, John ed. (1985a) Iconicity in Syntax, John Benjamins, Amsterdam. Haiman, John (1985b) Natural Syntax: Iconicity and Erosion, Cambridge University Press, Cambridge. Haspelmath, Martin (1995) “Diachronic sources of ‘all’ and ‘every’”, in E. Bach, E. Jelinek, A. Kratzer, and B. H. Partee eds., Quantification in Natural Languages, Kluwer Academic Publishers, Dordrecht, 363–382. Haspelmath, Martin (1997) Indefinite Pronouns, Oxford University Press, Oxford. Haspelmath, Martin (2003) “The Geometry of Grammatical Meaning: Semantic Maps and CrossLinguistic Comparison”, in M. Tomasello ed., The New Psychology of Language, Vol. 2, Erlbaum, Mahwah, 211–242. Haspelmath, Martin (to appear a) “Coordination”, in T. Shopen ed., Language Typology and Syntactic Description, Second edition, Cambridge University Press, Cambridge. Haspelmath, Martin (to appear b) “Nominal and Verbal Conjunction”, in M. Dryer, M. Haspelmath, D. Gil and B. Comrie eds., World Atlas of Language Structures, Oxford University Press, Oxford. Heine, Bernd (1997) Possession: Cognitive Sources, Forces and Grammaticalization, Cambridge University Press, Cambridge. Jakovlev, Nikolaj and D. Ashxamaf (1940) Grammatika adygejskogo liternaturnogo jazyka, Izdatel’stvo AN SSSR, Moscow. Jones, Russell (1984) “Loan Words in Contemporary Indonesian”, in J. W. M. Verhaar ed., Towards a Description of Contemporary Indonesian: Preliminary Studies, Part II, NUSA, Linguistic Studies of Indonesian and Other Languages in Indonesia, Volume 19, Universitas Katolik Indonesia Atma Jaya, Jakarta. Keenan, Edward L. (1978) “On Surface Form and Logical Form”, in B. B. Kachru ed., Linguistics in the Seventies: Directions and Prospects, University of Illinois, Champaign-Urbana, 163–203. Keenan, Edward L. and Leonard M. Faltz (1986) Boolean Semantics for Natural Language, Reidel, Dordrecht. Kemmer, Suzanne (1993) The Middle Voice, John Benjamins, Amsterdam. Kibrik, Aleksandr E. (1997) “Beyond Subject and Object: Towards a Comprehensive Relational Typology”, Linguistic Typology 1: 279–346. Kibrik, Aleksandr E. (2003) “Nominal Inflection Galore: Daghestanian, with Side Glances at Europe and the World”, in F. Plank ed., Noun Phrase Structure in the Languages of Europe, Empirical Approaches to Language Typology, Eurotyp 20–7, Mouton, Berlin and New York, 37–112. Kimball, Geoffrey D. (1991) Koasati Grammar, University of Nebraska Press, Lincoln. König, Ekkehard (1991) The Meaning of Focus Particles, A Comparative Perspective, Routledge, London and New York. Lakoff, George (1987) Women, Fire and Dangerous Things, University of Chicago Press, Chicago. Lakoff, George and Mark Johnson (1980) Metaphors We Live By, University of Chicago Press, Chicago. Lerdahl, Fred and Ray Jackendoff (1983) A Generative Theory of Tonal Music, MIT Press, Cambridge.

Riau Indonesian sama 423

Lichtenberk, Frantisek (2000a) “Inclusory Pronominals”, Oceanic Linguistics 39: 1–32. Lichtenberk, Frantisek (2000b) “Reciprocals without Reflexives”, in Z. Frajzyngier and T. S. Curl eds., Reciprocals, Forms and Functions, John Benjamins, Amsterdam and Philadelphia, 31–62. Maslova, Elena (2000) “Reciprocals and Set Construal”, in Z. Frajzyngier and T. S. Curl eds., Reciprocals, Forms and Functions, John Benjamins, Amsterdam and Philadelphia, 161–178. McGregor, William (1990) A Functional Grammar of Gooniyandi, John Benjamins, Amsterdam and Philadelphia. Michaelis, Susanne and M. Rosalie (2000) “Polysémie et cartes sémantiques: Le relateur (av)ek en Créole Seychellois”, Études Créoles 23: 79–100. Moravcsik, Edith (to appear) “A Semantic Analysis of Associative Plurals”, Studies in Language. Olawsky, Knut J. (1999) Aspects of Dagbani Grammar, With Special Emphasis on Phonology and Morphology, Lincom Europa, Munich. Rehg, Kenneth L. (1981) Ponapean Reference Grammar, University of Hawaii Press, Honolulu. Sanders, Gerald (1975) Invariant Ordering, Mouton, The Hague. Schwartz, Linda (1985) “Plural Pronouns, Coordination, and Inclusion”, in N. Stetson ed., Proceedings of the Tenth Annual Minnesota Conference on Language and Linguistics, University of Minnesota, Minneapolis, 152–184. Schwartz, Linda (1988) “Conditions for Verb-Coded Coordinations”, in M. Hammond, E. A. Moravcsik, and Jessica J. Wirth eds., Studies in Syntactic Typology, John Benjamins, Amsterdam and Philadelphia, 53–73. Sneddon, James Neil (2002a) “Diglossia in Indonesian”, Paper presented at the Sixth International Symposium on Malay/Indonesian Linguistics, Association for Linguistic Typology, Bintan Island, Indonesia, 4 August 2002. Sneddon, James Neil (2002b) “Variation in Informal Jakartan Indonesian, A Quantitative Study”, Paper presented at the 9th International Conference on Austronesian Linguistics, The Australian National University, Canberra, Australia, 10 January 2002. Stassen, Leon (2000) “AND-languages and WITH-languages”, Linguistic Typology 4: 1–54. Stassen, Leon (to appear a) “Noun Phrase Conjunction”, in M. Dryer, M. Haspelmath, D. Gil and B. Comrie eds., World Atlas of Language Structures, Oxford University Press, Oxford. Stassen, Leon (to appear b) “Predicative Possession”, in M. Dryer, M. Haspelmath, D. Gil and B. Comrie eds., World Atlas of Language Structures, Oxford University Press, Oxford. Stilo, Donald (this volume) “Coordination in Three Western Iranian Languages”. Stolz, Thomas (1996a) “Komitativ-Typologie, MIT- und OHNE-Relationen im crosslinguistischen Überblick”, Papiere zur Linguistik 54: 3–65. Stolz, Thomas (1996b) “Some Instrumentals Are Really Good Companions — Some Are Not, On Syncretism and the Typology of Instrumentals and Comitatives”, Theoretical Linguistics 23: 113–200. Stolz, Thomas, Cornelia Stroh and Aina Urdze (to appear) “Comitative and Instrumental”, in M. Dryer, M. Haspelmath, D. Gil and B. Comrie eds., World Atlas of Language Structures, Oxford University Press, Oxford. Tobin, Yishai (1990) Semiotics and Linguistics, Longman, London. Urtz, Bernadette (1994) The Syntax, Semantics and Pragmatics of a Nominal Conjunction — The Case of Russian “S”, PhD Dissertation, Harvard University, Cambridge. Van Valin, Robert D., Jr. (1999) “Generalized Semantic Roles and the Syntax-Semantics Interface”, in F. Corblin, C. Dobrovie-Sorin and J.-M. Marandin eds., Empirical Issues in Formal Syntax and Semantics 2, Thesus, The Hague, 373–389. Verma, Manindra K. and K. P. Mohanan eds., (1991) Experiencer Subjects in South Asian Languages, CSLI Publications, Stanford.

424 David Gil

Wälchli, Bernhard (1998) Co-compounds and Natural Coordination, Unpublished Paper, University of Stockholm, Stockholm. Wouk, Fay (1989) The Impact of Discourse on Grammar: Verb Morphology in Spoken Jakarta Indonesian. PhD Dissertation, UCLA, Los Angeles. Wouk, Fay (1999) “Dialect Contact and Koineization in Jakarta, Indonesia”, Language Sciences 2161–86. Zorc, R. David (1995) “A Glossary of Austronesian Reconstructions”, in D. T. Tryon ed., Comparative Austronesian Dictionary, An Introduction to Austronesian Studies, Part 1: Fascicle 2, Mouton, Berlin and New York, 1105–1197. Zuckermann, Ghil’ad (2000) Camouflaged Borrowing: ‘Folk-Etymological Nativization’ in the Service of Puristic Language Engineering, PhD dissertation, Oxford University, Oxford.


Chapter 15

Coordination in Lavukaleve* Angela Terrill Max Planck Institute for Psycholinguistics, Nijmegen

Introduction A brief glance at Lavukaleve Coordinators Structure of coordination 4.1 The coordinator 4.2 Number of coordinands 4.3 Ellipsis 5. Noun phrase coordination 6. Adjunct and adjective coordination 7. Coordination as one type of clause linkage in Lavukaleve 7.1 Independent coordinate constructions 7.2 More common types of clause linkage which are not coordination 7.3 Coordinated clauses within clause chains? 1. 2. 3. 4.



Coordination is one of a number of syntactic strategies available in Lavukaleve to combine various types of constituents. Coordination can be used to join noun phrases, nouns, postpositional phrases and other adjuncts and clauses. Coordination is generally achieved by means of coordinators. However, these are used mainly for joining noun phrases rather than other constituent types. It is certainly possible to coordinate clauses with or without overt coordinators, but far more common than coordination are other types of clause linkage, including subordination or cosubordination.

*I would like to express my gratitude to the following people for providing the data cited in this paper: Patteson Barua, Emily Fofokolin, Kilaev Gulum, John Hubert, Dora Iho, Sarah Kiko, Raymond Kolo, Georgina Makikifaria, Abel Moran, Mostin Nekuiga, James Nepolo, Wilson Ngara, Matthew Ngele, Donald Ofi, Clement Oiva, Melchior Pitu, Nicholas Sobo, Laurence Soitam, Janet Tokilo, Mase Toto and Edwin Vefele.

428 Angela Terrill

This paper begins with a short typological outline of Lavukaleve. The discussion of coordination begins with a description of what coordinators are available. Following this is an explanation of noun phrase coordination, then clausal and sentential coordination. The discussion then moves on to the issue of distinguishing coordination from cosubordination and serialisation, all of which also occur in Lavukaleve.


A brief glance at Lavukaleve

Lavukaleve is a Papuan isolate language spoken by about 1700 people in the Russell Islands, a small island group within the Solomon Islands. Constituent order is SOV, with head-marking in possessive phrases, postpositional phrases and verbs, but dependent-marking in adjectives and dependent clauses. Lavukaleve has three genders, marked on nominal dependents, postpositions and on some verbal morphology. Verbs mark their arguments affixally, by means of subject and object prefixes, and there is also a suffix that shows agreement in gender and number with a certain argument, the role of which depends on the construction in which the verb appears. Verbs (which can consist of simple verb stems or complexes including serial verb constructions, verb compounds and phrases of verb stem plus habitual auxiliary) can also mark by suffixation categories of tense, aspect and mood, as well as various categories of subordination, among other things. Serial verb constructions occur, and there are two main types of dependent clauses: cosubordinate clauses, occurring in clause chaining constructions, and subordinate clauses, including relative clauses, purposive clauses and three further subtypes of adverbial clauses. Focus is marked morphosyntactically, with focus markers using their inherent agreement properties to show the scope of the focus. There are three separate focus markers, which differ in terms of the sentence type with which they occur. More information on the grammar of the language can be found in Terrill (1999, 2003). The data on which this paper is based is a large corpus of recorded spontaneous narrative, supplemented by a small amount of elicited sentences.



There is a small number of coordinators (in Haspelmath’s (to appear) terms), as outlined in Table 1. Clauses (including full sentences) and noun phrases can be coordinated readily; postpositional phrases too can be coordinated readily, if only very rarely, but other constituent types are not coordinated. This is in contradiction to Payne’s (1985) implicational hierarchy of coordination types. The hierarchy is as follows:

Coordination in Lavukaleve 429

Table 1.Coordinators in Lavukaleve Coordinator

Joins clauses

aka ‘then, next, so, etc.’ hano ‘then’ taman ‘however, but’ leta ‘but’ olang ‘because’ o ‘and, or’ ve ‘or’ ne ‘with’ (formally identical to postposition ‘with’) no coordinator (=juxtaposition)

yes yes yes yes yes yes yes

Joins NPs, nouns Joins other constituents

yes yes yes

yes (adjuncts)


S-VP-AP-PP-NP, which is to be read that if the positions on the left are possible, the more rightward positions are also possible. Adjectival coordination does not occur in Lavukaleve; this is discussed further below (§ 6). To begin with, the next three examples show noun phrase coordination, clausal coordination and adjunct coordination respectively. The coordinators are highlighted in boldface throughout: (1) Lo-lav o lo-kilekile vo-ririgoi-ri-ne 3du.poss-bamboo(n) and 3du.poss-axes(pl) 3pl.O-prepare-caus-impf sia-re mola ga e-ke-vau-ri-re, … do-nf canoe(n) ‘Then the two prepared their bamboo and their axes and pushed the canoe out to sea, …’ (2) … ae lavea-vel ta aka roge e-o-kuru. go.up appear-compl just then one.sgn 3sgn.O-3sg.S-kill ‘… she comes up between two canoes, and she kills one.’ (3) Foiga e-na fi me-sa raul tuna pn.ntrl.sgn 3sgn.O-in 3sgn.foc really be.really kini me-nua tasi-n o fa*gi ga e-na. act sea-loc and island 3sgn.O-in ‘That’s why we all went amiss in the sea and on the island.’

430 Angela Terrill

4. Structure of coordination 4.1 The coordinator Where an overt coordinator occurs, the structure of coordination is always A co B (where co represents the coordinator). That is, the coordinator comes in between the coordinated items. The coordinator cannot easily be said to belong either to the first element coordinated or the second. Speakers can pause either before or after the coordinator. In the corpus, pauses before coordinators are roughly equal in frequency to pauses after the coordinator. For example (slashes in the next examples indicate pauses on level pitch): (4) Aka sia-re, lalo / o logo-logoa-e tin. then do-nf peace(n) and redup-be.happy-nomzr only ‘Because of that, [there was] only peace and happiness.’ (5) A-ham leo o kegu / o mina la 3sgm.O-for betel.nut(n) and leaf.for.betel.nut(f) and thing(f) o-ririgoi-ri-re, le’lenga / o fo’sal vo-ma-re 3sgf.O-prepare-caus-nf pudding(f) and fish(m) 3pl.O-take-nf o-hau. 3sg.S-go.shorewards ‘He prepared betel nut, leaf and everything for him, pudding and fish, and took them shorewards.’ (6) Kilikil na o / kilimarea la lako-v. bird.species(m) and white.seagull(f) cry-pl ‘The kilikil bird and the kilimarea bird cried.’ (7) Sa’sau o / mina man hin / hounio o/ armring(n) and thing(f) what(m) 3sgm.efoc necklace(f) and mina-mina akari ho’bea o-tin. redup.thing(f) then-psnv good-sgf 3sg.poss-only ‘They took arm-rings and whatever, necklaces, and anything that looked good.’ A rare example of a discontinuous phrase order suggests that at least in some cases the coordinator belongs to the first coordinand: (8) Tasi ga aka-ri sia le-me-ge foina sea(n) then-psnv do 3sgn.O.sbd-hab-ant pn.ntrl.sgm a-na ao-re tulav o homela-v o mina la. 3sgm.O-in children(pl) and woman-pl and thing(f) Fina o. belongings(m) and

Coordination in Lavukaleve 431

‘The sea was doing like that, and the children and women and everything went on board. And the bedding.’ Coordination can also be achieved by juxtaposition of the coordinands, with no coordinator. This is only possible for two coordinands, no more. It typically, although not obligatorily, occurs with semantically expectable pairings of items (9) or with a pronoun and a noun phrase in a relationship of accompaniment (10): (9) Nga-bakala nga-uia tula vo-ma-re, 1sg.poss-paddle(m) 1sg.poss-knife(f) small.sgf 3pl.O-take-nf a-vau. 1sg.S-go.seawards ‘I take my paddle and my small knife and go seawards.’ (10) I ngai Mima lava e-ko-ham ve lei um 1sg Mima(m) bamboo(n) 3sgn.O-throw-purp go exist fi filange-n. 3sg.n.foc weathercoast-loc ‘Um, Mima and I went to fish on the weathercoast.’ In (11) is an example of coordination by juxtaposition in which the two elements coordinated are postpositional phrases: (11) aram e-na keker e-na ground(n) 3sgn.O-in 3sgn.O-in ‘with land and reef ’ 4.2 Number of coordinands If a coordinator is used, it is rare to have more than two coordinands, and the maximum number found is four coordinands: (12) Belepui-verav o mama-kal o mavitu fela’koe catechist-pl and father-pl and congregation(pl) village(n) hoga e-na me-v nga-ma-lagari. mod.prox.sgn 3sgn.O-in spec-pl 1sg.O-3pl.S-choose ‘The catechists and priests and congregation in this village chose me.’ (13) Houl o kukulia-vil o tagio-vil o mina tu-tula trees(pl) and frog-pl and snake-pl and thing(f) redup-small kini vau-v. act go.seawards-pl ‘Trees and frogs and snakes and every little thing went seawards.’

432 Angela Terrill

In the case of coordination by juxtaposition, only two coordinands are possible. If there are more than two coordinands, the first two can be juxtaposed, while the third uses a coordinator: (14) … o-tau o-tin o mina la fele-re 3sg.poss-legs(n) 3sg.poss-body(n) and thing(f) return-nf o-e-laketei-ge ta aka… 3sgf.O-sbd-live-ant just then ‘… his legs, his body and everything came back to life…’ 4.3 Ellipsis Ellipsis does occur in noun phrase coordination; see for example (30) below. However, ellipsis in postpositional phrases and clauses is a more complex matter. In postpositional phrases, it is often difficult to tell if ellipsis has occurred, or if it is rather a case of coordination of two or more NPs, within a PP. For example: (15) Mina bilial o kovekukuam-aul o thing(f) and puddle-pl and loi-verav vo-na fi ngoa lo-me. 3pl.O-in 3sgn.foc stay 3sg.S-hab ‘It lived in things like pools of water in rocks, and puddles, and pools of water inside logs.’ (16) Solo-al o kofitaol vo-na foinala lioi-re mountain-pl and valleys(pl) 3pl.O-in those.mdu run-nf kelea me-mal. walk hab-du.m ‘Over mountains and valleys the two ran.’ For such postpositional phrases, it is not possible to tell if the bracketing is [[NP o NP] Postp]PP or [[NP Ø-Postp]PP o [NP Postp]PP]. However, in some cases it is possible to infer the correct structure from the number agreement of the postposition. Thus: (17) O-vulita o ki’kile vo-na o-fai 3sg.poss-shield(f) and axe(f) 3pl.O-in 3sg.poss-chin(m) a-kae-re vo-na iutia-re-ne. 3sgm.O-put.up-nf 3pl.O-in look.on-nf-impf ‘Putting his chin on his shield and axe, he [just] looked on at them.’

Coordination in Lavukaleve 433

(18) Leta aka oinala lo-toka-e o but then other.ntrl.mdu 3du.S-be.strong-nomzr and lo-laura-e vo-na fi, kini mima 3du.S-great-nomzr 3pl.O-in 3sgn.foc act hoga o-toka ke, foiga malav va mod.prox.sgn 3sg.S-be.strong emph pn.ntrl.sgn people(pl) kini hau ma-hoi. act go.shorewards ‘But through the strength and the power of those two, the church became strong, and people came in.’ In these examples, the coordinands are both singular, but the postpositions are plural. This shows that the postposition is agreeing with both constituents together, that is, the agreement is additive; thus, the bracketing here must be [[NP o NP] Postp]PP. If it was otherwise, the postposition would be in construction with only the second element, and thus should have singular agreement. (Note, incidentally, that additive agreement of two coordinands is marked as plural, not as dual, as one would expect.) Therefore, in those examples in which one can tell what the syntactic structure of the coordination is, the structure is that of coordinated NPs acting as the single argument of a postposition. Thus, it seems that ellipsis of material in coordinated PPs does not occur. Ellipsis does however occur in coordinated clauses. Nominal ellipsis is difficult to argue for, because while there is frequent NP ellipsis in non-coordinated structures, the presence or absence of affixal participant marking on predicates is dependent on a multitude of other morphosyntactic variables; in general, participant-marking affixes cannot be just left off. Therefore ellipsis of nominal arguments of clauses is not particularly significant. However, ellipsis of clausal material does occur in clausal coordination. Such constructions do not typically arise spontaneously; the following examples are elicited: (19) Ngai sa vo-liki-re a-lei o koi paenapol. 1sg banana(pl) 3pl.O-want-nf 1sg.S-exist and also pineapple(f) ‘I want bananas and pineapples.’ (20) Ngai gonu roa fin a-le-m o koi navula 1sg turtle(m) one.sgm 3sgm.foc 1sg.S-see-sgm and also whale(f) ro. one.sgf ‘I saw a turtle and a whale.’ These are both examples of forward ellipsis, or analipsis.

434 Angela Terrill


Noun phrase coordination

Following are some examples of noun phrase coordination using various overt coordinators: (21) Nga-kala o nga-ne me-v vo-ne ngoa-ne, 1sg.poss-mother(f) and 1sg.O-with spec-pl 3pl.O-with stay-impf o fo’foira o-i-ne a-ngoa. and work(f) 3sgf.O-do-impf 1sg.S-stay ‘I lived with my mother and siblings [lit: the ones with me], and stayed doing work.’ (22) O-le-re lo-re, kini o-holovem ve mina la 3sgf.O-see-nf finish-nf act 3sg.poss-devil(m) or thing(f) o-ne va’var o-hai. 3sgf.O-with talking 3sg.S-do ‘Having seen that, he saw a devil or something and talked with it.’ (23) Aka taragau na ne ruia la ni’kol then sea.eagle(m) with old.woman(f) first house tail fi lo-volori-ø hi. house(n) 3sgn.foc 3du.S-make-sgn 3sgn ‘So the sea eagle and the old woman built a house first.’ There is no special means of expressing neither…nor, but if one of the elements to be coordinated is a negated noun phrase, the meaning of the coordination construction is to be understood as negative. The following, literally the houses and nothing, is understood as neither the houses nor anything else. Note that negation is also marked on the verb; any NP containing roru ‘none’ must also carry a clausal negator, in this case the negative suffix on the verb. (24) Tail-av o mina ro-ru sikala-la. house-pl and thing(f) one.sgf-none be.bad-neg.sgf ‘Neither the houses nor anything was destroyed.’ Inclusory coordination is common. In this construction type, a non-singular pronoun is used to make reference to two or more participants who are also separately mentioned. The non-singular pronoun thus ‘includes’ the reference of both or all of the other participants. This is the typical way of constructing expressions such as ‘my husband and I’: (25) Aka el ngai nga-tua el le-sangiria. then 1du.ex 1sg 1sg.poss-wife(f) 1du.ex 1du.ex-mix ‘Then us two, my wife and I, came together.’

Coordination in Lavukaleve 435

(26) El Mima lava e-ko-ham ve lei. 1du.ex Mima(m) bamboo(n) 3sgn.O-throw-purp go exist ‘Us two, Mima and I, went to go rod fishing.’ (27) e-tairi-re, foiga e-u. Kanege e-sa. 3sgn.O-divide-nf pn.ntrl.sgn 3sgn.O-eat family 1pl.ex-group ‘… dividing it up, [we] eat it. My family and I.’

6. Adjunct and adjective coordination Adjuncts can readily be coordinated, although in practice it is rather uncommon that such constructions occur spontaneously. The term adjunct is used here in its most usual sense, to refer to optional clause-level modifiers. There are three types of adjuncts which noun phrases can occur in: postpositional phrases, adjuncts with the locative suffix, and bare noun phrase adjuncts. Whether a noun takes a postposition, the locative suffix, or appears as a bare adjunct, is lexically-determined. In any case, all of these adjunct types can be coordinated, typically with o ‘and’: (28) Fela’koe-n man man sia mem na, a-na koi village-loc what(m) what(m) do hab-sgm 3sgm.O-in also vo-ne sangine fo’foira o-i-ne a-ngoa: 3pl.O-with together work(f) 3sgf.O-do-impf 1sg.S-stay mima o-rara e-na, o fela’koe o-laketei church(m) 3sg.poss-side(n) 3sgn.O-in and village(n) 3sg.poss-life(n) e-na. 3sgn.O-in ‘Anything that happened in the village, I involved myself in: on the church side, and in village life.’ (29) O’ase-n, o tasi-n fi homela-v fo’foira o-i bush-loc and sea-loc 3sgn.foc woman-pl work(f) 3sgf.O-do ma-me. 3pl.S-hab ‘In the bush and in the sea women are doing their work.’ In Lavukaleve it is not possible to coordinate adjectives. This is in part because adjectives are a small word class, and for the most part modification is expressed by means of intransitive verbs or nouns rather than adjectives. For example, in (30) the modifiers ‘small’ and ‘thin’ are expressed as modifiers in noun phrases ‘small thing’ and ‘thin (thing)’ respectively, and it is the noun phrases which are coordinated:

436 Angela Terrill

(30) Ngai nga-vaisa la mina tula-a (o) 1sg 1sg.poss-younger.sister(f) thing(f) small-sgf and kasakasa feo. thin.sgf 3sgf.foc ‘My sister is small [lit: a small thing] and thin.’ Similarly in (31), the adjective korovisa ‘black’ is actually a headless noun phrase coordinated with the noun phrase consisting of the sole element kelekele, a noun meaning ‘white one’: (31) Korovisa o kelekele aka-ri. black.sgn and then-psnv ‘It (n) was black and white like that [lit: It was a black one and white one].’ In (32), the modifying expressions ‘alive’ and ‘dead’ are both expressed as intransitive verbs: (32) ika-la fi latena-v o kiu-v vo-lu-lu-ri-ham there-ext 3sgn.foc be.alive-pl and die-pl 3pl.O-redup-sort.out-purp hini o-vo-re. int 3sg.S-come-fut ‘… from there he will come to judge the living and the dead.’ (Barua n.d.)


Coordination as one type of clause linkage in Lavukaleve

Clauses in Lavukaleve divide into two major types: independent (main) and dependent. Dependent clauses themselves are of two types: subordinate and cosubordinate (Van Valin and LaPolla’s (1997) term). The following diagram outlines types of clauses in Lavukaleve. Lavukaleve clause types independent


subordinate adverbial


Figure 1.Clause types in Lavukaleve

cosubordinate relative

Coordination in Lavukaleve 437

In Lavukaleve there are clear morpho-syntactic distinctions between each of these clause types, and it is descriptively useful to consider them each separately. 7.1 Independent coordinate constructions Two or more independent clauses or sentences can be joined by the use of a coordinator. All the clauses joined using coordinators are independent; that is, they can freely stand on their own as complete utterances. The two most frequently used coordinators are aka ‘then, next, so, etc.’ and hano ‘then’. When hano is used to join two sentences, it is the last element of the first sentence (i.e. hano is a postpositive coordinator). However, when aka is used to join two sentences, it is the first element of the second sentence (i.e. aka is a prepositive coordinator). Lavukaleve has verb-final constituent order. The expectation, according to Haspelmath’s (to appear) predictions, would be that coordinators would be postpositive like hano, not prepositive like aka. However, aka is much more common than hano. Note that hano always occurs with a sentence-final drop in pitch, but many of the units which hano joins are actually not full sentences but cosubordinate or subordinate clauses, followed, after hano, by an independent clause. Coordinators are by no means obligatory, or even very common, in joining these clause types. They do not usually occur between cosubordinate or subordinate clauses, but as some of these examples show, the possibility exists. This is particularly the case when there is a pause between clauses; in these circ*mstances the coordinator can be seen as a filler. In the following examples, each numbered example is a single chunk of discourse: (33) i.

A-vala-re a-ma-re, kini vea-re, hano… 3sgm.O-pull-nf 3sgm.O-take-nf act emerge-nf then ‘Pulling him out, he emerged, then…’ ii. ikaika a-fo’foiri-nun hano… each 3sgm.O-pummell-dur then ‘they beat him up, then…’ iii. a-kuru-re ma-u-m. 3sgm.O-hit-nf 3pl.S-eat-sgm ‘killing him they ate him.’

(34) i.

O-hului-ri-ge ta (kalem ne kala!) 3sgf.O.3sg.S-go.round-caus-ant just (father(m) with mother(f) nato la e’rau oiga sago.palm(f) fall/jump other.ntrl.sgn ma-foto-n e’rau oiga ta hano… 3pl.poss-middle-loc fall/jump other.ntrl.sgn just then

438 Angela Terrill

‘Turning it around, oh my goodness!, the sago tree falls down in the middle of them all, then…’ ii. vo-kuru-ge kiu-re lo-v. 3pl.O.3sg.S-hit-ant die-nf finish-pl ‘it hitting them, they all died.’ (35) i.

Feu koi “Nam lake, tutu-m” go.inland also “ fire(n) grandparent-sgm a-e-re-ge hano… 3sgm.O-sbd-say-ant then ‘He went inland again. “Give me some fire, old man” he said, then…’ ii. o-laku-n e-ne-ne koi! 3sg.poss-hate-loc 3sgn.O-give-impf also ‘Once again he didn’t want to give it!’

Aka ‘then, next’ is the first element of the second clause: (36) Malaria la o-na fo’foira o-i-nun malaria(f) 3sgf.O-in work(f) 3sgf.O-do-dur o-i-nun o-i-nun o-i-nun 3sgf.O-do-dur 3sgf.O-do-dur 3sgf.O-do-dur fi aka ika-ri o-na fo’foira 3sgn.foc then there-psnv 3sgf.O-in work(f) o-i-re lo-re, aka koi ika-ri nun 3sgf.O-do-nf finish-nf then also there-psnv from lo a-i-ø ga e-na fi, kini koi finish 1sg.S-do-nomzr 3sgn.O-in 3sgn.foc act also Red Cross e-na fo’foira o-a-i. Red.Cross(n) 3sgn.O-in work(f) 3sgf.O-1sg.S-do ‘I continued working on malaria, then having finished working with malaria, then after I had worked there, I went and worked at the Red Cross.’ Aka also joins independent sentences: (37) i.

O-ae foiga. 3sg.S-go.up pn.ntrl.sgn ‘He climbed up [the tree].’ ii. Aka a-o-lai “Ae ngo-me-le ae then 3sgm.O-3sg.S-tell “go.up 2sg-hab-pot go.up ngo-me-le ae mina rara hoga e-hamail 2sg-hab-pot go.up um side(n) mod.prox.sgn 3sgn.O-facing mina matua feo ke.” um old.coconut(f) 3sgf.foc emph

Coordination in Lavukaleve 439

‘Then she told him: You go up, way up there, on this side, is a dry coconut.”’ (38) i.

“Aka kini houl vo-koroi-va!” hide ma-re ke. “then act trees(pl) thus 3pl.S-say emph ‘“(This is the tree we need!) So go and chop sticks!” they said.’ ii. Aka oiva airal lelemal houla airal lelemal then men.du(m) two(m) tree(f) men.du(m) two(m) houla airal lelemal houla oina o-koroi tree(f) men.du(m) two(m) tree(f) other.ntrl.sgm 3sgf.O-chop mem na o-koroi-ne fi ngoa lo-me hab-sgm 3sgf.O-chop-impf 3sgn.foc stay 3sg.S-hab nato la. sago.palm(f) ‘Then they were two men to a tree, two men to a tree, two men to a tree, one man was chopping away at the sago tree.’ iii. Aka kini hano fale-v. then act then stand-pl ‘Then they all stood up.’

While aka ‘then, next, so, etc.’ and hano ‘then’ are the main clausal and sentential coordinators, other coordinators can also be used, for instance o ‘and’ (39), taman ‘however, but’ (40) and olang ‘because’ (41): (39) Oia vau mo’sil savu-n fale me-a o other.ntrl.sgf go.seawards tide edge-loc stand hab-sgf and oina kini hau e’rau-m other.ntrl.sgm act go.shorewards fall/jump-sgm o-tum na. 3sg.poss-husband(m) ‘She went out to the sea shore, she was standing up, and he [the husband] came ashore.’ (40) Oina sou fale-re kini a-e-ve-meon taman other.ntrl.sgm rise stand-nf act 3sgm.O-sbd-go-surp but lake ga e-nua-ri-re… road(n) 3sgn.O-be.amiss-caus-nf ‘He stood up, but he took the wrong path…’ (41) Lugua-re fi ma-lei koi olang ruimal be.upset-nf 3sgn.foc 3pl.S-exist also because old.people.du(m) nala koi lo-vo’vou na olang lo-vo’vou also 3du.poss-boy(m) because 3du.poss-boy(m)

440 Angela Terrill

na a-e-kiu-ge… 3sgm.O-sbd-die-ant ‘Everyone was terribly upset also because the old couple’s child — because their child was dead…’ These structures notwithstanding, independent coordinate constructions are not common in Lavukaleve. It is certainly not the preferred way of joining clauses. Instead, speakers prefer to join clauses using clause-chaining constructions or subordinate constructions, as described in the next subsection. 7.2 More common types of clause linkage which are not coordination Much more common than linking clauses by coordination is using a subordinate construction. In Lavukaleve, verbs of subordinate clauses cross-reference their own arguments, but they can only specify other verbal properties (such as tense marking) in a very restricted way; essentially, they are only marked for their semantic relationship to their main predicate. Subordinate clauses are embedded within a main clause; they function as modifiers of particular parts of the main clause. Subordinate adverbial clauses, the type of subordinate clauses which typically express notions of a succession of events, have a split-ergative participant marking system in which first and second person subjects employ a nominative-accusative marking system (as in other clause types in the language) whereas third person subjects employ an ergative-absolutive marking system, unique to this type of subordinate clause. In the following example, there are two subordinate anterior clauses (with predicates esiage ‘doing’ and omafouge ‘putting’), dependent on main verb ove ‘he went’: (42) Malav va aunion e-sia-ge, ma-ruta people(pl) evening-loc 3sgn.O.sbd-do-ant 3pl.poss-lamp(f) la koa-n o-ma-fou-ge ta, taragau na door-loc 3sgf.O-3pl.S-put.on-ant just sea.eagle(m) hano o-ve ruta la o-ma-ham. then 3sg.S-go lamp(f) 3sgf.O-take-purp ‘Upon evening coming, upon the people putting out their light in the doorway, the sea eagle went to take the light.’ For ‘and’-type linkage, cosubordinate linkage is used. This type of clause linkage, a characteristic of many Papuan languages, is often known as clause chaining. It involves one or more cosubordinate verbs (usually referred to as ‘medial’) (in examples (43) and (44) they are marked with the Non-Finite suffix -re) and a final verb. Cosubordinate clauses are syntactically and semantically dependent on a main clause, but are not syntactically or semantically a part of that clause. Each cosubordinate

Coordination in Lavukaleve 441

clause is linked to the next cosubordinate clause “in a linear string … much like beads on a necklace” (Foley 1986: 194), and each one is dependent on the main clause. In Lavukaleve, verbs of cosubordinate clauses cannot cross-reference their subject arguments. Also, they cannot mark their own tense, aspect and mood; they rely on the main clause for this. However, they do carry morphology specifying their temporal relationship with the main clause. (43) O-ma-re, ma-taila-n vo-re ui 3sgf.O-take-nf 3pl.poss-house-loc come-nf food(n) e-u-re lo-re, ma-losi-v 3sgn.O-eat-nf finish-nf 3pl.poss-basket-pl vo-na mina o-ma-ho. 3pl.O-in thing(f) 3sgf.O-3pl.S-put.inside ‘Taking [everything], they came to their houses, and having finished eating food, everyone put something in their food baskets.’ (44) Kini daeva sia-re taalea la act goggles do-nf small.giant.clam(f) o-le-re o-a-te. 3sgf.O-see-nf 3sgf.O-1sg.S-get.shellfish ‘I put on my goggles and see the shellfish and collect it. [lit: Putting on my goggles, seeing the taalea, I collect it.]’ A further very frequent way of joining predicates, and thus another reason not to use coordination, is serial verb constructions: (45) Nei hoga-ri e-ma vau. coconut(n) mod.prox.sgn-psnv 3sgn.O-take go.seawards ‘Take this coconut seawards.’ (46) Ika-ri ta a-la a-lo-ri-re. there-psnv just 3sgm.O-cut 3sgm.O-finish-caus-fut ‘There I will cut it and (thus) finish it.’ 7.3 Coordinated clauses within clause chains? Haspelmath (1995, to appear) discusses the difference between coordination and subordination, but something often not discussed is the difference between coordination and cosubordination. It was stated above that clause chains consist of one or more cosubordinate clauses followed by one final clause. The relationship between the cosubordinate clauses and the final clause is a relationship of coordinate-dependency. But what of the relationship between the cosubordinate clauses themselves? What is their syntactic relationship to each other?

442 Angela Terrill

The cosubordinate clauses themselves follow each other sequentially. They receive the same marking as each other; they are mutually substitutable for each other. Their relationship to the final verb is indicated in Lavukaleve by the NonFinite suffix -re, or Completive -vel or Successive -vele. It is not the case that the use of one of these suffixes links the cosubordinate clauses to each other, because it occurs even if there is only one cosubordinate clause: Haspelmath (1995: 21) makes this suggestion when he says “each medial verb depends on the verb that follows it immediately”. This is not the case at least in Lavukaleve: rather, each medial verb depends only on the final verb. Instead, the role of the suffix on a cosubordinate verb in Lavukaleve is to link the cosubordinate clause to the final clause. There is no overt marker to link the cosubordinate clauses to each other. However, it is perhaps useful to consider that the cosubordinate clauses bear a coordinate relationship to each other, and they are linked by no overt morpheme but rather by juxtaposition.

Abbreviations act ant art compl du durimp efoc emph ex ext f foc group hab in int m mod

action particle anterior article completive dual durative imperative focus marker from heo paradigm emphatic exclusive extended feminine focus marker from feo paradigm referent forms a group habitual inclusive intention masculine demonstrative modifier

n nf nomzr NP ntrl O pctimp pl pn pot PP psnv redup S sbd sg spec surp

neuter non-finite nominaliser noun phrase distance-neutral demonstrative object punctual imperative plural demonstrative pronoun potential postpositional phrase Presentative reduplicated subject (both transitive and intransitive) subordinate singular specifier surprise

Coordination in Lavukaleve 443

References Barua, Patteson. n.d. [Lavukaleve translation of A Melanesian English Prayer book with hymns. 1965. Honiara: Church of Melanesia Provincial Press]. Ms, Mane Village, Russell Islands. Foley, William A. 1986. The Papuan languages of New Guinea. Cambridge: Cambridge University Press. Haspelmath, Martin. 1995. “The converb as a cross-linguistically valid category.” In Martin Haspelmath and Ekkehard König (eds.), Converbs in cross-linguistic perspective. Berlin: Mouton de Gruyter, 1–55. Haspelmath, Martin. to appear. “Coordination.” In Timothy Shopen (ed.), Language typology and syntactic description. Cambridge: Cambridge University Press. [second edition] Payne, John R. 1985. “Complex phrases and complex sentences.” In Timothy Shopen (ed.), Language typology and linguistic description 2. Cambridge: Cambridge University Press, 3–41. Terrill, Angela. 1999. Lavukaleve: a Papuan language of the Solomon Islands. Ph.D. thesis, Department of Linguistics, Research School of Pacific and Asian Studies, Australian National University, Canberra, Australia. Terrill, Angela. 2003. A grammar of Lavukaleve. Berlin: Mouton de Gruyter. Van Valin, Robert D. and Randy J. LaPolla. 1997. Syntax: Structure, meaning and function. Cambridge: Cambridge University Press.

Chapter 16

Coordination in Oceanic languages and Proto Oceanic* Claire Moyse-Faurie and John Lynch LACITO-CNRS, Paris / University of the South Pacific

1. Introduction 1.1 Typological preliminaries 1.2 Existing reconstructions 2. Nominal coordination 2.1 Tight and loose coordination 2.2 Common NPs vs proper NPs coordination 3. Verb phrase and clausal coordination 3.1 Phrasal (NP/VP) and clausal coordination marked differently 3.2 Phrasal and clausal coordination marked identically 3.3 Nominal coordination differs from verbal and clausal coordination 4. Numeral and end-of-list coordination 5. Coordination and comitative 6. Adverbial functions of coordinators 7. Discussion 7.1 Towards a typology of Oceanic coordination 7.2 The coordinator-to-TAM-marker shift in Western Oceanic 7.3 Proto Oceanic Coordinators 8. Conclusion

*An earlier version of this paper was presented at the Fifth International Conference on Oceanic Linguistics, held in Canberra, Australia, in January 2002. We are grateful to a number of colleagues who provided data, answered queries, and/or commented on versions of this paper: Niko Besnier, Bob Blust, Isabelle Bril, Terry Crowley, Robert Early, Alexandre François, Paul Geraghty, Even Hovdhaugen, Jean-Pierre Nirua, Françoise Ozanne-Rivierre, Jean-Claude Rivierre, Malcolm Ross, Hans Schmidt, Apolonia Tamata, Nicholas Thieberger and Jacques Vernaudon.

446 Claire Moyse-Faurie and John Lynch



This paper investigates coordination strategies in Oceanic languages from a typological, comparative and historical perspective. We aim to show that the primary distinction in Proto Oceanic seems to have been between coordination of items (whether nominal or verbal) at the phrase level as opposed to coordination of propositions at the clause level. A secondary distinction is also made in several Oceanic languages — and thus possibly in Proto Oceanic — between tight or loose coordination of nominal items or events. Following a typological analysis of Oceanic coordination strategies, we also attempt to reconstruct the Proto Oceanic coordinating morphemes and their functions.1 The term “Oceanic” refers to a large subgroup of the Austronesian family, whose members are spoken in all of Polynesia, most of Micronesia and parts of Melanesia. The general consensus among linguists today is that the Oceanic family has three major subgroups (see also Bril, this volume): i.

an Admiralties subgroup: languages of Manus Island and the Admiralty Islands, possibly also including Yapese; ii. a Western Oceanic subgroup, including the North New Guinea grouping, the Meso-Melanesian grouping (the islands off New Guinea plus the western Solomon Islands), and the Papuan Tip grouping (the southeast tip of New Guinea); and iii. a Central-Eastern Oceanic grouping, comprising five subgroups: a. Southeast Solomons; b. Utupua-Vanikoro (in the extreme Southeast of Solomon Islands); c. Southern Oceanic (North-Central Vanuatu, Southern Vanuatu, and New Caledonia); d. Micronesian; and e. Central Pacific (Fiji, Rotuma and Polynesia). 1.1 Typological preliminaries A thorough analysis of coordination in Southern Melanesian (Southern Vanuatu and New Caledonian) and Polynesian (the two Oceanic subgroups with which we happen to be most familiar), and a briefer analysis of coordination in a number of

1.Orthography and interlinear glosses generally follow those of the published sources, though we have partly standardised the latter. For Proto Austronesian orthography we follow Blust (in prep.) and for Proto Oceanic Ross (1988, inter alia). An equals sign indicates a cl*tic boundary. Data sources, genetic affiliations of languages, and morpheme-gloss abbreviations are given in the appendices.

Coordination in Oceanic languages and Proto Oceanic 447

other Oceanic languages, lead to the conclusion that the coordinative morphemes in Oceanic languages are not only used to coordinate NPs and VPs (either tightly or loosely) or clauses, but appear to cover other related semantic notions, all clustering around an “additive” perspective: – – – – – – –

coordination of NPs (and we need to further distinguish common and proper NPs); expression of the comitative (addition of participants) or, less often, the instrumental; coordination of VPs; additive focus particle ‘also’ (see, e.g., König 1991); coordination of clauses, formally more diverse; addition in numeration (‘plus’), such as is manifested in the expression of intermediate numerals linking units to higher numerals; “endless addition”: coordination marking non-restrictiveness or non-exhaustiveness at the end of a list or enumeration (‘etc.’, ‘and so on’, ‘and company’).

In fact, it is impossible, in many Oceanic languages, to isolate certain markers of coordination (in particular, those linking VPs and NPs) from, on the one hand, comitative markers, and on the other hand, “additive” markers which are used in numerals, ends of enumerations or the expression of adverbs like ‘also’. One thus has to deal with a whole constellation of values, translated, according to the language, i. by a single morpheme for the whole constellation (e.g. Iaai, Samoan); or ii. by several morphemes, each marking certain functions, but differently organised from language to language; e.g.: a. one morpheme coordinating NPs and VPs, also marking comitative and end of enumeration, differing from another morpheme used in numerals and marking the additive focus particle ‘also’ (e.g. Nêlêmwa); as opposed to b. one morpheme marking loose NP or VP coordination, used in numerals and at the end of enumerations, differing from another marking tight NP and VP coordination and the adverb ‘also’ (e.g. Xârâcùù); as opposed to c. one morpheme coordinating NPs and VPs, and marking comitative and the adverb ‘also’, differing from another morpheme used in numerals, also differing from a third morpheme used to coordinate clauses (e.g. East Futunan, Tongan); and so on. In discussing coordination in a range of Oceanic languages, subscript numerals will indicate ordering relationships between noun phrases, verbs, verb phrases and clauses (e.g. V1 comes before V2). We will refer to the following broad syntactic types:

448 Claire Moyse-Faurie and John Lynch

i. NP coordination — coordination of nouns or noun phrases as a single NP.2 ii. Additive coordination — this is our term for two types of coordination which are often similar in Oceanic languages, but marked differently in some languages: a. coordination of units with tens in numerals, referred to as numeral coordination; b. an “et cetera” marker at the end of lists, referred to as end-of-list coordination; the enumeration may be explicit (“yams and taros and sugarcane and”) or implicit (“Hina and [her family and relations, etc.] company”). iii. VP coordination — where the subjects of the conjoined VPs are identical, and where there is no overt subject marker in the non-initial VP (or, by another interpretation, where the conjunction replaces the subject marker); iv. Clause coordination (abbreviated CL) — where the conjunction does not replace the subject marker in the non-initial clause. In some languages we need to distinguish two sub-types: a. Same-subject clause coordination (abbreviated SS) — where, as in (iii), the subjects are identical, but where the subject is overtly expressed (i.e. the conjunction does not replace the subject-marker); and b. Different-subject clause coordination (abbreviated DS) — where the subjects of the conjoined clauses are not identical. Note that some languages do not have category (iii), while others have (iii) but not (iv-a). We use the term general coordination when a single conjunction covers (nearly) all of the functions described above (as, for example, in Samoan). The relationship between coordinators and comitatives is well known (cf. Stassen 2000, and § 5 below). However, as far as strict coordination (that is, NPs, VPs and clausal coordination) is concerned, the existence of several morphemes whose presence is conditioned by the nature of the coordinated elements is an interesting typological characteristic, to which Indo-European languages have not accustomed us. Oceanic languages, in fact, offer a great variety of markers of coordination (and we limit ourselves in this paper to coordination of the type ‘and’, ‘and then’); and if the coordination of phrases, whether nominal or verbal, is generally effected by the morpheme *ma/*me (see below § 3.1), that of clauses is generally effected by other markers, or by simple juxtaposition. We also show that there is a cline or continuum of closeness of coordination in certain Oceanic subgroups. We use terms like tight on the one hand and loose on the other to describe the ends of the continuum, the former implying a closer

2.We exclude from discussion here the “inclusory pronominal” construction (Lichtenberk 2000) of the type ‘we-two Peter = Peter and I’; although very widespread in Oceanic, this construction does not necessarily incorporate conjunctions, which are the focus of this paper.

Coordination in Oceanic languages and Proto Oceanic 449

connection between the conjuncts (whether nominal, verbal or clausal) than the latter. However, we note that there may be a number of intermediate stages, with different languages behaving differently in this regard, and we will more frequently use the terms tighter and looser to indicate this. Two summary tables (Appendix 1) present all of these coordinating or additive markers in several Southern Melanesian and Polynesian languages, according to the different types of coordination discussed. 1.2 Existing reconstructions A number of conjunctions have been reconstructed for various high-level protolanguages within Austronesian (Blust 1981, in prep., Grace 1969, Ross and Lithgow n.d.). The following is, we think, a complete list of those which have reflexes in at least some Oceanic languages coordinating NPs, VPs and/or clauses, with glosses as given in those sources (PAn = Proto Austronesian, PMP = Proto Malayo-Polynesian, PWOc = Proto Western Oceanic): (1) Pre-Proto Oceanic PMP *a ‘conjunction: and’ PMP *ba, *mpa ‘conjunction: or, if, perhaps, because’

PAn *ka ‘conjunctive particle: and’ ? PAn *maS ‘and’

PAn *may ‘and’ PAn *Na ‘and’

Proto Oceanic/Post-Proto Oceanic > POc *a ‘and, but’ ? > PWOc *be ‘subordinating/ irrealis conjunction’ ? > PWOc *bwa ‘sequence adverb: and, then’ > POc *ka ‘coordinating conjunction: and, with’ PWOc *ga ‘realis coordinating conjunction’ > POc *ma ‘coordinating conjunction: and, with’, POc *ma-i ‘comitative prepositional verb’ > POc *me ‘coordinating conjunction: and, with’ > POc *na ‘coordinating conjunction: and, with’

We return to these reconstructions in § 7. However, the POc form *me deserves particular, though brief, comment here. Although *ma is widely referred to as a Proto Oceanic coordinator, we are not aware of a doublet reconstruction *me. Nevertheless, many Oceanic languages have the form me rather than ma, among them Mussau (unclassified, possibly Admiralties), Malalamai, Kilenge, Nakanai,

450 Claire Moyse-Faurie and John Lynch

Nehan, Roviana (all Western Oceanic), Buma, Trukese, Ma¯ori and Hawaiian (Central-Eastern Oceanic). Lynch, Ross and Crowley (2002: 75) have suggested that the me-forms may derive from the conjunction + transitive suffix *ma-i. However, note from (1) that Blust (in prep.) reconstructs the Proto Austronesian doublets *maS and *may: the former would undergo regular development as POc *ma, the latter as POc *me.


Nominal coordination

In this section, we discuss nominal coordination in Oceanic languages. We show that a number of languages show a distinction (i) between tight and loose nominal coordination, and (ii) between coordination of common and proper NPs. 2.1 Tight and loose coordination In a few Southern Melanesian languages (a proposed subgroup including the nonPolynesian languages of Southern Vanuatu and New Caledonia — see Lynch 2000b), there is only one NP coordinator:3 (2) Sye Anejom ˜

nelat im noki ‘meat and coconut’ Natu im etwan ‘Natu and her sister’

This, however, is not the norm in Southern Melanesia. Quite a few of these languages have two nominal coordinators. In many cases, one of these is used to denote items which can be considered as “couples” or “pairs” — that is, two (and usually only two) nominals which are closely associated in the real world. The other coordinator is used with items which are less closely associated, at least in the particular context of the speech-event in process.4

3.Sye has the occasional variant form mi, and the more frequent procl*tic m-, but we take these to be simply variants of the same single coordinator deriving from POc *ma. 4.Haspelmath (to appear) draws a distinction between natural and accidental coordination (see also Mithun 1988: 332ff.), which corresponds to what we have called tight/tighter and loose/looser coordination. While his terms may seem more appropriate for nominal coordination, they are perhaps not so apt in reference to verbal and clausal coordination, and we will use the latter terms throughout this paper.

Coordination in Oceanic languages and Proto Oceanic


tighter: couples/pairs n6mataag m nihin ‘wind and rain’ Kwamera nes6n m n6matagi ‘rain and wind’ kut men daan Nemi ‘rain and wind’ pa¯ nÚajá mä párui Paicî ‘months and years’


Cèmuhî Xârâcùù

ni dà me bwØet ‘spears and clubs = arms’ ku mää mwè ‘yam and taro’

looser: other kuri m66ne pukas ‘a dog and a pig’ suka m66ne n6kava m66ne rais ‘sugar and kava and rice’ ngeli kuuk o ngeli hyo ‘the yams and the taros’ ¯ı nä-w·e bau ¯ı nä-wají â ‘the taro fields and the sugarcane fields’ ni u¯ ka¯ ni ujà â ‘yams and sugarcane’ ku mê dé mê mwè mê ‘yams and sugarcane and taro and so on’

In Cèmuhî, me indicates simultaneous association, may coordinate common nouns whose referents are seen to be very closely associated, and may also coordinate proper nouns and anthropomorphised nouns; the non-initial nouns in a conjunct with me may not be preceded by the article. In other words, me coordinates two nouns in tight coordination (and it is also the comitative marker). On the other hand, ka¯ indicates normal coordination or succession, and the non-initial nouns must be preceded by an article; in other words, it coordinates two noun phrases in loose coordination. Thus the structure of the Cèmuhî coordinate NPs given above is: (4) Cèmuhî ni dà me bwØet art spear and club ‘arms (spears and clubs)’

ni u¯ ka¯ ni ujà â art yam and art sugarcane ‘yams and sugarcane’ (Rivierre 1980: 171)

A similar situation occurs in the Paicî and Nemi examples above: there are no articles in the tight construction, but articles occur in the loose construction (Paicî i, Nemi ngeli in the example in (3)).5 In Xârâcùù, the coordinator mää marks the conjuncts as being indissociable (for example, yams and taros in traditional custom), or to designate the two members of a couple; whereas mê implies a looser relationship. Compare:

5.The Nemi situation is actually more complex than illustrated here, since there is also a proper coordinator ma whose use seems to be spreading. See § 2.2 and 3.2.2 below.


452 Claire Moyse-Faurie and John Lynch

(5) Xârâcùù gu mää gè ‘you and I’ (as a couple)

gu mê gè ‘you and I’ (no strong interpersonal relationship)

In the Lenakel examples in (6), the coordination labelled “tighter” which is used with paired nominal items is also used, with the same coordinator, with same-subject closely sequential VPs; that labelled “looser” is used with more randomly associated nominals or (with a different coordinator) with different-subject VPs which are also less closely sequential. (6) Lenakel tighter coordination nihin m n6mataag ‘rain and wind’ I-6m-va m-6m-apul. 1excl-past-come es-past-sleep ‘I came and slept.’

looser coordination suka m66ne n6kava ‘sugar and kava’ R-6m-va kani i-6m-apul. 3sg-past-come and 1excl-past-sleep ‘S/he came and I slept.’

We will have more to say about the relationship between nominal and verbal coordination later (see § 3). Among North-Central Vanuatu languages, we have found the distinction between tight and loose nominal coordination only in Mwotlap and Namakir, other North-Central Vanuatu languages which we have looked at normally having just a single NP coordinator. In Mwotlap, the distinction is between ‘and’ and ‘and also’: (7) Mwotlap No m-et Wilson (tiwag) mi Sera / Wilson, ba Sera. 1sg perf-see Wilson (together) with Sera / Wilson and Sera ‘I saw Wilson and Sera/Wilson, and also Sera.’ (Alexandre François, p.c.) In Namakir, ma coordinates human nouns where the relationship is very tight (and the article is not used on the second noun); (a)ne is used elsewhere in nominal coordination, and also coordinates clauses: (8) Namakir Ke-John ma Meri ri a dodomimir. art-John and Mary 3du recip love.them ‘John and Mary love each other.’ Ke-Harry ne ke-Noel ne ke-Wilson ne ke-David ru art-Harry and art-Noel and art-Wilson and art-David 3pl pu munum. int drink ‘Harry and Noel and Wilson and David are going out drinking.’

Coordination in Oceanic languages and Proto Oceanic 453

In Mangap-Mbula (Western Oceanic; Vitiaz linkage, North New Guinea Cluster), “there are strong tendencies for ma to be used when the combination of conjuncts is more formulaic …the conjunction mi, on the other hand, is used to conjoin more arbitrary lists of NPs” (Bugenhagen 1995: 214). For example: (9) Mangap-Mbula Tight/formulaic mbe] ma aigule ‘night and day (=all the time)’ moori ma tomooto ‘men and women, people’ tamaana ma naana ‘father and mother, parents’

More distant/arbitrary narabu lamata mi ye ru ‘five loaves and two fishes’ ]ge tai]gi mi ]ge tina ‘this pig and that pig’ lama-ta mi ru ‘one five and two = seven’

Tawala (Western Oceanic; Are-Taupota linkage, Papuan Tip Cluster) has the tight nominal coordinator po and the loose coordinator ma: (10) Tawala Tight/formulaic oloto po wawine ‘male and female’

More distant/arbitrary polo ma aniani dimdim ‘pork and trade food’

There appears to be no tight/loose distinction in nominal coordination in Polynesian languages. 2.2 Common NPs vs proper NPs coordination Some Oceanic languages make another distinction, using different coordinating morphemes depending on the categories linked. In Nêlêmwa, Nemi and Drehu, and also in languages of the Eastern Polynesian subgroup of Central Pacific, proper noun phrases are coordinated differently from common noun phrases. In Drehu (Moyse-Faurie 1983: 176), me is used with pronouns, proper nouns or common nouns referring to animates and culturally forming a tight/formulaic pair (and also with VPs, as we will show below § 3.1.1), and memin with common definite NPs (where it must be followed by the definite article la). (11) Drehu la kem me thin i angeic art father and mother poss 3sg ‘his father and mother’ angeic memin la thin i angeic 3sg and art mother poss 3sg ‘he and his mother’

454 Claire Moyse-Faurie and John Lynch

We showed in (3) above that, in common NPs in Nemi, men marks tight coordination and o marks loose coordination. With proper nouns (including anthropomorphised nouns), however, ma is used: (12) Nemi Kaavo ma Hixe khiny ma deny ‘Kaavo and Hixe’ ‘Wood-Swallow and Buzzard’ (Ozanne-Rivierre 1979 vol. 2: 53) Ma is also used in Nemi to link definite common nouns with human reference: (13) Nemi jama o ven kac ma ven hnook myth poss art man and art woman ‘the myth of the man and the woman’ (Ozanne-Rivierre 1979 vol. 2: 21) The situation in Nêlêmwa is also quite complex (see also Bril, this volume). Three coordinators can link NPs: i.

Ma links NPs referring, in principle, to animates, proper nouns, including anthropomorphised animals, etc. Ma is also used for end-of-list coordination with animates (‘and company’). ii. Me has significantly less restricted use, coordinating inanimate or animate NPs, and also being used in all other coordinating situations: in alternation with xa (see below § 3.2.1) in coordinating VPs and SS clauses; and exclusively for the coordination of DS clauses. Me also signifies ‘and so on’ at the end of enumerations of inanimates. iii. Xa ‘and also’ has above all an additive value; it can appear as a verbal determiner ‘also’, in numerals or, more rarely, in (loose) NP coordination. (14) Nêlêmwa kijin i Bodanu ma Fwâmaanu tan me taan shaamoâ me yameewu fwaago

‘the story of Bodanu and Fwâmaanu’ ‘night and day’ ‘Samoa bananas and the Fwaago type (of banana)’ Hla u oda mwa hlileny thaamwa xa ye. 3pl pfv go.up really these.2 woman also he ‘These two women and also him then went up.’ (Bril 2000: 356)

Nêlêmwa appears to be one of the very few languages which apparently retain both POc *ma and *me. It is not clear, however, if the distinction in Nêlêmwa reflects an earlier Proto Oceanic distinction, or whether Nêlêmwa has made an innovation. (In any case, this distinction is tending to disappear in Nêlêmwa, speakers more and more using one or the other indifferently, at least for VP and clausal coordination.) In Eastern Polynesian languages (e.g. in Rapanui, Tahitian and Ma¯ori), proper

Coordination in Oceanic languages and Proto Oceanic 455

NPs are also coordinated differently from common NPs. In Ma¯ori, me links NPs only when they contain common nouns. However, me cannot coordinate proper nouns;6 in this case, a dual or plural pronoun followed by the predicative ko (i.e. an inclusory pronominal construction) has to be used. (15) Ma¯ori Kei whea te waka me te hoe? tam where def canoe and def paddle ‘Where is the canoe and the paddle?’ (Biggs 1969: 102) Kei a Hone ra¯ua ko Mere o¯ra¯ua koti. (be).at pers Hone 3du pred Mere their coat ‘Hone and Mary have their coats.’ (Bauer 1997: 641) In modern Tahitian, NP coordination is effected solely by the clause coordinator ‘e/e, and the reflex of Proto Polynesian (PPn) *ma is no longer used except in the coordination of active VPs (see below). Nowadays, proper nouns have to be preceded by the predicative marker ‘o and coordinated by ‘e/e, the same marker which coordinates common nouns and clauses: (16) Modern Tahitian ‘o Ti‘a¯ouri ‘e ‘o Ti‘a¯oatea pred Tiaouri and pred Tiaoatea ‘Tiaouri and Tiaoatea’. te pua‘a ‘e te ‘urı¯ ‘e te ma¯moe art pig and art dog and art sheep ‘the pig and the dog and the sheep’ (Lazard and Peltzer 2000: 97) In old Tahitian, ma could link two proper nouns, while ‘e/e linked common nouns: (17) Old Tahitian o Tane ma Rua-nu‘u e te vahine a Tane o ‘Aruru … pred Tane and Rua-nu’u and def woman poss Tane pred ‘Aruru ‘Tane and Rua-nu’u and also ‘Aruru, Tane’s wife…” (Henry 1988: 473)


Verb phrase and clausal coordination

The Oceanic languages considered in this paper fall into three broad types as far as VP and clausal coordination are concerned:

6.Me is also the comitative marker, and in this function, it can link proper nouns as well as common nouns. There seems to be no real semantic difference between the comitative construction with me and the coordinating one with pronoun+ko (see Bauer 1997: 551–552).

456 Claire Moyse-Faurie and John Lynch


Those in which VP coordination is marked in the same way as nominal coordination, and differently from clause coordination: e.g. western Polynesian languages, Ajië, Drehu, Nengone. Languages which have two nominal coordinators generally use either the tighter of the two for VP coordination (e.g. the languages of Tanna and Anejom ˜ in Southern Vanuatu, Nemi and Paicî in New Caledonia), or the looser (e.g. Xârâcùù). ii. Those in which VP, nominal and clausal coordination are marked in the same way: e.g. Samoan, Sye, Iaai, and probably Drubéa. iii. Those which do not allow strict VP coordination, and where clausal coordination is different from nominal: e.g. Cèmuhî, Tîrî. Eastern Polynesian languages could also fall in this category, since strict VP coordination is not allowed, a definite article being required before the second verb (see § Type (i) has the widest distribution, and was probably the original system (see § 7 for more discussion of this). We exemplify these different types below. 3.1 Phrasal (NP/VP) and clausal coordination marked differently 3.1.1 In Southern Melanesian languages In most of the Southern Melanesian languages which fall into this category, person and number of the subject are marked by a preverbal morpheme, which in some languages also marks TAM. A reflex of *ma replaces this subject-marking particle when two same-subject VPs are coordinated.7 In languages which have the tight/loose distinction, it is often the tight coordinator which occurs here, as in Anejom: ˜ (18) Anejom ˜ Era pu hag aara m-amjeg. 3pl.aor fut eat.intr es-sleep ‘They will eat and sleep.’ Paicî È mwââ pá mê mä pu¯rö. 3sg then take dir and cook ‘She brings and cooks them.’ (Rivierre 1983: 138) but in Xârâcùù it is the loose coordinator:

7.In the literature on the Southern Vanuatu languages, this prefix has been referred to as the “echo-subject” prefix (abbreviated es), for reasons which are not relevant to this discussion.

Coordination in Oceanic languages and Proto Oceanic 457

(19) Xârâcùù Ru cha mê mara. 3du clear.bush and ‘They cleared the bush and worked in the fields.’ (Moyse-Faurie 1995: 124) while in Drehu it is the morpheme which is used to coordinate proper NPs which also links VPs: (20) Drehu Hnei angeic hna oleen me kapa la itre atr. ag he past thank and receive art pl man ‘He thanked and received the people.’ In these languages, if the subject of CL2 is overtly marked, then a quite different coordinator is used, whether or not the subject of CL2 is the same as that of CL1 (SS and DS being used before the examples below to mark same- and different-subject clauses). The examples below illustrate the coordinators Anejom ˜ am, ˜ Xârâcùù nä and Drehu nge: (21) Anejom ˜ SS: Is ecohos-pan aan ehele-n is am ˜ imy-ecej 3sg.past appear-there he dat-3sg 3sg.past and com-say.come y