Complex Predicates in Oceanic Languages: Studies in the Dynamics of Binding and Boundness 9783110913286, 9783110181883

Citation preview

Complex Predicates in Oceanic Languages

w DE

G

Empirical Approaches to Language Typology 29

Editors

Georg Bossong Bernard Comrie Yaron Matras

Mouton de Gruyter Berlin · New York

Complex Predicates in Oceanic Languages Studies in the Dynamics of Binding and Boundness

edited by Isabelle Bril Fran^oise Ozanne-Rivierre

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

© Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability.

ISBN 3 11 018188 6 Bibliographic

information

published

by Die Deutsche

Bibliothek

Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at < h t t p : / / d n b . d d b . d e > .

© Copyright 2004 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Printed in Germany.

Contents List of Contributors Introduction

vii ix

Complex nuclei in Oceanic languages: Contribution to an areal typology Isabelle Bril 1 What do we really know about serial verb constructions in Austronesian and Papuan languages? Gunter Senft 49 Core-layer junctures in Saliba (Saliba Island, Milne Bay province of Papua New Guinea) Anna Margetts

65

Serial and complex verb constructions in Teop (North Bougainville) Jessika Reinig

89

Chains of freedom: Constraints and creativity in the macro-verb strategies of Mwotlap (Vanuatu) Alexandre Frangois

107

Serial and compound verbs in Anejom (Vanuatu) John Lynch

145

Complex verbs and dependency strategies in Nelemwa (New Caledonia) Isabelle Bril

167

Complex predicate constructions in East Uvean (Wallis) Claire Moyse-Faurie

199

Serial verbs and complex constructions in Pileni (Solomon Islands) Ashild Ncess

225

Complex predicates in Tahitian: A particular case of qualitative modification Mirose Paia and Jacques Vernaudon

251

Complex predicates and Juxtapositional Constructions in Samoan Ulrike .'.,'·, sv/

263

The grammaticization of directional verbs in Oceanic languages Malcolm Ross

297

vi

Contents

The evolution of the verb 'take' in New Caledonian languages Frangoise Ozanne-Rivierre

331

Verbal compounds and lexical prefixes in the languages of New Caledonia Frangoise Ozanne-Rivierre and Jean-Claude Rivierre

347

References

373

Index

397

List of Contributors

Gunter Senft Max-Planck-Institut fur Psycholinguistik, PB 310, NL-6500 AH Nijmegen, Netherlands [email protected]

Anna Margetts Monash University Victoria 3800, Australia [email protected] Jessika Reinig University of Kiel [email protected] Alexandre Francois LACITO-CNRS, Centre Andre-Georges Haudricourt 7, rue Guy Möquet, 94800 Villejuif, France [email protected]

Claire Moyse-Faurie LACITO-CNRS, Centre Andre-Georges Haudricourt 7, rue Guy Möquet 94800 Villejuif, France moyse@vj f.cnrs. fr Ashild Naess Dept. of Linguistics University of Oslo P.O. Box 1102 Blindem 0317 Oslo, Norway [email protected] Mirose Paia INALCO, Paris [email protected] Jacques Vernaudon Universite de Nouvelle-Caledonie Noumea. vernaudon@univ-nc .nc

John Lynch Pacific Languages Unit, University of the South Pacific, Port Vila, Vanuatu [email protected]. AC.FJ

Ulrike Mosel University of Kiel. Olshausenstraße 40 D-24098 Kiel, Germany [email protected]

Isabelle Bril LACITO-CNRS, Centre Andre-Georges Haudricourt 7, rue Guy Möquet 94800 Villejuif, France [email protected]

Malcolm Ross Research School of Pacific and Asian Studies and Centre for Research on Language Change, The Australian National University Canberra ACT 0200, Australia [email protected]

viii

Contributors

Fran9oise O z a n n e - R i v i e r r e L ACITO-CNRS, Centre Andre-Georges Haudricourt 7, rue Guy Möquet 94800 Villejuif, France [email protected]

Jean-Claude Rivierre LACITO-CNRS, Centre Andre-Georges Haudricourt 7, rue Guy Möquet 94800 Villejuif, France [email protected]

Introduction

This book originated in a workshop on "Serial and Compound Verbs in Oceanic languages", organized in Paris by Fran - shows both the timeiconic ordering of actions and the [modifier-head] order of the quantifier. Thus, serial verb constructions often contradict the parametric settings of a language (see Carstens 2000) and favour time-iconic sequencing of serial verbs, whereas complement clauses are highly sensitive to the VO or OV parameters and could constitute a good distinctive criterion between both types. Durie (1997: 330-339) contends that the invariant sequencing of such serialized constructions proves the limits of the syntactic treatment of serialization and advocates a more cognitive or semantic approach. In support of this, Durie cites some language-specific semantic sequencing constraints: in Lahu, abstract verbs must precede concrete verbs (Matisoff 1969, in Durie 1997: 338); in Nelemwa (this vol.), the concrete verb always appears in Vi position and the abstract meaning of a verb often correlates

Complex nuclei in Oceanic languages: an areal typology

13

with its modifying function in the V 2 slot; thus, in the V 2 slot, the verb ä 'go, leave' may have the dispersive reading 'from place to place' and bwage 'go back, return' the reversive reading 'back'. Other languages specifically prefer verbs over directionals to express the abstract directions of verbs of perception or cognition.

3.2.2. Ordering of inchoative, terminative and modal verbs If no other morphosyntactic clue is available, the iconic ordering of inchoative and terminative Aktionsart verbs is a possible indicator of serialization, in contrast with their grammaticalization as aspectual operators or with complementizing constructions (as in English start/stop doing something). In most of the languages of the volume, their position is iconic, even though they display various degrees of specialization, incipient deverbalization, morphological erosion and grammaticalization. As for modality verbs ('try, able, hope, expect'), they tend to appear to the left of the main verb, while modifying verbs (of manner, result, direction) tend to appear to its right. Bi-directionality is thus common in asymmetrical constructions (see Bisang 1996: 580 for examples in Asian languages).

3.2.3. Ordering of arguments and argument structure Argument ordering in serial constructions results from a compromise between (i) the constraints of a matrix verb's argument structure (i.e. the number and type of arguments it may subcategorize for), (ii) the constraints of a complex verb's argument structure and (iii) more generally the Thematic Hierarchy 11 . Durie (1997: 330-40) contra Baker (1989: 541, 544) contends that the ordering of verbs is basically iconic and cannot be reduced to the Thematic Hierarchy; on the contrary, the Thematic Hierarchy "is an artifact of verb sequencing constraints". Argument ordering in the languages of this volume involves composition, sharing and fusion; none of them allow for multiple objects or contiguous objects such as mentioned by Senft (this vol.) and Crowley (1987: 39, 50).

14

Isabelle Bril

3.2.3.1. Argument-sharing vs argument fusion in contiguous nuclear constructions Argument-sharing is a powerful discriminating test for serial constructions (Foley and Olson 1985) though subject to much language-specific variation since neither subject nor object-sharing is universally obligatory. 3.2.3.2. From compositionality to sharing and fusion a) In core-layer serialization, argument structure is compositional, each verb has its own arguments and the complex VPs are ordered according to language-internal parametric settings including constituent order and conceptually iconic order. b) In contiguous nuclear constructions, argument structure ranges from shared arguments in symmetrical constructions to fused arguments in asymmetrical constructions. Foley and Olson (1985: 44) and more recently Durie (1997: 344-48) argue for argument fusion 12 with "an integrated set of semantic roles" when two or more verbs differing in the argument structure of their lexical entries constitute a complex nucleus. The "fused thetahierarchy is not the simple addition of the theta-roles for the two verbs", it is a different conceptual structure altogether (Durie 1997: 348). Table 5. Complex predicates with Same Subject (SS) and Different Subject (DS) constructions. Saliba

nuclear or contiguous type core type

Teop

Mwotlap

Anejom

Nelemwa Southern Pileni & NorthN.C. em N.C. languages languages

SS

SS or D S

SS or D S

infrequent

SS or D S

(+)

SS

SS

*

*

*

*

*

SS or D S

East Uvean

Samoan Tahiti an

SS

SS

*

*

- In symmetrical constructions, subject-sharing is obligatory in all the languages of this volume; it is the primary criterion which distinguishes them from asymmetrical modifying constructions. Object-sharing only applies to transitive verbs which have the same patient (SVtr.Vtr.o). If the sequence contains verbs with different valency, the intransitive verb (often a verb of motion) obligatorily precedes the transitive verb ( s v ^ Vtr.o), whether in v o languages (Nelemwa, Pileni) or in OV Saliba (SVintr. OsVtr.).

Complex nuclei in Oceanic languages: an areal typology

15

- In asymmetrical constructions, there are two main cases. In Nelemwa, Pileni or Saliba, the complex verb has a single set of arguments which is based on the argument structure of the head (V]), and the argument structure of the modifying verb (V2), be it intransitive or stative, fuses with it. Thus a transitive V] will trigger transitive concord on V 2 , which is marked by a transitive suffix [SVtr. V t r a n s 0 ] (as in Nelemwa, ex. (4) below or in Pileni) or an applicative affix as in Saliba [(SO) sVtr.V(V)apPi.o]. The inflectional concord in asymmetrical constructions is evidence that they constitute one complex predicative unit. ed

(4)

Nelemwa (Bril) [ex. 25b] Hla diya häähuux-e mwa 3 PL do be recent-TR house 'They built this house recently.'

eli that.ANAPH

The second case is illustrated by Mwotlap, with only asymmetrical nuclear "macro-verbs" and strict SVO order: "in case of conflict for the object position [...] the macro-verb adopts the primary orientation of Vi and the secondary orientation of V2" (Francois, 3.4. this vol.). Thus, in (5) two intransitive predicates make up one single complex transitive causative predicate with cause-effect or cause-result meaning within the SVO template and with a single argument structure. (5)

Mwotlap (Francois) [ex. 24] Ne-lefi {mi-yip hal-yak)

na-kat.

ART-wind

ART-cards

PFT-blow

fly-away

"The wind blew the cards away.' This is not a case of argument-sharing but of argument-restructuring and fusion, since in a non-serial construction, both verbs would be intransitive with different subjects. A non-compositional argument structure (i.e. different from that of its constituent parts) for a complex verb often signals some functional specialization, grammaticalization or lexicalization. This is not quite the case in Mwotlap, but this might well be a transitional stage in a general process of compression from looser core serialization with different subjects (such as < the wind blows the cards fly away >) towards compact nuclear transitive macro-verbs, in the process of specializing and grammaticalizing: V\ as a light causative verb or V 2 as an adverb, according to the structural slot which attracts grammaticalization for a specific function.

16

Isabelle Bril

3.2.4. Cross-linguistic

comparisons

Inflectional concord (in person, number) with the subject of the main verb or in transitivity with Vi signals one complex syntactic unit. In Tariana (Aikhenvald 1997: 476-78), causative serial constructions trigger subjectaffix agreement with the subject of the causation verb; this affix is cross-referenced on all verbs Ivou take you cause-cross you cause-stand], Tariana (Aikhenvald 1997: 476-78) (6)

Phia-nikha

phita

you-REC.p.iNFR

2sG+take 2sG-cross+CAUS

ha-ne-na

pi-thaketa

pi-eme 2SG-Stand+CAUS

hyapa-na-nuku

DEM-DISTAL-CL:VERT

hill-CL:VERT-TOP.NON.A/S

ha-ne-riku-ma-se DEM-DISTAL-CL:LOC-CL:PAIR-LOC

'Was it you who brought that mountain across (lit.you take you cross you put.upright) (the river) to the other side?' (REC.P = recent past, INFR = inferred, s = intransitive subject) In Öbölö (Niger-Congo, Faraclas 1984), subject-affix concord occurs in spite of switch-subject reading, but only in certain moods (positive imperative, hortative and subjunctive utterances) (7a.), and does not appear in the indicative (7b.), nor in the negative mood (Uche Aaron, in Durie 1997: 301). Durie, analyses such concord as a sign that they behave as co-heads and not as multi-headed VPs. Öbolo (Uche Aaron, cited in Durie 1997: 299, 301) (7) a E-gwen PL-call

emi

έ-ηύ

ISG

PL-come

' L e t t h e m call m e to c o m e . '

b. Emi ti-sa oge i-fieek ä-näm I lSG-use knife 3sG-cut meat Ί cut the meat with a knife.'

4.

Functions and semantics of serial constructions

As Table 7 shows and as cross-linguistic studies confirm, contiguous serial or complex predicates are polyfunctional and polysemous constructions whose interpretation is filtered by various factors: the type of predicate or

Complex nuclei in Oceanic languages: an areal typology

17

verb class, functional slots, types of collocations, as well as contextual inferences. Among their common functions and semantics are sequential or coincident actions, modification (or value), specification of circumstances or manner of action, causative, resultative, purposive, aspectual and less frequently, case-role marking (benefactive, comitative). But the mapping between construction, functions and meanings is highly variable and not easily generalizable cross-linguistically.

4.1. Distribution of types and functions The aforementioned functions may be mapped onto one to four different construction types (core or nuclear and their symmetrical or asymmetrical subtypes); yet few of the languages of the volume actually make use of them all. Some have three types (Saliba, Pileni), some two (Teop, Nelemwa), some only one (Mwotlap, Samoan, Tahitian). To the asymmetrical subtype is generally assigned modification (property, value) or specification of circumstances (manner, result, location, etc.), while the less frequent symmetrical subtype expresses sequential or purposive actions (Teop, Nelemwa, Pileni). Table 6. Serial constructions in Saliba and Pileni. nuclear asymmetrica 1 Saliba

(mostly) modification

Pileni

modification

symmetrical

sequential, purpose

core asymmetrical

symmetrical

(infrequent) Aktionsart, modal, causative

sequential, purpose

(infrequent) Aktionsart, action-result

Functions of ASYMMETRICAL CONTIGUOUS NUCLEI

ε

2

[direction, result, concomitance]

V

2

+ transitive concord

2

+ number concord (redup.)

[manner, concomitance sequential]

v

V

[value, manner, concomitance purposive, similative]

v2

2

[main action]

*

[value, manner, result concomitance purposive, similative, additive, etc.]

1 [main action, inchoative, modal]

[main action, inchoative imminence]

[main action]

£η

+ transitive concord

[value, manner, terminative]

V

1

v,

modification

V

ΐa1 >

modification

£

modification

13 σ

mostly modification

modification

modification

only modification

VI Ο

mostly modification

East Uvean

ο

Β

modification

Mwotlap

Teop

Pileni

Nelemwa and North. N.C. languages

Anejom

c υ 3 σ

e '5 ο

O. C c and strike dead with German totschlagen or < l i t pull come o u t > with pull out and herausziehen, or < lit scratch remove > with scrape away and wegschaben (Sebba 1987: 217; Rousseau 2000: 390-91). Serial verbs such as Nelemwa thege ulep du (lit. run pass down) are akin to complex predicates involving an adverb (run out) or a co-predicative participle (go out running, sortir en courant). They all refer to the circumstances of an action or to coincident actions and constitute one single event, in contrast with multiclause constructions which refer to sequential events such as / thege me i ulep du 'he runs and goes out', il sort et (il) court. There are many common points between asymmetrical serial verbs, medial verbs and converbs (König 1995; Haspelmath 1995: 5, 23-27, 3745; Bisang 1995: 137-88), they specify the circumstances of the main action (manner, intrumental, concomitance, locative, etc.), with a similar scale of desententialization (Lehmann 1988: 200). But there is a major difference: asymmetrical serial verbs may only be modifiers of verbs and do not display any dependency marker or nonfinite verb morphology, while converbs are nonfinite verb forms [such as adverbial participles they eat laughing or periphrastic converbs such as French il l'a fait en chantant] whose main function is to mark adverbial subordination, they are modifiers of verbs, clauses or sentences (Haspelmath 1995: 3, 6-9, 12-17). Besides, serial verbs and converbs alike may specialize and grammaticalize into adpositional or case-role markers (or "coverbs" Li & Thompson 1974a), applicative markers or conjunctions (Haspelmath 1995: 37-42).

7.

Structural compression

Languages show various degrees between two poles: towards more syndesis or more desententialization (Lehmann 1988: 210-13). Formal compression often correlates with economy of form, discourse, pragmatic functions and semantic proximity, as much as with syntactic binding. Croft (2001: 352-53) suggests that the pathways from less to more tightly integrated linkage and clause fusion are oriented from syndetic to asyndetic coordination, on to co-ranking serialization, leading some verbs to specialize and grammaticalize as governing verbs in complement

32

Isabelle Bril

constructions or as subordinating functors or conjunctions in subordinate constructions.

7.1. Tightening linkage and structural integration or compression The sole nuclear type exhibited by most of the languages in this volume might result from the reduction of the same-subject or different-subject core types, through gapping, ellipsis and argument fusion 20 . This would result in one single nuclear juncture with two different structural subtypes encoding different syntactic relations: (i) symmetrical/co-ranking nuclei referring to sequential actions and (ii) asymmetrical modifying nuclei displaying argument fusion and possibly transitive or number concord. Polyfunctionality would thus arise, as often, from gradually merging structures, with an intermediate period in which both types co-exist, one becoming dominant and the other receding. The centripetal tendencies towards integration and compression occur at all levels, from sentence and clause level to core and more desententialised nuclear juncture and further down on the scale of desententialization, favoured by contiguity (Lehmann 1988), at word level toward lexicalization, compounding, grammaticalization or morphologization.

7.2. Tightness of linkage and semantic or pragmatic integration Languages make different choices as to which functions (modification or sequential actions) are thus compressed (see 6.2.1 above). Tightness of linkage often correlates with semantic closeness and is subject to the same subject constraint and topic continuity. In Anejom, clauses with samesubjects allow some free choice between the four types of clause-linking: (i) asyndetic (parataxis), (ii) syndetic (conjunctive) clause-linking, (ii) "echo-subject" clause-chaining or (iv) infrequent nuclear serial21 constructions, whereas different subjects or different aspect-mood operators bar nuclear serialization, favouring parataxis or conjunctive aspect-markers (Section 2). This highlights some cross-linguistic tendencies of syntactic, semantic and pragmatic integration such as those analysed by Givon (1980) in his extensive research on English complementation and on the covariation between tight linkage, the semantics of the governing verb (desiderative, manipulative, perceptive, etc.) and contextual semantics. For instance, the

Complex nuclei in Oceanic languages: an areaI typology

33

type of complementation (0, to, that, -ing) is sensitive to aspectual semantics, as in I remembered taking the portrait (process) vs I remembered to take the portrait (action), or he began to dance (onset of action) vs he began dancing (initial phase of the process). The degree of integration thus covaries with the type of conceptual relations that languages conflate or distinguish. Some cross-linguistic features are however apparent. Foley & Van Valin (1984: chap. 6) have proposed a hierarchy of the correlation between tightness of linkage and semantic integration, which is only partly verified in this volume. They propose the following cline of semantic and functional domains, ordered from tight linkage (single event) to weak linkage (distinct events): causative > aspectual > psych action > purposive > jussive > perception > prepositional attitude (love, hate) > cognition > indirect discourse > temporal adverbial > conditionals > simultaneous actions > sequential actions (+/-overlapping) > action-action (unspecified) Tight nuclear juncture or looser core juncture in the languages of the volume (in bold italics) mostly concern adverbial modification {manner > circumstances), aspectual, time {simultaneous actions > sequential actions), purposive, causative, and resultative. Another common feature of serial constructions is their high polysemy, they encode sequential actions, actionpurpose, cause-effect and cause-result, which constitute a semantic continuum with no clear-cut boundary between them. Thus, Mwotlap's "macro-verbs" have a variety of closely related semantics (causative, purposive, resultative, depictive, manner of action) filtered by context. Samoan's juxtaposed constructions (Mosel) also have a variety of context-dependent meanings: [ex. 24] lamu fa'a-malü (lit. chew CAUS-soft) may be interpreted as purposive 'chew to soften' or as expressing manner of action 'soften by chewing', [ex. 25] tapena fa'a-lelei le fale (lit tidy up CAUS-good the house) 'tidy up the house so that it looks good' expresses purpose and cause-effect. A similar semantic continuum is apparent in the English conjunction 'so that' which may have consecutive, resultative or purposive reading depending on the sentence's ΤΑΜ information. Compare: he works hard so that he ΊΙ get the job {to get the job, purpose) and he worked hard so that he got the job {and he got the job, result).

34

Isabelle Bril

7.2.1. Integration of action-purpose~goal The choice of a tight serial construction vs syndetic subordination for purposives is cross-linguistically common (Croft 2001: 352) and shows a tendency to compression whenever semantically allowed. The main difference is structural (one vs two clauses) and conceptual (one vs distinct events). Syndesis vs serial constructions In Nelemwa, Tahitian and East Uvean, purposives can either be expressed by contiguous nuclei or as distinct events by a subordinate construction marked by a polyfunctional conjunction: in Nelemwa me (marking sequential coordination, action-goal dependency and complementation), in Tahitian no (marking subordination and complementation), in East Uvean mo (for coordination, additive properties, concomitant actions) or ο ('and/to' for coordination, purposive). In East Uvean, Moyse-Faurie shows that some contiguous nuclei [ex. 57, 58] with sequential or purposive meaning involving a few verbs of motion or stance (go, come, stand) might result from conjunction ellipsis of ο 'and/to'. Types of serial constructions Sequential actions and action-goal (such as 'he came and told me / he came to tell me') are often expressed by serial constructions: by core serialization in Saliba (Margetts, 6.2.) or by contiguous nuclei (Teop, Nelemwa, Pileni, East Uvean). In Mwotlap, Samoan and Tahitian only purpose or goal (but not sequential actions) is marked by contiguous nuclei. For an illustration of this in Mwotlap, see Francois [ex. 27] gengen maymay (lit. eat strong) 'eat well to strengthen your body'; [ex. 48] öl tog-yon magaysen (lit. callstay quiet-sad) 'we call (devil) to scare the kids silent'.

7.2.2. Integration of cause-effect and resultative constructions Cause-effect and resultative meanings are also frequently expressed by serial constructions across languages. They are tightly related concepts in that they constitute the boundaries of an event or action, either its origin/cause or its endpoint/result. In Nelemwa (Bril), these constructions also express the endpoint or goal of motion: [ex. 20] i caabwa-da kuut (lit. he get up stand) 'he gets up to his feet'. In Mwotlap (Francois), contiguous

Complex nuclei in Oceanic languages: an areal typology

35

nuclei may also express the temporal endpoint of the action: [ex. 21, 22] hag qöfi (lit. sit be night) 'stay until night/stay all day' or lak meyen (lit dance be day) 'dance until day / dance all night'. 7.2.2.1. Causality22 and causative constructions In the languages of this volume, causative constructions are dominantly expressed either by causative affixes (Nelemwa) or by complex clauses. In Mwotlap, with the decay of the causative affix (Frangois 3.2, fn. 13), causative constructions are expressed by complex verbs expressing action-cause(vi) - result^), in which Vj behaves as a "light" causative operator as in [ex. 23] (lit. sun shine melt ice), 'the sun melts the ice'. English shows similar constructions in which the first verb behaves as a light causative verb and the second one is a nonfinite secondary verb23; they may tighten into contiguous nuclei as in he let the blanket fall vs he let fall the blanket, he made two dancing images appear vs he made appear two dancing images. Such contiguous nuclei are licensed either by the Aux-V pattern of English, as in let fall where let behaves as a semiauxiliary, or by the head-adjunct pattern, as in pulled open in which open functions as an adverbial predicate, very much as in Mwotlap. Complex verbs are thus an intermediate stage between the highly integrated morphological expression of causation, semi-grammaticalized causative operators and multiclausal constructions. 7.2.2.2. Cause-effect and resultative constructions Notions of cause, means of action and result are closely related as in he hammered it soft. In Saliba (Margetts) [ex. 4] sikwa-he-beku-dobi (lit. poke-CAUS-fall-go.down) shows the same semantic ambiguity, '(he) poked making it fall' or '(he) made it fall by poking it', so does Mwotlap (Franc i s ) , see [ex. 9b] tit tenteri (lit. punch cry) 'he made (him) cry by punching him'. Cross-linguistically, resultative constructions may thus be expressed by contiguous nuclei as in Teop (Reinig) [ex. 48] tasu va-mate e kahi (lit. stone ADV-be.dead the dog) 'stone the dog to death', or by discontinuous complex predicates24 as in English he struck him dead, in which there is no internal role assignment by the adjectival predicate dead. Some languages distinguish direct result expressed by complex verbs and looser cause-result relations expressed by conjoined clauses (Lord 1975: 28). They are often marked by switch-subject core-layer construc-

36

Isabelle Bril

tions [SVO (s)v] with the possible evolution of V2 as a non-autonomous resultative marker. In Numbami (Austronesian, Huon Gulf family, Papua New Guinea, Bradshaw 1993: 138) for instance, former cause-result serial verb constructions have evolved into non-autonomous resultative predicates which must be used in conjunction with verbs. To a limited extent, English shows similar constructions which may tighten into contiguous nuclei [svov > SWO]: he pulled the flap open vs he pulled open the flap.

7.3. Secondary predication and serialization: some comparisons Larson (1991) pointed out strong similarities between "secondary predication" or adverbial "small clauses" in non-serializing languages and core or nuclear serial constructions, in terms of structural compression, polyfunctionality and polysemy. Various languages in this volume use contiguous nuclei with adjunct functions and meanings which are similar to English adjectival complements such as loom large, fall silent, plead innocent, run wild, spring open or similar to "secondary predication" in English. In English, "secondary predication" involves non-contiguous nuclei, the second one belongs to a closed set of adjectival predicates and mostly specifies the circumstances or manner of action with several possible readings: (i) resultative as in, he worried himself sick, she drank it empty, she beat him silly, he licked the plate clean; (ii) depictive as in she was born blind, they burnt her alive, he came home drunk, they married young, the sun shone bright, (iii) purposive as in kiss it better. The resultative and purposive types are delimiting, whereas the depictive type, which qualifies the subject or the manner of action {he scared the children stiff), is non-delimiting (Larson 1991: 201). Compare in this volume: Mwotlap [ex. 35] bow liwo (lit. bring up big) 'bring up till one is big', [ex. 36] vatne lolmeyen (lit. teach wise) 'teach to make wise', with Teop [ex. 40] hio tamaka (lit. sit be sad) 'she sits mourning', Nelemwa [ex. 21] not kuut 'watch standing', East Uvean [ex. 13] 'alu tagi 'arrive crying', Samoan [ex. 46] 'ai ola 'eat live', [ex. 47] lele tauaso 'fly blind', [ex. 50] tupu maualuga (lit. grow high). Samoan's depictive constructions can also be rephrased as adverbial clauses, as in [ex. 49] fanau mai 'o tauaso (lit. born blind) in which the short form of the progressive marker 'o marks the clause as subordinate.

Complex nuclei in Oceanic languages: an areal typology

37

7.4. Ellipsis vs structural compression Can monoclausal serial constructions be derived from underlying multiclauses? Foley and Olson (1985: 19-22) argue against this on semantic grounds, showing that, in various languages,25 serial constructions and two-clause constructions have different meanings. However, obligatory and optional serial constructions should be distinguished. The obligatory type is structural and correlates with the conceptualization of events (one event vs several events), it does not entail any loss of informative structure, and circumscribes semantic ambiguity within tolerable limits within a contiguous semantic domain. Adverbial nuclear serialization cannot be analysed in terms of ellipsis. By contrast, optional serialization resulting from conjunction ellipsis and pronoun gapping is tolerated as long as syntactic and semantic interpretation is not impaired, otherwise multiclausal strategies involving conjunctive or dependency markers are resorted to (see Mwotlap, Nelemwa, Tahitian, East Uvean). The syntactic constraints and limits imposed on nuclear serial verbs vary across languages; almost universally, different ΤΑΜ markers will trigger multiclausal constructions, whereas the same subject constraint is very variable. In Mwotlap, a macro-verb has only one subject and object, but while the subject of each component of the macro-verb may be similar or different, its object must be similar (Francois, 2.2; [L]): if the macro-verb comprises two transitive nuclei with a different object, one of them is left unmentioned if already referential, or a two-clause construction is required.

8. An overview of some evolutionary tendencies Compression and integration are cyclic processes which may affect several levels: (i) clause level (with loose or tight linkage and binding), (ii) lexical level and word form (compounding and boundness), or (iii) morphosyntactic level (cliticization and morphologization). Once a certain degree of compression and erosion is reached, it may in turn trigger lexical refection to make up for lost morphology (Vinogradov 1987) or the recourse to multiclausal constructions rather than contiguous nuclei. Generally, symmetrical contiguous nuclei are known to develop into co-lexicalized compounds whereas asymmetrical nuclei tend to grammaticalize. This is evidenced in the languages of this volume.

38

Isabelle Bril

8.1. Idiomaticization, lexicalization and lexical compounding Contiguous nuclei are commonly the source of idiomatic, co-lexicalized compounds which create new lexical entries and trigger categorial change. "[...] the degree of morphological compactness [...] reflects the degree of conventionalization of the ideas as a single unit." (Bruce 1988: 22). Most languages in this volume show such ongoing evolution (Anejom, Mwotlap, Nelemwa, New Caledonian languages, East Uvean), with various degrees of semantic predictability and transparency, functional restriction and autonomy. There are numerous such cases in Mwotlap: consider [ex. 58, 4.2] tot gal (lit. chop tell lies) 'sculpt, carve an image, portray'. Francois refrains from labelling them lexical compounds on two main criteria: morphophonological (they are distinct phonological words) and syntactic separability. Senft (this vol.) also mentions several such examples in Papuan languages 26 ; in Kalam and Tairora, Pawley (1993) describes serial verbs as an intermediate configuration between one and multiclausal structures, involving co-lexicalization (the most frequent option) or less frequently cogrammaticalization as tense-aspect markers. Manam (Lichtenberk 1983) is another such case of a language with no serial verbs but productive verb compounding, so are some New Caledonian languages (Ozanne-Rivierre & Rivierre, this vol.). In Anejom (Lynch, Section 6) the scarcity of serial verbs also correlates with the proliferation of 'deserialized' verb compounds which constitute one single stress unit in which the second item has preserved most of its modifying functions.

8.2. A cline from specialization to grammaticalization, morphologization and phonological erosion Apart from co-lexicalization, adjacency in the nuclear type provides a strong potential for specialization and co-grammaticalization in one of two possible directions, centripetal (as a verbal adjunct or transitive marker) or centrifugal (as an adposition marker) (see Rivierre, this vol.).

Complex nuclei in Oceanic languages: an areal typology

39

8.2.1. Functional specialization of verbs Functional specialization is an intermediate stage between lexical and grammatical categories or functions, varying with context and position. Yet, synchronically, both specialized and non-specialized functions coexist. In Nelemwa (Bril, 2.9.1), some verbs may have a slightly different meaning in V 2 position: ä 'go, leave' > 'from place to place', wälem 'walk' > 'quick(ly)', bwage 'return' > reversive meaning 'back'. In Mwotlap, various verbs also evolve into weak or strong causative verbs ('help', 'blow', etc.); Mwotlap also displays a verb veteg 'leave' [ex. 64] now mostly used to indicate the standard of comparison. It is the only example of such a function in this volume. Various Aktionsart verbs ('start, begin, last') and modality verbs ('hope, expect, be quiet') are intermediate categories, between full verbs and aspect or modal operators (see Mwotlap, Nelemwa, Pileni, East Uvean) and their syntactic function and status is difficult to assess (see 6.2.3 above). In East Uvean [ex. 35b] the verb 'amanaki 'hope' specializes as an aspect marker meaning 'be about to'. Categorial indeterminacy is often increased by the fact that, though they have undergone specialization, many entities preserve some residual verbal properties. Thus, deverbal adverbial adjuncts may preserve their transitive suffix in complex verb constructions (Nelemwa). In various languages (Mwotlap, Nelemwa, East Uvean), there are no clear-cut frontiers but a graded continuum between lexical units, functionally specialized entities and grammaticalized entities. This is enhanced in many Polynesian languages, particularly Samoan and Tahitian (this vol.), for words are not categorized prior to their syntactic constructions and environment and there is little or no distinctive finite vs nonfinite verbal morphology. Consequently, contiguous nuclei involve much categorial and functional indeterminacy, and their syntactic and semantic interpretation is context and position dependent.

8.2.2. Grammaticalization and morphologization Specialization may in turn lead to grammaticalization, sometimes entailing categorial change, modified semantic features, reduced or increased semantic field, development of more abstract conceptual and logical relations and meanings. It may also entail the reduction of inflectional

40

Isabelle Bril

morphology and loss of phonological form; such processes are ongoing in Anejom and in some New Caledonian languages, especially in southern languages (Ozanne-Rivierre & Rivierre this vol.) and in some Vanuatu languages; thus in Lewo (Epi Island, North and Central Vanuatu subgroup, Early 1993: 65-67), the nuclear type shows an advanced degree of incorporation, "most serialized verbs have become indistinguishable from a class of verb affixes and move towards verb incorporation"; a similar development is at work in Paamese (Crowley 1987: 82); whereas this process is much less advanced in Mwotlap. Considered in a typology of diachronic evolution, the phenomena explored by Ozanne-Rivierre and Rivierre in various New Caledonian languages illustrate the tendency that leads from isolation (serial verbs) to boundness and fusion. Paici has developed proliferating compounds in which the first element is phonologically reduced. These compounds display a high degree of morphophonological attrition, morphologization and fusion; they may even integrate a cliticized subordination marker such as mbänÄ 'so that, in order to' in a compound, as in [ex. 26] tü-mbä-tüu 'make fall, let fall from one's hands' (lit. put one's hand so that fall). The whole compound adopts the tone of the prefix. As Bisang (1996: 523-524) shows, grammaticalization is sensitive to "attractor positions", i.e. "catalyzer" slots in which linguistic items are attracted and grammaticalized. In the languages of this volume, four slots are dominant: direction/orientation, ΤΑΜ, adjunct and adposition slots. They also coincide with the most common historical paths of development of serial verbs. Various verbs tend to grammaticalize into adpositional or applicative markers, ΤΑΜ markers, conjunctive verbs or conjunctions: (i)

Cross-linguistically27, the most common verbs which grammaticalize as adpositional or applicative markers are 'give' (> benefactive), 'surpass' (> comparative), 'use' (> instrumental), 'come' (> ablative), 'go' (> allative), 'be at (> locative), 'take' (> comitative, instrumental, patient marker, see below). In Oceanic languages, Durie (1983) and Ross (this vol.) analyse the evolution of various serial verbs into adpositional coverbs or adpositions marking peripheral participants and case-roles, especially Proto Oceanic *pani 'give' > dative (benefactive), *su(ldR)i 'follow' > prolative 'about'. Thus, in Nelemwa, [ex. 50] oxuri 'follow' specializes with the meanings 'along', 'according to'.

Complex nuclei in Oceanic languages: an area! typology

41

(ii) In Oceanic languages, grammaticalization as ΤΑΜ markers essentially concerns deictic verbs of direction ('go','come'), posture verbs ('sit','stand') or Aktionsart verbs ('begin','finish'). In East Uvean [ex. 36b], lolotoga 'to last' has almost fully grammaticalized as a progressive aspect marker. (iii) In Oceanic languages, the main source of directional or deictic morphemes are grammaticalized directional or motion verbs. Ross (this vol.) shows that the Proto Oceanic verbs 'go' and 'come' have mostly grammaticalized as post-verbal clitics in V 2 position in Oceanic languages, (see Anejom Lynch, 7.2, this vol.). Their grammaticalization as preverbal clitics is much less common and stems from Proto Oceanic verb sequences such as 'go/come and...'; Teop (Reinig, 4.3) shows such an evolution. (iv) Ozanne-Rivierre and Rivierre (this vol.) analyse the gradual development of contiguous verbs involving gesture and action into compounds with a Vi shortened into a prefix and expressing manner of action. Lynch (Section 6) analyses similar cases in Anejom. (v) The grammaticalization of serial verbs into conjunctive verbs or conjunctions is analysed in 8.2.3 below. The verb 'take' shows various degrees of grammaticalization in several languages (this vol.). In Pileni (Naess, 4.3), in core serialization, 'take' as Vi has two functions: it gives prominence to a patient or specializes as an incipient or volitional marker; the difference lies in the position of the patient: between the verbs in the former [sV]0 sV2] or after both verbs in the latter [sVi sV2 O], The latter evolution offers an interesting parallel with the use of 'take' in English to mean 'start a new habit' (he took to drinking). Ozanne-Rivierre (this vol.) analyses the various types of grammaticalization and morphologization of 'take' in several northern New Caledonian languages, either (i) as a compound, with a strong tendency to fusion into one single transitive verb (as in Nemi tuut fe 'stand take') > tüe 'adopt (a child)', (ii) as an enclitic transitivizer in causative constructions (Nyeläyu pha 'take' > -va: pa-wec-va (CAUS-be born-TRANS) 'bring into the world') or (iii) as an enclitic associative case-marker28 (Nelemwa ftie 'take' > -ve (+inanimate), -vi (+animate): mu-ve 'stay with'). In Mwotlap, it is the verb tey 'hold' which grammaticalizes as a comitative case-role marker in [ex. 47] 'dance with me', as an associative case-role marker [ex. 60] 'dance with my shoes on' or to mark an even

42

Isabelle Bril

more abstract relation as in [ex. 61] 'slow with their work'. In contrast with the one-stage process 'hold', 'take' is a two-stage dynamic process (grab+hold), which might favour its common evolution as as transitiveapplicative marker or as a causative morpheme. But, if there are several instances of argument and valency adding serial constructions, there is no example of valency reduction in this volume.

8.2.3. From serial verbs to clause binding There are intermediate stages in the gradual evolution which leads from co-ranking to asymmetrical and subordinate, governed predicates belonging to different clauses (see Nelemwa, Bril, Section 4). Besides, some verbs specialize or grammaticalize as clause-linkers, conjunctive verbs or complementizers (see Nelemwa khabwe 'say' > 'that'; uya 'arrive' extends its semantic range from space to time boundary meaning 'until'). Binding and embedding increase as the matrix predicate gains more control and as the other predicate loses autonomy. This explains how former co-ranking Aktionsart or action verbs (lit. he starts he works) gradually become asymmetrical dependent constructions (as in 'he starts working'). In Nelemwa, the construction of thaaxa 'begin' is conditioned by the valency of V2: tight nuclear serialization correlates with an intransitive V2 while a two clause construction with the conjunction me is required with a transitive V2.

8.3. Cyclic evolutions: from isolation to boundness and fusion and from erosion to refection The data explored in this volume display various stages of some common evolutionary processes: compression from core to nuclear serialization, degrees of dependency from co-ranking constructions to asymmetrical dependent constructions (implying modification, governance). Saliba evidences such a process of compression; Margetts shows that sequential core serial verbs tend to evolve into the nuclear type by moving the nominal object (O) out of its VP to become the argument of a macro-verb within the basic SOV order; thus [sVmtr. OsVo] order as in [ex. 25, 26] evolve into [(S)O sVintr. sVo] order as in [ex. 29a, 30a, 31a], [ex. 29a] Osioni sVgosVseeO Ί go and see Sioni'. But such object movement cannot be ascribed to topicalization.

Complex nuclei in Oceanic languages: an areal typology

43

This volume also explores several evolutionary paths on different levels: degrees of grammaticalization (into operators, case-markers, conjunctions) and morphologization at the morphosyntactic level, and degrees of colexicalization and compounding at word level. At the extreme end of compression and fusion at word and morpheme level, phonological attrition, the decay and loss of bound morphology may occur and trigger the development of rigid word order to compensate for that loss and the refection of grammatical functions by recourse to the lexicon (see Lehmann 1986, Hagege 1993, to cite but a few). Thus, quoting Foley and Olson (1985: 51): "Phonological attrition causes syncope of segments or syllables, with the result that phonemic tone or complex vowel systems develop to compensate for phonemic distinctions being lost". [When] "valence-marking verbal morphology is lost, a new device for valence adjustment must be found. Verb serialization begins to be used in this function, provided serial constructions already exist in the language. Thus, no new pattern is being created. Rather, the function of an already existing pattern is simply extended. [...] the use of serial constructions to indicate valence increases is a function of the isolating structure of these languages." The great structural variety of serial and complex verbs in this volume and more generally in Oceanic languages could be seen as the synchronic manifestation of various phases and stages of this evolutionary process. This might also apply cross-linguistically and might account for the considerable similarities of functions and meanings of complex verb constructions across languages, in spite of the great variety of structural patterns and language-specific parameters and syntactic constraints.

44

Isabelle Bril

Oceanic languages cited and language sources. (Italicized names refer to subgroups or linkages) • Western Oceanic: - North New Guinea - Manam (Manam and Boesa Islands, Papua New Guinea; Lichtenberk 1983). - North Huon Gulf: Jabem (Bradshaw 1983, Bisang 1986, 1995); South Huon Gulf: Numbami (Bradshaw 1993). - Papuan Tip - Kilivila (Trobriand Island, Senft 1986, 1987); Saliba (Saliba Island, Milne Bay province of Papua New Guinea, Margetts 1999). - Meso Melanesian - South New Ireland subgroup: Tolai (East New Britain, Papua New Guinea; Mosel 1984). - North-West Solomonic linkage - Teop (North Bougainville, Mosel 1999, Reinig 2000). • Eastern Oceanic: - Southeast Solomonic - Toqabaqita (Malaita Island, Solomon Islands; Lichtenberk 2000). - Remote Oceanic - Southern Oceanic subgroup - North-Central Vanuatu: Mwotlap (Frangois 2001), Paamese (Crowley 1987), Araki (Frangois 2002), Lewo (Early 1993), Namakir (Sperlich 1993). - South Vanuatu·. Anejom (Lynch 2000), Lenakel (Lynch 1983), Sye (Erromango Island, Crowley 1998). - New Caledonian·. Far North of Mainland: Nelemwa (Bril 2000, 2002), Nyeläyu (Ozanne-Rivierre 1999); North of Mainland: Cemuhi (Rivierre 1986), Nemi (Ozanne-Rivierre 1979), Paid (Rivierre 1983); South of Mainland: Ajie (Leenhardt 1932, La Fontinelle 1976, Aramiou et al. 2001), Xäracüü (Moyse-Faurie 1995). - Central Pacific - Nuclear Polynesian·. East Uvean (Moyse-Faurie, to appear). - Proto Ellicean·. Samoan (Mosel and Hovdhaugen 1992); - Ellicean Outlier Pileni (Solomon Islands, Outlier Polynesian language; Naess 2000, Hovdhaugen et al. 2002). - Eastern Polynesian: Tahitian (Fortunel 1993, Lazard and Peltzer 2000).

Complex nuclei in Oceanic languages: an areal typology

45

Notes 1. 2. 3. 4.

5. 6.

7. 8. 9.

I am grateful to Alexandre Frangois, Frangoise Ozanne-Rivierre and Malcolm Ross for their insightful comments on an earlier version of this article. Or non-existent as in Manam (North New Guinea subgroup of the Western Oceanic branch, Lichtenberk 1983). As in some African and Creole languages. An event is defined as "a basic unit of thought or conceptualization [...] including causation, inception, state, deixis, termination" and "a complex event as two or more units of thought or conceptualizations" (Bodomo 1997). "Scope" is defined as the syntactic domain of a semantic operator (Byrne 1991: 210). Transitive concord is attested in various Oceanic languages, among which North and Central Vanuatu languages: Lewo (Epi Island, Early 1993) and Paamese (Crowley 1987). A neighbouring language. Frangois (2002) calls such constructions "clausechaining". In Igbo (Kwa family, Lord 1975: 25-26), nuclear serialization is not attested with stative and intransitive verbs (cited in Foley and Olson 1985: 42). The basic order is mixed: sov with nominal arguments and sVo with pronouns. There are other mixed ordering features: GEN-N; DEM-N; but N-ADJ; N-NUMERAL; V-ADV.

10. Universal 13: "If the nominal object always precedes the verb, then verb forms subordinate to the main verb also precede it. sov > s ο vsub vmain (Greenberg 1966: 11). Universal 15: "In expressions of volition and purpose, a subordinate verb form always follows the main verb [...] except in those languages in which the nominal object always precedes the veib." 11. Agent > instrument > patient > goal > locative > comitative > beneficiaiy, etc. 12. "To deal with this, we would need a theory of lexical semantics that will allow integration of the conceptual structures of such veibs to produce a single fused set of theta-roles for the combined predicate. Such a theory would offer a framework for a unified semantic treatment of the argument sharing properties of both contiguous and non-contiguous serialization." (Durie 1997: 303-304). 13. Though this is not recognized as a possible function of serialization in Baker's approach, as pointed out by Durie (1997: 308): "Baker's model demands that the overall argument structure of a serial chain be no more complex than that of the final verb in the chain".

14. Or fetch expressed as 'go take come back', or bring expressed as 'hold come', or tell expressed as 'speak give', or kill as 'hit die/be dead'.

46

Isabelle Bril

15. Formulas are "systems of knowledge that bind together pragmatic knowledge (of a world and discourse about that world) with semantic, syntactic and idiomatic-stylistic knowledge, and in which the pragmatic and semantic elements are primaiy." (Pawley 1993: 116). 16. There are only about a hundred verbal items in Kalam (Pawley 1993: 96). 17. "We speak of pseudo-complementation when we have to do with a clausal or sentential structure [...] which is treated syntactically as if it were a normal Scomplement, whereas its semantic role is not that of an S-complement but, rather, one of concomitant, resultative or purposive circumstance or event. A pseudo-complement is a suppositious sentential complement foisted on the syntax of a verb which either does not require such a complement semantically, or, if it does, does not allow for it on grounds of lexico-grammatical restrictions. [...] Veib serialization is typically ungoverned pseudo-complementation." (Seuren 1991: 196-97). 18. "If the (selectional) restrictions reside in the superordinate governing verb and the content of the pseudo-complementation is lexically unrestricted, [...] we are not facing [...] verb serialization. (Semen 1991: 203). 19. In Korafe (Papuan, Farr 1993), serialized verbs and clause-chaining coexist on a cline from tight to loose binding. 20. Core [sV(o) sV(o)] > nuclear [sW(o)]. 21. Mostly restricted to verbs of direction. 22. Causality encompasses both origin and causation. 23. Causative constructions such as I made him do it are also analysed as "predicate composition" rather than complementation (Andrews 1997: 9). 24. As in Sranan (Creole of Surinam) Kofi naki Amba kiri (lit. Kofi hit Amba kill) 'Kofi killed Amba', which Baker (1989: 524) analyses as a case of objectsharing. 25. Kwa languages of West Africa (Yoruba, Anyi - Ivoiy Coast), Lahu (TibetoBurman) and Yimas (Papuan language, Sepik region, Papua New-Guinea). 26. For Alamblak, see Bruce 1988. 27. See Foley and Olson (1985: 48); Seuren (1991: 203). For Oceanic languages: Pawley 1973, Lichtenberk 1985, Hamel 1993. For Asian languages: Li & Thompson 1974a, Hagege 1975, Clark M. 1978, Peyraube 1988, Bisang (1995: 534, 570). For West African languages: Ansre 1966, Lord 1973, 1993, Givon 1975, Heine & Reh 1984. 28. 'Take' may also grammaticalize as an applicative-transitive marker with instrumental case-role, as in Dagaare (North Ghana; Bodomo 1997).

Complex nuclei in Oceanic languages: an areal typology

47

Appendix Questionnaire

a)

Productivity: are there any reasons for the scarcity or proliferation of such constructions? b) Are there obligatory vs optional serial constructions; with what type of verbs/predicates and what semantics? c) Are there various (sub)types of complex predicates or serial verb constructions (symmetrical/multi-headed or asymmetrical) ? what are their functions? d) What are the syntactic or semantic categories of predicates/verbs prone to appear in serial verbs /complex predicates constructions? e) Is there any specific ordering constraint on such complex predicates/serial verbs? (syntactic, iconic, logico-semantic, etc.). What are the types of structural pattern(s) (number and type of positions/slots)? f) Overview of their functions (syntactic, semantic, discourse, etc.). g) What is the nature of the serial verbs/complex predicates? Do they evidence argument-sharing, fusion or composition of argument structures? Do they result from deleted constituents? h) What are the formal criteria that help discriminate juxtaposed strings of verb from serial verb constructions? (polarity, scope of ΤΑΜ markers, morphosyntactic properties and constraints, autonomy, semantic restriction, etc.). i) What are the available criteria to discriminate serial verb/complex predicate constructions from specialized or grammaticalized verbs? j) Do serial constructions fill a structural or a morphosyntactic gap or do they coexist with other morphological markers and constructions? If so, what is the difference of meaning or function? ( c f . valency increase, semantic roles, animacy hierarchy, comparison, causation, etc.)? k) What relationships do serial verbs or complex predicates entertain with coordination, subordination, complementation and other dependency-marking systems? What kind of choice or constraint is there in the selection of one of these strategies? Can serial constructions have complementizing function? 1) Does the language studied support or contradict typological predictions concerning serial/complex constructions? m) Do compound and serial verbs coexist in the language, with the same frequency? What are their distinctive criteria? (stress, prosodic and phonological criteria). Do they make up one or several phonological entities? Do they have specific functions?

48

Isabelle Bril

Diachronic perspective n)

o)

Is there any evidence of language evolution leading to such forms and constructions? Do serial constructions seem to be an unstable and transitional stage and mode of structuration? Are there verb types which are more prone to lexicalization or grammaticalization?

Cognitive and discourse perspective p)

q) r)

Do serial constructions show some choice between various clause conjoining strategies and dependency marking (from one to many clauses)? What are the differences? What is due to syntactic constraints and to discursive strategies of compression? If there is a choice, what is the kind of strategy involved? what are the semantic or discourse implications of such a choice and strategy?

What do we really know about serial verb constructions in Austronesian and Papuan languages? Gunter Senft T h e task of h a r m o n i z i n g a general account of verb serialization is a very m a j o r one w h i c h is hardly begun. (Durie 1997: 320)

Abstract This paper first presents a number of serial verb constructions (SVCs) found in Papuan and Austronesian languages and then raises the following questions: What are SVCs? What about typological research on SVCs? What about the syntactic description of SVCs? Is there a comprehensive definition of SVCs? Which types of SVCs do we find? What about the functions SVCs fulfill? Which kind of verbs constitute SVCs? What about the order of verbs within SVCs? What kind of lexicalization processes can we observe in SVCs? What is expressed as an event in a SVC and how is it expressed? Are there any language - and/or culture-specific rules for the combination of verbs in these constructions? Can we infer from SVCs to language - and/or culture-specific conceptualizations of events? The paper ends with a brief outline of the direction for a new approach to research on SVCs which may lead to answers for at least some of these crucial questions.

1. W h a t are serial v e r b constructions? - Features, types a n d functions 1 Speakers of Taba, the Austronesian language spoken on the islands M a k i a n , Moti and K a y o a in Indonesia (northern M o l u c c a s ) m a y describe their successful " h u n t i n g " of a mosquito as follows ( B o w d e n 1997: 339): (1)

Ν pun bobay npake sandal. n-pun bobay η-pake sandal 3.Sg-kill mosquito 3.Sg-use thong ' H e killed the mosquito with a thong.'

50

Gunter Senft

When the Paama islanders in Vanuatu kill a pig, they refer to this event in their Austronesian language Paamese as follows (see Crowley 1987: 43): (2)

Kail amuas vuas emat. kaile a-muasi vuasi emate 3.PI 3.Pl-real-hit pig 3.Sg-real-die 'They hit the pig and it died.'

In Yimas, the Papuan language of the Yimas people living at the Arafandi river in Papua New Guinea, the activity of making firewood can be described in the following way (Foley 1991: 331): (3)

Yanparkmpikapikmpiwarkt. ya-n-park-mpi-kapik-mpi-wark-t V-Pl 0-3Sg A-split-Seq-break-Seq-tie-Perf 'He split them, broke them into pieces and tied them together.'

Speakers of the Papuan language Kalam in the highlands of Papua New Guinea refer to the fact that a man has brought firewood to his house with the following utterance (Pawley 1993: 95): (4)

Β ak b man 'The

am mon pwk d ap ayak. ak am mon p-wk d ap ayak that go wood hit-break get come put-3.Sg.-Past man fetched some firewood.'

And in Kilivila, the Austronesian language of the Trobriand Islanders of Papua New Guinea, it is not uncommon to hear the question " A m b e y a l " 'Where (are you going) to?' when people are leaving their village. An adequate answer to such a question (which is ritualized as a form of greeting) may run: (5) Bala bakakaya baka 'ita basisu bapaisewa batai waga kevau} ba-la ba-kakaya ba-ka 'ita ba-sisu ba-paisewa l.Fut-go l.Fut-bath l.Fut-return l.Fut-be l.Fut-work Ί will go I will have a bath I will come back I will stay (in the village) I will work' Ba-tai waga ke-vau. 1 .Fut-cut canoe CP.wooden-new Ί will cut my new canoe.' Any reader of these five illustrative examples from different Austronesian and Papuan languages will notice immediately that all these sentences contain more than one verb. In the sentences (1) and (2) we find two verbs, sentence (3) is a verbal expression that consists of three verbs, sentence (4)

Serial verb constructions in Austronesian and Papuan languages

51

consists of 6 verbs (only the last of which (ay) is inflected), and sentence (5) also consists of 6 verbs (which are all inflected). 3 That such sentences with more than one verbs exist in the languages of the world is nothing new. In 1875, Christaller (1875: 69-73, 143f.) already discussed this phenomenon in his grammar of Twi (see: Sebba 1987: 5, Seuren 1991: 193), and Westermann, in his grammar of Ewe, pointed out ... a peculiarity of Ewe is that we often find a row of verbs one after the other. The chief features of this are that all the verbs stand next to each other without being connected, that all have the same tense or mood, and that in the event of their having a common subject or object, these stand with the first, the others remaining bare: should a conjunction stand between two verbs, the subject and object must be repeated ... In English these consecutive verbs are partly rendered by composite sentences. But very often several Ewe verbs may be expressed by a single verb in English. The explanation for this is that the Ewe people describe every detail of action or happening from beginning to end, and each detail has to be expressed by a special verb: they dissect every happening and present it in its several parts, whereas in English we seize on the leading event and express it by a verb, while subordinate events are either not considered or are rendered by means of a preposition, adverb, conjunction, or a prefix on the verb. (Westermann 1930: 126)4

In 1914 Hugo Schuchardt noted similarities with respect to these verb constructions between Suriname Creole on the one hand and Ewe on the other hand (see Muysken, Veenstra 1994: 289). The honour of being the first to describe the presence of more than one verb within a sentence for an Austronesian language - namely Jabem - is due to Otto Dempwolff (Dempwolff 1939; see also Bradshaw 1983; Bisang 1986). In 1957 Jan Voorhoeve used the technical term "verbal chains", however, according to Voorhoeve (1957), Sebba (1987: 2) and Zwicky (1990: 2) it was Stewart (1963) - again a scholar of African languages - who coined the term "serial verbs" to describe this phenomenon (see also Seuren 1991: 193). So far, most research on serial verbs and serial verb constructions (from now on abbreviated as "SVC") has been done on African languages and on pidgins and Creoles; however, s v c s are also to be found in Hmong-Mien, Mon-Khmer, Sino-Tibetan, Tai-Kadai, Semitic and Central-American languages as well as in Japanese. It should be noted, though, that the phenomena discussed as SVCs within the last mentioned languages are hotly and controversially debated and differently described and labelled. Linguists like Sebba (1987: 3), Seuren

52

Gunter Senft

(1991: 194), Bisang (1993: 71) and especially Pullum (1990: 219-222) also raise the issue whether there are SVCs in languages like English (or Dutch). Pullum refers to constructions like "Go get the paper.../Come get the paper.../Help get the paper.../Come go east with us" as "intransitive quasiserial verb constructions". There are only a few studies on SVCs in Austronesian and Papuan languages, and Durie emphasizes rightly that these languages are largely underrepresented in the present linguistic literature on the phenomenon: The range of construction types and grammatical properties that these languages encompass is rather greater than has been appreciated in much of the theoretical literature on serial verbs, which has tended to take as representative patterns occurring in a few restricted areal contexts, e.g. from West African and Creole languages, to the exclusion of data from Southeast Asia (except for a nod to 'Chinese') and the Pacific. (Durie 1997: 29If.).

This bias in typological studies on this phenomenon may also explain why there is no general definition of SVCs. Thus, Menick (1996: 41) laments that "there is little agreement as to how the phenomenon of serial verbs is to be defined", Comrie (1995: 25) points out that there is a "considerable controversy concerning the precise characterization of the serial verb construction", and in 1987 already Sebba pointed out that "... it is not at all clear that all authors are referring to the same thing when they speak of "serial verbs". Very few of them are actually explicit about what they mean by the term, usually applying it fairly indiscriminately to constructions in which there is a sequence of the form V NP V NP or V NP V ..." (Sebba 1987: 1). All in all we have to agree with Lord's (1993: 1) summary of the state of the art with respect to SVCs: Defining serial verb constructions is a sticky business". 5 Another reason for the difficulty in defining SVCs is certainly "the lack of a generally accepted framework in which to describe and compare these systems" (Crowley 1987: 36). Lane (1991: 31), for example, notes that linguists "have applied the term 'serial verb construction' to entities in a wide variety of languages, using a diverse set of analytic frameworks. Different analyses draw the boundaries between SVCS and other types of construction at different places". For many grammar theories SVCs are a special challenge: "What makes serial verbs interesting is the fact that they cut across established categories" (Zwicky 1990: 10). SVCs are especially difficult to deal with in theories "where sentences were expected to have exactly one main verb" (Sebba 1987: 6). In his excellent study on the problems of describing SVCs syntactically, Durie (1997: 294-320) has

Serial verb constructions in Austronesian and Papuan languages

53

shown that despite many attempts within the different variants of Chomsky's theory of grammar, representatives of GB theory (see e.g. Baker 1989; Dechaine 1993) have also failed to describe adequately the phenomenon of SVCs. Durie bases his prolegomena for the description of the grammatical structures of SVCs on his own research (see e.g. Durie 1988) and on the findings presented by Bruce (1986; 1988), Crowley (1987; 1990), Lane (1991), Pawley (1987; 1993), Sebba (1987), Foley and Van Valin (1984) and Foley and Olson (1985). At the moment his approach seems to be the most adequate for a syntactic description of SVCs. However, even Durie in his seminal contribution cannot give a general definition of what SVCs are. Like many other linguists he refers to a number of "key characteristics" (Durie 1997: 291) or "common features" (Lane and Pawley 1992: 5) or - (in a very cautious way) - "criteria" (Comrie 1995: 25), that are important for any description of SVCs. For Durie, these are the "key characteristics" of the SVCs: ... a single serial verb complex describes what is conceptualized as a single event: this is repeatedly reported to be a clear intuition of native speakers, and can be demonstrated through semantic analysis. It follows from this that a serial verb complex can often be best translated into a non-serializing language using a single, mono-verbal clause. - the serial complex has shared tense, aspect, modality, and polarity: this is often reflected in a single morphological realization of these operators ... or in obligatory concord across the verbs ... - serial verbs 'share' at least one and possibly more arguments. - one verb is not embedded within or as a complement of the other. - intonational properties of a clause with serialization are those of a mono-verbal clause (Givon 1990; 1991a, b). - the complex takes only one subject/external argument. - when serialization results in a complex of more than two arguments, the configuration of arguments corresponds closely to the kinds of configurations of arguments + adjuncts found for single clauses in non-serializing languages. - there is a very strong diachronic tendency to lexicalization and grammaticization of the meaning of serial complexes: this can involve treating the whole serial complex as a single lexical(ized) item, or 'demotion' of the meaning and grammatical status of one of the verbs to that of a modifier or case-marker. (Durie 1997: 291). 6

These features are characteristic of, and can be observed in, all SVCs and they certainly have to be incorporated into, or at least considered for, any definition of the phenomenon. However, some of these features are problematic themselves, especially the notion of "single event": What is a 'single event', how is it constituted, perceived, conceptualized, expressed

54

Gunter Senft

and reported? Can we define this notion, for example, as "the answer to the question: 'What happened?'" (Enfield, personal communication)? Moreover, it is also possible to differentiate different types of SVCs. Crowley (1987: 38-40) - "according to the nature of the relationships that hold between the nominal arguments associated with each of the verbs in question" (Crowley 1987: 38, 48) - proposes the following five subcategories for SVCs:

1.1. "Same-subject serialization" This is probably the most common and widespread type of SVCs; in this type "there is identity between the two subjects of the serialized verbs" (see examples (1) and (3)-(5) above).

1.2. "Switch-subject serial verbs" or "causative verbs" In this type "there is identity not between two subjects, but between the object of the first verb and the subject of the following verb" (see example (2) above).

1.3. "Multiple object serialization" This type is relatively rare in the languages of the world. We observe "'same-subject' or 'switch subject' conditions of identity between the subjects of the serialized verbs, each of which is transitive and each of which has its own object". Example (6) from Paamese and example (7) from Barai, a Papuan language the speakers of which live in the vicinity of Popondetta in Papua New Guinea, illustrate this type: (6)

Inau namun sin dal oai. (inau na-muni siini

dali

oai)

lsg

3sg-rea/-accompany

water

1 sg-real-drink gin

Ί drank gin with water'. (Crowley 1987: 39). (7)

Fu

burede

ije

he

bread

def knife take

sime

abe

ufu. cut

'He cut the bread with the knife.' (Foley, Olson: 1985: 44).

Serial verb constructions in Austronesian and Papuan languages

55

1.4. "Ambient serialization" This type is defined by Crowley (1987: 49) as "a construction in which a verb is serialized to another verb, but in which there is no specific referent associated with the subject of the serialized verb, and the verb simply describes a general predication". Example (8) from Paamese illustrates this fourth type of SVCs: (8)

Kihulin ato kail hemal. (ki-huli-nV atoo kaile he-malu) 2sg-dis-count-comm/obj chicken pi 3sg-dis-be.correct 'Count the chickens correctly'. (Crowley 1987: 40).

1.5. "Conjoined participant serialization" This type is only briefly discussed in Crowley (1987: 48). On the basis of this paper, Early (1993: 68, 89) proposes the term "conjoined participant serialization" and defines this type "for the situation where the subject and the object of the first verb become the combined subject of the second". Example (9) from Lewo, the Austronesian language of the Epi-Islanders in Vanuatu, illustrates this fifth type of SVCs: (9)

Ne-mio-la me-pano. lsS-with-3pO IpexclS-R.go 'We went together.' ( Ί with them we went') (Early 1993: 89).

On the basis of research by Foley and Van Valin (1984: 189-208) and Foley and Olsen (1985: 33-38), Crowley proposes yet another differentiation that is important for the description of SVCs. This differentiation does not refer to the argument structure of verbs, but rather to "the actual "layer" of the clause at which the serialization takes place" (Crowley 1987: 40). Crowley differentiates three such 'layers' that have specific 'operators': "The innermost layer is the nucleus, and nuclear operators include items such as aspect. The next layer is the core, and the operators at this level include the obligatory nominal arguments associated with a particular verb. The outermost layer is the periphery, and peripheral operators typically refer to things like the temporal and spatial setting on an event. It is argued that verb serialization can take place at either the nucleus or the core of the clause" (Crowley 1987: 40f.). Thus we differentiate between "Core Layer Serialization" (see example (2) above) and "Nuclear Layer Serialization"

56

Gunter Senft

(see example (3) above; see also Pawley 1993: 120). Crowley points out that there is a tendency for languages with an s o v word order pattern to prefer "Nuclear Layer Serialization" and a tendency for languages with SVO word order pattern to prefer "Core Layer Serialization" (Crowley 1987: 42). 7 In the various languages SVCs fulfill a number of grammatical functions; Early summarized some of the more important functions of SVCs in the following list: a. aspect b. temporal, spatial, or psychological movement or distance or location c. logical relations like cause-and-effect, and purpose d. various semantic roles, including Instrumental, Dative, Benefactive, Locative, Manner, Comitative, Accusative, Direction, Comparison... (Early 1993: 67f.). In what follows I illustrate these functions with a few examples: ASPECT (example from the Papuan-language (10)

Dani):

Wat-h-y-lak-ytyk. hit-REAL-DEP-stay-1SG PAST Ί was hitting him.' (Foley 1986: 144).

LOCATION (example from the Austronesian language Lewo): (11)

0-pa 0-tapolou 0-teke pulu-pälu. 3sS-go 3sS-hide 3sS-stay hole-creek 'He went and hid in the creek-bed.' (Early 1993: 68).

BENEFACTIVE (example from the Papuan-language (12)

Kalam):

Passkoy yp ag n-a-k. girl me (Obj) say give-3sg-PAST 'The girl told/confided (it) to me'. (Lane: 1991: 56).

MOTION (example from the Austronesian language Taba): (13)

Nhan ntono ni dawalat. n=han n=tono ni dawalat 3sg=go 3sg=look.at 3sg.POSS girlfriend 'He's gone to see his girlfriend.' (Bowden 1997: 354).

Serial verb constructions in Austronesian and Papuan

languages

PURPOSE (example from the Austronesian language Kilivila): (14)

Εma egimwali tokwalu. e-ma e-gimwali tokwalu 3.-come

3.-bargain.for

carving

'He came to sell carvings.' (Senft 1986: 41). CAUSATIVE (example from the Papuan-language Yimasj: (15)

Na-ka-tal-kwalca-t. 3sg0bj-1 sgSubj-hold-arise-PERF

Ί woke him up'. (Foley 1986: 154) RESULTATIVE (example from the Papuan-language Alamblak) (16)

Tat-noh-me-an-r. hit-die-R:PST-1 sg-3sgM

Ί killed him (by hitting him).' (Bruce 1986: 22) ABLATIVE (example from the Austronesian language Paamese): (17) Namual naumai en leiai. (na-muali nau-mai lsg-real-walk

lsg-real-come

en

leiai)

sp

bush

Ί walked from the bush.' (Crowley 1987: 53). ALLATIVE (example from the Austronesian language Paamese): (18) Namual namul en leiai. (na-muali na-mule lsg-real-walk

1 sg-real-exist

en

leiai)

sp

bush

Ί walked into the bush'. (Crowley 1987: 53). COMITATIVE (example from the Austronesian language Lewo): (19)

0-to

ma

a-mio

kana

3sS-stay

Cont

3pS-with his

mama

lala.

mother PL

'He stayed with his mother and the others.' (Early 1993: 69). MODALITY (example from the Papuan-language Hua): (20) Ke

hu-ko-mana.

talk do I-see-OTHER INCONSEQUENTIAL

Ί tried to talk (but to no avail).' (Foley 1986: 152, Haiman 1980 147).

58

Gunter Senft

In this section we have discussed different forms, types, and some of the main functions of SVCs. In what follows, we will take a closer look at the individual verbs within a SVC. Here we are first interested in the kind of verbs that are serialized and then in the position of the verbs with respect to their arguments and their internal order within the SVC. At the end of the section we will propose a new approach to the analysis of SVCs.

2. Event report and event conceptualization in SVCs: A proposal for a new approach to the analysis of serial verb constructions Besides the formal grammatical criterion "transitivity", semantic facts play a crucial role in answering the question which verbs can be found in SVCs (Foley and Olson 1985: 40ff). On the basis of their research on the phenomenon, Foley and Olson come up with a "serialization hierarchy" (Early 1993: 68), which is succinctly summarized by Crowley as follows: ... the verbs that are most frequently encountered in serial constructions in languages of the world are the basic motion verbs (e.g. come, go), which are followed by other active intransitive verbs (e.g. wander, disappear, crawl) and intransitive posture verbs (e.g. stand, lie), followed by any other active intransitive verbs (e.g. go hunting, speak, jump, etc.), and finally followed by the class of transitive verbs, which are therefore the verbs that are least liable to enter into serial constructions with other verbs. (Crowley 1987: 42). 8

With respect to the relationship between serialized verbs and their arguments, Durie observes the following: ... the phenomenon 'verb serialization' can manifest itself in two distinct patterns: either the verb series acts like a single verb for the purposes of placement of arguments - giving in an SVO language SV...V(0) type sequences - or an argument is located in the usual position with respect to the first verb that introduces it - giving in an SVO language S V ( 0 ) V ( 0 ) V ( 0 ) ... sequences. Some languages have either one pattern or the other, and some ... show both patterns. (Durie 1997: 307). 9

In languages like Kalam, SVCs can consist of up to nine or ten verbs (Lane 1991: 1); in this Papuan language constructions with five or six verb-stems are nothing special (see also Foley 1986: 113; 1997: 383). At first sight the sequence of verbs within the SVCs - especially within such complex SVCs - seems to follow "iconic" principles. Obviously

Serial verb constructions in Austronesian and Papuan languages

59

speakers of languages with SVCs want to describe an event (or a state) as minutely and as exactly as possible (see Escure 1991: 187, Senft 1986: 39). 10 In doing so, an event (or an action or a happening or a "Handlung" in Westermann's (1907: 94f.; 1930: 126) terminology) is classified, broken down, subdivided, ordered and arranged into its components and each of these components is expressed with a specific verb. In many languages this form of event report seems to follow specific rules, and these rules are constitutive for the well-formedness of the event report. Besides lexicalization processes that affect the combination of specific verbs, we also observe specific sequences of verbs that are crucial for the adequate expression of certain event types (Durie 1997: 322ff). These combination rules for verbs within SVCs seem to be largely culture dependent and culture specific. In what follows I will briefly illustrate these observations: For the Papuan language Kalam, Pawley (1993: 97f.) lists the following expressions that consist of a generic verb and of one or two preceding verb stems with one or more nominal or adverbial complements. In these expressions the verbs in combination are lexicalized so that they form new expressions: (21)

wsn sleeping

kn recline

amgo

'go to sleep, drift off to sleep'

OBJ

nb nnconsume perceive

'taste something'

OBJ

d untouch perceive

'feel something (by touching)'

OBJ

d hold

am go

'take something'

OBJ

d hold

apcome

'bring something'

OBJ

kby amleave go

'leave, abandon something'

In Alamblak, a Papuan language spoken in the East-Sepik area of Papua New Guinea (Bruce 1984), we find the following lexicalization: (22)

tu-fenah throw-arrive

'spear' (Bruce 1988: 33f.) and in Vanimo, a Papuan language spoken in the East-Sepik area of Papua New Guinea, we find the lexicalization:

60

Gunter Senft

(23)

Ηύη ha. 3SM:drink 3SM:go 'He laughs.' (Ross: 1980: 92).

In the last example the meaning of the lexicalized form cannot be deduced from the meanings of the two individual verbs. These lexicalized forms are especially problematic for a grammatical analysis of SVCs, because "one cannot assume that the serialized verb will have the same argument structure in serialization that it has when used alone" (Durie 1997:324). The Papuan language Kalam, however, shows yet another, even more complex pattern of productivity for SVCs. Pawley observes that event reports in Kalam follow general report schemata. These schemata can be represented in the following frame consisting of five event sequences (see Durie 1997: 325): I: movement to scene II: action at scene III: movement to next scene (taking something) IV action at scene V: movement away from scene Based on Pawley's (1987: 349) data, Durie (1997: 325f.) illustrates these schemata with the following sentence: (24)

Β man

tap sy stuff illegally

d-p-0 get-PERF-3SG

d-am get-go

kot g-y court do-SS:PRIOR

kalabws ay-p-wn. jail put-PERF-IPL 'We have put the thief in jail.' This event is subdivided into the event components "getting the t h i e f ' (II), "going to court" (III) and "putting chief in jail" (IV). Pawley (1987, 1993) Lane (1991) and Lane and Pawley (1992) present many such stereotyped schemata and "speech formulas" (Lane 1991: 72) that are defined as a construction type whose lexical content is partly fixed and partly variable. It is a much more complex bundle of elements than a lexical unit. Besides being a conventional pairing of form and meaning associated with a particular grammatical category, a speech formula is indexed for occurrence in particular discourse contexts and discourse functions, can be varied according to formula-specific and general grammatical and idiomaticity constraints, and is spoken with a particular intonation and rhythm. (Pawley 1997: 24)

Serial verb constructions in Austronesian and Papuan languages

61

This definition and the examples presented by Pawley and Lane show that the term of the "event" that is expressed with SVCs cannot be discussed if detached from the individual language and its speakers - and this means that analyses of SVCS cannot be purely syntactic, but have to be also - and maybe even above all - semantic and pragmatic (see Lord 1993: 239). Durie (1997: 327f.) expresses this demand in the following way: It is essential that my conception of event-type is based on what speakers habitually treat as unmarked complex events, so a rigorous investigation of the semantics and discourse properties of such sequences is required to develop a more convincing account for data of this kind. Thus, following Bruce (1988: 28) and Pawley (1997) we have to research what a speech community conventionalizes verbally within the frame of a SVC as an "event". Only then is it possible to decide (and to describe) whether a certain verb sequence within a SVC can be realized and will be accepted by the speech community, because it verbalizes an event type which is plausible and reasonable for the speakers of the respective language (see Enfield 1998, 2000b). As already stated above, the norms that mark what counts as an event within a speech community are certainly culture specific and culture dependent. Thus, SVCs ask for anthropological linguistic analyses. It is true that in many cases the sequence of verbs within SVCs is iconic. It is also true that in many cases parts of an event are listed in their temporal sequence within an event report (see Foley 1997: 384; Lane 1991: 176, Lord 1993: 237) - this is especially so for parts of an event that have a cause-effect relationship." But there are also many cases where iconic principles are completely irrelevant for the sequence of verbs within a SVC. This is illustrated with the following two examples from Kilivila, an Austronesian language spoken on the Trobriand Islands in Papua New Guinea: (25a) Ekebiga elivala makala... e-kebiga e-livala makala 3.-speak 3.-say like 'He said the following ...' (25b) Elivala ekebiga makala... e-livala e-kebiga makala 3.-say 3.-speak like 'He said the following ...'

62

Gunter Senft

These SVCs mark in a highly idiomatic way that the speaker's utterances that follow these introductory formulae represent verbatim the utterances of another speaker. I have no idea whether these two variants otherwise differ with respect to their semantics. 12 To find out how events that are expressed in SVCs are actually conceptualized by speakers of this language, we need discourse semantic/pragmatic and anthropological-linguistic analyses. This requires a completely new approach to the analysis and description of SVCs. Such an approach considers not only the syntactic facts but also the semantic-pragmatic as well as the cultural conditions of an event report and especially the common and the different event conceptualizations that are shared by the respective speech communities. Durie has outlined the direction for such a new approach: ... I wish to suggest that the lamp of syntactic analysis can only probe some protuberances of verb serialization, without allowing one to gain a full impression of the phenomenon. I propose that models of lexical conceptual structure and event-hood which are in any case needed to account for the properties of verbs in non-serializing languages, will also need to be deployed to deal properly with verb serialization. My main theme is that non-serial verbs and serial verb complexes are subject to ... many of the same constraints on conceptual structure, and also on syntactic linking ... An advantage of this approach ... is that we can hope to achieve some clarity about which properties of serialization are manifestations or projections of semantic structure, culture-specific constructions of eventhood, and tendencies of grammaticization and lexicalization ... (Durie 1997: 349). 13

Abbreviations A excl DEP Fut Hab CP 0 obj

Subject of a transitive verb exclusive dependent Future habitual Classificatory Particle (see. Senft 1996) Object of a transitive verb Object

Perf Pl/pl R/real S Seq Sg/sg/s V

Perfective Plural Realis Subject sequential Singular Verb

Serial verb constructions in Austronesian and Papuan languages

63

Notes 1.

2.

3. 4.

First of all I would like to thank the "Deutsche Forschungsgemeinschaft" (German Research Society) and the MPI for Psycholinguistics for financing and supporting my project on "Event conceptualization, event report, and serial verb constructions in Austronesian and Papuan languages" (DFG: SE 473/3-1). I would also like to thank the participants of "The Third European Meeting on Oceanic Linguistics", 9th-10th March 2001 at LACITO-CNRS in Villejuif/Paris, Teriy Crowley, and my colleagues at the MPI in Nijmegen, especially Miriam van Staden, Melissa Bowerman, Felix Ameka, Jürgen Bohnemeyer, Nick Enfield, and Steve Levinson for helpful, constructive and interesting comments and discussions on the topic of sves or rather - according to Nick Enfield (2000a) and Felix Ameka (2001) - multiverb constructions (MVCs). For further information on Kilivila and its orthography see Senft (1986). On ritual communication in Kilivila see Senft (1987). Evidence for the fact that the sentence quoted must be understood as a serial verb construction and not as parataxis comes from repairs that speakers of Kilivila make when producing sentences like this. If they realize that they have made a mistake in the sequence of veibs within a sve, they start their repair by going back to the veiy first verb in the serial construction. For a discussion of how many verbs can be found and produced within such constructions see below. This reads in the German original version as follows: "Eine Eigentümlichkeit des Ewe besteht darin, daß es gern eine Reihe von Verben unmittelbar aufeinander folgen läßt... Im Deutschen werden diese aufeinanderfolgenden Verba zum Teil durch zusammengezogene Sätze oder Satzgefüge wiedergegeben. Sehr oft können aber auch mehrere Verba des Ewe im Deutschen durch ein einziges ausgedrückt werden. Der Eweer beschreibt nämlich jede Handlung, jeden Vorgang in allen Einzelheiten von Beginn bis zum Ende and drückt jede solche Einzelhandlung durch ein besonderes Verbum aus; er zerlegt jede Handlung in ihre einzelnen Teile and bringt jeden Teil für sich zur Darstellung, während wir im Deutschen nur die Haupthandlung herausgreifen and sie durch ein Verbum ausdrücken, während alle Nebenhandlungen entweder ganz unberücksichtigt bleiben oder mittels einer Präposition, eines Adverbs, einer Konjunktion oder einer Vorsilbe des Verbum etc. wiedergegeben werden." (Westermann 1907: 94f.).

5.

See also Durie (1988: 3), Foley, Olson (1985), Foley (1997: 382f.), Givon (1990: 19), Lane (1991:v, 24, 31, 36ff) Lane, Pawley (1992: 5ff), Schüler (1990: 34ff), Sperlich (1993: 95f.), and Zwicky (1990: 2). The fact that "sve" is not an empirically grounded category may also explain these difficulties as Nick Enfield pointed out to me. The sequence of verbs within sve and the sequence of their arguments is discussed below.

64

6•

Gunter Senfi

For the "polarity" criterion see also Lane (1991:48): "It is typical in svcs for negation to be marked only once, and to have scope over the whole construction"; however, see also Dol (1996: 35) who states that for Maybrat: "the scope of the negator is unclear".

7.

8.

9.

10.

11.

12.

13.

Dune refers to Givon's research on the importance of the intonation contour and pauses for svcs. Givon first published his results on svc in 1990. The two papers which he published in 1991 (Givon 1991 a & b) are - with minor and marginal variations - identical with his article published in 1990. For a list of language specific characteristics of svc, see Lane (1991: 35f.); see also Dol (1996: 2If.) and Menick (1996: 42f.). On "Nuclear Layer Serialization" and its differentiation from "Compounding" see Crowley (1987: 59-62). By the way, it is rather plausible to find these preferences for languages with the word order patterns sov and svo: sov languages prefer nuclear layer serialization (sov[vvv...]), and languages with the svo word order pattern prefer core layer realisation (s v(o) [v(o)v(o)v(o)...]). See also the diagram in Crowley (1987: 69) that summarizes this hierarchy as follows: "1. Basic motion intrans. 2. Posture/active intrans. 3. Stative/process intrans. 4. Other intransitives. 5. Transitive". Durie discusses in detail the topic of svc and argument structure. Even a short discussion of this topic here would go far beyond the scope of this paper. Thus, 1 refer the interested reader to Durie (1997) and also to Crowley (1987), Foley and Olson (1985) and Bisang (1992). In what follows the term "event" is used as the most general cover term including states, actions, etc. (see Durie 1997: 320 fn.21). For the expressions "event" and "event report" see also Pawley (1993: 109). Durie (1997: 331-336) also presents and illustrates the following types of svc: "Causative Serialization, Goal/Benefactive Serialization, Motion Serialization, Instrumental Serialization". The first variant of this example is documented by many tokens in my Kilivila data corpora; examples for the second variant are rather rare. Besides such "synonymous serializations" Durie (1997: 336-339) lists the following types of serialization where iconic principles seem to be irrelevant for the sequence of verbs within the svc: "Coincident motion or posture serialization, Manner serialization, Comitative serialization". Miriam van Staden, Alex Dukers, and I have started a research project (funded by the German Research Society - DFG) on svcs at the MPI for Psycholinguistics in Nijmegen. We compiled a rather comprehensive questionnaire and have been asking colleagues via the Internet (Linguist List, Austronesian List, Papuan List) to please fill it in. The questionnaire can be found (and filled in) at the following web-page: http://www.mpi.nl/world/serial-verb/quest/lst-quest.html.

Core-layer junctures in Saliba Anna Margetts

Abstract Verb serialization is a well-known feature of Oceanic languages and commonly two types, nuclear v.v core-layer serialization, are distinguished. In Saliba, an Oceanic language of Papua New Guinea, nuclear-layer serialization constructions are well attested and a number of subtypes can be distinguished. The existence of core-layer serialization is less obvious. In this paper I investigate whether the language also has core-layer serial verbs and I address some of the problems in defining such constructions.

1.

Introduction

Verb serialization is a well-known feature of many Oceanic languages and commonly two types, nuclear and core-layer serialization, are distinguished. While structurally different, both types are documented to express similar kinds of functions across languages. Individual languages may show both types of constructions, such as Paamese discussed by Crowley (1987). In Saliba, an Oceanic language of Papua New Guinea, we find that nuclear-layer serial verbs are a prominent feature of the grammar (cf. Mosel 1994, Margetts 1999). They are very frequent and a number of different subtypes can be distinguished. The existence of core-layer serial constructions is less evident in Saliba and previous accounts make no reference to them. The question is whether there are no core-layer serial verbs in the language or whether they are just less prominent or simply less easily identified. In this paper I investigate whether Saliba does have corelayer serialization and I discuss a number of problems one encounters when addressing this question.

66

2.

Anna Margetts

Language background

Saliba is an Oceanic language of Papua New Guinea. It is a member of the Suauic family of the Papuan Tip Cluster, spoken in the Samarai area in Milne Bay Province. There are about 1,000 speakers living on Saliba Island, neighbouring islands, as well as on parts of the adjacent mainland and in the provincial capital Alotau. On Logea Island a very closely related dialect is spoken and together the number of speakers for Saliba and Logea is estimated at 2,500 (Oetzel and Oetzel 1997). Milne Bay English is the lingua franca of the area (rather than Tok Pisin as in most parts of PNG) and it is also the language of school education. Most children acquire Saliba as their mother tongue but they are also exposed to some English even before starting school as well as to other languages of Milne Bay Province such as Suau, Duau and Ware. Many older speakers of Saliba also speak and read Suau, which used to be the local mission language and the language of education. Saliba is mainly head-marking with the exception of some adnominal modifiers which carry a pronominal suffix indicating the number of the head noun. The language follows a nominative-accusative marking pattern and has fairly rigid SOV word order with nouns and S-V-0 order with pronominal arguments. Certain typological features of Saliba and other Papuan Tip Cluster languages - such as SOV word order - can be attributed to historical influence from non-Austronesian languages. Saliba has postpositions and the word order in nominal constituents is Genitive-Noun, Demonstrative-Noun, Noun-Adjective and Noun-Numeral. Inflected verbs minimally carry a subject prefix and if transitive also an object suffix and they constitute potential minimal clauses.1 Besides the pronominal affixes, there is little obligatory verbal morphology.

3.

The layered structure of the clause

Structurally, two types of serialization which have been labelled nuclear and core-layer serialization tend to be distinguished in the literature. The two types differ in the nature of the units that are serialized. The distinction is grounded in the notion of the clause as a layered structure as described by Foley and Van Valin (1984) and later work in Role and Reference Grammar (cf. Foley & Olson 1985, Van Valin 1993, Van Valin & LaPolla 1997).

Core-layer junctures in Saliba

67

On this view, the primary constituent units of the clause are the nucleus, which contains the predicate (usually a verb), the core, which contains the nucleus and the arguments of the predicate, and the periphery, which is an adjunct to the core and subsumes non-arguments of the predicate, e.g. setting locative and temporal phrases. (Van Valin 1993: 5) The layered structure of the clause is represented in figure 1 (adapted from Van Valin & LaPolla 1997).

Predicate

+ Arguments

Non-arguments

core

periphery

nucleus Figure 1. The layered structure of the clause The linkage between two units may be described in terms of its 'juncture' i.e. the level at which the linkage takes place. Instances of linkage may occur on each of the three levels and accordingly one can talk about three types: nuclear, core, and clausal juncture. In nuclear junctures, two or more nuclei are joined and they share all features of the core and periphery. In core junctures, two or more cores are joined and they share all features from the clause's periphery. In clausal junctures, the joined elements are clauses with their individual nucleus, core and periphery. Instances of both nuclear and core-layer junctures have been discussed in the literature as types of verb serialization (e.g. Foley & Van Valin 1984, Foley & Olson 1985, Crowley 1987, Van Valin & LaPolla 1997). Being a head-marking language, arguments are expressed in Saliba by pronominal affixes on the verb and/or as lexical nouns or free pronouns. Syntactic arguments are distinct from adjuncts in that they occur as bare NPs while adjuncts are generally marked by postpositions (with the exception of temporal expressions), and in that adjuncts may never be crossreferenced on the verb.2 In Saliba, the uninflected verb stem constitutes the nucleus; the inflected verb with its pronominal arguments plus the argument NPs constitutes the core. Non-arguments, which may precede and/or follow the verb, constitute the periphery of the clause. Consider the example in (1):

68 (1)

Anna Margetts Lahi yesterday

Maria ya Name lSG

kita-0 see-3SG.O

maketi unai. market POSTP.SG

nucleus core periphery periphery 'Yesterday, I saw Maria at the market.' In Saliba nuclear-layer juncture, the serialized units are verb stems which are joined together to form a single inflected verb. In core-layer juncture the joint units are inflected verbs.

4.

Defining serialization

Serialized verbs tend to share features with both single-verb predicates and multi-clausal constructions and a discussion of serialization should be grounded in some formal features distinguishing serialization from a mere sequence of verbs or clauses. Because of the wide range of constructions which have been discussed as instances of serialization, it proves to be difficult to provide a clear definition of serial verb constructions. For a review of some of the main criteria see the discussion in Senft (this volume). Durie (1988: 3) suggests the following working definition for serialization: ... serialization is what happens when two or more verbs are juxtaposed in such a way that they act as a single predicate ... The verbs are bound together syntactically and/or morphologically on the basis of sharing one or more core arguments, and neither verb is subordinate to the other. Typically in a serial construction there is no marker of subordination or coordination, no dividing intonational or morphological mark of a clause boundary, and the verbs cannot have separate scope for tense, mood, aspect, illocutionary force, and negation. Based on the criterion of argument sharing, Crowley (2002: 39-42) presents a typology of five structural patterns for serial verbs typically found in O c e a n i c . H e distinguishes SAME-SUBJECT, SWITCH-SUBJECT, INCLUSORY,

and AMBIENT serialization. In same-subject serialization, the subject of the first verb is coreferential with the subject of the second verb. In switch-subject serialization, the subject of the second verb is coreferential with the object of the first verb. In inclusory serialization, the MULTIPLE-OBJECT

Core-layer junctures in Saliba

69

subject of the second verb is coreferential with both the subject and the object of the first verb. In multiple-object serialization, both verbs are transitive and have independent objects. The subject of the second verb may be coreferential with either the subject or the object of the first verb (and therefore cases of multiple object serialization could in fact be described as subtypes of either same-subject or switch subject serialization). In the last type, ambient serialization, the subject of the second verb refers to the entire event expressed by the first verb (e.g. they dance IT is nice). In their typological overview of Oceanic languages, Lynch et al. (2002: 46-49) describe a number of functional types that are commonly attested in the Oceanic language group. They label these types DIRECTIONAL/POSITIONAL, SEQUENTIAL, CAUSATIVE, a n d MANNER s e r i a l i z a t i o n . 3

In

the

directional type, the first verb expresses an activity and the second verb expresses the directionality of the activity. In the sequential type, the first verb expresses movement and the second verb expresses an action that follows, often the purpose of movement. In causative serialization, the first verb expresses a cause and the second verb its result. Finally, in manner serialization, the first verb expresses an action and the second verb how it is performed. This list of functional types is clearly not exhaustive but it conveniently gives an inventory of types of serialization commonly found in Oceanic languages. 4 1 will come back to the types described by Crowley (2002) and Lynch et al. (2002) in the discussion of Saliba serial verbs.

5.

Nuclear-layer serialization

As mentioned, Saliba makes extensive use of nuclear-layer serial constructions. Since I describe these constructions elsewhere in more detail (Margetts 1999, Margetts in prep.), I present here merely a summary of the basic features. Saliba nuclear-layer constructions constitute single grammatical words with a complex stem (or nucleus). These complex verbs take a single set of inflectional affixes so that prefixes attach to the first verb stem and suffixes to the last stem of the sequence. The stems generally need to agree in their transitivity status. That means, if the first stem is transitive the following stems must be transitive too and an intransitive stem can only follow if it is transitivized by the causative or applicative affix. 5

70

Anna Margetts

Four structural slots can be distinguished within such complex verbs and each slot is associated with a certain type of function, as summarized in Figure 2. Vi (head) - V2 (result) - V3 (directionality) - V4 (adverbial/aspectual function) Figure 2. Structural slots in Saliba nuclear-layer serialization It is important to note that these structural slots are not necessarily identical to slots in the surface structure of a serial verb. (Examples of serialization with four stems do not actually occur in natural data.) Rather, the positional slots can be deduced from examples with three stems which show the ordering constraints that apply when certain stems co-occur, i.e. if they co-occur, V2 stems always follow Vi stems but precede V 3 and V 4 . V 3 stems always follow Vi and V 2 stems but precede V 4 , and finally V 4 stems can only occur in the last position of the complex verb. The first verb in a complex verb is generally the head of the construction while the second slot hosts stems expressing the result of an action expressed by V]. A same-subject constraint governs the combination of Vi and V 2 stems and such sequences can be described in Crowley's terms as same-subject serialization (in terms of logical subject, since structurally the stems share a single subject prefix). (2)

Galasi se koi-kesi-0. glass 3 PL hit-break-3sG.o 'They broke the glass.'

(3)

Ye

sikwa-he-beku-0.

3SG poke/hit-CAUS-fall-3sG.o

'He poked it and made it fall.' The stems in the third slot express directionality and they modify the action encoded by the preceding verb stem(s). Example (4) shows the sequence V1-V2-V3, example (5) a sequence of Vi and V3. (4)

Ye

sikwa-he-beku-dobi-ei-0.

3SG poke-CAUS-fall-go.down-APP-3sG.o

'He poked it down. / He made it fall down by poking it.' (5)

Ye yoga-sae. 3SG call-go.up 'He called up.'

Core-layer junctures in Saliba

71

The relation between V 3 or V 4 and preceding stems is one of ambient serialization in that the logical subject of V 3 or V 4 refers to the event described by the preceding verb(s) (except arguably for uyo 'go down' as V 4 ). Stems in the fourth slot again function as modifiers of the preceding stem(s). V 4 -stems express manner, or the end, repetition or unsuccessful attempt of an activity as well as other adverbial-like functions. Example (6) shows the sequence V1-V2-V4, (7) and (8) are examples of Vi plus V 4 . (6)

Ye tabe-he-dudulai-uyo-i-ya-ma. 3SG pull-CAUS-straight-back/again-APP-3SG.O-hither 'He pulled it straight again'

(7)

Ye 3SG

he-kata-namwa-namwa-i-gai. CAUS-learn-RED-good-APP-lEX.O

'She teaches us properly.' (8)

Se paisowa-gehe. 3PL work-finished 'They finished working.'

As can be seen in these examples, most of the functions described by Lynch et al. (2002) can be expressed by nuclear-layer constructions in Saliba. The directional type is illustrated by examples with V 3 stems as in (4) and (5). The causative type is demonstrated by other examples with V 2 stems, as in (2) and (3). Manner and other types of serialization are encoded by V 4 stems as in (6) to (8). The one function from the list discussed by Lynch et al. which is not covered by nuclear-layer constructions in Saliba is the sequential or purposive type. In terms of Crowley's typology of argument sharing between serial verbs, the Saliba nuclear-layer constructions show instances of same-subject and ambient serialization.

6.

Core-layer juncture in Saliba

While nuclear-layer serialization is a prevailing feature of Saliba grammar and easily identified through the single set of inflectional affixes, evidence of core-layer serialization is much harder to find. There are some problems in identifying constructions as core-layer serialization and distinguishing them from other core-layer constructions or even from a sequence of independent verbs or clauses. As there is little obligatory verbal morphology, marking for example tense or aspect, the bond between verbs in core-layer

72

Anna Margetts

juncture is not necessarily morphologically overt. As I will show, such bonds may however be identified through morphological tests. A second problem lies in the fact that - due to the pronominal affixes - every inflected verb constitutes a minimal, potentially complete clause. Therefore, even after a formal bond is established, we still need to identify the level of juncture as either core or clausal. Finally, when a core-layer juncture can be identified, the question remains whether it in fact constitutes serialization. Other constructions which may show a formal bond or dependency between verbs, such as coordination and subordination have to be considered in terms of their differences and similarities to the potential core-layer serial constructions. In 6.1,1 discuss instances of coordination and subordination, and in particular of complementation in Saliba to provide a base of comparison with these constructions. There are a number of constructions in Saliba which appear to be good candidates for serialization on the core-layer. In 6.2 to 6.4,1 discuss such verb sequences and present evidence for their analysis as formal constructions and junctures on the core-layer. In 6.6, I show that some of these constructions share features with both serialization and complementation.

6.1. Coordination and subordination There is comparatively little indication of hierarchical clause structures in Saliba and, while there are strategies for coordination of clauses, there is little evidence for structural subordination. The relation between clauses or sentences can be marked by a number of discourse particles. The conjunction na 'and' typically expresses temporal order and it often combines with the particle kabo which indicates future reference, sequence, or 'immediateness' (cf. Margetts 1999). The particle sequence na kabo in the text example in (9) can be translated into English as 'and then'. (9)

Ka

lao ka kai-gwali na kabo ka-mai ku lao-liga. go IEXCL INTR-spear CONJ ΤΑΜ CL2-1EXCL.P 2SG go-cook 'We'll go and spear fish and then you'll cook for us.' IEXCL

Clauses which are linked by na 'and', kabo 'then' or the sequence na kabo always occur in iconic temporal order and scrambling of clauses results in a change of temporal sequence of the expressed events. The particle ede functions as a connective, indicating the relation between clauses by marking the preceding discourse unit as presupposed.

Core-layer junctures in Saliba

73

In the text example in (10), the first sentence ends with a clause which constitutes new information and is introduced by na kabo 'and then'. The second sentence picks up this clause followed by ede, marking it as presupposed. (10)

Hewa-hewali-o

wa

se

RED-young.man-PL DET

ye 3SG

kai-kaikewa-i-0

na

kabo

3PL RED-look.at-APP-3SG.O CONJ ΤΑΜ

maliwai. vomit

'The young men were watching her and then she vomited.' Ye

maliwai

ede

kwateya

se

3SG

vomit

PRSP

yam

3PL go-appear

tau-masahala.

'As she vomited yams appeared.' As with na kabo 'and then' the clauses connected by ede always occur in iconic temporal order and scrambling results in a change of meaning. Also note that all of the clauses in the examples above can, in principle, stand alone. They are structurally like main clauses and there is no sign of formal dependency. Complement clauses in Saliba typically show no formal marking of subordination either and can be classified as 'sentence-like' complement clauses following Noonan (1985) since they tend to have the same structure as main clauses. In contrast to object arguments which precede the verb, complement clauses follow it, as shown in (11). (11)

Ya

kita-di

numa ne

LSG

see-3PL.0/P0SS house

unai

DET PP.SG

se

keno.

3PL

sleep.

Ί saw them sleeping in the house (lit.: I saw them, they slept in the house).' The particle bena generally expresses intention and/or obligation, as in (12) and (13). (12)

Bena ya lao Samarai. OBLI lSG go Place.Name Ί should/must go to Samarai.'

(13)

Waga wa bena ye uyoma. boat DET OBLI 3SG come.back 'The boat should come back.'

This particle also functions as an optional complementizer with verbs expressing desire or intention as in (14) to (18). The main verb and the

74

Anna Margetts

complement verb may share the same subject as in (14) and (15) or have different subjects as in (16) to (18). (14)

(15)

(16)

(17)

(18)

Ya

henuwa bena ya lao Samarai. LSG like OBLI/COMP ISG go Place.Name Ί want to go to Samarai.' Ye

henuwa bena keke like OBLI/COMP cake 'She wants to bake cake.'

ye

3SG

3SG

Ya

lao-ma. go-hither

henuwa bena ku ISG like OBLI/COMP 2SG Ί want you to come.'

gabu-0. bake-3sG.o

Kwa

gado bena ya tole-0 na kwa want OBLI/COMP 2SG put-3SG.o CONJ2PL 'Do you want me to put it on so you can hear it?'

lapu-lapui-0?

2PL

RED-hear-3SG.O

Ye

ku

henuwa bena yo-na peki 3SG like OBLI/COMP CL1-3SG.POSS backpack 'He wants you to buy a backpack for him.'

2SG

hemaisa-0. bui-3sG.o

The verbs gado 'want' and henuwa 'like' are the only verbs which may take nominal complements which, like complement clauses, follow the verb, as in (19) and (20). (19)

Ya henuwa keke. LSG like cake ' I want cake.'

(20)

Ye gado ti mo. 3SG want tea only ' She wants only tea.'

Gado 'want' and henuwa 'like' are also the only complement-taking verbs which may be followed by complement clauses that differ structurally from main clauses. Apart from occurring with full verbal clauses (with or without the particle bena), as in (14) to (18) above, they may also be followed by a bare verb stem without a subject prefix, as in (21) and (22). In these constructions bena cannot occur and the two verbs have to share the same logical subject which is expressed by a prefix on the main verb only. Because of the lacking subject prefix on the second verb, the complement differs formally from an independent verb or clause. (21)

Ye gado gelu. 3SG want board 'She wants to (get on) board.'

Core-layer junctures in Saliba (22)

75

Ya henuwa dobi Alotau. ISG like go.down Place.Name Ί want to go down to Alotau.'

In the following I present a number of constructions which are potential candidates for an analysis as core-layer serialization. I discuss evidence that the verbs in these constructions do not merely occur in sequence but rather share a formal bond or dependency relation which can be described as corelayer juncture following Foley and Van Valin (1984) and Van Valin and LaPolla (1997). These constructions share aspects with complementation in Saliba, which I investigate in 6.6.

6.2. Purposive constructions One type of construction to be considered in our search for core-layer junctures consists of two verbs, the first expressing movement and the second the purpose of the movement. Semantically as well as structurally, these sequences resemble the constructions discussed by Lynch et al. (2002: 47) as sequential serialization. This is interesting in light of the fact that sequential serialization is the only type from their list which is not covered by nuclear-layer constructions in Saliba (cf. section 5). Consider the examples in (23) to (26): (23)

Ya lao ya keno. ISG go ISG sleep Ί go and sleep.'

(24)

Ku lao ku dui! 2SG go 2SG wash 'Go and wash (yourself)!'

(25)

Ya dobi Sioni ya kita-0. ISG go.down Name ISG see-3SG.O Ί go down and see Sioni.'

(26)

Se sae kwateya se kuma-0. 3PL go.up yam 3PL plant-3sG.O 'They went up to plant yams.'

While the first verb in these constructions is always intransitive, the second verb may be intransitive, as in (23) and (24), or transitive, as in (25)

76

Anna Margetts

and (26). In the following sections, I discuss several features which suggest that these examples constitute constructions on formal grounds and junctures on the core layer.

6.2.1. Intonation contour, pauses and same-subject

constraint

One piece of evidence is the nature of the intonation contour. The two-verb sequences in (23) to (26) resemble mono-clausal utterances with a rise on the first verb and a falling tone at the end of the second verb. When appearing in a sequence with this intonation contour, the two verbs also have to share the same subject. In Crowley's (2002) terms, they constitute same-subject constructions. In order for the two verbs to have different subjects, the intonation contour would have to change to a falling intonation on each verb. In this case, a pause or a conjunction would also typically appear between the two verbs. The fact that the two verbs have to share the subject argument identifies these sequences as core-layer rather than clausal junctures.

6.2.2. Closed class of verbs in first slot A second point indicating that the two verbs build a formal construction is the fact that only a closed class of verbs may occur in the first slot of these expressions. This is the class of path-encoding motion verbs including lao 'go', sae 'go up', dobi 'go down', uyo 'go back', seuyo 'go back up' (a contraction of sae and uyo) as well as kawasi 'go across' (e.g. body of water), and dikwa 'go across' (e.g. hill). Manner-of-motion verbs such as heloi 'run' or kamposi 'jump' may not appear in this slot and example (27) is unacceptable. (27)

*Ya ISG

heloi

ya

run

ISG

kita-0. see-3SG.O

Ί run and see him.' In order to express the manner of motion within the purposive constructions the manner verb has to be followed by one of the path-encoding verbs, as in (28).

Core-layer junctures in Saliba (28)

Ya

heloi

ya

lSG

run

lSG go

lao

ya

77

kita-0.

lSG see-3SG.O

Ί run and see him.' Such three-verb sequences constitute core-layer junctures in themselves as discussed in 6.3.

6.2.3. Word order A third point speaking for an analysis of these sequences as mono-clausal units is a variation in word order. For sequences including a transitive verb in second position, the transitive object sometimes precedes the initial intransitive verb, as in (29a) to (31 a), rather than occurring in its canonical position preceding the transitive verb, as in (25) and (26) above. A sequence of object plus intransitive verb without the following transitive verb is ungrammatical as shown in (29b) to (31b). 6 (29) a. Sioni ya lao ya kita-0. Name lSG go lSG see-3SG.O Ί go and see Sioni.' b.

*Sioni ya lao. Name lSG go Ί go Sioni.'

(30) a. Ka-m

ti ku lao-ma ku tea 2SG go-hither 2SG 'Come and get your tea!' CL2-2SG.POSS

b.

*Ka-m CL2-2SG.POSS

ti

ku

tea

2SG

hai-0! get-3sG.o

lao-ma! go-hither

'Come your tea!' (31) a.

Wawaya-o

ya

lao

child-PL

LSG go

ya-wase-nei-di. LSG-search-APP-3PL.o/P

Ί g o and search for the children.'

b.

* Wawaya-o child-PL

ya

lao.

ISG go

Ί go the children.' It is not the case that the object in such clauses is merely topicalized and fronted, as the object NP is not intonationally prominent and, theoretically,

78

Anna Margens

a lexical subject could precede the object noun (practically clauses with two lexical arguments tend to be avoided). This word order pattern suggests an analysis as serialization. But it also poses an interesting problem for the analysis of these constructions in that the object occurs in a position preceding both verbs, which would be expected with nuclear-layer but not with core-layer constructions. 7 However, since the word order in (29) to (31) is only attested as a variation of the canonical position where the object occurs between the two verbs, an analysis as nuclear-layer juncture can be ruled out. Also, in contrast to the nuclear constructions described in section 5, both verbs take separate subject prefixes. Possibly, the option of fronting the object reflects an ongoing process of grammaticalization and an increasingly closer bond between the verbs which may ultimately result in a shift from a core to a nuclear layer construction. An argument against an analysis as serialization is the fact that the complementizer bena, described in 6.1, may optionally occur between the two verbs of the sequence, as in (32). (32)

Ye

lao-ma bena ye 3SG go-hither OBLI/COMP 3SG 'He came and wanted to jump.'

kamposi. jump

If more than two verbs occur in a sequence, as in example () above which encodes manner, path and purpose, the complementizer may occur between the path-encoding motion verb and the verb expressing purpose, as in (33). (33)

Na ye heloi ye lao bena bolo wa and 3SG run 3SG go OBLI/COMP ball DET 'And he ran to get the ball.'

ye

hai-0.

3SG

get-3SG.O

The complementizer is not attested between the first and the second position of such sequences, i.e. between the manner-of-motion verb and the path-encoding motion verb. I will return to this discussion in 6.6.

6.3. Manner of motion constructions As mentioned in 6.2, a sequence of three verbs is needed to express manner of motion within the purposive constructions, as the manner-of-motion verb may not replace the path-encoding motion verb but has to precede it.

Core-layer junctures in Saliba

79

Consider the examples in (34) and (35) which show the same pattern as (28) above. (34)

Ye

heloi

3SG run

ye

lao-ma

3SG go-hither

sina-na

wa

ye

kita-0.

mother-3SG.POSS

DET

3SG see-3SG.O

'She ran to see her mother.' (35)

Ye

ye

hai-0.

3SG

heloi ye lao-ma bena run 3SG go-hither OBLI/COMP 'He ran to him to get the ball.'

3SG

get-3SG.O

The complementizer bena may occur between the second and third verb, as in (35), but it is not attested between the manner-of-motion verb and the path-encoding verb of such sequences. The combination of a manner-of-motion verb and a path-encoding motion verb is not restricted to occur as part of the purposive construction. Sequences of these two verb types are very common and can be considered a construction in its own right. Examples such as (34) and (35) above where three verbs are combined are in fact a combination of two constructions which both constitute core-layer junctures. 8 More commonly mannerof-motion and path-encoding motion is expressed by a sequence of two verbs without a third verb expressing a purpose. In fact, manner-of-motion verbs are almost always followed by a verb expressing the path. 9 Semantically and structurally, these constructions resemble directional/positional serialization as described by Lynch et al. (2002) and both verbs are intransitive. Examples with heloi 'run' are most common, but other manner-ofmotion verbs are also attested. Consider (36) to (41). (36)

Se kamposi se sae. 3PL jump 3PL go.up 'They jumped up.'

(37)

Ye tapipili ye dobi. 3SG turn 3SG go.down 'He turned and went down.'

(38)

Ye nene ye 3SG crawl 3SG 'She crawled up.'

(39)

Se wose se dobi. 3PL paddle 3PL go.down 'They paddled down.'

sae-ma. go.up-hither

80

Anna Margetts

(40)

Se yona se lao-ma. 3pl walk.along.beach 3pl go.up 'They came along the beach.'

(41)

Ya mwalae ya lSG climb lSG Ί climbed back up.'

seuyo. go.back.up

The verbs in these constructions share the same-subject constraint and intonation contour of the purposive constructions discussed in 6.2. The examples further share with the purposive construction the fact that in one of the slots only the closed class of path-encoding motion verbs can appear. The manner-of-motion verbs in the first slot of the construction can (possibly) also be described as part of a closed class. Besides verbs expressing manner of motion, other items which may occur in this slot include verbs expressing the crossing of a boundary or departure from a location as in (42) to (45) (manner-of-motion construction is therefore only a working label). Further research is necessary to investigate the range of verbs which may occur in the initial slot of these constructions. (42)

Ye pesa ye dobi. 3sg exit 3SG go.down 'He exited and went down.'

(43)

Se lu se lao. 3pl go.into.bush 3pl go 'The went into the bush.'

(44)

Ye dahalai ye uyo 3sg leave 3sg go.back 'She left and went back.'

(45)

Se yabubu se dobi. 3pl go.away 3pl go.down 'They took off and went down.'

In summary, the manner-of-motion constructions constitute core-layer junctures and seem to qualify as verb serialization.

Core-layer junctures in Saliba

81

6.4. Constructions with hetubu 'start' and kaipate 'try' A third type of construction involves the stems hetubu 'start' or kaipate 'try' in initial position. They are translated into English as verb plus complement clause. The transitivity status of the construction depends on the second verb which can be transitive, as in (46) and (47), or intransitive, as in (48).10 (46)

Ya hetubu nogi ya hili-0. lSG start skirt lSG weave-3SG.O Ί started weaving a skirt.'

(47)

Ya

kaipate

keke

ya

gabu-0.

lSG try/learn cake lSG bake-3SG.O Ί tried to bake a cake.'

(48)

Ta-hetubu

ta-paisowa.

liNCL-start

liNCL-work

'We started working.' (49)

Kabo yama wa se ΤΑΜ fish DET 3PL 'Then the fish start to die.'

hetubu start

se mwaloi. 3PL die/dead

Structurally, the examples in (46) to (48) are quite similar to the purposive constructions discussed in 6.2 and the evidence for their analysis as a formal construction and core-layer juncture is of a similar nature. First, as in the purposive constructions, a same-subject constraint governs the relation between the two verbs. Second, the constructions also allow the word-order variation, where the object of the transitive (second) verb may precede the intransitive (first) verb, as in (50) and (51). The sequence of object plus intransitive verb without the following transitive verb is again ungrammatical, as shown in (50b) and (51b). (50) a. Nogi ya hetubu ya hili-0. skirt lSG start lSG weave-3SG.O Ί start weaving a skirt.' b. *Nogi ya hetubu-0. skirt lSG start-3SG.O Ί started a skirt.'

82

Anna Margens

(51) a. Keke ya cake

kaipate

ISG try/learn

ya

gabu-0.

LSG bake-3SG.O

Ί tried to bake a cake.' b. *Keke cake

ya

kaipate-0.

ISG try/learn-3SG.O

Ί tried a cake.' As with the purposive construction, the construction with kaipate 'try' allows the complementizer bena to intervene between the two verbs, as shown in (52). There are, however, no examples of constructions with hetubu 'start' featuring the complementizer. (52)

Se

kaipate

bena

se

hai-0.

3PL

try/learn

OBLI/COMP

3PL

take/get-3SG.O

'They tried to get it.'

6.5. Constructions with sagu-i 'help' On the search for Saliba core-layer serialization, there is a fourth type of construction to be considered featuring the transitive stem sagui 'help', which is composed of the noun stem sagu 'help' and the applicative suffix i. The examples in (53) to (55) show sagui 'help' followed by a second verb and the sequences translate into English as 'help' plus complement clause. As with the other constructions, each verb is inflected individually. The second verb in the sequence may be intransitive, as in (53) and (54) (in (54) it includes an incorporated object), or transitive, as in (55). (53)

Kwa

sagui-0

kwa

2PL

lao sina-da kwa go mother-1 INC.POSS 2PL 'Go and help our mother cook!'

help-APP-3sG.o

2PL

(54)

Lalaita ku sagu-i-0 kwa bosa-halusi! Name 2SG help-APP-3SG.O 2PL basket-weave 'Help Lalaita basket-weave!'

(55)

Kabo ya

sagu-i-go

ΤΑΜ

help-APP-2sG.ocake

ISG

keke ta IINCL

lao-liga! go-cook

gabu-0. bake-3sG.o

'I'll help you bake a cake.' There is again formal evidence that the two verbs form a juncture on the core layer. The constructions with sagui 'help' show a constraint on the

Core-layer junctures in Saliba

83

choice of subject. They can be considered inclusory constructions in Crowley's (2002) terms, as the subject of the second verb has to be coreferential with both the subject and the object of sagui 'help'. If this is not the case, the sequence is ungrammatical, as shown in (56) and (57) where the subject of the second verb is co-referential with only one of the arguments of sagui, the subject or the object. (56)

*Ya lSG

sagu-i-di

keke

help-APP-3PL.O cake

se

gabu-0.

3PL

bake-3sG.O

Ί h e l p t h e m bake a cake.'

*Ya sagu-i-di keke ya gabu-0. lSG help-APP-3PL.O cake lSG bake-3SG.O Ί help them bake a cake.' There are no examples of the complementizer bena following sagui 'help'. It is interesting to note that sagu-i 'help' can also engage in nuclearlayer serialization as in (53') to (55') below where the complex nucleus of two stems takes a single set of inflectional affixes. Note that neither the nuclear nor the core-layer constructions with sagu-i 'help' are common and it is unclear what the semantic differences between the example pairs are." (57)

(53')

Sina-gu

ya

lao-liga-sagu-i-0.

mother-lSG.POSS

lSG go-cook-help-APP-3SG.O

Ί h e l p e d m y mother c o o k . '

(54')

Lalaita ku bosa-halusi-sagu-i-0! Name 2SG basket-weave-help-APP-3SG.O 'Help Lalaita basket-weave!'

(55')

Kabo

keke ya

ΤΑΜ

cake

gabu-sagu-i-go.

lSG bake-help-APP-2SG.O

' I ' l l h e l p y o u bake cake.'

In these examples, what appears as the second stem of the corresponding core-layer constructions in (53) to (55) appears as Vi of the nuclearlayer verb while sagui 'help' appears as V 2 . Nuclear constructions with sagui 'help' are transitive or ditransitive. The transitivity status of the construction is determined by the status of the first stem which may be transitive or intransitive. To the argument of the first stem, sagui 'help' adds a helper argument in subject position, similar to the causer argument added by causativization. The form sagui 'help' is the only stem which

84

Anna Margetts

participates in this type of nuclear construction. Sagui 'help' does not fit easily into the discussion of structural slots in section 5 above. As it can neither follow nor precede verbs of the V 2 , V3 or V 4 slot it seems to form a different type of construction from the serial verbs discussed in section 5 even though the juncture is in both cases on the nuclear level.

6.6. Core-layer serialization vs complementation In the sections above, I discussed the Saliba purposive construction, the manner-of-motion construction as well as constructions with hetubu 'start', kaipate 'try' and sagui 'help' and presented these as potential candidates for core-layer serialization. I have shown that these expressions represent core-layer junctures and are constructions on formal grounds (rather than just based on their semantics or their English translation equivalents). The purposive construction presented in 6.2 corresponds to sequential serialization as discussed by Lynch et al. and the manner-of-motion constructions fit what they describe as directional/positional serialization. Both are commonly found across Oceanic languages. The constructions with hetubu 'start', kaipate 'try' and sagui 'help' do not correspond to common functional types of serialization. The similarities of these four constructions to Saliba complementation are worth exploring. Semantically, several of the constructions discussed in 6.2 to 6.5 seem compatible with an analysis as complementation and some of them have complement clauses as their English translation equivalents. Yet, the constructions are subject to more narrow constraints than most cases of complementation in Saliba in that they have clear argument sharing requirements (same subject or inclusory subject). However, similar (though not identical) constraints are shared by one particular subtype of complementation found with the verbs gado 'want' and henuwa 'like' as illustrated in (21) and (22) above and repeated here for convenience. (58)

gado

gelu.

3SG want

Ye

board

'She wants to (get on) board.' (59)

Ya henuwa

dobi

lSG like

go.down Place.Name

Alotau.

Ί want to go down to Alotau.'

Core-layer junctures in Saliba

85

These expressions share with the constructions in 6.2 to 6.5 that the two verbs have to share the same subject (with sagui constructions they share more generally that there is an argument-sharing constraint). However, in the expressions with gado 'want' and henuwa 'like', the subject prefix of the second verb may be dropped, which is in contrast to the constructions described in 6.2 to 6.5, where each verb needs to be inflected for its subject. For the purposive expression and the expressions with kaipate 'try', the strongest evidence against an analysis as serialization, and for complementation, is the fact that the complementizer bena may intervene between the two verbs. However, the purposive and kaipate constructions, like constructions with hetubu 'start', show a particular word order pattern that rather speaks for a serialization analysis. As discussed above, when the second verb in these constructions is transitive, the object may precede the first, intransitive verb rather than appear in the canonical position preceding the transitive verb, as in (60) to (62) (repeated from (29), (50) and (51)).12 (60)

Sioni

ya

Name

lSG go

lao

ya

kita-0.

lSG see-3SG.O

Ί go and see Sioni.' (61)

Keke ya kaipate ya gabu-0. cake lSG try/learn lSG bake-3SG.O Ί tried to bake a cake.'

(62)

Nogi ya hetubu ya hili-0. skirt lSG start lSG weave-3SG.O Ί start weaving a skirt.'

This word order is unusual but, moreover, it is in contrast with complementation constructions where the complement clause or nominal complement generally follows the main verb rather than precedes it (cf. 6.1 above). While semantically the notion of complementation seems compatible with the constructions with hetubu 'start' and sagui 'help', the complementizer bena is not attested with these construction types. Neither does it occur with the manner-of-motion constructions discussed in 6.3. In sum, the manner-of-motion constructions seem to be the best candidates for core-layer serialization in Saliba, followed by the expressions with hetubu 'start' and sagui 'help'.

86 7.

Anna Margetts Conclusions

In this paper, I investigated whether there is core-layer serialization in Saliba. Previous accounts have described nuclear-layer serial verbs as a common feature of Saliba grammar but do not make reference to core-layer serialization (Mosel 1994, Margetts 1999). Nuclear-layer constructions are easily identified in that they are inflected as single words. The question of whether there is core-layer serialization in the language involves a more detailed investigation as, due to the limited verbal inflection, it is more difficult to identify a sequence of two verbs as a mono-clausal construction. I have shown on the basis of naturally occurring data and some elicited material that there are a number of constructions in Saliba which can be analyzed as core-layer junctures and some of them as serial verbs. These constructions differ from a free sequence of verbs by a number of formal criteria including word order, intonation, shared arguments, as well as the fact that at least one position in each construction may only be filled by a closed class of verbs. Some of these constructions, particularly the purposive expressions and those with kaipate 'try', share features with both serialization and complementation. Constructions with hetubu 'start' and sagui 'help' are better candidates for serialization in that they do not seem to allow the complementizer to intervene. These constructions are however quite rare in naturally occurring language. The verb sequences which conform most closely to the notion of core-layer serialization are the constructions expressing manner of motion, followed by path-encoding motion. These manner-of-motion constructions are very frequent in everyday speech. It seems therefore that core-layer serialization in Saliba is restricted to only a few constructions, but at least one of them must be considered a core part of Saliba grammar.

Core-layer junctures

in Saliba

87

Orthographic conventions and abbreviations The Saliba working orthography shows quite straightforward correspondence between graphic symbols and phonemes. APP CAUS CLl CL2 DET EXCL INCL INTR NEG Ο

applicative causative possessive classifier (general) possessive classifier (food, etc.) determiner exclusive inclusive detransitivizer negative object

OBLl/ COMP obligation marker/ complementizer possessive POSS plural PL postposition POSTP presupposition PRSP reduplication RED singular SG tense/aspect/mode ΤΑΜ

Notes 1. 23.

45. 6.

7. 8.

The obligatory subject prefixes on verbs are written separately in the Saliba working orthography which I follow here (cf. Oetzel and Oetzel 1997). There seem to be no verbs in Saliba that subcategorize for a postpositionally marked NP. Lynch et al. also list ambient serialization (cf. Crowley's types above). In contrast to their other types, this type is characterized by formal rather than functional criteria and fits better into Crowley's (2002) structural list of construction types. See e.g. Early (1993) and the overview by Senft (this volume) for discussion of further types. There are exceptions to this tendency, see Margetts (1999) for discussion. Some speakers reject clauses like (29a) and (30a) as ungrammatical, but crucially, examples with this word order are not uncommon and are attested in natural speech - even by speakers who reject them in elicitations. Thanks to Bill Palmer for pointing this out and to Robert Van Valin for comments on these constructions. Although more research on these complex constructions is called for, one can probably describe the manner-of-motion construction as embedded into the purposive construction, i.e. as taking the place of the single path-encoding verb in the simple purposive constructions described in 6.2. A representation of the complex purposive constructions with brackets may look like this: [[Vi manner - V 2 path] manner V 3 purpose] purposive

88

Anna Margetts

9.

Some manner-of-motion verbs can be followed by path-encoding motion verbs in either nuclear or core-layer juncture while other verbs only seem to allow one of the two juncture types. More research is needed to investigate this choice. 10· Some speakers suggested constructions with nominalizations as alternatives to the transitive clauses in (46) and (47). In these examples the second, transitive verb appears as a nominalized form carrying the third person possessive suffix as in (i) and (ii) (i) Keke gabu-na ya kaipate-0. cake

bake-3sG.O

lSG

try/leam-3SG.O

'I'm trying to make a cake.' (lit.: The baking of cake I try) (ii)

Gita

koi-na

ya

hetubu ya

guitar hit-3sG.O lSG start

kaipate-0.

lSG learn-3sG.O

'I'm starting to learn guitar.' (lit: The hitting of guitar I start I learn) 11. In (ii) and (iv) I present some constructed negative clauses with both types of constructions. It is possible that the negative particle may alternatively precede the subject. (Note that these examples have not been verified with a native speaker.) (ii) Sina-gu nige ya sagu-i-0 ka lao-liga. mother-3sG.POSS

NEG

LSG

help-APP-3SG.O IEXCL

go-cook

Ί didn't help my mother cook.'

(iv)

Sina-gu

nige

ya

lao-liga-sagu-i-0.

mother- lSG.POSS

NEG

lSG

go-cook-help-APP-3sG,o

Ί didn't help my mother cook.'

12· It is unclear whether bena could occur in clauses (60) and (61) with a preposed transitive object; there are no instances of this in my data. The question whether the clauses with and without bena are in fact instances of the same construction will require further investigation.

Serial and complex verb constructions in Teop

Jessika Reinig

Abstract Verbal sentences in Teop, an Austronesian Oceanic language, have highly complex predicates. Besides ΤΑΜ and negation particles, adverbial and nominal modifiers and cross-referencing object-markers, the verb complex may contain serial verbs. These serial verb constructions operate at the nuclear layer. They can be classified into two different subtypes, and there are a few constructions constituting cases of specialization and grammaticalization.

1.

Introduction

Teop is an Austronesian Oceanic language which is spoken on the northeast coast of Bougainville. Geographically, Bougainville is part of the Solomon Islands and politically, it belongs to Papua New Guinea. The language belongs to the North-Nehan network of the North-West Solomonic group of the Meso-Melanesian cluster (Ross 1988: 25) These languages are known for their complex verbal morphology (Mosel/Spriggs 1999b: 321). Teop is an accusative language, and the word order in clauses with verbal predicates is SVO (if the subject is the topic) or OVS (if the object is the topic) (Mosel/Spriggs 1999a: 45). The object is not part of the verb complex. This paper examines serial and complex verb constructions in Teop 1 . Serial verbs in Teop all operate at the nuclear layer and are contiguous. Section 2 gives a short overview of word class distinction in Teop. Section 3 analyses the structure of the verbal complex. Section 4 deals with different types of serial and complex verb constructions in Teop in relation with the cross-linguistic criteria used for the identification and classification of serial and complex verb constructions.

90 2.

Jessika Reinig Word class distinction in Teop

With the exception of the adjective-verb distinction (Schwartz 2001), word class in Teop has not yet been investigated in detail. No morphological marking allows any clear definition of word classes, but using a combination of distributional, functional and semantic criteria, a preliminary definition of lexical word classes in Teop can be reached (see Mosel 2002 and Schwartz 2002).

2.1. The verb Teop has lexemes denoting actions and processes which correspond to the semantic description of the prototypical verb. The verb is the nucleus of the verb complex. In Teop, a verb cannot be distinguished by its morphological form. In addition to their function as nuclei of the verb complex, verbs can also function as modifiers in verb complexes (see Section 3) and noun phrases. The verb complex constitutes the predicate of the verbal sentence.

2.2. The noun Lexemes which correspond to the semantic characteristics of prototypical nouns are words denoting people, kin, animals and other living things, objects and places. Just like verbs, nouns in Teop have no morphological features by which they could be distinguished from other word classes. With the exception of place names, each noun belongs to one of three nominal classes. These classes are marked by means of three different articles (Mosel & Spriggs 1999b). Typically, the noun functions as the head of a noun phrase. A noun can also be incorporated into the verb complex (see Section 3.1). Note that verbs can also function as the head of a noun phrase.

2.3. The adjective In Teop, there are property words which denote different types of characteristics (e.g. dimension, value and colour) and which are semantically defined as adjectives. Their prominent syntactic function is to modify the

Serial and complex verb constructions in Teop

91

head noun of an NP. In this case, the adjective functions as the head of an adjectival phrase, which follows the noun phrase. The adjectival phrase is marked by an article which agrees with the article of the head noun.

3. The verb complex 3.1. The elements and the structure of the verb complex In addition to a verb, which is the nucleus (see Section 2.1), the Teop verb complex may contain one or more of the following elements (Mosel & Reinig 2000): - TAM-particles (pre- and postverbal) - two negation particles which form a kind of bracket around part of the verbal complex - an incorporated noun - a serial verb - several adverbs (pre- and postverbal) - object markers which cross-reference nominal and pronominal object arguments - a directional particle Sentence (1) shows a relatively simple verb complex: (1)

S VC Ο Kakato na ani nana bona overe. ART Kakato REAL eat IMPF:3SG ART coconut 'Kakato is eating a coconut.' 2 (Mosel 1998)

Ε

The verb complex consists of the nucleus, which is the verb ani 'eat', and two ΤΑΜ particles, the preverbal na and the postverbal nana. The particle na is a realis marker and alternates with the TAM-markers paa (PAST), kahi, pasi, tau (FUTURE) and toro (MUST). The particle nana is the third person singular form of the imperfective aspect marker. It has different forms for different persons 3 . ΤΑΜ markers are not obligatory in Teop. Apart from the nucleus, several other elements can occur within the verb complex as shown in (2), (3) and (4):

92

Jessika Reinig

(2)

A

otei na ani overe nana. boy REAL eat coconut IMPF:3SG 'The boy is eating coconut.' (Mosel 1998) In addition to its nucleus, the verbal complex (in bold) in (2) contains an incorporated noun overe 'coconut', which loses the argument status that it has in sentence (1). It is within the verb complex and does not have an article. In contrast to (1), sentence (2) does not have a direct object, but a verb and an incorporated noun which modifies the nucleus (Reinig 2000). ART

Example (3) shows the distribution of the discontinuous negation particles sa ... haa (Mosel & Spriggs 1999a). The first negation morph, sa, precedes the preverbal ΤΑΜ particle and the second morph, haa, follows the nucleus. (3)

Ean

sa

2SG

NEG PAST

a

paa

rake like te-naa.

haa bata

maa

NEG SIMUL DIR

nom IMPF:2SG

beiko child PREP- 1SG 'At the same time you did not like my child.' (Mosel & Spriggs 1999a)

ART

Moreover, the verb complex also contains the adverb bata and the directional particle maa 'hither, towards me or the deictic centre'. Sentence (4) shows a verb complex which contains the adverb bata and the incorporated preposition ki plus the second person singular object marker -u: (4)

Enaa pasi

suusue

bata

ki-u

1 SG

tell:RED

SIMUL

PREP-OBJ:2SG

FUT

nom-an-ie IMPF:1SG-2SG-3SG

Ί will tell you about it.' (Purupuru 1, 134) ki is classified as an incorporated preposition and not as an applicative marker, because it also occurs outside the verb complex as a particle introducing a prepositional phrase (PP), as sentence (5) clearly shows. (5)

Ε

Sovavi vakakao vai vo inu Sovavi run then GOAL home tea huvi bona maa kaukau ki bene subunae. COMP peel ART PL sweet potatoe PREP ART grandpa-p0ss-3sc 'Then Sovavi ran home to peel sweet potatoes for her grandpa.' (Mosel 1998) ART

Serial and complex verb constructions in Teop

93

Sentence (6) shows the object marker ri inside the verb complex; ri does not have the status of an argument, but merely cross-references the pronominal object bari 'them', which stands outside the verb complex. (6)

Eori

he

na

vaa-hovo

3PL

CONJ

REAL

CAUS-enter OBJ

ri

tei

rori

komana

inu.

stay

IMPF:3PL

inside

house

rori

bari,

IMPF:3PL

them

'But they let them in, they stay in the house.' (Katavara 1, 146) The sequential order of the elements inside the verb phrase is fixed and can be summarized as follows: (7)

NEG ΤΑΜ ADV V NEG ADV APPL PREP ADV DIR ΤΑΜ

Very little variation is allowed: The adverb bata 'simultaneously' has two possible positions in negated predicates. It can occur before or after the second negation morph haa. The particle koa 'just' also has two possible positions: before or after the APPL/PREP slot. The rules for these variations are still not quite clear (Reinig 2000).

3.2. The position of a serialized verb As already mentioned in Section 1, verb serialization in Teop operates at the nuclear layer. Serial verbs are defined as a sequence of two or more verbs which act together as a single predicate (Durie 1997, Crowley 1987, Aikhenvald 1999). This definition contradicts that of Foley & Olson (1985:18) who seem to have a different notion of predicate: "Serial verb constructions always contain two or more predicates". This paper is based on the first definition. In Teop, a serialized verb appears within the verb complex before or after the nuclear slot, according to the type of serialization (see Section 4). This is shown by the simplified schema in (8): (8) ΤΑΜ ... V V V ... ΤΑΜ Teop has no core-layer serialization: if another verb occurs outside this verb complex, it is the nucleus of another predicate, and hence not a serial verb. In Teop, serial verbs are contiguous - no other elements intervene between the two verbs, e.g. (9), (10) and (11): (9)

A

ruene paa hapusu oha-u. water PAST flow go over-IMM 'The water has flowed over.' (TD4 ο ha)

ART

94

Jessika Reinig

(10)

Earn toro5 peha must

2PL

climb

kaveru

ο

karuu

te

Kitata.

steal

ART

karuu nut

PREP

Kitata

'You must climb and steal Kitata's karuu nuts.' (TD be) (11)

No gotto go

get

ο

rapisi.

ART knife

'Go get the knife.' (TD no) Such serialized constructions correspond to the defining criteria of verb serialization (Crowley 1987, Durie 1997, Aikhenvald 1999; see introduction to this volume): They share at least one argument and take only one subject; they share a single ΤΑΜ operator or show concord for the same marking; neither verb is a complement to the other. Not all of these features have to be present at the same time, but on the other hand, a single feature would not be sufficient to classify a construction as a serial verb construction. The notions of single event and intonation pattern, I consider highly problematic in Teop and, therefore, do not use them to identify serial verb constructions. First, how could someone know exactly what is conceptualized as a single event for a native speaker of Teop (or any other language)? Secondly, to my present knowledge, there has not been any thorough phonetic analysis of intonation patterns with respect to verb serialization for any of the languages discussed by Crowley, Durie or Aikhenvald. Furthermore, it is unrealistic to assume that there is only one intonation pattern in a monoverbal clause. A verb serialization can occur in a verb complex which also contains adverbials, directionals and other elements and therefore is a complex predicate. Example (12) shows a verb complex containing a directional proclitic ma 'come' (see Section 4.3), the serialized verbs kao 'go' and hovo 'go into' (see Section 4.1) and two adverbs vatotobin and koa. (12)

Me-a

otei

and-ART man

nomaa,

me ma kao

hovo

come

and come go

go into straight

inu

te-nam.

house

PREP-lEXCL

vatotobin

koa just

'And the man came and he came and went straight into our house.' (Purupuru 2, 152)

Serial and complex verb constructions in Teop

4.

95

Types of serial verbs

Teop has two main subtypes of serial verbs. The first type consists of an activity verb that involves some kind of motion, followed by a motion directional verb. The second type consists of a sequence of two activity verbs which occur at the same time. In addition, there is a construction which obviously originates from a serial verb construction but which is not classified as such. It includes a preverbal direction clitic and an activity verb (see Section 4.3). The preverbal direction clitic has grammaticalized from a former verb, it only occurs in this type of construction and never as an autonomous nucleus.

4.1. Serialized direction verbs The first type comprises two verbs, each of which can be the nucleus of a verb complex. The first verb is a motion verb which is unspecified with respect to the 'Path' of the motion 6 . The second verb is a directional motion verb. (13)

Ean tow kao gunaha nao kasuana. 2SG must go go down DIR beach 'You must go down to the beach.' (TD gunaha)

kao 'go' expresses an undirected motion, and gunaha 'go down' denotes a directed motion. Both verbs can be the nucleus of a verb complex, as examples (14) and (15) clearly show. (14)

(15)

Kao te-a sisibu! go PREP-ART bath 'Go have a bath!' (TD kao) Ο

torovaagina na gunaha bata maa-na. airplane REAL go down SIMUL DIR-IMPF:3SG 'The airplane is coming down (landing).' (TD gina)

ART

Verb serialization as in (13) is frequent in Teop. The verbs in second position form a closed class which seems to consist of ten items: seven of them are directional motion verbs, and three of them express a motion towards a particular goal, i.e. home, the bush or the water: gunaha 'go down', oha 'go over, pass sth.', hiava 'go u p \ pahana 'go across, cross', hovo 'go in, enter', taviri 'go around, go in circles', tavus 'go out, leave'; takin 'go out of the water', voos(u) 'go home', tagihu 'go to the bush'. The

96

Jessika Reinig

examples below illustrate the use of these verbs first as a serialized verb and secondly as an autonomous nucleus. (16)

A

ruene

ART water

paa

hapusu

oha-u.

PAST

flow

go over-lMM

'The water has flowed over.' (TD oha) (17)

A

beiko moon

ART child

vaa of

na

woman

te-a

oha

te-o

REAL go over

vamokomoko

PREP-ART pass

skul.

PREP-ART school

'The young girl passed the school exam.' (TD oha) (18)

A

anohi

na

hiri

hiava

ART black ant REAL crawl go up

bata-na

moo-naa.

SIMUL-IMPF:3SG leg-LSG

'The black ant is crawling up my leg.' (TD anohi) (19)

Εan toro hiava Namatoa. 2SG must go up Namatoa 'You must go up to Namatoa.' (TD hiava)

(20)

Ε

kahi

ART d o g

vo

vai

a

DEM

ART black

paru

na

ena

pa/tana

REAL Swim cross

bata SIMUL

Vapahana. GOAL IMF:3SG Vapahana 'The black dog is swimming across to Vapahana.' (TD ena) (21)

nana

Naa

na

LSG

REAL CAUS-finish

vaa-kavara

ο

vaa-dorana

ART ORD-four

ο

vinu

ART year

enaa

he

LSG

CONJ

sikuuru

nom

me

naa

paa

pahana

ο

school

IMPF:1SG and

lSG

PAST

cross

ART ORD-five

ο

vinu, enaa tau

ART year

lSG

FUT

vaa-taonim

vaa-sikuuru. CAUS-school

Ί finished the fourth year when I was in school and I crossed over to the fifth year when I was about to teach.' (Siimaa 1, 96) (22)

Ausu e help

tama-man

ART father-2SG

tea

deedee

COMP carry

hovo a go in

maa

ART PL

ta naono. timber

'Help your father carry the pieces of timber inside (the house).' (TD ausu) (23)

Hovo bata-na a toogusu huan he siko viravira. go in SIMUL-IMPF:3SG ART large quantity rain CONJ fall very much 'When he entered, the rain came down in buckets.' (TD hovo)

Serial and complex verb constructions in Teop Ε... ο

uhu

vai

well ART decoration DEM

to

kahi

va-taviri

REL

FUT

MULT-go around with

ri-ori

te-o

piuvu.

IMPF:3PL-3PL

PREP-ART

dance

97

mi

'Well, this is the decoration they are going around in the dance.' (Kaetavara 1, 179) Kahi bae FUT

taviri

rori

fence go around

ri-ori

to

bono hum

IMPF:3PL ART

kahi

tii

IMPF:3PL-3PL REL FUT

to

kahi

tei

place REL FUT

ri-a

tobara

stay

upee.

stay IMPF:3PL-ART group

upee boys

'They will fence around the place where the young upee boys will be staying.' (Kaetavara 3, 68) Pasi

tavus

paa

paku

FUT

come out DIR-IMM ART time 3SG CONS past

make

bono

paku,

ART

feast

ma-u

ο

vuri ο

re

'There will come a time when they would make a feast...' (Buasiana 2, 52) ...o

tapuhaka

re

oro

tavus

maa.

pus CONS flow come out DIR ' . . . so that the pus flows out.' (TD hum) ART

Pita

takin

vai maa

eove

he

tea

pita

va-

walk

go out of the water

then DIR

3SG

CONJ

COMP

walk

ADV-

hata

kurusu

bata

bad

very

SIMUL DIR

maa.

'He came out of the water walking very badly.' (TD kurusu) Takin! go out of the water

'Come out of the water!' (TD takin) A

maa

ART PL

hum

he

tara

place CONJ see

meori

paa

and-3PL

PAST paddle

hua

voosu

vamanemanee

bata

koa,

invisible

SIMUL

just

maa.

go home DIR

'When the place became invisible, they paddled back home.' (TD voosu, voos)

98

Jessika Reinig

(31)

Me-nam

paa

voos

and-lEXCL PAST

vaha

go home again

maa

vaan.

DIR

village

'And we came home here to the village.' (TD voosu, voos) (32)

(33)

Me-nam paa rosin tagihu. and-lEXC PAST flee go to the bush 'And we fled into the bush.' (TD tagihu) Ο

vahara beiko pasi

ART group

child

FUT

tagihu

rori

go to bush

IMPF:3PL COMP get

tea

gono

ohita. galip nuts

'The children will go to the bush to collect galip nuts.' (TD tagihu) These serial constructions do not comprise more than two verbs and their order is fixed. The nuclear verb, most often an activity verb, comes first; the directional verb comes second and may not occur as V*. Such constructions follow the cross-linguistic defining criteria of verb serialization (Aikenvald 1999, Durie 1997, Crowley 1987), namely: the two verbs form one predicate, are part of the same verb phrase, and have the same subject. As shown by the examples above, they are two juxtaposed, morphologically independent words. Grammatically, they have equal status and both of them can be autonomous nuclei. The transitivity of such constructions is determined by the first verb, for the second verb is always intransitive. If the first verb is intransitive, the whole complex is intransitive, and if the first verb is transitive, the whole complex is transitive. Since verbs in Teop are generally not morphologically marked for transitivity, there is no morphological evidence that this construction is asymmetric or has a head-modifier relationship. But there are some interesting examples which justify the assumption that in sentences with a transitive Vi, might have the role of a modifier, and that is when Vi and V 2 do not share the same subject. Examples (34) and (35) can be analysed as if the object of Vi is the subject of V 2 : (34)

Hoa

sunaha

ο

push

go down

ART canoe

sinivi

vo

tahii.

GOAL sea

'Push the canoe down to the sea.' (TD hoa2) (35)

At ο

hiava

ni-a

guvi.

hold

go up

TR-ART water container

'Hold up the water container.' (TD hiava) The contrast with intransitive sentences or sentences with a bare locative noun as locative adjunct is evident in sentences (16) and (36) where the verbs share the same subject:

Serial and complex verb constructions in Teop (16)

99

A

ruene paa hapusu oha-u. water PAST flow go over-lMM 'The water has flowed over.' (TD oha)

ART

(36)

Ean toro peha hiava inu. 2SG must climb go up house 'You must climb up to the house.' (TD hiava)

In order to determine whether this modifying function is a regular fact with a transitive Vi and different underlying subjects, we need more data. However, the absence of the prefix va- which derives adverbs from verbs in Teop, could be considered as evidence for equal status.

4.2. Serial verb constructions vs. modification with a derived adverb 4.2.1. Serial verb constructions with two independent verbs of actions/activities The second type of verb serialization in Teop consists of a sequence of two juxtaposed verbs which describe two actions happening at the same time and making up one event, as in (37) and (38), or an action done with a purpose (39) and not an activity verb and a directed motion verb as in Section 4.1. (37)

A

kara

aatoa na hoa taono ni aatoa creepers REAL grow up press down APPL bona mode mohina. ART watermelon garden 'The aatoa creepers extended and covered the watermelon in the garden.' (TD hoa) ART QUANT

(38)

Ε

bubuu na hio ihuana nana bona nahu kaukau. granny REAL sit wait for IMPF:3SG ART pot sweet potato 'Granny is sitting and waiting for the pot of sweet potatoes.' (pr. 4.27) ART

(39)

Earn toro peha kaveru ο karuu te Kitata. 2PL must climb steal ART karuu nut PREP Kitata 'You must climb and steal Kitata's karuu nuts.' (TD be)

It would be interesting to contrast (37), (39) and (38) with constructions which have two independent verb complexes which are connected by means of a coordination marker (e.g. SUBJ ΤΑΜ peha ΤΑΜ 'and' ΤΑΜ

100

Jessika Reinig

kaveru ΤΑΜ . . . O B J ) and a purpose marker (e.g. SUBJ ΤΑΜ peha ΤΑΜ 'so that' ΤΑΜ kaveru ΤΑΜ ... OBJ). Interestingly, all three sentences have a direct object; while the first verb is intransitive, the second verb is transitive. This kind of verb serialization is very infrequent in our Teop data. It is not used when the action or state expressed by the first verb has to be modified in some way. In this case, the modifier will be marked with the prefix va-.

4.2.2. Adverbial modification vs serialization To modify a verb, Teop does not use a serial verb, but an adverb derived from a verb by the prefix va-. This appears in the following pair of sentences. Sentence (40) shows a serial verb construction comprising a concomitant activity hio 'sit' and a state tamaka 'be sad'. (40)

Ε

amoba

ART w i d o w

na

hio tamaka

nana.

REAL

sit

IMPF:3SG

be sad

'The widow is sitting and mourning.' (Mosel/Spriggs 1999c) (40) does not mean 'the woman is sitting in a sad way'; this would be expressed as (41), which is not a serial verb construction, with a va- derived adverb. (41)

Ε

amoba

ART w i d o w

na

hio va-tamaka

nana

REAL

sit

IMPF:3SG

ADV-be s a d

'The widow is sitting in a sad way.' (Mosel/Spriggs 1999c) Examples (42), (43) and (44) also contain a verb modified by an adverb. (42)

A

kaveo

ART nail

kuri-n-e

bubuu

hand-POSS-ART g r a n n y

na

hoa

REAL

grow

va-bebeahu ADV-be l o n g

kurusu. very

'The fingernails of granny grew very long.' (TD kaveo) (43)

(44)

Ε

kahi na mimi va-hata nana. ART dog REAL urinate ADV-be bad IMPF:3SG 'The dog is urinating unbecomingly.' (TD mimi) Ο

nana.

ART

tae kahi na avuhu va-hata kurusu excrement dog REAL smell ADV-be bad very 'The dog's excrements smell very bad.' (TD tae)

IMPF:3SG

Serial and complex verb constructions in Teop

101

This seems to be quite a regular derivation used for modification. But there are some Teop sentences where serialization seems to be used instead of modification. Examples (45), (46) and (47) contrast with (44). (45)

Ο

mimi te kahi na avuhu hata urine PREP dog REAL smell be bad 'The dog's urine smells awful.' (TD mimi)

ART

(46)

Ο

buraa

te

kahi

to

avuhu hata

ART

vomit

PREP

dog

REL

smell

be bad

maa-na. DIR-IMPF:3SG

maa-na. DIR-IMPF:3SG

'The dog's vomit smells bad.' (TD buraa) (47)

A

nahu toa na avuhu neenesi kurusu ART pot chicken REAL smell be lovely very 'The pot of chicken smells lovely.' (TD avuhu)

maa-na. DIR-IMPF:3SG

There is no explanation yet why (45), (46) and (47) use hata without derivation, whereas (44) is a sentence with a similar meaning, using the derived form. The only difference between the two constructions is that (45), (46) and (47) contain the directional particle maa. Cause-effect states of affairs are also expressed by the adverbial prefix να-, as shown in the examples below: (48)

Tasu va-mate

e

kahi,

stone ADV-bedead ART dog

ha! INTERJ

'Stone the dog to death!' (TD ha) (49)

Ει

be

naa

asun va-mate

e

taan

ei,

DEM

CONJ

LSG

kill

ART

person

DEM

a

banoas-na-e

ADV-bedead

repaa

ART relatives-POSS-3SG CONS

gavagava. angry

'Here, if I kill someone to death, his/her relatives will then get angry.' (Vosunana 1, 167) They are distinct from causative constructions which are marked by the prefix vaa-, as in (50). (50)

Ean

toro

vaa-mate

va-mataa

2SG

must

CAUS-be dead

ADV-be good ART chicken

e

toa.

'You must kill the chicken properly.' (TD mate) Some languages use serial verb constructions to express causatives (Durie 1997), but this is not the case in Teop. It would be interesting to find out if there is some kind of evolution towards more adverbial derivation or

102

Jessika Reinig

towards more nuclear serialization, but at the current stage of our research, this is not possible.

4.3. Directional proclitics Another type of complex verb construction involves former deictic direction verbs. This class consists of two members, no 'go' and ma 'come', which occur in the first slot of the sequence and specify an activity verb (V2), which is the main verb. The words no and ma do not have the grammatical status of verbs, because they cannot appear as autonomous nuclei. They seem to have developed from the verbs nao 'go' and nomaa 'come' through a process of grammaticalization and developing into a sort of auxiliary expressing direction. This kind of complex verb differs from the other types in that the construction does not consist of two grammatically equal members, but of a directional proclitic specifier (Vi) and a head (V2). The preverbal clitics are used to express the movement - 'come' or 'go' - that is necessary to fulfil the action which is expressed by the main verb. This kind of serialization is cross-linguistically common (Durie 1997). Examples (51) and (52) show the use of ma, examples (53) and (54) the use of no. (51)

Ean

torο

kao kahi

2SG

must

go

pasi

ma

gono

FUT

come get

ο

pauna

away ART banana

vu-a

too

ο

vuri,

ART ripe

vaahito.

IMM-ART person

own

'You must leave the ripe bananas, the owner will come and get them.' (TD maa) (52)

Dadao

ri-a

maa

moon,

call:RED

3PL-ART

PL

woman 3PL

eori

ma

ani.

come

eat

'Call the women over so that they come and eat.' (TD rid) (53)

No gono

ο

go

ART knife

get

rapisi.

'Go and get the knife.' (TD no)

Serial and complex verb constructions

(54)

A

booboo

ART booboo

pasi FUT

eori

re

nahu

3PL

CONS cook

no

hee

u

naa

go

give

IMM lSG

in Teop

103

bono vahara

beiko

ART

children

group

bari. 3SG

Ί will go and give the booboo [kind of fish] to the children so that they will cook it.' (TD bari) Grammaticalization of former serial verbs is not unusual in the languages of the world. This kind of complex verb could be compared with the asymmetric serial constructions described by Aikhenvald (1999). The transitivity of this kind of construction with ma and no is determined by the transitivity of the main verb. These directional proclitics cannot be considered as directional particles, as their presence does not block the usage of the postverbal directionals maa and nao. (55)

No dadao

maa

e

go

DIR

ART grandmother-2sG

call:RED

subu-man.

'Go and call your grandmother.' (TD dadao)

4.4. Cases of specialization The Teop data provide some examples of serial verb constructions which do not belong to one of the types described in sections 4.1.-4.3. They involve Aktionsart verbs. The verb taneo 'begin, start' can appear as the first or as the second verb in a serial verb construction. When taneo comes first, the following verb behaves like a complement in (56): (56)

Enaa na

taneo

katnisi

nanava.

lSG REAL start be sick yesterday Ί b e c a m e sick yesterday.' (TD kamisi)

However, this construction is not regular. The verb taneo can also take a "real" complement, which is introduced by the complementizer tea and which is not part of the verb complex: (57)

Ε Sovavi taneo vai tea oga. ART Sovavi begin then COMP cry 'Sovavi began to cry.' (TD taneo)

It is not yet clear under which conditions a serial construction or a complement clause is used with taneo. When taneo appears as the second

104

Jessika Reinig

verb in a serial sequence, it describes the starting point of the action expressed by the first verb: (58)

Ε

Kakato paa kuu taneo ma-u te-o overe. ART Kakato PAST fall start DIR-IMM PREP-ART coconut 'Kakato fell from a coconut tree.' (TD taneo)

(59)

A

maa

hum paa takae taneo ma-u storm PAST rise start DIR-IMM 'The storm rose from the ocean.' (TD hum)

ART PL

namana. ocean

In (58) and (59), I would not analyse taneo as an ablative marker, but as a serialized verb (but this time with spatial and not Aktionsart value). First, it keeps its meaning 'start', and second, the verb complex contains the directional marker maa (shortened to ma- in the sentences above). Maa expresses that the motion of Kakato and the storm is towards the speaker or the deictic centre. The opposite of maa is nao which means away from the speaker or from the deictic centre. The verb kavara 'finish, be entire, be finished, end up' can also be used in a serial verb construction, as in (61). In this case, kavara is in first position, it does not refer to the end of an activity, a process or an event, but to the entirety of the subject, to the fact that all the referents of the subject NP perform the action expressed by the second verb. (60)

Evehe

pasi

kavara

pahi-u.

Kahi

but

FUT

finish

however-LMM

FUT

kavara-u. finish-LMM

'But it will finish. Will finish.' (Purupuru 2, 101) (61)

Ο tao bar ο puisi ae ο kahi ART chicken and ART cat and ART dog paa kavara rosin vuru rasuu te-a bon. PAST be entire flee a long time ago bush PREP-ART night 'The chickens, the cats and the dogs all ran away into the bush last night.' (TD ae)

The verb momohu 'be first', 'go ahead', 'be the leader' can be used alone or as the second verb in a serial sequence. In this case, it does not describe two simultaneous actions as the serialization described in Section 4.2.1, but in (63) it serves as a temporal/chronological modifier. (62)

Enam

momohu ναι, earn pasi vamurina ma-u. go ahead now 2PL FUT come later DIR-IMM 'We are going ahead, you shall all come later.' (TD murina) IEXCL

Serial and complex verb constructions in Teop (63)

105

Na

pete bata koa nao-ri te-a inu vai add SIMUL just DIR-IMPF:3PL PREP-ART house DEM to paku momohu roho ori ei. REL build first ANTE 3 PL DEM 'They only add (extra sago palm leaves) to the house that they built first.' (Kaetavara 2, 80) REAL

5.

Summary

Serial verb constructions in Teop are contiguous and operate on the nuclear layer. They consist of not more than two verbs. There are different subtypes of serial verbs. The first type consists of an intransitive or transitive activity verb that implies a kind of motion and an intransitive direction verb. The serialized verb in such sequences specifies the direction of the first verb. The second kind is a serialization of two verbs that express two states of affairs that happen at the same time or a purposive action. Only one of these verbs - namely the second one - can be transitive, which makes the whole complex transitive. There is no example of two transitive verbs in a serial complex. My assumption is that in such a case, Teop would use a complex sentence containing two predicates, i.e. two verbal complexes. The serialization of two co-ordinate events contrasts with a construction consisting of a verb which is modified by means of a derived adverb. A third type of complex verb consists of a directional proclitic that only appears in combination with another verb. The transitivity in such constructions is determined by the transitivity of the second verb. The last cases studied seem to be cases of specialization and grammaticalization varying with position. More research with native speakers will be necessary to clarify them.

106

Jessika Reinig

Abbreviations ADV the prefix να- which derives adverbs from veitos APPL the applicative particle ni ART article CAUS the causative prefix vaaCOMP complementizer CONJ conjunction CONS consequential or purpose conjunction reDEM demonstrative DIR the directional particles maa and nao EXCL first person plural exclusive FUT the future markers kahi and pasi GOAL the preposition VO IMM the immediateness marker u IMPF the imperfective markers nom, nana, raara

INTERJ MULT

NEG OBJ PAST

PL PREP QUANT REAL

SG SIMUL

interjection the verbal prefix va- which indicates multiple action, i.e. reciprocal action, a movement to and fro or an action done by a group of people. negation particles saka.., haa cross-referencing object suffix the past tense marker paa plural preposition quantifier the realis-marker na singular simultaneous adverbial marker bata

Notes 1.

2.

3. 4. 5.

6.

I want to thank those who have assisted me in the process of writing this paper: Ulrike Mosel for her most helpful advice and comments and the editors of this volume for their constructive criticism and editorial woik. The English translations of the example sentences and the additions given in parentheses were done by the informant and are not changed even if they do not conform to Standard English. 3PL: rori\ IINCL: raara·, ISG, 2SG, EXC, 2PL: nom. The source reference refers to the entry in the Teop-English Dictionary (TD) under which the sentence is found. The element toro 'must', which translates as a modal auxiliaiy in English or German, is not considered as a serialized veib, as it alternates with the preverbal TAM-markers (see Section 3.1). For the notion of 'Path', see Talmy 1985: 61.

Chains of freedom: Constraints and creativity in the macro-verb strategies of Mwotlap Alexandre Francis

Abstract In Mwotlap (north Vanuatu), most verb phrases consist of two or more verb roots chained together (Vi-V 2 ...), acting like a single verb. Although a clear case of verb serialization, such phrases reveal a strong asymmetry between their free verb head (Vi), and what appears to be little more than a post-head modifier (V 2 ). Because the resulting "macro-verb" can only refer to a single action, its internal structure has to obey strict rules; this paper analyses the way the valencies of both component verbs are capable of consistently merging into that of the whole macroverb, avoiding such things as conflicts between competing objects. Constraining though they may be, these syntactic rules turn out to be a powerful tool serving the speaker's creativity: indeed, this paradoxical "chained freedom" brings about spectacular paths of evolution in the history of Mwotlap macro-verb strategies.

1.

Introduction

Quite diverse kinds of structures have been placed under the term "verb serialization", which might well deserve different analyses. From the formal point of view, a contrast must be made between what has been acknowledged (Foley & Olson 1985) as "Nuclear-layer serialization" - of the type I hit-die-d your brother - and "Core-layer serialization" - of the type I hit your brother he died. And even after such formal precautions, it might well be also that within a single type, two different languages use similar structures to encode different semantic values - in such a way that it may not be prudent to generalize to all serializing languages the observations made for one of them. For instance, supposing nuclear-layer serialization in one language should be proved to refer to single events, it remains possible that the same syntactic device be used, in another language, to refer to several, contiguous events.

108

A lexandre

Frangois

Mwotlap is an Oceanic language spoken by approximately 1,800 speakers in northern Vanuatu, Banks Is. (Frangois 2001, 2003). More precisely, it belongs to the genetic subgroup which was labelled 'North-Central Vanuatu' [NCV] by Ross Clark (1985). Most of the NCV languages which have been already described have shown at least one, and sometimes two kinds of serial verb constructions: Paamese (Crowley 1987), Lewo (Early 1993), Namakir (Sperlich 1993), Ambae (Hyslop 2001) all combine nuclear-layer and core-layer serialization patterns; Araki, a language spoken in south Santo by a handful of speakers ( F r a n c i s 2002), also has both patterns, but shows a very strong tendency for core-layer SVC. Finally, moving further northwards shows the latter structure to be less preferred than nuclear-layer SVC, in such a way that Mwotlap has virtually no example of core-layer SVC [see fn.6]. The latter facts are summarized in the following chart: Table 1. Different distribution of serial-verb constructions across some NorthCentral Vanuatu languages some NCV languages

Core-layer SVC e.g. I hit him he died

Paama, Lewo Araki Mwotlap

+ +

Nuclear-layer SVC e.g. I hit-die-d him +

(+) +

-

The present paper will thus analyse exclusively Nuclear-layer serialization patterns in Mwotlap. In this language, it is very common about twenty per cent of the clauses in spontaneous speech - that a single verb phrase contains not only one verb lexeme, but two or three, and up to four verb radicals, chained together within a single syntactic phrase: (1)

Tö then

ke (ni-hö 3SG AO-paddle

ihöl return

lok) again

höw. down

'So he paddled his way back to the west.' After placing these structures in their syntactic context, we will analyse more specifically the way they handle such issues as valency and argument structure. This should not only allow for a better insight into the formal mechanisms of serialization, but also help formulate hypotheses on the semantic and pragmatic functions fulfilled by this strategy.

Constraints and creativity in the macro-verb strategies of Mwotlap

2.

109

The internal syntax of VPs and Nuclear-layer serialization

It may be useful to present the basic syntax of Mwotlap clauses, so that the reader can appreciate serializing strategies in their proper context.

2.1. The Verb phrase Mwotlap obeys a strict order of constituents SVO1. It is unusual within the North Vanuatu subgroup, in that neither the subject nor the object is crossreferenced on the verb form: both arguments are represented by a noun phrase or a free pronoun (sometimes a zero anaphora), with their function only indicated by their position in the clause. Mwotlap has lost the transitivizing morphology of its ancestors (e.g. POc suffix *-i): (2) a. No ISG

m-et

nek.

b. Nek

PFT-see

2SG

2sg

Ί saw you.'

m-et

no.

PFT-see

ISG

'You saw me.'

As far as the verb is concerned, it is obligatorily marked as finite by means of a Tense-Aspect-Mood marker. These ΤΑΜ markers, of which Mwotlap has no less than twenty-five ( F r a n c i s 2003), take the form either of a prefix, a pre-clitic, or a post-clitic. A few of these markers are discontinuous, being a combination of a prefix and a post-clitic: e.g. et-... te 'realis negative', te-... veh 'potential'; they can embrace several elements, which together form the VP. These 'bracketing' morphemes turn out to have a crucial role in the syntactic analysis, since they provide an efficient test for the delimitation of VPs in Mwotlap. For example, they make it easy to observe that Mwotlap VPs do not include the object: (3) a. Key 3 PL

{et-galeg NEGrmake

te)

n-ein.

NEG2

ART-house

'They don't build houses.' b. Key 3 PL

(ta-galeg POT ι-make

veh)

n-em.

POT2

ART-house

'They can build houses.'

110

Alexandre Frangois

2.2. Adjuncts: both a category and a function Thanks to the same discontinuous morphemes, it is possible to characterize lexemes or phrases in Mwotlap, whether they can surface inside or outside the VP. Whereas direct objects, as well as oblique complements, are excluded from the VP, several lexemes systematically appear within this verb phrase, immediately following the verb head: (4)

Kömyö

ν eh) talöw le-mtap. tomorrow in-morning 'You will be able to have a walk together once again tomorrow morning.' 2DU

(ta-tatal

tiwag

lok

se

POT,-walk together back again

POT2

Although all words in bold correspond, in English translation, to a single category called 'adverbs', it is obvious that they must be clearly distinguished for a language like Mwotlap: distributionally speaking, a difference must be made between those words which only fit VP-internal positions (e.g. tiwag 'together'), and those which cannot enter the VP, and take the slot of oblique complements (e.g. talöw 'tomorrow'). We reserve the term 'Adverb' for this second category, while the VP-internal words correspond to a specific function which we label 'Adjunct'. Semantically speaking, adjuncts have the function of a head-modifier, in very much the same way as adjectives modify the nominal head of an NP. Sentence (4) suggests that while a single verb phrase can perfectly include several adjuncts, only one verbal head is allowed at a time. To sum up, the structure of a verbal clause in Mwotlap obeys the following pattern: Subject

(ΤΑΜI verbal

head + Adjuncts TAM) Object + Complements

A crucial remark to be made, is that the syntactic slot of adjunct is not reserved to a few lexemes specialized in this function ("pure adjuncts"), like tiwag or lok in (4). In fact, several versatile lexemes, as well as whole categories, do fit the same verb-modifying position. For example, virtually all lexical adjectives, besides their function as noun-modifiers, can also modify semantically a verbal head, taking the slot of an adjunct : (5) a. na-lqöven qaqa ART-woman stupid 'a foolish woman' b. Imam (ma-hag qaqa ewe) l-em. Dad PFT-sit stupid just in-house 'Dad is staying idly at home.'

Constraints and creativity in the macro-verb strategies of Mwotlap

111

(6) a. nu-qul lawlaw ART-lamp bright 'a red lamp / a shining lamp' b. (Tog etet lawlaw) ke! proh look:DUP bright 3sg 'Stop watching her with those bright [i.e. greedy] eyes!' Although this is less frequent, some nouns appear in the same position of adjunct: (7)

Ke

{ma-hag

3SG PFT-sit

tuvusmet)

höw.

high.chief

down

'He is sitting cross-legged.' [lit. He is sitting high-chief] (8)

Tigsas ke (et-wot vu te), ke {mo-wot et). Jesus 3SG Ν EG!-born spirit NEG2 3SG PFT-born person 'Jesus Christ was not born a spirit, he was born a man.'

Through this use in the adjunct position, the set of lexical adjectives and the set of nouns both provide a stock of possible verb-modifiers, allowing new combinations to be built in order to express semantically complex processes.

2.3. Nuclear-layer serialization and the status of V 2 It is now possible to present the serializing sentence (1) above, repeated below: (1)

To ke (ni-hö möl lok) höw. then 3SG AO-paddle return again down 'So he paddled his way back to the west.'

This sentence may be seen as a typical instance of nuclear-layer serialization: a single verb phrase includes more than one verb radical here two. Tense-aspect-mood markers only appear once, affecting the complex verb phrase as a whole: prefixes come before the first verb, and post-clitics appear after the last verb or the last adjunct. Nothing can intervene between two serialized verbs, especially no object phrase; the object of the first verb Vi, if any, is either left implicit, or becomes the object of the whole verb phrase:

112

A lexandre Frangois

(9) a. Tali T.

{mi-tit PRET,-punch

tö)

Kevin.

PRET 2

K.

'Tali punched Kevin.' b. Tali T.

(mi-tit

teilten

to)

Kevin.

PRET,-punch

cry:DUP

PRET 2

K.

'Tali made Kevin cry by punching him.' The issue of argument structure and valency will be addressed in Section 3. At this point of the discussion, it may be relevant to ask the following question: in serializing structures like (1) and (9), should we consider that the verb phrase contains two verbal heads? or is the first verb still the unique head, whereas the second one would be described as its adjunct? The right answer to this question seems to be the second one: in a sequence (Vi+V2), the serialized verb V2, far from being a second head, is nothing more than an adjunct to the preceding verb Vi. The first reason for this claim is the high parallelism between serialized verbs and adjuncts: they take the same syntactic slot in the verb phrase, and both have the semantic function of modifying the first verb Vi. Whatever its internal complexity, the whole VP (Vi+adjunct, or V1+V2) can be understood as a mere development of Vi, with the same basic lexical meaning: for example, hag qaqa ewe 'sit just idly', as well as hag tuvusmel 'sit cross-legged', are nothing more than two possible ways to hag 'sit'; etet lawlaw 'watch greedily' refers to a way of watching; hö möl 'paddle back' is an instance of paddling; and tit teilten 'knock to make cry' is a sub-type of tit 'knock', but not a sub-type of ten 'cry'. 2 The latter remark can be reworded in more narrowly syntactic terms: the verbal head of a VP necessarily takes the same subject as the whole phrase of which it is the head. Thus Tali in (1) is both the subject of tit 'knock' and of the whole VP mi-tit teilten tö 'knocked in such a way to make cry'. On the contrary, the following verbs in an SVC are not subject to the same syntactic constraint regarding their subject: although both verbs in (1) do have the same subject, this is not the case in (9), where the only suitable head is the first verb V,. Among other arguments which will not be detailed here, the asymmetry which we claim exists between Vi and V2 is confirmed by a difference in their lexical inventory. Whereas all verb lexemes of Mwotlap can be the head (Vi) of a serializing VP, the subsequent position (V2) is restricted to a much smaller set of verbs, probably a few dozen; for example, such common verbs as van 'go', lep 'take', et 'see' never appear in the position of

Constraints and creativity in the macro-verb strategies of Mwotlap

113

V 2 . And even the verbs which can fit both positions (Vi or V 2 ) sometimes show traces of asymmetry in their forms: e.g. the verb 'know' has the form eglal when found in Vi, but veglal when in V 2 ; the duplicated form of the verb sok 'search' is soksok when in Vi, but sosok when in V 2 , and similarly tey 'hold' duplicates regularly as teytey when it is the head of the VP, but as tetey when it is used as a head-modifier. All these remarks tend to demonstrate that what looks, at first sight, like a simple chain of verbs [Vi-V 2 V 3 ...] placed on the same level, involves in fact an asymmetrical relationship, that of a head followed by its modifiers. Consequently, the best way to analyse serial verbs in Mwotlap follows the model Head + Adjunct. 3 This pattern fits well in the category identified by Durie (1997) as asymmetrical serial constructions, whereby a single verbal head (Vi) is modified by a limited set of verbal modifiers (V2).

2.4. One or several actions? It is perfectly possible that this analysis of Mwotlap does not match the structures of other serializing languages, in which SVCs would basically allow for a string of successive actions performed by the same subject. For example, the following sentence in the Papuan language Barai 4 could well lead to the opposite conclusion, i.e. one VP having several heads: BAR Ε man

ije fu the

a-nafa-fu-o

kan-ia buvua

3SG child-PL-3SG-POSS kill-3PL cut.up

i. eat

'The man killed, cut up (and) ate his children.' Now, the reader must realize that such a string of actions 5 would never be coded by a serial structure in Mwotlap. Contrary to what is suggested by the general label 'serial verbs', this language will combine verbs in a single VP to refer to a single action, and hardly ever more. When Mwotlap needs to describe a series of actions, it does not use serialization, but coordination, by means of such conjunctions as ba 'and' or to 'so, then' - in a way very similar to European languages. The following example should help fix this important point. It is the narration of a series of successive actions undertaken by an individual (a healer called Boyboy) within a short period of time. Although this is typically the kind of context in which many serializing languages would make use of serial verbs, it is remarkable that Mwotlap codes all these actions by means of distinct clauses, separated by prosodic pauses 6 and/or coordinators (underlined): all these devices are typical of non-serializing languages.

114

Alexandre

(10)

Böyböy

Frangois

me-yem, ma-hap, me-hew tey wonwon; PFT-climb PFT-pick PFT-descend hold intact:DUP 'B. climbed (up the coconut-tree), picked (some coconuts), brought them down intact',

B.

to

ke

ni-ey,

to

ke

ni-van

tey

me

l-em;

then

3sg

AO-husk

then

3sg

AO-go

hold

hither

in-house

then he husked them, then he brought them home; ke ni-tot ne-tenge nan, ke ni-van

tey

me,

3SG

hold

hither

AO-chop

ART-leaf

ANA

3SG

AO-go

he cut the relevant (medicinal) leaves, he brought them here; tö ni-böl madamdaw nö-göygöyi qetenge nan, ... then

AO-hammer

soft

ART-roots

plant

ANA

then he crushed their roots soft, ... to lep me to nok in

to

nok

we

then

then

lSG

AO:good

take

hither

then

lSG

AO:drink

then gave it to me, then I drank it, then I got better, to ni-bah. then

AO-finish

and this is it.' Nevertheless, serialization patterns are not totally absent from the last citation; they appear in bold. In each case, what we observe is a single VP, inflected with a single ΤΑΜ prefix {me-, ni-), and including more than one verb radical - which corresponds well to the formal definition of 'serialization'. But from the semantic point of view, we claim that each of these serializing VPs points to a single, minimal action, with no possibility to split it into distinct phases in time. Here they are repeated: (11)

Böyböy

{me-hew

B.

PFT-descend hold

tey

wonwon). intact:DUP

'Boyboy brought them down without-breaking-them.' (12)

ke

{ni-van

3SG AO-go

tey)

me

hold

hither in-house

l-em.

'He brought them home.' (13)

tö

(ni-böl

then AO-hammer

madamdaw)

nö-göygöyi

qetenge

nan

soft

ART-roots

plant

ANA

'then he softened the roots of the plant by hammering them' Without going into too much detail here, it is instructive to notice that the internal structure of serializing VPs in Mwotlap confirms our claim that

Constraints and creativity in the macro-verb strategies of Mwotlap

115

they cannot be interpreted as a series of actions. Indeed, it is common in most serializing languages, to translate such a verb as Eng. 'bring' by a string of two actions Vi = 'take' + V2 = 'come'; e.g. pidgin Bislama follows iconically the order of the two phases: BSL

Hem

i

3SG

PRD carry

karem

kokonas

i

coconut

PRD come

kam.

'He brought a coconut.' In these languages, it is still possible to consider this string of two verbs as reflecting two successive (phases of) actions. On the contrary, Mwotlap codes the same idea using a non-iconic order of terms [V| = verb of movement + V2 = 'hold'], in such a way that the "« verbs, η actions" interpretation becomes impossible. The only reading possible for sentences (11) and (12) corresponds not to successive actions, but to simultaneous facets of a single action: Boyboy descends (from the tree) holding (the coconuts) and keeping them in one piece.

2.5. Summary: An optical illusion We can now summarize the results of these first observations about Mwotlap. In this language, a single verb phrase may include more than one verbal lexeme at a time, with no other element intervening. The surface pattern (Vi+V2+V3...) vp recalls similar strings of verbs in certain languages - like Tariana (Aikhenvald 1999) - and suggests the term of '(nuclearlayer) verb serialization' for Mwotlap. However, a deeper analysis shows that the term 'serial verb' may well be an optical illusion. First, there is a formal and semantic asymmetry between V,, the unique head of the verb phrase, and the following verbs, whose basic role is to modify this head. In this sense, serialized verbs enter a syntactic slot we have called 'adjunct'; far from being exclusive to verbs, the position of adjunct is also open to adjectives and nouns, plus many lexemes ("pure adjuncts") exclusive to this function. From the semantic point of view, a serial verb string in Mwotlap cannot refer to several distinct actions - in which case, coordination is used - but to a single action, undertaken by one subject at a given point in time. The internal complexity of these SVCs allows coding this single action under several of its facets. Focusing on the issue of valency and argument structure, the second section of this paper will examine the way this complexity is handled by the socalled 'serializing' strategy.

116 3.

Alexandre Frangois Serial verbs and transitivity issues

After this first exploration of Mwotlap SVCs, we would like to focus our attention on the relations existing between the argument structure of a serialized VP and that of its components. Indeed, we have already seen that a multi-verb VP behaves externally like any simple verb, having one subject on its left, and no more than one object on its right. Knowing that Mwotlap does not allow for double-object structures (Eng. I gave him a pen), nor is it possible to insert any object NP between two serialized verbs, syntactic conflicts may arise in the combination of two transitive verbs (see [o] below). In reality, cases of valency conflict seldom occur in Mwotlap SVCs. Most of the time, what is observed is a regular capacity to merge the argument structures of two verbs into that of a new, composite 'macro-verb'. The following paragraphs will try and establish a syntactic classification of Mwotlap SVCs, according to the argument structure of their input elements, and of the output verb phrase. Basically, the main relevant opposition is that between intransitive and transitive verbs: for instance, we will see in which cases the combination of two intransitive verbs leads to the formation of an intransitive or a transitive VP. But for this analysis, finer criteria will be needed, such as the sameness or difference between, say, the object of Vi and the subject of V2, etc. As a consequence, each type in the following classification will be presented with a simple formula, using small letters (x, y, z) for arguments, and an SVO convention; e.g. x-Vi means Ύι is an intransitive verb having a subject x'; x-V2-y means 'V 2 is a transitive verb having a subject χ and an object y\ A short note is necessary here about the category "adjective". In Mwotlap as in many Austronesian languages, adjectives follow the same syntactic patterns as intransitive verbs, in most contexts: both categories are directly predicative, they combine with the same aspect-mood markers, and so on; the only position where they can be distributionally contrasted is inside the noun phrase, since only an adjective can modify a noun directly. Because our study is concerned with serial constructions in predicative phrases, where adjectives and verbs are merged, it will here be legitimate to regard adjectives as a sub-class of intransitive verbs, following the pattern x-V

Constraints and creativity in the macro-verb strategies of Mwotlap

117

3.1. Both verbs are intransitive

x-Vi + j c - V 2 ->-x-[VP]

[a]

Both verbs can be intransitive. If their individual subject is semantically the same, then the output is normally an intransitive macro-verb. Semantically, the subject can be said to perform both 'actions' - or, to be more precise, both facets of the same action - at the same time. (14)

Nok lSG

(taq

mitiy

tusu).

AO:bend.down sleep

a bit

'Let me have a nap.' (15)

Gen

(yow

llN:PL

AO:jump stand

tig)

höw

le-be!

down

in-water

'Let's jump into the river!' (16)

Inti

{ma-kal

son:2SG PFT-crawl

qelen). disappear

'Your baby's crawled away.' As was mentioned before, nothing prevents us from adding to the present list those cases in which the adjunct V 2 is an adjective (or "adjectival verb"), with the same semantic subject as Vi: (17)

Ke

(me-ten

3SG

PFT-cry

magaysen). sad

'He was crying miserably.'

[b]

s - V | + x-V2 -> x-[VP]-x

A rather strange pattern provides an exception to [a] above, since it shows two intransitive verbs with the same subject, resulting in a formally transitive macro-verb. The object of this VP has the same reference as its subject, which corresponds, incidentally, to the coding of reflexive verbs in Mwotlap. In fact, this pattern [b] occurs only in familiar speech, with basically two verbs in adjunct position: mat 'die' and its slangish counterpart mem 'piss'. The basic idea is that the subject χ is performing an (intransitive) action V, in such an intense manner, that it makes him metaphorically die... or, less seriously, urinate. Practically, this structure is used as a jocular intensifier for certain intransitive actions:

118

Alexandre Frangois

(18)

Nek

(me-yeye

mat)

nek

ae

egen!

2SG

PFT-laugh

die

2SG

ANA

now

'You're laughing yourself to death!' (18)' Nek (me-yeye mem) nek ae egen! 2SG PFT-laugh piss 2SG ANA now {slang) 'This makes you piss with laughter!' In order to be understood, this isolated pattern [b] should be compared to others, more productive. For instance, it can be interpreted with reference to [e] below - except that in this case x = y . Or it may also be compared with the causative structure ([j] below), with which both verbs mat and mem are usually associated, in phrases like { χ Vi maty } 'x kills y (through the action Vi)' or { χ Vi mem y } 'x makes y piss (through a violent action V b e.g. knock down or scare)'. What is particular in (18)(18)', is that V, is intransitive, and the patient coincides with the actor 8 . [c]

U-Vi + =V 2 ->x-[VPl

It sometimes happens that the logical subject of V 2 is not just χ (the subject of Vi), but rather corresponds to the predicate structure x-Vi as a whole. The verb comments on the manner in which the action (x-)Vi is carried out. 9 This is often the case when the serialized element is an adjective, which may then be said to work as an "adverb": (19)

Na-day

nono-n

(me-plag

ART-blood

of-3SG

PFT-run

lililwo). big:DUP

'His blood flowed abundantly.' (20)

No-qo

e

ART-pig ANA

ke

(ma-mat

3SG

PFT-die

hiywe). be.true

'The pig was well and truly dead.' In example (20), it is clear that the subject of V 2 'be true' is not the pig itself, but 'the dying of the pig': the logical structure of these sentences is thus {V 2 (V,(x))}, involving a second-order predicate. See also [h]. [d]

U-Vi + 0 - V 2 - > x - i V P l

Sometimes, an intransitive Vi is combined with an impersonal verb V2, with zero-valency 10 , e.g. qön 'be night', myen 'be daylight'. The latter does not affect the valency of the main verb, and adds only the meaning '(do Vi) until it is night [resp. day]'.

Constraints and creativity in the macro-verb strategies of Mwotlap (21)

Tita (ni-hag qön) l-em. Mum AO-sit be night in-house 'Mum will stay at home all day long.'

(22)

Kimi

119

(ma-lak meyen tö) ? P R E T J -dance be day PRET2 'Did you dance all night long?'

2pl

[e]

|X-V,+j;-V2

x-\VP]-y

Two intransitive verbs may merge into a transitive macro-verb: this happens when the semantic subject of V2 is different from that of Vi. In this pattern, the general meaning is "x performs/undergoes an intransitive action Vi, which results in another element y undergoing in turn a transformation (V2)"; the syntactic output of this combination is a transitive macro-verb (V1-V2) with a causative meaning. Notice that the present pattern normally does not concern animate actors, but rather natural forces: with an animate subject, the feature [control] would normally result in the choice of a transitive Vi. (23)

Na-lo

(ni-hey

simsim)

n-aes.

ART-sun AO-shine melt:DUP ART-ice

'The sun melts the ice (by shining).' (24)

Ne-len {mi-yip hal-yak) na-kat. ART-wind PFT-blow fly-away ART-cards 'The wind blew the cards away.'

(25)

Ni-yiy {mi-yiy sisisgoy) na-mtig. ART-quake PFT-quake fall:DUP ART-coconut 'The earthquake made the coconut trees fall down.'

Even when it is animate, the subject is generally not agentive - but note ex. (27): (26)

Nek (mi-tig melemleg) na-lo den kemem. 2sg PFT-stand black ART-sun from 1ex:pl 'Standing as you are, you're hiding the sun from us.' [lit. You're standing dark the sun from us.]

(27)

(Gengen AO:eat:DUP

may may) strong

na-taybe! ART-body:2SG

'Eat well, to strengthen your body.' [lit. Eat strong your body.]

120

A lexandre

Frangois

To our knowledge, pattern [e] was never witnessed in any other serializing language which has been described to date. It must be distinguished from the so-called 'switch-subject serialization' ([j]), since the first verb is intransitive: we would call it "low agency causative serialization". Most remarkably, this structure contradicts the claim usually made about serialization, that "serial verbs share at least one (...) argument" (Durie 1997: 291): Mwotlap proves that two predicates, having no argument in common, can perfectly merge into a single serial verb construction. 11 The structure, however, is rare: our corpus shows little more than the few examples cited here. [f]

Combination of several verbs, none being transitive

Any combination of the above formulas, involving more than two verbs, leads to the expected result. We will give here only one example, which combines [c] and [d] above: the combination of an intransitive verb + an adjunct commenting on this first action + an impersonal adjunct, results in an intransitive macro-verb. (28)

Köyö (S-S.P.R. qaqa qöii). 3DU AO:roam:DUP stupid be night 'They spend the whole day aimlessly wandering.' 1 2

The formula corresponding to this example would be: jc-Vi + =V 2 + 0-V3 ^>x-[VPl Such combinations are very common in everyday speech, and there even seems to be a preference for this kind of multi-verb serialization in colloquial discourse and slang.

3.2. Only one verb is transitive When only one verb is transitive, then the result of the combination is invariably a transitive VP. [g]

x-Vi-y + x-V2

- > x-[VI']-y

Despite its simplicity, this pattern seldom occurs in Mwotlap: usually, the combination of a transitive head with an intransitive adjunct is interpreted as a 'switch-subject serialization' ([j] below), i.e. a causative structure in

Constraints and creativity in the macro-verb strategies of Mwotlap

121

which the subject of V2 is understood to be y, notx. However, the following example is compatible with the interpretation [g]: (29)

Ye

(ti-tiok

magaysen)

nek?

who

FUT-see.off

sad

2SG

'Who will have the sad role to see you off?' [x is 'sad'] (or maybe: 'who will see you off, you poor fellow...') [y is 'sad'] [h]

x-V,-y + = V2 -> x-[VP]-y

An intransitive verb, or often an adjective, can describe the manner of a transitive action Vi. In comparison with other patterns, what appears here is that the logical argument of predicate V2 is neither χ nor y, but the whole event (x-Vi-y): this is another case of so-called 'ambient serialization' (see [c]). (30) Na-bago ( x-[VP]-z

A less frequent variant of the causative pattern we have just presented also combines a first transitive verb Vi with an intransitive verb V 2 ; but the subject of V 2 is by itself a new element (z), corresponding neither to the subject (x) nor to the object (y) of Vj. Since there is only one object slot available for two distinct applicants (y / z), we are facing here the first case of syntactic conflict between arguments. The solution adopted by Mwotlap is normally to drop the object y of Vi, thus leading to a transitive verb oriented towards z, the experiencer of V 2 : {x-[VP]-z }. Often, y either features as a topic in the same sentence [see also ex. (48) below], or is easily reconstructed from the discourse context: (40)

Tita nonon mi-gil ne-qyö-n, {mi-gil wawah) na-taqme-n. mother his PFT-dig ART-grave-3SG PFT-dig clean ART-body-3SG 'His mother dug his grave, cleaning (the earth) off his body.' [lit. she dug his grave, dug clean his body]

(41)

Ke

ma-yah

ne-sem

nen

3SG PFT-scrape ART-cowrie that

etö

(ni-yah

then AO-scrape

ölöt) call:DUP

ne-yedep egen. ART-Pritchardia now 'And as she was scraping those cowries (to make shell-money), her scraping made the palm leaf screech.' [lit. she scraped screech the leaf]

124

Alexandre Frangois

(42)

Ke

jc-[dance calmly with]-j

rule [h]

x-[dance calmly with]-y + = good -> jc-[just dance calmly with]-j

127

The result of this combination of rules, as expected, is a transitive macro-verb, whose subject is the subject of the head (by essential property of head), and whose object is the object of the only transitive verb of the string, namely V2 'hold'. A similar combination involves the rare pattern [k] above, in which three semantic arguments had to share only two syntactic slots. Once again, the rejected element is y (the object of Vi), which only appears as a sentence-initial topic: (48)

Yebek en, kemem (öl tog-yon magaysen) ige susu. devil ANA 1 E X : P L AOxall stay-quiet sad PL children [lit. That Yebek (legendary monster), we call silent sad the kids.] 'That Yebek, we call (him) to scare the kids silent.'

This sentence can be read as a development of (37) above: instead of the all-purpose verb ak 'make', a more specific verb is used to express the action which is performed to achieve the goal tog-yon '(the children) be quiet'. But contrary to ordinary causative sentences [e.g. (13)], the object of Vi is here different from the global undergoer (the "causee"): parents are naming the monster in order to calm the children down. In other words: rule [k]

x-call-y + z-be.quiet

—> jc-[shut up]-z

Now, notice the third predicate in the same series. Evidently, once a complex transitive verb has been construed, it is possible to add an extra verb V3, and thus build a new transitive macro-verb, according to rule [j] above: rule [j]

x-[shut up]-z + z-sad

—» x-[terrorize]-z

The reader will appreciate how much concision is made possible by the use of a serial strategy: this is how a single VP can combine such unrelated notions as 'call a name', 'be silent', and 'be sad'. The sentence itself illustrates a common case of adult cruelty towards their children... but this is another issue.

128

Alexandre Frangois

3.3. Both verbs are transitive We shall now mention the last combination patterns, those which involve two transitive verbs. [n] JC-VI-y + x-V2-y

x-[VP]-j

No conflict arises if both verbs share the same subject and the same object: the result is a similar macro-verb. The function of V2 is obviously not a syntactic one - e.g. valency-increasing- but consists in adding some semantic features to Vi. (49)

(50)

[O]

Key

(et-

no.

3PL

NEGR

et veglal te) see know NEG 2 'They did not recognize me.'

LSG

Nok

en.

LSG

{tivig veteg bah) ke bury (leave) PRIOR 3SG 'Let me first bury (and leave) him.'

ANA

x-Vi-y + x-V2-z

JT-[VP]-z

The last case we will be facing does involve a typical conflict between arguments, in a way similar to pattern [k] above. Each transitive verb has its own object, which results in two items applying for the single object-slot of the macro-verb. Once again, the usual strategy used by Mwotlap is to select the object of V 2 (z) to that position, thus leaving implicit the object of V,(y)· In (51) below, 'beer' is the semantic object of V; 'drink'. But because V 2 'accompany' also has its own object (z = 'kava'), the word 'beer' is formally excluded from the clause, and can only be retrieved from the immediately preceding context: (51) Nek ta-kale veh mi ni-bia... a so {in biyin 2SG

POT r chock

ewe)

na-ga.

just

ART-kava

POT2

with ART-beer that is drink

accompany

' [when drinking kava] one may "force it down" with beer: that means nothing but (drink together with) kava.' Such a syntactic constraint, which makes it necessary for the speaker to introduce an item (here y = 'beer') in a preceding sentence, shows that

Constraints and creativity in the macro-verb strategies of Mwotlap

129

Mwotlap Nuclear-layer SVCs must be analysed in a very different way from Core-layer SVCs, even in closely-related languages. For example, Paamese has the possibility, thanks to its Core-layer SVCs, to express each verb with its own object, in a structure which Crowley (1987) labels 'multiple-object serialization'. Obviously it does not have the same consequences for syntax and discourse strategies. Compare Mwotlap (51) with this Paamese sentence (1987: 39): PAA Inau na-mun sin dal oai. lSG lSG:REAL-drink gin 3SG:REAL:accompany water Ί drank gin with water.' [lit. I drank gin it accompanied water] We do not think the "rules" we are defining for Mwotlap have to be conceived in terms of a purely formal constraint, e.g. Mwotlap forbidding (vs Paamese allowing) the insertion of an object inside SVCs. It seems more convincing to consider that these two languages, despite their vicinity, employ two drastically different structures, including in semantic terms. In Paamese, the verb mun 'drink' remains perfectly oriented towards its own, expected object, without being much affected by the presence or absence of a serial structure - in a way typical of Core-layer SVCs. Conversely, Mwotlap SVCs do considerably affect the diathetic orientation of its verbs, in such a way that in 'drink' in (51) could be said to be no longer oriented towards its "semantic object" ('beer'), but rather has become part of a macro-verb, the only function of which is to specify it semantically. Considering in biyin as a whole, we must admit that the serializing operation has blurred the syntactic link between 'drink' and 'beer', in such a way that the macro-verb (drink-accompany) is now semantically oriented towards the 'kava': it describes a certain action that the actor performs in relation with kava, e.g. increasing its effect, improving its taste, etc. In the framework of this new action, the item 'beer' is neither an object nor - w e claima patient, but has now a sort of semi-presence, in the same way as a peripheral argument (instrument, locative...) could have. The hypothesis we are suggesting, and which has theoretical consequences, is that when an SVC pattern has the effect of demoting an object from its position, then this demotion does not only take place at a syntactic ('surface-structure') level, which would leave intact its semantic role as a patient. In Mwotlap SVCs, any change altering the syntactic status of an object also makes it necessary to modify its semantic properties, in such a way that what was once a patient is redefined as a peripheral case role (e.g. instrument). Thus compare this sentence, in which na-hat is object and patient:

130

Alexandre Frangois

(52) a. (Hey) na-hat AO:wear ART-hat 'Put that hat on.'

anen. that

...with the following one, in which the same action (hey 'put on, wear') has been integrated as the first element of a macro-verb hey goy, and has therefore lost its object: b. (Hey

goy)

ni-qti

MI

na-hat

AO:wear (cover) ART-head:2SG with ART-hat 'Cover your head with that hat.'

anen.

that

Since it has been rejected from the position of object of Vi, the item 'hat' is no longer a patient, but has been redefined as an instrument (preposition mi) within a new action: lit. 'You (wear-(something)-on) your head with that hat' 16 . Such sentences illustrate how Mwotlap SVCs can involve a complete reorganization of argument structure and semantic roles in the clause17. [p] Combination of several verbs, two of which are transitive Finally, the reader will not be surprised to learn that our corpus shows examples of serial chains with more than one transitive verb. (53)

Key

(lam

3PL

AO:beat die

tö then

mat veteg) höw

nö-lömgep

(leave) down ART-boy

en, ANA

qeleqlen. AO:disappear:DUP

'They got rid of the boy by beating him to death, and escaped.' This example apparently brings about a new, complex formula, in which the same argument (y= 'the boy') is simultaneously the object of V|, the subject of V2, and again the object of V3. But as usual, we find no formal clue whatsoever which would encode the internal structure of the macroverb. How can the hearer cope with such complexity, and consistently assign the right semantic role to the right argument? As was stated in [m], this kind of multi-verbal combinations can in fact always be analysed as the result of several binary expansions, starting from the head. It is then easy to retrieve the patterns we established earlier in this study: rule [j]

x-beat-y + y-diQ

—> x-[kill]-_y

rule [n]

x-[kill]-y + jc-leave-y

—» jc-[get rid of]-j

Constraints and creativity in the macro-verb strategies of Mwotlap

131

3.4. Summary: Few rules, strict rules The function of Mwotlap macro-verbs is to define a meaningful representation of a single action, performed by a given subject at a particular point in time, exactly the same way as a simple verb would do. This is why, far from linking together any two predicates the same way as would do coordination, Mwotlap serial strategies obey strict rules regarding the semantic compatibility between its components, as well as their syntactic organization. We have shown it was possible to reduce the high diversity of all SVC instances to a finite number of thirteen regular patterns, each one being easily formalized. These patterns all consist in observing how the argument structure of a verb Vi and a verb V 2 regularly combine, in order - f o r the linguist as well as the speaker- to calculate what the argument structure of the resulting 'macro-verb' will be. All these rules 18 are summarized in the following chart, each one corresponding to a single square. In bold, italic letters we represent those cases which are most frequent or productive in Mwotlap discourse, the other ones being much rarer. The two rows correspond to the valency (intransitive vs transitive) of Vi; each column informs on the status of V 2 . Notice that the use of letter ζ is only relevant when it contrasts with y (= two distinct applicants for one object slot): hence the shaded squares. Table 2. Combination of intransitive distribution of arguments X-V2

0-V2

and

=v2 X-\VP\

x-V,

X-[VP\

X-[VP]

x-Vry

x-[VP]-y

x-[VP]-y x-{VP\-y

/ or transitive verbs: Effects upon the

y-v2

x- V2-y

x-[vp]-y

x-[vp]-y

x-\VP\-y

x-[VP]-y

z- V2

x- V2-z

x-[VP]-z

x-[VP]-z

While detailed discussions have already been given throughout the preceding pages, we will only add a few general comments to this chart. The combination of two verbs in Mwotlap follow a small number of very regular principles, some of which are obvious, while others are less so: -

The number of arguments of the macro-verb is never higher than two.

-

The number of arguments of the macro-verb is never lower than the sum of distinct arguments associated with the input verbs.

-

The subject of the macro-verb is systematically the same as for V] (i.e. χ).

132

Alexandre

Frangois

-

The object of the macro-verb is any argument other than x, associated to the input verbs (i.e. y or z).

-

In case of conflict for the object position, any argument introduced by V 2 (i.e. z, either its semantic subject or object) will have priority over the object of Vj.

-

In other words, the macro-verb adopts the primary orientation 19 of Vi; its secondary orientation depends mainly on V2.

These rules are most probably operated by the speaker, while creating new combinations and building his discourse - but also by the hearer, while endeavouring to associate the right referents with the right syntactic slots. The high number of attested patterns does not mean that any combination of two verbs will be possible in Mwotlap. First, as we said already, the inventory of the adjunct V2 is limited, even for the patterns which are regular (e.g. no example of SVC with V2= 'take', 'see', 'say', etc.). Secondly, one can observe that several "potential" combinations are impossible in Mwotlap: *{ jc-V, + _y-V2-;c }; *{x-V, + >·-V2-z }; *{x-Vry + z-V 2 -y}; *{ x-Vi-y + y-V2-z }· The last one, in particular, is worthy of notice, since it would have corresponded to a plausible scenario of X acting upon Y, in order for this Y to act upon Z; this corresponds to an "agentive causative" structure (Fr.factitif, opp. causatif), e.g. { Ix commandyouY + youY shut the door ζ * I command-shut... }. This pattern, which is perfectly possible with Core-layer SVCs20, is impossible with the serial structures of Mwotlap: subordination structures will be used instead. In brief, serialization strategies are not so free and random as one could believe at first sight: not only is the position of adjunct restricted to a finite inventory of verbs, but also the combinations themselves are governed by strict principles. A noteworthy paradox is that, whilst all these constraints and limitations tend to restrain haphazard innovations, they also provide efficient tools to invent novel macro-verbs, with a lesser chance of ambiguity, and a higher guarantee that the communication process will be successful.

Constraints and creativity in the macro-verb strategies of Mwotlap 4.

133

Co-lexicalization and categorial change

4.1. Synchronic heterogeneity and language dynamics The last section of this paper will be concerned with diachronic change and the language dynamics implied by the serial patterns of Mwotlap. Indeed, one would get too simplistic an idea of these structures if they were to be described as a homogeneous, entirely productive set of transformative rules, applying to any lexical unit in synchrony; and conversely, it would also be erroneous to point to the other extreme, as if every 'macro-verb' were already lexicalized and learnt as such by the speakers. After observing the way these SVCs work in Mwotlap, it seems much more accurate to synthesize these two possible interpretations. Serial verbs in Mwotlap are better defined as a heterogeneous linguistic device, involving several layers in the lexicon, and different depths in time: -

Many combinations are "already lexicalized" from the speaker's point of view, in such a way that the behaviour of the bundle cannot be unambiguously derived from the meaning of its components. These combinations are learnt as they are, as would be the case for any lexeme or idiom - cf. Eng. give up. - » e.g. matmöl 'faint' < mat 'die' + möl 'return'.

-

A great number of combinations, despite being already there in the language, remain relatively transparent to the hearer, i.e. could be interpreted as the "free" association of a Verb (known separately) + an adjunct (whose meaning is easily reconstructible, and thus ready to be reused) - cf. Eng. climb up. —» e.g. yow tig 'jump or dive while remaining upright' < yow 'jump' + tig 'stand'.

-

Out of attested combinations, especially the transparent ones, the speaker draws a set of rules. These emerging principles can be syntactic, as we saw earlier, or semantic; they may be general laws governing all serial structures, or more specific rules associated with a particular subset of verbs (e.g. verbs of movement, of speech...), or even with a single lexeme. —> e.g. The lexeme mat 'die, dead' can be used as an adjunct: . combined with a transitive, agentive verb, it will mean that the object dies; thus the macro-verb refers to a way of killing: e.g. nit mat /bite+die/ ' x bite to death Y \

134 .

.

-

Alexandre Frangois combined with an intransitive, active verb, it will mean that the subject metaphorically kills himself in doing an action, i.e. performs it in an intense manner (jocular use); the subject will be repeated as an object of the same verb ([b]): e.g. lakmat /dance+die/ 'χ dance furiously χ'. combined with an intransitive, non-active verb, mat is reduplicated, and adds the meaning 'quietly'; the result is intransitive: e.g. hag matmat /sit+die:DUP/ ' x sit still'.

A certain degree of freedom is allowed within the limits of (and also thanks to) these rules. Far from damping down linguistic creativity, the high specificity of these constraints, both in syntactic and semantic terms, makes it easy for the speaker to create - and for the hearer to understand - novel Verb-Adjunct combinations. —» e.g. Regarding mat 'die, dead', any new combination (Vi + mat) can easily be associated with one of its already attested meanings: woh mat /hit+die/ could be created for 'χ hit to death Y ' teii mat /cry+die/ could be created for ' x cry hard x ' taq matmat /stoop+die:DUP/ could be created for ' x stoop still'.

These novel combinations will in turn take part in the definition of slightly different rules, which will be the cause of syntactic changes and semantic shifts through time. The remainder of this paper will illustrate the two evolution paths most commonly found in Mwotlap: -

the co-lexicalization of serial verbs, close to lexical compounding;

-

the emergence of productive adjuncts.

4.2. The lexicalization of macro-verbs The tendency for serialized verbs to co-lexicalize, i.e. eventually behave like a single lexeme, has long been acknowledged (e.g. Crowley 1987: 61). This is a natural phenomenon, especially when one realizes that languages are more than just a list of isolate lexemes, and always involve a high number of idioms, phrases or even whole clauses as part of their resources (Pawley & Syder 1983). Indeed, it would be clearly artificial to see all macro-verbs as the analysable combination, performed in situ by the speaker, of two autonomous lexemes: in the same way as show off or give up have become opaque to the speakers of English and are just memorized

Constraints and creativity in the macro-verb strategies of Mwotlap

135

as a single unit, SVCs in Mwotlap also consist massively in "fixed" combinations that owe nothing to the creativity of the modern speaker. Although this is an informal figure, we would personally estimate that in one day of linguistic interaction in Mwotlap, about 90 per cent of (VerbAdjunct) combinations are nothing but the repetition of combinations already heard by the speaker - whereas only 10 per cent(?) are novel associations which are created by applying productive rules on this matter 21 . Among all these 'already heard' combinations, many will eventually carry some semes which are not necessarily present or predictable from their components. A small selection of them is given below: (54)

ak 'make'+go_y '(cover, obstruct...)'[see fn. 17] —» akgoy 'reserve s.th. for o.s., put a taboo on s.th.; reserve (a woman) as o.'s future wife, get engaged to'

(55)

dem 'think' + liwo 'big' —» dem liwo 'consider with respect; be worried about; boast, show o f f

(56)

dem 'think' + veteg 'leave' —» dem veteg 'forget on purpose: give up, forgive; omit'

(57)

tog 'live somewhere, stay for several days' + qön 'be night' [see 21] -» tog qön 'go for a picnic, spend the afternoon somewhere'

(58)

tot 'chop' +gal 'lie, tell lies' —> tot gal 'notch wood; sculpt, carve an image; portray, take a photo or video o f .

Obviously, such semantic shifts make it necessary to consider these socalled "serial verbs" as single lexemes (written in one word?), and thus treat them as separate entries in a dictionary of Mwotlap (in preparation). The question arises whether serial verbs in Mwotlap, or more generally (Verb-Adjunct) combinations, should be considered as a kind of verb compounding. We do not exclude this interpretation, considering it at least to be better - as far as Mwotlap is concerned - than the purely serializing interpretation [see fn.3]. Nevertheless, three reasons prevent us from speaking of lexical compounding: 1 )A phonological criterion: Whatever their semantic evolution, Verb and Adjunct are always treated as two distinct phonological words, as is shown by the syllabic template mapping on each element separately 22 . Compare ni- 'prefix 3SG' + v(e)teg 'leave' —> ni-pteg ('he leaves'); but

136

Alexandre Franqois

hö 'paddle' + v(e)teg 'leave' —» hö veteg (*hö-pteg) 'paddle away from, quit'. 23 2) A syntactic criterion: Under certain conditions, it happens that some other adjunct may insert between the two elements of a "compound". 3) A semantic criterion: As the reader will have noticed throughout this paper, it is often difficult to draw the line between those cases in which it is 'clear' that (Verb + Adjunct) form a single lexical unit, and those cases in which it is 'clear' that it consists of two autonomous notions. In order to describe the facts of Mwotlap, it is not necessary to follow the old-fashioned definition of the 'lexeme' as coinciding with a single word; what we have here are whole phrases (Verb + Adjunct) which are more or less (co-)lexicalized as a fixed formula.

4.3. Intransitive vs transitive adjuncts The second diachronic path which is often followed by Mwotlap SVCs, is for the second verb V 2 either to undergo a radical semantic shift as opposed to its use as a verbal head, or sometimes completely lose the latter, and specialize as a pure adjunct.

4.3.1. Semantic split Verb / Adjunct A good example of a semantic shift affecting a serialized verb, is provided by the verb tey: as a main verb, it normally means 'hold, have in o.'s hands'. Used as an adjunct with a movement verb Vi, tey will help express such meanings as 'bring, take away (s.th. somewhere)'. The basic meaning 'hold' is maintained in these examples [see (44)], but widens to cover any value ο f X m o v e s with Y, including 'accompany (s.o. somewhere)': (59)

Hiqiyig

(ni-hö

someone AO-paddle

tey)

tital

hold

Mum

'Someone takes Mum in their canoe!' This combination is not restricted to movement verbs; with other verbs too, tey takes a general comitative meaning: (60)

No

(ta-lak

lSG FUT-dance

tey

qiyig)

hold

IMM

na-savat. ART-shoes

Ί will dance with my shoes on.' (= on my feet, not *in my hands)

Constraints and creativity in the macro-verb strategies of Mwotlap

137

Finally, the same verb tey sometimes encodes a more abstract relation f(x,y), with no trace of the original, lexical meaning 'hold in o.'s hands': (61)

Key

(ne-mle).

—» Key

3PL

STA-SIOW

3PL

(ne-mle tey) na-mwumwu namuy. STA-slow hold ART-work their 'They are slow.' 'They are slow AT their work.' The kind of 'semantic bleaching' involved here suggests the possibility of a shift from a lexical, semantically specific use of tey (x holds y in χ ' s hands), to a 'grammatical use' as a "transitivizer morpheme" (by performing an action V1( χ affects y in some manner) 24 .

4.3.2. Emergence of pure

adjuncts

Ultimately, the two uses (as a verb Φ as an adjunct) of the same lexeme will be felt to belong to two distinct linguistic units; and the decay which may affect, say, the verb will not affect the homophonous adjunct. As a consequence, the lexeme eventually leaves the category of verbs, and resembles the 'pure adjuncts' already existing in the language [cf. tiwag in ex. (4)]. The same evolutionary path was described for Paamese nuclearlayer SVCs, by Crowley (1982: 167): The meanings of these forms as verb phrase heads and as adjuncts diverged so widely that they were no longer felt to represent the same morpheme. Subsequently, the normal processes of lexical loss and replacement resulted in the situation where some of these forms as verb phrase heads ceased to exist. From the syntactic point of view, new adjuncts proceed either from an intransitive or from a transitive (ex-)verb. In the first case, the result will be the same as those dozens of adjuncts which have a purely semantic function, and no effect on valency. Although they do not behave as verbs in synchrony, it would be perfectly plausible that former verbs are the source of modern adjuncts like qet '[do s.th.] completely', vatag 'already', ege 'hastily', teqel '[go] down', woy '[split] lengthwise'... 2 5 In other cases, however, an intransitive adjunct may originate in a word class other than verbs, e.g. ewe 'just' < adjective 'good' [ex. (47)]; or 15 'out' < POc *lua, apparently an adjunct since early stages of Oceanic. In the second case, a transitive verb has become what may be named a "transitive adjunct". Contrary to the latter, these adjuncts have the power to alter the argument structure of a verb Vi, either by increasing its valency if

13 8

Alexandre Frangois

Vi is intransitive, or by modifying its secondary orientation if it is already transitive (following the same rules as in the table p. 131). For example, the adjuncts sas '(find)' and day '(await)', although they are not verbs in synchrony, not only change the meaning of the preceding verb, but also transitivize it: (62)

Köyö

(ma-van

SAS)

ni-tilto.

2DU

PFT-go

(find)

ART-egg

'As they were walking, they found an egg.' (63)

Ν ok (tigtig DAY) nek. ISG AO:stand:DUP (await) 2SG Ί was (standing) waiting for you!'

Sometimes, a lexeme is already totally specialized as a transitive adjunct in younger people's speech - but suddenly some archaic or literary phrase uses it as a VP-head, betraying its true origin. This is the case with yak, which is employed as an adjunct '[take, go...] away' in about ninety-nine per cent of its uses, but appears as a verb 'pick up' in a ritual phrase connected with marriage {yak ne-sem '[the bride's father] picks up the money [given by the groom]'). Similarly, the word v(e)teg which we have been glossing 'leave' is very seldom used as a verb; most of the time it forms a transitive adjunct, expressing a physical or abstract separation between the subject and the object [ex. (45)-(50)-(53)-(57)]. This includes the coding of comparison, a metaphorical extension of the idea of separation: (64)

Ke (ne-mnay VETEG) 3SG STA-clever (leave) 'He is cleverer than you.'

nek. 2SG

As is shown in this last example, the so-called "transitivizing" function of some adjuncts does not necessarily affect verbs as such, but also adjectives or other categories; it would be better defined in terms of logical relations (Lemarechal 1998), as a device allowing combining a first unary predicate - h e r e the adjective f(x)= ' c l e v e r ' - to a binary, relational predicate f(x,y)= 'be superior to'. This is how the comparative structure, par excellence relational, happens to be coded by a "transitive adjunct", originally a verb. These transitive ex-verbs look as if they had transformed, one may argue, into prepositions: but what is true for other languages does not fit the structures of Mwotlap, where the strong VP-final boundary preserved V2 from becoming a preposition 26 . One could then think of another possible

Constraints and creativity in the macro-verb strategies of Mwotlap

139

category, i.e. grammaticalizing into an applicative morpheme 27 ; however, contrary to applicatives, we have already seen that these adjuncts are capable of modifying the diathetic orientation of the verb Vi, in such a way that the direct complement of the macro-verb eventually takes over the semantic role of patient. Consequently, the new 'grammatical unit' arising from the diachronic specialization of transitive verbs in the adjunct-position, should no longer be described as a verb any more (contra Crowley 1987: 61), nor as a preposition or an applicative. Rather, the most reasonable analysis would certainly consist in sticking to the "vernacular" category of (intransitive vs transitive) adjunct, to which are associated certain syntactic and semantic properties, different from other word classes. Although this approach seems to partially hamper cross-linguistic comparison, it has the paramount advantage of allowing a deeper understanding of the synchronic structures, as well as the historical evolution, of Mwotlap grammar. Thanks to this methodological caution, it becomes obvious why some nouns, some adjectives, some verbs (etc.) seem to be merging into a single category, obeying the same rules, whenever they follow the head of a VP. This is the only way a vernacular-oriented analysis of the language becomes feasible, as a necessary requirement before any cross-linguistic statement is made.

5.

Conclusion: Serial verbs or Adjuncts?

Although, at first sight, Mwotlap is a good example of what we expect a "serializing language" to be, the present paper showed several reasons why such a tag would mislead linguistic analysis rather than help it. The case where a verbal head V, is indeed followed by one or more verb roots, rather than being interpreted as a serial verb construction, appears to be merely a subcase of a more general and productive pattern in this language, i.e. the syntactic function of adjunct. As far as these adjuncts are concerned, the only grammatical boundary that really matters is not so much the distinction between parts of speech (adjectives, verbs...), but the contrast between unary and binary predicates:

140

A lexandre

Franqois

Table 3. Unary v.v binary predicates in the adjunct position Type of predicate

Syntactic categories

Argument slots

Unary predicate

1 slot:

Binary predicate

2 slots: /(*,y)

/(x)

adjectives + nouns + intransitive verbs + intransitive pure-adjuncts transitive verbs + transitive pure-adjuncts

While unary predicates have little effect upon the argument structure of the macro-verb, binary adjuncts do influence its syntactic behaviour, following complex rules which were detailed here. The fact that some verbs, along with other word classes, are taking part in this mechanism, is indeed worthy of notice; but the quest for data on "serial verbs" as such should not obscure the main pivot of the system, which is not so much the verb as a lexical category, but rather the adjunct as a syntactic function in the clause.

Orthographic conventions and abbreviations. The spelling conventions adopted for Mwotlap include the following: e = [Ο; δ = [υ]; g=[y\,b = [ m b]; d = [ n d]; q = [ V ] ; m = [ W ] ; ή = [q]. The abbreviations used in literal translations include: ANA anaphoric marker PRD predicative PRET AO Aorist Preterite ART Prioritive article PRIOR DU dual PROH Prohibitive DUP reduplicated form PRSP Prospective IMM STA Stative Immediate future NEG negation lEX first exclusive PFT Perfect 1 IN first inclusive POT Potential

Constraints and creativity in the macro-verb strategies of Mwotlap

141

Notes 1.

In this regard, Mwotlap contradicts the tendency proposed by Foley and Olson (1985) that an svo language should make use of Core juncture rather than Nuclear juncture. Other svo languages, like Paamese (Crowley 1987: 82) and Lewo (Early 1993: 88), have also been mentioned in this respect. 2. This description of V2 as being essentially a "modifier" to Vi should be taken for what it is: a first attempt to describe Mwotlap serial verbs in broad, nontechnical terms. The following pages will show that the function of V2 is obviously more complex, and cannot be reduced to a role of lexical specification (see for example the case of causative serialization). The relevant point at this stage of our presentation is to underline the syntactic asymmetry between Vi (the head) and V2 (the adjunct). 3. By preferring the 'adjunct' interpretation rather than the 'serializing' one, we support - at least for Mwotlap - the analysis of Early (1993: 80-81) and even Crowley (1982: 166), contra Crowley (1987: 59). 4. Foley (1986: 117), from Olson (1981). See also the Kalam examples cited by Senft (this volume), after Pawley (1993: 95). 5. A sentence like this one clearly refers to several distinct actions ('kill, cut up, eat'...). Saying this does not contradict the hypothesis, often formulated (Givon 1991a; Durie 1997: 291), that globally only one event is involved. 6. The clear presence of these pauses, as well as coordinators, makes it impossible to talk about Core-layer serialization. The only case which shows Corelayer svc is reserved to purpose clauses in an Irrealis context, e.g. following a general statement, a prospective clause or an order (Aorist): Lep me ne-be nok in. 'Give me some water [so that] I drink' (Francois 2003: 187). Even though the absence of pause between both clauses strongly recalls Core-layer serialization, it must be noted that this structure is functionally very limited. 7. Likewise, Ross (1998: 35) chooses to talk about "adjectival verbs" rather than "adjectives" for Proto Oceanic. Frangois (2001) discusses in detail the issue of Mwotlap parts of speech, including the adjective vs verb contrast. 8. A similar phenomenon occurs in English: compare You hit him to death (with transitive hit), and You danced yourself to death (with intransitive dance). 9. This corresponds to what Crowley (1987: 40, 49) labelled 'ambient serialization' (e.g. /I hit you it was hard/ = Ί hit you hard'); but his description of V2 as a "general" predication, in our opinion, somewhat lacks precisioa 10. The corresponding statements make use of an "impersonal" subject mahe ('place') for this kind of weather sentence: Mahe mö-qön 'It is night' [lit. The place is night]. That mahe is an empty argument, not a true one, is proven

142

11. 12.

13.

14.

15.

16.

17.

18.

19.

20· 21.

22. 23.

Alexandre Frangois

precisely by the impossibility of finding it as an object of a (Vi-V2) bundle: thus [d] * [e], Further evidence against this general prediction will appear in [k]. The verb S.P.R. is a loanword from pidgin bislama S.P.R. [espiar], which stands for Sperem Pablik Rot, lit. 'hit the public road', i.e. 'roam, wander aimlessly all day long'. This has become a full verb in Mwotlap, including the possibility of root reduplication, hence S-S.P.R. [esespiar], Mwotlap has lost the POc causative prefix *pa(ka)-, except for two residual, unanalysable, verbal lexemes wot 'be born' > vawot 'give birth to'; eh 'live' > vaeh 'save s.o.'s life, rescue, cure'. For other "potential" formulas, the risk of semantic ambiguity of a serialized vp would be so high that subordination is always the rule: see the case of {x-Vx-y+y-V2-z}v. 132. The notion of relational seme appears to be particularly fertile in the theory developed by Alain Lemarechal, regarding the semantic and syntactic mechanisms of verb serialization (1998: 206-233). Very similar sentences are reported for neighbouring Mota: saru GORO natarapefha mun ο siopa 'clothe over your body with a garment' (Codrington 1885: 293). The last example, along with many others, is commented in detail in a paper (Frangois 2000 b) dedicated to the numerous valency-effects of this adjunct (not verb) goy [V+goy = 'cover, obstruct, occupy, forbid, react, disturb, dominate, reserve, fetch, protect, assist...' ( En haßß-ir-ay ledcfe ßoggol '(lit.) we shall bind-with the wood a cord' (Labatut n.d.: 126). On Applicatives, see Palmer (1994), Lemarechal (1998: 189).

Serial and compound verbs in Anejom John Lynch

Abstract Like other members of the Southern Vanuatu subgroup, Anejom manifests rather less verb serialization than do many other Oceanic languages. One of the reasons for this is the development of an 'echo-subject' construction, in which the Proto Oceanic conjunction *ma 'and' has developed as a preverbal clitic indicating that the subject of the clause to which it is cliticized is the same as the subject of the previous clause. This paper documents the contexts in which echo-subject marking may operate, and outlines the very productive process of verbal compounding. In doing so, it shows how both of these processes have assumed many of the roles played by serialization in other Oceanic languages.1

1.

Introduction

Anejom, the sole indigenous language of Aneityum (the southernmost inhabited island in Vanuatu), is a member of the Southern Vanuatu subgroup of Oceanic. Crowley (1998: 268-269) makes the following comment on verb serialization in Sye (spoken on Erromango) and other languages of the Southern Vanuatu family: The Southern Vanuatu languages in general appear to differ from most other western Oceanic languages in that there is relatively little evidence of serial verb constructions. Many of the functions expressed by serial verb constructions in other Oceanic languages are expressed in Sye by means of the echo verb construction [i.e. echo-subject prefix + verb] ... [T]he directional orientation of a verb can be expressed in Sye by means of the echo verb construction. In many other Oceanic languages, the directional orientation of a verb is normally expressed by means of a serial verb construction. Anejom is typical of other members of the family in having very little strict verb serialization. Three factors seem to have given rise to this. First, Anejom makes great use of what I have called elsewhere an "echo-subject" construction, somewhat similar to switch-reference structures in languages of other families. Second, compound verbs proliferate in the language, and

146

John Lynch

these also replace serialization in a number of contexts. And third, what may once have been serialized directional verbs have become directional suffixes to verbs (and other morpheme-categories as well).

2.

Clause-chaining and conjunctions

There are a number of ways in which two or more propositions can be conjoined in Anejom. These are: (a) strict paratactic conjoining; (b) the use of conjunctions; (c) the use of conjunctive aspect-markers (am and pat)\ (d) the echo-subject proclitic m-\ and (e) verb serialization. Since (a), (b) and (c) are only tangentially germane to the main topic of this paper, I will deal with them very briefly here. I will cover (d) and (e) in Sections 3 and 4 respectively, discuss the contexts in which different conjoining strategies may or may not be used in Section 5, and will then examine verb-compounding in Section 6. Throughout the paper, clauseboundaries are marked by square brackets.

2.1. Simple clause chaining Simple clause chaining involves strict parataxis - i.e. simply stringing the clauses one after another, with no conjunction or other overt marking of coordination, apart from a brief pause, marked here with a comma. Noninitial clauses in a chain must include subject-tense and/or aspect-mood marking, and they also include overt NP subjects if these are different from the subject of the initial clause. Such conjoining may occur when the subjects of the clauses involved are different, and may occur with both verbal and verbless clauses: (1)

[Ekrau

edou ajamrau}, [ek apahni anak era-i roam we.EX.DL 1 SG.AOR go.everywhere I LOC-CS iji-teptag asga]. COL-nakamal all 'We wandered around and I went to every single nakamal (meetinghouse, men's house).' 1EX.DL.AOR

Serial and compound verbs in Anejom

(2)

[Nida-i

nataman Nagesga], [nida-i

name-CS man

Nagesga

natahen iyiiki

name-CS woman

147

Inmohoc]

that

Inmohoc

'The man's name was Nagesga (Sun), and the woman's name was Inmohoc (Moon).' Clause-chaining is perhaps more frequent when the subjects are identical. In verbless clauses, this simply involves juxtaposition: (3)

[Inyipal Anejom story

ninki], [inyipal

Aneityum this

story

era-i

Nismocop im

REF-CS Nismocop

Nijvan]

and Prawn

'This is an Aneityumese story, a story about Nismocop and Nijvan (Prawn).' 2

With this kind of conjoining in verbal clauses with identical subjects, the VP in clauses other than the first may repeat the subject-tense and aspect-mood marker.3 (4)

[/v

lep

amen a nomrag

again live

3SG.P

ici

old.man

s

aen], [/v

be.like.this he

[/.v

lep

[/Is

lep

3SG.P

echei

is

lep

different

3SG.P

again

amen Anouyac],

3SG.P

live

ici

Anouyac

aen], [/'s

again be.like.this he

3SG.P

is 3SG.P

amen natimi]

again live

3SG.P

omrag

upni\,

old

good

[m-amenjina-i

person

yin].

ES-look.after-TR

him

'There was another old man like this, he lived at Anouyac; he was like this, very old, and he had someone looking after him.' More frequently, however, subject-tense marking occurs only in the first clause, with only aspect markers appearing in subsequent clauses. This occurs very commonly with the perfective marker man, as in (5), but it also occurs (though less frequently) with other aspect markers, as shown in (6). (5)

[Is 3SG.P

[man

man

atgaawun aan], [man imy-auru],

PF

walk.away he

yet-pan

PF

a

PF

COM-hurry

nemnem].

arrive-there LOC h o m e

'He walked off and hurried with it and arrived home.' (6)

[An

le

2SG.INCP

take.SG LOC-its POSS.G

era-η

[pu

mas

aak],

FUT

die

you.SG

[pu

uwu mas

FUT die

intak atimi INDEF

aek],

people

you.SG

aak

a noupan],

you.SG

Τ

time

'(If) you take fruit from someone else's branch, you will die, you will die on the spot.'

148

John Lynch

2.2. Conjunctions Anejom has three clause-level coordinating conjunctions - ka 'or', and two adversative conjunctions jai and jam 'but'. The alternative conjunction ka conjoins two clauses: (7)

[Et

man apan Umej aan] ka [et amen a nemnem] ? PF go Umej (s)he or 3SG.AOR stay LOC home 'Has he/she already gone to Umej or is he/she (still) at home?' 3SG.AOR

Of the two adversative conjunctions, jai (which has an abbreviated variant ja used in fast, casual speech) is used for adversative coordination when the subjects of the conjoined clauses are different. For example: (8)

[Is 3SG.P

man

atgaawun aan], [man yet-pan

PF

walk.away (s)he

jai [eris but

PF

a

nemnem],

reach-there LOC home

par apitac ohowat o-mapo-ri] ...

NSG.P SEQ follow

both

DL-grandchild-3SG

'He walked off and went home, but both of his grandchildren followed him ...' (9)

[Eris

akrou m-alpa-i eama], jai [/s par han] ... share ES-give-TR us.EXC.PL.o but 3SG.PSEQ enough 'They shared it out to us, but there was enough ...' Jai is also used when the subjects of the conjoined clauses are identical but where the verb phrase in the second clause contains a subject-tense marker. This occurs in cases where either a change of tense is involved, as in (10), or where there is overt aspect marking in the second clause, as in (11), introduced by the abbreviated form ja: 3PL.AOR

(10)

[El

ek

COND

par apan anak], jai [ki

lSG.AOR SEQ go

I

but

itiyi

apan anak],

lSG.INCP NEG

go

I

Ί should go, but I won't.' (11)

[Is lep Ρ

atii

n-imehe], [am

again not.be.PL N-sick

and

itiyi NEG

imehe natimi] sick

person

[m-amen] [m-omrag

upni] ja

[par

emesmas],

ES-stay

good

SEQ

die.PL

ES-old

but

'And again there were no illnesses, and people didn't get sick; they lived to a ripe old age but then died.' The other adversative conjunction, yam, is used when the subjects of the conjoined clauses are the same, and where there are no intervening subjecttense or aspect particles (and thus no change of tense-aspect):

Serial and compound verbs in Anejom

(12)

[Is

amen

3SG.P stay

[ala-n

aan] [m-ala-i he

pikad uwu-n],

ES-feed-TR pig

149

jam

P0SS.G-3SG but.SS

era-i iji-sjupura].

feed-3SG LOC-CS COL-evening

'He stayed and fed his pigs, but he fed them every evening.' (13)

[Ens 3PL.P

ago

kava luma-n

m a k e kava

aara]

jam

POSS.D-his they.PL but.SS

[ago is make3SG.P

erou]. two

'They made his kava, but they made two (bowls).' It is possible that jam may actually derive historically from jai plus the echo-subject proclitic m- (see Section 3 below), which also does not allow a following subject-tense or aspect marker. However, while m- is cliticized to the first word of the verb phrase, there is no evidence that m- is encliticized to jai.

2.3. The conjunctive aspect-markers am and par There are two other markers which are involved in conjunctive coordination. The semantics of these are similar to those of conjunctions in other languages, as their translations imply: am 'and', par 'and then, so'. However, their morphosyntactic behaviour is different from the conjunctions just discussed. They behave as if they are aspect markers rather than conjunctions in that they come within the verb phrase following the subject-tense marker, although in both cases the subject-tense marker may be deleted if the subject is the same as the subject of the previous clause. I thus refer to them as CONJUNCTIVE ASPECT-MARKERS. The marker am is simply a coordinator with no overt implication of temporal sequence, and translates best as 'and'. The example below shows the use of am with a preceding subject-tense marker, which appears when the subjects are different. (14)

[Ekris lecse-i 3DL.P

take.PL-TR

u-rau

aaraü], [/s

POSS.G-3DL they.DL

am atpu tah aarau].

3SG.P and hide o n e they.DL

'The two of them took theirs, and one of them hid.' However, there are also examples where the subject-marker is omitted, even though the subjects of the conjoined clauses are different. Usually in these cases, the NP subject of the following clause(s) is overtly marked. For example:

150

John Lynch

(15)

[... Icp again

[lep

adia

a tijiraaki\

leave

s

[m-alau-jhou]

this.TL

[m-amen

ES-go-outside

atec] [im-yipal\ [am asjec a nomrag

ES-stay

ijhou] outside

a-nlii-i

niom\.

again sit ES-tell-story and lie S old.man LOC-middle-CS house ' . . . these three again went outside and stayed out there telling stories, and the old man w a s lying d o w n inside the house.'

When the subjects of the conjoined clauses are identical, the subjecttense markers in non-initial clauses are normally deleted (probably only being retained to avoid any possible confusion or ambiguity): (16)

[Is

ecohos-pan aan ehele-n] [am imy-ecej yiii\. DAT-3SG and COM-say.come him 'He appeared before him and told him to come with him.' 3SG.P appear-there he

It will be seen below that verb serialization may occur when a verb of motion expresses the directional orientation of the preceding verb. I will argue, however, that serialization is not required in this context; and indeed am can occur within a clause which expresses such directional or locational orientation: (17)

[Lecse-i isji-tal ajourau] [am imy-apan take.PL-TR shoots-taro you.DL and COM-go 'Take the taro shoots home ...'

a nemnem] ... LOC home

(18)

[Alau-jhou aak] [am amen ijhou] [m-alek nagesga.] go-outside you.SG and stay outside ES-look.for sun 'Go and stay outside and look for the sun.'

Similar comments apply to the other conjunctive aspect-marker par, a sequential marker which indicates that the event of the clause in which it occurs follows the previous event in temporal sequence. Again, the subjecttense marker generally occurs in a non-initial clause when the subjects of the conjoined clauses are different: (19)

[Is Ρ

wat ici\, T.P

[/s par ein

be.like.this

Ρ

inmap

SEQ eat.TR chestnut

a natimi asga]. S

person

all

'It went on like this, and everyone ate the Tahitian chestnuts.' When the subjects are the same, the subject-tense marker sometimes occurs in a following clause: (20)

[Ek [ek

par

apan anak imran\, go I tomorrow le inpas uwu-n\.

lSG.AOR

SEQ

take.SG axe

lSG.AOR

wut T.F

POSS-3SG

'When I go tomorrow, then I will get his axe.'

Serial and compound verbs in Anejom

151

However, it is more usual for the subject-tense marker to be deleted in a non-initial clause if it is the same as the subject of a preceding clause; thus: (21)

[Er/s NSG.P

lecse-i

u-rau

aarau] [par imy-apaii].

take.PL-TR

POSS.G-their.DL

they.DL

SEQ

COM-go

'The two of them took theirs and went.' While sentences like (20) and (21) occur in my corpus, it is definitely more common to mark all the clauses in a sequence, including the initial clause, with par, as in (22): (22)

[Is

gar add

n-epei ingeje-n a Nismocop], \par

3SG.P SEQ finish N-shave chin-his

[par elad-jai-koü\ \par SEQ look-up-DLST

alek

SEQ search

S Nismocop

alau-jhou],

SEQ go-outside

nagesga]. sun

'Nismocop finished shaving and went outside, and looked out to the east searching for the sun.'

3. The echo-subject proclitic mLike the other languages of the Southern Vanuatu family, Anejom has an echo-subject proclitic m- {im- before a consonant),4 which indicates that the subject of the verb is the same as the subject of the previous verb. Compare: (23)

[Et

apam aan], [ek come s/he ISG.AOR 'S/he came and I slept.' 3SG.AOR

(24)

[Et 3SG.AOR

amjeg anak]. sleep I

apam

aan] [m-amjeg],

come

s/he

ES-sleep

'S/he came and slept.' (25)

[Et

apam aan], [et amjeg aan], come s/he 3SG.AOR sleep s/he 'S/he came and s/he [a different person] slept.' 3SG.AOR

This proclitic attached to the first word in the verb phrase of a noninitial clause which has the same subject as the preceding clause. So while (26) and the final clause in (27) show the most frequent pattern of mcliticized directly to the verb, in the second clause in (27) and in (28) m- is cliticized to a preverbal particle.5

152

John Lynch

(26)

[Ekris apan aarau] [m-ago 3DL.P

go

they.DL

nupui]

ES-make k.o.pudding

[m-ago

ihnii\.

ES-make f i n i s h

'They two went and made nuputand finished making it.' (27)

[Et

amen aan] [im-lep

3SG.AOR

stay

(s)he

tas-pujhou

ES-again talk-outside

ehele-i etwa-ri\ DAT-CS

brother-3sG

[m-ika...] ES-say

'He again talked to his brother outside and said ...' (28)

[/s

itiyi

ege-ktit nitinin is asan aan], 3SG.P NEG hear-well something 3SG.P say (s)he [m-itiyi atou intas-apatinin is asan aan], ES-NEG know word-dark this 3SG.PAST say (s)he 'He didn't hear clearly what he said, and so didn't know this secret word that he said.'

Although the echo-subject proclitic can be followed by a negative or adverbial particle, it cannot occur with aspect-mood markers (like man perfective and par sequential in the examples below). In cases of samesubject coordination when the aspect-mood marker has to be specified in a non-initial clause, then only clause-chaining can be used, as the examples in previous sections have shown. Thus compare (29) with (30), which is ungrammatical with m- despite having the same subject. (29)

[Et

man

apan aen], [man le injap agkou], (s)he PF take.SG salt there 'He went and got salt there.' 3SG.AOR PF

go

(30)

\Et

man

apan aen], [im-fnan le injap agkou], 3SG.AOR PF go (s)he ES-PF take.SG salt there I have called this proclitic the "echo-subject" proclitic, since this is the term used for cognate morphemes in other languages of the Southern Vanuatu family (Lynch 1983). The term was originally used with reference to the Tanna languages, in which the marker m- may cross-reference not only the subject of the previous clause but also, under certain conditions, some other nominal in that clause. The following examples from Lenakel are instructive (note that a is inserted between m- and a following consonant): (31)

[I-em-ho

kiiri] [m-akamw].

1-P-hit

dog

ES-run.away

Ί hit the dog and ran away.'

Serial and compound verbs in Anejom

(32)

[I-em-ho

kuri

mil]

1-P-hit

dog

DL

153

[m-u-akamw]. ES-DL-run.away

Ί hit the two dogs and they ran away.' (33)

[I-em-ho kuri] [rna-kas ramo-k]. 1-P-hit dog ES-bite father-my Ί hit the dog and (then) it bit my father.' In (31) we have the simple situation of the subjects of both clauses being identical, and m- marking this identity. In (32), the subject of the first clause is singular, but the object is dual; the echo-subject marked verb in the second clause is marked for dual number, and therefore m- refers to i.e. "echoes" - an identifiable dual NP in the previous clause: in this case, the object kuri mil 'two dogs'. In (33), both NPs in the first clause are singular; however, the only logical subject of the verb in the second clause is kuri 'dog', not io T , so again we have an echoing situation. In Anejom, however, the prefix m- is a same-subject prefix. The Anejom examples given earlier in this section parallel the Lenakel example (31), where the subjects of both clauses are the same. No Anejom parallel can be given for (32): number is not marked independently of person and tense in Anejom (as it is in Lenakel), and m- cannot be used with a subject-tense marker. Thus the Anejom equivalent of (32) would have to involve some other type of coordination: (34) [Kis awod kuri is erou] [ens par adia aarau]. lSG.P

hit

dog

Ρ

two

NSG.PSEQ go.away

they.DL

Ί hit the t w o d o g s and they ran a w a y . '

Similarly, m- cannot be used in the Anejom parallel of (33) - or if it is used, it can only have a same-subject interpretation, with the meaning Ί bit my father'. The two alternative readings are: (35)

[Kis

awod kuri] [m-aces

lSG.P hit

dog

ES-bite

etma-k]. father-my

Ί hit the dog and (I) bit my father.' (36)

[Kis

awod kuri] [/s par aces

lSG.P hit dog Ρ SEQ bite Ί hit the d o g and it bit m y father.'

etma-k], father-my

154

John Lynch

4. Verb serialization Anejom manifests more frequent verb serialization than does Sye or Lenakel (though not nearly as much as many Oceanic languages). In fact, I am not even sure if there is any strict verb serialization in Anejom. Since this is a VOS language, one would probably expect a serial construction with no change of subject to exhibit V i ( 0 ) V 2 S order; instead, the subject N P always occurs in the first clause before any 'serialized' verb - thus W\ (O) S V2.

What I refer to here as verb serialization corresponds roughly to what others have referred to as nuclear serialization (see e.g. Foley and Olson 1985, Crowley 1987).6 That is, I define a clause as occurring in a serial construction if: (a) (b) (c)

there is no overt subject nominal in non-initial verb phrases; there is no subject-tense marker; and there is no aspect-mood marker.

A serialized verb may, however, take verbal prefixes, almost exclusively the comitative prefix imy(i)-, and may occur with one or more adverbial phrases. Most of the examples of serial verbs in my corpus (with or without the comitative prefix imy(i)-) supply the directional orientation of the preceding verb; for example: (37)

... [am adia aarau] \apart] [m-edou] and go.out they.DL go ES-walk. around [m-arehed Anejom\. ES-go.right.around Aneityum '... and the two of them went off and walked right around Aneityum.'

(38)

... [am and

lecse-i

u-rau

aarau]

take.PL-TR POSS.G-their.DL

\imy-apan

ika va-n Inmanjapitac]. call GOAL-TR Inmanjapitac '... and they took theirs to the place called Inmanjapitac.' COM-go

(39)

[Is

man

a-nworen

iyiiki

they.DL

LOC-place that

nitai eris ika va-n thing 3PL.P call GOAL-TR Nelcau-Eri-Nisyeg] Γ imy-alau-jhou] [m-ati-i-se ijhou}. Nelcau-Eri-Nisyeg COM-go-outside ES-put-TR-down outside 'He took something called Nelcau-Eri-Nisyeg outside and put it down outside.' 7 3SG.P PF

le

aan

ens

3PL.P

take.SG (s)he

tah

INDEF

Serial and compound verbs in Anejom (40)

[/v man P

lep

PF

rectidai

again get.up

aattaj\ \apan a-nlii-i

niom\.

they.TL

house

go

LOC-inside-CS

155

'They three got up again and went inside the house.' There are relatively few examples of serial constructions in my corpus where the verb in the non-initial clause is not a verb of directional motion. One such example is: (41)

...[/s

am lecse-i

3SG.P and take.PL-TR

[imy-apan a-nworen COM-go

LOC-place

uwu-n

aan] \imv-athut]

POSS.G-his

he

COM-run

\awapawap\ tiptoe

iyiiki\. that

'... and he took his and ran on tiptoe and brought it to that place.' The last clause in (41) includes a serialized directional verb. The second and third clauses, however, are also serialized verbs but are non-directional (although they are verbs of motion). Other examples: (42)

... [par imy-apan] [a/an\ SEQ COM-go

go.along

[etha-n-se

a-nwai\ ...

put-TR-down

LOC-water

'... and went and put it in the water...' (43)

[Is 3SG.P

amen aeri\ [omrag

upni].

live

good

he

old

'He was very old.' (More literally, 'He lived and was very old.') (44)

... [m-amen ES-stay

aen] jam he

but.SS

[imy-amen] COM-stay

\alanaheni\. pray

'... and he prayed repeatedly.' (More literally,'... and he stayed and stayed and prayed.') I will discuss the factors influencing the choice of serialization as a conjoining strategy in the next section.

5.

Choice of conjoining strategies

Anejom speakers have a number of different conjoining strategies available to them. Here I attempt to provide some discussion relating to which strategies are used in which syntactic-semantic contexts. 8 In particular, I will try to outline why serialization is not used in cases where it might be expected to be. Only two options seem to be available when the subjects of the conjoined clauses are different: these are clause chaining and the use of the conjunctive aspect-markers. (Neither echo-subject marking nor verb serialization is an available strategy here, since both involve the same sub-

156

John Lynch

jects.) Hence this discussion focuses on conjoined clauses with the same subject. When the subjects of the conjoined clauses are the same, then all four conjoining strategies are possible. This is illustrated in numerous examples above, as well as the following; (47) illustrates both the echo-subject proclitic and verb serialization. Clause-chaining: (45)

[Eris man lep er-atga aara], [man lep apan a nemnem], 3PL.P PF again MUT-walk they.PL PF again go LOC home 'Then again they walked off and went home.'

Conjunctive (46)

aspect-markers:

[/v par ein inmap a natimi asga] 3SG SEQ eat.TR Tahitian.chestnut s person all [am ein inma a natimi asga.] and eat.TR breadfruit s person all 'All the people ate Tahitian chestnut and all the people ate breadfruit.'

Echo-subject and Verb serialization: (47)

... [am adia and go.out

aarau] [apan] [m-edou] they.DL

go

ES-roam

[m-arehed Anejom\. ES-go.right.round Aneityum '.. .and the two of them left and went off and wandered all around Aneityum.' It is difficult to characterize the contexts in which each strategy occurs, and it seems to me that there is a reasonable amount of freedom here. When sequence needs to be expressed overtly then the conjunctive aspect-marker par is used. Similarly, when the second clause requires a tense or aspect marker, then the only option is clause-chaining. I did mention above that serialization generally occurs with verbs of motion to express directional orientation (as with apan in (47), for example), and serialization is rarely used in other contexts. However, the reverse is not true: directional orientation can be expressed by other means. Contrast the following two sentences, both taken from different parts of the

Serial and compound verbs in Anejom

157

same story, and both virtually identical in meaning; the first uses serialization, the second the conjunctive aspect-marker am: (48)

[Lecse-i take.PL-TR

uni-mirau

ajourau] \imy-asuoI\ ...

POSS.G-your.DL

you.DL

COM-go.down

'Take yours and go down with i t . . . ' (49)

[Lecse-i take.PL-TR

isji-tal

ajourau] [am

shoots-taro you.DL

and

imy-apan a COM-go

nemnem] ...

LOC home

'Take the taro shoots home Note also the following sentences from the same story. In both cases, the people referred to have heard the old man calling out and they come in to ask him what he was saying. The sentences are virtually identical, except that different conjoining strategies are used - echo-subject in (50), serialization in (51). (The relevant clauses are underlined.) (50)

[Rectidai aattaj\ [m-athut] [m-apan a-nlii-i niom\ ... get.up they.TL ES-run ES-go Loc-inside-cs house 'They got up and ran inside the house ...'

(51)

[/s man rectidai aattaj\ [apan a-nlii-i

nioin\ ...

P PF get.up they.TL go LOC-inside-CS ' T h e y g o t up and w e n t inside the h o u s e . . . '

house

Apart from the few restrictions noted above, therefore, it appears as if speakers have a relatively free choice as to which conjoining strategy to use when the subjects of the conjoined clause are the same. In terms of frequency of occurrence in text, note the following: (a) (b) (c) (d)

echo-subject marking is by far the most frequent strategy used; the next most frequent strategy is to use a conjunctive aspect-marker; clause-chaining is relatively infrequent; and serialization is the least frequent strategy of all.

There appear to be no semantic constraints on (a) or (b), and there are probably also none on (c). Serialization, however, is almost totally restricted to cases where the second verb expresses directional orientation, and even then other options are possible. In Section 7 below I will examine why this is the case. First, however, I turn to verbal compounding in Anejom.

158

John Lynch

6. Verbal compounding Compounding is, and was, quite productive in Anejom: many compounds are quite transparent, but others are fossilized, which suggest productive compounding at some earlier stage in the history of the language. For example, a number of fairly recent introductions to Aneityum culture and society are expressed as compounds, indicating the productive nature of the process: (52)

nepjed-eromaga 'mandarin orange' citrus-Erromango nadiat-atuihap 'Sunday' day-rest inta-alainod-hat 'oxy torch, blowtorch' thing-break.by.burning-stone On the other hand, there are many words which look like compounds, but where one or both parts do not occur independently. The verbs in (53) seem to contain a second element (c)lyck with the idea of 'badly, wrongly, unsuccessfully'; but neither (e)Iyek (with this meaning) nor the first part of any of the forms in (53) occurs independently in modern Anejom: (53)

acjalyek 'spear inaccurately or randomly' ehvelyek 'drop accidentally, lose through death' torelyek ' affect strongly, damage' woyelyek 'stumble, trip' Similarly, the apparent compounds in (54) all involve the idea of 'breaking', all have an identifiable verb as the first element and all involve a second element of the form amod.; however, there is no free-form verb amodin the modern languages. (54)

acsamod ahvamod asvamod atamod atcamod

'cut ' 'break by squeezing' 'wring, twist off 'cut with knife raised' 'break by hitting'

cf. cf. cf. cf.

acsei ahvii asvii atai

'cut with sawing motion' 'press w. finger' 'bend' 'slice, cut without raising knife' cf. atcai 'hit'

In contrasting serialization and compounding in languages generally, it is often difficult to decide whether any case of Verb ι followed by Verb2 is a compound or a serial construction. In Anejom (as the discussion in 6.2 will show), however, it is usually quite simple: compounds very frequently exhibit distinct morphophonemics - in particular, the loss of a vowel across

Serial and compound verbs in Anejom

159

the morpheme boundary - which do not occur in other environments. In the discussion which follows, then, I do not believe that there is any reason to treat what I call compound verbs as serialized verbs.

6.1. The semantics of compound verbs Compound verbs normally consist of two (occasionally three) identifiable morphemes: the first is always a verb, the second is most frequently also a verb (though there are some compounds of the shape verb-noun, which we will ignore here). In many cases, phonological reduction has taken place, and I discuss that topic specifically in 6.2. In these compounds, the second verb usually modifies, limits, or further describes the action expressed by the first: ama-i

+

chew-TR

alde-i

>

asvi-i

+

bend-TR

alde-i

>

athe-i

+

ipni-i

asgi-i

>

direct. water-TR

+

smell-TR

upni

isvi-i

+

athasgi-i 'dam'

>

good

count-TR

asvalde-i 'break in two'

CUt-TR

cover.w.earth-TR

amalde-i 'bite one's tongue'

CUt-TR

ipnupni 'fragrant'

adumoj

>

return

isvidumoj 'count backwards'

I have not done a statistical count as to the nature of the second verb in verbal compounds. My guess would be that fewer than ten percent perhaps even fewer than five percent - have as their second verb one expressing the directional orientation of the first. There are some examples, like the following: (56)

isvi-i count-TR

ade go.down

ayid defecate

+ adumoj return

+ apok go.seawards

+ arehed go.around

> isvidumoj 'count backwards'

> adapol·? 'go down to the sea'

> ayidarehed 'defecate all over the place, have no proper toilets'

However, these are very few compared to those in which the second verb is non-directional.

160

John Lynch

6.2. Morphophonemics An investigation of the morphophonemics of verbal compounds helps to shed light on the processes involved, and to explain why compounding has 'taken over' many of the roles of serialization. Before proceeding with this discussion, however, two other points need to be made briefly here. The first is that the majority of Anejom verbs are vowel-initial - 87% of verbs begin with a vowel, the commonest being a (52% of all verbs) and e (20%) - due to the historical accretion of a vowel onto verb roots for reasons which need not concern us here. The second point is that transitivity is marked by a suffix -i{-n for inanimate objects with one subclass of verbs), but many verbs are morphologically unmarked for transitivity - e.g. awod 'hit', etjem 'dive for', ahese 'rain on', ayik'lie to', akred1,scrape', etc. When the first verb ends in a consonant (which is quite rare), then there is usually no morphophonemic change. However, the two verbs form a single phonological unit (in terms of stress placement), which is why I treat such forms as compounds rather than serial constructions - e.g. tan 'cry' + tas 'talk' > 'tantas 'alternately cry and talk (as when mourning)', with stress on the penult of the compound. When the first verb is vowel-final (which is true of the vast majority of cases), the first thing to note is that the transitive suffix does not occur on the first verb in the compound, whether or not that verb - and/or the whole compound - is transitive: the examples in (55) and (56) illustrate this. If the second verb is transitive, then it will take the suffix if it belongs to that class of verbs which overtly mark transivity: see the first three examples in (55) above, and also: (57)

aihi-i

+ alde-i

begin.to.plait-TR

ala-i

cut-TR

+ agre-i

blow-TR

break-TR

> aihalde-i 'trim '

> alagre-i belch

If, however, the second element is a transitive verb which does not take a transitive suffix, then no transitive suffix appears anywhere in the compound, even it if is actually transitive: (58)

almo-i

+ ajidjid

see-TR

inspect

ecse-i chip-TR

+ akred scrape

> almajidjid 'inspect'

> ecsekred 'sharpen by paring or shaving'

Serial and compound verbs in Anejom

161

Following the deletion of any transitive suffix, if two vowels come together across a morpheme boundary, one is deleted. In most cases, it is the first vowel which is deleted: (59)

asvi-i bend-TR

+ acihi-i pull-TR

>

asvacihi-i 'straighten, stretch'

acse-i cut-TR

+ alge-i turn-TR

>

acsalge-i 'cut w. sawing motion'

athe-i + asgi-i > cover w. earth-TR direct.water-TR

athasgi-i 'dam'

However, there are some cases where the second vowel is deleted: (60)

isvi-i + adumoj count-TR return

>

isvidumoj 'count backwards'

ecse-i + akred > ecsekred chip-TR scratch 'sharpen by paring' and also a few where the combination of vowels results in a third vowel which takes on some features of each of the adjacent vowels (e.g. i+ a> e in the example below): (61)

ecji-i + apela-n ram-TR turn-TR

> ecjepela-n 'stuff with material'

I cannot at this stage account for which vowel would be deleted in any particular context. I have a feeling that stress may be (or may once have been) a factor, but I will not pursue this here.

7.

Discussion

The classic serialization environment in a VOS language is where two clauses have the same subject, and where the tense-aspect, polarity, and other participants remain constant. That is, (62) f r « , , V, (O,) SJ

[ΤΑΜ, V. (Ο,) Sqi] >

v|

v! S ü }

I mentioned earlier that, in those relatively rare cases when serialization does occur in Anejom, it is almost always when the second verb expresses the directional orientation of the first. An examination of compound verbs in Anejom, of which those given as examples in the previous section are typical, shows virtually no cases where the second verb in a compound is a

162

John Lynch

directional verb. Rather, the prototypical compound verb is one which expresses a single complex non-directional action as a combination of its parts: thus chew + cut = 'bite one's tongue' cover with earth + direct water = 'dam (a river)' chip + scrape = 'sharpen by paring or shaving' That is, we find the following patterns, where the category on the left refers to the orientation of the second verb: (63)

directional: non-directional:

m-Y (echo-subject) m-W (echo-subject)

or V + V (serialization) or V-V (compounding)

The questions that need answering in the Anejom context, then, are (i) why has non-directional serialization virtually disappeared from the language, and (ii) why is even directional serialization so relatively infrequent? Why, in both cases, does echo-subject marking play such a dominant role? I will look at this last question first.

7.1. The rise of echo-subject marking In all of the languages of the Southern Vanuatu family, the use of the echosubject prefix is extremely frequent. This prefix almost certainly developed from an historical conjunction, deriving from POc *ma 'and'. In most of these languages, the echo-subject prefix is phonologically identical or very similar to the conjunction which coordinates noun phrases (this is im in Anejom), but phonologically different from the conjunction which coordinates clauses with different subjects (e.g. am in Anejom).10 It thus appears to have derived from a nominal coordinator. Similar statements can be made about at least some New Caledonian languages. In Xäräcüü, for example, where clauses with different subjects are conjoined by nä, Deux verbes peuvent etre coordonnes par le relateur coordonnant de syntagmes me 'et'. Comme les verbes series [my emphasis], les verbes coordonnes partagent un meme sujet, et les modalites aspecto-temporelles s'anteposent au premier verbe, ou se postposent au second, selon leurs positions respectives (Moyse-Faurie 1995: 124)." Moyse-Faurie clearly likens this kind of coordination to that expressed by serial verbs. There may well have been something about the semantics of *ma, at least in these languages, which implied some kind of 'tight' rather than

Serial and compound verbs in Anejom

163

'loose' coordination. That is, just as Ν ma Ν formed a single unit, a noun phrase [ NP Ν ma N]NP, SO V ma V may have been seen in the same way in these languages as a single tight unit [ v V ma V] v . This would have been reinforced by the fact that *ma has replaced the preverbal subject marker in not only the Southern Vanuatu languages but also in at least some New Caledonian languages, like Xäräcüü: (64)

ru

cha

me

3DL clear.bush

mara

and.SS work.in.fields

'They cleared the bush and worked in the fields.' When directional serialization does occur in Anejom, it probably does so most frequently with the comitative prefix imy(i)-. This prefix can be used in non-serial constructions, where it implies an action carried out together by a group: (65)

Is man

imyi-tas-tas

aarau,

P

COM-talk-talk

they.DL ES-say ...

PF

m-ika...

'They talked together (or to each other), saying ...' It is most frequently, however, used with directional verbs in a serial construction, as illustrated in (38), (39), (41) and (48) above, and also: (66)

Lc

inpas una-k

take.SG axe

aak

POSS.G-my you.SG

imy-apam. COM-come

'Bring my axe.' (i.e. 'Take my axe and come with it.') This prefix could possibly be a compound of the echo-subject prefix m(which occurs as im- before a consonant) and the 3SG inceptive preverbal particle yi. That is, imyi-apam in (66) might be, or might have been, im-yiapam 'and it will come'. (And if this is so, then we have a true echo construction, parallel to the Lenakel example (33) in which the object of the first clause is the subject of the second.) Thus what appear to be serial verbs with the prefix imyi- might in fact have once been echo-subject constructions.

7.2. Compounding and directional suffixes In addition to the 'takeover' by the echo-subject prefix m-, we also need to note the development of directional suffixes to verbs. These were probably once directional verbs themselves, and thus a kind of compounding may have originally taken place. However, these forms are now true suffixes (although there are verbs which are phonologically related to some of

164

John Lynch

these). The following are some of the directional suffixes which may occur on verbs: (67)

Vertical

Horizontal

Distance

-jai

-pok

'seawards'

-ki

'near'

'landwards, inland'

-kou

'distant'

'up, south, east'

-se(h) 'down, north, west' -pahai -pam

'hither, towards speaker/focus'

-pan

'thither, away from speaker/focus'

A verb may take one, two or three suffixes, one from each set. If it takes one, it can only take a vertical or horizontal suffix. These may be attached to non-directional verbs (following the transitive suffix if there is one, as in the second example below); they may also be suffixed to the empty base au- which forms locative demonstratives. Thus: (68)

elad-se-pam-kou ati-i-pan-kou au-se-pahai-ki

'look down towards me in the distance' 'put it down over there' 'here down inland nearby'

The horizontal suffixes are matched phonologically by four directional verbs apok ' c o m e / g o seawards', apahai 'come/go landwards/inland', apam 'come' and apan 'go'. It is not improbable that an original serial construction V + VDIR b e c a m e V-DIR.

There is a further point at issue here, however. POc *p develops regularly as Anejom h, while Anejom ρ derives from POc *b, a prenasalized voiced stop. 12 The regular development of POc *pano 'go' is the verb han 'arrive there' (which has a semantic pair ham 'arrive here, come', possibly a compound of *pa(no) + *mai 'hither'). The verbs more commonly used with the meanings 'go' and 'come', however, are apan and apam. These suggest an earlier prenasalized stop: i.e. if han < *pano, then apan < *a-bano or *a-m-pano. So it is possible that even these verbs, and the directional suffixes which derive from them, have an historical echo-subject prefix. If what I have suggested is correct, then the development of directional suffixes probably began as just one form of compounding (with or without m-). However, while compounded directional verbs became grammaticalized as directional suffixes (which could precede and follow other suffixes and could be added to other bases), compounded non-directional verbs formed a far more heterogeneous set, did not participate in grammaticalization in the same way because they were less susceptible to it, and

Serial and compound verbs in Anejom

165

remained simply compounds. These underwent phonological reduction: most such compounds involve a vowel-final verb followed by a vowelinitial verb, and vowel loss or fusion converted these vowel sequences into a single vowel - and thus converted the two verbs into a single phonological unit.

7.3. Conclusions What can we conclude from this discussion? It may be that serialization disappeared from Anejom because of a combination of three factors: (i) the development of a special use of the conjunction *ma to mark a close relationship between clauses (possibly at first only with directional verbs?); (ii) phonological reduction, which converted two serialized (nondirectional?) verbs into one compound verb; and, possibly later, (iii) the development of directional suffixes from m-marked directional verbs. Once these processes had become entrenched in the grammar, and as serialization became less and less frequently used, both echo-subject marking and compounding took on broader roles.

Orthographic conventions and abbreviations in and ρ represent labiovelars ; d = [Θ], c = [γ], g = [rj], j = [tj ~ d3]. The following abbreviations are used in interlinear glosses: AOR COL COM COND CS DAT DIR DIST DL ES EX INCP MUT Ν

aorist collective comitative conditional construct suffix dative directional distant dual echo subject exclusive inceptive mutual/multiple action nominalizer

NSG Ρ PF PL POSS.D POSS.G

s SEQ

ss τ ΤΑΜ T.F T.P TR

non-singular past perfect plural drink possessive general possessive subject sequential aspect same subject temporal marker tense-aspect-mood marker temporal conjunction, future temporal conjunction, past/present transitive

166

John Lynch

Notes 1.

This paper has been adapted from various sections of my grammar of Anejom (Lynch 2000). An earlier version was presented as a seminar at LACITO, and I am grateful for comments made by various researchers based there, as well as comments from Teny Crowley. 2. A nismocop is a kind of small freshwater fish. 3. The first clause in (4) includes two relative clauses, while the last clause has echo-subject marking. 4. See for example Crowley (1998: 246-262) for a discussion of echo-subject marking in Sye and Lynch (1978: 45-47, 1983) for Lenakel. Claire Moyse-Faurie has suggested to me that the echo-subject marker in Anejom might be better analysed as a free particle rather than as a clitic. While the grounds for analysing cognate markers in the languages of Erromango and Tanna as clitics are compelling (e.g. m- occurs in the same grammatical slot as person prefixes to verbs), these arguments do not apply so well in Anejom, where m- replaces person-tense particles. However, the phonological facts accord better with the treatment of m- as a clitic than as a free particle. 5. Note that each of the clauses in (28) contains a relative clause. 6- See also vol. 32/1 (1993) of Oceanic Linguistics, which includes a number of articles on serial verbs in Oceanic languages. 7. Nelcau-Eri-Nisyeg is the name of an individual canoe: 'the canoe (nelcau) of the leaves (eri) of the barrel tree (Acacia simplex) (nisyeg)\ 8- I exclude from this discussion the alternative conjunction ka and the adversative conjunctions ja(i) and jam, since these are both semantically and syntactically distinct from all other conjoining strategies. 9. There is a directional suffix -^oArwith a similar meaning as the verb apok (see §7.2). However, if this were the suffix, the verb would be ade-pok. 10. Something similar can be said for other languages in the family. In Lenakel, for example, ma(ne) coordinates NPS, kani coordinates clauses whose subjects are different, and m- is the echo-subject prefix. In Sye, however, m- is the echo-subject prefix, while im coordinates both NPS and different-subject clauses. 11. Translation: "Two verbs may be coordinated by me 'and'. As for serial verbs [my emphasis], the coordinated verbs share the same subject, and the ΤΑΜ markers stand before the first verb or after the second one, according to their own position." (Moyse-Faurie 1995: 124). 12- Examples: *patu 'stone' > in/hat, *tapine 'female' > na/tahen; *baga 'banyan' > in/pak, *boni 'smell' > i/pni-i.

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

Isabelle Bril

Abstract This paper analyses the various types of serial and complex verbs and the criteria that help restrict this phenomenon in Nelemwa. Complex verbs belong to three main types involving various types of verb and different hierarchies: (i) time-iconic, co-ranking active verbs; (ii) asymmetric verbs expressing adverbial modification or aspectual and modal specification; (iii) semi-grammaticalized verb strings with argument-expanding or conjunctive functions. This paper also investigates the role of complex verbs in a typology of clause-linkage and assesses the syntactic, semantic and discourse differences between one-clause and multi-clause strategies: i.e. between complex verbs and syndetic or asyndetic coordinate or subordinate VPs and clauses. It also assesses the semantic notions which are conceptualized as complex predicates rather than as interdependent clauses.

1.

Introduction

Nelemwa is one of the twenty-eight Kanak languages of New Caledonia, it is spoken by approximately a thousand speakers in the Far North of the Mainland. These languages belong to the Remote Oceanic subgroup of the Austronesian family. Here is a brief overview of the syntactic organization and typological features of Nelemwa. Nelemwa has a split constituent order and split accusative-ergative case-marking system varying with argument category. Personal pronouns, which only refer to human entities (inanimates are zero-marked) have an accusative system, while nominal arguments are case-marked as ergative or absolutive; the ergative case-markers are (e)a (for animate agents) and ru (for inanimate agents) and the absolutive case is zero. The neutral order is sVo with subject and object pronouns referring to human entities (otherwise 0V0), whereas nominal arguments come after the predicate,

168

Isabelle Bril

either as VS (where [S] is the absolutive nominal argument of an intransitive verb) or VOA (where Ο stands for the second argument/patient and A for the agent of a transitive verb). The subject index cross-references the postverbal nominal prime argument (if human), irrespective of its absolutive or ergative case-marking and agrees in number with it, thus (s)VS, or (s)VOA; whereas the object pronoun does not occur with the nominal second argument.

2.

"Nuclear layer" serial constructions [(s)W(o)] in Nelemwa

Serial constructions in Oceanic languages are generally described as belonging to three main types - nuclear, core and adjunct serialization 1 according to the level and type of nexus (see, Bril introduction, this vol.). Some languages display several types varying with verbal category 2 or syntactic function. New Caledonian languages differ in their constituent order, SVO, VOS, and sometimes mixed SVO and VOS order according to tense-aspect; however they all display the nuclear type of serial construction, i.e. with contiguous verbs [sVVo]. They are thus counterexamples to Foley and Olson's (1985) generalization that "nuclear layer" serialization is mostly found in SOV languages, while SVO languages have predominant core layer serialization. In Nelemwa, serial verb constructions have one single primeargument and one single patient [sW(V)(o)]; [sV sV] patterns would constitute two independent clauses with a pause inbetween, not a serial construction. Thus, a verb string does not necessarily constitute a serial verb construction, and a fmer-grained analysis is required. Different terms will be used to refer to various types of verb strings: serial verb construction will only be used for co-ranking constructions that follow the criteria and rules listed in 2.1 and 2.4 below; a complex predicate will refer to an asymmetric verb string comprising a verbal head and a verbal adjunct (see 2.7 and 2.8) or to an Aktionsart or modal verb (2.10 and 2.11) or to verbal compounds whose meaning is somewhat different from the meaning of each verb (2.9). Any other sequence of verbs, especially those involving functional specialization and grammaticalization or which belong to different underlying clauses will be referred to as verb strings (see 4). This will also involve assessing the type of syntactic or semantic relation between verbs in serial constructions, and more broadly, an analysis of dependency in a typological perspective of clause-linkage (see 5 and 6).

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

169

2.1. Criteria of serial verb constructions in Nelemwa Serial verbs are generally defined by a set of criteria which also apply to Nelemwa: they are verbs or VPs constituting one single predication/clause, one single prosodic entity, sharing syntactic arguments (with a single set of pronominal affixes or nominal arguments), sharing tense, aspect, mood, illocutionary force, polarity and referring to aspects of a single event. Specific to Nelemwa is the requirement that the verbs be contiguous, only directionals may intervene between V] and V 2 . Each verb should be a full lexical predicate with no loss of semantic, morphosyntactic properties or phonological form. Anything that does not meet these criteria will not be considered as a serial or complex predicate, but as a mere verb string. Thus, in (1), only ä and kuut are serialized, while the adverb mwadu between kuut and axi is evidence that axi belongs to another clause with an ellipted subject index and does not constitute a serial construction: (1)

I

u

ä

kuut mwadu

axi axamalileny

3SG PFT leave stand down there see these2.DELCT

ebai

nu

then

coconut

malileny. these2.DELCT

'He leaves, stands down there, sees these two coconuts.'

2.2. The criterion of lexical autonomy: the case of transitivized adjuncts The requirement of lexical autonomy excludes a number of deverbal adjuncts from serial verb constructions. They cannot be used as autonomous verbs, even though they have preserved some morphosyntactic features from their former verbal category such as their transitive concord 3 with a transitive V, in complex predicates (cf. (2)), in contrast with true adverbial adjuncts (hulima 'hastily, disorderly', maya 'slowly, correctly') which never display any transitive concord. (2)

Hla paare pexagiix-e paara. 3PL tell separate-TR story 'They tell a story in turns.'

Constructions such constructions. Among hayu(li) 'haphazardly', constantly'; hada, hade

as (2) constitute complex predicates, not serial those deverbal adjuncts are: neo(li) 'quietly', pwamapwan(ili) 'not directly'; gela(li) 'always, 'only';pexagiik 4 'each'.

170

Isabelle Bril

2.3. Common patterns and types of verbs prone to serial construction Serial constructions are frequent in Nelemwa, but they are restricted to some types of verbs and functions. The most common are active and motion verbs as well as stative verbs. Table 1. Type of verbs. FREQUENCY TYPE OF VERBS

frequent


infrequent —»

sequential or simultaneous actions

SEMANTICS

. , stative verbs modification, manner, qualification

Aktionsart, , , , modal verbs

—>

inchoative, terminative (un)able, know, try

manner purpose, endpoint result (infrequent)

The most common pattern involves two verbs; sequences of three verbs are infrequent and generally comprise a verb of motion or direction. - V, may host any active or stative predicate and is the open slot, - V2 or V 3 as the modifier slot may also host any active or stative predicate, but it is more restricted. Table 2. Type of verb slots.

v, stative V

active V (+/- transitive)

v2 ?? stative V intransitive V [* ungramm. transitive v ] stative V (+/- transitivized) intransitive or transitive V

The only absolute restriction bars the specification of a stative Vi by a transitive V2; no example of serial verbs consisting only of stative verbs could be found, all other combinations are possible. Serial constructions in Nelemwa display most of the common rolemarking functions of such constructions (motion, trajectory and direction, manner of action, purpose, modality, Aktionsart), but they exclude causa-

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

171

tive or resultative functions, as well as benefactive or comitative case-role marking. - V2 may also have argument-expanding function, adding case-roles (such as locative, associative, etc.) which are not part of the verb's semantic structure or which do not have any other morphosyntactic expression; - V2 may also mark clausal dependency, such as complementation (with the quotative verb khabwe 'say') or subordination (with uya 'arrive' grammaticalized as a conjunction meaning 'until'). This marks the frontier between serial constructions and grammaticalized conjunctive verbs with possible semantic bleaching.

2.4. Syntactic representation and syntactic functions of SVCs in Nelemwa In Nelemwa, a verb sequence may have various underlying syntactic structures and functions, correlated with various types of predicates. Here are the most common: (i) sequences of active, co-ranking verbs referring to sequential actions and constituting a multi-headed construction, with verbs sharing the same subject / agent (2.6); (ii) asymmetric secondary predication in which the verbs share the same subject, but in which an active V2 specifies the circumstances of Vi (2.7); (iii) hierarchized, asymmetric head-adjunct constructions, in which the head V) is modified by a fully verbal V2 expressing qualification; this follows the general pattern of modification in Nelemwa [modified modifier]. V[ and V2 share the same patient, but not the same agent. Transitivity is determined by the head Vi and V2 shows transitive concord with a transitive V] (2.8); (iv) complex predicates involving inchoative Aktionsart (2.10) or modality (2.11), with verbs such as shelä 'know, be able', mwemweli 'know', shaya 'move, work, try', kuat 'refuse', thaaxa 'begin'. This other type of asymmetric structure reverses the usual pattern of modification, since Vi has scope on V2. Yet, none of these verbs may be analysed as auxiliaries, and a different analysis will be proposed. Table 3. Patterns of specification. V,v2

modality, inchoative Aktionsart

172

Isabelle Bril

These four types involve different types of verb, in different slots and positions; they have different underlying structures, different patterns of argument sharing, different morphosyntactic characteristics and different properties.

2.5. Argument sharing and negation Argument sharing is the prime distinctive criterion for complex verb constructions. In Nelemwa, the nuclear [sVV(V)o] pattern has a unique set of arguments (only directionals may intervene between the verbs) and the negation marker kio has scope over the whole complex (see ex. (3)). Of the four types mentioned above, only the third one, which comprises a modifying stative verb [such as V3 yoo-ll in (3)], displays only patientsharing, as shown by the topicalization of the patient vhaa nelemwa in (3b), but no subject-sharing. (3) a.

Kio i

tälä

mwemweli yoo-ll vhaa Nelemwa. be good-TR talk Nelemwa 'He doesn't understand the Nelemwa language very well.'

NEG 3SG hear

b.

know.TR

Vhaa Nelemwa xe kio ζ tälä mwemweli yoo-ll talk Nelemwa TOP NEG 3SG hear know.TR be good-TR 'The Nelemwa language, he doesn't understand very well.'

The other types share the subject/agent, but not necessarily the patient. In (4), only V3 'look for' has a patient, V\ 'go down' and V 2 'fish' being intransitive. (4)

Na tu pwe yaagi caaxi caa-hi. lSG go down fish look for side-dish starch food-POSS.lDU.lNCL Ί went down (to) fish (and) look for food as a side-dish to our starch-food.'

2.6. Equally ranked constructions with active verbs This is commonly found in narratives. A chain of active verbs, all sharing the same subject, expresses a tight sequence of actions in logical and chronological order and makes up one single event. In this multi-headed construction, with co-ranking verbs, there is an order constraint: a transitive verb (if any) comes last and the verbs do not necessarily share the

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

173

patient. Their order is both pragmatically iconic and consistent with the VO word order of Nelemwa and the s(V)VO order of the nuclear serial construction: the intransitive verbs come first and the transitive verbs come last, before the patient. Semantically, they refer to sequential actions, varying with the semantics of the verbs involved in the string. (5)

Mo flie-dame pae-da bwa on na hmwiny. 2DU take-up here throw-up on sand LOC here 'Bring it up here (and) throw it up onto the shore here.'

(6)

Ma

u

oda uya tu axaleny Boovalan. go up arrive find this man Boovalan 'We went up, arrived, found this Boovalan.' LDU.EXCL PFT

(7)

Na hla xam oome tuuli hi shaam hleny xe xam but 3PL ASS come here find this plate this.DELCT TOP ASS maat-maat. empty 'And they arrive (and) find the plate which is empty.'

2.7. Complex predicates with specifying active V 2s As in 2.6, the verbs share the same subject, but they do not necessarily share the same patient, and if there is a transitive verb in the string, it comes last. But in contrast with 2.6, the V 2 specifies the circumstances of the main action expressed by Vi (concomitance, direction, manner of action, result, etc.). Yet, the construction analysed in 2.7 is different from the asymmetric head-modifier complex predicates analysed in 2.8, which evidence fusion of argument structure.

2.7.1.

Specification of circumstances (concomitance, manner of action, result, purpose, etc.)

Such complex predicate constructions refer to one single event, comprising a second predicate which expresses concomitance or more generally the circumstances of the main action (manner of action, purpose, result). Their intermediate status between co-ranking and asymmetric structure appears in their English translation, either as coordinate or as participial verbs, but

174

Isabelle Bril

apart from semantics, nothing distinguishes them from the co-ranking serial verbs (in 2.6). (8) Me i u gaa gi kääleng, gi kääleng. DEPEND 3SG PFT PROG cry circle

cry circle

'And she keeps going around and cries/crying, going around crying.' (9)

Hla

thege oga

3PL

run

hi

pwiak.

leave this.DElCT net

'They run and leave/leaving the (fishing) net.' V 2 specifying manner of motion (10)

yha cap tha coot

'escape swimming' (yha 'swim' + cap 'run away') 'fish on foot' (tha 'spear fish with assagai' + coot 'cross a ford') 'run dragging' (thege 'run' + pwiidi 'drag')

thegepwiidi (11)

Hla thege oga 3PL run

hi

pwiak.

leave this net

'They run leaving this fishing-net.' V2 specifying manner or result of action: (12)

oda täimi khaayi mode tilixääc nap

(13)

I

'go up grabbing (a rope)' 'break (by) tugging' (khaayi 'pull', mode "break') 'pretend to be angry' (tilixääc 'be/get angry', nap 'lie, deceive')

magothucong.

3SG sleep do be stubborn

'He keeps on sleeping.' In (14a) nooli specifies the manner of action, but also undergoes some degree of co-lexicalization (since it also means 'ignoring' or 'not paying much attention to them'); whereas (14b) expresses sequential or purposive action and could be rephrased as a dependent clause with me 'and/to': (14) a. Hla wälem 3PL walk

nooli. see.TR

They x pass watching them.' b. Tu go down

nooli. see.TR

'Go (and/to) see them.'

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

175

The sequence may consist of two (or more) transitive verbs sharing the same patient (15a). It may also be rephrased as a dependent clause with me (15b): (15) a. Hä

tami kaamwa-le

lPL.lNCLopen

dohow?-TR

bwat

hleny?

box

this.DElCT

'How do we open that box?' b. Co kaamwa-le me

yo

payaili mashin hleny? set in movement-TR machine this.DElCT 'How do you do to start this machine?"

2SG dohow?-TR DEPEND 2SG

Compare the complex verb construction with harat, harar-ili 'hurry, do in a hurry' (16b) and the dependent clause with me (16c): (16) a. axo

yo

harar-ili!

INJ.NEG 2SG

hurry-TR

'Don't speed it!' b. I

diya harar-ili shaya eli.

3SG d o

hurry-TR

work

this.ANAPH

'He did the work in a hurry.' c. Na

harat

lSG h u r r y

me

na toven.

DEPEND lSG f i n i s h

'I'm hurrying to finish.'

2.7.2.

Specification of direction, trajectory and motion, position

This use of the complex verb construction is restricted to cases when no other morphological or lexical marker is available, whether directionals, modifiers {bwabwali 'sideways') or prepositional nouns (nanamwa-t 'middle'). V 2 specifies trajectory or direction: (17) ο biiyet 'go around' (o 'go', biiyet 'go around') o-xääleng 'go around' (kääleng 'circle, surround') fuk-xääleng 'fly around' (fuk 'fly') (18)

Hlamu kääleng i aayo. 3SG stay surround CONN chief 'They stay around the chief.'

176

(19)

Isabelle Bril

I fuk ulep daxi ni ßvaa-mwa. 3SG fly cross threshold up away in hole-house 'It flies into the house.' (lit. through the door)

V2 specifies the endpoint of the motion or trajectory expressed by Vi: (20) I caabwat-da kuut. 3sg get-up

stand

'He gets up to his feet.' V2 specifies position: (21)

Hli u not kuut mwadu. 3 DU PFT look stand down there 'They watch standing down there.'

2.8. Asymmetric, head-adjunct constructions: manner, qualification As a general rule, the verbal adjunct, whichever its category (active or stative verb), comes last and expresses a wide range of meanings. The Vi and V2 constitute one complex predicate with fused argument structure inherited from the Vi's argument structure. Consider (22a, b): (22) a. I

diya mwang.

3SG do

be bad

'He acted badly.' b. I

hnawe mwang-ili.

3SG drop

be bad-TR

'He dropped it akwardly.' In (22a, b), the V, and the V2 evidence argument fusion not argument sharing: they have different underlying subjects and in (22b), the stative verb mwang inherits the patient of Vj; this is signalled by its transitive concord with V]. Argument fusion and head-driven transitive concord on V 2 (marked by one of various transitive suffixes -e, -le, -(i)lf) signal asymmetric constructions and complex predicates rather than serial verbs (see Aikhenvald 1999: 482; Foley & Van Valin, 1984: 208-210). The tight integration of the constituents of such complex predicates stands in contrast with the looser syntactic dependency of multi-headed serial constructions in which no such concord ever occurs (see 2.6).

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

177

Among the most common stative verbs to appear as V 2 in such constructions are: sho 'be good', mwang 'be bad', khare 'be different', ßvaat 'be clear', bööm 'be soft, gentle', häähuuk 'be recent, new', shuva? 'be how?'; shumweli 'be thus'. Compare the following, in which the b. examples display transitive concord: (23) a. Mee-yo6 (mee 'be ripe', sho 'be good') 'It's fully ripe.' b. Co gaa tili-yoo-li na le kak. 2SG PROG sweep-be good-TRLOC LOC.ANAPH forest 'Go on cutting the undergrowth.' (24) a. Hla

pe-xare.

3PL

REC-be different

'They are different from one another.' b. Hla ο 3PL go

pe-xare. REC-be different

'They go their own way.' c. Na pe-xare-IV. put REC-be different-TR

'Putthem apart!' In (24c), the reciprocal predicate pe-xare7 'be different from each other', derived from the verb khare 'be different', shows transitive concord with a transitive Vi. Here are some additional examples of such sequences: a ni-xare 'go away separately' (a 'leave, go');flie-xare 'take somewhere else' {fhe 'take, carry'); mu-xare 'live separately' (mu 'stay'); na-xare 'put separately' (na 'put'); ni-xare 'send s.o. elsewhere' (ni 'send'); oxare 'go away separately'. The Aktionsart verb häähuuk

'be recent, new' may be transitivized as

V2: (25) a. Häähuuk be

recent

ßvähuux-eli. stoiy-that.ANAPH

"This story is a recent one.' b. Hla diya häähuux-e mwa eli. 3PL do

be recenl-TR house that.ANAPH

'They built this house recently.' So can the similative verb shumweli 'be thus':

178 (26)

Isabelle Bril Ehleena na khabwe shumwelt-le. these lSG say in this way.TR 'These are (the words) I thus said.'

2.9. Complex verbs, specialization, co-lexicalization or compounding? The specification of a Vi by a verbal adjunct V2 (as in 2.8) does not imply any categorial change, just a functional change induced by its syntactic position, in keeping with the general pattern of determination [determiner determined]. Some of these constructions may gradually gain some degree of semantic specialization or may even become grammaticalized, entailing some loss of autonomy, a reduced semantic range or some morphosyntactic modification. Here are a few cases which range from specialization to co-lexicalization, with some fuzzy frontiers between them.

2.9.1.

Functional and semantic specialization

Some verbs of motion or direction (ä 'go, leave', wälem 'walk' 8 ) may specialize as modifiers expressing manner of action in V2 position, with a slightly different meaning, like pavange wälem 'get ready quickly' (27). These verbs are not grammaticalized, they only specialize as modifiers in V2 position and retain their full lexical meaning in other collocations and sometimes even in V2 position (ä wälem 'go walking'). (27)

wälem tabö pavange wälem ä wälem wälem ä ämu mu ä ('stay'+'leave') äpexare (pexare 'be different') pexare ä

'walk (and) go downhill' 'get ready quickly' 'go walking' 'go from place to place' 'go settle' 'live from place to place' 'go separately' 'be different from place to place'

In V2 position, ä 'leave, go' may specialize as an adverbial adjunct with dispersive meaning 'from place to place' or remain a full motion verb as in (28a); example (28b) exemplifies both positions and functions (motion verb and modifier). Function and meaning are thus context-dependent and vary with the degree of idiomatic conventionalization or lexification of a specific collocation:

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

(28) a. Ena

xe co ku

a

me

yo ä

179

oda

now

TOP 2SG stand leave DEPEND 2SG leave go up

ni

hoona

in

this.DEICT

mwa. house

'Now, get up, go and get into that house.' b. Hla ä

mu

a.

3PL leave stay

leave

'They left to live from place to place.' In some cases, a verb has more abstract meaning when used as a V2 in a serial construction. This is the case of bwage 'turn around, go back' which has abstract, reversive meaning as V2: (29) nobwagi 'lookback, look behind' fhe bwage 'give back' (Jhe 'take, carry'). (30)

Wa axe dau hleny, hla wälem bwage agu na bwaa-t. 2PL see island this.DEICT 3PL walk return people LOC top 'You see this island, people stopped there on their way back.' (from a trip to the neighbouring islands).

Similarly, after a Vi of perception, uya 'reach, arrive' specializes as an endpoint marker with the meaning of 'up/down to' showing incipient grammaticalization, but less so after a V] of motion: (31)

/

a

uya

awölö

3SG leave arrive home

Soop. Soop

'He leaves, arrives at Soop's house (or) he goes up to Soop's house.' (32)

Käälek

ο

i

no-du

uya

fwa-jahoot.

be impossible VIRT 3SG see-down reach mouth-river

'It is impossible for him to see down to the mouth of the river.'

2.9.2.

Incipient grammaticalization and categorial change

The verb khoon 'carry sth on one's shoulders, bear', whose t r a n s i t i v e f o r m is khöli, may appear as V2 in a complex verb construction, under its lenited formxööli, with the more abstract reading of 'grudgingly'. (33) I diya xööli. 3SG do

bear.TR

'He did it grudgingly.'

180

Isabelle Bril

The two verbs v . i . f u u n g , v.t. fuuge 'gather, heap up', and v . i . f i i n g , v.t. viing(e)9 'join, put end to end' have similar constructions. As Ν2, fuung has its lenited form wuung, wuuge10 meaning 'together'. (34) a. Hla fuuge nu. 3 PL gather coconut 'They're heaping up coconuts.' Compare the scope of wuuge, on the patient in (34b) and on the agent in (34c) which suggests that it is being grammaticalized as an adjunct, since it can have scope over the subject or the patient according to context. b. Hla hnaxi wuuge shälaga malaaleny. 3PL tie

together-TR

crab

these.DElCT

'They tied these crabs together.' c.

Hla dlya wuuge fagau mwa eli. 3PL do together-TR body house this.ANAPH 'They built the wall of the house together.'

Examples (34a, b) are reminiscent of the scope of the secondary predication, as in she ate her meat raw and she left the party angry, which are close equivalents to core serial construction in other languages, as pointed out by Larson (1991: 201). Thus, the syntactic structure alone is underspecified and it is contextual semantics that orients the secondary predication towards either the agent or the patient. There are a few cases when an intransitive verb, such as niiva 'err, be wrong, be lost, make a mistake', which normally requires a causative or factitive derivation to be transitivized (fa-niiva-(v)i(li') 'deceive, lose s.o.'), may show transitive concord under a slightly different form (niivali) which only appears in complex verb constructions (35c). It starts behaving like Stative qualifying verbs (mwang, sho, etc.), thus signalling a change of category in one of its usages. Its meaning varies with the category of Vi, but generally expresses absence of control and detrimental meaning. With verbs of motion, niiva(li) means 'by mistake, aimlessly', with other verbs, it means 'without any reason, wrongly'. (35) a. I niiva. 3SG m a k e a mistake

'He made a mistake.' b. Hla peäi niiva. 'They fight without any reason.'

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

c. I

diya niiva-li

3sg do

181

shay a eli.

make a mistake-TR work that.ANAPH

'He did this work wrong.'

2.9.3.

Incipient lexification and morphological

change

In the case of not 'see', the loss of the final consonant signals incipient co-lexicalization, as opposed to the serial construction which retains its full form (see ex. (21)). Here are a few examples: no-xiluuk 'look from under' (kiluuk 'bend'); no oxuri 'follow with one's eyes' (oxuri 'follow, go along'); no mwimwi 'recognize on seeing' (,mwimwi 'know'); no huu 'follow with one's eyes, look for' (huu(ri) 'follow'); no-yelä 'recognize' (shelä 'know'). But most often, there is no morphological change to distinguish co-lexicalized verbs from serial verbs and only semantic notions hold, such as the fact that co-lexicalization creates more complex verbal concepts to make up for a lexical or morphological gap. Nor should intervocalic sandhi 11 be considered as a sure sign of co-lexicalization, but it does signal asymmetric serial constructions with modifying function. po-radi 'make noise when tearing' {pot 'make noise', thadi 'tear'); tili-xibwaat 'scatter' (tili 'sweep', kibwaat 'throw away'); khaayi mode 'break (by) tugging' (khaayi 'pull, tug, hoist', mode 'break'); o-rurua 'pass by hiding' (ο 'go', thurua 'hide'); shaya thurue / shaya-rurue12 'do sthg on the sly' (shaya 'work' + thurue 'hide, bury sthg'); shaya ku-(x)äyoot or shaya kwäyootn 'do sthg on the sly' (kwäyoot 'hide'); tUixääc nap 'pretend to be angry' (ülixääc 'get, be angry', nap 'lie, deceive'); yaat äbwe 'thatch with the roots of the straw downwards' (yaat 'thatch' + äbwe 'deviate, send back'); kole pwa 'decant' (kole 'pour', pwa 'fold, bend, break, snap'); pwa nibi 'fold again' (nibi 'fold');pwa khabwe 'answerback' (khabwe 'say'). That such specifications should be expressed by verbs rather than by prepositions, directionals, participial verb forms, nominal forms, etc., is a language-specific phenomenon. Nelemwa and English have similar strategies with motion verbs, though they use different morphemes, whereas French uses participial or nominal forms. Compare: fuk ulep 'fly out'/ 'sortir en volant'. The degree of predictability and conventionalization of such expressions may in turn lead to gradual semantic specialization and true compounding.

182

Isabelle Bril

2.10. Aktionsart verbs in serial constructions Aspect is mostly marked by preverbal morphemes in Nelemwa, except for inchoative and terminative aspects which are expressed by Aktionsart verbs. Among them, only thaaxa 'begin, start' and küüli 'finish, complete' admit serial constructions, toven 'finish, stop' requires a dependency marker. This is evidence that their construction is constrained by the syntactic and argument structure of the verb and does not just correlate with semantics. Similarly, the position of thaaxa and küüli is iconic (with inchoative in V| position and terminative in V 2 position), unlike that of toven, which always appears as Vi with a complementizer. Table 4. Construction of Aktionsart verbs order Inchoative

SVC

Terminative

SVC NO

Terminative (36)

Wa thaaxa 2PL start

SVC

thaaxa + V 2 Vi + küüli toven ο + V

V + rove« 14

shay a! work

'Start working!' (37)

I shay α xüü i f 5 . 3SG work finish

'He worked at it till completed.' Could the verb string in (36) be interpreted as a complementizing construction of Vi and could (37) be a modifying serial construction? In (37), V 2 küüli, with its weakened form xüüli, might undergo some degree of specialization as an aspectual modifier; as yet, it is a secondary predicate that specifies the end of the process marked by Vi. In the case of thaaxa, the nuclear serial construction is restricted to intransitive \ i s , while a transitive V2 triggers an obligatory two-clause construction, marked by the general dependency marker (coordinator or subordinator) me. Thus, (38) with obligatory coreferential subject pronouns for the two verbs can either be interpreted as a coordinate construction (in a.) or as a complement clause (inb.).

Complex verbs and dependency strategies in Nelemwa (New Caledonia) (38) a. [Ix thaaxa] me [ix khuxi α älö hleny kuvic hlenyj. b. [Ix thaaxa [me ix khuxi α älö hleny 3SG start this.DEICT

DEPEND 3SG eat.TR

kuvic

183

hleny]].

AGT child this.DEiCT y a m

'The child starts eating the yam.' The two constructions of thaaxa according to the valency of V 2 could be unified as being two expressions of dependency, either as a tight complex predicate referring to one event (36), or as a dependent clause (38a, b). Example (39a) shows that a verb sequence may contain various hierarchical levels: a loose co-ranking relation between the motion Vj tuume and the other verbs, since it can be autonomized as two sentences referring to two events (as in 39b), in contrast with the tight syntactic and semantic dependency between V 2 and V 3 , thaaxa specifying the initial phase of haga. (39) a. Hli tuume thaaxa haga 3DU go down start fish 'They go down (and) start fishing.' b.

Hli tuume, hli thaaxa haga 'They go down, they start fishing.'

2.11. The case of modal V| A few full verbs with modal semantics appear as Vi with scope on V 2 : kuat 'refuse', shelä 'know, can, be able', mwemweli 'know', shaya 'move, work, try, attempt'. These are distinct from modal modifiers which also appear in preverbal slot, but which are not autonomous verbs and might be former verbs grammaticalized as modal predicates: koni 'unable, impossible', jar a 'desire, feel like'. in (40a), Vi shelä has scope over V 2 and V 3 , and all the verbs share the same subject. Since such constructions reverse the usual order of modification, what kind of underlying syntactic structure can account for this? (40) a. Ε hla shelä [tuume taabwa] shi-wa ru FUT 3PL know

yameewu

avat.

clan

some

go down here

sit

side-POSS.2PL AGT

'Some clans will be able to go north and settle among you.' There seems to be a conflict between the syntactic and the semantic structure of such sequences. V\ shelä behaves as the syntactic head of a

184

Isabelle Bril

complementizing structure, with V 2 and V 3 as its syntactic complements; yet, the semantic node is expressed by V2 and V3, and shelä is an optional modal specifier (compare 40a, b). There are different hierarchies as well, V 2 and V 3 are co-ranking but are under the scope of V^ b.

Ε

hla

FUT 3PL

[tuume

taabwa] shi-wa

go down here sit

ru

yameewu

avat.

side-POSS.2PL AGT clan

some

'Some clans will go north (and) settle among you.' The verbs share the same subject pronoun and the scope of negation bears on the whole sequence: (41) a. Kio na NEG

mwemweli bwagi-na

ISG know

return-LSG

ο VIRT

khabwe say

kiya there is no

Kaavo Kaavo

mudi-ny. company-poss. ISG

Ί can't go back without Kaavo in my company.' b.

Kio na bwagi-na ο khabwe kiya Kaavo Ί won't go back without Kaavo in my company.'

mudi-ny.

Nor are these verbs modal auxiliaries, they remain full lexical verbs whose modifying function as modal predicates is context-dependent or position-dependent. In (42a), shaya 'move, work' is the head specified by V 2 harat 'hurry', whereas in (42b), shaya is the specifier meaning 'try': (42) a. I

shaya

3SG move

harat. be in a hurry

'He works in a hurry.' b.

I

shaya

3SG move

ä-yayap16. go-RED.escape

'He's trying to escape.' Thus, the polysemic verb shaya specializes as a modal predicate in some collocations, but shows, as yet, no sign of grammaticalization as a modal auxiliary. Only once they have reached an advanced degree of grammaticalization and divergence from the original lexical root do they become homophonic elements with possibly fairly different meaning. The complex verb construction is thus one of the two possible complementation strategies for a few verbs with modal semantics. One of the verbs is then construed as the argument of another verbal head, according to such factors as control of coreference (43) or transitivity of V2. Thus kuat 'refuse' admits a tight complex verb construction (43a) when the verbs share the same subject/agent, but a "deranked" complementation marked by

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

185

the irrealis marker ο otherwise. In (43a), the verbs are not co-ranking, since Vi modifies V 2 , but at least, there is no morphological imbalance as in (43c). (43) a. Hla kuat axe fwamwa pwagiik ο i oda. LSG refuse see country CLASS-one VIRT 3SG progress 'They refuse to see a place that's making progress.' b.

c.

Ix kuat yagei-ey a Pwa-Kebox. 3SG refuse help-3so AGT Pwä-Kebö 'Pwä-Kebö refuses to help him.' Ix

kuat

F'wä-Kcböy. Pwä-Kebö 'He refuses that Pwä-Kebö (should) help him.'

3SG refuse

3.

ο

iy

VIRT

3SG help-3sc

yagei-ex

α

AGT

Syntactic and semantic constraints to serial verb construction

Complex verb constructions (CVCs) often play the role of a morphosyntactic gap-filler, expressing directional 17 , aspectual or case-marking functions by lexical rather than morphological means. In Nelemwa, syntactic constraints on the use of CVCs are of two types: (i) to license an argument that is not part of a verb's argument structure; (ii) to license an additional argument to an otherwise intransitive verb.

3.1. Verbal semantic structure The role of the CVC is to license an argument whose semantic features are not compatible with the main verb's. Thus, in (44), uuri 'carve, sculpt, scrape' requires an inanimate patient and the complex verb licenses an animate patient (uuri uyamwili: lit. carve imitate): (44) a. I uyamwi kääma-n. 3SG imitate father-POSS.3SG 'He imitates his father.' b. I uuri uyamwili fagau-t agu. 3SG carve imitate body person 'He's sculpting a person's body.' The same goes with Hi 'write, draw': Iii uyamwili 'draw (a person)'.

186

Isabelle Bril

3.2. Valence increase and case-marking function: locative complements It is a common cross-linguistic fact in Austronesian and African languages that serial constructions may have valency-increasing function and integrate more participants than the basic arguments which are part of a verb's argument and syntactic structure. Thus, Austronesian languages, which generally do not allow ditransitive verbs or constructions18, make use of serial constructions to integrate and case-mark such participants as beneficiary, comitative, instrumental and even directional or locative complements. In Nelemwa, participants other than the basic arguments are usually marked as prepositional phrases and infrequently by complex verb constructions. Yet there are a few such cases with the verbs oxo (+animate) or oxuri (+inanimate) 'follow' (45, 46), which are both direction markers and argument-expanding devices licensing a patient (thaamwa, (45)) or a locative complement (hi mat, (46)) to an otherwise intransitive Vi such as thege 'run', ο 'go', wälem 'walk' or teewot 'sail before the wind'. Again, they serve as lexical and morphological gap-fillers, since there is no other preposition or directional marker to express such notions as 'after' or 'along'. (45)

I

thege oxo

3SG run

follow

thaamwa. woman

'He runs after the woman.' (46)

I u 3SG pft

haga ο oxuri-da hi mat, [...] fish go follow-up this low tide

oda-xi uya Fhalap. go up away reach Fhalap ' She goes fishing following the (low) tide, goes up away, reaches Fhalap.' (47)

I u wälem oxuri hnawa horaamwaleny Kaavo. 3SG pft walk follow trace this woman Kaavo 'He walks following Kaavo's footprints.'

(48)

I u oot oxuri ve on. 3sg pft sing follow dir sand 'He sings along the beach.' (* i oot ve on is ungrammatical)

Complex verbs and dependency strategies in Nelemwa (New Caledonia) (49)

I

u

teewot

oxuri

kuru.

3SG PFT sail b e f o r e w i n d

follow

wind

187

'He sails following the wind / before the wind.' In some of its usages, oxuri 'follow' may (though more infrequently) also specialize as a specifier meaning 'according to' and express quantification and degree, as in (50), where N j j a x e 'measure' normally requires a concrete patient. V 3 oxuri licenses the quantification of an abstract notion (money) and thus bypasses the constraint on the semantic frame oijaxe: (50)

Na iyult jaxe oxuri jaxa mwani i na. lSG buy.TR measure.TR follow.TR measure money CONN lSG Ί bought enough of it according to what money I had.' (lit. according to the amount of my money."

The verb string in (50) is not a serial construction since the verbs have different patients: the patient of V! is the thing bought, whereas the patient of V 2 is the amount of money required for that; as for V 3 , it evidences functional specialization as a valence-expanding device. The specialized use of a verb 'follow' as a case-marking or valency-expanding device is also attested in Ajie (New Caledonia), where two verbs (veri 'follow' and xara 'be at') have grammaticalized as prepositional associative markers. See also example (51) in Mwotlap (Francis, this vol.) and Paamese (Crowley 1987: 67).

4.

Complex verb constructions and clause dependency

This section analyses complex verbs in the overall system of VP or clause linkage (coordination, complementation or subordination) with a typological perspective. It will focus on the syntactic function of two verbs 19 used as clause-dependency markers in complex verb constructions in Nelemwa: uya 'reach, arrive', which is grammaticalized as a time boundary conjunctive marker with the reading 'until', and the quotative verb khabwe 'say' which is the complementizer of discourse, cognition and more rarely perception verbs: tälä khabwe 'hear, understand that', shelä khabwe 'know that', nanami khabwe 'think, believe that', häuk khabwe 'not know that', hangi khabwe 'suppose that'.

188

Isabelle Bril

(51)

Na u hangi khabwe kebuk. lSG PFT suppose say be true Ί supposed it was true.' The following tries to assess the frontier between the lexical use of khabwe in a serial verb construction and its grammaticalization as a complementizer. But this is a context-dependent continuum, with synchronically coexistent functional layers. In (52), all verbs share the same subject and khabwe retains its full lexical meaning 'say', thus counting as a case of serialization. (52) Co mwa ä 2SG ASS

shinook

leave glance

ni hoona in

khabwe je

a

say

POTENT what?

be.LOC

da

mwa.

this.DElCT house

'Go, take a look (and) tell us what is in that house.' By contrast, in (53), the verbs do not share the same subject, tälä is a verb of perception and khabwe undergoes some semantic bleaching as the complementizer of tälä: (53) Hli u tälä khabwe aax-iik agu xe i hää 2DU PFT hear

say

CLASS-un person

uya

mwada

awölö.

arrive

up there

home

TOP 3SG recent

'They hear that someone has arrived at their place.' Similarly, in (54), the verb sequence is hierarchized, V2 thabwi is the modifier of V^ and constitutes a complex predicate, while khabwe is a partially grammaticalized complementizer, since most of its meaning is preserved ('able to say'): (54)

Kia

ho

i

mwemweli thabwi

there is not this 3SG know

hooli

khabwe je

take care say

pa ...

be(LOC) where?

foliix-eli.

that.ANAPH thing-ANAPH

'Nobody really knows where that object is.' (lit. nobody really can say where...) By the criteria used so far, the four-verb string in (55) is not a serial construction, for it contains various types of hierarchized constituents, some of them with functional specialization:

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

189

(55)

I no than khabwe käälek ο i no-du 3SG see close say be impossible viRT 3SG see-down uya fwä-jahoot. reachmouth-river 'He can't see anything, for it is impossible for him to see down to the mouth of the river.' (i) The verbs do not share the same subject; (ii) V 2 than specifies V] and constitutes a co-lexicalized sequence or a compound verb 2 0 meaning 'have one's view blocked, see nothing'; (iii) V 3 khabwe 'say' is grammaticalized into an explicative marker meaning 'that is to say' 2 1 ; (iv) finally, V 4 käälek 'be impossible' has scope on the following clause and the verb string belongs to different underlying clauses. This is evidence that a surface collocation is misleading and requires closer analysis. It also reveals the limits of prosody as a criterion, since no perceptible pause can help dissociate the verbs in (55).

5.

Paratactic, serial or dependent constructions

Serial constructions may be syntactically constrained, but there are some discourse options between various types of dependencies: paratactic or serial or dependent constructions with dependency morphemes.

5.1. Serial verbs vs dependent construction: a discourse choice Such verb strings are sometimes analysed as resulting from ellipted dependency markers. In Nelemwa, this could only hold true with active verbs, as in (56), but its explicative value is not very far-reaching, since it does not explain why such condensed structures exist at all in a particular system and not in all languages, nor does it explain the semantic, discourse or conceptual differences between a serial construction and a dependent clause.

5.1.1.

Serial vs coordinated

verbs

Corpus analysis of narratives shows that serial verbs or VPs and clauses conjoined with me and xa 'and' are more common than paratactic constructions. Serialized active verbs refer to one single event comprising

190

Isabelle Bril

several immediately sequential or simultaneous actions (56a); whereas dependent multi-clause constructions refer to two different events (56b) or stress argumentative or logical links between them. (56) a. Na CONTR

yo

thege

ulep

me

yo

2SG

run

cross threshold

DEPEND

2SG go

maa-on

bai

mwexi.

point-sand

that.ANAPH

there

ä

bwa on

'Run out and go to the tip of sand over there.' thegela22 me i ulep. CONTR 3SG ASS.PONCT run DEPEND 3SG cross threshold 'And she runs and goes out.' The same goes for (57a, b), the coordination marker xa 'and also, too' refers to two distinct events, although xa signals a more immediate sequence than me\ (57) a. Hla oda kuut ni hooli yang. 3PL go up stand in that.ANAPH festival 'They go (south) to take part in this festival.' b. Na

i

xau

b. I tu axaleny dagiiny xa kuut ni hooli civa. 3PL go down this man messenger also stand in that.ANAPH dance 'The messager goes down and takes part in the dance.' (58) a. I

u

no-da

axe

3SG PFT look-up see

mweyaayec xe

pwa-giik.

house

CLASS-one

TOP

'He looks up (and) sees a house.' b. I

u

no-du

me

i

u

axe.

3SG PFT look-down DEPEND 3SG PFT see

' She looks down and sees it.'

5.1.2.

Serial vs dependent purposive clauses

The choice of a serial construction with a Y] of motion and an active V2 (59), rather than a dependent clause, is a discourse strategy meant to increase rhythm and compress information. Most of the time, the serial construction can be rephrased as a dependent clause with me, with the same ambiguity between a coordinate reading 'and' and a subordinate purposive reading 'in order to'. But the complex clauses stress argumentative or logical links.

Complex verbs and dependency strategies in Nelemwa (New Caledonia) (59) a. Na

191

i

khabwe: "tu thoogi-e dame awölo". COORD 3SG say go down call-3SG up here home 'And he says: "go down (and/to) invite her up here at home".'

b. Na

hla aa

tu me hla thoogi-e da awölö. go down DEPEND 3PL call-3SG up home 'And they go down again and / to invite her up there to their home.' COORD

3PL ITER

(60) a. Wa oome thu da? 2PL come do what? 'What did you come here for?' b. Wa oome me wa thu da? 2PL come DEPEND 2PL do what? 'What did you come here to do?' c.

Wa oome me

pwaxa

thu da? do what? ' W h a t ' s the purpose of your visit here?'

2PL come DEPEND necessity

It is a cross-linguistic tendency for serial verbs comprising a motion verb and an active verb to have this double (coordinate or purposive) reading. Whereas complex verbs comprising a stative V 2 (as in 61a) impose an asymmetric, modifying reading, and exclude the purposive reading; nor can they be interpreted as resulting from deleted dependency markers. Compare: (61) a. I

diya mwang-ili

3SG do

bebad-TR

mwa. house

'He built the house in the wrong way.' b. I

diya me mwang mwa. do DEPEND be bad house 'He made it so as to spoil the house.' (lit. he did things in such a way that the house was spoiled) Example (62) also questions the notion that serial constructions result from ellipted dependency markers and gapping, for the ellipted coreferential pronouns in the second part of (62) do not result in serial constructions, but in paratactic clauses: 3SG

192

lsabelle Bril

(62)

Oda mwa ye, gi ye mwaida go up ASS 3SG.FR be.LOC 3SG.FR up there 0 fhe-me hooli cet, take-here

that.ANAPH

awölö, home

pot

'She goes up, she's up there at home, brings that pot, 0 na-daxi bwa to, 0 kole-duxi wi. put-up away on stones pour-down away water puts it down on to the hot stones, pours water into it.' Compare (62) with the coordinated clauses (63), which slow down the narrative rhythm and convey a feeling of pain and difficulty: (63)

me

i

/

oda-me me come up here DEPEND pe-kälap mwa bwabuk.

3SG

INTENS-lie

DEPEND 3SG

ASS

i

pe-thalic

3SG I N T E N S - s t u m b l e

me DEPEND

below

'And she comes up here and she keeps stumbling and she keeps falling to the ground.' And similarly in (64a, b): (64) a.

Oöxi Pwä-Kebö 0 khure shi Pwä-Hivic. approach Pwä-Kebö touch hand Pwä-Hivic 'Pwä-Kebö approaches, takes Pwä-Hivic's hand.'

b. Pwa ööxi me yo shaavic mwa. a little approach DEPEND 2SG warm ASS 'Come closer and/to get warm by the fire.' Serial verbs express immediately sequential or simultaneous actions, whereas the dependency marker refers to a distinct event: (65) Yo xau coot-ve täimi shi horaamweeli Kaavo 2SG ASS.PONCTjump-DIR take hold hand this woman Kaavo me

hä

u

ulep! DEPEND IPL.INCL PFT cross threshold 'Jump away, take Kaavo's hand and let's get out!' All this indicates that, rather than reducing serial or complex verb constructions to other structural types of dependency through ellipsis, such constructions should be analysed as one possible configuration on a cline of various types of dependencies, going from tight co-ranking, argumentsharing predicates to asymmetric modifying predicates, or to looser asyndetic and syndetic coordination and subordination strategies.

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

193

Complex predicates represent the highest degree of compactness and morphological economy to express syntactic and semantic specification (Aktionsart, aspect, modality, qualification, manner, direction, result, purpose, etc.) and more infrequently complementation. It is thus analogous to the cline of dependency-marking for noun determination (juxtaposed, direct or indirect marking) but in ways that are related to the verbal or predicate sphere (i.e. as complex construction vs coordination or subordination). But while the possible choice between syndetic coordination or coranking serial verbs signals some discourse strategy, the type of complementation is submitted to syntactic constraints and verb subcategorization.

6.

Conclusion: synchronic and diachronic perspectives

Serial and complex verbs are thus a fairly underspecified structural pattern, on which are mapped a great variety of cross-linguistically common syntactic and semantic operations. Yet, their structural type is very varied, even among closely related languages. Several criteria help distinguish genuine serial constructions from verb strings with semi-grammaticalized syntactic-marking functions, as well as serial constructions from conventionalized and semantically less predictable co-lexicalizations. In Nelemwa, semi-grammaticalized verb strings with argumentexpanding functions and case-role marking functions are morphosyntactic gap-fillers and never coexist with productive morphemes with a similar function. Case-roles are generally marked by prepositions (such as the instrumental marker ο) or by prepositional nouns (such as the recipient marker shi); serial verbs infrequently have case-marking functions and when they do, they mostly mark peripheral participants which are lower in the semantic hierarchy. Similarly, the very productive factitive and causative prefixes pa-, fa- might explain the noticeable absence of causative or resultative serial constructions in Nelemwa, such as I made them go or I struck him dead. Resultative notions are expressed by independent clauses (66) or by factitive derivation (67a, b), with focus on the cause(r) rather than on the result: (66)

Hli u khiibo hörn, maak. 3DU PFT hit the mute one die 'They hit the mute one, (she) died.' (with a pause before maak)

194

Isabelle Bril

(67) a. I

khi

3SG hit

pa-maaxa-e. FACT-die.TR-3sG

'He beat him dead.' (lit. he strike cause-die-him) b. Hla ta-pa-maaxa-e

ο

do.

3 PL pierce-FACT-die.TR-3SG CONN assagai

'They killed her with assagais.' (lit. they spear cause-die-her) Since, in Oceanic languages, cause-result or cause-effect notions are very frequently expressed by switch-subject core-serialization, with its [sVo sVo] pattern ( Ί caused them they do it'), the lack of core-layer serialization in Nelemwa might account for this. But Mwotlap (this vol.), with its nuclear-layer causative serial construction invalidates this hypothesis. Interestingly, Mwotlap and Nelemwa have made different choices, Mwotlap has developed the serial construction strategy over the now decayed causative prefix, which only survives in a few unanalysable verb forms (see Franfois, this vol.: fn. 13), while Nelemwa favours the morphological marking of causative notions and case-roles such as beneficiary, locative, etc.

6.1. Grammaticalization A distinction must be made between grammaticalization and specialization. Grammaticalization refers to cases involving some morphophonological or categorial change, signalled for example by loss of autonomy. Specialization refers to cases which evidence synchronic functional layering of an item, either as a full lexical verb or as a syntactic operator, without any morphophonological or semantic change (only involving semantic narrowing), its change of function being context- and position-dependent. In spite of the typological prediction that nuclear serial constructions with their VV pattern favour the erosion and drift of V 2 as either a transitive morpheme or an adposition, there is only one such case of divergent evolution and grammaticalization of a verb into an applicative associative marker in Nelemwa: this is the case o f f l i e 'take', which has grammaticalized under the form ve (+inanimate), vi (+animate), while retaining its full verbal form flie. (68) a. I

thege-vi-e

ru

hooc.

3SG run-ASSOC-3sG AGT horse

'The horse runs with her (on its back).'

Complex verbs and dependency strategies in Nelemwa (New Caledonia) b. I

hi para nu na mwada. 3SG PROG crawl-ASSOC this crumb coconut LOC up there 'It (ant) crawls across with this crumb of coconut inside (the house).'

c.I

gaa

195

yuup-ve

u flie dume mwa taa-ve na bwa kavebu-n. take down here ASS sit-ASSOC LOC on kingdom 'He brought it here to the north (and) settled with it in his kingdom.' 3SG PFT

Cross-linguistically, a few verbs 'go, give, say, make, take' commonly grammaticalize or specialize as syntactic or case-marking functional morphemes. There are also a few cases of a centripetal drift of V2s into adverbial adjuncts, with various degrees of grammaticalization, such as loss of autonomy or delexicalization (i.e. loss of free lexemic status, Vinogradov 1990: 120) and categorial change, yet they retain their full meaning and some of their verbal properties, since they still show transitive concord (see 2.2). Finally, there are also a few cases of delexicalization without desemanticization (i.e. semantic loss), with the non-autonomous modal verbs koni 'be unable' and jara 'feel like' which behave as quasiauxiliaries; it is a common cross-linguistic tendency for verbs such as 'wish, know, be powerful, think' to grammaticalize as modality operators.

6.2. Specialization On the whole, most Aktionsart or modality verbs have not undergone complete categorial change or loss of autonomy; they just evidence functional layering as full lexemes, syntactic operators or grammaticalized markers, according to context and position. Besides, specialization or grammaticalization are contiguous phenomena with somewhat fuzzy frontiers; their identification is contextdependent, varying with the degree of semantic abstraction or narrowing. Thus, in argument-expanding serial constructions, the motion verb oxuri 'follow' tends to specialize as an adposition with the spatial reading 'along', but it is grammaticalized with the more abstract reading 'according to' (see 3.2.). Similarly, uya 'arrive' specializes as a spatio-temporal endpoint marker (meaning 'up to') in some contexts, but is fully grammaticalized as a temporal conjunction, meaning 'until' in other contexts.

196

Isabelle Bril

Finally, the quotative verb khabwe 'say' also evidences a cline in its complementizing functions, from an operator that preserves its lexical meaning 'say' to an utterly desemanticized complementizer. As long as there is no clear morphophonological change, the meaning and function of such items varies from lexical and context-free (non positional) to context-bound (combinational, Vinogradov 1990: 119), with possible semantic narrowing. Its interpretation requires some contextual computation, activating or desactivating some features to retain sometimes only abstract semic features. In time, formal divergence may separate the lexical item from the syntactic operator or morpheme; but synchronic stratification of functions and meanings is one stage in the universal diachronic tendencies that Hopper (1990) labeled «layering, specialization and divergence». Does the stratification of functions and meanings always result from some evolutionary and diachronic process? Might it not also be analysed as a synchronic strategy for economy of form and function? This would allow the mapping of several functional layers onto one form, without loss of meaning, their specific function being context-dependent.

6.3. Co-lexicalization and idiomaticization Finally, serialization is not correlated with scarcity of lexical derivation in Nelemwa, verb compounds are numerous and favoured by verb contiguity [VV], They differ from serial verbs by their degree of morphological erosion and conventionalization. In Nelemwa, only a few verbs of posture (lie, stand, sit), and gesture (strike, pierce) have evolved as verbal prefixes in verbal compounds, whereas some languages further south on the Mainland have developed in the opposite direction with proliferating compounds and infrequent serial verbs (Ozanne-Rivierre and Rivierre, this vol.).

Complex verbs and dependency strategies in Nelemwa (New Caledonia)

197

Orthographic conventions and abbreviations The main orthographic conventions used in Nelemwa are as follows: c = [c]; th = [ ή ; kh = [χ]; χ = [γ]; vh = [β]; b = [mb]; d = f d ] ; j = ft]; g = [,Jg]; ng = [q]· AGT AN ΑΡΗ ASS ASSOC CLASS COORD CONN CONTR DEICT DEPEND DIR DU FACT FR

FUT INCL

agent anaphoric assertive associative classifier coordinator connector contradiction deictic dependency marker directional dual factitive free form future inclusive

INTENS ITER LOC NEG PFT PONCT POSS POTENT PROG RED TOP TR VIRT

V VP

intensifier iterative locative negation perfective punctual possessive potential progressive reduplication topicalization marker transitive virtual/irrealis verb verb phrase.

Notes 1. 2.

See Foley and Van Valin 1984, Crowley 1987, Durie 1997. In Paamese (Crowley 1987), "core-layer serialization" is productive and associated with intransitive verbs (basic motion or posture verbs, Stative, intransitive verbs), while "nuclear layer serialization" is more restricted and mostly associated with transitive verbs. 3. Various Oceanic languages display such concord on modifiers (see Bril, this vol. "Complex nuclei in Oceanic languages: contribution to an areal typology"). 4. pe-xa-giik (reciprocal mkr-distributive mrk-one). 5. -(i)lt is cognate with Proto Oceanic (POC) *(aki)ni. 6. Intervocalic sandhi: sho /J-/ > yo 1)1. 7. Intervocalic sandhi: khare /x/ > -xare /γ/. 8. The general verbs of motion ο 'go' and tu 'go down' never occur as V2 with modifying function. 9· Intervocalic sandhi /f/ > /v/ with front vowels. 10- Intervocalic sandhi: /f/ > /w/ with back vowels.

198

Isabelle Bril

11. Intervocalic sandhi is not obligatory, it is just frequent and does not constitute evidence of compounding. Ex.: bwa wone or bwa pone 'on the rocky land', vhaa (f)waat 'speak clearly' (vhaa 'speak' + fwaat 'be clear'). / p / o r / f / > [v]ou[w] Ν > [r], [1] /k/or/x/ > [x] > 0 /fw/ > [w] Λ*/ > [r] /c/or/J - / > [j] 12. Intervocalic sandhi and loss of aspiration: 1^1 > Irl. 13. Loss of the intervocalic velar stop and labiovelarization: ku-(x)ayoot > kwäyoot. 14- In post-veibal position roven 'finished entirely, completed' is an aspectual modifier verging on quantification, not a verb. I xau tii me foro roven. 'He painted it entirely white.' 3SGASS

paint

DEPEND be white finish/all

15. Intervocalic sandhi: Ik/ (küüli) > /γ/ (χύΰΐΐ). 16. Reduplication of cap 'run away' and intervocalic sandhi /c/ > /j/. 17. Directions, other than those expressed by some directionals 'up, down, across, away', are often lexicalized by a verb. 18. Foley states that Austronesian languages do not have ditransitive verbs except by contact with other languages ("A typology of information packaging in the clause". In Shopen, forthcoming). 19. In Nelemwa, most other clause-dependency markers are conjunctions or relational nouns expressing logical relationships: puxe-t 'reason, origin, cause' > 'as, since';pwaxa-t 'necessity' > 'so that', etc. 20. This is evidenced by other such expressions mu than 'stay, live closed up' with literal meaning (in a place), or figurative meaning (in one's thoughts). 21. In sentence initial position, it is also an explanatory conjunctive locution meaning 'that is to say'. 22. Thegela is the intransitive form and thege the form found in serial and complex constructions.

Complex predicate constructions in East Uvean (Wallis) Claire Moyse-Faurie

Abstract Strict serial verb constructions, denoting one event including simultaneous actions, are quite rare in F a k a ' u v e a , mainly for two reasons. First, one o f the elements o f the series tends to e v o l v e either into a modal or aspectual morpheme ( V i ) or into a manner adjunct (V2)· A f e w lexemes specialize in both positions. Second, coordinate and complementizing complex predicative structures are often preferred to SVCs, referring to quasi simultaneous events for the former and to successive actions o f a single event for the latter.

East Uvean 1 belongs genetically to the Nuclear Polynesian subgroup of the Polynesian family, but displays a number of Tongic-type features as a result of Tongan colonization. It is an ergative language, with a P(redicate)OA or PAO word order where A and Ο are lexical items or independent pronouns, or a/.vP(0) word order (where a or 5 are preposed clitic pronouns, a standing for the ergative argument of a transitive predicate, s for the unique argument of an intransitive predicate or the absolutive argument of an indirect transitive predicate). It shows no inflection for tense, aspect or case, all these grammatical markers being independent morphemes. For less than thirty intransitive predicates, number is marked by partial reduplication 2 . Although marginal, this fact will be used as a criterion to enlighten cases of concord in coordinate or serialized constructions.

1.

Word categorization and complex predicates

As in the other Polynesian languages, word categorization is a tricky problem to handle. First of all, no formal distinction can be made among lexemes out of context, that is, at the lexical level. Second, any lexeme may function as a predicate, with tense-aspect markers, or as the head of a

200

Claire

Moyse-Faurie

non-predicative phrase, preceded by an article or a possessive pronoun (cf. Broschart, 1997). Hence, grammatical morphemes select different syntactic contexts: tense-aspect markers define a "predicative context" (elsewhere called "verbal context"), whereas articles, classifiers, deictics and possessive pronouns define a "non-predicative context" (the so-called "nominal context"). However, semantic restrictions do exist and some lexemes may not combine with any classifier, directional, quantifier or with any article. Indeed, although any lexeme describing an action may be the head of a nonpredicative phrase, some of them, and especially transitive ones, appear in most cases only preceded by the specific singular article te or by a specific singular possessive determiner. They may also retain directional adjuncts and arguments expressed either as possessive determiners or preceded by a case-marking preposition. Also, although it is true that some lexemes have exactly the same meaning in both syntactic contexts, some show a broader or reduced meaning in one of them 3 . Though the problem of the verb/noun opposition (its existence, or the linguistic level where it may function) is not the topic of this paper, word categorization appears to be a crucial issue, prior to any consideration of "serial verb constructions". Indeed, how are we to determine if we are dealing with two serialized "verbal predicates", or with a succession of a head + determiner, since no morphological indications are given? I shall illustrate this problem by contrasting the following sentences: (1) a.

(2)

Έ fa'u malimali e te tamaliki NS build smile ERG DEF children 'The children build a boat smiling.'

te vaka. DEF boat

Έ fa 'u hiva te kau 'aliki. build song DEF CLAS chief 'The chiefs are composing songs.'

NS

Fa 'u is a potentially transitive (ergative) predicate, referentialized by the non-past (non-specific) tense-aspect marker 'e. In (la), it is followed by two arguments, one in the absolutive case (marked by the preposition ia or 0 ) and one in the ergative case (always marked by the preposition e). The adjunction of malimali 'smile' (which may elsewhere occur as a sole indirect transitive predicate, with two arguments, one in the absolutive case, and one in the oblique case marked by the preposition ki or i) does not change the valency of fa 'u. We are dealing here with a complex predicate, referring to a single event including two simultaneous actions.

Complex predicate constructions in East Uvean

201

Transitive predicates may also occur with only one expressed argument. Depending on their semantic orientation, this unique argument will refer to a patient or to an agent. Fa 'u is agent-oriented. If only one argument (hence in the absolutive) is expressed, it will refer to an agent: (l)b.

Έ fa'u

malimali

NS build smile

te

tamaliki.

DEF children

'The children are building [something] smiling.' The complex predicate fa 'u malimali still expresses the same single event including two simultaneous actions (smiling and building) as it did in (la). In ( 2 ) , f a 'u has also only one argument (hence in the absolutive), te kau 'aliki, referring to the agent. The complex predicate fa 'u hiva does express a single event, but not two simultaneous actions. Here, hiva is a kind of determiner - the role that incorporated object has in Oceanic languages and takes a generic meaning. Of course, in other contexts, the lexeme hiva may occur as an ergative predicate: (3) a.

Έ

hiva

NS sing

te

kätiko

e

Paulo.

DEF canticle ERG Paulo

'Paulo sings a canticle.' Its absolutive argument {te kätiko) may be incorporated (3b); it loses its article, the predicate becomes intransitive, its unique argument is now in the absolutive case instead of in the ergative case in (3a), and the complex "predicate + incorporated object" takes a generic meaning: (3) b.

Έ hiva kätiko ia Paulo. NS sing canticle ABS Paulo 'Paulo sings canticles.'

Kätiko in (3b) has the same function as hiva in (2), and both (2) and (3b) refer to an event consisting of a single action. So it is not only the occurrences, in the predicative slot, of successive lexemes, which may occur elsewhere as autonomous predicates, that will tell us that we are dealing with serial constructions, but the meaning itself of the lexemes in this particular context. In a few cases, a change of valency will allow us to differentiate object incorporation from a serial construction. But most often, only the semantic notions of event (state, affect or action) and parts of an event (consisting of simultaneous actions, affects or states) will give us a chance to differentiate a succession like fa 'u malimali in (1) from another type of succession like fa 'u hiva or hiva kätiko in (2) or (3b). In most cases nothing can help for-

202

Claire Moyse-Faurie

mally to make a difference between a compound "V+incorporated object" and a SVC. Why successions of lexemes evolved towards an idiomatic formula, as fa 'u hiva which means 'compose songs' and not 'build while singing', whereas fa'u malimali means 'build while smiling', may have semantic or pragmatic explanations but not syntactic ones. Like malimali, hiva may refer to an action; but, contrary to malimali, hiva is also perceived as an "object" that may be built and that allows it to form a compound with fa 'u. Before turning to the analysis of what can be called strict SVCs in East Uvean, we must also consider another type of succession of lexemes occurring in the predicative slot, as shown in example (4a): (4) a.

Έ

lanu

kula

te

NS

colour

red

DEF CLAS

mo 'i kie. loin-cloth

'The loin-cloth is red.' (lit. is colour red) A similar meaning can be obtained without any change in valency, but with a complex argument instead of a complex predicate: (4) b.

Έ

kula

NS red

te

lanu

ο

te

mo 'i kie.

DEF colour POSS DEF CLAS loin-cloth

'The colour of the loin-cloth is red.' Only a few lexemes belonging to specific semantic fields (body-parts, inherent qualities, words meaning 'attitude' or 'manner') can occur as first element of the predicative complex, instead of the head of the argument. Most of them (especially body-parts) seldom occur as predicate, but some do, like lanu in the example below: (4) c.

Έ lanu te lopa. NS (have a) colour DEF pearl 'The pearl has a [nice] colour.'

Followed by a colour name, lanu forms a kind of generic compound with the predicate, similar to the English "long-legged" construction. With body parts, we get the same kind of Stative compounds, most often expressing chronic or congenital diseases, as for examples ihu pe'e /nose/soft/ 'to have a cold', kili felea /skin/spotty/ 'to have scabies', va'e ketu /foot/limp/ 'to limp', mata hali /eye/trickle/ 'to have conjunctivitis', and so on. In all these cases, the predicate is formally complex, but on the semantic level, the first element just specifies the localization or the type of intrinsic quality of the state expressed by the second element. Hence, here again, we are not dealing with a complex event, but with a compounding process denoting a single action.

Complex predicate constructions in East Uvean

203

I will now turn to the main topic of this article, that is how East Uvean expresses complex events. To make it short, I will label "verbs" or "verbal predicates" any lexeme that may be referentialized by a tense-aspect marker and which refers to a state, an affect or an action. East Uvean presents different types of complex predicates expressing complex events: a) The predicative phrase consists of two adjacent lexical elements; these lexical elements may be autonomous, as independent predicates, without any morphological modification. I will first consider successions of lexemes which refer to a complex event, keeping in mind that each of its facets occurring alone could be an event in itself. Then, I will examine diverse cases leading to specialization of one of the lexemes, which gets us out of the scope of strict serialization, and back to a situation close to the ones discussed above, with a succession of the type "predicate + modifier" or vice versa. Note that if these successions of lexemes referring to a complex event most often occur as complex predicates, they are also found as heads in a nominal context (see example 13 below). b) The predicative phrase consists of two predicates coordinated by the coordinative and comitative marker mo\ each predicate may retain its own object or determiners but they usually share the same subject, whatever the position of this subject. c) The predicative phrase consists of two predicates coordinated by the conjunction o. Constraints are generally the same as with mo although it is sometimes possible to insert a pronominal subject in the second conjunct coreferential 4 with the subject of the first conjunct. Each of these three complex predicative structures has significant semantic correlates, but only the first one can be called "serial verb construction" although, as I just said, there are various degrees of grammaticalization that push some of these constructions out of the scope of strict SVCs.

2.

Verb serialization

Verb serialization (in the strict V,V 2 sense) is a fairly rare phenomenon 5 in East Uvean, although it is quite grammatical and accepted as such by native speakers. It is rare, because there is a strong tendency in this language to prefer one predicate utterance followed by a nominalization, rather than two

204

Claire Moyse-Faurie

head predications (see 2.3). In my entire text corpus (more than thirty hours of recordings), I only found some thirty occurrences of SVCs, most of them involving specialization (see 2.2, 2.4 and 2.5). Besides, the distinction between nuclear-layer and "modifying" SVCs is often hard to make. Here again, we will find no syntactic criteria and we will have to rely upon semantics or pragmatics: are we dealing with simultaneous actions belonging to the same event, or to a single-faceted event? The strict SVCs fall into three subcategories, each having important correlates, due to various grammaticalization processes: V1V2 without specialization (nuclear-layer SVC) V1V2 withV 2 specialization V1V2 with Vi specialization The last two cases correspond to what Gunter Senft [this volume] considers as grammaticalization when used in an SVC.

2.1. V! V 2 without specialization Prototypical nuclear-layer SVCs showing no specialization generally consist of active predicates of various valencies. The respective status and original meaning (that is when not in a SVC) of each predicate are always preserved. Semantically they refer to a single event comprising simultaneous, not successive, actions without any notion of purpose. In this type of structure, the two predicates cannot be separated by any constituent, be it a coreferential pronoun or the included object of Vi, that is, core-layer SVCs do not exist in East Uvean. Nothing can separate the two constituents even though we are not dealing with frozen expressions or compounds. Examples 5 to 9: (5)

Pea nä ma 'u logo leva 'ae ne'e mate tanä tamai. and 3D find listen EMPH DEIC PAST die their father 'Then they learned that their father was dead.' (lit. found listening)

(6)

Talamohe, ko totatou fagona 'e tou 'alu sao 7 pe. Talamohe PRED our trip NS IPINCL go play RESTR 'Talamohe, we are going to go there playing a game.'

We will see in Section 3 that this construction is different from 'to go to play', which requires the conjunction ο between the two predicates.

Complex predicate constructions in East Uvean (7)

(8)

Kae but 'But (Seo Έ NS

205

toe opea taei ai pe te ki'i kumä. again be dragged cry really RESTR DEF small rat the rat, crying all the time, was dragged again by the flow.' p.72)

nä feolo 'aki totolo 3D go hither and thither crawl

i

te

fai

ο

tanä

gäue.

OBL

DEF

do

POSS

their

work

pe RESTR

foki also

lä EMPH

'They both went crawling here and there in order to do their work.' (9)

Ko

te

tama

nei

'e

TOP

DEF

child

DEIC

NS j u m p

ο

te

POSS

DEF

hopo

malimali

i

loto

mala 'e

smile

OBL

inside

yard

faleako. school

'This child jumps smiling in the school yard.' Tense-aspect markers, always preposed to the first predicate 6 , as well as arguments referring to the agent (either clitic or independent) are shared by the two predicates. The argument structure of the sentence is usually determined by the valency of the first predicate. However, this is not the case in the following example, in which the second verb si'aki 'to leave, to abandon' shares one argument (the preposed pronominal argument referring to the agent) with the first predicate, but has its own, Futa'e Pakupaku, in the absolutive case and referring to the patient. Hence, the focused argument täua stands for both the unique (S) argument of the intransitive predicate olo, and for the ergative (A) argument of si'aki: (10)

Io,

ko

täua

yes

PRED

IDINCL EMPH NS

ia

'e

olo

si 'aki pe

go(PL) l e a v e

RESTR

Füta 'e Pakupaku ke nofo ai. Futa'e Pakupaku in order to stay ANAPH 'All right, we will go and leave Futa'e Pakupaku so that she remains here.' Indeed, each verb of the series may have its own argument, as long as one - the agent - is shared. Hence in (11), V 2 malimali shares the agent (te tamaliki) with fa'u; the agent is in the ergative case because of the transitive valency of Vi; but malimali has its own oblique argument te haha 7, 'people' and fa 'u its own absolutive argument te vaka 'the boat': (11)

Έ fa'u malimali ki te haha'i te vaka e te tamaliki. NS build smile OBL DEF people DEF boat ERG DEF children 'The children are building a boat smiling to people.'

206

Claire Moyse-Faurie

However, the scope of the negative marker mole (from the verb mole 'disappear') is semantically restricted to V2 in East Uvean serial constructions, as shown in (12): (12)

Έ mole gao'i malimali toku foha. NS NEG play smile my son 'My son doesn't smile while playing.'

SVCs, as any lexeme, may occur inside nominal phrases. The following example shows how they can occur as heads of nominal phrases (defined here by the article te), keeping their adjunct (directional age), but also determined by a possessive phrase (o tona foha): (13)

Pea

sio

mai pe

and

see

DIR

ο

tona

POSS his

Fakahega

RESTR Fakahega

ki

te

'alu

tagi

OBL DEF go(SG) cry

age DIR

foha. son

'Fakahega saw his son coming in tears.' (lit. Fakahega saw the crying arrival of his son) In this type of structure, where there is no clear specialization of one of the predicates in the series, it should be possible to invert the position of the two predicates, and indeed, it is in some cases: lele kata 'run laughing' or kata lele 'laugh while running' are both accepted, as well as tagi palalau vs palalau tagi 'speak crying' or 'cry speaking'. Actually, however, the order is often fixed, so people will say: 'e moe mimi te toe (NS/sleep/pee/DEF/child/) 'the child pees in his bed', and never * 'e mimi moe te toe; 'e lele tagi 'he runs crying', whereas * 'e tagi lele puzzled my informants. Yet, in all these examples, both events are perfectly simultaneous. V, seems to be preferentially a motion or posture verb, that is a dynamic intransitive verb: lele hiva 'run while singing', lele kai 'run while eating', nofo tagi 'stay crying', nofo moe 'stay sleeping', etc. while V 2 allows more varied semantic fields, often denoting a continuous action. However, I also found serial constructions with two verbs of affect, as in (14) and (15): (14)

Ko

tana ha'u

PRED his

pe

te

'aia ο

ha'u kae tagi hoha 'a

come there CONJ come but

cry

be worried

kiu.

RESTR DEF sea-bird

'On his way, he met a sea-bird who was crying in anxiety.'

Complex predicate constructions in East Uvean (15)

Ne'e

mole

sio

jta

ki

te

PAST

NEG

look at

be angry

OBL

DEF

207

tamasi'i. child

'He didn't look angrily at the child.' Transitive predicates seem to be the least prone to enter SVCs. The fact that we are dealing here with strict serial verb constructions (without specialization) is evidenced by reduplication. As already mentioned, in East Uvean and other Western Polynesian languages, a few intransitive predicates are partially reduplicated when their argument is plural. In the V1V2 structure without specialization, predicates are inflected for plural in the serial construction as they would be individually: (16) a. Έ

matou

lele

kakata.

1PEXCL

run

PL.RED.laugh

NS

'We run laughing.' (16) b. Έ

au

lele

kata.

NS

IS

run

laugh

Ί run laughing.' (17)

Ne'e

matou

nonofo

momoe/kakai

i

PAST

1PEXCL

PL.RED.stay

PL.RED.sleep/PL.RED.eat

OBL

fale. house

'We stayed at home sleeping/having lunch.' Note that nofo, moe, kai and kata are the singular forms; lele does not vary for number. Setting aside the SVCs that have been analysed so far, which express simultaneity or manner, other successions generally imply the specialization of the first or the second predicate of the series. Still, it is often hard to determine the cut-off point between lexemes still to be considered as potential predicates and lexemes that appear as adverbial modifiers in the majority of their occurrences. Here again, we can only rely upon the semantic interpretation, which will tell us if we are dealing with one single action event or with two simultaneous actions belonging to the same event. Indeed, it is not always possible to use SVCs to express simultaneous actions. For instance, sentence (18a) requires a two-clause construction: (18) a.

'Aua

na'a

inu

pe

ke

should not

lest

drink

or

2S

ha'ele! walk

'You shouldn't drink while walking!' (lit. don't drink or you walk) This is because the sequence inu 'to drink' + ha'ele 'to walk' has a different meaning, as shown in (18b) and does not express simultaneity:

208 (18) b.

Claire Moyse-Faurie 'Aua

na'a

inu

should not

lest

drink

ha'ele! walk

'You shouldn't drink in random places!' 7 Very often then, one of the two predicates semantically specializes as an aspect or adverbial modifier, as ha 'ele in (18b).

2.2. V 2 specialization: V 2 tends to be an adverbial modifier Quite often, V 2 specializes as an adjunct even when V 2 can still occur autonomously. This specialization requires no derivation on V 2 . Hence, for example, the predicate lelei means 'be well, be good' in (19a) and 'well, good' as an adjunct in (19b): (19) a. Ko PRED

te

ava

ο

Fatumanini

'ae

'e lelei

age

DEF

channel

POSS

Fatumanini

there

NS

DIR

ai

pe

te

gelu.

ANAPH

RESTR

DEF

fishing

be g o o d

'It is in the Fatumanini channel that fishing is the best.' (19) b. Ne'e PAST

he'eki

'ilo'i

lelei

pe

ko

he

tama

mälama

not yet

know

be g o o d

or

PRED

INDEF

boy

world

pe

ko

he

or

PRED

INDEF

temonio. devil

'It was not yet (well) known if this young man was a human being or a devil.' In examples 20 and 21, vale 'foolish, idiot' and fakaholoholo 'be successive' (formed by the causative prefix faka- + 'to fall in, to slip') tend to give an adverbial (manner) meaning to the V l s although they can both occur as autonomous predicates elsewhere: (20)

Έ tukuna vale te janau a Soane. NS be abandoned be foolish DEF children POSS Soane 'Soane leaves his children without supervision.' (lit. Soane's children are left foolishly)

(21)

Kua

'u'ua

fakaholoholo

leva

te

PERF

sing

be successive

EMPH

DEF CLAS

te

temi

DEF

moment

w

moa

i

rooster

OBL

'aia. DEIC

'At that moment, roosters began to sing successively.' (Seo p. 76)

Complex predicate constructions in East Uvean

209

The fact that predicates cannot be inverted is a sign of asymmetry and, in some cases, of specialization with modifying function. For instance, in sentence (22), ma w8 'to be fixed, to be held' cannot appear in Vi position and generally specializes as an adverb meaning 'firmly': (22)

Pea koutou pikipiki ma 'u kae 'alu te vaka. and 2P cling be fixed but go DEF boat 'Hold on firmly before the boat sails off!'

V2 specialization as an adjunct is also evidenced by the fact that it does not agree in number with the absolutive argument, as shown by the singular form of mate in examples (23a) and (23b): (23) a. Ne'e momoe mate nätou. PAST

PL.RED.sleep

die(SG)

3P

'They went to sleep very soundly.' (23) b. Ne'e

lavevea

mate nätou. wounded die(SG) 3P 'They were wounded to death.' As it specializes, the second verb in the series loses its reduplication (*momoe mamate9, *lavevea mamate) when the argument is plural. In this case, the specialization of V 2 is not only semantic but syntactic as well. The second verb loses its status as a co-predicate, becoming an unvarying postposed adjunct 10 . It also loses the possibility to have its own argument. Specialization sometimes requires quite subtle evidence. Compare the two examples below. In (24a) both predicates reduplicate, showing number agreement with the plural argument nätou; we are thus dealing with a strict serial construction: PL.RED.be

PAST

(24) a.

Έ

natou

nonofo

NS

3P

PL.RED.stay

mäsisiva. PL.RED.poor

'They are living in poverty.' (lit. live being poor) My informant added the following comments: in (24a), all the members of the family are poor, from the grandfather to the grandson. In (24b), masiva does not agree in number, as nonofo does. According to the same informant, the meaning of (24b) is slightly different: poverty does not concern every single member of the family, but the family as a whole. In this case, masiva behaves as an adjunct: (24) b.

Έ

natou

NS 3P

nonofo

masiva.

PL.RED.stay poor(SG)

'They are living in poverty.'

210

Claire Moyse-Faurie

The occurrence as a V2 adjunct sometimes entails a change in the meaning of predicates. I already mentioned the case of ha 'ele which means 'to walk' as a predicate and 'in random places' as an adjunct. Similarly, the ergative verb fakahä 'to show' (causative prefix + hä 'to appear') will mean 'unexpectedly' as a V2: (25)

Oiaue, ne'e ke ha 'u fakahä alas PAST 2S come show 'Alas, you came unexpectedly!'

age! DIR

There is certainly a semantic predisposition that leads some lexemes to specialize as adjuncts in V 2 position. However, the characterization of an adverbial part of speech (lexemes occurring only as adjuncts) is as problematical in East Uvean as the characterization of noun and verb categories. There seem to be a few lexemes that only occur as adjuncts: nonderived terms such as noa 'in vain', täfito 'precisely for', gafua 'easily', gata 'a 'with difficulty' or derived terms, either by reduplication (mämälie 'slowly, softly' < mälie 'be well') or with the prefix faka- 'after the manner o f (fakamanu 'beastly' < manu 'animal'). However, this category is unstable and uncertain. On the other hand, most of the intransitive predicates may be derived by this prefix faka-, which is also the causative prefix, and these derived forms often occur as adjuncts, describing qualities or posture, as fakaholoholo above in example (21), or as fakahagatonu 'straightforward' (from haga 'to face', tonu 'just, exactly') in (26) below, in a rare case of a more than twoterm serial construction, here in a nominal context: (26)

Kua

ina fuafua ko tana maliu lele fakahagatonu. 3S measure PRED his go back run straightforward 'He decides to turn around and run normally.'

PERF

Amongst predicates which tend to often occur in adjunct function (besides faka- derived forms), we find statives such as lelei '(be) good, suitable', mälie '(be) well', lahi '(be) tall', and so on. A peculiar case of specialization is attested with verbs in V 2 position used to express intensity in idiomatic phrases. For example, the verbs 'elo 'to stink', vale 'be foolish' and momoho 'be cooked' are used to express depreciative intensity after some stative verbs: piko 'elo 'very lazy'; kovi 'elo 'very bad'; fula 'elo 'very fearful', vale 'elo 'very stupid'; kula momoho /red/be cooked/ 'crimson (by the sun)'; legalega momoho /yellow/be cooked/ 'very yellow (yellow-orange), very ripe'; kai vale /eat/be foolish/ 'to guzzle'.

Complex predicate constructions in East Uvean

211

2.3. "Deserialized" constructions Although it is the most frequent case of specialization, V 2 as an adjunct is still not very much used. In fact, East Uvean clearly prefers constructions with Vi "nominalization" as in (27b), (28b) and (29b) instead of serial verb constructions (or, more precisely, +adverbial modifier constructions) as in (27a), (28a) and (29a): (27) a.

Έ

ina kai logo 'ä

NS 3S

eat

tana

mo Ί

make noise his

pipi.

CLAS meat

'He eats noisily his piece of meat.' (27) b.

Έ

logo a

tana kai

ο

te

mo 'i

pipi.

NS make noise his eat[ing] POSS DEF CLAS meat id. (lit. his eat[ing] o f a piece o f meat is noisy)

(28) a.

Έ popoto NS

tatau

pe

ia

PL.RED.clever be same RESTR ABS

Soane Soane

mo Petelö. and Petelo

'Petelo and Soane are equally clever.' (28) b.

Έ

tatau

pe

te

poto /popoto

ο

NS be same

RESTR DEF intelligent / PL.RED.intelligent POSS

Soane

mo

Petelö.

Soane

and

Petelo

id. (lit. Soane's and Petelo's intelligences are equivalent) Notice that in (28b) poto may be reduplicated; the non-reduplicated form poto is the nominal or the singular verb form while the reduplicated form popoto is clearly the plural predicative form, but still allowable in a nominal context. There is no significant difference in meaning between the two options. In this type of construction where Vi is nominalized, V2 often belongs to the intransitive class, generally stative predicates describing properties: (29) a. Ne 'e mole PAST

NEG

hiva

kovi

te

sing

bad

DEF young girl

ta 'ahine.

'The young girl did not sing badly.' (29) b. Ne'e

mole

kovi te hiva a te ta'ahine. PAST NEG bad DEF song POSS DEF young girl id. (lit. The young girl's song was not bad) The formal parallelism between the compound constructions described at the very beginning of this paper (example 4a) and the examples of modifying serialization just above - both types of constructions that may be

212

Claire

Moyse-Faurie

decomplexified in a nominalizing processus - is of course striking. However, in modifying serialization, the first element may occur elsewhere as an autonomous predicate, while in part-whole compounds, the predicative use of the first element is the exception. Moreover, there is no semantic restriction in modifying serialization as there is in part-whole compounds.

2.4. V, specialization: V! tends to become an aspect or modal auxiliary Vi specialization only concerns a few lexemes and, perhaps for this reason, it shows a stronger tendency to grammaticalization. V! specialization leads to aspect or modal auxiliaries. As all East Uvean tense-aspect markers but one are preposed to the predicate, it is not surprising that a Vi (rather than a V 2 ) specializes as an aspect or a modal auxiliary. The grammaticalization varies in degrees: kamata 'to begin' and 'amanaki 'to hope' occur as often as aspectual auxiliaries and as predicates; lolotoga 'to last' seldom occurs autonomously, and mä'ua 'to happen' seems completely specialized in an impersonal existential meaning. a) The ergative verb kamata 'to begin', used as an autonomous predicate in (30a) tends to specialize as an inceptive aspect in Vi position, meaning 'begin to' as in (30b) and (30c): (30) a. Pea kamata atu ai leva te tau. and begin DIR ANAPH EMPH DEF war 'Then the war began.' (30) b. Kua

kamata mafoa begin rise 'It begins to dawn.' PERF

(30) c. Ne'e

leva

te

EMPH

DEF

kamata

mataku

PAST

begin

be afraid really 3S and

ai

ia

really 3S

ai

'aho. day

ia pea

ki

te

motu

ko

OBL

DEF

island

PRED

ne'e PAST

fia

hola

feel like run away

Nukufotu. Nukufotu

'He began to be really afraid and he wanted to run away to Nukufotu island.' (Seo p.29) The predicate kamata may be reduplicated with a diminutive meaning. This derivation is also possible in Υ,/auxiliary position, as in (31):

Complex predicate constructions in East Uvean (31)

Kua

kamakamata

PERF b e g i n s l o w l y

hopo

ake.

jump

DIR

ma 'olusa

lä

te

la'a

i

tanä

b e high

EMPH

DEF

sun

OBL

their

213

'The sun barely began to rise when both appeared.' Like any other verbal series, kamata + V2 may be found in nominal contexts, preceded in (32) by the singular specific article: (32) Ko'e pe te kamata hivahiva atu a Hina... there

RESTR

DEF begin

sing slowly DIR POSS

Hina

'Now Hina begins to sing s l o w l y . . ( l i t . here is the beginning singing slowly of Hina), (Seo, p. 38) Here too, nominalized constructions are often preferred to complex predicates, but it is now the V2 of the SVC which becomes the head of the absolutive argument. The focus varies depending on the construction: while it is the action which is focused in the nominalized construction (33b), the focus is on the agent in the complex predicate construction (33a): (33) a.

Έ

kamata

NS

begin

'My mother

(33) b.

lalaga

te

fala

weave

DEF m a t

e

taku fa 'e.

ERG

my

mother

starts w e a v i n g a/the m a t . '

Έ kamata te lälaga ο te fala e taku fa 'e. NS begin DEF weaving POSS DEF mat ERG my mother 'My mother starts weaving the mat.' (lit. The weaving of the mat by my mother starts)

The scope of the negation is on the whole complex verb: (34)

Ne'e mole kamata fa'u e Soane tona PAST NEG begin build ERG Soane his 'Soane has not begun to build his house.'

fale. house

b) Used as an autonomous predicate as in (35a), 'amanaki means 'to hope, to expect'; in a serial construction (35b et 35c), it takes on an aspect meaning 'be about to': (35) a.

Έ

'amanaki

e

te

NS

hope

ERG

DEF p e o p l e

haha 'i te/ke DEF/that

'People are hoping it rains.' (35) b.

Έ 'amanaki mavae te sitima. NS expect leave DEF steamer 'The steamer is about to leave the harbour.'

'ua. rain

214

Claire Moyse-Faurie

(35) c. Tuna,

'e iai

eel

te

me'a

kovi

'e

'amanaki hoko

NS exist DEF thing

bad

NS expect

kiä

koe.

happen OBL 2S

'Eel, something bad is going to happen to you.' (Seo, p.32) c) The grammaticalization is almost complete with lolotoga 'to last' which is rarely used as an autonomous predicate, only to describe a continuous event that serves as background for another event, as in (36a): (36) a. Έ lolotoga te misa kae au 'alu au ki te falekoloa. NS last DEF mass but IS go IS OBL DEF store Ί am going to the store during mass.' (lit. mass is going on but I go to the store) This lexeme tends to grammaticalize as a progressive aspect marker: (36) b.

Έ

matou

NS

IPEXCL last

lolotoga lalasa te weave

fala

DEF mat

ο

tomatou

POSS our

falefono. meeting

house 'We are weaving the mat for our meeting house.' d) Slightly different is the grammaticalization of mä'ua 'happen, turn out' (derived from the ergative verb ma'u 'find, obtain, catch') which now has an impersonal existential meaning and only occurs as Vi, in front of posture verbs: (37)

Kae ne'e mä'ua nofo ai te matu'a i Vaimalau nei. but PAST happen stay really DEF old man OBL Vaimalau here 'But the old man turned out to stay in Vaimalau.'

(38)

Ne'e

kua

au mä'ua

tu 'u pe

he

kua

au

PAST

PERF

IS

stand

because

PERF

IS

mälie'ia

tau

appreciate

your

happen

RESTR

gao'i. play

'It so happened that I stopped because I appreciate your play.'

2.5. The same verb specializes either in Vi or in V 2 In fact it adverbial when the with 'osi present.

seems that V t or V 2 position is crucial in assigning an aspect or meaning to the components of the series. This is clearly seen words are precoded with several semantic values, as it is the case 'to be finished' and vave 'to be quick', which I am now going to

Complex predicate constructions in East Uvean

215

The aspectual V] position and the adverbial V 2 position seem to be entirely pre-programmed as is shown by the following examples including one and the same lexeme which is aspectual in V|, whereas it behaves as a manner adjunct in V 2 . And of course, these lexemes can still occur as autonomous predicates. a) vave as a predicate means 'to be quick' (39a); as a Vi it means 'soon' (39b) and as a V 2 , it means 'quickly' as in (39c): (39) a.

Έ

vave te gäue ko tona kaugamälie. be quick DEF work PRED its good participation 'The work is quickly done because there are many people.'

NS

(39) b. Kua

vave

'alu ia

PERF be quick

go

Petelö.

ABS Petelo

'Petelo is going to leave soon.' (39) c.

Έ

lalaga

vave

e

NS

weave

be quick

ERG DEF woman her

te

fafine

tona

fala. mat

'The woman is weaving her mat quickly.' b) As an autonomous predicate 'osi means 'be finished'; it means 'already' in Vi position and 'quite, definitely' in V 2 position. - 'osi is the sole predicate: (40) a. Ne 'e PAST

tona his

'osi be finished

pe

ni

ki 'i

'aho si 'i

ki

RESTR

INDEF.PL

small

day

OBL

EMOT

fänau'i... give birth

Ά few days after the birth, ...' (lit. a few days are finished after his birth) (Seo, n.d.: 37) (40) b. Kua

teitei

'osi

pe

te

gäue

PERF almost be finished RESTR DEF work

ο POSS

ο

te

'ekelesia

POSS DEF church

Vaitupu. Vaitupu

'The works in Vaitupu church are almost finished.' -

in Vi position 'osi specializes as an aspectual marker, here in front of kamata 'to begin':

(41) a. Kua

'osi

kamata

PERF be finished begin

te

me'e.

DEF dance

'The dance has already begun.'

216

Claire Moyse-Faurie

The pronominal argument is usually anteposed to 'osi as in (41b), but it can also separate 'osi from the verbal predicate, as in (41c). (41) b. Kua ke 'osi lau te tohi 'ael PERF 2S be finished read DEF book DEIC 'Have you already read this book?' (41) c. Kua

'osi

matou

tahi

inu

kafe

i

tona

'apiu.

be finished I P E X C L each drink coffee OBL his house 'Each of us already had coffee at home.' However, these constructions are seldom used and 'osi occurs more often as a predicate followed by a prepositional phrase as in (42b): (42) a. Kua au 'osi mamata te fenua. PERF IS be finished visit DEF country Ί have already visited the country.' PERF

(42) b. Kua

'osi

te

fenua

i

taku

mamata.

PERF be finished DEF country OBL my visit id. (lit. the country is finished concerning m y visit)

-

in V 2 position 'osi specializes as an adjunct with an adverbial value, meaning 'completely': (43) a. Kua fetogi 'osi tana aga talu te 'alu ο PERF change

be finished his

way

since

DEF go

POSS

Petelö ki te Use. Petelo OBL DEF high school 'Petelo's behaviour has completely changed since he started high school.' (43) b.

Έ

vevela

NS

very hot be finished DEF day

'osi

te

'aho

nei. DEIC

'It is very very hot today.' (compared to kua 'osi vevela 'it is already hot') 'Osi comes last when several adjuncts are expressed: (43) c. Kua ma 'a lelei 'osi taku PERF clean well be finished my Ί completely cleaned up my car.'

mötokä. car

Complex predicate constructions in East Uvean 3.

217

Strong coordination

Let us now turn to the second type of structure in which the predicate phrase comprises two predicates, connected by the phrasal coordinator mo 'and, with', and referring to two quasi simultaneous events or properties; most often, both predicates are separated from each other only by mo; however, it is also possible for Vi to have its own object (generally incorporated as in example (44), or its own adjuncts. (44)

Ne'e

'alu ia

Talapili

ο

tä

PAST

go

Talapili

CONJ

cut leaf pannel and string

ABS

kauafo

mo

tui

lau. leaf

'Talapili went to cut leaves and string them (for the roof).' In contrast, Vi and V2 share the same agent or experiencer argument whatever its position. They also always share the same tense-aspect marker and the negative marker, which are positioned before Vi. The topicalized argument is shared and occurs before the complex predicate as in (45): (45)

Ko

te

TOP

DEF

ta 'ahine 'aia, ne 'e pakoko lua mo matakovi. young girl DEic PAST skinny very and bad looking 'This young girl, she was very skinny and looked bad.' (Seo p. 37)

or after Vi, either in the ergative case as in (46) or in the absolutive as in (47): (46)

Ne 'e tali PAST

answer

ο

te

POSS

DEF

leva

e

Lupelutu mo ßa

EMPH

ERG

Lupelutu and feel like laugh be what news

kata: koteä

logo

fenua? country

'Lupelutu felt like laughing and answered: what news from home?' (Seo, n.d.: 68) (47)

Pea

ioio

pe

te

ki'i

kumä

mo

heka

ki

and

agree mildly

RESTR

DEF

small

rat

and

sit

OBL

te

tu 'a ο

DEF b a c k

POSS

te

tafola 'a.

DEF

whale

'The rat agreed and sat down on the whale's back.' (Seo p. 72) It may also occur after V 2 : (48)

Έ me'e lelei mo kata le'o lahi NS dance be well and laugh voice big 'The boy dances well and laughs loudly.'

te tama. DEF boy

218

Claire Moyse-Faurie

When there is plural agreement, it occurs on both coordinated verbs as in strict serial verb constructions: (49)

Ne'e

ina tuku

ai

tona kupega,

mo'o

'ä

PAST

3S

leave

ANAPH

his

for

enclose

ο

te

'ü

POSS

DEF

ki 'i

motu

net

na 'a mavetevete

small island lest

CLAS

takatakai mo

surround

täfefea.

PL.RED.scatter and PL.RED.drift

'He [Tagaloa] left his net there as a barrier for the islands so that the islands would not scatter and drift.' The semantic difference between coordinate predicates and strict serial verb constructions is small. SVCs make up one single event including two facets which occur simultaneously, while coordinate predicates consist of two quasi simultaneous events. Syntactically however, the difference is more consistent since the scope of negation is not the same. Although the negative marker is always preposed to the complex predicate, negation applies to both coordinate predicates (and not only to V 2 as in SVC constructions, as in (12)): (50)

Έ

mole

NS NEG

lele

mo

kata.

run

and

laugh

'He doesn't run nor laugh.' (Compare with: Έ mole lele kata. 'He doesn't laugh while running.') (51)

Έ

mole

NS NEG

kai mo

inu

ia

eat

drink

ABS Petelo

and

Petelö

he

'e

mahaki.

because

NS

sick

'Petelo doesn't eat nor drink because he is sick.' Moreover, exceptionally, each coordinated element may have its own agent argument, as in (52), where the whole complex occurs as the absolutive argument of the predicate sio 7: (52)

Sio Ί e

nätou

mole

tafea

ai

he

lau

togo

see

ERG

3PL

NEG

drift

ANAPH

INDEF

leaf

mangrove

ia

pe

EMPH

RESTR

mo hopo ai he kanahe pea fakalaka. and jump ANAPH INDEF mullet and go through 'They saw neither the mangrove leaves floating nor the mullets jumping, but kept on their way.'

4.

Conjunctive or purposive complex predication

Let us now consider the last structure mentioned at the beginning, that is conjunctive complex predication. At first sight, the sequential conjunctive ο

Complex predicate constructions in East Uvean

219

structure has nothing to do with svcs nor with predicate coordinate constructions except that it is the most common and frequent way of associating two predicates and, moreover, it corresponds semantically to nuclear juncture such as described in Pileni [Naess, this volume], another Polynesian language (hano moe 'go to sleep'). Like the already described constructions (strict SVCs and coordinate predicates), the sequential structure with ο implies sharing an argument, as well as double number agreement: (53)

Vakai mai hau tahi ia fa 'a uku mo taku tokotolu beware DIR your one OBL a lot dive with my CLAS+three 'aerii 'e οίο ο nonofo i lalo. DEIC N S g o ( P L ) CONJ PL.RED.Stay OBL bottom 'Choose somebody who is a good diver to compete with my three men [as] they will go and stay underwater.' This construction very often consists of an intransitive predicate of motion or posture in Vi, associated with any active predicate in V2, but there can also be a transitive predicate in Vi. In the latter case, the shared subject, ina in (54), is the ergative argument: (54)

Ne'e

ina to 'o te toki ο tu 'usi 'aki te fu'u niu. 3S take DEF axe CONJ cut with DEF CLAS coconut tree 'He took the axe and cut the coconut tree with it.' In this type of construction, haga 'to face, to turn to' occurs very often in Vi position. It then has the aspectual meaning 'to be at', as in (55a) even combined with kamata 'to begin' in (55b): PAST

(55) a. Kua PERF

haga ia ο kai te mo'i pane i tana to face 3S CONJ eat DEF CLAS bread OBL his

kua PERF

fia kai want eat 'He began eating the bread because he was very hungry.' (55) b. kua

hasa ia to face 3S i tana kua OBL his PERF 'He began to bite PERF

ο

kamata kai te mo'i pane begin eat DEF CLAS bread fia kai. want eat at the bread because he was very hungry.' CONJ

The negative marker may occur either before haga, and then the aspectual meaning is maintained:

220

Claire Moyse-Faurie

(56) a.

haga ia Paulo ο fa'u te fale. NS NEG to face ABS Paulo CONJ build DEF house 'Paulo is not starting to build the house.' or before V2, and in this case, haga keeps its basic non-aspectual value 'to face, to turn':

(56) b.

Έ mole

Έ

haga ia Paulo ο mole fa 'u te fale. to face ABS Paulo CONJ NEG build DEF house 'Paulo is turning [back] in order not [to have] to build the house.' On the other hand, we may wonder whether this type of conjunctive structure could be at the origin of some cases of serialization - nuclear juncture as in Pileni - after dropping of the conjunction o, as occasionally happens in fast speech after predicates of motion, as in (57) and (58): NS

(57)

Pea and

tali mai answer DIR

e Havea ERG Havea

'e ha 'u

(o)

tau!

NS

(CONJ)

fight

come

he 'e ä? ko'erii because NS what here is

näua 3D

'Havea answered: why is this? Here they come fighting (they come and fight)!' (58)

Pea

fakatu 'u

(o)

lele

leva

te

tala

mo

te

and

to stand u p

(CONJ)

fly

EMPH

DEF

sea-gull

and

DEF

veka

ki

'uta.

Rallus sp. OBL bush

'Then the sea-gull and the banded rail were about to fly off towards the bush.' (Seo p. 72) However, this serialized construction is semantically quite different from what we have seen in 2.1, as it usually still describes two successive or purposive actions rather than two simultaneous actions of a single event.

5.

Conclusion

East Uvean shows a morphosyntactic continuum of complex predicates, with the following semantic correlates: Strict Vi + V 2 one event with simultaneous actions (rare), evolving towards V + Adverbial modifier or Modal/Aspectual + V Vi mo V 2 two quasi simultaneous events or properties.

Complex predicate constructions in East Uvean

221

V, ο V2

a) one event with two successive actions. b) one action + purpose, intentionality, inceptive value; very frequent with a posture (haga 'to face' > 'to be at') or motion predicate ('alu 'to go', ha 'u 'to come') as V\. What can we retain as being true SVCs? If we only define SVCs as the use of two or more verbs to express a semantically-complex concept, then East Uvean is not a very good example of a serial verb language, since it prefers to deal with one predicate, showing different strategies to avoid serialization: -

SVCs may be replaced by constructions in which Vi or V2 is nominalized.

SVCs evolve into aspectual or adjunct specialization of one of the predicates, so we may wonder if we are dealing with modifying serialization, or grammaticalization, or lexicalization. Finally, structures requiring a coordinating or subordinating morpheme are much more frequently used than strict SVCs. In V1V2 structure, we are dealing with one event with simultaneous actions (or one action with different facets, depending on the terminology we adopt). In V, ο V2, the event is also unique though complex, with two successive actions, except when the first verb indicates purpose in relation to the second verb, rather than an action in itself. On the contrary, coordination with mo refers to two different events, although quasi simultaneous ones.

222

Claire

Moyse-Faurie

Orthographic conventions and abbreviations As for most Polynesian languages: the apostrophe indicates the glottal stop, the macron indicates vowel length and the letter 'g' stands for the nasal velar consonant. ABS INDEF singular indefinite/non-specific absolutive case marker article ANAPH anaphoric NEG negative marker classifier CLAS NS non-specific tense-aspect marker CONJ conjunction OBL oblique case marker D dual PERF perfective aspect DEF singular definite/ specific article P, PL plural DEIC deictic PL.RED plural reduplication DIR directional POSS possessive marker EMOT emotional article PRED predicative marker EMPH emphatic particle RESTR restrictive particle ERG ergative case marker S, SG singular EXCL exclusive TOP topic marker INCL inclusive

Notes 1.

2.

3.

East Uvean is spoken by approximately 10,000 people in Wallis/East Uvea, to which must be added an equivalent number of East Uvean emigrant speakers in New Caledonia. East Uveans call their language Faka'uvea. All data given in this paper come from my own fieldwork, except when indicated as Seo, extracted from Seo (n.d.). Reduplication is a ubiquitous synchronic process in the Polynesian languages; it also evidences an historical process, as the semantically related unreduplicated root is not always attested in contemporary East Uvean, as for example lalaga 'plait', mamata 'visit', lelei 'good', malimali 'smile'. Total reduplication has only a semantic diminutive value (io 'agree', ioio 'agree mildly') whereas partial reduplication marks number agreement in verbs that allow such marking (nofo 'stay' (SG), nonofo 'stay (PL)'), or has a semantic intensive value (vela 'hot', vevela 'very hot'). In the word for word translation, I only indicated plural reduplication, which is the only relevant one for the topic. The meaning of lexemes referring primarily to entities changes with the syntactic context. For example, papa means 'flat stone, board' in a nominal context, but 'stony' as a predicate; hina 'bottle' means 'use as a bottle, share a

Complex predicate constructions

in East Uvean

223

bottle' as a predicate; the place name 'Uvea means 'originate from Uvea' as a predicate, and so on. 4. The pronominal subject of the second predicate may also be non-coreferential with the subject of the first predicate, but in this case, we are dealing with two different clauses and not anymore with complex predicates. 5. The same observation has been noted in other Polynesian languages. See for example Besnier's Tuvaluan grammar (2000:538) in which he states: "The existence of verb serialization, or verb chains, is well attested among Oceanic languages other than members of the Polynesian family' [my italics], 6. The exception is ana/, the future tense-aspect marker which is postposed to the verb phrase and which is the grammaticalized form of the temporal adverb 'änai 'soon'. 7. However, new ways of life may restore the SVC interpretation. Let us reconsider the two lexemes kai 'eat' et ha 'ele 'walk': - in a strict SVC, each predicate keeps its basic meaning: kai ha 'ele 'eat while walking', as in a self-service, recently introduced. The succession ha 'ele kai is also possible (but less elegant), with the same meaning; this interpretation is recent, although it may have been the first meaning of this series, before the evolution towards modifying series; - in a modifying s v c (the more usual interpretation), with specialization of V 2 , kai ha 'ele means 'eat anywhere'. 8. There are two homophonous lexemes ma 'u in East Uvean: one is an ergative verb and means 'have, obtain, find', and can appear as V, as in example (5); the other is an intransitive verb and means 'be fixed, be held', has to be derived (fakama 'u 'fix, hold') in order to become ergative and tends as V 2 to grammaticalize as an adjunct, meaning 'firmly' as in (22). 9. *mate moe or *mamate momoe are not accepted; 'die while sleeping' will be expressed either by two clauses: ne'e mate kae moe /PAST/die/but/sleep/ 'he died while sleeping' or by a single predicate followed by a prepositional phrase: ne'e mate i tana moe /PAST/die/oBL/his/sleep/ lit. 'he died in his sleep'. 10· V 2 specialization may lead to compounds, as for example the ergative verb tämate'i 'kill', literally 'hit dead', compounded of tä 'hit' and mate 'die', plus the transitivizing suffix - 7. 11. Other word orders are also accepted, with 'osi preceding immediately the predicate: kua matou tahi 'osi inu kafe i tona 'api or kua tahi 'osi inu kafe mätou i tona 'api.

Serial verbs and complex constructions in Pileni1 Äshild Nasss

Abstract The Polynesian Outlier Pileni presents a number of challenges with respect to identifying possible serial verb constructions. Since argument ellipsis is common, and since temporal sequences of events are frequently expressed by juxtaposition of clauses or verb phrases without any intervening conjunctions, it is difficult to decide when a construction might be analysable as an SVC and when it should rather be considered a sequence of clauses with some or all arguments omitted. Core-layer SVCs are particularly difficult to identify, for the reasons outlined above. There are a number of constructions which resemble core-layer SVCs, but which on the basis of currently available evidence are impossible to conclusively distinguish from other types of complex clauses or clause sequences. Some constructions do occur, however, where it is possible unambiguously to ascribe a meaning or function to the construction which would not arise with simple juxtapositions of clauses. Pileni appears to show "ambient serialization" (Crowley 1987) as well as two types of SVCs with toa "take" as V,; one where toa functions to "introduce an object noun" (Lord 1993), and one where it contributes an inceptive meaning to the clause. The formal difference between these two construction corresponds to the distinction between "contiguous" and "noncontiguous" serialization; in the first instance a shared object will occur between the two verbs of the SVC while in the latter it will follow both verbs. Pileni SVCs illustrate the importance of distinguishing between core-layer and nuclear-layer SVCs. Many accounts take as a criterion for serialization that the verbs in the construction should show the same tense and aspect. There is a difference between these two features, however, in that aspect is a nuclear-layer operator while tense is a peripheral one; and so, in Pileni, core-layer SVC must have the same temporal frame but may differ in aspect, while nuclear-layer SVCs must share both tense and aspect.

1.

Introduction

Pileni is a Polynesian Outlier spoken by roughly 1,500 people in Temotu province, the easternmost province of the Solomon Islands. The area is a fairly isolated one located at a considerable distance from the capital and the major islands of the Solomons, and within Temotu province, the Pileni

226

Äshild Ncess

speakers again reside mainly on small, remote and inaccessible islands. The language is spoken on the islands of Pileni, Nifiloli, Matema, Nukapu and Nupani in the Reef Islands group as well as in a couple of fairly recent settlements on the rather larger island of Santa Cruz, and also on the Duff Islands some distance away from the Reefs 2 . On the Reefs and Santa Cruz live speakers of the so-called Reefs-Santa Cruz languages, with whom the Pileni speakers have had regular contact for centuries for purposes of trade and intermarriage. Of the Reefs-Santa Cruz languages, little is known other than that they are probably not Austronesian; they are generally classified as being of Papuan origin. In addition to this long-standing language contact, the use of the heavily English-based Solomon Islands Pidgin is increasing on the islands. Pileni has basic SVO word order and a mostly accusative alignment of syntactic processes. There is no morphological case-marking, and word order is the chief means of marking grammatical relations. However, clauses with two full NP arguments are relatively rare, since any argument which is retrievable from context may be omitted. This raises special problems with respect to serial verb constructions, since it may be difficult to distinguish genuine SVCs from strings of clauses with all arguments omitted. In addition to full NPs, arguments of predicates may be reflected in two ways in a Pileni clause. Firstly, transitive verbs with a 3rd person singular subject obligatorily take a suffix -i on the tense-aspect marker. The suffix, which is glossed 3SG in the examples, never appears on intransitive verbs, nor with any arguments in persons or numbers other than 3rd person singular. The 3rd singular suffix must be distinguished from the set of clitic pronouns which frequently appear prefixed to the tense-aspect markers. These clitics may appear with both transitive and intransitive verbs; but unlike the -i suffix, they are never obligatory with either. As is the case with most Polynesian languages, a clear-cut distinction between nouns and verbs is not easy to make. Many lexemes may occur both in nominal and verbal phrases, and what appear to be verbal phrases may occur in nominal constructions with a possessive (te noho huarahi a tatou 'our sitting together'). However, while such alternations are certainly not uncommon, most lexemes appear to have a predominant use either as a noun or as a verb. For some lexemes, such a predominant use may be difficult to establish, but a large group of lexical items shows a clear tendency to occur mainly as

Serial verbs and complex constructions in Pileni

227

nouns or mainly as verbs. In the following, I will assume that lexemes can indeed be classified as "mainly nominal" or "mainly verbal". I will refer to a lexical item with predominantly verbal semantics and distribution (occurring with tense-aspect and modal markers, transitive suffixes, directional particles, etc.) as a verb, and one with predominantly nominal semantics and distribution (occurring with prepositions, determiners etc.) as a noun. This paper builds on data collected during two brief field trips to Pileni island in 1997 and 1998. The bulk of the material consists of traditional narratives tape-recorded and transcribed with the aid of native speakers (see Hovdhaugen et al. 2002). While the material is fairly rich, there are clear restrictions inherent in the lack of genre variation and in the fact that I have not had access to native speakers while examining the topic of serial verbs. Although the account presented below is supported by the available data, the full picture is therefore likely to be more complex.

2.

Serial verbs vs complex clauses

2.1. Complex clauses in Pileni The problems of defining and identifying serial verbs are discussed in detail elsewhere in this volume. Much of the difficulty lies in distinguishing serial verb constructions (SVCs) from other complex constructions in a given language. To facilitate this enterprise, I will begin with a brief discussion of various types of complex clauses in Pileni. The contrast between SVCs and other types of complex constructions supposedly lies in the absence of co- or subordinating morphology in the former; in the words of Foley and Olson, serial verbs are "constructions in which verbs sharing a common actor or object are merely juxtaposed, with no intervening conjunctions" (Foley and Olson 1985: 18). Pileni has a number of conjunctions that function to conjoin verb phrases or clauses. For instance, ο conjoins verb phrases and expresses a relation of temporal sequence and/or purpose, while a conjoins clauses that describe sequences of events: (l)a.

Ko

ai

ka

kave-a

ο

saki

na

ivi?

TOP

who

TA

carry-TR

CONJ

throw

3SG.POSS

bone

'Who will take its bones and throw them away?'

228

As hi Id Ncess

b.

Ko-i

taka-hia

TA-3SG

stamp-TR

te

na

a

ko

ART

lau-plaka leaf-giant.taro apulu-oho loa

DEM

CONJ

TA

fotu, a ko i moana. hole CONJ ΤΑ sink-DIR EMPH LOC sea '(The pig) stamped on the taro leaf and it broke and sank down into the sea.' However, one of the most frequent ways of expressing a temporal sequence of events is by a simple linear sequence of verb phrases or clauses, with no intervening conjunctions. In addition to this, the overt expression of arguments as NPs is not required in Pileni as long as their referents are retrievable from context; and so one gets long strings of verb phrases which "share a common actor or object" and have "no intervening conjunctions": (2)

Ioko te

magki

no noho i

CONJ ART m o n k e y TA

Ne

lele-ho

TA

Jump-DIR DEM

na

live

ko-i

luga ο

LOC t o p

vakapua

TA-3SG g r a b

te

lakau.

POSS ART

tree

te

lole

i

te

lima

ART

candy

LOC

ART

hand

ο

te

lepü

ko lele-ake

ma

ia

ko

POSS

ART

rat

TA j u m p - D I R

with

3SG

TA

vetheki. run

'And there was a monkey living at the top of the tree. (He) jumped down and grabbed the candy from the rat's hand, (he) jumped up with it and ran away.' Subordinating morphology is rather rarer in Pileni than in, for instance, English. The language has a few lexemes that introduce subordinate adverbial clauses, such as pelä ' i f , ndekina 'because', and katoma 'although'. The complementizer po introduces complement clauses of a number of verbs such as tukua 'say', iloa 'know', kutea 'see' and others: (3)

Ko

ko

kute-oho

po

ko

a

hina

lha-no

2SG

TA

see-DIR

COMP

TOP

ART

who

3DU-TA

vethaki

ma

nohine-au

na.

fight

CONJ

wife-2SG.POSS

DEM

'You will see who is fighting with your wife.' Po is also frequently used to introduce purpose clauses: (4)

Lhat-ko

ua-kina

po

lhatu

3PL-TA

paddle-TR

COMP

3PL

'They paddled him off to bury him.'

thamu-nia. bury-TR

Serial verbs and complex constructions in Pileni

229

But even the verbs which usually take po with complement clauses sometimes occur without it, giving rise to what might be considered either a complement clause without any subordination marking or, alternatively, a relative clause: (5)

Lharou

ko

kute-age

te

3PL

TA

see-DLR

ART hermit.crab

i LOC

uga

no

tholo

mai

TA

crawl

DIR

haupe. beach

'They saw the hermit crab crawling towards them on the beach.' If (5) is interpreted as containing a relative clause, a more appropriate translation might be 'they saw the hermit crab which was crawling towards them on the beach'. There is no formal relativization marker, and only relativized prepositional phrases give rise to a proform in the relative clause, so most relative clauses are formally indistinguishable from simple main clauses: (6)

U-e

tulia

loa

tuku

gakau

po

ku

LSG-TA

like

EMPH

LSG.POSS

intestines

COMP

LSG.MOD

kute-a

te

tai

ne-ί

te-ia

a

see-TR

ART

man

TA-3SG

kill-TR

ART

nohine-aku. wife-LSG.POSS

Ί would very much like to see the man who killed my wife.' The NP te tai is at the same time the object argument of kutea 'see' and the subject of teia 'kill'. These verbs, then, appear to be sharing an argument, with no subordinating morphology. There are no formal distinctions between (6) and, on the one hand, simple sequences of clauses with coreferent arguments omitted, and on the other, some of the structures that are discussed below as possible SVCs. The only criterion that can be applied to distinguish (6) from these other clause-types is a functional one: the VP nei teia a nohineaku functions as a restricting clause modifying the nominal head te tai, and the head plus the restricting clause together function as a nominal argument of the main clause with the verb kutea (see Keenan 1985). For this particular instance, there is also a difference in mood marking between the two VPs involved, an argument against analysing it as an s v c (see 4.4. below); but if both VPs carry the same tense-aspect or modal marker, which is frequently the case, there simply is no formal criterion that separates putative relative clauses from the possible SVCs discussed in 4 below. Nor is there any formal argument for why (6) should not be analysed as a sequence of simple clauses, with a meaning like Ί would very much like to see the man, he killed my wife'.

230

Äshild Ν CESS

2.2. Criteria for serialization We see, then, that one existing definition of serial verbs, that of verbs being "juxtaposed with no intervening conjunctions", is inadequate for Pileni, since conjunctions are often omitted between clauses that denote sequential events, and since such juxtaposition may arise in constructions which are not normally considered as SVCs, such as relative clauses. Another definition found in the literature is that of Sebba (1987), who defines an SVC as "a surface string of verbs or verb-phrase items which occur within what appears to be a single clause". As far as Pileni is concerned, the problem with this definition is that when arguments may be freely omitted, it is not always a simple matter to determine whether a string of verb phrases constitute components of a single clause (and so might qualify as an SVC) or a sequence of distinct clauses. Foley and Olson (1985: 47) define a clause as "a grammatical structure consisting of one and only one periphery". This definition avoids the question of whether core arguments are overtly present or not, but instead depends on the presence of overt peripheral material for the identification of clauses - see 4.1. below. Faced with the difficulty of coming up with reliable formal criteria for SVCs, semantic ones are often invoked. Most discussions of SVCs appeal to the notion of event, it is said that the verbs in an SVC refer to subparts of an event (Crowley 1987; Lord 1993), or that SVCs refer to what is conceptualized as a single event (Durie 1997). The problem is, of course, that defining what constitutes a "single event" may turn out to be at least as difficult as defining SVCs (see also Senft, this volume). This paper will not employ the notion of event as a defining criterion for SVCs, but will occasionally make use of the term in comparing putative SVCs with other types of constructions; it will be claimed that SVCs present clusters of actions as having a greater degree of internal cohesion than other kinds of complex constructions (3.2.). The distinction between nuclear-layer and core-layer serialization (Foley & Van Valin 1984) will be relevant in what follows. Foley and Van Valin assume a clause structure consisting of three "layers" - the nucleus, consisting of the predicate, the core, which includes the valence-bound arguments of the predicate, and the periphery, which "contains arguments expressing the spatio-temporal setting of the event, as well as the secondary participants in the event, e.g. beneficiaries" (Foley & Van Valin 1984: 77). For each layer there is a set of grammatical operators which have scope

Serial verbs and complex constructions

in Pileni

231

over the layer in question. For instance, aspect is a nuclear-layer operator, modality a core-layer operator, while tense is a peripheral one (Foley & Van Valin 1984: 208ff). Serialization may take place either on the level of the nucleus or of the core. Two predicates may be joined together to form a complex nucleus (nuclear-layer serialization), or two cores, each with its own nucleus and core arguments, may be joined together to form a larger, complex core. Crucially, both verbs in an SVC must share all operators of the layer in which they are joined, and of higher-level layers. The discussion in this paper assumes the following criteria for identifying possible instances of SVCs: 1. The verbs share one or more core arguments, not only semantically, but also syntactically. That is, the shared argument is expressed only once (if at all) within the SVC; constructions with coreferential but separately expressed arguments will not be counted as SVCs, but rather as sequences of clauses. 2. Both verbs in the construction (I have no clear examples of SVCs involving more than two verbs) also occur as independent verbs, that is, they occur outside of serialization constructions. This criterion distinguishes SVCs from constructions where verbs are modified by particles which do not occur independently, or do not function as verbs when occurring independently. 3. The verbs/verb phrases in an SVC share the relevant operators for the clause layer in question, for instance, aspect and negation for nuclear-layer SVCs, tense and mood for core-layer SVCs (Foley & Olson 1985: 33ff) 4. If the shared argument(s) is/are expressed separately for each verb in the construction, then the construction is either ungrammatical, unidiomatic, or takes on a different meaning. The last criterion would appear to provide a useful test for SVCs. Unfortunately, its practical application is somewhat limited in the current context, since I did not at the time of writing have access to a native speaker who could provide this kind of judgement. In some cases, it is clear from the available material that alternative constructions with separately expressed arguments would in fact be ungrammatical or express a different meaning; but in others, the question will have to remain unresolved.

232 3.

Äshild Ν cess Nuclear-layer serialization

3.1. Nuclear-layer serialization and verb-medial languages Nuclear-layer serialization creates a complex nuclear unit, and all arguments are arguments of this complex unit (Foley & Van Valin 1984). Foley and Olson (1985) and Crowley (1987) assume that nuclear-layer serialization is mainly a feature of verb-final languages, being strongly disfavoured by SVO languages. This is explained by assuming the following basic serializing structures: "a. verb medial: man take book go; b. verb final: man book take go" (Foley & Olson 1985: 45), where the shared argument intervenes between the two verbs in the verb-medial languages and "prevents the erasing of the individual core layer boundaries and the merger of the two nuclei into one complex nucleus" (Foley & Olson 1985: 27). Foley and Olson assume that verb-medial languages which do have nuclear-layer serialization must at some previous stage have been verb-final. While the fairly rigid SVO word order found in Pileni is rather unusual for a Polynesian language, it is highly unlikely that Pileni has ever been verb-final, since a characteristic of Polynesian languages is that they tend to be verb-initial. The presence of nuclear-layer serialization in verb-medial Pileni, highly typologically unlikely according to Foley and Olson, therefore cannot be explained away by appeal to diachronic facts. On the other hand, the analysis in Foley and Olson seems to assume an ordering of the verbs in the SVC which does not apply to Pileni nuclearlayer serialization. They write: "Any nuclear juncture consists of two or more verbal slots. One of these slots, the open slot, may be filled by a large number of verbs drawn from a wide variety of semantic classes. The other slot, the restricted slot, is much more selective. Only certain verbs or classes of verbs are allowed in this position. (...) In general, all open slots precede all restricted slots in linear order." (Foley and Olson 1985: 40, my emphasis). While a Pileni nuclear-layer SVC does consist of one "restricted slot", where only hano 'go' and a few other intransitive verbs of motion are found, and an "open slot" allowing a wider variety of verbs, they are not ordered as described in Foley and Olson; it is the V, which represents the restricted slot and the V2 the open slot. Therefore, the construction 'take X go' on which Foley and Olson base their analysis does not exist in Pileni; rather, we get an SVC of the type 'go take X'. This type of nuclear-layer serialization is much more consistent with the syntax of SVO languages, and

Serial verbs and complex constructions in Pileni

233

is not at all uncommon; according to Durie "every serializing language I have encountered includes a category of motion serialization, where a verb of motion is combined with some other verb in such a way that the motion verb comes first and the moving argument is the Agent of the second verb" (Durie 1997: 310).

3.2. Nuclear-layer SVCs in Pileni Nuclear-layer SVCs in Pileni, then, consist of a sequence of two verbs, with only one marker of tense-aspect preceding the whole sequence: (7)

A

hiai

ne

hano

lavaki,

na

e

CONJ

NEG

TA

go

disappear

3SG

ΤΑ

thaili-age. wait-DiR

'(He) did not go away, he waited for him.' Both hano 'go' and lavaki 'disappear' are lexical verbs which occur independently as heads of verb phrases. This distinguishes example (7) from (8) below, where ke 'away' is a directional particle which does not occur independently: (8)

Tu-tetuä ha-na po ke fano 2PL-chase woman-DEM COMP MOD go 'Turn this woman out so she'll go away.'

ke. DIR

Aspect markers such as ne "perfective" in (7) are nuclear-layer operators (Foley & Olson 1985: 33) and have scope over the entire verb sequence in nuclear-layer SVCs. Another nuclear-layer operator is negation, present here in the form of the negation particle hiai. The fact that the negation here has scope over both verbs, treating them as a unit, is evidence that we are indeed dealing with a case of nuclear-layer serialization. Hano 'go' (and its dual/plural form ö) is the most frequent verb found in Vi position of nuclear SVCs, although a few others also occur. All of them are intransitive, and most are verbs of motion, such as ua 'paddle' in (9): (9)

Na

no

ua hehega na ko matu tuohine paddle search DEM TOP IPLEXCL.POSS sister 'He has paddled here in search of our sister.'

3SG TA

na. DEM

This type of construction is used for expressions of motion, location, and purpose, translating loosely as English 'go and do something', 'go to do something'. There are other constructions that may express a relation of purpose between two verbs; compare (9) with (10), where the 'go and do...' meaning is expressed by means of the conjunction ο 'and':

234

(10)

Äshild Ncess

Ko

te

ag-ona

ko

te

hano

ki

mouku

ο

aga.

TOP

ART

job"3SG.POSS

TOP

ART

go

to

bush

CONJ

work

'His job was to go to the bush and work.' The difference between (7) and (9) on the one hand, and (10) on the other, seems to be that in the serialized instances, there is a more direct causal connection between Vi and V 2 . In (7), the disappearing is construed as a direct consequence of the going; in (9) the searching takes place through the act of paddling, that is, the searching consists of paddling from place to place with the purpose of finding something. In (10), however, the connection is of a different kind; the going is not the cause or means of working, just a prerequisite for it. (10) describes a sequence of events: first going, then working. (7), on the other hand, presents a single event of disappearing by means of going, and (9) a single event of searching by means of paddling.

3.3. Nuclear-layer serialization vs complex verbs One question which must be addressed is whether such nuclear-layer serializations can be formally distinguished from complex constructions where a verb is combined with another verb or adverbial element; fie kai 'want eat' = 'be hungry' 3 . In other words, is there really a separate formal construction type which merits the label "SVC", distinct from other kinds of verb sequences in the language, or should they be considered to belong to the same category as what might be called "complex verbs" of the fie kai type? The evidence for SVCs as a formal type distinct from complex verbs can be found by examining the possibility of material intervening between the two elements of the construction. Pileni has a set of postverbal particles which occur very frequently, with various semantic or discourse-pragmatic functions. These particles treat the complex verbs as an indivisible whole and so occur after the whole sequence in such cases; whereas in a serialization construction a particle may intervene between the two verbs in the construction. This is illustrated in the following examples, where the particle mua, which functions as a politeness marker and for the "softening" of requests and commands, is found between the verbs in the serializing construction, but after both elements (and the incorporated object) of the complex verb:

Serial verbs and complex constructions in Pileni (11) a.

Τα-δ

mua

1 DU.INCL-gO.PL PART

tahao

i

Stroll

LOC ART

te

235

taon. town

'Let's go for a walk in the town.' b. A-koe

fie kai poi mua ο hiai. ART-2SG DEM ΤΑ want eat pig PART CONJ NEG 'Would you like to eat pork or not?' It is possible that mua might also be placed after the verb sequence in (11a); but interposing it between the two elements of the complex verb in ( l i b ) would certainly be ungrammatical. This indicates that there is indeed a formal distinction between the nuclear-layer serialization construction and complex verbs. There are only very few attested examples of such SVCs with intervening material between the verbs, and so it is difficult to say exactly which items may occur in this position. The particle ala, with a meaning similar to mua, and the emphatic particle loa, are found in a few instances, as are the directional particles. Combinations of particles occur very occasionally: (12)

na

no

Tu-ö-atu

loa

ta-ia

mbulumakao

e

2PL-go.PL-DlR

EMPH

kill-TR

cattle

TA stand

päpoi

na.

fence

DEM

thü

/

te

LOC ART

'Go and kill the cows standing inside that fence.'

3.4. Asymmetrical nuclear-layer serialization There is another, formally similar construction in which the V 2 denotes the manner or path of the action denoted by Vi. Since Pileni has no formally distinct class of adverbs, this kind of information can only be specified by another verb following the verb describing the actual action: (13) Nekepo akuaku vakao loa i thaug-ona na. PART weed encircle EMPH LOC house-3SG.POSS DEM 'She weeded around her house.' Semantically, this is somewhat different from the SVCs described above in that the V 2 seems to modify the Vi; that is, it does not so much contribute to a complex event (cf. 'go+disappear' in (7) above) as it specifies a property of an event which in itself is already described by the Vi. This resembles what Aikhenvald (1999) calls an asymmetrical serial verb

236

Äshild Ν cess

construction, described as consisting of one verb from a large, open class and another from a small, closed set, and referring to "a single event described by the verb from a large class; while the verb from a closed class provides an additional directional or aspectual specification" (Aikhenvald 1999: 472). It is not clear whether there is a restricted set of verbs which may occur as V 2 in this construction in Pileni, but it is only attested with a small set of verbs such as vakao 'encircle', tahao 'stroll', Ιανοί 'good' and tapeo 'bad' in the V 2 position. The meaning contributed by the V 2 is generally one of manner rather than necessarily direction or aspect, as seen from examples such as piki-a lavoi-na (keep-TR good-TR) 'take good care o f . Aikhenvald writes that "the transitivity value of an asymmetrical serial verb construction as a whole is usually the same as that of the verb from an open class" (Aikhenvald 1999: 472), i.e. of the Vi. This is reflected in Pileni by a formal restriction: the V 2 in the asymmetrical construction must agree with the V, in transitivity. That is, if the V] carries one of the set of transitive suffixes which are obligatory on all transitive verbs, the V 2 takes such a suffix as well, although there is no reasonable way of interpreting the V 2 as semantically transitive: (14)

Ko-i

mot-ia

TA-3SG cut-TR

na

mara

3SG.POSS eye

na,

mot-ia

DEM cut-TR

vakao-ina. encircle-TR

'He cut off its (the coconut's) end, he cut around it.' The allomorph -ina is the productive form of the transitive suffix and occurs for example on loanwords in transitive contexts as well as on verbs modifying transitive heads. This form of the suffix invariably occurs on the second verb of the modifying construction, even when this verb used in isolation takes another form of the transitive suffix (see Nasss 2002 for more on the transitive suffixes in Pileni): (15)

Kolu-no

maoli la khoulua kip-ina true DEM 2DU keep-TR 'If you are telling the truth, keep it quiet.'

2DU-TA

themu-ina. quiet-TR

The verb themu 'still, quiet' here takes the productive suffix form -ina, although in other contexts such as under causativization, the suffix appearing on this verb is -akina: hathemuakina 'to silence (someone)'. This difference further suggests that the suffix in (15) is not an indicator of the semantic transitivity of this verb, but rather marks agreement in transitivity between a main verb and its modifying manner verb. The fact that the transitivity of the V 2 depends on that of the Vi in this construction, the V 2 taking a regularized form of the transitive suffix in

Serial verbs and complex constructions in Pileni

237

agreement with V t whenever Vi is transitive, is a further indication of the "asymmetry" of this construction: the V t provides both main lexical content and transitivity value while the V 2 contributes an additional property of the event described by the Vi and must adopt the same transitivity value as the Vi. It is difficult to be certain that this is formally different from the other cases of nuclear-layer SVCs cited above, since, as already stated, these SVCs mostly occur with 'go' and other motion verbs as Vi, and these are all intransitive. There is no attested case of such SVCs with a transitive Vi, and so no way of being sure that such a construction would not behave identically to the asymmetrical construction. However, it is possible for this construction to have an intransitive Vi and a transitive V2: (16)

Ta-o

umu a earth.oven POSS 'Let's go open our earth oven.'

LDU.INCL-go.PL

fuke-a

open-TR

te

raua.

ART

1DU.INCL.POSS

This would seem to indicate that the transitivity of the V 2 in this type of SVC does not depend on the Vi in the way it does in the asymmetrical construction. Seen in conjunction with the semantic difference between a construction describing a complex event on the one hand and a manner construction on the other, this suggests that we may here be dealing with two formally different constructions. One is a case of symmetrical serialization, where both verbs contribute equally to the meaning and formal properties of the construction, while the other is an asymmetrical SVC where one verb determines the meaning and transitivity of the construction while the other verb just modifies it.

4.

Core-layer serialization

4.1. The basic structure of core-layer SVCs Identifying core-layer SVCs in Pileni is a rather more complex matter than in the case of nuclear-layer SVCs, since core-layer serializations consist of two (or, in principle, more) distinct nuclei with separate tense-aspect markers, which are joined together at the level of the core, that is, the predicate + argument level. As noted in the introduction, Pileni allows omission of any argument NPS whose referents may be retrieved from context. This means that it is not at all unusual to find long sequences of verb phrases without any overt arguments at all. Where one overt argument appears to be shared between two or more VPs, it is often impossible to

238

Äshild Ncess

determine whether this is a syntactic requirement (separate expression of the argument for each VP being impossible or unidiomatic), or whether it simply results from the optional ellipsis of coreferent NPs. A further problem concerns the status of the subject clitics as regards argument omission. The precise distribution of subject clitics in Pileni is at present poorly understood; they are frequent, but apparently never obligatory. The question is, if two verb phrases each have overt, coreferent subject clitics, do these count as overt arguments, which would mean that such a construction consists of two juxtaposed clauses? Or are the clitics just formal reflexes of an omitted argument which the VPs might both share, making such constructions potential SVCs? The problem is illustrated by (17): (17)

Lha-ko

toa lha-ko mot-ia te pakola la na. take 3DU-TA cut-TR ART giant DEM DEM 'They cut the giant (to pieces) (lit. they-took they-cut the giant).'

3DU-TA

Here, both verbs have a subject clitic on their ΤΑΜ marker, and so might be said to have their subject argument expressed separately. On the other hand, the semantics of this construction strongly suggests that this is an instance of serialization with toa as V* (see 4.3.2. below), in which case the clitics presumably cannot be taken to represent independent expression of arguments. More research is needed into the properties of subject clitics in Pileni. According to Foley and Olson (1985: 47), a structural characteristic of core-layer serializations is that the verb phrases in the construction share their peripheral layer, that is, "any core units bound together in a core juncture must share a common set of arguments of location and time, as well as other peripheral features like tense and mood." This shared periphery distinguishes them from conjoined clauses, which are conjoined at the peripheral layer and so may have separate locational/temporal arguments and so forth. The difference is illustrated in (18): (18) a. Lha-ko 3DU-TA

laka

i

pass

LOC ART i s l a n d

te

motu

lha-ko

u-ake

3DU-TA p a d d l e -

i

te

LOC ART

vai. water

DIR

b.

'They passed an island and paddled up a stream.' Nekepo hua-mua mai a te tupua na PART CAUS-front DIR PART ART creator DEM e-i pnepen-age a hihiga ne-i tuhämac. T A - 3 S G prepare-DLR ART RED.thing T A - 3 S G distribute 'In the beginning the creator made all things and distributed them.'

Serial verbs and complex constructions in Pileni

239

In (18a), the locative phrases i te motu and i te vai have scope over only one each of the conjoined verb phrases, lhako laka and lhako uake respectively. In (b), however, the temporal phrase nekepo huamua mai is shared by both verb phrases; bothpnepenage 'prepare' and tuhä 'distribute'. This may not in itself be an adequate criterion for SVCs, as there is of course no restriction against conjoined clauses referring to events taking place in the same location and time, and so sharing one or more peripheral arguments. To the extent that shared peripheral arguments may be used as a diagnostic for core-layer SVCs, then, it is strictly a negative criterion: if two VPs have different peripheral arguments, the construction is not an SVC; but if they do not, it may or may not be. Pileni shows constructions which resemble core-layer SVCs, such as (19): (19)

Ko

nh-age

te

thua

la

ko-i

tapena-kina

kaikai.

TA

go-DlR

ART

spirit

DEM

TA-3SG

ready-TR

food

'The spirit went and prepared some food.' Here two verb phrases with separate tense-aspect marking share a common argument {te thua) which is expressed only once. This is a very common pattern in Pileni and is found with both transitive and intransitive verbs in all possible combinations. The shared argument is always the subject of the V 2 but may be either subject or object of the Vi; if it is the subject, then it is moved from its normal preverbal position and postposed to the Vi, so that it is in the position between the two verbs. This can be done without further marking if the Vi is intransitive, as in (19); if the V2 is transitive, such postposing of the subject requires marking it with the preposition e (Naess 2000, 2002, ex. 10). What is the evidence for or against considering (19) an instance of serialization? There is not very much. The fact that the subject of nhage has been postposed may be an indication that this is a construction specifically geared towards the syntactic sharing of an argument by two verbs - the argument is placed in this position so that it can be "taken over" by the V 2 . However, the verb-subject order might also be a result of stylistic or discourse-pragmatic factors - more study is needed of the factors that trigger alternative word orders in Pileni. Nor is it known whether the meaning or grammaticality of (19) would change if both verbs had their subject argument expressed by separate NPs. (19) is certainly the most idiomatic way of expressing the meaning in question; but I have not been able to test whether the sentence is equally

240

Äshild Noess

good with an overt subject NP for each verb (in which case (19) would be an instance of argument ellipsis rather than serialization). Similar constructions to (19), but where the shared argument is the object rather than the subject of the V], also occur, and under a serialization analysis these would qualify as "switch-subject" SVCs: (20)

Ko

ko

te-ia

nohine-aku

ko

2SG TA kill-TR wife-LSG.POSS TA ' Y o u killed m y w i f e and she died.'

mate. die

Again, it is uncertain whether this sentence would be equally felicitous with overt expression of the shared argument for both verbs (i.e. with an overt pronominal subject for the V 2 ). It is also formally similar to the relative construction in (6), and it is not possible on the basis of the currently available data to conclusively classify (20) as SVC, relative construction, or conjoined clauses with argument ellipsis.

4.2. VP modification constructions There are some cases where more reliable criteria may be applied to argue for the classification of a construction as a core-layer SVC in Pileni. One such case is a construction where a VP modifies another VP, with no possibility of overt expression of arguments for both VPs separately. Since there cannot be an overt NP subject for each VP, this type of construction cannot be interpreted as resulting from argument ellipsis. This is the kind of construction referred to by Crowley (1987: 49) as "ambient serialization", which he defines as "a construction in which a verb is serialized to another verb, but in which there is no specific referent associated with the subject of the serialized verb, and the verb simply describes a generalized predication". A construction which seems to fit this description is found in Pileni with either of three verbs, lävoi 'good', mbarava 'good', and osi 'finish', as V2: (21) a. Lha-ko

lävoi. good 'They ate until they were full (lit. they ate, it was good).'

3DU-TA

b.

khaikhai

ko

RED.eat

TA

A

hihiga

mharu-ko

ART

RED.thing

IPLEXCL-TA

pepenä ko mbarava. prepare TA good 'We have prepared everything properly (lit. we have prepared the things, it is good).'

Serial verbs and complex constructions in Pileni c.

241

A meme fonu ko ö k-osi. ART small turtle TA go.PL TA-finish 'The small turtles all went off (lit. the small turtles went, it finished).'

These constructions function at the level of the core as the verbs each have separate tense-aspect markers, in contrast to the nuclear constructions where a single tense-aspect marker covers both verbs of the construction (ex. 7, 9). There is no possibility in (21a-c) of introducing an overt NP as the subject of lävoi, mbarava, or osi. This is because these verbs do not predicate anything of a specific referent; rather, they modify the main VP (V, + subject) and refer to aspects of the situation described by the Vi. Since these verbs in this construction cannot have independently expressed subjects, this is not an instance of argument ellipsis, but rather must be counted as an SVC where V2 is a modifier of the head Vi, which means that the two verbs are not on the same syntactic level. Unlike the asymmetrical nuclear-layer SVCs described in 3.4. above, it seems clear that the set of V2S for core-layer VP modification is extremely limited. All three of the attested V 2 s refer in this construction to the (satisfactory) completion of an action, with readings such as 'enough', 'properly', 'finished'. It seems that this construction is used to denote that an action or process has been brought to its conclusion, with lävoi and mbarava in addition implying that the event was concluded satisfactorily or in the appropriate manner.

4.3. Grammaticalized functions of core-layer SVCs: the case of toa 'take' Another indication that a construction should be classified as a grammaticalized core SVC is that the construction has a specific meaning that would not arise with a simple sequence of clauses. Such constructions occur in Pileni with the verb toa 'take' as V]. There are two such constructions with different semantic and syntactic properties: one where toa introduces an object argument, and another where it contributes a volitional or inceptive meaning to the clause.

242

Äshild Ncess

4.3.1. toa serialization 1: Argument

introduction

The occurrence of the verb 'take' in SVCs is known to be a common phenomenon in serialising languages (see also Ozanne-Rivierre, this volume). Lord (1993) devotes a whole chapter to serial constructions with the verb 'take' in various languages, and claims that 'take' in these constructions tends to evolve into object markers of various kinds. She says: "In many languages, a verb meaning 'take', occurring as the first verb in a serial verb construction, serves to introduce an object noun functioning as semantic Instrument, Patient, means, manner, or material" (Lord 1993: 65). A great majority of the instances of constructions resembling SVCs in Pileni involve the verb toa 'take' as V]. For a number of these cases, Lord's description fits very well: (22)

Lha-ko

to-age

3du-ta

take-DiR a r t

a

ika ko thun-age. fish t a cook-DiR 'They took the fish and cooked it.'

In (22), toa could very well be considered to introduce the NP which is the object argument of the V2. The text immediately preceding this example speaks of a man returning from the sea, but does not explicitly mention the fish he has caught; and so the construction with toa may be a means of introducing a ika 'the fish' into the discourse before anything further is predicated of it. It is also quite possible to construe the contribution of toa in (22) as referring to an actual act of taking; the fish is first taken, then cooked, and so on. There are instances, however, where no actual taking seems to be involved: (23)

Te

kuli ko-i toa na pihoulu ko-i lulu-ia. dog TA-3SG take 3SG.POSS head ta-3sg shake-TR 'The dog shook his head (lit. The dog took his head shook).' Presumably, toa is not to be interpreted as referring to a physical act of taking here; the dog does not actually grab hold of its head to shake it by means of its paws. Rather, toa serves here to introduce the object argument of luluia 'shake'. I am not certain why this construction is used here rather than the simple sentence koi luluia na pihoulu. Possibly it is a means of emphasizing the head as an argument central to the event in question, by introducing it through serialization with toa before mentioning the actual act involved. The construction appears to be common in contexts where the agent uses a specific part of his body to perform an action, see also (24): art

Serial verbs and complex constructions in Pileni (24)

243

Ko-i

toa na noko ko-i mue-ia i te paua. take 3SG.P0SS bottom TA-3SG slam-TR LOC ART clam 'He slammed into the clam with his bottom.'

TA-3SG

4.3.2. toa 2:

Volitional/inceptive

In a number of instances of apparent serialization with toa, however, it is impossible to construe the verb as "introducing an object noun". In (25) below, it is the subject (of Vi as well as V 2 ) which follows directly after toa, so if toa is to be said to "introduce" anything, it must rather be the subject: (25)

Ko-i

toa

e

te

hahine

tau

athua

la

ko-i

TA-3SG

take

AG

ART

woman

of

spirit

DEM

TA-3SG

tuku-age

ki-a

te

nonohine

la

po...

say-DIR

to-ART

ART

old.woman

DEM

COMP

'The spirit woman said to the old woman... (lit. took the woman-ofspirits said to the old woman)' Again, toa does not refer to any actual event of taking. But neither does it introduce any kind of object argument; and it is difficult to explain why a complex construction such as an SVC would be needed to introduce a subject argument, which would be just as grammatical and at least as prominent if it were just placed in the regular subject position of tukua 'say'. The point becomes even clearer if we consider a sentence like (26): (26)

Ko-i

toa

TA-3SG t a k e

ko-i

taka-hia

te

lauplaka

TA-3SG stamp-TR ART l e a f . t a r o

na

a

ko

DEM CONJ ΤΑ

fotu. break

'(The pig) stamped on the taro leaf, and it broke.' Here, toa does not appear to introduce anything at all - the subject has no overt expression, and the shared object argument follows the entire verb sequence rather than being "introduced" by toa. Nor is there any kind of taking event involved; the example is taken from a story where a group of animals use a taro leaf as a boat to sail out to sea, so the pig is here standing on the leaf and does not 'take' it in any sense of the word. Rather, in a broad range of similar examples found in the data, toa 'take' appears to contribute a kind of inceptive meaning to the clause; it emphasizes the volitional initiation of the act described by the V 2 . It is not unusual for 'take' to acquire an inceptive meaning in certain contexts; the English expression 'take to V-ing' is an example, as is the Norwegian ta og

244

Äshild Ncess

V 'take and V ' , which has a strong intentional meaning with inceptive connotations (Faarlund et al. 1997: 535). The pig in (26) is acting willfully, almost spitefully (he is being teased by his companions for his bad singing and responds by stamping angrily, with disastrous results). The examples in (27) show similar uses of toa+V2· (27) a.

Ko

te

TOP

ART mother-3SG.POSS

te

memei

ART small

hina-na

na

ko-i

toa

ko-i

DEM

TA-3SG

take

TA-3SG

tätai-age carve-DIR

loku. bow

'His mother (finally) carved him a small bow.' b.

Ko-i

ko-i

usi-a

na

TA-3SG

toa e te paua take AG ART clam 'The clam bit his leg.'

TA-3SG

bite-TR

3SG.POSS

vae. leg

The context in (27a) is that the woman's son has been asking her to make a bow for him, but she has declined, saying that she does not know how; however, he keeps on asking and in the end she does as he asks; so (27a) has a sense of "she finally gave in and did it". Example (27b) presents the act of biting as a sudden and rather hostile action by a volitional agent, which takes the patient (a heron wading in the shallows) by surprise. There is, then, an emphasis on the volitionality of the agent in performing the act; and since the agent's volitional acting is what starts off a typical transitive event, by implication this amounts to a focus on the initiation of the act, which in turn may give rise to an inceptive reading.

4.3.3. Word order in core-layer SVCs: contiguous vs serialization

non-contiguous

Durie (1997) makes a distinction between "contiguous verb sequencing, where any arguments are placed outside the verb string, and non-contiguous sequencing, where arguments can intervene between the verbs." (Durie 1997: 302). There are two possible patterns for verb serialization, Durie says; "either the verb series acts like a single verb for the purposes of placement of arguments (...) or an argument is located in the usual position with respect to the first verb that introduces it. (...) Some languages have either one pattern or the other, and some (...) show both patterns." (Durie 1997: 307).

Serial verbs and complex constructions in Pileni

245

As will be apparent from the examples in the preceding sections, Pileni does in fact exhibit two patterns of argument placement in core-layer SVCs, although a slight modification must be made to Durie's definition: the difference is relevant only with respect to the placement of the shared object. Whether this object is inside or outside the verb sequence, the subject may always be postposed to the Vi; if the Vi is transitive (as is the case with the toa construction), such postposing triggers the agent marker e on the subject NP. This is a possibility in all types of core-layer SVCs regardless of their meaning or function. However, the two SVCs described in 4.3.1.-2., where toa as V] serves two different functions, are distinguished by a difference in the position of the object. In the first construction, where toa serves to introduce the object argument, the object is, as might be expected, placed directly after toa, between the two verbs of the construction - what Durie refers to as noncontiguous serialization. In the second construction type, however, where toa has an intensifying or inceptive meaning, the object follows both verbs, as seen in examples (26-27). In other words, Pileni exploits the contrast between contiguous and non-contiguous serialization to encode two different meanings with the same V,: in the pattern where "the verb series acts as a single verb for the purposes of placement of arguments", toa as a Vi cannot be separated from V 2 by an intervening object argument, and it fullfils a more grammaticalized function with respect to this V 2 - it functions to emphasize or intensify the actor's volitional involvement in the action the V 2 describes. On the other hand, where the object argument "is located in the usual position with respect to the first verb that introduces it", in other words where the order is (S)VOV, toa as Vi shows a greater semantic independence from the V 2 - it either contributes its own semantics to the situation, implying that something is actually, physically taken; or at the very least it introduces into the discourse an argument which will then be "taken over", so to speak, by the V 2 . To the extent that Pileni can be considered to have core-layer SVCs with other types of verbs as well, it is this second pattern that they follow; the constructions discussed in 4.1. as possible SVCs show the non-contiguous word order, with the shared S or Ο of the Vi interposed between V, and V 2 . This might imply a kind of iconicity in SVC patterns: in an SVC where both verbs contribute semantically to the description of a complex event, the verbs are treated as separate entities in that they may be separated by an object argument; but where the V, is grammaticalized into something like an inceptive or intensifying marker, there is a tighter cohesion between the

246

Äshild Ncess

Vi and the V 2 , and the shared object argument may not intervene between them.

4.4. Tense-aspect marking in core-layer SVCs The majority of core-layer SVCs in Pileni have the same tense-aspect marker on both verbs in the construction. This is frequently cited as a defining characteristic of SVCs; Durie (1997: 291), for instance, writes that "the serial complex has shared tense, aspect, modality, and polarity". In Pileni, however, we find some instances of constructions resembling core-layer SVCs where the tense-aspect marker is different for the two verbs: (28)

Ne-i

toa

no-i

kenukenu-a

TA-3SG take TA-3SG decorate-TR

a

gau-ana

ART arrow-3SG.POSS

na. DEM

'He decorated his arrows.' One might argue that a failure to satisfy the "shared tense-aspect" criterion simply disqualifies this construction as an SVC. On the other hand, it strongly resembles, both semantically and formally, the constructions in 4.3.2. which we have taken to be SVCs; the verb toa appears to contribute an inceptive meaning rather than denoting an actual event of taking. It would seem, then, that by some criteria, (28) is an SVC, but by other criteria it is not. The solution to this problem lies in the distinction between nuclear-layer and core-layer serialization. As pointed out by Foley and Olson (1985: 33), aspect is a nuclear operator, which "indicates the intrinsic nature of the action of the verb". Tense, on the other hand, is a feature of the periphery which defines the temporal setting of the core as a whole. In other words, claiming that serial verbs must have shared tense and aspect obscures the difference between nuclear-layer and core-layer serialization. In Foley and Olson (1985), where this difference is set out, the only claim is that "all verbs in a serial construction must have the same tense and mood' (47, my emphasis). If the core- vs nuclear-layer SVC distinction is valid, then we would expect nuclear-layer SVCs obligatorily to share both tense and aspect, while for core-layer SVCs only tense (and mood, a core-layer operator) would be required to be the same for all verbs involved. The tense-aspect markers in Pileni are primarily aspectual (with the exception of the future marker ka), although they may also give rise to

Serial verbs and complex constructions in Pileni

247

temporal interpretations. The most common are e "stative", ne "perfective", no "imperfective", and ko "inceptive". In (28), there is no difference in the temporal interpretation of the two verb phrases. If the two verbs were to be read as differing in tense, the meaning would be something like 'He took the arrow and is now decorating it'. But this is not what (28) means; the construction as a whole refers to a past event, and in fact there is no clear distinction between a process of taking and one of decorating. They both constitute a single event and so cannot be assigned different temporal references. Rather, the difference between the verb phrases in (28) is one of aspect. While the act of decorating is presented as continuing over a certain expanse of time, taking imperfective aspect, toa here fills a grammaticalized function, and so there is no actual act of taking involved that might be construed as imperfective, and the verb rather takes the perfective marker. The construction consists of two nuclei with separate aspectual specifications which are combined into a complex core - a core-layer SVC where toa 'take' has acquired a specialized aspectual meaning. Similar examples are found where the Vi describes an act and the V2 its resulting state: (29)

Ko-i

kei-na

tokoto loa i na mnava. lie EMPH LOC 3SG.POSS stomach 'He ate it, and it lay in his stomach.'

TA-3SG eat-TR

e

ΤΑ

Again, the verbs in this construction do not differ in temporal reference. It is possible to conceive of an interpretation of (29) as meaning 'He ate it, and (now) it is lying in his stomach'. But this is not what (29) in fact means in the context from which is it taken. Later in the same text, the fish in question has its stomach cut open and the ring which it has eaten is removed; so tokoto 'lie' cannot be ascribed a present-time reference. The event as a whole is situated in the past; but the initial process, the eating, is marked with the inceptive aspect marker while the resulting state takes stative aspect. By contrast, nuclear-layer SVCs consist of a complex nucleus where the tense-aspect marker has scope over both verbs, as seen in the example in (7). Thus nuclear-layer SVCs necessarily share both tense and aspect, while core-layer SVCs may differ in aspect, but not in tense.

248

5.

Äshild Ν cess

Conclusion

Despite the difficulties inherent in distinguishing SVCs from complex clauses or clause sequences in a language with widespread argument ellipsis and little subordination marking, a number of constructions can be isolated in Pileni to which it seems appropriate to apply the label "serial verb construction". Serialization occurs both at the level of the nucleus and of the core, and the formal and semantic properties of these constructions to a great extent correspond to some of the main types of SVCs recognized for other serializing languages: nuclear-layer SVCs with intransitive motion verbs as Vi, serialization with the verb toa 'take' with various more or less grammaticalized functions. An interesting twist to the latter is the way the difference between contiguous and non-contiguous serialization is exploited to achieve two different functions with the same V|. Due to syntactic properties such as argument ellipsis and lack of subordination markers in many instances, there are a number of constructions for which it is impossible to draw any firm conclusions as to their status as SVCs or some other type of complex or conjoined clause. There are constructions which are certainly SVCs, and constructions which are certainly not; as for the "grey area" in between, further research might resolve the question for some instances, but others may simply be impossible to categorize conclusively.

Orthographic conventions and abbreviations g represents the voiced velar nasal. An h following an oral or nasal stop marks (phonemic) aspiration. Long vowels are marked with a macron. AG ANAPH ART CAUS COMP CONJ DEM DIR DU EMPH EXCL INCL

agentive preposition anaphoric particle article causative prefix complementizer conjunction demonstrative directional particle dual emphatic particle exclusive inclusive

LOC MOD NEG PART PL POSS RED SG TA TOP TR

locative preposition modal marker negation particle postverbal particle (meaning unclear) plural possessive reduplication singular tense-aspect marker topicalizing particle transitive suffix

Serial verbs and complex constructions

in Pileni

249

Notes 1.

2.

3.

In addition to the editors of this volume, whose insightful comments prompted a radical revision of the initial version of this paper, I thank Even Hovdhaugen for additional comments and discussion. I would also like to acknowledge the invaluable help which our main Pileni consultant, Charles Bice Mate, has provided with the analysis and interpretation of the Pileni material. I am of course entirely responsible for any remaining mistakes. The 1999 census (Beimers 2002) treated the language of the Duffs separately from that of Pileni and surrounding islands; but they are mutually comprehensible, and speakers describe them as "the same language". The formal status of fie 'want' in Pileni is somewhat uncertain. It does not occur independently as a main verb, and its cognates in related languages are often classified as belonging to a class of preverbal adverbs (e.g. Mosel and Hovdhaugen 1992, Hooper 1996). However, no such class has been established in Pileni; with the exception of fie there is little evidence of any prenuclear elements of the verb phrase other than purely grammatical particles such as tense/aspect and negation. It is tempting to classify fie as something like an auxiliary verb, although this is not a class that is usually recognized in Polynesian languages.

Complex predicates in Tahitian: A particular case of qualitative modification Mirose Paia and Jacques Vernaudon

Abstract Although the concept of "veib" is not relevant in our metalinguistic framework of Tahitian1, we come closer to the topic of "serial verbs" in focusing on two-lexeme strings 'x Y', with constituents X and Y usually denoting processes. In such compounds, χ and Y never refer to two successive or concomitant processes, but υ always modifies x. The string 'χ γ' refers to a unique action described by the first lexeme x, while the second lexeme Y provides additional qualifications of x. It is a particular case of qualitative modification, used both in predicate and argument phrases.

1. A short presentation of morphosyntax and metalinguistic options The distinction between lexemes on the one hand (e.g. 'amu, 'eat'; fare, 'house'; uteute, 'red'; etc.) and grammatical morphemes on the other (e.g. ua, tense-aspect-mood particle; te, "article"; ia, anaphoric; etc.) is relevant and useful at an early stage of analysis. Yet, all items do not fall neatly into one of these two categories. Some of them are traces of abstract operations (like grammatical morphemes) and yet appear in the same paradigms as lexemes (e.g.pau, hope, tupu, vaiho, etc.). As in other Polynesian languages, there is no morphological criterion to classify isolated Tahitian lexemes. From a semantic point of view, the same lexeme can refer either to an entity, a process2, a property, etc., depending on the context in which it occurs. (1) Tari'a mai na i to tari'a i te fa'ateni a te ear DIR DEICT PREP your ear PREP DET lament PREP DET mauri tupuna. ghost ancestor 'Listen to the lament of the ancestors.' (J. Mairai, Maropütoto, 1987)

252

Mirose Paia and Jacques Vernaudon

(2)

Ε mea varavara

ato'a te

ΤΑΜ

also

DET PLUR man

ΤΑΜ go

te

i'a, no te mea,

aita

e

NEG

ID

rare

titi'amata diving mask

(3)

i PREP

te

mau

tane e

DET fish because

titi'amata

no

diving mask

PREP DET to fish

haere

e ΤΑΜ

täi'a.

'Also the men would seldom go fishing underwater as there was no diving mask to do so.' (TM: 62) Ε 'ahu rätou i te 'ahυ 'uo'uo. ταμ wear 3 PL prep det dress white 'They wear white dresses.' (TM: 41)

Most linguists working on Tahitian define noun and verb categories on distributional grounds. Briefly, an environment is considered as verbal or nominal according to the kind of grammatical morphemes it contains, those morphemes being classified α priori as specifically "verbal" or "nominal". For example, te is labelled "article"; it is therefore considered as a "nominal particle". Then 'te X' is a "noun phrase" and X, as the nucleus of this phrase, is a "noun". On the other hand, ua, a tense-aspect-mood morpheme (ΤΑΜ), is traditionally classified as a "verbal particle". Therefore 'ua x ' is a "verb phrase" and x is a "verb" in this particular context. However, it can easily be shown that some lexemes appearing with ΤΑΜ markers still point to a class of entities in which the referent of the subject is included. From the semantic point of view, those lexemes are closer to the nouns of Indo-European languages. (4)

(5)

(6)

Ua hina'aro ato'a rä 'oia ia riro Matatura ei 'aito ταμ want also but 3sg ΤΑΜ become Matatura ΤΑΜ hero mai ia na iho i 'aito i roto i to na va 'amata 'eina 'a. prep 3sg dir tam hero prep in prep his clan 'But he also wanted Matatura to become a hero, as he himself used to be a hero for his clan.' (HPR2: 170) Ε ua tamaiti a'e ra Taaroa. and TAM young man dir DEICT Taaroa 'And Taaroa became a young man.' (TH: 343) Ua mou'a rua Papara. TAM mountain two Papara 'Now, Papara has two mountains.' (TH: 147)

On the other hand, lexemes referring to processes are frequently preceded by the "article" te, and, still, can be used with grammatical

Complex predicates in Tahitian

253

morphemes and complements (directional, frequency indicators, "passive voice" suffix, object, etc.) that are typical "verb" indicators in classical descriptions. It is then tempting to establish a semantic connection between those lexemes and verbs in Indo-European languages. (7) Ε na hea te tätara i te reira mau fifi ia ΤΑΜ PREP where DET untie PREP DET ANAPH PLUR problem ΤΑΜ tupu no a mai ? occur only DIR

'How can we solve these problems when they occur.' (RAU: 30) (8)

Ua mo'ehia ia Vaitä te ha'apapü ato'a mai e, e lost-hia PREP Vaita DET make sure also DIR that ΤΑΜ tupu fa'ahou mai teie rä'au.

TAM

grow

again

DIR this tree

'Vaita has forgotten to make clear that this tree will grow up again.' (PNM: 31) (9) Ia roa'a mai te pöreho i te hopuhia, mea tanu ΤΑΜ reached DIR DET shellfish PREP DET dive-hia ΤΑΜ bury ta Tetauru i raro i te repo. DET PREP Tetauru PREP beneath PREP DET soil 'When he obtains a shellfish after diving for it, Tetauru buries it beneath the soil.' (HPR2: 73) In order to avoid a gap with the Indo-European classical metalanguage, most of the descriptors using a distributional approach keep the traditional labels "verb" and "noun" to distinguish two kind of phrases in Tahitian. This terminological choice frequently leads to label as "verbal" items which are semantically similar to Indo-European nouns, and, in reverse, as "nominal", items which are nearest to the traditional verbs. To find a way out of this paradox, we prefer not to use these labels for our own description of Tahitian. When necessary, we would rather say that the lexeme refers to a "process", an "entity" or a "quality". Now, let us briefly consider the main syntactic constituents of a declarative sentence. The predicate phrase (Pred) is the only obligatory constituent: (10)

Ua reva. TAM leave

'Let's go.' (PAA: 7)

254

Mirose Paia and Jacques Vernaudon

(11)

Parau

mau.

word

fixed

'It's true.' (PAA: 11) The predicate phrase can be followed by arguments, which can be proper nouns, personal indexes, a word or group of words preceded by a determiner. When it is explicitly mentioned, the first argument following the predicate phrase refers most of the time to the subject. It comes without any specific preposition, contrary to the other oblique arguments which are introduced by prepositions i, no, na, etc. (12) Ε riri te päpä. ΤΑΜ anger DET father

'The father is becoming angry.' (PAA: 3) (13) Mea varavara roa te parau tahiti. ΤΑΜ rare HD DET speak Tahitian 'It very seldom happens that [we] speak Tahitian.' (PAA: 3) The most frequent oblique argument is introduced by i (i with phrases with determiner, ia (i+a) with proper nouns and personal indexes). It corresponds to very different semantic roles, as the following examples partly show: (14) Ε 'amu mätou i te mä'a tahiti. ΤΑΜ eat IPL-EXCL PREP DET food Tahitian 'We eat Tahitian food.' (PAA: 4) (15) Ua roa'a te feiä 'eiä i te müto'i. ΤΑΜ reached DET people steal PREP DET police 'The thieves have been caught by the police.' (16)

Ua fa'a'ite Tauhiti i to na metua täne ΤΑΜ show Tauhiti PREP his parent male i to na 'öpuara'a. PREP

(17)

(18)

his

decision

'Tauhiti told his father about his decision.' (HPR2: 36) Ε piti 'äva'e au i te fa'aeara'a i te fare. ΤΑΜ two month LSG PREP DET stayRA'A PREP DET house Ί have stayed at home for two months.' (MeT: 59) Ε piti metera i te roa. ΤΑΜ two meter

PREP DET long

'It is two meters long.' (TM: 30)

Complex predicates in Tahitian (19)

255

Ua hope rätou i te reva. ended 3 PL PREP DET leave 'They have all gone.' (DFV: 200)

ΤΑΜ

When the predicate phrase refers to a state, the experiencer of this state is made explicit by the following complement without preposition (see ex. (12) and (13)). The basic order is then: Pred 0 Argument If the predicate phrase refers to a process with two participants, the agent is usually expressed by the first bare argument without preposition, and the oblique argument introduced by i refers to the patient (see ex. (14)). The order is: Pred 0 Agent i Patient However, a handful of lexemes indicating processes with more than one participant impose the reverse disposition (e.g. roa'a, mutu, fati; see ex. (15)): Pred 0 Patient i Agent Comparing the second order (Pred 0 Agent i Patient) with the first one (Pred 0 Argument) with only one participant, some authors have classified Tahitian as an accusative language3. It should be noticed however that there are constructions which would be typically called "ergative" in a classical actancy role analysis in this language. The patient of a transitive process and the sole subject of an intransitive process receive the same grammatical treatment: (20)

Emea

'amu na rätou te mä'a tahiti. eat PREP 3 PL DET food Tahitian 'They eat Tahitian food (i.e. they appreciate it)'.

ΤΑΜ

In the example above, the second argument (without preposition) refers to the patient and not the agent (as would be expected) of the process 'amu, 'eat'. Without expanding this topic too far, we find it interesting to qualify the common statement about the "accusativity" of Tahitian.

2.

Nature of the relation between two juxtaposed processes

As seen above, we do not use the concept of "verb" in our metalinguistic framework. Nethertheless, to come closer to the topic of "serial verbs", we have tried to focus on two-lexeme strings of the type 'X Y', so that (i) the

256

Mirose Paia and Jacques Vernaudon

lexemes X and Y individually refer to a process, and so that (ii) the string 'X Y' globally refers to a process. In strings involving two processes 'Χ Υ', Y always modifies X. In example (21), huna qualifies how the action inu is being done. Such strings are common in Tahitian: (21)

Ε

inu huna 'ona i muri mai i te fare. drink hide 3SG PREP behind DIR PREPDET house 'He drinks secretly behind the house.'

ΤΑΜ

It is our speaker's intuition, confirmed by paraphrases, that allows us to assert that inu and huna individually both refer to a process in this string. As mentioned above, there is no morphological or distributional criterion to assure this. But (21) can be paraphrased as (22): (22)

Ε

inu 'ona ma te huna atu i te reira, i muri drink 3SG with DET hide DIR PREP DET ANAPH PREP behind mai i te fare. DIR PREP DET house 'He drinks hiding (for it) behind the house.' (lit. he drinks with the hiding it behind the house). ΤΑΜ

Notice that if we reverse the order within the string {inu huna inu), inu no longer refers to a process: (23)

huna

Ε

huna inu 'ona i muri mai i te fare. hide drink 3SG PREP behind DIR PREP DET house 'He hides drinks behind the house.'

ΤΑΜ

This time, inu refers to an entity (i.e. a liquid which someone drinks). This is a typical example of "object incorporation". The non-specific patient inu is incorporated into the predicate phrase just after the lexeme referring to the process. The following paraphrase of (23) confirms that inu is indeed an "incorporated patient": (24)

Ε

huna 'ona i te mau inu i muri mai i ΤΑΜ hide 3SG PREP DET PLUR drink PREP behind DIR PREP te fare. DET house 'He hides drinks behind the house.'

Consider the next example:

Complex predicates in Tahitian

257

(25)

Ua pure onoono ratou i te Atua ia tauturu 'oia ΤΑΜ pray insist 3pl p r e p d e t god ΤΑΜ help 3sg ia rätou. prep 3pl 'They have insistently prayed God that He help them.' If we consider pure and onoono independently, they can both refer to a process, respectively 'pray' and 'insist'. But in the example above, they do not designate two successive actions. Onoono simply qualifies how the process of praying (pure) is achieved (i.e. insistently). The corresponding paraphrase would be: (26) Ε mea onoono ta rätou pure i te Atua. ΤΑΜ insistent their prayer PREP DET god 'Their prayer to God is insistent.' On the other hand, juxtaposition of two lexemes X and Y referring to two processes cannot express succession or concomitance. In this case, other constructions are used, as in (30) to (32). Thus, the next strings cannot refer to two successive or concomitant actions: (27)

*fa'ahoro drive

(28)

*tai'o read

täniuniu telephone

'amu'amu nibble

(29)

*hJmene ta'i sing cry But one would say: (30) Ε fa'ahoro 'onama te täniuniu. ΤΑΜ drive 3SG with DET to telephone 'He telephones while driving.' (31)

(32)

Ε tai'o 'ona a 'amu'amu noa ai. ΤΑΜ read 3SG ΤΑΜ nibble only ANAPH 'He nibbles while reading.' Ε himene rätou na roto ι te ta'i. τ α μ sing 3pl prep inside p r e p d e t cry 'They cry while singing.'

Additional morphemes are here necessary to indicate that the two actions are concomitant. The first construction (i.e. x ma te Y) uses the preposition ma 'with'. Notice that the lexeme Y referring to the second process must be preceded by the "article" te. The second construction (i.e.

258

Mir ose Paia and Jacques

Vernaudon

X a Y ai), as in (31), suggests that the onset of process X does not interrupt process Y. The ΤΑΜ morpheme a is an inceptive marker, the anaphoric morpheme ai (lit. 'then') roughly indicates the concomitance of the two processes X and Y, and noa suggests a continuous Y: process Y is ongoing even though X occurs at the same time. The third construction (i.e. X na roto i te Y), quite common in contemporary Tahitian, may be a loan construction from French en + V[gerund], Other constructions refer to several successive actions. However, the lexemes designating the different actions cannot be simply juxtaposed, without at least a pause between them. Moreover, arguments (agent, patient, etc.) can be made explicit at any time. Consequently, those structures are not "serial verbs" but multi-clause constructions referring to chains of actions, as in (33). (33)

Ε

höhore

'oe i

ΤΑΜ p e e l

na

'δ

te

pa'a,

2SG PREP DET bark

noa e

mä,

roto

PREP inside

piti hepetoma, tätara,

PREP thus only ΤΑΜ two week

ia

tu'u i put

remove

i

te

vari,

PREP DET m u d

hororoi, rinse

tära 'i.

ΤΑΜ clean spread out in the sun

'You peel the bark, then put (it) into the mud, thus two weeks, then (you) remove (it), rinse (it), (and) when (it) is clean, (you) spread (it) out in the sun.' (PAA: 10) To express dependency between clauses, Tahitian uses the imperfective aspect marker e as in (34) or the preposition i followed by the "article" te to introduce the subordinate clause (see section 4 for other cases of subordination): (34)

A

haere e

ΤΑΜ go

ti'i

i

te

faraoa.

ΤΑΜ fetch PREP DET bread

'Go and get some bread.' (but never *A haere tii i te faraoa) (35)

Ε

hina'aro

TAM want

vau i

te

lSG PREP

DET leave

reva

Ί want to leave.' (but never *E hinaaro reva vau.)

3.

Types of qualitative modification

Lazard & Peltzer (2000: 25) distinguish a particular syntactic function in Tahitian which they call "qualificatif' (in French). This concept labels any lexeme which follows and modifies another lexeme. In the present paper,

Complex predicates in Tahitian

259

we would rather use the term "modifier" to refer to this particular function, by which the modifier qualifies the preceding term. (36) te ta'ata rahi 'the man who is big'

(37)

DET human Modified

big Modifier

te

moni

fare

DET house Modified

'the bank'

money Modifier

In contrast to the string ta'ata rahi, which is not lexicalized, /are moni is a compound, just like fare rä'au 'pharmacy' (rä'au 'vegetal, medicine') or fare rata 'post office' (rata < Engl, 'letter'). However, in both examples, the second lexeme (i.e. rahi and moni) modifies the first one (i.e. ta'ata and fare). Some modifications can be associated with a gradient while others cannot. Thus, the man to which te ta'ata rahi refers can be more or less rahi 'big', but the building te fare moni is not more or less moni, even though the lexeme moni (< Engl, 'money') may be compatible with a gradient in other contexts, then meaning 'expensive': (38)

Ε mea moni roa teie fare. ΤΑΜ expensive HD this house 'This house is very expensive.'

But in (37), moni refers exclusively to 'money'. In short, a fare moni is not a fare which is expensive, but a fare containing money. In strings comprising more than two lexemes, the additional lexeme can either modify the one before (see (39)), or the entire preceding phrase (see (40)). (39)

te

ta'ata

DET human

horo

vitiviti

run Modified

fast Modifier

Modified

(40)

te

fare

DET house Modified

'the guy who runs fast'

Modifier

moni

'äpl

money Modifier

new

Modified

'the new bank'

Modifier

It is commonly admitted that the modifiers do not belong to particular lexical classes. Any lexeme can have this function, regardless of what it refers to (process, entity, quality, etc.).

260

Mirose Paia and Jacques Vernaudon

Since strings of two processes in Tahitian always seem to be a case of modification of the first one by the second one, we can then consider huna in inu huna and onoono in pure onoono as modifiers, just like rahi in ta 'ata rahi and moni in fare moni. In other words, strings of two processes in Tahitian are just a special case of a larger procedure of modification which concerns both predicate phrases and arguments. inu huna 'to drink secretly.' drink hide

4.

huna hide

inu drink

'to hide drinks.'

mä'iti choose patu build

rima hand timä cement

'to vote by a show of hands.'

piri stick

fe'e octopus

'to stick like a limpet (lit. like an octopus).

'to build with cement.'

Evidence for the modifying relation between X and Y

Rewording and paraphrases of utterances containing strings of two processes 'X Y' reveal different types of modification, expressed by different markers, depending on the exact nature of modification. -

' x ma te Y', where Y qualifies the manner of process x (see (21) and (22)): inu huna —> inu ma te huna i te reira drink hide

-

drink with

DET hide

PREP

DET ANAPH

'e mea Y te X', where Y evaluates the way the process X is done (see (25) and (26)): pure onoono —> \ e mea \ onoono te pure pray

insist

ΤΑΜ

insist

DET pray

- 'X no te Y', where Y explains the reasons or purpose for starting X: (41) Ua vävähi fa'a'äpl 'ona i to na fare. ΤΑΜ destroy

renew

3sg

PREP his

house

'He has pulled his first house down to build a new one on the same foundations.' (lit.: He has destroyed-renewed his house)

Complex predicates in Tahitian (42)

Ua vavahi 'ona i ΤΑΜ destroy

3SG

to na fare

PREP his

no te

fa'a'api i

h o u s e PREP DET r e n e w

261

te

PREP DET

reira. ANAPH 'He has pulled his house down to renew it.' -

'X ia Y', w h e r e Y indicates the result aimed at b y the agent-initiator of the process X:

(43)

Ua taparahi pohe rätou ia na. beat die 3 PL PREP 3SG 'They beat him to death.'

ΤΑΜ

(44)

-

Ua taparahi rätou ia na [ia] pohe roa ΤΑΜ beat 3PL PREP 3SG ΤΑΜ die HD 'They beat him so that he will die.'

'ona. 3SG

'X i te Y'

The lexeme ha'avare, 'lie, deceive', can occur in a string with another process with the meaning 'pretend'. For example: ha'avare 'amu 'pretend to be eating', ha'avare ta'i 'pretend to be crying'. The only reformulation of such a string is with the ha'avare i te Y construction as in (46). (45) Ε TAM

ha'avare

ta'oto

lie

sleep

'o Teri'i. Terii

'Terii pretends to be asleep.' (46) Ε ΤΑΜ

ha'avare 'o

Teri'i

lie

Terii

i

te

ta'oto.

PREP DET

sleep

'Terii feigns sleep.' Sometimes, no paraphrase can reformulate the string 'X Y', as in (47) when the lexeme tä'ue is used as a modifier. (47) Ua topa tä'ue te ua. TAM fell suddenly DET rain 'It suddenly started to rain.' In this context, tä'ue means 'suddenly'. But, by itself, tä'ue means 'throw'. (48) Α tä'ue atu i tena vz, ua pe. TAM

throw

DIR

PREP

this

mango

TAM

'Throw away this mango, it is rotten.'

rotten

262

Mirose Paia and Jacques Vernaudon

In the example above, tä'ue alternates with fa'aru'e 'to get rid o f . Such a substitution is impossible in (47) where tä'ue alternates with the few lexemes which can only be used as modifiers of processes, such as 'oi'oi 'quickly, early', or vave 'hastily'. Thus, even if tä'ue used as a modifier comes from tä'ue 'throw', they have now two different meanings with only the seme of suddenness in common. This is a case of grammaticization.

5.

Conclusion

Strings of two lexemes 'X Y' globally referring to a single process are frequent in Tahitian. Each constituent, X and Y, can individually denote a process. Such strings always refer to a unique action described by the first lexeme X, while the second lexeme Y provides additional qualifications and specifications about X. It is a particular case of modification following the order modified - modifier. This procedure is used both in the predicate phrase and in argument phrases.

Orthographic conventions and abbreviations The grapheme " ' " represents the glottal stop and macron on vowel is for length (vowels are either short or long). anaphoric determiner

high degree marker plural

ANAPH:

HD:

DET:

PLUR:

preposition identification marker

PREP: ID:

Notes 1.

2. 3.

The Tahitian language (reo tahiti or reo mä'ohi) is part of the Eastern group of Polynesian languages. Although it is mainly spoken in the Society Islands archipelago, in French Polynesia, it is also used together with French as a lingua franca throughout this territory (245,400 inhabitants). The number of people who speak Tahitian fluently is difficult to determine. According to the most optimistic figures, they represent 70% of the total population. An indepth linguistic survey would probably reveal a more complex picture of the situation. "Process" is used as a generic designation for action, event or state. See for example Lazard, 1994: 260.

Complex predicates and Juxtapositional Constructions in Samoan

Ulrike Mosel

Abstract In Samoan, complex predicates can contain constructions of two or more contiguous lexical words which correspond to noun incorporation, serial verb constructions, depictive secondary predicates, and endocentric modifier, possessive and copulative constructions or compounds in other languages. There are, however, two remarkable differences: Firstly, while noun phrases and verb complexes are well defined by the presence of articles and tense/aspect and mood particles, respectively, nouns, adjectives and verbs cannot be distinguished by morphological or distributional criteria of classification; put differently, noun phrases and verb complexes cannot be defined as projections of nouns and verbs. Consequently, the various types of the constructions of juxtaposed content words, henceforth called Juxtapositional Constructions, are not defined by morphosyntactic characteristics, but are solely established by the lexical features of the content words involved. Secondly, Juxtapositional Constructions not only function as predicates, but also as the nucleus of noun phrases, as modifiers in noun phrases and as components of higher ordered Juxtapositional Constructions. This multifunctionality of Juxtapositional Constructions within noun phrases and verb complexes shows that they form independent syntactic units below phrase level.

1.

Introduction

S a m o a n is a Polynesian ergative language with s o m e derivational, but n o inflectional m o r p h o l o g y . T h e u n m a r k e d constituent order of simple verbal clauses is: predicate plus arguments. T h e order of a r g u m e n t s after the predicate is variable. Samoan is a strictly configurational language in w h i c h different types of phrases - n o u n phrases, verb c o m p l e x e s and prepositional phrases - are defined b y the occurrence of certain grammatical particles and their position. The word order within phrases is fixed. W h i l e tense/ aspect and m o o d markers precede the nucleus of verb c o m p l e x e s and

264

Ulrike Mosel

articles the nucleus of noun phrases, most modifiers follow the nucleus in both types of phrases. The syntactic functions of noun phrases are expressed by prepositions; only the primary argument of intransitive verbal clauses and the undergoer argument of transitive verbal clauses are unmarked. Complex predicates 1 can contain two or more contiguous lexical words. These constructions are similar to what is called noun incorporation, serial verb constructions, depictive secondary predicates, and modifier, possessive and coordinate constructions in other languages. Table 1. Juxtapositional Constructions Example tausi take.care.of

Gloss pepe baby

Type of construction

'take care of babies, baby-sit'

noun incorporation

'cry and talk at the same time'

coordinate construction

lamu fa 'a-malu chew CAUS-soft

'chew sth. so that it becomes soft'

SVC

nofo sit

'sit properly'

modifier construction

'eat s.o. alive'

depictive construction

'have a flat nose'

possessive construction

tagi cry

'ai eat isu nose

tautala speak

lelei good ola live papa flat

For Samoan, however, the analysis and description of these constructions faces the problem that we cannot establish the word classes or syntactic categories of noun, verb, and adjective on the basis of morphological or distributional syntactic criteria. Consequently, to speak of noun incorporation or serial verb constructions would be inadequate. 2 Nevertheless, collocations of two words as those in Table 1 do not represent idiomatic lexicalized constructions, but productive types of constructions which are definable in terms of the lexical properties of their component parts. Each type is formed by words of particular lexical features, so that the meaning of a complex expression of this kind can be derived from the lexical features of its parts. Such constructions are, however, not restricted to predicates. They can also function as the nucleus of argument phrases or as modifiers. The present paper will first discuss the dilemma of word classification in Samoan in section 2 and then, in section 3, it will offer a tentative

Complex predicates and Juxtapositional Constructions in Samoan

265

classification of words on the basis of lexical properties. Section 4 describes the structure of complex predicates. In section 5, we present a detailed analysis of the six types of Juxtapositional Constructions as they are used in complex predicates. Section 6 shows that Juxtapositional Constructions not only occur in predicates, but are as multifunctional as single lexical words in that they can also function as the head of noun phrases, as modifiers, and as components of superordinate Juxtapositional Constructions. These findings suggest that they form linguistic units below phrase level which in spite of the lack of any overt morphological marking can be called constructions, i.e. Juxtapositional Constructions.

2.

Problems of word classification in Samoan

The distinction of word classes, in particular the noun/verb distinction, is a notorious problem of Polynesian linguistics 3 and cannot be discussed here in detail. In Samoan, lexemes do not inflect for grammatical categories such as tense or case. The only grammatical category which is morphologically marked is number, but the marking of number is restricted to a closed class of lexemes and seems to be idiosyncratic. The basic syntactic unit below clause level is the phrase, which consists of a lexical nucleus and grammatical particles such as tense/aspect markers and articles. These grammatical particles determine the boundaries of the phrases within the clause and also allow one to classify them. For Samoan we distinguish: 1. verb complexes (VCs) which are formed by a nuclear lexeme preceded by tense, aspect and mood particles and followed by modifiers, e.g. in (1) sä tagi 'cried', in (2) 'ua leaga '(is/are/became) bad'; 2. noun phrases (NPs) which are mostly 4 introduced by an article, e.g. in (1) le teine, in (2) le laupapa

3.

'the p l a n k ( s ) ' ;

prepositional phrases (PPs) which consist of a preposition and a noun phrase, e.g. in (2) ο le fale 'of the house'. For practical reasons we maintain the traditional terms of verb complex and noun phrase, although theses types of phrase are not defined as a projection of their nucleus, but by the grammatical particles they take as satellites. (1)

Sä

tagi

le

teine.

ΤΑΜ cry art girl 'The girl cried.'

266

Ulrike Mosel

(2)

'Ua leaga le laupapa ο le fale. ΤΑΜ bad ART plank5 of ART house 'The planks of the house are bad.'

On the basis of their syntactic function, lexemes like leaga, laupapa and fale cannot be classified as verbs and nouns, because one and the same lexeme can function as the head of a verb complex and a noun phrase. When writing the Samoan teachers' grammar Ο le Kalama ο le Gagana Samoa and the Samoan monolingual students' dictionary Utugagana, the Samoan authors (Fosa Siliko, Ainslie So'o, Agafili Tuitolova'a) discussed the classification of words in various language panels and workshops 6 with Samoan teachers and arrived at the conclusion that all content words can be used as the lexical heads of noun phrases and verb complexes, and also as modifiers. Compare (1) and (2) with (3) - (5). (3)

Ν a vave 'uma le tagi a Susana. ΤΑΜ quick finished ART cry of Susana 'The crying of Susan was quickly finished.' (Mosel et al 1999: 29)

(4)

lona fale laupapa his house plank 'his wooden house' (lit. 'his plank house')

(5)

Ε laupapa lona fale. ΤΑΜ plank his house 'His house is made of planks.' (lit. 'His house is plank') (Mosel et al. 1999: 30).

Consequently, a classification of words by the criterion of syntactic function will not lead to word classes comparable to nouns, verbs, and adjectives in European languages. In the description of Samoan, the terms 'verb', 'noun' and 'adjective' can only be used for content words in a particular syntactic function, i.e. in the function of the head or the modifier in a verb or a noun phrase. But in this sense, 'noun', 'verb' and 'adjective' do not denote word classes, but syntactic functions. 7 This not to say that all content words allow the same constructions. For example, content words differ in their valency. When functioning as the head of a verb complex, only particular words, most of which express actions, can govern an ergative argument expressing the agent and an absolutive argument expressing the patient, e.g.

Complex predicates and Juxtapositional Constructions in Samoan

(6)

Sa

tausi

e

ΤΑΜ take.care

le

teine

ERG ART girl

le

267

pepe.

ART baby

'The girl took care of the baby.' When functioning as the head of a noun phrase, the ergative N P can be maintained or be replaced by a possessive phrase: (7)

'o

le

tausi

e

le

teine

PRES

ART

take.care

ERG

ART

girl

ο

le

POSS

ART

/ /

a le teine poss ART girl

pepe baby

'the taking care of the baby by the girl' Since the very same (lexical) word can function as the head of a noun phrase, we should not call it a transitive verb, but a transitive content word or lexeme. The corresponding class of intransitive content words contains all other content words, which cannot combine with an ergative argument, including the translation equivalents of English nouns like laupapa 'plank' in (2, 4, 5). In other words, it is possible to identify a word class corresponding to English transitive verbs, but in Samoan these words do not form a subclass of 'verbs'. In the same way, the selection of tense/aspect markers or the possibility to combine with directional markers allows us to identify subclasses of content words, but again, these subclasses do not relate to the traditional European classes of nouns and verbs. The same holds true for derivational morphology. Even words carrying the so-called nominalization suffix can occur in the head position of a verb complex, e.g. (8)

'Ua

to-gä-niu

'ätoa

le

ΤΑΜ

plant-NOM-coconut

whole

ART place

mea

maupu 'epu 'e. hill

'The whole hilly place was now a coconut plantation.' (Aiavao 1987: 10) Thus, Samoan is different from European languages in that the noun phrase and the verb complex cannot be considered as a projection of the word classes noun and verb.

268

3.

Ulrike Mosel

Lexical classes

Sequences of two words like those listed in Table 1 are not arbitrarily lexicalized constructions, but represent distinctive and productive types of construction. Compare: Table 2. Examples of six types of Juxtapositional Constructions Example

Literal translation

Free translation

fa 'atau ipu fufulu ipu fufulu lima

buy cup

'buy cups' 'wash cups', 'wash the dishes'

'umi uliuli moe tagulu nofo tatalo

long black

'long and black'

sleep snore sit pray

'sleep and snore' 'sit and pray'

3

fau fa 'amaualuga 'elifa 'amaualalo tipi fa 'apa w

build make-high dig make-deep cut make fall

'build high' 'dig deep' 'cut (trees) down'

4

iloa lelei

know good

'know well'

ola lelei tali sa 'o

live good

'live well'

answer straight

'answer truthfully'

5

moe väiväi folo 'ätoa nofo to 'a-tasi

sleep exhausted swallow whole stay (human)-one

'sleep exhausted' 'swallow s.th. whole' 'stay by oneself

6

loto väiväi nifo tele 'ofu pa 'epa 'e

heart weak tooth big dress white

'have/with a weak heart' 'have/with big teeth' 'have a white dress, dressed in white'

1

2

wash cup wash hand

'wash the hands'

These types of construction are not distinguished from one another by any formal means of expression but by the various kinds of semantic relationships between the juxtaposed words, e.g. the relationship between a transitive action and its patient (type 1) or an action and the bodily condition or the number of participants (type 5). The various kinds of relationship seem to be determined by the lexical properties of the juxtaposed words. Therefore, I try to determine these lexical properties and classify the words accordingly. The resulting lexical classes bear some similarity with Dixon's "semantic types":

Complex predicates and Juxtapositional Constructions in Samoan

269

"The words of any language can be grouped into a number of lexical classes called SEMANTIC TYPES, which have a common meaning component and some shared grammatical properties." 8 (1995: 175)

As I only want to show the principles which govern juxtaposition, the classification suggested in Table 3 does not cover the whole lexicon of the Samoan language. This semantic classification is supported by lexical and grammatical characteristics of various kinds such as: - only ACTION words can enter the negative imperative construction 'aua -

-

-

le ACTION (DON'T DO ART ACTION) ' D o n ' t do ACTION!' 9 TRANSITIVE A c T i o N - w o r d s , but not INTRANSITIVE ACTION w o r d s can

take the so-called Cia-sufflx (Mosel & Hovdhaugen 1992:741 ft); PERSON words are distinguished from all other words by the use of the numeral classifier to 'a, which is prefixed to numerals and tele 'many' when persons are counted 10 ; only NUMBER-words can take the prefix ta 'i- 'each'; interrogative words and constructions, e.g. ai 'who' for PERSON, fia ' h o w m a n y ' for NUMBER- and QUANTITY; fea

' w h e r e ' f o r PLACE,

anafea 'when in the past', 'äfea 'when in the future' for TIME, 'O le ä lau mea na fai? (PRES ART what your thing ΤΑΜ do) 'What did you do? (lit. What is your thing (you) did?) for ACTION, 'o le ä le lanu? 'What is the c o l o u r ? ' for COLOUR, etc.

-

only PART, BODY-PART and CLOTHING words are used in possessive constructions (cf. below).

For the description of some rules of combination, the PROPERTY words are subclassified into COLOUR, DIMENSION, VALUE and SHAPE, and a residual class of PROPERTY words which have not been subclassified yet. There are also some classifications for which we have not yet found any justification which would be independent of Juxtapositional Constructions. A case in point are the QUANTITY words 'uma 'all' and 'ätoa 'whole' and the PART-OF words (including body-part terms and words for clothing) which are distinguished from other words by their occurrence in Depictive and Possessive Constructions, respectively.

270

Ulrike Mosel

Table 3. Lexical classes Examples

Semantic Class (in small caps) ACTION

MENTAL ACTIVITIES

Intransitive Action

'ai 'eat', tagi 'cry', masüsü 'sob', nofo 'sit',

Transitive Action

'ai 'eat s.th.', tipi 'cut s.th.', tusi 'write s . t h . \ f u f u l u 'wash (s.th.)'

Perception

va 'ai 'see',Τα 'alogo 'hear'

Cognition

iloa 'know, recognize', mafaufau 'think, reflect', manatu 'think (of s.th.)', mälamalama 'understand'

BODILY CONDITION

tauaso 'blind', väiväi 'exhausted', ola 'alive'

PROCESS

tupu 'grow, happen', pa 'ü 'fall', pe 'die' Colour

samasama 'yellow', mumu 'red', pa 'epa 'e 'white', uliuli 'black'

Value

lelei 'good', leaga 'bad', mänaia

Dimension

'umi 'long', pu 'epu 'e 'short', läpo 'a 'large, big', tele 'big, much', läitiiti 'small, little', maualuga 'high', maualalo 'low'

Shape

läpotopoto 'round'

Unclassified

ta'e 'broken',palapalä 'dirty'

PROPERTY

'nice'

QUANTITY

'ätoa 'whole, 'uma 'all'

NUMBER

tasi 'one', lua 'two', tolu 'three',...

SPATIAL RELATION

mamao 'distant', latalata 'near'

TEMPORAL RELATION

latalata 'near, soon', muamua 'first', mulimuli 'later'

TIME

pö 'night'

THING (physical object)

laupapa 'plank, board', lima, 'hand, arm', ulu 'head', 'ofu 'piece of clothing', pülou 'hat'

P A R T ( o f a w h o l e ) , B O D Y PART

(of a person or animal)

lima, 'hand, arm', ulu 'head'

CLOTHING (s.th. which is on your body and is like a body part)

'ofu 'piece of clothing', pülou 'hat'

PERSON

tagata 'person',pepe 'baby', lo 'omatua 'old woman'

Complex predicates and Juxtapositional Constructions in Samoan

271

Our classification differs from classical categorization and Dixon's classification" in that it allows multiple membership of elements. An instance of multiple membership are, for example, the words belonging to PART-OF class, because they also belong to the THING class. PART-OF words can enter the same constructions as other THING words (8), but not all THING words can enter Possessive Constructions (10). (9)

fufulu lima fufulu ipu

'wash hands' 'wash dishes'

(10)

lima lelei ipu lelei

'good hand, have a good hand, be skilled' 'good cup, *have a good cup'

As for the subclasses, we find multiple membership with a number ACTION words which like 'ai 'eat' can denote intransitive and transitive actions and, correspondin+gly, occur in both intransitive and transitive constructions (cf. labile verbs in Mosel & Hovdhaugen 1992: 108 ff). In addition to the classes listed above, three classes of morphologically and semantically derived lexical units are found as the second element in Juxtapositional Constructions. All of them are derived by fa'a- which represents three distinct homophonous morphemes: the causative prefix, the similative prefix and the associative prefix. Table 4. Words derived by fa 'aSemantic class

CAUSATIVE

Subclass

Example

ACTION CAUSATIVE

fa 'a 'ata

'make laugh'

PROCESS CAUSATIVE

fa 'a-pa u

'make fall'

Value

fa 'a-lelei

'make good'

PROPERTY

Dimension fa 'aläpo 'a

'make big'

Shape

fa 'a-läpotopoto

'make round'

fa 'a-lua

'make two' 'a second time'

fa 'a-tamaitiiti fa 'a-mauga

'like a child' 'like a mountain'

fa 'a-le-atunu 'u

'having to do with the country, national'

CAUSATIVE

FREQUENTATIVE (NUMBER CAUSATIVE) SIMILATIVE

ASSOCIATIVE

LIKE PERSON / THING

ASS-ART-country

272 4.

Ulrike Mosel The structure of complex predicates

In addition to their lexical head and a ΤΑΜ particle, verb complexes can contain a prenuclear clitic pronoun, a prenuclear negation, a postnuclear anaphoric pronoun, a postnuclear directional particle, and pre- and postnuclear adverbs and postnuclear emphatic particles. Adverbs are words of a small closed class which exclusively occur in various fixed positions in the verb complex. Due to their fixed positions, the particles, pronouns and adverbs mark the boundaries between the verb complex and adjacent phrases, e.g. the directional particle mai 'hither' in (11). ΤΑΜ

(11)

ACTION

...a Jo tagi but ΤΑΜ cry

ACTION

masüsü sob

ACTION

fa'alogologo lean

PARTICLE

mai hither

Saraiatea i lo 'u falafata. Saraiatea at my chest '... Saraiatea leaned crying and sobbing against my chest.' (Tuitolova'a 1985: 30) When, as in (11), a verb complex contains more than one lexical word, we speak of a complex predicate or verb complex. In such verb complexes, the lexical words are contiguous. By definition, all lexical words can also occur independently and form the nucleus of phrases. There are two kinds of constructions which at first sight look similar to verb complexes: 1. complement clause constructions of words like 'ämata 'begin', iloa 'know', and 2. the construction of alu 'go' with a purposive clause. In (12) for example iloa 'a 'au 'know (how to) swim' looks like a verb complex, but (13) shows that the two lexical words can be separated by an adverb and consequently do not form a single verb complex according to the definition given above. (12)

... 'ou te le iloa 'a 'au. I ΤΑΜ not know swim '... I don't know how to swim.' (Milner 1966: 48)

(13)

e le iloa fo'i ta 'u le 'upu. TAM not know also say ART word '[But if they do not know the meaning] they also do not know how to say the word.' (Faugagana, June/July 1999: 3)

Rather, the second A C T I O N word (here 'a'au 'swim' and ta 'u 'say') functions as a second predicate which independently of the first predicate

Complex predicates and Juxtapositional Constructions in Samoan

273

(iloa 'know') can have its own arguments (le 'upu 'the word'). An example of the construction of alu 'go' with a juxtaposed purposive clause is (14). (14)

Alu

fa'atau

se

go

buy

ART sugar

suka.

'Go and buy some sugar.' This construction is an elliptical variant of alu e fa 'atau (go ΤΑΜ buy), in which the second ACTION word (fa 'atau) is marked by a ΤΑΜ particle (e). Therefore, alu fa 'atau is not considered as a complex verb complex, but as a sequence of two verb complexes, the second one (e fa'atau) being the predicate of a purposive clause.

5.

Types of Juxtapositional Construction in complex predicates

In order not to complicate things, we will start the classification of constructions of only two lexical words. The criteria of classification are the various kinds of syntactic and semantic relations that hold between the first and the second element. The nature of these relations is determined by the lexical properties of the words and becomes evident when the complex constructions are broken up and replaced by simple constructions. Since the present volume is mainly concerned with serial verb constructions, we shall only briefly touch upon constructions with argument incorporation and possessive constructions. The comparison of all construction types and their components reveals that each construction type is formed by a particular combination of lexical classes.

5.1. Constructions with argument incorporation Most examples of constructions with argument incorporation have an ACTION word as their first component and as their second one a PERSON or THING word denoting the patient, PLACE or TIME of the action. Which kind of thematic role the incorporated word plays is determined by the valence and the semantics of the ACTION word. The incorporated words differ from noun phrase nuclei in that they cannot be combined with determiners such as demonstratives or possessive pronouns.

274

Ulrike Mosel

Table 5. Argument Incorporation. Lexical class

Thematic role

fa 'atau i 'a

'sell fish'

THING

PATIENT

fa 'amätagä tagata

'disfigure people'

PERSON

PATIENT

fatu pese

'compose song'

THING

PATIENT

ti 'eti 'e solofanua

'sit (on a) horse, ride'

THING (ANIMAL)

PLACE

ti 'eti 'e ta 'avale

'sit in a car'

THING

PLACE

tu 'u 'apa

'put s.th. (into) tins, tin'

THING

PLACE

The incorporation of the PATIENTS differs from that of PLACES in that the former changes the valence of the verb complex, whereas the latter does not. In contrast to (6), the verb complex sä tausi pepe 'took care of babies' in (15) changes the argument structure of the clause. 12 Compare (6) with (15): (15)

Sä tausi pepe le teine. ΤΑΜ take.care.of baby ART girl 'The girl took care of babies / was a baby-sitter.'

(15) is intransitive and consequently has an unmarked, absolutive noun phrase {le teine 'the girl') as its agent (cf. Mosel & Hovdhaugen 1992: 392396). Verb complexes with an incorporated PLACE adjunct or complement, on the other hand, require a PATIENT argument if the nucleus of the verb complex is transitive: ( T R ACTION + THING PLACE ) TR

(16)

... le fale

e

tu 'u

'apa

ai

i'a

house ΤΑΜ put can ANAPH fish: PL 1 3 '... the house where the fish is canned.' (lit. 'the house put (into) can there the fish' (Leauga n.d.: 7) ART

5.2. Coordinate Constructions Coordinate Constructions consist of two lexical words of equivalent syntactic status which express two simultaneous states of affairs. They are frequently found in dictionary definitions describing the colour, size and shape of animals and plants. The sequential order of property words can vary e.g.:

Complex predicates and Juxtapositional Constructions in Samoan

275

COLOUR DIMENSION

(17)

'o

le

'au

PRES ART stalk

ο

le

lau

POSS ART leaf

e

uliuli

ΤΑΜ black

pa 'e 'e. thin

'the stalk of the leaf is black and thin.' (Mosel & So'o, 2000: 146) DIMENSION COLOUR

(18)

'o

le

itu'äiga i'a e 'umT uliuli kind fish ΤΑΜ very.long black '...a kind of fish that is very long and black ...' (Mosel & So'o, 2000: 99)

...

PRES ART

In (19) COLOUR words are put together to denote a single mixed colour, e.g.: COLOUR + COLOUR

(19)

Ugä v.,

'e 'ena

ugä

brown: PL yellow

samasama

nifo. teeth: PL

'Ugä (v.)14: the teeth are yellow-brown.' (Mosel & So'o, 2000: 28) Other words of the same subcategory of PROPERTY words are usually not juxtaposed, but coordinated by the conjunction ma 'and' e.g.: DIMENSION + DIMENSION

(20)

...e

mäfiafia

ΤΑΜ thick

ma

läpo 'a

le

'afa,

and

large

ART string

tauaga,

m.f.

strainer

etc.

'... the string, strainer, etc. is thick and large.' 15 (Mosel & So'o, 2000: 87) While coordinated PROPERTY words frequently occur in the dictionary, we have only very few examples of ACTION + ACTION constructions. Here are two more in addition to (11): ACTION + ACTION

(21)

Ό

lea

na

tagi tautala ai loa le tuna PRES DEM ΤΑΜ cry talk ANAPH then ART eel 'And so, the eel cried and spoke to Sina ...' (Mosel & So'o, 1997: 135)

iä

Sina DEM Sina

276

Ulrike Mosel

ACTION + ACTION

(22)

ο

le

leo

PRES ART sound

e

fai

i

ΤΑΜ

make

LOC ART

fa 'aläpotopoto

laugutu

CAUS-round

lip:PL

le

gutu mouth

'[Whistle]: the sound made in the mouth when protruding the lips and making them round...' (Mosel & So'o, 2000: 81). With words denoting actions, the sequential order of constituents can be determined by the sequence of events. The word nofo 'sit, sit down, stay', for instance, always occurs as the first component as people first sit down and then do something while being seated. (23)

Olo'o 'ua taufai nofo tatala tagata e to'atele. ΤΑΜ ΤΑΜ many sit pray person: PL ΤΑΜ many 'Now many people were sittting there /had sat down and were praying.' (Petaia 1996: 62)

This pattern has been conventionalized, it is also used to express new concepts, e.g. nofo totogi 'stay (and) pay', i.e. 'stay in a house or flat and pay rent, be a tenant'.

5.3. Cause-Effect Constructions Cause-Effect Constructions express that a transitive action causes a person or thing to have a certain property, e.g. TRANSITIVE ACTION + PROPERTY CAUSATIVE

(24)

... e

fa'a-aogä

e

lamu fa 'a-malü ai mea 'ai. τ α μ CAUS-used ΤΑΜ chew make-soft ANAPH thing eat ' [molars are broad and big teeth] which are used to chew the food soft.' (Mosel & So'o, 2000: 62) The anaphoric pronoun ai in the example above marks the right border of the v c . Thus fa 'amalü holds the second position in a Juxtapositional Construction and does not form a verb complex on its own comparable to fa'atau 'buy' in alu fa'atau 'go to buy' (cf. 14). All examples of this type have a TRANSITIVE ACTION word as their first element and a CAUSATIVE word as their second element.

Complex predicates and Juxtapositional Constructions in Samoan

211

ACTION + CAUSE-VALUE

(25)

Suga,

alu

e

tapena

fa 'a-lelei

le

fale

girl,

go

ΤΑΜ

tidy.up

CAUS-good

ART

house

'Girl, go and tidy up the house so that it looks good.' (lit. 'tidy up (to) make good') (Taulelei 1986: 7) ACTION TR + CAUS-DIMENSION

(26)

'oti 1. v.,

sele

fa 'a-pupu 'u

le

lauulu

i

se

seleulu

...

cut CAUS-short: PL ART hair with ART scissors ' 'oti (l.v.): cut the hair short with scissors ...' (lit. 'cut (to) make short') (Mosel & So'o, 2000: 26) Further typical examples of the ACTION + CAUSE-PROPERTY type are (Mosel & So'o 2000): (27)

tipifa 'anini 'i 'cut into small pieces' (p. 73) su 'ifa 'a 'umi 'sew and make long (a necklace)' (p. 28) fau fa 'amaualuga "build high' (p. 90) 'elifa'amaualalo 'dig deep' (p. 131) fau fa'aläpotopoto 'build round' (p. 13) su 'ifa 'amau 'sew fast' (p. 65) There is one example in which an action does not cause the affected participant to have a new property but to undergo a process: ACTION + PROCESS CAUSATIVE

(28)

Tipi

fa 'a-pa 'ü

cut

CAUS-make.fall ART tree

le

lä'au.

'Cut the tree down.' (Mosel & So'o, 2000: 143) In Cause-Effect Constructions, both components are transitive and share their arguments. Both could be used by themselves with the absolute argument yielding clauses with similar meanings, while with Modifier Constructions such a splitting would result in semantically completely different clauses (cf. section 5.4). Compare (25) and (29): (29)

Tapena le fale, fa 'a-lelei le fale. tidy.up ART house CAUS-good ART house 'Tidy up the house, make the house g o o d '

With respect to argument sharing, Cause-Effect Constructions are similar to Coordinate Constructions and can be regarded as a subtype of the latter.

278

Ulrike Mosel

5.4. Modifier Constructions These constructions are most frequent and show the greatest variety of combinations. They refer to an event, done in a certain way, near or far away, or a certain number of times, e.g. ACTION + VALUE

(30)

Ό

le

toea 'ina

PRES ART old.man

ma

le

and

ART old.woman

olomatua

Samoa, Samoan

e le gata e tausi lelei e le 'äiga, ΤΑΜ NOT only ΤΑΜ care.for good ERG ART family 'The old Samoan man and woman are not only well cared for by the family (but also by the neighbour).' (Aiavao 1987: 102) PROCESS + VALUE

(31)

Ό

le

aso manino e susulu lelei ai le lä... PRES ART day clear ΤΑΜ shine good ANAPH ART sun Ά clear day on which the sun shines nicely.' (Mosel & So'o, 2000: 68) The VALUE word lelei 'good' is frequently found with COGNITION words, e.g. iloa lelei 'know well', manatua lelei 'remember well', etc. ACTION + DIMENSION

(32)

Sä moe 'umi le tama. ΤΑΜ sleep long ART boy 'The boy slept long.'

The next example shows two ACTION + DIMENSION/ QUANTITY constructions; the first one is a VC without ΤΑΜ marking, while the second one is marked by an article and hence forms a NP. ACTION + DIMENSION/ QUANTITY

(33)

Miti 6. v., inu-inu la 'itiiti 'ae le 'o se inu tele. sip drink-drink little but not PRES ART drink much 'Sip (6. v.): drink repeatedly a little, but [it is] not [a] drinking much.' (Mosel & So'o, 2000: 89)

Compare the frequently used phrase 'ai tele 'eat much', e.g. ACTION + DIMENSION/ QUANTITY

(34)

Sä

[ai

ΤΑΜ eat

tele

le

much

ART boy

'The boy ate a lot.'

tama.

Complex predicates and Juxtapositional Constructions in Samoan

279

At first sight, one could think that la'itiiti 'little' and tele 'much' are incorporated arguments, but the test for argument incorporation does not work: (33) and (34) cannot be transformed into clauses in which la'itiiti or tele would function as an argument, e.g. * 'ai le tele, * 'ai i le tele. FREQUENTATIVE numerals modify ACTION and PROCESS words and express the number of times that someone does something or that something happens. TRANSITIVE ACTION + FREQUENCY (35)

...e

tusi

fa 'a-afe

le

fuai'upu

ΤΑΜ write CAUS-thousand ART sentence '...to write the sentence a thousand times.' (lit. write making a thousand the sentence) (Aiavao 1987: 66) PROCESS + FREQUENCY (36)

'Ua 0 fa 'a-lua ... ΤΑΜ explode CAUS-two ... '[The thing called atomic bomb] exploded twice.' (Aiavao 1987: 35)

Juxtaposition can also be used for the expression of spatial relations, as the following examples of modification by mamao 'distant, far away' and latalata 'close' show: ACTION + SPATIAL RELATION (37)

...'ua

alu

mamao...

ΤΑΜ go far.away '... have gone far away' (until it can hardly be recognized and vanishes from sight) (Mosel & So'o, 2000: 95) (38)

Sa tatau ona tau mamao le Talune. ΤΑΜ appropriate that anchor far.away ART Talune 'The Talune should have anchored far away (from the beach).' (Aiavao 1987: 82)

PROCESS + SPATIAL RELATION (39)

'o le lä'au PRES ART plant

tupu grow

maualalo low

e ola TAM live

latalata close

i le sami. LOC ART sea Ά low growing plant which lives close to the sea.' (Mosel & So'o, 2000: 68)

280

Ulrike Mosel

Latalata 'close' can also denote a temporal relationship , e.g.: PROCESS + TEMPORAL RELATION

(40)

Ε

fua latalata i le taimi ο le paolo. ΤΑΜ bear, fruit close LOC ART time POSS ART paolo 'It bears fruit close to the time of the paolo.' (Mosel & So'o, 2000: 68) In contrast to the incorporation of peripheral arguments with the thematic role of PLACE and TIME, the modifiers latalata 'close, near' and mamao 'distant, far away' cannot be replaced by locative prepositional phrases, e.g. *i le latalata, *i le mamao,16 Finally I would like to give an example for modification by a SIMILATIVE and an ASSOCIATIVE word derived from a noun phrase, e.g. ACTION + SIMILATIVE

(41)

...e

mäfaufau

fa'a-tamaitiiti

le

ΤΑΜ

think

SIM-child

ART person

tagata

matua. adult

'... the adult person thinks like a child.' (Mosel & So'o, 2000: 92) ACTION + ASSOCIATIVE

(42)

Usu 2. v., fa 'afeiloa 'i fa 'a-le-aganu 'u ni mälö. usu welcome ASS-ART-custom ART guest 'Usu (2. v.): welcome guests according to the custom.' (Mosel & So'o, 2000: 81)

Modifier Constructions differ from Coordinate Constructions in that splitting up the clauses would result in ungrammatical or semantically completely different clauses: (43)

Ε

iloa

ΤΑΜ k n o w

lelei

le

good

ART thing

mea

'(She/he) knows this well.' (44)

Ε

iloa

le

mea

ΤΑΜ

know

ART thing

lea this

'(She/he) knows this' (45)

Ε lelei le mea lea. ΤΑΜ good ART thing this 'This is good.'

lea this

Complex predicates and Juxtapositional Constructions in Samoan

281

5.5. Depictive Constructions In Depictive Constructions 17 the second element expresses a certain bodily or mental condition of a core participant or the quantity of participants. Typical English examples for such constructions are George left the party angry. Carol drinks her coffee black. In our corpus, we could identify a number of predicative Juxtapositional Constructions which are semantically similar to such secondary depictive predicates. In the second position, i.e. in the function of a secondary predication we find BODILY CONDITION, NUMBER and QUANTITY words. In the following example TRANSITIVE ACTION + BODILY CONDITION

(46)

Ole 'ä ju ola 'oe. ΤΑΜ eat live 2SG 'You are going to be eaten alive.' (Moyle 1981: 52)

The word ola 'live, life' obviously does not modify the action of eating (it is not a 'lively eating'); nor does it denote the patient ('life') of the action of eating, but says something about the patient. Our next example is taken from a story in which a scientist covered the eyes of a flying fox and let him fly: ACTION + BODILY CONDITION (OF PARTICIPANT)

(47)

...e

le

se

aitu

ΤΑΜ NOT ART spirit

ο

lona

'äiga

le

of

his

family

ART thing

mea

lea

na

DEM ΤΑΜ

lele tauaso. a 'o le pe 'a moni lava. fly blind but ART flying-fox real EMPH '...it was not a spirit of his family, this thing that flew blind, but a real flying-fox.' (Aiavao 1987: 21) Here it is not the flying which is blind, but the flying fox. One could not say (48)

*na tauaso le lele 'the flying was blind'

An alternative construction for depictive constructions are adverbial clauses: (49)

...le tagata na fänau mai '_o tauaso ART person ΤΑΜ be.born DIR ΤΑΜ blind '... the person who was born blind ..' (Aiavao 1987: 76)

282

Ulrike Mosel

The sequence 'ο tauaso is marked by the short form of the progressive marker which marks the clause as being subordinate. It makes a predication about le tagata 'the person' and expresses the fact that the state of being blind is simultaneous with the event expressed by the superordinate clause. In contrast, Modifier Constructions allow the promotion of the subordinate second component, i.e. the PROPERTY word, to predicate position. Compare: PROCESS + PROPERTY (DIMENSION)

(50)

'o

le

itü'äiga

PRES ART kind

niu

e

coconut

ΤΑΜ not

le

lupu

maualuga

tele

grow

high

very

'a k i n d o f c o c o n u t w h i c h d o e s n o t g r o w v e r y h i g h '

(Mosel & So'o, 2000: 95) (51)

'o

le

laau

PRES ART plant

lana

tupu

its

grow

ο

le

niu...

POSS ART coconut

e

le'i

ΤΑΜ not.yet

maualuga high

'a germinating coconut... it has not yet grown high ' (lit. the plant of the coconut... its growing is not yet high) (Mosel & So'o, 2000: 95) In (51) maualuga 'high' makes a predication about tupu 'grow' as it does in (50) where it functions as a modifier. There is also an aspectual difference between lele tauaso, fänau tauaso on the one hand and tupu maualuga on the other: while the blindness existed independently and before the flying and the birth, the height of the plant develops through its growth; the growing ends in being high. A good example for a NUMBER word in the second position of a Depictive Construction is: ACTION + NUMBER (OF PERSON)

(52)

...ona

mä

sisi

to 'a-lua

atu

lea

'o

'oe

7

lugä.

then IEXC.DU lift HUM-two DIR then PRES 2SG LOC up '... then, the two of us together can lift you up on top.' (Moyle 1981:96) Here the NUMBER word to 'alua does not quantify the number of events but the number of participants. In fact, the human classifier to 'a explicitly indicates that the counted item is a human being. If the number of events is

Complex predicates and Juxtapositional Constructions in Samoan

283

counted, the second element would be a frequentative numeral in the function of a modifier. ACTION + FREQUENTATIVE

(53)

'Ua

tagi

fa 'a-lua

le

tama'ita'i.

ΤΑΜ

cry

CAUS-two ART lady

'The lady's crying a second time.' (lit. cries (by) making (it) two)' (Moyle 1981: 230) Similar to (51), the alternative construction promotes the modifying FREQUENTATIVE to predicate position. In an example provided by our corpus, the argument of the predicative FREQUENTATIVE is expressed by a complement clause. (54)

Ε

fa 'a-tolu

ona

ΤΑΜ CAUS-three that

tatalo

si

pray

ART old.man.

toeaina.

'The old man prayed three times.' (lit. 'It was made three that the old man prayed.') (Leauga 1974: 13) Two QUANTITY words are frequently used in Depictive Constructions: 'ätoa 'complete, whole' and 'uma 'all', e.g. TRANSITIVE ACTION + QUANTITY

(55)

'Ua folo 'ätoa Fitilo'ilo'i. ΤΑΜ swallow whole Fitilo'ilo'i 'She swallowed Fitilo'ilo'i whole.' (Moyle 1981: 116)

An alternative construction which nicely shows that 'ätoa relates to the patient occurs in the same context: (56)

... ma

le

manava

'ä

mäsae

i

lenei

ART

stomach

ΤΑΜ

burst

LOC

this

'ätoa

'ua

na

foloina.

whole

ΤΑΜ

3SG

swallow

with

tasata person

'... [going around] with her stomach about to burst from this person she had swallowed whole' (lit. 'from this whole person she had swallowed') (Moyle 1981: 116) The quantifier 'uma 'all' is used in the same way. It can be incorporated in the verb complex or function as an attribute in the noun phrase. The following example shows both constructions:

284

Ulrike Mosel

(57)

... 'ο

toea 'i 'ina 'uma lava ο l-o tatou 'äiga PRES old.men all EMPF poss ART-POSS 1INC.PL family '... all the old men of our family, tai

ta 'u

'uma

lava

e

tätou

nearly

call

all

EMPH

ERG

1INC.PL PRES

'ο

ο

tätou

tamä

POSS

llNC.PL

father

we call nearly all (of them) our father.' (Aiavao 1987: 13)

5.6. Possessive Constructions These constructions are similar to Indo-European bahuvrihi compounds of the redskin-type, which are also called exocentric compounds or possessive compounds in the English literature.18 They consist of a word denoting a body-part, some other part of a whole, or a piece of clothing which is modified by a PROPERTY or N U M B E R word. In contrast to redskin, they are mainly used in predicative and attributive function. PART/THING + PROPERTY

(58)

'o ΤΑΜ

mälamalama

ta 'e

pea

le

glass

broken

still

ART lamp

mölitü

ο

le

POSS

ART f a m i l y

'äiga.

'... the lamp of the family is still having a broken glass.' (Aiavao 1987: 108) PART/THING + N U M B E R

(59)

Ε

aß.

tolu

le

setima.

ΤΑΜ

funnel

three

ART

steamer

'The steamer has three funnels.' (Milner 1966: 5) B O D Y PART + PROPERTY

(60)

Ε

isu

papa

le

teine.

ΤΑΜ

nose

flat

ART

girl

'The girl has a flat nose.' (Milner 1966: 175) The possessive relationship becomes overt in an alternative construction in which the construction is split up, e.g. (61)

Ε

papa le isu ο le teine. flat ART n o s e POSS ART girl 'The nose of the girl is flat.'

ΤΑΜ

Complex predicates and Juxtapositional

(62)

Ε ΤΑΜ

tolu afi ο le three funnel 19 POSS ART

Constructions in Samoan

285

setima. steamer

'The steamer has three funnels.' (lit. 'The funnels of the steamer are three.')

6.

Juxtapositional Constructions in other functions

The constructions discussed in the preceding paragraphs are not confined to verb complexes, but also occur as modifiers in verb complexes and noun phrases and as the head of NPs without undergoing any morphological changes. Furthermore, they can also be embedded in other Juxtapositional Constructions and thus form a sequence of three or more juxtaposed lexical words which are organized in hierarchical structures.

6.1. Juxtapositional Constructions in the function of modifiers As mentioned in section 2, Samoan does not distinguish between verbs, nouns, adjectives and adverbs derived from adjectives as for instance do English and French. All kinds of content words are used as modifiers, and so are all kinds of Juxtapositional Constructions. Here we give first a few examples of simple modifiers and then of Juxtapositional Constructions. Apart from the examples and their translations, Table 7 shows the lexical class of the nucleus and the modifier and the thematic role of the nucleus in the action, process or state denoted by the modifier. Table 6. Simple modifiers. Example

Literal translation

Free translation

Lexical classes

Thematic role

tagata täfafao

people wander about

'tourists'

PERSON A C T I O N

AGENT

le lä goto

the sun sink

'the sinking sun'

THING PROCESS

UNDERGOER

le ta 'avale mümü

the car red

'the red car'

THING

B E A R E R OF

PROPERTY

PROPERTY

le potu moe

the room sleep

'the bed-room'

THING ACTION

PLACE

'the pay day'

TIME A C T I O N

TIME

le aso totogi the day pay

286

Ulrike Mosel

The following examples of Juxtapositional Constructions illustrate selected types of construction; further examples are found in chapter 6 and 7 of the Samoan Reference Grammar (Mosel & Hovdhaugen 1992). 6.1.1. Argument Incorporation Construction as a modifier in a NP (63)

le

teine

ART girl

(tausi

ma 7)

(take.care.of

patient)

(ACTION

PERSON)ARG.INC.

'the nurse' 6.1.2. Argument Incorporation Construction as a modifier in another Argument Incorporation Construction (64)

'Ua

(su 'e

ΤΑΜ

look.for

(ma 'a

(ACTION

(taotao

stone

(put.weight.on

(THING (ACTION

ta'inamu))), mosquito.net))) T H I N G ) A R G . I N C . ) M O D . ) ARG.INC.

'We/they looked for the stones [which were used] to put weight on [the lower edges of] the mosquito net [to keep them fast on the floor].' (lit. 'the putting weight on mosquito nets stones') (Aiavao 1987: 74) 6.1.3. Coordinate Construction as a modifier in a NP (65)

...le loli ('ena 'ena ART lorry (brown (PROPERTY

pe so o) a Eti die often) of Eti PROCESS)CO0RD.

'...Eti's often dying brown lorry' (Aumaga 1984: 1) 6.1.4. Possessive Construction as a modifier in a NP (66)

...fia want

maua

la'u

atu

(mata

tasi).

find

my

bonito

(eye

one)

( B O D Y - P A R T NUMBER)POSS.

'... I want to find me a one-eyed bonito.' (Moyle 1981: 222)

Complex predicates and Juxtapositional Constructions in Samoan 6.1.5. Possessive (67)

Construction

287

as modifier in a VC

la,

'ole'ä

atu

(mata

well,

ΤΑΜ

bonito

(eye

tasi)? one)

' W a s that going to be the one-eyed bonito?' (Moyle 1981: 222) Note that in (67) the head of the VC is the name of an animal; in fact, head and modifier are formed by exactly the same words as in the NP in (66).

6.2. Juxtapositional Constructions in the function of a N P nucleus plus satellites Similar to the large variety of content words which are found in the nucleus position of NPs, w e find all kinds of Juxtapositional Constructions in this position. In the Samoan Reference Grammar, these constructions were called nominalizations when they contained an ACTION, PROCESS, PROPERTY word as their nucleus because w e understood them as "noun phrases which are derived from verbal clauses" (Mosel & Hovdhaugen 1992: 513). Today I prefer not to speak of derivation as these noun phrases are not marked by any derivational morphology. A simpler way of description is just to say that ACTION, PROCESS and PROPERTY words can function as the nucleus of noun phrases. 2 0 The fact that they can be combined with negative particles and adverbs can be interpreted as a lexical property which distinguishes these words in any position, i.e. as lexemes, f r o m PERSON and THING words. 6.2.1. Argument Incorporation (68)

le

fa 'a-fefe

ART CAUS-fear

Construction

tasata person

'the frightening of people' (Aiavao 1987: 42)

288

Ulrike Mosel

6.2.2. Depictive (69)

Construction

'ο

Ιο 'u

PRES

my

'äiga family

fia want 21

pule

to 'a-tasi

ο

mea

a

le

control

HUM-one

POSS

thing

POSS

ART

potopoto extended

' M y wish to alone control the affairs o f the extended family.' (Taulogo 1986: 1) 6.2.3. Possessive (70)

construction

'Ua

oso

loa

le

loto

leasa

i

lona

uso.

ΤΑΜ

arise

then

ART

heart

bad

LOC

his

brother

' H e became jealous o f his brother.' (lit. 'the having a bad heart against his brother arose') (Moyle 1981: 122)

6.3. Complex Juxtapositional Constructions Since Juxtapositional Constructions have the same functional status as single words, it is no surprise that a Juxtapositional Construction can take another one as one of its components. Below w e give examples for various construction types: 1. Coordinate

Construction

containing an Argument

Incorporation

Construction (71)

...e

(mänaia

ΤΑΜ (beautiful

(fa 'a-momo 'i (CAUS-break

loto)) heart))

ΤΑΜ (PROPERTY (ACTION THING) ARG. INC.)COORD.

le

tetetete

ο

le

leo

ο

si

tama 'ita 'i.

ART

quivering

POSS

ART

voice

POSS

ART

lady

'...the quivering o f the voice o f the lady was heart-breakingly beautiful.' (Larkin 1967: 7)

Complex predicates and Juxtapositional Constructions in Samoan 2. Coordinate Construction containing two Possessive (72)

Ε

((lanu

ΤΑΜ ((colour

Constructions

mümü)

(mata

läpopo α")).

red)

(eye

big: PL))

((PART PROPERTY) Poss

289

PART PROPERTY) P o s s . )

Coord.

'It (the malau fish) has a red colour and big eyes.' (Mosel & So'o, 2000: 77) 3. Coordinate Construction containing a Possessive (73)

'o

le

metale

PRES ART metal

(vaivai

(föliga

(soft

(appearance

construction mümü)) red))

( PROPERTY ( PART PROPERTY ) 'a soft metal with a red appearance (copper)' (Mosel & So'o, 2000: 151) 4. Modifier Construction containing a Modifier (74)

'olo 'o

tu

fa 'a-le-päleni

ταμ

(stand

(SiM-not-balanced good))

ACTION (PROPERTY

P O S S.)COORD.

Construction

lelei

ai

lo 'u

vae

ANAPH

my

foot

PROPERTY) P R 0 P E R T Y )

ACTION

'... where my foot stood somewhat not well balanced' (Vili 1984: 1) 5. Depictive Construction containing a Cause-Effect (75)

... 'ua

((tatipi

ΤΑΜ ((cut: PL

o'u

'ofu

fa 'a-läiti) CAUS-small)CAUSE.EFFECT

Construction

'uma) all) DEPICrlVE

ä'oga...

my: PL dress:

school

'...my school dresses were all cut to make them small (for my sister) ...' (Falealili 1970: 20) 6. Possessive Construction containing a Coordinate (76)

Ε

{fua

(läpotopoto

ΤΑΜ (fruit (round

Construction

nini Ί). tiny)Co0RD.)poss.)

'It has tiny round fruits.' (Mosel & So'o, 2000: 95)

290

Ulrike Mosel

6.4. Summary The preceding sections showed that Juxtapositional Constructions of the Form (AB) can have the position of a nucleus plus satellite in VCs, of a modifier in a VC or NP, and of a component of another, superordinate Juxtapositional Constructions, i.e. the posi4. tion on any of the two components X and Y of a Juxtapositional Construction (XY) can be filled by another Juxtapositional Construction (AB) or (CD) resulting in ((AB) Υ), (X (AB)) or ((AB) (CD)). This variation of functions shows that Juxtapositional Constructions form a linguistic unit below phrase level.

7.

Conclusion

Samoan content words form lexical classes, but these do not correspond to the word classes or syntactic categories of noun, verb, and adjective in European languages. In principle, any content word can be used as the nucleus of a verb or noun phrase and as a modifier. Content words can be juxtaposed and form so-called Juxtapositional Constructions of various types. Five or six types of such constructions can be distinguished, depending on whether Cause-Effect-Constructions are considered as a class by itself or a subclass of Coordinate Constructions. Table 8. Overview of Juxtapositional Constructions with examples Name of construction

Example

Gloss

Argument Incorporation Constructions (AIC)

tausipepe tu 'u 'apa

'take care of babies' 'put into cans'

Coordinate Constructions (cc)

'umi uliuli nofo tatalo

'long (and) black' 'sit (and) pray'

Cause-Effect Constructions (CEC)

sele fa 'apupu 'u

'cut (and make) short'

Modifier Constructions (MC)

nofo lelei

'sit well (properly)'

Depictive Constructions (DC)

'ai ola

'eat alive'

Possessive Constructions (PC)

isu papa

'(have a) flat nose'

The classification of Juxtapositional Constructions is based on their meaning, but supported by the fact that each construction is semantically and syntactically related to a structurally distinct construction, as summarized in Table 9. To show how the different constructions correspond to

Complex predicates

and Juxtapositional

Constructions in Samoan

291

each other, we use X and Y as labels for the first and the second component of simple Juxtapositional Constructions and their equivalents in the related constructions. Where necessary for a better understanding, we introduce (Z) as a label for the primary argument. Each type of Juxtapositional Construction is formed by words of particular lexical classes which, on the one hand, allow us to derive the meaning of a Juxtapositional Construction from the lexical class of its components and on the other, to productively form new expressions of each construction type. Two of the distinctive lexical properties pertain to semantic-syntactic relations, namely valence and relationality: -

The valence o f TRANSITIVE ACTION words requires the expression o f a patient, whereas all other words do not; the patient is either expressed by a noun phrase or by a content word incorporated into the verb complex.

-

PART-OF words inherently carry the information that they function as the possessum in a possessive relationship. A PART is always a part of a whole, and a BODY-PART always belongs to a person or animal.

In other cases, particular semantic relations manifest themselves in "lexical solidarities" 22 , i.e. inherently given syntagmatic semantic affinities between certain lexemes. Such lexical solidarities exist for example between the following lexical classes of our sample: Table 9. Juxtapositional Constructions and related constructions Type of construction AIC

Related constructions

(ΤΑΜ) XY do X to PATIENTs of the kind Y

(TAM) X ART Y

do X to the PATIENT Y

(ΤΑΜ) XY do X in PLACEs of the kind Y

(ΤΑΜ) X LOC ARTY

do X in the PLACE Y

CC

(ΤΑΜ) XY do X do Y be X be Y

(ΤΑΜ) X and (ΤΑΜ) Y

do X and do Y be X and be Y

CEC

(ΤΑΜ) X CAUS-Y

(ΤΑΜ) X and (TAM) CAUS-Y

do X and cause (Z) to be Y

MC

(ΤΑΜ) XY do X in a Y way; do Χ Y times; etc.23

(TAM) Y ART X

X is Y; X is Y times

DC

(ΤΑΜ) XY do X while being Y; do X to (Z) while (Z) is (Y)

(ΤΑΜ) X (Z); (ΤΑΜ) Υ (Ζ)

(Z) does X / X is done to (Z) while (Z) is Y

PC

(ΤΑΜ) XY have an X of the kind Y

(TAM) Y X POSS (Z)

the X of (Z) is Y

do X to (Z) cause (Z) to be (Y)

292

Ulrike Mosel

In other cases, particular semantic relations manifest themselves in "lexical solidarities", i.e. inherently given syntagmatic semantic affinities between certain lexemes. Such lexical solidarities exist for example between the following lexical classes of our sample: -

-

-

BODILY CONDITION a n d PERSON: a BODILY CONDITION w o r d is o n l y

used in connection with a PERSON or ANIMAL word; NUMBER and PERSON, THING: with the exception of abstract mathematical statements, the use of a cardinal numeral presupposes the existence of an expression referring to countable entities, i.e. persons and things; FREQUENCY and ACTION, PROCESS: frequentative numerals imply the

presence of a ACTION or PROCESS word as they denote how many times something happens. The six types of construction can now be distinguished in terms of valence, relationality and lexical solidarity in the following way: - in Argument Incorporation Constructions (AIC), e.g. tausi pepe 'babysit', ti'eti'e solofanua 'sit on a horse, ride', the first element is an ACTION word; the second element denotes a person or thing which, in accordance with the valence of the first element, plays the thematic role o f a PATIENT o r PLACE;

-

in Coordinate Constructions (CC) both elements must belong to the same class, i.e. either to the lexical class of ACTION or to the lexical class PROPERTY, e.g. 'umi uliuli 'long (and) black', nofo tatalo 'sit (and) pray'; they have the same valence and share their arguments in predicative function; - Cause-Effect Constructions (CEC) are a subtype of Coordinate Constructions; both elements denote transitive actions; the only difference is that the second transitive action is a causative action, i.e. a transitive action which causes the patient to have a certain property or undergo a certain process, e.g. tapena fa 'alelei 'tidy up and make good, tipifa 'apa 'ü 'cut and make fall'; - in Modifier Constructions (MC), the second element is a PROPERTY or FREQUENCY word; irrespective of the valence of the first element, the second one is intransitive and modifies the first one by characterizing the quality, dimension or frequency of the state of affairs, e.g. iloa lelei 'know well', nofo lelei 'sit properly', 'inuinu la'itiiti 'drink a little', tagi fa 'alua 'sing twice, a second time';

Complex predicates and Juxtapositional

Constructions in Samoan

293

Table 10. Distribution of lexical classes in Juxtapositional Constructions Type

Lexical classes First element

AIC

CC

CEC

Examples Second element

TR. ACTION

PERSON, THING

tausipepe 'care for babies'

ACTION

inherently related to a place

THING

PROPERTY

PROPERTY

ACTION

ACTION

ti 'eti 'e nofoa 'sit on a chair' 'umi uliuli 'long and black' nofo tatalo 'sit and pray'

TR. ACTION

CAUSEPROPERTY

TR. ACTION

CAUSEPROCESS

MC

ACTION

PROPERTY

ACTION

DIMENSION

ACTION

ACTION

FREQUENCY

PROCESS

FREQUENCY

ACTION

SPATIAL RELATION

ACTION

TEMPORAL RELATION

DC

ACTION

SIMILATIVE

ACTION

ASSOCIATIVE

ACTION

BODILY CONDITION

PC

ACTION

QUANTITY

BODYPART

DIMENSION

tapena fa 'alelei 'tidy up and cause to be good' tipifa 'apa u 'cut and make fall' tausi lelei 'take care in a good way' moe 'umi 'sleep long' inuinu la 'itiiti 'drink little' tusifa 'aafe 'write a hundred times' pä fa 'alua 'explode two times tau mamao 'anchor far away' alu muamua 'go first' mäfaufau fa 'a- tamaitiiti 'think like a child' fa 'afeiloa 'ifa 'a-le-aganu 'u 'meet according to the custom' 'ai ola 'eat alive' fola atoa 'swallow whole' lauulu 'umi 'have long hair'

294 -

Ulrike Mosel

in contrast to Modifier Constructions, the second element of Depictive Constructions (DC) is not related to the first element, but through lexical solidarity to an argument denoted by an NP with a PERSON or THING word as its nucleus, e.g. 'ai ola 'eat alive', folo 'ätoa 'swallow whole', nofo to'atasi 'sit (by) one(self)'. Typically, the second element of a Depictive construction belongs to the lexical class of BODILY CONDITION or to the quantifiers 'uma 'all' and 'ätoa 'whole' and the cardinal numerals which only quantify PERSON or THING, but not ACTION or PROCESS;

-

Possessive Constructions (PC) have a relational PART-OF word as their first element which always denotes a part of a participant. This can be a body-part of a person or animal, a piece of clothing worn by a person, or any part of a thing.

Similar to single content words, Juxtapositional Constructions can occur in four functions: 1. as the nucleus plus satellite in verb complexes 2. as the nucleus plus satellite in noun phrases 3. as modifiers in verb and noun phrases 4. as components of superordinate Juxtapositional Constructions This multifunctionality supports our claim that they form independent syntactic units below phrase level. To sum up, juxtaposed content words can form distinct types of construction. Each type of such Juxtapositional Constructions is characterized by the particular lexical features of its components which determine the relations between the two components or between a component and a participant outside the Juxtapositional Construction.

Complex predicates and Juxtapositional Constructions in Samoan

295

Orthographic conventions and abbreviations Vowel length is indicated by a macron and the glottal stop is indicated by the apostrophe. AN ΑΡΗ ART ASS CAUS DEM DIR DU EMPH ERG EXC HUM

anaphoric pronoun article associative causative demonstrative directional particle dual emphatic particle ergative preposition exclusive numeral classifier for human beings

INC LOC NEG PL POSS PRES Q SIM SG ΤΑΜ

inclusive locative-directional preposition negation plural possessive preposition presentative preposition question particle similative singular tense-aspect-mood particle

Notes 1. 2.

3. 4. 5. 6.

7.

For comments on earlier versions of this paper I am most grateful to Nikolaus Himmelmann, Gunter Scnft and the editors and reviewers of this book. See the Samoan Reference Grammar (Mosel & Hovdhaugen 1992: 397, 738) and the critique of our treatment of the noun/verb distinction in Vonen 1993: 112 ff., Haspelmath 1994. See Broschart 1997, Vonen 2000. The only exceptions are plural specific noun phrases, e.g. teine 'the girls' and noun phrases which are inherently specific like proper names. laupapa refers to the material in general here. Therefore, it is translated by 'planks'. The grammar and dictionaiy project was funded by AusAid (The Australian Agency for International Development) for three years. I participated in the project as a consultant for a total of seven months during five visits. This is the way the Kalama ο le Gagana Samoa and the dictionary Utugagana (Mosel et al. 2000) use the terms veape 'verb' and nauna 'noun'. Veape (v.) is the head of a veib complex (fui 'upu veape) and nauna (n.) is the head of a noun phrase {fui'upu nauna). Consequently, in dictionaiy entries, the abbreviation n. after the headword means: "when used as the head of a noun phrase, the headnoun has the following meaning"; and correspondingly the

296

8.

9.

10. 11. 12. 13. 14. 15. 16.

17.

18.

19. 20. 21. 22.

23.

Ulrike Mosel

abbreviation v. means: "when used as the head of a veib complex ...". (see examples (17) ff). For the need to distinguish between lexical and syntactic categories, see Sasse 1993. I do not agree with his claim that "each semantic type will, in a given language, be associated with a particular word class" (1995: 175). Cardinal numerals, for instance, certainly constitute a semantic type, but in many languages they belong to more than one word class. Note that in this construction the ACTION word is the head of a NP. There are no other classifiers of this kind in Samoan. In Dixon's classification "... the lexical items of a language fall into a number of 'semantic types' (each item belonging to just one type)" (1977: 9). See Mosel & Hovdhaugen (1992: 392-396). /'A is an absolutive plural NP. The right border of the vc is marked by the anaphoric pronouns ai 'there'. In the function of nucleus, ugä means 'yellow-brown'. The meaning 'thick and large' in this context is not quite clear to me. Note that latalata does not modify the preposition as English 'close' does. This becomes evident when the absolutive argument le lä'au 'the tree' is inserted: Ε fua latalata le lä 'au i le taimi ο le paolo. 'The tree bears fruit close to the time of the paolo.' Semantically similar constructions in other languages are called depictive secondary predicate, predicative attribute, copredicate or copredicative, see the thorough investigation of depictive secondary predicates by SchulzeBerndt & Himmelmann forthcoming. In the Samoan Reference Grammar we called them Isu Mamafa Compounds. The compound isu mamafa literally means '(having) a heavy nose' and is used metaphorically in the sense of 'having a cold' (Mosel & Hovdhaugen 1992: 88f, 300f). Afi is a specific plural noun phrase here. With a possessive pronoun the clause would be Ε tolu ona afi. It has three funnels; its funnels are three.' Compare Gil's arguments for a similar interpretation and way of description of Riau (Gil 1994). fia 'want' is not a verb, but an adverb or modal particle. Lexical solidarities are syntagmatic relations of lexemes in which the content of one lexeme functions as a distinctive semantic feature of a second lexeme, so that this second lexeme entails the content of the first one and determines selectional restrictions like, for example, 'bark', which is only said of dogs, and 'neigh' of horses (Coseriu 1967). I find it difficult to describe the various types of modifier constructions and could not express the meaning of all types in such an abbreviated manner.

The grammaticization of directional verbs in Oceanic languages Malcolm Ross

Abstract In Oceanic languages we often find directional elements which were originally directional verbs in serial verb constructions. The various grammaticization paths that have led to the modern constructions are described in this article. The grammaticized directional elements in modern languages are of three kinds: postverbal directional clitics ('directionals'), pre-verbal directional clitics, and prepositions and relators. Post-verbal directional clitics are the most widely attested outcomes of this grammaticization. The other two categories are much less common. However, the third category, prepositions and relators which have resulted from the grammaticization of directional verbs, has received relatively little attention in the literature and so is a focus of attention here.

1.

Introduction

Oceanic languages often have directional elements which were originally directional verbs in serial verb constructions (SVCs). In this paper I examine the various grammaticization paths that have led to the modern constructions.1 The morphemes that occur in these constructions fall into three categories: (1)

a. post-verbal directional clitics ('directionals') (Section 3); b. pre-verbal directional clitics (Section 3.2); c. prepositions and relators (Section 3.3).

The most widely represented of these categories is (a). Category (b) is found in only a few languages. Although (c) is much less widespread than (a), it receives more attention here because its history has received little mention in the Oceanic literature. The term 'directional' could logically be applied to both (a) and (b), but is conventionally used only for (a), and I maintain that convention here, referring to (b) simply as 'pre-verbal clitic'.

298

Malcolm Ross

The grammaticization paths described in this paper begin with certain SVCs which can be reconstructed for Proto Oceanic (POc), and in some cases with combinations of SVCs and locative phrase (LocP) constructions. Because a LocP formed a constituent of certain SVCs, I reconstruct LocP constructions first (1.1), then SVCs (2.2). The language classification adopted in this paper is that used by Ross, Pawley and Osmond (1998: 7-11) for the purposes of reconstructing Proto Oceanic. Briefly, they assume Oceanic to consist of three major primary subgroups, Admiralties (Adm), Western Oceanic and Eastern Oceanic. Western Oceanic consists of the North New Guinea (NNG), Papuan Tip (PT) and Meso-Melanesian (MM) clusters and the Sarmi-Jayapura (SJ) group. Eastern Oceanic includes Southeast Solomonic (SES), North-Central Vanuatu (NCV), South Vanuatu (sv), New Caledonia (NCal), Nuclear Micronesian (Mic), and Central Pacific (divided for convenience into Fijian and Polynesian (Pn)).

2.

Some Proto Oceanic constructions associated with location

2.1. Proto Oceanic locative phrases I use the term 'locative phrase' as a cover term for a phrase indicating location, whether the LocP is licensed by the verb or is an adjunct, and whether it is structurally an adpositional phrase, a noun phrase or an adverb. Elsewhere I have reconstructed two LocP constructions that must have occurred in POc (Ross to appear). One was the local construction, formed simply with the preposition *i,2 the other the general adjunct prepositional construction formed with the sequence *i ta-. Some nouns, e.g. *Rumaq 'house', occurred in both constructions: (2)

local construction: general adjunct construction:

*i Rumaq *i ta-fia Rumaq

'at home' 'at a/the house'

Two groups of noun occurred in the local construction, labelled for convenience 'local nouns'. The first group included placenames, demonstrative pronouns and nouns which denoted familiar places like 'home', '(own) village', '(own) garden', 'bush', 'beach', i.e. locations whose reference would be clear to the addressee without further specification or whose exact reference was irrelevant (like the noun in the English construction with at and no article: at home, at school, at work). With *i these nouns

The grammaticization of directional verbs in Oceanic languages

299

behaved morphosyntactically and semantically like a proper placename. Hence *i Rumaq meant 'at home', *i qutan 'in the bush' etc. This group also included nouns denoting regions, either in relation to, say, the island home of the interlocutors or in relation to the speaker. These nouns denote geographic directions such as 'up above', 'down below', 'inland', 'at sea' and so on. There is no sharp distinction between these and nouns for familiar places: 'inland' and 'bush', for example, both POc *i qutan, are often synonymous in Oceanic languages. Nouns reconstructible in this construction are listed in (3), together with their meanings as common nouns and as local nouns. Some of them also formed a locative adverb with *qa- ( ' — ' indicates that the item is not reconstructible, ' . . . ' that the evidence is insufficient to decide whether it is reconstructible). (3) POc local nouns:

*qutan *lorja *laur *tasik *Rumaq *tanoq *atas *layit *laka *liway *ni *no

as common noun

as local noun

adverb

'bushland, hinterland'

'inland' 'inland' 'seawards' 'at sea' 'home' 'down below' 'up above' 'up top, high up' 'up above' 'middle' 'here, near speaker' 'there, near third person'...

*qa-qutan

—

'sea, salt water' 'house' 'earth, soil' 'high country, uplands' 'sky, weather' 'open space' — —

—

*qa-laur —

*qa-Rumaq *qa-tanoq — —

*qa-laka —

*qa-ni

The second group of local nouns were directly possessed 'relational' local nouns which occurred with a possessor noun phrase, as in *i lalo-na Rumaq 'at the inside of the house', i.e. 'in the house'. These are only peripherally relevant to the topic of this paper and are not further discussed here. Local nouns of both kinds are reconstructed with supporting evidence in Ross (2003). Although the set of local nouns seems to have been small (but probably not closed), at least some of its members also functioned as common nouns with the meanings noted in (3). Common nouns occurred as heads of noun phrases which functioned as subject or object or possessor. In the general adjunct prepositional construction formed with the sequence */ ta- they

300

Malcolm Ross

functioned - originally, at least - as possessor, since *ta- was apparently a semantically empty directly possessed noun. 3 Thus */ ta-na Rumaq meant 'at a/the house' (rather than 'at home'). 4 It seems likely that by the time POc broke up, however, *i ta- alternated with *ta-, and both *i ta- and *ta- had been reanalysed as prepositions (*i ta- as a composite preposition). This may well have reflected a more general process whereby *i was omissible when the prepositional phrase it initiated (whether a local or a general adjunct construction) was licensed by the verb. This process still applies in Longgu (SES) (Hill 1992: 239-240).

2.2. Some Proto Oceanic serial verb constructions SVCs are common in Oceanic languages, and can be divided into types according to the semantic categories of verb that occur in them and according to the ordering of these verbs. Before we can describe the SVC types that concern us here, three categories of verb need to be recognized. They are exemplified from Yabem (NNG). The POc etyma reflected by the Yabem terms are given where these are reconstructible: 5 (4) Verbs of movement in a geographic direction (Vgeog). Typically this category includes verbs of movement along the horizontal and vertical axes and may also include verbs like 'enter', expressing movement relative to a specific location: -sa -pi -so

'ascend' (< POc *sake) 'climb up' 'enter' (< POc *surup)

-sep -mil -deij

'descend' (< POc *sipo) 'go back' (< POc *mule) 'move to, reach'

These verbs are intransitive but optionally license a LocP (2.1.). (5) Verbs of movement in a deictic direction (Vdeic), i.e. verbs which make reference to one of the three persons (the same three-way distinction is made by the demonstratives of many Oceanic languages): -mey -wä? -ya

'come to speaker[s]' (< POc *mai) 'go to addresseefs]'(glossed'go:2') (< POc *watuf 'go to her/him/them'(glossed'go:3') (< POc *yaku)

These verbs are intransitive and do not license a LocP.

The grammaticization of directional verbs in Oceanic languages

301

(6) Locomotion verbs (Vloco), entailing no directionality. They include both intransitive and transitive verbs, such as: intransitive transitive 'insert' -selerj 'walk, travel' -•pwane? 'push' -ne 'sink, submerge' -ü? 'paddle (a canoe)' -millη 'go first' -sit) 'pull, drag' -lob 'fly' -e Geographic and deictic direction verbs occur in three svcs. I label these types geographic directional, deictic directional, and sequential svcs. Each of these is widely enough represented in 'well-behaved' Oceanic languages for us to be reasonably confident that it occurred in POc. (By a 'wellbehaved' Oceanic language, I mean one in which the constituent order is Verb-Object, not Object-Verb.) This means that the geographic direction and deictic direction verb categories are also reconstructible in POc. I present the lexical reconstructions below without supporting evidence, as this is provided in detail by Ross (2003). POc deictic direction verbs made the same three-way person-oriented distinctions as POc demonstratives (Ross 2003). Many of their reflexes are directionals, and, I reconstruct them as both verbs and directionals. The major deictic direction verbs were: (7) POc deictic direction verbs: as verb *mai, *ma (VF) 'come' *watu, *ua (VF) 'go towards addressee (= go:2)' *lako, *la (VF) 'go (to) (= go:3)' *pano, *pa (VF) 'go away'

as directional 'towards speaker' 'towards addressee' 'away from speaker' 'away from speaker'

Of these verbs, *mai/*ma, *lako/*la and *pano/*pa, have long and short forms. It is difficult to reconstruct their distribution with any certainty, but there seem to be two svc-related contexts in which the short forms may have emerged. One was as the final verb of a deictic directional SVC. The other was as the first verb of a sequential SVC. POc *watu and *ua also look superficially like a long/short pair, but they tend towards a regional distribution, unlike the other three pairs.7 Reflexes of both *lako/*la and *pano occur in modern languages as directionals with the meaning 'away from speaker'. However, there is evidence that as verbs they had different meanings. Most verbal reflexes of *lako 'go (to)' license - optionally, at least - a destination (expressed, for

302

Malcolm Ross

example, as a prepositional phrase), whilst *pano apparently did not and was used, as a deictic directional verb, in the sense 'go away (from speaker), depart'. It seems that *pano, however, was also used as a geographic direction verb meaning 'move in a transverse direction, move across the valley'. In this sense it contrasted with 'go up, go inland' and 'go down, go seawards'. The two different uses of *pano are reconstructive for POc. However, it is reasonable to assume that at some earlier stage they were simply implicatures from a single meaning. The main semantic domains of geographic direction verbs are those of direction in relation to a valley - uphill, downhill and across - and of vertical direction, downward and upward. The directions relating to valleys were expressed by the three verbs in (8). It is likely that *sipo and *sake were also adverbs in POc. (8) POc geographic direction verbs (valley-related) *sipo *sake *pano

'go downward' 'go upward' 'move in a transverse direction, move across the valley'

These verbs often have the secondary horizontal senses 'seaward' and 'inland' or, from the sea, 'landward', because Oceanic speakers have often dwelt on mountainous islands (and still do), as well as 'to the west' and 'to the east', as these are the directions in which the sun goes down and comes up.

2.2.1. The geographic

directional serial verb

construction

In a geographic directional SVC, the first verb denotes among other things the manner of the movement, the second its geographic direction. Typically, the geographic direction is up or down. If the locomotion verb is intransitive, the moving object/person is the subject of both verbs. The construction is: (9) Geographic directional SVC (intransitive) Vloco, TR + Vgeogi...„ ( + LocP) This construction is illustrated in (10}—(13). Other examples occur in (31), the second clause of (34), (40) and (41).

The grammaticization of directional verbs in Oceanic languages (10)

(11)

(12)

Arop-Lokep (NNG) Vloco Vgeog LocP Am-dada am-du pang ni S:lEP-run S:lEP-go.down to place 'We ran (and we went) down to our place.' N.E. Ambae (NCV) Vloco Vgeog No = mo dige no = mo S:1S=REAL walk S:1S=REAL Ί walked to Longana.'

vano go

ki-am. POSS-P:1EP

LocP Longana. Longana

N.E. Ambae (NCV) Vloco soi Mwere, gato mo like hermit.crab S:3S:REAL fall 'Like, the hermit crab fell down

(13)

303

Vgeog mo S:3S:REAL

hivo... go.down

Tamabo (NCV) (Jauncey 1997)

Vloco Vgeog Vgeog LocP Dae-na mo kolo mo jivo mo kakau alau. blood-P:3S S:3S flow S:3S go.down S:3S reach coast.direction 'His blood flowed down as far as the coast.' If the locomotion verb is transitive, the moving object/person is object of the first verb and subject of the second. The construction is: (14)

Geographic directional SVC (transitive) Vloco TR (+ Obj) + Vgeogi...« (+ LocP)

Examples are (15)—(17). The verbs 'say' and 'hit' are apparently conceptualized as transitive locomotion verbs here: they cause their object to move to a location. (15)

Yabem(NNG)

Vloco Obj Ya-pwane? mo S:lS-lRR:insert taro 'I'll put the taro into

Vgeog LocP e-sep ad-?-sur]. S:3S-lRR:descend mouth-P:lS-hole my mouth.'

304

Malcolm Ross

(16)

Yabem(NNG) Vloco

Obj

Vgeog

LocP

Ya-som birj e-nderj lau. S:lS-lRR:say word S:3S-lRR:reach people 'I'll say something to the people.' (17)

Paamese(NCV) Vloco

Vgeog

LocP

Kaik ko-muasi-nau nau-vä netan. you S:2S:REAL-hit-0:1S S:lS:REAL-go ground 'You hit me down.' Verbs of stationary posture ('remain', 'sit') also occur in this construction: (18)

Kele(Adm) Vloco Obj Yu u-ruei pdlet S: 1S S:lS-put dish Ί put the dish onto the

'remain' LocP i-so dta-n keaw. S:3S-remain top-P:3S table table.'

In these examples the LocP consists of a noun phrase without an adposition. I assume that (in some cases, at least) this reflects the omission process mentioned above: POc *i was probably omissible when the prepositional phrase was licensed by the verb. The intransitive and transitive variants of the geographic directional SVC, shown in (9) and (14), are both reconstructive for POc and are exemplified by the reconstructions in (19) and (20) respectively. With regard to these and succeeding reconstructed clauses, I should emphasize that they are a presentational device: I do not wish to claim that their component forms are necessarily correct representations of POc.8 (19)

POc geographic directional SCV (intransitive) Vloco Vgeog LocP *au = laka au-sipo [ i] laur. S:lS=walk S:lS=go.down PREP coast Ί walked down to the beach.'

The grammaticization of directional verbs in Oceanic languages (20)

305

POc geographic directional SVC (transitive)

Vloco

Obj

Vgeog

LocP

*au=papa-i = a natu-gu i-sipo [i] laur. S: 1 S=carry-TR-0:3S child-P:lS S:3S=go.down PREP coast Ί carried my child down to the beach.'

2.2.2. The deictic directional serial verb

construction

The intransitive deictic directional SVC is: (21)

Deictic directional SVC (intransitive) Vlocoitr/Vgeog + Vdeic

The first verb denotes the manner or geographic direction of movement, whilst the second is deictic, i.e. denotes direction in relation to the speaker and addressee. If the first verb is intransitive, the moving object/person is the subject of both verbs, as in (22)-(24). Further examples are (40), (41), (29)—(31) and the first clause of (34) below. (22)

Kele(Adm) Vloco Vdeic I i-htuat i-ew. he S:3S-run S:3S-go.away 'He ran away.'

(23)

Gumawana (PT) (Olson 1992) Vgeog Dokanikani-ya-na bogina giant-DET-P:3S PERF 'The giant came down.'

(24)

Vdeic

i-ma. i-sou S:3S-go.down S:3S-come

Paamese (NCV) Vloco

Vdeic

Meatin kail a-valus au-mai. person PL S:3P:REAL-row S:3P:REAL-come 'The people rowed hither.' If the first verb is a locomotion verb, it may be transitive. In this case, the moving object/person is object of the first verb and subject of the second. The construction is:

306

Malcolm Ross

(25)

Deictic directional SVC (transitive) VlocoTR (+ Obj) + Vdeic

Examples are (26) and (27), as well as (42), (35) and (36) below. (26)

Kele(Adm) Vloco Obj Vdeic Su we he-guni yotu we at-la. S:3P POT S:3P-take S:1EP POT S:lEP-go 'They will take us away.'

(27)

N.E. Ambae (NCV) Vloco Obj Vdeic Go = wali na loko mo vanai. S:2S=carry ACC pudding S:3S:REAL come 'Bring the pudding.'

In Yabem deictic directional SVCs with a geographic directional (not a locomotion) verb, a LocP occurs immediately after the geographic direction verb, giving the construction in (28). This reflects the fact that it is the geographic direction verb that licenses the LocP. (28)

Vgeog (+ LocP) + Vdeic

Examples are given below (I have so far found examples in no language other than Yabem). There are no corresponding transitive examples as a geographic direction verb is always intransitive. (29)

(30)

Yabem (NNG) Vgeog LocP Ke-pi lom S:3S-go.up men's.house 'S/he climbed up to the

Vdeic ge-ya. S:3S-go:3 men's house.'

Yabem (NNG) Vgeog LocP Ya-nderj am-a? nem apomtau S:lS-lRR:go.to 2-COL POSS:2 chief Ί will go to your chief.'

Vdeic ya-wa?. S:lS-lRR:go:2

The grammaticization of directional verbs in Oceanic languages (31)

307

Yabem(NNG) Vloco

Vgeog

LocP

Vdeic

Way ge-ne ke-sep gwe? canoe S:3S-sink S:3S-descend sea 'The canoe sank into the sea.'

ge-ya. S:3S-go:3

When there is no geographic direction verb to license a LocP, a LocP may occur as an adjunct following the whole SVC, giving the construction below: (32)

Vloco,TR/TR + (+ Obj) + Vdeic + LocP

Examples (33) and the first clause of (34) are intransitive, whilst (35) and (36) are transitive. (33)

Yabem (NNG) Vloc

Vdeic

LocP

Balosi ge-lob ge-ya lo?. dove S:3S-fly S:3S-go:3 mountain 'The doves flew off to the mountain.' (34)

Bali (MM) Ε ART

voruko giant

Vloco ki SEQ:S:3

pete run

Vdeic ki SEQ:S:3

Vloco

ua go

LocP na lo-na ke-na. PREP inside-P:3S P O S S - P : 3 S

Vgeog

rumaka ki liki pete ki duatalata. house SEQ:S:3 do.again run SEQ:S:3 go.out 'The giant ran into his house again, then he ran out...' (35)

Bali (MM) Vloco Ε ART

Obj

Vdeic

Litjei te hola a pazi ki = to ua Lingei REAL:S:3 carry ART betelnut SEQ:S:3=CONT go

LocP virι Galiki. PREP Galiki 'Lingei took betelnut to Galiki.'

308

Malcolm Ross

(36)

Tamabo (NCV) (Jauncey 1997) Vloco

Vdeic

LocP

Ku tau = a a mai Vomarada. S:1S place=0:3S S:3S come Vomarada Ί sent it (here) to Vomarada.' The variants of the deictic directional SVC may be reconstructed for POc. Sample reconstructions are given below (the numbers on the right refer to the constructions presented above). (37)

POc deictic directional SVC (intransitive)

Vgeog

[LocP]

Vdeic

(21), (28)

a. *i = sipo [[i] laur] i = ua. S:3S=go.down [PREP coast] S:3S=go:2 'S/he went down (towards you) [to the beach].' Vloco™

Vdeic

[LocP]

(21), (32)

b .*i = laka i-ua [fij ta-na Rumaq]. S:lS=walk S:lS=go:2 [PREP ta-P:3S house] 'S/he walked away (towards you) [to the house].' (38)

POc deictic directional SVC (transitive)

VlocoTR

Obj

Vdeic

[LocP]

*au=papa-i = a natu-mu i-ua [fij S: 1 S=carry-TR-0:3S child-P:2S S:3S=go.2 [PREP Ί brought your child (to you) [to the beach].'

2.2.3. The combined construction

geographic

and

deictic

(25), (32) laur], coast]

directional

serial

verb

Often geographic and deictic directional serialization are combined, giving the three-verb sequence in (39). (39)

Geographic-cwm-deictic SVC VlocoITR/TR (+ Obj) + Vgeogi...„ (+ LocP) + Vdeic

This construction is illustrated in (40)-(42).

The grammaticization of directional verbs in Oceanic languages (40)

309

Roviana (MM) Vloco

Vgeog

Vdeic

Haqala pule mae sa boko. run return come ART pig. 'The pig ran back towards [them].' (41)

Roviana (MM) Vloco Vgeog Vdeic Totoso ene nuquru la garni... when walk enter go D:1EP 'When we walked in ... '

(42)

Yabem (NNG)9 Vloco

Obj

LocP

Vgeog

Vdeic

Lau s-e war) ayga salsrj ke-sa ge-metj. people S:3P-pull canoe from forest S:3S-ascend S:3S-go:l 'The people pulled the canoe(-log) up here out of the forest.' The combined geographic/deictic reconstructive for POc: (43)

construction

in

(39)

is

also

POc geographic-cwm-deictic SVC VlocoITR

Vgeog

[LocP]

Vdeic

a. *i = laka ko — sipo [fij ta-na Rumaq] S:lS=walk S:2S=go.down [PREP ta-P:3S house] 'S/he walked down (towards you) [to the house].' VlocoTR

Obj

Vgeog

[LocP]

i-ua. S:lS=go:2 Vdeic

b. *au=papa-i = a natu-mu *i = sipo [ f i j laur] i-ua. S:lS=carry-TR-0:3S child-P:2S S:3S=go.down [PREP coast] S:3S=go.2 Ί brought your child down (to you) [to the beach].'

2.2.4. The sequential serial verb

construction

In this SVC the first verb is a geographic or deictic direction verb expressing 'go and ...' or 'come and ...', the second a verb (phrase) expressing the main event of the predication. The verbs have the same subject, and a purposive relationship between the actions is usually implied. The construction is:

310

Malcolm Ross

(44)

Sequential SVC V g e o g / V d e i c + EVENT

Examples are: (45)

Bali (MM) Vgeog

EVENT

Hizi mi = ri zio ki they IRR=HYP:3 go.down SEQ:3 'They will go and catch fish.' (46)

vahi-arja get-PL

iharja. fish

Roviana (MM) Vdeic

EVENT

Kote mae hupul-i-a sa boko. FUT come gut-TR-0:3S DEF pig '[The men] come to gut the pig.' (47)

N.E. Ambae (NCV) Vgeog

Gai-rue

EVENT

ra — ru

NUM-two S : 3 P = D

mo

vano

S:3S:REAL g o

ra — ru mo S:3P=D

rivu

S:3S:REAL plant

talu. garden

'Two went to plant the garden.' (48)

Paamese (NCV) Vgeog

EVENT

Ki-ro-vä-tei

ki-hol?

S:2S: IRR-NEG1 -go-NEG2

S:2S:IRR-dance

'Won't you go dancing?' Again the construction may be reconstructed for POc, as illustrated in (49): 10 (49)

POc sequential SVC Vgeog

EVENT

a. *ra = sipo ra=paqus-i-a na waga. S:3P=go.down S:3P=bind-TR-0:3S ART canoe 'They went down to bind up the canoe.'

The grammaticization of directional verbs in Oceanic languages Vdeic

311

EVENT

b. *ra = mai ra=paqus-i-a na waga. S:3P=come S:3P=bind-TR-0:3S ART canoe 'They came to bind up the canoe.'

3.

Grammaticization paths

3.1. From directional verb to directional As I indicated in Section 1, directional elements in the verb phrases of modern Oceanic languages are often no longer verbs. Following the first of the three identifiable kinds of grammaticization path, they form a morpheme class of 'directionals' (glossed DIR), whose members follow the verb phrase and are often enclitics to it. Directionals are descended both from geographic direction verbs in geographic directional SVCs and from deictic direction verbs in deictic directional SVCs. In (50), the directional sa 'upwards' is identical in form to the verb root -sa 'ascend', and it is easy to see that directional sa is derived from the geographic direction verb -sa by deletion of the subject marker on the latter (cf. (42)): 11 (50)

Yabem (NNG) Vloco Obj DIR Se-balarj bo? sa. S:3P-carry.on.shoulders pig upwards 'They lifted the pole with the pig tied to it.'

It is not clear from descriptions of Yabem exactly what the functional difference is between the directional sa 'upwards' and a directional SVC with -sa 'ascend', but it is clear enough from nearby Mangap-Mbula how they arose. In (51) the subject marker on the second or further verb in an SVC is optional (Bugenhagen 1995: 153-154). (51)

Mangap-Mbula (NNG) Vloco

Obj

Vgeog

Ti-ur mburu [ i- ]su toono. S:3P-put cargo [S:3S-]go.down ground 'They put the cargo down on the ground.'

312

Malcolm Ross

This must once also have been the situation in Yabem, and the directional sa resulted from the deletion of the marker. The Yabem construction in (50) reflects the POc geographic directional SVC in (20), and the directional sa reflects the POc geographic direction verb *sake 'go upward'. More often, perhaps, directionals reflect POc deictic direction verbs. Thus in the Babatana (MM) examples in (52) and (53) the directionals la and me reflect POc *lako 'go:3' and *mai 'come'. Example (52) reflects the POc intransitive construction in (37a), whilst (53) reflects the transitive construction in (38). (52)

Sisiqa (MM) Vloco= DIR

LocP

yöi ma-zo = la Susuka. s/he S: 3 S: IRR-walk=thither Susuka 'S/he is going to Susuka village.' (53)

Sisiqa (MM) Vloco= DIR Obj Ra ko-gisu = me kavia I S:lS:REAL-carry=hither some Ί have brought some coconuts.'

kuda. coconut

It is even clearer here than in (50) that the directionals are no longer verbs, as they are encliticized to the verb phrase in both intransitives like (52) and transitives like (53). This is different from the pattern of transitive directional SVCs, where the object usually intervenes before the direction verb, as in (25). In N.E. Ambae, grammaticization has become lexicalization. Reflexes of the deictic directionals *mai and *watu have combined with the geographic directional verbs *pan, *sipo and *sake to form a set of directional morphemes which may function as deictic verbs (see (27)), adverbs or demonstratives. (54)

N.E. Ambae (NCV) (Hyslop 2001: 292)

away from speaker towards speaker towards addressee, or past/future deictic centre

— *mai *watu

across/ traverse *pano vano van-ai van-alu

up/ landward *sake hage ha-mai hag-atu

down/ seaward *sipo hivo hi-mei hiv-atu

The grammaticization of directional verbs in Oceanic languages

313

In Manam (NNG) the geographic directional suffixes -oti 'seawards', -oro 'inland', -ra?e 'up, to one's right when facing sea' (< POc *sake), -ria 'down, to one's left when facing sea' (< POc *sipo) always precede the deictic directional suffix la?o 'away' (< POc *lako), and never vice versa (Lichtenberk 1983: 576-582), reflecting the sequence of the deictic directional SVC.12 For example: (55)

Manam (NNG) Ogi i-do?-i-ra?e-la?o. axe S:3S-take-0:3P-upward-away 'He took the axes away upward.'

The sequence -ra?e-la?o reflects POc *sake 'go up' + *lako 'go:3'. This sequence is also reflected in Nyeläyu and Nelemwa (NCal) directional enclitics. The Nyeläyu geographic directional clitics =da 'upwards' ( > > > >

tnee- (Nl), tnoo- (N2) tatumo te- 'advance, walk'

In Nemi 2, where intervocalic stop consonants are generally weakened, there is an increasingly opaque incorporation o f f e in V 2 position: (14)

-fe 'take' (Nemi 1)

>

-ve > -e (Nemi 2)

Thus we find the following correspondences between the two Nemi variants (ex. 15): (15)

Nemi 1 (west coast) tnee-fe 'run with something' (run-take) ta-fe 'flee with something' (flee-take) te-fe 'carry out to sea (current)' (advance-take) mo-fe 'take care of, bring up (a child)' (stay-take) tu-fe 'stand with, adopt (a child)' (stand-take)

Nemi 2 (east coast) tnoo-ve ta-ve te-ve ~ tee moe tiie (< tu-vef

In Nemi 1, the compound verb structure is still clear even though the meaning of certain groups (ex. mo-fe 'take care of, bring up', tu-fe 'adopt (a child)') is no longer easily deduced from the verbs' respective meanings. In Nemi 2, the verb 'take' is still identifiable with its associative use in the first two compounds, but its fossilization is complete in the last verbs in example (15): te-ve ~ tee 'carry out to sea (current)', moe 'take care of, bring up', tüe (hyaok) 'adopt (a child)' are reanalyzed as simple transitive verbs with well specialized meanings. Compared to Nemi 1, Nemi 2 is characterized by a general weakening of intervocalic stop consonants and the former disyllabic compounds have been aligned on the phonological pattern. The monosyllabic structure of the verb 'take' has certainly favoured this morphological erosion since, as we can see in example (16), in the same compositional context, di- or trisyllabic V 2 s remain stable: (16)

tuut 'be standing' + hoa 'watch over' + mwaji 'attack' but: + fe 'take'

Nemi 2 > tu-hoa 'watch for' > tu-mwaji 'block in an ambush' > tüe (hyaok) 'adopt (a child)'

338

Frangoise Ozanne-Rivierre te- 'advance, move' + thaning 'free' + thuaom 'shelter o.s.' + thio 'be on tiptoe' but: + fe 'take'

> > > >

te-thaning 'walk burdenless' te-thuaom 'walk sheltering o.s.' te-thio 'advance on tiptoes' te-ve ~ tee 'carry out to sea'

In Nemi, it is therefore a process of lexicalization which is at work in the evolution of the verb fe 'take'. We will now take a look at a different evolution of this same verb in Nelemwa and in Nyeläyu, two languages of the far north, taking place in the same compound structures.

5.

Grammaticalization of the verb βhe 'take' in Nelemwa

Nelemwa (Bril 2000, 2002) is a split ergative language: of the accusative type sVo when the agents are person markers and of the ergative type, with unmarked VOA word order, when the arguments are nouns. If the nominal agent is human, the ergative marker is (e)a, if it is not human, the ergative marker is ru. The verb flie 'take, carry', related to the Nemi verb fe, can also be used as an autonomous transitive verb: (17)

Na i fhe ädaga i ye. and 3Sg take comb of her 'And she takes her comb.' (Bril 2000: 508)

(18)

I fhe thaamwa ena ea Pwayili. 3SG take woman this ERG P. 'Pwayili married this woman.' (Bril 2000: 144)

(19)

I fhe-e. 3SG take-him 'He carries him.'

It can also be found in V] position in complex verb structures: (20)

flie-bwage 'return (to owner)' fhe-xare 'carry away, move'

< fhe+bwage /take/return/ < flie+khare /take/different/

In Nelemwa, one also finds an associative morpheme -ve, enclitic in posture and motion verbs, most probably derived from the verb fhe in former serializations (V +fhe>V-ve) (Bril 1994:379). Semantically and formally, it appears in compound verb structures closely resembling those we just described for Nemi, with the syllabic reduction of V, as in (21):

The evolution of the verb 'take' in New Caledonian languages (21)

ku-ve 'stand with, support' mu-ve 'stay with, adopt' kä-ve 'sleep with' ca-ve 'flee with, abduct' thege-ve 'run away with' or without formal change in Vi as in

< kuut 'be standing' < muuvi 'stay' < käläp 'be lying down' < cap 'escape' < thegela 'run' example (22):

(22) fuk-ve 'fly away with' yuup-ve 'crawl away with' wälem-ve 'walk with'

< fuk 'fly' < yuup 'crawl' < wälem 'walk'

339

The resulting sequence is transitive and uses an ergative construction (ex. 23): (23)

I

thege-ve

khayoot ru loto ena. 3SG run-ASSOC barrier ERG car this.DELCT 'This car carried the barrier away.' (Bril 2002)

But, contrary to Nemi, the transitivizing morpheme -ve can only introduce non-human patients. When the patient associated with the process is human, -ve is inflected as -vi, by analogy with the most common transitive verb inflection, illustrated in example (24): (24)

(-human) axe khage yaage

(+human) axi khagi yaagi

pwe

ru

nok.

net

ERG

fish

'to see' 'to push' 'to look for' etc. Compare examples (25) and (26), as well as examples (27) and (28): (25)

I

thege-ve

3SG run-ASSOC

'The fish carried the net away.' (26)

I

thege-vi

3SG run-ASSOC

ye

ru

hooc.

him

ERG

horse

'The horse dragged him.' (Bril 2000: 327) (27)

I mu-ve hlaaleny nanamiwo 3SG stay-ASSOC the thought 'He stays with his thoughts.'

(28)

I

mu-vi

na.

3SG stay-ASSOC

me

'He adopted me.' (Bril 2002)

i ye. of him

340

Frangoise Ozanne-Rivierre

We saw in example (19) that ftie is not inflected before human objects, it is therefore clear that the link between this verb and the associative morpheme which stems from it is no longer perceived in Nelemwa. Grammaticalization is ongoing and as shown by Bril (this volume), the morpheme -ve /-vi is grammaticalized as an applicative associative case marker transitivizing suffix. We will see that in the neighbouring Nyeläyu language, the grammaticalization is even more advanced, as the associative morpheme -va, derived from the verb pha 'take' is becoming the obligatory transitive marker of certain causative constructions.

6.

Grammaticalization of the verb pha 'take' in Nyeläyu

Nyeläyu (Ozanne-Rivierre 1998), like Nelemwa, is a split ergative laguage. When the arguments are person markers, the construction is accusative (sVo). When the arguments are nouns, the construction is ergative but the unmarked word order is VAO, contrary to Nelemwa where, as we saw earlier, the word order is VOA. Nyeläyu ergative markers vary according to the agent's grammatical category: ru (+ common noun), an (+ proper name), a- (+ personal pronoun). Compare the Nelemwa example (29) with the Nyeläyu example (30): (29)

Nelemwa (Bril 2000: 66) (I)

thäli

(3Sg) close V

daan ru path Ο

ciiy-eena.

ERG tree-this.DElCT A

'This tree cut the road off.' (30)

Nyeläyu (Ozanne-Rivierre 1998: 53) Ta pavara

ru

uru

ti

3Sg break V

ERG

wind

the A

hada-yeek. branch-tree O

'The wind broke the branch.' In Nyeläyu, as in all the languages of the north, the transitive verb pha 'take' may be used autonomously (ex. 31), but when the object is a personal pronoun or a proper noun, the verb pha is inflected as phae (ex. 32):

The evolution of the verb 'take' in New Caledonian languages

(31)

Lha

pha

ca

pwa-ru

3P1

take

each

CLASS-two mat

341

dep.

'They each took two mats.' (32)

Iya! iya! taa-me phae-no! octopus octopus rise-towards take-me Octopus! Octopus! come take me!' (Ozanne-Rivierre 1998: 219) The verb 'take' can also be used in SVCs of the most common type in Nyeläyu: those containing a motion verb ('go', 'go up', 'go down', etc.) (tu tau 'to go down/to fish', ta yalaap 'to go up/to get food', etc.). This use of pha is illustrated in (32). See also example (33): (33)

Kam

ron

charemwa

ta

so

she+PERF

run

go up take

pha

nae-n. child-her

'So she ran up to get her child.' (Ozanne-Rivierre 1998: 213) Nyeläyu also has the applicative associative marker -va, derived from the verb pha 'take, carry', enclitic in the same posture and motion verbs as in Nelemwa: (34)

mo 'stay, reside' taabwa 'be sitting' cur 'be standing' äläp 'be lying down' charemwa 'run' hyao 'swim' taar 'flee' phuc 'fly'

mo-va 'be in charge of, adopt' taa-va 'take on one's lap' cu-va 'hold' (standing with)' äba [ämba] 'sleep with' ( < an + va) charee-va 'run with' hyao-va 'transporting swimming' taar-va 'flee with, abduct' phuc-va 'fly away with'

The resulting sequence is transitive and the nominal agent is in the ergative case. We see clearly in examples (35) and (36) that the morpheme -va cannot be interpreted as a "preposition": the newly introduced argument follows the agent in the ergative case and therefore occupies the position of grammatical object, since the word order in Nyeläyu is VAO (see 30). (35)

Ta taa-va an Carta nae-n 3Sg sit-TRANS ERG Rosana child-her 'Rosana is sitting with her child on her lap.'

(36)

Ta tha-phuc-va ru uru dao-yeek 3Sg PREF-fly-TRANS ERG wind leaf-tree 'The wind makes the leaves fly.'

342

Frangoise

Ozanne-Rivierre

But the grammaticalization of the morpheme -va is even more advanced than in Nelemwa since, as shown below, this enclitic "transitivizing" morpheme can also introduce the patient in causative constructions. In Nyeläyu, causative derivation is marked by one of two prefixes: a prefix pa- which is the reflex of the Proto-Oceanic *paka-, and a prefix pha-, more recently introduced, which is now becoming generalized. Syntactically, causative derivation increases the valence of intransitive (stative or active) verbs. The verbs thus derived receive different transitive markers (the suffixes -a or -en for example) but, in many cases, the transitive marker on derived verbs will be the suffix -va. Thus, in the examples (37) and (38), from the stative verb 'extinguished (referring to a fire)' one finds the following causative constructions: (37)

Ta pha-bo-va

a-ri

3Sg CAUS-extinguished-TRANS

ERG-who the

ti

naap? fire

'Who put the fire out?' (38)

Ta pha-bo-va

an

Coceep

ti

naap.

3Sg CAUS-out-TRANS

ERG

Joseph

the

fire

'Joseph put the fire out.' Here are a few examples of causative transitivization: hwek 'twisted' mavic 'broken' geek 'dirty' küüxü 'suckle' huu 'eatjntr' kam 'deviatejntr' wee 'be born' etc.

> > > > > > >

pha-hwek-va 'twist something' pha-mavic-va 'break something' pha-geek-va 'dirty something' pha-küüxü-va 'breastfeed' pha-huu-va 'make eat' pha-kam-va 'deviate^' pa-wec-va 'bring into the world'

If it is true that the transitive marker -va in causative constructions has the same origin as the transitivizing morpheme -va, of which we gave examples in (34), it is clear that, in this context, it has lost all semantic links with the verb 'take' from which it would have issued. Its grammaticalization to a transivitizing marker is complete. The shift may have been influenced by situations where the causative construction implied a "manipulation", in the true sense of the word, of an "object" held or carried by an Agent acting upon it. From that viewpoint, the verbs 'twist something', 'break something' in (39) imply such a direct physical manipulation, as does the last example in the list 'bring into the world (a child)' which

The evolution of the verb 'take' in New Caledonian languages

343

implies the intervention of a midwife (a-pa-wec vau 'midwife' (lit. who brings into the world)). Another argument pleads in favour of the verbal origin of the transitivizing suffix -va, which is that it receives the same case marker before proper nouns and personal pronouns as the verb pha 'take'. Compare the inflection of the verb 'take' in (32) (phae-no /take-me/) to the inflection of the transitive morpheme -va in example (40): (40)

Ta pha-käm-vae-no

naa

3Sg CAUS-deviate-TRANS-me from

tina

ma-nameru

li no.

the

NOMlN-think

of me

'He made me change my mind.' A last remark that may be made on the transitive marker -va in Nyeläyu is that it is spreading to all causative constructions. During my fieldwork I observed that young speakers had a tendency to regularize these forms on the basis of older derived causative forms. Here are the examples I came across: (41)

tabo 'fall'

'make fall' pa-raboa (elderly speakers) pha-tabo-va (young speakers)

caleng 'turn'

'make turn' pa-yalenga (elderly speakers) pha-caleng-va

cian 'disappeared'

7.

'lose'

(young speakers)

pa-ciäleen (elderly speakers) pha-cian-va (young speakers)

A special case: Cemuhi

All the languages of the north have a construction that expresses the association of an "object" with certain posture and motion verbs and, in all the languages examined so far, this syntactic relation is expressed by the transitive verb 'take, carry'. One might think that the association between the verb 'take' and an associative value is attributable to the entire northern group. However, one of the languages in the group, the tonal language Cemuhi (Rivierre 1980, 1994), neighbour of Paid, contradicts this hypothesis and is distinguished from all the others in that it has chosen a different lexeme to express the same relation: the transitive verb imwf 'seize, hold'. Nevertheless, Cemuhi also possesses a verb pe 'take, carry', obviously related to the other verbs 'take' attested in the other languages (ex. 42):

344

Frangoise

Ozanne-Rivierre

(42)

Kä

pe

έ

äli

bwels-gi

ίε-η

kä

ε

cipwä.

and 3Sg take the club-axe of him and 3Sg leave 'And he takes his ceremonial axe and he leaves.' (Rivierre 1994: 42) However, it is the atonal form imi, derived from the mid tone verb fmwf 'seize, hold', combined with the same posture and motion verbs as those of the other languages of the north, that allows the introduction of a concomitant accessory in Cemuhi: (43)

cü-fmi teb\v5 fmf tets imi ίέέίέ imi (älfbwät) cilä fmf

'standing holding sth.' 'sit with, sit holding sth.' 'run carrying sth.' 'leap with (the club)' 'flee with, abduct'

(< cüut 'stand') (sit/hold) (run/hold) (jump/hold) (flee/hold)

The languages of northern New Caledonia share a certain number of phonological innovations which have allowed their classification as a subgroup within the Neo-Caledonian group (Ozanne-Rivierre 1995). It is clear that through diversification, these languages have undergone independent changes in the domain of phonology as well as that of syntax and, likewise, they have come up with different solutions for the same conceptual needs. This is what has happened with the choice either of the verb 'take', or of the verb 'seize', to express the same associative relation and we must conclude that these choices have been made independently, after the languages had started separating. In the same way, languages which chose the verb 'take' to express this semantic relation, evolved separately in pushing this verb, as yet attested as a V2 serial verb in Paid, either towards lexicalization, or towards grammaticalization, the two sides of the same process called "morphogenesis" by Hagege (1993). From serialization, Nemi has gone on to verbal composition with a strong tendency for Vi and V2 to fuse into a simple transitive verb. In two languages of the far north, Nelemwa and Nyeläyu, the morpheme arising from the verb 'take, carry' has been reanalyzed and grammaticalized as an enclitic transitive marker.

8.

Conclusion

At the beginning of this paper, I pointed out that the grammaticalization of the verb 'take' into an object case marker in serial constructions has taken place in several language families and I mentioned the example of

The evolution of the verb 'take' in New Caledonian languages

345

Mandarin. According to Lord (1993: 96), this development of the verb 'take' can be explained by its highly transitive nature which implies an Agent voluntarily acting on concrete and movable objects. In so far as this verb marks a "prototypical" transitive action, with its two arguments Agent and Patient, one can understand how a term derived from it could end up as a transitive construction marker and may further be grammaticalized as a syntactic object marker. In all the examples cited by Lord, the verb 'take' is the first verb in a series. The sequence is iconic 'X takes Y and does something with it'. In the languages of New Caledonia, the verb 'take' is in final position and the sequence does not express successive but simultaneous events 'X is in such or such position (or is moving) taking (or carrying) Y'. Which is why, as opposed to the cases described by Lord, the verb 'take' has not evolved into a "preposition" introducing a patient, but rather into an applicative transitivizing morpheme, enclitic on the verb it formerly specified. The promotion to direct object case-marking of an accessory held or carried by an agent is common in Oceanic languages. Pawley (1986) gives other such examples in Fijian (Eastern Oceanic language) and Margetts (1999) mentions similar cases in Saliba (Oceanic language of Papua New Guinea). But in both cases, such promotion is marked by a transitivizing suffix which is a reflex of a Proto Oceanic applicative transitivizing suffix *-aki(ni)A, whose function was also to promote various peripheral arguments (such as instrument, comitative, cause, result, accessory, etc.) to direct object function. In Fijian and Saliba as in New Caledonian languages, it is verbs of motion and posture (sit, stand, run, climb, swim, etc.) that are most often associated with an accessory argument. What is original in New Caledonian languages is that, instead of using a reflex of the Proto Oceanic applicative transitivizing suffix, they make use of a semantically transparent transitive verb 'take, carry' for associative case-marking.

Abbreviations ASSOC CAUS CLASS

associative causative classifier

DEICT ERG

deictic ergative

NOMIN

nominalizer

PERF REV TRANS

perfective reversive transitive

346

Frangoise Ozanne-Rivierre

Notes 1.

2. 3.

4·

The critical examination of the criteria which are most commonly used to define [such constructions] show that, even within the very limited domain of African languages, all of them are contradicted by exceptions, except one, the most general, the criterion of the unique subject. Note that the object marking with ba requires the object to be definite (Lord 1993: 116). The development of a front vowel /ü/ results from the fronting of /u/ preceding the bilabial consonant /v/. Thus in contemporary Nemi 2 there is an opposition between tua 'to detach' and tiia 'to be straight'. In the case of tüe (< tu-ve), it is evident that the former compound is aligned on this phonological pattern. Two transitivizing suffixes are reconstructed in Proto Oceanic the suffix *-i which marks close objects corresponding to semantic patients and the suffix *-aki(ni) which marks remote objects such as instrument, source, result, accessory, etc. (cf. Pawley 1986).

Verbal compounds and lexical prefixes in the languages of New Caledonia Frangoise Ozanne-Rivierre and Jean-Claude Rivierre

Abstract The rise of "classifying" verbal prefixes specifying manner of action is not just a typological feature characteristic of some Austronesian languages of Papua New Guinea. A parallel development occurred in the languages of New Caledonia, but it is not due to word order shift (vo to OV) as in the languages of the Papuan Tip cluster or Madang Province (Bradshaw 1982, 1985). Serial verbs of the v v type, which are diachronically unstable, evolve into verbal compounds. In the languages of the Mainland of New Caledonia, the "classifying" verbal prefixes, whose verbal origin is often transparent, derive from compound constructions in which the first verb is reduced to its first syllable (or first mora). There is a continuum between the languages of the north, where such verbal prefixes are infrequent, and the languages of the south, where they proliferate. Analyzing the development of "classifiers" and co-lexicalizations, which contribute to the renewal of the lexicon, helps understanding this evolution. More specifically, the two tonal languages of the center-north of the Mainland (Cemuhi and Paid) point to the prosodic aspect of this evolutive process.

1.

Introduction

T h e languages of N e w Caledonia (including the Mainland and the Loyalty Islands) are often cited as being complex, irregular or even aberrant within the Oceanic group. Although it is true that the languages of N e w Caledonia, located in the far south of Melanesia, are not a m o n g the most canonical of the Oceanic group, the wealth of their phonological innovations should not o v e r s h a d o w the fact that they also contain very important conservative features for the reconstruction of Proto Oceanic (POc). Thus, in several languages, final consonants inherited f r o m Proto Austronesian are preserved, and in most languages the f o r m e r oral stops (uvular, velar, bilabial stops) do not undergo lenition in initial position but remain stops, etc. 1

348

Frangoise Ozanne-Rivierre and Jean-Claude Rivierre

Verbal compounds and lexical prefixes in the languages of New Caledonia

349

Being both conservative and innovative, New Caledonian languages are therefore important to historical linguistics as well as general linguistics, but these two aspects have mostly, up until now, been illustrated in the domain of phonology by specialists of these languages. We propose here to broach a morphosyntactic problem concerning the existence and development of lexical verbal prefixes (traditionally called "classificatory prefixes") within this language group. These prefixes give information on how an action is performed (with the hand, foot, mouth, by hitting, spearing, etc.). The existence of such prefixes has been noted in different places within the Austronesian family. In Formosa, Tsuchida (1990) has catalogued many in Tsou and Nojima (1996) describes their usage in Bunum, underlining the formal resemblance between certain prefixes and independent verbs. More recently, Adelaar (forthcoming) has suggested a diachronic explanation for their development in Siriya. In the Austronesian languages of Papua New Guinea, these prefixes are considered a typological feature of the Papuan Tip Cluster languages (Capell 1943; Ezard 1978, 1997). They are also attested in certain languages of Madang Province (Bradshaw 1982, 1985; Lichtenberk 1983). In both cases, the word order changed from SVO to SOV through contact with surrounding non-Austronesian languages where the SOV order is predominant. Also found in the New Caledonian Mainland languages are verbal prefixes which are semantically very close to the Western Melanesian classificatory prefixes. Nevertheless, their syntactic context is different as all New Caledonian languages have VO order (VOS or SVO) (Moyse-Faurie and Ozanne-Rivierre 1983). As suggested by Osumi (1995: 128), these prefixes, of often transparent verbal origin, probably stem from former nuclear-layer serializations. These constructions, diachronically unstable, frequently evolve towards verbal composition, and, in the languages of the Mainland, the lexical prefixes expressing the manner in which an action is performed clearly stem from these complex verbal constructions. We shall begin by presenting the process of verbal composition in the Northern Kanak languages where these lexical prefixes (henceforth LPs) are still infrequent. Next we will examine the situation in the Southern languages where these prefixes proliferate; at the same time we will try to determine which factors led to the multiplication of compound verbs within this group. But first of all, a rapid reminder of the constitution of "classi-

350

Frangoise Ozanne-Rivierre and Jean-Claude Rivierre

ficatory prefixes" attested in Western Melanesia and the hypotheses concerning their genesis.

2.

"Classificatory prefixes" in the Austronesian languages of Papua New Guinea

The term "classificatory prefixes" appeared for the first time in Capell (1943) to designate prefixes in the Austronesian languages of the Milne Bay Province which "define the manner in which an action is carried out by hand, by word of mouth, by poking holes, etc." (p. 252). Today these languages are classified by Ross (1988) with the Papuan Tip cluster, and most of them have verb final (SOV) clause order 2 . Classificatory prefixes are also found in certain Austronesian languages of Madang Province (such as Manam). These languages have also gone from SVO to SOV word order, influenced by the neighbouring Papuan languages. Between these two language groups, the Morobe Province languages have kept SVO order. These languages do not have classificatory prefixes but are rich in SVOV type "core layer serializations" (Foley & Olson 1985, Crowley 1987) which typically express a cause-result relationship. According to Bradshaw (1982, 1985), these causative serial constructions were the source of the resultative compound verbs - and the classificatory prefixes derived from them - which are found in the languages having changed from VO to OV order. Bradshaw (1985: 26) reconstructs two types of causative serial construction in the Papua New Guinea Oceanic languages: a) the "switch subject type" in which the second verb is intransitive and the object of the first verb is the subject of the second: SV t OVj ('they hit pig die'). b) the "same subject type" in which both verbs are transitive and both have the same subject and object: SVtOVt ('they hit pig kill/causedie'). In both constructions, the first verb describes the manner in which an action is performed and the second indicates the result. According to Bradshaw (1985: 28), these SVOV serial causatives can give rise to different analyses according to whether VO or OV order is considered the norm, and two evolutions are therefore possible:

Verbal compounds and lexical prefixes in the languages of New Caledonia

351

1. In SVO languages, the second verbs of the series may lose their verbal status and become resultative particles: SVmanner OVresult >

SV m a in Ο RESULT

This is what happened in Numbami for example where the element uni (< POc *punu(q) 'kill, hit') no longer functions as an independent verb. It may still mark the result of the action as in example (1). (1)

Numbami

Ti-lapa 3Pl-hit

bolo pig

uni. dead

'They killed the pig.' (Bradshaw 1985: 26) But in example (2) one sees that this particle only denotes the intensity of an action: (2)

Numbami

Awa

ni-nggo

aiya

uni.

mother Ir.3Sg-say 2Sg dead 'Mother is going to scold you to death.' (Bradshaw 1993: 138) 2. In languages where the word order has gone from SVO to SOV, it is the first verbs in the series which tend to lose their verbal status to constitute a class of "classificatory prefixes": SVmanner OVresult >

S Ο MANNER-V ma in

This is what happened in Manam, a language spoken in Madang Province, for which Lichtenberk (1983: 214-217) notes ten "classificatory prefixes". All of these prefixes except one have corresponding independent transitive verbs but, as we can see in list (3), these prefixes cannot take the thematic consonant attached to the corresponding independent verb. (3)

Manam classificatory prefixes ?ara-

'with the teeth' (< ?arat 'bite')

?in?oro-

'with the fingers' (< ?int 'pinch')

tara-

'a cutting action' (< ?orot 'cut') 'by hitting with the sole of the foot' (< dua 'hit with the sole of the foot') 'a chopping action' (< tara 'chop')

zarj-

'a hitting and breaking action' (< zarj 'pound')

rau-

'a hitting (not necessarily breaking) action' (no independent verbal counterpart) 'a throwing and breaking action' (< tata 'hit sth. against sth. else')

dua-

tata-

nagu- 'a pricking, piercing action' (< nagur 'prick') ro?a-

'a throwing action' (< ro?a? 'throw')

352

Frangoise Ozanne-Rivierre and Jean-Claude Rivierre

These prefixes, combined with action verbs, indicate how an action is performed as shown in example (4): (4)

Manam

Moli i-?ara-sisi?-i. orange 3Sg.realis-CLAS-peel-3Sg.Object 'He peeled the orange (with his teeth).' (Lichtenberk 1983: 215)

In the Papuan Tip languages, several classificatory prefixes are clearly linked to full transitive verbs attested in the same language. But others have lost all link with the corresponding independent verb and their original meaning can only be recaptured through comparison with other languages. This is the case for the prefix /w-'*hit' in Iduna in example (5): (5)

Iduna

Bawe hi-lu-ve-'alika-na. pig 3Pl-*hit-cause-die-3Sg 'They killed the pig.' (Bradshaw 1985: 26)

In the Mainland languages of New Caledonia, one finds LPs which are semantically very close to those mentioned by Lichtenberk in Manam but, as we noted in the introduction, it is impossible to correlate the development of these prefixes with a change in word order. The syntactic context is different. All New Caledonian languages have a VO word order [VOS or SVO] and "nuclear-layer serializations" [V1V2] are predominant. These languages therefore constitute a counter example to Foley and Olson's prediction (1985) according to which VO languages are prone to developing "core layer junctures" between serialized verbs, whereas "nuclear layer serializations", without any intervening arguments between serialized verbs 3 , are prevalent in o v languages. In New Caledonia, as in Western Melanesia, these LPs are most probably derived from former compound constructions indicating the manner of action, but unlike the languages of Papua New Guinea where LPs combine with free form verbs, we will see that in the Kanak languages of New Caledonia, these prefixes are preferably linked to "bound verb stems", generally expressing the result of the action initialized by the prefix. In other words, both elements forming the compound have often lost their status of independent verbs 4 .

Verbal compounds and lexical prefixes in the languages of New Caledonia 3.

353

C o m p o u n d verbs in the Mainland languages of N e w Caledonia

All the languages of the Mainland show a small set of lexical verbal prefixes (LPs) whose forms are related. Originally, these prefixes probably acted both as independent verbs and as the first element of c o m p o u n d verbs expressing the gesture accompanying the action. The correspondences illustrated in (6) confirm the c o m m o n origin of some of them: (6)

a ) ' W i t h the teeth' [< POc *kati 'husk with the teeth' (Ross et al. 1998: 259) 5 ] Northern languages:

cai- (Nemi, Fwäi, Jawe, Pije, P w a p w ä ) yai-lzai- (Pwaamei) ce(e)- (Nelemwa) yee- (Nyeläyu) έ- (Cemuhi), e v e r b /e- (Paici)

Southern languages:

ka- (Ajie), ks- (Xäräcüü, Xärägure) e- (Tin) γα verb/γα- (Numee), γα- (Dumbea)

b) 'With the h a n d ' [< PEO *taur(a) ' h o l d ' (Grace 1969)] Northern languages:

to- (Nemi, Fwäi, Pije) co- (Jawe, Pwaamei, P w a p w ä , K o n e dialects)

Southern languages: tuWTb/tu- (Ajie) c) Ά hitting action' [< PEO *sasa 'hunt, thrash, a w h i p ' (Grace 1969)] Northern languages:

Southern languages:

tha- (Nyeläyu, Kone dialects) the- (Nemi, Fwäi, Pije) tä- (Paici, Cemuh!) yavcrh/ya- 'hunt, hit' (Ajie) savab/sa-

'hunt, hit' (Xäräcüü)

to verb//α- (Tlri) j e verb 'hit' (Drumbea, N u m e e ) Furthermore, one sees in example (7) that certain bound forms, attested as the second component in different languages, stem f r o m former independent action verbs:

3 54

Frangoise Ozanne-Rivierre and Jean-Claude Rivierre

(7) POc*piti(k) 'crack' (Ross et al. 1998: 276) Northern languages: -'"biyi (Nelemwa, Nixumwak) -mbic (Nyeläyu) -mbii (Nemi, Fwäi, Jawe, Kone dialects) -ii (Cemuhi) Southern languages: -vii (Ajie) POc *pwalaq 'split' (Ross et al. 1998: 260) Northern languages: -vgwana (Pwapwä) -vgane (Nemi, Fwäi, Jawe, Pwaamei) -mbwana (Kone dialects) -mbw5ne (Cemuhi) POc *sua 'spear (V, N)' (Ross et al. 1998: 265) Northern languages: -"dua 'pierce' (Nyeläyu, Nemi, Fwäi, Jawe, Pwaamei) POc *qatu(rj) 'strike from above, pound' (Ross et al. 1998: 271) Southern languages: -yai 'pound, crush' (Ajie, TM), -kai ~ -vgai (Xäräcüü) Our hypothesis is that there existed in Proto Mainland a small number of verbal compounds expressing the relation gesture + action having a shortened transitive verb as the first element. Stemming from this common nucleus, the languages of the Northern group and the languages of the Southern group developed this formal compound pattern in two different ways, which is what we shall now look at.

4. Verbal compounds in the Northern languages One finds in all the Northern languages a small paradigm of LPs expressing movement, as in Nemi for example, where four prefixes of this type are attested (Ozanne-Rivierre 1979: 47), as shown in example (8). The etymology is known for three of these prefixes, and the one expressing the action 'by hand' to- has the corresponding independent verb thoot 'hold'. 6 (8) Nemi LPs cai-

'with the teeth'

to-

'by hand (without percussion)' [< PEO *taur(a) 'hold']

[< POc *kati 'husk with the teeth']

tee-

'action from above, with the foot or a tool (stick, pestle, spear)'

the-

'by hitting, projecting'

[< PEO *sasa 'hunt, thrash, a whip']

Verbal compounds and lexical prefixes in the languages of New Caledonia

355

Certain prefixes may be combined with independent transitive verbs, but more often they are associated with verbal forms having lost all lexical autonomy. Example (9) illustrates the combinations of the prefix to- 'by hand' with bound verbal forms: (9) Nemi to- 'by hand' (without percussion)' + bound verb forms + -mbalii 'drive in' + -'"bii 'break' [< POc *piti(k)] + -mbune 'cut' + -"dua 'pierce' [< POc *sua] + -'Jgane 'split' [< POc *pwalaq]

to-mbalii 'drive in by hand' to-mbii 'crush by hand, grind (coffee in a mill)' to-mbune 'break (by hand), tear in two (fabric)' to-ndua 'pierce with the finger' to-"gane 'separate, incise, split (with a knife)'

+ -leei 'push, make fall' to-leei 'push with the hand' The same prefix may be combined with a few independent transitive verbs as in (10): (10) Nemi to- 'by hand'+ independent transitive verbs + hyuu 'break (by folding)'

to-hyuu 'break with both hands (brushwood)'

+ uri 'pull, remove (from a bag, hole)' to-uri 'extract by hand (ex. a root)'

The four prefixes indicating the mode of action can therefore be combined with the same verb form to express different activities of daily life as in example (11), with the bound forms -'"bii 'crack, break' and -mbune

(11)

'cut':

Nemi a)

b)

cai-mbii 'break with the teeth (ex. a nut)' to-mbii 'crush by hand, grind (coffee)' tee-mbii 'break by foot, crush with a pestle' the-mbii 'break by percussion (ex. glass, shell, coconut, etc.)' cai-mbune 'cut with the teeth, gnaw to cut' to-mbune 'break (by hand), tear in two' tee-mbune 'cut (wood on the ground)' the-mbune 'cut with an axe (ex. a tree)'

These Nemi examples of transitive compound verbs illustrate well what is found in the Northern languages on the whole. The same prefixes

356

Frangoise Ozanne-Rivierre

and Jean-Claude

Rivierre

associated with the same verbal roots (bound or free) are found in all the central Northern languages (Hienghene and Kone regions). In the far north, in Nelemwa, Bril (2000, 2002) mentions three LPs which specify the manner of an action (ce(e)-, 'with the teeth', ta(a)- 'by pricking', khii- 'by hitting') and the same prefixes are attested in Nyeläyu (Ozanne-Rivierre 1998) (yee-, 'with the teeth', tee- 'by pricking', tha- 'by hitting'). All the Northern languages therefore have in common a small stock of transitive compound verbs where the first element specifies the manner of action. But another compound sequence is also common to this group of languages, where the first element is a shortened posture verb (standing, sitting, lying down) which may be combined with either independent verbs (active or Stative) or modifiers. This is illustrated in example (12). (12) Nemi posture verbs tuut 'be standing' (< POc *tuqud) > tu+ hoa 'watch over' tu-hoa 'watch for' + mwanji 'attack' tu-mwafji 'block in an ambush' + halam 'measure with arm span' tu-halam 'standing arms spread'

ta'igo 'be sitting' > ta+ teea 'lean' + oomba 'in hiding, in secret' + k«oneetj 'be immobile'

ta-teea 'sit leaning' ta-oomba 'sitting in watchfulness' ta-k'ioneerj 'sitting without moving'

k'ionap 'be lying down' (< POc *qenop) > k''o+ uren 'be black, dark' + thi-hi 'lean on one's elbow' + kala"da 'upside down'

kno-uren 'go to bed without eating'7 ko-thi-hi 'lying leaning on one's elbow' ko-kalanda 'lying on the back'

This pattern with abbreviated posture verbs is very productive in all the Northern languages without exception, whereas it has not been attested in any of the Southern languages. This may be a shared innovation in the Northern group. Another composition structure, much rarer and limited to certain languages, concerns incorporated object constructions involving a transitive verb reduced to its first syllable or mora. This formal type of construction, unattested in the Far Northern languages, is more frequent in the Hienghene region languages, and is especially developed in Paid. Some Nemi examples are given in (13):

Verbal compounds and lexical prefixes in the languages of New Caledonia (13)

357

Nemi incorporated object constructions thi"'be 'peel' thi-hyoonik 'peel niaoulis (Melaleuca sp.)' kaai 'hit' ka-nuk 'fish' (hit-fish) cevi 'pick (by pinching)' ce-"do 'pick leaves' (pick-leaf) pai 'plait, weave' (< POc *patu) pa-k'Jam 'weave mats' (plait-mat) hwambwe 'uproot' hwa-yaat 'uproot hay' (uproot-Imperata sp.) hwii 'eat meat' hwi-aman 'eat (intransitive)' (eat-thing)

To finish, some verbal compounds where the first element is an intransitive verb reduced to its first mora are sporadically found in certain languages. But these reduced verb forms have no productivity, and to call them "classificatory" prefixes is meaningless. It is the formal pattern which is pertinent and it is the productive compound forms described above which serve as matrix. (14)

Nemi t"oon 'run' > t"opala 'speak' >pa-

t"o-pila 'trot (horse)' (run-dance) t"o-ve 'run away with' (run-take) pa-mbway 'speak through the nose' (speak-tight) pa-hyaayo 'murmur' (speak-noiselessly)

pila 'dance'>pipi-taniy 'surround while dancing' (dance-surround) Compound verbs of this type are not very numerous in Nemi, but it is clear that the sequence run-dance in the first example for 'trot (for a horse)' is a relatively recent creation. It is entirely plausible that this composition pattern with a shortened first verb, still limited in the 1980s, continues to develop. To conclude, one may say that the number of LPs common to all the Northern languages is limited - those that specify the manner of an action (which number between four and six) - as well as those stemming from abbreviated posture verbs. On the other hand, in certain languages, the formal process of abbreviation of the first verbal element making up these compounds has spread to certain active intransitive verbs and incorporated object constructions.

5.

Verbal compounds in the Southern languages

The situation radically changes when one examines the Southern languages as a whole, as well as Paid, an intermediate tonal language genetically related to the Northern group (Ozanne-Rivierre 1995) but which shares

358

Frangoise Ozanne-Rivierre

and Jean-Claude

Rivierre

certain typological features with the Southern languages. Among these features are the loss of the final consonants still attested in most of the Northern languages, and a well developed verb composition process (Rivierre, nd.). Furthermore, most of the Southern languages no longer reflect POc consonants in medial position except for apicals 8 . These languages have only open syllables and words have undergone considerable attrition. Monosyllables are very frequent, and are only distinguished by vowel timbre, length, orality or nasality. This tendency towards monosyllables, historically attested in the Southern group, seems to have favoured the multiplication of compound verbs, as though numerous short verb forms were formally apt to generate compound verbs, as the first element of a compound. The following observations concern above all the best documented Southern languages. They are based upon several recent descriptions: Osumi (1995) on Tin, Moyse-Faurie (1995) on Xäräcüü and the recently published dictionary of a group of Ajie speakers (Aramiou et al. 2001) 9 . These monographs highlight the importance of verbal compounding and the proliferation of LPs, of which we saw a few examples from the Northern languages. Contrary to what we have just seen for the Northern languages, the Southern languages have not developed compound verbs from posture verbs, nor from incorporated object compounds. On the other hand, they have multiplied those compounds where the first element specifies the manner of action. We will now speak of this stock of compounds.

5.1.

LP multiplication in the Southern languages

As early as 1935, Leenhardt underlined the importance of compound verbs in his study of the Houailou language. He wrote: "... nombre de radicaux evoquent des gestes precis de la main et du pied... Parmi les verbes, une serie remarquable de verbes monosyllabiques exprime les mouvements des membres. lis deviennent les radicaux de la plupart des autres verbes polysyllabiques" (Leenhardt 1935: 7-8). 10

He gives a list of monosyllables expressing different arm movements and adds a group of figures illustrating the accompanying gestures. These prefixes, according to Leenhardt, form the basis of most Ajie compounds.

Verbal compounds and lexical prefixes in the languages of New Caledonia

359

Arm movements and corresponding monosyllables in Ajie (M. Leenhardt 1935: 9)

The Northern languages have only two or three prefixes involving the hand or arm, whereas Leenhardt presents almost fifteen. This remarkable enrichment also spreads to other movements. Those implicating the mouth for example, where one not only finds a prefix for 'biting' as in the north, but also 'holding in the mouth', 'chewing', 'sucking', 'spitting out', etc. Many prefixes concern technical gestures, either with a tool (awl, spear, knife, etc.) or without. The corpus of compound verbs also applies to perception and judgement verbs ('see', 'measure', 'count'). It also includes verbs expressing the action of natural elements such as water, fire, sun, wind, thunder and plants. In all, several dozen prefixes indicate, in each of the Southern languages, the manner in which a result is obtained. The growth of the stock of compounds therefore took place in the same manner + result semantic field. It also strictly conforms to the formal framework which implies an initial monosyllabic prefixed element since each of these LPs is reduced to a single mora. Contrary to Leenhardt's beliefs, not all LPs function as independent verbs. Some, from among the most productive, do not have (or no longer have) lexical counterparts. One may suppose that they were autonomous

360

Frangoise Ozanne-Rivierre and Jean-Claude Rivierre

verbs which were rivalled by the compounds they gave rise to, and were finally eliminated by these compounds.

5.1.1. LPs without a lexical

counterpart

In the Southern languages, LPs without a lexical counterpart are mostly movement verbs involving body parts. This is the case, for example, of the prefix 'by teeth' in (15), stemming from the POc *kati 'husk with the teeth" 1 . (15)

LPs involving body parts

Ajie 'by teeth' ka'with fingers' ka'by pressure (with the body)' ko'by leg' Φ'by fist' J>ja'by beak" 2 "do-

5.1.2. LPs having a lexical

Tiri Xäräcüü ekeokewa- kwafo-"je"do- "di"do- J'jo- 'by pecking, by pricking'

counterpart

The LPs stemming from free verbs are very numerous but unequally productive. They stem from two sources: First, from the abbreviation of verbs having the structure CVI, CVV or CVCV reduced to their initial mora when they become the first element of a compound. Examples of these reductions are given in (16). (16)

a) CV: > CV-

Paid Ajie Xäräcüü TM

ρ Γι 'attach' cii 'knock' ζεε 'pull' yii 'push sth. forward' yii 'grate' fuu 'pound' koo 'vomit'

> >

ρΐct-

> >

yi-

> > >

γίfuko-

Verbal compounds and lexical prefixes in the languages of New Caledonia

CVV > c v Ajie Xäräcüü Tiri

poi 'attach' yio 'cut (with knife)'

>

ß.i 'attach' cue 'put'

>

hoi 'attach' tia 'tear up' etc.

>

CVCV > CVpweti 'chew' Paici [ri 'dig' Ajie cari 'rub, polish' ogwere 'throw' Xäräcüü pira 'drill' TM po(p 'turn' te^e 'tie a knot' etc.

> >

poyiAeu-

>

hoti-

>

pwe-

>

Γ-

>

ca-

>

ogwepi-

> > >

361

pote-

Secondly, LPs stem from short, monomoraic independent verbs, which are, in a way, preformated to fill the first slot of a compound - as if their form and their meaning predisposed them for this usage. A few examples are given in (17) (17)

cv>cvPaici Ajie Xäräcüü TM

ηί 'collapse' ogf move'

>

ki- 'dig'

>

m

bA 'stir, move' •T'jE 'hang' rju 'move, fidget'

>

kim bA-

>

fije-

>

"de 'hang' ogu 'move' β 'pull out (taro)' etc.

>

w"degoP-

>

> >

ηΐOgf-

There is a contrast between the Northern languages where the LPs, with the exception of Paid, are still infrequent and the Southern languages where they proliferate. Osumi (1995) lists thirty-four prefixes in TM but Moyse-Faurie (1995) lists at least twice as many in Xäräcüü. Both authors seem to agree that their inventory is exhaustive, or nearly so. In Paici and

362

Frangoise Ozanne-Rivierre and Jean-Claude Rivierre

Ajie where documents (dictionaries, texts) are abundant, it seems that the inventories are open-ended. These two languages have a plasticity that enables them either to shorten verbs (as in Paici) or to promote any short verb to the role of LP (in Paici and Ajie), thereby multiplying the possibilities of forming compound verbs.

5.2. The second element in compound verbs The second element in a compound may be more or less tightly bound to the LPs we have just examined. First, let us see the most frequent, which constitute a single phonological word.

5.2.1. Atonal or unstressed second elements In all the Southern languages, the second element of a compound can be chosen among twenty or thirty unstressed forms, with a resultative value. The same notions are found across languages, even when the words are not related. They may be grouped around two axes: - action on the form, integrity of the object (break, cut, split, crush, tear, etc.) - action on the place, position, or trajectory of the object (return, block, tear off, detach, etc.) In all the languages, the combinatory possibilities of the two components are highly variable. In Ajie, for example, the prefix ta- 'with the foot' may occur in around twenty compounds (example 18), whereas the prefix "da- 'evaluate, measure' only occurs in three compounds (example 19) [Aramiou et al. 2001]. (18)

Ajie

ia- 'with the foot' (< tea 'put the foot down') ta-^uru trampling' td-vari td-ogewe 'turn over with the foot' ta-verii ta-luuwiri 'push underwater with ta-νία the foot' ta-vii ts-mwaa 'gather with the foot' td-vio

ta-uru ~ tddru 'make appear by

ta-tia ~ td-ria

'rip with the foot,

wound one's foot'

td-vui

'pierce with the foot' 'scrape with the foot' 'trip, slide' 'block with the foot' 'break with the foot' 'crush with the foot (as with plants)' 'use the foot as a lever'

Verbal compounds and lexical prefixes in the languages of New Caledonia ta-fpra ta- [argt3-i;iri

'break, cut in two with the foot' 'prick one's foot with something' 'pick with the foot (coconuts)'

ta-wiri ta-yai ta-γενε ta-yui

363

'press, hold down with the foot' 'crush with the foot' 'put the foot in the right spot' 'spear through with the foot'

(19) Ajie "da- 'action of evaluating, measuring' (< "dai 'aim') "da-rhara 'establish the limits of a lot' [measure-cut in two] "da-ngewe 'establish false limits' [measure-turn over] "da-wiri 'denigrate someone' [measure-keep down]

Most of these enclitic elements do not have (or no longer have) a corresponding independent verb, but some of them reflect former verb forms. Thus the bound forms -yai (Ajie, Tin), -kail-"gai (Xäräcüü) 'crush, pound' reflect the POc verb *qatu(y) 'pound, strike from above' (see example 6). In some cases, the verbal counterpart of a bound form is still attested in one or another of the neighbouring languages (ex. - / a ß 'untie (bound form)' in Tin, but fa 'untie (free form)' in Xäräcüü). In other cases, both forms (bound and free) coexist within the same language. The bound verb form is then unstressed. Such is the case of the verb 'split, tear' in example (20): (20)

'split, tear' free form bound form

Ajie tia -tial-ria

Xäräcüü tia -tia ~ -"dia

TM tia -ria

LP+bound root forms a single phonological word stressed on the first syllable, that is on the LP. In the tonal languages (Cemuhi and Paid) situated on the border between north and south, the whole compound takes the tone of the prefix (ex. 21): (21)

Paid

Mid tone LP c5- 'with the foot' + -"dari 'split' (atonal) —> c5-"dÄri 'split by stepping on' High tone LP CA- 'exert pressure on' n

+ -"ώπ 'split' (atonal)

-> cÄ- di(ri 'split by pressure, crack something by sitting on it' The integration of the second component may be further marked on the segmental level. For example:

364

Frangoise Ozanne-Rivierre and Jean-Claude Rivierre

1) Prenasalization of stops in Xäräcüü If the prefix contains a nasal vowel, the bound form's oral stop becomes prenasalized. Moyse-Faurie (1995: 239-43) cites approximately twenty forms which follow this rule: (22)

Xäräcüü -tia ~ -ndia 'split, tear' ke-tia 'tear with the teeth' (by teeth-tear) but: k§-"dia 'tear, split with the fingers' (with the fingers-tear) -koro ~ -vgoro 'break, shatter' m bi-koro 'grind (ex. coffee in a mill)' (turn-break) ke-koro 'break with the teeth' (by teeth-break) but: ke-ogoro 'shatter with the fingers' (with the fingers-break) etc.

2) Lenition of oral stops in Ajie In bound verb forms, the initial stop of an independent verb undergoes lenition when unstressed : (23)

Ajie 'split, tear' 'peel, skin' 'lift up' 'touch, run into sth.' etc.

independent verb tia pari pui ke

5.2.2. Stressed or tonic second

bound form -tia ~ -ria -vari -vui -γε

elements

Free form verbs (stressed or tonic), without any segmental linking marker, may also be combined with different LPs to form compounds. This is attested in all of the languages and, even if limited, still further enriches compound possibilities. Thus in Ajie (example 24), the initial consonant of free form verbs combined with the prefix td- 'with the foot' does not undergo lenition. (24) Ajie ta-piyari 'knock over with the foot' ta-pere 'push away with the foot' to-coo 'detach with the foot'

\piyari 'capsize'] [pere 'set aside'] [coo 'untie']

Verbal compounds and lexical prefixes in the languages of New Caledonia

365

Stress hierarchy of such juxtaposed units is attested in Xäräcüü (see Rivierre 1978). It is probable that this kind of hierarchical organization is also attested in the various Southern languages, but prosodic documentation is as yet insufficient. In the tonal Northern languages (Cemuhi and Paid), each element of the compound remains tonic. In Paid (Rivierre 1974), for example, when both the prefix and verb have mid tones, there is even a boundary between these units: the first mora of the second element takes a high tone in this context. Thus, in example (25) two mid tone elements such as the prefix c5- 'with the foot' and the independent verb üra 'project, make fall' are therefore pronounced c5-urä 'make fall with the foot' when they form a compound 13 . (25)

Paid co- 'with the foot' + üra 'make fall' —> cö-ürä 'make fall with the feet' cä- 'with the flat of the hand' + äta 'rub' —> cä-ätä-ri 'massage'

5.2.3. Consecutive

constructions

Finally, certain constructions take on a consecutive or purposive meaning, and may be joined to different prefixes to make up compounds of a new type. Let us take a quick look at this last compound pattern in several languages. 1) Paici In Paid, these compounds are constructed as follows: prefix + mba 'so that' followed by a verb showing the result: [LEXICAL PREFIX + mba + VERB]

This construction may be considered as the compact expression of a consecutive subordinate clause containing, in its explicit form, a full, tonic subordinator (mbänÄ 'so that, in order to'), as in example (26): (26)

Paid

tü '"bäm e nye tüu put one's hand so that it assertive fall —> tü-mbä-tüu 'make fall, let fall from one's hands' In this type of construction, mba- (instead of mbäm) loses its own tone to adopt that of the prefix. In principle, the verb introduced by mba- remains tonic as in example (27):

366

Frangoise Ozanne-Rivierre and Jean-Claude Rivierre

(27)

Paid tä-mbä-püu 'hit make lean' cÄ-mbä-tüu 'make fall by exerting pressure' i-'"bd-tyu 'saw to make fall' s-":ba-piiu 'saw and make lean' tü-mbä-nÄmbwe 'put the last hand to' tü-mbä-iri 'empty out' etu 'bail, empty water' > εi-'"bä-mw,\ri 'empty, dry' ti'pi 'untie, detach' > tfti~-'"bä-ewa 'loosen the hold'

[bump-so that-lie down] [with pressure-so that-fall] [saw-so that-fall] [saw-so that-lie down] [with the hand-so that-finish] [with the hand-so that-empty] [bail-so that-dry] [untie-so that-take off]

However, in certain cases, the verb itself becomes atonal. Furthermore, some P a i d atonal forms have arisen from these constructions. Such is the case of the fossilized consecutive forms -mbau (still considered by speakers as the equivalent of mba-küu 'so that-make noise') and -"'boaa (which comes from mba-wäa 'so that-open'). See example (28) (28)

Paid cä-mbäύ 'produce sound by slapping hands' cö-mbäü 'produce sound by tapping the feet'

[by slapping-make noise] [with the foot-make noise]

cä-mboäa 'split by slapping' £-mboaa 'open with the teeth'

[by slapping-open] [with the teeth-open]

2) Ajie In Ajie, numerous compounds are constructed according to the same pattern, with the particle νε 'so that, in order to': [LEXICAL PREFIX + νε + V E R B ]

The form νε is described as a particle with a purposive or consecutive meaning. This construction therefore seems close to that in Paid 1 4 . Following LPs, it makes up numerous set compounds and constructions. The authors of the Ajie dictionary [Aramiou et al. 2001] show several hundred entries in their dictionary. A few examples are given in (29): (29)

Ajie (LP + νε + V) o- (< pa 'water') ρ-νε-mii 'put out by watering'

[water-ve-kill]

Verbal compounds and lexical prefixes in the languages of New Caledonia ka- 'with the fingers' ka- ve-ndayi 'reduce to crumbs' ka- ve-mbA 'stir' ka- νε-γΛ 'snap fingers' ta- 'with the foot' ts-ve-ko 'make resonate by walking' ta-vE-yu 'make collapse with the foot' etc.

367

[with fingers-ve-crumble] [with fingers-ve-move] [with fingers-ve-snap] [by foot-ve-resonate] [by foot-ve-collapse]

The combination of νε with a stative verb tends to take on an adverbial meaning (example 30): (30)

Ajie ka- 'with the fingers' ka- νε-3 'to arrange carefully' ta- 'with the foot' td-ve-ngiri 'to stand firmly'

[with fingers-ve-good] [with the foot-ve-strong]

3) Xäräcüü In Xäräcüü, as in the other languages of the far south (Xärägüre, Numee, Drumbea), one finds the same type of compounds. In all of these languages, the morpheme a, which introduces the second element of a compound, has the same values (consecutive and adverbial) as in Ajie 15 . [ L E X I C A L PREFIX + a + V E R B ] ,

Once again, various binding markers show that the elements tend to merge. Thus in Xäräcüü (Moyse-Faurie 1995: 200), the adjunction of the sequence a + V E R B very often triggers the lengthening of the prefix vowel (example 31): (31)

Xäräcüü bi- (< mbiri 'twist, turn')

m

m

bi-a-ke 'screw tightly' mo- 'with the foot' mDD-te 'tap with the foot'

[twist + a + tight] [with the foot + a + tap]

fi- (