Topics in Oceanic Morphosyntax 9783110259919, 9783110259896

Citation preview

Topics in Oceanic Morphosyntax

Trends in Linguistics Studies and Monographs 239

Editor

Volker Gast Founding Editor

Werner Winter Editorial Board

Walter Bisang Hans Henrich Hock Heiko Narrog Matthias Schlesewsky Niina Ning Zhang Editors responsible for this volume

Volker Gast Walter Bisang

De Gruyter Mouton

Topics in Oceanic Morphosyntax

Edited by

Claire Moyse-Faurie Joachim Sabel

De Gruyter Mouton

ISBN 978-3-11-025989-6 e-ISBN 978-3-11-025991-9 ISSN 1861-4302 Library of Congress Cataloging-in-Publication Data Topics in Oceanic morphosyntax / edited by Claire Moyse-Faurie, Joachim Sabel. p. cm. ⫺ (Trends in linguistics. Studies and monographs; 239) Includes bibliographical references and index. ISBN 978-3-11-025989-6 (alk. paper) 1. Austronesian languages ⫺ Morphology. 2. Austronesian languages ⫺ Syntax. 3. Austronesian languages ⫺ Grammar. I. MoyseFaurie, Claire. II. Sabel, Joachim, 1962⫺ PL5033.T67 2011 4991.5⫺dc23 2011031600

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” 2011 Walter de Gruyter GmbH & Co. KG, Berlin/Boston Typesetting : PTP-Berlin Protago-TEX-Produktion GmbH, Berlin Printing: Hubert & Co. GmbH & Co. KG, Göttingen ⬁ Printed on acid-free paper Printed in Germany. www.degruyter.com

Table of contents

Introduction Claire Moyse-Faurie and Joachim Sabel

1

Part one: Sentential syntax and sentence types Deriving linear order in OV/VO languages: evidence from Oceanic languages Joachim Sabel Questions and answers in Niuean Diane Massam, Donna Starks and Ofania Ikiua

27

65

Questions and word order in Polynesian Eric Potsdam and Maria Polinsky

107

Nominalization and exclamation in Oceanic languages Claire Moyse-Faurie

135

Part two: Nominal morphosyntax Two or three things in the Unua noun phrase Elizabeth Pearce

163

Noun incorporation in Saliba Anna Margetts

203

Noun-phrase conjunction in Austronesian languages: additive, inclusory and comitative strategies Isabelle Bril

235

vi

Table of contents

Part three: Historical developments Neither accusative nor ergative: an alternative analysis of case in Eastern Polynesian Yuko Otsuka

289

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect Jacques Vernaudon

319

Subject index Language index

341 343

Introduction

The Austronesian language family consists of some 1200 genetically related languages dispersed over an area encompassing Madagascar, Taiwan, Southeast Asia, and islands of the Pacific. The Oceanic language group, comprising at least 450 languages, is a large subgroup within this language family. The speakers of these languages inhabit the geographical area encompassing Melanesia, Micronesia and Polynesia. Oceanic languages exhibit a number of typologically and theoretically interesting morphosyntactic properties, many of which have been relatively little studied and are still under-explored. In this introduction, we will discuss some relevant phenomena that are also addressed in the articles in this volume, with particular reference to the form and function of sentences, matters of nominal morphosyntax and historical developments in several domains. For further discussion of morphosyntactic aspects of Oceanic languages, the reader is referred to Lynch (1998), Lynch, Ross, and Crowley (2002), and Ross (2004b). For comparative and typological studies on specific domains such as serial verb constructions, complex predicates and negation see Bril & Ozanne-Rivierre (2004), Crowley (2002), Senft (2008), Hovdhaugen & Mosel (1999), and for a discussion of the linguistic representation of space and deixis see Bennardo (2002) and Senft (2004). Oceanic languages display VSO, VOS, SVO and SOV constituent order. As stated by Ross (2004b:495), almost all of the verb-final languages “are located on or near the mainland of New Guinea and belong to the North New Guinea and Papuan Tip linkages.” These verb-final Oceanic languages have been in contact with SOV non-Austronesian (Papuan) languages and their non-canonical constituent order is undoubtedly due to this contact. Furthermore, Oceanic languages have subject/object agreement markers and additional verbal markers for transitivity, special coordination strategies, free word order and special possessive marking. Oceanic languages make use of noun-incorporation, use several possibilities of forming wh-questions such as wh-in situ, wh-ex situ and partial wh-movement and they display interrogative verbs. Other notable features include the formation of serial verb constructions, extensive use of nominalizations, and sentence type markers. As far as their alignment systems are concerned, Oceanic languages have nominative-accusative as well as ergativeabsolutive case marking.

2

Introduction

The articles published in this volume bring together comparative work on some of these morphosyntactic features of Oceanic languages, which are either subject to variation or show uniformity within the family. Most of them are revised versions of papers given at the Seventh Conference on Oceanic Linguistics (Numea, New Caledonia, July 2007). The book is divided into three parts. The first part discusses the form and function of sentences in Oceanic languages, the second addresses selected topics in nominal morphosyntax and the third deals with historical aspects of the morphosyntax of Oceanic languages. The data in the articles come from many endangered languages, the majority coming from the authors’ own fieldwork. Not all of the articles are written in a specific theoretical framework. Some articles provide empirical and/or typological generalizations, whereas others also interpret the data discussed with respect to implications for current theoretical developments. We hope that the volume will lead to further investigations into Oceanic languages from descriptive as well as theoretically oriented linguists.1

Part one: Sentential syntax and sentence types The first four papers of the present volume deal with sentence types in Oceanic languages. Sentence types like interrogatives, declaratives, and imperatives constitute functional classes that appear in all languages of the world (Saddock and Zwicky 1985; K¨onig and Siemund 2006). Their surface realization varies substantially, as is shown by empirical cross-linguistic studies. From a theoretical point of view, the constitution of sentence types is based on the interaction between various grammatical means such as word order, inflectional morphology, particles, and intonation, which jointly determine the unique semantic (functional) object corresponding to the relevant sentence type. The notion of ‘sentence type’, however, is not always understood in an identical manner in the relevant literature. Foley and Van Valin (1984) divide sentence properties into three categories: illocutionary role (speech act mode), realis/irrealis, and mode. Bybee (1985, chapter 8) and Cinque (1999) classify sentence types into speech act properties (imperative, interrogative, optative), epistemic modality (evidential degree), and grammatical mode (subjunctive, indicative, imperative). A sentence type is understood here as a concept, which relates speaker attitudes and propositions in systematic ways, an intuition that was expressed as early as the beginning of the previous century by the philoso1. We would like to thank Volker Gast for his comments on the introduction and Jean-Michel Roynard for his help in turning the manuscript into a book.

Introduction

3

pher and mathematician Gottlob Frege (1879) in his investigations into the modal force of sentences. An analysis of the connection between speech acts and syntactic structure was furthermore proposed in generative semantics (cf. Ross 1970). There is one complication, however, in that a one-to-one correlation does not always hold between syntactic structures and speech acts/sentence types. A statement, for example, may be expressed with a declarative sentence but also with an interrogative sentence in the case of rhetorical questions. An order may be expressed with imperative sentences but also with declarative or interrogative sentences. Sentence types may be analyzed from a synchronic, comparative, and historical point of view. Recent syntactic analyses in the framework of generative grammar focusing on the left sentential periphery assume a small set of universal grammatical features and structural positions (Rizzi 1997) that are supposed to allow for a derivation and a unified treatment of several sentence types (especially y/n-interrogatives, wh-interrogatives, declaratives, together with focus and topicalization phenomena). The split of the left peripheral complementizer system into a structural hierarchy of functional categories ‘force’ > ‘topic’ (Top) > ‘focus’ (Foc) > ‘finiteness’ (Fin) as proposed by Rizzi (1997) can be illustrated with the following example from the Western Austronesian language Malagasy. The overt topic head dia and the focus head no, as well as a lower silent topic head, are realized, providing evidence in favor of Rizzi’s projection hierarchy: Malagasy (1) Force > Top > Foc > . . . (fa) i Ketaka dia tany Betafo no that det Ketaka top there Betafo focus Top > Fin matetika, n-iasa. often pst-work ‘(. . . that) as for Ketaka, it is in Betafo that she often worked.’ The analysis of the left periphery of sentences leads to the further assumption that focus/topic features, usually connected to discourse conditions, and grammatical features (for example, interrogative) are closely related on the syntactic level. This opens up the perspective that these features depend on each other to a large degree, thereby strongly affecting sentence type constitution. The complementizer system is seen as the interface between a propositional content (expressed by the area below the complementizer) and the superordinate structure (either a higher clause in cases of an embedded sentence or the articulation

4

Introduction

of discourse, if we consider a root clause). Finiteness represents the interface between the superordinate structure and the proposition. Given that propositional content is independent of speaker attitudes, the grammatical realization of sentence types is restricted to left peripheral clausal positions. If we turn to a characterization of sentence types from a comparative perspective we observe that natural languages differ with respect to the morphosyntactic marking of different sentence types (Sadock and Zwicky 1985; K¨onig and Siemund 2006). Focusing on Oceanic languages, this book sheds more light on the grammatical mechanisms, features and processes relevant for sentence type constitution on the different levels of grammatical description, and in so doing, demonstrates that Oceanic languages have substantial contributions to make to linguistic theory. The articles in the section on sentential syntax and sentence types deal with declarative, interrogative and exclamative sentences. Declarative sentences are the most basic and frequent form of sentence in language. There are two main ways of realizing declaratives. Either some obligatory formal marking is added to the basic clause structure or nothing special is added (Saddock and Zwicky 1985). For example, declarative sentences in Germanic languages such as English and German are realized by a certain verb morphology and a certain position of the verb. In a language such as Hidatsa, a declarative sentence needs to contain one of five different particles. Languages such as Japanese, Korean, and Turkish, among others, use suffixes to mark declaratives (Sohn 1994). We have already seen in (1) that declarative sentences realize topic and focus positions in the left periphery of the sentence. It is also well-known that languages have clause internal positions for topic-comment and focus-background structures. Sabel (this volume) discusses declaratives in the Oceanic free word order languages Kiribati and Fijian. These verb-initial languages do not use any special markers for declarative sentences. Based on information structure effects in declarative sentences, it is shown that the sentences in the Oceanic VOS languages Fijian and Kiribati make use of a focus position in the left periphery and in the postverbal domain. The postverbal focus position precedes the postverbal position for background. In this respect, Fijian and Kiribati differ from SOV and SVO languages where focus in the postverbal domain is realized in sentence final position following background. Adjunct positions in VOS languages are realized likewise in an inverse order compared to SVO and SOV languages. Although linearization is homogeneous with respect to adjuncts and focus/background positions in these languages, this does not hold for nominal objects, which appear in different unmarked orders in VOS languages. For ditransitive constructions the order is recipient-theme in Fijian and Kiribati but

Introduction

5

theme-recipient in the VOS language Malagasy. Some theoretical implications of the derivation of these linear orders are discussed. This article also addresses the phenomenon of free word order (scrambling) in Kiribati and Fijian.The examination of scrambling within generative grammar goes back to Ross (1967) and Hale (1983, 1992). Hale discussed free word order in “non-configurational” languages such as Warlpiri and observed that those languages display a cluster of properties. Besides free word order, they allow for pro-drop and discontinuous constituents. On the basis of this observation Hale proposed analyzing free word order as a base property, i.e. as a result of various base-generated word orders. In contrast, it has been argued that non-standard word orders in languages such as German and Japanese are derived from canonical SOV basic word order displacement operations (see, for example, Mahajan 1990 for Hindi, and Grewendorf and Sabel 1999 for German and Japanese). Hence “free word order” is not analyzed as a homogeneous phenomenon, i.e. there is no single macro-parameter responsible for the absence/presence of the phenomenon. Sabel argues that Kiribati and Fijian are configurational free word order languages (see also Otsuka 2005 for a similar claim concerning Tongan). The reader is referred, for example, to Hale (1992), Baker (2001) and Pensalfini (2004) for analyses of typologically different scrambling languages, i.e. configurational and non-configurational free word order languages. The papers on interrogatives in this volume deal with yes-no and content questions. Interrogative sentences vary in their realization in natural languages. Polar questions, i.e. questions requesting a yes-no answer, may differ from the corresponding declarative sentences by the absence of a declarative marker, or by the use of a question particle and/or an interrogative verbal marker that is added to the corresponding declarative sentence to indicate its interrogative character. Some languages use different word order, for example there is verb fronting in Germanic languages, and in some languages polar questions are realized by a rising intonation pattern. In Jacaltec, for example, this is the only formal criterion that distinguishes these sentences from declarative sentences (Craig 1977; Sadock and Zwicky 1985), whereas in Greenlandic, yes-no questions are not associated with a special intonation at all. In Yiddish, yes-no questions are marked in one of three ways, by a sentence-initial particle, a specific word order or rising intonation. Massam, Starks and Ikiua (this volume) discuss yes-no questions in the Oceanic language Niuean. Yes-no questions in Niuean may be signaled by intonation alone or by one of several particles. The particles are particularly interesting in that some appear to have more than one form, each has its own particular syntactic distribution, and each can appear in various semi-fixed expressions. The authors consider the grammatical, discourse-related, and pragmatic factors

6

Introduction

that contribute to the choice of particular forms, such as questions and answers, factors such as negation, focus, expected response, and written vs. spoken language. With respect to content (information) questions, we observe that the Micronesian VOS language Kiribati, for example, realizes wh-questions (with a single wh-element) in two ways without any change in meaning. The wh-element can either appear in sentence-initial position (2a) or it remains in situ, as illustrated in (2b). In the first case, a particle (ae), glossed here as left-peripheral marker LM, is obligatory, whereas the particle cannot appear with wh-in situ. Kiribati (2) a. b.

Teraa *(ae) e noor-i-a Rui? what lm 3sg see-tr-3sg Rui (*Ae) e noor-a teraa Rui? lm 3sg see-3lsg what Rui ‘What does Rui see?’

Both options are also found in Western Austronesian languages, for example in Malagasy, Tagalog (Schachter and Otanes 1972; Aldridge 2004, among others), Tukang Besi (Donohue 1999), Seediq (Holmer 1996) (see Himmelmann 2005; G¨artner et al. 2006 and Chung and Polinsky 2009 for a general description of core grammatical characteristics of the Western Austronesian languages). Turning to the forty or so Polynesian languages which belong to the Oceanic branch of Eastern Malayo-Polynesian, we observe similar options for the formation of wh-questions. Like Kiribati, these languages are predicate-initial. In Tongan, for example, we find wh-in situ as well as wh-ex situ. Yes-no questions are marked with the question particle nai (3). This particle may also appear after the sentence initial wh-element in a content question (4). Initial wh-phrases and some predicates in numerous Polynesian languages are obligatorily preceded by the particle ko (see Bauer 1993; Massam, Lee, and Rolle 2006): Tongan (3) Na’e lau tohi nai ’a e leka? past read book Q abs det child ‘Did the child read?’ (4)

Ko hai nai na’e ’alu? prtcl who Q past leave ‘Who left?’

Introduction

7

The word order of the wh-questions in (2a) and (4) is structurally ambiguous. These examples from Austronesian languages have been analyzed as resulting from a leftward displacement operation that affects the wh-element (as in What did you see?), or from either a cleft sentence, i.e. a biclausal impersonal construction in which the wh-phrase is a focused part of the predicate and the subject is an expletive ([What] is it [that you saw]?) or a pseudo-cleft sentence, i.e. a biclausal equative construction in which the wh-phrase is the predicate and the subject is a nominalized relative clause (What is [the thing you saw]?) (see Polinsky and Potsdam, this volume). Irrespective of the analysis, the wh-element appears in sentence-initial position. Potsdam and Polinsky (this volume) address this structural ambiguity in detail, arguing that the relevant constructions in most Polynesian languages are pseudo-clefts or clefts but that, at least in the case of Rapanui, wh-questions with a sentence initial wh-element seem to be derived from leftwards displacement of the wh-element. They show that it is often difficult to decide between a cleftand a pseudo-cleft analysis and that a broader range of examples needs to be considered. It turns out, for example, that adjunct and argument questions are not derived in the same way in one and the same language. Comparative analyses have shown that exclamatives can be expressed by a wide variety of formal means and constructions (cf. Michaelis 2001; K¨onig and Siemund 2006), such as assertive and interrogative clauses, relative clauses, inversion or subordination or as specific nominal expressions (Portner and Zanuttini 2004). Moyse-Faurie (this volume) discusses exclamatives in Oceanic languages that are realized as nominalized constructions associated with a specific intonation such as a flat contour at the beginning of the sentence with a high pitch on the first syllable of the last word of the sentence. She discusses these exclamatives using nominalisations mainly from Kanak and Polynesian languages, comparing their properties and specificities to other nominalized constructions such as adverbial, complement and relative clauses. Nominalized clauses expressing exclamations are often truncated clauses with a nominalising prefix, lacking either a possessive adjunct, a head noun, or a predicative/focus marker.

Part two: Nominal morphosyntax The three articles in the second part of the volume deal with nominal phrases in Oceanic languages. Much research on nominal structures within the last 20 years has dealt with finding parallel structures in clauses and nominals and with the search for a universal set of positions in nominal phrases. The parallels between noun phrases (or determiner phrases) and clauses involve, for example,

8

Introduction

subject-like relations between possessors, i.e. subjects of DPs compared to the corresponding structures with subjects of verbs. An important starting point was Abney’s (1987) “The English Noun Phrase in its Sentential Aspect.” Further important research areas focus on the make-up of nominal phrases, i.e. finding universal structures on the basis of comparative analyses. NP-analyses include, for example, the position of articles, demonstratives and proper nouns, as well as the position of modifiers such as attributive adjectives, the ordering restrictions of these elements, and also the realization of (genitive) arguments and (in)alienable possessors. This research branch is mainly based on the analysis of word and morpheme order. Concerning word order in noun phrases, Greenberg (1963: 87) claimed, for example, that “when any or all of the items (demonstrative, numeral, and descriptive adjective) precede the noun, they are always found in that order. If they follow, the order is either the same or its exact opposite.” Cinque (2005, 2009) and Abels and Neeleman (2006) try to show how this generalization can be derived from different assumptions about a universal DP-structure. Taking Greenberg’s generalization and the aforementioned analyses as a starting point, Pearce (this volume) shows that a dual-numeral preceding a demonstrative in the Unua noun phrase (and other Malakula languages) is unexpectedly dispreferred in contrast to other numerals in postnominal position. Another topic within the morphosyntax of nominals that has gained much attention and is attested in Oceanic languages is the phenomenon of noun incorporation, i.e. the integration of a noun into a verb resulting in the derivation of a complex verb. In some languages the process is obligatory, whereas in others it is optional. Mithun (1984), working in the functional descriptive framework, develops an influential typology of four different types of noun incorporation on the basis of an examination of over 100 noun-incorporating languages. Type I, i.e. lexical compounding, is found in connection with institutionalized concepts. Type II incorporation changes transitive to intransitive verbs which has an effect on Case, for example in ergative languages such as Tongan, or on verbs in languages such as Drehu, Iaai, and Nengone, which are spoken on the Loyalty Islands. In type III incorporations, the effect of noun incorporation is a manipulation of discourse structure, i.e. a first mentioned constituent is realized as not incorporated whereas subsequent references are part of the background when being incorporated. In type IV (classifying) incorporations, an object is “doubled” on the verb. The four types form an implicational hierarchy, i.e. type IV implies the existence of types I-III, type III implies the existence of types I and II, and type II implies the existence of type I. A central question with respect to noun incorporation has been whether it is a lexical (see also DiSciullo and Williams 1987) or syntactic process. Baker

Introduction

9

(1988, 1996, 2009) analyses incorporation as involving syntactic (adjunction) movement of the head noun from the object position targeting the verb. Restrictions on incorporation, for example the impossibility of incorporating noncomplements, are derived from restrictions on movement. Rosen (1989) argues that a lexical (non-movement) analysis might alternatively derive the distribution of noun incorporation structures. Either an incorporated noun saturates a verb’s argument structure through lexical compounding, or the incorporated noun is a classifier that restricts the interpretation of a verb-externally-realized syntactic argument. The conflict between analyses regarding noun incorporation as a purely syntactic or lexical/morphological process, however, disappears in the light of modern theories about the interaction of morphology and syntax such as Distributed Morphology (Halle and Marantz 1993; Marantz 1997), where it is claimed that an autonomous morphological component operating within a lexicon is unnecessary (see, for example, Johns 2007). Another controversial issue in the domain of nominal syntax is whether verb + noun sequences in different languages are always instances of the same homogeneous noun incorporation phenomenon or whether different phenomena may be involved to varying degrees. Turning to noun incorporation in Oceanic languages, Massam (2001) argues for the concept of “pseudo noun-incorporation.” In Niuean, an incorporated object can be phrasal, as it may contain relative clauses and coordinate structures. Moyse-Faurie (1997) differentiates various incorporation types attested in New Caledonian and Polynesian languages, such as (i) morphological incorporation involving position constraints, indissociability, and formal verb variations, but without any consequences on the argument marking or the verb valency, which remains transitive; (ii) syntactic incorporation, by contrast involving an intransitivation process with a change in argument structure, and being constrained by semantic and pragmatic conditions. The different types of incorporation depend on the category of the incorporated element (nominal, pronominal, proper noun, preposition) and its grammatical function (argument, complement, nominal modifier). Noun incorporation structures may serve as input for relexicalizations such as the formation of nominal compounds, which can in turn be retransitivized and take a new object argument. In this context, Margetts (this volume) presents important evidence for variation in noun incorporation. She demonstrates that in the Oceanic language Saliba the incorporated noun may appear on either side of the verb. What is worth noticing is that the possibility of different word ordering of incorporation in Saliba is limited to a few verbs and is not semantically or grammatically predictable. This possibility is also attested in Polynesian languages such as Samoan or East Futunan. Preverbal incorporation, however, is restricted to a few nouns (mainly body parts, other parts of a whole or inherent characteristics of entities), giving

10

Introduction

rise to isu mamafa ‘heavy nose’ compounds, as Mosel & Hovdhaugen (1992: 88) label them. Using six different tests, Margetts illustrates that the noun is morphologically bound to the verb, and that noun and verb constitute a semantic unit. One striking finding of these tests is the fact that it is the reduplication of the incorporated noun which marks the progressive aspect, instead of reduplication of the verb normally used in such contexts. Saliba also features very interesting cases of incorporation with intransitive base verbs. The morphosyntax and semantics of noun coordination has also gained a lot of interest in current research. The term coordination refers to constructions in which two or more elements, for example nouns/noun phrases, verbs (or VPs), or sentences of the same type are combined with a coordinator to build a coordination phrase. In English, the class of coordinators includes and (conjunctive coordinator), or (disjunctive coordinator), and but (adversative coordinator), among other items. A coordinator is either overt (syndetic coordinations) or covert (asyndetic coordinations), as in a beautiful, young girl. When it comes to determiners and coordinated nouns, agreement becomes a complex phenomenon because unexpected agreement patterns can be observed (Dalrymple and King 2004). Even though the article in [this(/*these) man and woman] are(/*is) nice has to be singular, the whole conjunct behaves like a plural, as can be seen from the plural verb agreement. Furthermore, a coordinated structure such as two mathematicians and linguists has at least three readings. It can refer to two persons, each of them is a mathematician and linguist, or to four persons, or it can mean [two mathematicians] and [linguists]. The question is how these phenomena can be explained on the basis of coordination and NP-structure. Although all languages appear to display coordination structures, much crosslinguistic variation can be observed (Haspelmath 2004). For example, different languages may require different coordinators depending on the syntactic type of the coordinated elements. Austronesian languages such as Yapese (an Oceanic language of Micronesia, Jensen 1977: 311) and Malagasy, for example, use sy (Malagasy), ngea (Yapese) ‘and’ for NP conjunction, but ary (Malagasy), ma (Yapese) ‘and’ for sentential conjunction. Even among similar conjuncts, we observe different coordination strategies. In Nakanai (Oceanic), the asyndetic strategy is used with NP conjuncts that are discourse topics, whereas an overt coordinator me is obligatory for subjects. This observation is presented in Bril (this volume). Bril discusses morphosyntactic and semantic properties as well as pragmatic effects of noun-phrase coordination in a wide range of Oceanic languages.

Introduction

11

Part three: Historical developments The last two articles discuss the historical source of the divergence between ergative and accusative Oceanic languages and, in addition, an example of grammaticalization. Many Oceanic languages have been argued to be ergative. An “ergativity pattern” is attributed to a language if the sole core argument of intransitive clauses and the object of transitive clauses behave similarly with respect to certain grammatical features (such as case or agreement), while differing from the agent argument of transitive clauses in this respect. In nominativeaccusative languages, by contrast, the subject of intransitive clauses has the same grammatical features as the agent argument of transitive clauses, whereas objects are different. Ergativity has an impact on many aspects of the grammar of a language, such as grammatical functions, case (marking), agreement, binding, extraction, transitivity, and aspect (see Dixon 1994; Manning 1996; Johns 2000, among others). However, ergativity phenomena show variability between languages (cf. Bittner and Hale 1996). From a theoretical point of view, the question arises as to which parametric options are responsible for the observed differences. Ross (2004b) postulates a canonical Proto Oceanic type with nominativeaccusative marking. However, ruling out the Oceanic ergative-absolutive marking languages from the canonical Oceanic type is problematic. Most, if not all Oceanic languages exhibiting ergative patterning also allow or require nominative-accusative marking; moreover, originally ergative languages such as some Polynesian Outliers (West Futunan, Vaeakau-Taumako, Ifira-Mele, Kapingamarangi, etc.) recently shifted from VOS to SVO word order. Different ergative splits are found which may be conditioned by the semantic or syntactic nature of the noun phrases, by tense-aspect-mood markers or by the semantic class of the verb. These recurrent features among Oceanic languages belonging to different subgroups should be taken into account in the reconstruction of a Proto Oceanic canonical sentence type. Ergative constructions in Oceanic languages are correlated with several (morpho-)syntactic features (mainly verbal agreement, coreference, verb classes, pronominal vs. nominal arguments, and tense-aspect); they are also correlated with semantic facts (human/non human arguments, degree of agentivity or intentionality) and finally with pragmatic considerations (choice, when possible, between an ergative construction and something else –a possessive or auxiliary construction–, as first shown by Duranti and Ochs 1990; Duranti 1994). From a diachronic point of view, there has been much debate in Oceanic linguistics about the structure of Proto Polynesian and the direction of change between the three ergative, passive, and accusative patterns in recent decades.

12

Introduction

Phonologically and lexically, Polynesian languages manifest an evident genetic affiliation but the existence of different case marking patterns has raised the following questions: What happened historically? Did a drift take place from accusative to ergative or was the accusative construction a late evolution? And what about the higher branching? What was the structure of Proto Oceanic, given that quite a number of Oceanic languages still have ergative structures? If Proto Polynesian (PPn) was ergative, one must explain how languages such as M¯aori, Tahitian, Hawaiian became accusative. If PPn was accusative, one must explain how East Futunan and other Western Polynesian languages became ergative. Hohepa and Hale’s explanation (1969) is that PPn was an accusative language with a passive construction (as in M¯aori). A drift inside PPn favored the passive at the expense of the active construction. In Western Polynesian languages, the passive became compulsory for all verbs except middle verbs. Afterwards, the passive suffix was no longer useful to distinguish passive from active; it became optional and, finally, the drop of the passive suffix gave rise to the ergative structure. The advantage of this hypothesis is that it implies the following parsimonious reanalysis: – Reanalysis of passive marking, which becomes compulsory; – Reanalysis of passive clauses, which become transitive active clauses. Clark (1976) underlines one drawback of this analysis, namely that it supposes a parallel change in Samoan and Tongan, i.e. languages which split at the Proto Polynesian level. In addition, he disputes the notion of ‘drift’, proposing a model close to Western Polynesian languages, with two transitive verb classes: – Class A: direct transitives, “canonical” (‘eat’, ‘drink’, ‘kill’, ‘wash’), which are ergative; – Class B: indirect transitives (‘see’, ‘call’, ‘love’), that is, the so-called ‘middle verbs’. He argues that the innovation concerned only the Eastern group (M¯aori), whose contemporary alignment properties have resulted from an overgeneralization of the accusative marker for all transitive imperfective clauses and from a reanalysis of perfectives as passive clauses. Chung (1978) proposes a more complex argument structure for PPn, reconstructing an accusative marker *i different from the oblique markers *i and *ki. According to her analysis, PPn had two verb classes: (i) Canonical transitives [V S i(accusative) O] and (ii) middle transitives [V S i/ki(oblique) O] with a passive V-(C)ia e Agent S for both verb classes. Western Polynesian languages reanalyzed this passive form as an active one, (V-suf e A O), with progressive loss of the -(C)ia suffix, giving birth to the present day ergative type construction.

Introduction

13

Hence, like Hohepa and Hale, Chung is in favor of a passive towards ergative evolution. Gibson and Starosta (1990), however, are convinced that M¯aori used to be an ergative language. They are both specialists of Philippine languages and of Proto Austronesian. They state that all linguists, whatever theory they adopt (Lexicase Grammar, Lexical Functional Grammar, Relational Grammar, Government and Binding or Categorial Grammar), come to the same conclusion: Tagalog, Ilokano, etc. and Proto Austronesian are all ergative languages. Gibson and Starosta put forward several criteria: argument marking, productivity, morphological identity, existence of middle verbs, semantic (degree of affectiveness, aspect), in order to decide which is the canonical transitive model. They conclude that in M¯aori, the -(C)ia suffixed construction corresponds closely to this model, and that therefore M¯aori is an ergative language. They consequently assume that Proto Polynesian as well as Proto Austronesian must have been ergative. Otsuka (this volume) not only presents a very convincing hypothesis for the case marking patterns of Eastern Polynesian languages, she also offers a comprehensive view of the entire Polynesian languages family, with fundamental insights into their general evolution and transitional stages, explaining the current situation for all spoken Polynesian languages. Her paper might put an end to 40 years of a heated debate opposing supporters of an ergative > passive evolution and supporters of opposite directionality for languages of the Eastern branch of the Polynesian family, such as M¯aori or Tahitian. On the basis of semantic and syntactic arguments, she demonstrates very clearly that Proto Polynesian must have been ergative, as previously stated by Clark (1973) and Gibson & Starosta (1990), and that Eastern Polynesian languages developed a symmetric voice system (neither accusative nor ergative). This indicates that the patient is directly affected when the pseudo-passive suffix PPn *-Cia occurs, while on the other hand, the use of the pseudo-accusative preposition PPn *i marks a low degree of affectedness of the patient. Grammaticalization is the final topic discussed in this introduction and in the volume. Grammaticalization processes involve morphemes, lexical items and word clusters (Lehmann 1995; Hopper and Traugott 2003). These elements lose some of their lexical meaning and become functional elements (categories). Discussing the grammaticalization paths of three motion verbs (go, come and return) in Fijian, in some Solomon and Vanuatu languages, and in M¯aori, Lichtenberk (1991) points out that meaning extension is not arbitrary since metaphor and metonymy play an important role in this process. He also shows that an extension of meaning is motivated by the relation perceived by speakers between the new and the old item. Concerning the verb ‘return’, Lichtenberk identifies different paths of grammaticalization, as ‘reditive’ directionals, encoding the direction

14

Introduction

of an event back to the deictic center, and as repetitive, additive and reflexive markers. The verb ‘come’ was grammaticalized in centripetal directionals, continuative, inchoative aspect, prepositions, comparative marker, etc., while ‘go’ was grammaticalized as centrifugal directional, conditional, aspectual or temporal marker. Ross (2004a) and Moyse-Faurie (2010) also explore the grammaticalization of directional verbs or spatial notions in some Oceanic languages. The grammaticalization of the noun ‘thing’ has already been discussed elsewhere. Heine and Kuteva (1992) list three grammaticalization options for lexical equivalents of ‘thing’: complementizer (Japanese, Ik (Uganda), etc.), indefinite pronoun (Swahili, Nahuatl, Albanian) and A-possessive (Thai, Khmer, Proto Central Khoisan). Vernaudon (this volume) gives convincing examples that Tahitian mea ‘thing, matter’ is a semantic wildcard or pro-form that is not attached to a specific lexical category. Firstly, it grammaticalized as a qualifying marker, taking on a stative aspect. This peculiar evolution of ‘thing’ is not mentioned in Heine & Kuteva (1992). The author also points out a feature that could be reminiscent of an ergative structure in Tahitian, involving the use of mea ‘thing’. Other grammaticalization paths for the noun ‘thing’ are attested in Oceanic languages, for example in Xˆarˆac`uu` (New Caledonia), in which d¨ou ‘thing’ is part of the group of relative clause markers and conjunctions.

The contributions Isabelle Bril Noun-phrase conjunction in Austronesian languages: additive, inclusory and comitative strategies This is an investigation of various types of NP conjunctive strategies in Austronesian languages, taking in additive, inclusory and comitative strategies. The focus is on asymmetrical conjunction involved in the inclusory and comitative types; the aim is to delineate the semantic parameters of these constructions and the syntactic constraints on their use, as well as their pragmatic effects. Bril also includes an analysis of the etymology of some of the conjunctive morphemes (comitative marker, noun “fellow”, verb “accompany, be with”). Anna Margetts Noun incorporation in Saliba Saliba contributes some interesting examples to the discussion on noun incorporation. It has external possession and lexical compounding constructions, but also a construction featuring morphologically intransitive verbs with phrasal

Introduction

15

objects. This paper is concerned primarily with the second type, lexical compounding, where an object noun is morphologically incorporated into the verb and loses its syntactic independence. There are a number of features in the Saliba constructions which are typologically unusual. The morpheme ordering in the compound stem can be NV but also VN. In both cases the construction underlies the same semantic and pragmatic constraints and the verb and the noun stem form a morphological unit, i.e. a complex verb stem. The variation in the order of stems is an unusual feature for incorporation constructions crosslinguistically and has not been described for other languages of the Papuan Tip cluster or the wider Oceanic language group. The position of the incorporated noun seems to be an idiosyncratic feature of the verb, which is not predictable on formal or semantic grounds. A second typologically unusual feature of the Saliba constructions is that certain intransitive verbs can incorporate a noun much in the same way as transitive verbs can. The noun in these constructions can be argued to be the semantic object even though the verb is morphologically intransitive. This paper describes the different patterns found with Saliba object incorporation and discusses the constraints governing these constructions. Diane Massam, Donna Starks and Ofania Ikiua Questions and answers in Niuean This paper examines polar questions and their answers in the Niuean language, with a focus on the various question particles found in interview data collected as part of the Pasifika Languages of Manukau Project. Niuean has a rich variety of ways to both ask and answer questions. The first part of the paper addresses questions, which may be signalled by intonation alone, or by one of several particles. The particles are particularly interesting in that some appear to have more than one form, and each has its own particular syntactic distribution. The second part of the paper focuses on answers, which also take a variety of forms. Massam, Starks and Ikuia consider the grammatical, discourse, and pragmatic factors that contribute to the choice of particular forms of questions and answers. These include negation, focus, expected response, and written vs spoken language. We also consider the relative frequency of each type of question and answer. Claire Moyse-Faurie Nominalization and exclamation in Oceanic languages Exclamations can be expressed by a wide variety of formal means and constructions, such as assertive and interrogative clauses, relative clauses, inversion or subordination. In Oceanic languages, further options are found, most of them

16

Introduction

consisting of nominalized constructions, associated with a specific intonation, such as a flat contour at the beginning of the sentence and a high pitch on the first syllable of the last word of the sentence. This article presents several constructions, all based on different nominalization strategies which express an emotional stance, i.e. the surprise of the speaker (and supposedly, of the hearer) in the face of unexpected events, in several Oceanic languages (mainly Kanak and Polynesian languages). Yuko Otsuka Neither accusative nor ergative: an alternative analysis of case in Eastern Polynesian Eastern Polynesian (EP) languages have traditionally been regarded as accusative. “Active” constructions have the unmarked agent with the patient marked by i/ki. In “passive” constructions, the verb has a suffix -Cia, the patient is unmarked, and the agent is marked by e. Comparison with western Polynesian (WP) languages makes this analysis questionable. Verbal constructions in WP are strikingly similar to those in EP: one with e-marked agents and the other with i/ki-marked patients. Yet in WP, the former is considered a canonical transitive (i.e., ergative) construction and the latter, a dyadic one with an oblique object. This paper argues that EP is neither ergative nor accusative, but has a symmetric voice system similar to that of Western Austronesian. The study shows (a) that i-marked objects in EP behave like oblique objects in WP (hence, not accusative); and (b) that e-marked agents in EP show characteristics of core arguments (hence, not passive). The study therefore supports the theory that Proto Polynesian (PNP) was ergative like WP, and proposes that in PNP, the ergative pattern was used with verbs entailing affected patients and that the objects of verbs entailing less/indirectly affected patients were marked as oblique. The difference between WP and EP is accounted for by positing two Proto-EP innovations: (a) extended use of i with all dyadic verbs to indicate less/indirectly affected patients and (b) obligatory -Cia suffixation in constructions with affected patients. In this analysis, e-marked agents and i-marked patients are both core arguments in EP and -Cia can be viewed as theme voice morphology. Elizabeth Pearce Two or three things in the Unua noun phrase Unua, an Oceanic language of Malakula, Vanuatu, has an unusual restriction on the cooccurrence of demonstratives with the cardinal numeral “two”. When the numeral is higher than two, nominal expressions in Unua have the “inverse” ordering identified in Greenberg’s Universal 20: Noun – Adjective – Numeral –

Introduction

17

Demonstrative. The numeral ‘two’ xeru occurs at the right edge of the nominal expression after adjectives and possessives, but, when a demonstrative is present, “two” is expressed through the use of the dual third person pronoun following the demonstrative. The investigation in this article shows that a loosening of the restriction on the placement of xeru ‘two’ before a demonstrative in more liberal speaker dialects is associated with ongoing synchronic erosion of morphological number marking in the forms of the demonstratives. In the account that is presented, the alternative strategy using the dual pronoun (in the restricted dialect) is a means of avoiding a mismatch in Number features between a numeral with a [+dual] feature and a structurally higher demonstrative with an incompatible non-dual Number feature. In the less restricted dialects, the loss of number marking morphology means that demonstratives with unspecified Number are no longer incompatible with a dual numeral. Data from some other Malakula languages give indications of comparable morpho-syntactic behavior in the expression of “two” in the presence of a demonstrative. Eric Potsdam and Maria Polinsky Questions and word order in Polynesian This paper uses the Polynesian languages as a stepping stone for the typological investigation of wh-question formation strategies in verb-initial languages. It is shown that a number of structural options are available for wh-question formation: displacement, clefting, and pseudo-clefting. All these options yield Wh-word-first order in a verb-initial language. This article suggests that in some cases it may be difficult to determine which structure is being employed, and offers some morphosyntactic diagnostics to help decide the issue. The relevant diagnostics include mono- or bi-clausality of the question, the status of the whphrase as predicate, and the constituent in the subject position of the question. Although these diagnostics are presented here for Polynesian languages they should also be applicable to other verb-initial and wh-word-first languages. Joachim Sabel Deriving linear order in OV/VO languages: evidence from Oceanic languages The investigation in this article demonstrates that the Austronesian VOS languages Malagasy, Kiribati, and North-West Fijian differ with respect to linearization of arguments in the predicate phrase, whereas adjuncts are uniformly ordered. The linearization of adjuncts depends on their position relative to the head they modify. Adjuncts following the head appear in inverse order in all discussed Austronesian verb-initial languages compared to SVO/SOV languages, and in addition, focus precedes background in the postverbal domain of all of

18

Introduction

the discussed V-initial languages. Nominal arguments appear in normal or inverse base word order. Sabel discusses the implications of these findings for the mechanisms that have been proposed in the literature for deriving inverse linear order in VOS languages such as Malagasy, i.e. roll-up movement analysis and (emptied) VP-remnant movement analysis. Finally, the author points out that the data from (the non overt Case-marking, rich pronominal, and free word order languages) Fijian and Kiribati are not compatible with an analysis of these languages as non-configurational languages, as could be expected from the “pronominal argument hypothesis.” Jacques Vernaudon Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect In Tahitian, one and the same lexeme can denote either an entity or a property, depending on its position in a phrase: e ta’ata ‘that is a human being’; e reo ‘that is a voice’; e ta’ata reo ‘that is a person with a voice’; e reo ta’ata ‘that is a human voice’.This paper examines the grammmaticalization of mea, originally a lexical morpheme whose semantic content is only weakly determined; normally translated as ‘thing, matter’, it may develop into an introducer of qualifiers. It occurs in first position (that of modified), relegating the other lexemes to second position, as property-denoting expressions: E moni ‘That is money’ > E mea moni ‘That is expensive’. In the absence of combinatorial restriction, the lexemes that refer to processes may also appear in this context. Finally, mea enters the paradigm of aspectual markers as a marker of stative aspect-evoking actions of an individual or a group.

References Abels, Klaus and Ad Neeleman 2006 Universal 20 without the LCA. Ms. University of Tromsø / UCL. Abney, Steven 1987 The Noun phrase in its sentential aspect. Massachusetts Institute of Technology. Ph.D. Dissertation. Cambridge MA. Aldridge, Elisabeth 2004 Ergativity and word order in Austronesian languages. Ph.D. Dissertation, Cornell University. Baker, Mark 1988 Incorporation. A Theory of Grammatical Function Changing. Chicago and London: The University of Chicago Press. Baker, Mark 1996 The polysynthesis parameter. New York: Oxford University Press.

Introduction Baker, Mark 2001

Baker, Mark 2009

19

The Natures of Nonconfigurationality. In: Baltin, Mark and Chris Collins (eds.), The handbook of contemporary Syntactic Theory, 407– 438. Oxford: Blackwell Publishers.

Is head movement still needed for noun incorporation? Lingua 119: 148–165. Elsevier. Bauer, Winifred (with William Parker and Te Kareongawai Evans) 1993 M¯aori. London/New York: Routledge. Bennardo, Giovanni (ed.) 2002 Representing space in Oceania. (Pacific Linguistics 523.) Canberra: The Australian National University. Bittner, Maria and Ken Hale 1996 Ergativity:Toward aTheory of a Heterogeneous Class. Linguistic Inquiry 27: 531–604. Cambridge, MA: Massachusetts Institute of Technology Press. Bril, Isabelle and Fran¸coise Ozanne-Rivierre (eds.) 2004 Complex predicates in Oceanic languages: studies in the dynamics of binding and boundedness. (Empirical Approach to Linguistic Typology 29.) Berlin: Mouton de Gruyter. Bybee, Joan 1985 Morphology. A Study of the Relation between Meaning and Form. Amsterdam: John Benjamins. Chung, Sandra 1978 Case marking and grammatical relations in Polynesian. Austin: University of Texas Chung, Sandra and Maria Polinsky 2009 Introduction. Natural Language and Linguistic Theory 27-4: 659–673. Dordrecht: Springer. Cinque, Guglielmo 1999 Adverbs and functional heads. Oxford: Oxford University Press Cinque, Guglielmo 2005 Deriving Greenberg’s Universal 20 and Its Exceptions. Linguistic Inquiry 36: 315–332. Cambridge, MA: Massachusetts Institute of Technology Press. Cinque, Guglielmo 2009 The Fundamental Left-Right Asymmetry of Natural Languages. In: Sergio Scalise, Elisabetta Magni and Antonietta Bisetto (eds.), Universals of Language Today, 165–184. Berlin: Springer. Clark, Ross 1973 Transitivity and case in Eastern Oceanic Languages. Oceanic Linguistics 12: 559–605. Honolulu: University of Hawaii Press.

20

Introduction

Clark, Ross 1976

Aspects of Proto-Polynesian Syntax. (Te Reo Monographs.) Wellington: The Linguistic Society of New Zealand. Craig, Colette G. 1977 The structure of Jacaltec. Austin: University of Texas Press. Crowley, Terry 2002 Serial Verbs in Oceanic: A Descriptive Typology. Oxford: Oxford University Press. Dalrymple, Mary and Tracy Holloway King 2004 Determiner agreement and noun conjunction. Journal of Linguistics 401: 69–104. Cambridge: Cambridge University Press. DiSciullo, Anne-Marie and Edwin Williams 1987 On the definition of word. Cambridge, MA: Massachusetts Institute of Technology Press. Dixon, Robert Malcolm Ward 1994 Ergativity. Cambridge: Cambridge University Press. Donohue, Mark 1999 A Grammar of Tukang Besi. Berlin: Mouton de Gruyter. Duranti, Alessandro 1994 From Grammar to Politics. Linguistic Anthropology in a Western Samoan Village. University of California Press. Duranti, Alessandro and Elinor Ochs 1990 Genitive constructions and agency in Samoan discourse. Studies in Language 14-1: 1–23. Amsterdam: John Benjamins Publishing Company. Foley, William A. and Robert D. Van Valin 1984 Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Frege, Gottlob 1879 Begriffsschrift: eine der arithmetischen nachgebildete Formelsprache des reinen Denkens. Halle. G¨artner, Hans-Martin, Paul Law, and Joachim Sabel 2006 Clause Structure and Adjuncts in Austronesian Languages: A Critical Introductory Survey. In: G¨artner, Hans-Martin, Paul Law and Joachim Sabel (eds.), Clause Structure and Adjuncts in Austronesian Languages, 1–42. Mouton de Gruyter: Berlin/New York. Gibson Jeanne D. and Stanley Starosta 1990 Ergativity east and west. In: Philippe Baldi (ed.), Linguistic Change and Reconstruction Methodology. (Trends in Linguistics. Studies and Monographs 45.) Berlin, New-York: Mouton de Gruyter. Greenberg, Joseph H. 1963 Some universals of grammar with particular reference to the order of meaningful elements. In: Joseph Greenberg (ed.), Universals of lan-

Introduction

21

guage, 73–113. Cambridge, MA: Massachusetts Institute of Technology Press. Grewendorf, G¨unther and Joachim Sabel 1999 Scrambling in German and Japanese: Adjunction versus Multiple Specifiers. In: Natural Language and Linguistic Theory 17: 1–65. Dordrecht: Springer. Hale, Kenneth L. 1983 Warlpiri and the Grammar of Non-configurational Languages. Natural Language and Linguistic Theory 1: 5–47. Dordrecht: Springer. Hale, Kenneth L. 1992 Basic Word Order in Two “Free Word Order” Languages. In: Doris L. Payne (ed.), Pragmatics of Word Order Flexibility, 63–82. (Typological Studies in Language 22.) Amsterdam: John Benjamins Publishing Company. Halle, Morris and Alec Marantz 1993 Distributed Morphology and the Pieces of Inflection. In: Kenneth Hale and S. Jay Keyser (eds.), The View from Building 20, 111–176. Cambridge, MA: Massachusetts Institute of Technology Press. Haspelmath, Martin (ed.) 2004 Coordinating constructions. Amsterdam/Philadelphia: John Benjamins, Typological Studies in Language 58. Heine, Bernd and Tania Kuteva 1992 World lexicon of grammaticalization. Cambridge/New York: Cambridge University Press. Himmelmann, Nikolaus 2005 The Austronesian Languages ofAsia and Madagascar: Typological Characteristics. In: Alexander Adelaar and Nikolaus Himmelmann (eds.), The Austronesian Languages of Asia and Madagascar, 110–181. London: Routledge. Hohepa Patrick W. 1969 The Accusative-to-Ergative Drift in Polynesian Languages. Journal of the Polynesian Society 78. Auckland: Polynesian Society. Holmer, Arthur 1996 A Parametric Grammar of Seediq. Lund: Lund University Press. Hopper, Paul J. and Elizabeth Closs Traugott 2003 Grammaticalization. Cambridge: Cambridge University Press. Hovdhaugen, Even and Ulrike Mosel (eds.) 1999 Negation in Oceanic Languages. (Typological Studies.) Munich: Lincom Europa. Jensen, John Thayer 1977 Yapese Reference Grammar. Honolulu: The University Press of Hawaii.

22

Introduction

Johns, Alana 2000

Johns, Alana 2007

Ergativity: A Perspective on recent work. In: Lisa Cheng and Rint Sybesma (eds.), The First GLOT International State-of-the-Article Book: The Latest in Linguistics, 47–73. The Hague: Mouton de Gruyter.

Restricting noun incorporation: root movement. Natural Language and Linguistic Theory 25: 535–576. Dordrecht: Springer. K¨onig, Ekkehard and Peter Siemund 2006 Speech act distinctions in grammar. In: Timothy Shopen (ed.), Language Typology and Syntactic Description I: 276–324. Cambridge: Cambridge University Press. Lehmann, Christian 1995 Thoughts on grammaticalization. (Lincom Studies in Theoretical Linguistics I.) Munich : Lincom Europa Lichtenberk, Frantisek 1991 Semantic change and heterosemy in grammaticalization, Language 673: 475–509. Lynch, John 1998 Pacific Languages. An Introduction. Honolulu: University of Hawai’i Press. Lynch, John, Malcolm Ross and Terry Crowley 2002 The Oceanic languages. Richmond, Surrey: Curzon Press. Mahajan, Anoop 1990 The A/A-bar distinction and movement theory. PhD dissertation. Cambridge, MA: Massachusetts Institute of Technology. Manning, Christopher D. 1996 Ergativity: Argument Structure and Grammatical Relations. Stanford, CA: Center for the Study of Language and Information Publications. Marantz, Alec 1997 No Escape from Syntax: Don’t Try Morphological Analysis in the Privacy ofYour Own Lexicon. In: Alexis Dimitriadis, Laura Siegel, Clarissa Surek-Clark and Alexander Williams (eds.), University of Pennsylvania Working Papers in Linguistics vol. 4.2: 201–225. Philadelphia, PA: University of Pennsylvania. Massam, Diane 2001 Pseudo Noun Incorporation in Niuean. Natural Language and Linguistic Theory 19: 153–197. Dordrecht: Springer. Massam, Diane, Josephine Lee and Nicholas Rolle 2006 Still a preposition: the category of “ko”. Te Reo 49: 3–38. Wellington: Linguistic Society of New Zealand.

Introduction

23

Michaelis, Laura 2001 Exclamative constructions. In: Martin Haspelmath, Ekkehard K¨onig, Wulf Oesterreicher and Wolfgang Raible (eds.), Language Typology and Language Universals. Berlin: de Gruyter. Mithun, Marianne 1984 The Evolution of Noun Incorporation. Language 60-4: 847–894. Mosel, Ulrike and Even Hovdhaugen 1992 Samoan Reference Grammar. Oslo: Scandinavian University press. The Institute for Comparative research in Human Culture. Moyse-Faurie, Claire 1997 Ph´enom`enes d’incorporation dans quelques langues oc´eaniennes. Studi italiani di linguistica teorica e applicata XXVI 1997-2: 227–246. Rome. Moyse-Faurie, Claire 2010 (D´e)Grammaticalisation d’expressions spatiales dans des langues oc´eaniennes. In: Choi-Jonin, Injoo, Marc Duval and Olivier Soutet (eds.), Typologie et Comparatisme. Hommages offerts a` Alain Lemar´echal, 295– 314. (Orbis Supplementa vol. 28.) Leuven: Peeters. Otsuka, Yuko 2005 Scrambling and Information Focus: VSO-VOS Alternation in Tongan. In: Joachim Sabel and Mamoru Saito (eds.), The free word order phenomenon: Its syntactic sources and diversity, 243–279. Berlin and New York: Mouton de Gruyter. Pensalfini, Robert 2004 Towards a typology of configurationality. Natural Language and Linguistic Theory 22: 359–408. Dordrecht: Springer. Portner, Paul and Raffaella Zanuttini 2004 The Semantics of Nominal Exclamatives. In: Reinaldo Elugardo and Robert J. Stainton (eds.), Ellipsis and Nonsentential Speech, 57–67. (Studies in Linguistics and Philosophy 81.) Dordrecht: Springer. Rizzi, Luigi 1997 The fine structure of the left periphery. In: Liliane Haegeman (ed.), Elements of Grammar: Handbook of Generative Syntax, 281–337. Dordrecht: Springer. Rosen, Sarah Thomas 1989 Two types of noun incorporation: a lexical analysis. Language 65: -2: 294–317. Ross, John R. 1967 Constraints on Variables in Syntax. Ph.D. dissertation. Cambridge, MA: Massachusetts Institute of Technology. Ross, John R. 1970 On declarative sentences. In: Jacobs, Roderick A. and Peter S. Rosenbaum (eds.), Readings in English transformational grammar, 222–277. Waltham, MA: Blaisdel.

24

Introduction

Ross, Malcolm 2004a The grammaticization of directional verbs in Oceanic languages. In: Isabelle Bril and Fran¸coise Ozanne-Rivierre (eds.), Complex Predicates in Oceanic Languages. Studies in the Dynamics of Binding and Boundness. Berlin: Walter de Gruyter. Ross, Malcolm 2004b The morphosyntactic typology of Oceanic languages. Language and Linguistics 5.2: 491–541. Taipei: Institute of Linguistics, Academia Sinica. Sadock, Jerrold M. and Arnold M. Zwicky 1985 Speech act distinctions in syntax. In: Timothy Shopen (ed.), Language Typology and Syntactic Description I: 155–196. Cambridge University Press. Schachter, Paul and Fe Otanes 1972 Tagalog Reference Grammar. Berkeley, CA: University of California Press. Senft, Gunter (ed.) 2004 Deixis and Demonstratives in Oceanic Languages. (Pacific Linguistics 562.) Canberra: The Australian National University. Senft, Gunter (ed.) 2008 Serial verb constructions in Austronesian and Papuan languages. (Pacific Linguistics 594.) Canberra: The Australian National University. Sohn, Ho-Min 1994 Korean. London: Routledge.

Part one: Sentential syntax and sentence types

Deriving linear order in OV/VO languages: evidence from Oceanic languages Joachim Sabel

1. Introduction VOS/VSO languages often show mirror image ordering with respect to word order phenomena in the predicate phrase and in DP compared to SOV/SVO languages. This will be illustrated with respect to adverb/PP/adjective order, double objects and on the basis of data on focus/background (information) structure. It will be shown that a unified analysis for these phenomena in SOV/SVO/VOS languages based on the parametric option of having either roll-up or head movement, as has been proposed in the literature, is incompatible with data found in Oceanic VOS languages such as Fijian (i.e., the North-West Viti Levu variant) and Kiribati (Gilbertese). Although many languages linearize arguments and adjuncts in the predicate phrase in a unitary way, Kiribati and North-West Fijian use different strategies for arguments in double object constructions and for adjuncts. The following generalizations concerning adjunct/argument order are formulated: Adverbs, PP-adverbials and adjectives following the head they modify appear in inverted order compared to adjuncts preceding the head. The order of arguments (in double object constructions) is determined independently of the order of adjuncts. The theoretical consequences of the empirical generalizations for analyses deriving linear order in SOV/SVO and verb-initial languages are discussed. Another new generalization of the present study of word order is that all Austronesian verb-initial languages that will be discussed realize informational focus in a position preceding background in the post-verbal domain. In this respect, the verb-initial VO-languages differ from SVO and SOV languages which are all focus-final. Finally, the data show that the “free” word order languages Kiribati and (Standard and North West) Fijian are incompatible with an analysis involving non-configurationality, as has been proposed for the so-called “pronominal argument languages.“

28

Joachim Sabel

2. Mirror order syntax: SOV vs. VOS languages and adverb order For SVO languages like English and Italian and SOV languages like German, Cinque (1999) has diagnosed the continuously descending pattern of adverb types in 1 . . . 6 > 7 > 8 > 9 > 10 > (V◦ ), ranging from speech act markers [1] via aspectual adverbs like already [4] to frequency adverbs like always to manner adverbs like well [10]. As illustrated in the following examples, in SOV languages, for example in German and Japanese, frequency adverbs such as immer and itsumo ‘always’ precede manner adverbs such as gut and yoku ‘good’. (The same holds for SVO languages such as English, as will be discussed in (14) below.)1 German (SOV) (1) a. dass er immer die Hemden gut w¨ascht that he always the shirts good washes ‘He always washes the shirts well.’ b. *dass er gut die Hemden immer w¨ascht that he good the shirts always washes Japanese (SOV) (Yuko Otsuka, pers. comm.) (2) a. Kare-wa itsumo huku-o yoku arau. he-top always clothes-acc well washes ‘He always washes the clothes well.’ b. *Kare-wa yoku huku-o itsumo arau. he-top well clothes-acc always washes

1. The data in this paper were elicited during field work in Fiji, Kiribati, and in Madagascar in 2008–2009, unless otherwise indicated. Special thanks to my informants on Malagasy, Kiribati and Fijian, i.e., Mereisi Kamoe, Meeribwa, Elizabeth Ravaorimalala, Tekarei Russell, Apolonia Tamata, Teeang (Miss Kiribati 2006), Tereau, among others, to Paul Geraghty and Shelly Harrison for very helpful discussions on Fijian and Kiribati, and also to Elizabeth Pearce and Claire Moyse-Faurie for comments on an earlier version of this article. I also thank the audiences at the University of the South Pacific (Fiji), UCLA (Los Angeles), UCL (London) and Frankfurt University for discussions, especially to Daniel B¨uring, Annabelle Cormack, Hans-Martin G¨artner, G¨unther Grewendorf, Richard Kayne, Ed Keenan, Hilda Koopman, Hans van de Koot, Ad Neeleman, Yuko Otsuka, Dominique Sportiche, and Ede Zimmermann. Field work on Kiribati (Tarawa) and Fiji (Suva and Nadi) in 2008–2009 was funded by the FNRS (Fonds National de la Recherche, Belgium). Abbreviations: abs-Absolutive, acc-Accusative, appl-Applicative, b-Background, dat-Dative, dem-Demonstrative, det-Determiner, dir-Directional, do-Direct Object, f-Focus, io-Indirect Object, link-Linker, lm-Leftperipheral Marker, nom-Nominative, pl-Plural, po-Prepositional Object, pres-Present, rec-Recipient, sg-Singular, th-Theme, topTopic, tr-Transitive.

Deriving linear order in OV/VO languages: evidence from Oceanic languages

29

Based on the work of Cinque (1999); Rackowski and Travis (2000) develop one of the now most influential studies of Austronesian adverb placement. The inverse order of adverbs is found in VOS languages, for example in Malagasy. In the domain following the verb, we find the inverse order V ◦ 10 < 9 < 8 < 7 < 6. . . This is illustrated in (3)-(4) with frequency and manner adverbs like foana (‘always’) [8], and tsara (‘well’) [10] from Malagasy and for wasoma ‘often’ and cala ‘wrong’ from Standard (Bauan) Fijian. VSO languages will not be discussed in detail, but Tongan (5), for example, behaves like the previously mentioned VOS languages in this respect. Based on these examples it can be concluded that VOS languages show the inverse order of adverbs. Malagasy (VOS) (3) a. Manasa tsara foana ny lamba i Fafa. washes good always det clothes det Fafa ‘Fafa always washes the clothes well.’ b. *Manasa foana tsara ny lamba i Fafa. washes always good det clothes det Fafa Fijian (VOS) (4) a. E sava-t-a cala wasoma na isulu o 3sg washes-tr-3sg wrong often det clothes det John. John ‘John often washes the clothes wrong.’ b. *E sava-t-a wasoma cala na isulu o 3sg washes-tr-3sg often wrong det clothes det John. John Tongan (VSO) (Yuko Otsuka, pers. communication) ipu ma’u p¯e. (5) a. ’Oku ne fufulu lelei ’a e pres 3sg wash good abs det cup always ‘He always washes the cup well.’ b. *’Oku ne ma’u p¯e fufulu ’a e ipu lelei. pres 3sg always wash abs det cup good Let us next turn to the linear order of nominal arguments. Beside the inverse order of subjects and objects (SO vs. OS) in these languages, it has been ob-

30

Joachim Sabel

served that inverse orders of nominal arguments are found in the predicate phrase as well. The variation concerns the order of objects in ditransitive constructions.

3. The order of objects in the double object construction 3.1. General remarks on ditransitives A ditransitive construction contains a verb that takes an agent, a theme and a recipient argument. The ditransitive constructions in different languages compared in this paper all contain a verb that expresses the idea of a physical or mental transfer, for example, the verbs give, send, sell, borrow, bring or verbs like tell, show and explain. The verb offer is also understood here as a verb of transfer. Ditransitive constructions can be realized in different ways, for example, as double object constructions (6a) or as prepositional ditransitive constructions (6b): (6)

a. b.

John gives the teacher the book. John gives the book to the teacher.

The languages of the world do not behave uniformly with respect to the realization of the ditransitive construction. First of all, languages do not all have the same set of ditransitive verbs, and secondly, the ditransitive verbs that they have in common may have different properties. In English, German (see (7) below), and Japanese, for example, the double object construction contains a dative and an accusative object, whereas other languages, such as for example, Malagasy (see (8) below) realize the construction with two accusative objects. Thirdly, English allows for both possibilities in (6), other languages, such as for example Malagasy, allow only for either (6a) or (6b), depending on the ditransitive verb. Standard (Bauan) Fijian allows for (6b) and not (6a). North-West Fijian allows for (6a) and (6b). Fourthly, in some languages, such as Standard Fijian, object incorporation may also be involved (see Heine and K¨onig 2007; Margetts and Austin 2007, Shopen 2007 for an overview and more possibilities of realizing ditransitives, such as for example, the serial verb strategy in a broad range of languages). In languages in which the alternation (6) is found, (6a) is often the result of an applicative construction, whereby an applicative construction is usually analyzed as involving a grammatical function changing process in which the addition of an applicative morpheme on the verb has the effect that an oblique

Deriving linear order in OV/VO languages: evidence from Oceanic languages

31

becomes the (applied) object of the verb. We find recipient/goal (see the discussion of Kiribati and North-West Fijian below), benefactive, and instrumental applicatives. 3.2. Focus/background and word order in the double object construction Let us now consider the linear orders of objects that are found in the double object construction in VOS and SOV languages. Before I turn to the examples from Oceanic languages, I will introduce some basic concepts for the analysis of double object constructions on the basis of a comparison between the SOV language German and the (Western Austronesian) VOS language Malagasy. Given that languages such as German and Malagasy show variable word order of both DP objects in the VP (i.e., scrambling), cf. (7) and (8), we have to use some tests on the basis of which to decide whether a normal, unmarked word order exists (Theme > (precedes) Recipient or Recipient > Theme), assuming, in addition, that the unmarked word order represents the basic word order. German (SOV) (7) a. dass er den Kindern das Brot that he the children-dat(rec) the bread-acc(th) gegeben hat given has ‘He gave the children the bread.’ b.

dass er that he gegeben given

das Brot den Kindern the bread-acc(th) the children-dat(rec) hat has

Malagasy (VOS) (8) a. Nanolotra ny mofo ny ankizy offer the bread-acc(th) the children-acc(rec) izy. he-nom ‘He offered the children the bread.’ b.

ny mofo Nanolotra ny ankizy offer the children-acc(rec) the bread-acc(th) izy. he-nom

32

Joachim Sabel

Let us first consider German. It can be shown on the basis of several tests that the recipient (indirect object (io)) precedes the theme (direct object (do)) in the unmarked case (7a). The same tests reveal that the inverse order is the unmarked order in VOS languages such as Malagasy, i.e. here the theme > recipient order is unmarked (8a). Hence, we also find a mirror order effect with respect to the word order of DP arguments in SOV and VOS languages.2 One test for establishing the basic word order relies on the focus (new) / background (old) information structure in the middle field of a German sentence. A wh-question/answer pair can function as the basis for establishing the informational structure of a sentence. The answer repeats part of the question (or from the context), the background information, and the wh-word asks for new information, the (information) focus. In (9a-b), for example, the dative den Kindern ‘the children’ represents the focus and the accusative object das Brot ‘the bread’ represents the background with respect to both objects. Focus/ background ordering determines the unmarked word order. Assuming that the unmarked word order is the basic word order we can find out what the basic word order of both objects is in double object constructions. In German, focus/background ordering determines the unmarked word order IO (Rec.) > DO (theme) on the basis of the following two word order constraints (cf. Lenerz 1977, Schweikert 2005). Firstly, non-focused DPs precede focused DPs (B > F, Background > Focus). Secondly, the recipient > theme order is the basic order. I assume that a violation of at least two word order constraints 2. In other Western Austronesian VOS languages such as Palauan (VOS) (Georgopoulos 1992: 168) and in Oceanic languages such as Nˆelˆemwa (VOS) (Bril 2004: 117), Drehu (VSO/SVO), Xˆarˆac`uu` (VSO/SVO) (see Moyse-Faurie (1983: 162); (1995, 87–88), Tongan (VSO), the theme > recipient word order is also cited as a possible word order in the VP. However, whether this order represents an unmarked option needs still to be clarified for these languages. Palauan (VOS) (i) Ng-mils-terir a buu’ a rengalek a Sabino. 3sg-gave-3pl det betelnut det children det Sabino ‘Sabino gave the kids some betel nuts.’ Nˆelˆemwa (VOS) (ii) Na na ve keet shi agu ali. 1sg give dir basket ben person dem ‘I gave the basket to that person.’ Drehu (VSO/SVO) (iii) Eni a ham¨eën la itus koi angeic. 1sg pres give det book to him ‘I give the book to him.’ Tongan (VSO) (Yuko Otsuka, pers. communication) (iv) Na’e ’oange ’e Mele ’a e tohi ki he tamasi’i. past give erg Mary abs det book to det boy ‘Mary gave the book to the boy.’

Deriving linear order in OV/VO languages: evidence from Oceanic languages

33

results in a marked sentence (= #), whereas a violation of only one constraint has no influence on the acceptability of a sentence. Based on this assumption, it can be shown that the order IO > DO is less restricted than the order DO > IO. This can be taken as evidence that IO (dative) – DO (accusative) – V is the base word order. In (9), the recipient is questioned, and we see that both orders of objects are perfectly acceptable in the answer. In (10), where the theme is questioned, only the order recipient > theme is unmarked, assuming throughout our discussion neutral intonation. In (10b), two word order constraints are violated. We can conclude that the recipient > theme is the basic word order in the double object construction in German. German (SOV) (9) Q: Wem hast du das Brot gegeben? whom-dat have you the bread-acc given ‘Whom did you give the bread?’

(10)

a.

das BrotB gegeben. Ich habe den KindernF I have the children-dat the bread-acc given (*B>F, IO>DO)

b.

den KindernF gegeben. Ich habe das BrotB I have the bread-acc the children-dat given (B>F, *IO>DO)

Q: Was hast du den Kindern gegeben? what-acc have you-nom the children-dat given ‘What did you give to the children?’ a.

das BrotF gegeben. Ich habe den KindernB I have the children-dat the bread-acc given (B>F, IO>DO)

den KindernB gegeben. b. #Ich habe das BrotF I have the bread-acc the children-dat given (*B>F, *IO>DO) The test provides only information on linear order. Other work on German double objects has shown that the dative recipient argument in the double object construction is in a structurally higher position in German than the theme. The relevant evidence comes from extraction and binding asymmetries (see, for example, Sabel 2002 and the literature cited there).

34

Joachim Sabel

Let us next consider focus/background ordering in the VOS language Malagasy. Focus/background ordering determines word order in Malagasy in the opposite way from SOV languages: Focused DPs precede non-focused DPs. This is another important aspect of inverse ordering because SOV and SVO languages are both focus-final. As was illustrated with examples (9)-(10), German has a focus final position. Strictly verb final languages, for example Turkish, Malayalam, Persian, also have a focus final position immediately to the left of V. The mentioned strict verb-final languages are wh-in situ languages and, for example, wh-elements have to occur in a sentence final focus-position. SVO languages also have a focus position at the end of the clause. We will see in the following sections that verb-initial Oceanic languages show the same inverse ordering of focus > background as Malagasy. For Malagasy, I assume that the constraints are F > B, Theme (TH) > Recipient (REC). Based on the generalization on the focus position and on the assumption that the theme precedes the recipient in the unmarked case, it can be derived that in Malagasy, the order theme – recipient is less restricted than the order recipient – theme. This can then be interpreted as one argument for the fact that theme > recipient is the basic word order in the double object construction.3 Malagasy (VOS) (11) Q: Iza no notolorany ny mofo? who lm offer-3sg det bread-acc ‘Whom did he offer the bread?’ a.

Nanolotra ny mofoB ny ankizyF izy. offer det bread det children 3sg-nom ‘He offers the children the bread.’ (*F>B, TH>REC)

b.

Nanolotra ny ankizyF ny mofoB izy. offer det children det bread 3sg-nom (F>B, *TH>REC)

3. Other factors may potentially influence the unmarked order of objects such as the pronominal/nominal character of the objects, their specificity/definiteness, their structural complexity, and animacy/special thematic relationships with certain verbs (Wasow and Arnold 2003). I abstract here from these aspects, keeping the examples constant with respect to properties. Note however that, with respect to other factors, mirror order effects are also attested. Malagasy has, for example, non-specific DPs preceding specific ones, whereas German has the unmarked order of definites preceding indefinites.

Deriving linear order in OV/VO languages: evidence from Oceanic languages

(12)

35

Q: Inona no natolony ny ankizy? what lm offer-3sg det children ‘What did he offer to the children?’ a.

Nanolotra ny mofoF ny ankizyB izy. offer det bread det children 3sg-nom ‘He offers the children the bread.’ (F>B, TH>REC)

b. #Nanolotra ny ankizyB ny mofoF izy. offer det children det bread 3sg-nom (*F>B, *TH>REC) In addition, Pearson (2000) discusses object shift and weak crossover effects in order to motivate the TH > REC base structure in Malagasy. On the basis of the tests, it is also demonstrated that the indirect object is in a structurally higher position than the theme in Malagasy.4 To sum up, we have seen that SOV languages show the order recipient-DP > theme-DP and frequency adverb > manner adverb, whereas the order in VOS languages is manner adverb > frequency adverb and theme-DP > recipient-DP. Based on this correlation, a typological classification has been proposed, which will be discussed in the following section.

4. A typological generalization on VO/OV languages The fact that the word order of adverbs and (theme/recipient DP) objects in the above-discussed SOV and VOS languages is exactly the opposite is astonishing and seems to have systematic reasons. Pearson (2000) expresses the correlation by classifying the languages into direct and inverse order languages. Direct order languages are the SOV languages, whereas VOS languages are indirect order languages. SVO languages show the same word order patterns as SOV languages with respect to DP object order and adverb order (and VSO languages pattern with VOS languages). In the SVO language English, for example, we

4. In the following discussion on double objects in Oceanic VOS languages, I will use other tests in order to argue for a certain linear order and hierarchical structure. Besides informationstructuring (focus-background order), I will use passivization and wh-extraction properties as diagnostics. These can be more easily applied as diagnostics to the sample of Oceanic languages discussed below.

36

Joachim Sabel

find the orders IO (REC) > DO (TH) and FA (frequency adverb) > MA (manner adverb), as illustrated in (13)–(14): (13)

a. John gave the children the books. b. *John gave the books the children.

(14)

a. John has always washed the clothes well. b. *John has well washed always the clothes.

This gives rise to the following classification, where we find two types of VO languages, i.e. direct VO languages, for example English, and inverse VO languages like Malagasy: (15)

Two types of VO languages Direct order

Inverse order

Arguments

SOV IO > DO

SVO IO > DO

VOS DO > IO

Adverbs

FA > MA

FA > MA

MA > FA

In the VO-languages with inverse order, the subject follows the verb, whereas in the direct order languages, the subject precedes the verb. Pearson (2000), following Rackowski (1998) and Rackowski and Travis (2000), derives the 3 correlations from the Universal Base Hypothesis (Kayne 1994) and Cinque’s (1999) Universal Hierarchy of Adverbs. The assumption is that SVO is the base structure for every language, as well as the IO > DO and FA > MA base orders but that different transformations apply in SOV/SVO and VOS/VSO languages, producing direct order and inverse order languages. For example, SVO languages have verb movement as illustrated in (16), which implies that the order of adverbs (16a) and arguments (16b) is retained (according to Pearson, the same holds for SOV languages; a different analysis for SOV languages on the basis of the Universal Base Hypothesis, however, is proposed in Kayne (1994)). Verb movement is motivated as a mechanism for endowing every phrase on the main projection path of the sentence with a “verbal” feature. In VOS (i.e. inverse order) languages, verb movement is blocked, therefore roll-up movement, i.e. predicate (or VP-) movement has to apply (Pearson 2000: 342). This movement results in a change of word order for adjuncts (17) and nominal arguments (18), the same holding for VSO languages.

Deriving linear order in OV/VO languages: evidence from Oceanic languages

37

Direct order (16)

a.

[FA V [MA
Object ]]

6

b.

[Subject V [Goal [
Theme ]]]

6

Inverse order (17)

a.

[FA [MA [V Object ]]]

b.

FA [ [V Object] MA
] FA


6

(18)

a.

[Subject [Goal [V Theme ]]]

b.

[Subject [[V Theme] Goal
]]

6

c.

[[V Theme] Goal
] Subject


6

The analysis predicts that in VOS languages, DO > IO order / MA > FA and IO > DO / FA > MA order should cooccur. Just two types of VO languages, i.e. direct and indirect languages should exist. However, as will be illustrated in the next section, Oceanic VOS languages do not fit into the typology in (15). Kiribati and Fijian represent a third type of VO-language where the order of the arguments is the same as in English, i.e., as in direct VO languages, but the order of adjuncts is the same as in Malagasy, i.e., as in inverse VO languages. The relevant data are discussed in sections 5 and 7.

38

Joachim Sabel

5. Linearization of adverbs and nominal objects in Oceanic VOS languages 5.1. Double objects in Kiribati (Gilbertese)5 Kiribati is an Oceanic (Micronesian) VOS language (Lynch 1998: 154; Lynch et al. 2002, chapter 3) with nominative-accusative Case marking. It is a free word order language without overt Case and number marking on nouns and an optional wh-in situ language. In wh-ex situ structures, the wh-element precedes a marker that is very similar to a demonstrative-pronoun that also introduces relative clauses (Cowell 1951). Embedded argument clauses are introduced by the pure subordinator bwa (sometimes written as b’a) that occurs in many different contexts, for example with declarative and interrogative argument clauses, as well as with reason adverbial clauses. It probably derives from a neuter demonstrative in the language (te b’ai aei ‘Art this thing’).6 Like most other Oceanic languages, Kiribati uses a formal marking for transitive verbs (a transitive marker), and has a set of pre-verbal clitic pronouns/agreement markers, which indicate the person and number of the subject. Furthermore, we find postverbal clitic pronouns/agreement suffixes agreeing in person, number, and, in the third person plural, with respect to animacy with the object. A syntactic passive construction is realized with the verbal suffix -aki. In the present context, Kiribati is interesting because it does not fit into the typology in (15). It cannot be classified as a pure inverse or direct order language. Object order in ditransitives patterns with direct order languages whereas adjunct order patterns with inverse order languages. Let us first consider ditransitives. Kiribati displays the double object construction. We find the prepositional ditransitive (see (19a)) and the double object (dative shift or applicative) construction, as shown in (19b). A natural analysis for (19b) is to assume that preposition-incorporation (P-to-V) has taken place (see the discussion below). Te is the article indicating singular, and taan is the plural article (Appl. = applicative). 5. ‘Kiribati’ is the rendition of the name ‘Gilberts.’ Captain Thomas Gilbert was the European who discovered the Gilbert Islands in 1788. The former name of the islands was ‘Tungaru’ (Sabatier and Oliva 1971). Kiribati has about 105.000 speakers and is spoken in Kiribati (formerly the Gilbert Islands). The Kiribati speech area comprises the Gilbert Islands and the island of Bwaanaba, now parts of the Republic of Kiribati, the island of Niu and the Ellice Islands (now Tuvalu). It is also spoken on parts of the Solomon Islands, Fiji, the Line Islands, and on Nauru – see Harrison (1995) for a detailed description. 6. In addition, Kiribati uses feminine and and masculine demonstrative pronouns/particles that are used to refer to people (for example, teuaei ‘this man’; neiei ‘this woman’), cf. Cowell (1951: 41).

Deriving linear order in OV/VO languages: evidence from Oceanic languages

(19)

a.

A anga te boki nakon taan reirei ataei. 3pl give det book to det teachers children ‘The children give the book to the teachers.’

b.

A anga-n taan reirei te boki ataei. 3pl give-appl det teachers det book children

39

There are at least two double object verbs in Kiribati, anga/angan ‘give’ (19) and tua/tuanga ‘tell’ (20). These verbs behave similarly with respect to the syntactic restrictions that will be discussed below. Other ditransitive verbs, such as kanako ‘send’, uota ‘bring’, kore ‘write’, kaboo ‘buy’, kaota ‘show’, kaitiboia ‘introduce’, appear only in the prepositional ditransitive construction. (20)

a.

A tua te rongorongo nakon Rui taan reirei. 3pl tell det news to Rui det teachers ‘The teachers tell the news to Rui.’

b.

A tua-nga Rui te rongorongo taan reirei. 3pl tell-appl Rui det news det teachers ‘The teachers tell Rui the news.’

Only the applied (recipient) argument can be realized as clitic pronoun/agreement marker in the double object construction. – A in (21b-c) is the 3rd person singular masc. object marker (see Sabel 2011 for discussion):7 (21)

a.

I anga-n teuaarei te m’ane. 1sg give-appl det-man det money ‘I give the man the money.’

b.

I anga-n-n-a te m’ane. 1sg give-appl-tr-3sg(him) det money ‘I give him the money.’

c. *I anga-n-n-a teuaarei. 1sg give-appl-tr-3sg(it) det-man ‘I give it the man.’

7. Noun incorporation in ditransitives is possible but very restricted. The theme can be incorporated in the prepositional ditransitive construction: (i) I angam’ane [nakon teuaarei]. 1sg give-money to det-man ‘I give the money to the man.’

40

Joachim Sabel

As already mentioned, Kiribati is a language with free word order (and without overt Case marking on nouns). It allows for clause-internal scrambling but not for scrambling out of finite clauses.The following variants of postverbal ordering of objects are found with the prepositional object construction (19a): (22)

a.

A anga [nakon Rui] te boki taan reirei. 3pl give to Rui det book det teachers All: ‘The teachers give the book to Rui.’ (V PO DO S)

b.

A anga te boki [nakon Rui] taan reirei. 3pl give det book to Rui the teachers (V DO PO S)

c.

A anga te boki taan reirei [nakon Rui]. 3pl give det book det teachers to Rui (V DO S PO)

d.

A anga [nakon Rui] taan reirei te boki. 3pl give to Rui det teachers det book (V PO S DO)

e.

A anga taan reirei [nakon Rui] te boki. 3pl give det teachers to Rui det book (V S PO DO)

f.

anga taan reirei te boki [nakon Rui]. A 3pl give det teachers det book to Rui (V S DO PO)

The examples in (23) show that scrambling out of finite clauses is impossible: (23)

a. *I iango-i-a [nakon Rui] bwa a anga taan 1sg think-tr-3sg to Rui comp 3pl give det reirei te boki. teachers det book ‘I think that the teachers give the book to Rui.’ b. *I 1sg taan det

iango-i-a [te boki] bwa a anga nakon Rui think-tr-3sg det book comp 3pl give to Rui reirei. teachers

Deriving linear order in OV/VO languages: evidence from Oceanic languages

41

Let us now come back to the double object construction in (19b).The information structure test for the basic/unmarked word order cannot be applied in the double object construction in Kiribati. The double object construction only allows for the order IO > DO, as in English (cf. (13)): (24)

a.

A anga-n taan reirei te boki ataei. 3pl give-appl det teachers det book children ‘The children give the teachers the book.’

b. *A anga-n te boki taan reirei ataei. 3pl give-appl det book det teachers children However, in contrast to English, several phenomena show that in Kiribati, as in Malagasy, focus precedes background.8 Obviously, the order in (24a) is the unmarked/basic word order. Other orders in which the recipient Rui is in postverbal position but not adjacent to the verb are impossible: (25)

a.

A anga-n Rui taan reirei te boki. (V IO S DO) 3pl give-appl Rui det teachers det book All: ‘The teachers give Rui the book.’

b. *A anga-n taan reirei Rui te boki. (V S IO DO) 3pl give-appl det teachers Rui det book c. *A anga-n te boki taan reirei Rui. (V DO S IO) 3pl give-appl det book det teachers Rui d. *A anga-n taan reirei te boki Rui. (V S DO IO) 3pl give-appl det teachers det book Rui Both objects can be realized as wh-phrases, giving rise to a multiple wh-question. In multiple wh-questions with wh-in situ, we observe the same restriction, i.e. the recipient DP has to be adjacent to the verb. Although both object orders can be found with the prepositional ditransitive construction (26), only one word order is possible in the double object construction (27): 8. This is illustrated in section 7 for the order of PPs and also in (i) below for the prepositional dative construction on the basis of the information structure test: (i) Teraa ae a ang nakon te ataei taan reirei? what lm 3pl give to det child det teachers a. A anga te boki nakon te ataei taan reirei. 3pl give det book to det children det teachers b. #A anga nakon te ataei te boki taan reirei. 3pl give to det children det book det teachers

42 (26)

(27)

Joachim Sabel

a.

Ko anga teraa nakon antai? 2sg give what to whom ‘What did you give to whom?’

b.

Ko anga nakon antai teraa? 2sg give to whom what

a. *Ko anga-n teraa antai? 2sg give-appl what whom ‘What did you give to whom?’ b.

Ko anga-n antai teraa? 2sg give-appl whom what

The applied/shifted object must immediately follow the verb. The double object construction in Kiribati resembles English in this respect. Furthermore, English and Kiribati are similar with respect to passivization in the double object constructions. Firstly, recipients are more easily passivized than themes: (28)

a. The teachers were given books (by the children). b. ??The books were given teachers (by the children).

(29)

a.

A anga-n-aki te boki taan reirei (irouia 3pl give-appl-pass the book the teachers (by ataei). children) ‘The teachers were given the book (by the children).’

b. *E anga-n-aki taan reirei te boki (irouia 3sg give-appl-pass the teachers the book by ataei). children ‘The book was given the teachers (by the children).’ Secondly, passivization of the theme becomes acceptable when the recipient is extracted (LM = left peripheral marker): (30)

Whom were the books given (by the children)?

(31)

Antai ae e anga-n-aki te boki (irouia ataei)? whom lm 3sg give-appl-pass det book (by children) ‘Whom was given the book (by the children)?’

Deriving linear order in OV/VO languages: evidence from Oceanic languages

43

Finally, the recipient acts like the structural object with respect to extraction. Whmovement in double object constructions is more restricted for themes than for recipients. (32b) is only grammatical with a special (non-information question) echo-reading (cf. Barss and Lasnik 1986). (32)

a. Who did you give which book? b. *Which book did you give who ?

(33)

a. b.

To whom did you give what ? What did you give to whom?

Chomsky’s (1973) “Superiority Condition” captures the extraction facts in (32): When two wh-phrases are present, the structurally higher (“superior”) one must move. Assuming that the recipient is in a structurally higher position, as, for example, in the analysis of double objects in Larson (1988). . . V [VP RecipiTheme]] (where marks the base position of the raised verb), ent/Goal [V  the theme cannot be extracted over the recipient. In the more recent literature, this movement constraint has been reinterpreted in terms of the Minimal Link Condition or as a Relativized Minimality/Intervention Effect (Rizzi 1990; Chomsky 1995, 2000, 2001). In the prepositional object structure. . . V [VP Theme [V  [PP Recipient]]] (33), however, the theme is higher than the prepositional object and can be extracted. An effect similar to the one observed in (32)-(33) is observed in Kiribati, i.e. the theme cannot be extracted over the recipient, see (32)/(34). As in English, no asymmetry is found with prepositional ditransitives in Kiribati, see (33)/(35):9 (34)

a.

Antai ae ko angana [te boki ra]? who(m) lm 2sg give det book which ‘Whom did you give which book?’

b. */??[Te boki ra] ae ko angan antai? det book which lm 2sg give who(m) ‘Which book did you give whom?’

9. Extraction from the prepositional object is possible as in English. However, given that Kiribati does not allow for preposition-stranding, a resumptive pronoun appears obligatorily on the preposition, irrespective of the position of the PP: (ii) a. Antai ae ko anga te boki nakoina? who lm 2sg give det book to-3sg ‘Who did you give the book to?’ b. Antai ae ko anga nakoina te boki? who lm 2sg give to-3sg det book

44 (35)

Joachim Sabel

a.

[Nakon antai] ae ko anga teraa? to who lm 2sg give what ‘To whom did you give what?’

b.

Teraa ae ko anga [nakon antai]? what lm 2sg give to who ‘What did you give to whom?’

A wide range of similar facts has been observed with double objects in English and Kiribati. Note also that the double object construction in Kiribati (and in North-West Fijian, cf. the discussion below) behaves in many respects like the locative applicative construction in Bantu languages such as Kinyarwanda (Baker 1988, 1992; Bresnan and Moshi 1990; Marantz 1993; Nakamura 1997). As pointed out in Bresnan & Moshi (1990), the applied object can be realized as an object marker in Kinyarwanda, however, object marking of the theme is impossible (compare (21)). In addition, the applied object obligatorily precedes the theme in locative applicatives (compare (24)). Furthermore, the applied object can be extracted, whereas extraction of the theme is ruled out (compare (34)). The applied object of a locative applicative can be passivized, whereas passivization of the theme is impossible (cf. (29)). Finally, Zeller and Ngoboka (2006) have observed that the theme can be passivized in those contexts where the applied object has itself been extracted (cf. (31)). Baker (1988, 1992) and Nakamura (1997) assume that in contrast to other types of applicatives, the applicatives with asymmetrical object behaviour are derived syntactically, i.e. by preposition incorporation, as I have also assumed in connection with the discussion of (19). Some of the asymmetric properties of the objects in the double object construction are accounted for in terms of Case theory. The idea is that the applied object in a locative applicative has structural Case, whereas the theme has inherent Case. Given that only structural Case can be absorbed in a passive construction, the applied object may be passivized but not the theme.10 To sum up, in this section, we have seen that with respect to the order of nominal objects in the VOS language Kiribati, the recipient precedes the theme and is also in a structurally higher position than the theme. Kiribati shows the 10. Another type of applicative exists in Kinyarwanda that also has an analytical variant with an independent preposition. In the so-called symmetrical applicatives in Kinyarwanda, for example with instrumental applicatives, the two objects do not behave differently from each other in the above-mentioned respects. Baker (1988, 1992) and Nakamura (1997) assume that this type of construction is not derived syntactically by incorporation but lexically, in that an applicative morpheme is added to the verb in the lexicon introducing a new element into the verb’s argument structure (see also McGinnis 2001 for relevant discussion).

Deriving linear order in OV/VO languages: evidence from Oceanic languages

45

direct order of objects like SVO and SOV languages and not the inverse order that was observed in VOS languages such as Malagasy. The unmarked order of objects in Malagasy is an impossible word order in Kiribati. With respect to nominal object order, Kiribati behaves like a VO-language with direct order (such as English (13)) and not like a VO-language with inverse order such as Malagasy (cf. (11)-(12)). 5.2. Adverb order in Kiribati Unexpectedly, the order of VP-internal adverbs is MA > FA in Kiribati. In this respect, the language patterns with inverse order VO-languages like Malagasy (cf. (3)) and not with English (cf. (14)). The examples in (36) show that the manner adverb precedes the frequency adverbial and that the object may intervene between both, cf. (36a). (37) illustrates that it is impossible for the frequency adverbial to precede the manner adverb (internal structure of verbs omitted).11 (36)

(37)

a.

E uatii raoi kunnikai n tai nako Rui. 3sg washes well clothes in time every Rui ‘Rui always washes the clothes well’

b.

E uatii kunnikai raoi n tai nako Rui. 3sg washes clothes well in time every Rui

c.

E uatii raoi n tai nako kunnikai Rui. 3sg washes well in time every clothes Rui

a. *E uatii n tai nako kunnikai raoi Rui. 3sg washes in time every clothes well Rui ‘Rui always washes the clothes well.’ b. *E uatii kunnikai n tai nako raoi Rui. 3sg washes clothes in time every well Rui

The manner adverb has to follow the verb: compare (36) with (38). (38)

kunnikai n tai nako Rui. *E raoi uatii 3sg well washes clothes in time every Rui ‘Rui always washes the clothes well.’

11. One might wonder whether n tai nako has special properties because it is an adverbial PP and not an adverb. However, data from Fijian in the next section will provide similar relevant examples with adjuncts that are doubtlessly adverbs.

46

Joachim Sabel

As already mentioned, Kiribati is a counterexample for the above mentioned typology (15). The data in this section have shown that VOS languages may use independent mechanisms to linearize adverbs and arguments of the verb. Let us next consider Fijian, another Oceanic VOS language. 5.3. Double objects and adverb order in Fijian Fijian is spoken by about 330.000 people in the South Pacific. It is a CentralEastern Oceanic language with nominative-accusative Case marking and VOS word order (Lynch et al. 2002, chapter 3; Ross 2004). The language has free word order without overt Case and number marking on nouns like Kiribati. It is an optional wh-in situ language. Wh-ex situ is realized as wh-preposing into the pre-verbal position.12 Like other Oceanic languages, Fijian uses formal marking for transitive verbs, and has a set of pre- and post-verbal clitic pronouns/agreement markers, which indicate the person and number of the subject and object. There are over thirty “Fijian” dialects. The extreme western and extreme eastern varieties being mutually unintelligible (see Geraghty 1983 for details). They can thus be seen as two languages (Lynch 1998: 33). In the following, I will discuss two variants of Fijian, i.e. Standard (Bauan, i.e. the eastern dialect of the island Bau) Fijian and North-West Fijian spoken in the North-West of Viti Levu in the region of Nadi and Lautoka. However, only the latter has the double object construction and is therefore relevant for our discussion of direct and inverse order languages. Standard Fijian does not have the double object construction. It has only the oblique/prepositional ditransitive construction. (The article na is used for common nouns and o is the article used for proper nouns): (39)

(40)

a.

E soli-Ø-a na ivola vei Mary o Jone. 3sg give-tr-3sg det book to Mary det John ‘John gives the book to Mary.’

b.

E soli-Ø-a vei Mary na ivola o Jone. 3sg give-tr-3sg to Mary det book det John

*E soli-Ø-a Mary na ivola o Jone. 3sg give-tr-3sg Mary det book det John ‘John gives Mary the book.’

12. It is a general characteristic of Austronesian verb-initial languages to allow for wh-in situ as well. See Polinsky and Potsdam, this volume, for a discussion of different wh-ex situ strategies in Oceanic languages.

Deriving linear order in OV/VO languages: evidence from Oceanic languages

47

The question test shows that the unmarked order of the objects is TH > REC in Fijian and that focus precedes background, as in the other verb-initial languages that I have already discussed (see also footnote 8, for similar Kiribati examples): (41)

Q: [Na cava] e solia vei Mary? det what 3sg give to Mary or

E solia [na cava] vei Mary? 3sg give det what to Mary Both: ‘What does he give to Mary?’

a.

E solia na ivolaF vei MaryB . (F>B, DO>IO) 3sg give det book to Mary ‘He gives the book to Mary.’

b. #E solia vei MaryB na ivolaF . (*F>B, *DO>IO) 3sg give to Mary det book (42)

Q: [Vei cei] e solia na ivola? to whom 3sg give det book ‘What did he give to Mary?’ a.

E solia na ivolaF vei MaryB . (*F>B, DO>IO) 3sg give det book to Mary ‘He gives the book to Mary.’

b.

E solia vei MaryB na ivolaF . (F>B, *DO>IO) 3sg give to Mary det book

As shown in (43a), the adverb order in the following example is as in Kiribati and Malagasy: MA > FA. In contrast to Kiribati, the object has to follow both adverbs (cf. (43b), (44)):13 (43)

a.

E sava-t-a cala wasoma na isulu o Jone. 3sg wash-tr-3sg wrong often det clothes det John ‘John often washes the clothes wrong.’

b. *E sava-t-a wasoma cala na isulu o Jone. 3sg wash-tr-3sg often wrong det clothes det John

13. The same ordering restrictions hold for the manner adverb vinaka ‘well’. Compare (43) with (i) and (44) with (ii).

48

Joachim Sabel

(44)

a. *E sava-t-a cala na isulu wasoma o Jone. 3sg wash-tr-3sg wrong det clothes often det John ‘John often washes the clothes wrong.’ b. *E sava-t-a na isulu cala wasoma o Jone. 3sg wash-tr-3sg det clothes wrong often det John

The same order restrictions are found if we replace wasoma ‘often’ with the adverb t¯ug¯a ‘always.’ T¯ug¯a obligatorily precedes the manner adverb. Standard Fijian also allows for N-incorporation. When the object is incorporated, the verb is obligatorily realized in its intransitive form. We get the same adverb order as with verbs containing a transitive marker and an object suffix: (45)

a.

E sava-isulu cala wasoma o Jone. 3sg wash-clothes wrong often det John ‘John often washes the clothes wrong.’

(i) a. *E sava-t-a wasoma vinaka 3sg wash-tr-3sg often well ‘She often washes the clothes well.’ b. E sava-t-a vinaka wasoma 3sg wash-tr-3sg well often

na det

isulu. clothes

na det

isulu. clothes

(ii) a. *E sava-t-a vinaka na isulu wasoma. 3sg wash-tr-3sg well det clothes often ‘She often washes the clothes well.’ b. *E sava-t-a na isulu vinaka wasoma. 3sg wash-tr-3sg det clothes well often However, the bare adverb vinaka opens up an interesting alternative. The prefix vaka- derives adverbs from adjectives and some of the bare adverbs may also appear as vaka-adverbs (see Sch¨utz 1985: 406). There is no change in meaning involved, for example, with vinaka ‘well’ and vakavinaka ‘well’. However, the word order is different. Interestingly, vakavinaka is in the same position as well in direct order languages such as English, German or Japanese, i.e. the adverbs appear in the opposite order from that found with vinaka (iiia-b). In addition, the object can intervene between both adverbs (iii c.): (iii) a. E sava-t-a wasoma vakavinaka na isulu. 3sg wash-tr-3sg often well det clothes b. *E sava-t-a vakavinaka wasoma na isulu. 3sg wash-tr-3sg well often det clothes c. E sava-t-a wasoma na isulu vakavinaka. det clothes well 3sg wash-tr-3.sg often ‘She often washes the clothes well.’ A related adverb in Standard Fijian is dau with the meaning ‘habitually, usually, normally’ and sometimes ‘always’. It appears only in pre-verbal position, as in (iv): (iv) E dau sava-t-a cala na isulu o Jone. 3sg usually wash-tr-3sg wrong det clothes det John

Deriving linear order in OV/VO languages: evidence from Oceanic languages

49

b. *E sava-isulu wasoma cala o Jone. 3sg wash-clothes often wrong det John Standard Fijian is compatible with (15), i.e. the order of objects and the order of adverbs is the same as in inverse word order languages (but see footnote 11). However, ditransitives are always realized as prepositional ditransitives and not as double objects. These examples contain a different construction from the one I have discussed in Kiribati and Malagasy. The North-West Fijian dialect (VOS, nominative-accusative) is more interesting with respect to ditransitive constructions. In this dialect, we find the verb vagania ‘give’ that is used with both the prepositional object and the double object construction.14 (46)

a.

A vagataki na ivola ’i vo qasenivuli ’o Jone. 3sg give det book to det teachers det John ‘John gives the book to the teachers.’

b.

A vagan-i-a ’o qasenivuli na ivola ’o Jone. 3sg give-tr-3sg det teachers det book det John ‘John gives the teachers the book.’

The order of objects in (46a-b) can be inverted, retaining the same meaning: (47)

a.

A vagataki ’i vo qasenivuli na ivola ’o Jone. 3sg give to det teachers det book det John ‘John gave the book to the teachers.’

b.

A vagan-i-a na ivola ’o qasenivuli ’o Jone. 3sg give-tr-3sg det book det teachers det John ‘John gives the teachers the book.’

Given (46b), (47b), extraction (focusing) of the recipient (either of the DP ’o qasenivuli ‘the teachers’ or its interrogative counterpart o cei ‘who’) is possible but not of the theme na ivola ‘the book’ or its interrogative counterpart na cava ‘what’: (48)

a.

’O det o det

qasenivuli a vagan-i-a i vola teachers 3sg give-tr-3sg det book Jone. John

14. Thanks to Paul Geraghty for bringing this fact to my attention.

50

Joachim Sabel

b. *Na det o det

ivola a vagan-i-a ’o qasenivuli book 3sg give-tr-3sg det teachers Jone. John

The same extraction asymmetry has been already discussed for English, Kiribati and Bantu in section 5.1. It provides evidence for the fact that the recipient is in a structurally higher position than the theme. North-West Fijian differs from English and Kiribati in allowing the word orders in (46b) and (47b). Given that object scrambling is possible, the questionanswer test can be applied to find out the basic/unmarked linear order of objects in the double object construction. But given that extraction (focusing) of the recipient (either in the form of the DP ’o qasenivuli ‘the teachers’ or its interrogative counterpart o cei) is possible but not extraction of the theme or its interrogative counterpart na cava ‘what’ (cf. (48)), the questions are formed with the wh-element in situ. As with the other VOS languages, focus precedes background. In contrast to Malagasy, but in line with Kiribati and given the extraction facts in (48), I assume that the “normal” word order of the two objects is recipient > theme. On the basis of both assumptions (focus > background, IO (REC) > DO (TH)) we can explain that only answer (50b) is marked. Only in this case, both relevant constraints on word order are violated: (49)

(50)

Q: A vagania ’o qasenivuli na cava ’o Jone? 3sg give det teacher det what det John ‘What does John give to the teacher?’ a.

A vagania ’o qasenivuliB na ivolaF ’o Jone. (IO>DO, *Foc>Backg.)

b.

A vagania na ivolaF ’o qasenivuliB ’o Jone. (*IO>DO, Foc>Backg.)

Q: A vagania ’o cei na ivola ’o Jone? 3sg give det whom det book det John a.

A vagania ’o qasenivuliF na ivolaB ’o Jone. (IO>DO, Foc>Backg.)

b. #A vagania na ivolaB ’o qasenivuliF ’o Jone. (*IO>DO, *Foc>Backg.)

Deriving linear order in OV/VO languages: evidence from Oceanic languages

51

To conclude, the unmarked word order for objects is recipient > theme in this Fijian dialect. Concerning the order of adverbs, we observe the same phenomenon as in Kiribati. Adverb order is as in the inverse order languages Malagasy, Kiribati, and Standard Fijian. However, the order depends on the position of the adverbs relative to the verb. North-West Fijian allows for both adverbs either to follow or to precede the verb. In the following examples the first option is chosen. Both adverbs appear post-verbally. In this position, they can appear only in the ‘inverse’ order MA > FA: (51)

a.

Sa sava leya i sulu vakacau ’o Jone. tns/prog wash wrong det clothes often det John

b.

Sa sava i sulu leya vakacau ’o Jone. tns/prog wash det clothes wrong often det John

c. *Sa sava vakacau i sulu leya ’o Jone. tns/prog wash often det clothes wrong det John d. *Sa sava i sulu vakacau leya ’o Jone. tns/prog wash det clothes often wrong det John The same adverbs can precede the verb. In this case, only the ‘direct’ FA > MA order is grammatical: (52)

a.

sava i sulu ’o Jone. Sa vakacau leya wrong wash det clothes det John tns/prog often

leya vakacau sava i sulu ’o Jone. b. *Sa tns/prog wrong often wash det clothes det John Ungrammaticality results when one of the adverbs precedes while the other follows the verb: (53)

a. *Sa vakacau sava leya i sulu ’o tns/prog often washes wrong det clothes det Jone. John b. *Sa leya sava vakacau i sulu ’o tns/prog wrong washes often det clothes det Jone. John

52

Joachim Sabel

To sum up, North-West Fijian allows the direct as well as the inverse order of adverbs, depending on whether the adverbs are located in preverbal or postverbal position. Adverb and object order in the North-West Fijian dialect is like in Kiribati. In the postverbal region of the sentence, manner adverbs precede frequency adverbs as in indirect order languages and indirect objects precede direct objects in their unmarked/base position as in direct order languages. The typology in (15) is confronted with empirical counter-evidence from at least Kiribati and North-West Fijian. These languages do not show a one-to-one correlation between adverb order and the order of nominal objects. The question arises as to whether there is an alternative empirical generalization that covers the data just described. We saw that the order of objects varies. Objects following the verb appear in inverse or direct order in VOS languages. The order of adverbs, however, shows a similar word order pattern with respect to the position of the verb in all the verb-initial VO languages discussed. Adverbs following the verb appear in inverse order and adverbs preceding the verb appear in direct order. It might be possible that the word order typology in (15) should refer only to adverbs, or more generally to adjuncts, also including adjectives. In order to see whether this idea is promising, I will consider how adjective order in the DP is represented in OV/VO languages in section 7. Before doing so, I briefly discuss the empirical results in light of the analysis that has been proposed for deriving the relevant word order facts in VO/OV languages in the following section.

6. Deriving inverse word order One important generalization to be drawn from the preceding discussion is that arguments of the verb and adjuncts may be independently linearized. The mechanism that was proposed in (16) for deriving direct order languages and the derivations in (17)-(18) for deriving inverse order languages is too strong if it is applied to all OV/VO languages. Let us assume that the universal base hypothesis is assumed in conjunction with strict cyclicity and a general freezing constraint (no extraction from dislocated constituents), and that the base (or merged) order is (FA) subject (recipient) (MA) V theme. Then one way to deal with the facts would be to assume that verb raising in the sense of (16b) is permitted in the VOS languages Kiribati and North-West Fijian that have the IO > DO order, as well as predicate movement, to yield the inverse MA > FA order. The simplified example (36a) (with a transitive verb) from Kiribati would be derived by (16a), i.e. verb movement ([subject V[ VP MA theme]]), followed by (string-vacuous) subject movement out of ν P subject [ V MA theme], followed by merger of the FA and fur-

Deriving linear order in OV/VO languages: evidence from Oceanic languages

53

ther roll-up movement in front of the FA as in (17c). After the ν P-remnant moves in front of the already dislocated subject, we get [ V MA theme] FA subject. In (36b), the derivation is similar, except that the theme argument has undergone object shift in front of MA. However, the analysis undermines the initial motivation for predicate raising as a mechanism for generating reverse adverb orders as in (17). Recall that the motivation for predicate raising comes from the impossibility of having verb movement in a language. (Cinque (2005) also assumes that only XP-movement applies in DPs to generate all possible word orders between demonstrative, numeral, adjective and noun. N-movement is explicitly excluded.) Furthermore, given that examples (44) (and (iia.-b.) in footnote 13) are ungrammatical, the outlined derivations are excluded in Standard Fijian. In Standard Fijian, only the order V MA FA theme (see (43a)) is possible. In order to avoid the simultaneous application of head and roll-up movement, we could assume that starting from a base structure (FA) subject (recipient) (MA) V theme, the subject raises out of ν P and then the object(s) is (are) moved in front of the subject, followed (or preceded) by merger of FA above the raised subject. Finally, ν P-remnant movement applies (see Massam 2000). The derivations result inV (recipient) MA theme FA subject orV (recipient) theme MA FA subject, as in (36a-b). But the question remains as to why this derivation should be impossible in North-West Fijian. North-West Fijian has scrambling. Scrambling is also possible in Standard Fijian, cf. examples (39a-b). The alternative derivation that avoids simultaneous application of head and roll-up movement is to assume that in the discussed Oceanic VOS languages, word order is derived without predicate fronting. The observed facts are compatible with a base-generated structure involving right adjunction and/or rightperipheral specifiers, as assumed, for example, in Ernst (2002). (See also the introduction in G¨artner et al. (2006); Kaufmann (2006) and Abels and Neeleman (2007) for an analysis of DP-internal structure along these lines). In the Oceanic languages with double object constructions, we would have the structure [T [ν P subject [ν  ν [VP recipient [V  [V theme] MA]]] FA] ]. The verb is moved into the position of ν and the subject to Spec TP. Adjuncts are right adjoined and object shift of the theme argument applies (optionally in Kiribati, cf. (36)) to the right, targeting a position between MA and FA (cf. (36a)) or to the right of FA. In North-West Fijian, however, object shift applies obligatorily to a position to the right of FA (43a).15

15. Given the variation of IO > DO and DO > IO order in VOS languages, we would also expect to find both orders in SVO and SOV languages. Most of the SVO/SOV languages show the unmarked order IO > DO. However, Heine and K¨onig (2010) report that the unmarked order DO > IO can be found in the double object construction of, for example, Eskimo (SOV),

54

Joachim Sabel

Let us next consider the examples (22), (24), and (25) from Kiribati. Given the proposed structure for ditransitives, (24a) represents the canonical order. (25a) involves extraposition of VP, whereas (24b) and (25b-d) cannot be derived (due to an independent adjacency requirement). In the prepositional ditransitive (22), I assume that the theme is promoted and moved to Spec VP and the recipient is a demoted PP that is right-adjoined to Spec VP giving rise to (22b). In addition to this, (22c) involves PP-extraposition over the subject, (22d) extraposition of the lower VP-segment (stranding PP), (22e) PP-extraposition over the subject followed by extraposition of the lower VP-segment, (22a) “short” VP-extraposition over PP, and (22f) is derived by extraposition of VP. This is the explanation for the scrambling data on the basis of a configurational sentence structure.

7. Adjunct order again: adjectives and PP-adverbials It has already been established, even with respect to Oceanic languages, that there is a systematic ordering relation between the elements in the DP to the left and to the right of the nominal head (see, among others, Greenberg 1966: 87; Cinque 2005; Abels and Neeleman 2007; Kahnemuyipour and Massam 2006; Pearce, this volume). Let us start with head final structures. Adjectives in DP are ordered similarly to adverbs in German and Japanese. The following example contains a quantitative and a qualitative adjective. The order is the same as the order of frequency and manner adverbs. Japanese (Yuko Otsuka, pers. comm.) (54) a. takusan-no yoi kangae many-link good idea ‘many good ideas’ b. *yoi takusan-no kangae good many-link ideas German (55) a.

viele gute Ideen many good ideas ‘many good ideas’

Nivkh (SOV), Huichol (SOV), Danish (SVO), Cantonese (SVO), Bukiyip (SVO), Ewe (SVO), Mapuche (SVO).

Deriving linear order in OV/VO languages: evidence from Oceanic languages

55

b. *gute viele Ideen good many ideas V-initial languages are also N-initial. In head-initial DPs we find the inverse adjunct order. Consider the following examples from English and Standard Fijian. (56)

the red leaf-made Fijian skirts

The Fijian order of adjectives is the mirror image of the order found in English. Other orders are excluded in English as well as in Standard Fijian. (57b) is one example of an impossible ordering in Fijian. The ordering facts cannot be demonstrated with examples from Kiribati. The language has only predicative adjectives. Standard Fijian (57) a. Na liku vakaviti draunikau damudamu det skirt Fijian leaf red b. *Na liku vakaviti damudamu draunikau det skirt Fijian red leaf Note also that adjectives that precede the head in Standard Fijian appear in the same order as in English: (58)

levu na ikarua ni gone many det second poss child ‘many second babies’

Based on the ordering facts observed with adjectives and adverbs, we can assume that (59) is empirically adequate (where V = full verb): (59)

Generalization on the order of adjuncts Adjuncts (adverbs and adjectives) following the head they modify appear in inverted order. a. 1 > 2 > 3 > 4 . . . N/V . . . 4 > 3 > 2 > 1 b. 1 > 2 > . . . N/V . . . 4 > 3 . . .

Note that (59) does not say anything about structures in which one adverb/ adjective precedes and the other follows the head. In fact, these cases are easily misinterpreted. Consider example (60a). It seems that we are dealing with two

56

Joachim Sabel

adverbs in postverbal position. Then according to (59a) we would expect the inverse order bien > souvent to be the only possible order. However, as shown in (60b-c), this order is impossible. (Note that (60b) is ungrammatical in the sense of the indicated translation. Although this is irrelevant to our discussion, it is grammatical with a reading in which bien modifies souvent.) French (60) a.

Marie lave souvent bien le linge. Mary washes often well the clothes ‘Mary often washes the clothes well.’

b. *Marie lave bien souvent le linge. c. *Marie lave bien le linge souvent. In fact, the examples are not problematic for (59). They are instances of (59a), where both adverbs are generated in a preverbal position. In order to understand why this is the case, we have to consider (61). (61)

Marie a souvent bien lav´e le linge. Mary has often well washed the clothes ‘Mary has often washed the clothes well.’

The example shows that in its base position, the full verb is located to the right of both adverbs. The verb raises from this position below the two adverbs over the frequency adverb in (60a). So far, we have only considered the order of adverbs and adjectives. Let us next turn to the order of PP-adjuncts. They show similar ordering patterns as adverbs and adjectives. Consider, for example, the order of temporal and locative PPs in an SOV and VOS language. In the German examples (63), based on (62), the only violation of two word order constraints at the same time is found in (63b) and this is the most marked sentence. We can conclude that in German, the base order is temporal > locative (see Schweikert 2005: 102). (62)

Hans hat am Sonntag in M¨unchen geschlafen. Hans has on Sunday in Munich slept

(63)

Q: Wo hat Hans am Sonntag geschlafen? where has Hans on Sunday slept a.

Hans hat am Sonntag in M¨unchen geschlafen. Hans has on Sunday in Munich slept (T.>L., Back.>Foc.)

Deriving linear order in OV/VO languages: evidence from Oceanic languages

b.

(64)

57

#Hans hat in M¨unchen am Sonntag geschlafen. Hans has in Munich on Sunday slept (*T.>L., *Back.>Foc.)

Q: Wann hat Hans in M¨unchen geschlafen? when has Hans in Munich slept a. Hans hat am Sonntag in M¨unchen geschlafen. Hans has on Sunday in Munich slept (T.>L., *Back.>Foc.) b.

Hans hat in M¨unchen am Sonntag geschlafen. Hans has in Munich on Sunday slept (*T.>L., Back.>Foc.)

In Kiribati, the order is locative > temporal. Again, we have to assume that focus precedes background in Kiribati in order to be able to derive the marked variant (see also footnote 8, for focus/background order in the prepositional ditransitive construction in Kiribati). (66) and (67) illustrate the wh-question in its wh-ex situ and wh-in situ variants. The only violation of two word order constraints is found in (66a). (65)

E matu i Betio n te tabati Rui. 3sg slept in Betio prep det Sunday Rui ‘Rui slept on Sunday in Betio.’

(66)

Q: E matu ia Rui n te tabati? 3sg slept where Rui prep det Sunday or

Ia ae matu n te tabati Rui? where lm slept prep det Sunday Rui Both: ‘Where did Rui sleep on Sunday?’

a. #E matu n te tabati i Betio Rui. 3sg slept prep det Sunday in Betio Rui (*L.>T., *Foc.>Back.) b.

(67)

E matu i Betio n te tabati Rui. 3sg slept in Betio prep det Sunday Rui (L.>T., *Foc.>Back.)

Q: E matu n ningai Rui i Betio? 3sg slept prep when Rui in Betio

58

Joachim Sabel

or

N ningai ae matu Rui i Betio? prep when LM slept Rui in Betio

a.

E matu n te tabati i Betio Rui. 3sg slept prep det Sunday in Betio Rui (*T.>L., Foc.>Back.)

b.

E matu i Betio n te tabati Rui. 3sg slept in Betio prep det Sunday Rui (T.> L., *Foc.>Back.)

The data from Kiribati illustrate that in this language, focus precedes background in the postverbal domain, that locatives precede temporals in unmarked order, and that the broad range of adjuncts show the inverse order in Kiribati, i.e. adjectives, adverbs and PP-adverbials.

8. Adjuncts, arguments, focus/background and configurationality Finally, I would like to mention another implication of the present findings. In order to classify the typological properties of languages within a restrictive theory for syntactic parameters, it has been claimed that for a range of languages, the use of pronominals has consequences concerning the structural positions of nominal expressions, giving rise to a cluster of syntactic properties that are characteristic for a certain type of free word order language, i.e. the “pronominal argument languages” (Jelinek 1984; Baker 1995, among others). The idea is that free word order is not a homogeneous phenomenon and that there is no single macroparameter that is responsible for the free word order phenomenon, see Hale (1983); Baker (1995, 2001), and Pensalfini (2004); Sabel and Saito (2005) for analyses of different types of free word order languages, i.e. configurational and nonconfigurational free word order languages. As regards nonconfigurational free word order languages, the presence of subject/object-markers as pronominal arguments has led to an analysis of the full DPs as adjuncts, for example, in languages such as Mohawk (Baker 2001). Being adjoined to the predicate phrase, the DPs may be generated in different linear order, and they may be freely dropped. Extraction asymmetries among DPs do not exist. Furthermore, because the adjuncts are all dominated by the same nodes, the absence of Case marking on full DPs is explained. Case is usually assigned under a Spec-Head (or within a head complement) relationship that is missing here. Note that languages such as Kiribati and (Standard as well as North-West) Fijian also have a rich

Deriving linear order in OV/VO languages: evidence from Oceanic languages

59

pronominal system, free word order, absence of overtly Case marked nominal constituents and the possibility of allowing freely for pro-drop. Therefore, they have been claimed to be pronominal argument languages (see Alderete 1998). However, the preceding discussion has provided evidence for the conclusion that these languages are not instances of the non-configurational “pronominal argument languages” (for more discussion, see Sabel 2011). The discussion in the preceding sections has shown that (Standard as well as North-West) Fijian and Kiribati display clear configurational properties. We have observed asymmetries between theme and recipient arguments with respect to the ordering of objects in the double object construction, concerning wh-extraction in the left peripherpy of the sentence, and with respect to passivization. Furthermore, the restrictions for licensing focused arguments can be explained if we assume a structural focus position in the post-verbal domain (Jayaseelan 2008). Finally, the fact that the order of adjuncts is fixed, as in other configurational languages, suggests that sentences and DPs have the full set of functional categories and a fully-fledged hierarchical structure.

9. Conclusion In this paper, I have shown that the Austronesian VOS languages Malagasy, Kiribati, and North-West Fijian differ with respect to linearization of arguments in the predicate phrase. Adjuncts, however, are uniformly ordered. I have shown that focus precedes background in the postverbal domain of all discussed Austronesian verb-initial languages, even though the order of nominal objects varies in these VOS languages. The linearization of adjuncts depends on their position relative to the head they modify. Furthermore, an empirical generalization on adjunct order, i.e. that head-initial languages/structures show inverse adjunct order, was also shown to be stable. I have discussed the mechanism that has been proposed in the literature for deriving the mirror orderings in VOS languages such as Malagasy, i.e. the roll-up movement analysis and the VP-remnant movement analysis, and I have pointed out some potential problems for these analyses when they are applied to Kiribati and North-West Fijian.

References Abels, Klaus and Ad Neeleman 2007 Linear Asymmetries and the LCA. Manuscript, University College London.

60

Joachim Sabel

Alderete, John 1998 Canonical types and noun phrase configuration in Fijian. In Matt Pearson (ed.), UCLA Occasional Papers in Linguistics 21, Proceedings of the third and fourth meetings of the Austronesian Formal Linguistic Association (AFLA III-IV), 19–44. Los Angeles, CA: UCLA Department of Linguistics. Baker, Mark 1988 Incorporation: A Theory of Grammatical Function Changing. Chicago: Chicago University Press. Baker, Mark 1992 Thematic conditions on syntactic structures: evidence from locative applicatives. In: Iggy M. Roca (ed.),Thematic Strucuture: Its Role in Grammar, 23–46. Berlin: Foris. Baker, Mark 1995 The Polysynthesis Parameter. Oxford: Oxford University Press. Baker, Mark 2001 The natures of nonconfigurationality. In: Mark Baltin and Chris Collins (eds.), The Handbook of Contemporary Syntactic Theory, 407–438. Oxford: Blackwell Publishers. Barss, A. and H. Lasnik 1986 A note on anaphora and double objects. Linguistic Inquiry 17: 347–354. Cambridge, MA: Massachusetts Institute of Technology Press. Bresnan, Joan and Lioba Moshi 1990 Object Asymmetries in Comparative Bantu Syntax. Linguistic Inquiry 21/2: 147–181. Cambridge, MA: Massachusetts Institute of Technology Press. Bril, Isabelle 2004 Deixis in Nˆelˆemwa (New Caledonia). In: G. Senft (ed.), Deixis and demonstratives in Oceanic languages, 99–127. (Pacific Linguistics 562.) Canberra: Pacific Linguistics. Chomsky, Noam 1973 Conditions on transformations. In: Stephen Anderson and Paul Kiparsky (eds.), A Festschrift for Morris Halle, 232–286. New York: Academic Press. Chomsky, Noam 1995 The Minimalist Program. Cambridge, MA: Massachusetts Institute of Technology Press. Chomsky, Noam 2000 Minimalist inquiries: the framework. In: R. Martin, D. Michaels and J. Uriagereka (eds.), Step by Step, 89–155. Cambridge, MA: Massachusetts Institute of Technology Press.

Deriving linear order in OV/VO languages: evidence from Oceanic languages

61

Chomsky, Noam 2001 Derivation by phase. In: M. Kenstowicz (ed.), Ken Hale. A Life in Language, 1–52. Cambridge, MA: Massachusetts Institute of Technology Press. Cinque, Guglielmo 1999 Adverbs and Functional Heads. A Cross-linguistic Perspective. Oxford University Press: Oxford Cinque, Guglielmo 2005 Deriving Greenberg’s universal 20 and its exceptions. Linguistic Inquiry 36: 315–332. Cambridge, MA: Massachusetts Institute of Technology Press. Cowell, Reid 1951 The structure of Gilbertese. Beru, Gilbert Islands: Rongorongo Press. Ernst, Thomas 2002 The Syntax of Adjuncts. Cambridge: Cambridge University Press. G¨artner, Hans-Martin, Law, Paul and Sabel, Joachim 2006 A critical introductory survey. In: H.-M. Gärtner, P. Law, and J. Sabel (eds.), Clause Structure and Adjuncts in Austronesian Languages, 1–42. Berlin: Mouton de Gruyter. Georgopoulos, Carol 1992 Syntatic Variables, Resumptive Pronouns and A’-Binding in Palauan. New York: Kluwer. Geraghty, Paul 1983 The history of the Fijian languages. (Oceanic Linguistics special publication 19.) Honolulu: University of Hawaii Press. Greenberg, Joseph 1966 Language Universals: With Special Reference to Feature Hierarchies. The Hague: Mouton. Hale, Kenneth L. 1983 Warlpiri and the grammar of non-configurational languages, Natural Language and Linguistic Theory 1: 5–47. Dordrecht: Springer. Harrison, Shelly 1995 Kiribati. In: D. T. Tyron (ed.), Comparative Austronesian Dictionary: An introduction to Austronesian studies, 879–893. Berlin: Mouton de Gruyter. Heine, Bernd and Christa K¨onig 2010 On the linear order of ditransitive objects. Language Sciences 32: 87– 131. Jayaseelan, K. A. 2008 Topic, Focus and adverb positions in clause structure. Nanzan Linguistics 4: 43–68.

62

Joachim Sabel

Jelinek, Eloise 1984 Empty categories, case, and configurationality, Natural Language and Linguistic Theory 2: 39–76. Dordrecht: Springer. Kahnemuyipour, Arsalan and Diane Massam 2006 Patterns of phrasal movement: The Niuean DP. In: Hans-Martin Gartner, Paul Law and Joachim Sabel (eds.), Clause Structure and Adjuncts in Austronesian Languages, 125–150. Berlin: Mouton de Gruyter. Kaufmann, Daniel 2006 Adverb ordering in Tagalog. In: Hans-Martin G¨artner, Paul Law and Joachim Sabel, Clause Structure and Adjuncts in Austronesian Languages. Berlin: Mouton de Gruyter. Kayne, Richard 1994 The Antisymmetry of Syntax. Cambridge, MA: Massachusetts Institute of Technology Press. Larson, Richard K. 1988 On the double object constructions. Linguistic Inquiry 19: 335–391. Cambridge, MA: Massachusetts Institute of Technology Press. Lenerz, J¨urgen 1977 Zur Abfolge der nominalen Satzglieder im Deutschen. (Studien zur deutschen Grammatik 5.) T¨ubingen: Niemeyer. Lynch, John 1998 Pacific Languages. Honolulu: University of Hawaii Press. Lynch, John, Malcolm Ross and Terry Crowley 2002 The Oceanic languages. (Curzon Language Family Series.) Richmond, Surrey: Curzon Press. Marantz, Alec 1993 Implications of asymmetries in double object constructions. In: Sam A. Mchombo (ed.), Theoretical Aspects of Bantu Grammar, 113–148. Stanford: Center for the Study of Language and Information Publications. Margetts, Anna and Peter Austin 2007 Three-participant events in the languages of the world: towards a crosslinguistic typology. Linguistics 45-3: 393–451. Madrid: Complutense University of Madrid. Massam, Diane 2000 VSO and VOS: aspects of Niuean word order. In: A. Carnie and E. Guilfoyle (eds.), The Syntax of Verb Initial Languages, 97–117. Oxford: Oxford University Press. McGinnis, Martha 2001 Variation in the phase structure of applicatives. In: Pierre Pica and Johan Rooryck (eds.), Linguistic Variation Yearbook 1: 101–142. Amsterdam: John Benjamins Publishing Company.

Deriving linear order in OV/VO languages: evidence from Oceanic languages

63

Moyse-Faurie, Claire 1983 Le drehu. Langue de Lifou (Iles Loyaut´e): Phonologie, morphologie, syntaxe. Paris: SELAF (Soci´et´e d’´etudes linguistiques et anthropologiques de France). Moyse-Faurie, Claire 1995 Le xˆarˆac`uu` , langue de Thio-Canala (Nouvelle-Cal´edonie). Paris: SELAF (Soci´et´e d’´etudes linguistiques et anthropologiques de France). Nakamura, Masanori 1997 Object extraction in Bantu applicatives: some implications for minimalism. Linguistic Inquiry 28-2: 252–278. Cambridge, MA: Massachusetts Institute of Technology Press. Pearce, Elizabeth this volume Two or three things in the Unua noun phrase. Pearson, Matthew 2000 Two types of VO languages. In: Peter Svenonius (ed.), The Derivation of VO and OV, 327–363. Amsterdam and Philadelphia: John Benjamins. Pensalfini, Robert 2004 Towards a typology of configurationality. Natural Language and Linguistic Theory 22: 359–408. Dordrecht: Springer. Polinsky, Maria and Eric Potsdam this volume Questions and word order in Polynesian. Rackowski, Andrea 1998 Malagasy adverbs. In: Ileana Paul (ed.), The structure of Malagasy, Volume II. UCLA Working Papers in Linguistics, 11–33 Rackowski, Andrea and Lisa Travis 2000 V-initial languages. X or XP movement and adverbial placement. In: Andrew Carnie and Eithne Guilfoyle (eds.), The Syntax of Verb-Initial Languages, 117–141. Oxford: Oxford University Press. Rizzi, Luigi 1990 Relativized Minimality. Cambridge, MA: Massachusetts Institute of Technology Press. Ross, Malcolm 2004 The morphosyntactic typology of Oceanic languages. Language and Linguistics 5-2: 491–541. Taipei: Academia Sinica Sabatier, Ernest and M. Oliva 1971 Gilbertese-English Dictionary. Tarawa: Sacred Heart Mission. Sabel, Joachim 2002 Die Doppelobjekt-Konstruktion im Deutschen. Linguistische Berichte 190: 229–244. Sabel, Joachim 2011 Configurationality, long dependencies, and object agreement in Kiribati and Fijian. In: Proceedings of the Seventeeth Meeting of the Austronesian. Formal Linguistics Association (AFLA XVII).

64

Joachim Sabel

Sabel, Joachim and Mamoru Saito 2005 The Free Word Order Phenomenon. Berlin/New York: Mouton de Gruyter. Sch¨utz, Albert J. 1985 The Fijian Language. Honolulu: University of Hawaii Press. Schweikert, Walter 2005 The Order of Prepositional Phrases in the Structure of the Clause. Benjamins: Amsterdam Shopen, Timothy 2007 Language Typology and Syntactic Description. Vol. I: Clause Structure. Cambridge: Cambridge University Press. Wasow, Thomas and Jennifer Arnold 2003 Post-verbal constituent ordering in English. In: G. Rohdenburg and B. Molndorf (eds.), Determinants of Grammatical Variation in English, 119–154. Berlin: Mouton de Gruyter. Zeller, Jochen and Jean Paul Ngoboka 2006 Kinyarwanda locative applicatives and the minimal link condition. Southern African Linguistics and Applied Language Studies, Volume 24, Number 1, February 2006, 101–124.

Questions and answers in Niuean1 Diane Massam, Donna Starks and Ofania Ikiua

Introduction In this paper we consider polar questions in Niuean, a Polynesian language listed as endangered, with 5481 speakers in New Zealand, and 1230 in Niue (Siosikefu and Haberkorn 2008; New Zealand Ministry of Education, 2009). Given its status, urgent attention needs to be placed on understanding and documenting its structures (Sperlich 1995, 1996, 2005). Niuean is a language that uses three separate particles to form polar questions. This paper looks at the particles in polar questions, their function and grammar, as well as how polar questions containing each of the three particles are answered. We base our study on questions and answers in interviews, a question rich context. Polar questions include various types of non-content questions such as yes-no questions and polar alternative questions. Cross-linguistically, polar questions can be formed by intonation, changes in word order, or particles. Typologies of languages (i.e. Cheng 1997; Haspelmath et al. 2001) categorize languages that use particles to form polar questions versus languages that use other devices. The first group would include languages such as Mandarin, Hindi, Palauan (Cheng 1997) and Qu´ebec French (Picard 1992; Vinet 2001; Morin 2006, 2009), for example, while the second group would include English, Polish, and Standard French (Cheng 1997). In the first group of languages, particles may be optional or obligatory and can occur in different positions (Haspelmath et al. 2001). Only some question particles are used to form canonical questions, whereas others appear to have a tag-like function. There has been relatively little research on polar questions in Austronesian languages, with few detailed studies in the area. Most available material is to be found in grammars or descriptions of individual languages, which typically contain a small section or note stating 1. We would like to thank our research assistants, Colin Gorrie, Isaac Gould, and Nicholas Rolle for help at various stages of this project. Funding for this research was provided by a Social Sciences and Humanities Research Council of Canada Standard Research Grant “Featural Variation in the Left Periphery of a Predicate Initial Language”, a Royal Society of New Zealand Marsden Grant “Pasifika Languages of Manukau”, and a University of Auckland internal research grant on “Niuean Discourse Particles”.

66

Diane Massam, Donna Starks and Ofania Ikiua

that non-content questions are formed by a particle, intonation, or both. Closely related languages may differ as to whether they form questions by intonation or a particle (see Otsuka 2006 for a discussion of Niuean and Tongan). The relationship, if any, between yes-no question particles in related languages is often unclear (see Paul 2001 for Malagasy and Tagalog). In some Austronesian languages, question particles have a marked function (i.e., Tetun ka, which is used for disjunctive questions, see van Klinken 1999). Although it is possible to have more than one question particle in Austronesian languages, researchers who note the use of more than one question particle often refer to the ambiguous or dubious status of one of the particles because of its marked semantic reading (see Bauer 1997 on New Zealand M¯aori ), or its occurrence in wh-questions (e.g Malagasy moa, see Paul 2001). Finally, some grammars that state that yes-no or polar questions are formed by intonation alone, give examples that contain adverbs or clitics that express uncertainty (see Churchward 1953 for Tongan). These items often express functions similar to those expressed by question particles in other languages, and are in need of further analysis. Part of the difficulty in researching polar questions has to do with the exact nature of the structure under consideration, and the complexity of information-seeking. The terminology can vary (see Haspelmath et al. 2001). Some researchers use yes-no questions in a broad sense. For these researchers, yes-no questions contrast with content questions (i.e. wh-questions), and do not contain a wh-question word. For others, this broad grouping is labelled “polar questions”. We follow the latter usage in this paper. For a finer-grained analysis, the focus is on response type. Yes-no questions are limited to those that elicit a yes or no response, while polar alternative questions (“Will you go or not?”) are considered separately. Questions that elicit other response types are labelled differently (e.g. echo, rhetorical questions etc., see for example, Huddleston 1994). Part of our ongoing research is to determine the question types relevant for Niuean, but in this paper we have not yet made finer-grained distinctions among the various polar question types, rather, we have included all non-wh questions in our examination. Before turning to a discussion of the Niuean question particles, we provide here a brief overview of basic sentence structure. Niuean has predicate-subjectobject or predicate-initial word order. The predicate is usually preceded by a tense particle. Negation, when it occurs, appears between the tense particle and the predicate. The predicate is followed by a series of particles, discussed further below (see Figure 2). The very last of these particles is the question particle. The particles in turn are followed by the arguments. The subject appears first, followed by the object, then by the indirect object and obliques. Generally, each noun phrase in Niuean is preceded by a case article, which encodes both the case (absolutive, ergative, locative or etc.) and the proper or common status of the

Questions and answers in Niuean

67

following noun. The three positions where we find question particles in Niuean are first, after the predicate at the end of the sequence of post-verbal particles, second, at the very end of the entire sentence, and third, after negation. The basic word order, as well as the possible positions for question particles, is shown in Figure 1. Tense > Neg > [Q] > Predicate > Particles > [Q] > Subject > Object > Indirect Object > Oblique > [Q] Figure 1. Niuean Word Order Showing Possible Question Particle Positions

One more point about Niuean grammar will be mentioned here. In forming content questions, a wh-word in Niuean can remain in situ, functioning as a subject, object, indirect object or oblique, or, alternatively, the wh-word, with the addition of the predicate-forming preposition ko, can form the predicate of the sentence, appearing in the position of the predicate (Seiter 1980; Massam 2003). The existence of question particles has been noted in previous work. Research on Niuean states that nakai (glossed herein as Q1) occurs in neutral or non-focused yes-no questions, and that the particle has a relatively fixed post-predicate position (Seiter 1980; Sperlich 1997; Massam 2003). Most researchers concerned with the syntax of Niuean have also remarked that this question particle does not appear in wh-questions (Massam 2003; Seiter 1980; Sperlich 1997), lending support for the claim that nakai is a true polar question particle. Massam (2003) also claims that nakai questions do not appear with a negative predicate. Various researchers have commented on the similarity in form between the question particle nakai and the negative particle n¯akai (Seiter 1980; Sperlich 1997). We hope to return to this issue at a later date. In the literature, a second polar question particle, ka (glossed as Q2 in this paper), can appear post-verbally and in other environments, such as at the end of a sentence. It can co-occur with negation, both after the negative particle and after the verb (Massam 2003). Although ka is mentioned in several works on Niuean grammar (e.g. Sperlich 1997), ka as a question particle is not mentioned in Seiter, except as an abbreviation of kia + proper absolutive case marker a. Other possible related items are ka ‘but’, ka ‘conditional/future’(Sperlich 1997). ka may have a number of forms such as ka ha/ ka h¯a2 and ka: (with a lengthened vowel, indicated here with a colon, distinct from phonemically long vowels indicated with macrons), and might also be the form occuring in collocations such as ka e koe ‘what about’. 2. In our data sources we find both ka ha and occasionally ka h¯a. Since Niuean orthography inconsistently utilizes macrons to mark long vowels, at the present time we cannot clearly determine if these are two different constructions, or the same. We set this issue aside in this paper.

68

Diane Massam, Donna Starks and Ofania Ikiua

A third particle kia (glossed as Q3 in this paper), is said to be able to substitute for nakai in the usual post-predicate position but is also claimed to occur in several additional syntactic positions, including sentence final. Although McEwen (1970: xix) considers kia, ka and ka h¯a to occur only sentence finally, kia may mark a negative question either by appearing after the negative or a verb where it conveys a non-neutral expectation. It can also appear after a content question constituent in information questions (McEwen 1970; Seiter 1980; Sperlich 1997; Massam 2003), in which position its function is unclear. In addition, Sperlich claims that kia is related to the imperative kia and thus imparts the notion that the polar questions are expected to be answered in the positive and demand an answer. Whittaker (1982: 27) notes that, in the latter position [sentence finally], kia may be pronounced as “a throaty k¯a”, and speculates this form might be the result of “either as shortening of kia or kah¯a”. In addition, and possibly in conjunction with the above particles, questions may be marked by the use of intonation. McEwen (1970: xix) claims that questions may be formed by a sharp rise in pitch. Both Whittaker (1982: 27) and Kaulima and Beaumont (2002) state that the tone of the voice rises at the end of a question. Intonation rises appear on polar questions with and without particles. Sperlich (1997: 10) notes that declarative sentences may have final rising intonation. This is mentioned here to highlight the fact that intonation appears to be quite complicated in Niuean and is in need of much more study. Table 1 provides an overview of the available information on polar questions in Niuean. It gives a visual representation of what is known about the function of the question particles. The table points to gaps in our knowledge of the polar question particles, and even greater gaps in our knowledge of intonation questions. It also shows that we know relatively nothing about answers to these questions. This paper examines polar question particles in interview data, a question-rich genre, to see how interviews may help enlighten our understanding of the various functions and uses of Niuean polar questions. An updated version of this table appears at end of this paper.

Methodology The texts and materials in this paper come from spontaneously produced sources, rather than through elicitation. This has both advantages and disadvantages. The former include the authenticity of the material. The data represents native speakers’ language use when interacting with other native speakers in a naturally produced linguistic context. This has the effect of creating rich linguistic resources which might not have been produced under elicitation. This type of data brings

Questions and answers in Niuean

69

Table 1. Polar questions in Niuean: an overview nakai

ka

kia

INTONATION

Information status

neutral/ non-focussed

?

non-neutral expectation

?

Position

post predicate

post predicate/ sentence final

post predicate/ sentence final

sharp final rise

Occurs in content (wh-) questions

No

No

Yes

?

Occurs with NEG

No

Yes

Yes

?

Possible variant forms

–

ke ka: kaha kah¯a

ka(?) k¯a

?

Related forms

n¯akai ‘NEG’

ka ‘but’ ka ‘conditional’ ka ‘future’

kia ‘imperative’

–

ANSWERS

e¯ ‘yes’, n¯akai ‘no’

with it the spontaneous use of discourse particles (Schriffrin 1987) as well as discourse phenomena typically categorised as “performance errors” (but see note 4). These are included here as they represent “real” language use. Some of the hesitation and self-corrections in the data create a body of information on the underlying grammar of these speakers and for this reason, they are not edited out of the final product. While this results in additional reading difficulty for linguists focussing on the data, it also creates a body of potentially interesting information for those who may mine the data for alternative purposes at a later date. The materials are thus less structured than elicited data, and do not contain intuitional judgements. Instead we rely on the frequency of different question particles and the contexts in which they occur to provide information about the function and form of the Niuean particles. The dataset contains structured information, but the structure is different from that typically considered by formal syntacticians. The questions are produced in three different contexts, in two oral interviews and one written interview

70

Diane Massam, Donna Starks and Ofania Ikiua

schedule. As the content and sequencing of the talk is identical in the three genres, differences and similarities in the questions in these contexts provide useful information about the structure of polar questions and their potential answers in these genres. Language documentation has a long tradition of collecting and analysing natural texts for dictionary preparation and in grammatical descriptions. These types of data collection procedures are increasingly becoming the main data collection tool for indigenous languages worldwide. This paper presents a description of polar questions in Niuean from this perspective in the hopes that others will see the advantages of this methodology for the development of their own research agendas. We use as the primary basis of this paper data collected from two oral interviews recorded as part of the Pasifika Languages of Manukau Project (PLMP). The PLMP project focusses on the language maintenance and use of Pasifika Languages in the Manukau region, and interview topics cover life histories, social networks, and language proficiency, use and attitudes. Thirty individuals were interviewed, 15 in the Niuean language. These oral interviews were recorded on a TCM-5000 tape recorder with a lapel microphone, most in the homes of the participants. Two of the recorded interviews are fully transcribed and translated and consist of approximately 200 pages of text. The first interview is just over an hour in duration, the second slightly longer. The interviewer (Q) and her participants (P) were born in Niue but reside in New Zealand. One interview is with an older female (coded as NOFIP), the other with a middle-aged male (coded as NMMIP). The interviewer was in her late twenties at the time of the interviews. All three participants claim Niuean as their primary language. The older female claimed to be monolingual in Niuean. The interviewer used a 27-page written interview schedule to complete the interviews. The written interview schedule consisted of fully scripted sentences and questions in point-format and is available on the project website.3 The two oral interviews are similar to the written schedule but often differ subtly in their structure as the interviewer adapted topics to fit the situation.4 Although at times 3. The Pasifika Languages of Manukau Project website is available at http://www.aut.ac.nz/ research/research institutes/icdc/projects/pasifika languages of manukau/ See also Bell, Davis and Starks (2000) and Starks, Tuhipa, Williams, Ikiua, and Lui-Heka (2003). 4. The interview schedule was prepared initially in English, translated into Niuean [by the interviewer] and then back-translated into English by a second Niuean speaker. The two translators met to discuss differences in the translations and to finalise the Niuean version of the interview schedule. Outright ungrammaticalities (e.g. wrong cases used) were noted as such in the transcriptions, so that, while the data in this paper reflects oral language, it should not contain truly ungrammatical utterances.

Questions and answers in Niuean

71

the translations are not in idiomatic English, they closely reflect the meaning and spoken style of the Niuean sentences and we have not standardized them. In some instances, we rely on supplementary data drawn from other genres, most noteably children’s storybooks. Our final point concerns one of our co-authors. We were fortunate to have the Niuean interviewer for the PLMP project as a co-author. Ofania Ihiua was also responsible for the translation of the interview from English into Niuean. In addition Ofania worked with Diane Massam on some of her syntactic analyses of Niuean where she has commented on the structure of polar questions, and provided indications of the differences between the forms.

Polar questions in Niuean Our interview data was rich in questions in contrast with other text types where questions either did not occur, or did so very infrequently. The structure of this type of genre is limited in that almost all questions are produced by the interviewer. This paper concerns itself only with matrix questions.5 Our data contained 328 examples of polar questions marked by particles, and many others that were based solely on intonation. The intonation-only questions were difficult to classify and count. We comment briefly on these constructions later in the paper. The polar question particle, by far the most frequent, is nakai. As illustrated in Table 2, there were 222 examples of nakai questions in the corpus. ka questions were the next most frequent type, with 85 tokens. kia questions were the least frequent in the database, with only such 20 examples. There were a greater number of nakai questions in the oral interviews. The fact that nakai questions are more frequent in the oral interviews (see table 2) than in the interview schedule is simply an artefact of the interview schedule where questions are presented in the form of one single complex question. This is illustrated by examples (1) and (2) below. Example (1), taken from the interview schedule, contains one nakai question. Example (2) is one part of this same question extracted from one of the oral interviews. In the oral interview, questions such as (2) must be asked repeatedly to cover Niuean, English, and other languages, whereas in the written interview schedule these multiple questions can be formatted as a single question, as in (1). This results in fewer instances of the question particle in the written interview compared to the oral interview. The differing number of nakai questions in the two oral interviews 5. Virtually no research has been done on embedded questions in Niuean, and we do not include them in this paper, in part because there were very few examples in our data.

72

Diane Massam, Donna Starks and Ofania Ikiua

in contrast with each other, is due to the differing number of responses by the two interviewees (e.g. some questions asked about the language proficiency of each family member).6 Table 2. Polar question particles in the oral interviews and written interview schedule nakai

Ka

Kia

TOTAL

Interview I NMMIP

100

50

6

156

Interview II NOFIP

78

35

13

126

Interview schedule

45

0

1

46

TOTAL

222

85

20

328

(1) Ka fakanogonogo a koe ke he tau tagata ne t¯utala (i) faka-Niue, (ii) faka-P¯alagi, (iii) falu a vagahau foki, maeke nakai ia koe ke: i Niue ii P¯alagi iii Falu foki 1 maama e tau mena oti ne vagahau e lautolu 2 maama falu a mena ne vagahau e lautolu 3 maama e tau vala tala mukamuka 4 maama fakatote falu a kupu po ke tau vala tala 5 ai maama ha mena When you listen to people speaking, can you: i Niuean ii Eng 1 2 3 4 5

iii Other

understand everything they say understand most of what they say understand simple conversations understand only a few words or phrases not understand it at all

6. (1) is not glossed, as it is the layout of the question that is important here, not the grammatical structure. In the data glossing, we follow the Leipzig conventions. The abbreviations used are as follows: abs: absolutive, adv: adverb, ben: benefactive, c: common, com: comitative, dem: demonstrative, dir: directional marker, du: dual, emph: emphatic, erg: ergative, excl: exclusive, fut: future, gen: genitive, gl: goal, hab: habitual, incl: inclusive, ins: instrument, loc: locative, lig: ligature, man: manner, neg: negative marker, nfut: nonfuture, p: proper, perf: perfect, pers: personal, pl: plural, pst: past, pred: predicative, pron: pronoun, prt: particle (exact meaning uncertain), q: question, quant: quantifier, sbjv: subjunctive, sg: singular, 3: third person, 2: second person, 1: first person. More information about most of these morphemes can be found in Seiter (1980). We do not use punctuation in the Niuean transcribed spoken language data, but data taken from written sources appears with the punctuation from the source.

73

Questions and answers in Niuean

(2)

Q: liu a taua mogonei ke lata ma e return abs.p 1.du.incl now sbjv right ben abs.c vagahau p¯alagi, ka fakanogonogo a koe ke language English while listen abs.p 2.sg gl.c he tau tagata ne t¯utala fakap¯alagi, maeke loc.c pl person nfut converse English possible nakai a koe ke maama oti e tau mena q1 abs.p 2.sg sbjv understand all abs.c pl thing ne vagahau e lautolu, aa. . . nfut language erg.p 3.pl mmm ‘Again we will now, for the English language, when you listen to the people speaking in the English language, can you understand everything they say? Mm. . . ’ P:

aa. . . maeke ‘Mm. . . yes.’

Nakai questions Nakai is the unmarked question particle in the written interview schedule. There were 45 examples of nakai questions in the interview schedule, one kia question, and no ka questions (see table 2). In our dataset, the nakai question particle “Q1” only occurs in post-predicate position, in the position indicated in Figure 2 below, at the end of the postpredicate particles after directional and manner particles, the instrumental applicative aki, the universal quantifier oti, the locative resumptive pronoun ai, various adverbs and emphatics, and the perfect particle tuai. (See Seiter 1980; Massam 2000, 2001; Massam and Roberge 1996 for more about Niuean predicate structure and these particles.) dir

man

instr

quant

resumptive loc pron ai

adv

emph

perf

q

Figure 2. Order of Post-Predicate Adverbials and Particles

The data in the interviews confirms what is already known in the literature. When there are no other post-verbal particles in the utterance nakai appears after the bare predicate, as in (1) and (2) where it appears after the predicate maeke ‘can/able to’. In more complex structures, as in (3), nakai appears after the complex predicate fai tagata Niue, and in (4) it appears in post-predicate

74

Diane Massam, Donna Starks and Ofania Ikiua

position after the directional particle atu. The particle nakai only appears in sentence final position as in (5) when the predicate phrase is in final position, due to there being no (unincorporated) nominals in the utterance. There are only two pieces of new information about nakai we can draw from the interviews. The first concerns the relative infrequency of constructions such as (5). Most cases of predicate-only nakai structures either involve relatively fixed expressions such as mooli nakai? ‘is that true?’ or occur in tag constructions such as pihia nakai as in (6). The second piece of new information concerns the nature of the predicate, which can be in Niuean as in (1) to (6), or in English, as in (7). (3)

Q: homo, fai tagata Niue nakai ne, he tapu ne good be person Niue q1 nfut loc.c church nfut finatu a koe ki ai come.over abs.p 2.sg gl.p there ‘Good, is there a Niuean that, at the church in which you attend?’ P:

(4)

fai, tokologa be many ‘Yes there is, many.’

Q: f¯a fenoga atu nakai a koe ke he motu hab journey dir2 q1 abs.p 2.sg gl.c loc.c island ko Niue pred Niue ‘Do you usually travel to Niue?’ P:

lavea ‘sometimes’

tau po ke Q: e, f¯a laga fiha, laga taha he yes, hab times how.many times one loc.c year or falu n¯ı a tau some emph lig year ‘Yes, about how many times, about once a year or every few years?’ (5)

Q: e, ti fai mena foki nakai yes so be thing also q1 ‘Yes, so, is there anything else?’ P:

ai fai foki n¯ı ka ha, ko e not be also emph that’s.it pred C ‘There is nothing ay, the . . . ’

Questions and answers in Niuean

(6)

(7)

75

Q: e, ti ko e, hau mogoia [a] koe ki yes, so pred c come at.that.time abs.p 2.sg gl.p hinai, fano mogoia koe ke he aoga he here go at.that.time 2.sg gl.c loc.c school gen.c tau f¯anau ikiiki i hinei, pihia nakai pl children small loc.c here thus q1 ‘Yes, and, you then came here, and then you went to the school for little children here, is that correct?’ P:

yeh, fano n¯ı he aoga ia i hinei he yeh, go emph loc.c school dem gl.p there gen.c primary ka ha primary that’s.it ‘Yeh, just went to that primary school here.’

P:

include nakai ka peh¯e ko e internete haau ne include Q1 if thus pred c internet gen.2.sg nfut totou read ‘does it include if you are reading off the internet?’

Q: n¯akai, ai l¯a hoko a taua ke he neg neg emph up abs.p 1.du.incl gl.p loc.c vala ia piece dem ‘No, we are not up to there yet.’ Our corpus contained few examples of nakai that had other notable characteristics. There were no examples of the nakai particle in wh-questions, and, as also expected, no examples of negative polar interrogatives containing nakai. There is one interesting expression that co-occurs with nakai questions. The expression e po ke n¯akai ‘yes or no’ can occur by itself, but usually occurs as a sentence final tag, as in (8) and (9). In these examples, the question particle nakai and the negative particle n¯akai co-occur, though notably they are not directly modifying the same predicate. (8)

Q: f¯a o¯ mai nakai a lautolu he motu ko Niue hab come q1 abs.p 3.pl gen.c island pred Niue nonofo mo koe, e¯ po ke n¯akai stay com 2.sg yes or no ‘Do people in Niue come and stay with you, yes or no?’

76

(9)

Diane Massam, Donna Starks and Ofania Ikiua

P:

e¯

P:

‘yes’

Q: mitaki, f¯a fano nakai a koe ke he tau good, hab go q1 abs.p 2.sg gl.c loc.c PL fakafetuiaga he tau Niue, e po ke n¯akai gathering gen.c pl Niue yes or no ‘Right, do you go to other Niuean events, yes or no?’ P:

fano ‘Go.’

Ka questions The next most frequent question particle in our corpus is ka. This particle is more difficult to count for a number of reasons, primarily because it may have different forms. When ka is followed by an argument, the case marker is usually not visible. In (11) it seems clear that the case marker is absorbed into the question particle, since the question particle undergoes a phonological change, but in (10) we could consider the case particle to be either absorbed or deleted, since there is no phonological change to the question particle. (10) Ka + a (ka) Q: tala ka koe a ia ke he tau mamatua talk q2 2sg abs.p 3sg gl.c loc.c pl parent ‘Are you talking about the parents?’ P:

tau mamatua e, kamata mai he tau mamatua ke pl parent yes begin dir1 loc.c pl parent gl.c he tau f¯anau hoko ke reach gl.c loc.c pl children ‘The parents, yes, starting from the parents right up to the children.’

Q: e ‘Yes.’

Questions and answers in Niuean

77

(11) Ka + e (ke) Q: pihia oti ke tau S¯a ka hea mai ki a thus all q2 pl Sa(moan) when call dir1 gl.p abs.p koe Ha ha ha 2.sg [laughter] ‘Are all the Sa’s the same when they call you?’ A second difficulty is the number of related forms not the result of simple morpho-phonemic processes, such as ka: and ka ha. These are described later in this section. A third and final issue relates to the number of homophonous and possibly polysemous forms. ka is listed in dictionaries (e.g. McEwen 1970) with various meanings, such as ka ‘future’, ka ‘conditional’, and ka ‘but’, which are frequently occurring forms in the present database. The possible relationship/overlap between ka ‘but’ and ka ‘polar question particle’ is also commented on later in the section. In our oral interviews, ka occurs in expected positions, that is, in postpredicate position as in (12) and sentence final position, as in (13). When there are no nominal arguments, post-predicate position vacuously equates with sentence final position as in (14). Somewhat interestingly, unlike nakai questions, this default position is fairly common. The position of ka is not as fixed as the literature would indicate. In five instances in our corpus, ka occurs after a nonpredicate fragment as in (15). There are also a few instances where it is followed by a discourse particle such as pihia ‘is it so?’as in (16) and in six cases it appears on its own, as illustrated in (17). We have no instances of nakai or kia on its own. (12)

Q: ko e h¯a ne malona ai e na e, kua pred c why nfut broken there prt prt prt perf malona leva ka e long.time q2 abs.c broken ‘Why is this broken? Has it been broken for long?’ P:

(13)

e ‘Yes.’

Q: e, pihia he kitia e koe ka yes, thus loc.c see erg.p 2.sg q2 ‘Yes, is that how you see it?’ P:

mm ‘Yes.’

78

Diane Massam, Donna Starks and Ofania Ikiua

(14)

Q: mooli ka true q2 ‘Is that true?’

(15)

Q: matini hila ka machine electric q2 ‘Electric machine aye?’

(16)

Q: ti peh¯e ka ko e fai vala Niue ka pihia so thus whether pred c be part Niue q2 thus ‘Are you saying that, there is part Niuean?’ po ke peh¯e n¯a haau a manatu ai ko e or sbjv thus emph gen.2sg lig thought neg pred c tau Niue oti a mautolu pl Niue all abs.p 2pl.excl ‘or your thoughts are we are not all Niuean?’

(17)

P:

ti mamate e tau kulene ia haaku ai so die abs.c pl grandparents dem gen.1sg neg kitia e au see erg.p 1sg ‘and died, those grandparents of mine, I did not see them.’

Q: ka ‘Really?’ P:

ia ‘Yes.’

Q: e Q: ‘Right.’ Other points concerning ka questions relate to their function. As indicated in Table 2, ka is the only particle which does not appear in the interview schedule. This may indicate that ka is considered as more appropriate in the spoken register. It is however not the case that ka is restricted to truly oral spoken discourse. Although ka does not appear in the written interview schedule, we have noted examples of both nakai and ka in published children’s storybooks. While the use of the question particle nakai appears to be unmarked and information seeking in this genre (used, for example, in a story where a policeman asks a little girl for information), this is not the case for ka questions. In a story about a man fishing, three ka questions occur, all with the same form, and in all cases, they

Questions and answers in Niuean

79

are used rhetorically, and within quoted speech.7 A child and a father have been fishing, and on the way home they meet a pastor, a matai, and a neighbour, all of whom ask the question in (18). The father gives them all fish, and the father and child end up with hardly any fish left when they get home. A second rhetorical question occurs in (19) where the mother in the story asks, Ko e tau ika n¯ı ka ia ha mua ne h¯ı ? ‘Is this all you caught?’. (18)

“Ko e o¯ mai ka mua i tahi?” pred c go dir2 q2 2du gl.p sea ‘Finished for the day?’

(19)

“Ko e tau ika n¯ı ka ia ha mua ne h¯ı ?” he pred c pl fish emph q2 dem gen 2du nfut catch as vagahau he Matua Fifine mo e malimali. language gen parent woman com c smile “‘Is this all you caught?” asks Mum with a smile.’

There are no examples of rhetorical questions in the interview data, but there are some other observations to be made from the dataset that shed light on the function of ka questions. In the interviews, ka questions usually follow on from other questions (as in 20) and are used to check information as in (20) and (21). (20)

e Q: ka e iloa nakai e koe ka peh¯e ko but lig know q1 erg.p 2sg if thus pred c Tonga Tongan? ‘But would you know if it is a Tongan?’ P:

kehe foki ha Tonga leo different also gen.p Tongan voice ‘Tongans pronunciation is also different’

Q: e, iloa e leo Tonga ka yes know abs.c voice Tongan q2 ‘Yes, you know the Tongan pronunciation aye?’ P:

e iloa e au e leo Tonga. yes know erg.p 1sg abs.c voice Tongan ‘Yes I know the pronunciation of Tongan.’

7. This is from the story Kua Tufa E Oa Ha Maua ‘Sharing Our Catch’ by Tia Aluni Taylor, translated into Niuean by Aiao Kaulima, published for the Ministry of Education by Learning Media Ltd. Wellington (1993).

80 (21)

Diane Massam, Donna Starks and Ofania Ikiua

Q: e, tuga e tau pepa fakapuke ka yes, like abs.c pl paper fill q2 ‘Yes, like the questionnaires aye?’ P:

haia, tuga e tau pepa fakapuke fakaNiue, that’s.it like abs.c pl paper fill Niuean fakap¯alagi Pihia e falu motu he kitia e au English like abs.c some island as see erg.p 1sg ka ha he f¯a totou au he falu pepa ne that’s.it as hab read 1sg loc.c some paper nfut f¯a e tamai. Fai fakamaamaaga tei he hab abs.c buy be explanation prt loc.c fakamotu ha foaia, toe e tau Niue native.language left.out abs.c pl Niuean ‘That’s right, like the fill in papers in Niuean, English. Some countries are like that in what I see when I read the papers that are brought. There is an explanation in their language, none for Niueans.’

When ka appears on its own, as in example (22), it serves as a request for clarification of topics previously introduced into the discourse, or as in the earlier example in (17), it simply serves as an aid to maintain conversational flow. (22)

P:

ko e Niue e matua taane ka e fanau pred c Niuean abs.c parent male but lig born i Ausetalia. loc.p Australia ‘The father is Niuean but born in Australia.’

Q: ka Q: ‘Aye?’ P:

umm Niue matua taane Mauri e matua fifine mmm Niuean parent male Maori abs.c parent female ‘Mmm, father is Niuean and mother is Maori.’

Q: e Q: ‘Right.’ As noted earlier, ka is particularly challenging because of the number of possible variants. In one instance, an extra long ka: occurs as in (23). This form may be an emphatic, perhaps related to the sentence final particle described by Whittaker

Questions and answers in Niuean

81

(1982).8 Since we only have one such example, little more can be deduced from this example. In a few instances, the sequence ka ha ka occurs as in (24), while in other instances ka ha appears to signal a polar question in its own right, as in (25) and (26). In its use as a question particle ka ha can be used in post-predicate position when the predicate is a wh-word. In most contexts however, ka ha ‘that is so’ has no interrogative meaning, as in the second occurrence in (26) and in (27). (23)

Q: e vagahau p¯alagi n¯ı ka ha t¯umau ka: yes speak English emph q2 as always aye ‘Right, speak in English always aye?’

(24)

Q: uafulumahiva e tau haau a ia, ka peh¯e hau twenty-nine lig year gen.2sg then if thus come a koe he tau onogofulu, uafulumahiva e tau abs.p 2sg loc.c year sixty twenty-nine lig year a ia haau ka ha ka kaeke kua fanau abs.p 3sg gen.2sg that’s.it aye if perf born hiva e a koe he tau taha e afe abs.p you loc.c year one lig thousand nine lig teau uafulumahiva, ko e tau haau he hundred twenty-nine pred c year gen.2sg loc.c taha e afe hiva e teau onogofulu kua, one lig thousand nine lig hundred sixty perf uafulumahiva tei twenty-nine prt ‘You were twenty nine years old then, if you came in the year nineteen sixty, you were twenty nine years old then, right? If you were born in the year one thousand nine hundred and twenty nine, your age in the year one thousand nine hundred and sixty, you were twenty nine years old?’ P:

ia, tau tepu ia yes pl area that ‘Yes, about there.’ Q: e ‘Right.’

8. The exact status of this extra long vowel is unclear. It may be simply a case of stylistic lengthening. However, it is also possible that the vowel length is a relic of the existence of a longer original form, perhaps kaha ‘that’s it’ or even kia.

82 (25)

Diane Massam, Donna Starks and Ofania Ikiua

Q: ka e lahi he tau magaaho ko e vagahau but lig most gen.c pl time pred c language Niue Niuean ‘But most of the time speak Niuean.’ P:

ko e lahi n¯ı ko e vagahau Niue pred c most emph pred c language Niuean ‘mostly spoke in Niuean’

Q: ne mukamuka a koe ka ha pst easy abs.p 2sg right ‘Because easy for you aye?’ P:

(26)

yeh ‘Yes.’

Q: e, tino malol¯o a koe ka ha Ha ha ha ha. . . yes body well abs.p 2sg right [laughter] ‘Yes, your body is well aye? Ha ha ha. . . ’ P:

malol¯o na au ne nofo ai fano he aoga well pst 1sg nfut stay not go loc.c school ia a abs.p dem ‘Well I was the one that stayed home and did not go to school then.’

P:

mo e tau gahua lima ha mautolu, tau and lig pl work hand gen.c. 1pl.excl pl community ka ha community right ‘And our handcrafts, the communities aye.’

Q: e ‘Right.’ (27)

Q: aa ka e heigoa e gahua haau mm but lig what abs.c work gen.2sg ‘Mm what is your job?’

Questions and answers in Niuean

P:

83

ko e leveki tagata gahua, tuga ko e pule gahua pred c care.for person work like pred c supervisor ka ha right ‘Take care of working people, like a supervisor.’

There are a few other collocations with ka. ka appears clause initially in expressions such as ka i ko ‘what about’, as in (28). The collocation in Whittaker (1982) ka e kua ‘And how about?’ does not occur in our database. (28)

Q: ka i ko e haau a tama what.about C gen.2sg lig child ‘What about your child?’ P:

fakap¯alagi fakaNiue ‘English Niuean’ Q: e ‘Yes.’ The latter use of ka may be related to the clausal conjunction ‘but’ which can appear as a discourse marker, where it can co-occur with nakai, as in (29). In other cases ka ‘but’ occurs without a question particle as in (30) where it has the illocutionary force of a question [repeated in part from (25) above]. We have not examined this connection in enough detail to comment on it further. We have also not examined the prosodic structure of utterances such as these examples to determine the possible effects of intonation or pitch on the illocutionary force of these utterances. (29)

Q: ka e iloa nakai e koe ka peh¯e ko e but lig know q1 erg.p 2sg if thus pred c Tonga Tonga ‘But would you know if it is a Tongan?’

(30)

tau magaaho ko e vagahau Q: ka e lahi he pred c language but lig most gen.c pl time Niue Niue ‘But most of the time speak Niuean.’ P:

ko e lahi n¯ı ko e vagahau Niue pred c most emph pred c language Niue ‘Mostly spoke in Niuean.’

84

Diane Massam, Donna Starks and Ofania Ikiua

Kia questions Our third question particle is kia (glossed as Q3). In all of our examples in the corpus, kia occurs in post-predicate position as in (31) and (32), although again, if the sentence consists only of the predicate, kia occurs vacuously in sentence final position, as in (33). Kia is rare in the latter contexts. This distributional difference provides one source of evidence that ka and kia are distinct particles. In our dataset, there are no sentence final kia questions and no negative questions with kia even though the interviewer produced these structures in elicitation, and such occurrences are noted to occur in the literature. We did have instances of kia in sentences which also contain wh-words, (see 34), but examples of kia directly after a wh-predicate were not found in our data, although they have been accepted in elicitation (Massam 2003). (31)

Q: e, ti ka liogi a koe, ka liogi tokotaha yes, so when pray abs.p 2sg when pray alone a koe, ko e ko e vagahau Niue t¯umau kia abs.p 2sg pred c pred c language Niue always q3 a koe abs.p 2sg ‘Right, so when you pray, when you pray alone, do, do you speak always in Niuean.’ P:

(32)

vagahau. . . vagahau Niue n¯a au pihia n¯a au speak speak Niue emph 1sg thus emph 1sg liogi a mautolu ka kai he laulau ka when pray abs.p 1pl.excl when eat loc.c table ka nonofo auloa, ko e fakaNiue n¯a au when stay together pred lig Niuean.way emph 1sg ‘Speak. . . I just speak Niuean and the same when we pray when we eat on the table, when we sit together, I just speak Niuean.’

Q: manatu tuga e tau manamanatuaga hake mai thought like abs.c pl thoughts up dir1 he loto haau ka ha tau manamanatuaga hokulo loc.c inside gen.2sg right pl thought deep haau mukamuka kia a koe ke fakakite e gen.2sg easy q3 abs.p 2sg sbjv show abs.c tau manatu ia haau ke he vagahau Niue pl thought dem gen.2sg gl.c loc.c language Niue

Questions and answers in Niuean

85

‘Thoughts like thoughts that come up from inside you. Your deep thoughts. Is it easier for you to show those thoughts in Niuean?’ P:

(33)

mukamuka atu ka fakakite mai ke he vagahau easy more if show dir1 gl.c loc.c language Niue Niue ‘Easier if shown in Niuean.’

Q: haia ti ko e motu f¯e e tupuaga haau right so pred c island which abs.c birth gen.2sg ko e motu ko Niue kia pred c island pred Niue q3 ‘Right, so which country/island were you born in? Was it Niue so?’ P:

ko e motu ko Niue. Ko e, Ko e matua pred c island pred Niue pred C pred c parent fifine haaku fanau i Samoa female gen.1sg born loc.p Samoa ‘Island of Niue. My, my mother was born in Samoa.’

There are some differences across the interviews. Although the interview with the middle-aged male was longer and overall contained more questions than the interview with the older female, the interviewer produced more kia questions in the interview with the older female. There were 16 examples of kia questions from the interview with the older female, 8 in the interview with the middle-aged male. The greater number of kia questions in the interview with the older female may be due in part to the interviewer’s perception (based on her comments in an elicitation task) that such forms are “are used more commonly” by older speakers. There are also differences across the genres. There is only one kia question in the written interview schedule. The topic of the kia question in the written interview schedule concerned language death. It was a complex question, represented as two separate questions in the oral interviews. The interviewer used kia in this question in both oral interviews. It is interesting to speculate whether the relative importance of the language death question in an interview on language maintenance may have resulted in the use of kia rather than nakai given that, as Sperlich (1997) notes, in other contexts kia is an imperative marker.

86 (34)

Diane Massam, Donna Starks and Ofania Ikiua

F¯ef¯e haau a manatu, kua tei what gen.2sg lig thought perf almost vagahau Niue language Niue i. i Niu Silani E ii. i Niue E Ko e ha?

galo kia e lost q3 abs.c

N¯akai N¯akai

Do you think the Niuean language is in danger of being lost? No i. In New Zealand Yes ii. In Niue Yes No There are a number of features in the kia questions which support the claim that it is somehow marked. In a quarter of our examples, kia is preceded by a predicate with the predicate focus particle n¯ı, as in (35). This suggests that, unlike nakai questions, kia questions may have a focus component, again perhaps related to its use as an imperative marker. (35)

Q: e, kua vagahau p¯alagi tei a ia. Ka ko e yes perf speak English almost then But pred c tapu Niue ne finatu a koe ki ai. . . ka church Niue nfut attend abs.p 2sg gl.p there if peh¯e ko e magaaho n¯ı he tau Niue ko e thus pred C time emph loc.c pl Niue pred c heigoa e vagahau ne fakaaoga e lautolu ke what lig language nfut use erg.p 3pl sbjv tapu vagahau Niue fakatutala ka hili e talk when finish abs.c church language Niue n¯ı kia emph q3 ‘Yes, then speak English then. But the Niue church that you attend. . . if it is only the time for the Niueans, what is the language they use to speak when the church is finished? Just speak in Niuean only?’ P:

vagahau Niue, vagahau Niue ‘Speak Niuean speak Niuean?’

Q: vagahau Niue t¯umau ‘Speak Niuean always?’

Questions and answers in Niuean

P:

87

ia, ko e tau youth n¯a ha mautolu ne yes pred c pl youth dem gen 1pl.excl nfut fakap¯alagi speak.English ‘Yes, just our youth that speak in English.’

Q: fakap¯alagi e Q: ‘Speak English, yes.’ Kia occurs when the interviewer appears to be trying to seek additional information, as in the example below. It appears to have a more information seeking purpose than the confirmation function typical of ka questions, described below. (36)

P:

o¯ hui n¯ı ti liliu mai go foot emph then turn dir1 ‘Just walked and come back.’

Q: mamao koa ka ha far indeed right ‘Very far though!’ P:

pihia mautolu he maaga ko Liku ka h¯a he thus 1pl.excl loc.c village pred Liku right gen.c falu tama ne o¯ ki Hakupu some child nfut go gl.p Hakupu ‘Like us in the village of Liku, of some children that went to Hakupu.’

Q: e ‘Right.’ P:

ki Hakupu ‘To Hakupu.’

Q: e ‘Right.’ Q: o¯ hui n¯ı kia go foot emph q3 ‘Just walked?’

88

Diane Massam, Donna Starks and Ofania Ikiua

P:

o¯ hui n¯ı he mogo pogipogi to liliu mai go foot emph loc.c time morning fut return dir1 he mogo afiafi ka hili e aoga pihia loc.c time evening when finish abs.c school thus n¯a pogi emph tomorrow ‘Go by foot in the morning time and come back in the evening time after school, same tomorrow.’

We might also add that kia sometimes gives the feel of a leading question, as in (37). (37)

Q: fef¯e haau a kitiaaga ki a lautolu ia what gen.2sg lig view gl.p pers 3pl dem he magaaho nei kua kua pihia kia haau a loc.c time dem perf perf thus q3 gen.2sg lig manatu ha kua kitekite atu a koe ke thought because perf look dir3 abs.p 2sg gl.c he tau f¯anau he magaaho nei, n¯akai loc.c pl children gen.c time dem neg makaukau lahi a lautolu ka ha ke peh¯e fakaako determined very abs.p 3pl right sbjv thus learn e vagahau Niue po ke fakaaoga e vagahau abs.c language Niue or sbjv use abs.c language Niue Niue ‘What is your view about them at this time, do you feel that way because you look at the children these days, they are not very determined to say, learn the Niue language or use the Niue language?’ P:

mooli lahi, mooli lahi. . . true very true very ‘Very true, very true. . . ’

This claim is further supported by the interviewer in her use of final openended information seeking devises such as po ke in an attempt to elicit further information or alternative suggestions, as in (38) and (39). This may be partial support for Sperlich’s comment that such questions insist upon an answer.

Questions and answers in Niuean

(38)

89

Q: e, ti ko e vagahau p¯alagi n¯ı kia t¯umau yes so pred c language English emph q3 always po ke. . . or ‘Right, so speak always in English or. . . ’ P:

vagahau p¯alagi t¯umau n¯ı ka ha ha ko e speak English always emph right because pred c mena ai maeke au ke vagahau fakaLalotonga, thing not possible 1sg sbjv speak Raratongan ai maeke a ia ke vagahau fakaNiue not possible abs.p 3sg sbjv speak Niuean ‘Speak only in English because I am not able to speak in Cook Island Maori, he cannot speak in Niuean.’

Q: e mitaki ‘Right, ok.’ (39)

he tau magaaho haau ka ita Q: fakamua ke first gl.c loc.c pl time gen.2sg if angry ai a koe ko e vagahau Niue t¯umau kia then abs.p 2sg pred c speak Niue always q3 a koe po ke, po ke tatai n¯ı e vagahau abs.p 2sg or or equal emph abs.c language Niue haau mo e vagahau p¯alagi Niue gen.2sg com c language English ‘Firstly, at times when you are angry, do you speak in always Niuean or, or is your use of Niuean the same as English?’ P:

vagahau Niue n¯ı falu magaaho, vagahau p¯alagi speak Niue emph some time speak English falu magaaho some time ‘Speak in Niuean only sometimes, speak in English sometimes.’

Another potential related collocation is between kia and po ke n¯akai, as in (40) and (41). This same collocation appears with nakai questions, but this never appears in ka questions.

90

Diane Massam, Donna Starks and Ofania Ikiua

(40)

Q: ka i ko he motu ko Niue tei galo kia what.about loc.c island pred Niue almost lost q3 e vagahau Niue he motu ko Niue, e po abs.c language Niue loc.c island pred Niue yes or ke n¯akai sbjv no ‘What about the island of Niue? Is the Niue language almost lost, yes or no?’

(41)

Q: e, mitaki kitekite atu ke he vagahau Niue yes good look dir2 gl.c loc.c language Niue ke he magaaho nei, f¯ef¯e haau a manatu gl.c loc.c time dem what gen.2sg lig thought kua tei galo kia e vagahau Niue, e po ke perf almost lost q3 abs.c language Niue, yes or Niu Silani n¯akai He motu nei ko no loc.c island dem pred New Zealand ‘Yes, looking at the Niue language at this time, what do you think, is the Niue language almost lost? Yes or no? In this country of New Zealand?’ P:

e na haaku a mena ne kua toka ki yes emph gen.1sg lig thing nfut perf leave gl.p ai ha ko e tau f¯anau mogonei kua there because pred c pl children now perf uka ke vagahau Niue lahi ke vagahau p¯alagi difficult sbjv speak Niue very sbj speak English ha ko e tau mamatua n¯ı ka ha ne ai because pred c pl parent emph right pst not fakaako e tau f¯anau he tau kaina ke teach abs.c pl children loc.c pl home sbjv vagahau Niue speak Niue ‘Yes is what I leave that with, because the children nowadays, it is difficult to speak in Niuean, mostly speak in English, because the parents do not teach the children in the home how to speak in Niuean.’

Questions and answers in Niuean

91

Intonation questions In all of the above question types it is possible that intonation is used to reinforce the status of the utterance as a question. Polar questions often appear without any question particles and many examples, such as (42), appear in the oral interviews. Intonation questions are difficult to analyse without a detailed acoustic analysis of the recordings, and we have not made any attempt to do so. Informal impressionistic analysis of selected polar questions with and without particles suggests a high degree of variability in the degree and type of intonation in such questions. It is possible, as in Bauer’s (1997) description of NZ M¯aori, that there may be many different types of intonation used in polar questions. The variability suggests that this issue needs to be explored in much greater detail (as in Gunlogson 2001 for English). (42)

Q: nofomau stay.permanent ‘Stayed for good?’ P:

nofomau, haia ‘Stayed for good, that’s right.’

There is one possible example of an intonation only question in the written interview schedule. It occurs in the final section on proficiency where three prior sections had a nakai particle as in example (1). It is possible that the question in (43) and others in the oral interviews are simply due to ellipsis. (43)

Ka tohi a koe (i) fakaNiue (ii) fakaP¯alagi maeke when read abs.p 2sg (i)Niuean (ii)English possible i a koe k¯e. . . loc.p pers 2sg sbjv ‘When you read (i) in Niuean (ii) in English, can you:. . . ?’

In the interviews, intonation-only questions tend to be most like ka questions. They tend to follow on from other questions, as in (44). As in ka questions, an intonation question is often used to confirm information, also seen in (44). (44)

Q: fakamua, i i, he fano he aoga i first loc.p loc.p when go loc.c school loc.p Matala, i Niue Matala loc.p Niue ‘Firstly, in in, when you went to the school of Matala, in Niue?’

92

Diane Massam, Donna Starks and Ofania Ikiua

P:

oi vagahau Niue n¯ı oh speak Niue emph ‘Oh, just spoke in Niuean all the time.’

Q: vagahau Niue t¯umau speak Niue always ‘Speak in Niue all the time?’ P:

mm vagahau Niue t¯umau n¯ı mm speak Niue always emph ‘Mm, just spoke Niuean all the time.’

Answers and responses in Niuean Questions tend to be followed by answers, yet little has been written on answers to questions in Niuean. Our study shows that there are several types of answers possible, and that there is some correlation between the particle used in a question and the type of answer the question receives, as we elaborate below. In our data most questions, both wh- and polar, occur as three part questionanswer-response (Q-A-R) sequences as in (45) or more complicated sequences with embedded structures such as Q-A-[Q-A]-R in (46). In both cases, the third part of the sequence, the response, is e. The latter response particle may appear alone as in (45) or as the first part of a new turn as in (47).9 (45)

Q: vagahau Niue n¯ı t¯umau ka speak Niue emph always q2 ‘Was it always Niue Language?’ P:

vagahau Niue t¯umau speak Niue always ‘Always Niue Language.’

Q: e ‘Yes.’

9. The particle e also occurs as minimal feedback in extended turns and thus, is very frequent in interviews.

Questions and answers in Niuean

(46)

93

Q: ka i ko e laulahi n¯ı ka he tau magaaho what about c mostly emph q2 loc.c pl time ia ke feleveia a koe mo lautolu e dem sbjv meet abs.p 2sg com 3pl lig mamatua motua atu ka ha ia ko e pihia nakai? parent old more right dem pred c thus q1 ‘what about, mostly at that time you meet with those parents older than you, is that right?’ P:

yeah ia motua au ia lautolu yeah, yeh old 1sg loc.p.pers 3pl ‘Yeah, yeh I’m older than them.’

Q: ka, ha ha ha fai nakai ne motua atu ia when [laughter] be q1 nfut old more loc.p.pers koe you ‘Mm, ha ha ha, is there any that are older than you?’ P:

tapu ne motua atu fai a mautolu he more be abs.p 1pl.excl loc.c church nfut old ia au ka ha kua fitulima fitufitu loc.p.pers 1sg right perf seventy.five seventy.seven ‘There are some of us at church that are older than myself that is seventy-five, seventy-seven.’

Q: e ‘Right.’ P:

(47)

ia ‘Yes.’

Q: to fakafano nakai e koe e tau f¯anau fut send q1 erg.p 2sg abs.c pl children haau ke he aoga pihia gen.1sg gl.c loc.c school thus ‘Would you send your children to that type of school?’ P:

fakafano ‘Send.’

94

Diane Massam, Donna Starks and Ofania Ikiua

Q: e, ko e h¯a ne pihia ai haau a yes pred c why nfut thus then gen.2sg lig manatu thought ‘Yes, why do you think that way?’ McEwen (1970) is one of the few researchers to comment on answers. He states that the answer to an interrogative is either e¯ ‘yes’ or n¯akai ‘no’. Our data shows a fuller array of options. Affirmative answers may begin with either a particle e ‘yes’ as in (48–49)10 , a minimal response a ‘mhmm’ as in (50), or another affirmative particle such as ia in (51). A question with a negative answer has slightly fewer available options. Answers may begin with either n¯akai as in (52) or a¯ kai as in (53). Another possible answer is lavea ‘sometimes’ as in (54). (48)

Q: mitaki, to fakafano nakai e koe e tau good fut send q1 erg.p 2sg abs.c pl f¯anau ke he aoga tokoluga ne taute e children gl.c loc.c school high nfut do abs.c tau fakaholoaga fakaakoaga ke he vagahau Niue pl program teaching gl.c loc.c language Niue ‘Right, would you send the children to the High School that runs programmes of teaching in Niuean?’ P:

(49)

Q: Niue n¯ı ka Niue emph q2 ‘Was it only Niue?’ P:

(50)

e ‘Yes.’

e ‘Yes.’

koe Q: fai aoga nakai ke ua aki ne finatu a be school q1 sbjv two ins nfut attend abs.p 2sg ki ai he aoga tokoluga gl.p there loc.c school high ‘Was there a second school that you attended at High School?’ P:

a ‘Mm.’

10. Note the response is often with a short vowel.

Questions and answers in Niuean

(51)

Q: mooli ka true q2 ‘Is that true?’ P:

(52)

n¯akai n¯akai ko e. . . ‘No!, no!, that’s. . . ’

Q: e ai pihia a koe ka yes neg thus abs.p 2sg q2 ‘Yes, you are not like that aye?’ P:

(54)

ia ‘Yes.’

Q: pihia nakai thus q1 ‘Is that right?’ P:

(53)

95

a a¯ kai ‘Mm, no.’

Q: haia. manako foki a mautolu ke h¯uh¯u atu right want also abs.p 1.pl.excl sbjv ask dir2 hagaao ia ke he tau fenoga o¯ mai mo e o¯ about dem gl.c loc.c pl journey from and lig go koe ke Niue. f¯a fenoga atu nakai a ki abs.p 2sg gl.c gl.p Niue. hab journey dir2 q1 he motu ko Niue loc.c island pred Niue ‘Right. We also want to ask you about the travellers arriving and travelling to Niue. Do you usually travel to Niue?’ P:

lavea ‘Sometimes.’

Negative questions are rare in the present corpus. Based on our limited data, it appears that when the speaker disagrees with the negative proposition, the answer is n¯akai or more commonly a¯ kai as in (53); when the speaker agrees, the answer is e, as in (55). This point has been noted in McEwen (1972). When a speaker wants to correct a previous response to a nakai question, as in (56) where the interviewer initially responds e ‘yes’, the particle naha ‘that is not

96

Diane Massam, Donna Starks and Ofania Ikiua

correct?’ may be used. This word is notable because it is not recorded in any of the Niuean dictionaries, or noted in grammars of the language. (55)

(56)

Q: ai lata n¯ı ka ha ke vagahau fakaNiue po not should emph right sbjv language Niuean or ke vagahau fakap¯alagi n¯ı hokoia ka sbjv language English emph only q2 ‘should not learn only the Niue language or the English language only aye?’ P:

e ‘Yes.’

P:

maama nakai understand q1 ‘Understood?’

Q: e, naha ko e vala fakamui ia haau, ui yes no pred c part last dem gen.2sg say a koe fakauaua a koe ke vagahau Niue abs.p 2sg hesitate abs.p 2sg sbjv language Niue ha ko e mena t¯utala p¯alagi, ka koe, ko because pred c thing converse English but 2sg pred e lata ka koe ke peh¯e, fakauaua a koe ke c should if 2sg sbjv thus hesitate abs.p 2sg sbjv fakaaoga. . . use. . . ‘Yes, no it is that last bit of yours, you say you hesitate to use the Niue language because you speak in English, should you say, you hesitate to use the. . . ’ As illustrated in (56) above, and (57–58) below, an answer may contain both a particle and a predicate or full sentence. (57)

Q: e, iloa e leo Tonga ka yes know abs.c sound Tonga q2 ‘Yes, you know the Tongan pronunciation aye?’ P:

e iloa e au e leo Tonga yes know erg.p 1sg abs.c sound Tonga ‘Yes I know the pronunciation of Tongan.’

Questions and answers in Niuean

(58)

R:

97

fakamolemole l¯a ko e tala koe ke he totou please emph pred c talk 2sg gl.c loc.c read he tohitala include nakai ka peh¯e ko e gen newspaper include q1 if thus pred c internete haau ne totou internet gen.2sg nfut read ‘Please, are you talking about reading newspaper, does it include if you are reading off the internet?’

Q: n¯akai, ai l¯a hoko a taua ke he no not emph up abs.p 1du.incl gl.c loc.c vala ia place dem ‘No, we are not up to there yet.’ In other instances, the answer may simply be a predicate or sentence with no particle, as in (59–61) and elsewhere in the paper. (59)

talafakap¯alagi a koe manako nakai Q: fakamua ka first when speak.English abs.p 2sg want q1 tau kupu ne a koe ke vagahau oti e all abs.c pl word nfut abs.p 2sg sbjv speak manako a koe ke vagahau want abs.p 2sg sbjv speak ‘Firstly, when you speak English, can you say anything you want to say?’ P:

(60)

maeke ‘Possible.’

Q: aa. . . maeke nakai a koe ke maama e mm possible q1 abs.p 2sg sbjv understand abs.c tau mena ne tau mena oti ne vagahau e pl thing dem pl thing all nfut language erg.p lautolu 3pl ‘Mm. . . can you understand the things that, everything they say?’ P:

ai maeke not possible ‘Can’t.’

98 (61)

Diane Massam, Donna Starks and Ofania Ikiua

Q: oih. . . tote lahi ka oh small very q2 ‘Oh, . . . very small aye?’ P:

tote lahi ‘Very small.’

When the answer to a polar question contains a negative predicate, the predicate is negated with the negative particle ai as in (61). Although it is possible for the particle n¯akai to negate a predicate as in (62), there are no examples in the corpus where the particle n¯akai occurs in answer to a polar question. This may be a simple omission in the dataset, or may be of more interest. Further data would be needed to determine this. (62)

Q: e mitaki. vagahau p¯alagi nakai a koe ka yes good speak English q1 abs.p 2sg when ha ha i ai e tagata Niue ne n¯akai be.located gl.p there abs.c person Niue nfut not maama e vagahau p¯alagi understand abs.c language English ‘Yes, good. Do you speak English if there is a Niuean that does not understand English?’

Table 3 presents a quantitative summary of the various answers. As illustrated in Table 3, simple e or n¯akai is a dispreferred answer. The most common way of answering a polar question is to use a predicate or sentence, either by itself (49.1%), or together with a polar question particle (20.1%). When a particle occurs with a predicate or sentence, it may precede the predicate or sentence (20 examples), or more commonly, follow it (36 cases). We have one instance, where the particle both precedes and follows. The polar answer particles occur on their own in only 11.7% of all cases, not much more frequently than other particles such as ia ‘yeah’, haia ‘that’s it’ (at 9.5%)11 . We now turn to the answers to the different question types. There are 178 nakai questions, of which 156 receive audible answers.12 The most common answer to a nakai question is one that contains a predicate or sentence. Of the answers, 103 are simple predicates, a further 29 consist of a predicate or sentence 11. The non audible responses in Table 3 refer to cases when the interviewer did not wait for a response from the participants and asked another question. Other answers refer to instances where the participant answered with a question. 12. In several instances, it is likely that the answer was non-verbal as the interviewer did not rephrase her question, but instead continued on with the interview.

99

Questions and answers in Niuean Table 3. Distribution of answers to polar questions in the two interviews Number

% TOTAL

139

49.1

e/n¯akai alone

33

11.7

other particles

27

9.5

e/n¯akai predicate combination

57

20.1

no audible response

24

8.5

3

1.1

predicate/sentence alone

other answers TOTAL

283

100

plus a particle e or n¯akai. These two types of answers represent 74.2 per cent of the data, see Table 4. The remaining answers are simple particles, but notably none are ia or haia. Since the latter are confirmation particles, it is of interest that they do not occur in response to nakai questions, demonstrating that these questions are more information-seeking than confirmational. Table 4. Percentage of response types by polar question particle Nakai [n=178] predicate/sentence alone

Ka [n=85]

Kia [n=20]

57.9

23.5

80.0

e/n¯akai alone

8.5

21.2

0

other particles

5.6

18.8

5.0

e/n¯akai predicate combination

16.3

29.4

15.0

no audible response

11.8

3.5

0

0

3.5

0

other answers

Answers to ka questions are varied. Of all the question particles, they are the most likely to elicit a simple response, either e/n¯akai (21.2%) or some other particle such as ia (18.8%), as illustrated in (63). They are also the most likely question type to use e/n¯akai, either alone (21.2%) or together with a predicate or sentence (29.4%) as in (64), and the least likely to answer only with a predicate or sentence (23.5%). Although we did not examine length of response in any detail, predicate answers appear much shorter in ka questions, including less information repeated from the question, and less elaboration. Simple particle responses are particularly frequent when ka occurs as a completed utterance (see 17) or in a tag question, as in pihia ka?. Although these answers provide

100

Diane Massam, Donna Starks and Ofania Ikiua

some support for an analysis of ka questions as confirmation seeking rather than information seeking, the distribution of answer types is not as definitive as we would have hoped.13 (63)

Q: mooli ka true q2 ‘Is that true?’ P:

(64)

ia ‘Yes.’

Q: kua leva te nofo mai ki hinei ka perf long.time prt stay dir1 gl.p here q2 ‘Lived here for a long time aye?’ P:

e, haia, kua leva te nofo mai i yes that’s.right perf long.time prt stay dir1 loc.p Niu Silani nei ka ha kua motua tei New Zealand dem right perf old now ‘Yes, that’s right, have lived here in New Zealand for a long time and old now.’

The answers to kia questions are more systematic. None of our kia questions are answered with a simple e or n¯akai. There were three instances in which the answer to a kia question contains e or n¯akai, and in all cases, the particle occurs after the predicate as in (65). There was also one answer with the discourse particle haia ‘that’s it’. The most common response is a predicate or sentence (80%). The answer to a kia question is generally quite detailed with (66) being one of the shorter responses. This may be due to kia imparting a sense of obligation on the part of the speaker to provide a justification for their response. (65)

kia a ia Q: e ko e vagahau p¯alagi n¯ı English emph q3 abs.p 3sg yes pred c speak t¯umau always ‘Yes, does he mostly speak English all the time?’

13. We checked to see if ka questions could be answered with a negative. Although there were far more negative responses to nakai questions than ka questions, there were a few examples, particularly with ka ha.

Questions and answers in Niuean

P:

(66)

101

aa. . . kehekehe lahi he tau magaaho e mm different very loc.c pl time yes ‘Mm. . . different at times, yes’

Q: e ti ka liogi a koe, ka liogi tokotaha yes so when pray abs.c 2sg when pray alone a koe ko e ko e vagahau Niue t¯umau kia abs.p 2sg pred c pred c language Niue always q3 a koe abs.p 2sg ‘Right, so when you pray, when you pray alone, do, do you speak always in Niuean?’ P:

vagahau. . . vagahau Niue n¯a au pihia n¯a au speak speak Niue emph 1sg thus emph 1sg laulau ka liogi a mautolu ka kai he when pray abs.p 1pl.excl when eat erg.c table ka nonofo auloa ko e fakaNiue n¯a au when stay together pred abs.c Niuean emph 1sg ‘Speak. . . I just speak Niuean and the same when we pray when we eat on the table when we sit together, I just speak Niuean.’

The correlation of answers to questions would point to the predominant use of ka in true yes-no questions, in that questions with ka are more often answered with a yes or no response, while nakai and kia questions would line up differently. However, given the fact that the correlations between question type and answer type are not rigid, it would seem premature to label the particles according to answer type.

Conclusion The paper examines questions, answers and responses in Niuean, focussing on polar questions (broadly construed), which contain a polar question particle. The results are based on a corpus of interview data which contains relatively large numbers of questions, a rich source of data for questions and answers. On occasion, we also supplement our data with other datasets. The findings confirm what is already known about polar questions, and build on this. Our findings confirm that nakai questions are unmarked, that the question particle occurs in post-predicate position, and that nakai does not occur in wh-

102

Diane Massam, Donna Starks and Ofania Ikiua

questions or with negation. Our findings confirm known facts about ka and kia questions. The ka and kia particles, for example, occur in sentence final and post-predicate position. Our findings also question some prior assumptions, and provide new insights. In particular, we see that although Seiter (1980) considers Niuean to have only two polar questions particles, there are in fact three distinct particles. Seiter does not mention ka, possibly because he considered it to be a variant of kia, or perhaps alternatively simply a variant of ka ‘but’, or another of the many homophonous or near homophonous forms. Such assumptions are problematic on a number of fronts. ka ‘Q2’ is more flexible in its syntactic positions than kia, the two questions have different functions and different types of answers. ka ‘but’ and ka ‘Q2’ occur in different positions, with only ka ‘Q2’ appearing after the predicate. The interviews provide some of the first information on the use and information status of the various question particles. nakai questions are unmarked, while ka and kia questions are less neutral in their use, and the interview data documents how these three may be distinguished. Our interview data shows that the polar question particles occur in different frequency and in different genres. nakai questions were the most frequent type of polar question, and the preferred question type in the written questionnaire. ka questions did not occur in the written questionnaire, and may be more typical of spoken Niuean, although their use may not be absolutely restricted to this genre. kia questions were rare in both our written and spoken genres. The current analysis indicates considerable and interesting overlaps between related forms. Although we know that n¯akai ‘Neg’ does not occur in a nakai question, our data points to one other possible restriction. n¯akai is dispreferred in a negative predicate in answer to a polar question. Such findings highlight a relationship between negative and question particles. This relationship has been observed for some other languages (e.g. Qu´ebec French and Gascon), which also rule out or restrict such co-occurences (Vinet 2001; Morin 2006, 2009; Han and Romero 2001). The overlap between ka Q2 and ka ‘but’ is also of interest. Utterances with ka ‘but’ have the illocutionary force of a question in some instances, but not in others and intonation may also play a role here. As there is no question particle ka in other Polynesian languages, it is possible that the question particle ka may be derived from a discourse particle. The overlap between kia ‘imperative’ and kia Q3 is also of interest, as many of the kia questions in our database appear to have greater force than the other question types, a relationship first observed by Sperlich (1997). Our findings on answers add to the available knowledge on Niuean. In our interviews, we found that questions are followed by answers and responses. The response is e. The particle translates roughly as ‘yes’, but answers to polar

Questions and answers in Niuean

103

Table 5. Polar question in Niuean: a revised overview nakai

ka

kia

INTONATION

Information status

neutral/nonfocussed

confirmation seeking or rhetorical

non-neutral expectation

confirmation seeking

Genre

written & spoken

spoken, limited written

spoken NumP > AP > NP

In accordance with the Linear Correspondence Axiom (LCA) of Kayne (1994), Cinque (2005) adopts the (37) hierarchical ordering, placing the modifying XPs of (37) in Specs on left branches but with successive complement constituents as right-branches. In the terms of Cinque’s account the surface mirror-image ordering (36b) is derived through iterative phrasal movement. In an alternative approach to the placement of Specs, Abels and Neeleman (2006) also propose

188

Elizabeth Pearce

right-branching placement for complements but with the parametric possibility of the Specs occurring variably on right or left branches. These two alternative aproaches to left/right Spec placement in the base structure for the mirror-image ordering are shown schematically in (38a) and (38b). (38)

Mirror-image ordering: Two approaches a. Specs on the left plus iterative phrasal movement (Cinque 2005) XP DemP

YP NumP

ZP AP

NP

The implementation of the iterative movement in the derivation of the mirrorimage surface ordering (36b) from (38a) is demonstrated schematically in (i)– (iv). (i)

AP

NP

Merge of AP above NP followed by raising of [NP]

[ NP AP ]

Merge of NumP followed by raising of [NP AP]

[ NP AP NumP ]

Merge of DemP followed by raising of [NP AP NumP]

[ AP NumP DemP ]

Surface ordering

6

(ii)

NumP 6

(iii)

DemP 6

(iv)

NP

In the case of the (35c) surface ordering, the derivation starts out as for (i)-(ii) and then proceeds as in (v). (v)

DemP 6

[ [ NP AP ] NumP ]

Merge of DemP followed by raising of [ NP AP ]

Two or three things in the Unua noun phrase

189

The partial phrasal movement option selected in (v) is marked in that it fails to Pied-pipe the whole content of the larger phrase, stranding NumP in the lower part of the structure. b.

Specs on the right and no movement (Abels and Neeleman 2006) XP DemP YP NumP ZP AP NP

For Abels and Neeleman (2006), there is no movement in the derivation of the (36b) mirror-image surface ordering and, in the derivation of the (35c) ordering, it is the ZP of (38b) which raises to the higher position above XP. It is beyond the purposes of the present paper to argue as to which of (38a) and (38b) is to be preferred.14 But whichever one of these approaches is adopted (and, similarly, with respect to other approaches15 ) we will focus here on the failure of xeru ‘two’ to be the occupant of the NumP. In the discussion to follow, I will assume the correctness of the hierarchical ordering common to (38a) and (38b), abstracting away from whether the Specs are on the left or the right in the base structure. The focus will therefore be on the compatibility of the content of successive Specs. 4.3. Demonstratives and cardinal numerals To begin with, we restrict our attention to what I have inferred is the conservative grammar, lacking the [ØNumber] possibility with deictic demonstratives. With only [+plural] and [plural] available, our analysis of the data suggests that strict matching in this grammar of Number features between numerals and demonstratives presents the following schema:16 14. Cinque (2007) extends on his account of the asymmetrical approach to DP-internal structure in applications to the hierarchical ordering of a greater range of sentence constituents. 15. For discussions of other approaches, see: Longobardi (2001), Giusti (2002) and Alexiadou, Haegeman and Stavrou (2007). 16. What has not been included in this discussion is the patterns of cooccurrence of demonstratives with soxa ‘one’. Data that I have collected on “one” shows (variable) acceptance of the two or-

190

Elizabeth Pearce

Table 4. Numerals and [+plural]/[-plural] deictic demonstratives17 (Restricted dialect) Numeral

Deictic demonstrative [− plural]

[+ plural]

[dual]

NO

NO

[plural]

NO

YES

Under strict feature matching only higher numerals are compatible with the merge of a DemP, which must have a [+plural] feature. If the NumP contains a dual numeral it will fail to merge a DemP above it because of Number feature incompatibility since there are no [+dual] demonstratives. To express this in Minimalist terms (Chomsky 2001), it could be proposed in these cases that the demonstrative has a [-interpretable] Number feature which must be valued by [+interpretable] Number. Numerals higher than two will have the [+interpretable] Number [plural] and can therefore check [-interpretable] [plural] on the demonstrative. These derivations converge in the Phonology producing successful spell-out of morphologically plural demonstratives. In the case of the [dual] numeral, however, there is no possible morphological realization of a [dual] demonstrative and the derivation fails to converge. This scenario seem relatively straightforward. What is then more puzzling is that, with the determiner-like, previous mention demonstratives, we have compatibility failure with the [dual] numeral, but not with [plural] numerals (6), (7). The relevant cooccurrences are shown in Table 5 Table 5. Numerals and previous mention demonstratives (Restricted dialect) Numeral

Previous mention demonstrative ngo

nga

[dual]

NO

NO

[plural]

YES

YES

What is unexpected in the distribution shown in Table 5 is that ngo/nga are compatible with [plural], but not with [dual]. If these determiner-like demonstratives simply have default Number, we might expect that they would cooccur with any derings, soxa before or after a [-plural]/[ØNumber] demonstrative. These data are not probative for the central focus of this paper. 17. For ease of comparison with examples, Tables 4–7 have left/right placement of the columns for numerals and demonstratives in accordance with the respective surface orderings.

Two or three things in the Unua noun phrase

191

Number specification, but it seems that “default” number is compatible with any Number but [dual]. We might thus conclude that [dual] as a marked feature is incompatible even with a previous mention demonstrative with [ØNumber]. In the less restricted dialects, leakage in the system is manifested in the extension of default realizations of both kinds of demonstratives (deictic and previous mention): both [dual] and [plural] numerals come to be compatible to varying degrees with non-plural demonstratives of both kinds [(13)–(15), (29), (30), (32)]. The available cooccurrence patterns for the less restricted dialects are shown in Table 6. Table 6. Numerals and [+ plural]/[− plural]/[ØNumber] demonstratives (Less restricted dialects) Numeral

Demonstrative Deictic

Previous mention

[− plural]

[+ plural]

[ØNumber]

[dual]

YES

NO

YES

[plural]

YES

YES

YES

In the less restricted dialects, xeru ‘two’ joins up with higher numerals in having the ability to cooccur with a previous mention demonstrative (30a), the latter represented in Table 6 as having [ØNumber]. Another innovation within the less restricted dialects is that both classes of numerals may cooccur with a non-plural deictic demonstrative. The contrasts in the cooccurrence possibilities shown in Table 6 versus in Tables 4 and 5 suggest that the less restricted dialect gives evidence of two innovations: (i) non-plural deictics come to have a default Number specification option (they become more like the previous mention demonstratives); and (ii) the [dual] feature of xeru weakens to the extent that it becomes compatible with a [ØNumber] demonstrative (with this feature now applicable to both deictics and previous mention demonstratives).18

18. It is interesting that, as reported in Baerman, Brown and Corbett (2005: 95), the most common syncretism in agreement systems merges dual with plural as against dual with singular. That is, for instance, in a language which has a three-way contrast singular/dual/plural in its pronoun system, but only a two-way contrast plural/non-plural in its subject-verb agreement paradigms, the verb agreement morphology with a [dual] subject is [+ plural] rather than [− plural]. In the present analysis of the Unua demonstratives, however, I am suggesting not that there is syncretism between dual and non-plural/singular, but rather that the morphological contrast evolves to become a contrast between plural Number versus un(der)specified Number.

192

Elizabeth Pearce

Overall, the scenario of feature matching that is suggested with respect to the outcomes displayed in Tables 4–6 goes some way to accounting for the data. The leakage applies to the resistance of the dual numeral to the acceptance of cooccurrence with a demonstrative which fails to bear a [+ dual] number specification. 4.4. Demonstratives and the dual pronoun At the high end of the structure, we have seen there are two types of acceptable sequences including deictic demonstratives: either (i) a quantifier occurs after a plural demonstrative or (ii) a dual pronoun occurs after a non-plural demonstrative. As we consider these constructions, we need to look to an increase in the available projections above the XP housing the DemP of (38a/b) because, as seen in (39), a partitive and a pronoun can cooccur. (39)

a.

namar ngaro rate ingot chief the.pl 3pl(they) many ‘many of the chiefs’

b.

namar xevej ngaro rate xeter chief four the.pl 3pl(they) three ‘three of the four chiefs’

In (39a) and (39b), rate is the 3rd person plural pronoun and, in both cases, it precedes a partitive, the quantifier in (39a) and the cardinal numeral in (39b).19 We might suppose that the inclusion of a pronoun provides overt referential content in the Spec of the DP and that the quantifier is housed in a projection above the DP (see also Giusti 1997 and Longobardi 2001 on quantifier position). The more extended “DP-internal” structure at the top end is then as follows: (40)

WP QuantP DP PronP D

XP DemP

19. If we compare (39a,b) with the earlier examples of (28), we see that the inclusion of the 3rd person pronoun is an optional additional specification.

Two or three things in the Unua noun phrase

193

From the earlier discussion in section 2.3, we saw that, with both classes of demonstratives, the only possible cooccurrences when raru is in PronP are those shown in Table 7. Table 7. Dual pronoun and [+ plural]/[− plural] demonstratives (All dialects) Demonstrative [− plural]

[+ plural]

YES

NO

Pronoun [dual]

The dual pronoun is incompatible with a plural demonstrative, but it cooccurs freely with a non-plural demonstrative, either deictic or previous mention. The freedom of cooccurrence of the dual pronoun with a non-plural demonstrative is unexpected in that it goes against the inability of the dual numeral to cooccur with a non-plural demonstrative: why should [dual] Number features have different effects at these different levels within the DP? The identification of the actual referent(s) of a definite DP is a function of a relation between that DP and a discourse (or in-the-world) antecedent/referent. The D/DP level of the structure is the location for the marking of the external link. We could suppose that the D therefore bears the phi-features (person and number) of its referent. All of the content in the structure below D can thus be viewed as additional descriptive specifications. If we suppose that, at the point at which D merges with the XP of (40), any (uninterpretable) Number feature on XP must be valued by D. If we further suppose that the XP may carry a [+ plural] specification (such as when a DemP in its Spec is [+ plural]), but that in other cases it does not have any Number features, then a [+ plural] XP can merge only a D which is also [+ plural] and an XP with no Number specification is free to merge a D of any Number specification. This kind of analysis of the role of Number features at this level of the DP structure accounts for the cooccurrence of plural pronouns with plural demonstratives, as in (39) and for the availability of the dual pronoun only with nonplural demonstratives. However, what it also entails is that a plural pronoun should not be incompatible with a non-plural demonstrative. This possibility requires further testing, but it is however supported by the following data from the author of the New Testament translations: (41)

No-ngva namar ngo/ngaro rate re-b-vena vex re 1sg-think chief the/the.pl 3pl 3pl-irr-come to loc xenen. feast ‘I hope those chiefs will come to the feast.’

194

Elizabeth Pearce

Overall, the data that we have seen has shown a robust pattern in which the dual numeral is dispreferred in the mirror-image ordering N – Num – Dem, which is however freely available when the Num is a numeral higher than two. Whilst evidence for such a restriction has not to date been shown in the literature on other languages which distinguish singular, dual and plural, the analysis of the Unua data that has been presented here naturally leads us to ask if languages closely related to Unua present evidence of a similar phenomenon. Whilst there are some difficulties in considering data from the neighbouring languages, in the section to follow I will show that three Malakula languages present evidence that they may have a restriction similar to that which applies in Unua.

5. Demonstratives and Number in languages related to Unua There are now in existence published descriptions of eight of the 40 or so Malakula languages,20 but the information included in these descriptions on the relative ordering of demonstratives and numerals is generally not as comprehensive as would be needed for a full comparison with the data that we have seen for Unua.21 However, there is data from three of these languages (Port Sandwich, Uripiv and Naman) that give indications that they may include a restriction that is similar to that found in Unua. Port Sandwich is one of nine, or possibly ten, languages which Cinque (2005: 319, fn.12) cites as having the DP-internal ordering: N – A – Dem – Num. Cinque’s analysis of the Port Sandwich ordering is based on information contained in the sketch grammar of Crowley (2002). The more extended description

20. These are of V’¨enen Taut (Fox 1976), Port Sandwich (Charpentier 1979), N¯ati (Crowley 1998), Naman (Crowley 2006a),Avava (Crowley 2006b), Nese (Crowley 2006c),Tape (Crowley 2006d) and Neve’ei (Musgrave 2007). I have also had access to unpublished materials on other Malakula languages kindly supplied to me by their authors. Input from these descriptions will be identified where relevant in the discussion following on in the text. Lynch (2006, 2007) presents hypotheses as to the subgroupings of the Malakula languages. For Lynch (2007), the major divide is between a Western and an Eastern linkage. Among the languages for which data is presented in this section, Unua-Pangkumu, Port Sandwich and Tirax belong in Lynch’s proposed Eastern linkage and Naman and Neve’ei in the Western linkage. For earlier accounts of the Malakula subgroupings, see Pawley (1972), Tryon (1976), Clark (1985), and Lynch (2000, 2001, 2003). 21. Some of these descriptions suffer from the limitation that data could be accessed only from a very small number of remaining speakers where the languages are moribund. Crowley (2006a: 90), for instance, states that he was uncertain about the precise functions of four demonstrative forms in Naman as information on spontaneous use of these demonstratives was not available to him (there remaining at the time not more than 20 fluent speakers of Naman, Crowley 2006a: 11).

Two or three things in the Unua noun phrase

195

of Charpentier (1979) claims the following ordering of constituents within the Port Sandwich DP (in which “N” can be understood as “nominal expression”): (42) N

Port Sandwich (Charpentier 1979: 169): N

Dem

Particle Numeral Verb

Comparative “other”

PL

Total

Poss delimiter

The data given in Charpentier (1979) do not include examples of noun phrases containing both a demonstrative and a numeral, but, in testing forms with two speakers,22 I was offered the following forms as translations of the English given in (43) and (44): Port Sandwich (43) a. n¨oangg xavoi amb-ixa e-u¨ 23 canoe local ?-prox it-two ‘those two canoes’ b. *n¨oangg xavoi e-u¨ amb-ixa canoe local it-two ?-prox (44)

n¨oangg xavoi ali-moreu¨ al-ixa canoe local pl-three? pl-prox ‘those three canoes’

Port Sandwich has a rich pronoun system, distinguishing a total of 15 personal pronouns, including four trials along with four duals (Charpentier 1979: 49). The data in (43) and (44) include two characteristics which appear to correspond to what we have seen in the data from Unua: the contrast in ordering for the demonstrative with “two” versus with “three” and the use of an apparently nonplural demonstrative with “two” versus the use of a plural form with “three”. Although further evidence from Port Sandwich would be needed to support the data shown in (43) and (44), on the present evidence, the Port Sandwich forms 22. My thanks to Flora Batick and Toula Telessi for contributing the data shown in (43) and (44). 23. In the orthography, the morpheme divisions and the glossing applied in (43) and (44), I have included information given by the speakers and information contained in and adapted from Charpentier (1979: 62–64, 164–173).

196

Elizabeth Pearce

are comparable to forms produced in Unua (the extensively demonstrated (1a) versus (1b) contrast). In an as yet unpublished account of the grammar of Uripiv, McKerras (2005) provides the following statement about DP-internal ordering: (45)

Uripiv: DP-internal ordering (McKerras 2005: 9) N – Modifier – Dem – Emphatic – Particles – Number

According to McKerras, numerals may appear in the Modifier position preceding Dem and in the further Number position following the Dem slot. An example that McKerras provides is the following: Uripiv (McKerras 2005: 10) (46) apu se kem nen nuru grandparent gen 1excl.du deic 3u ‘those two grandparents of ours’ In (46), the dual number specification is provided by the third person dual pronoun following the deictic demonstrative. Like Unua, Uripiv has an 11-way distinction in its personal pronoun system and its subject agreement morphology. It therefore seems possible that Unua and Uripiv may be comparable with respect to the use of the dual pronoun in DPs including a demonstrative.24 In the case of Naman, Crowley (2006a) does not present any formal statement on DP-internal ordering, but his account includes the following examples including a demonstrative and a “two” specification: Naman (Crowley 2006a: 81, 87) (47) a. nev¨ens nge iru banana dem two ‘these two bananas’ b.

birav nge aru lesser.yam dem du ‘those two lesser yams’

Like Unua, Naman has 11 personal pronouns (Crowley 2006a: 56), including the third person dual form of example (47b). The data in (47) is of considerable interest with respect to the analysis of the present discussion in that we see 24. However, it is also the case that, in my testing of Uripiv forms, the data supplied by one speaker have not provided confirmation for the pattern shown in (45). My thanks to Ansen Veremaito for this data.

Two or three things in the Unua noun phrase

197

in (47a) and (47b) the alternate use of the dual numeral and the dual pronoun following the demonstrative. Although we do not have data on the ordering of other numerals with respect to demonstratives, it is possible that the forms in (47) indicate that the grammar of Naman shows effects similar to those of Unua in that it may include a similar restriction on the use of the dual numeral before a demonstrative within the DP. Thus, even on the basis of very limited information, we do see indications in the data from Port Sandwich, Uripiv and Naman that the Number matching requirements that we have posited for Unua may be present in at least these languages as well. Descriptions of three further Malakula languages also provide information on the DP-internal ordering of numerals or number with demonstratives. Musgrave (2007) gives the examples from Neve’ei shown in (48) in which a demonstrative cooccurs with a number specification, the ordinal “second” in (48a) and cardinal “one” in (48b). Neve’ei (Musgrave 2007: 69, 71) (48) a. I-tokh ran niyim nge aruan. 3sg.real-live loc house dem second ‘(S)he lives at the second house.’ b.

I-takh lowi nevwene-n nakhankhan nge 3sg.real-take away fruit-constr pawpaw dem sevakh. one ‘She took away one fruit of the pawpaw tree.’

Musgrave also claims the following as the most frequently occurring order for noun phrases with a noun head: (49)

Neve’ei: DP-internal ordering (Musgrave 2007: 59) Nominal + (Adjective) + (Possessive postmodifier) + (Determiner) + (Plural postmodifier) + (Numeral) + (Quantifier) + (Relative clause)

The data shown in (48) conform to the pattern given in (49) in that in both (48a) and (48b) a number specification follows the determiner/demonstrative. But it is not clear if the available variability in ordering covers any restrictions on the ordering of specific cardinal numerals with the determiner/demonstrative. According to Crowley’s (2006b) claim as to DP-internal ordering in Avava, Avava has a “standard” mirror-image ordering for its DP constituents:

198 (50)

Elizabeth Pearce

Avava: DP-internal ordering (Crowley 2006b: 24) Noun + Adjective + Numeral + Demonstrative/Relative clause

In the case of Avava, therefore, we do not have any evidence for the possibility that a number specification may be expressed after a demonstrative (data shown in Crowley 2006b do not include examples of noun phrases containing both a numeral and a demonstrative). In the case of Tirax, Brotchie (2007) proposes the following DP-internal ordering: (51)

Tirax: DP-internal ordering (Brotchie 2007: 29) N – (AP*) – (ANum) – (Poss NP) – (Dem/Art) – (Plural marker xnEr) – (Quant/VNum) – (RelCl)

In Tirax the “ANum” close to the head N has reduplicative morphology like that of adjectives which can immediately follow the N. The “VNum” located in the higher structure of the DP bears verbal morphology. Brotchie includes the following examples showing that the “two” specification may occur in the ANum position preceding a demonstrative: Tirax (Brotchie 2007: 63) (52) a. Resan ru-ru xar nte hxal sentence dup-2 dx2 thing one ‘These two sentences mean the same thing.’ b.

Resan i-ru xar nte hxal sentence 3sg.realis-2 dx2 thing one ‘These two sentences mean the same thing.’

Brotchie describes the “dx2” of Tirax as an intermediate demonstrative, closest to the proximate in terms of relative distance in a set of four demonstratives. As Brotchie points out (personal communication 18/10/2007), the numeral “two” appears to be less restricted in its distribution than other numerals in that it may occur in the VNum form before a demonstrative, as is shown in (52a). The very intriguing data from all of these languages indicate that many details in the morphological composition may have an effect on DP-internal orderings. This brief review has only touched the surface and a much deeper examination of the data is required before we can gain a fuller understanding of the processes at work. But we have nevertheless been able to observe three additional cases, Port Sandwich, Uripiv and Naman, for which we have evidence of the use of the dual pronoun following a demonstrative within a DP. These data indicate that it is possible that further investigation of DP-internal ordering in the languages

Two or three things in the Unua noun phrase

199

of Malakula may show that the restrictions that I have demonstrated for Unua may be manifested in similar ways in yet other languages.

6. Conclusions The findings with respect to Unua have brought to light a marked restriction on the use of cardinal numeral “two” in the grammar of a language which has a three-way distinction singular/dual/plural in its pronoun system. I have attributed the Unua restriction against the cooccurrence of the cardinal numeral “two” with a following demonstrative to the unavailability of a [+ dual] specification on demonstratives. Although there are many languages in the world which distinguish singular, dual and plural in their pronoun systems, I have not found any reference to comparable restrictions concerning DP-internal numeral positions in accounts of any other languages. However, some data from other Malakula languages that has been shown here brings out the possibility that the Unua phenomenon may not be unique to Unua. The analysis of the restriction favouring the use of the dual pronoun following a demonstrative in Unua has led to an account of processes that give rise to a marked DP-internal surface constituent ordering. Such an account is of particular interest for the understanding of the syntax of available orderings of constituents within DPs. I hope that the analysis that I have presented here may make a contribution to such continuing investigation.

References Abels, Klaus and Ad Neeleman 2006 Universal 20 without the LCA. Glow Newsletter 56: 16–17. Berlin: Mouton de Gruyter. Alexiadou, Artemis, Liliane Haegeman and Melita Stavrou 2007 Noun Phrase in the Generative Perspective. Berlin: Mouton de Gruyter. Baerman, Matthew, Dunstan Brown and Greville Corbett 2005 The Syntax-Morphology Interface: A Study of Syncretism. Cambridge: Cambridge University Press. Bembe, Kalangis 2005 Rejen nga mifo hise Iesu Kresto Mark iri [The Gospel of Mark]. Ms. Ruhumbo, Malakula. Bembe, Kalangis 2006 Rejen ga mifo hise Atua Matiu iri [The Gospel of Matthew]. Ms. Ruhumbo, Malakula.

200

Elizabeth Pearce

Brotchie, Amanda 2007 Nouns and noun phrases. Ms. University of Melbourne. Charpentier, Jean-Michel 1979 La langue de Port-Sandwich (Nouvelles-H´ebrides): Introduction pho´ nologique et grammaire. Paris: Soci´et´e d’Etudes Linguistiques et Anthropologiques de France (SELAF). Chomsky, Noam 2001 Derivation by Phase. In: Michael Kenstowicz (ed.), Ken Hale: A Life in Language, 1–52. Cambridge, MA: Massachusetts Institute of Technology Press. Cinque, Guglielmo 2005 Deriving Greenberg’s Universal 20 and its exceptions. Linguistic Inquiry 36: 315–332. Cambridge, MA: Massachusetts Institute of Technology Press. Cinque, Guglielmo 2007 The fundamental left-right asymmetry of natural languages. University of Venice Working Papers in Linguistics 17: 77–107. Venezia: Universit`a Ca’ Foscari Venezia Clark, Ross 1985 Languages of North and Central Vanuatu: Groups, chains, clusters and waves. In: Andrew Pawley and Lois Carrington (eds.),Austronesian Languages at the 15th Pacific Science Congress, 199–236. (Pacific Linguistics C-88.) Canberra: Pacific Linguistics. Crowley, Terry 1985 Common noun phrase marking in Proto-Oceanic. Oceanic Linguistics 24: 135–193. Honolulu: University of Hawai’i Press. Crowley, Terry 1998 A salvage sketch of N¯ati (Southwest Malakula, Vanuatu). In: Lynch, Ross and Crowley, 101–148. Crowley, Terry 2002 Port Sandwich. In: Lynch, Ross and Crowley, The Oceanic Languages, 650–659. Richmond, Surrrey: Curzon Press. Crowley, Terry 2006a Naman: A Vanishing Language of Malakula (Vanuatu) [ed. by John Lynch]. (Pacific Linguistics 576.) Canberra: Pacific Linguistics. Crowley, Terry 2006b TheAvava Language of Central Malakula (Vanuatu) [ed. by John Lynch]. Canberra: Pacific Linguistics. Crowley, Terry 2006c Nese: A Diminishing Speech Variety of Northwest Malakula (Vanuatu) [ed. by John Lynch]. Canberra: Pacific Linguistics.

Two or three things in the Unua noun phrase

201

Crowley, Terry 2006d Tape: A Declining Language of Malakula (Vanuatu) [ed. by John Lynch]. Canberra: Pacific Linguistics. Giusti, Giuliana 1997 The categorial status of determiners. In: Liliane Haegeman (ed.), The New Comparative Syntax, 95–123. London: Longman. Giusti, Giuliana 2002 The functional structure of noun phrases. In: Functional Structure in DP and IP: The Cartography of Syntactic Structures, Vol. 1: 54–90, ed. by Guglielmo Cinque. Oxford: Oxford University Press. Greenberg, Joseph H. 1966 Some universals of grammar with particular reference to the order of meaningful elements. In: Universals of Language 2nd Edition: 73–113. ed. by Joseph H. Greenberg. Cambridge, MA: Massachusetts Institute of Technology Press. Kayne, Richard S. 1994 The Antisymmetry of Syntax. Cambridge, MA: Massachusetts Institute of Technology Press. Longobardi, Giuseppe 2001 The structure of DPs: Some principles, parameters and problems. In: Mark Baltin and Chris Collins (eds.), The Handbook of Contemporary Syntactic Theory, 562–603. Oxford: Blackwell. Lynch, John 2000 Linguistic subgrouping in Vanuatu and New Caledonia. In: Bill Palmer and Paul Geraghty (eds.), Proceedings of the Second International Conference on Oceanic Linguistics, 155–184. (Pacific Linguistics 505.) Canberra: Pacific Linguistics. Lynch, John 2001 Article accretion and article creation in Southern Oceanic. Oceanic Linguistics 40: 224–246. Honolulu: University of Hawai’i Press. Lynch, John 2003 Low vowel dissimilation in Vanuatu languages. Oceanic Linguistics 42: 359–406. Honolulu: University of Hawai’i Press. Lynch, John 2006 Some notes on the linguistic history of Malakula. Paper presented at the Terry Crowley Memorial Workshop on Vanuatu Languages. Wellington, 13–14 November 2006. http://www.vuw.ac.nz/lals/about/news-events. aspx. Lynch, John 2007 The Malakula linkage of Central Vanuatu. Paper presented at COOL 7: Seventh International Conference on Oceanic Linguistics, 2–6 July 2007, Noumea.

202

Elizabeth Pearce

Lynch, John and Terry Crowley 2001 Languages of Vanuatu: A New Survey and Bibliography. (Pacific Linguistics 517.) Canberra: Pacific Linguistics. Lynch, John, Malcolm Ross and Terry Crowley (eds.) 2002 The Oceanic Languages, Richmond, Surrey: Curzon press. Lyons, Christopher 1999 Definiteness. Cambridge: Cambridge University Press. McKerras, Ross 2005 Uripiv phonology and grammar. Ms. SIL. Musgrave, Jill 2007 A Grammar of Neve’ei, Vanuatu. (Pacific Linguistics 587.) Canberra: Pacific Linguistics. Ozanne-Rivierre, Fran¸coise 1992 The Proto-Oceanic consonantal system and the languages of New Caledonia. Oceanic Linguistics 31: 193–207. Honolulu: University of Hawai’i Press. Ozanne-Rivierre, Fran¸coise 1995 Structural changes in the languages of New Caledonia. Oceanic Linguistics 34: 45–72. Honolulu: University of Hawai’i Press. Pawley, Andrew 1972 On the internal relationships of Eastern Oceanic languages. In: R.C. Green and M. Kelly (eds.), Studies in Oceanic Culture History Vol. 3: 1–142. Honolulu: Bernice Pauahi Bishop Museum. Pearce, Elizabeth 2007 The reflexes of Proto-Oceanic *na in Unua. In: Jeff Siegel, John Lynch and Diana Eades (eds.), Language Description, History and Development: Linguistic Indulgence in Memory of Terry Crowley, 327–339. Amsterdam: Benjamins. Pearce, Elizabeth 2010 Possession syntax in Unua DPs. In: Raphael Mercado, Eric Potsdam, and Lisa deMena Travis (eds.),Austronesian andTheoretical Linguistics, 141–162, Amsterdam: Benjamins. Tryon, Darrell T. 1976 New Hebrides Languages: An Internal Classification. (Pacific Linguistics C-50.) Canberra: Pacific Linguistics. Vanuatu National Statistics Office 1999 National Population and Housing Census. Port Vila: National Statistics Office.

Noun incorporation in Saliba Anna Margetts

1. Introduction1 Saliba, an Oceanic language of the Papuan Tip cluster spoken in Papua New Guinea, shows some interesting patterns in noun incorporation constructions. The language has two types of incorporation, lexical compounding (type I in Mithun’s 1984 typology of incorporation) and external possession (Mithun’s type II) as well as a related construction sometimes considered as pseudo incorporation. In both types of incorporation, lexical compounding and external possession, a noun is morphologically incorporated into the verb and, as a consequence, it is semantically backgrounded and loses its syntactic independence. In the case of external possession, a body part term which functions as the subject of an intransitive verb is incorporated and forms a compound with the verb.2 (1)

a.

Boga-na ye sese. belly-3sg.poss 3sg swollen ‘His belly is full.’

b. Ye boga-sese. 3sg belly-swollen ‘He’s full.’

In lexical compounding, it is the object argument which is incorporated into the verb. In this paper, I investigate the semantic and morpho-syntactic patterns in lexical compounding constructions and discuss how the Saliba data can contribute to the ongoing discussion of the lexical vs. syntactic nature of noun incorporation cross-linguistically. I will not be concerned with Saliba external possessive constructions here. For a discussion of those the reader is referred to Margetts (1999).

1. I am grateful to Felix Ameka, Ulrike Mosel, Eric Pederson and Steve Levinson for feedback on a much earlier version of this paper, as well as to two anonymous referees and to the editors of this volume who gave very helpful comments. 2. The data discussed here come from a corpus of about 15 hours of transcribed spoken language, recorded between 1995 and 2005. Information in parentheses following the free translation of examples indicates the source text. Examples without such information are either elicited or were noted from conversations. Abbreviations follow the Leipzig Glossing rules and in addition the following are used: ANA, anaphoric, COMP, complementizer, PP, postposition, RED, reduplication.

204

Anna Margetts

The term “noun incorporation” has been used in a rather heterogeneous way in the literature and definitions differ as to what constitutes incorporation. Before turning to the Saliba data, a brief definition of incorporation is therefore called for. I apply a structurally based definition which distinguishes incorporation from other constructions that involve the backgrounding of nouns without formally binding them to the verb. According to this definition, a noun is considered to be incorporated only if it is structurally part of the verb and is not morpho-syntactically independent. This definition allows us to distinguish between true morphological incorporation and other instances of intransitive verbs with objects. Such constructions are cross-linguistically not uncommon and are at times considered as incorporation in the literature, as in the Mayan examples discussed by Mithun (1984: 852) or in Fran¸cois’ Araki grammar (2002: 60, 138).3 Margetts (2008) discusses clauses with intransitive verbs and backgrounded but syntactically independent objects as “discord” constructions in Saliba and other Oceanic languages (drawing on Sugita 1973; and Lichtenberk 1982 among others).4 These constructions differ from noun incorporation in that the object occurs in the canonical object position, rather than inside the verb, and in that the nouns can take modifiers, albeit drawn from a more restricted set than for regular object nouns. Objects in discord constructions are typically low in some features of object individuation (such as definiteness, specificity, affectedness, etc.) as discussed by Hopper and Thompson (1980). Similar constructions have been discussed as “noun stripping” (Miner 1986, 1989). Gerdts (1998: 93) describes the difference between incorporation and noun stripping as follows. [. . . ] a “stripped” noun does not have the usual case marking associated with its grammatical function . . . Noun stripping differs from incorporation, however. Incorporation is morphological: the two elements involved are part of the same word in surface structure. In noun stripping, the two elements remain as separate words according to phonological criteria [. . . ]

Similar constructions have also been discussed as pseudo noun incorporation or phrasal noun incorporation in Polynesian languages like Niuean, M¯aori and Tongan (Massam 2001; Chung 2004; Ball 2005a, 2005b) where nouns are “stripped” of their case marking. In the Oceanic examples discussed in Margetts (2008) it is not obvious that the nouns are in fact stripped of any marking as the languages 3. The Araki examples may in fact be cases of morphological incorporation but Fran¸cois (2002) does not present evidence for this. The defining criteria for what he calls incorporation are the intransitive status of the verb and the non-referential status of the object noun. 4. Such clauses can be described as showing discord between verb-level transitivity (intransitive verb forms) and clause-level transitivity (lexical object arguments).

Noun incorporation in Saliba

205

do not have case marking and the nouns can take a number of modifiers. It is rather the verb which is stripped of its transitivity marking. While Sugita’s (1973) study suggests that languages have either noun incorporation or clauses with discord (“semi-transitive verbs” in his terms), some languages, like Saliba, show both types of constructions. (Similarly, Miner 1986 shows that noun stripping and incorporation both exist in Zuni.) An example of Saliba noun incorporation is given in (2a), one of discord in (3a). The incorporation example shows a noun-verb compound. The discord example show the noun in the canonical object position but the verb is morphologically intransitive; it does not carry the applicative and object suffix. The corresponding canonical transitive clauses, without incorporation or discord, are given in the examples in (b). Elicitations have not been able to determine semantic differences between incorporation and discord examples. They may well exist but be difficult to pin down. (2)

(3)

a.

Se koya-tudai. 3pl garden-dig ‘They dig a garden.’ (Boneyawa 26DZ 0030)

b.

Koya se tudai-Ø. garden 3pl dig-3sg.obj ‘They dig a garden.’ (Garden 05CY 0007)

a.

Puwaka ta bahe. pig 1incl carry ‘We carry pigs.’ (oldtime1:127)

b.

Puwaka ta bahe-i-di. pig 1incl carry-appl-3pl.obj ‘We carry the pigs.’

Having established that incorporation and discord constitute distinct constructions, the remainder of this paper will focus on the nature of Saliba lexical compounds. For a more detailed discussion of discord constructions the reader is referred to Margetts (2008). In many respects, Saliba noun incorporation follows cross-linguistically common and well attested patterns. There are however some typologically uncommon features which make the Saliba constructions particularly interesting. Apart from the coexistence of discord and noun incorporation, an interesting feature is the ordering of stems within incorporation constructions which can be either N-V or V-N. A third point of interest is the fact that, in some cases, the incorporating base verb is morphologically intransitive rather than transitive

206

Anna Margetts

and would therefore not be expected to incorporate a noun (see section 3.4). In section 2 I provide some information on Saliba morpho-syntax as a background for the discussion of noun incorporation which follows. In section 3 I give an overview of lexical compounding in Saliba from a semantic and a morphosyntactic perspective. Section 4 addresses the question of the lexical versus syntactic nature of incorporation and section 5 provides some conclusions.

2. Notes on Saliba morpho-syntax Saliba is a nominative-accusative head-marking language. The intransitive subject (S) and the transitive subject (A) pattern identically and contrast with the object of transitive verbs (O) both on morphological and syntactic grounds. Saliba has a rigid Subject-Object-Verb constituent order and also shows related word order characteristics like postpositions and genitive-noun ordering (but note that the order of the pronominal affixes is s-V-o). There is no case marking on nominals. Adjuncts can occur before and after the verb and are generally marked by postpositions. There is no verb in the language that subcategorizes for a postpositionally marked argument. All Saliba verbs obligatorily carry a subject prefix and, if they are morphologically transitive, an object suffix. It is the subject prefix which defines a word as an inflected verb. (The subject prefix is written as a separated word in the Saliba orthography developed by the Summer Institute of Linguistics.) The pronominal affixes occur regardless of whether there is a lexical subject or object noun phrase in the clause.5 There is a full person/number paradigm for both affixes. The subject and object markers are analyzed as affixes based on the following criteria: (a) Intonationally the object pronoun is part of the word; it can trigger a stress shift onto the new penultimate syllable. (b) The object suffix of the third person singular has a word-final (-Ø) and a word-internal allomorph (-ya). The latter appears if the object suffix is followed by a directional suffix (-ma or -wa) or the perfect suffix (-ko). And finally, (c) only bound derivational morphemes can intervene between the subject or object suffixes and the verb stem. For the subject prefix these derivational morphemes include reduplication, the causative, resultative, and reciprocal prefix, and the reduced transitivity marker kai-. The only derivational morpheme that can intervene between the object suffix and 5. The affixes appear independent of whether there is a co-referential lexical noun present in the clause. However, as I discuss further below, not all objects are cross-referenced on the verb. So, while a clause can contain a lexical object that is not cross-referenced, the object suffix is not absent because the lexical object is present but because of the status of the object argument (e.g. indefinite).

Noun incorporation in Saliba

207

the verb stem is the applicative suffix. The applicative suffix is always followed by an object suffix. In turn, verbs that derive a transitive stem by means of the applicative suffix can never take the object suffix directly on the stem; they need to be applicativized first. It can therefore serve as a test for the presence of the zero allomorph of the object suffix. In the following I use the term “transitive verb” to mean a verb which is morphologically transitive, i.e. one which carries an object suffix (and, depending of the verb class, also the applicative suffix). In contrast to subjects, which are always cross-referenced on the verb, only certain types of objects are marked on the verb by the object suffix. The class of objects which are cross-referenced is hard to capture by a rigid definition. Specific, definite objects tend to be cross-referenced (although there are counter examples, cf. Margetts 2008). Sugita (1973) describes this class as highly individuated, as compared to less individuated objects which are not cross-referenced.

3. Lexical compounding in Saliba In Mithun’s (1984) type I incorporation or lexical compounding, a noun stem and a verb stem are compounded to build a new lexical item which functions as a verb stem. Noun incorporation of this kind is a common and productive process in Saliba. Loanwords may be incorporated and incorporation constructions can express relatively novel activities. Consider (4) and (5) where the English loans gita ‘guitar’ and leta ‘letter’ are incorporated. (4)

Ye leta-kuli. 3sg letter-write ‘He wrote letters.’ (PairedFilms 01AW 0301)

(5)

Se koi-gita. 3pl hit-guitar ‘They played guitar.’

The semantics of noun incorporation have been widely discussed in the literature and Mithun (1984: 850) summarizes them as follows: [. . . ] these constructions are generally used to describe activities or events whose patients are neither specific nor countable – e.g. habitual, ongoing, or projected activities; those done by several people together; or those directed at a non-specific part of a mass.

In terms of discourse function, incorporation generally results in the backgrounding of the incorporated noun in some ways (cf. Heath 1976; Giv´on 1990).

208

Anna Margetts

Foley and VanValin (1985: 344) consider it as a type of antipassive construction. The basic semantic/pragmatic condition for the creation of lexical compounds is that a noun stem and a verb stem together constitute what Mithun (1984: 848) considers a “recognizable, unitary concept” in the language. In many cases, the compounds are semantically transparent and the rough meaning of the constructions can be predicted from its parts. However, there is a tendency for lexicalization of incorporation constructions in the sense that they tend to acquire a more specialized meaning than the corresponding analytic clauses. The semantics of Saliba incorporation constructions including the choices of nouns and the tendency for lexicalization are discussed in section 3.1. Within type I incorporation, Mithun (1984) distinguishes between morphological compounding and composition by mere juxtaposition. While she discusses a number of Oceanic languages as examples of compounding by juxtaposition, Saliba exhibits characteristics of morphological compounding. But as opposed to some languages with morphological compounding, in Saliba there are no word-internal processes that could identify incorporation constructions as single units on phonological grounds. Rather, Saliba incorporation constructions are identified as single units on the basis of bound morphology. The base verb and the incorporated noun together build a morphologically complex verb stem which takes a single set of inflectional affixes. A number of morphological tests demonstrating the dependent status of the noun are discussed in 3.2. In Saliba the ordering of stems within the compound is most commonly nounverb, as in (4) above, but there are also constructions with the reverse order, as in (5). The choice between N-V versus V-N ordering within the compound is discussed in 3.3. The base verb is typically transitive and, through the process of incorporation, it becomes intransitive. In some cases, however, the base verbs are intransitive and therefore should not be able to participate in the process of incorporation, as discussed in section 3.4. 3.1. Semantics of incorporation In Saliba, as cross-linguistically, the incorporated noun has the function of specifying the activity expressed by the verb and it generally constitutes the object argument in a corresponding analytic clause. Incorporation constructions typically refer to habitual and/or repetitive activities and they often denote activities performed by several people directed at multiple objects (although this is not a requirement). In the following, I discuss the types of objects which can be incorporated in Saliba, the tendency of the compounds to lexicalize and the criteria of habitual activities, multiple objects and multiple subjects.

Noun incorporation in Saliba

209

3.1.1. Types of objects The incorporated nouns in Saliba type I incorporation are almost exclusively patients of the activity denoted by the verb (some exceptions featuring incorporated locations are discussed in 3.4 below but there are no examples, e.g., of incorporated instruments). There is no specific formal class of nouns that can be incorporated into the verb. Proper nouns are never incorporated which is not surprising given that incorporation is generally a backgrounding device whereas proper nouns express highlighted information and definiteness. In Saliba, incorporation occurs especially when people talk about central areas of their life such as eating, drinking, gardening, and activities around the house. Impressionistically, among the most common nouns to be incorporated are koya ‘garden’, niu ‘coconut’ and puwaka ‘pig’. Most commonly, incorporated nouns denote inanimate objects in Saliba, but animates are also attested as in (6) where the incorporating verb is hekai ‘feed’ (which is derived from kai ‘eat’ by means of the causative prefix). The incorporated version in (a) is lexicalized with the meaning ‘to raise pigs’ (i.e. habitually feeding pigs). (6)

a.

Se puwaka-he-kai. 3pl pig-caus-eat ‘They raise pigs.’ (Giyahi 01AA 0009)

b.

Puwaka se he-kai-di. pig 3pl caus-eat-3pl.obj ‘They fed the pigs.’(Giyahi 01AA 0010)

There is a tendency in Saliba to incorporate superordinate terms rather than subordinates as demonstrated in (7). In (7a) the noun kaleko ‘clothes’ is incorporated into the verb deuli ‘wash’. Incorporation of the subordinate terms lulu ‘shirt’ or pilipou ‘trousers’ was considered unacceptable by speakers. (7)

a.

Ya kaleko-deuli. 1sg clothes-wash ‘I do the laundry.’

b. ?Ya lulu-deuli. 1sg shirt-wash ‘I shirt-wash.’ Body parts can also be incorporated as in the following examples. In the analytical constructions, these body part terms are inalienably possessed and need to carry a possessive suffix.

210

Anna Margetts

(8)

a.

Ya mata-kabi. 1sg eye-touch ‘I wash my face.’ (lit.: I eye-touch.) (Conversation 06B0 0022)

b.

Mata-gu ya kabi-di. eye-1sg.poss 1sg touch-3pl.obj ‘I touch/wash my eyes.’

a.

Ya nima-deuli. 1sg hand-wash ‘I wash my hands.’

b.

Nima-gu ya deuli-di. hand-1sg.poss 1sg wash-3pl.obj ‘I wash my hands.’

a.

Nige ye gayagaya-yali. neg 3sg chin-shave ‘He doesn’t shave.’

b.

Gayagaya-na nige ye yali-Ø. chin-3sg.poss neg 3sg shave-3sg.obj ‘He didn’t shave his chin.’

(9)

(10)

3.1.2. Habituality, repetition and multiple participants As noted, Saliba Type I incorporation typically denotes habitual and/or repetitive activities performed by several people and directed at multiple objects. But habituality alone is not a sufficient condition for incorporation and not everything which is habitual can be expressed by incorporation. An activity which might be habitual for a single person does not constitute a good candidate for an incorporation construction. The activities which are expressed by noun incorporation are typically recognized by speakers as “unitary, institutionalized activities” (Mithun 1984: 850). Incorporation constructions create labels for stereotypical activities which are considered salient enough in the language community to have their own name. These conditions can pragmatically restrict the choice of incorporated noun. The examples in (7) above showed the tendency to incorporate superordinate terms rather than subordinate ones. The following examples show that there can also be clear preferences in the choice between hyponyms. The verb numa

Noun incorporation in Saliba

211

‘drink’ can incorporate the noun ti ‘tea’. The clause in (11a) is lexicalized with the meaning to have breakfast.6 (11)

a.

Ka ti-numa. 1excl tea-drink ‘We drank tea / we had breakfast.’

b.

Ti ka numa tea 1excl drink ‘We drank tea.’

Nouns like kopi ‘coffee’ or kordiyal ‘cordial’, both well known but less common drinks (and less affordable), cannot be incorporated by numa ‘drink’, as shown in (12). (12)

b. ?Se kopi-numa. 3pl coffee-drink ‘They coffee-drank.’

In some instances, a further prerequisite for incorporation, besides habituality, is repetitive activity and, as an effect of that, the involvement of multiple objects and potentially multiple subjects. This is the case for a number of constructions denoting group activities in the garden such as koya-daibi ‘garden-clean’ and koya-deula ‘garden-terrace’, and also for the incorporation of niu ‘coconut’ into a number of verbs. Certain compound stems with niu ‘coconut’ are lexicalized as referring to the production of copra (smoked coconut meat). Examples are compound stems such as niu-hesulu ‘coconut-pile’, niu-pulisi ‘coconut-husk’, niu-isi ‘coconut-split’, and niu-tutu ‘coconut-pound’. These compound stems refer to the collective activity of gathering and husking piles of coconuts which are then opened and smoked. Later the smoked meat is taken out of the shell and stuffed tightly into copra bags. This work typically goes on for several days and involves a whole group of people and masses of coconuts. The text example in (13a) refers to husking coconuts for making copra. The corresponding analytic construction in (b) also refers to the activity of husking coconuts, but is typically interpreted as referring to a smaller number of coconuts such as used for cooking or drinking. (13)

a.

Se niu-pulisi. 3pl coconut-husk ‘They coconut-husked (to make copra).’ (fishdial107)

6. To my knowledge this expression can be used even if people do not have tea during breakfast.

212

Anna Margetts

b.

Niu ye pulisi-di. coconut 3sg husk-3pl.obj ‘He husked the coconuts (e.g. to drink).’

The same contrast holds for the clauses in (14) where (a) refers to breaking and stuffing the dried coconut meat into copra sacks but (b) refers to breaking open coconuts for cooking or drinking. (14)

a.

Se niu-tutu. 3pl coconut-pound ‘They pound and stuff coconut meat into copra bags.’ (Fishing 01BQ 0670)

b.

Niu ya tutu-di. coconut 1sg pound-3pl.obj ‘I break open the coconuts.’

In this example, the compound stem is lexicalized to express a different activity (stuff into bags by pounding) from the verb in the analytic counterpart (break shell open). 3.2. Morpho-syntax of incorporation As mentioned, there is no phonological evidence that Saliba lexical compounds resulting from incorporation constitute single units. It is on the basis of bound morphology that they can be identified as such. Incorporation constructions create grammatical words. This is to say that while noun and verb in these cases can alternatively occur as independent words with their respective inflections, they are not independent of each other in incorporation constructions in the sense that the incorporating construction could not be cut in half and result in two grammatical words. (E.g. in the case of N-V incorporation, the subject prefix and the noun do not constitute a grammatical word and neither does the verb stem without the subject prefix). In this section, I discuss a number of tests which demonstrate this dependent and morphologically bound status of the incorporated noun. As detailed in 3.3, there are both N-V and V-N compounds in Saliba. Some of the tests apply to both N-V and V-N constructions while others are useful specifically to one of the two types. An overview of the tests and their functions is presented in Table 1. All tests show the morphologically bound status of the noun. Note that if a noun classifies as incorporated by one test it will classify as incorporated by all the tests, i.e. all tests give the same result about the status of a construction (if all tests can be applied.) The nominalization test also demonstrates that the lexical compounds of noun and

Noun incorporation in Saliba

213

verb build a semantic unit. In addition, the complex verb test and the -ko suffix test show that the compound verb is morphologically intransitive. Table 1. Morphological tests for noun incorporation Test

Description

most useful for

Subj prefix test Reduplication test Modifier test Complex verb test -ko suffix test Nominalization test

N appears between Subj prefix and V. N reduplicates instead of V. N cannot take modifiers. Serialized verb attaches to N not to V. Suffix attaches to N not to V. N and V are nominalized as a unit.

N-V N-V N-V and V-N V-N V-N N-V and V-N

3.2.1. Subject prefix test For N-V incorporation, where the noun stem precedes the verb, the incorporated status of the noun is immediately obvious from its position inside the inflected verb, between the obligatory subject prefix and the verb stem. (15)

Ka huni-keli. 3pl taro-dig ‘We dug up taro.’ (Giyahi 01AA 0666)

(16) Ya kaigalu-yagu. 1sg wild.greens-pick ‘I picked kaigalu greens.’ 3.2.2. Reduplication test In Saliba, the progressive aspect is expressed by reduplication of the first two syllables of the verb stem. In incorporation with N-V order, the incorporated noun stem reduplicates instead of the verb. In (17a), the verb incorporates the loan word leta ‘letter’ and it is this incorporated noun which reduplicates to express the progressive aspect. It is ungrammatical to reduplicate the verb stem instead, as shown in (b). In (17c) without incorporation, the verb stem kuli ‘write’ is reduplicated. (17)

a.

Sinebada ye leta-leta-kuli. old.woman 3sg red-letter-write ‘The woman was letter writing.’ (PairedFilms 01AW 0926)

b. *Sinebada ye leta-kuli-kuli. old.woman 3sg letter-red-write

214

Anna Margetts

c.

Leta ye kuli-kuli-Ø. letter 3sg red-write-3sg.obj ‘She was writing a letter.’

A text example is presented in (18). (18)

Se koya-koya-paisowa. 3pl red-garden-work ‘They were working in the garden.’ (BudoiNualele 01CY 018)

3.2.3. Modifier test Saliba incorporated nouns cannot be modified. Modifiers can neither occur adjacent to the noun nor can they be stranded preceding or following the verb. In (19a) the bare noun is incorporated into the verb. Example (19b) shows an unincorporated object followed by the anaphoric marker wa. (19)

a.

Se niu-yaga. 3pl coconut-scrape ‘They scraped coconuts.’

b.

Niu wa se yaga-di. coconut ana 3pl scrape-3pl.obj ‘They scraped the coconuts.’

Even inalienably possessed nouns, which in analytic constructions may not occur without a possessive suffix are incorporated as bare nouns. In (20a) the noun kawa ‘mouth/teeth’ is incorporated (parallel to (8a), (9a) and (10a) above). In the analytic construction in (b) the noun must carry a possessive suffix (parallel to (8b), (9b) and (10b) above). The clause in (c) without this suffix is ungrammatical. (20)

a.

Ya kawa-deuli. 1sg tooth-wash ‘I brushed my teeth.’

b.

Kawa-gu ya deuli-Ø. mouth-1sg.poss 1sg wash-3sg.obj ‘I brushed my teeth.’

c. *Kawa ya deuli-Ø. mouth 1sg wash-3sg.obj

Noun incorporation in Saliba

215

3.2.4. Complex-verb test Further proof that the lexical compounds form a morphological unit is provided by cases of incorporating verbs followed by other verbs in nuclear-layer serialization (cf. Margetts 2004). For V-N incorporation, the test verifies that the noun is part of the verb rather than occupying a position following it. Consider the example in (21a) where the compound kabi-kabole ‘make-sago’ is followed by the verb gehe ‘finished’. The example in (b) with gehe ‘finished’ intervening between the verb and the noun stem is unacceptable. (21)

a.

Se kabi-kabole-gehe. 3pl make-sago-finished ‘They finished making sago.’

b. *Se kabi-gehe kabole. 3pl make-finished sago Complex verbs can also be employed to test the morphologically transitive or intransitive status of a verb, which will be relevant in section 3.4. In nuclearlayer serialization, certain adverb-like verbs with meanings such as ‘quickly’, ‘properly, ‘do again’, or ‘finished’ have to agree with the transitivity status of the preceding verb stem and occur either in their transitive or in their intransitive form. This provides a test for verbs without overt transitivity marking which may or may not carry the zero object suffix. The example in (22a) demonstrates that the lexical compound niu-hesulu ‘coconut-pile’ is intransitive because the final verb stem appears in its intransitive form uyo ‘do again’. The example in (b) shows that niu-hesulu cannot be followed by the transitive stem uyo-i (with the applicative suffix). The example in (c) shows that the verb stem hesulu ‘pile’ is transitive because the final verb stem occurs in its transitive form uyo-i. (22)

a.

Se niu-hesulu-uyo. 3pl coconut-pile-again ‘They piled coconuts again.’

b. *Se niu-hesulu-uyo-i-Ø. 3pl coconut-pile-again-appl-3sg.obj c.

Niu se hesulu-uyo-i-di. coconut 3pl pile-again-appl-3pl.obj ‘They piled the coconuts again.’

216

Anna Margetts

3.2.5. -ko suffix test The perfect suffix -ko marks the final word boundary on verbs. The -ko suffix test is most useful to V-N type incorporation because, like the complex verb test, it shows that the noun is part of the verb rather than occupying a position following it, since the suffix attaches to the incorporated noun, as in (23a). Example (23b) demonstrates that the suffix cannot intervene between the verb stem kabi ‘touch/make’ and the noun stem maketi ‘market’. The example in (c) shows that the noun maketi cannot follow the verb plus -ko suffix, even if the verb is transitive, and (d) shows that the verb kabi is transitive when it occurs without incorporation. (23)

a.

Se kabi-maketi-ko. 3pl touch-market-prf ‘They already prepared their market goods.’

b. *Se kabi-ko maketi. 3pl touch-prf market c. *Se kabi-ya-ko maketi. 3pl touch-prf market d.

Se kabi-ya-ko 3pl touch-3sg.obj-prf ‘They touched/grabbed it already.’

Like the complex-verb test, the -ko suffix test can establish a verb’s morphological transitivity status. If the verb is intransitive the perfect suffix attaches directly to the compound stem, as is the case in (23a). If a verb carries the word-final zero object suffix, adding the -ko suffix triggers the non-final object suffix -ya, as in (d). 3.2.6. Nominalization test The nominalization test shows that the verb stem and the incorporated noun behave like a morphological unit and a semantically unitary concept. In this test, which applies to both N-V and V-N constructions, the compound stem is nominalized as a unit by a possessive classifier, by the morpheme tau ‘man/person’, or by kaba ‘place/instrument’, as in (24) to (26).7 7. It is not obvious whether tau and kaba are prefixes or independent nouns. While tau can occur as an independent nominal ‘man’ and as a modifier ‘male’, kaba only ever occurs with a following verb stem which specifies the use of the place or instrument.

Noun incorporation in Saliba

(24)

a.

Ka niu-pulisi. 1excl coconut-husk ‘We husked coconut.’

b.

Yo-ma niu-pulisi. clf1-1excl.poss coconut-husk ‘Our coconut husking.’

217

The noun tau ‘man/person’ derives agent nouns from verbs. (25)

a.

Se kabi-kabole. 3pl make-sago ‘They made sago.’

b.

Tau kabi-kabole se lage. person make-sago 3pl arrive ‘The sago makers arrive.’

The stem kaba ‘place/instrument’ derives a noun denoting a location where or an instrument with which the activity expressed by the verb stem takes place. (26)

a.

Ka niu-tutu. 1excl coconut-pound ‘We pound coconuts.’ (Fishing 01BQ 0670)

b.

Kaba niu-tutu. place coconut-pound ‘place or instrument for pounding coconuts.’

After this overview of structural tests for incorporation, I now turn to the issue of the ordering of stems within the lexical compounds. 3.3. N-V versus V-N incorporation Saliba type I incorporation is attested with two different internal structures: the ordering of stems can be N-V or V-N, as shown in examples (4) and (5) above. I will call the position of the noun pre-nuclear vs. post-nuclear, reflecting its position with respect to the verb stem. Most incorporating verbs allow only one of the two orderings. Incorporation into the pre-nuclear slot is attested with a larger number of verbs (and is the only order allowed with external possession incorporation), but examples of post-nuclear incorporation are also common in that they occur especially with high-frequency verbs. For the few verbs which allow both orderings, the different order of stems in some cases correlates with a difference in meaning. Overall, the same semantic and pragmatic constraints

218

Anna Margetts

apply to N-V andV-N constructions: the incorporated nouns are patients and they specify and narrow the scope of the expressed activity. The difference in order is an idiosyncratic feature of certain verbs and the order of stems is not semantically or grammatically predictable. It has to be learned for each individual verb, just like the fact whether the verb allows incorporation at all. As most incorporating verbs take the noun stem in the pre-nuclear position and examples of this type are provided throughout the paper, I will focus on post-nuclear incorporation in the remainder of this section. Seven verbs are so far attested which may incorporate into the post-nuclear position. They do not constitute a formal or semantic class but the stems share that they are high-frequency lexical items. The list of roots is given in (27).8 (27)

he-kai kabi kai kaibwada kaiheya koi numa

‘feed (cause-eat)’ ‘touch, make, hold’ ‘eat’ ‘ask for, beg’ ‘play’ ‘hit, cut’ ‘drink’

Of this list, kai ‘eat’, kaiheya ‘play’ and kaibwada ‘ask for’ can only occur with V-N ordering (section 3.3.1.); numa ‘drink’ is essentially attested with N-V ordering but some speakers also accept V-N ordering; and kabi ‘touch/make’, koi ‘hit’ and he-kai ‘feed’ are attested with both N-V and V-N ordering resulting in semantic differences, as I discuss in 3.3.2. In section 3.3.3 I briefly look at cases which resemble V-N incorporation but which are in fact better analysed as cases of complementation. 3.3.1. Verbs which only allow V-N ordering The verb kai ‘eat’ only allows V-N type incorporation. It is attested with several different incorporated nouns, including puwaka ‘pork’, laisi ‘rice’ and simsim ‘watermelon’. In (28a) the noun puwaka is incorporated following the verb stem. The reverse order of stems is ungrammatical as shown in (28b). The V-N order is only allowed with incorporation, not with regular phrasal objects.9 8. The fact that most of the verbs are k-initial appears to be coincidental. Apart from he-kai ‘feed’ all verbs in the list are monomorphemic. 9. The examples with phrasal objects in (c) and (d) are given with the perfect suffix -ko on the verb (which triggers the non-zero allomorph of the third singular object suffix) because otherwise examples (a) with incorporation and (d) with a phrasal object could not be distinguished.

Noun incorporation in Saliba

(28)

219

a.

Se kai-puwaka. 3pl eat-pig ‘They ate pork.’

b. *Se puwaka-kai. 3pl pig-eat ‘They ate pork.’

c.

Puwaka se kai-ya-ko pig 3pl eat-3sg-prf ‘They already ate the pig.’

d. *Se kai-ya-ko puwaka 3pl eat-3sg-prf pig ‘They already ate the pig.’

A text example where simsim ‘water melon’ is incorporated is presented in (29). (29)

Sola, ye kai-simsim baguna! still 3sg eat-watermelon first ‘Wait, she’ll eat watermelon first!’ (Emadial122)

Similarly, the stem kaiheya ‘play’ may only incorporate with V-N order. The incorporated noun denotes the game that is played. In both attested examples the incorporated noun is an English loan. (30)

Se kaiheya-bolo. 3pl play-ball ‘They play ball.’ (Mouse3 02DA 0008)

(31)

Se kaiheya-mabolo. 3pl play-marble ‘They played the marble game.’ (Mosel 1994: 33)

The stem kaibwada ‘ask for’ can incorporate the theme, i.e. what is asked for. In (32) laisi ‘rice’ is incorporated, in (33) it is the noun moni ‘money’. (32)

Se kaibwada-laisi. 3pl ask.for-rice ‘They asked for rice.’

(33)

Se kaibwada-moni. 3pl ask.for-money ‘They asked for money.’

In analytical constructions, the applicativized transitive stem kaibwada-i ‘ask for’ can take either the theme or the addressee of the request as its object but in contrast to the theme, the addressee cannot be incorporated. 3.3.2. Verbs which allow V-N and N-V ordering The verb stem numa ‘drink’ normally shows N-V incorporation, as in (34a) and (35a). But some speakers allow both N-V and V-N ordering and accepted

220

Anna Margetts

the examples in (a) and (b). Other speakers rejected the (b) examples. For the speakers who allowed incorporation into either position there was no difference in meaning. (34)

a.

Ta ti-numa. 1incl tea-drink ‘We drank tea.’

b.

?Ta numa-ti. 1incl drink-tea ‘We drank tea.’

(35)

a.

Se gulai-numa. 3pl soup-drink ‘They drank soup.’

b.

?Se numa-gulai. 3pl drink-soup ‘They drank soup.’

There are also three verbs which can occur with either N-V or V-N ordering and for which the order of stems reflects a semantic difference. The verb kabi ‘touch/make’ allows both N-V and V-N type incorporation. A range of different nouns may appear in the post-nuclear position as shown in (36) to (39). In all of these examples, the reversal of stems is unacceptable. (36)

Se kabi-kabole. 3pl touch-sago ‘They made sago.’

(37)

Se kabi-maketi. 3pl touch-market ‘They prepared their market goods.’

(38)

Se kabi-numa. 3pl touch-house ‘They built a house.’

(39)

Ye kabi-noi. 3sg touch-nest ‘She made a nest.’

Only a single noun stem is attested in the pre-nuclear slot, shown in (40). Again, the reversal of stems is unacceptable. The expression mata-kabi is lexicalized to mean to wash one’s face. (40)

a.

Ya mata-kabi. 1sg eye-touch ‘I wash my face.’

b.

*Ya kabi-mata. 1sg touch-eye

The difference between the constructions with kabi ‘touch/make’ lies, firstly, in the fact that in (40a) the incorporated nouns is a body part and the construction is an instance of external possession incorporation. Secondly, the verb translates with a slightly different meaning in the two constructions: as ‘build’, ‘make’, ‘prepare’ or ‘produce’ in the V-N examples but as ‘wash’ or ‘touch’ when the order is N-V (although washing one’s face may also be considered preparing it). A semantic difference is also found with koi ‘hit’. The musical instruments bwayatu ‘kundu drum’and gita ‘guitar’only occur in the post-nuclear slot, while

Noun incorporation in Saliba

221

the noun kaiwa ‘tree/wood’ only occurs in the pre-nuclear position. Consider the examples in (41) to (43) none of which may show the reversed order of stems (example (41) is repeated from (5) above). (41)

Se koi-gita. 3pl hit-guitar ‘They played guitar.’

(43)

Se kaiwa-koi. 3pl tree-hit ‘They cut trees.’

(42)

Se koi-bwayatu. 3pl hit-kundu.drum ‘They played kundu drum.’

In both cases the incorporated noun denotes the patient of the expressed activity but the nature of the activity is different. The N-V compound in (43) refers to the activity of tree-cutting while the V-N compounds in (41) and (42) refer to playing a musical instrument. The final example of a stem which allows incorporation into either position is different in nature. For the derived causative stem he-kai ‘feed (cause-eat)’ both N-V and V-N order is only allowed with a single noun stem, puwaka ‘pig’. For this verb, the different incorporation positions correspond to different semantic roles of the object noun. In analytic constructions, he-kai ‘feed’ may occur as the head of a ditransitive clause with two objects, a recipient (the causee) and a theme. The noun puwaka ‘pig’ can be incorporated into the verb with either role, recipient or theme, but into different positions (and not simultaneously). Consider the examples in (44) and (45) where the pre-nuclear noun in (44) (repeated from 6a above) is a recipient, the post-nuclear noun in (45) is a theme. (44)

Se puwaka-he-kai. 3pl pig-caus-eat ‘They raise pigs.’ (i.e. they habitually feed pigs)

(45)

Kwabuli se he-kai-puwaka-Ø. widow 3pl caus-eat-pig-3sg.obj ‘They fed the widow pork.’

The two examples differ morpho-syntactically in that in (44) the recipient of the feeding event is incorporated and the resulting verb is intransitive. In (45), the incorporated noun denotes the theme of the feeding event and the resulting verb is morphologically transitive, with the recipient cross-referenced on the verb. This is not transparent because of the zero allomorph of the third singular object suffix but it can be established by the transitivity tests discussed in 3.2. Example (45) is a lexicalized expression referring to a custom where a widow is released from her duties to her in-laws (she is given pork to smell but is

222

Anna Margetts

not actually fed). If the order of stems in (45) is changed to N-V, as in (46), the interpretation can only be one where the widows are the agent and the incorporated noun denotes the recipient of the feeding event, parallel to (44) above, rather than to (45). (46)

Kwabuli se puwaka-he-kai. widow 3pl pig-caus-eat ‘The widows feed/raise pigs.’

The origin of the two kinds of internal orderings within Saliba type I incorporation is not obvious. Of interest is possibly the historical word order change when ancestor languages to the Papuan Tip cluster shifted from Proto Oceanic VO as basic word order to OV, presumably under Papuan influence (Bradshaw 1982; Ross 1988). In Oceanic, as well as cross-linguistically (cf. Mardirussian 1975), languages with VO word order typically incorporate into the post-nuclear position, while languages with OV as basic order tend to show incorporation into the pre-nuclear slot. Given this background, V-N type incorporation could possibly constitute a reflex of the historical VO word order while N-V ordering could be an innovation following the word order change to OV. However, both noun positions can host relatively recent English loanwords and the V-N compounds are not fossilized. A further problem with such a historical explanation is that one would expect to find similar patterns among related languages. However, to date, variations in the ordering of stems in incorporation have not been reported for any other language of the area. Those languages of the Papuan Tip cluster for which there is some grammatical description seem to have no incorporation at all, but compounds involving a verb and a noun stem tend to have N-V order, as attested, e.g., in Tawala (Ezard 1997: 100) and Iduna (Huckett 1974: 67). The Papuan Tip cluster is one of the less well described groups within Oceanic and further grammatical descriptions will possibly shed more light on this issue. 3.3.3. Post-verbal complements versus V-N incorporation Even though Saliba SOV word order is rather rigid, a noun’s position following the verb stem is not sufficient proof for its incorporated status. There are three verbs, henuwa ‘like’, gado ‘want’, and hemala ‘become’, which allow or require objects to occur in post-verbal position. These constructions are not instances of incorporation since the nouns do not form a single unit with the verb. They are morpho-syntactically independent. Consider the sentences in (47) and (48). The examples in (a) resemble V-N incorporation but, as shown in (b) by the modifier test, the nouns are not incorporated. The nouns in these constructions can take any modifier, including determiners and demonstratives. In (47a) the verb stem henuwa ‘want/like’ is followed by the object ti. In (b), the possessive

Noun incorporation in Saliba

223

classifier kadi ‘their’ intervenes between the verb and the noun stem. In (48a) the stem gado ‘want’ is followed by the object simsim ‘watermelon’. In (48b) the object is modified by the demonstrative teina ‘this’. While henuwa ‘want/like’ allows object nouns to precede or follow, gado ‘want’ can only take objects in postverbal position. (47)

(48)

a.

Se henuwa ti.10 3pl like/want tea ‘They want tea.’

b.

Se henuwa ka-di ti. 3pl like/want clf2-3pl.poss tea ‘They want their tea.’

a.

Se gado simsim. 3pl want watermelon ‘They want watermelon.’

b.

Se gado teina simsim. 3pl want prox.dem watermelon ‘They want this watermelon.’

The constructions in (47) and (48) can be analyzed as reduced complement clauses. Complement clauses generally follow the verb and the nouns following henuwa ‘like/want’ and gado ‘want’ can be extended to full clauses, as in (49) and (50). (The complementizer bena is not obligatory.) (49)

Se henuwa (bena) ti se numa. 3pl like/want comp tea 3pl drink ‘They want to drink tea.’

(50)

Se gado (bena) ti se numa. 3pl want comp tea 3pl drink ‘They want to drink tea.’

Like gado ‘want’, the stem hemala ‘become’ allows objects only postverbally and not in the canonical position preceding the verb. In (51a) the verb is followed 10. Note that none of the morphological transitivity tests discussed in 3.2. can clarify whether or not henuwa ‘like’ carries a zero object suffix when it is followed by a noun or complement clause. No overt object suffix (e.g. third person plural) is allowed in such constructions and none of the other tests apply. The same holds for constructions with gado ‘want’ and hemala ‘become’. For simplicity, I write these forms without an object suffix even though the absence of an object suffix has not been established beyond doubt.

224

Anna Margetts

by the place name Kwato. As demonstrated in (51b), the modifier test shows that, unlike incorporated nouns, the object following hemala ‘become’ can be modified (but it is in fact Kwato modifying tau). (51)

a.

Ya hemala Kwato. 1sg become Place.Name ‘I became Kwato.’ (FirstBoat 01CK 0072)

b.

Ya hemala Kwato tau-na. 1sg become Place.Name man-3sg.poss ‘I became a Kwato man.’

3.4. Intransitive base verbs In a number of cases, Saliba lexical compounding occurs with intransitive base verbs. As intransitive verbs, these lexemes should not have an object in their argument structure which could be incorporated. Consider (52) to (54) which show the underived intransitive verbs in (a), and the incorporating verbs in (b). (52)

a.

Se wase. 3pl search ‘They searched.’

b. Se sada-wase. 3pl betelnut-search ‘They searched for betelnuts.’

(53)

a.

Se kuma. 3pl plant ‘They planted.’

b. Se kwateya-kuma. 3pl yam-plant ‘They planted yams.’

(54)

a.

Se deula. b. Se koya-deula. 3pl terrace 3pl garden-terrace ‘They made terraces.’ ‘They terraced a garden.’

The incorporated nouns are semantically the patients of the activity expressed by the verb. The verbs are members of a class of intransitive verbs which can derive transitive stems by means of the applicative suffix. When the verbs are transitivized, the incorporated objects can occur as cross-referenced object arguments in analytic clauses, as in (55) to (57). (55)

Sada se wase-nei-di. betelnut 3pl search-appl-3pl.obj ‘They searched for the betelnuts.’

Noun incorporation in Saliba

(56)

Kwateya se kuma-i-di. yam 3pl plant-appl-3pl.obj ‘They planted the yams.’

(57)

Koya se deula-i-Ø. garden 3pl terrace-appl-3sg.obj ‘They terraced the garden.’

225

The class of intransitive verbs which can be applicativized divides into two subclasses: those which take a remote object as their applied object, i.e. an object with a peripheral semantic role such as location, addressee or concomitant, and those which take a close object, with the semantic role of patient or theme. (For the notion of close and remote object cf. Pawley 1973, 1986; Pawley and Reid 1980.) The intransitive verbs which are attested to incorporate belong to the subclass which take close objects as their applied objects. These intransitive verbs can be analyzed as semantically transitive and as having a semantic object argument which may or may not surface morpho-syntactically (cf. Margetts 2007). For example, they behave like transitive verbs in serialization and trigger the transitive version of final stems like uyo ‘do again’ (cf. the complex verb test discussed in 3.2.4. above). Verbs of this class are also attested in the discord constructions mentioned in section 1 above, as in (58) to (60) where intransitive verbs co-occur with backgrounded but morpho-syntactically independent objects. (58)

Sada se wase. betelnut 3pl search ‘They searched for betelnut.’

(59)

Kwateya se kuma. yam 3pl plant ‘They planted yams.’

(60)

Koya se deula. garden 3pl terrace ‘They terraced gardens.’

Morphologically, the underived stems are clearly intransitive but distributionally, they share characteristics with transitive verb stems. It seems that noun incorporation is among the constructions which are sensitive not only to the syntactic but also to the semantic arguments of the verb. Cross-linguistically, cases are also attested where an incorporated noun does not correspond to an argument of the verb in an analytic clause but to an adjunct

226

Anna Margetts

with a semantic role other than patient or theme (see the references in section 4 below). In Saliba, there are examples of this but this type of incorporation is not productive. In (61) the noun keda ‘path/way’ is incorporated into the verb baguna-i ‘go ahead of ’ (derived by the applicative suffix from the stem baguna ‘be first’). The incorporated noun denotes a location. In (62) keda ‘path/way’ appears in a postpositional phrase following the verb. (61) Taumana ya keda-baguna-i-di. visitor 1sg way-go.first-appl-3pl.obj ‘I lead the visitors the way.’ (62) Taumana ya baguna-i-di keda unai. visitor 1sg go.first-appl-3pl.obj path pp.sg ‘I lead the visitors on the way.’ A similar example is presented in (63) (repeated from (18) above) where the noun koya ‘garden’ is incorporated. In (64) it occurs as an adjunct introduced by a postposition. (63)

Se koya-koya-paisowa. 3pl red-garden-work ‘They were working in the garden.’ (BudoiNualele 01CY 018)

(64)

Se paisowa koya unai. 3pl work garden pp.sg ‘They worked in the garden.’

4. Lexical and syntactic accounts of incorporation There has been ongoing discussion in the linguistic literature about the nature of noun incorporation. The fundamental question which lies at the center of the discussion is whether noun incorporation is a lexical or a syntactic process. The lexical approach has been argued most prominently by Mithun (1999, 1984, 1986) but has also been supported by a range of other authors (e.g. Sapir 1911; DiSciullo and Williams 1987; Rosen 1989; Vel´azquez-Castillo 1995a, 1995b; Evans 1997). The lexical approach basically treats incorporation as a type of compounding, that is, as a word formation process rather than as syntactic generation of speech. The compound stems constitute new lexical items which are stored in the mental lexicon. Evidence in support of this claim is that, crosslinguistically, such constructions classify as single, morphologically complex verbs by language-internal criteria. In addition, unlike for the on-line production

Noun incorporation in Saliba

227

of sentences, speakers tend to show some awareness of creating novel combinations when they utter such noun-verb compounds (Mithun and Corbett 1999). Among the advantages of a lexical approach is that it successfully accounts for the lexicalization of the compound stems. The most prominent representative of the syntactic approach is Baker (1988, 1995, 1996) but syntactic analyses of incorporation have also been proposed by Postal (1962) and Sadock (1980, 1985, 1986) among others. Following Baker’s account, noun incorporation is a purely syntactic process and it “proves to be no more than the result of applying standard movement transformation to words rather than to full phrases” (Baker 1988: 1). In particular, he suggests that “the generalized transformation Move-Alpha applies to the head noun of a noun phrase, adjoining it to the verb and leaving behind a coindexed trace . . . ” (1995: 6). In support, he cites evidence that the incorporated noun can be specific and referential, that the noun can be modified by “stranded modifiers”, and that incorporation constructions can be paraphrased by corresponding analytic clauses.11 A further point which he brings forward in support of his analysis is that: This syntactic account of NI has the further advantage of accounting for one of the most important properties of noun incorporation: that fact that objects incorporate but subjects do not. (Baker 1995: 7)

A major criticism of Baker’s treatment of noun incorporation is that it cannot account for the fact that, cross-linguistically, incorporated nouns tend to be nonspecific and non-referential, and that, for example, proper nouns are generally not incorporated at all. In addition, the syntactic account fails to explain the phenomenon of lexicalization, i.e. that incorporation constructions and their analytic counterparts are often not semantically equivalent. A further problematic fact is that, in some languages, certain verbs must incorporate and that there are thus incorporation constructions which have no analytic counterparts. (Although Baker 1988 deals with this by arguing for morphological subcategorization frames of verbal roots which may either allow or require incorporation.) Besides this, describing incorporation as a result of the generalized transformation Move-Alpha is argued to make the wrong predictions about which nouns can be incorporated into the verb: this transformation would allow only objects of transitive verbs and subjects of unaccusative (patient-oriented) verbs to be incorporated. (The theory-internal claim is that only arguments that are sisters to the V can be incorporated but not those which are sisters to the VP or which are headed by a preposition.) But, while it is clearly a cross-linguistic tendency that incorporation is restricted to such cases, incorporation of adjuncts, subjects 11. See Evans’ discussion of Mayali (1997: 401, 1996) in response to this claim and also Rosen (1989).

228

Anna Margetts

of unergative (agent-oriented) verbs, and even subjects of transitive verbs has been reported from a variety of languages (cf. Sapir 1911; Allen et al. 1984; Sasse 1984; Axelrod 1990; Wilhelm 1992; Polinsky 1993; Spencer 1995; Evans 1997; Cook and Wilhelm 1998).12 While incorporation in Saliba largely follows cross-linguistically well attested patterns, there are certain aspects of the data which present a challenge for a syntactic analysis a` la Baker. The Saliba examples of morphologically intransitive base verbs incorporating their semantic objects, as discussed in 3.4, constitute a problem along the same lines as the cross-linguistic examples of incorporation of adjuncts and subjects of unergative and transitive verbs. Other challenging aspects of the Saliba data are the interaction between incorporation and complex verbs, and the different slots in which incorporated nominals can occur. As discussed in 3.2, compound stems resulting from incorporation can enter into complex verb constructions by nuclear-layer serialization. Consider the examples in (65) to (67) (and also (21) above) where an incorporating verb occupies the initial slot followed by another verb stem in a complex verb. Together the stems build a single inflected word. In example (67) the incorporating stem is transitivized by the final stem sagu-i ‘help’ in the verb series (not all final stems have to agree in transitivity with the initial stem). (65)

Se niu-pulisi-gehe. 3pl coconut-husk-finished ‘They finished husking coconut.’

(66) Ye kwateya-kuma-uyo. 3sg yam-plant-again ‘He planted yams again.’ (67) Ya nui-tutu-sagu-i-Ø. 1sg coconut-pound-help-appl-3sg.obj ‘I help him pound copra (into a bag).’ The fact that complex verb stems and noun incorporation can be combined in this way suggests that these are processes of a similar nature. Both produce morphologically complex verb stems and the semantics of both may or may not be derived from the sum of their parts, depending on the degree of lexicalization of the compound stem. The fact that, in Saliba, noun incorporation and complex verbs interact in the way they do suggests that either both processes are syntactic or both are lexical. The derivation of complex verbs by nuclear-layer serialization 12. Baker’s solution to incorporation of nouns which are marked by adpositions in the corresponding analytic constructions is assuming that they appear without the adposition in the D-structure.

Noun incorporation in Saliba

229

is, however, typically discussed as a lexical process and is considered as such in Baker’s (1989) account of verb serialization. (Baker 1988 considers at least some such cases as multiple incorporation, i.e. V-V incorporation followed by Nincorporation. Lexicalization may pose a problem in some cases for an analysis with this ordering of processes.) The final challenge in the Saliba data for a syntactic approach are the two positions into which a noun can be incorporated, preceding or following the verb stem, as discussed in 3.3. This may prove problematic for an approach in which incorporating verbs are derived from analytic constructions by syntactic transformation. The choice into which slot a verb stem can incorporate is not predictable on grammatical or semantic grounds. For a few verbs which allow incorporation into either slot there is a meaning distinction between the resulting constructions but the semantic distinctions are idiosyncratic and different for each verb. There is no pattern that would allow us to predict which position the noun will take (e.g. based on semantic role). It is a feature which needs to be learned and which suggests that it is ultimately a lexical matter.

5. Conclusion Saliba contributes some interesting examples to the discussion on noun incorporation. It has external possession (Margetts 1999) and lexical compounding constructions, but also a construction featuring morphologically intransitive verbs with phrasal objects (Margetts 2008). In this paper I presented an overview of lexical compounding, which in Saliba can feature N-V but also V-N compound stems. In both cases the constructions underlie the same semantic and pragmatic constraints and the verb and the noun stem form a morphological unit, i.e. a complex verb stem, as can be seen from a number of morpho-syntactic tests. The variation in the order of stems is an unusual feature for incorporation constructions cross-linguistically and is not described for other languages of the Papuan Tip cluster or the wider Oceanic language group. The position of the incorporated noun is an idiosyncratic feature of the verb, which is not predictable on formal or semantic grounds. Saliba also features incorporation with morphologically intransitive base verbs. Such verbs would not normally be expected to incorporate, yet they incorporate what can be analyzed as a semantic object argument. The Saliba data present a challenge to some treatments of noun incorporation suggested in the literature and appear to support a lexical rather than syntactic account of incorporation.

230

Anna Margetts

References Allen, Barbara J., Donna Gardiner and Donald Frantz 1984 Noun incorporation in SouthernTiwa. International Journal ofAmerican Linguistics 52: 388–403. Chicago, IL: The University of Chicago Press. Axelrod, Melissa 1990 Incorporation in Koyukon Athabaskan. International Journal of American Linguistics 65: 179–95. Chicago, IL: The University of Chicago Press. Baker, Mark C. 1988 Incorporation: a Theory of Grammatical Function Changing. Chicago: University of Chicago Press. Baker, Mark C. 1989 Object sharing and projection in serial verb constructions. Linguistic Inquiry 20: 513–553. Cambridge, MA: Massachusetts Institute of Technology Press. Baker, Mark C. 1995 Lexical and nonlexical noun incorporation. In: Urs Egli, Peter E. Pause, Christoph Schwarze, Arnim Stechow and G¨otz Wienold (eds.), Lexical Knowledge in the Organization of Language, 1–34. (Amsterdam Studies in the Theory and History of Linguistic Science, Series 4: Current Issues in Linguistic Theory 114.) Amsterdam, Philadelphia: John Benjamins. Baker, Mark C. 1996 The Polysynthesis Parameter. New York: Oxford University Press. Ball, Douglas 2005a Phrasal Incorporation in Tongan. Proceedings of the XII Conference of the Austronesian Formal Linguistics Association. Los Angeles, CA: University of California. Ball, Douglas 2005b Tongan Noun Incorporation: Lexical Sharing or Argument Inheritance? Proceedings of the 12th International Conference on Head-Driven Phrase Structure Grammar, University of Lisbon. Bradshaw, Joel 1982 Word order change in Papua New Guinea Austronesian languages. Unpublished Ph.D. dissertation, University of Hawai’i, Honolulu. Chung, Sandra and William A. Ladusaw 2004 Restriction and Saturation. Cambridge, MA: Massachusetts Institute of Technology Press. Cook, Eung-Do and Andrea Wilhelm 1998 Noun incorporation: new evidence from Athabaskan. Studies in Language 22: 49–82. Amsterdam, Philadelphia: John Benjamins. DiSciullo, Anna Maria and Edwin Williams 1987 On the Definition of Word. Cambridge, MA: Massachusetts Institute of Technology Press.

Noun incorporation in Saliba

231

Evans, Nicholas 1996 The syntax and semantics of body part incorporation in Mayali. In: Hillary Chappell and William McGregor (eds.), The Grammar of Inalienability, 65–109. Berlin: Mouton de Gruyter. Evans, Nicholas 1997 Role or Cast? Noun incorporation and complex predicates in Mayali. In: Alex Alsina, Joan Bresnan and Peter Sells (eds.), Complex Predicates, 397–430. Stanford: CSLI Publications. Ezard, Bryan 1997 A Grammar of Tawala: an Austronesian Language of the Milne Bay Area, Papua New Guinea. (Pacific Linguistics C-137.) Canberra: Pacific Linguistics. Foley, William and Robert D. Van Valin, Jr. 1985 Information packaging in the clause. In: Timothy Shopen (ed.), Language Typology and Syntactic Description, 282–364. Cambridge: Cambridge University Press. Fran¸cois, Alexandre 2002 Araki, a Disappearing Language of Vanuatu. (Pacific Linguistics 522.) Canberra: Pacific Linguistics. Gerdts, Donna B. 1998 Incorporation. In: Andrew Spencer and Arnold M. Zwicky (eds.), The Handbook of Morphology, 84–100. Blackwell Handbooks in Linguistics. Oxford: Blackwell. Giv´on, Tom 1990 Syntax. A Functional-Typological Introduction. Amsterdam, Philadelphia: John Benjamins. Heath, Jeffrey 1976 Antipassivization: a functional typology. Berkeley Linguistics Society 2: 202–211. Berkeley, CA: Berkeley Linguistics Society. Hopper, Paul and Sandra Thompson 1980 Transitivity in grammar and discourse. Language 56-2: 251–299. Huckett, Joyce 1974 Notes on Iduna grammar. In: Alan Healey (ed.), Workpapers in Papua New Guinea Languages, 63–133. Ukarumpa: Summer Institute of Linguistics. Lichtenberk, Frantisek 1982 Individuation hierarchies in Manam. In: Paul Hopper and SandraThompson (eds.), Studies in Transitivity (Syntax and Semantics) 15: 261–76. New York, San Francisco, London: Academic Press. Mardirussian, Galust 1975 Noun incorporation in universal grammar. In: Papers from the 11th Regional Meeting of the Chicago Linguistic Society, 383–389. Chicago, IL: Chicago Linguistic Society.

232

Anna Margetts

Margetts, Anna 1999 Valence and Transitivity in Saliba, an Oceanic Language of Papua New Guinea. (MPI Series in Psycholinguistics 12.) Nijmegen: Max Planck Institute for Psycholinguistics. Margetts, Anna 2004 Core-layer junctures in Saliba. In: Isabelle Bril and Fran¸coise OzanneRivierre (eds.), Complex Verbs and Serialization in Oceanic Languages, 69–89. (Empirical Approaches to Language Typology.) London, New York: Mouton de Gruyter. Margetts, Anna 2007 Close and remote objects in a language with a single transitive suffix. In: Jeff Siegel, John Lynch and Diana Eades (eds.), Language Description, History and Development: Linguistic Indulgence in Memory of Terry Crowley, 69–78. Amsterdam and Philadelphia: John Benjamins. Margetts, Anna 2008 Transitivity discord in some Oceanic languages. Oceanic Linguistics 47-1: 31–44. Honolulu: University of Hawaii Press. Massam, Diane 2001 Pseudo Noun Incorporation in Niuean. Natural Language and Linguistic Theory 19: 153–97. Dordrecht: Springer. Miner, Kenneth L. 1986 Noun stripping and loose incorporation in Zuni. International Journal of American Linguistics 52: 242–54. Chicago, IL: The University of Chicago Press. Miner, Kenneth L. 1989 A note on noun stripping. International Journal of American Linguistics 55: 476–477. Chicago, IL: The University of Chicago Press. Mithun, Marianne 1984 The evolution of noun incorporation. Language 60: 847–94. Mithun, Marianne 1986 On the nature of noun incorporation. Language 62: 32–37. Mithun, Marianne and Greville Corbett 1999 The effect of noun incorporation on argument structure. In: Lunelle Mereu (ed.), Boundaries of Morphology and Syntax, 49–72. Amsterdam: John Benjamins. Mosel, Ulrike 1994 Saliba. (Languages of the world/Materials 31.) M¨unchen, Newcastle: LINCOM Europa. Pawley, Andrew 1973 Some problems in Proto Oceanic grammar. Oceanic Linguistics 12: 103– 188. Honolulu: University of Hawaii Press. Pawley, Andrew 1986 A reanalysis of Fijian transitive constructions. Te Reo 29: 81–112.

Noun incorporation in Saliba

233

Pawley, Andrew and Lawrence Reid 1980 The evolution of transitive constructions in Austronesian. In: Paz B. Naylor (ed.), Austronesian studies: Papers from the Second Eastern Conference on Austronesian Languages, 103–130. Ann Arbor: The University of Michigan Center for South and Southeastern Asian Studies. Polinsky, Maria 1993 Subject inversion and intransitive subject incorporation. In: Katherine Beals, Gina Cooke, David Kathman, Sotaro Kita, Karl-Erik McCullough and David Testen (eds.), Papers from the Annual Regional Meeting of the Chicago Linguistic Society 29-1: 343–362. Chicago, IL: Chicago Linguistic Society. Postal, Paul M. 1979 Some Syntactic Rules of Mohawk. New York: Garland Press. Rosen, Sara 1989 Two types of noun incorporation: a lexical analysis. Language 65: 294– 317. Ross, Malcolm 1988 Proto Oceanic and the Austronesian Languages of Western Melanesia. (Pacific Linguistics C-98.) Canberra: Pacific Linguistics. Sadock, Jerrold 1980 Noun incorporation in Greenlandic. Language 62: 300–319. Sadock, Jerrold 1985 Autolexical Syntax: a proposal for the treatment of noun incorporation and similar phenomena. Natural Language and Linguistic Theory 3: 379–440. Dordrecht: Springer. Sadock, Jerrold 1986 Some notes on noun incorporation. Language 62: 19–31. Sapir, Edward 1911 The problem of noun-incorporation. American Anthropologist 13: 250– 283. Arlington, VA: American Anthropological Association. Sasse, Hans-J¨urgen 1984 The pragmatics of noun incorporation in Eastern Cushitic languages. In: Frans Plank (ed.), Objects: Towards a Theory of Grammatical Relations, 243–268. New York, London: Academic Press. Spencer, Andrew 1995 Incorporation in Chukchi. Language 71: 439–489. Sugita, Hiroshi 1973 Semitransitive verbs and object incorporation in Micronesian languages. Oceanic Linguistics 12: 393–407. Honolulu: University of Hawaii Press Vel´azquez-Castillo, Maura 1995a Noun incorporation and object placement in discourse: the case of Guaran´ı. In: Pamela Downing and Michael Noonan (eds.), Word Order in Discourse, 555–579. Amsterdam, Philadelphia: John Benjamins.

234

Anna Margetts

Vel´azquez-Castillo, Maura 1995b Noun incorporation in Guaran´ı: a functional analysis. Linguistics 33: 673–709. Wilhelm, Andrea 1992 Noun incorporation in Northern Athabaskan. MA thesis, Calgary: University of Alberta.

Noun-phrase conjunction in Austronesian languages: additive, inclusory and comitative strategies Isabelle Bril

Introduction1 In the past two decades, cross-linguistic evidence of asymmetrical coordination at all levels (clauses, VPs or NPs), have led to revision of its definitions as a symmetrical type of conjunction, and to propose that the conjuncts are actually syntactically asymmetrical, with the conjunctive marker (if any) heading one of the conjuncts (Van Valin 1997; Johannessen 1998). Cross-linguistically, additive, but also comitative and inclusory conjunction2 display cases of asymmetrical coordination. The first extensive analysis of inclusory constructions in Oceanic languages was authored by Lichtenberk (2000), who outlined a typology of such constructions, more specifically in Toqabaqita (Solomon Islands), elaborating on their phrasal and split subtypes. The present analysis will centre on noun-phrase conjunction in Austronesian languages and the Oceanic subgroup of this family, focusing on asymmetrical conjunction of the inclusory and comitative types, and delineating the syntactic and semantic parameters on the use of these conjunctive constructions, as well as their pragmatic effects. Section 1 is an overview of various conjunctive strategies and their distinctive features. Section 2 analyses the syntactic and semantic features of inclusory constructions and the constraints on their use. Section 3 is an analysis of the conjunctive morphemes, their functions and etymology. Section 4 sketches a typology of the strategies used in a sample of Austronesian languages for standard coordination, inclusory constructions, or comitative coordination, and the semantics of these conjunctive strategies.

1. I wish to thank two anonymous reviewers, as well as Claire Moyse-Faurie and Joachim Sabel for comments and suggestions which have helped improve this paper. 2. For comitative coordination, see Stassen 2000, 2001, 2003. Inclusory coordination is found in Mayan (Tzotzil), Australian, Tchadic (Hausa), Bantu (Nkore-Kiga), Finno-Ugrian (Hungarian, Finnish), Germanic (Old Norse, Old English), Slavic (Russian, Polish, Czech, etc.) languages (see Schwartz 1988, Aissen 1989, Singer 2001, Evans 2003, Moravcsik 2003, Faarlund 2004).

236

Isabelle Bril

1. Additive, comitative and inclusory conjunction: some distinctive features Let us briefly define the three conjunctive types under consideration: the additive, inclusory and comitative types. In additive NP coordination, the conjuncts are in a semantic relation of addition, and the conjunctive morphemes may also be used in counting (“and/plus”). Comitative coordination, on the other hand, makes use of a comitative morpheme, which also behaves as a coordinator (“with/and”) and triggers agreement in number with the conjuncts. The conjuncts of inclusory constructions stand in a relation of inclusion, typically between a superset pronoun which includes a subset NP in its reference. These three types are illustrated by examples (1) in Nˆelˆemwa: (1a) is a case of additive coordination with xa ‘and, plus’3 , (1b) is the standard NP coordination for animate conjuncts marked by the comitative coordinator ma, and (1c–d) are cases of inclusory conjunction. When syndetic (i.e. marked by a conjunctive marker), inclusory constructions often use a comitative coordinator (“with/and”) (1c–d). Nˆelˆemwa (vos, Eastern Oceanic, New Caledonia) – Additive coordination ye]. oda mwa [hlileny thaamwa xa (1) a. Hla u 3 PL pfv go.up act these.DU woman conj 3sg.fr4 ‘These two women and/plus him go up.’ (Bril 2002: 428) – Standard NP coordination (with a comitative coordinator) b. Hli ulep [dagiiny ma axaleny Teˆa Paak]. 3du go.out messenger conj this.m Teˆa Paak ‘The messenger and Teˆa Paak go out.’ (Bril fieldnotes)

3. Compare with tujic xa bwaat pwagiik ‘eleven’ (lit. ten and/plus top one) (Bril 2002: 428). 4. Abbreviations follow the leopzig glossing rules and in addition the following are used: act actualiser; af actor focus; anaph anaphoric; aor aorist; ass assertive; assoc associative; bf buffer; coll collective; conj conjunctive mkr; conn connector; coord coordinator; dir directional; do direct object; dx1 proximal deictic; dx2 medial deictic; emph emphatic mkr; fr free (pronoun); genr general tam marker; id identification mkr; idx indexer; n.fut non-future; n.vis non visual; part particle; pc paucal; p.clf possessive classifier; pers.mkr personal marker; prosp prospective; rev reversive; rl realis; red reduplication; seq sequential; spec specifying preposition; sr same reference; transl translative aspect mkr; tri trial; unc uncertain mood; vent ventive; virt virtual.

Noun-phrase conjunction strategies in Austronesian languages

237

– Non-phrasal inclusory construction pe-vhaa ma axomoo-n. c. Ma 1 DU.excl recp-speak conj mother-poss.3sg ‘We and his mother speak to each other.’ (Bril fieldnotes) – Phrasal inclusory construction ma Polie. d. Yaman 1 DU.excl.fr conj Polie ‘Polie and I.’ (Bril fieldnotes) The inclusory construction in (1c) is non-phrasal since the dual subject index ma (indicating the superset) is part of the VP5 and does not make up a conjunctive phrase of the [NP and NP] type with the subset axomoo-n. In (1d) on the other hand, the inclusory construction is phrasal, the superset free6 pronoun yaman and the subset NP are a conjunctive phrase [NP and NP]. In both cases, the reference to the subset is included in the superset (an index pronoun in (1c) or a free pronoun in (1d)), it is not added up to it (see ex. (26) below for further analysis). In Nˆelˆemwa, conjunction is syndetic; in other Austronesian languages, these three conjunctive types (additive, inclusory, comitative) may be syndetic or asyndetic; some languages may even use both strategies, varying with syntactic constraints or semantic properties (detailed in sections 1.1 & 1.2 below). Example (2a) in Aji¨e illustrates asyndetic inclusory construction and (2b) syndetic standard (additive) coordination. Aji¨e (Eastern Oceanic, New Caledonia, A. Boehe, p.c.) (2)

a.

Paul. G¨ovu 1 DU.excl.fr Paul ‘Paul and I.’ (asyndetic inclusory construction, lit.: we2 Paul)

b.

G`enyˆa ma7 Paul. 1 SG.fr conj Paul ‘Paul and I.’ (syndetic additive coordination)

The semantics of additive and comitative coordination is slightly different: additive coordination has a distributive (“each”) or a collective reading, while comitative coordination tends to have a collective reading only. 5. This is often referred to as “verb-marking strategy” in the literature on this question. 6. Free pronouns are full pronominal forms, with NP status, as opposed to personal indexes which belong to the VP. 7. Ma in Aji¨e is an additive coordinator which never has the inclusory function that similar coordinators have in northern New Caledonian languages (such as Nˆelˆemwa or Zuanga).

238

Isabelle Bril

1.1. NP additive coordination Various types of additive coordination and their semantic properties will now be analysed, including the summation pronoun strategy. 1.1.1. Asyndetic vs. syndetic coordination In some languages, syndetic and asyndetic coordination are two possible options, with syntactic and semantic correlates. For instance, asyndetic coordination may be specific to NPs with human reference or expressing greater semantic proximity between nominals. In Nakanai, the asyndetic strategy is commonly (but not obligatorily) used when the NP conjuncts are in the discourse topic position of the sentence, as in (3a–b). But when the NP conjuncts are in the subject position (3c–d), the coordinator me is then obligatory. Nakanai (Western Oceanic, Meso-Melanesian cluster, New Britain) (3)

a.

[E Marisa e latu-la], egira goata e Malalia. art Marisa art child-3poss 2du go.up art Malalia ‘Marisa and her child, they’ve gone up to Malalia.’ (Johnston 1980: 185)

b.

egira goata e Malalia. [E Marisa (me)8 latu-la], art Marisa conj child-3poss 2du go.up art Malalia ‘Marisa and her child, they’ve gone up to Malalia.’ (Johnston 1980: 185)

c.

La bolo me la paia ogala-le. art pig conj art dog at.outside-there ‘The/a pig and the/a dog are outside there.’ (Johnston 1980: 185)

d.

E tau me rutu ilua. art man conj wife two ‘A man and his wife.’ (Johnston 1980: 186)

The syndetic strategy may be reinforced by a dual numeral, as in (3d). 1.1.2. Summation free pronoun strategy vs. coordinator Another frequent conjunctive strategy in Austronesian languages is to use a summation free pronoun referring to all conjuncts as in (4, 5). In most languages, 8. The coordinator me assimilates the noun marker e.

Noun-phrase conjunction strategies in Austronesian languages

239

this is restricted to higher animates, or to humans with close relationship, as in Tetun or Kaulong. Other types of NPs are conjoined by coordinators.9 Tetun-Fehan (svo, Central Malayo-Polynesian, West Timor) (4)

Fu’a [a ina-n sia ruas a ama-n] b´a. waken 2sg mother-gen 3pl.fr two 2sg father-gen imp ‘Wake up the two of them, your mother and father.’ (van Klinken 1999: 157)

Kaulong (svo, Western Oceanic, North New-Guinea cluster, Southwest New Britain) (5)

Hiang ngong titie-n. 3sg.m 3sr.DU wife-3sg ‘He and his wife.’ (Throop & Ross, in Lynch et al. 2002: 396)

In Manam or Kairiru, the summation pronoun strategy is also restricted to higher animates with close relationship. Other nouns are either juxtaposed or conjoined by a standard coordinator, be ‘and’in Manam (6b), bo in Kairiru (7c) (inanimates are juxtaposed in Kairiru). In Manam, the coordinator be is optional with nouns referring to humans and with tight relationship. Note that the position of the summation free pronoun is not always medial; in Manam, it comes after the conjuncts (6a): Manam (sov, Western Oceanic, North New-Guinea cluster, Manam and Boesa Islands, PNG) (6)

a.

di-l´a/o-ru. [Tam´oata r´oa-ø d´ı-a-ru] man spouse-3sg 3 PL-bf-DU.fr 3 PL.rl-go-DU ‘The man went with his wife.’ [lit.: man spouse they2 they went] (Lichtenberk 1983: 430)

b.

mo´an-(be) a´ ine di-p´ura. man-and woman 3pl.rl-arrive ‘The men and the women arrived.’ (Lichtenberk 1983: 365)

9. See examples (60) in Tetun. In Kaulong, the medial coordinator u ‘and’ conjoins NPs or VPs; the comitative coordinator to conjoins animate NPs, see examples (52).The coordinator ma conjoins VPs, si ‘and, but, and then, and so’ is used for clause coordination (Throop & Ross, in Lynch et al. 2002: 399, 408).

240

Isabelle Bril

The summation pronoun is medial in Kairiru (7a); but when more than two conjuncts are involved, it appears only once, between the first two conjuncts, the other NP conjuncts are juxtaposed (7b): Kairiru (sov, Western Oceanic, North New-Guinea cluster, East Sepik, PNG) (7)

a.

¯ Cawor]. Cuok [Smowai ru boundary Smowai 3 DU.fr Cawor ‘Smowai and Cawor’s boundary.’ (Wivell & Ross, in Lynch et al. 2002: 209)

b.

¯ awo-˜n natu-˜n. Wocul rı Wocul 3 PL.fr spouse-3sg child-3sg ‘Wocul, his wife and children.’ (Wivell & Ross, in Lynch et al. 2002: 208)

c.

moin bo nat ‘woman and children’ (Wivell & Ross, in Lynch et al. 2002: 208)

The summation free pronoun is also medial in Mangap-Mbula (8) or Teanu (9): Mangap-Mbula (svo, Western Oceanic, North New Guinea cluster) (8)

zin pikin] am-sa yom [Nio niam-Na-n 1sg 1PL . EXCL-nmlz-poss.3pl pl child 1PL . EXCL-wait 2pl ma som. and do.in.vain ‘I together with the children waited for you in vain.’ (Bugenhagen 1995: 212)

Teanu (svo, Central Eastern Oceanic, Vanikoro, Solomon Islands) (9)

mwalik’ iape. Emele da 3sg.poss woman 3 DU man ‘a woman and her husband.’ (Fran¸cois p.c.)

In M¯aori or Tahitian, the summation pronoun strategy is restricted to conjoining NPs referring to humans (represented by personal names or common nouns). The inclusory construction is required to conjoin a pronoun and a noun (see also (24, 25) below).

Noun-phrase conjunction strategies in Austronesian languages

241

M¯aori (Eastern Polynesian, New Zealand) (10)

tana m¯okai]. Ka noho nei [a Ponga r¯aua ko tam sit near art Ponga 3 DU spec his slave ‘Ponga and his servant went on sitting.’ (Bauer 1997: 548)

Tahitian (Eastern Polynesian) (11)

’o Terii]. ’Ua m¯omo’e [’o Tama r¯aua pfv disappear(pl) id T. 3 DU.excl id Terii ‘Tama and Terii disappeared.’ (Lazard & Peltzer 2000: 179)

In Tahitian, common nouns may also be conjoined by the standard coordinator ’e ‘and’, or by the comitative coordinator m¯a ‘and, with’. Summation pronouns are often reanalysed as coordinators or comitative markers (Bril 2004: 526–527). This mostly (but not exclusively) occurs with 3rd person pronouns. In Mwotlap, only k¯oy¯o, the 3rd person dual inclusory pronoun, is reanalysed as a coordinator of NPs referring to humans with a tight semantic link (Fran¸cois, 2000: 389–390). Compare the inclusory construction (12a) and the reanalysed coordinator k¯oy¯o (12b). Mwotlap (svo, Eastern Oceanic, Banks Is., North Vanuatu) (12)

a.

Imam k¯oy¯o it¯ok father 3du be.good ‘Are father (and him) well?’ [lit.: father theydual well] (Fran¸cois 2000: 386)

b.

Mayanag k¯oy¯o e¯ gn¯o-n. chief 3du spouse-3sg ‘The chief and his wife.’ (Fran¸cois 2000: 389)

In Aji¨e, the occurrence of a non-inclusory subject pronoun (na ‘he’ in (13)), in what was formerly an inclusory construction (wak`e curu y`ev`u [lit.: work they2 chief]) has triggered the reanalysis of the dual pronoun as a comitative marker (Bril 2004: 526–528). Aji¨e (svo, Eastern Oceanic, South New Caledonia, A. Boehe, p.c.) (13)

Na wak`e curu y`ev`u. 3 SG work 3du chief ‘He works with the chief.’

242

Isabelle Bril

1.1.3. Summation free pronoun strategy plus a conjunctive marker The summation pronoun strategy may be reinforced by a comitative coordinator, as in Atayal (Formosan) or Samoan (15). In Atayal, conjunction with the comitative coordinator ci/ or with laha/ ci/ (the summation pronoun and the comitative coordinator respectively) is restricted to animate entities (14a–b). Otherwise, the standard coordinator lu/ ‘and’, which is compatible with all types of NPs (± animates, see (14c)), is required. Inclusory constructions, which are used to conjoin a pronoun and a noun, are presented in examples (50) below. Atayal (Formosan) (14)

a.

ci/ Nyaw]. M-pa-siliy-cu [ci/ hu|il laha/ af-irr-beat-1sg.nom acc dog 3 PL . FR COM cat ‘I will beat the dog and the cat.’ (Huang 2006: 228)

b.

m-awas [ka/ naßkis laha/ ci/ ule/. Nyel-ø10 prog.prox af-sing nom old.man 3 PL . FR COM child ‘The old man is singing here with the child.’ (Huang 2006: 224) [af = actor focus]

c.

Cyel ma-sa-siliy [ka/ Watan lu/ Temu]. prog.dist recp-red-beat nom Watan and Temu ‘Watan and Temu are beating each other (with sticks).’(Huang 2006: 226)

Samoan (Polynesian) (15)

Fai le mafaufau tele faiva o [Peni laua ma do not think much fishing.trip poss Peni 3 DU and Ruta]. Ruta ‘Peni and Ruta did the fishing trip rather thoughtlessly.’ (Mosel & Hovdhaugen 1992: 680)

1.1.4. Associative plural constructions A slightly different strategy, sometimes called “associative plural constructions”, is instanciated by a plural pronoun referring to an indefinite group of humans and meaning “and the rest, and followers, and companions”. In Papuan Malay, this construction is mostly used with subject and object arguments and is less acceptable with obliques; thus, *(b) is ungrammatical. 10. The 3rd person (sg. or pl.) bound nominative pronoun on the VP is zero-marked.

Noun-phrase conjunction strategies in Austronesian languages

243

Papuan Malay (svo, Eastern Province of Papua11 ) (16)

a.

Kitong pi lia Yohannes dong. 1pl go see Yohannes 3pl ‘We went to see Yohannes and his mates.’ [lit.: Yohannes them] (Donohue forthc.)

b. *Sa-su-masa banyak untu sa-pu-ana dong. 1sg-pfv-cook much for 1sg-poss-child 3pl [Intended meaning: ‘I’ve already cooked a lot for my daughter and her friends.’] Toqabaqita (Central Eastern Oceanic, Malaita, Solomon Islands) (17) Tha Gerea kera. art.pers Gerea 3pl.fr ‘Gerea and them.’ (Lichtenberk 2000: 29) In Nakanai, the associative plural pronoun fuses the coordinator me and a nonsingular pronoun, yielding the plural mite (me+ite 3pl), or the dual mira (me+ira 3du). Nakanai (Western Oceanic, Meso-Melanesian cluster, New Britain) (18)

a.

E Tubu, mite. art Tubu and.3pl ‘Tubu and the rest.’ (Johnston 1980: 186)

b.

E Tubu, mira. art Tubu and.3du ‘Tubu and the other fellow.’ (Johnston 1980: 186)

c.

rutu-la. E Tubu, mira me e art Tubu and.3du and art wife-3poss ‘Tubu together with his wife.’ (or:) ‘Tubu and his wife.’ (Johnston 1980: 186)

Note that while (18a) might be interpreted as a standard additive coordination, (18b) is clearly inclusory, since the dual pronoun mira includes “Tubu” in its reference. Compare with standard coordination in Nakanai (3). In (18c), the presence of the coordinator me suggests that mira has been reanalysed as a collective-associative marker.

11. Formerly Irian Jaya.

244

Isabelle Bril

In Nˆelˆemwa, the suffix -ma is an associative plural marker referring to members of a group or fellows. Nˆelˆemwa (Eastern Oceanic, New Caledonia) (19)

Cawi-ma ma Dahoot-ma. Cawi-assoc.pl and Dahoot-assoc.pl ‘The Cawi clan and the Dahoot clan.’ (Bril fieldnotes)

1.2. Inclusory constructions In contrast with additive coordination, inclusory constructions comprise a superset NP (generally marked by a dual or plural pronoun), which includes the reference to the subset NP. These constructions may be ± syndetic, ± phrasal. They are always restricted to higher animates with tight relationship. The syntactic constraints and semantic properties of inclusory constructions, in comparison with additive or standard coordination, will be analysed in section 2. 1.2.1. Asyndetic inclusory constructions 1) Phrasal construction with free pronoun strategy In the phrasal construction involving a superset free pronoun and an included subset NP, the superset free pronoun is the head of the construction and the included subset stands as its specifier. In Aji¨e (20), the first person dual pronoun curu ‘we2’ is the verb’s subject index, while the second occurrence of curu belongs to the inclusory construction, it is the superset pronoun which includes the subset “maternal uncle” in its reference. Aji¨e (Eastern Oceanic, New Caledonia, Leenhardt 1932: 190) (20)

[curu ka≠a χ i-´]. Curu bOri we vi-t´v´ 3 DU seq pvf recp-speak 3 DU maternal.uncle poss-3sg ‘Then they talked, him and his maternal uncle.’ (or:) ‘Then he talked with his maternal uncle.’ [lit.: theydual talked theydual maternal uncle].

2) Non-phrasal or verb-marking strategy Asyndetic, non-phrasal inclusory constructions (labelled “split inclusory” by Lichtenberk 2000) are illustrated by (21a–b) in Toqabaqita. The dual subject indexes meki or mere ‘we’ also mark tense-aspect on the verb and do not make up a conjunctive phrase with the subset NPs mature woman in (a), or brother in (b). In some approaches, this is sometimes called verb-marking strategy. Yet,

Noun-phrase conjunction strategies in Austronesian languages

245

though the subject index is part of the VP, it includes the other NP conjunct in its reference and thus behaves referentially as the superset. Toqabaqita (Central Eastern Oceanic, Malaita, Solomon Islands) (21)

a.

lae uri Honiara qana [ø kukeqe-nau] meki at mature.woman-1sg 1 DU.excl.fut go to H. wiki loo. week that.up ‘My wife and I will go to Honiara next week.’ (Lichtenberk 2000: 21)

b.

ngata. [ø doqora-mu mere brother-2sg 1 DU.excl.nonfut speak ‘Your brother and I spoke (to each other).’ [lit.: your brother wedual spoke] (2000: 21)

The examples (21a–b) stand in contrast with the phrasal construction in (21c), in which the superset free pronoun kamareqa ‘we2’ makes up a conjunctive phrase with the subset “brother” included in its reference. The dual subject index meki marks future and number agreement on the verb. c.

lae ma-i. [Kamareqa doqora-ku] meki 1 DU.excl.fr brother-1sg 1 DU.excl.fut go dir-at ‘I and my brother will come tomorrow.’[lit.: wedual my brother wedual spoke] (Lichtenberk 2000: 3)

When comparing the non-phrasal (21a–b) and phrasal constructions in Toqabaqita (21c), the non-phrasal construction contains a potential position whose gap is symbolised by Ø, and which is filled by an inclusory free pronoun (such as kamareqa ‘we2’in (21c)) in the phrasal construction, not by a 1st person singular pronoun “I”. The same two constructions, non-phrasal (21d) and phrasal (21e) are also found with object arguments. d.

Kera thaitoqoma-mareqa wela-nau. child-1sg 3pl know-1 DU.excl ‘They know me and my child.’ (Lichtenberk 2000: 22)

e.

maka-nau]. Wane e laba-taqi [kamareqa man 3sg.n.fut harm-tr 1 DU.excl.fr father-1sg ‘The man harmed me and my father.’ (Lichtenberk 2000: 14)

Many languages display the non-phrasal pattern, Papuan Malay for instance.

246

Isabelle Bril

Papuan Malay (svo, Eastern Province of Papua) (22) Maria kitong-dua su-tara-bisa lae. pfv-neg-able again Maria 1 PL -two ‘Maria and I can’t do it anymore.’ [lit.: Maria wedual . . . ] (Donohue forthc.) On the other hand, Tetun displays the phrasal pattern. Tetun-Fehan (svo, Central Malayo-Polynesian, West Timor) ruas P´ak T´om b´a Laran. (23) Ami 1 PL.excl two Mr Tom go Laran ‘We two Mr Tom (and I) went to Laran.’ (van Klinken 1999: 157) In Tahitian (24) and M¯aori (25), inclusory constructions are only used to conjoin pronouns and nouns referring to humans; see (ex. 10–11) for other conjoining strategies. Tahitian (Eastern Polynesian) ’o Peu. (24) ’Ua reva a¯ tu m¯aua pfv leave away 1 DU.excl id Peu ‘Peu and I left.’ (Tryon, 1970: 67; in Payne 1985: 35) M¯aori (Eastern Polynesian) (25) Kei te aha k¯orua ko Tame? tam what 2 DU spec Tame ‘What are you and Tame doing?’ (Bauer 1997: 548) 1.2.2. Syndetic inclusory constructions In other languages, such as Nˆelˆemwa and many northern languages of New Caledonia (Nyelˆayu, Nemi, Bwatoo, C`emuhˆı, etc.), inclusory constructions are syndetic12 and involve a conjunctive morpheme, often a comitative conjunction. – Phrasal vs. non-phrasal constructions In Nˆelˆemwa, the phrasal construction (26a) with a superset free pronoun (yaman ‘we2’) and an included subset is emphatic, while the non-phrasal construction 12. While inclusory constructions or their traces tend to be asyndetic in southern languages (Aji¨e, Tˆırˆı (Osumi 1995), etc.)

Noun-phrase conjunction strategies in Austronesian languages

247

(26b) is neutral. In (26b), the subject index ma ‘we2’ includes Polie in its reference. The potential position symbolised by ø in (26b) cannot license a pronoun standing for only one of the subset conjuncts (a 1st person singular pronoun would thus be ungrammatical), it only allows an inclusory free pronoun as in (26a). Inclusory constructions are thus syntactically obligatory to conjoin a pronoun with a noun; no other construction is available in Nˆelˆemwa. The NP headed by the conjunctive ma in (26b) is a conjunct, not an adjunct, nor an afterthought. The notion of collective action may be stressed by an adverb wuung ‘together’, as in (26c). Nˆelˆemwa (vos, Eastern Oceanic, New Caledonia) tu haga [yaman ma (26) a. Io ma fut 1DU.excl go.down fish 1DU.excl.fr conj Polie]. (emphatic) Polie ‘We’ll go fishing, Polie and I.’ (Bril 2000) b.

tu haga Ø ma Polie. (neutral) Io ma conj Polie fut 1DU.excl go.down fish ‘Polie and I will go fishing.’ [lit.: wedual will go fishing including Polie] (Bril fieldnotes)

c.

Hˆa shaya wuung ma co. 1pl.incl work together conj 2sg.fr ‘We and you work together.’ (Or:) ‘We work together with you.’ (Bril fieldnotes)

Non-contiguity results from the fact that subject indexes with superset reference (ma ‘we2’, or hˆa ‘we’ in (26b–c)) must be in pre-verbal position, while nominal arguments and free pronouns are post-verbal. Compare with standard coordination (26d), in which ma conjoins two animate NPs. d.

I tu mwa uya shi [hlileny kˆaaˆ ma-n ma 3sg go.down act arrive at these.2 father-his conj axomoo-n]. mother-his ‘He arrives at his father and mother’s place.’ (Bril fieldnotes)

1.2.3. Inclusory interrogative constructions The subset may be an interrogative pronoun with argument function; the same ± syndetic and ± phrasal constructions are found.

248

Isabelle Bril

1) Asyndetic phrasal Manam (sov, Western Oceanic, North New-Guinea cluster, PNG) (27)

N´aita /´an-ru? who? 2 PL.fr-DU ‘Who are you with?’ (lit.: who you2?) (Lichtenberk 1983: 431)

Toqabaqita (Central Eastern Oceanic, Malaita, Solomon Islands) (28) Tei keeroqa? who? 3 DU.fr ‘Who (is) he with?’ (Lichtenberk 2000: 28) Mwotlap (Eastern Oceanic, Banks Is., North Vanuatu) (29) K¯omy¯o y¯e Wilson ma-van me? – Kamy¯o 2 DU who? pfv-go vent 1 DU.excl Wilson ‘Who did you come with?’ – ‘With Wilson.’ (Fran¸cois 2000: 391) 2) Phrasal vs. non-phrasal syndetic constructions The phrasal conjunctive construction in sentence-initial position in (30a) is a cleft, emphatic construction; the superset free pronoun yamon ‘you2’ includes the interrogative pronoun in its reference. On the other hand, the non-phrasal construction (30b) is pragmatically neutral, the 2nd person dual subject index mo (30b) also includes the interrogative pronoun in its reference. Nˆelˆemwa (Eastern Oceanic, North New Caledonia) (30)

a.

Yamon ma ti me mo tu yhalap? 2 DU.FR coord who? conj 2 DU go.down gather.food ‘Who is it that you went food-gathering with?’ [lit.: (it’s) you and/ with whom that you went food-gathering?] (Bril 2002: 265)

b.

Mo perui ma ti? 2du.excl meet conj who? ‘Who have you met?’ [lit.: youdual met including whom?] (Bril fieldnotes)

Tolai also displays a syndetic phrasal inclusory construction.

Noun-phrase conjunction strategies in Austronesian languages

249

Tolai (svo, Western Oceanic, Meso-Melanesian cluster, New Britain) (31) Amur ma ia? 2 DU.fr conj who? ‘Who is with you?’ [lit.: youdual including whom?] (Mosel 1984: 41) 1.2.4. Inclusory possessive constructions Superset pronouns may be expressed by possessive pronouns in asyndetic (32–33) or syndetic (34–35) constructions. The possessive pronoun includes the reference to the other conjoined possessor. Mwotlap (Eastern Oceanic, Banks Is., North Vanuatu) (32) Inti-mamy¯o Wilson. son-1 DU.excl Wilson ‘It’s our son (with) Wilson.’ (or:) ‘It’s the son of Wilson and me.’ (Fran¸cois 2000: 479) M¯aori (Eastern Polynesian, New Zealand) (33) Kei whea t¯a k¯orua moenga ko t¯o t¯ane? at where your.DU bed spec your husband ‘Where is your and your husband’s bed?’ (Bauer 1997: 549) Zuanga/Yuanga (Eastern Oceanic, North New Caledonia) (34) Mw˜o-[bi ma a˜ baa-nu]. house-poss.1 DU.excl conj brother-poss.1sg ‘It’s my and my brother’s house.’ (Bril fieldnotes 2006) Samoan (Polynesian) (35) ’i te fa’atoaga a [l¯a’ua ma l-o-na tin¯a. loc art plantation poss 3 DU conj art-poss-3sg mother ‘. . . to the plantation belonging to her mother and herself.’ (Mosel & Hovdhaugen 1992: 680)

2. Syntactic and semantic constraints on inclusory constructions Several questions crop up at this point. Do inclusory constructions result from some elliptical pronoun or referent? Are such constructions optional? What are the constraints on inclusory constructions?

250

Isabelle Bril

2.1. Obligatory inclusory strategy: syntactic constraints on the type of conjuncts In Oceanic languages, inclusory constructions usually obey a number of syntactic constraints, and have specific semantic properties and pragmatic effects related to topicality. Inclusory constructions, when obligatory, do not result from constituent ellipsis, but rather, from a syntactically obligatory null instantiation, as discussed in §2.1.1. about Nˆelˆemwa. 2.1.1. Constraint on pronominal conjuncts and person hierarchy In Nˆelˆemwa, only nouns may be phrasally conjoined as in (36a), pronouns may not (*you and me), nor may nouns be conjoined with pronouns (*she and her girl), such conjuncts trigger obligatory inclusory constructions as in (36b), which is the neutral inclusory construction. In (36b), the conjunct position marked by ø must remain empty if represented by a pronominal conjunct, hence the notion of “obligatory null instantiation”. Only an inclusory free pronoun is admitted in this position, as in (36c), yet the reference to the non-instantiated pronoun is included in the verb’s subject index (the verb-marking strategy) acting as the superset pronoun. The emphatic construction would involve an inclusory free pronoun, as in (36c). Nˆelˆemwa (Eastern Oceanic, North New Caledonia) (36) a. Hli u peeva [Teˆa Pwayili ma horaamwaleny Kaavo. 3du pfv argue Teˆa Pwayili conj this.woman Kaavo ‘Teˆa Pwayili and this woman Kaavo argue.’ (Bril fieldnotes) b.

Hli u peeva Ø ma horaamwaleny Kaavo. 3du pfv argue conj this.woman Kaavo ‘He and this woman Kaavo argue.’ [lit.: theydual argue including this woman Kaavo]

c.

ma horaamwaleny Kaavo. Hli u peeva hli Kaavo 3du pfv argue 3du.FR conj this.woman ‘They argue, he and this woman Kaavo.’ (Bril fieldnotes)

The superset pronoun abides by person and referential hierarchy: 1st includes 2nd person, which includes 3rd person. According to such hierarchy, “I and you” yields a “we2 incl. you” form; “I and him” yields “we2 excl. him” and “you and him” yields a “you2 him” form. No difference of inclusory pattern was observed for 1st , 2nd or 3rd persons. In contrast with languages which restrict inclusory constructions to 1st and 2nd person pronouns, 3rd person inclusory pronouns are

Noun-phrase conjunction strategies in Austronesian languages

251

also allowed, possibly because pronouns have distinct forms for dual, sometimes trial, and plural forms, thus preventing any referential ambiguity. As already pointed out, in some Northern New Caledonian languages such as Nˆelˆemwa or Nemi (37), inclusory constructions are obligatory for conjuncts involving a noun and a pronoun (there is no available alternative): Nemi (Eastern Oceanic, North New Caledonia) am me ti-nga-me ma Hixe. (37) a. Yelu li 3DU virt ass together go.down-rev-prox and/with Hixe ‘She’ll go back with Hixe.’ (Ozanne-Rivierre 1979: 124) b.

daama. me hen ma-m13 Yelu li 3 DU virt together go and/with-def chief ‘He’ll go with the chief.’ (Ozanne-Rivierre 1979: 124)

On the other hand, Southern languages such as Aji¨e, have an alternative construction: a noun and pronoun may be conjoined by an inclusory construction or by a standard additive coordinator (as in (1a–b)), though the latter construction might be a more recent development. The constraints bearing on pronominal conjuncts is a widespread feature of languages with inclusory constructions: conjunction of pronouns ranges from dispreferred (in Russian for instance, McNally 1993) to ungrammatical (Nˆelˆemwa). Though widespread, the ban on pronominal conjuncts is not universally criterial for inclusory constructions; various Oceanic languages, among which Mangap-Mbula (8), southern New Caledonian languages (38), Samoan (39), Manam (43a), Toqabaqita (44a), Fijian (45), also allow pronominal conjuncts. Xˆarˆac`uu` (Eastern Oceanic, South New Caledonia) (38) Mwˆaaˆ =g`e mˆe d`ee` ri Mˆıaˆ . house=2sg.fr coord people Mˆıaˆ ‘Your (sg.) and the people of Mˆıaˆ ’s house.’ (Moyse-Faurie 1995: 30) Samoan (Polynesian) (39) . . . ou te le alofa ia te [’oe ma tamaiti]. 1sg genr neg love loc 2sg.fr coord children ‘. . . [that] I do not care for you and the children.’ (Mosel & Hovdhaugen 1992: 680) 13. The clitic -m is a singular definite nominal determiner.

252

Isabelle Bril

2.1.2. Higher animacy hierarchy Inclusory constructions are most generally restricted to humans or higher animates with close relationship as in Mwotlap for instance, while inanimates require other types of coordination or comitative adjunct strategies. No exceptions have been found in the sample. Mwotlap (Eastern Oceanic, Banks Is., North Vanuatu) ¯ el ¯ el ¯ non? ¯ y¯e so van vonop Ent¯ (40) a. Ent¯ 1 TRI.incl.fr 1 TRI.incl.fr who? prosp go fishing Wilson. Wilson ‘Who are we2 going to fish with?’ [lit.: we3 who go fishing] ‘With Wilson.’ (or:) ‘Wilson and us.’ [lit.: we3 Wilson] (Fran¸cois 2000: 392) There are subtle semantic features relating to the position of the inclusory pronoun, before or after the subset NP (40b), signalling differences in topicality, anaphoricity and empathy, the first conjunct is more topical14 . b.

k¯oy¯o Wilson Wilson k¯oy¯o 3du Wilson Wilson 3du ‘(S)he and Wilson.’ ‘Wilson and the other.’ (Fran¸cois 2000: 390)

Other NPs are conjoined by the coordinators ba ‘and, but’ (40c), wa (40d), but also by the instrumental preposition mi (+ inanimates) or the comitative preposition tiwag mi ‘together with’ (± animates) which also has conjunctive function (40e), even for pronouns (40f). c.

NataN mi/ba naVasel? basket with/coord knife ‘basket and knife’ (Crowley, in Lynch et al. 2002: 592)

d.

imam wa ige tita. ige art.pl father and art.pl mother ‘the fathers and mothers’ (Fran¸cois 2002: 263)

e.

ige susu tiwag mi ige lililwo. art.pl small.one together with art.pl big.one ‘the children and the adults’ (Fran¸cois 2002: 263)

14. A similar phenomenon is described in Toqabaqita by Lichtenberk (2000: 9–10, 27).

Noun-phrase conjunction strategies in Austronesian languages

f.

253

Ino, tiwag mi n¯ek. 1sg together with 2sg ‘me and/with you’ (Fran¸cois 2002: 263)

Compare coordination in (40g) with a comitative adjunct in (40h), without number agreement: g.

Na-mango ¯ tiwag mi n¯o-w¯oh, k¯oy¯o v¯el¯es neneh art-mango together with art-coconut 3du only sweet v¯el¯es. only ‘The mango and the coconut are as sweet.’ (Fran¸cois 2002: 262)

h.

vanvan tiwag mi ige h¯ow en! N¯ek tog together with pl down dx 2 SG proh go ‘Do not go with the people down there (to the north)!’ (Fran¸cois 2002: 262)

2.2. Properties of inclusory constructions: sameness of syntactic function, semantic role and topicality In most languages, inclusory constructions appear with most syntactic functions (analysed in § 2.2.2. below). 2.2.1. Sameness of syntactic function, semantic role and topicality Like other types of conjuncts (additive or comitative), conjuncts in inclusory constructions must have the same syntactic function, independently from their morphological marking which may sometimes be different due to their asymmetrical head-specifier relation. They must also have similar semantic roles. A] New Caledonian languages For instance, the conjunct NPs are co-agents or co-patients, collective or reciprocal participants, or they display similar properties (41) (same age, size, etc.); they are joint possessors or participants with kinship and part-whole relationship, and they are included in the agreement pattern. Semantic asymmetry triggers syntactic hierarchy, which is marked by associative or comitative adjuncts, without number agreement.

254

Isabelle Bril

Zuanga/Yuanga (Eastern Oceanic, North New Caledonia) (41) Pe-kau-li ma Hiixe. recp-year-3du conj Hiixe ‘She and Hiixe are the same age.’ (Bril fieldwork 2007) A tentative construction such as *pe-kau-n ma Hiixe [recp-year-3sg], with a 3rd singular pronoun and an adjunct ma Hiixe, with the intended meaning “she’s the same age as/with Hiixe” would be ungrammatical. Various other New Caledonian languages, among them Bwatoo, C`emuhˆı15 and Aji¨e (Bril 2004: 509–511), show similar contrasts. Oceanic languages generally distinguish comitative conjuncts restricted to higher animate NPs and marked by an “and/with” coordinator (with number agreement), from associative adjuncts marked by a preposition “with”, without number agreement. They usually express differences in agency, animacy, and topicality. B] Nakanai In Nakanai, (± syndetic) inclusory coordination is obligatory with pronominal conjuncts (42a-b) and stands in contrast with a comitative adjunct marked by the preposition le, without agreement (42c): Nakanai (Western Oceanic, Meso-Melanesian cluster, New Britain) (42) a. Eia ge voku la luma [tamutou me la valalua 3sg irr work art house your.PL and art men taume]. your.SG ‘He will build a house for you (sg.) and your men.’ (Johnston 1980: 269) b.

me tabara-bara-gu], La baa taku sesele, [amite art area my true 1 PL . EXCL and brother-red-my isasa. [egite la valua isahari] la baa tamite art area our.pl.excl one 3 PL art men some ‘My section, belonging to me and my brothers and some others, that’s one section.’ (Johnston 1980: 273)

15. In C`emuhˆı (Rivierre, 1980: 194–95) the inclusory construction with the coordinator me triggers number agreement and is distinct from the noncoagentive preposition imi ‘with, carrying’, derived from imwi ‘seize, hold’, without number agreement (see Ozanne-Rivierre 2004).

Noun-phrase conjunction strategies in Austronesian languages

c.

255

Amite umala ge pou vikapopo le-gite 1pl.excl proh irr sit be.together with-3pl Siapanipani. Japanese ‘We were not to stay with the Japanese.’ (Johnston 1980: 35)

C] Manam In Manam, conjunction of nouns with pronouns is allowed and may be ± syndetic. Asyndetic noun-pronoun conjunction (43a) expresses different empathy and discourse prominence from an inclusory construction (43b), which puts more stress on the subset/ specifier NP standing in sentence initial position. Manam (sov, Western Oceanic, North New-Guinea cluster) to/´a-gu /´e-ru] /i-r´eba-ru. (43) a. [N´au 1SG.fr older.brother-1sg.poss 1excl.fr-DU 1excl-sail-DU ‘I and my older brother sailed.’ or ‘I sailed with my older brother.’ (Lichtenberk 1983: 430) b.

/i-reba-r´eba-ru. [To/´a-gu /´e-ru] older.brother-1sg 1excl.fr-DU 1excl.rl-sail-red-DU ‘My older brother and I were sailing.’ (Lichtenberk 1983: 275)

D] Toqabaqita In Toqabaqita, nouns and free pronouns may be conjoined with ma if equally topical (44a), while inclusory constructions such as (44b) express different empathy and discourse prominence between entities. Toqabaqita (Central Eastern Oceanic; Malaita, Solomon Islands) too (44) a. [Nau ma wela-nau ki] mili 1 SG.fr and child-1sg pl 1pl.excl(n.fut) stay siafaqa. be.not.well.off ‘I and my children, we are not well off.’ (Lichtenberk 2000: 8) b.

lae bii kamiliqa? [Kamaroqa tha Bita] moki art Bita 2 DU.fut go with 1pl.excl.fr 2 DU.fr ‘Will you (sg.) and Bita come with us?’ (Lichtenberk 2000: 9)

256

Isabelle Bril

E] Fijian In Fijian, the construction with or without the first person singular (yau) in (45) profiles arguments differently and carries different stress and different pragmatic effects. A phrasal coordination between a pronoun and a noun stresses the pronoun (yau). The unmarked construction excludes the pronoun (yau), yielding a non-phrasal inclusory construction. Boumaa Fijian (Central Pacific) (45) ’Eirau aa sota vata (o yau) ’ei Jone mai 1du.excl past meet together art 1sg and John at Viidawa. Viidawa ‘John and I met at Viidawa.’ (Dixon 1988: 157) Nˆelˆemwa displays similar constructions (26a–b) with similar stress on the inclusory free pronoun (26a), but there is one major difference, the free pronoun must be inclusory in Nˆelˆemwa. 2.2.2. Syntactic function: subject vs object function In most languages, inclusory constructions appear with most syntactic functions, as in Mangap-Mbula, Toqabaqita, Nakanai or Nˆelˆemwa. Some languages tend to prefer or to restrict them to subjects and co-agents and to use comitative adjunction for other syntactic functions. Mangap-Mbula (Western Oceanic, North New Guinea cluster) pa [niamru Bob uraata aN-pit (46) Nio ko 1 SG unc 1 SG-recount about 1 DU.excl Bob work tiam].16 LOC .1 PL .excl ‘I will tell a story about my work together with Bob.’ (Bugenhagen 1995: 114)

16. Dual forms only exist for nominative pronouns, otherwise only sg vs pl pronouns.

Noun-phrase conjunction strategies in Austronesian languages

257

Toqabaqita (Central Eastern Oceanic, Malaita, Solomon Islands) (47) Toqa loo kera ngata taqaa suli [kamareqa people that.up 3pl.n.fut speak be.bad about 1 DU.excl.fr wela nau]. child 1sg ‘The people up there speak badly about me and my child.’ (Lichtenberk 2000: 15) In Nakanai and Nˆelˆemwa, the inclusory construction is obligatory for direct objects. Nakanai (Western Oceanic, Meso-Melanesian cluster, New Britain) harua-la]. (48) Eau alalavi hilo [egira (me)17 e 1sg yesterday see 3 DU conj art husband-3poss ‘I saw her and her husband yesterday.’ (Johnston 1983: 185) A tentative construction of (48) with a 3rd singular pronoun would be ungrammatical: *Eau alalavi hilo eia (3SG) me e harua-la. In Nˆelˆemwa, inclusory constructions are also used with prepositional conjuncts. Nˆelˆemwa (Eastern Oceanic, North New Caledonia) (49) a. Kiya-a pe-mwang ni pe-wooxa[-va there.is.not-at.all recp-bad in recp-middle-poss.1 PL . EXCL ma Teˆa Belep]. conj Teˆa Belep ‘There is not the least conflict between us and Teˆa Belep.’ (Bril fieldnotes) b.

ma Kiya-a pe-pwa-wo shi[-vaak18 there.is.not-at.all recp-break-coll side-poss.1 PL .excl conj axamaleena Polo]. these.m.dx2 Polo ‘There is not the least reason of discord between us and the Polo clan.’ (Bril fieldnotes).

17. The coordinator me is optionally deleted when the NP is not in subject/topic function and position. 18. The possessive form -va is neutral, -vaak is emphatic

258

Isabelle Bril

3. Syndetic constructions: the status of the conjunctive morphemes Let us now turn to the syntactic status of the conjunctive morphemes used in syndetic constructions. 3.1. Etymology of the conjunctive morpheme There are two frequent etymological sources for the conjunctive morpheme used in the various types of NP conjunction: a coordinator or a comitative adposition, often (but not exclusively) expressing co-agentivity19 , and less frequently nouns meaning “friend, companion”, and verbs meaning “accompany, follow”. 3.1.1. Source: a comitative adposition The use of comitative conjunctions is frequent in Austronesian languages and is often restricted to animate NPs. In Atayal (Formosan) for instance, inclusory constructions use a comitative morpheme ci/ heading the subset noun. Constructions may be non-phrasal (50a) or phrasal (50b) with an inclusory free pronoun (cami). For other NP conjoining strategies using ci/ or laha/ ci/ (the summation pronoun + ci/), see (14). Atayal (Formosan) m-awas ci/ Temu. (50) a. Nyel-min prog.prox-1PL.excl.nom af-sing com Temu ‘I’m singing with Temu.’ (Huang 2006: 223) [af = actor affix]

19. In French, nous irons avec Michel (lit. we’ll go with Michel) is ambiguous: it generally refers to more than two participants with the standard adjunctive function and reading of avec, but if the context is clear, it may also refer to two participants (Michel and I, understood as a couple) with the inclusory reading. In Chilean Spanish (Schwartz 1988), the comitative preposition con ‘with’ may be used with an inclusory reading, but only for co-agentive animates with close relationship (as in a–b below). With inanimates (as in c), the inclusory reading is impossible, con only has the instrumental adjunct reading; with distantly related animates (d), the reading cannot be inclusory either and must be that of a comitative adjunct: a. fuimos a casa con mi madre ‘I went home with my mother’ [lit. we went home with my mother]; b. fuimos al mercado con mi perro [inclusory, 2 participants] ‘I went to the market with my dog’; c. fuimos al mercado con mi auto [instrumental reading]: ‘we went to the market with my car’ [impossible inclusory reading: *I went to the market with my car]; d. fuimos al mercado con el Mayor de la ciudad [comitative] ‘We went to the market with the Mayor of the city’.

Noun-phrase conjunction strategies in Austronesian languages

b.

259

m-awas [cami ci/ Temu]. Nyel-min prog.prox-1 PL.excl.nom af-sing 1 PL.excl.fr com Temu ‘I’m singing with Temu.’ (Huang 2006: 228)

In Takia (Oceanic), NPs with human reference are conjoined with the comitative conjunction da ‘and/with’ triggering plural agreement on the verb, whereas nonhuman NPs are simply juxtaposed. Takia (sov, Western Oceanic, North New-Guinea cluster, Madang province (PNG)) (51) Ab a oN w-abi ya-n [Nai Meit da] lo house dx2 2sg build rl-def 1sg.fr Meit COM 3.LOC.in mu-mado da. 1 PL . EXCL-stay ipfv ‘The house that you built, Meit and I are living in it.’ (Ross, in Lynch et al. 2002: 230) In Kaulong, the comitative conjunct headed by to and triggering number agreement is restricted to animates. Other (± inanimate) NPs may also be conjoined by the standard coordinator u ‘and’ (see footnote 6). Kaulong (Western Oceanic, North New-Guinea cluster, Southwest New Britain) to (52) Hiang me tik [po vio-n 3sg.m come call.name 3 PL father-in-law-3sg COM elaio-n]. mother-in-law-3sg ‘He came and uttered the names of his father-in-law and mother-in-law.’ (Throop & Ross, in Lynch et al. 2002: 405) 3.1.2. Source: a comitative coordinator with mV form Comitative coordinators exist cross-linguistically; they are found in many Slavic languages (Polish (z), Russian (s), Bulgarian), in Old Norse, in Hungarian, etc. In many Oceanic languages, comitative coordinators with a mV form are common; they may originate from former comitative verbs, reconstructed as Proto-Oceanic *ma-i or *ma-ni ‘be with’. They may head comitative conjuncts with agreement, and, in some languages, they also head comitative adjuncts without agreement, as in Tolai (Mosel 1984: 94), M¯aori, Niuean, East Futunan (Moyse-Faurie & Lynch 2004).

260

Isabelle Bril

In many Northern Vanuatu languages, mV comitative coordinators have conjunctive and inclusory functions. Hiw (Eastern Oceanic, Torres, North Vanuatu) mi Edga peon t¯o. (53) Kemare ¯ 1 DU.excl and/with Edga fut go.indiv ‘Edgar and I will go.’ (Fran¸cois p.c.) Lo-Toga (Eastern Oceanic, Torres, North Vanuatu) (54) Kemor mi 20 Edga t¨e o¯ r-v¯en? 1 DU.excl and/with Edga prosp aor.du.go ‘Edgar and I will go.’ (Fran¸cois p.c.) In M¯aori, the conjoining function of the comitative/associative preposition me ‘with’ appears in older texts and is thus not an innovation, but the conjoining function has increased in modern speech (Bauer 1997: 550). Me conjoins humans (with group or generic reference) and inanimates, in all syntactic functions (subjects, direct or oblique objects). The choice between the comitative coordinator me (55a-b) or a summation dual pronoun (55c) does not carry any semantic difference. M¯aori (Eastern Polynesian) me te t¯ut¯ua¯ ] e (55) a. E kore [te tangata teina younger com the low.born tam tam NEG the man tae atu ki reira noho ai. arrive away to there sit part ‘People of junior rank and low-born people could never go and sit there.’ (Bauer 1997: 547) b.

I haere atu [a Hone me Piri] ki te tam move away pers John com Bill to the tangihanga. tangi ‘John and Bill went to the tangi.’ (Bauer 1997: 552)

20. In Lo-Toga, mi has extended its conjunctive function to clause conjunction (Fran¸cois, p.c.).

Noun-phrase conjunction strategies in Austronesian languages

c.

261

Piri] ki te I haere atu [a Hone r¯aua ko tam move away art John 3 DU spec Bill to the tangihanga. tangi ‘John and Bill went to the tangi.’ (Bauer 1997: 552)

The comitative preposition me also expresses set inclusion in M¯aori: d.

Ng¯a t¯angata katoa i kata me te rangatira atu. the.pl people all tam laugh com the chief away ‘All the people, including the chief, laughed .’ (Bauer 1997: 219)

e.

Kua tika [katoa ng¯a kai me ng¯a m¯ıti atu]. tam right all the.pl food com the.pl meat away ‘All the food, including the meat, should be ready now.’(Bauer 1997: 219)

3.1.3. Source: a noun “companion, fellow” Another source may be nouns meaning ‘companion, fellow’ grammaticalised into a comitative marker. This is found for instance in many languages of North Vanuatu (Bank Islands); the marker is a reflex of a Proto-North Vanuatu etymon *tua ‘fellow’ (Fran¸cois p.c.). Mwerlap (Eastern Oceanic, Banks Islands, North Vanuatu) i Edga s-van. (56) To-k COM -1 SG art Edgar aor-go ‘Edgar and I will go.’ (Fran¸cois p.c.) In Dorig, ta- appears in an inclusory construction (57a), or functions as a comitative coordinator (57b). Dorig (Eastern Oceanic, Banks Islands, Gaua Island, North Vanuatu) i s¯e s-van? (57) a. Kmur ta-n¯ 2 DU COM-2sg art who prosp-go ‘Who will you (sg) go with?’ [lit.: you2 your companion who will go?] (Fran¸cois p.c.) b.

i tma kmur. I vve kmur ta-n¯ art mother 2du COM-3 SG art father 2du ‘your father and your mother’ (Fran¸cois p.c.)

262

Isabelle Bril

In Vur¨es, the comitative marker (58a) has been further grammaticalised as an instrumental preposition tek with inanimate nouns (58b). Vur¨es (Eastern Oceanic, Banks Islands, North Vanuatu) i Womtelo ma-van me. (58) a. (K¯om¯or¯ok) te-k pfv-go hither 1 DU.excl COM-1sg art W. ‘I came with Womtelo.’ [lit.: (we2) my-fellow Womtelo came here] (Fran¸cois p.c.) b.

vin m¯ot¯o. N¯e m¯o-b¯ol no tek o 3 SG pfv-hit 1 SG ins art skin palm-tree ‘He hit me with the bark of a palm-tree.’ (Fran¸cois p.c.)

3.1.4. Source: a verb “accompany, be with” In various Austronesian languages, these comitative markers are verbs or deverbal morphemes, reanalysed as coordinators triggering number agreement, such as o` ra ‘be with’ in Leti. Leti (Central Malayo-Polynesian, Timor, Southwest Maluku) p¨uata=didi] ma-saava=m`eka=e. (59) a. [A¨u-`ora 1 SG-be.with woman=dx1 1 PL . EXCL-marry=only=idx ‘I just marry this woman here.’ (van Engelenhoven 2004: 254) b.

[A¨u-`ora=o] tata-lt¨ıeri-nena=o. 1 SG-be.with=2 SG neg-1 PL . INCL-speak-rl=ind ‘We cannot speak with each other.’ (van Engelenhoven 2004: 255)

c.

Vat¨um¨ıaa] ra-ara=e. [Tutk¨ue` i n-`ora 3 PL-war=idx T. 3SG-be.with B. ‘Tutukei and Batumiau were at war (with each other or someone else)’ (van Engelenhoven 2004: 254)

In Tetun (Timor), the verb21 hoo ‘accompany, be with (person)’ has been reanalysed as a coordinator of co-actor NPs under the form no ‘and’ (from n-o [lit.: 3sg-be with], a reduced third person singular form of hoo).

21. Another verb, hodi ‘bring, take, use’ has been reanalysed as a prepositional verb introducing instrument noun phrases, and as a coordinator of clauses (van Klinken 2000).

Noun-phrase conjunction strategies in Austronesian languages

263

Tetun-Fehan (svo, Central Malayo-Polynesian, West Timor) (60) a. Ha’u k-oo emi. 1sg 1sg-be.with 2pl ‘I am with you.’ (van Klinken 2000: 358) b.

Maromak la n-abusik lerek [emi no ha’u]. God neg 3sg-leave.free forsake 2pl conj 1sg ‘God does not just abandon you and me.’ (van Klinken 2000: 360)

Several other conjoining strategies have previously been mentioned in Tetun: two of them (the summation pronoun (4) and inclusory constructions (23)) are restricted to NPs with human reference; the comitative coordinator hoo is also more frequent than the coordinator no ‘and’ for human NP conjunction. 3.2. The status of ma in Nˆelˆemwa: a case study This section will focus on the syntactic status of the connector ma in Nˆelˆemwa, based on my own data. Ma is one of several coordinators, among which two others (me and xa) have NP conjoining functions and will briefly be mentioned for comparison (see Bril 2002, 2004 for further details). Ma may mark standard coordination (62a) or inclusory coordination (61a), while xa is only additive (and also used in counting). – Additive xa vs. inclusory ma Nˆelˆemwa (61) a. Yaman ma axaleny. 1du.excl.fr conj this.one ‘Me and that man.’ (inclusion) b.

Yaman xa axaleny. 1du.excl.fr conj this.one ‘The two of us and/plus that man.’ (addition)

– Me and ma The comitative coordinator ma conjoins nouns referring to animates with close relationship (62), and triggers obligatory number agreement. Only ma may be used in inclusory constructions (61a) (never me, nor xa); ma never has any adjunctive function.

264

Isabelle Bril

Nˆelˆemwa (Bril fieldnotes) (62) Coxaada-me [Aloe ma Mudae] me hli u fhe [waja go.up-vent Aloe conj Mudae conj 3du pfv take boat [Kaavo ma Teˆa Pwayili]]. Kaavo conj Teˆa Pwayili ‘Aloe and Mudae (two waves) rise up and take Kaavo and Teˆa Pwayili’s boat.’ By contrast, the general coordinator me ‘and’ conjoins ± human NPs (as well as VPs and clauses, as in (62)). When conjoining NPs, it never has any inclusory function, and it may head a comitative adjunct without number agreement as in (63a) (me might be the reflex of POc *mai ‘be with’). Compare the NPs conjoined with ma which trigger number agreement in the dual (63a), and the comitative adjunct (pwaxi ‘child’) headed by me, which is excluded from agreement. Similarly, in (63b), me hlaaleny shˆalaga is an adjunct. Nˆelˆemwa (Bril fieldnotes) muuvi mwˆa [Kaavo Dela ma Teˆa (63) a. Na [hli u then K. D. conj T. conj 3 DU pfv stay Pwayili]] me aaxiik pwaxi-hli. P. conj one child-poss.3du ‘And Kaavo Dela and Teˆa Pwayili lived there and/with their child.’ b.

Hli u oda-me hli me hlaaleny shˆalaga 3du pfv go.up-vent 3du.fr conj these.dx1 crab i hli. conn 3du ‘They(2) come back, they(2) with their crabs.’

Several tests will be used to assess the conjunctive functions of ma in Nˆelˆemwa. A] Extraction Extraction (e.g. topicalization, focus or wh-movement) is commonly used to distinguish a coordinator from an adposition. An adposition may be stranded, not a coordinator: who did you go with?, but *who did you go and? is ungrammatical. An NP may be extracted together with its adposition with whom did you go?, not with a coordinator *and who did you go? The following test will bear on the properties of the comitative coordinator ma in comparison with the deverbal associative preposition ve∼vi ‘with’ (from fhe ‘take, carry’). As far as focus movement goes, the preposition ve∼vi and the comitative coordinator ma behave

Noun-phrase conjunction strategies in Austronesian languages

265

similarly, with a resumptive anaphoric pronoun in situ, the main distinctive feature between them is number agreement (see Bril 2004: 512–513, for details). But in topicalisation (64b), only ma behaves as a phrasal coordinator, ve∼vi may not, as shown in (64c). Nˆelˆemwa (Bril fieldnotes) (64) a. Mo pˆıˆıla wuung yamon ma aˆ lˆo-raamwa 2 DU stroll together 2 DU.fr conj young-woman eli. that.anaph ‘You took a walk together, you and that girl.’ (emphatic) b.

[Yamon ma aˆ lˆo-raamwa eli] xe mo pwe 2du.fr conj young-woman that.anaph top 2du fish wuung. together ‘You and that girl, you took a walk together went fishing together.’

c.

I fuk vi Kaavo ‘he flew with Kaavo’; but *ye vi Kaavo xe hli fuk (intended meaning: ‘he with Kaavo, they flew’) would be ungrammatical.

Agreement and extraction thus show that ma is a comitative coordinator, behaving differently from the comitative preposition ve∼vi. Phrasal inclusory constructions such as (64a–b) are structurally asymmetrical, with an inclusory free pronoun including the reference to the ma-headed subset NP, yet number agreement highlights their semantic symmetry (they are co-agents with close relationship). Non-phrasal (split) inclusory constructions (such as (64d) or (26b)), without a superset free pronoun (contiguous to the ma-headed NP, as in (64a)) are structurally even more asymmetrical, yet the reference to the ma-headed NP (64d) is included in the number of the subject pronoun, as a conjunct not an adjunct. d.

tu haga ma na xe yo haga O hˆı if 1DU.incl go.down fish conj 1 SG top 2sg fish co, bu fo idaama-m . . . 2sg.fr as there.is eye-poss.2sg ‘When we go fishing, (you) and/with me, you will be able to fish since you have your eyes . . . ’ (but what about me who have lost mine?)

266

Isabelle Bril

Example (64d) displays the strange occurrence of a 1st person subset pronoun (na), instead of the expected 2nd person (co, yo) in constructions with standard person hierarchy (see (64e): e.

O hˆı tu haga ma co ‘when we go fishing, you (and me)’

This is contextual and has some pragmatic effect: in (64d) the speaker is blind and helpless, thus not in a leading position, but in the position of a subservient co-agent, which accounts for the occurrence of the ma-headed 1st person pronoun na. Otherwise, example (64d) is not structurally different from the standard inclusory construction in (64e) (see a similar construction in (26b-c) above), it just reverses the standard person hierarchy for semantic and pragmaticargumentative (possibly contrastive) effect. In any case, the ma-headed subset NP is not an adjunct. B] Iteration Comitative adpositions are usually not iteratable in listing, while coordinators may be. Ma in Nˆelˆemwa can be iterated (65a), exactly like the standard coordinator me (65b). Nˆelˆemwa (Eastern Oceanic, North New Caledonia) hlaaleny aavak thaxamo i ye: (65) a. Hla oda-me conn 3sg 3 PL go.up-vent these.dx1 four wife horaamwaleny Naan ebai ma Naabuc ma this.woman.dx1 fly anaph conj mosquito conj Deedan ma Naalˆooˆ p. horsefly conj small.fly ‘His four wives come up here: this woman Fly, and Mosquito, and Horsefly, and small Fly.’ (Bril 2004: 506) b.

I xau mwˆemwˆelˆı roven hleena yada-va 3sg ass know all these.dx2 customs-poss.1pl.incl me yada fwamwa-va me conj customs country-poss.1pl.incl conj kedoxa-va me. medicine-poss.1pl.incl conj ‘She knows all our customs, and all the customs of our country, and our medicine and so on.’ (Bril fieldnotes)

Noun-phrase conjunction strategies in Austronesian languages

267

By contrast, the comitative coordinator me in M¯aori cannot be iterated more than twice and seems to preserve more of its prepositional properties. Listing usually starts with juxtaposition and ends with one or two final me: M¯aori (Estern Polynesian, Bauer 1997: 552) (66) . . . e whiu ana i te taonga nei, i te h¯apuku tam collect tam do art treasure near do art groper [. . . ], i te huahua kuku me te [. . . ] do art preserved.game pigeon conj art huahua k¯ak¯a, me te huahua k¯uaka. preserved.game kaka conj art preserved.game godwit ‘. . . they collected these treasures, groper [. . . ] and preserved pigeon, kaka and godwit.’ 3.3. Other tests of the syntactic status of the conjunctive morphemes Some other distinctive properties will be considered. The first one concerns conjunct position and phrasal vs. non-phrasal properties. 3.3.1. Position: Contiguous vs. non-contiguous conjuncts Are conjuncts obligatorily phrasal and contiguous, by contrast with adjuncts? NP conjuncts are often contiguous, but cross-linguistic data show that there is no universal constraint on the contiguity of conjuncts, and that non-contiguity is not an a priori argument against coordination. On the other hand, non-contiguous conjuncts may be banned by some language specific rules. In many Oceanic languages (among them Nˆelˆemwa), conjuncts in split inclusory constructions are non-phrasal and non-contiguous; and in Tolai, even standard coordinate NPs may be non-contiguous (67). Tolai (Western Oceanic, Meso-Melanesian cluster, New Britain) (67) Telengai dir rovoi ma ra pap. Telengai 3 DU hunt conj art dog ‘Telengai and the dog hunted.’ (Mosel 1984:176) In Nakanai, contiguous subject conjuncts indicate the concomitant involvement of the participants in an action or state, while non-contiguous subject conjuncts as in (68b) seem to indicate a slight shift of emphasis and topicality22 , the men being slightly less topical in (b); 22. Johnston is not more specific in his description.

268

Isabelle Bril

Nakanai (Western Oceanic, Meso-Melanesian cluster, New Britain) (68) a. E hatavile, la valalua, egite o-io po-pou. art women art men 3pl at-there red-sit ‘The women and the men are there waiting.’ (Johnston 1980: 243) b.

E hatavile o-io po-pou, la valalua. art women at-there red-sit art men ‘The women are there waiting, and the men.’ (Johnston 1980: 243)

Rather than position, the most reliable criterion for conjunction in Austronesian languages is number agreement with the conjuncts; though number agreement is known not to be a universal constraint either, since various language types choose single conjunct agreement with the nearest or furthest conjunct (in Slavic languages, for instance Russian or Polish (Corbett 2006), in Welsh (Sadler 2003)). – Comitative conjunctions and the positional criterion If contiguity is not fully criterial for standard conjunction, on the other hand, in Oceanic languages, contiguity and agreement are important factors for comitative adpositions to acquire conjoining functions and thus be disambiguated from their adjunctive function. In Mwotlap, for instance, the NP conjuncts must be contiguous for the comitative preposition tiwag mi23 ‘together with’ to function as a coordinator (see (40e-f-g)). In Manam, the comitative postposition z´aiza has conjoining functions with number agreement if NPs are contiguous (69a), in contrast with (69b) where it is adjunctive. Manam (sov, Western Oceanic, North New-Guinea cluster, PNG) (69) a. [/´ai/o n´aita z´aiza] /a-pile-p´ıle? 2sg who com 2 PL.rl-speak-red ‘Who are you speaking with?’ (Lichtenberk 1983: 432) b.

An´ua ne-Ø i-l´a/o [n´atu-Ø z´aiza]. village poss-3sg 3 SG-go child-3sg com ‘She went to her village with her child.’ (Lichtenberk 1983: 376)

c.

[[r´oa-ø-be n´atu-ø] z´aiza] spouse-3sg.poss-and child-3sg.poss com di-sin-s´ın-ø-to. 3 PL-drink-red-3pl-PC ‘He used to drink them (coconuts) with his wife and child.’ (Lichtenberk 1983: 376)

23. Tiwag originally meant ‘one’ (Fran¸cois p.c.).

Noun-phrase conjunction strategies in Austronesian languages

269

In (69c), z´aiza is also conjunctive as shown by the plural subject index referring to the three participants (including an unexpressed 3rd person singular participant), instead of the expected singular subject index if z´aiza were an adjunctive comitative postposition. Z´aiza functions as a set inclusion morpheme, while the standard coordinator be conjoins the two members of the subset (spouse and child). – Negative comitative constructions The negative comitative construction (“without”) expressing lack of accompaniment follows the same principle. Conjuncts are contiguous and the negation marker t´ago has scope over the comitative conjunct headed by z´aiza (69d); note again the plural subject pronoun on the verb. Negative comitative may also be expressed by a negated coordinate construction marked by a summation pronoun (d´ı-a-ru in (69e)). The scope of the negation marker t´ago is over the second conjunct. d.

z´aiza]] di-p´ura-to. [N´ai [t´ago to/´a-ø 3 PL.rl-arrive-PC 3sg.fr NEG older.brother-3sg com ‘He came without his older brothers.’ (lit.: he did not arrive with his older brothers) (Lichtenberk 1983: 432)

e.

d´ı-a-ru] [N´ai t´ago to/´a-ø 3sg.fr NEG older.brother-3sg 3 PL .fr-bf-DU di-p´ura-ru. 3 PL.rl-arrive-DU ‘He arrived without his older brother.’ (Lichtenberk 1983: 275)

Both constructions trigger number agreement with the conjuncts in spite of the lack of co-agency or collective action. This is an additional argument to suggest that syntactic coordination overrides semantics. Similarly in French or in English, the actor NPs conjoined by et trigger plural number agreement even though the predicate is negated, son fr`ere et lui ne sont pas venus (ensemble/s´epar´ement) ‘his brother and him have not come (together/separately)’, whereas the comitative adjunct of a negated predicate is not included in number agreement il n’est pas venu avec son fr`ere ‘he has not come with his brother’24 . 24. Similarly, in Chilean Spanish, inclusory construction with con ‘with’ may be discontinuous (as in (a)), while in its conjoining function, the conjuncts are contiguous (b): (a) fuimos al mercado con mi amiga ‘I went to the market with my friend’ [lit. went.we to the market with my friend]; (b) [yo con mi amiga] fuimos al Mercado ‘I with my friend we went to the market’ (Schwartz 1988).

270

Isabelle Bril

3.3.2. Semantic tests: collective or distributive readings Additive and comitative coordination tend to have slightly different semantics: additive coordination has a distributive (‘each’) or a collective reading, while comitative coordination tends to have a collective reading only, unless the comitative morpheme has undergone reanalysis as a full-fledged coordinator. A] Collective and distributive readings of the conjunctive construction Coordinators are compatible with a collective or a distributive reading and with adverbs such as ‘together’ or ‘separately’, whereas comitative/associative morphemes only have the collective reading. In Zuanga (neighbour of Nˆelˆemwa), the inclusory construction with the comitative coordinator ma is compatible with predicates with a collective or a distributive reading (70a–b). Compare with standard coordination in (70c). Zuanga/Yuanga (Eastern Oceanic, North New Caledonia) (70) a. Bi a pe-haze ma a˜ baa-nu. 1du.excl go recp-separate conj brother-poss.1sg ‘My brother and I went separately.’ (Bril fieldnotes 2006) b.

Bi vara kiba`o a-k`o bw`o ma 1du.excl each kill clf-three bat conj a˜ baa-nu. brother-poss.1sg ‘My brother and I killed three bats each.’ (Bril fieldnotes 2006)

c.

Li a kaze [Kaavo ma Hiixe]. 3du go fish Kaavo conj Hiixe ‘Kaavo and Hiixe go fishing.’ (Bril fieldnotes 2006)

The use of inclusory constructions for NP conjunction involving a pronoun and a noun is receding among younger Zuanga speakers, who tend to use standard coordination with a non-inclusory free pronoun in sentence topic position as in (70d); this type of construction is ungrammatical in neighbouring Nˆelˆemwa. Thus inclusory constructions with comitative conjunctions tend to evolve towards standard coordination in some northern New Caledonian languages. d.

ma moo-je] li uda no mwa. [Ije 3 SG.FR conj wife-3sg 3 DU enter in house ‘He and his wife, they went into the house.’ (Bril fieldnotes 2007)

Noun-phrase conjunction strategies in Austronesian languages

271

In Fijian, the comitative conjunction kei ‘and, with’ is also compatible with the collective or distributive readings: Fijian (Central Pacific) (71) a. Keirau lako kei Samu. 1du.excl go conj Samu ‘Samu and I are going.’ (Milner 1967: 67) b.

Erau s¯a dui kania na kona ika [ko Jone kei 3du emph each eat art his fish art Jone conj Filipe]. Filipe ‘Jone and Filipe are each eating their own fish.’ (Milner 1967; in Payne 1985: 22)

B] Predicate semantics Predicate semantics is another possible discriminating factor: motion and action verbs favour the collective reading of a coordinate expression such as John and I ran away, interpreted by default as ‘ran together’, and which is also expressible as a comitative construction I ran away with John in English. On the other hand, property predicates tend to favour the distributive reading of a coordinate expression such as my brother and I are tall, and exclude the collective reading or a comitative construction (such as *I am tall with my brother). The comitative construction becomes slightly more acceptable with inclusory pronouns (?we are tall with my brother) provided that the property is conceived as a shared or symmetrical property; “with” then signals set inclusion, not standard comitative adjunction. English existential-possessive predicates allow both the distributive and the collective reading of a coordinate expression, as in my brother and I have a house (each or in common); the collective reading is also expressible as a comitative construction I/we have a house with my brother. In Zuanga (72) and Nˆelˆemwa (73), inclusory constructions with the comitative conjunction ma, are compatible with property predicates25 and accept a distributive or a collective reading when they refer to some shared or similar property. The important point is that ma expresses set inclusion, not comitative adjunction.

25. Schwartz (1988: 69–70) also notes that in Chilean Spanish property predicates may appear in inclusory constructions with con ‘with’ expressing set inclusion: somos altos con mi hermano ‘my brother and I are tall’ [lit. we are tall with my brother], nos gusta el jazz con mi hermano ‘my brother and I like jazz’ [lit. we like jazz with my brother].

272

Isabelle Bril

Zuanga/Yuanga (Eastern Oceanic, North New Caledonia) (72) a. Bi en˜o ma a˜ baa-nu. 1du.excl young conj brother-poss-1sg ‘My brother and I are young.’ (Bril fieldnotes 2006) b.

Kixa mwani-bi ma a˜ baa-nu. there.is.no money-poss.1du.excl conj brother-poss-1sg ‘My brother and I are poor [lit.: there is no money of ours including my brother].’

Nˆelˆemwa (Eastern Oceanic, North New Caledonia) (73)

Pe-kau-ma ma Polie. recp-year-1du.excl conj Polie ‘Polie and I are the same age.’ [lit.: (it is) ourdual reciprocal age including Polie] (Bril 2002)

4. Competing strategies in Austronesian languages: standard coordination, inclusory constructions, comitative coordination Many Austronesian languages display various competing strategies for (i) standard coordination, (ii) inclusory constructions, (iii) comitative conjunction, and (iv) comitative adjunction.All four types may be marked by different morphemes and different constructions; otherAustronesian languages make use of one single morpheme whose various functions are distinguished by agreement or position. 4.1. Languages with multiple morphemes and constructions In Austronesian languages, the chosen strategy is sensitive to the category of the conjuncts (pronouns or nouns), to semantic properties (± animate, tight vs. loose relationship) and to pragmatic properties (± even saliency). In Kairiru, the summation pronoun strategy (see ex. (7)) is restricted to humans with close relationship; NPs referring to humans, like other NPs, may also conjoined by the standard coordinator bo ‘and’. If unequal control is involved, a comitative adjunct is then used (74). Kairiru (sov, Western Oceanic, North New Guinea cluster, East Sepik, PNG) nau qeq-au kyau. (74) Nat a-fwat child 3 SG-bathe sea COM -1 SG 1sg ‘The child bathed in the sea with me.’ (Implied meaning: I hold him) (Wivell & Ross, in Lynch et al. 2002: 213)

Noun-phrase conjunction strategies in Austronesian languages

273

Here are further examples of competing strategies in a sample of (mostly Western) Oceanic languages. 4.1.1. Maleu In Maleu, free pronouns and nouns may be conjoined by me ‘and’ (75a), or expressed as inclusory constructions (75b–c) marking close relationship between persons or kin. Maleu (Western Oceanic, North New Guinea cluster) me Vulum. (75) a. Iou 1 SG .fr and Vulum ‘Vulum and I.’ (Haywood 1996: 168) b.

Vulum. Iami-lua 2 PL.fr-two Vulum ‘You and Vulum.’ (Haywood 1996: 169)

c.

tivu-a. Ire-lua 3 PL.fr-two grand-father-poss.3sg ‘He and his grand-father.’ (Haywood 1996: 169)

Inclusory construction or coordination with me require conjuncts with identical properties, otherwise triggering the use of comitative verbs (in complex verb constructions, (75d)) and instrumental strategies (75e). These strategies express differences of rank, topicality or other semantic differences. d.

ta-reNe ta-ualiu. Ku-moi 2 SG-come 1 PL.incl-be.with 1 PL.incl-wash ‘Come and bathe with us.’(lit.: come be with us we bathe) (Haywood 1996: 180)

This comitative verb is reanalysed as a deverbal preposition, sometimes involving metathesis, from tireNe (3sg be.with) to tiNere ‘with’. e.

na-melamela tiNere tna-re nm-children com mother-poss.3 PL ‘the children with their mother’ (Haywood 1996: 180)

274

Isabelle Bril

4.1.2. Mangap-Mbula In Mangap-Mbula, pronouns and nouns may be conjoined by a summation pronoun (76a) when the participants are co-agents or co-experiencers of the event, though with a slight topical asymmetry and a slightly less topical conjunct to the right of the conjunctive summation pronoun. Mangap-Mbula (svo, Western Oceanic, North New Guinea cluster) wae-N] am-la. (76) a. [Nio niamru 1sg.fr 1 DU.EXCL associate-poss.1sg 1 PL.EXCL-go ‘I, together with my wife, went.’ (Bugenhagen 1995: 114) Nouns and pronouns may be conjoined by the NP coordinator ma or mi ‘and’26 (76b), expressing equal prominence and topicality. b.

[Nio ma abu mi Matias] am-la 1sg.fr conj grand-mother conj M. 1 PL.excl-go mokleene tiama. taro.place loc.1 PL.excl.n.vis.prox. ‘I and my grandmother and Matias went to our garden.’ (Bugenhagen 1995: 395)

c.

ti-re yam. Ata ma NamoNo ti-mar Ata conj NamoNo 3 PL -come 3 PL-see 1 PL . EXCL ‘Ata and NamoNo came to see us.’ (Bugenhagen 1995: 213)

Unequal syntactic and semantic status triggers a comitative adjunct, marked by raama ‘together, with’, without number agreement (76d–e). d.

Ni i-mbot raama zin buzur saNsaN-Na-n. PL animal wild-nmlz-gen.3pl 3sg.fr 3sg-stay com ‘(S)he stayed with the wild animals.’ (Bugenhagen 1995: 213)

e.

raama yom totomen. Ni ko i-mbot 2pl always 3sg.fr unc 3 SG-stay com ‘He will always be with you.’ (Bugenhagen 1995: 114)

26. Ma is used for formulaic conjuncts of equal topicality, or for commonly or naturally associated entities; mi is used for more arbitrary conjuncts or lists, and for clauses (Bugenhagen, 1995: 214).

Noun-phrase conjunction strategies in Austronesian languages

275

4.1.3. Tawala In Tawala, there are three main coordinating strategies, using two standard coordinators, po and ma ‘and’(77a), or a comitative conjunction mite- ‘with’(77b–c). Po is more commonly used for NP conjuncts with tight relationship, while ma conjoins NPs with looser relationship, and clauses (Ezard 1997: 155, 199). Ma has additive and distributive meanings. Compare (i) and (ii): (i)

amau po hinau ‘my father and mother, my parents’ (there is no other term for ‘parents’)

(ii)

amau ma hinau ‘my father and/plus my mother’.

Nouns and pronouns are also conjoined by these coordinators, without requiring an inclusory construction. Tawala (SVO, Western Oceanic, Papuan Tip cluster, Milne Bay, New Guinea) (77)

a.

Tau ma Mika. 1sg.fr conj Mika ‘Mika and I.’ (Ezard 1997: 157)

b.

Taumi mite-u ta-kaoha. 2pl.fr with-1sg 1 PL.incl-happy ‘Let’s be happy together.’ (Ezard 1997: 159)

c.

Tau mite-mi ta-memae. 1sg.fr with-2pl 1 PL.incl-stay ‘I and you are staying together.’ (Ezard 1997: 160)

The comitative conjunction mitehi (from mite-hi ‘with-3pl’) functions as a postpositional conjunction and as a set inclusion morpheme indicating co-agentive conjuncts. It may be combined with po and ma (77e). d.

Bada natu-na mitehi hi-nei. man child-3sg together 3pl-came ‘The man came with his child.’ (Ezard 1997: 158)

e.

Tau ma Bryan a manago mitehi 1sg.fr conj B. poss.3sg family together to-gelu. 1 PL.excl-embark ‘I and Bryan and his family got on board.’ (Ezard 1997: 159)

276

Isabelle Bril

There are further differences: po and ma are medial coordinators [NP po ∼ ma NP], while mitehi is postposed to the second conjunct [NP NP mitehi]. The negative comitative ‘without’ is expressed by the negation marker ega which has scope on the comitative conjunct [NP [ega NP mitehi]] (lit.: [NP [not NP together]]), as in Manam (69d). Comitative conjunction and set inclusion with mitehi is further distinguished from comitative adjunction which is marked by a preposition a ‘with’, without number agreement and referring to subservient actors (1997: 158). 4.1.4. Num`ee` (New Caledonia) New Caledonian languages generally distinguish a comitative conjunction ‘and/ with’ triggering number agreement with the conjuncts and expressing coagentivity, from an associative preposition ‘with’ marking associative adjuncts. Comitative conjuncts and adjuncts express not only differences in agency, but also in topicality and animacy (see -vi vs. ma in Nˆelˆemwa (64b–c)). Num`ee` has several coordinators, the focus here will be on two of them, mˆe and mˆo. Mˆe is the standard coordinator which conjoins (± animate) NPs, possibly with listing effect (78a–b), as well as clauses. Mˆo ‘and, with’ is both conjunctive and adjunctive27 , its functions are distinguished by number agreement, as shown by (78c vs.78e–f). As a comitative conjunction, it conjoins all types of nouns, proper or common, ± animate. Num`ee` (Far South New Caledonia) (78)

a.

`e kai..i mˆe yàre Nê` ngˆe wè` mêˆ mˆe. 3pl transl do vent food and fishing and ‘They bring food and fish.’ (Rivierre p.c.)

b.

Nê` mwâ` t`uu` vˆa niikwˆerˆe mˆe v¨ue¨` mˆe 3pl then board group child and woman and ` e kô. ` wawèr´ everyone on.top ‘They board the children women and everyone on them (boats).’ (Rivierre p.c.)

27. Also marking complement of comparative: ` Yaà gu mê` v´e-t´et´e mˆo nê. neg 2sg fut be.similar with 3pl, ‘You won’t be similar to them.’

Noun-phrase conjunction strategies in Austronesian languages

277

c.

ngê` trii´e? [Kwaraa mˆo Kwˆajoro] w´ee´ ngˆaâ` xìro K. and K. interj 2 DU pole(boat) with who? ‘Kwaraa and Kwˆajoro w´ee´ e´ ! who are you(2) travelling with?’ (Rivierre p.c.)

d.

`e . . . p`akanˆo-é` w`e-karé´ mˆo kumwà mˆo corn-this full.maturity and sweet.potato and d´e. sugar-cane ‘. . . full of fully mature corn-cobs and sweet potatos and sugar-cane.’ (Rivierre p.c.)

e.

´ erˆe nyaar`e mê` gu vê` bw`e mˆo e´ kwˆerˆe Ekwˆ comp 2 SG come with 1 DU.excl to 1 DU.excl want vˆıô` wˆe e´ kwˆerˆe. be.wife of 1du.excl ‘We(2) would like you to come with us to become our wife.’ (Rivierre p.c.) › mwâ` nyˆı ba t`ro mˆo nyô-ê` . . . then 3 SG ass stay with woman-this ‘. . . then he stays with this woman . . . ’ (Rivierre p.c.)

f.

Comitative ∼ associative (78c) (and instrumental) adjuncts (78g) are also marked by the preposition ngˆe ‘with’, the adjunct has subservient or patientive features. g.

`e ngê` n` uu a` nyˆı ngˆe t`uu` mwˆeeˆ mêˆ and 3sg transl go.down again vent with knife this ` ngˆe mˆo wâbutrii ngˆe mwˆeeˆ . and basket this again ‘and she runs down again towards the shore with her knife and basket.’ (Rivierre p.c.)

4.2. Languages with a unique morpheme distinguished by number agreement On the other hand, many other Oceanic languages display a single morpheme for this range of functions: (i) standard coordination, (ii) comitative conjunction, (iii) comitative ∼ associative, and sometimes instrumental adjunction. Number agreement, position or a lexical device (such as collective adverbs ‘together’) may disambiguate the conjunctive from the adjunctive functions. Among such languages are Samoan (Mosel & Hovdhaugen 1992: 148), Fijian, Drehu and

278

Isabelle Bril

Iaai (Loyalty Islands, New Caledonia), Xˆarˆac`uu` 28 (South New Caledonia), most Polynesian languages (Lynch & Moyse-Faurie 2004). The functions of this polyfunctional morpheme sometimes extends to inclusory constructions, then expressing set inclusion (Tolai, North New Caledonian languages), and possibly to VP or clausal coordination29 as an ‘and’ coordinator. In Tolai, ma displays this full range of functions, which are distinguished by agreement (Mosel, 1984: 94, 175–76): coordinator (of NPs,VPs and clauses), inclusory conjunction, adjunctive marker with comitative, instrumental or associative meaning, also expressing part-whole relationship (he ate the banana with the skin), or causal relationship (they suffered with/from hunger, full of/with fish). In Fijian, kei ∼’ei ‘with, and’ also has a wide range of functions: standard NP coordination (79a), inclusory conjunction (set inclusion) (79b), comitative adjunction without agreement (79c). Inclusory conjunction (79b) is used for noun-pronoun conjunction, but Dixon (1988: 157–160) notes that it seems to be receding among younger speakers who increasingly use the comitative adjunct construction as in (79c). Standard Fijian (Central Pacific) na watina]. (79) a. Erau lako mai [na turaga kei 3 DU move here art chief conj art wife ‘The chief and his wife are coming.’ (Payne 1985: 29) b.

a vei-vosa-ki kaya30 . Keirau 1 DU.excl past distr-talk-tr kei.3sg ‘I talked with him.’ (Churchward, 1941: 42; in Sch¨utz 1995: 354)

c.

Au aa vei-vosa-ki kaya. 1 SG past distr-talk-tr kei.3sg ‘I talked with him.’ (Milner, 1972: 68, in Sch¨utz 1995: 354)

Dixon (1988: 157–160) also notes a growing tendency to use vata ’ei instead of ’ei alone: Boumaa Fijian (Central Pacific) sota vata ’ei Jone mai Viidawa. d. Au aa 1 SG past meet together with John at Viidawa ‘I met up with John at Viidawa.’ (Dixon 1988: 159) 28. The coordinator coexists with coagentive comitative prepositions or associative noncoagentive, w´er`e (+animate), k`er`e (+/animate) without agreement (Moyse-Faurie, 1995: 140). 29. Like me in Iaai (Ozanne-Rivierre 1976) and ma in Nemi (Ozanne-Rivierre 1979). 30. The form kaya is analysable as kei + 3sg pronoun.

Noun-phrase conjunction strategies in Austronesian languages

279

Because of these two coexisting patterns (inclusory and adjunctive), Dixon (1988) and Sch¨utz (1995:354) point out that sentences such as (79e) become ambiguous as to the number of people involved, two (in the inclusory reading) or three (in the adjunctive reading)? Standard Fijian (Central Pacific) lako kei Samu. e. Drau aa 2 DU past go conj Samu ‘You and Samu went.’ (Milner 1972: 68; in Sch¨utz 1995: 354) (or: you.2 and Samu went) Inclusory constructions are unstable and often undergo reanalysis. I have outlined some of these changes in a previous article (Bril 2004: 528–529) and shown either the reanalysis of the inclusory pronoun as a comitative marker, or the decay of the construction once the inclusory pronoun has been replaced by a non-inclusory free pronoun in a standard phrasal coordinate construction, as in Fijian (79c). Similarly, in Zuanga, younger speakers increasingly use a comitative adjunct with the singular pronoun e ‘(s)he’ (80a), instead of the expected inclusory dual pronoun li ‘they2’ found in the inclusory construction still in use among older speakers (80b): Yuanga/Zuanga (North New Caledonia) uda n˜o mwa m˜a ti? (80) a. E 3 SG enter in house com who? ‘Who did he enter the house with?’ (Bril fieldnotes 2007) b.

uda n˜o mwa m˜a ti? Li 3 SG enter in house com who? ‘Who did he enter the house with?’ (Bril fieldnotes 2007)

A similar evolution was noted in Nyelˆayu where the inclusory construction (81a) is increasingly replaced by standard coordination (81b): Nyelˆayu (North New Caledonia) (81)

a.

ri? Hor ma 2 DU and/ with who? ‘You (sg.) and/ with whom?’ (Ozanne-Rivierre 1998: 119)

280

Isabelle Bril

b.

Yo ma ri? 2 SG and/ with who? ‘You (sg.) and/ with whom?’ (Ozanne-Rivierre 1998: 119)

5. Final discussion All the discussed Austronesian languages show coordination and display various conjoining strategies, often within the same language (as shown in table 2 below), varying with (i) topicality or pragmatic effects, (ii) syntactic constraints, or (iii) semantic properties (higher animacy, close relationship, co-agency). Coordination may be ± syndetic, according to the type of conjuncts (± animate, nouns or pronouns). When syndetic, the markers may be standard additive coordinators, comitative coordinators or summation pronouns. If contiguity of the conjuncts is not an absolute prerequisite for standard (± syndetic) coordination – provided that discontinuity is compensated on by some other device like number agreement –, on the other hand, contiguous conjuncts are a prerequisite for a comitative adposition to be reanalysed and to function as a full coordinator (like z´aiza in Manam). Inclusory constructions with their set inclusion pattern stand in some medial position on the cline between prototypical NP coordination and prototypical comitative constructions, and are compatible with ± contiguous conjuncts. Contiguity and number agreement generally express close relationship and similar roles or properties, such as co-agency, otherwise adjunctive constructions are used. Inclusory constructions display specific properties and may be subject to various semantic and syntactic constraints. Among the semantic constraints is their universal restriction to NPs referring to higher animates: some languages restrict their use to co-agentive animates (though sometimes with a slight topical asymmetry), other languages extend their use to other semantic roles and syntactic functions. Among the syntactic constraints is the frequent (but non-universal) ban on noun-pronoun conjunction, as in Nˆelˆemwa and other northern New Caledonian languages. In languages without any such ban, the choice between inclusory constructions and standard coordination mostly correlates with discourse or pragmatic effects: coordination is chosen to express equal topicality, saliency or emphasis, otherwise inclusory constructions are used, expressing set inclusion. Coordinate constructions may be ± symmetrical, as evidenced cross-linguistically by various types and degrees of conjunct asymmetry. Inclusory constructions display one type of asymmetry between the superset and its specifying subset which is often headed by a comitative coordinator.

Noun-phrase conjunction strategies in Austronesian languages

281

Cline of asymmetry between conjuncts: Lesser asymmetry → Higher asymmetry Standard NP coordination → phrasal ± syndetic inclusory → non-phrasal ± syndetic inclusory

The comitative conjunction used in syndetic inclusory constructions stand in a medial position on the cline between prototypical coordination and prototypical comitative adjunction. Its conjoining function is clearly marked by set inclusion agreement. Compatibility with distributive or collective adverbs, or with verbs with similar semantics, help disambiguate the function and meaning of these comitative conjunctive markers. Table 1. Functions of the pV, bV, mV or kV morphemes in some Oceanic languages Inclusory Coordinating Comitative adjunct Nakanai me me Maleu me Mangap-mbula ma, mi Manam -be Tawala po, ma Tolai ma ma ma Toqabaqita ma Mwotlap ba tiwag mi tiwag mi me ma, me ma Nˆelˆemwa mani m˜a, mani Zuanga/Yuanga m˜a ma ma ma Nyˆelˆayu ma ma ma Nemi ma ma Bwatoo ma ma C`emuhˆı ma Aji¨e m¨aa¨ mˆe mˆe Xˆarˆac`uu` mˆo mˆo, mˆe Num`ee` me me me Drehu, Iaai Fijian kei kei kei Samoan ma ma ma Tahitian m¯a m¯a M¯aori me me

Associative

Instrumental

me

ma

ma mi

tiwag mi mani ma

mani ma

mˆo

Table 1 below sums up the conjunctive or adjunctive functions of the morphemes used in the Oceanic languages of the sample, concentrating on morphemes with

282

Isabelle Bril

the form pV, bV, mV and kV 31 . Table 2 sums up the various conjunctive or adjunctive constructions used in some Austronesian languages of the sample. Table 2. Construction Types Inclusive asyndetic Atayal Leti Tetun Papuan Malay Kairiru Kaulong Nakanai Maleu Mangap-Mbula Manam Tawala Toqabaqita Mwotlap Nˆelˆemwa Zuanga/Yuanga Nyˆelˆayu Nemi Bwatoo Aji¨e Num`ee` Tolai Fijian Samoan Tahitian M¯aori

+ +

+ + + + + +

+

+ +

Inclusive syndetic +

Coordination Coordination Summation asyndetic syndetic pronoun + + + + + + + + + + + + (+pronouns) + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

31. These morphemes are often cognates (see Lynch et al. 2002; Moyse-Faurie & Lynch 2004).

Noun-phrase conjunction strategies in Austronesian languages

283

References Aissen, Judith 1989 Agreement controllers and Tzotzil comitatives. Language 65: 518–36. Bauer, Winifred 1997 The Reed Reference Grammar of M¯aori. Auckland: Reed Publishing, Bril, Isabelle 2002 Le nˆelˆemwa (Nouvelle-Cal´edonie) : Analyse syntaxique et s´emantique. (Coll. Langues et Cultures du Pacifique 16.) Paris: Peeters . Bril, Isabelle 2004 Coordination strategies and inclusory constructions in New Caledonian and other Oceanic languages. In: Martin Haspelmath (ed.), Coordinating constructions. Amsterdam/Philadelphia: John Benjamins, Typological Studies in Language 58: 499–534. Bugenhagen, Robert D. 1995 A Grammar of Mangap-Mbula:AnAustronesian Language of Papua New Guinea. (Pacific Linguistics A-82.) Canberra: The Australian National University. Corbett, Greville G. 2006 Agreement. Cambridge: Cambridge University Press. Churchward, C. Maxwell 1941 A new Fijian Grammar. Suva: Government Printer. Dixon, Robert M.W. 1988 A grammar of Boumaa Fijian. Chicago: The University of Chicago Press. Evans, Nicholas 2003 Bininj Gun-Wok: a pan-dialectal grammar of Mayali, Kunwinjku and Kune (2 vol.), 419–420. (Pacific Linguistics 541.) Canberra: Australian National University. Ezard, Bryan 1997 A Grammar of Tawala, An Austronesian language of the Milne Bay area, Papua New Guinea. (Pacific Linguistics C-137.) Canberra: Australian National University. Faarlund, Jan Terje 2004 The Syntax of Old Norse, 89–90. Oxford University Press. Fran¸cois, Alexandre 2001 Contraintes de structures et libert´e dans l’organisation du discours: Une description du mwotlap, langue oc´eanienne du Vanuatu. Doctoral dissertation, Universit´e Paris-IV Sorbonne. 3 vol. Haspelmath, Martin (ed.) 2004 Coordinating constructions. Amsterdam/Philadelphia: John Benjamins, Typological Studies in Language 58.. Haywood, Graham 1996 A Maleu grammar outline and text. In: M. Ross (ed.), Studies in Languages of New Britain and New Ireland. Vol. 1: Austronesian languages

284

Isabelle Bril of the North New Guinea Cluster in Northwestern New Britain, 145–196. (Pacific Linguistics C-135.) Canberra: Australian National University.

Huang, Lilian 2006 Case-marking system in Plngawan Atayal, in Hon. of Prof. Li, Streams converging into an ocean. (Language and Linguistics Monograph Series W-5: 205–238.) Taipei: Institute of Linguistics, Academia Sinica. Johannessen, Janne Bondi 1998 Coordination. New York: Oxford University Press. Johnston, Raymond L. 1980 Nakanai of New Britain. The Grammar of an Oceanic language. (Pacific Linguistics B-70.) Canberra: Australian National University. Lazard, Gilbert and Louise Peltzer 2000 Structure de la langue tahitienne. (LCP 15, SELAF 391.) Paris, Louvain: Peeters. Leenhardt, Maurice 1932 Documents n´eo-cal´edoniens. (Travaux et m´emoires de l’Institut d’Ethnologie IX.) Paris: Institut d’Ethnologie. Lichtenberk, Frantisek 1983 A grammar of Manam. (Oceanic Linguistics Special Publications 18.) Honolulu: University of Hawaii Press. Lichtenberk, Frantisek 2000 Inclusory pronominals. Oceanic Linguistics 39-1: 1–32. Honolulu: University of Hawaii Press. Lynch, John, Malcolm D. Ross and Terry Crowley 2002 The Oceanic Languages. Richmond, Surrey: Curzon Press. McNally, Louise 1993 Comitative coordination: a case study in group formation. Natural Language and Linguistics Theory 11-2: 347–379. Dordrecht: Springer. Milner, George B. 1967 Fijian grammar. Suva, Fiji: Government Press. [3rd edition 1972]. Mithun, Marianne 1986 Disagreement: the case of pronominal affixes and nouns. In: Deborah Tannen and James E. Alatis (eds.), Proceedings of the Georgetown University Round Table Conference on Languages and Linguistics 1985: 50–66. Washington DC: Georgetown University Press. Mithun, Marianne 1988 The grammaticization of coordination. In: Haiman, J. and Sandra A. Thompson (eds.), Clause combining in grammar and discourse, 331– 359. Moravcsik, Edith 2003 A semantic analysis of associative plurals. Studies in Language 27: 469– 503.

Noun-phrase conjunction strategies in Austronesian languages

285

Mosel, Ulrike 1984 Tolai syntax and its historical development. (Pacific Linguistics B-92.) Canberra: The Australian National University. Mosel, Ulrike and Even Hovdhaugen 1992 Samoan reference grammar. Oslo: Scandinavian University Press. Moyse-Faurie, Claire 1983 Le drehu, langue de Lifou (Iles Loyaut´e). (LCP 3.) Paris: SELAF. Moyse-Faurie, Claire 1995 Le xˆarˆac`uu` , langue de Thio-Canala (Nouvelle-Cal´edonie). (LCP 10.) Paris: SELAF. Moyse-Faurie, Claire and John Lynch 2004 Coordination in Oceanic languages and Proto Oceanic. In: Haspelmath, Martin (ed.), Coordinating constructions. Amsterdam/Philadelphia: John Benjamins, Typological Studies in Language 58: 445–497. Osumi, Midori 1995 Tinrin grammar (New Caledonia). (Oceanic Linguistics Special publication 25.) Honolulu: University of Hawaii Press. Ozanne-Rivierre, Fran¸coise 1976 Le iaai. Langue m´elan´esienne d’Ouv´ea (Nouvelle-Cal´edonie). (SELAF 30.) Paris: SELAF. Ozanne-Rivierre, Fran¸coise 1979 Textes nemi, Nouvelle-Cal´edonie (Vol.1). (SELAF 31.) Paris: SELAF. Ozanne-Rivierre, Fran¸coise 1998 Le nyelˆayu de Balade (Nouvelle-Cal´edonie). (LCP 12.) Paris: Peeters. Ozanne-Rivierre, Fran¸coise 2004 The evolution of the verb “take” in New Caledonian languages. In: Bril, Isabelle and Ozanne-Rivierre, Fran¸coise (eds.), Complex predicates in Oceanic languages: Studies in the dynamics of binding and boundness. (Empirical Approaches to Language Typology 29.) Mouton de Gruyter. Payne, John R. 1985 Complex phrases and complex sentences. In: Timothy Shopen (ed.), Language typology and syntactic description II: Complex constructions, 3–41. Cambridge: Cambridge University Press. Rivierre, Jean-Claude 1980 La langue deTouho. Phonologie et grammaire du c`emuhˆı. (TO 30.) Paris: SELAF. Rivierre, Jean-Claude, Sabine Ehrhart and Raymond Di´ela 2006 Le bwatoo et les dialectes de la r´egion de Kon´e (Nouvelle-Cal´edonie). (LCP 17, SELAF 435.) Paris: Peeters Sadler, Louisa 2003 Coordination and Asymmetric Agreement in Welsh. In: Miram Butt and Tracy Holloway King (eds.), Nominals: Inside and Out, 85–118.

286

Isabelle Bril

Chicago: Center for the Study of Language and Information Publications. Sch¨utz, Albert J. 1985 The Fijian Language. Honolulu: University of Hawai’i Press. Schwartz, Linda 1988a Conditions on verb-coded coordinations. In: Michael Hammond, Edith Moravcsik and Jessica Wirth (eds.), Studies in syntactic typology, 53–73. Amsterdam: John Benjamins. Schwartz, Linda 1988b. Asymmetric feature distributions in pronominal ‘coordinations’. In: Michael Barlow and Charles A. Ferguson (eds.), Agreement in Natural Language, 237–249. Stanford: Center for the Study of Language and Information. Singer, Ruth 2001 Inclusory constructions in Australian Languages. Honours thesis, Department of Linguistics and Applied Linguistics, University of Melbourne. Stassen, Leon 2000 AND-languages and WITH-languages. Linguistic Typology 4: 1–54. Berlin: Walter de Gruyter. Stassen, Leon 2001 Noun phrase coordination. In: Language Typology and Language Universals. Vol. 2. Berlin: Mouton de Gruyter. Stassen, Leon 2003 Noun phrase conjunction. In: F. Plank (ed.), Noun Phrase Structure in the languages of Europe, 761–817. Tryon, Darrell 1970 Conversational Tahitian. Berkeley: University of California Press. Van Engelenhoven, Aone 2004 Leti, a Language of Southwest Maluku. Leiden: KITLV Press. Van Klinken, Catharina L. 1999 A grammar of the Fehan dialect of Tetun, An Austronesian language of West Timor. (Pacific Linguistics C-155.) Canberra: Pacific Linguistics. Van Klinken, Catharina L. 2000 From verb to coordinator in Tetun. Oceanic Linguistics 39-2: 350–363. Van Valin, Robert D. Jr and Randy J. LaPolla 1997 Syntax. Structure, meaning and function. Cambridge: Cambridge University Press.

Part three: Historical developments

Neither accusative nor ergative: an alternative analysis of case in Eastern Polynesian Yuko Otsuka

1. Introduction1 The Polynesian language family consists of five subgroups (Tongic, NuclearPolynesian, Ellicean-Outlier, Easter Polynesian, and Central Eastern Polynesian), as illustrated in Figure 1 (Marck 2000).

Figure 1. Subgrouping of Polynesian languages (based on Marck 2000)

1. I thank Claire Moyse-Faurie, Andrew Pawley, Joachim Sabel, and an anonymous reviewer for helpful comments and suggestions on various earlier versions of this paper. Thanks are also due to Albert Sch¨utz for his help with the Fijian data and Kaliko Baker for providing me with the Hawaiian data.

290

Yuko Otsuka

In terms of syntax, a major divide lies between Eastern Polynesian (EP) and the rest. Traditionally, EP languages such as M¯aori and Hawaiian have been considered to be accusative languages with passives (Bauer 1997; Chung 1978; Clark 1973, 1976; Hohepa 1969; Elbert and Pukui 1979; Ota 1999). At first glance, this appears to be the only appropriate description of these languages. The sole argument of a monadic verb (S) and the agent of a dyadic verb (A) are unmarked, whilst the patient argument of a dyadic verb (O) is marked by a prenominal marker i. Furthermore, these languages have what appears to be a passive construction, in which the verb is suffixed by -Cia (C representing a variable thematic consonant), the patient is u+nmarked, and the agent is marked by a reflex of Proto Polynesian (PPn) *e. The paradigm is schematically represented in (1). (Here and below, I use terms “agent” and “patient” for the sake of simplicity. Strictly speaking, dyadic verbs may involve experiencer instead of agent.) Examples from Hawaiian are provided in (2). (1)

Verbal constructions in Eastern Polynesian intransitive transitive passive

V V V-Cia

e

agt/pat agt agt

i

pat pat

Hawaiian2 (2)

a.

Ua hele au. pfv go 1sg ‘I went.’

b.

Ua ’ai ke kanaka i ka poi. pfv eat def man acc def poi ‘The man ate the poi.’ (Elbert and Pukui 1979: 39)

c.

Ua ho’iho’i ’ia ke ka’a e ia. pfv return pass def car agt 3sg ‘The car was returned by him.’ (Hopkins 1992: 157)

In contrast, outside EP, an ergative pattern is found. In languages such as Tongan and Samoan, S and O are unmarked (or marked by a reflex of PPn *qa), while A is marked by a reflex of PPn *e. 2. Interlinear glosses are mine for the examples cited from Biggs [1969] 1998, Harlow 2001, Hawkins 1982, Hopkins 1992, and Hovdhaugen et al. 1989. In other examples, interlinear glosses are based on the original sources, but some items are slightly modified (e.g., the > def, tam > ns).

Case in Eastern Polynesian

291

Samoan (Mosel and So’o 1997:106, 107) (3) a. S¯a siva le teine. pst dance def girl ‘The girl danced.’ b.

Na pu’e e le leoleo le p¯agot¯a prf catch erg def police def culprit ‘The police caught the criminal.’

Passivization as a syntactic operation does not exist in these languages. Instead, there is a third type of dyadic verbal constructions, called “middle” in the modern Polynesian linguistic literature. It should be noted that what is called “middle” in Polynesian linguistics is different from the more commonly known middle constructions in languages like English (e.g., This book reads well). “Middle” in Polynesian refers to a construction that is semantically transitive (i.e., taking two core arguments), but syntactically intransitive (i.e.,A is marked as absolutive, not ergative). A class of verbs that exclusively occur in the middle pattern includes verbs of perception and emotion and also some other dyadic verbs: e.g., ‘to wait,’ ‘to follow,’ ‘to visit,’ ‘to speak,’ ‘to arrive,’ and ‘to call.’ In middle constructions, A is unmarked (i.e., absolutive), while O is marked by a preposition i ‘in, on, at’ or ki ‘to’, as shown in (4). The relevant paradigm is schematically represented in (5). Tongan (4) a.

b.

(5)

Na’e sio ’a e tamasi’´ı ki he ’akau. pst see abs ref boy.def to ref tree ‘The boy saw a tree.’ ’oku sai’ia ’a e tamasi’´ı ’i he mango. pst like abs ref boy.def in ref mango ‘The boy likes mangoes.’

Verbal constructions outside Eastern Polynesian intransitive ergative middle

V V V

(’a) e (’a)

agt/pat agt agt

i

pat pat

Notice the striking morphological similarity between EP “transitive” and NonEP middle on the one hand and between EP “passive” and non-EP ergative on the other. In this paper, I argue that the difference between EP and non-EP languages is not that the former is accusative and the latter is ergative. Rather,

292

Yuko Otsuka

what separates EP from the rest of Polynesian is an innovation to allow all dyadic verbs to occur in the middle pattern, which resulted in a symmetrical two-way system. In the rest of Polynesian, the ability to occur in the middle pattern is lexically determined. In what follows, I first put forth a historical account of how the two-way system in EP was developed. The proposed analysis supports Clark’s (1973, 1976) view that PPn was ergative, contra the passive-to-ergative hypothesis (Hale 1968; Hohepa 1969; Chung 1978). Section 3 argues that EP languages are not accusative by showing that constructions with i-marked patients in EP are in fact middle constructions. Section 4 argues that EP languages are not ergative, either. Specifically, it will be shown that what has been called “(active) transitive” and “passive” in EP cannot be analyzed as antipassive and canonical transitive (i.e., ergative), respectively. In Section 5, I propose that EP should be analyzed as having a symmetric voice system similar to that of western Austronesian.

2. Two dyadic constructions in Polynesian Clark (1973, 1976) reconstructed two “transitive” constructions for PPn, as given in (6).3 (6)

Clark’s (1973: 569) reconstruction of PPn transitive constructions Pattern 1 Pattern 2

V V

(*-Ci/-Cia)

*e

agt agt

*i/*ki

pat pat

In Tongic and Nuclear Polynesian (excluding EP), whether a verb occurs in Pattern 1 or Pattern 2 is lexically determined. One group of verbs can only occur in Pattern 2 (“ergative” verbs).4 The other group of verbs (“middle” verbs) mainly occurs in Pattern 1, but may also occur in Pattern 2. When middle verbs occur 3. It may be more accurate to call them dyadic constructions, for it is not entirely clear whether these constructions are transitive or intransitive in any or all Polynesian languages. It should also be noted that a class of verbs called “neuter verbs” (Hooper 1984, Maunsell 1894) or “stative verbs” (Biggs [1969] 1998, Pawley 1973) occur in another dyadic pattern, in which the agent or experiencer is marked by i and the patient is unmarked (e.g., (19c) below). See Hooper (1984) for a comprehensive study of neuter verbs in Polynesian. 4. Strictly speaking, some ergative verbs can also occur in Pattern 1 when a partitive meaning is intended, as will be discussed in Section 3.3 below. However, the use of ergative verbs in Pattern 1 is considerably more limited than the use of middle verbs in Pattern 2. Note also that in East Uvean (i) and East Futunan (ii), the oblique marker used with these ergative verbs is ki (Claire Moyse-Faurie, pers.comm.)

Case in Eastern Polynesian

293

in Pattern 2, they are affixed by a reflex of PPn transitive suffix *-Ci (Clark 1973, 1976; Pawley 2001). Thus, Pattern 2 has two variants, one with a suffix and the other without. In EP, there is no clear distinction between ergative and middle verbs. All dyadic verbs may freely occur in either pattern. Based on this observation, Clark proposed that an innovation that defines the EP subgroup is the extension of Pattern 1 to the class of verbs which had been earlier confined to Pattern 2 (i.e., ergative verbs). This innovation gave rise to the apparent accusative pattern, in which A is unmarked and O is marked morphologically. While Clark remained agnostic about the exact nature of the relation between the middle verbs in Pattern 1 and their -*Ci suffixed counterparts in Pattern 2, data from languages like Tongan (Tongic) and East Futunan (Nuclear Polynesian) suggest that -Ci verbs imply that the patient is directly (or more) affected (Biggs 1974; Moyse-Faurie 1992).5 East Futunan (Moyse-Faurie 1992: 220) (7) a. E tusi a Kalala ki le toe. ns point.to abs Kalala obl def.sg child ‘Kalala points out the child.’ b.

E tusi-’i le toe e Kalala. ns point.to-Ci def.sg child erg Kalala ‘Kalala designates the child.’

(i) ’e inu ia Paulo ki te fo’i niu o Soane. npst drink abs Paulo obl textscart cl coco poss Soane ‘Paulo drinks a little of Soane’s coconut.’ (ii) e kai le toe ki f¯a putete. impf eat art enfant obl cl potato ‘The children ate (some of) the potatoes.’ The partitive use of ki may be an independent innovation in these languages. That is, the fact that the less affected patient occurs in the middle pattern (marked by ki) was extended to partitive constructions. Thus, the derived middle pattern is morphologically identical to the inherent middle pattern. This hypothesis is compatible with Moyse-Faurie’s (pers.comm) observation that ki-marked patients in such examples behave like core arguments as are middle objects in these languages. 5. According to Moyse-Faurie (2003: 15), -(C)i suffixation is possible with inherently ergative verbs. When this happens, “O is simply semantically more affected and A is more agentive, but there is no change in valency.” This suggests that in East Futunan, the semantic difference between middle and -Ci transitive verbs has been reanalyzed as the function of -Ci. That is, the suffix has acquired the function to indicate a higher degree of transitivity in the sense of Hopper and Thompson (1980) (i.e., more volitional agent and more affected patient).

294

Yuko Otsuka

Tongan (8) a.

b.

’oku sio ’a e kaka´ı ki he ta’ahin´e. prs see abs ref people.def to ref girl.def ‘The people saw the girl.’ ’oku sio-’i ’e he kaka´ı ’a e ta’ahin´e. prs see-Ci erg ref people.def abs ref girl.def ‘The people stared at the girl (in an offensive way).’

Thus, PPn had a morphological means to derive a transitive verb from a middle verb to indicate a high degree of affectedness of the patient. In this paradigm, less affected patients are morphologically marked by *ki, while affected patients are unmarked. On the other hand, PPn did not have a productive morphosyntactic means to derive a middle verb from a lexically ergative verb to indicate a low degree of affectedness. (See below on the partitive use of *i, however). I propose that Proto Eastern Polynesian (PEP) developed this alternative, namely, the use of *i with lexically ergative verbs to indicate a low degree of affectedness of the patient. It should be noted that Gibson and Starosta (1990) has proposed a similar analysis for M¯aori, which they claim to be ergative (contrary to the commonly held view). Note also that in pairs of sentences like those in (7) and (8), the verb has a suffix when the patient is directly affected. I propose that this, too, was generalized in PEP and led to another innovation, i.e., obligatory use of *-Cia in Pattern 2. The reason why *-Cia is used instead of *-Ci is twofold. First, the reflex of PPn *-Ci was not available in PEP, for *-Ci had no longer been a productive suffix in Proto Nuclear Polynesian (Clark 1973).6 Second, the function of PPn *-Cia was primarily semantic, i.e., indicating that the patient is directly affected. Synchronically, reflexes of *-Cia appear to be very flexible in selection of bases to which it is attached. It can attach to nouns to create stative verbs (e.g., Samoan lago ‘fly’ > lago-ia ‘covered with flies’, af¯a ‘storm’ > af¯a-tia ‘struck by storm’), stative verbs to create active verbs (e.g., Tongan sai ‘good’ > sai-’ia ‘to like’), middle verbs to create ergative verbs (e.g., Tokelauan alofa ‘to love’ > alofa-gia ‘to love’). What these diverse instances of -Cia have in common is a semantic effect, namely, the affectedness of the patient. Because of its semantic property, *-Cia was employed in PEP to mark constructions which had an affected patient. To summarize, I propose the paradigm in (9) for PPn and the one in (10) for PEP. The relevant PEP innovations are (a) the obligatory use of the suffix -*Cia 6. An exception to this generalization is the suffix -’i in East Futunan and East Uvean. In these languages, -’i remained highly productive, although other reflexes of *-(C)i (with C being other than the glottal stop) were no longer productive (Claire Moyse-Faurie, pers.comm.)

Case in Eastern Polynesian

295

in the ergative pattern (to indicate the high degree of affectedness of the patient), which resulted in the loss of the distinction between inherent and derived ergatives, and (b) the ability to derive middle constructions from ergative constructions (to indicate the low degree of affectedness of the patient) by removing -*Cia from the verb and marking the patient with *i. (9)

Dyadic constructions in PPn middle derived ergative inherent ergative

(10)

V V-*Cia V

(*qa) *e *e

agt agt agt

*ki

agt agt agt

*ki

*e

(*qa)

pat [-affected] pat [+affected] pat [+affected]

Dyadic constructions in PEP inherent middle ergative derived middle

V V-*Cia V

*i

pat [-affected] pat [+affected] pat [-affected]

3. Evidence for the present analysis There is substantial evidence to support the current analysis. First, even within EP, there is a clear distinction between inherent middle constructions (with a ki-marked patient) and derived middle constructions (with an i-marked patient). Second, languages such as Rapanui and Pukapukan demonstrate a transitional stage of the proposed historical development: i.e., constructions in which the agent is marked by e, but verbs are without -Cia. Third, both internal and external evidence show that the preposition *i was multifunctional in PPn, making it the logical source of the patient marker in the derived middle construction. Fourth, i-marked “direct objects” in EP exhibit the syntactic characteristics of patients of middle verbs. 3.1. Inherent vs. derived middle constructions in EP As seen in the Tongan and East Futunan examples above, patients of middle verbs are predominantly marked by a preposition ki ‘to’. In Tongan, the majority of middle verbs take ki and only a few verbs take ’i ‘in’ instead: e.g., sai’ia ‘to like’, ifo’ia ‘to like (find tasty)’, fehi’a ‘to dislike’, and m¯alie’ia ‘to enjoy’. It is intriguing that all of these verbs are verbs of liking/disliking and are derived from an adjective: e.g., sai ‘good’, ifo ‘tasty’, m¯alie ‘pleasant’. Adjectives of emotion/sensation generally take ’i-marked cause/source: e.g., fiefia ‘happy’,

296

Yuko Otsuka

ilifia ‘afraid’, ’ita ‘angry’. I assume that PPn middle verbs used *ki to mark the patient and that the use of ’i with some middle verbs in Tongan is an extension of the use of ’i to mark cause/source. (See section 3.2 for various uses of PPn *i and its reflexes.) The distinction between inherent and derived middle verbs can be found in EP languages like M¯aori and Rapanui, in which PPn *k is retained. In these languages, we find ki-marked patients in addition to i-marked patients. I consider the former to be historical relics of PPn middle constructions. M¯aori (11) a.

b.

E m¯ohio ana au kia ia. npst know impf 1sg to.pers 3sg ‘I know him.’ (Biggs [1969]1998: 113) Ka p¯ırangi ia ki ng¯a mea katoa. ns want 3sg to def.pl thing all ‘He wants all the things.’ (Bauer 1997: 197)

Rapanui (Weber 1988: 36) (12)

a.

E haNa era a ia ki te uka . . . impf desire then pers 3sg to def girl ‘When he desired a girl . . . ’

b.

He piri a Taparahi ki te tahi Na poki asp join pers Taparahi to def other pl child ‘Taparahi joined other children.’

3.2. Transitional stage: Rapanui and Pukapukan Data from languages like Rapanui and Pukapukan also support the current analysis. In Rapanui, dyadic verbs may occur in two patterns. In one pattern, A is unmarked and O is marked by i; in the other, A is marked by e and O, unmarked (Alexander 1981; Weber 1988; Du Feu 1996). Unlike other EP languages, however, the verb is not affixed by -Cia in the second pattern. This situation may have arisen if only the first rule (i.e., marking less affected patients with *i) came into effect, but not the second one (i.e., marking affected patients with *-Cia). Rapanui facts suggest that the obligatory use of *-Cia may have been an innovation in Proto Central Eastern Polynesian rather than PEP.

Case in Eastern Polynesian

297

Rapanui (Alexander 1981: 132) Derived middle: (13)

a.

He tiNa’i te vi’e i te taNata prs hit def woman ob def man ‘The woman beats the man.’

Ergative: b.

He tiNa’i te taNata e te vi’e prs hit def man agt def woman ‘The woman beats the man.’

A similar pattern is found in Pukapukan, a Polynesian Outlier language belonging to the Nuclear Polynesian subgroup. In Pukapukan, verbs may or may not be suffixed by -Cia in the ergative pattern, as shown in (14b) and (14c).7 Pukapukan (Chung 1978: 323–324) Derived middle: (14)

a.

Na patu m¯atou i te tamaiti. pst hit 1pl acc def child ‘We hit the child.’

Ergative without -Cia: b.

Na patu te tamaiti e m¯atou. pst hit def child erg 1pl ‘We hit the child.’

Ergative with -Cia: c.

Na patu-a te tamaiti e m¯atou. pst hit-Cia def child agt 1pl ‘We hit the child.’

On the other hand, lexically middle verbs must be suffixed with -Cia when occurring in the ergative pattern, as illustrated in (15). This is expected in the 7. The authors have little to say about the semantic differences between these patterns. For Rapanui, Du Feu (1996) states that the accusative pattern (13a) is neutral and the ergative pattern (13b) is “emphatic”. Rapanui also permits i and e to occur in a single sentence, which, according to De Feu (1996), is also “emphatic”. As for Pukapukan, Chung (1978) notes that there is no clear semantic difference, but the accusative pattern (14a) is formal and polite, the ergative pattern (14b) is informal and casual, and the passive pattern (14c) is neutral.

298

Yuko Otsuka

present analysis. At any stage in the scenario proposed above, middle verbs may not occur in the inherent ergative pattern.8 Pukapukan (Chung 1978: 323–324) Middle: (15)

a.

Ko mina i a-na i te ika prog want nom pron-3sg acc def fish ‘He wants the fish.’

Ergative with -Cia: b.

Ko mina-ngia te yua e-ku. prog want-Cia def water agt-1sg ‘I want the water.’

Ergative without -Cia: c. *Ko mina e-ku te yua. prog want erg-1sg def water ‘I want the water.’ 3.3. Multiple functions of PPn *i and its reflexes If it was *ki that marked the patient in PPn middle constructions, as proposed in (9) above, why did PEP choose *i instead of *ki to mark the patients in derived middle constructions? While the function of PPn *ki was to mark dative ‘to’, instrumental ‘with’, and patients of middle verbs, *i was used to indicate (a) locative ‘in’, (b) source ‘from’, (c) cause ‘because of ’, and (d) partitive. Clark (1976) included “objects of some verbs” as another function of *i, but I consider this function to be subsumed under “cause” or “source” for the reasons 8. Tokelauan is similar to Pukapukan in that a number of middle verbs must be affixed with -Cia when occurring in the ergative pattern: e.g., alofa-gia ‘to love’ > alofa ‘to love’, pelo-gia ‘to deceive’ > pelo ‘to deceive’, and tago-fia ‘to remove’ > tago ‘to touch’ (Hovdhaugen et al. 1989, Sharples 1976). On the other hand, some middle verbs can be used in the ergative pattern without a suffix: e.g., kikila ‘to see’ vs. ‘to supervise’ and teteke ‘to oppose’ vs. ‘to reject’ (Hovdhaugen et al. 1989: 86). (i) a. Na teteke ia ki te fuafuaga. pst oppose 3sg to the plan He opposed the plan.’ b.

Na teteke e pst oppose erg ‘He rejected the plan.’

ia 3sg

te the

fuafuaga. plan

Case in Eastern Polynesian

299

stated above (Section 3.1). In other words, *i was already a multifunctional preposition in PPn and the range of its functions was more readily expandable than that of *ki. Clark (1973: 600) suggests that the use of *i in derived middle constructions in PEP was modeled on the PPn partitive construction, illustrated in (16). In this construction, an inherently ergative verb takes an oblique *i NP to denote that the action partially or incompletely affects the object. (Also see note 3 for examples from East Futunan and East Uvean.) Tongan (Clark 1973: 600) (16) a. Na’e kai ’a e ika ’e he tamasi’i. pst eat abs ref fish erg ref boy ‘The boy ate (up) a fish.’ b.

Na’e kai ’a e tamasi’i ’i he ika. pst eat abs ref boy in ref fish ‘The boy ate some of a fish.’

Although Clark (1973) did not explicitly relate this semantic fact to the PEP innovation of the extension of the middle pattern, his speculation that the partitive construction may have been the source of this innovation is resonant with the current proposal. External witnesses, namely, the Fijian languages, provide further support for the scenario postulated here. In comparison with the relatively limited usage of Proto Oceanic *(q)i (locative and temporal) (Pawley 1973, Ross 1988), its reflexes in Fijian have expanded their range of functions.9 Standard Fijian e and Western Fijian i cover a range of oblique case roles: (a) locative ‘at’, (b) source ‘from’, (c) inanimate cause/instrument/effecter, (d) experiencer, and (e) material (Pawley 1980; Sch¨utz 1985). Standard Fijian (17) a. S¯a coko na ika e na lawa.

asp be.caught def fish prep def net ‘The fish was caught in/by the net.’ (Andrew Pawley, pers.comm. 2007) b.

Âu vana-i koya e na dak¯a¯ı. 1sg shoot-tr 3sg prep def gun ‘I shot him with the gun.’ (Sch¨utz 1985: 341)

9. I thank Andrew Pawley for bringing this point to my attention.

300

Yuko Otsuka

c.

S¯a oca o Filipe e na s¯ausau-mi taro. asp tired pers Filipe prep def answer-tr question ‘Filipe was tired of/from answering questions.’ (Sch¨utz 1985: 341)

d.

S¯a caka e na k¯au¯ na waqa. asp made prep def wood def boat ‘The boat is made of wood.’ (Geraghty 1976: 508)

Wayan Fijian (Andrew Pawley, pers.comm. 2007) (18) Ei macala i-vu¯a.10 prep-3sg asp clear ‘It’s clear to him/He understands.’



These various usages are attested in Polynesian languages, as exemplified by the Tongan data below. Tongan (19) a.

’oku ne ’ita ’i he tamasi’´ı. prs 3sg angry in ref boy.def ‘He is angry with the boy.’

b.

’oku nau puke ’i he mofi teng´ı. prs 3pl sick in ref fever dengue.def ‘They are sick because of dengue fever’

c.

’oku mahino ’iate au. prs clear in.pron 1sg ‘I understand.’ (lit.: ‘it is clear to me.’)

The preposition *i was thus on its way to further expanding the range of its functions at a pre-Polynesian stage.As noted by Clark (1973), PPn *i additionally marked partitives (see (16b) above). Finally, in PEP, the preposition *i acquired another function to mark indirectly (or less) affected patients. 3.4. Syntactic similarities between i-marked patients and patients of middle verbs In EP languages, ki-marked patients of middle verbs and i-marked patients are treated alike in various syntactic phenomena such as passivization and ActorEmphatic (AE) constructions. Oblique NPs cannot participate in these syntactic 10. In Wayan Fijian, when the cause/beneficiary/experiencer is animate, the preposition i becomes i-vu (before third person pronouns) or iva (elsewhere).

Case in Eastern Polynesian

301

operations. This suggests that patients of middle verbs and i-marked patients form a syntactic class separate from obliques. “Passivization” is available for i-marked patients (20) as well as patients of middle verbs (21). In contrast, oblique NPs such as ki-marked instruments may not undergo passivization (22) (Bauer 1997: 197; Chung 1978: 173–174). M¯aori (Bauer 1997: 197) (20) a. Ka pupuhi ia i te manu. tam shoot 3sg obj def bird ‘He shot the bird.’

(21)

(22)

b.

Ka p¯u-hia e ia te manu. tam shoot-pass by 3sg def bird ‘The bird was shot by him.’

a.

Ka p¯ırangi ia ki ng¯a mea katoa. ns want 3sg to def.pl thing all ‘He wants all the things.’

b.

Ka p¯ırangi-tia e ia ng¯a mea katoa. ns want-pass by 3sg def.pl thing all ‘All the things are wanted by him.’

a.

r¯akau ki te toki. Ka tua a ia i te ins def axe ns chop pers 3sg obj def tree ‘He will chop the tree with the axe.’

b. *Ka tua-ina te toki e ia i te r¯akau. ns chop-pass def axe by 3sg obj def tree Intended: ‘The axe was chopped down the tree by him.’ In AE constructions, S or A appears sentence-initially with a genitive marker n¯a (for past tense) or m¯a (for future tense) followed by a Tense/Aspect marker and the rest of the sentence (23a). When a transitive verb (i.e., a form without -Cia) or a middle verb occurs in an AE construction, O is unmarked (23b, c).11 Chung

11. Bauer (1997: 505) claims that the AE construction cannot be used with middle verbs (“experience verbs” in her terms) such as hiahia ‘to want’. She argues that the AE construction can be used with kite if it is used as a transitive verb, meaning ‘to find’, but not if it is used as an experience verb, meaning ‘to see’. Since Chung (1978) provides only one example of the AE construction with a middle verb and it involves kite ‘to find’, her argument on this point may not be taken to be conclusive.

302

Yuko Otsuka

(1978) considers this to be an instance of promotion of direct object to subject.12 Again, this operation treats i-marked patients on a par with patients of middle verbs, but distinguishes these from obliques. As shown in (23d), oblique NPs such as i-marked locatives may not undergo promotion in AE constructions. M¯aori (23) a.

M¯a Pita e haere. poss Peter fut go ‘Peter will go.’ (Biggs 1974: 87)

b.

N¯a H¯one i hanga he whare. poss John pst build a house ‘John built a house.’ (Chung 1978: 179)

c.

. . . ki te wahine n¯a-na i kite taua poro r¯akau. to the woman poss-3sg pst see that piece wood ‘. . . to the woman who had found that piece of wood.’ (Chung 1978: 182)

d. *N¯a H¯one i oma te whare. poss John pst run def house Intended: ‘John ran in the house.’ (Chung 1978: 182) 4. EP is neither accusative nor ergative The preceding discussion has shown that what has been considered to be the canonical transitive construction in EP is parallel to the middle construction in non-EP languages such as Tongan and Samoan. In other words, we do not have a basis to assume that EP languages are accusative, with i being the accusative case marker. Can we regard EP languages as ergative, however? In the ergative analysis of EP, the construction with an i-marked patient (Pattern 1) is regarded as antipassive and the construction with a verb suffixed by -Cia (Pattern 2) is considered to be canonical transitive (i.e., ergative). In this section, I argue that EP languages are not ergative either. What has been regarded as canonical transitive in EP cannot be analyzed as antipassive; nor can what has been called “passive” in EP be regarded as ergative.

12. The relevant promotion rule does not exist in all EP languages. For example, in Hawaiian, Mangarevan, and Rarotongan, the patient argument is still marked by i in AE constructions. See Harlow (1986) for an analysis of AE constructions in EP.

Case in Eastern Polynesian

303

4.1. EP constructions with i-NPs are not antipassive The ergative analysis of EP claims that Pattern 1 with the i-marked patient is antipassive. That is, it is an intransitive construction whose sole core argument is the agent. The patient of antipassives is an optional, oblique NP in a way analogous to the agent of passives. Thus, the ergative analysis predicts that i-marked patients behave like obliques, and not as core arguments.13 The homophony between the “object” marker i and the multifunctional preposition i makes this hypothesis plausible. Gibson and Starosta (1990) argue that Pattern 1 is antipassive in M¯aori and that therefore, M¯aori is ergative. Semantically, i-marked patients are less affected, which is a typical property of antipassives. Syntactically, i-marked patients are treated like obliques in terms of relativization (also see Sinclair 1976). Throughout Polynesian, relativization of oblique NPs typically requires the ai strategy (Chapin 1974; Chung 1978; Massam and Roberge 1997). Relative clauses formed by this strategy contain a resumptive element ai in lieu of a gap. In M¯aori, the ai strategy is used not only with obliques (24a), but also with i-marked patients of verbs like pupuhi ‘to shoot’ (24b). M¯aori (24) a.

b.

te w¯ahi e t¯u ai te kaituki def place tam stand ai def chanter ‘the place where the chanter stands’ (Bauer 1997: 575) te poaka [i pupuhi ai taku matua] def pig pst shot ai poss.1sg father ‘the pig my father shot’ (Biggs [1969] 1998: 159)

Another relativization strategy used for oblique relatives in M¯aori is the genitive strategy (25a), in which the thematic subject of the relative clause is realized as a genitive, seemingly modifying the head noun, and the relative clause contains ai. The genitive strategy is also used to relativize i-marked patients of verbs like k¯ohuru ‘to kill’ in (25b).

13. An anonymous reviewer points out that the patient of antipassives is semantically selected by the verb and therefore is a core argument. Although I acknowledge this position, in this study, I take antipassive to be a syntactic analogue of passive and hence consider it to be an intransitive construction in which the thematic patient is demoted to the oblique status and hence not a core argument.

304

Yuko Otsuka

M¯aori (Bauer 1997: 570, 577) (25) a. te whare a te m¯okai r¯a [i moe ai] def house of def slave dist tam sleep ai ‘the house where the servant slept’ b.

te tangata a Hone [i k¯ohuru ai] def man of John tam murder ai ‘the man that John murdered’

Thus, on the surface, it appears that i-marked patients are treated as oblique. However, careful examination reveals that these two relativization strategies actually distinguish i-marked patients from obliques. While the ai strategy is the preferred method of relativization for obliques, its use with i-marked patients is very limited. According to Bauer (1997: 569–72), it is accepted only in nonpast contexts. Harlow (2001) also notes that many speakers consider the use of ai strategy to relativize i-marked patients as incorrect. On the other hand, the genitive strategy is commonly used to relativize i-marked patients, but its use for oblique relatives is limited. Only some dialects permit its use for oblique relatives and even in those dialects, it is a much less common method than the ai strategy (Bauer 1997: 577).14 Furthermore, another relativization strategy, the AE strategy clearly distinguishes i-marked patients from obliques. As shown in (26), i-marked patients of verbs like wh¯angai ‘to feed’ can be relativized through this strategy, but oblique NPs such as ki-marked goals cannot. M¯aori (Bauer 1997: 198–199) (26) a. Ko t¯enei te k¯otiro n¯a Pani [i wh¯angai top this def girl belong Pani tam feed ‘This is the girl Pani fostered.’

]

b. *te kuia n¯a r¯aua [i hari te harakeke def old.woman belong 3.du tam carry def flax Intended: ‘the old woman they carried the flax to’

]

Ota (1999) argues that constructions with i-marked patients are not antipassive in Hawaiian because i-marked patients are syntactically different from obliques. Despite their identical morphology, emphatic fronting may apply to obliques

14. The genitive strategy is commonly used for i-marked patients as well as obliques in Hawaiian (Hopkins 1992: 232–233, Hawkins 2000). Thus, in Hawaiian, i-marked patients are not distinguished from obliques in this respect.

Case in Eastern Polynesian

305

such as locative i-phrases, as shown in (27a), but not i-marked patients (27b).15 Similarly, in M¯aori, emphatic fronting of i-marked patients is prohibited (Bauer 1997: 49, 57).16 Hawaiian (Ota 1999:119) (27) a. I kahakai i ’ike ai ke keiki i ke kohol¯a. at beach pst see ai def child acc def whale ‘It was on the beach that the child saw the whale.’ b. *I ka mo’o i ’ike ai ka wahine. acc def gecko pst see ai def woman Intended: ‘It is the gecko that the woman saw.’ Recall also that obliques and i-marked patients are treated differently with respect to passivization and AE constructions (section 3.4). Altogether, these facts suggest that i-marked patients are syntactically different from obliques in EP, thereby arguing against the antipassive analysis of Pattern 1. 4.2. EP constructions with e-marked NPs are not ergative The other claim that the ergative analysis makes is that Pattern 2, in which the agent is marked by e, is canonical transitive, with the e-marked NP being a core argument. Chung (1978) provides compelling evidence against this hypothesis. 15. An anonymous reviewer suggests that the inability to front has nothing to do with argument status, but is due to another constraint observed in a number of Austronesian languages, namely, non-NPs front freely, while NPs are more restricted. In this view, we would have to assume two homophonous items in Hawaiian, i as a preposition and i as a case marker (or at least not a preposition) in order to distinguish i-marked locatives, which are PPs and can be freely fronted, from i-marked patients, which are not PPs and hence cannot be fronted. Even if that is the case, the relevant fact would still support the current claim that i-marked patients are different from obliques (by not being PPs). 16. Moyse-Faurie (2004) notes that a similar difference between middle objects and obliques in East Futunan. While oblique i-NPs can be topicalized by fronting (ia), ki-marked middle objects (ib) and ABS objects (ic) must be marked by the predicative marker ko when topicalized. (i) a. I le asotapu e maponopono a koloa. obl def Sunday impf closed abs shop ‘On Sunday, the shops are closed.’ (Moyse-Faurie 2004: 316) b.

Ko le tosi leia, na kau kole atu ki ai. pred def book this pst 1sg ask dir obl anaph ‘This book, I asked for (it).’ (Moyse-Faurie 2004: 311)

c.

Ko le ’aga, na futi e Petelo. pred def shark pst fish.with.a hook erg Petelo ‘The shark, Petelo caught with a hook.’ (Moyse-Faurie 2004: 308)

306

Yuko Otsuka

In M¯aori, the subject of the complement clause of a verb of volition, effort, motion, sending, or command may be deleted by the rule of ki te Equi. This rule applies to subjects of intransitives (28a) and subjects of transitives (28b), but interestingly, not derived subjects of passives (28c). M¯aori (28) a.

b.

haere. Ka whakaaro au ki te ns think 1sg comp17 go ‘I decided to go.’ (Chung 1978:107) E hiahia ana a H¯one ki te patu i ng¯a manu. ns want prog p John comp kill acc def.pl bird ‘John wants to kill the birds.’ (Chung 1978: 112)

c. *I hiahia au ki te patu-a e Rewi. pst want 1sg comp hit-pass agt Rewi Intended: ‘I wanted to be hit by Rewi.’ (Chung 1978: 113) These data suggest that the rule of ki te Equi does not make reference to case or grammatical relation, but selectively applies to agents. Intriguingly, however, e-marked agents cannot be deleted by the same rule, as shown in (29). M¯aori (29) *I hiahia au ki te patu-a te poaka. pst want 1sg comp hit-pass def pig Intended: ‘I wanted to kill the pig.’ (Chung 1978: 114) A possible, and the simplest, explanation for this fact would be that what distinguishes e-marked agents from unmarked agents (as in (28a, b)) is that the former are not core arguments unlike the latter, and that the rule of ki te Equi may target only the agents that are core arguments. In non-EP languages, verbs of volition take a subjunctive clausal complement, in which e-marked agents may undergo equi deletion, as shown in the Tongan example in (30). This suggests that unlike in M¯aori, e-marked agents are core arguments in non-EP languages.

17. Although Chung (1978) treats ki te as a complementizer, it is possible to analyze this sequence as two separate morphemes, the preposition ki and the definite article te. Given that the matrix verb seem to belong to the same class of verbs as middle verbs, we may analyze these sentences as middle constructions in which the object is a nominalization (hence preceded by the definite article te).

Case in Eastern Polynesian

307

Tongan (30) ’oku loto ’a Sione ke langa ’a e fale. prs want abs John sbjv build abs ref house ‘John wants to build a house.’ Additional evidence against the ergative analysis of EP concerns nominalization, where we find a stark contrast between e-marked agents in EP and those in nonEP languages such as Tongan. In Tongan, for example, a possessive pronoun can be used in nominalization to refer to S (31a), e-marked A of an ergative verb (31b), unmarked A of a middle verb (30c), as well as unmarked O of an ergative verb (30d).18 Crucially, however, obliques cannot be replaced with a possessive pronoun (30e). Therefore, the use of possessives in nominalizations can be used as a diagnostic for core arguments and according to this test, e-marked NPs are core arguments in Tongan. (See Moyse-Faurie 2007 for similar data from East Uvean.) Tongan (31) a.

’eku ’alu ki Tonga poss.1sg go to Tonga ‘my going to Tonga’

b.

’eku kai ’a e ika poss.1sg eat abs ref fish ‘my eating a fish’

c.

’eku sio ki he ika poss.1sg see to ref fish ‘my seeing a fish’

d.

hono kai ’e Sione poss.3sg eat erg Sione ‘its being eaten by Sione’

e. *’ene/hono ’alu ’a Sione (ki ai) poss.3sg go abs Sione to ai Intended: its being gone (to) by Sione

18. In Tongan, a separate set of possessive pronouns is used for O (Churchward 1953). Hawaiian also has two sets of pronouns, but they can be used interchangeably in nominalization of nonstative verbs (Wilson 1976). See Chung (1973) for a comparative study of the use of possessives in nominalizations in various Polynesian languages.

308

Yuko Otsuka

In EP, e-marked agents are not treated as core arguments in terms of nominalization. In Hawaiian, for example, possessives may represent S (32a), unmarked A of a transitive verb (32b), as well as unmarked O of a “passive” verb (32c). However, e-marked agents cannot be realized as a possessive in nominalizations (32d). Hawaiian (Kaliko Baker, pers.comm. 2008) (32) a. kona hele ’ana poss.3sg go nmlz ‘his going’ b.

kona a’o ’ana i ke keiki poss.3sg teach nmlz acc def child ‘his teaching the child’

c.

kona a’o ’ia ’ana poss.3sg teach pass nmlz ‘his being taught’

d. *kona a’o ’ia ’ana ke keiki poss.3sg teach pass nmlz def child Intended ‘his teaching the child’ Thus, nominalization treats e-marked agents on a par with obliques in EP. Altogether, the data discussed above suggest that Pattern 2 in EP is not canonical transitive (i.e., ergative). 5. Concluding remarks: an alternative system? To recapitulate, a crucial observation in analyzing verbal constructions in EP is that all dyadic verbs may occur in both Pattern 1 (in which the patient is marked by i) and Pattern 2 (in which the agent is marked by e). In non-EP languages such as Tongan and Samoan, which are generally regarded as ergative, only a certain class of verbs (i.e., “middle” verbs) may occur in Pattern 1. There are a few exceptional verbs that may occur in Pattern 1 with the patient marked by ’i (Tongan) or ki (East Futunan and East Uvean) when partitive meaning is intended To account for this difference, I have posited the following historical development. PPn had two classes of dyadic verbs (ergative and middle) and two dyadic constructions, Pattern 1 in which the patient is marked by *ki, and Pattern 2, in which the agent is marked by *e. The ability to occur in a certain construction was lexically determined: ergative verbs occurred in Pattern 2 and middle verbs

Case in Eastern Polynesian

309

occurred in Pattern 1. PPn also had a morphological means to derive an ergative verb from a middle verb, i.e., -*Cia suffixation.Thus, middle verbs could occur in Pattern 2 when suffixed by -*Cia. Crucially, however, ergative verbs could never occur in Pattern 1, except for a very limited use of partitive constructions.This paradigm, as shown in (32), is retained in most of the languages outside EP. (33) Verb types and construction patterns in PPn

Pattern 1

Pattern 2

V middle

(*qa)

agt

*ki

V ergative

(*qa)

agt

*i

V middle-*Ci(a)

*e

agt

(*qa)

V ergative

*e

agt

(*qa)

pat [-affected] pat [partitive] pat [+affected] pat [+affected]

In PEP, however, the following innovations took place: (a) the productive use of *i to mark less affected patients of ergative verbs; and (b) in Pattern 2, the obligatory -*Cia suffixation to all verbs, including ergative verbs, which were unmarked in PPn. This resulted in the paradigm in (33). (34) Verb types and construction patterns in PEP

Pattern 1

Pattern 2

V middle

agt

*ki

V ergative

agt

*i

V ergative-*Cia

*e

agt

V middle -*Cia

*e

agt

pat [-affected] pat [-affected] pat [+affected] pat [+affected]

In a nutshell, PEP lost the distinction between inherent and derived ergative verbs in Pattern 2 and acquired a syntactic means to allow ergative verbs to occur in Pattern 1. In this paradigm, neither Pattern 1 nor Pattern 2 is morphosyntactically more basic than the other. In that sense, the relationship between the two patterns differs from that between active and passive, or between ergative and antipassive. This observation has led to the conclusion that EP is neither clearly accusative nor unequivocally ergative. As discussed in section 4, empirical data do not provide any conclusive evidence that EP languages (at least M¯aori and Hawaiian) are either accusative or ergative.

310

Yuko Otsuka

The conclusion drawn from the present study is in line with Pucilowski’s (2006) observation about M¯aori, that is, the language has both accusative (Pattern 1) and ergative (Pattern 2) constructions. Pucilowski (2006), however, considers M¯aori to be split ergative and suggests that the controlling factor of the split ergativity in M¯aori is transitivity in the sense of Hopper and Thompson (1980): verbs of high transitivity (e.g., affected patient, dynamic event, perfective aspect) are more likely to occur in the ergative pattern, while the verbs of low transitivity (e.g., non-dynamic event, imperfective, future, unaffected patient) tend to occur in the accusative pattern. She observes that middle verbs occur more frequently in the accusative pattern (Pattern 1) than the ergative pattern (Pattern 2), although the reverse is true with ergative verbs (Pucilowski 2006: 49). The present study identifies three morphological factors, each of which contributes to a different aspect of transitivity. First is the noun marker i. The present analysis proposes that this marker acquired a new function in PEP, that is, to mark the low degree of affectedness of the patient. Second is the suffix -Cia. This suffix was originally used to indicate a high degree of affectedness of the patient in PPn. The often observed fact that -Cia gives rise to the perfective interpretation (Clark 1973; Milner 1973) may be a byproduct of this initial function. Because the affected patient is a factor of high transitivity, this is associated with another factor, telicity. A third factor is the marker e. The original PPn use of this marker was to mark A in ergative constructions. The absence of the marker i in this construction indicates a high degree of affectedness of patient. On the other hand, the presence of e indicates a higher degree of control the agent has (Besnier 1986; Duranti 1990; Duranti and Ochs 1990). What we find in EP languages is curiously reminiscent of the so-called voice/focus system of western Austronesian languages. In these languages, any dyadic verbs may occur in at least two types of constructions freely. One is called the actor voice construction, in which the actor has a special marker (e.g., ang in Tagalog, glossed below as NOM). In the other, patient voice construction, the patient receives this special marking. Verbs have special morphology in both constructions. In other words, morphologically, neither construction is more basic than the other. Some languages also have locative voice, in which the locative has the special marker. Tagalog (Manning 1996: 11) (35) a. B-um-ili ang=lalake ng=isda sa=tindahan prf.av-buy nom=man gen=fish dat=store ‘The man bought fish at the store.’

Case in Eastern Polynesian

b.

B-in-ili-Ø ng=lalake ang=isda sa=tindahan prf.pv-buy gen=man nom=fish dat=store ‘The man bought fish at the store.’

c.

B-in-ilh-an ng=lalake ng=isda ang=tindahan prf.lv-buy gen=man gen=fish nom=store ‘The man bought fish at the store.’

311

It is known that ang-marked NPs in Tagalog (and their equivalents in other languages) demonstrate syntactic prominence. Only ang-marked NPs may undergo certain syntactic operations such as relativization (Keenan 1976; Schachter 1976, among others). Thus, relativization of the patient is permissible if the relevant construction is in patient voice, but not in actor voice. EP languages show a similar constraint on the relativization of patients in that the “passive” strategy is strongly preferred (Bauer 1997; Harlow 1996, 2001). In this regard, Pattern 1 (“active”) in EP is similar to the actor focus in that the actor is the prominent NP. Likewise, Pattern 2 (“passive”) in EP is analogous to the patient focus in that the patient is the prominent NP. This alternative view explains why middle verbs as well as intransitive verbs involving directly affected locatives can be “passivized” in EP. The latter could be regarded as very limited application of locative focus. It should be noted that some have proposed that western Austronesian languages with this type of voice system should be regarded as ergative, with the patient focus being canonical transitive and the actor focus, antipassive (Aldridge 2004; Gerdts 1988; De Guzman 1988; Liao 2004; Payne 1982; Reid and Liao 2004; Starosta 1988, 1999, among others). In this approach, the aforementioned syntactic prominence of ang-NPs is understood as a manifestation of syntactic ergativity: ang-NPs are absolutive and hence some syntactic rules exclusively apply to them. In EP, unmarked NPs manifest similar syntactic prominence compared to e-marked NPs. For example, unlike unmarked NPs, e-marked NPs cannot undergo raising (Chung 1978) or be questioned or relativized (Ota 1999). These facts could be understood as instances of syntactic ergativity: only absolutive (unmarked) NPs and not ergative (e-marked) NPs may undergo certain syntactic operations. Thus, EP languages show a remarkable parallel to western Austronesian languages also in this respect. Given the contrast between EP and non-EP languages within Polynesian, however, it seems premature to conclude that EP languages are ergative. As discussed above, EP languages are clearly different from non-EP languages such as Tongan and Samoan, which are generally considered to be unarguably ergative. In particular, e-marked NPs in the latter clearly exhibit syntactic characteris-

312

Yuko Otsuka

tics of core arguments, including the ability to undergo raising and relativization (Chung 1978). Such difference between EP and non-EP languages should somehow be accounted for. The data from EP point toward the need of an alternative system, which is neither accusative nor ergative.

Abbreviations abs = absolutive, acc = accusativem anaph = anaphor, art = article, asp = aspect, av = actor voice, c = common, cl = classifier, comp = complementizers, dat = dative, def = definite, dir = directional, dist = distal, erg = ergative, exp = experiencer, fut = future, gen = genitive, impf = imperfective, ins = instrument, lv = locative voice, nfut = non-future tense, nmlz = nominalizer, nom = nominative, npst = non-past tense, ns = nonspecific tense/aspect, obj = object, obl = oblique, p = proper, pers = personal, pl = plural, poss = possessive, pred = predicate, prep = preposition, prf = perfect, prog = progressive, pron = pronominal, pv = patient voice, ref = referential, sg = singular, tam = tense/aspect/mood, top = topic, tr = transitive, 1 = first person, 2 = second person, 3 = third person.

References Aldridge, Edith 2004 Ergativity and word order inAustronesian Languages. Ph.D. dissertation, Cornell University. Alexander, James D. 1981 Case-marking and passivity in Easter Island Polynesian. Oceanic Linguistics 20: 131–49. Honolulu: University of Hawai‘i Press. Bauer, Winifred 1997 With William Parker, Te Kareongawai Evans and Te Aroha Noti Teepa. The Reed Reference Grammar of M¯aori. Auckland: Reed. Besnier, Niko 1986 Word order in Tuvaluan. In: Geraghty et al. 1986, FOCAL I: Papers from the Fourth International Conference on Austronesian Linguistics, 245–268. (Pacific Linguistics C-93.) Canberra: Pacific Linguistics. Biggs, Bruce [1969] 1998 Let’s Learn Maori. Auckland: Auckland University Press. Biggs, Bruce 1974 Some problems of Polynesian grammar. Journal of the Polynesian Society 83: 401–426. Auckland: Polynesian Society.

Case in Eastern Polynesian

313

Chapin, Paul G. 1974 Proto-Polynesian *ai. Journal of the Polynesian Society 83: 259–307. Chung, Sandra 1973 The syntax of nominalizations in Polynesian. Oceanic Linguistics 12: 641–686. Honolulu: University of Hawai‘i Press. Chung, Sandra 1978 Case Marking and Grammatical Relations in Polynesian. Austin: University of Texas Press. Churchward, C. Maxwell 1953 Tongan Grammar. London/New York: Oxford University Press. Clark, Ross 1973 Transitivity and Case in Eastern Oceanic languages. Oceanic Linguistics 12: 559–605. Honolulu: University of Hawai‘i Press. Clark, Ross 1976 Aspects of Proto-Polynesian Syntax. Auckland: Linguistic Society of New Zealand. De Guzman, Videa P. 1988 Ergative analysis for Philippine languages: An analysis. In: McGinn 1988, Studies in Austronesian Linguistics, 323–345. Athens, Ohio: Center for Southeast Asia Studies, Center for International Studies, Ohio University. Du Feu, Veronica 1996 Rapanui. London/New York: Routledge. Duranti, Alessandro 1990 Politics and grammar: Agency in Samoan political discourse. American Ethnologist 17-4: 646–666. Davis, CA: American Ethnological Society. Duranti, Alessandro and Elinor Ochs 1990 Genitive constructions and agency in Samoan discourse. Studies in Language 14: 1–23. Amsterdam and Philadelphia: John Benjamins. Elbert, Samuel H. and Mary Kawena Pukui 1979 Hawaiian Grammar. Honolulu: University of Hawai‘i Press. Geraghty, Paul 1976 Fijian Prepositions. Journal of the Polynesian Society 85: 507–520. Auckland: Polynesian Society. Geraghty, Paul, Lois Carrington and S.A. Wurm (eds.) 1986 FOCAL I: Papers from the Fourth International Conference on Austronesian Linguistics. (Pacific Linguistics C-93.) Canberra: Pacific Linguistics. Gerdts, Donna B. 1988 Antipassives and causatives in Ilokano: Evidence for an ergative analysis. In: McGinn 1988, Studies in Austronesian Linguistics, 295–321. Athens, Ohio: Center for Southeast Asia Studies, Center for International Studies, Ohio University.

314

Yuko Otsuka

Gibson, Jeanne and Stanley Starosta 1990 Ergativity east and west. In: Philip Baldi (ed.), Linguistic Change and Reconstruction Methodology, 195–210. Berlin: Mouton de Gruyter. Hale, Kenneth 1968 Review of Hohepa 1967. Journal of the Polynesian Society 77: 83–99. Auckland: Polynesian Society. Harlow, Ray 1986 The actor emphatic construction of the Eastern Polynesian languages. In: Geraghty et al. 1986, FOCAL I: Papers from the Fourth International Conference on Austronesian Linguistics, 297–308. (Pacific Linguistics C-93.) Canberra: Pacific Linguistics. Harlow, Ray 1996 M¯aori. M¨unchen and Newcastle: LINCOM Europa. Harlow, Ray 2001 A M¯aori Reference Grammar. Auckland: Longman. Hawkins, Emily 1982 Pedagogical Grammar of Hawaiian: Recurrent Problems. Honolulu: Hawaiian Studies Program, University of Hawai‘i at M¯anoa. Hawkins, Emily 2000 Relative clauses in Hawaiian. In: Steven Roger Fischer and Wolfgang B. Sperlich (eds.), Leo Pasifika: Proceedings of the Fourth International Conference on Oceanic Linguistics, 127–141. Auckland: Institute of Polynesian Languages and Literatures. Hohepa, Patrick 1969 The accusative-to-ergative drift in Polynesian languages. Journal of the Polynesian Society 78: 295–329. Auckland: Polynesian Society. Hooper, Robin 1984 Neuter verbs, stative aspect, and the expression of agency in Polynesian. Journal of the Polynesian Society 93: 39–70. Auckland: Polynesian Society. Hopkins, Alberta Pualani 1992 Ka lei ha‘aheo: Beginning Hawaiian. Honolulu: University of Hawai‘i Press. Hopper, Paul J. and Sandra A. Thompson 1980 Transitivity in grammar and discourse. Language 56: 251–299. Hovdhaugen, Even, Ingjerd Ho¨em, Consulata Mahina Iosefo andArnfinn MuruvikVonen 1989 A Handbook of the Tokelau Language. Oslo: Norwegian University Press, Institute for Comparative Research in Human Culture. Distributed by Oxford University Press. Keenan, Edward L. 1976 Remarkable subjects in Malagasy. In: Li, Subject and Topic, 247–301. New York: Academic Press

Case in Eastern Polynesian

315

Li, Charles (ed.) 1976 Subject and Topic. New York: Academic Press. Liao, Hsiu-chuan 2004 Transitivity and ergativity in Formosan and Philippine languages. Ph.D. dissertation, University of Hawai‘i at M¯anoa. Manning, Christopher D. 1996 Ergativity: Argument Structure and Grammatical Relations. Stanford, CA: CSLI Publication. Marck, Jeff 2000 Topics in Polynesian Language and Culture History. Canberra: Pacific Linguistics. Massam, Diane and Yves Roberge 1997 Operator-bound clitics and Niuean ai. In: Jim Black and Virginia Motapanyane (eds.), Clitics, Pronouns and Movement, 273–300. (Current Issues in Linguistic Theory 140.) Amsterdam and Philadelphia: John Benjamins. Maunsell, Robert 1894 Grammar of the New Zealand Language (4th edition). Auckland: Upton and Co. McGinn, Richard (ed.) 1988 Studies in Austronesian Linguistics. Athens, Ohio: Center for Southeast Asia Studies, Center for International Studies, Ohio University. Milner, George 1973 It is aspect (not voice) which is marked in Samoan. Oceanic Linguistics 12: 621–639. Honolulu: University of Hawai‘i Press. Mosel, Ulrike and Ainslie So’o 1997 Say It in Samoan. (Pacific Linguistics D-88). Canberra: Pacific Linguistics. Moyse-Faurie, Claire 1992 Verb classes and argument structure variation in Futunan. Oceanic Linguistics 31: 209–228. Honolulu: University of Hawai‘i Press. Moyse-Faurie, Claire 2003 The ergative features of Papuan and Austronesian languages. In Ergatividae na Amazˆonia (II), Centre d’Etudes des langues indig`enes d’Am´erique (CNRS, IRD), Laborat´orio de L´ınguas Ind´ıgenas (UnB). Moyse-Faurie, Claire. 2004 Convergence entre th`eme et focus dans les langues polyn´esiennes. Faits de Langues 23–24: 303–320. Paris: Ophrys Moyse-Faurie, Claire. 2007 Les formes nominalis´ees du verbe dans quelques langues oc´eaniennes. Faits de langues 30: 97–116. Paris: Ophrys

316

Yuko Otsuka

Ota, Katsuhiro J. 1999 Aspects of case-marking and transitivity in Polynesian. PhD dissertation. University of Hawai‘i at M¯anoa. Pawley, Andrew 1973 Some problems in Proto-Oceanic grammar. Oceanic Linguistics 12: 103–188. Honolulu: University of Hawai‘i Press. Pawley, Andrew 1980 On the history of Polynesian transitive clauses. Unpublished manuscript. Pawley, Andrew 2001 Proto-Polynesian *-CIA. In: Joel Bradshaw and Kenneth Rehg (eds.), Issues in Austronesian Morphology, 193–216. Canberra: Pacific Linguistics. Payne, Thomas E. 1982 Role and reference related subject properties and ergativity in Yup’ik Eskimo and Tagalog. Studies in Language 6: 75–106. Amsterdam and Philadelphia: John Benjamins. Pucilowski, Anna 2006 Split-ergativity in M¯aori. MA thesis. University of Canterbury. Reid, Lawrence A. and Hsiu-chuan Liao 2004 A brief syntactic typology of Philippine languages. In: Elizabeth Zeitoun (ed.), Language and Linguistics 5: 433–490. (Linguistics Papers from the International Symposium on Austronesian Cultures: Issues Relating to Taiwan). Taipei: Academia Sinica Ross, Malcolm D. 1988 Proto Oceanic and the Austronesian Languages of Western Melanesia. (Pacific Linguistics C-98). Canberra: Pacific Linguistics. Schachter, Paul 1976 The subject in Philippine languages: Actor, topic, actor-topic, or none of the above. In: Li, Subject and Topic, 491–518. New York: Academic Press. Sch¨utz, Albert J. 1985 The Fijian Language. Honolulu: University of Hawai’i Press. Seiter, William J. 1980 Studies in Niuean Syntax. New York: Garland. Sharples, Peter R. 1976 Tokelauan syntax: Studies in the sentence structure of a Polynesian language. Ph.D. thesis, University of Auckland. Sinclair, Michael B.W. 1976 Is Maori an ergative language? Journal of the Polynesian Society 85-1: 9–26. Auckland: Polynesian Society. Starosta, Stanley 1988 A grammatical typology of Formosan languages. Bulletin of the Institute of History and Philology 59: 541–576. Taipei: Academia Sinica.

Case in Eastern Polynesian

317

Starosta, Stanley 1999 Transitivity, ergativity and the best analysis of Atayal case marking. In: Elizabeth Zeitoun and Paul Jen-kuei Li (eds.), Selected Papers from the Eighth International Conference on Austronesian Linguistics (8ICAL), 371–392. Taipei: Academia Sinica. Weber, Nancy L. 1988 Case marking in Rapa Nui, the Polynesian language of Easter Island. MA thesis, University of Texas at Arlington. Wilson, William H. 1976 The o and a possessive markers in Hawaiian. M.A. Thesis, University of Hawai‘i at M¯anoa.

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect Jacques Vernaudon

Like other Polynesian languages, Tahitian, an Eastern Polynesian language of the Austronesian family, spoken primarily in the Society Archipelago in French Polynesia, presents very weak lexical pre-categorization. A priori no morphological index allows a specific class to be allocated to a lexical morpheme. As Moyse-Faurie (2005: 20) stresses, in Polynesian languages, all lexemes can appear either in a predicative function or a non predicative function1 . For example in Tahitian, actantial functions are also accessible to every lexeme (including those that refer to processes), more particularly thanks to the article te (Vernaudon and Rigo 2004). Based on a distributional analysis combined with a frequency of use study according to the syntactic environment, it is indeed possible to identify contexts where such or such a lexeme is more likely to occur. Furthermore, if it is acknowledged that the meaning of lexemes contributes to their typology, it may be observed that some of them refer more to processes and others to entities. For example, tamaiti ‘boy’ is found more often in non-predicative environments with a non-processive value. However, such an attribute remains contingent; the majority of the lexemes are highly labile and can, without any prior morphological modification, appear in another syntactic context and refer more to a process while previously corresponding to an entity, or vice versa. Thus, in the example given below, tamaiti is found in a predicative context accompanied by an aspect marker to refer to the process of “becoming a boy”. (1)

E ’ua tamaiti a’e ra Ta’aroa. and trs boy dir dx Taaroa ‘And Taaroa became a boy.’ (TH: 343)2

1. I wish to sincerely thank Claire Moyse-Faurie for her precious help in preparing this article. 2. The source of data indicated by abbreviations (e.g., TH, GF, PAA) refers to our personnal corpus of authentical utterances. The texts are not published. When there is no source indicated, the example is generated by the author.

320

Jacques Vernaudon

In both the production and comprehension of language, the semantic coding and decoding of a lexeme’s referential value therefore relies on additional devices: The use of grammatical morphemes that mark out the interpretation. For example, aspectual determinations are typically associated with processes (ex. 1), in both predicative and non-predicative contexts. The position of lexical items in a syntactic construction. For example, as will be seen in more detail later in this text, ta’ata ‘human being’ when positioned second in a syntactic construction is immediately interpreted as expressing a property (ex. 3) and not as denoting an entity (ex. 2): (2)

E ta’ata. cls human

‘It is a human being.’

(3)

E reo ta’ata. cls voice human

‘It is a human voice.’

We are going to demonstrate how mea, that is at first a lexical morpheme the semantic content of which is only weakly determined, grammaticalised into an introducer of qualifiers; and then how it enter into the paradigm of aspect markers to evoke actions characteristic of an individual or a group.

1. The lexical value of mea The lexeme mea occurs frequently in the Tahitian language. The Dictionary of the Tahitian Academy (1999: 266) gives the following definition for it: “1. 2. 3. 4.

Thing, object; Used in front of an adjective to imply the noun it qualifies; Such-and-such (to replace a proper noun or a number); Doing/Making, replacing a verb that you do not wish to say or do not know how to say.”

This is a term that belongs to the common lexical stock of Polynesian languages, as witnessed in the extracts below: • Cook Islands Maori (Buse & Taringa 1995: 244–245) 1. (a) Thing, not necessarily physical object (often used because of lack of information instead of a more precise word). (b) Used essentially as a proform, or as a vague place holder for a noun where the following attribute is semantically more important.

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect

321

2. So-and-so, what’s-it-called, what’s-his-name. Also used (impolitely) in addressing unidentified people. 3. Used with wide range of meanings, almost as a pro-verb (do, make). • Tuamotuan (Stimson 1964: 299) “Predicative usage: to do, deal with, perform ; a non-descript predicative substituting for any action mentioned, understood, or implied. To keep on doing an act [which is] understood Substantive usage: A thing, object A supposition; a fact, event, case; a reason, cause. A quantity. A person; so-and-so; the one referred to; the other, others; the unnamed-, unmentioned- one.” • Marquesan (Le Cl´eac’h 1997: 86) “A term that replaces a common noun, proper noun or verb. Thing, matter, substance, object of some sort.” • East Uvean (Rensch 2002: 160), for the cognate me’a: “s. Thing, matter v. To come, to go, to do/make” • Samoan (Milner 1993: 142) “Thing, object” • East Futunan (Moyse-Faurie 1993: 296), for the cognate ne’a: “thing, individual.” • In Tongan (Yuko Otsuka, p. c.), too, the cognate me’a is basically a noun meaning ‘thing’, but can be used to replace nouns, adjectives, or verbs, just as in Tahitian. (4)

Na’e ha’u ’a me’a. pst come abs thing ‘What’s-his-name came.’

(5)

Na’e me’a ’a Sione. pst thing abs John ‘John did it/whatever.’

Mea is frequently translated as ‘thing’ and this is indeed the meaning that can be given to it in a number of Tahitian examples such as:

322

Jacques Vernaudon

(6)

Aore te ho’¯e mea i toe. neg art one thing pfv remain ‘There is nothing left.’

(7)

I te mahana hope’a, ’ua ineine te mau mea ato’a. loc art day last pfv ready art pl thing all ‘The last day, everything was ready.’

It can be noted that in the possessive form with o, mea tends to be interpreted as denoting the male genital organs. Conversely, with the possessive marker a no specific value is given more importance3 . (8)

t-o ’oe mea art-poss.o 2sg thing ‘your thing = your sex’

(9)

t-a ’oe mea art-poss.a 2sg thing ‘your thingumajig, the thingumajig you are talking about’

As the majority of the preceding definitions indicate, ‘thing’ is only a partial translation equivalent to explain some of the lexical usages of mea. In fact, mea is first and foremost characterised by its semantic vagueness. Owing to such vagueness, it can take the place of any lexical item in an utterance from the moment the utterer does not wish to or is unable to give an accurate qualitative contour to whatever he or she is talking about, whether this be a person, a thing or a process. Mea is a prime example of lability in lexemes in Tahitian. Thus, mea may be used in place of a proper noun: (10)

E mea m¯a, eita e nehenehe e fa’ahuru vitiviti hey thing coll neg ao can ao make speed t-a t¯atou paraparaura’a? art-poss 1pl.incl discussion ‘Hey you, can we speed up our discussion?’ (GF: 4)

3. There are two distinct possessive classes (a and o) in Tahitian. Generally, o indicates a relation that is either seen as essential, inherent, natural, or permanent and where the possessor may be subordinate. A expresses a contingent relation with a more agentive possessor.

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect

(11)

323

Nehenehe ’o mea e pata i te tahi h¯oho’a no can pres thing ao release obj art one picture of ’outou? 2pl ‘Can “what’s-his-name” take a picture of you?’ (PAA: 11)

The previous example can be compared with: (12)

Nehenehe ’o Teva e pata i te tahi h¯oho’a no can pres Teva ao release obj art one picture of ’outou? 2pl ‘Can Teva take a picture of you?’

Mea may refer to an action that nothing is being said about. It is the utterance’s context and the co-text that gradually enables a specific value to be inferred for mea in the utterance. i mea roa i te reira parau. (13) Aita m¯aua neg 1du.excl pfv thing very obj art ana say ‘We did not really pay attention to this story.’ (PAA: 2) Compare example (13) with the following: (14) Aita m¯aua i t¯au’a roa i te reira parau. neg 1du.excl pfv pay.attention very obj art ana say ‘We did not really pay attention to this story.’ Mea also occurs in spoken language, with a phatic function, when the utterer hesitates. It is then used as a device to compensate for a memory blank or lapse and signals that the utterer is going to correct his/her own discourse. There again we see proof of its role as a provisional substitute. (15)

E tamari’i m¯atou, haere e parau far¯an . . . mea . . . parau cls children 2pl go ao speak French thing speak tahiti m¯atou. Tahitian 2pl ‘We were young, we used to speak Fren . . . er . . . speak Tahitian.’ (PAA: 12)

324

Jacques Vernaudon

We can also note the occurrence of mea in expressions such as mai te mea e¯ . . . ‘if . . . ’, no te mea e¯ . . . ‘because . . . ’, etc., where it is a cataphoric substitute for thought content clarified in a postposed clause. (16)

’Ua ho’i mai ’¯ona no te tauturu mai. pfv return all 3sg purp art help all ‘He came back to help (us).’

(17)

’Ua ho’i mai ’¯ona no te mea e¯ , e hina’aro pfv return all 3sg purp art thing compl ao want ’¯ona i te tauturu mai. 3sg obj art help all ‘He came back because (lit.: for the thing that is) he wanted to help (us).’

In accordance with Buse & Taringa (1995) in their Cook Islands Maori Dictionary, mea may be considered as a pro-form. Nonetheless, it should be stressed that such a role is perfectly transversal. In a language such as English, pro-forms whose semantic content is only weakly determined are attached to specific lexical classes. For example, do is used in the case of verbs and thing in the case of nouns. On the other hand, in Tahitian, mea occupies indiscriminately gaps where it commutes either with lexemes that refer to entities (ex. 6 to 9), or with proper nouns (ex. 10 and 11), or with lexemes that denote processes (ex. 13). From a syntactic point of view, it can equally occupy predicative and non-predicative functions. This is why we have no hesitation in presenting mea as the lexical “joker” of the Tahitian language.

2. Attributing a Quality In the absence of extensive lexical pre-categorization, the Tahitian language has to resort to other devices to attribute specific functions to lexemes in an utterance. One such device, that is here called a positional device, is based on the position of the lexeme in the syntactic construction. Thus, when there are two lexemes in succession in a syntactic construction, the second one qualifies the first one. On this subject, in theirTahitian Grammar Lazard and Peltzer (2000: 22) introduce the concept of a “qualificative form” giving it the following definition: “A qualificative form is a lexeme that directly follows another lexeme (without any particle).”

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect

325

It is therefore the position of a lexeme following another lexeme that gives it its qualificative function, as is the case in maita’i, tauturu and moni in the following examples: (18)

E ta’ata maita’i ’¯ona. cls human good 3sg ‘He is good.’

(19)

E ta’ata tauturu ’¯ona. cls human help 3sg ‘He is helpful.’

(20)

E ta’ata moni ’¯ona. cls human money 3sg ‘He is rich.’

Conversely, introducing, for example, moni straight off in the first lexical position in a syntactic construction leads to it being interpreted as denoting class inclusion (ex. 22) and no longer as expressing a quality (ex. 21). (21)

E ta’ata moni. cls human money ‘(He) is a rich person.’

(22)

E moni. cls money ‘(It) is money.’

Furthermore, permutating the terms in a syntactic construction brings about a reversal of the qualification relation. (23)

E reo ta’ata. cls voice human ‘(It) is a human voice.’

(24)

E ta’ata reo. cls human voice ‘(He) is a person who has a good voice.’

326

Jacques Vernaudon

In example (23), ta’ata qualifies reo by making it clear that it is a human voice rather than an animal’s cry that is being referred to. In example (24), it is reo that qualifies ta’ata by attributing voice qualities to such and such a person. The canonical structure of the syntactic unit is therefore the following: modified noun – qualifier. It is important to stress that the status of the qualifier is not pre-determined (contrary to the case of adjectives in English, for example) but results in the lexeme being juxtaposed. The juxtaposed qualifying element may express a physical or moral quality (e.g. fare rahi ‘big house’, ta’ata itoito ‘brave person’), the content (e.g. mohina pape ‘bottle of water’), the matter or the components (e.g. ’¯afata r¯a’au ‘wooden box’, hei tiare ‘floral wreath’), a species (e.g. tumu v¯ı ‘mango tree’), the origin (e.g. reo tahiti ‘Tahitian language’), gender (e.g. ’¯ı’¯ıt¯a ’¯ot¯ane ‘male papaya tree’), purpose or use (e.g. fare ma’i ‘hospital’), the means of propulsion (e.g. va’a t¯a’ie ‘sail-propelled canoe’), etc. If we focus our attention on the attributive form in Tahitian (i.e. the form that allows a quality to be attributed to the subject of an utterance), this appears as a particular case in the inclusive form (i.e. the form that allows a subject to be included in a class). e X : ‘this is an X’ e X Y : ‘this is an X that has the quality Y’ This phenomenon is clearly illustrated by example (18) above, for which the exact gloss is ‘he is a good human being’, and not simply ‘he is good’. The e particle in this inclusive/attributive structure does not determine any tense, aspect or modal relations, or definiteness, gender or number. It simply introduces the class in which the subject is included. The time interval for which this identification is valid is deduced from the context, unless it is clarified by means of an explicit temporal adverbial. (25)

I tera area matahiti, e ihitai ’¯ona. loc dem area year cls seaman 3sg ‘In those years, he was a seaman.’

(26)

I ter¯a ra tau, e ta’ata maita’i ’¯ona. 3sg loc dem deict time cls human good ‘In those times, he was a good person.’

The positional device presented here does not appear to be applied so regularly in another Polynesian language from the Eastern Branch such as M¯aori.

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect

327

Indeed, in this language the equivalent of the Tahitian inclusive marker e is he. In her M¯aori Grammar, Winifred Bauer (1997: 28–29) gives the name “classifying particle” to the marker he (marked “cls” in the line-by-line translation): “The he-phrase identifies the class to which the Subject of the sentence belongs.” Thus, for example, people say: M¯aori (27) He kaiako ia. cls teacher 3sg ‘She is a teacher.’ However, in M¯aori, qualifiers can also be introduced immediately after he. M¯aori (28) He nui te whare. cls big art house ‘The house is big.’ Such a structure is impossible in Tahitian. Therefore, one has to say: Tahitian (29) E mea rahi te fare. cls thing big art house ‘The house is big’ Conversely, people cannot say, taking the model of the preceding M¯aori example: Tahitian (30) *E rahi te fare. cls big art house Example (30) can only be uttered provided e is no longer interpreted as the inclusive particle, but as the homonymous morpheme e, the aorist marker. In this case, the meaning obtained is ‘this house will get bigger’. Such an utterance can only be envisaged in a supernatural story. This time rahi no longer refers to a gnomic quality, but a future state, achieved subsequent to a growth process. In order to convince ourselves that we are indeed dealing with two e homonyms, one that marks inclusion in a class and the other the aorist, we may

328

Jacques Vernaudon

observe that the first e commutes with ei, which corresponds to the irrealis form of inclusion: (31)

E uaina. cls.rea wine ‘It’s wine.’

(32)

Ei uaina. cls.irr wine ‘It shall be wine.’

(33)

E ta’ata itoito ’oe. cls.rea human brave 2sg ‘You are a brave one.’

(34)

Ei ta’ata itoito ’oe. cls.irr human brave 2sg ‘Be a brave one.’

Conversely, the aspectual e commutes with the other aspectual particles. (35)

E reva te pah¯ı i te hora pae. ao leave art boat loc art hour five ‘The boat leaves at five.’

(36)

’Ua reva te pah¯ı i te hora pae. pfv leave art boat loc art hour five ‘The boat has left at five.’

A comparison with M¯aori upholds this analysis. The Tahitian e and the ei, realis and irrealis inclusion markers, have the cognates he and hei respectively in M¯aori. M¯aori (37) He kaiako ia. cls.rea teacher 3sg ‘She is a teacher.’

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect

(38)

329

Hei kaiako ia. cls.irr teacher 3sg ‘She is going to be a teacher.’

Conversely, the cognate for the Tahitian aorist e in M¯aori is the imperfective particle e (Bauer 1997: 91). M¯aori (39) E

tangi he p¯u i ng¯a p¯o katoa. sound a flute at art.pl night all ‘A flute plays every night.’ IPFV

3. The role of mea as a saturator in the attributive form Mea, which as we have seen has only weak semantic value, can opportunely take the place of the first qualified lexeme in an attributive structure. It thereby allows the place of the modified lexeme to be saturated and is juxtaposed with the second qualifying lexeme: (40)

E ta’ata maita’i. cls human good ‘(He) is (a) good person.’

(41)

E mea maita’i. cls thing good ‘(He/it) is good.’ (this form can be used either for human or non-human, animate or inanimate)

The same phenomenon can be found in a non-predicative context, where mea makes it possible to maintain a qualificative interpretation of the lexeme which juxtaposes it. In examples 42 and 43, ’uo’uo ‘white’, juxtaposed, provides information about the colour in its role as a complement. However, in the absence of mea (ex. 44), the lexeme ’uo’uo loses its qualifying function and ends up referring to the colour itself: (42)

te puta ’uo’uo art book white ‘the white book’

330

Jacques Vernaudon

(43)

te mea ’uo’uo art thing white ‘the white one’ (i.e. the one that is white)

(44)

te ’uo’uo art white ‘white’ (i.e. the colour itself)

C. Moyse-Faurie (p.c.) points out the equivalent role of ne’a ‘thing’ in East Futunan; thus, le ne’a tea ‘the white thing’ is in contrast to le tea ‘white’ (i.e. the colour itself). As seen previously, the realis inclusive particle e can commute with the irrealis inclusive particle ei. (45)

Ei mea maita’i. cls.irr thing good ‘It shall be a good thing.’

Commutation with aspectual markers is unlikely in this context: *’Ua mea maita’i. Therefore, it is still a question of the inclusive structure e XY ‘this is an X that has the qualityY’, where mea takes the place of the term X. The form e mea, that associates the inclusive particle e with the lexeme mea, has therefore become the most productive attributive pattern in Tahitian. Any lexeme introduced by e mea is interpreted as a qualifier. (47)

E mea moni teie puta. cls thing money dem book ‘This book is expansive.’

(48)

E mea tupa i te p¯o toru. cls thing crab loc art night three ‘There is a lot of crabs the third night after the full moon.’

(49)

E mea ’amu te fe’e. cls thing eat art octopus ‘Octopus is edible.’

We have seen that the inclusive particle e does not contribute to specific temporal determination. The attributive form e mea can be used in a past or present

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect

331

temporal context when it is a question of a stable property in a significant time interval. (50)

I to ’u na’ina’ira’a, e mea i’a roa te loc art-poss 1sg youth cls thing fish very art ava. channel ‘In my youth, the channel was really full of fish.’

Other turns of phrase may be preferred to express a more isolated qualification. (51)

Inap¯o, ’ua i’a mai i roto i te ava. yesterday.night pfv fish all loc inside loc art channel ‘Yesterday night, the channel was really full of fish.’

In short, four additional structures can be identified in Tahitian that correspond to affirmative and negative forms of inclusion or attribution. Table 1.

Classifying form Attributive form

Affirmative Negative Affirmative Negative

Tahitian eX e’ere i te X e mea X e’ere i te mea X

English ‘It is (a) noun’ ‘It is not (a) noun’ ‘It is adjective’ ‘It is not adjective’

In each of these structures, it is the position of the lexeme X in the syntactic construction that confers on it a classifying or a qualifying value according to whether it is either immediately after the inclusive particle e, or in second position, after mea (or any other lexeme) respectively. (52)

E i’a. cls fish ‘It is (a) fish.’

(53)

E’ere i te i’a. neg loc art fish ‘It is not (a) fish.’

(54)

E mea i’a. cls thing fish ‘It is full of fish.’

332

Jacques Vernaudon

(55)

E’ere i te mea i’a. neg loc art thing fish ‘It is not full of fish’.

The use of a morpheme signifying ‘thing’ as a saturator in itself is nothing unusual. Equivalent phenomena can be found in numerous languages (cf. the English anything, something, nothing, etc.). In Xˆarˆac`uu` , a Kanak language spoken in the South of the New Caledonian Mainland, d¨ou ‘thing’allows the position of a dependent noun modifier that does not have a free form to be saturated. E.g. kw´ee´ -d¨ou ‘drawing’ (lit.: image-thing) (Moyse-Faurie 1995: 26). However, d¨ou ‘thing’ is also used optionally between a modified term and a qualifier: x¨ou d¨ou chˆaba ‘thick clothing’ (lit.: thick thing clothing) (ibid., p. 42). Closer to Tahitian, in Samoan, Mosel and Hovdhaugen (1992: 516) present the turn of phrase ‘’O (le) mea + modifying attribute ’o + nominalised verbal clause’, in which mea introduces a qualifier. (56)

’O le mea sili ’o le inu fua. pres art thing best pres art drink without.restriction ‘The best thing is to drink without any restriction.’

Though it is not a common pattern in Tongan, Yuko Otsuka (p.c.) indicates one exception in a fixed expression, me’a pango ‘it is unfortunate that . . . ’: Tongan (57) Me’a pango kuo ’osi ’alu ’a Sione. thing unfortunate pfv already go abs John ‘It’s too bad that John has already left.’ 4. Transformation of e mea into mea ∼ [me:] In spoken language, the Tahitian attributive form e mea tends to become [me:], or even [me]: (58)

E mea moni cls thing money ‘It is expensive’ is frequently pronounced [me:moni].

This transformation into [me:], that is a grammaticalization index, only occurs in this context. When mea is used with a lexical meaning, it is always pronounced [mea]. The written language is more conservative and a transcription

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect

333

of the grammaticalised mea cannot yet be found in the form *me. However, the inclusive particle e is very often made to disappear. Thus, e mea moni will be transcribed as mea moni. Simplification phenomena, in spoken language, can also be observed in the negative form e’ere i te mea that often takes the form [e’ereme:]. The following table presents the various oral contemporary forms: Table 2.

Classifying form Attributive form

Affirmative Negative Affirmative Negative

Base Forms eX e’ere i te X e mea X e’ere i te mea X

Oral Contemporary Forms [e X] [e’ere e X] [me: X] [e’ere me: X]

In the oral forms, e’ere, that originates from the negative verbe ’ere, behaves like a negative particle, which is added to the corresponding affirmative sentence. This generalization extends to attributive forms; e’ere is simply added to the corresponding affirmative form [me: X]. An interesting phenomenon results from the truncation from e mea to mea, often pronounced [me:] – the emergence of a specific stative marker in Tahitian. Placed in front of a lexeme, the morpheme mea ∼ [me:] allows the latter to act like a qualifier (without necessarily being an adjective a priori). It is therefore in contrast to the inclusive marker e that introduces a class showing inclusion. e X : ‘this is (a) noun’ (X refers to the class the subject belongs to) mea X : ‘this is adjective’ (X refers to a quality attributed to the subject) Evidence of grammaticalization equivalents to mea can be found in other Polynesian languages of the Eastern Branch. Thus, Cablitz (2006: 61), in Marquesan, has no hesitation in making a synchronic contrast between the lexical mea that means ‘thing’ and the aspectual mea that she defines as follows: “Mea occurs in the position of TAM-particles. It forms verbal phrases in particular together with state and experience verbals and expresses that something or someone is in a certain state.” We have transcribed one of the Marquesan examples cited by the author, where mea appears as both an aspectual marker and a lexeme. Marquesan (59) Mea meita’i oko te-na mea. good very art-dem thing ‘That thing is very good.’

334

Jacques Vernaudon

In a note (2006: 626), Cablitz adds: “The particle mea is probably cognate with mea ‘thing’.”

5. Aspectualization of the Attributive Form and Diathesis Alternations While (e) mea ∼ [me:] specialises as an attributive form, it ends up entering paradigms of aspectual particles. This is what we are now going to observe. As we have seen previously, the form (e) mea ∼ [me:] can introduce all lexemes, including those that a priori refer more to processes. It then confers a qualifying value on them. (60)

Mea haere ’¯ona i te purera’a. stat go 3sg at art church ‘He habitually goes to church.’

(61)

Mea ’amu te rana. stat eat art frog ‘Frogs are edible.’

If it is a transitive process, like in 61, the syntactic subject of the utterance corresponds to the semantic role of patient. In the above example, frogs are edible; they are not the entities doing the eating. It is possible to clearly define the agent; it is introduced by the preposition n¯a ‘by’ and more often than not it is postposed immediately after the predicate. (62)

Mea p¯o’ara n¯a te ’orometua te tamari’i. stat punch by art teacher art children ‘Children used to be punched by the teacher.’

(63)

Mea mana’ona’o n¯a ’u te reo ta’ata tahiti. stat think.about by 1sg art language human Tahiti ‘The language of Tahitian people is my thing of concern.’ (PAA: 3)

(64)

Mea ’amu n¯a te Far¯ani te rana. stat eat by art French art frog ‘French people eat frogs.’ (lit.: Frogs are edible things by French people)

This is an ergative type structure, since the patient is unmarked, as the sole subject of a monovalent process, while the agent is introduced by n¯a ‘by’. In this type

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect

335

of structure, (e) mea ∼ [me:] cannot commute with the aspectual markers. This is, as far as we are aware, the only trace of ergativity in Tahitian. One might compare this ergative pattern with actor-emphatic construction, which also include n¯a. The n¯a-marked agent occurs sentence initially and is followed by an aspect marker: (65)

N¯a te pi’ifare i ’amu i te moa. by art cat pfv eat obj art fowl ‘It is the cat that ate the fowl.’

Through an analogy phenomenon, we also observe, both in the spoken and written language, utterances introduced by (e) mea ∼ [me:] followed by a process, but the structure of which is active –the agent occupies the subject function postposed immediately after the predicate without a compulsory introductory4 morpheme, and the patient finds itself occupying the object function, preceded by the morpheme i. (66)

Mea ’amu te Far¯ani i te rana. stat eat art French obj art frog ‘French people eat frogs.’

Shown below are two possible structures: Ergative form: mea process n¯a agent ø patient subj Active form:

mea process ø agent i patient subj

(67)

Mea ti’aturi n¯a Pito teie r¯u’au. stat trust by Pito dem old ‘This old man has the confidence of Pito.’

(68)

Mea ti’aturi Pito i teie r¯u’au. stat trust Pito obj dem old ‘Pito trusts this old man.’

(ex. 67) (ex. 68)

In the active pattern, mea has completely lost the nominal nature that is still remnant in ergative forms like 62 to 64 with the possessive preposition n¯a. Nouns (agent and patient) are marked in the same way as in verbal constructions. Furthermore, consistent with Tahitian’s actantial canonical structure, this time 4. In fact, the syntactic subject can be preceded by ’o, particularly when it is a proper noun.

336

Jacques Vernaudon

(e) mea ∼ [me:] can commute with the paradigm of aspecto-temporal markers (’ua, e, t¯e...nei/na/ra, etc.). Thus, we can compare example 66 with: (69)

’Ua ’amu te Far¯ani i te rana. stat eat art French obj art frog ‘The Frenchman ate the frog.’

The aspectual value of the form (e) mea ∼ [me:] is the result of its attributive usage. This is an expression of a process envisaged as a characteristic property of the Subject. Utterance 66 does not refer to a one-off occurrence of a process “eating frogs”, but allows this activity to be envisaged as one of the defining properties of the group “Frenchmen” (as one would describe the British as a beer drinking nation, for example). In this aspectual usage, (e) mea ∼ [me:] may appear redundant with the aorist e that is used, among other things, for habitual actions. However, a nuance does exist. (70)

E ’amu Teva i te honu. ao eat Teva obj art turtle ‘Teva eats turtle.’

(71)

Mea ’amu Teva i te honu. stat eat Teva obj art turtle ‘Teva likes to eat turtle.’ or ‘Teva is a big consumer of turtle.’

In the second utterance, more than habitual action, the process of “eating turtle” is presented as the characteristic trait of a person, an inherent property of the subject. This pattern is more emphatic. The action is not only habitual, it is also intensive in quantity or in quality. If the predicate is stressed, one can even perceive an assertive modality in this use of mea: ‘Teva is really a big consumer of turtle.’ Compare this example with sentences 48, 50 and 54 which intensify the quantity of fish or crab predicated. By way of comparison, equivalent use of the stative ka in Drehu, the language of Lifou (Loyalty Islands, New-Caledonia) (Sam 2007: 377–378), that is used both to introduce properties and also the characteristic actions of the subject. Drehu (72) Ka tru Wala. stat tall Wala ‘Wala is tall.’

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect

(73)

337

Ka iji kofi Wala. stat drink coffee Wala ‘Wala drinks coffee.’ (ie. He is a coffee drinker)

Conclusion As we come to the end of this development, we can sum up the major stages of grammaticalization of the morpheme mea. Firstly, a lexical morpheme characterised by weakly determined semantic content, mea is used as a “joker” in the Tahitian language, when an utterer does not wish to or is unable to give an accurate qualifying contour for what he/she is talking about (whether this be an entity or a process). Mea therefore perfectly illustrates the lexical lability of Tahitian where the majority of lexemes indiscriminately and without any morphological variation occupy predicative or non predicative functions, sometimes referring to entities and sometimes to processes. It is as a “joker” that it occupies the place of a modified lexeme in the attributive turn of phrase e mea X by thus maintaining X in the second lexical position that corresponds to a qualifier. This turn of phrase becomes fossilized and undergoes morphological transformations in spoken and written language (particularly the loss of the inclusive particle e). Mea thus functions like a new grammatical morpheme; any lexeme it introduces is interpreted as a qualifier. In the absence of combinatorial restriction, the lexemes that refer to processes may also appear in this context. This process is then envisaged as a characteristic property of the subject. If the process is transitive, two actantial structures are possible in contemporary language. One is ergative (mea process n¯a agent ø patient). The other, active (mea process ø agent i patient), follows by analogy the structure of processive utterances in Tahitian. In the second context, mea commutes with Tahitian aspectual markers the paradigm of which it enhances as a stative marker. The two forms of mea (mea the lexical “joker” and mea the aspectual stative) co-exist in contemporary Tahitian and it is helpful to distinguish them synchronically, even if, in our eyes, there can be no doubt as regards their common origin. We have seen that grammaticalization of lexical equivalent of ‘thing’ as an attributive marker has also developed in other Oceanic languages, which cognitively is no surprise. More unusually, the developing of mea into a stative marker observed in Tahitian is also attested in Marquesan. In the “World Lexicon of Grammaticalization” (Heine and Kuteva 1992), only three options of grammaticalization are listed for lexical equivalents of ‘thing’: (1) Complementizer (Japanese, Ik, etc.), (2) indefinite pronoun (e. g. in Swahili, Nahuatl, Albanian) and (3) A-possessive (Thai, Khmer, proto-central

338

Jacques Vernaudon

Khoisan). With at least two occurrences in Polynesian languages, we can thus add a fourth possibility with stative aspect.

Abbreviations abs ao ana coll deict du incl irr neg pfv poss pst rea stat

absolutive aorist anaphora collective deictic dual inclusive irrealis negation perfective possessive past realis stative aspect

all art cls compl dem excl ipfv loc obj pl pres purp sg

allative article classifying or inclusive particle completive introducer demonstrative exclusive imperfective locative object marker plural presentative purposive singular

References Acad´emie tahitienne 1999 Dictionnaire tahitien-fran¸cais. Papeete: STP Multipress. Bauer, Winifred 1997 The Reed reference grammar of M¯aori. Auckland: Reed. Buse, Jasper and Taringa, Raututi 1995 Cook islands Maori dictionary. Canberra: Australian University Press. Cablitz, Gabriele H. 2006 Marquesan: A Grammar of Space. Berlin/NewYork: Mouton de Gruyter. Heine, Bernd and Tania Kuteva 1992 World lexicon of grammaticalization. Cambridge/new York: Cambridge University Press. Lazard, Gilbert and Louise Peltzer 2000 Structure de la langue tahitienne. Paris: Peeters. Le Cl´eac’h, Herv´e 1997 Lexique marquisien-fran¸cais. Papeete: STP Multipress. Milner, Georges B. 1993 Samoan dictionary. Auckland: Polynesian Press.

Grammaticalization of Tahitian mea ‘thing, matter’ into a stative aspect

339

Mosel, Ulrike and Even Hovdhaugen 1992 Samoan reference grammar. Oslo: Scandinavian University Press. Moyse-Faurie, Claire 1993 Dictionnaire futunien-fran¸cais. Paris: Peeters. Moyse-Faurie, Claire 1995 Le xˆarˆac`uu` , langue de Thio-Canala (Nouvelle-Cal´edonie), El´ements de syntaxe. Paris: Peeters. Moyse-Faurie, Claire 2005 Probl`emes de cat´egorisation syntaxique dans les langues polyn´esiennes. In: Gilbert Lazard et Claire Moyse-Faurie (eds.), Linguistique typologique, 161–192. Lille: Presses du Septentrion. Rensch, Karl 2002 Dictionnaire wallisien-fran¸cais. Canberra: Archipelago Press. Sam, L´eonard 2007 Marques aspecto-temporelles et modales et structures d’actance du drehu, langue de Lifou (Nouvelle-Cal´edonie). Ph.D. dissertation, University of New Caledonia. Stimson, J. Frank 1964 A dictionary of some Tuamotuan dialects of the Polynesian language. The Hague: Martinus Nijhoff. Vernaudon, Jacques and Bernard Rigo 2004 De la translation substantivante a` la quantification: Vers une caract´erisation s´emantique de l’article te en tahitien. Bulletin de la Soci´et´e de Linguistique de Paris, Tome XCIX: 457–480. Paris: Peeters.

Subject index

accusative case marking 1, 11ff., 16, 30, 38, 46, 49, 206, 290ff. adjective order 27, 32, 54ff., 58, 166, 185f., 197 affective stance 140 agent (-ive) 11f., 16, 30, 117, 136, 150, 154, 217, 222, 228, 253, 256, 258, 265f., 274ff., 278, 280, 290, 292f., 295, 303ff., 322, 334ff. answers 6, 15, 65, 68ff., 92ff. antipassive 208, 292, 302ff. applicative 30f., 38, 44, 73, 205, 207, 215, 219, 224ff., associative plurals 242ff. asymmetrical coordination 14, 265ff. background 4, 8, 17, 27f., 31f., 34f., 41, 47, 50, 57ff., 203f., 207, 209, 225 cleft 7, 17, 107ff., 248 comitative coordination 14, 153, 235ff. configurationality 5, 18, 27, 54, 58f. coordination 1, 10, 235ff. demonstrative 8, 16f., 38, 53, 156, 164ff., 222f. determiner 7, 10, 110, 125, 127, 143, 166ff., 222, 251 discord 204f., 225 discourse particle 65, 69, 77ff. ditransitive verb 4, 30f., 38ff., 221 double object construction 27, 30ff. dual 8, 17, 163, 165, 174ff., 182ff., 237f., 241, 243ff., 256, 260, 264, 272, 279 dyadic verb 16, 290ff., 295f., 308, 310 ergative 1, 8, 11ff., 16, 66, 116, 118, 139, 289ff., 334f., 337 exclamative 4, 7, 15, 135ff. expletive (subject) 7, 109ff., 123f., 128 external possession 14, 203, 217, 220, 229

focus 3f., 6ff., 15, 17, 27, 31ff., 35, 41, 47, 49f., 57ff., 66ff., 86, 103, 109, 112, 122f., 125, 127, 135, 142, 144, 152f., 242, 264, 310f. frequency adverb 28f., 35f., 45, 52, 54, 56 grammaticalization 264f., 319ff.

11, 13f., 18, 166f.,

imperative 2f., 68f., 85f., 102f., 140 inclusory construction 14, 235ff. incorporation 9, 30, 44, 149, 203ff. – noun 1, 8ff., 14f., 39, 48, 74, 203ff. – preposition 38, 44 – pseudo 9, 110 information structure 4, 27, 32f., 35, 41, 69, 79, 87f., 91, 99f., 102ff., 108, 141, 209, 320 interview (oral) 15, 65, 68ff., 79, 84f., 87f., 91f., 95, 98f., 101f., 104 intonation 2, 5, 7, 15f., 33, 65f., 68f., 71, 83, 91, 102f., 140, 142, 151, 153f., 206 intransitive verb 8ff., 11, 14f., 48, 146, 148, 150, 203ff., 208, 213, 215f., 221, 224f., 228f., 290ff., 303, 306, 311 lexical compounding 8ff., 14f., 203, 205ff. linear correspondence axiom (universal base hypothesis) 36, 52, 187 locative 44, 56ff., 66, 73, 151, 155, 298f., 302, 305, 310f. manner adverb 28f., 35f., 45, 47f., 52, 54 mirror-image ordering 27f., 32, 34, 55, 59, 166, 182, 185ff., 194 negative – affect 140 – answer 94, 100, 331, 333 – comitative construction 269, 276 – marker/particle 67, 75, 98, 102, 155, 333 – predicate 67, 98, 102 – question 68, 75, 84, 95

342

Subject index

nominalization 1, 7, 15f., 109, 116, 135ff., 212f., 216, 306ff., 332 nominative Case marking 1, 11, 38, 46, 49, 206 non-expectation 68f., 103 noun stripping 204f. number agreement 17, 38, 46, 206, 236, 245, 253f., 259, 262ff., 265, 268f., 274, 276f., 280, 326 numerals 8, 163ff., 167, 170, 172, 179, 182, 187, 189, 191, 194, 196ff., 279, 320 object shift 35, 53 oblique 12, 16, 30, 46, 66f., 117, 152, 242, 250, 292, 299ff. passive 11f., 13, 16, 38, 44, 290ff., 297, 302f., 306, 308f., 311 person and animacy hierarchy 250ff. polar questions 5, 15, 65ff. predicate fronting 36, 52f. pseudo-cleft 7, 17, 107ff. qualifying 14, 326, 329, 331, 334, 337 question particle 5f., 15, 65ff., 114f., 120 relative clause 7, 9, 14f., 38, 113, 116ff., 119, 121, 123, 126, 138, 141, 143, 153, 168, 186, 197f., 303 – nominalised 7, 109 – headless 112f., 124f., 128f. response types 66, 99 resumptive pronoun 43, 73, 117, 126f., 265, 303 rightwards adjunction 53f. roll-up movement 18, 27, 36, 53, 59 scrambling (free word order) 1, 4f., 18, 27, 31, 38, 40, 46, 50, 53, 58f.

speech act 2f., 28, 135, 140, 142 stative aspect 14, 18, 150, 292, 294, 319, 333, 336f. summation pronoun strategy 238ff., 242, 272 superiority condition 43 temporal PP 56ff., 326 “thing” 14, 319ff. universal – 12 107 – 20 16, 166 verb- (predicate-) initial 4, 17f., 27, 34, 46f., 52, 55, 59, 65f., 107ff., 123, 128 wh-in situ 1, 6, 34, 38, 41, 46, 57, 107f., 129 wh-movement 1, 35, 46, 59, 112, 121, 127, 264 wh-question 1, 3, 5ff., 17, 32, 34f., 38, 46, 50, 57, 59, 66f., 69, 75, 82, 92, 103, 107ff., 264 – multiple 41, 43, 71, 121 word order – SOV 1, 4f., 17, 27f., 31ff., 45, 53f., 56, 222, 239f., 248, 255, 259, 268, 272 – SVO 1, 4, 11, 17, 27f., 32, 34ff., 45, 53f., 121, 239ff., 243, 246, 249, 263, 274f. – VOS 1, 4ff., 11, 17f., 27ff., 31f., 34ff., 44ff., 49f., 52f., 56, 59, 107, 110, 236, 247 – VSO 1, 27, 29, 32, 35f., 107, 110, 112 yes-no question

5f., 65ff.

Language index

Aji¨e 237, 241, 244, 246, 251, 254, 281f. Albanian 14, 337 Atayal 242, 258, 282 Avava 194, 197f.

Iaai 8, 152, 155, 278, 281 Ifira-Mele 11 Ik 14 Italian 28, 152

Bwatoo 246, 254, 281f.

Jacaltec 5 Japanese 4f., 14, 28, 30, 48, 54, 108, 143, 255, 337

C`emuhˆı 137, 147, 246, 281 Cook Islands Maori 320, 324 Dorig 261 Drehu 8, 32, 137, 140ff., 146, 277, 281, 336 East Futunan 9, 11f., 110f., 135f., 139, 144, 149ff., 153ff., 259, 292ff., 299, 305, 308, 321, 330 East Uvean 110, 114, 127, 135, 138ff., 144, 151ff., 292, 294, 299, 307f., 321 English 4, 10, 28, 30, 35ff., 41ff., 48, 50, 55, 65, 71, 74, 91, 108, 121, 123, 126f., 141f., 152, 155, 167, 181, 185f., 195, 207, 219, 222, 235, 269, 271, 291, 324, 326, 331f. Fagauvea 135 Fijian (Boumaa) 4f., 13, 18, 27ff., 37, 45f., 58f., 136, 256, 278 Fijian (North-West) 17, 31, 44, 51f., 58f. French (Standard) 56, 65, 108, 123, 142ff., 152, 154, 258, 269 Gascon 102 German 4f., 28, 30ff., 48, 54, 56, 141ff. Germanic 4f., 143, 235 Greenlandic 5 Hawaiian 12, 289f., 302, 304f., 307ff. Hidatsa 4 Hindi 5, 65 Hiw 260

Kairiru 239f., 272, 282 Kapingamarangi 11 Kaulong 239, 259, 282 Khmer 14, 338 Kinyarwanda 44 Kiribati 4ff., 17f., 27f., 31, 37ff. Korean 4, 143 Leti 262, 282 Lo-Toga 260 Lolovoli 152 Malagasy 3, 5f., 10, 17f., 28ff., 34ff., 41, 45, 47, 49ff., 59, 66, 123, 143, 156 Malakula languages 8, 16f., 163f, 166 187, 194, 197, 199 Malay 141, 152 Maleu 273, 281f. Manam 239, 248, 251, 255, 268, 276, 280ff. Mandarin 65 Mangap-Mbula 240, 251, 256, 274, 281f. M¯aori 12f., 66, 91, 114, 125, 146, 152, 154, 204, 240f., 246, 249, 260f., 267, 282, 290, 294, 296, 301ff., 327ff. Marquesan 110, 126, 151, 321, 333, 337 Mohawk 58 Mokilese 147f., 156 Mwerlap 261 Mwotlap 241, 248f., 252, 268, 281f. Nahuatl 14, 337 Nakanai 10, 238, 243, 254, 256f., 267f., 281f.

344

Language index

Naman 194, 196ff. Nˆelˆemwa 32, 236f., 240, 244, 246ff., 250f., 256f., 263ff., 270ff., 276f., 279ff. Nemi 246, 251, 278, 281f. Nengone 8, 154 Nese 193 Neve’ei 194, 197 Niuean 5, 9, 15, 65ff., 107, 110f., 114, 116ff., 121, 125, 204, 259 Num`ee` 276, 281f. Nyelˆayu 246, 279 Palauan 32, 65 Pangkumu 163, 194 Papuan Malay 242f., 245f., 282 Papuan Tip Cluster languages 15, 203, 222, 229, 275 Polish 65, 259, 268 Polynesian 6f., 9, 11ff., 16f., 65, 102, 107ff., 135f., 138, 145, 148f., 151, 154, 156f., 204, 241f., 246, 249, 251, 260, 262f., 267, 278, 289ff., 319f., 326, 333, 338 Port Sandwich 187, 194f., 197f. Proto Central Khoisan 14, 338 Proto-Oceanic 11f., 167, 184, 222, 259, 299 Proto-Polynesian 12f., 289f. Pukapukan 295ff. Qu´ebec French

65, 102

Rapanui 7, 119ff., 126, 129, 149f., 153, 295ff. Rarotongan 302 Saliba 9f., 14f., 203ff. Samoan 9, 12, 111, 114f., 125, 145, 149, 242, 249, 251, 277, 281f., 290f-, 294, 302, 308, 311, 321f. Southern Oceanic 163, 184 Swahili 14, 337

Tagalog 6, 13, 66, 310f. Tahitian 12ff., 18, 124, 138f., 145, 240f., 246, 281f., 319ff. Takia 259 Tape 194 Tawala 222, 275, 281f. Teanu 240 Tetun 66, 239, 246, 262f., 282 Tetun-Fehan 239, 246, 263 Thai 14, 337 Tirax 194, 198 Tokelauan 115, 145, 294, 298 Tolai 248f., 259, 267, 278, 281f. Tongan 5f., 8, 12, 29, 32, 66, 107, 110ff., 114f., 201, 204, 290f., 293ff., 299f., 302, 306ff., 311, 321, 332 Toqabaqita 144, 156, 236, 243f., 245, 248, 251f., 255ff., 281f. Tuamotuan 321 Tukang Besi 6, 144 Turkish 4, 34 Tuvaluan 114, 124ff., 129, 145, 147 Unua 8, 16, 164ff. Uripiv 194, 196ff. Vaeakau-Taumako 11 V’¨enen Taut 194 Vanuatu languages 13, 260 Vur¨es 262 West Uvean 135, 148f., 152, 155 West Futunan 11 Xˆarˆac`uu` 14, 32, 137ff., 147, 152, 251, 278, 281f., 332 Yapese 10 Yiddish 5 Yuanga 249, 254, 270, 272, 279, 281f. Zuanga 237, 249, 254, 270, 271f., 279, 281f.