Valency Classes in the World’s Languages: Volume 2 Case Studies from Austronesia, the Pacific, the Americas, and Theoretical Outlook 9783110429343, 9783110438444

Table of contents :
Contents
Leipzig Questionnaire on valency classes
Part II: Case Studies (cont.)
Austronesia and the Pacific
22 Balinese valency classes
23 Valency classes in Jakarta Indonesian
24 Case frames in Sri Lanka Malay
25 Valency classes in Xârâcùù (New Caledonia)
26 Valency in Nen
27 Complex verbs, simple alternations: valency and verb classes in Jaminjung
Americas
28 Valency classes in Central Alaskan Yupik, an Eskimoan language
29 Transitivity in Ojibwe
30 Valency Classes in Hoocak? (Ho-Chunk)
31 Valency classes in Sliammon Salish~1~
32 Valency classes in Yaqui
33 Valency classes in Zenzontepec Chatino
34 Valency classes in Yucatec Maya
35 Valency classes in Bora
36 Valency classes in Mapudungun
Part III: Theoretical Outlook
37 Situation types, valency frames and operations
38 The hierarchy of two-place predicates: its limitations and uses
39 Verb classes within and across languages
40 Verb classes and valency alternations (NSM approach), with special reference to English physical activity verbs
Language index
Subject index

Citation preview

Andrej Malchukov and Bernard Comrie (Eds.) Valency Classes in the World’s Languages Vol. 2

Comparative Handbooks of Linguistics

Edited by Edith Moravcsik and Andrej Malchukov

Volume 1.2

Valency Classes in the World’s Languages

Volume 2 Case Studies from Austronesia and the Pacific, the Americas, and Theoretical Outlook

Edited by Andrej Malchukov Bernard Comrie

DE GRUYTER MOUTON

ISBN 978-3-11-043844-4 e-ISBN (PDF) 978-3-11-042934-3 e-ISBN (EPUB) 978-3-11-042937-4 ISSN 2364-4354 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2015 Walter de Gruyter GmbH, Berlin/Boston Cover image: Jupiterimages/PHOTOS.com/thinkstock Typesetting: Meta Systems Publishing & Printservices GmbH, Wustermark Printing and binding: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Contents Volume 2 Andrej Malchukov and the Leipzig Valency Classes Project team Leipzig Questionnaire on valency classes xi

Part II: Case Studies (cont.) Austronesia and the Pacific Masayoshi Shibatani and Ketut Artawa 22 Balinese valency classes 877 Thomas Conners, John Bowden and David Gil 23 Valency classes in Jakarta Indonesian 941 Sebastian Nordhoff 24 Case frames in Sri Lanka Malay

987

Claire Moyse-Faurie 25 Valency classes in Xârâcùù (New Caledonia) Nicholas Evans 26 Valency in Nen

1015

1069

Eva Schultze-Berndt 27 Complex verbs, simple alternations: valency and verb classes in 1117 Jaminjung

Americas Osahito Miyaoka 28 Valency classes in Central Alaskan Yupik, an Eskimoan language Richard Rhodes and Rand Valentine 1205 29 Transitivity in Ojibwe

1165

vi

Contents

Iren Hartmann 30 Valency Classes in Hoocąk (Ho-Chunk)

1265

Honoré Watanabe 31 Valency classes in Sliammon Salish~1~

1313

Zarina Estrada-Fernández, Mercedes Tubino and Jesús Villalpando 1359 32 Valency classes in Yaqui Eric Campbell 33 Valency classes in Zenzontepec Chatino Christian Lehmann 34 Valency classes in Yucatec Maya Frank Seifart 35 Valency classes in Bora

1391

1427

1481

Fernando Zúñiga 36 Valency classes in Mapudungun

1515

Part III: Theoretical Outlook Christian Lehmann 37 Situation types, valency frames and operations

1547

Tasaku Tsunoda 38 The hierarchy of two-place predicates: its limitations and uses Beth Levin 39 Verb classes within and across languages

1597

1627

Cliff Goddard 40 Verb classes and valency alternations (NSM approach), with special 1671 reference to English physical activity verbs Language index Subject index

1703 1709

Contents

Volume 1 Abbreviations

ix

Acknowledgments List of authors

xi xiii

Part I: The Leipzig Valency Classes Project: Introducing the Framework Bernard Comrie, Iren Hartmann, Martin Haspelmath, Andrej Malchukov, and Søren Wichmann 3 1 Introduction Andrej Malchukov and the Leipzig Valency Classes Project team 27 2 Leipzig Questionnaire on valency classes Martin Haspelmath and Iren Hartmann 3 Comparing verbal valency across languages

41

Andrej Malchukov 4 Valency classes and alternations: parameters of variation Martin Haspelmath 5 Transitivity prominence

73

131

Damian E. Blasi 5a Assessing transitivity prominence from a statistical perspective: A 149 commentary on Martin Haspelmath’s “Transitivity prominence” Søren Wichmann 6 Statistical observations on implicational (verb) hierarchies

Part II: Case Studies Africa Martina Ernszt, Tom Güldemann, Alena Witzlack-Makarevich 7 Valency in Nǁng 185

155

vii

viii

Contents

Denis Creissels 8 Valency properties of Mandinka verbs

221

Ronald P. Schaefer and Francis O. Egbokhare 9 Emai valency classes and their alternations Joseph Dele Atoyebi 10 Valency classes in Yorùbá

261

299

Csilla Kász 11 Valency properties of verbs in Modern Standard Arabic

327

Eurasia Jóhanna Barðdal 12 Icelandic valency classes: oblique subjects, oblique ambitransitives and the 367 actional passive Michela Cennamo 13 Valency patterns in Italian

417

Michael Daniel and Victoria Khurshudian 14 Valency classes in Eastern Armenian

483

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova 15 Valency and valency classes in Bezhta 541 Andrej Malchukov and Igor V. Nedjalkov 16 Valency classes in Even (North Tungusic) in a comparative Tungusic perspective 571 Edward J. Vajda 17 Valency properties of the Ket verb clause

629

Robert Schikowski, Netra Paudyal, and Balthasar Bickel 18 Flexible valency in Chintang 669 Bingfu Lu, Guohua Zhang and Walter Bisang 19 Valency classes in Mandarin 709 Hideki Kishimoto, Taro Kageyama and Kan Sasaki 20 Valency classes in Japanese 765

Contents

Anna Bugaeva 21 Valency classes in Ainu Language index Subject index

855 861

807

ix

Andrej Malchukov and the Leipzig Valency Classes Project team

Leipzig Questionnaire on valency classes

*

Preliminaries The present questionnaire was compiled by Andrej Malchukov with feedback from the other Leipzig Valency Classes Project members (Bernard Comrie, Martin Haspelmath, Iren Hartmann and Søren Wichmann) at an early stage of the project and distributed to contributors to the volume. Since 2010 the conception of the project as reflected in the questionnaire has undergone certain changes, as described in the Database Manual (Haspelmath & Hartmann 2013, available at: http://www.eva.mpg.de/ lingua/valency/files/database_manual.php). The Database Manual can be seen as a follow-up to the questionnaire but is intended for contributions to the database rather than for book chapters. Moreover, it differs somewhat in scope (see in particular the advanced part of the Leipzig Questionnaire, which addresses a broader set of questions not implemented in the database) and also provides the database contributors with a succinct introduction to the framework and terminological conventions used in the project. The present questionnaire does not pursue this latter goal, which is largely fulfilled within the present volume by the chapter by Haspelmath and Hartmann (this volume). Thus, the Leipzig Questionnaire largely follows the original format reflecting its use to guide contributions to the volume rather than database contributions, except for one important update. It includes an updated list of 70 core meanings rather than the original list of 64 verbs. The definitive list was agreed on the basis of the input from the project participants, as well as the feedback from the contributors.

Introduction The present questionnaire deals with a typology of valency classes, or verb types, in terms of Levin (1993). Levin (1993) is a seminal study of syntactic classes of verbs in English, which shows that a semantic classification of verbs can be achieved through applying syntactic diagnostics. Yet, this study, as well as an earlier study by Apresjan (1969) on Russian, has not been followed up cross-linguistically, which leaves open the question of which aspects of these classifications are universal and which are language particular. Similarly, valency dictionaries are few in number and mostly deal with European languages, thus they cannot fill the gap. The questionnaire has been compiled by participants of the DFG funded project on valency classes1 and is designed to obtain a consistent set of data from a representative set * For convenience, we repeat the questionnaire chapter from volume 1 in identical form with adjusted page numbers. 1 See http://email.eva.mpg.de/~haspelmt/ValencyClasses.pdf for the project description.

xii

Andrej Malchukov and the Leipzig Valency Classes Project team

Tab. 1: The 70 verb meanings. meaning label

role frame

typical context

RAIN BE DRY BURN SINK ROLL BE A HUNTER BE HUNGRY BE SAD DIE FEEL COLD FEEL PAIN

(it) rains S is dry S burns S sinks A rolls S is a hunter E is hungry E is sad S dies S is cold E feels pain in X

SCREAM LAUGH PLAY LIVE LEAVE GO SING JUMP SIT DOWN SIT RUN CLIMB COUGH BLINK SHAVE DRESS WASH EAT HELP FOLLOW MEET HUG SEARCH FOR THINK

S screams S laughs S plays S lives somewhere (L) A left L S goes somewhere (L) S sings A jumps S sits down (somewhere (L)) S sits somewhere (L) A runs A climbs (up L) S coughs S blinks A shaves (his beard/hair) A dresses P A washes P A eats P A helps X A follows X A meets X A hugs P A searches for X A thinks about X

KNOW LIKE FEAR FRIGHTEN SMELL LOOK AT SEE TALK

A knows P E likes X E fears X A frightens P E smells X A looks at P E sees X A talks (to X) (about Y)

It rained yesterday. The ground is dry. The house is burning. The boat sank. The ball is rolling. This man is a hunter. The baby is hungry. The little girl was sad. The snake died. I’m cold. My arm is hurting. = I’m feeling pain in my arm. The man screamed. The little girl laughed. The child is playing. The old people live in town. The boy left the village. The woman went to the market. The boy sang (a song). The girl jumped. The children sat down on the bench. The children sat on the floor. The horse is running. The men climbed (up) the tree. The old man coughed. I blinked (my eyes). The man shaved his beard/cut his hair The mother dressed her daughter The mother washed the baby. The boy ate the fruit. I helped the boys. The boys followed the girls. The men met the boys. The mother hugged her little boy. The men searched for the women. The girl thought about her grandmother yesterday. The girl knew the boy. The boy liked his new toy. The man feared the bear. The bear frightened the man. The bear smelled the boy. The boy looked at the girl. The man saw the bear. The girl talked to the boy about her dog.

Leipzig Questionnaire on valency classes

xiii

meaning label

role frame

typical context

ASK FOR SHOUT AT TELL SAY NAME BUILD BREAK KILL BEAT HIT TOUCH CUT

A A A A A A A A A A A A

asks (X) for Y shouts at X tells (X) Y says “...” (to X) name X (a) Y builds P (out of X) breaks P (with I) kills P (with I) beats P (with I) hits P (with I) touches P (with I) cuts P (with I)

TAKE TEAR PEEL HIDE SHOW GIVE SEND CARRY THROW TIE

A A A A A A A A A A

takes P (from X) tears P (from X) peels (X off) P hides T (from X) shows T (to R) gives T to R sends T (to X) carries T (to X) throws T somewhere (L) ties P (to L) (with I)

PUT POUR COVER

A puts T somewhere (L) A pours T somewhere (L) A covers P (with X)

FILL LOAD

A fills P (with X) A loads T (onto L)

The boy asked his parents for money. The woman shouted at the children. The girl told the boy a funny story. They said “no” to me. The parents called the baby Anna. The men built a house of wood. The boy broke the window with a stone. The man killed his enemy with a club. The boy beat the snake with a stick. The boy hit the snake with a stick. The boy touched the snake with a stick. The woman cut the bread with a sharp knife. The man took the money from his friend. The girl tore the page from the book. The boy peeled the bark off the stick. The boy hid the frog from his mother. The girls showed pictures to the teacher. We gave the books to the children. The girl sent flowers to her grandmother. The men carried the boxes to the market. The boy threw the ball into the window. The man tied the horse with a rope to the tree. I put the cup onto the table. The man poured water into the glass. The woman covered the boy with a blanket. The girl filled the glass with water. The farmer loaded hay onto the truck. = The farmer loaded the truck with hay.

of languages to be described in contributions to the edited volume. It starts with the study of a list of 70 verb meanings (Vs, for short) taken as representative of the verbal lexicon, as well as Levin’s taxonomy (but see below, in particular § 6)2. Since Vs might allow for different uses, the meanings are narrowed down through the use of example sentences to be translated into the respective languages (see the reference to ‘typical contexts’ in Table 1 above). The subsequent parts of the questionnaire address coding and syntactic properties of Vs. It begins with questions about coding properties in constructions formed by a V (in particular, case-marking of arguments) in order to determine the basic valency pattern. In

2 This study focuses on lexical verbs rather than auxiliary verbs (with modal, aspectual and other uses).

xiv

Andrej Malchukov and the Leipzig Valency Classes Project team

accordance with a conventional usage the valency of a V is understood here as the list of its arguments with their coding properties (referred to as coding frames; see Haspelmath & Hartmannn this volume, for definitions of terms). § 3 deals with case alternations, which do not involve voice morphology (cf. different ‘transformations’ used to cross-classify the English verbal lexicon in Levin’s work). § 4 addresses diathetic alternations (or verb-coded alternations), asking about availability of certain valency operations (like passives and applicatives) for Vs, as well as for the meanings expressed by the valency changing markers with Vs, in case they are polysemous. § 3 and § 4 can be seen as largely complementary, as what is a case alternation in one language will be coded as a diathetic alternation in another language (for example, many case alternations in English will be expressed by different valency operations in languages with richer morphology). The last (advanced) part of the questionnaire explores to what extent Vs are representative of lexical classes, i.e., which other verbs belong to the same valency class. This latter part cannot be fully reflected in individual contributions to the volume, which due to size limitations will just offer a summary of verb taxonomies starting from these 70 verb meanings.

I Basic Questionnaire 1 Valency patterns basic examples Please provide glossed examples of sentences containing the relevant Vs (see the prompt typical contexts in the Table 1 above).

The examples sentences exemplifying typical contexts are intended to elicit the verb meanings (Vs) introduced above. The author of a chapter is asked to provide either translational equivalents of the sentences above or other comparable constructions found with the Vs (possibly extracted from corpora). In either case, of special interest is the coding of verbal arguments, for this reason the arguments should be overtly expressed (at least in dependent-marking languages). For headmarking languages, constructions with pronominal (1st/2nd person) subjects and objects should also be considered, as 3rd person arguments are often not indexed on the verb. Alternatively the corresponding constructions with pronominal arguments should be described in § 2.2 under indexing. Examples might be somewhat modified to reflect cultural realities. In cases where a V has different translational equivalents, please choose the verbal lexeme which is more basic (i.e. more frequent and/or morphologically less complex); in the case of several basic items, please include all.

Leipzig Questionnaire on valency classes

xv

The following sections will provide further explanation of the glossed examples with respect to coding and behavioral properties of Vs.

2 Coding properties of valency patterns Coding properties involve the following techniques (Haspelmath 2005; Malchukov et al. 2010): – flagging (case or adposition marking) – indexing (agreement, cross-referencing) – word order (in the absence of other kinds of marking)

2.1 Flagging How are the arguments of the verbs flagged (by a case or adposition)?

NB degree of differentiation will naturally depend on the number of cases available. Thus, German makes finer coding distinctions among verb classes than English, which is due to the availability of verb-specific dative and genitive case selection in German (Sauerland 1994). Further questions about flagging, which may be addressed in case they interact in an interesting way with verb classification. Does flagging differ for different kinds of nominals (animate/inanimate, definite/indefinite)?

Some languages show variation in case marking depending on nominal features such as animacy and definiteness; most commonly it has been observed for (direct) objects (cf. differential object marking in languages like Hindi; where P is marked if human, or definite in the case of inanimates). Other arguments may also differ in case marking depending on nominal features; cf. different cases for animate/ inanimate locations, as in Dyirbal. Do free pronouns show the same valency pattern? (Bound pronouns are considered in § 2.2).

This need not be the case, as witnessed by split-ergative languages of the Australian type. What are other relevant factors affecting argument marking here?

In some languages, alignment patterns further depend on TAM-features, as is familiar from split ergative languages like Hindi and Newari which have ergative

xvi

Andrej Malchukov and the Leipzig Valency Classes Project team

alignment in the perfective/past and accusative in the imperfective. Such alternations are relevant for the project to the extent they are sensitive to verb classes. Thus, in Newari, availability of an ergative pattern depends on tense, on the one hand, and on lexical class of the verb, on the other hand, so that 2-argument verbs deviating from the transitive prototype can take this pattern optionally.

2.2 Indexing How are the arguments of Vs indexed (by agreement/cross-referencing)?

Some languages (head-marking languages), may distinguish valency patterns through indexing rather than flagging. Thus, Tlapanec has 4 different patterns: ergative vs. absolutive indexing patterns are used for canonical transitives, and the ‘pegative’-dative alternation is used for less canonical transitives (like ‘fear’ and ‘meet’). Note that the discussion of indexing (conjugation) markers that also signal voice distinctions (like the middle voice in Greek) should be postponed to § 4 dealing with diathetic alternations. Further questions about indexing; which need to be addressed to the extent these patterns reveal verb classification. Does indexing depend on the features of the nominal (see above)?

Indexing, like flagging, may depend on nominal features. For example, in many languages with object agreement only prominent (animate/definite) objects are indexed. Again, such cases will be relevant for our project insofar as these features further interact with the verb type. For example, in the Austronesian language Manam some experiencer verbs (‘like’, ‘know’, ‘be bad at’) use object indexing only when the object is prominent, while canonical transitives (like ‘break’) invariably index the object. Thus, here we observe effects of differential object marking (i.e. differential object indexing) for verb-types deviating from canonical transitives.

2.3 Word order What are the word order patterns associated with Vs?

Most often word order depends on syntactic transitivity, but some languages make further distinctions depending on the valency class. Thus, in (Gao) Songhai, canonical transitives (‘break’, ‘kill’), have SOV order, while less canonical transitives (‘see’, ‘follow’, ‘love’), have SVO order. Questions of word order need to be ad-

Leipzig Questionnaire on valency classes

xvii

dressed to the extent word order interacts in an interesting way with verb classification.

3 Argument alternations In this section only argument alternations (or uncoded case alternations) are considered; discussion of verb-marked diathetic alternation is postponed to the next section. The two sections should be seen as largely complementary, as alternations will be coded by dedicated markers in some languages with a richer morphology and left uncoded in other languages (like English). Do Vs allow for an alternative construction (valency pattern)?

For example, for English, Levin (1993) mentions, in particular, the following alternations. (The list below mentions only fairly productive alternations; and does not include verb-coded diathetic alternations like the passive alternation): a) the inchoative-causative alternation (John broke the stick ~ The stick broke) b) the “middle” alternation (John cut the bread ~ The bread cuts easily) c) the reflexive deletion (John washed himself ~ John washed) d) the reciprocal transformation (John married Mary ~ John and Mary married) e) the dative alternation (Mary gave the book to John ~ gave John the book) f) the locative alternation (John loaded the truck with hay ~ the hay onto the truck) g) the conative alternation (John cut the bread ~ cut at the bread) h) the object deletion alternation (John ate the bread ~ John ate). i) the preposition dropping alternation (John climbed up the hill ~ climbed the hill). Given that Levin’s list includes a number of other more lexically restricted alternations, it is clear that alternation types need to be generalized before they can be applied cross-linguistically. We will distinguish between the following general types of case alternations, which also find equivalents among voice-alternations to be considered in the next section: a) subject-demoting/deleting case alternations This type will include subject-demoting alternations frequently discussed under the heading of differential subject marking (e.g., genitive of negation in Russian, or ergative/oblique alternation related to volitionality in some ergative languages), but also subject-deleting alternations, as in the case of S/P labile verbs (like break; see Levin’s “inchoative-causative alternation”).

xviii

Andrej Malchukov and the Leipzig Valency Classes Project team

b) object-demoting/deleting case alternations Object-demoting alternations refer, in particular, to varieties of differential object marking (e.g., asymmetric alternations such as the ACC/NOM alternation in Hindi, but also symmetric ones, such as the ACC/PART alternation in Finnish). Object-deleting alternations will refer, in particular, to A/S-labile verbs (like eat; see Levin’s “object deletion alternation”). c) object rearranging case alternations These include, in particular, varieties of dative and locative alternations (cf. (e, f) above), also found in other languages. So the first question to be addressed is: What are the major varieties of case alternations in your language (subject-demoting/deleting; object-demoting/deleting; object rearranging)?

Of course, there may be further varieties of argument alternations, not listed above (for example, object incorporation). These varieties, as well as the alternations listed above, are relevant to the project to the extent they are sensitive to verb classification (e.g., some varieties of differential object-marking apply to any transitive verb, which does not yield an interesting clustering of verb types). More generally for our purposes most relevant are those alternations which are fairly productive (not restricted to a few lexical items), but – most importantly – are sensitive to lexical classes. That is, we are interested in alternations which are distinctive for the verbal lexicon (as sampled here) rather than in those which apply across the board or apply to just few items. After delimiting in this way the set of most relevant case-alternations in your language, the question is to be addressed is: To which Vs in your language do these alternations pertain?

For example, if your language features labile verbs, which of the verbs from the list are labile: S/P labile (cf. the causative-inchoative alternation), and S/A labile (cf. the unspecified object-deleting transformation)?

4 Diathetic alternations and valency changing operations As mentioned above, languages with richer morphology use diathetic alternations for many argument alternations left uncoded in English. Thus in Even, a Tungusic language, the “middle alternation” is signaled by the mediopassive marker, the “inchoative-causative alternation” is signaled by the causative marker (in competi-

Leipzig Questionnaire on valency classes

xix

tion with the mediopassive), while equivalents of English verbs allowing for a “reciprocal alternation” commonly involve a lexicalized sociative marker (e.g. bakalda [find-SOC] ‘meet’). It is convenient to use the same (or similar) taxonomy for the domain of verbcoded diathetic alternations, as we adopted for (uncoded) case alternations. Also in this case we will distinguish between the following types: a) Subject demoting/deleting (anticausatives, etc) b) Subject-Object rearranging (passives) c) Object demoting/deleting (antipassives) d) Object-Object rearranging (applicatives) In addition we address valency increasing alternations: e) Subject addition (causatives) f) Object addition (applicatives) It should be noted that in some cases the distinction between these subvarieties may be problematic (especially between subvarieties of valency rearranging vs. valency increasing applicatives). It is also convenient to treat Subject-Object rearranging operations (passives) together with subject demoting/deleting (anticausatives; reflexives, etc), as they frequently employ the same markers. From this perspective, the major distinction will be between valency reducing vs. valency increasing diathetic alternations. Obviously, the set of valency/voice markers varies across languages; some of these languages distinguish between several such markers (e.g., anticausatives vs. reflexives), while other languages use the same polyfunctional marker. For such cases it is important to state both availability and the meaning of particular markers for certain Vs.

4.1 Valency-reducing operations Valency reducing operations come in several subtypes, as illustrated below: a) Subject demoting/deleting voice alternations – anticausative (cf. a) in § 3 above) – middle (cf. b) in § 3 above) – reflexive (cf. c) in § 3 above) – reciprocal (cf. d) in § 3 above) b) Object demoting/deleting voice alternations – antipassive (cf. g), h) in § 3 above) c) Subject-Object rearranging voice alternations – passive (differs from anticausative in that A may be expressed, or is implied)

xx

Andrej Malchukov and the Leipzig Valency Classes Project team

Does your language have the voice alternations listed above? If not, how are these functions expressed (this information should be given in § 3 if the functions are expressed through argument alternations). If so, what Vs do these markers apply to?

What are the functions of these markers when applied to different Vs?

For example, in Russian the “reflexive” suffix -sja can be used in a reflexive function with some verbs (myt’-sja ‘wash’), while with other verbs it has anticausative (slomat’-sja ‘break’), passive (stroit’-sja ‘be built’), or reciprocal (vstrečat’-sja ‘meet’) function.

4.2 Valency-increasing and valency-rearranging operations The most important valency-increasing operations are causatives and applicatives; the latter however may rearrange rather than increase the valency. In some languages, the same marker is used both in causative and applicative functions; also for such cases it is important to determine which Vs select for which function. If your language has causative markers, which Vs can they apply to? Does the meaning of the causative marker differ with the V involved?

Note that some languages have several causative markers, for example, for building intransitive vs. transitive causatives. These can be used to test for transitivity of less prototypical transitive verbs. If your language has applicative markers, which Vs can they apply to? Does it have several applicative markers used with different Vs? Does the meaning of the applicative marker differ with the V involved?

There may be several subtypes of applicatives, depending on which object is promoted (for example, in Hoocąk (Siouan), there are 4 different applicative markers, including the benefactive applicative, the instrumental applicative and two types of locative applicatives). On the other hand, the general applicative in Salish has been claimed to have different meanings depending on the verb’s class. Applicatives may be used to render many of the alternations listed in § 3, including the dative (cf. e) in § 3 above), locative (cf. f)), but also preposition dropping (cf. i)). Other languages may use directional markers to code some of these alternations (cf. Russian na-gruzit’ seno na telegu [PREF-load hay on cart] ‘to load the hay on the cart’ vs. za-gruzit’ telegu senom [PREF-load cart hay.INSTR] ‘to load the cart with hay’; German laden vs. be-laden).

Leipzig Questionnaire on valency classes

xxi

4.3 Other valency/voice categories Of course, it is impossible to foresee (let alone, list here) all language-particular voice categories. While choosing to address certain voice alternations, one should again be reminded that voice constructions will be relevant to the project to the extent that they interact in an interesting way with the verb lexicon (in particular, are neither restricted to few lexical items, nor apply across the board to all verbs). Does your language have other voice categories? Which Vs do they apply to?

For example, some (Austronesian) languages show a variety of “voice” (or “focus”) forms (“actor focus”, “goal focus”, etc), used for ‘promotion’ of different objects to the subject position; for these languages it will be relevant which Vs allow for which voice constructions. On the other hand, head-marking languages of the “hierarchical type” show a direct-inverse alternation triggered by the relative prominence of the A and P arguments. In that case it is relevant to study the use of direct-inverse alternations with different groups of two and three argument verbs (in the latter case, it is also relevant which of the object arguments takes part in the alternation; e.g., Theme or Recipient of a ditransitive verb). But also for the domain of monotransitives some languages may show further differentiation; e.g., some languages (like Tlapanec) may have different inverse forms for different subtypes of 2-argument verbs.

II Advanced Questionnaire 5 Further properties of individual verbs 5.1 Morphological issues: complexity Indicate which of the Vs are morphologically complex?

In § 1 the contributors were prompted to select for the basic (nonderived) equivalent of verbs on the list. In some cases, however, this is impossible, as in the case where all ditransitives including GIVE are derived (e.g. applicative, as in Tzotzil). Therefore it is important to provide information about morphological complexity of Vs. This question is relevant insofar as morphological make-up may determine availability of a certain valency pattern. For example, in Malayalam, only derived ditransitives (causatives of transitives), take a double object construction, while basic (underived) ditransitives take a dative construction.

xxii

Andrej Malchukov and the Leipzig Valency Classes Project team

5.2 Semantic issues: polysemy Do Vs have other meanings?

This question is relevant insofar as the valency pattern may be motivated through one of the meanings of the polyfunctional item. For example, in some languages, which use the same verbs for both ‘hit’ and ‘throw’, this verb follows the allative pattern, as expected for caused motion verbs.

5.3 Semantic issues: etymology Do you know etymology of Vs?

The motivation for this question is the same as for the previous one: a verb may inherit the valency pattern from its original meaning.

6 Further properties of lexical classes 6.1 Lexical issues: open and closed valency classes Which other verbs belong to the same valency pattern as individual Vs? Is it an open or a closed class? For an open class, please specify which verbs belong to this class (in terms of relevant semantic or formal features). For a closed class, please list other verbs in this class.

For example, if some of the Vs are labile (see § 3 above), please give the list of other (S/P and S/A) labile verbs. Do other verbs, semantically similar to a V, participate in the same alternations as this V?

Do other verbs, semantically similar to a V, show the same diathetic alternations as this V?

References Apresjan, Jurij D. 1969. Èksperemental’noe issledovanie russkogo glagola. Moskva: Nauka. Haspelmath, Martin. 2005. Ditransitive Constructions: The Verb ‘Give’. In Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.), The World Atlas of Language Structures, 426–429. Oxford: Oxford University Press. Levin, Beth. 1993. English Verb Classes and Alternations. Chicago: University of Chicago Press.

Leipzig Questionnaire on valency classes

xxiii

Malchukov, Andrej, Martin Haspelmath & Bernard Comrie. 2010. Ditransitive constructions: A typological overview. In Andrej Malchukov, Martin Haspelmath & Bernard Comrie (eds.), Studies in Ditransitive Constructions, 1–65. Berlin: Mouton de Gruyter. Sauerland, Ulrich. 1994. German diathesis and verb morphology. In Douglas A. Jones (ed.), Working Papers and Projects on Verb Class Alternations in Bangla, German, English, and Korean, 37–92. MIT AI Memo 1517.

Part II: Case Studies (cont.) Austronesia and the Pacific

Masayoshi Shibatani and Ketut Artawa

22 Balinese valency classes

1

1 Introduction Balinese (Basa Bali) is one of the eastern-most members of the Western MalayoPolynesian language family. According to Adelaar (2005), Balinese forms the Malayo-Chamic-BSS (Balinese-Sasak-Sumbawa) group together with Malayic, Chamic, Sasak, and Sumbawa. Like many other Western Malayo-Polynesian languages in Indonesia, Balinese shows a remnant of the Austronesian focus system, which has undergone a gradual attrition such that, while many Formosan and Philippine languages rather faithfully reflect the original four-way morphological contrast (Actorfocus , Patient-focus *-ən, Locative-focus -an, Circumstantial-focus -*(S)i), Balinese and many other focusing Austronesian languages of Indonesia retain only a two-way morphological opposition of a nasal vs. zero verbal prefix, which corresponds to the proto-Austronesian AF and PF morphology. This morphological contrast correlates with the structural opposition of the Actor-focus (AF) and the Patient-focus (PF) construction, where the former aligns the Subject/Actor with the Topic relation and the latter the Object with the Topic, as seen below:2 (1) a. Standard Indonesian (Bahasa Indonesia) AF construction (N-) Saya me-motong daging. I af-cut meat ‘I cut some meat.’ b. PF construction (0̸-) Daging itu saya potong. meat that I pf.cut ‘I cut the meat.’

1 I Wayan Arka helped the preparation of this paper greatly by providing us with a number of relevant examples as well as his grammaticality judgments on many of the examples contained in this paper. Our sincere thanks also extend to Bernard Comrie for reading an earlier version of this paper closely and providing us with comments we found highly valuable in improving both the content and the style of this paper. The support by the U.S. National Science Foundation grant BCS0617198 is also gratefully acknowledged. 2 See Blust (2002), Ross (2009), and Starosta et al. (1982) for details on the Austronesian focus morphology and its syntactic correlates.

878

Masayoshi Shibatani and Ketut Artawa

(2) a. Balinese (Basa Bali) AF construction (N-) Tiang n-godot be. I af-cut meat ‘I cut some meat.’ b. PF construction (0̸-) Be=ne godot tiang. meat=def pf.cut I ‘I cut the meat.’ A distinction between the grammatical relations Subject and Topic must be made for Austronesian languages because many belonging to the Western Malayo-Polynesian family in the Philippines and Indonesia as well as Formosan languages exhibit syntactic phenomena controlled by the Subject, namely the union of the Agent (A) nominal of a transitive sentence and the intransitive subject (S), on the one hand, and those that are controlled by Topic nominals, irrespective of their Subject and Object status, on the other. Because the difference between the Subject=Topic alignment in the AF construction and the Object=Topic alignment in the PF-construction is not directly relevant to the issues of valency in this language, we shall primarily present the data in AF form and concentrate on the coding patterns at the semantico-syntactic level, where argument types are organized in terms of the syntactic relations of Subject, Object, and Oblique. That is, in the following discussion we shall ignore the characteristic Austronesian pragmatico-syntactic phenomenon of Subject=Topic and Object=Topic alignment alternation, except for illustration purposes. Balinese shows three levels of register that are known as Basa Bali sor (Low Balinese), Basa Bali kepara (Ordinary Balinese) or Basa Bali madya (Medium Balinese), and Basa Bali alus/singgih (High Balinese). Low Balinese is used only when speaking derogatively about the subject referent or to a socially lower addressee. Ordinary (or Medium) Balinese is used when speaking to a familiar person and is the most commonly heard variety, while High Balinese, with different degrees of formality, is used in addressing a respected person or in a formal setting like a ceremony. These are illustrated below, where the three different forms for the verb ‘eat’ (ngamah, ma-daar, and ngajeng), for ‘I’ (icang, tiang, and titiang), and for ‘child’ (panak, oka, and putra), in addition to the two different forms for ‘dog’ (cicing and asu) and the two adverbs for ‘still’ (nu and kari) should be noticed. (3) Low Balinese Speaking about a lowly thing to a familiar person: a. Cicing icang=e ngamah nasi. dog I=poss af.eat rice ‘My dog is eating a meal.’

Balinese valency classes

879

b. In a quarrel Panak cai=ne ngamah dogen. child you=def af.eat only ‘Your son is only eating (without doing anything else).’ (4) Ordinary/Medium Balinese Speaking to a friend a. Cicing tiang=e ngamah nasi. dog I=poss af.eat rice ‘My dog is eating a meal.’ b. Panak cai=ne nu naar nasi. child you=poss still af.eat rice ‘Your child is still eating a meal.’ c. Panak tiang=e nu naar nasi. child I=poss still af.eat rice ‘My child is still eating a meal.’ (5) High Balinese Speaking to a teacher a. Cicing tiang=e ngamah. dog I=poss af.eat ‘My dog is eating.’ a′. Asu-n guru=ne ngamah. dog-poss teacher=poss af.eat ‘Your dog is eating.’ / Lit. ‘The teacher’s dog is eating.’ b. Oka=n guru=ne kari ngajeng. child=poss teacher=def still af.eat ‘Your child is still eating.’ / Lit. ‘The teacher’s child is still eating.’ c. Speaking to the Indonesian Presidentip Asu=n titiang=e ngamah nasi. dog=poss I=poss af.eat rice ‘My dog is eating a meal.’ d. Putra=n ratu=ne kari ngerayunang. child=poss caste.title=poss still af.eat ‘Your child is still eating.’ / Lit. ‘Sir’s child is still eating.’ The following discussion is based on the data from Ordinary/Medium Balinese provided primarily by Ketut Artawa, who was born and grew up in Penatahan Village, Bali Island. Additional data come from several speakers from the capital city of

880

Masayoshi Shibatani and Ketut Artawa

Denpasar, whose dialect is less conservative than the varieties spoken in the villages away from Denpasar. Notable differences are pointed out in the relevant sections of the paper.

2 Basic properties of the Balinese verbal lexicon One peculiar feature of the Balinese verbal lexicon, which is also seen in Standard Indonesian and perhaps in other major Indonesian languages of Austronesian stock, is the existence of what Artawa (1994) and others call “precategorial” roots. While there are a fair number of verb roots that are associated with one or two basic valency patterns without derivation, there are also a large number of roots that cannot be used as verbs without a derivational affix and as such whose valency value and alignment pattern are not determined until a derivational affix is selected. Take the form uruk ‘learn/teach’. In order for this to function syntactically, it must take the middle prefix m(a)-, or one of the transitive suffixes -in or -ang, as below, where neither the AF marking (6a) nor the PF (zero) marking (6b) alone is sufficient.3 (6) Underived verb form a. *Tiang ng-uruk basa Inggeris (ka anak=e cenik ento). I af-learn language English to person=def small that ‘I am teaching English (to the child).’ b. *Basa Inggeris uruk tiang (ka anak=e cenik ento). language English pf.learn I to person=def small that ‘I am teaching English (to the child).’

(AF)

(PF)

(7) m(a)-derived form Tiang m-uruk (basa Inggeris). I mid-learn (language English) ‘I am learning (English).’ 3 Our definition of precategorials differs from that offered in Artawa (1994). When a root form is usable only after a verbal derivation, it is precategorial. AF and PF marking is not a derivational process and hence if a root is usable with just AF or PF marking, it is not precategorial. The verb root tegak ‘sit’, for example, is not precategorial and can be used as an intransitive verb without a derivation by simply marking it AF (i.e. by nasalizing the initial consonant) as I Nyoman negak di bataran ento ‘Nyoman sits on the floor’. There are also noun-based precategorial verbs such as suah ‘comb’ and sikat ‘tooth’, which can be used as nouns by themselves, but which must undergo a derivation in order for them to function as a verb (see § 5.3). There is one other way, besides ma-/-in/-ang derivations, in which some of these precategorial roots can be realized, namely reduplication. The precategorial verb kijep ‘blink’, for example, can be used as intransitive either via ma-derivation (ma-kijep-an) or in the form of kijep-kijep via reduplication (see § 7). As this example shows, ma-derivation produces a circumfixed ma…an form with certain roots.

Balinese valency classes

(8) -in derived form a. Tiang ng-uruk-in anak=e cenik ento (basa Inggeris). I af-learn-in person=def small that (language English) ‘I am teaching the child (English).’ b. Anak=e cenik ento uruk-in tiang (basa Inggeris). person small that pf.learn-in I (language English) ‘I am teaching the child (English).’ (9)

-ang derived form a. Tiang ng-uruk-ang basa Inggeris (ka anak=e cenik I af-lean-ang language English to person=def small ento). that ‘I am teaching English (to the child).’ b. Basa Inggeris uruk-ang tiang (ka anak=e cenik language English pf.learn-ang I to person=def small ento). that ‘I am teaching English (to the child).’

881

(AF)

(PF)

(AF)

(PF)

Compare the above with the pattern shown by the non-precategorial verb tegul ‘tie’ below, where the AF (10a) and the PF (10b) forms without a derivational affix are both grammatical. (10) Underived basic verb form a. Tiang negul 4 jaran=ne (ka punyan kayu=ne). I af.tie horse=def to trunk tree=def ‘I tied the horse to the tree trunk.’ b. Jaran=ne tegul tiang (ka punyan kayu=ne). horse=def pf.tie I to trunk tree=def ‘I tied the horse to the tree trunk.’

(AF)

(PF)

(11) ma-derived form Jaran=ne ma-tegul (ka punyan kayu=ne). horse=def mid-tie to trunk tree=def ‘The horse is tied (to the tree trunk).’

4 In the Denpasar dialect, where the -in suffix is widely used as a transitivizing suffix, negul-in is also usable here.

882

Masayoshi Shibatani and Ketut Artawa

(12) -in derived form a. Tiang negul-in punyan kayu=ne jaran. I af.tie-in trunk tree=def horse Lit. ‘I tied the tree trunk (with) a horse.’ b. Punyan kayu=ne tegul-in tiang jaran. trunk tree=def pf.tie-in I horse Lit. ‘I tied the tree trunk (with) a horse.’ (13) -ang derived from a. Tiang negul-ang jaran=ne ka punyan kayu=ne. I af.tie-ang horse=def to trunk tree=def ‘I tied the horse to the tree trunk.’ b. Jaran=ne tegul-ang tiang ka punyan kayu=ne. horse=def pf.tie-ang I to trunk tree=def ‘I tied the horse to the tree trunk.’

(AF)

(PF)

(AF)

(PF)

The presence of the precategorial roots indicates that valency patterns5 and derivational processes are much more tightly integrated in Balinese than in other languages, where the basic verb valency patterns and the derived patterns (that is, valency alternations) are relatively clearly delineated in terms of the absence vs. the presence of derivational affixes. In addition, Balinese valency alternation has two aspects to it. One is valency increase and the other is argument realignment or rearrangement (without valency increase). These two aspects of valency alternation are effected by the same derivation affix. Compare the basic pattern and the -in derived pattern in (14a)–(14b) and (15a)–(15b) below, where one usage pattern of the verb GIVE (baang), i.e. with an animal goal, is illustrated. (14) SLEEP (pules) anyar). a. Ia pules6 telung jam (di umah=ne he pf.sleep three hour at house=3sg.poss new ‘He slept three hours (in his new house).’ b. Ia mules-in umah=ne anyar telung jam. he af.sleep-in house=3sg.poss new three hour ‘He slept three hours in his new house.’

5 By the term “valency pattern” we refer to two argument properties. One is valency value, which relates to the number of obligatory nominal expressions that a verb requires, and the other is the argument alignment pattern, especially the alignment of the Figure and Ground expressions with the grammatical relations of Object and Oblique. See below on the Figure/Ground distinction. 6 See Section 7 below on the focus marking on intransitive verbs.

Balinese valency classes

883

(15) GIVE (baang) a. Basic verb form Tiang maang banyu ka celeng=e. I af.give scrap to pig=def ‘I gave food scraps to the pig.’ b. -in derived form Tiang maang-in celeng=e banyu. I af.give-in pig=def scrap ‘I gave the pig food scraps.’ The verb pules ‘sleep’ is a monovalent intransitive verb and as such the specification of a goal location is optional, as in (14a) above. When the -in form is used, a location must be specified, as in (14b). The -in derivation here, therefore, counts as a case of valency increase. In the case of the trivalent verb baang ‘give’, however, a goal specification is obligatory, as in (15a). The -in derived form in (15b), therefore, does not increase valency since the goal specification is part of the basic valency of the verb. Thus the -in derivation in (15b) constitutes a case of argument realignment rather than valency increase. In addition to the cases where alternations involve no change in the valency value, -in and -ang derivations of precategorial roots cannot be characterized in terms of valency increase/decrease because precategorials, by definition, do not have a basic valency value. Because of these facts, it is essential to first understand the basic functions of the derivational suffixes -in and -ang and then to compare the valency patterns across the Balinese verbal lexicon, whether or not the derived forms/valency alternations result in valency change.

3 Figure and Ground in argument alignment In considering valency patterns within and across languages, a useful notion is the perceptual construct of the Figure-Ground distinction that has been invoked by a number of linguists in the descriptions of argument coding in several languages (Talmy 2000; Croft 1991; DeLancey 2001). In terms of the thematic roles, the Theme corresponds to the Figure and represents an entity that is situated at a location or in a state that moves from one location to another in physical space or from one state to another in the construal of a change-of-state as an abstract motion. The Ground is the background against which a Figure is delineated, and it subsumes various locative expressions in language such as a stationary location (expressed as at, on, in in English), a source location (from), a goal location (to, against), etc. with respect to which a Figure expression is predicated as being located or moving. Locations can be both physical or human, and thus both so-called goal locations

884

Masayoshi Shibatani and Ketut Artawa

(as in John walked to the station) and human recipients (as in John gave Bill the book) count as instances of the Ground. What is generally characterized as Patient in the study of thematic roles is construed to be an instance of the Ground=Object in this paper. The two main alignment patterns that figure importantly in the following discussion are the Ground=Object (GR=OBJ) alignment, seen in (16), and the Figure=Object (FIG=OBJ) alignment, as in (18) below. (16) GR=OBJ construction a. John loaded the wagon with hay. b. John hit the fence with the stick. (17) FIG=OBJ construction a. John loaded the hay onto the wagon. b. John hit the stick against the fence.

3.1 -in locative applicative As some of the earlier examples show, Balinese -in derived forms align a Ground expression with the Object, and this is consistent with the role this suffix plays in deriving locative applicative forms with a variety of locational meanings such as a locus of both static and dynamic events, a goal location (including a human recipient), and a source location. (18) COME (teka) a. Ia teka ka kantor=e ene. s/he pf.come to office=def this ‘S/he came to this office.’ b. Ia neka-in kantor=e ene. s/he af.come-loc office=def this ‘S/he came to this office’ / ‘S/he visited this office.’ (19) SIT (tegak) a. Ia negak di kursi=ne. s/he af.sit in chair=def ‘S/he sits in the chair.’ b. Ia negak-in kursi=ne. s/he af.sit-loc chair=def ‘S/he sits in the chair.’ (20) a. Tiang mula biyu di tegal=e. I af.plant banana in garden=def ‘I planted bananas in the garden.’

Balinese valency classes

885

b. Tiang mulan-in tegal=e biyu. I af.plant-loc garden=def banana ‘I planted the garden with bananas.’ (21) a. Cai nyilih pipis uli/sig bapa=ne. you af.borrow money from father=3sg.poss ‘You borrowed money from his father.’ b. Cai nyilih-in bapa=ne pipis. you af.borrow-loc father=3sg.poss money ‘You borrowed money from his father.’ Like locative applicatives in other languages, the Balinese -in locative derivation construes the conceptualization of the specified location as being somehow affected or more interactive with the activity of the described event. Thus, the -in derived ditransitive construction in (20b) implies that the entire garden has been planted with banana trees.7 Earlier we noted a situation where -in derivation increases valency and another in which it simply realigns arguments without valency increase – cf. (14) and (15). The latter realignment case results in the same alignment pattern as the valencyincreasing case; namely, in both cases the -in derived form has a locative Ground expression aligned with the Object. In other words, the role of the -in suffix as a valency-increasing operator is simply a consequence of the basic function of this suffix, namely that of aligning a Ground argument with the Object. If the basic verb valency frame includes a Ground expression, as in the case of baang ‘give’, the -in form selects it as the Object. If the basic valency does not contain a locative element construable as a Ground, then the -in derivation introduces a new locative argument and aligns it with the Object. This is the essence of the valency-increasing effect of -in derivation.

3.2 -ang causative In contrast to the -in locative applicative, which aligns a Ground expression with the Object, -ang derivation aligns a Figure with the Object. There are two types of Figure involved in this derivation. The typical causative construction aligns the Theme Figure of a caused motion with the Object, as in I pushed the chair across the room and I made him roll down the hill.

7 See Section 8 below for a fuller discussion on this sort of restriction.

886

Masayoshi Shibatani and Ketut Artawa

(22) a. Anak=e cenik ento menek ka gedebeg=e. person=def small that af.climb to cart=def ‘The child climbed onto the cart.’ b. Ia menek-ang anak=e cenik ento ka gedebeg=e. s/he af-caus person=def small that to cart=def Lit. ‘S/he loaded the child onto the cart.’ (23) a. Lumur=e ento ulung. glass=def that pf.fall ‘The glass fell down.’ b. Tiang ng-ulung-ang lumur=e ento. I af-fall-caus glass=def that ‘I dropped the glass. Balinese -ang morphological causatives are also semantically restricted, and they generally express direct causation (see Section 8). As such, favorite inputs to -ang causativization are inactive processes or stative predicates, where a Theme/ Figure undergoes either a concrete motion in space, as in (22) and (23), or an abstract motion from one state to another, as in (24) and (25) below. (24) a. Tembok=e tegeh. wall=def pf.high ‘The wall is high.’ b. Tiang negeh-ang tembok=e. I af.tall-caus wall=def ‘I made the wall high.’ (25) a. Celeng=e mati. pig=def pf.dead ‘The pig is dead.’ b. Anak=e ento nge-mati-ang celeng=e. person=def that af-dead-caus pig=def ‘The man killed the pig.’ The other type of Figure involved in -ang derivation is an instrument that moves in space and that typically comes into contact with a Ground element. The following illustrates the typical use of -ang aligning an instrumental Figure with the Object. (26) a. Ia ng-lempag cicing=e aji sampat. s/he af-hit dog=def with broom ‘S/he hit the dog with a broom.’

Balinese valency classes

887

b. Ia ng-lempag-ang sampat ka cicing=e. s/he af-hit-caus broom to dog=def ‘S/he hit the broom against the dog.’ Lit. ‘S/he caused the broom to hit the dog.’ (27) a. Ia nyikut natah=e aji tungked. s/he af.measure yard=def with stick ‘S/he measured the yard with a stick.’ b. Ia nyikut-ang tungked ka natah=e. s/he af.measure-caus stick to yard=def ‘S/he used a stick to measure the yard.’ The above use of -ang is normally analyzed as a case of instrumental applicative apart from causativization (Artawa 1999: Chap. 3; Arka 2003: 195 ff.). But, compare the alignment patterns of the earlier causative forms in (22b)–(23b) and the -ang instrumental applicative forms in (26b)–(27b) above. In both, the Figure expressions are aligned with the Object, and the Ground expressions are coded as a prepositional phrase. As discussed below, Balinese instrumental applicatives are based on the coding of an event as a caused motion involving an instrument as a causee Theme. In the following descriptions, the -ang suffix, which aligns a Figure expression with the Object, will be glossed ‘CAUS’. The following transitive/ditransitive sentences illustrate the contrastive functions of the locative -in and the causative -ang suffix with regard to the Object alignment. The valency phenomena involving bi-/trivalent verbs discussed next center around these two alignment patterns. (28) GR=OBJ alignment (-in locative suffix) Tiang ng-uruk-in anak=e cenik cenik (basa Inggeris). I af-learn-loc person=def small small language English ‘I teach the children (English).’ (29) FIG=OBJ alignment (-ang causative suffix)8 Taing ng-uruk-ang basa Inggeris (ka anak=e cenik cenik). I af-learn-caus language English to person=def small small ‘I teach English (to the children).’

8 The -ang causative, as a caused motion expression, normally requires or strongly implicates a Ground (goal) expression. Like the English counterpart, uruk ‘learn/teach’ can be used as monotransitive with a GR=Object or a FIG=Object, as in (29) and (30).

888

Masayoshi Shibatani and Ketut Artawa

4 The valency classes of transitive/ditransitive verbs As noted above, the locative -in and causative -ang suffixes play an integral role in the valency structures of Balinese verbs. In order to see this, observe first the transitive/ditransitive verb classes according to the basic valency patterns and their alternants. As is clear from Table 1, -in and -ang derivations supply the forms with the specific valency patterns where basic (underived) verbs displaying the relevant patterns are not available. Also recognized from the table is that the precategorial/ non-precategorial distinction in Balinese has an important implication for the study of valency due to the fact that a large number of verb forms do not have a basic valency pattern associated with them. -in and -ang derivations fix the valency value and argument alignment for them, without which they cannot function as a verb. On the basis of the Object alignment pattern, Balinese (di-)transitive verbs group themselves into four large classes: (i) a very small minority of verbs (e.g., baang GIVE) display unmarked (un-coded) alternating patterns of SU > GR=OBJ > FIG=OBJ/OBL (the GR=OBJ pattern) or SU > FIG=OBJ > GR=OBL (the FIG=OBJ pattern) (ii) a fair number of verbs (e.g., ejang PUT) are FIGURE-oriented and display the basic FIG=OBJ pattern, which alternates with the GR=OBJ pattern via -in derivation (iii) another class (e.g., lempag HIT) has GROUND-oriented verbs displaying the basic GR=OBJ pattern, which alternates with the FIG=OBJ pattern via -ang derivation

Tab. 1: Transitive/Ditransitive. Verb Classes and Alternation Patterns. Alignment:

GR=OBJ > FIG=OBL/OBJ (Locative pattern)

FIG=OBJ > GR=OBL (Causative pattern)

Basic verb class (A) Basic verb class (B) Basic verb class (C) Basic verb class (C′): Basic verb class (D)

basic forms basic forms -in forms _______

basic forms -ang forms basic forms basic forms -ang forms basic forms -ang forms ________

Precategorial class 1 Precategorial class 2

basic forms -in forms -in forms -in forms

(The dash mark for Class (C′) and Precategorial 2 indicates that no form is available for the respective alignment pattern even via -in or -ang derivations.)

Balinese valency classes

889

(iv) a considerable number of precategorial verbs (e.g., uruk LEARN/TEACH) have no basic valency pattern associated with them, their valency pattern being determined by -in (for the GR=OBJ pattern) and -ang derivation (for the FIG= OBJ pattern). It is in this manner that -in and -ang derivations fill the gap in the two transitive/ditransitive valency patterns, namely the GR=OBJ and the FIG=OBJ pattern. Other coded alternations effected by benefactive, middle, and passive operations do not revolve around the GR=OBJ/FIG=OBJ alternation per se, and, accordingly, their contributions to the Balinese valency patterns are discussed separately below. With this proviso, the Balinese transitive/ditransitive alternation patterns can be most conveniently discussed in terms of the seven classes of verbs on the basis of derivational possibilities, as summarized in Table 1 above.

4.1 Basic transitive verb class (A) Basic locative pattern: GR=OBJ > FIG=OBJ Basic causative pattern: FIG=OBJ > GR=OBJ/OBL GIVE (baang), FILL (isinin)9 The only basic verbs allowing the two alternate alignment patterns without coding/ derivation are GIVE (baang) and FILL (isinin). (30) GIVE (baang)10 a. Basic locative pattern: GR=OBJ > FIG=OBJ Guru=ne (nge-)maang anak=e cenik ento buku. teacher=def af.give person=def small that book ‘The teacher gave the child the book.’ b. Basic causative pattern: FIG=OBJ > GR=OBL Guru=ne (nge-)maang buku ka anak=e cenik ento. teacher=def af.give book to person=def small that ‘The teacher gave the book to the child.’

9 In this paper, verbs are cited in their root forms, which are identical with the PF forms for nonprecategorials. 10 The verb baang is interesting in that it allows an optional -in derivation when the goal is an animal as opposed to a human, permitting Valency Pattern C and indicating that animals are treated as if they were a locational Ground. There may be a possibility that ang of baang is related to the causative -ang. See next footnote.

890

Masayoshi Shibatani and Ketut Artawa

(31) FILL (isinin)11 a. Basic locative pattern: GR=OBJ (> FIG=OBJ/OBL) Anak=e ento ng-isinin lumur=e ento ((aji) yeh). person=def that af-fill glass=def that (with) water ‘The man filled the glass (with water).’ b. Basic causative pattern: FIG=OBJ > GR=OBL Anak=e ento ng-isinin yeh ka lumur=e. person=def that af-fill water to glass=def Lit. ‘The man filled water into the glass.’ GIVE (baang) is the only true ditransitive verb that requires both Figure and Ground expressions in both the GR=OBJ and the FIG=OBJ pattern. The verb FILL (isinin) does not require a Figure nominal in the GR=OBJ construction, as in (31a), while both Figure and Ground nominals are required in the FIG=OBJ construction (31b). Unlike the English verb put, the Balinese counterpart ejang does not obligatorily require a Ground expression. Sentence (32b) below, without a Ground specification, is perfectly well-formed and has a meaning similar to ‘put down’ in English. It is only when the Ground expression is aligned with the OBJ that both Figure and Ground nominals are required, as in (c). This makes ejang difficult to classify, but we consider it a member of Class (C), which would then contain both trivalent and bivalent verbs that optionally take a locative phrase. (32) PUT (ejang) a. Ia ng-ejang buku di meja=ne. s/he af-put book on table=def ‘S/he put the book on the table.’ b. Ia ng-ejang buku. s/he af-put book ‘S/he put down the book.’ c. Ia ng-ejang-in meja=ne (aji) buku. s/he af-put-loc table=def (with) book Lit. ‘S/he put the table (with) a book.’

11 The (n)in ending of isinin ‘fill’ is likely to have historically come from the locative -in suffix. Observe that it disappears in the middle-marked resultative form of this verb: Lumur-e ma-isi yeh ‘the glass is filled with water’. However, this ending does not seem to be functioning as the LOC suffix any longer. The true LOC -in suffix disappears, unlike ...(n)in in (32b), when a FIG expression is aligned with the Object.

Balinese valency classes

891

4.2 Basic transitive verb class (B): Basic locative pattern: GR=OBJ (> FIG=OBL/OBJ) -ang derived causative pattern: FIG=OBJ > GR=OBJ HIT/BEAT (lempag), TOUCH (tundik), STAB (tusuk), COVER (rurub) (CUT (godot), SAW (regaji)) These surface-contact verbs are Ground-oriented and have the GR=OBJ alignment pattern as basic. They yield the FIG=OBJ pattern through -ang derivation. (33) HIT/BEAT (lempag)12 a. Anak=e ento ng-lempag lalipi-ne (aji tungked). person=def that af-hit snake=def with stick ‘The man hit the snake (with a stick).’ b. Anak=e ento ng-lempag-ang tungked ka lalipi=ne. person=def that af-hit-caus stick to snake=def ‘The man hit the stick against the snake.’ / Lit. ‘The man caused the stick to hit the snake.’ In the basic pattern of this type, an instrumental Oblique is optional, whereas in the derived pattern it is aligned with the OBJ and is obligatory, rendering the -ang causative derivation here a valency-increasing operation. Notice that the Oblique ka-phrase in (33b) is a required argument. The -ang form here obtains most readily when the movement of a Figure in space is clearly perceived, as in the case of HIT/BEAT (lembag) and STAB (tusuk). When a surface contact does not result from a clearly perceivable directed motion, as in the case of slicing and sawing, it is rather difficult to apply the -ang derivation. The (b) form below is deemed not as well-formed as (33b) above. Similarly, while STAB (tusuk) and THROW (timpug) are perfect in the -ang derived form, SAW (regaji) is not. (34) CUT/SLICE (godot) a. Anake ento ngodot poh (aji tiuk). person that af.cut mango with knife ‘The man cut the mango with a knife.’

12 This verb permits an -in derived form as well, which adds a repetitive meaning to the basic meaning. The repetitive meaning associated with many action verbs is perhaps related to the transitivity effect associated with the -in suffix.

892

Masayoshi Shibatani and Ketut Artawa

b. *?Anake ento ngodot-ang tiuk=ne ka poh=e. person that af.cut-caus knife=def to mango=def ‘The man cut the mango with the knife.’ / Lit. ‘The man caused the knife to cut the mango.’ The verb COVER (rurub) is somewhat unusual and behaves differently from the other members of this class in allowing the -in locative derivation despite the fact that the basic form already has the locative GR=OBJ pattern associated with it. The -in derived form, however, is different from the basic form in that it allows both OBJ > OBJ (double Object) and OBJ > OBL configurations, though the FIG expression remains optional. (35) COVER (rurub) a. Basic locative pattern: GR=OBJ (> FIG=OBL) Ia ngerurub anak=e cenik ento (aji saput). s/he af.cover person=def small that (with blanket) ‘S/he covered the child with a blanket.’ b. -in locative pattern Ia ngerurub-in anak=e cenik ento (saput)/(aji saput). s/he af.cover-loc person=def small that (blanket)/(with blanket) ‘S/he covered the child with a blanket.’ c. -ang causative pattern Ia ngerurub-ang saput ka anak=e cenik ento. s/he af.cover-caus blanket to person=def small that Lit. ‘S/he covered the blanket to the child.’ Surface-contact verbs contrast with verbs of creation such as tulis ‘write’ and bangun ‘build’, which may take an instrumental nominal but which do not allow -ang derivation. (36) WRITE (tulis) a. Anak=e ento nulis surat (aji pulpen=ne ene). person=def that af.write letter with pen=def this ‘The man wrote a letter (with this pen).’ b. *Anak=e ento nulis-ang pulpen=ne ene ka surat. person=def that af.write-caus pen=def this to letter ‘The man wrote a letter with this pen.’ It is not clear how verbs of creation like these should be classified since the created objects seem to be neither a Figure nor a Ground, which classify those elements denoting things that exist independently from the activities expressed by the verbs. Interestingly, WRITE (tulis) also behaves somewhat like a member of Precategorial

Balinese valency classes

893

Class 1 and allows both -in locative and -ang causative derivations when a Ground is specified with respect to which an instrumental Figure is perceived to move. (37) a. Ia nulis aksara Bali (di tembok=e) (aji pulpen). s/he af.write characters Balinese on wall=def with pen ‘S/he wrote Balinese characters (on the wall) (with a pen).’ b. Ia nulis-in tembok=e (aksara Bali) (aji pulpen). s/he af.write-loc wall=def characters Balinese with pens ‘S/he wrote (Balinese characters) on the wall (with a pen).’ c. Ia nulis-ang pulpen=e aksara Bali ka tembok=e. s/he af.write-caus pen=def characters Balinese to wall=def ‘S/he wrote Balinese characters on the wall with the pen.’ Notice that the basic form of WRITE (tulis) is bivalent, with a location and an instrumental specification being optional (37a). The -in derivation, while it aligns a GR expression with the Object, does not increase valence, as seen in (37b), where the affected patientive entity and the instrument Figure are both optional. The -ang causative derivation in (37c), on the other hand, increases valence by two, requiring specifications of an instrument Figure aligned with the first Object (the first of the double Objects defined in terms of word order) and a Ground expression coded as an OBL. The effected patientive entity is then coded as a second Object as in (37c). (Cf. (37b, c) with the patterns exhibited by precategorial verbs below. The status of the first and second Objects of the double Object construction is discussed in Section 6.)

4.3 Basic transitive verb class (C) Basic causative pattern: FIG=OBJ (> GR=OBL/OBJ) -in derived locative pattern: GR=OBJ > FIG=OBL/OBJ PUT (ejang), SEND (kirim), STEAL (maling), ASK (idih), GIVE (baang with an animal recipient) Basic verb class (C) primarily contains caused-motion verbs exhibiting the trivalent FIG=OBJ > GR=OBL pattern with the -in derived alternate pattern. (38) SEND (kirim) a. Basic causative pattern: FIG=OBJ > GR=OBL Anak-e ento ngirim buku=ne ka sekolah. person=def that af.send book=def to school ‘The man sent the book to the school.’

894

Masayoshi Shibatani and Ketut Artawa

b. -in derived locative pattern: GR=OBJ > FIG=OBJ/OBL Anak=e ento ngirim-in sekolah buku=ne. person=def that af.put-loc school book=def Lit. ‘The man sent the school the book.’ (39) STEAL (maling) a. Basic causative pattern: FIG=OBJ (> GR=OBL) Tiang nge/ma-maling pipis (uli guru=ne). I af/mid-steal money from teacher=def ‘I stole money from the teacher.’ b. -in derived locative pattern: GR=OBJ > FIG=OBJ Tiang nge-maling-in guru=ne pipis. I af-steal-loc teacher=def money Lit. ‘I stole the teacher money.’ The verb GIVE (baang) shows the above pattern when the goal Ground is an animal as noted earlier (see (15)).

4.4 Basic transitive verb class (C′) Basic causative pattern: FIG=OBJ (> GR=OBL/OBJ) *-in derived locative pattern: GR=OBJ > FIG=OBL BRING (aba), CARRY (tenteng) GET (baan), TEAR (uek), WIPE (sapuh) This class is a subclass of C and contains those basic caused-motion verbs but that do not easily permit the alternate -in locative pattern. The GR=OBL expressions of the verbs belonging to this class take either the preposition ka ‘to’ or uli ‘from’. The verbs that introduce the GR=OBL by ka include BRING (aba) and CARRY (tenteng), and the other verbs listed use uli to mark the OBL. Compare the following examples with (38) above: (40) BRING (aba) a. Basic causative pattern: FIG=OBJ (> GR=OBL/OBJ) Tiang ng-aba buku=ne ka kantor pos. I af-bring book=the to office post ‘I brought the book to the post office.’ b. *Tiang ng-aba-in kantor pos buku=ne. I af-bring-loc office post book=def Lit. ‘I brought the post office the book.’

Balinese valency classes

895

When the Ground is a person, the -in derived form is possible with this verb, but it, in addition, has the benefactive meaning of carrying something for someone. (41) BRING (aba) a. Anak luh ento ng-aba buku=ne ka guru=ne. person female that af-bring book=def to teacher=def ‘The girl brought the books to the teacher.’ b. Anak luh ento ng-aba-in guru=ne buku.13 person female that af-bring-loc teacher=def book ‘The girl brought the books to the teacher.’ ‘The girl carried the teacher’s books (while walking with the teacher).’ The verb tenteng ‘carry’, on the other hand, does not permit the usual (i.e. the GR= OBJ) -in derivation whether the Ground is human or not. (42) CARRY (tenteng) a. Anak luh ento nenteng buku-ne ka sekolahl/guru=ne. person female that af.carry book=def to school/teacher=def ‘The girl carried the book to the school/teacher.’ b. Anak luh nenteng-in guru=ne buku. person female af.carry-loc teacher=def book ‘The girl carried the book for the teacher. ’ ‘*The girl carried the book to the teacher.’ c. *Anak luh nenteng-in sekolah buku. person female af.carry-loc school=def book Lit. ‘The girl carried the school a book.’ The verbs GET (baan), TEAR (uek), and WIPE (sapuh) introduce a locative Oblique expression with uli ‘from/with’, but -in suffixation is unavailable, preventing the GR=OBJ pattern to obtain with them. Compare the pattern of STEAL (maling) in (39) above with that of GET (baan) below. Perhaps the difference lies in the fact that STEAL implicates the affectedness of the Ground more strongly than GET. (43) GET (baan) a. Basic causative pattern: FIG=OBJ (> GR=OBL) Anak luh ento maan buku ento (uli guru=ne). person female that af.get book that (from teacher=def) ‘The girl got a book from the teacher.’

13 Both this example and (42b) below can replace -in with the benefactive -ang suffix. See Section 5.1 on -ang benefactives.

896

Masayoshi Shibatani and Ketut Artawa

b. -in derived GR=OBJ pattern *Anak luh ento maan-in guru=ne buku. person female that af.get teacher=def book

4.5 Basic transitive verb class D (a) Basic locative pattern: GR=OBJ (> FIG=OBL) -ang derived causative pattern: FIG=OBJ > GR=OBJ (b) Basic causative pattern: FIG=OBJ > GR=OBL -in derived locative pattern: GR=OBJ > FIG=OBJ TIE (tegul), NAIL (pacek), SMEAR (uap), GLUE (elim) The semantics of verbs of fixing contained in this class allows them to behave both like the surface-contact Class B verbs and like the caused-motion Class C verbs in allowing the basic forms to exhibit both the locative GR=OBJ and the causative FIG=OBJ pattern, respectively. These basic patterns are then altered to the reverse pattern via -in and -ang derivations. (44) TIE (tegul) a. Basic locative pattern: GR=OBJ (> FIG=OBL) Tiang negul(-in) jaran=e (aji tali). I af.tie horse=def with rope ‘I tied the horse (with a rope).’ b. -ang derived causative pattern: FIG=OBJ > GR=OBL Tiang negul-ang tali=ne ka jaran=e. I af.tie-caus rope=def to horse=def ‘I tied the rope to the horse.’ (45) TIE (tegul) a. Basic causative pattern: FIG=OBJ (> GR=OBL)14 Tiang negul jaran=e (di/ka punyan kayu=ne). I af.tie horse=def in/to trunk tree=def ‘I tied the horse (to the tree trunk).’

14 Without a Ground specification, we would not know whether the Object (the horse) is construed as a Figure or a Ground. The optional -in suffix in (44a), on the other hand, specifically marks the Object to be Ground.

Balinese valency classes

897

b. -ang derived causative pattern: FIG=OBJ > GR=OBL Tiang negul-ang jaran=e ka punyan kayu=ne. I af.tie-caus horse=def to trunk tree=de ‘I tied the horse to the tree trunk. c. -in derived locative pattern: GR=OBJ > FIG=OBJ Tiang negul-in punyan kayu=ne jaran. I af.tie-loc trunk tree=def horse Lit. ‘I tied the tree trunk (with) the horse.’ Example (45b) with -ang has the same FIG=OBJ alignment as the basic alignment pattern in (45a). But (45a) and (45b) differ crucially in that while a locative expression is optional in the former, it is obligatory in the latter due to the valency-increasing effect of the -ang suffix. (44b) and (45b) both align the Figure expressions with Object. In the former, an Instrumental Figure is aligned with the Object and has a valency-increasing effect due to the promotion of an Oblique to the Object position. In (45b) the valency is increased without promotion, the basic Theme/ Figure=OBJ alignment remaining unaffected. The verbs showing the alternation patterns above are either trivalent ditransitive verbs requiring both Ground and Figure expressions or those transitive verbs that optionally allow these. Those basic transitive verbs, such as KNOW (tawang), that do not take any optional Ground or Figure expression do not show the derived/ coded alternation patterns under study. (46) KNOW (tawang) I Made nawang anak=e cenik ento. I Made af.know person=def small that ‘I Made knows the boy.’ Other transitive verbs like EAT (ajeng/daar) and SLAUGHTER (tampah) may take an optional locative and/or an instrumental expression, but since these optional expressions do not satisfy the semantic requirements for either -ang or -in derivation (see § 4.2 and § 9), they do not show the relevant alternation patterns either. These verbs should perhaps be considered members of a separate class of transitive verbs that fail to, ones that fail to alternate due to the semantic restrictions pertaining to the alternate GR=OBJ construction. (47) EAT (ajeng) a. Anak=e luh ento ng-ajeng roti di kamar=ne. person=def female that af-eat bread in room=def ‘The girl is eating bread in the room.’

898

Masayoshi Shibatani and Ketut Artawa

b. *Anak=e luh ento ng-ajeng-in kamar=ne roti. person=def female that af-eat-loc room=def bread ‘The girl is eating bread in the room.’ (48) a. Anak=e ento nampah celeng=e aji parang. person=def that af.slaughter pig=def with cleaver ‘The man slaughtered the pig with a cleaver.’ b. *Anak=e ento nampah-ang parang=e ka celeng=e. person=def that af.slaughter-caus cleaver=def to pig-def ‘The man slaughtered the pig with a cleaver.’

4.6 Precategorial Class 1 No Basic pattern: -in derived locative pattern: GR=OBJ > FIG=OBJ -ang derived causative pattern: FIG=OBJ > GR=OBL (i) SHOW (edeng), GIVE (enjuh), SAY/TELL (orah), INSERT (selek) (ii) POUR (turuh), THROW (entung), HIDE (engkeb) (iii) DRESS SOMEONE (payas) Precategorial verbs form a large class, and, lacking a specified valency pattern, many of them allow both -in and -ang derivations with the respective alignment pattern, as shown below. (49) SHOW (edeng) a. -in deried locative pattern: GR=OBJ > FIG=OBJ Anak=e luh ento ng-edeng-in guru=ne buku. person=def girl that af-show-loc teacher=def book ‘The girl showed the teacher the book.’ b. -ang derived causative pattern: FIG=OBJ > FIG=OBL Anak=e luh ento ng-edeng-ang buku ka guru=ne. person=def girl that af-show-caus book to teacher=def ‘The girl showed the book to the teacher.’ (50) POUR (turuh) a. -in derived locative pattern: GR=OBJ > FIG=OBJ Anak=e ento nuruh-in lumur=e yeh. person=def that af.pour-loc glass=def water Lit. ‘The man poured the glass water.’

Balinese valency classes

899

b. -ang derived causative pattern: FIG=OBJ (> GR=OBL) Anak=e ento nuruh-ang yeh=e ento (ka lumur=e ento). person-=def that af.pour-caus water=def that to glass=def that ‘The man poured the water into the glass.’ (51) DRESS SOMEONE (payas) a. -in derived locative pattern: GR=OBJ (> FIG=OBL) Tiang mayas-in anak=e cenik ento (aji baju). I af.dress-loc child=def small that (with shirt) Lit. ‘I dressed the child (with a shirt).’ b. -ang derived causative pattern: FIG=OBJ > GR=OBL Tiang mayas-ang baju=ne ento ka anak=e cenik ento. I af.dress-caus shirt=def that to child=def small that ‘I put the shirt on the child.’ While the alignment patterns displayed by the -in and -ang forms above are similar, the three sets of the examples above represent three different argument patterns. SHOW (edeng), GIVE (enjuh), SAY/TELL (orah), and ENTER/INSERT (celep) all require both a Figure and a Ground expression as arguments in either the -in and the -ang pattern. With POUR (turuh), THROW (entung), and HIDE (engkeb), the -in LOC=OBJ pattern requires both Figure and Ground arguments, while in the -ang FIG=OBJ pattern, an OBL Ground expression typically shows up, but it is not a requirement. DRESS SOMEONE (payas) reverses this pattern and calls for both Figure and Ground arguments in the -ang FIG=OBJ form, while an OBL Figure expression is optional in the -in GR=OBJ pattern. Compared to POUR (turuh), the basic verb FILL (isinin) aligns with DRESS SOMEONE (payas) in that a Figure expression is optional in the GR=OBJ construction with them, whether the GR=OBJ pattern is basic as in FILL (isinin) or -in derived as in DRESS SOMEONE (payas). There are two GIVE verbs in Balinese. One is the basic verb baang, which allows the GR=OBJ and the FIG=OBJ alignment without derivation, as seen in the earlier examples in (30). The other is precategorial enjuh, which yields the GR=OBJ pattern through -in derivation and the FIG=OBJ pattern through -ang derivation, as shown below. The basic verb baang is more general in meaning, whereas the precategorial enjuh specifically means giving/delivering something to someone by hand. (52) GIVE (enjuh) a. -in derived locative pattern: GR=OBJ > FIG=OBJ Ia ng-enjuh-in timpal=ne buku. s/he af-give-loc friend=3.poss buku ‘S/he gave his/her friend a book.’ b. -ang derived causative pattern: FIG=OBJ > GR=OBL Ia ng-enjuh-ang buku=ne ento ka timpal=ne. S/he af-give-caus book=def that to friend=3.poss ‘S/he gave the book to his/her friend.’

900

Masayoshi Shibatani and Ketut Artawa

The -in derived form here contrasts with the -ang derived form in that the former has the double Object pattern, while the latter has the OBJ > OBL pattern, reflecting the semantics of caused motion characteristic of the -ang causative construction. A number of notionally bivalent precategorial verbs show somewhat irregular patterns. There are some that allow both -in and -ang derived forms with the same alignment pattern. In the case of LOOK (telektek), the -ang version is more commonly used than the -in version. (53) LOOK (telektek) Tiang nelektek-ang/-in anak=e cenik ento. I af.look-caus/-loc person=def small that ‘I look at the child.’ The verb FOLLOW (tuut) has a basic alignment pattern, but it still allows both -in and -ang derived forms that differ in meaning. The basic form tuut means to copy or to imitate, but the -in derived form means to follow or to obey, and the -ang derived form means to obey. (54) FOLLOW (tuut/tutut) a. Basic Tiang nuut anak=e luh ento. I af.follow person=def female that 1. ‘I copied (the work done by) the girl.’ 2. ‘I imitate the girl.’ 3. ‘I obeyed the girl.’ b. -in derived Tiang nuut-in anak=e luh ento. I af.follow-loc person=def female that 1. ‘I emulate the girl.’ 2. ‘I obeyed the girl.’ c. -ang derived Tiang nuut-ang anak=e luh ento. I af.follow-caus person=def female that 1. ‘I copied the girl.’ 2. ‘I obeyed the girl.’

4.7 Precategorial Class 2 No Basic pattern: -in derived locative pattern: GR=OBJ *-ang derived causative pattern: FIG=OBJ

Balinese valency classes

901

HELP (tulung), SEE/MEET (tepuk) CALL X Y (kauk) There are several precategorial verbs that have -in derived forms with no corresponding -ang derived forms. The verb CALL X Y (kauk) belongs to this class. (55) HELP (tulung) a. Tiang nulung-in anak=e luh ento. I af.help-loc person=def female that ‘I helped the girl.’ b. Tiang nulung-in guru=ne ngae umah. I af.help-loc teacher=def build house ‘I help the teacher to build a house.’ (56) CALL X Y (kauk) Luh Sari ngauk-in anak=e ento I belog. female Sari af.call-loc person=def that art fool ‘Sari called the man a fool.’

5 Other verb-coded valency alternations In addition to -in and -ang derivations, there are several other processes that provide verb-coded valency alternations in Balinese. The most pertinent ones are -ang benefactives, ma-middles, and -a passives. All these apply to (di)transitive verbs, provided the relevant semantic restrictions are satisfied. ma-middles play the important role of filling the lexical gap of missing intransitives and, thereby, feed the intransitive repertoire of the language.

5.1 Valency-increasing -ang benefactive applicative As in many other languages, the benefactive applicative is limited to situations involving the transfer of an object to a beneficiary, and, as such, it typically involves basic transitive verbs of procurement and production; however, it does not apply to intransitive verbs and those transitive situations where the transfer of an object to a beneficiary does not materialize, as e.g., throwing away garbage on someone’s behalf (Shibatani 1996). The following form with the verb BUY (beli) is a typical Balinese benefactive applicative, by which we mean a construction that aligns a Beneficiary with the Object and whose meaning conveys an intended or successful transfer of a Theme/Figure to the Beneficiary. Syntactically, the benefactive applicative is ditransitive, and semantically, it contrasts with expressions in which a Beneficiary is coded in Oblique form, as in the contrast between English I

902

Masayoshi Shibatani and Ketut Artawa

bought Made this book and I bought this book for Made – the former is a benefactive applicative without morphology (see below). (57) BUY (beli) a. Basic bivalent form Tiang meli buku=ne ene. I af.buy book=def this ‘I bought this book.’ b. Benefactive trivalent form Tiang meli-ang Made buku=ne ene. I af-ben Made book=def this ‘I bought Made this book.’ Involvement of the suffix -ang and of a transferred object would make us wonder whether the benefactive alignment pattern has the Theme/Figure argument in Object position. But this is not the case. As in (57b) above, the benefactive applicative has what looks like a Ground (recipient Beneficiary) expression in Object position, the pattern that -in derivation normally entails. Upon closer examination, however, it becomes clear that the Beneficiary expression in Object position is not a Ground. First of all, a true Ground expression cannot be aligned with the Object via benefactive derivation. Unlike the English dative-shift counterpart (as in the translation in the parentheses below), benefactive applicative (58b) below does not mean that the Beneficiary “my mother” was the goal to which the letter was to be sent. In other words, (58a) and (58b) are not synonymous as are the English translations I wrote a letter to my mother and I wrote my mother a letter. What grammatically corresponds to the English sentence I wrote my mother a letter would be (58c), but this sentence is not very well-formed in Balinese because of the semantic requirement on the -in derivation that the Ground be highly interactive with the relevant action (see Section 9, and compare (58c) with the earlier example in (38b) and with (60) below). (58) a. Tiang nulis surat ka meme=n tiang=e. I af.write letter to mother=poss I=poss ‘I wrote a letter to my mother.’ b. Tiang nulis-ang meme=n tiang=e surat. I af.write-ben mother=poss I=poss letter ‘I wrote a letter for my mother.’ (≠ I wrote my mother a letter.) c. *?Tiang nulis-in meme=n tiang=e surat. I af.write-loc mother=poss I=poss letter ‘I wrote my mother a letter.’ Secondly, a true Ground expression can be overtly specified in addition to a Beneficiary nominal, as in (59b) below.

Balinese valency classes

903

(59) a. Tiang ngirim buku ka sekolah. I af.send book to school ‘I sent books to the school.’ b. Tiang ngirim-ang guru=ne buku ka sekolah. I af.send-ben teacher=def book to school ‘I sent books to the school for the teacher.’ The above sentence still entails that the teacher is to ultimately receive the books that are sent to the school, but ka sekolah ‘to the school’ is the true Ground in this expression. Contrast this with the -in locative construction, where a true Ground is coded as the Object. (60) Tiang ngirim-in sekolah=e buku. I af.send-loc school=def book ‘I sent the school books.’ The Ground and the Beneficiary are also distinguished in Standard Indonesian when they occur as Obliques (see (61a, b) below). The benefactive applicative -kan form in (61c) is not exactly synonymous with (61b) in that the former implies that while “his/ her father” is not the addressee of the letter, he nevertheless temporarily obtains the letter written on his behalf. That is, (61c) means that s/he wrote the letter on his/her father’s behalf and handed it to him (to mail to the ultimate recipient).any other languages, the benefactive applicative (61) a. Standard Indonesian Dia me-nulis surat kepada ayah=nya. s/he af-write letter to father=3poss ‘S/he wrote a letter to his/her father.’ b. Dia me-nulis surat untuk ayah=nya. s/he af-write letter for father=3poss ‘S/he wrote a letter for his/her father.’ c. Dia me-nulis-kan ayah=nya surat. s/he af-write-ben father=3poss letter ‘S/he wrote a letter for his/her father. (≠ S/he wrote his/her father a letter.) (Sneddon 1996: 81) An OBL Beneficiary expression corresponding to (61b) is not available in Balinese (see (62c) below for an alternative). The -ang benefactive applicative, which aligns a Beneficiary expression with the Object, is, thus, different from the GR=OBJ construction effected via -in derivation in its alignment pattern. While synchronically we may want to distinguish the causative -ang, which aligns a causee Theme/Figure with the Object, from the -ang

904

Masayoshi Shibatani and Ketut Artawa

benefactive, which aligns a Beneficiary with the Object, the causative/benefactive applicative polysemy finds an explanation in the historical development of the benefactive applicative from the causative via direct causative construal of a causative semantics following the derivation pattern of the GIVE verb. The causative semantics [X CAUSE Y to HAVE Z] under the direct causative construal may lexicalize as [X HAVE-CAUSE Y (the child) Z (a book)], where the abstract causative predicate is realized as a causative suffix, as in Ainu kor-e (have-caus) ‘give’. Similarly, [X CAUSE Y to HEAR Z] is lexicalized as nu-re (hear-caus) ‘inform’ in Ainu. In some other languages, like English, lexicalization of these causative semantics incorporates the causative component CAUSE into a hosting verb, giving rise to expressions like “X gives Y (the child) Z (a book)” and “X informs Y of Z”. The rise of benefactive applicatives from causatives follows these lexicalization patterns. [X CAUSE Y to BUY Z], for example, may give rise to a lexical benefactive applicative “X buys Y Z” following the “X gives Y Z”, or it may give rise to a morphological benefactive applicative in the form of “X buy-CAUS Y Z” following the GIVE pattern “X have-CAUS Y Z”, where the causative suffix is subsequently reanalyzed as a benefactive applicative suffix. In Balinese the verb of giving baang follows the English lexicalization pattern, while benefactive applicatives follow the Ainu pattern of realizing CAUSE morphologically.15 Since the basic function of the locative -in, the causative-ang, and the benefactive -ang is to align respectively a Ground location, a Theme Figure, and a Beneficiary with the Object, rather than to increase valency per se, these do not combine. In other words, they compete for the (first) Object position and once Object alignment is fixed by a derivation, it cannot be altered by another derivation. Thus, even semantically plausible -ang/-in derived verbs such as turuh-ang ‘pour-CAUS’ and kirim-in ‘send-LOC’ do not yield -ang benefactives. Serial verb benefactive constructions involving the verb baang ‘give’ must be used instead, as in (62c). (62) a. Tiang nuruh-ang yeh ka lumur=e. I af.pour-caus water to glass=def ‘I poured water into the glass.’ b. *Tiang nuruh-ang-ang guru=ne yeh ka lumur=e. I af.pour-caus-ben teacher=def water to glass=def ‘I poured water into the glass for the teacher.’ c. Tiang nuruh-ang yeh ka lumur=e baang guru=ne. I af.pour-caus water to glass=def give teacher=def ‘I poured water into the glass for the teacher.’ Compare (62a–b) and (59a–b). Both (59a) and (62a) have the semantically comparable alignment pattern of FIG=OBJ > GR=OBL, but only the former yields a benefactive -ang form (59b). Similarly, while both (63a) and (64a) below have the same 15 Also note the interpretation “I carried the girl” of the causative form in (121b).

Balinese valency classes

905

alignment pattern of GR=OBJ > FIG=OBJ, only the underived structure in (63a) yields a correct -ang benefactive counterpart. (63) a. Tiang ngodot poh aji tiuk. I af.cut mango with knife ‘I cut the mango with a knife.’ b. Tiang ngodot-ang anake=e cenik poh aji tiuk. I af.cut-ben person=def small mango with knife ‘I cut the mango with a knife for the child.’ (64) a. Tiang ngebat-in meja=ne aji taplak meja. I af-spread-loc table=def with cloth table Lit. ‘I spread the table with a tablecloth.’ b. *Tiang ngebat-in-ang meme=ne meja=ne aji taplak meja. I af.spread-loc-ben mother=def table=def with cloth table Lit. ‘I spread the table with a tablecloth for mother.’ c. ?Tiang ngebat-in meja=ne aji taplak meja baang meme=ne.16 I af.spread-loc table=def with cloth table give mother=def Lit. ‘I spread the table with a tablecloth for mother.’ The precategorial verb HUNT (boros) is interesting with regard to the above discussion. As seen above, a transitive derived from the locative -in or the causative -ang suffixation cannot undergo the benefactive derivation. What is interesting about the verb boros ‘hunt’ is that (i) its ma-middle form takes an indefinite Object (see (65a) and the discussion on this topic in § 5.2), and (ii) the locative-based benefactive is directly built on the verb root, thereby avoiding the -in-ang combination. (65) HUNT (boros) a. Anak=e ento ma-boros kidang. person=def that mid-hunt deer ‘The man hunted (himself) deer.’ b. Anak=e ento moros-in kidang. person=def that af.hunt-loc deer ‘The man hunted deer.’

(Does not allow a definite Object)

(Allows a definite Object)

c. *Anak=e ento moros-in-ang tiang kidang. person=def that af.hunt-loc-ben I deer ‘The man hunted me deer.’ 16 The oddity of this sentence is due to the difficulty of construing the table with a tablecloth on it as a transferred entity. Compare this with (62c), where the glass of water is easily construable as a transferred entity.

906

Masayoshi Shibatani and Ketut Artawa

d. Anak=e ento moros-ang tiang person=def that af.hunt-ben I kidang. deer ‘The man hunted me deer.’

(Semantic equivalent of (65c))

Usually the (b) to (d) derivation above is not possible with -in transitive verbs. (66) HEAT (panas-in) a. Ia manas-in yeh=e ento. s/he af.heat-loc water=def that ‘S/he heated the water.’ b. *Ia manas-in-ang tiang yeh=e ento. s/he af.heat-loc-ben I water=def that ‘S/he heated me the water.’ c. *Ia manas-ang tiang yeh=e ento. s/he af.heat-ben I water=def that ‘S/he heated me the water.’ d. Ia manas-in yeh=e ento baang tiang. s/he af.heat-loc water=def that give I ‘S/he heated the water for me.’ We will look at the combinability of the -in locative with the -ang causative in Section 8 below.

5.2 Valency-decreasing ma-middle and -a passive Both ma-middle and -a passive reduce syntactic valency (but see below). While the resultative ma-middle and the passive are similar in syntactic transitivity, they differ in semantic transitivity; the former is syntactically as well as semantically monovalent, while the latter is syntactically monovalent but semantically bivalent, allowing the possibility of encoding the Agent as an Oblique nominal. Compare the following examples: (67) a. Anake ento negul jaran=e dipunyan kayu=ne. person that af.tie horse=def to trunk tree=def ‘The man tied the horse to the tree trunk.’ b. Jaran=ne ma-tegul di punyan kayu=ne (*teken anak=e horse=def mid-tie to trunk tree=def by person=def ento). that ‘The horse was tied to the tree trunk.’

(Active)

(Middle)

Balinese valency classes

c. Jaran=ne tegul-a di punyan kayu=ne (teken anak=e horse=def tie-pass to trunk tree=def by person=def ento). that ‘The horse was tied to the tree trunk (by the man).’

907 (Passive)

Together with a separate periphrastic transitive reflexive construction, Balinese morphological middle ma-forms cover the central middle semantics of body actions (grooming and changing body posture), translational and non-translational motions, reciprocals, resultatives, spontaneous processes, and antipassive middles (Shibatani & Artawa 2007). Many roots yielding grooming verbs are either precategorial (sugi ‘wash (face)’, baseh ‘wash (limbs)’, ambuh ‘wash (hair)’) or noun-based precategorials (suah ‘comb’, sikat ‘(brush) teeth’, pupur ‘powder (face)’, which can be used as nouns without any derivation but which must undergo a derivation in order for them to function as a verb. Similarly, many body-posture middles (malingeb ‘lie face down’, ma-sila ‘sit down cross-legged’, ma-jujuk ‘stand up (straight, not bending)’, ma-tangi ‘stand up’), non-translational and translational motion middles (ma-kecog ‘jump’, ma-lincer ‘spin’, ma-jalan ‘walk’, ma-laib ‘run’), as well as some miscellaneous types (ma-suryak ‘shout’, ma-takon ‘ask’, ma-bangkes ‘sneeze’, ma-kecuh ‘spit’, ma-kenyir ‘smile very briefly’) are all precategorials, and they can be used only with the middle marker. The middle derivation here does not constitute a case of valency reduction or intransitivization since these roots are precategorial without a basic valency value. There are, in addition, transitive grooming verbs that reduce valency via maderivation. (68) a. Wayan nyukur Ketut. Wayan af.shave Ketut ‘Wayan is shaving Ketut (his head hair).’ b. Wayan ma-cukur. Wayan mid-shave ‘Wayan is cutting his hair.’17 ‘Wayan had his hair cut (by somebody).’ Other valency-reducing middle constructions are also transitive-based. (69) Reciprocal middle a. I Nyoman ngaplug I Ketut. art Nyoman af.bump art Ketut ‘Nyoman bumped into Ketut.’

17 This reading is not available for some speakers.

908

Masayoshi Shibatani and Ketut Artawa

b. I Nyoman lan I Ketut ma-kaplug. art Nyoman and art Ketut mid-bump ‘Nyoman and Ketut bumped into each other.’ (70) Resultative middle: COVER (rurub) a. Ia ngerurub anak=e cenik ento (aji) saput. s/he af.cover person=def small that (with) blanket ‘S/he covered the child with a blanket.’ b. Anak=e cenik ento ma-rurub (aji) saput. person=def small that mid-cover (with) blanket ‘The child is covered with a blanket.’ (71) Antipassive middle: EAT (daar) a. Tiang naar nasi=ne. I af.eat rice=def ‘I ate the rice.’ b. Tiang ma-daar (nasi). I mid-eat rice (Cannot have the definite nasi=ne ‘the rice/meal’) ‘I ate (a meal).’ A ma-middle form may enter the active (transitive)/antipassive (intransitive) opposition with a derived transitive construction when the root is precategorial, as below. (72) Active/antipassive opposition based on a precategorial a. -in derived transitive Ia ngeneh-in tiang. s/he af-think-loc I ‘S/he loves me.’ b. ma-antipassive middle Ia ma-keneh teken tiang. s/he mid-think with I ‘S/he is in love with me.’ While the valency-decreasing property of these middle constructions is clear, the nature of the syntactic transitivity of some middle expressions is not quite straightforward. That is, certain middle constructions have what looks like an Object. We have already seen instances of this in (65a) and (71b) above, where an indefinite Object may occur in ma-middle constructions. While it is true that a definite Object cannot occur in such a form, the definiteness restriction itself is not clear. Sometimes definite noun phrases such as basa Inggeris “English language” may co-occur with a middle verb, as in the following example (73a):

Balinese valency classes

909

(73) a. Anak=e cenik cenik ento m-uruk basa Inggeris. Person=def small small that mid-learn language English ‘The children are learning English.’ b. *Basa Inggeris m-uruk-a teken anak=e cenik cenik ento. language English mid-learn-pass by person=def small small that ‘English is learned by the children.’ As indicated by (73b), the Object of the middle verbs cannot be turned into a passive Subject. This means either that the definite Object in (73a) is not a true Object syntactically or that middles, even if they had a syntactic Object, would not undergo passivization, which is the case in other languages. Notice that a (regular) transitive Object, whether it is basic or derived via the -ang causative or the -in locative, turns into a passive Subject. (74) a. Guru=ne ng-uruk-ang basa Inggeris ka anake cenik cenik teacher=def af-learn-caus language English to person small small ento. that ‘The teacher teaches English to the children.’ b. Basa Inggeris uruk-ang-a ka anake cenik cenik ento (teken language English learn-caus-pass to person small small that by guru=ne). teacher=def ‘English is taught to the children (by the teacher).’ c. *Basa Inggeris suba m-uruk-ang ka anake cenik cenik ento. language English already mid-learn-caus to person small small that ‘English is already taught to the children.’ (75) a. Guru=ne ng-uruk-in anake cenik cenik ento basa Inggeris. teacher=def af-learn-loc person small small that language English ‘The teacher teaches the children English.’ b. Anake cenik cenik ento uruk-in-a basa Inggeris (teken person small small that learn-loc-pass language English by guru=ne). teacher=def ‘The children are taught English (by the teacher).’ c. *Anake cenik cenik ento suba m-uruk-in basa Inggeris. person small small that already mid-learn-loc language English ‘The children are already taught English.’

910

Masayoshi Shibatani and Ketut Artawa

One of the major differences between the ma-resultative middle and the -a passive is that the former is far more restricted in its applicability, as indicated by the ungrammatical forms in (74c) and (75c) above. There are two factors relevant to the formation of Balinese middle constructions. One is a morphological consideration. Like some other derivational affixes, the middle prefix ma- does not combine with the locative -in suffix. Some causative forms can be made from either the -ang or -in suffix, as in (a) example below. But only the ma … ang combination, which is realized as ma … an, is permitted.18 (76) a. -ang(/-in) causative Tiang mutih-ang/in tembok-e. I af.white-caus/loc wall-def ‘I whiten the wall.’

(AF)

b. Tembok=e suba ma-putih-an. wall=def already mid-white-caus(?) ‘The wall is already whitened.’ c. *Tembok=e suba ma-putih-in. wall=def already mid-white-loc ‘The wall is already whitened.’ (77) a. -in derived transitive Ia mules-in umah=ne anyar. s/he af.sleep-loc house=3poss new ‘S/he slept in his new house.’ b. ma-resultative middle *Umah=ne anyar suba ma-pules-in. house=def new already mid-sleep-loc ‘His new house has already been slept in.’ (78) a. -ang derived causative Tiang nylampar-ang sandal ka cicing=e ento. I af.throw-caus sandal to dog=def that ‘I threw a sandal at the dog.’ b. ma-resultative middle Sandal=ne ento ma-slampar-an ka cicing=e ento. sandal=def that mid-throw-caus(?) to dog=def that ‘The sandal is thrown at the dog.’ 18 Some younger speakers from Denpasar permit both ma … in and ma … ang/an forms; that is, for them (76c) and (76b) are also acceptable as well as the ma … ang version of (76b). Some, on the other hand, do not even accept (76b), indicating the general resistance for ma- to combine with either -ang and -in.

Balinese valency classes

911

The other restriction on the derivation of ma-resultatives is semantic in nature. Specifically, the subject of a ma-resultative middle must be in a perceptibly new state characterized in terms of a change of state or location resulting from the previous event. The best qualified resultative subjects, accordingly, are the following type: (79) a. Active with underived verb with GR=OBJ Anake ento ngodot poh aji tiuk. person that af.cut mango with knife ‘The man cut/sliced the mango with a knife.’ b. Resultative middle Poh=e suba ma-godot aji tiuk. mango=def already mid=cut with knife ‘The mango is already cut/sliced with a knife.’ (80) a. Active with underived verb with FIG=OBJ Anak=e ento ngejang buku di meja-ne. person=def that af.put book on table=def ‘The man placed the book on the table.’ b. Resultative middle Buku=ne suba ma-ejang di meja=ne. book=def already mid-put on table=def ‘The book is already placed on the table.’ Neither of the double Objects of GIVE (baang) and -ang benefactives qualifies as an entity in a new state, nor do they turn up as the subject of ma-resultatives. (81) GIVE (baang) a. Tiang maang anak=e cenik ento buku. I af.give person=def small that book ‘I gave the child a book.’ b. *Anak=e cenik ento suba ma-baang buku. person=def small that already mid-give book ‘The child is already given a book.’ c. *Buku suba ma-baang anak=e cenik ento. book already mid-give child=def small that ‘The book is already given to the child.’ (82) a. Benefactive -ang ditransitive Wayan meli-ang anak=e cenik ento buku. Wayan af.buy-ben person=def small that book ‘Wayan bought the child a book.’

912

Masayoshi Shibatani and Ketut Artawa

b. *Anak=e cenik ento suba ma-beli(-ang) buku. person=def small that already mid-buy(-ben) book ‘The child has already been bought a book.’ c. *Buku=ne ento suba ma-beli(-ang) anake=e cenik ento. book=def that already mid-buy-ben person=def small that Lit. ‘The book is already bought the child.’ Again, the -a passive is not as restricted as the ma-resultative, and the ungrammatical forms in (81b, c) and (82b, c) can be made passive grammatically by replacing the middle prefix with the passive suffix – see (83) and (85) below.

6 Object symmetry As seen above, Balinese passivization converts a wide variety of Objects into passive Subjects. Some of the second Objects of double Object constructions, however, cannot be converted in this way.19 Double Objects may be sanctioned by underived ditransitive verbs or by -in/-ang derivations. While some double Objects are symmetrical in the sense that both behave similarly to the monotransitive Object, some are not. The double Objects of the underived verb GIVE (baang) are symmetrical at least with respect to passivization and topic formation, such that both the first and the second Object can be made a passive Subject as well as the Topic of a PF construction. Passivization, therefore, provides two types of coded valency alternation, as shown below. (83) Lexical double Object constructions: GIVE (baang) a. Guru=ne maang anak=e cenik ento buku=ne. teacher=def af.give person=def small that book=def ‘The teacher gave the child the book.’

(AF)

b. Anak=e cenik ento baang=a buku=ne (teken guru=ne). person=def small that give=pass book=def by teacher=def ‘The child was given the book (by the teacher).’ c. Buku=ne ento baang=a anak=e cenik ento (teken guru=ne). book=def that give=pass person-def small that by teacher=def ‘The book was given to the child (by the teacher).’ 19 We avoid the terms Primary and Secondary Objects since determining which of the double Objects is syntactically primary requires a thorough investigation beyond the checking of the syntactic status of the two relevant Objects against a couple of syntactic phenomena. We use the terms “first Object” and “second Object” in reference to their order of occurrence in double Object constructions.

Balinese valency classes

913

Double Object constructions similar to GIVE in meaning, such as -in derived SEND (kirim-in) and SHOW (edeng-in), as well as benefactive double Object constructions, have symmetrical Objects. (84) -in derived double Object construction: SHOW (precategorial: edeng) a. Wayan ng-edeng-in Made poto=ne. Wayan af-show-loc Made photo=def ‘Wayan showed Made the photo.’ b. Made edeng-in-a poto=ne (teken Wayan). Made show-loc-pass photo=def (by Wayan) ‘Made was shown the photo (by Wayan).’ c. Poto=ne edeng-in-a Made (teken Wayan). photo=def show-loc-pass Made (by Wayan) Lit. ‘The photo was shown Made (by Wayan).’ (85) Benefactive double Object constructions a. Guru=ne beli-ang Made buku=ne ene. teacher=def buy-ben Made book=def this ‘The teacher bought Made this book.’ b. Made beli-ang-a buku=ne ene (teken guru=ne). Made buy-ben-pass book=def this (by teacher=def) ‘Made was bought this book (by the teacher).’ c. Buku=ne ene beli-ang-a Made (tekan guru=ne). book=def this buy-ben-pass Made (by teacher=def) Lit. ‘This book was bought Made (by the teacher).’ In contrast, the second Objects of many other double Object constructions involving -in derived verbs do not behave like the first Object, showing the nonsymmetric nature of the double Objects involved. (86) POUR (turuh) a. Made nuruh-in lumur=e yeh. Made af.pour-loc glass=def water Lit. ‘Made poured the glass (with) water.’ b. Lumur=e ento turuh-in-a yeh (teken Made). glass=def that pour-loc-pass water (by Made) Lit. ‘The glass was poured water (by Made).’ c. *Yeh=e turuh-in-a lumur=e (teken Made). water=def pour-loc-pass glass=def (by Made). ‘The water was poured (into) the glass (by Made).’

914

Masayoshi Shibatani and Ketut Artawa

(87) HIDE (engkeb) a. Anak=e luh ento ng-engkeb-in guru=ne buku=n person=def female that af-hide-loc teacher=def book=3.poss dane=ne. dear/he=def Lit. ‘The girl hid the teacher his book.’ b. Guru=ne engkeb-in-a buku=n dane=ne (teken teacher=def hide-loc-pass book=3.poss dear/he=def by anak=e luh ento). person=def female that ‘The teacher had his book hidden by the girl.’ c. *Buku=n dane=ne engkeb-in-a guru=ne (teken book=3.poss dear/he=def hide-loc-pass teacher=def by anak=e luh ento). person=def female that ‘His book was hidden (from) the teacher by the girl.’ As with many other phenomena, object symmetry is not entirely clear-cut. The -in double Object construction with STEAL (paling) exhibits the following pattern, where the status of the second Object is unclear with regard to passivization: (88) STEAL (paling) a. Anak=e ento nge-maling-in guru=ne pipis spp. person=def that af-steal-loc teacher=def money school Lit. ‘The man stole the teacher the school money.’ b. Guru=ne paling-in-a pipis spp (teken anak=e ento). teacher=def steal-loc-pass money school (by person=def that) ‘The teacher had the school money stolen (by the man).’ c. ?Pipis spp paling-in-a guru=ne (teken anak=e ento. money school steal-loc-pass teacher=def (by person=def that) Lit. ‘The school money was stolen the teacher (by the man).’

7 Intransitive verbs On morphological grounds, basic (underived) Balinese intransitive verbs are divided into two classes: (i) those calling for the PF zero-marking and (ii) those taking the AF nasal-marking. While there are several eventive verbs such as ulung ‘fall’ and teka ‘come’ that are PF zero-marking, PF zero-marking verbs are generally

Balinese valency classes

915

stative predicates denoting uncontrollable states of affairs, and the AF nasalmarked verbs are largely activity verbs, suggesting split intransitivity (Arka 2003).20 We shall, accordingly, consider the subject of AF-marked intransitives to be Agent (A), or Agent/Theme (A/TH) when motion is involved, and that of PFmarked intransitives to bear the Theme (TH) role. (89) PF (0̸-prefix) monovalent predicates a. Anak=e cenik ento seduk. person=def small that pf.hungry ‘The boy is hungry.’ b. Tembok=e putih. wall-def pf.white ‘The wall is white.’ c. Raket tiang=e usak. racket I=poss pf.broken ‘My racket is broken.’ (90) AF (N-prefix) monovalent predicates a. Anak-e cenik ento negak di bataran=e. person=def small that af.sit on ground=def ‘The child sits on the ground.’ b. Anak=e cenik ento nongos di desa=ne. person=def small that af.live in village=def ‘The child lives in the village.’ c. Anak=e cenik ento ng-eling. person=def small that af-cry ‘The child cried.’ The intransitive verbs in the sample verb list group in the following way on the basis of the marking they take.

20 These intransitives are all non-precategorial. A similar intransitive split is seen in Philippine languages as well (see Shibatani 1988). Arka (2003) gives a list of verbs showing the relevant meaning contrast between the PF zero form and the AF nasal form; e.g., joh ‘far’: nge-joh ‘make oneself far’, enceh-enceh ‘involuntarily urinate repeatedly’: ng-eceh ‘urinate’. We know that the nasal marking in these intransitive verbs is the same AF marking that occurs in transitive verbs from the fact that the intransitive nasal marking remains the same when intransitive verbs are transitivized via -ang derivation, and also from the fact that it is lost in PF transitive form (see (94)). Similarly, the PF intransitive verb has the same zero marking as in transitive PF verb forms (cf. (63a) and (93c)).

916

Masayoshi Shibatani and Ketut Artawa

(91) 0̸-marked COME (teka), LAUGH (kedek), SLEEP (pules), FALL (ulung), BURN (puun), BREAK/BROKEN (belah), DRY(tuh), DIE/DEAD (mati), HUNGRY (seduk), SAD (sebet), FEEL PAIN (sakit), LIKE (demen), FRIGHTENED (takut) (92) N-marked 21 SIT (negak), CLIMB (menek), ROLL (ngelilik), LIVE (nongos), APPEAR (ngenah), FEEL COLD (ngedged), and all other intransitives in the sample list. Many of these basic monovalent verbs undergo either -ang or -in derivation, or both, and become transitive, exhibiting the AF/PF alternation. Observe the alternation in the initial consonant between the transitive AF form (93b) and the transitive PF form (93c), which has the same zero marking as the basic PF intransitive form (93a). (93) Basic PF intransitive with a TH subject a. Tembok-e putih. wall-def pf.white ‘The wall is white.’ b. Tiang mutih-ang tembok-e. I af.white-caus wall-def ‘I whitened the wall.’

(AF)

c. Tembok-e putih-ang tiang. wall-def pf.white-caus I ‘I whitened the wall.’

(PF)

(94) Basic AF intransitve with an A/TH subject a. Tiang negak di dampar=e. I af.sit on bench=def ‘I sit on the bench.’ b. Cai negak-ang tiang di dampar=e. you af.sit-caus I on bench=def ‘You sat me on the bench.’

(AF)

c. Tiang tegak-ang cai di dampar=e. I pf.sit-caus you on bench=def ‘You sat me on the bench.’

(PF)

21 The root forms of these all have an oral/zero initial consonant, which reveals itself when the forms are used in PF of the derived transitive constructions.

Balinese valency classes

917

d. Tiang negak-in dampar=e. I af.sit-loc bench=def ‘I sit on the bench.’

(AF)

e. Dampar=e tegak-in tiang. bench=def pf.sit-loc I ‘I sit on the bench.’

(PF)

Besides these basic (underived) intransitive verbs, there are two other types of derived intransitive verbs. They are: (i) ma-derived middles, discussed earlier in Section 6, and (ii) reduplicated intransitive verbs, which may be derived from precategorial roots like JUMP (kejong-kejong < kejong), from the roots of nasal-marked intransitives like ROLL (gelilik-gelilik < ngelilik), or from zero-marked intransitives like COME (teka-teka < teka). These reduplicated forms, which do not show the AF/ PF/MID marking contrast, may co-exist with basic or ma-derived forms, as below. (95) Basic intransitive: ROLL (gelilik) a. Bola=ne ngelilik. ball=def af.roll ‘The ball rolled.’ b. Bola=ne gelilik-gelilik. ball=def roll-roll ‘The ball is rolling.’ (96) Precategorial: JUMP (kejong) a. Anak=e cenik ento ma-kejong. person-def small that mid-jump ‘The child jumped/jumps.’ b. Anak=w cenik ento kejong-kejong. person-def small that jump-jump ‘The child is/was jumping.’ In our discussion of the transitive/ditransitive alternation earlier, we noted that in Balinese the notion of valency increase is difficult to maintain consistently with the derivational processes, partly because of the presence of precategorial verbs, which do not have a basic valency value, and partly because the relevant derivation may simply alter the alignment patterns without affecting the number of required nominal expressions. A similar problem is seen with some semantically bivalent but syntactically intransitive verbs which require two nominal expressions in completing a predication but which code the second nominal as an OBL. Compare example (97) and (98):

918

Masayoshi Shibatani and Ketut Artawa

(97) Bivalent transitive verb Tiang nyepak cicing=e. I af.kick dog=def ‘I kicked the dog.’ (98) Bivalent intransitive verb: DISTRUST (sangsaya) a. Tiang sangsaya teken anak=e ento. I pf.distrust with person=def that ‘I distrust that man.’ b. Tiang nyangsaya-in anak=e ento. I af.distrust-loc person=def that ‘I distrust that man.’ As shown in (98b) above, the -in derivation makes DISTRUST (sangsaya) syntactically transitive, but without increasing valency (since the basic form of the verb is semantically bivalent, requiring two arguments). This contrasts with the valency-increasing transitivization by the -in derivation shown below. (99) Monovalent intransitive: STUDY (plajah) a. Tiang mlajah. I af.study ‘I am studying.’ b. Tiang mlajah-in basa Inggeris. I af.study-loc language English ‘I am studying English.’ On the basis of the realization pattern of the Oblique expression marked teken ‘with/about’ seen in (98a), we can distinguish several classes of (syntactically) intransitive verbs.

7.1 Basic monovalent disallowing a teken OBL This class of intransitive verbs does not permit an Oblique nominal marked teken ‘with’ and shows the following valency alternation pattern: Basic: Intransitive (*teken OBL) -in derived: Transitive -ang derived: Transitive (100) CRY (eling) a. Tiang ng-eling (*teken Ketut). I af-cry with Ketut ‘I cried.’

Balinese valency classes

919

b. Tiang ng-eling-in Ketut. I af-cry-loc Ketut ‘I cry over/about Ketut (because e.g., he is leaving / he is suffering).’ c. Ketut ng-eling-ang tiang. Ketut af-cry-caus I ‘Ketut made me cry.’ There do not seem to be many verbs that disallow a teken OBL nominal but that can introduce an OBJ via -in/-ang derivation.

7.2 Basic monovalent intransitive verbs with an optional teken OBL These verbs allow an optional teken OBL nominal, dispalying the following alternation pattern: Basic: Intransitive (teken OBL) -in derived: Transitive -ang derived: Tranistive ANGRY (pedih), CRAZY ABOUT (buduh), APPEAR (enah) (101) ANGRY (pedih) a. Nadi pedih (teken Ketut). Nadi pf.angry (with Ketut) ‘Nadi is angry (at Ketut).’ b. Nadi medih-in Ketut. Nadi af.angry-loc Ketut ‘Nadi is angry at Ketut / Nadi scolded Ketut.’ c. Ketut medih-ang Nadi. Ketut af.angry-caus Nadi ‘Ketut made Nadi angry.’

7.3 Basic bivalent intransitive verbs with an obligatory teken OBL These verbs are semantically bivalent, requiring a teken Oblique nominal, but are syntactically intransitive, and they display the following alternation pattern: Basic: Intransitive *(teken OBL) -in derived: Transitive (-ang derived: Transitive)

920

Masayoshi Shibatani and Ketut Artawa

In addition to DISTRUST (sangsaya), illustrated above in (98) above, LIKE (demen) and a few others belong to this class. LIKE (demen) is odd in that it does not nasalize the initial consonant under -in conversion and the derived -in form only accepts an indefinite Object, similar to ma-middles. Many of these verbs, e.g., LOVE (tresna) and BELIEVE (percaya), do not allow -ang forms, and even those that do have irregular meanings, as in the cases of REMEMBER (inget) and HATE (gedeg). (102) REMEMBER (inget) a. Tiang inget teken anak=e ento. I pf.remember with person=def that ‘I remember the man.’ b. Tiang ng-inget-in anak=e ento. I af-remember-loc person=def that 1. ‘I recognize the man.’ 2. ‘I reminded the man (to do something).’ 3. ‘I made the man remember.’ c. Tiang ng-inget-ang anak=e ento. I af-remember-caus person=def that 1. ‘I made the man remember.’ 2. ‘I reminded him (to do something/no to do something).’ (103) HATE (gedeg) a. Tiang gedeg teken anak=e luh ento. I pf.hate with person=def female that ‘I hate that girl.’ b. Tiang ngedeg-in anak=e luh ento. I af.hate-loc person=def female that ‘I hate that girl.’ c. Tiang ngedeg-ang anak=e luh ento. I af.hate-caus person=def female that ‘I am angry at that girl.’ While the verbs allowing the classification above in terms of a teken Oblique are mostly psychological/cognitive verbs, other intransitive verbs can also be classified on the basis of the realization pattern of a locational Oblique nominal. The following expemplify the two patterns – one involving a verb that calls for an obligatory locational Oblique (104) and the other with a verb taking an optional locational Oblique (105):

Balinese valency classes

921

(104) CLIMB (penek)22 a. Anake cenik ento menek ka gedebeg=e. person small that af.climb to cart=def ‘The child climbed onto the cart.’ b. Anake cenik ento menek-in gedebeg=e. person small that af.climb-loc cart=def ‘The child climbed onto the cart.’ c. Tiang menek-ang anake cenik ento ka gedebeg=e. I climb-caus person small that to cart=def ‘I made the child climb onto the cart.’ (105) APPEAR (enah) a. Anake cenik ento ng-enah. person small that af-appear ‘The child is visible.’ b. Tiang ng-enah ka anak=e cenik ento. I af-appear to person=def small that ‘I became visible to the child.’ / ‘I emerged in front of the child.’ c. Tiang ng-enah-in anak=e cenik ento. I af-appear-loc person=def small that Lit. ‘I appeared to the child.’ / ‘I became visible to the child.’ d. Tiang nge-nah-ang anak=e cenik ento. I af-appear-caus person=def small that ‘I made the child visible.’

7.4 Precategorial intransitives Many precategorial roots yield intransitive verbs via ma-derivation, exhibiting the following pattern, where ma-forms fill the gap in the basic intransitive repertoire. This is actually the complete paradigm, including the intransitive version, for the transitive/ditranstive alternation pattern of precategorial roots that we saw earlier in § 4.6 and § 4.7. Examples (106)–(109) below show the triad pattern, consisting of a ma-derived middle intransitive, an -in derived transitive, and an -ang derived transitive form. As can also be seen in the following examples, the derivation patterns and the semantics involved are irregular with some of these verbs. *Basic: Intransitive ma-derived: Intransitive 22 This verb also allows a transitive version without the preposition ka.

922

Masayoshi Shibatani and Ketut Artawa

-in derived: Transitive -ang derived: Transitive LEARN/TEACH (uruk), DRESS (payas), COMB (suah), THINK ABOUT (keneh), SWEAR (sumpah) (106) GET DRESSED (payas) a. Precategorial: no basic intransitive/transitive form *Anak=e cenik ento mayas. person=def small that af.dress ‘The child dressed.’ b. ma-derived monovalent intransitive Anak=e cenik ento ma-payas. person=def small that mid-dress ‘The child dressed.’ c. -in derived transitive Tiang mayas-in anak=e cenik ento baju. I af.dress-loc person=def small that shirt Lit. ‘I dressed the child (with) a shirt.’ d. -ang derived transitive Tiang mayas-ang baju=ne ka anak=e cenik ento. I af.dress-caus shirt=def to person=def small that ‘I put the shirt on the child.’ (107) ASK ABOUT (takon) a. Precategorial *I Made nakon teken tiang unduk sekolahan. I Made af.ask with I about school ‘I Made asked me about the school.’ b. ma-derived intransitive I Made ma-takon teken tiang unduk sekolahan. I Made mid-ask with I about school ‘I Made asked me about the school.’ c. -in derived transitive I Made nakon-in tiang unduk sekolahan. I Made af.ask-loc I about school ‘I Made asked me about the school.’ d. -ang derived transitive I Made nakon-ang (unduk) sekolahan teken/ka tiang. I Made af.ask-caus (about) school with/to I ‘I Made asked me about the school.’

Balinese valency classes

923

(108) SWEAR (sumpah) a. Precategorial *Ia nyumpah. he af.swear ‘He is swearing.’ b. ma-derived intransitive Ia ma-sumpah unduk Nyoman teken/ka tiang. he mid-swear about Nyoman with/to I ‘He is swearing to me about Nyoman.’ c. -in derived transitive Ia nyumpah-in tiang (*unduk Nyoman). he af.swear-loc I (about Nyoman) ‘He is swearing to me.’ d. -ang derived transitive Ia nyumpah-ang Nyoman teken tiang. he af.swear-caus Nyoman with I 1. ‘He swears about Nyoman to me.’ 2. ‘He makes Nyoman swear to me.’ (109) THINK ABOUT (keneh) a. Precategorial *Ia ngeneh teken tiang. s/he af.think with I ‘S/he is in love with me.’ b. ma-derived bivalent intransitive Ia ma-keneh teken tiang. s/he mid-think with I ‘S/he is in love with me.’ c. -in derived transitive Ia ngeneh-in tiang. s/he af-think-loc I ‘S/he loves me.’ d. -ang derived transitive Ia ngeneh-ang tiang. s/he af.think-caus I ‘S/he loves me.’ Most of these precategorial roots have both ma-middle intransitive and -ang causative transitive forms, but some lack -in locative forms largely for the reason

924

Masayoshi Shibatani and Ketut Artawa

discussed in the next section. In other words, a few items are like COUGH (kokohan) in that they are usable only as ma-middle intransitives. (110) COUGH (kokohan) Anak-e cenik ento ma-kokohan. person=def small that mid-cough ‘The child coughed.’

8 Semantic and morphological restrictions on valency-increasing processes As is clear from the previous discussions, two productive valency-increasing operations allowing coded valency alternation in Balinese are the -in locative applicative and the -ang causativization. Both processes, however, are semantically constrained. As for the locative applicative, the Ground expression to be placed in the Object position must be construable as something highly interactive with the action, such as being fully affected or being integral and connected to the denoted action in terms of a function conventionally defined. Observe the following contrast between grammatical and ungrammatical forms: (111) SIT (tegak): di-marked location a. Ia negak di kursi=ne. s/he af.sit in chair=def ‘S/he sat in the chair.’ b. Ia negak-in kursi=ne. s/he af.sit-loc chair=def ‘S/he sat in the chair / S/he occupies the chair – a professional post.’ (112) a. Ia negak di taman=e. s/he af.sit in garden=def ‘S/he sat in the garden.’ b. *Ia negak-in taman=def. s/he af.sit-loc garden=def ‘S/he sat in the garden.’ (113) COME (teka): ka-marked goal location: a. Ia teka ka rumah sakit (ng-adep nasi/bakal ma-preksa gigi). s/he come to house sick af-sell food/to mid-exam tooth ‘S/he came to the hospital (to sell food/to get his/her teeth examined).’

Balinese valency classes

925

b. Ia neka-in rumah sakit (?ng-adep nasi/bakal ma-preksa gigi). s/he af.come-loc house sick af-sell food/to mid-exam tooth ‘S/he visited the hospital (to sell food/to get his/her teeth examined).’ (114) a. Ia teka ka Bali jani. s/he pf.come to Bali today ‘S/he came to Bali today.’ b. *Ia neka-in Bali jani. s/he af-come-loc Bali today ‘S/he came to Bali today.’ (115) STEAL (maling): uli-marked source location a. Tiang nge-maling pipis uli guru=ne. I af-steal money from teacher=def ‘I stole money from the teacher.’ b. Tiang nge-maling-in guru=ne pipis. I af-steal-loc teacher=def money Lit. ‘I stole the teacher money.’ (116) GET (baan): uli-marked source location a. Anak luh ento maan buku ento uli guru=ne. person female that af.get book that (from teacher=def) ‘The girl got a book from the teacher.’ b. *Anak luh ento maan-in guru=ne buku. person female that af.get teacher=def book ‘The girl got a book from the teacher.’ Since ma-middles do not generally allow a definite object, they do not undergo the -in locative derivation, which creates a transitive structure. For example, while the lexical/underived middle tegak ‘sit’ can undergo -in derivation, the ma-derived middle verb ma-sila ‘sit cross-legged’ cannot. (117) SIT (tegak) a. Ia negak di kasur=e. s/he af.sit on bed=def ‘S/he sat on the bed.’ b. Ia negak-in kasur=e. s/he af.sit-loc bed=def ‘S/he sat on the bed.’

926

Masayoshi Shibatani and Ketut Artawa

(118) SIT CROSS-LEGGED (ma-silla) a. Ia ma-sila di kasur=e. s/he mid-sit.cross-legged on bed=def ‘S/he sat cross-legged on the bed.’ b. *Ia ma-sila-in kasur=e. s/he mid-sit.cross-legged bed=def ‘S/he sat cross-legged on the bed.’ The -ang causativization is also restricted both semantically and morphologically. The morphological causative with -ang typically requires a causee bearing a Theme role construed as a Figure in the Figure-Ground framing of an event. This applies to both intransitive-based and transitive-based constructions, as below. (119) WHITE (putih) a. Tembok=e putih. wall=def white ‘The wall is white.’ b. Tiang mutih-ang tembok=e. I white-caus wall=def ‘I whitened the wall.’ (120) STAB (tusuk) a. Tiang nusuk anak=e ento aji temutik. I af.stab person=def that with knife ‘I stabbed the man with a knife.’ b. Taing nusuk-ang temutik ka baong anak=e ento. I af.stab-caus knife to neck person=def that Lit. ‘I caused the knife to stab to the neck of that person.’ Where the causee of an -ang causative is understood to be agentive, the causation involved is of the sociative type, where the causer is directly involved in and is totally in control of the activity denoted (Shibatani & Pardeshi 2002). In (121b) below, the interpretation is either that the causer also ran along with the girl holding her hand and assisting her or that the causer carried the girl. (122b) expresses a situation where the causer brought the causee’s face closer to the younger sibling so that s/he may be able to kiss him/her (Arka 2003: 187). “Normal” indirect causation with an agentive causee calls for a periphrastic construction with the causative verb derived from the verb ngae ‘make’, as in (122c). (121) RUN (palaib) a. Anak=e luh ento malaib. person=def female that af.run ‘The girl ran.’

Balinese valency classes

927

b. Tiang malaib-ang anak=e luh ento. I af.run-caus person=def female that ‘I made the girl run / I carried the girl and ran / I ran carrying the girl.’ (122) KISS (diman) a. Ia niman adi=ne. s/he af.kiss younger.sibling=3.poss ‘S/he kissed his/her younger sibling.’ b. Tiang niman-ang ia sig adi=ne. I af.kiss-caus s/he to younger.sibling=3.poss ‘I made him/her kiss his/her younger sibling. ’ (Arka 2003: 187) c. Tiang ngae ia niman adi=ne. I af.caus s/he af.kiss younger.sibling=def ‘I made him/her kiss his/her younger sibling.’ The -ang derivation is restricted morphologically as well, such that it cannot apply to an -in derived transitive structure even if the valency structure itself satisfies the condition for it. (123) PUT (ejang) a. Basic causative pattern: FIG=OBJ (> GR=OBL) Anak=e ento ng-ejang buku=ne (di meja=ne). person=def that af-put book=def on table=def ‘The man put (down) the book (on the table).’ b. -in derived locative pattern: GR=OBJ > FIG=OBJ/OBL Anak=e ento ng-ejang-in meja=ne (aji) buku. person=def that af.put-loc table=def (with) book Lit. ‘The man put the table with the book.’ c. -ang derived pattern: FIG=OBJ > GR=OBL *Anak=e ento ng-ejang-in-ang buku ka/di meja=ne. person=def that af-put-loc-caus book to/on table=def ‘The man put the book on the table.’ / Lit. ‘The man caused the book to put to/on the table.’ The -ang derived form in (123c) may strike one as superfluous (and accordingly ungrammatical) since that entails a similar alignment pattern with the basic form of the verb PUT (ejang), as in (123a). But the -ang derivation typically creates a valency pattern in which an OBL Ground is obligatory. -ang derivation here is thus functional, as the comparison between (123a) and (123c) indicates (cf. earlier exam-

928

Masayoshi Shibatani and Ketut Artawa

ples (8) and (11)). The ungrammaticality of (123c) is again due to the ban against tampering with the valency pattern fixed by a previous derivational process. The -ang causative derivation cannot apply to -ang derived causatives either. (124) BREAK (belah-ang) (GR=OBJ (> FIG=OBL)) a. Tiang melah-ang jendela aji tungked=e. I af.break-caus window with stick=def ‘I broke the window with a stick.’ (FIG=OBJ > GR=OBL) b. *Tiang melah-ang-ang tungked=e ka jendela.23 I af.break-caus-caus stick=def to window ‘I broke the window with a stick.’ / Lit. ‘I caused the stick to break the window.’ Compare the above with the following pattern, where the FIG=OBJ (> FIG=OBL) argument pattern by an underived basic verb is shown to undergo -ang causativization. (125) TOUCH (tundik) a. Anak=e ento nundik lelipi=ne (aji tungked=e). person=def that af.touch snake=def with stick=def ‘The man touched the snake with the stick.’ b. Anak=e ento nundik-ang tungked=e ka lelipi=ne. person=def that af.touch-caus stick=def to snake=def ‘The man touched the snake with the stick.’ / Lit. ‘The man caused the stick to touch the snake.’

9 Summary and theoretical implications A major Balinese valency phenomenon revolves around argument coding in Object position. Non-agentive, secondary event participants are divided into two groups according to the Figure/Ground distinction. Instrument, Theme, Causee-Theme, and Beneficiary-Theme are grouped as Figure, while locations of various types including a human recipient as well as the contact point of a surface-contact action are grouped as Ground. It is the alternation between the Figure=Object and the

23 The ungrammaticality of this sentence may be due to the possibility that the derived verb melahang ‘break’ does not count as a surface contact verb.

Balinese valency classes

929

Ground=Object alignment that largely determines the valency patterns underlying the classification of transitive and ditransitive verbs. Except for the verbs GIVE (baang) and FILL (isinin), the alternation patterns in Balinese are morphologically coded in the verb, such that the locative -in suffix aligns a Ground expression with the Object, and the causative -ang suffix aligns a Figure expression with the Object, effecting the two alternative alignment patterns corresponding to John loaded the truck with hay (GR=OBJ pattern) and John loaded the hay onto the truck (FIG=OBJ pattern). When the basic (underived) verb is Ground-oriented and countenances the GR=OBJ pattern as a non-coded alternant, the -ang derivation typically supplies the FIG=OBJ pattern as a coded alternant. Surface-contact verbs CUT (godot), TOUCH (tundik), HIT/BEAT(lempag), and COVER (rurub) are all Ground-oriented. On the other hand, caused-motion verbs such as PUT (ejang), SEND (kirim), TAKE (jemak), STEAL (paling), etc. are Figure-oriented, and the basic verb form shows the FIG=OBJ pattern, which alternates with the GR=OBJ pattern through -in derivation. Fixing verbs, e.g., TIE (tegul), NAIL (pacek), SMEAR (uap), have neutral orientation and allow both GR=OBJ and FIG=OBJ patterns without derivation. But unlike GIVE (baang) and FILL (isinin), these verbs also allow coded valency patterns through -in and -ang derivations. The two versions of the GR=OBJ construction, one uncoded (e.g., “TIE (negul) the horse (with a rope)”) and the other, the -in coded pattern (e.g., “TIE (negul-in) the tree trunk with a horse”), are different in valency value in that, while in the former, the instrumental Figure expression is optional, both the Ground expression (tree trunk) and the Figure expression (horse) are obligatory in the latter. The same observation holds for the two versions of the FIG=OBJ construction. In the uncoded version (e.g., “TIE (negul) the horse (to the tree trunk)”), the Ground expression is optional, whereas in the -ang derived form (e.g., “TIE (negul-ang) the rope to the horse”), both the Figure expression (rope) and the Ground expression (horse) are obligatory. Similar valency-increasing properties of -in and -ang derivations are observed between underived intransitive constructions and the transitive counterparts resulting from these derivations. These observations about the -in locative applicative and the -ang causative are consistent with the general understanding about them as valency-increasing processes. Balinese, however, presents two cases where these processes cannot be characterized as valency-increasing. One is where a Ground expression is obligatory as in “SEND (ngirim) the books to the school”. The -in derived form aligns the Ground expression with the Object and yields a form like “SEND (ngirim-in) the school the book” without increasing valency. This represents the case of argument realignment without valency increase. More unusual is the existence of so-called precategorial roots that do not have a basic valency value and that need to be derived in order to function as a verb. These may undergo ma-middle, -in locative, or -ang causative derivation (as well as reduplication for some). The ma-middle derivation typically reduces valency. However, with precategorials, one cannot really speak of valency reduction for this

930

Masayoshi Shibatani and Ketut Artawa

or valency increase for -in and -ang derivations because precategorials by definition do not have a basic valency value. Yet, even with precategorials, ma-derivation derives various types of intransitive middle expressions, and -in/-ang derivations align a Ground expression with the Object and the Figure expression with the Object, respectively; e.g., ma-uruk ‘to learn’, nguruk-in ‘teach’ children (English), nguruk-ang ‘teach’ English (to children). What Balinese shows with these two cases where the -in locative applicative and the -ang causative do not increase valency is that the essential function of these processes is that of aligning a Ground or a Figure expression with the Object rather than what our customary characterization of them as valency-increasing processes suggests.24 The valency-increasing property associated with these processes is simply a consequence or a side effect of their fundamental function; when there is a Ground or Figure argument in the specified valency frame of the verb root, -in/-ang derivations align them with the Object, and if neither of them is countenanced by the verb root, they introduce them in Object position, regardless of whether it results in actual valency increase or not. On the basis of the following kind of pattern, causatives are usually understood as introducing a new agent/Subject. (126) a. Lelipi=ne mati. nake=def dead ‘The snake is dead.’ b. Anak=e ento nge-mati-ang lelipi=ne. person=def that af-dead-caus snake=def ‘The man killed the snake.’ (127) a. Anak=e cenik ento menek ka gedebeg=e. person=def small that af.climb to cart=def ‘The child climbed onto the cart.’ b. Ia menek-ang anak=e cenik ento ka gedebeg=e. s/he af-caus person=def small that to cart=def ‘He made the child climb onto the cart.’ But we have argued above that the -ang causative derivation places a causee Figure in Object position as if it were a causee that was being introduced. Indeed, the pattern below suggests this.

24 Our characterization of the functions of applicativization and causativization must be generalized in order to accommodate those languages that may not have the grammatical relation Object, as in ergative-alignment languages.

Balinese valency classes

931

(128) a. Ia menek-in gedebeg=e somi. s/he af.load-loc cart=def hay ‘S/he loaded the cart with hay.’ b. Ia menek-ang somi=ne ka gedebeg=e. s/he af.load-caus hay=def to cart=def ‘S/he loaded the hay onto the cart.’ Moreover, (128b) has no corresponding non-causative intransitive like (127a); the following is not possible. (129)* Somi=ne menek ka gedebeg=e. hay=def af.climb to cart=def. Lit. ‘The hay climbed onto the cart.’ Actually the Subject of (127a), which bears the A/TH (Agent/Theme) double roles, and does not semantically correspond to the causee Object in (127b), which bears only the Theme role. It is only when there is a strict thematic correspondence between the Subject of an intransitive sentence and the causee Object of a causative expression, as in (126a) and (126b), that we can speak of introducing a causer in the formation of direct causatives of the type that Balinese -ang causatives represent. We have also suggested that the causative-instrumental applicative and the causative-benefactive polysemy, also observed in a fair number of other languages, should be analyzed as resulting historically from the direct causative construal of causative structures, where a parellism was drawn between the HAVE-CAUS and BUY-CAUS lexicalization patterns: (i-a) I give (HAVE-CAUS) him a book / (i-b) I buy (BUY-CAUS) him a book vs. (ii-a) I have-CAUS him a book / (ii-b) I buy-CAUS him a book. While the Balinese verb baang ‘give’ opts for the (i-a) lexicalization pattern, its benefactive applicatives are based on the reanalysis of the (ii-b) pattern effected by direct causative construal of the relevant structure; e.g., a direct causative situation of making someone buy a book can be brought about by buying a book and giving it to the person in question. There is an interesting debate going on regarding whether the kind of alternation shown by Balinese -in and -ang forms involves derivation at all. The problem is acute in a case like English, where the relevant alternation is not accompanied by morphological derivation (see Iwata (2008) on the controversy over the derivational vs. non-derivatonal accounts). Balinese presents a clear case where the direction of derivation is indicated morphologically. Apart from the precategorial roots, which show a bipolar derivation pattern, Figure-oriented verbs (caused-motion verbs) are morphologically unmarked for the FIG=OBJ alignment pattern, and they realize the alternate GR=OBJ pattern through -in derivation. Surface-contact Ground-oriented verbs, on the other hand, are unmarked in realizing the GR=OBJ

932

Masayoshi Shibatani and Ketut Artawa

pattern, while they must undergo -ang derivation in realizing the alternate FIG= OBJ pattern. In ascertaining the role of derivation in the alignment alternations of this type, it is necessary that we look into both cognitive and functional underpinnings of morphological derivation. Crosslinguistic investigations of the markedness patterns and examination of the usage patterns of alternating forms provide a useful avenue of inquiry into this problem.25 Balinese presents some clear evidence that morphological derivation matters. For example, in contrast to basic verb forms, those derived via ma-, -in, and -ang derivations generally do not undergo further derivations. It is noteworthy that derived forms that have the same alignment pattern as underived forms do not undergo (further) derivation, while the latter may undergo the relevant derivation, indicating that the grammatical status of derived forms is different from that of underived basic forms. The three central derivations effected by the ma-middle prefix, the -in locative applicative, and the -ang causative are best considered as lexical derivational processes, which create new verbs. These are highly productive processes, but they on occasion create verbs whose meanings are irregular and are unpredictable from the forms themselves. ma-middle formation and a-passivization both typically reduce valency, and both involve morphology in Balinese. Yet, they differ in that the former is more severely constrained by semantic and morphological restrictions than the latter. Though ma-forms as a whole constitute a middle-voice domain, they vary considerably in the specific meanings they express (see the summary table below). This differs significantly from passivization, which does not create new verbs with different referential meanings. While Balinese ma-middle formation and -a passivization can be both considered voice phenomena, they differ in a way similar to the way lexical aspect and grammatical aspect do. The summary table in the appendix includes information on the basic valency patterns for representative verbs and their derivational possibilities for alternate valency patterns. The table does not indicate the Actor-focus/Patient-focus and the active/passive alternation patterns since they are generally predictable; those forms with a definite Object typically show alternations over these oppositions.

25 See Shibatani (forthcoming).

Meaning label

RAIN

BLINK

COUGH

RUN

JUMP

SING

LAUGH

SCREAM

FEEL COLD

DIE

PLAY

BE SAD

BE HUNGRY

ROLL

#

69

46

47

49

52

53

57

58

60

61

62

63

64

65

ngelilik

seduk

sebet

macanada

mati

dingin

makraik

kedek

magending

makejog

malaib

makokohan

ngedipang mata

ujan

Verb form

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

V

Coding frame schema

Appendix: Summary table

–

–

–

–

–

–

–

–

Causative with instrument causee

+

–

–

–

+

+

–

–

–

m

–

–

–

–

Causative with Theme causee

–

–

–

–

–

–

–

+

–

+

–

–

–

Locative alternation (argument adding)

–

–

–

–

–

–

–

–

–

–

Locative alternation (argument rearanging)

–

–

–

–

–

–

–

–

–

–

Locative alternation (oblique to object)

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Passive -a alternation

–

–

–

–

–

–

–

–

–

–

–

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Passive ka- Resultative alternation ma- alternation

Balinese valency classes

933

Meaning label

SINK

BURN

BE DRY

BE A HUNTER

EAT

HUG

LOOK AT

SEE

SMELL

FEAR

FRIGHTEN

LIKE

KNOW

THINK

SEARCH FOR

WASH

DRESS

SHAVE

#

66

67

68

70

 1

 2

 3

 4

 5

 6

 7

 8

 9

10

11

12

13

14

nguris

majunin

ngayehang

ngalihin

ngenehang

nawang

nemenin

nakutin

nakutin

ngadekin

nepukin

nlektekang

ngelut

ngajeng

tukangboros

tuh

puun

nyilem

Verb form

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V

1V

1V

1V

Coding frame schema

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Causative with instrument causee

–

–

–

–

–

–

–

–

–

–

–

–

+

–

–

+

+

+

Causative with Theme causee

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Locative alternation (argument adding)

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Locative alternation (argument rearanging)

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Locative alternation (oblique to object)

+

+

+

+

+

+

+

+

+

+

+

+

+

+

–

–

–

–

Passive -a alternation

+

+

+

+

+

+

+

+

+

+

+

+

+

+

–

–

–

–

+

+

+

–

–

–

–

–

–

–

–

–

+

+

–

–

–

–

Passive ka- Resultative alternation ma- alternation

934 Masayoshi Shibatani and Ketut Artawa

HELP

FOLLOW

SHOUT AT

PEEL

CLIMB

LEAVE

FEEL PAIN

ASK FOR

TELL

NAME

BUILD

BREAK

KILL

BEAT

HIT

TOUCH

CUT

COVER

FILL

15

16

20

33

48

55

59

19

21

23

24

25

26

27

28

29

30

43

44

ngisinin

ngerurub

ngodot

nundik

nglempag

nglempagin

ngematiang

melahang

mangunang

ngelunin

ngorahin

nagihin

nyakitang

ngalahin

menekin

melutin

ngerakin

ngikutin

nulungin

1 V 2 aji+3

1 V 2 aji+3

1 V 2 aji+3

1 V 2 aji+3

1 V 2 aji+3

1 V 2 aji+3

1 V 2 aji+3

1 V 2 aji+3

1 V 2 aji+3

1V23

1V23

1V23

1V2

1V2

1V2

1V2

1V2

1V2

1V2

+

+

+

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

+

+

+

+

+

+

+

+

+

+

+

+

–

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

–

–

–

–

–

–

–

–

–

–

+

–

–

–

Balinese valency classes

935

Meaning label

TIE

PUT

SHOW

GIVE

SEND

CARRY

THROW

POUR

LOAD

TAKE

TEAR

HIDE

TALK

SIT

SIT DOWN

LIVE

#

40

41

35

36

37

38

39

42

45

31

32

34

18

50

51

56

1 V 2 uli+3

1 V 2 uli+3

1 V 2 ke+3

1 V 2 ke+3

1 V 2 ke+3

1 V 2 ke+3

1 V 2 ke+3

1 V 2 ke+3

1 V 2 ke+3

1 V 2 di+3

1 V 2 aji+3 ke+4

Coding frame schema

nongos

negak

negak

ngomongin

1 V di+2

1 V di+2

1 V di+2

1V2 unduk+3

ngengkebang 1 V 2 uli+3

nguek

nyemak

menekang

nuruhang

ngentungang

ngaba

ngirimang

maang

ngedengang

ngejang

negul

Verb form

–

–

–

–

–

–

–

–

–

–

–

–

+

Causative with instrument causee

+

+

+

–

–

–

–

–

–

–

–

–

–

–

–

Causative with Theme causee

–

–

–

Locative alternation (argument adding)

–

–

+

+

+

–

Locative alternation (argument rearanging)

+

+

+

–

+

–

+

–

+

+

+

+

Locative alternation (oblique to object)

–

–

–

+

+

+

+

+

+

+

+

+

+

+

+

+

Passive -a alternation

–

–

–

+

+

+

+

+

+

+

+

+

+

+

+

+

–

–

–

–

+

+

+

–

+

+

+

+

–

+

+

+

Passive ka- Resultative alternation ma- alternation

936 Masayoshi Shibatani and Ketut Artawa

MEET

SAY

17

22

ngorahang

matemu

luas

1 V UTT2 teken+3

1 V teken+2

1 V ke+2

–

–

–

–

–

–

Legend: + occurs regularly; m occurs marginally; − occurs never; _ no data

GO

54

–

–

–

–

–

–

–

–

–

+

–

–

+

–

–

–

–

–

Balinese valency classes

937

938

Masayoshi Shibatani and Ketut Artawa

Abbreviations AF ANG FIG GR IN MID PF

actor focus -ang marked derivation figure ground -in marked derivation middle patient focus

References Adelaar, Alexander. 2005. Malayo-Sumbawan. Oceanic Linguistics 22. 357–388. Artawa, Ketut. 1994. Ergativity and Balinese syntax. Doctoral thesis, La Trobe University. Arka, I Wayan. 2003. Balinese Morphosyntax: A Lexical-Functional Approach (Pacific Linguistics 547). Canberra: Australian National University. Blust, Robert. 2002. Notes on the history of “focus” in Austronesian languages. In Fay Wouk & Malcolm D. Ross (eds.), The History and Typology of Western Austronesian Voice Systems, 63–78. Canberra: Pacific Linguistics, Research School of Pacific and Asian Studies, Australian National University. Croft, William. 1991. Syntactic Categories and Grammatical Relations: The Cognitive Organization of Information. Chicago: University of Chicago Press. DeLancey, Scott. 2001. Lectures on Functional Syntax. 2001 LSA Summer Institute, UC Santa Barbara. Iwata, Seizi. 2008. Locative Alternation: A Lexical-Constructional Approach. Amsterdam: John Benjamins. Ross, Malcom. 2009. Proto Austronesian verbal morphology: a reappraisal. In Alexander Adelaar & Andrew Pawley (eds.), Austronesian Historical Linguistics and Culture History: a Festschrift for Robert Blust, 295–326. Canberra: Pacific Linguistics, Research School of Pacific and Asian Studies, Australian National University. Shibatani, Masayoshi. 1988. Voice in Philippine languages. In M. Shibatani (ed.), Passive and Voice, 85–142. Amsterdam: John Benjamins. Shibatani, Masayoshi. 1996. Applicatives and benefactives: A cognitive account. In M. Shibatani & S. A. Thompson (eds.), Grammatical Constructions: Their Form and Meaning, 157–194. Oxford: Oxford University Press. Shibatani, Masayoshi. forthcoming. The role of morphology in valency alternation phenomena. In T. Kageyama & W. M. Jacobsen (eds.), Transitivity and Valency Alternations: Studies on Japanese and Beyond. Berlin: De Gruyter Mouton. Shibatani, Masayoshi & Prashant Pardeshi. 2002. The causative continuum. In M. Shibatani (ed.), The Grammar of Causation and Interpersonal Manipulation, 85–126. Amsterdam: John Benjamins. Shibatani, Masayoshi & Ketut Artawa. 2007. The middle voice in Balinese. In S. Iwasaki, A. Simpson, K. Adams & P. Sidwell (eds.), SEALS XIII: Papers from the 13th Meeting of the Southeast Asian Linguistics Society, 241–263. Pacific Linguistics, Canberra: Australian National University. Sneddon, James. 1996. Indonesian: A Comprehensive Grammar. London: Routledge. Starosta, Stanley, Andrew K. Pawley & Lawrence A. Reid. 1982. The evolution of focus in Austronesian. In Amran Halim, Lois Carrington & S. A. Wurm (eds.), Papers from the Third

Balinese valency classes

939

International Conference on Austronesian Linguistics, Vol. 2: Tracking the Travellers, 145–170. Canberra: Australian National University. Talmy, Leonard. 2000. Toward a Cognitive Semantics. Cambridge, Mass.: MIT Press.

Thomas Conners, John Bowden and David Gil

23 Valency classes in Jakarta Indonesian 1 Introduction Jakarta Indonesian is an Austronesian language belonging to the Malayic subgroup. It is the general colloquial language used in Jakarta, the capital and largest city of Indonesia, by well over 10 million people, in most everyday contexts, for inter-ethnic and increasingly also intra-ethnic communication. Jakarta Indonesian has been described in a reference grammar by Sneddon (2006), in articles by Wouk (1989, 1999) and Sneddon (2003), and more recently in publications associated with the Max Planck Institute for Evolutionary Anthropology Jakarta Field Station, including Cole & Hermon (2005), Cole et al. (2005, 2006), Tjung (2006), Gil (2006b, 2010) and Hidayat (2010, 2011). Within the city itself, Jakarta Indonesian is distinct from albeit very similar to Betawi Malay, the now endangered native dialect of the city’s indigenous ethnic community, described by Kähler (1966), Abdul Chaer (1976), Ikranagara (1980), Muhadjir (1981), Grijns (1991) and others. Jakarta Indonesian is much more different from Standard Indonesian, used in more formal contexts, and familiar to many general linguists from an extensive literature. However, as is often the case in situations of diglossia, there exists a continuum of language varieties between Jakarta Indonesian and Betawi Malay, and between Jakarta Indonesian and Standard Indonesian. Further afield, Jakarta Indonesian is one of a number of varieties of Malay and Indonesian used as regional contact varieties throughout the archipelago, among which are Riau Indonesian, as described in Gil (2001a, b, 2002, 2004, 2005b, 2006a, 2013, to appear), Manado Malay, as in Stoel (2005), and Ambonese Malay, as in van Minde (1997). However, Jakarta Indonesian is itself gaining in currency throughout Indonesia as an informal lingua franca, especially amongst younger and more upwardly mobile populations, where it is used as a mesolectal variety alongside and in competition with the more basilectal regional varieties of Malay/Indonesian such as those mentioned above. (For this reason perhaps, Jakarta Indonesian is sometimes referred to simply as “colloquial Indonesian”.) In general, Jakarta Indonesian shares a common typological profile with most other regional contact varieties of Indonesian, though differences in grammar, phonology and lexicon often lead to reduced mutual intelligibility.1

1 In contrast, Jakarta Indonesian and other varieties of Malay/Indonesian spoken in insular Southeast Asia are typologically very distinct from the genealogically closely related Sri Lanka Malay discussed in Nordhoff (this volume).

942

Thomas Conners, John Bowden and David Gil

2 Basics of Jakarta Indonesian morphosyntax Jakarta Indonesian morphosyntax may be insightfully described not from the vantage point of Standard Average European, as is tacitly presupposed by many current theoretical approaches, nor from the frame of reference of Standard Indonesian, as is explicitly assumed by Sneddon (2006) and others, but rather from the perspective of Isolating-Monocategorial-Associational (IMA) Language, an idealized language type introduced in Gil (2005a), and defined as follows: Isolating-Monocategorial-Associational Language a) Morphologically Isolating No word-internal morphological structure; b) Syntactically Monocategorial No distinct syntactic categories; c) Semantically Associational No distinct construction-specific rules of semantic interpretation (instead, compositional semantics is exclusively associational). The three defining properties of IMA Language pertain to three different linguistic domains, morphology, syntax and semantics, representing the limiting points of maximal simplicity within each of these domains. The first property, morphologically isolating, asserts that in IMA language there is no word-internal structure: every word contains exactly one morpheme, and every morpheme constitutes a word. The second defining property, syntactically monocategorial, says that in IMA language, all words belong to the same syntactic category: there is no distinction between categories such as noun, verb, adjective and preposition, or between lexical categories and their phrasal projections. The third defining property, semantically associational, pertains to the compositional semantics, the ways in which the meanings of complex expressions are derived from the meanings of their constituent members; what it says is that when a set of expressions E1 ... En, with meanings M1 ... Mn respectively, are brought together, the meaning of the whole is anything that is associated in some way with the meanings of the individual expressions – using A to denote the Association Operator, the resulting meaning may thus be represented as A (M1 ... Mn). While no naturally occurring language meets the definition of IMA Language in full, some languages come closer to the idealized IMA Language type than others. As argued in Gil (2005a), the closely related Riau dialect of Indonesian is one language that comes close to the ideal type, thereby justifying its appellation as a Relative IMA Language. In fact, many of the arguments cited for Riau Indonesian carry over also to the Jakarta dialect of Indonesian, which may thus also be characterized as a Relative IMA Language, if perhaps less so than Riau. Our description of Jakarta Indonesian thus proceeds in two stages: first showing how in

Valency classes in Jakarta Indonesian

943

broad outline it exhibits the patterns of IMA Language, then focusing in on those respects in which it goes beyond the ideal IMA Language type. Jakarta Indonesian is a highly isolating language; along with other colloquial varieties of Malay/Indonesian, it is perhaps among the world’s most extreme exemplars of the isolating type. Case marking and agreement morphology are completely absent, while tense-aspect-mood morphology is limited to subtle aspectual effects associated with the prenasalization marker N- discussed in Section 5 below. In contrast to the more well-known isolating languages of Mainland Southeast Asia, compounding is also relatively infrequent. Thus, a typical naturalistically occurring utterance might consist entirely of monomorphemic words. There are, however, a small number of other affixes, some of which occur quite frequently; a few of these are discussed in Section 5 below. Two other relatively common morphological processes are reduplication and truncation.2 Jakarta Indonesian is for the most part monocategorial: with just a very few exceptions, all words and larger expressions exhibit similar syntactic behavior and may accordingly be said to belong to a single open syntactic category. All members of this category exhibit similar distributional properties; accordingly, any combination of such words and larger expressions is grammatical (albeit not necessarily semantically felicitous). Morphologically, words differ with respect to the derivational affixes that they may take: whether a word can take a particular affix is partly arbitrary and partly based on semantic criteria. However, no two affixes select the same (or even roughly the same) class of host words; there are thus no morphological word classes either. Exceptions to the monocategoriality of Jakarta Indonesian are provided by a few dozen “functional” items whose grammatical properties are more highly constrained, and are accordingly assigned to a distinct closed syntactic category. However, the role of such items in the overall grammatical organization of the language is relatively minor, compared to most other languages. Jakarta Indonesian is characterized by a compositional semantics that is based largely on associationality, the principle whereby, when two or more expressions are juxtaposed, the meaning of the whole is anything that has to do with the meanings of the juxtaposed expressions. Associationality is universally present in the compositional semantics of all languages; where languages differ is in the extent to which associationality is supplemented by additional semantic rules making reference to specific morphosyntactic configurations involving properties such as word order, case marking, agreement, and various other functional items. In Jakarta Indonesian, the role of such additional semantic rules is substantially more limited than in many other languages, thereby supporting its characterization as a highly associational language.

2 In the interlinear glosses that follow below, reduplication and truncation are glossed as distr and fam respectively, reflecting their prototypical functions.

944

Thomas Conners, John Bowden and David Gil

To get a feel for how the IMA structure of Jakarta Indonesian works in practice, consider the following three typical naturalistic utterances:3 (1) Hape gembira. mobile.phone happy ‘If you have a mobile phone you’re happy.’ [speaker talking enviously about his friends who had recently acquired mobile phones] A ( mobile phone, happy ) (DGD) (2) Tembak ma bencong. shoot prt transgendered.person ‘Shooting is for sissies.’ [speaker talking dismissively about his friends who had their ears pierced by a “shooting” method] A ( shoot, sissy ) (DGD) (3) O, dia ujan ni, Timo. Oh 3 rain dem.prox Timo ‘Oh, he’s getting rained on, Timo.’ [adult and child playing with colouring books; adult talking about Mickey Mouse, who, in one of the pictures, is walking in the rain] A ( he, rain ) (JFS 613025161552120601) Each of the above three utterances consists of a core of two words, underlined, plus, in the latter two examples, one or more additional items which have no bearing on the grammatical structure of the utterances and hence may be ignored. The isolating nature of Jakarta Indonesian is evidenced in the fact that the underlined words (as well as, for that matter, all of the other forms occurring in these examples) are monomorphemic. The monocategorial character of Jakarta Indonesian is

3 Most of the data cited in this paper are presented in 5 lines: (i) a transcription in the semiconventionalized orthography of Jakarta Indonesian; (ii) an interlinear gloss; (iii) the context in which the utterance occurred, enclosed in square brackets; (iv) a translation into English; and (v) the source of the data, also enclosed in square brackets. Most of the data is taken from the Jakarta Field Station corpora, and is accordingly marked with “JFS” followed by the unique ID number of the utterance in the corpus database. A smaller number of examples are from the third author’s ongoing diary study and are marked with “DGD”. One example comes from the internet and is marked accordingly. In examples (1)–(3) below, an additional line inserted between the context and the English translation presents the logical formulas representing the meaning of the utterances in question.

Valency classes in Jakarta Indonesian

945

reflected in the fact that the underlined words belong to variegated semantic classes, such as thing, property and activity, which, in other languages (see, for example, their English glosses), would likely belong to different syntactic categories, such as noun, adjective and verb. As suggested, however, by the above examples, in a simple two-word sentence, each of the two positions can be occupied by a word belonging to any semantic class whatsoever. The associational semantics of Jakarta Indonesian is represented with logical formulas involving the association operator A, which may be read ‘Mobile phones are associated with happiness’, ‘Shooting is associated with sissies’, and ‘He is associated with rain’. Beneath these formulas, the more natural-sounding translations reflect the actual contexts of these utterances. Semantically, these English translations are much more specific than the formulas, introducing additional semantic relations such as conditionality in (1), habituality in (2), and affectedness in (3). However, these additional semantic relations are present only in the English translations, not in the Jakarta Indonesian source sentences, whose semantic representations are completely parallel, involving nothing more than the relationship of association, as represented in the logical formulas. As suggested by the above examples, the IMA structure of Jakarta Indonesian provides no grammatical mechanisms for the assignment of thematic roles. Still, words in Jakarta Indonesian have semantic frames similar to those of corresponding words in other languages, and very often, these semantic frames are satisfied by other expressions occurring in the same construction. Naturalistic examples (4)–(6) present three occurrences of two-word utterances in which the second word is bawa ‘carry’, whose semantic frame contains at least an agent and a patient.4 (4) Untung bawa. fortune carry ‘Fortunately you brought it.’ [Playing a game involving a tea set] A ( fortune, carry ) (JFS 223278083757270503) (5) Tante bawa. aunt carry ‘Yes I brought one.’ [Interlocutor asks speaker if she brought a gift] A ( aunt, carry ) (JFS 659236135415100502)5 4 In examples (4)–(6) and subsequently throughout this chapter, the word whose valency properties are at issue is underlined, while other words whose semantic relationships to the underlined word are under discussion in the text are indicated in boldface. 5 The reader may note that several of the JFS examples come from a child-language corpus and involve interactions between adults and young infants; however, all of the cited examples are from

946

Thomas Conners, John Bowden and David Gil

(6) Kompor bawa. stove carry ‘I brought a stove.’ [Discussing plans for cooking] A ( stove, carry ) (JFS 373687155731130607) Whereas in (4), untung ‘fortune’ is not understood as bearing either of these roles, in (5), tante ‘aunt’ is understood as the agent of bawa, while in (6), kompor ‘stove’ is understood as its patient. Thus, in (5) and (6), the semantic frame of bawa is satisfied by the word preceding it, albeit via different thematic roles. However, there is nothing in the structure of (5) and (6) to suggest the presence of differential grammatical processes of thematic role assignment involving agent and patient respectively. Rather, (5) and (6) share with each other, as well as with (4), an identical grammatical structure, as well as identical compositional semantics, involving a relationship of association between bawa and the word preceding it. As in (1)– (3) previously, the meaning of these sentences is most appropriately represented in terms of the association operator, as suggested by the logical formulas provided. Thus, example (5) says ‘I am associated with a carrying’, while (6) says ‘A stove is associated with a carrying’, any further inference of thematic roles being left to the pragmatics.

3 Valency patterns As a mostly monocategorial language, Jakarta Indonesian has an exceedingly simple system of valency patterns. With the exception of the small closed class of functional items, all words in Jakarta Indonesian exhibit the same valency patterns and therefore constitute a single valency class. This is true not only of words expressing the 70 meanings specified by the Leipzig Valency Classes Project, most of which correspond in other languages to verbs, but also of words expressing other meanings, of the kinds which in other languages are typically encoded by adjectives or nouns. To illustrate this, we shall contrast the behavior of four example words, satpam ‘security guard’, gede ‘big’, tidur ‘sleep’, and beli ‘buy’, all typical members of the single open lexical class of Jakarta Indonesian. These four words represent four

the adults’ speech, and there is no reason to believe that these utterances exhibit grammatical properties that are peculiar to a specific child-directed “motherese” register. In particular, the expression of first or second person reference by means of kinship terms (eg. tante ‘aunt’ as in (5) above) or names (eg. Ica as in (16a) below), although particularly common in child-directed speech, is not limited to such speech, and occurs also in other registers of adult-to-adult conversation.

Valency classes in Jakarta Indonesian

947

different semantic categories, thing, property, monovalent activity and bivalent activity; accordingly, in many other languages, the counterparts of these four words belong to different grammatical categories: noun, adjective, intransitive verb, and transitive verb respectively. However, in Jakarta Indonesian all four words exhibit identical valency patterns. This is illustrated in the (artificially constructed) paradigms in (7)–(9) below: (7) a. Satpam security.guard ‘He is a security guard.’ security guard b. Gede big ‘He is big.’ big c. Tidur sleep ‘He is sleeping.’ sleep d. Beli buy ‘He bought it.’ buy (8) a. Rahmat satpam. Rahmat security.guard ‘Rahmat is a security guard.’ A ( Rahmat, security guard ) b. Rahmat gede. Rahmat big ‘Rahmat is big.’ A ( Rahmat, big ) c. Rahmat tidur. Rahmat sleep ‘Rahmat is sleeping.’ A ( Rahmat, sleep ) d. Rahmat beli. Rahmat buy ‘Rahmat bought it.’ A ( Rahmat, buy )

948

Thomas Conners, John Bowden and David Gil

(9) a. Satpam rumah. security.guard house ‘He is a security guard of a house.’ A ( security guard, house ) b. Gede rumah. big house ‘He is as big as a house.’ A ( big, house ) c. Tidur rumah. sleep house ‘He is sleeping in a house.’ A ( sleep, house ) d. Beli rumah. buy house ‘He bought a house.’ A ( buy, house ) In examples (7)–(9) above, the meaning of each sentence is provided first in terms of a logical formula representing, where appropriate, the general associational semantics, and then in terms of a more specific English translation corresponding to one of the most likely contexts in which the sentence might be uttered, though other quite different translations might also be available. Example (7) shows that all four words may occur as independent sentences without any additional words or grammatical markers present. Whereas (7b–d) are typical of so-called “pro-drop” languages, (7a) is somewhat more exceptional from a cross-linguistic perspective, since many “pro-drop” languages exhibit a noun/ verb asymmetry whereby if the predicate is nominal, an overt subject is required.6 Examples (8) and (9) show that all four words exhibit similar behavior with regard to their ability to co-occur with an additional expression, without any further grammatical markings. Whereas in (8) the additional expression, Rahmat ‘Rahmat’, occurs before the word in question, in (9) the additional expression rumah ‘house’ occurs after it; all of the combinations shown are grammatical. In particular, examples (9c) and (9d) reflect the absence of any structural distinction between words constituting the translational equivalents of intransitive and transitive verbs in other languages. Of course, words such as tidur ‘sleep’ and beli ‘buy’ differ with regard to their semantic frames in much the same way as their English glosses; accordingly, rumah in (9d) has the option of being understood as the pa-

6 For some discussion of this generalization in the context of Malay/Indonesian and other languages of the region, see Gil (2003: 497–498).

Valency classes in Jakarta Indonesian

949

tient of beli, as per the chosen English translation, whereas the corresponding option is not readily available with tidur in (9c). However, there is no difference in the grammatical structures of (9c) and (9d), nor is there any difference in their semantic representations: just as (9c) denotes a sleeping associated with a house, so (9d) refers to a buying associated with a house. Thus, examples (7)–(9) above show that satpam, gede, tidur and beli exhibit identical valency patterns. In general, all words belonging to the single open syntactic category of Jakarta Indonesian, that is to say, almost all words in the language, exhibit identical valency properties. Stated simply: there are no valency classes in Jakarta Indonesian. A description of Jakarta Indonesian concerned exclusively with distinguishing the possible from the impossible in the realm of valency classes would stop here. All combinations of expressions belonging to the single open syntactic category are grammatical, and their compositional semantics is as provided for by the Association Operator – end of story. However, there is another story to be told, one that is concerned with distinguishing not just the possible from the impossible, but also the common from the rare, and the more frequent from the less frequent. It is this additional story that constitutes the focus of the remainder of this chapter and of the accompanying electronic database.7 All languages distinguish between constructions that are more common and others that are less. One way of representing such differential frequency patterns is by means of preference rules of the kind proposed in another cognitive domain, tonal music, by Lerdahl & Jackendoff (1983). Some proposals for preference rules governing the grammatical patterns of Riau Indonesian are provided in Gil (2005b: 258–260, 2012); these preference rules are equally applicable to the Jakarta dialect of Indonesian, as further discussed below. One such rule proposed in Gil (2012), Semantic Frame Satisfaction, specifies a preference for interpretations in which, given two expressions in construction with one another, one of the expressions satisfies the semantic frame of the other one. This rule is in effect a weakened version of Chomsky’s (1981) “theta criterion”. For example, in the JFS Valency Corpus, the word beli ‘buy’ occurs in construction with a following unmarked expression (as in (9d) above) a total of 599 times: of these, the second expression satisfies the semantic frame of beli as agent, patient, benefactive or source in 566 cases (94 %), while failing to do so in 33 cases (6 %), most involving a locative expression. The effect of Semantic Frame Satisfaction is thus

7 The description of the differential frequency patterns of Jakarta Indonesian that follows is supported by a subcorpus of the Jakarta Field Station naturalistic corpus containing 23,040 word tokens that were coded for their valency properties; these word tokens were chosen to express the 70 meanings specified by the Leipzig Valency Classes Project plus a number of additional meanings of specific interest for Jakarta Indonesian. This subcorpus is referred to henceforth as the JFS Valency Corpus.

950

Thomas Conners, John Bowden and David Gil

evident in example (9) above. Although all four constructions in (9) are grammatical, not all occur with equal frequency. In particular, whereas simple expressions without further grammatical marking are very common after words such as beli, they are somewhat less common after words such as tidur, and considerably less common after words such as gede. That they do, nevertheless, occur, is evidenced by naturalistic examples such as (19a) in Section 4.2 below, corresponding to (9c), and the following example, corresponding to (9b): (10) Dulu ma asal ada perawan namanya segede juga, gede si before prt as.long.as exist virgin name:assoc one:big also big pers Kori, udah dipagerin. Kori prf g.pass:fence:g.appl ‘Before if a girl grew up to be that big, as big as Kori, they’d be secluded.’ [Discussing customs in the old times] (JFS 412497032528240209) However, the relative infrequency of examples such as (19a) and (10) is due to the Semantic Frame Satisfaction preference rule and the different semantic frames of the respective words; specifically, whereas the semantic frame of beli includes, inter alia, both agent and patient, that of tidur and gede includes only a theme. A complete description of Jakarta Indonesian must take cognizance of the fact that such constructions occur with less frequency than alternative constructions available in the same contexts to express similar meanings. In order to represent such frequency facts, reference can be made to valency preference classes, sets of lexical items which differ with respect to the ease and frequency with which they enter into constructions associated with various valency patterns. For example, words such as beli and tidur would belong to distinct valency preference classes by dint of the greater frequency of occurrence of a following grammatically unmarked expression such as rumah with beli, as in (9d), in contrast to tidur, as in (9c). The contents of the accompanying electronic database are, in general, most appropriately construed as making reference to such valency preference classes.

4 Order and flagging alternations Order and flagging alternations are the two major kinds of alternations involving the arguments of the words whose valency properties are under consideration.

4.1 Order alternations Linear order is flexible; expressions can occur in any order with respect to one another. For example, tidur ‘sleep’ may be either preceded or followed by its associated theme, as in (11):

Valency classes in Jakarta Indonesian

951

(11) a. Ais tidur. Ais sleep ‘Ais is sleeping.’ [Small talk] (JFS 820946082433020201) b. Oh Tidur Intan. Oh sleep Intan ‘Oh, Intan is sleeping.’ [Somebody asks where Intan is] (JFS 283860014242281007) Similarly, bawa ‘carry’ may be either preceded or followed by its associated agent, as in (12); patient, as in (13); or locative, as in (14): (12) a. Tante nggak bawa. aunt neg carry ‘I didn’t bring it.’ [Child asks speaker for something] (JFS 830645105327220501) b. Nggak bawa Tante. neg carry aunt ‘I didn’t bring one.’ [Child asks speaker whether she brought a pencil] (JFS 729646104010061101) (13) a. Kuartet bawa. quartet carry ‘I brought the quartet.’ [Speaker is asked if he brought the quartet (a card game), and answers] (JFS 523254120651180603) b. Bawa koper. carry briefcase ‘He’s carrying a suitcase.’ [Playing a lego game with a child] (JFS 826944102619050701) (14) a. Sekolah bawa payung? school carry umbrella ‘Do you take an umbrella to school.’ [Small talk] (JFS 374775141419120701)

952

Thomas Conners, John Bowden and David Gil

b. Kalo pegang-pegang, aku bawa kamar, lho. top distr~hold 1sg carry room excla ‘If you touch it I’ll take it into the other room.’ [Warning a child who is playing with an electric fan] (JFS 303211142018290503) Such flexibility is not limited to constructions with only one argument present, but is equally characteristic of constructions with two or more arguments expressed. For example, when occurring with both agent and patient expressions, bawa may be preceded by the agent and followed by the patient, as in (15a), or preceded by the patient and followed by the agent, as in (15b): (15) a. Tapi Om Okki nggak bawa duit-duitannya. but uncle Okki neg carry money:assoc ‘But I didn’t bring the toy money.’ [Playing with child] (JFS 751790145633060803) b. Piso-pisoan nggak bawa Om Okki. knife neg carry uncle Okki ‘I didn’t bring the toy knife.’ [Playing with child] (JFS 486192154806260603) In fact, all 6 logically possible orders are grammatical, as shown by the following example, involving suka ‘like’ and its experiencer and theme arguments: (16) a. Nggak suka Ica bukunya? neg like Ica book:assoc ‘Don’t you like the book?’ [Looking through picture book; adult asks child] (JFS 431431101132090501) b. Oh, suka kacang tupainya. excla like bean squirrel:assoc ‘Oh, the squirrels like peanuts.’ [Watching cartoons on TV] (JFS 792852083710261101) c. Omnya suka dangdut. uncle:assoc like dangdut ‘He likes dangdut.’ [Talking about a prospective visitor’s musical tastes (dangdut is a genre of Indonesian pop music)] (JFS 691423081100010199)

953

Valency classes in Jakarta Indonesian

d. Semua suka Mami, ya. all like mother yes ‘I like everything.’ [People discussing what kinds of food they like to eat] (JFS 584137094010290501) e. Kalo Indri ayam suka. top Indri chicken like ‘But Indri likes chicken.’ [People discussing what kinds of food they like to eat; interlocutor says she doesn’t like chicken; mother, talking about her child, says] (JFS 272105063705120901) f. Kalo Joshua Tante nggak suka. top Joshua aunt nrg like ‘I don’t like Joshua.’ [Discussing music (Joshua is a famous child singer)] (JFS 450047132340210901) Nevertheless, although word order is extremely flexible, not all word orders are of equal frequency. Table 1 shows the relative frequencies of the six logically-possible word orders for beli ‘buy’ and makan ‘eat’ with both agent (A) and patient (P) arguments present, in the JFS Valency Corpus: Tab. 1: Word Order Preferences in Jakarta Indonesian, for beli and makan. VAP

VPA

AVP

PVA

APV

PAV

beli ‘buy’ 145 tokens

1% 1

 7 % 10

 88 % 128

0% 0

0% 0

 4 %  6

makan ‘eat’ 355 tokens

0% 0

 2 %  8

 94 % 333

1% 2

>0% 1

 3 % 11

As evident from Table 1, AVP order is by far the most common, which is why Jakarta Indonesian is often thought of as exhibiting basic SVO word order. While VPA and PAV are the most common variants, the remaining three orders occur much more rarely. The relative frequencies of the various orders can be captured in terms of the preference rules proposed for closely related Riau Indonesian in Gil (2005b: 258– 260). As in most or all languages, principles of information flow make reference to notions such as topic and comment, effecting a preference for topic-first order, while permitting focus-first order as a more highly-marked variant. Semantic factors are also responsible for differential frequencies. Again as in most or all lan-

954

Thomas Conners, John Bowden and David Gil

guages, activities, especially “highly transitive” ones, are commonly conceptualized as following a metaphorical trajectory beginning with the agent, proceeding through the activity itself, before culminating in the patient or other associated participants. In many languages, this conceptualization is then iconically reflected in linear order, yielding a preference for agents to precede activities, and activities to precede patients and other participants. Finally, but this time language specifically, Jakarta Indonesian exhibits a preference for heads to precede modifiers, which in the case of clauses containing an activity word and associated arguments translates into a preference for the activity word to precede all of its associated arguments. In conjunction, these and additional principles account for the flexibility of word order and also the existence of preferences amongst available orders. For example, whereas an activity-precedes-patient order is consistent with both iconicity and headedness, an opposite patient-precedes-activity is inconsistent with these two principles, and accordingly less frequent. Still, patients do sometimes precede activities, and when they do, it is generally because these two principles have been overridden by principles of information flow, as is clearly the case in (13a) with its salient topic-initial character. It should be noted, however, that while some linear orders are more common than others, there is no evidence to suggest that the less frequent orders are derived in any sense (e.g. by “movement”, “scrambling”, or some other kind of unidirectional alternation) from their more frequent counterparts.

4.2 Flagging alternations So far, all of the examples considered have involved “bare” participants, that is to say, participants that do not contain any kind of morphosyntactic flagging expressing their relationship to the associated activity. However, Jakarta Indonesian does possess a variety of morphosyntactic devices whose primary function is that of flagging, among the most common of which are di, expressing locative-related notions; ke, a rather typical dative marker; dari ‘from’; and (s)ama, a macrofunctional marker whose basic meaning is comitative but also encompassing a range of other functions which can be subsumed under the general rubric of non-absolutive.8 Unlike most other words, which belong to a single open syntactic category, these four items belong to a closed class of items whose behavior is more restricted; in particular, they must always precede their hosts, never follow them. In fact, two of these items,

8 The representation of the form in question as (s)ama reflects its alternative pronunciations as sama or ama. Gil (2004) presents a detailed analysis of the corresponding form in Riau Indonesian, plus, in the concluding section, some discussion of minor differences with respect to its function in the Riau and Jakarta varieties of Indonesian.

Valency classes in Jakarta Indonesian

955

di and ke, are probably more appropriately analyzed as proclitics rather as separate words (but this is of no concern in the context of the present chapter). In keeping with the basic design of the language, as presented in Sections 2 and 3 above, these markers are always optional: there is no morphosyntactic environment in which any one of them is obligatory. Their role is simply to add semantic specificity to a construction whenever this is deemed necessary. In this respect, their behavior is more like that of attributive adverbs or adjectives than that of typical adpositions or case markers in other languages. Unlike verbs in many languages, which are often said to “govern” a particular adposition or case marker, in Jakarta Indonesian, a given activity word typically offers a choice between no flagging expression whatsoever and one or more flags, making possible the expression of occasionally fine shades of meaning. In examples (17)–(20) below, an (a) example without any flagging is contrasted with one or more subsequent examples illustrating various flagging possibilities. Example (17) shows that for kasi ‘give’, a preceding agent may occur either without any flagging or with the non-absolutive marker (s)ama: (17) a. He-eh, saya nggak kasi. uh-huh 1sg neg give ‘Uh-huh, I won’t let her.’ [About a friend who wanted to take a certain job] (JFS 784562142936300606) b. Tapi ama bapaknya nggak kasi. but with father:assoc neg give ‘But his father didn’t let him.’ [About a friend who wanted to take a certain job] (JFS 770444090205170707) Example (18) shows that for the same kasi, a following goal may also occur bare or with (s)ama; however, it also has the option of taking the dative marker ke: (18) a. Kasi Tante Bety. give aunt Bety ‘Give it to Bety.’ [Mother telling child to give friend a drink] (JFS 495302083525290801) b. Kasi dulu sama Tante. give before with aunt ‘Give it to her.’ [Telling child to give transmitter to experimenter, so she can then fasten it to child’s pants] (JFS 907183153320220601)

956

Thomas Conners, John Bowden and David Gil

c. Dah, kasi ke Om Okki. prf give to uncle Okki ‘Alright, give it to Okki.’ [Father telling child to give a puzzle to the experimenter] (JFS 724923114124050603) In conjunction, (17) and (18) show that for kasi ‘give’, neither word order nor marking with (s)ama serve to differentiate between agents and goals; only the dative marker ke possesses this particular function. Example (19) shows how for tidur ‘sleep’, a following locative may occur either bare, or with one of two different markers, locative di or dative ke: (19) a. Wah, pada selalu tidur rumah ni. excla pl always sleep house dem.prox ‘Wow, they always sleep at home, then.’ [Talking about some TV characters who go places with a camper van] (JFS 378758154328170901) b. Mulai ntar malem, tidur di rumah Ayu. beginning later night sleep in house Ayu ‘Starting from tonight, you sleep at Ayu’s house.’ [To child who was fighting] (JFS 340124105028230801) c. Tidur ke apotik. sleep to pharmacy ‘He slept at the pharmacy.’ [Explaining why her father had not come home since yesterday] (JFS 227294130017060802) The contrast between (19b), with di, and (19c), with ke, conveys a fine semantic distinction not readily expressible in English: whereas locative di provides the usual means for expressing location, dative ke contributes an additional shade of meaning, highlighting the location as being somewhere that the actor had to go to in order to be in. Finally, example (20) illustrates an embarras de richesses of possibilities for beli and a following source or locative, which may occur either bare, or with one of four different flags, di, ke, dari, or (s)ama: (20) a. Eh Rizka, Rizka beli apa di Pasar Minggu, beli apa Pasar excla Rizka Rizka buy what in Pasar Minggu buy what Pasar Minggu. Minggu

Valency classes in Jakarta Indonesian

957

‘Hey, Rizka, what did you buy at Pasar Minggu?’ [Mother asking child about her recent shopping trip to Pasar Minggu (Sunday Market)] (JFS 874342075524250903) b. Saya dulu ma punya duit lima puluh rebu udah segala 1sg before prt have money five ten thousand prf all macem kita beli ke pasar, ya. sort 1pl buy to market yes ‘Before, if I had fifty thousand, we could buy all kinds of things in the market, right.’ [Talking about how expensive things have become] (JFS 299165093032241108) dari sana. c. Kata dia capek-capek beli pala word 3 distr~tired buy nutmeg from there ‘He said he was tired buying nutmeg from there.’ [About shopping at the market] (JFS 626931142219150403) d. Tu Mbak War beli ama Nenek? dem.dist ma’am War buy with grandmother ‘So War bought it from you?’ [Talking about a plot of land] (JFS 923865071849290508) Example (20a) illustrates alternation between bare source and source marked with locative di within a single utterance, highlighting how difficult it is to define precise conditions on the alternation (other than the rather trivial observation that, once the locative relation is specified in the first part of the utterance, there is no need to repeat it in the second). Example (20b) with dative ke conveys a shade of meaning analogous to that in (19c), in the present case focusing on the activity of going to the market. Example (20c) with dari ‘from’ is arguably that which gives most explicit expression to the inherent semantic relationship between the activity of buying and its source participant, though it is perhaps the least frequently occurring of all the available options. Finally, example (20d) with nearly all-purpose (s)ama emphasizes the active participation of the source participant, here in human form, in the act of buying (or selling). Thus, examples (17)–(20) underscore the optional nature of flagging alternations in Jakarta Indonesian. A quantitative view of this optionality is provided in Table 2 below, showing the distribution of flagging across the arguments of a selected word, beli ‘buy’, in the JFS Valency Corpus:

958

Thomas Conners, John Bowden and David Gil

Tab. 2: Flagging Preferences in Jakarta Indonesian, for beli. Patient

Agent

Benefactive

Source

Locative

bare

100 % 630

100 % 366

22 %  2

  0 %   0

44 % 52

flagged

  *

  0 %   0

78 %  7

100 %  21

56 % 66

In Table 2, the columns represent the four main participants of beli, plus a fifth semantic role, locative (which is often hard to distinguish from source), while the rows represent the two options, bare vs. flagged. Whereas the grammar of Jakarta Indonesian does not offer any option for the flagging of patients, all other semantic roles may optionally be flagged, however, the frequency of such flagging varies across different semantic roles. In general, the weaker the syntactic and semantic nexus between the participant and its semantic role assigner, the more likely it is to be flagged. In Table 2, this tendency is evident in the contrast between agents, which are never flagged in the restricted sample of the JFS Valency Corpus (although such flagging, with (s)ama, is a grammatical option, as exemplified in (17b) above), and the more weakly bound benefactives, sources and locatives, which are all flagged with over 50 % frequency.9 Nevertheless, even though in many cases the presence of some kind of morphosyntactic flag is more frequent than its absence, from a formal point of view the bare option is the more basic one, with the presence of a flag conforming to the general principle whereby morphemes and words are added in order to achieve greater semantic specificity. In this section, we considered linear order and flagging as independent alternations, as is indeed the case. However, they clearly interact in myriad ways. To cite just one example, the non-absolutive marker (s)ama occurs much more frequently on agents when they follow the activity word than when they precede it – as is the case in (17b). However, a more detailed account of such interactions lies beyond the scope of this chapter.

5 Generalized-voice and other similar alternations This section deals with alternations that are encoded, for the most part morphologically, on the words whose valency properties are under consideration; such alter-

9 The flags used for arguments of beli were as follows: for benefactives, buat (all 7 tokens); for sources, di (11), (s)ama (8), ke (1), dari (1); and for locatives, di (59), ke (7).

Valency classes in Jakarta Indonesian

959

nations involve primarily the three markers N-,10 di- and -in, plus, secondarily, a handful of additional markers, including -an, ke-, ke--an, and kena.

5.1 Generalized voice alternations The forms N-, di- and -in 11 may be characterized as marking generalized voice, in accordance with the following definition from Gil (2002: 276): “A generalized voice marker is a marker M which, when applied to a form X, marks the argument of X bearing the thematic role T as having a set of properties P”. For example, a prototypical active voice marker in a language would be one which marks the argument of its host bearing the thematic role of agent as having a set of properties associated in that language with subjects: such properties might include nominative case marking, control of agreement, pre-verbal position, and others. The larger the set of properties associated with the argument in question, the stronger the generalized voice marker is said to be. However, in Jakarta Indonesian, the set of properties that N-, di- and -in associate with their respective arguments is significantly smaller; accordingly, they may be viewed as weak generalized voice markers. The markers N- and di- constitute a pair of agent-oriented and patient-oriented generalized voice markers respectively (see Gil (2002) for a detailed description of the corresponding markers in Riau Indonesian plus also a brief discussion of Jakarta Indonesian). The marker N-, when applied to a given word, simply asserts that the word in question has an agent in its argument structure: in terms of the above definition, the set of properties P associated with the agent consists just of the single property of existence. Analogously, the marker di-, when applied to a given word, asserts that that word has a patient in its argument structure. Accordingly, N- and di- may be characterized as generalized active and passive markers respectively; they differ from prototypical active and passive markers in other languages

10 The symbol N- (commonly referred to in speech as “prenasalization”) represents a morpheme whose realization depends on the phonological properties of the host word. For monosyllabic words (a small minority in Jakarta Indonesian) it assumes the form nge- (eg. cat > ngecat ‘paint’), while for polysyllabic words, its realization depends on the initial segment of the host word: for l and r it is nge- (eg. liat > ngeliat in (23)); for b, d, j and g either nge- (eg. buka > ngebuka in (22)) or, in a different subdialect, a homorganic nasal (eg. buka > mbuka); for p, t, c, k and s the initial consonant is replaced by a homorganic nasal (eg. senter > nyenter in (21)); for vowels it is ng- (eg. ambil > ngambil ‘take’), while for m, n, ny and ng there is but a single form of the word with no overt marker of prenasalization, this form occurring both in contexts where N- would be likely to occur and contexts where it would not. 11 Of these three markers, two, N- and -in, have cognates in Balinese, discussed in Shibatani & Artawa (this volume), with related but not identical functions. Historically, whereas Jakarta Indonesian and Balinese N- are probably both retentions from a common ancestral language, Jakarta Indonesian -in comes from the corresponding Betawi Malay form which was borrowed from Balinese.

960

Thomas Conners, John Bowden and David Gil

in that they make no reference to specific morphosyntactic properties associated with the agent and patient arguments. The words to which N- and di- may apply constitute two partially overlapping sets lacking any general systematic defining features; whether or not it can take Nand/or di- thus forms part of the grammatical information associated with each individual word. For example, whereas puter ‘revolve’ can take both generalized voice markers, pergi ‘go’ can take neither. The agent-oriented marker N- also exhibits a significant degree of arbitrariness with regard to the frequency with which it actually occurs on words to which it may attach. For example, whereas it applies readily to kopi ‘coffee’ to yield ngopi ‘drink coffee’, it attaches less commonly to teh ‘tea’ to yield ngeteh ‘drink tea’, even though Indonesians are no less fond of tea than they are of coffee.12 Because of such frequency effects, words marked with N- have the potential of undergoing reanalysis, with the initial nasal being reinterpreted as part of the stem. For example, the form tangis ‘cry’, familiar to many speakers of Jakarta Indonesian from the standard language, no longer occurs in Jakarta Indonesian, since it was marked so often with N- that the resulting form, nangis, became reanalyzed as a new stem, which, like other nasal-initial stems, cannot be marked with N-. Constructions containing N- or di- are of two types, depending on the basic semantic frame of the host word. In the first construction type, the host word, in its unmarked form, lacks the relevant thematic role in its semantic frame; in such cases, the effect of applying N- or di- is to enrich the semantic frame of the host word by adding the relevant thematic role, a process typically involving a semantic type shift, such as from thing to activity. Example (21) below illustrates this for the case of host word senter, whose basic meaning is ‘torch’ (or in American English ‘flashlight’): (21) a. Ini kan senter bagus, Timo. dem:dem.prox q torch good Timo ‘This is a good torch, Timo.’ [Mother suggesting to her son to use a torch] (JFS 989892173255021003) b. Oo cuma senter kepalanya Ica tuh. aunt only torch head:assoc Ica dem.dist ‘I’m only shining it on your head.’ [Playing with a torch] (JFS 351516185632260303)

12 Thus, in the complete Jakarta Indonesian corpus, N- applies to kopi with 4.2 % frequency (in 8 out of 191 utterances containing host word kopi), but to teh with just 0.5 % frequency (in 3 out of 593 utterances containing host word teh).

Valency classes in Jakarta Indonesian

961

c. Enggak, Oo ... Oo cuman nyenter ... neg aunt aunt only g.act:torch ‘No, I’m just shining it.’ [Playing with a torch] (JFS 167799185458260303) d. Ica ntar disenter lho. Ica fut g.pass:torch prt ‘They’re going to film you.’ [To a child about to be recorded] (JFS 314245080600010199) In (21a), senter occurs morphologically unmarked, and with its basic thing-denoting meaning. In (21b), senter is also morphologically unmarked; however, it has undergone semantic type-shifting, in order to denote an activity, ‘shine on with torch’; such zero-marked type-shifting is widespread in Jakarta Indonesian (see Gil 2013 for extensive discussion of the same phenomenon in Riau Indonesian). In the following two utterances, a similar type shifting is achieved by means of the generalized voice markers. In (21c), senter is marked with N-, the presence of the agent-oriented generalized voice marker being to introduce an agent into the semantic frame of senter, and, in so doing, convert its ontological category from thing to activity. Similarly, in (21d), senter is marked with di-, the patient-oriented generalized voice marker here introducing a patient into the semantic frame of senter, and, again, converting its meaning from thing to activity. As suggested by the above example, when the application of N- or di- induces a semantic type shift, some additional semantic feature of an arbitrary, unpredictable nature may be added; thus, nyenter and disenter mean ‘illuminate with torch’, and not, say, ‘ignite with torch’ (like the English verb torch), or ‘hit with torch’. In the second construction type, the host word, in its unmarked form, already contains the relevant thematic role in its semantic frame; in such cases, applying N- or di- does not alter the semantic frame of the host word, or result in semantic type-shifting. Rather, its effect is merely to emphasize and highlight the presence of a participant bearing the appropriate thematic role, generally for purposes of discourse coherence. Example (22) below illustrates this for the case of host word buka ‘open’: (22) a. Daud, Daud juga uda jarang buka. Daud, Daud also prf seldom open ‘Daud also seldom opens it any more.’ [Discussing Friendster] (JFS 606276082702230508)

962

Thomas Conners, John Bowden and David Gil

b. Dia ngebuka aura. 3 g.act:open aura ‘He can open the aura.’ [Discussing mysticism] (JFS 256297085410101008) c. Orang ma pada dibuka. person prt pl g.pass:open ‘People were opening them.’ [About people at railway station buying food bags and eating] (DGD) Whereas in (22a) buka occurs in bare form, in (22b) the agent-oriented generalized voice marker N- emphasizes the existence of an agent in the semantic frame of buka, while in (22c) the patient-oriented generalized voice marker di- highlights the presence of a patient in the same word’s semantic frame. As agent- and patient-oriented generalized voice markers, N- and di- resemble active and passive voice markers, but don’t quite make the cut. Comparing (21c) and (21d), it looks as though they could in fact be active and passive markers: while in (21c) Oo, preceding nyenter, is its agent, in (21d), Ica, preceding disenter, is its patient. Indeed, it is contrasts such as these which underlie the widespread characterization of these forms (and their cognates in other dialects of Malay/Indonesian) as expressing an active/passive contrast. However, whereas (21c) and (21d) represent the most frequent pattern in Jakarta Indonesian, alternative patterns are also in evidence. In actual fact, the presence of N- or di- does not rule out any assignment of thematic role to a preceding (or following) expression. Thus, for example, in (22b) and (22c), ngebuka and dibuka are both preceded by an expression bearing the thematic role of agent. Whereas constructions like (22b) are common, constructions such as (22c) are much less frequent, but nevertheless still possible. Paradigms similar to (22) are presented in (23)–(25) below: (23) a. Saya dulu bisa liat. 1sg before can see ‘Before, I could see them.’ [Talking about visiting their kids] (JFS 363772095322071004) b. Eh, si Mita ngeliat. excla pers Mita g.act:see ‘Mita happened to see it.’ [Small talk] (JFS 467738093826050607)

Valency classes in Jakarta Indonesian

963

c. Terus Ica diliat, punya nggak yang jumlahnya empat. continue Ica g.pass:see have neg rel total:assoc four ‘So see if there are any that add up to four.’ [Playing a card game] (JFS 402707141333260902) (24) a. Tangannya coba liat. hand:assoc try see ‘Let me see your hands!’ [Adult to infant] (JFS 804307101317290301) b. Poto putih kan udah ngeliat. photo white q prf g.act:see ‘She’s already seen the white photo.’ [Child asks to see white photo, mother says to friend] (JFS 596710140953200404) c. Om diliat pake itu. uncle g.pass:see use dem:dem.dist ‘Look at me using that.’ [asking child to look at him with a pair of binoculars] (JFS 153689122431270202) (25) a. Tutupnya kasi Oma. close:assoc give grandma ‘Give granny the lid.’ [Adult to child] (JFS 274323092313021000) b. Sisah ngasi saudara. remainder g.act:give relative ‘I gave the rest to my family.’ [About the money that he took on a trip] (DGD) c. Obatnya cuma dikasi buat lima hari. medicine:assoc only g.pass:give for five day ‘The medicine was only given for five days.’ [Talking about the medicine for her high blood-pressure] (JFS 899547064124080109) Example (23) shows liat ‘see’ preceded by its experiencer, (24) shows the same form liat ‘see’ preceded by its theme, and (25) shows kasi ‘give’ preceded by its patient;

964

Thomas Conners, John Bowden and David Gil

Tab. 3: Generalized Voice Markers N- and di- in Jakarta Indonesian. A before V

P before V

total

N-V

 98.2 % 266.0

  1.8 %   5.0

100 % 271

di-V

  1.5 %   2.0

 98.6 % 137.0

100 % 139

within each example, the activity word is bare in (a), marked with N- in (b), and marked with di- in (c). These examples further support the observation that the generalized voice markers N- and di- do not unequivocally determine the thematic role of a preceding argument. Nevertheless, the generalized voice markers N- and di- do make for different preferences: with N- a preceding argument is more likely to be an experiencer as in (23b), than a theme as in (24b), or a patient as in (25); conversely, with di- a preceding argument is less likely to be an experiencer as in (23c), than a theme as in (24c), or a patient as in (25c). The effect of these two markers on the preferred interpretations is shown in Table 3. Table 3 looks at all words in the JFS Valency Corpus marked with either N- or di-, in which the closest preceding argument is either its agent (A) or its patient (P), and examines which of the two it is. The preferences are overwhelming: whereas for words marked with N- the preceding argument is understood as its agent in over 98 % of the cases, for words marked with di- the preceding argument is understood as its patient in a similarly large proportion of cases. It is of course these very strong preferences that are in large part responsible for the widespread characterization of these forms as active and passive markers. However, the existence of the dispreferred interpretations, albeit at a frequency of under 2 %, shows that N- and di- cannot be characterized straightforwardly as active and passive markers. Another reason why N- and di- are not proper markers of active and passive voice is that they do not constitute a paradigm of forms in complementary distribution. As evidenced in many of the above examples, words denoting activities frequently occur in bare form, containing neither N- nor di-. Conversely, it is also possible for the two markers, N- and di-, to co-occur on a single word: (26) Udah dinempel di sini. prf g.pass:g.act:attach in loc:dem:dem.prox ‘It got stuck here.’ [Talking about some food in the freezer] (JFS 457652161136221001) In (26) above, tempel ‘attach’ is marked by both N- and di-. Such cases of combined agent- and patient-oriented generalized voice marking are rare, but they do occur –

Valency classes in Jakarta Indonesian

965

see also example (35) below. Examples such as these show that the two markers, N- and di-, cannot be considered as regular markers of active and passive voice. With respect to N-, at least, yet another way in which it differs from a prototypical marker of active voice is that it is also associated with an aspectual function, sometimes inducing a preference for progressive and other atelic interpretations, occasionally even forcing such interpretations; see Hidayat (2011); Conners (2012) and Conners & Brugman (to appear). In summary, though, the semantic function of N- and di- involving agent and patient thematic roles, and the frequency effects associated with the clause types in which they occur, underscore the affinity of these two markers with more prototypical actives and passives, as captured in their proposed characterization as generalized voice markers. Alongside N- and di-, an important role is also played by a third generalized voice marker, the suffix -in. Whereas N- and di- select agent and patient roles respectively, the semantic role associated with the marker -in is less readily amenable to a single uniform characterization. In general, words marked with -in denote activities which are construed as highly transitive, in the sense that they involve metaphorical motion from an initial source to a final target or end-point where the activity terminates. Target may be understood as a kind of thematic role, though, as shown below, it typically overlaps with other thematic roles, such as stimulus, patient, instrument, benefactive, and the like. The suffix -in may thus be characterized as a target-oriented generalized voice marker. Like N- and di-, it is a weak generalized voice marker, in that it merely asserts the existence of a target in the semantic frame of the host word; it does not associate the target expression with any further grammatical properties. Whereas N- and di- are considered to be markers of generalized active and passive voice, -in may thus be characterized as a generalized applicative marker. Compared to N- and di-, the usages of -in involve a greater amount of lexical idiosyncrasies, pertaining to whether it can attach to a given word, and if it can, what kind of effect it has. For a few words, -in seems to function as a marker of iterativity, for example, pukul ‘hit’, pukulin ‘hit repeatedly’; such cases would appear to fall outside the function of -in as a generalized applicative marker. For some other words, addition of -in seems to have no obvious syntactic or semantic effect; for example, denger ‘hear’ and its counterpart dengerin ‘listen to’ are associated with identical grammar and a subtle and rather idiosyncratic difference in meaning. In such cases, -in would seem to be marking the existence of a target participant already present in the semantic frame of the host word, namely the stimulus; in this respect its function is analogous to that of N- and di- in examples (22)–(25). Most commonly, however, the presence of -in serves to introduce a new target participant into the semantic frame of the host word, though even here there is a considerable amount of idiosyncratic variation, not least with regard to the themat-

966

Thomas Conners, John Bowden and David Gil

ic role associated with the target participant. One such example is kulkas ‘refrigerator’, which forms the basis for kulkasin ‘put in the refrigerator’ (cf. English refrigerate with different meaning), thereby introducing a new target semantic role, that of patient, as well as a new agent role; however, lemari ‘cupboard’ does not allow an analogous *lemariin ‘put in the cupboard’. A quite different example is tutup ‘close’, which forms the basis for tutupin ‘cover’; here, the function of -in is to introduce a new target participant bearing the role of instrument – rather than making use of a completely different word, a more perspicuous English gloss might be ‘close with’. Nevertheless, in spite of examples such as the above, the majority of usages of the suffix -in fall into two major productive construction types: causative and benefactive. In the causative construction, addition of -in characterizes the theme of the host word as the target of a derived activity, thereby entailing the introduction of an additional participant into the semantic frame, namely the causer. For example, gede ‘big’ forms the basis for gedein ‘enlarge’; similarly, in (27) and (28) below, masuk ‘enter’/‘go in’ forms the basis for masukin ‘enter’/‘put in’: (27) a. Jarinya masuk. finger:assoc enter ‘Put your fingers in.’ [Adult to child] (JFS 626985091916080900) lagi, Na. b. Gasingnya masukin spinning.top:assoc enter:g.appl again fam\Hana ‘Put the top in, Hana.’ [Playing with child] (JFS 374255083417190203) (28) a. Kamu kok nggak mau masuk? 2sg how.come neg want enter ‘How come you can’t score.’ [Father to son playing football] (JFS 927778153506061101) yok. b. Yok, Dedek masukin exhrt younger.sibling enter:g.appl exhrt ‘Go on, put one in.’ [Playing a cooking game] (JFS 483073101518181202) While the (a) examples contain the bare form masuk, the (b) examples contain the derived form masukin. And whereas in (27) masuk is preceded by its theme, in (28) it is preceded by an expression denoting its causer. Note that the constructions

Valency classes in Jakarta Indonesian

967

with and without -in are formally parallel. What makes -in causative in the above examples is a semantic feature: while masukin is always interpreted as causative, the bare form masuk is seldom understood as causative, though as shown in example (28a), such an interpretation is nevertheless possible. Thus, unlike prototypical causative constructions in other languages, the causative use of the generalized applicative marker -in bears no consequences vis-à-vis the syntactic organization of the clause; its function is entirely semantic. In the benefactive construction, addition of -in introduces a new target participant into the semantic frame, and associates it with the semantic role of benefactive. For example, cabut ‘pull out’ forms the basis for cabutin ‘pull out for’; similarly, in (29)–(31) below, beli ‘buy’ forms the basis for beliin ‘buy for’. (29) a. Beli saya. buy 1sg ‘I bought one.’ [Talking about a shopping trip] (JFS 364366153158130607) b. Beliin bos Tante. buy:g.appl bos aunt ‘My boss bought it.’ [Talking about a toy] (JFS 418291130943140105) (30) a. Udah beli baju Lebaran? prf buy clothes Lebaran ‘Have you bought the clothes for Lebaran yet.’ [Talking about preparations for Lebaran (the most important Muslim holiday, also known as Idul Fitri)] (JFS 450823143141210601) b. Oh, nanti Oo beliin baterai. oh fut aunt buy:g.appl battery ‘I’ll buy the batteries.’ [Playing a game, the batteries run out] [JFS 436762104820080201] (31) a. Mau beli lu baju ni. want buy 2sg clothes dem.prox ‘I want to buy you a shirt.’ [Small talk] (https://www.facebook.com/permalink.php?story_fbid= 128176783908958&id=161146877260600, accessed 1 May 2012)

968

Thomas Conners, John Bowden and David Gil

b. Nanti saya beliin Bang kalo dah waras. fut 1sg buy:g.appl fam\elder.brother top prf recover ‘I’ll buy you some once you’re better.’ [Discussing food] (JFS 298246084300131009) Again, while the (a) examples contain the bare form beli, the (b) examples contain the derived form beliin. And whereas in (29) beli is followed by its agent, in (30) it is followed by its patient, and in (31) by its benefactive target. As with the causative examples, here too, the constructions with and without -in are formally parallel, and are distinguished only by the benefactive interpretation of the derived form beliin. What makes the benefactive function of -in resemble applicative constructions in other languages are the differential frequencies of the various possible patterns. For example, if both patient and benefactive expressions are present, then with beli the most common pattern is for the patient to precede the benefactive and for the benefactive to be flagged with buat ‘for’, as in (32), whereas with beliin the most common pattern is for the benefactive to precede the patient without any flagging, as in (33): (32) He, Oo mau beli mainan buat Timo. fill aunt want buy play:nmlz for Timo ‘Hey, I’m going to buy a toy for Timo.’ [Adults playing with child] (JFS 742330153105100501) (33) Oo kan beliin kamu es krim sama biskuit? aunt q buy:g.appl 2sg ice cream with biscuit ‘But I bought you ice cream and biscuits, didn’t I?’ [Adult playing with child] (JFS 391080143454100903) As suggested by their English translations, the above pair of sentences present a close superficial parallel to, among others, the dative-shift alternation in English. The crucial difference, however, is that in Jakarta Indonesian such patterns represent mere preferences which can be violated. For example, in (31a), an unflagged benefactive lu precedes a patient but without being licensed by a benefactive -in. Conversely, in (34) below, a flagged benefactive follows a patient although the benefactive -in is present. (34) Oo kan pernah beliin buku cerita buat kamu. aunt q exp.asp buy:g.appl book story for 2sg ‘But I’ve already bought you story books.’ [Adult playing with child] (JFS 377391095321121000)

Valency classes in Jakarta Indonesian

969

Thus, with respect to flagging and word order, (34) with -in presents a complete parallel to (32) without -in. Examples such as this show that, unlike at least some of the corresponding constructions in other languages, the benefactive use of -in bears no consequences vis-à-vis the syntactic organization of the clause. Like its causative counterpart, the benefactive usage of the generalized applicative marker is thus of an essentially semantic nature. Whereas the generalized active and passive markers N-and di- co-occur very infrequently, the generalized applicative marker -in commonly occurs with either of these two markers; one such example was dipagerin in (10). In fact, as shown in the following example, all three markers may co-occur on a single word: (35) diminggirin dulu rambutnya. g.pass:g.act:edge:g.appl before hair:assoc ‘Move your hair aside.’ [Adult to child] (JFS 667645134006290601)

5.2 Other similar alternations In addition to the three generalized voice markers considered above, there are a number of additional markers which, to varying degrees, involve alternations of relevance to valency and valency classes. One such marker is the suffix -an, which occurs with great frequency and with a wide range of functions which are not obviously related to one another. One common function takes a word denoting an activity and derives a word denoting a conventionally associated thing, for example goreng ‘fry’ > gorengan ‘fritter’, tulis ‘write > tulisan ‘script, main ‘play’ > mainan ‘toy’ (see example (32) above). Another, somewhat more systematic function takes a word denoting a property and yields its comparative counterpart, for example cakep ‘beautiful’ > cakepan ‘more beautiful’. Yet another function, somewhat more idiosyncratic, but applying in a few high-frequency activity words, involves the semantic augmentation of the activity and its associated arguments, for example jual ‘sell’ > jualan ‘engage in habitual selling’, udah ‘finish’ (usually used to denote the more abstract tense-aspect notion of perfect) > udahan ‘all finished’. Whereas none of the above functions involve voice alternations, in some cases -an is used to mark reciprocals, as for example peluk ‘hug’ > pelukan ‘hug each other’, ganti ‘change’ > gantian ‘swap’, and in a very recent innovation internet ‘internet’ > internetan ‘go online’ (reflecting the interactive reciprocal nature of the two probably most common online activities, namely chatting and gaming). Another marker, also associated with several apparently unrelated functions, is the form ke/ke-, which, depending on context, is written either separately or

970

Thomas Conners, John Bowden and David Gil

joined to the following word (though it is not clear whether this orthographic contrast reflects any substantive difference with respect to the strength of the bond between it and its host). Written as a separate word, ke is the dative marker discussed in Section 4.2 above. Written joined on to the following word, ke- has a number of additional functions. From cardinal numerals it forms ordinals, for example tiga ‘three’ > ketiga ‘third’. From at least one word, beli ‘buy’, it adds an abilitative meaning, kebeli ‘can afford’. Its main function, however, is to mark certain words as being associated with an agent of decreased level of volitionality or involvement; for example, bawa ‘carry’ > kebawa ‘get taken away accidentally’, jepit ‘pinch’/‘squeeze’ > kejepit ‘get squeezed’, setrum ‘electricity’ > kesetrum ‘get an electric shock’. As suggested by the above examples, this function of ke- resembles that of a passive. However, it differs from the generalized passive marker didiscussed in Section 5.1 in that whereas di- highlights the salience of the patient, what ke- does is to downplay the salience of the agent, by negating the characteristic agentive features of volitionality or control – it might therefore be characterized as a marker of de-agentivity. Another related difference is that, as suggested by the above examples, this function of ke- is often associated with an adversative meaning. Yet another difference between ke- and di- is that ke- is much less commonly used, and is more idiosyncratic in its application. In particular, in some words with an inherently deagentive meaning, ke- may apply without obvious semantic effect, for example temu ~ ketemu ‘find’/‘meet’. In fact, for at least one such word, the presence of ke- is almost obligatory: ?*tawa > ketawa ‘laugh’. A third marker, the circumfix ke--an, appears to constitute a combination of the two preceding markers, ke- and -an. While there are indeed some affinities between the functions of ke--an and the two markers that form its constituent parts, the properties of ke--an are not predictable from those of ke- and -an; it is thus most appropriately analyzed as a unitary circumfix. One common function of the circumfix ke--an is to form words denoting abstract entities, for example bakar ‘burn’ > kebakaran ‘fire’ (a spatio-temporal event), jadi ‘become’ > kejadian ‘event’, putus ‘disconnect’ > keputusan ‘decision’. Of relevance here is its function as a deagentive marker, resembling the corresponding function for ke- described above. Generally speaking, the use of ke--an is of somewhat greater generality than that of ke-, though it is also conditioned in lexically idiosyncratic ways. One class of cases involves property words, for which ke--an introduces an adversely affected experiencer, which may often be translated into English with ‘too’, for example sempit ‘tight’/‘narrow’ > kesempitan ‘be in an unpleasantly narrow enclosure’, gede ‘big’ > kegedean ‘too big’, pendek ‘short’ > kependekan ‘too short’. Another class of cases also introducing an adversely affected experiencer is that of meteorological and other similar expressions, for example ujan ‘rain’ > keujanan ‘get rained on’, banjir ‘flood’ > kebanjiran ‘get flooded’, siang ‘daytime’ > kesiangan ‘experience time pressure by getting going too late in the morning’. Yet another class involves activity words, where the resulting meaning would seem to be a little more idiosyn-

Valency classes in Jakarta Indonesian

971

cratic, though still preserving the basic deagentive meaning, for example tinggal ‘stay’/‘leave’ > ketinggalan ‘get left behind’, tau ‘know’ > ketauan ‘get found out’, liat ‘see’ > keliatan ‘be visible’. Until now, all of the markers considered in this section have been bound forms. However, the final form which we shall consider here, kena, is an independent content word whose closest translation into English is perhaps ‘undergo’. As a content word it has a semantic frame, consisting of an experiencer and a theme. The word kena is a member of the single open syntactic category of Jakarta Indonesian; like almost all other such words, it can stand alone as a complete non-elliptical sentence – on its own, kena might be translated into English as ‘He got hit’, ‘Something’s happening to it’, or whatever, depending on the context. However, when occurring with an associated theme denoting an activity, the resulting construction resembles a periphrastic passive. Consider the following examples: (36) a. Kena mata nanti. undergo eye fut ‘It’ll get into your eyes.’ [Warning child playing with fish] (JFS 541816085900060505) b. Kena duri? undergo thorn ‘Did she get stuck by a thorn?’ [About her friend who was walking] (JFS 225025125344140904) c. Emang Lily kena tembak? indeed Lily undergo shoot ‘Did Lily get shot?’ [Playing doctor with a child] (JFS 619450091651230402) While in (36a) kena is followed by its experiencer mata ‘eye’, in (36b) and (36c) it is followed by its theme. However, whereas in (36b) the theme is a thing, duri ‘thorn’, in (36c) it is an activity tembak ‘shoot’, yielding what looks just like a periphrastic passive. But as suggested by (36a, b), there is nothing “grammatical” about (36c); it is the straightforward product of the juxtaposition of three content words belonging to the single open syntactic category of Jakarta Indonesian, kena and its experiencer and theme arguments.13 13 Note how example (36c) differs crucially from its counterparts in many other languages, e.g. the English Lily got shot. As for kena in kena duri in (36b) and kena tembak in (36c), it might be suggested that English offers a parallel between got in got a thorn and in got shot. However, in English there is plenty of evidence that the got in got shot is a “different” got to that in got a thorn; specifically, that it is of a grammatical rather than a lexical nature. Formally, the got in got shot

972

Thomas Conners, John Bowden and David Gil

6 Conclusion Having surveyed the valency patterns of Jakarta Indonesian in the preceding sections, we may now address the more general question of how these valency patterns compare with those of other languages, or in other words, how Jakarta Indonesian fits into typological space. The answer depends crucially on whether we consider valency classes as categorical black-or-white entities, or as preferencebased entities in various shades of grey. Adopting the former, categorical perspective, Jakarta Indonesian, as a Relatively IMA Language with but a single valency class, is very much a typological outlier – at least with respect to the other languages considered in this volume. However, following the latter, preference-based approach, represented in the accompanying database, Jakarta Indonesian turns out looking substantially more like other, more familiar languages. The contrast between these two different perspectives is illustrated schematically in Figures 1 and 2 below. These two figures both make reference to a sample set of 15 coding frames, involving a bivalent word V, its A(gent) and P(atient) arguments, the non-absolutive marker (s)ama, and the generalized active and passive voice markers N- and di-. Figure 1 shows these 15 coding frames ranked in terms of formal complexity, while Figure 2 shows the same coding frames ranked in terms of frequency of occurrence.

N-V

V

di-V

A N-V P P N-V A

AVP PVA

A di-V P P di-V A

(s)ama A N-V P P N-V (s)ama A

(s)ama A V P P V (s)ama A

(s)ama A di-V P P di-V (s)ama A

Fig. 1: 15 coding frames ranked by formal complexity.

governs a particular form of the verb, the past participle, whereas the got in got a thorn is actually assigning accusative case (as can be seen by replacing the NP with a pronoun, got him). And semantically, its meaning is different: whereas the got in got shot has an abstract passive meaning, the got in got a thorn possesses a more concrete meaning involving the notion of acquisition and various extensions thereof – without further modification it cannot mean ‘get stuck by a thorn’. Similar arguments extend to many similar cases in other languages. In contrast, kena has the same grammatical and semantic function in both kena duri and kena tembak.

Valency classes in Jakarta Indonesian

N-V

V

di-V

A N-V P

AVP

P di-V A P di-V (s)ama A

(s)ama A N-V P

(s)ama A V P PVA P V (s)ama A

P N-V A P N-V (s)ama A

973

A di-V P (s)ama A di-V P

Fig. 2: 15 coding frames ranked by frequency.

Figure 1 ranks the 15 coding frames in terms of formal complexity. The simplest coding frame consists just of V; more complex frames are derived by the sequential introduction of additional items. Moving left or right adds a generalized voice marker, while moving down introduces A and P arguments and then the flagging marker (s)ama. The perspective adopted in Figure 1 is additive, or incremental; coding frames are viewed as being derived from simpler coding frames by the introduction of meaning-bearing expressions, all of which are, from a strictly grammatical point of view, completely optional. This perspective is reminiscent of the usual way of looking at stacked modifier constructions, for example, English sentential adverbs, as in, say, John met Bill → John met Bill unexpectedly → John met Bill unexpectedly in the park → John met Bill unexpectedly in the park on Saturday. Both involve the incremental introduction of grammatically-optional expressions, each expression adding to the semantic specificity of the entire construction. However, while such a perspective is commonplace for stacked modifier constructions, it is less familiar in the context of basic clause structure and its associated alternations. Nevertheless, to the extent that Jakarta Indonesian is a relatively IMA Language with but a single valency class, this is the most appropriate way of representing the relationship between the various coding frames present in the language. In contrast, Figure 2 ranks the same 15 coding frames in terms of frequency of occurrence, classifying them into four groups, which may be assigned the informal labels very frequent, frequent, infrequent and very infrequent. The ranking is of a rough and approximate nature, averaging across different choices of V associated

974

Thomas Conners, John Bowden and David Gil

with different relative frequencies. The first group consists of the V with no additional lexical items, reflecting the fact that in Jakarta Indonesian, like in many other languages, the default and most frequent case is for arguments not to be expressed. (Within this group, the relative frequency of the bare V and the V marked with generalized active or passive markers varies considerably across different choices of V, but for very many, all the possibilities are very frequent.) The second group, consisting of coding frames that are also frequent, though less frequent than those of the first group, contains two active-like A V P constructions, with the V either bare or marked by the generalized active voice marker N-, and two passive-like P V A constructions, with the V marked by the generalized passive marker di- and the A either bare or flagged with (s)ama. The third group, consisting of coding frames that are infrequent, contains two A V P frames where the A is flagged with (s)ama and the V is either bare or marked with N-, and also two P V A frames, where the V is bare, and the A either bare or flagged with (s)ama. And the fourth group, consisting of coding frames that are very infrequent, involves cases where the V is flanked by A and P arguments but marked with the generalized voice marker that is dispreferred with regard to the given relative order of A and P: two P V A frames where the V is marked with the generalized active voice marker N-, with the A either bare or flagged with (s)ama, and two A V P frames where the V is marked with the generalized passive voice marker di-, and the A, again, either bare or flagged with (s)ama. Viewed from the perspective of Figure 2, Jakarta Indonesian looks less of a typological outlier, and more like many other languages, in particular other Austronesian languages. Focusing on the second group, containing the four most frequently occurring coding frames with both arguments expressed, we see that two of the four coding frames, A N-V P and P di-V (s)ama A, look just like garden variety active and passive constructions in a typical SVO language, with voice morphology marking the verb, and overt flagging of the agent in the passive construction. A third coding frame in this group, P di-V A, is somewhat less typical cross-linguistically; however, similar frames in other varieties of Malay/Indonesian and in other languages of the region have been argued, by Himmelmann (2005 and elsewhere), to support the characterization of such languages as involving so-called symmetric voice systems, in which active and passive are formal mirror-images, as is the case for the A N-V P and P di-V A coding frames here. Symmetric voice systems thus stand in opposition to the more cross-linguistically widespread state of affairs in which the passive construction is more formally complex than its active counterpart, and therefore, arguably, derivable from it: in symmetric voice systems, there is no evidence that one of the two members of an active-passive pair is more basic, and the second member derivable from the first. However, it is the fourth coding frame in this same group, namely the bare A V P construction, that poses the most serious challenge to various other approaches to voice in Jakarta Indonesian and related languages: if the A N-V P and P di-V A frames represent symmetric active

Valency classes in Jakarta Indonesian

975

and passive respectively, how, then, is the A V P frame to be characterized? In fact such three-way oppositions between bare V and Vs marked with generalized active and passive voice are not specific to Jakarta Indonesian but rather quite common across the region. In summary, then, if attention is limited to the top half of Figure 2, showing just the frequent coding frames, Jakarta Indonesian emerges as a relatively unexceptional language. It is only when the infrequent coding frames in the bottom half of Figure 2 are added to the picture that Jakarta Indonesian appears to be more of a typological outlier. In conjunction, Figures 1 and 2 show that formal complexity and frequency do not always coincide. Sometimes they may; for example, the bare V is both the simplest and one of the most frequent coding frames. But in other cases the two criteria conflict; for example, the P di-V (s)ama A coding frame, marked with both di- and (s)ama, is more formally complex but more frequent than its P V A counterpart, marked with neither. Such clashes between complexity and frequency pose a serious challenge to any attempt to organize the various coding frames into a small set of basic frames plus a larger set of frames derived from the basic frames by means of alternations. For example, whereas an incremental complexity-based approach might derive the P di-V (s)ama A coding frame from its P V A counterpart by means of the introduction of the two markers di- and (s)ama, a frequency-oriented approach would probably opt to derive the P di-V (s)ama A coding frame from its A N-V P counterpart by means of a “passive” alternation. We leave it to the reader to choose the approach that is most amenable to his or her needs and theoretical persuasions. The contrast between Figures 1 and 2 highlights the nature of Jakarta Indonesian as a wolf in sheep’s clothing. The wolf is the basic IMA structure of the language, with the absence of more than a single valency class. The sheep’s clothing is the frequency effects and associated system of valency preference classes, which conspire to make Jakarta Indonesian seem more similar to many other languages than it really is. In the context of the Leipzig Valency Classes Project, what is perhaps most striking about Jakarta Indonesian is that although the basic architecture of the language is much closer to the IMA prototype than is the case for most other languages, the superimposed frequency effects lead in the direction of a language that is quite ordinary from a cross-linguistic point of view, rather than in any number of other logically possible directions. As evident from the accompanying database, the valency preference classes and more frequent coding frames of Jakarta Indonesian fall within the range of variation of valency classes and grammaticalized coding frames observed in most other languages. Diachronically, the Jakarta Indonesian system of valency preference classes is perhaps most appropriately understood as the outcome of a lengthy process in which a prior and more categorical system of valency classes undergoes weakening, driven by some combination of contact-induced simplification and assimila-

976

Thomas Conners, John Bowden and David Gil

tion to various substrate languages with a prior presence in the region: see Gil (2015) for detailed discussion. Phylogenetically, though, the Jakarta Indonesian system of valency preference classes may be viewed as a kind of “living fossil”, providing a model for an intermediate stage in the evolution of valency classes, whereby the typically categorical valency classes of most languages might have emerged out of ancient IMA language origins by means of the conventionalization and grammaticalization of the kinds of preference rules that are still in evidence in Jakarta Indonesian.

pikir

teriak

ng(k)edip

batuk

lari

loncat

lompat

nyanyi

pergi

berangkat

ketawa

 10 THINK

 20 SHOUT AT

 46 BLINK

 47 COUGH

 49 RUN

 52 JUMP

 52 JUMP

 53 SING

 54 GO

 55 LEAVE

 57 LAUGH

Verb form

ujan

Meaning label

 69 RAIN

#

Appendix

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

V

Coding frame schema

–

–

–

–

–

–

–

–

–

–

+

–

Accidental Passive

–

–

–

–

+

–

–

–

–

–

+

+

Adversative Passive

–

–

–

–

–

–

–

–

–

–

–

–

Applicative Allative

–

–

–

+

–

–

–

–

–

–

–

–

Applicative Benefactive

+

–

–

+

+

+

–

–

+

+

+

–

Bare Applicative Transitive

–

–

–

+

–

–

–

–

–

–

–

–

Ditransitive Alternation

–

–

–

–

–

–

–

–

–

–

+

–

Impersonal Passive

–

–

–

–

–

–

–

–

–

–

–

–

Intransitive alternation

–

–

–

–

–

–

–

–

–

–

–

–

Obl. Subj. Passive Agentive byphrase

–

–

–

–

–

–

–

–

–

–

–

–

Obl. Subj. Passive Ditransitive

–

–

–

–

–

–

–

–

–

–

–

–

Passive Transitive

–

–

–

–

–

–

–

–

–

–

–

–

Thematic Adversative Passive

–

–

–

+

–

+

–

–

–

–

+

–

Transitive Alternation

Valency classes in Jakarta Indonesian

977

Verb form

teriak

sakit

dingin

mati

tinggal (2)

sedih

laper

gelinding

tenggelem

nyala

kering

pemburu

mendidih

nongol

sakit

Meaning label

 58 SCREAM

 59 FEEL PAIN

 60 FEEL COLD

 61 DIE

 61 DIE

 63 BE SAD

 64 BE HUNGRY

 65 ROLL

 66 SINK

 67 BURN

 68 BE DRY

 70 BE A HUNTER

 80 BOIL

 81 APPEAR

 82 BE ILL

#

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

1V

Coding frame schema

–

–

–

–

–

–

–

+

–

–

–

–

–

–

–

Accidental Passive

+

–

–

–

+

–

–

–

+

–

+

–

+

+

–

Adversative Passive

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Applicative Allative

–

–

–

–

+

+

–

–

–

–

+

–

+

–

–

Applicative Benefactive

+

+

–

–

+

+

–

+

–

–

+

–

+

+

+

Bare Applicative Transitive

–

–

–

–

–

+

–

+

–

–

+

–

+

–

–

Ditransitive Alternation

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Impersonal Passive

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Intransitive alternation

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Obl. Subj. Passive Agentive byphrase

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Obl. Subj. Passive Ditransitive

–

–

–

–

–

–

–

+

–

–

+

–

–

–

–

Passive Transitive

–

–

–

–

–

–

–

–

–

–

+

–

–

–

–

Thematic Adversative Passive

–

–

–

–

–

–

–

+

–

–

–

–

–

–

–

Transitive Alternation

978 Thomas Conners, John Bowden and David Gil

n(t)angis

jatoh

dingin

omong

makan

peluk

liat

liat

cium

takut

takutin

suka

seneng

kenal

cari

mandiin

cuci

cukur

tolong

bantu

ikut

 83 CRY

 84 FALL

 91 BE COLD

 22 SAY

  1 EAT

  2 HUG

  3 LOOK AT

  4 SEE

  5 SMELL

  6 FEAR

  7 FRIGHTEN

  8 LIKE

  8 LIKE

  9 KNOW

 11 SEARCH FOR

 12 WASH

 12 WASH

 14 SHAVE

 15 HELP

 15 HELP

 16 FOLLOW

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1 V (UTT2)

1V

1V

1V

–

+

+

+

+

–

–

–

–

–

–

–

+

–

–

+

+

–

–

–

–

–

–

–

–

–

–

–

–

+

–

–

+

–

+

+

–

–

–

+

–

–

+

–

–

–

–

–

–

+

–

–

–

–

–

+

+

–

–

–

–

–

–

–

–

–

–

+

–

+

–

–

–

–

–

–

+

+

–

–

–

+

–

–

+

+

(m)

+

+

–

–

–

+

–

–

–

–

+

+

+

+

+

+

+

+

+

+

–

–

+

–

+

–

–

–

–

–

–

+

+

–

–

–

+

–

–

–

+

+

+

–

–

+

+

+

+

–

+

–

–

–

–

–

–

–

–

–

–

+

–

+

+

–

–

–

+

–

–

+

–

–

–

–

+

–

–

–

–

–

+

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

+

+

+

+

+

+

(m)

–

–

+

–

–

+

+

+

+

–

–

–

–

–

–

–

–

–

–

–

–

–

+

–

+

–

+

+

–

–

–

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

+

–

–

–

Valency classes in Jakarta Indonesian

979

temu

minta

bangun

pecahin

patahin

bunuh

matiin

pukul

pukul

gebuk

sentuh

colok

pegang

potong

 17 MEET

 19 ASK FOR

 24 BUILD

 25 BREAK

 25 BREAK

 26 KILL

 26 KILL

 27 BEAT

 28 HIT

 28 HIT

 29 TOUCH

 29 TOUCH

 29 TOUCH

 30 CUT

Verb form

ketemu

Meaning label

 17 MEET

#

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

Coding frame schema

+

+

+

–

–

+

+

–

+

–

–

+

–

+

–

Accidental Passive

–

–

–

–

–

–

–

–

–

–

–

–

–

+

–

Adversative Passive

–

–

–

–

–

–

–

–

–

–

–

–

+

–

–

Applicative Allative

+

+

–

–

–

–

–

–

–

–

–

+

–

–

–

Applicative Benefactive

–

+

+

–

+

+

+

–

+

–

–

+

+

+

+

Bare Applicative Transitive

+

–

+

–

–

–

–

–

–

–

–

+

+

–

–

Ditransitive Alternation

+

–

–

+

+

+

+

–

+

–

–

+

+

–

–

Impersonal Passive

–

–

+

–

–

–

–

–

–

–

+

+

–

–

+

Intransitive alternation

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Obl. Subj. Passive Agentive byphrase

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Obl. Subj. Passive Ditransitive

+

+

+

+

+

+

+

+

+

+

+

+

–

–

–

Passive Transitive

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Thematic Adversative Passive

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Transitive Alternation

980 Thomas Conners, John Bowden and David Gil

sobek

kupas

kulitin

umpetin

bawa

ikat

tutup

penuhin

panjat

naik

main

giling

lap

gali

dorong

bawa

colong

ajar

dengar

masak

bikin

 32 TEAR

 33 PEEL

 33 PEEL

 34 HIDE

 38 CARRY

 40 TIE

 43 COVER

 44 FILL

 48 CLIMB

 48 CLIMB

 62 PLAY

 71 GRIND

 72 WIPE

 73 DIG

 74 PUSH

 75 BRING

 76 STEAL

 77 TEACH

 78 HEAR

 79 COOK

 85 MAKE

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

1V2

–

+

+

–

+

+

+

+

+

+

–

–

+

–

+

–

+

–

–

+

+

–

–

+

–

+

–

–

–

–

–

–

–

–

+

–

–

–

–

–

–

–

–

–

–

+

–

–

–

–

–

–

–

+

–

–

–

–

–

–

–

–

–

+

+

–

(m)

+

+

–

+

+

–

–

–

–

–

+

–

+

–

–

+

+

–

+

+

–

+

–

+

+

+

+

+

+

–

–

+

+

–

–

–

+

+

–

+

–

+

+

+

–

+

–

+

–

–

–

–

–

–

+

–

–

+

+

–

+

–

–

+

+

+

+

+

+

–

–

–

–

+

+

+

–

–

–

+

–

+

+

+

–

–

–

–

–

–

+

(m)

–

–

+

–

–

+

–

–

+

–

–

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

+

–

–

+

+

+

(m)

+

+

–

–

–

+

+

+

+

–

+

+

+

–

–

–

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Valency classes in Jakarta Indonesian

981

pengen

dapet

bakar

panggil

kasitahu

kasi

kirim

ambil

curi

pakein

tunjukin

lempar

tuang

 87 WANT

 89 GET (= obtain)

116 BURN (tr)

140 BE NAMED

 21 TELL

 36 GIVE

 37 SEND

 31 TAKE

 76 STEAL

 13 DRESS

 35 SHOW

 39 THROW

 42 POUR

Verb form

terima

Meaning label

 86 GET

#

1 V 2 ke+3

1 V 2 ke+3

1 V 2 ke+3

1 V 2 ke+3

1V2 dari+3

1V2 dari+3

1V23

1V23

1V23

1V2

1V2

1V2

1V2

1V2

Coding frame schema

–

–

–

–

–

+

+

–

–

+

+

–

+

+

Accidental Passive

–

–

–

–

+

+

–

–

–

–

–

–

–

–

Adversative Passive

–

–

–

–

–

–

+

+

–

–

–

–

–

–

Applicative Allative

+

+

–

–

–

+

+

+

–

+

+

+

–

+

Applicative Benefactive

–

–

–

–

–

+

–

–

–

+

+

–

–

–

Bare Applicative Transitive

+

+

–

–

–

+

–

–

–

+

+

+

–

–

Ditransitive Alternation

+

+

–

+

+

+

+

+

+

+

+

–

+

+

Impersonal Passive

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Intransitive alternation

–

–

–

–

–

–

–

+

+

+

–

–

–

–

Obl. Subj. Passive Agentive byphrase

–

–

+

–

–

–

–

+

+

+

–

–

–

–

Obl. Subj. Passive Ditransitive

+

+

–

+

+

+

(m)

+

+

–

+

–

–

+

Passive Transitive

–

–

–

–

–

–

–

–

–

–

–

+

–

–

Thematic Adversative Passive

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Transitive Alternation

982 Thomas Conners, John Bowden and David Gil

masuk

isi

taro

taro

bicara

cerita

bilang

duduk

duduk

tinggal (1)

muncul

 45 LOAD

 45 LOAD

 41 PUT

 45 LOAD

 18 TALK

 21 TELL

 22 SAY

 50 SIT

 51 SIT DOWN

 56 LIVE

 81 APPEAR

+

–

+

1 V LOC2

1 V LOC2

1 V LOC2

1 V LOC2

1 V ke+2 UTT3

1 V ke+2 soal+3

1 V ama+2 soal+3

–

–

–

–

–

–

–

1 V 2 LOC3 –

1 V 2 LOC3 –

1 V 2 ke+3

1 V 2 ke+3

1 V 2 ke+3

–

+

+

+

–

–

–

–

–

–

–

–

–

–

+

+

–

–

+

–

–

–

–

–

–

+

–

–

–

–

–

+

+

–

+

–

Legend: + occurs regularly, m occurs marginally, − occurs never, _ no data

isi

 44 FILL

–

+

+

+

–

–

+

–

–

+

+

+

–

+

–

–

+

+

–

–

–

–

–

–

–

–

–

–

–

–

–

+

+

+

+

+

+

+

+

+

–

+

+

–

–

–

+

–

–

–

–

–

–

–

–

–

–

+

–

+

–

–

–

–

–

–

–

–

–

+

–

+

–

+

–

–

–

–

–

(m)

(m)

+

–

+

–

+

+

+

–

–

–

–

–

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

Valency classes in Jakarta Indonesian

983

984

Thomas Conners, John Bowden and David Gil

Acknowledgements The Jakarta Indonesian valency database was completed by Thomas Conners. The accompanying chapter was written by David Gil. The database and the chapter are products of a joint project at the Jakarta Field Station of the Max Planck Institute for Evolutionary Anthropology, in which the naturalistic corpora of Jakarta Indonesian were coded for valency patterns; this project was led by Thomas Conners, with the participation of John Bowden, Erni Farida Sri Ulina Ginting, David Gil, Dalan Mehuli, Poppy Siahaan, Nadia Sulistiono, and Tessa Yuditha, and with technical assistance from Brad Taylor. This material was presented at two workshops of the Leipzig Valency Classes Project held at the Max Planck Institute for Evolutionary Anthropology in Leipzig; we are grateful to several participants at these events for helpful comments and suggestions. Finally, we are indebted to the leaders of the Leipzig Valency Classes Project, Bernard Comrie, Iren Hartmann, Martin Haspelmath, and Andrej Malchukov for their ongoing advice, assistance and patience, and their extensive comments on earlier versions of the database and chapter.

Abbreviations assoc circ distr excla exhrt exp.asp fam fill g.act g.appl g.pass imit pers prt

associative [second half of] circumfix distributive exclamation exhortative experiential aspect familiar filler generalized active generalized applicative generalized passive imitative personal particle

References Abdul Chaer. 1976. Kamus Dialek Jakarta. Jakarta: Nusa Indah, Ende. Cole, Peter & Gaby Hermon. 2005. Subject and non-subject relativization in Indonesian. Journal of East Asian Linguistics 14. 59–88. Cole, Peter, Gaby Hermon & Yassir Nasanius Tjung. 2005. How irregular is WH in situ in Indonesian? Studies in Language 29. 553–581.

Valency classes in Jakarta Indonesian

985

Cole, Peter, Gaby Hermon & Yassir Nasanius Tjung. 2006. Is there pasif semu in Jakarta Indonesian? Oceanic Linguistics 45. 65–90. Chomsky, Noam. 1981. Lectures on Government and Binding: Studies in Generative Grammar. Dordrecht: Foris. Conners, Thomas. 2012. More on the aspectual role of the nasal prefix in Jakarta Indonesian. Paper presented at the 16 th International Symposium on Malay/Indonesian Linguistics, Kelaniya, Sri Lanka, 24 June 2012. Conners, Thomas & Claudia Brugman. (to appear). The event semantic role of the nasal prefix in Jakarta Indonesian. In Wayan Arka (ed.), Proceedings of the 12 th International Conference on Austronesian Linguistics. Gil, David. 2001a. Creoles, complexity and Riau Indonesian. Linguistic Typology 5. 325–371. Gil, David. 2001b. Escaping Eurocentrism: Fieldwork as a process of unlearning. In Paul Newman & Martha Ratliff (eds.), Linguistic Fieldwork, 102–132. Cambridge: Cambridge University Press. Gil, David. 2002. The Prefixes di- and N- in Malay / Indonesian Dialects. In Fay Wouk & Malcom Ross (eds.), The History and Typology of Western Austronesian Voice Systems, 241–283. Canberra: Pacific Linguistics. Gil, David. 2003. English goes Asian: number and (in)definiteness in the Singlish noun-phrase. In Frans Plank (ed.), Noun Phrase Structure in the Languages of Europe, Empirical Approaches to Language Typology, 467–514. Eurotyp 20–7. Berlin & New York: Mouton. Gil, David. 2004. Riau Indonesian sama, explorations in macrofunctionality. In Martin Haspelmath (ed.), Coordinating Constructions (Typological Studies in Language 58), 371–424. Amsterdam: John Benjamins. Gil, David. 2005a. Isolating-monocategorial-associational language. In Henri Cohen & Claire Lefebvre (eds.), Categorization in Cognitive Science, 347–379. Oxford: Elsevier. Gil, David. 2005b. Word order without syntactic categories: how Riau Indonesian does it. In Andrew Carnie, Heidi Harley & Sheila Ann Dooley (eds.), Verb First: On the Syntax of VerbInitial Languages, 243–263. Amsterdam: John Benjamins. Gil, David. 2006a. Intonation and thematic roles in Riau Indonesian. In Chungmin M. Lee, Matthew Gordon & Daniel Büring (eds.), Topic and Focus, Cross-Linguistic Perspectives on Meaning and Intonation, Studies in Linguistics and Philosophy 8, 241–268. Dordrecht: Springer. Gil, David. 2006b. The acquisition of voice morphology in Jakarta Indonesian. In Natalia Gagarina & Insa Gülzow (eds.), The Acquisition of Verbs and Their Grammar: The Effect of Particular Languages, 201–227. Dordrecht: Springer. Gil, David. 2010. The acquisition of syntactic categories in Jakarta Indonesian. In Jan Don & Roland Pfau (eds.), Parts of Speech, Empirical and Theoretical Advances, 135–167. Amsterdam: John Benjamins. (First appeared in Studies in Language 32. 637–669, 2008.) Gil. 2012. Where does predication come from? Canadian Journal of Linguistics 57(2). 303–333. Gil, David. 2013. Riau Indonesian: a language without nouns and verbs. In Jan Rijkhoff & Eva van Lier (eds.), Flexible Word Classes, 89–130. Oxford: Oxford University Press. Gil, David. 2015. The Mekong-Mamberamo linguistic area. In Nick J. Enfield & Bernard Comrie (eds.), Languages of Mainland Southeast Asia: The State of the Art, 262–351. Berlin: De Gruyter Mouton. Gil, David (to appear) What is Riau Indonesian? In Scott Paauw & Peter Slomanson (eds.), Studies in Malay and Indonesian Linguistics [tentative title]. Berlin: De Gruyter Mouton. Grijns, C. D. 1991. Jakarta Malay: A Multidimensional Approach to Spatial Variation, Leiden: KITLV Press. Hidayat, Lanny. 2010. The acquisition of verb argument structure in Basilectal Jakarta Indonesian. PhD Dissertation. Newark, DE, USA: University of Delaware.

986

Thomas Conners, John Bowden and David Gil

Hidayat, Lanny. 2011. The analysis of the semantic function of the prefix N- in Basilectal Jakarta Indonesian. Proceedings of the Workshop on Tense, Aspect, Mood and Evidentiality in Indonesian Languages. Tokyo: ILCAA. Himmelmann, Nikolaus P. 2005. The Austronesian languages of Asia and Madagascar: typological characteristics. In K. Alexander Adelaar & Nikolaus P. Himmelmann (eds.), The Austronesian Languages of Asia and Madagascar, 110–181. London: Routledge. Ikranagara, Kay. 1980. Melayu Betawi Grammar (NUSA Linguistic Studies in Indonesian and Languages in Indonesia, vol. 9) Jakarta: Universitas Atma Jaya. Kähler, Hans. 1966. Wörterverzeichnis des Omong Djakarta. Berlin: Dietrich Reimer. Lerdahl, Fred & Ray S. Jackendoff. 1983. A Generative Theory of Tonal Music. Cambridge: MIT Press. Muhadjir. 1981. Morphology of Jakarta Dialect, Affixation and Reduplication. NUSA Linguistic Studies in Indonesian and Languages in Indonesia, Volume 11. Jakarta: Universitas Atma Jaya. Nordhoff, Sebastian. (this volume). Valency Classes in Sri Lanka Malay. DeGruyter Mouton. Shibatani, Masayoshi & Ketut Artawa. (this volume). Balinese Valency Classes. DeGruyter Mouton. Sneddon, James Neil. 2003. Diglossia in Indonesian. Bijdragen tot de Taal-, Land- en Volkenkunde 159. 519–549. Sneddon, James Neil. 2006. Colloquial Jakartan Indonesian, Canberra: Pacific Linguistics. Stoel, Ruben. 2005. Focus in Manado Malay, Grammar, Particles and Intonation, Leiden: CNWS Publications. Tjung, Yassir Nasanius. 2006. The formation of relative clauses in Jakarta Indonesian: a subjectobject asymmetry. PhD Dissertation. Newark, DE, USA: University of Delaware. van Minde, Don. 1997. Melayu Ambong, Phonology, Morphology, Syntax, Leiden: CNWS Publications. Wouk, Fay. 1989. The impact of discourse on grammar: verb morphology in spoken Jakarta Indonesian. PhD Dissertation, Los Angeles: UCLA. Wouk, Fay. 1999. Dialect contact and koineization in Jakarta, Indonesia. Language Sciences 21. 61–86.

Sebastian Nordhoff

24 Case frames in Sri Lanka Malay 1 Introduction Sri Lanka Malay is a high contact variety with Malayic lexicon and South Asian grammar. It is spoken by the descendants of mercenaries and exiles, who were brought by the colonial powers, the Dutch and the British, roughly between 1650 and 1850. During the last three centuries, the language has radically restructured its grammar. It has, for instance, shifted to head-final word order and developed a number of morphosyntactic categories unheard of elsewhere in the Malay world (case markers, infinitives, participles). The grammatical changes are nearly all traceable to influence from the local languages Sinhala and Tamil (see Nordhoff (2009) and references therein for a more thorough discussion). An area where patterns of the adstrate languages have not been copied is valency: Sinhala and Tamil have distinct conjugation classes for transitive and (certain types of) intransitive verbs; this is not found in Sri Lanka Malay. Sri Lanka Malay is a language with direct linking between semantic role and morphosyntactic expression. This direct link is established by means of postpositions, and these postpositions neatly map onto semantic roles. For instance, the clitic =yang is used to express patient, and patient can generally not be expressed by any other means. Similarly, =nang expresses recipient, and recipient cannot be expressed by any other means. This is different from languages like English, where the mapping between semantic roles and morphosyntactic expression is more involved. Typically, direct objects are patients or themes in English, but this general rule does not hold in a number of cases. For instance, a patient can be found in subject position, typically reserved for agents, under passivization. (1) The city was destroyed by Caesar. In addition, a recipient can be found in direct object position under dative shift. (2) John gave me the cake. Finally, a goal can be found in direct object position with locative alternation. (3) John loaded the truck with hay. In languages like English, it is therefore necessary to keep track of the other morphosyntactic marking applied in a particular sentence in order to successfully decode the propositional content. This is not necessary in Sri Lanka Malay, where no

988

Sebastian Nordhoff

such alternations exist. Sri Lanka Malay is a WYSIWYG language (Hengeveld 2011; Nordhoff 2011b) as far as the encoding and decoding of argument structure is concerned: what you seen in morphology (the enclitic postposition) is what you get in semantics (the thematic role). This has also been observed for the adstrate Sinhala by Gair (1991), who coined the term ‘Strong Lexical Case Assignment Hypothesis’.

2 Basics of Sri Lanka Malay morphosyntax Sri Lanka Malay is a language based on Malay vocabulary and Dravidian grammar. Its grammar differs dramatically from any other variety of Malay, so that it is considered ‘a language in its own right’ (Adelaar 1991), rather than a dialect of Malay/ Indonesian, e.g. Jakarta Indonesian (Gil, this volume). Among the features which distinguish Sri Lanka Malay from Malay/Indonesian, we find: − retroflex consonants, prenasalized stops, vowel length, consonant length − relatively abundant morphology, e.g. infinitives and participles − SOV word order − postpositional case markers − preposed relative clauses − grammaticalized indefiniteness marking − copula constructions (Nordhoff 2011a) − solid word class distinctions (Nordhoff 2013) − serial verbs (Nordhoff 2012) It is thus about as different from Malay/Indonesian as English is from Marathi as far as grammar is concerned. Its lexicon, on the other hand, is solidly Malay with about 90 % cognacy with other Malay varieties (Paauw 2004). In the context of valency, word order, head marking, and dependent marking are the most interesting domains. These will briefly be discussed here.

2.1 Word order Sri Lanka Malay word order is generally head final, with SOV being the normal case for pragmatically neutral sentences.1 A model sentence is given in (4).

1 SOV is understood to mean Agent Patient V with no claim as to the existence of grammatical relations like subject and object in Sri Lanka Malay.

Case frames in Sri Lanka Malay

989

(4) Itthu baathu=yang incayang Seelong=dering laayeng nigiri=nang dist stone=acc 3s.polite Ceylon-abl other country=dat asà-baapi. cp-bring ‘These stones, he brought them from Ceylon to other countries.’ (K060103nar01) Note that the realization of four NPs is very atypical; it is more common to only have one overt NP, with the other NPs being dropped. The nature of the NPs has no bearing on the possibility to drop them; any NP can be dropped if recoverable from context. The order of the prenominal NPs is governed by pragmatic considerations of topicality, in this case itthu baathu ‘those stones’ is topicalized. As far as morphosyntax is concerned, word order is free. In pragmatically neutral contexts, all NPs are preverbal. For focalization purposes, right extraposition of one NP is possible. Example (4) also shows the main device Sri Lanka Malay uses for the indication of semantic roles: postpositional case markers. We find the accusative marker =yang, the dative marker =nang, and the ablative marker =dering. The pronoun incayang ‘he’ does not take a postposition in this sentence. In the context of this paper, I will consider NPs like incayang to be in the nominative.

2.2 Head marking While the encoding of semantic roles/valency is quite rich as far as dependents are concerned, such is not the case for verbs: there is no indexing, and the existence of any diathesis marking on the verb is debatable (see below).

2.3 Dependent marking The question of ‘who did what to whom’ is answered by postpositions in Sri Lanka Malay. These morphemes are also often analyzed as clitics, a theoretical problem common in head-final languages. Nordhoff (2009) adduces evidence for their sta-

Tab. 1: Case postpositions in Sri Lanka Malay and their glosses. =yang =nang =dering =ka =sàsaama

accusative dative ablative locative comitative

990

Sebastian Nordhoff

tus as clitics, rather than affixes or independent words. Table 1 gives an overview of the most common morphemes and their meanings. Examples (5)–(8) show prototypical instances of the use of these postpositions.2 (5) Itthusubbath deram pada jaalang arà-kijja butthul ruuma therefore 3pl pl road non.past-make correct house pada=yang arà-picca-kang. pl=acc non.past-broken-caus ‘Therefore, they build the street, they demolish many houses.’ (K051222nar04) (6) Kithang lorang=nang baaye mliiga athi-kaasi. 1pl 2pl=dat good palace irr-give ‘We will give you beautiful places.’ (K051213nar06) (7) Kaaya oorang=dering=jo arà-cuuri. rich man=abl=emph non.past-steal ‘It was from rich people that he stole.’ (K051206nar02) (8) Minnyak klaapa=ka gooreng. coconut.oil coconut=loc fry ‘Fry it in coconut oil.’ (K060103rec02) In certain cases, the postpositions can be dropped. This is most notably found in topicalizations as in (9), where the locative marker =ka is dropped. (9) Seelong=0̸ samma thumpath=0̸ mlaayu aada. Ceylon all place Malay exist ‘In Sri Lanka there are Malays everywhere.’ (K051222nar04)

2 Most examples in this paper come from naturalistic discourse. The source of the examples follows the English translation. ‘K051222nar04’ for instance refers to the fourth narrative of the 22 nd of December 2005 from the city of K(andy). All media files and texts are in principle available from the Dobes archive (http://www.mpi.nl/tools/imdi_browser). Examples come from a variety of speakers whom I recorded during my stays in Sri Lanka (7 months between 2006 and 2011) and the visit of some speakers to Europe (three weeks in 2008). Examples with the source XXXXXXXeliXXX are from elicitation sessions. Some examples in the introductory passages of sections are constructed by me. These do not have source attributions.

Case frames in Sri Lanka Malay

991

The accusative marker =yang is furthermore often dropped if the context of the sentence already makes it clear which referent is the agent (baapa ‘father’) and which referent is the patient (pohong ‘tree’), as reverse roles are very unlikely.3 (10) Baapa derang=pe kubbong=ka hatthu pohong=Ø nya-poothong. father 3pl=poss garden=loc indf tree past-cut ‘My father cut a tree in their garden.’ (K051205nar05) The particular parameters conditioning the occurrence of =yang are poorly understood, and no hard and fast rule can be given. Animacy, definiteness, topicality, number, affectedness, and idiolect all play a role, but none of these parameters is sufficient on its own to predict the occurrence of =yang (Nordhoff 2009: 286–291). Taking a look at a number of semantic roles and a number of postpositions, one can map the ways in which they relate to each other. This is shown in Table 2. Above, I have argued that there is a clear and direct mapping between semantic roles and morphosyntactic expression. Taking a look at Table 2, we see that this overstates the regularities. It is common that a morpheme can be used for more than one function, and, more problematically, several functions can be encoded by more than one morpheme. This is the case for the coding of patient by both =yang and =nang, as well as the coding of goal by both =nang and =ka. Looking more closely into these cases, it appears however that they are not of the same type as passivization, dative shift or locative alternation in English. The use of =nang for marking arguments normally seen as patients is restricted to a very small set of three verbs: puukul ‘hit’, theembak ‘shoot’, thiikam ‘stab’. In all of these three cases, we are dealing with transfer or impact, so that the semantic role could as well be interpreted as goal or recipient. This shows that what we first took as a patient is indeed not a patient in Sri Lanka Malay, but rather a recipient. This thus relieves us of the unfortunate one-to-many mapping of roles and cases. There is a lexical specification of which highly transitive verbs take a recipient (the three just discussed), and which take a patient (all the others). There is no way to alter this and to make a verb from group 1 take =yang, or to make a verb from group 2 take =nang. The second one-to-many mapping concerns the encoding of goal. This is discussed below in Section 4.1 To sum up, even with the two minor glitches just presented, it seems fair to say that Sri Lanka Malay is very close to a direct mapping of semantic roles onto morphosyntax.

3 A similar argument can be made for jaalang ‘street’ in (5) and mliiga ‘palace’ in (6), but in those cases it could be argued that the absence of the accusative marker has to do with the lack of affectedness. This is clearly not the case in (10), where the tree is highly affected.

992

Sebastian Nordhoff

Tab. 2: Mapping of postpositions on semantic roles in Sri Lanka Malay. yang patient theme

nang

ka

dering

locative

instrumental source path

goal recipient benefactive purpose experiencer

2.4 Some complications: modals, institutions and involitive derivation Verbs are the prototypical assigners of case in all morphosyntactic theories. In Sri Lanka Malay, there are two additional processes which lead to the expression of case on an NP, and which are unrelated to verbs. These are modals and institutional actors. A third process, involitive derivation, changes the feature [+volition] of a verb to [−volition]. As a result, a former actor will lack volition and will instead be coded as an experiencer. These three cases will now be discussed and illustrated in turn. Sri Lanka Malay has a lexical category of modals, whose main members are boole ‘can’, thàrboole ‘cannot’, maau ‘want’ and thussa ‘want.not’. These words have in common that they assign the infinitive mà- to their clausal complement, and the dative =nang to the agent of the verb of their clausal complement. This is exemplified in (12) and (13). In (11), we see a plain clause without case marking on the NP. In (12), where a modal is added, we see that the verb is marked for infinitive with mà-, and the NP takes the dative marker =nang. (11) Tony arà-nyaanyi. Tony nonpast-sing ‘Tony sings.’ (12) Tony=nang mà-nyaanyi boole. Tony=dat inf-sing can ‘Tony can sing.’ When the modal is procliticized, the infinitive marker is dropped. (13) Tony=nang boole-nyaanyi. Tony=dat can-sing ‘Tony can sing.’

Case frames in Sri Lanka Malay

993

In cases like (13), the coding of the actor is thus not as expected. Given that the propositional content changes as far as its epistemic value is concerned, we are, however, not dealing with a morphosyntactic alternation expressing the same semantic content, but with a difference in semantic content mirrored by a difference in morphosyntactic encoding. Still, it is important to point out that not all instances of case marking in Sri Lanka Malay are due to the 1 : 1-mapping argued for above. Another instance of an erstwhile actor bearing the dative marker after the addition of some material is the so-called ‘involitive derivation’. The prefix kànà-, when applied to a verb, indicates lack of volition and makes the new verb assign the dative to a former agent. This is shown in examples (14)−(15). In (14), the singer is volitional and is 0-̸ marked, whereas in (15), he is non-volitional, and is marked with the dative. One can argue that kànà-changes the semantic role a verb assigns from agent to experiencer, the latter unintentionally taking part in the action. (14) Tony arà-nyaanyi. Tony nonpast-sing ‘Tony is singing.’ (15) Tony=nang arà-kànà-nyaani. Tony=dat nonpast-invol-sing ‘Tony is singing involuntarily/against his will.’ Involitive derivation can also be found in predications with more than one argument (16). (16) Tony-nang naasi anà-kànà-maakang. Tony=dat rice past-invol-eat ‘Tony ate rice against his will.’ Modals and involitive derivation both belong to the verbal domain, and act upon the arguments of the verb. As such they are close to the traditional idea of case being assigned by verbs, although they operate outside of lexical semantics. A third process that complicates case assignment in Sri Lanka Malay is unrelated to verbs: institutional actors. Institutional actors (e.g. governments, armies or law enforcement) in Sri Lanka Malay take ablative marking instead of the normal 0-̸ marking. This does not depend on the number of arguments the verb takes, and it also does not alter the coding of the remaining arguments. Examples (17)–(18) show the difference in encoding of non-institutional and institutional actors. In (17), we are dealing with a non-institutional actor, where we find 0-̸ marking. The institutional actor in (18) receives ablative marking. Again, it is not possible to swap these markings, as this would lead to ungrammaticality.

994

Sebastian Nordhoff

(17) Tony arà-dhaathang. Tony nonpast-come ‘Tony is coming.’ (18) Police=dering arà-dhaathang. police=abl nonpast-come ‘The police are coming.’ Like the modals, institutional actors obscure the direct link between thematic role and morphosyntactic expression, but do not give rise to a free alternation of case frames, which could be exploited for stylistic purposes for instance. To sum up this section, case marking in Sri Lanka Malay mainly depends on the lexical verb and the thematic roles it licenses. There is generally a transparent relation between the case selected and the underlying thematic role. This can be obscured by processes unrelated to argument structure, namely modals, involitive derivation, and institutional actors. We will now see how case markers are employed in predicates with between zero and four arguments.

3 Valency patterns One result of this study is that Sri Lanka Malay verbs can hardly be categorized based on the number of arguments they take. In order to make this article comparable to the other articles in this volume, a practical approach based on the arity of the English translational equivalent is chosen. I will start with 0-place predicates and end with 4-place predicates. I have used my best judgment to decide which verb goes where, but different choices would always have been possible.

3.1 Zero-place predicates There is one zero-place predicate, uujang ‘rain’. Other types of precipitation (snow, hail, sleet) are unknown in Sri Lanka. One could probably add a beneficiary or a source to this predicate, but this is outside of the prototypical context this predicate is used in. (19) 0̸ arà-uujang. nonpast-rain ‘It is raining.’

3.2 One-place predicates Sri Lanka Malay is a role-dominated language with no grammatical relations. The semantic role of a referent is directly reflected in morphosyntax through the case

Case frames in Sri Lanka Malay

995

marker selected (see Section 2.3). One-place-predicates can have their argument marked with either nominative, accusative, dative, or ablative.4 The discussion of case frames will therefore focus on these four cases here. Eventually, the locative will also enter the picture. Cases not assigned by the verb, but by a modal or an ‘institutional actor’ (see Section 2.4), are marked by an asterisk. (20) IO nom SL J P P M J P P M J P P M J P P M J P dat J P P M JP P PM M JQ acc TM J J P P M JP P dat*P PM M P J M P PM JP M P P J KR abl UM N With one-place predicates, nominative is the most common case (21). Dative is assigned for experiencers (22), or if a modal is present in the clause (25). Accusativemarking is very restricted in one-place predicates and has only been found on two verbs: thìnggalam ‘sink’ (23) and jaatho ‘fall’ (24). The ablative finally is not conditioned by semantic role, but by the nature of the referent: if it is an institution, such as a government or a committee, use the ablative, otherwise, stick with the semantic role the verb assigns (26). (21) Itthukapang Tony Hassan=Ø su-pii. then Tony Hassan past-go ‘Then Tony Hassan left.’ (K060116nar09) (22) Go=dang karang bannyak thàràsìggar. 1s.familiar=dat now very sick ‘I am now very sick.’ (B060115nar04) (23) Titanic kappal=yang su-thìnggalam. Titanic ship=acc past-sink ‘The ship “Titanic” sank.’ (K081104eli05) (24) Piiso=yang meeja=dering baawa=nang asà-jaatho seppe kaaki knife=acc table=instr down=dat cp-fall my foot su-poothong kìnna. past-cut advers ‘The knife fell down (from the table) and cut my foot.’ 4 To the best of my knowledge, there is no way to distinguish arguments from adjuncts in Sri Lanka Malay. I therefore treat all participants of a verb alike and refer to them as ‘arguments’. Note

996

Sebastian Nordhoff

(25) Kithang=nang ... two o’clock=ke=sangke bole=duuduk. 1pl=dat two o’clock=simil=until can-stay ‘We can stay up until two o’clock.’ (K061026rcp04) (26) Police=dering su-dhaathang. police=abl past-come ‘The police came.’ (K081105eli02) Overall, nominative marking is by far the most frequent, followed by the well-developed dative for actors lacking volition (see below), mainly experiencers. The accusative is rare, and the ablative is not conditioned by the verb. To my knowledge, the only one-place predicates taking the accusative are thìnggalam ‘sink’ and jaatho ‘fall’, but there might be more. Both cases can be explained by influence from Sinhala (Gair 1991).

3.3 Two-place predicates With two place predicates, we have to distinguish the actor argument and the undergoer argument. The actor can be described in exactly the same way as for oneplace predicates: normally, it is encoded in the nominative, but lack of volition or modals in the clause trigger dative marking. Institutional actors take the ablative. Note that accusative marking for actors is not attested in Sri Lanka Malay.5 The non-actor argument will normally be in the accusative, although the accusative marker =yang is often dropped. Recipients and beneficiaries, as well as some rare patients (of the verbs puukul ‘hit’, thiikam ‘stab’ theembak ‘shoot’) will be in the dative. Askees are in the locative. (27) I O nom S J P J P P J P dat P P J J JQ T JP dat* P P P P P J J J JR abl U J J K

O nomS L P P M P M P P M P acc P P M P P P M P PM M Q dat TM P P M P M P P M abl P P P M P P M P PM R loc UM N

Note that the forces operating on the actor argument (modals, institutions) are independent of the forces operating on the undergoer argument, which is only

that this might lead to more than four arguments for a predicate, although I limit myself to up to four arguments in this paper. 5 Accusative marking for actors is found in the adstrate Sinhala (Gair 2003: 791), so it is significant that it is not found in Sri Lanka Malay.

Case frames in Sri Lanka Malay

997

conditioned by the thematic roles the verb assigns. As a result, any of the 3×5 possible combinations in (27) is grammatical. I will not illustrate the different possibilities for actors here, as these are identical to what has been said above for one-place arguments. As far as the undergoer argument is concerned, it can be marked as nominative (28) or accusative (29). (28) Baapa derang=pe kubbong=ka hatthu pohong=0̸ nya-poothong. father 3pl=poss garden=loc indf tree past-cut ‘My father cut a tree in their garden.’ (K051205nar05) (29) Ithukapang lorang=pe leher=(yang) kithang=Ø athi-poothong. then 2pl=poss neck=acc 1pl irr-cut ‘Then we will cut your neck.’ (K051213nar06) Dative marking of the undergoer argument is also found, but refers to three semantically distinct concepts: Patients of some verbs are marked with the dative, as explained above. An example for this is given in (30). (30) Rose-red=Ø buurung=nang su-puukul. Rose-red bird=dat past-hit ‘Rose-red hit the bird.’ (K070000wrt04) If a modal is present in a clause with a verb which already assigns dative to the non-actor, both actor and non-actor will be marked with the dative, giving rise to ambiguity. (31) Se=dang Farook=nang bole=puukul. 1s=dat Farook=dat can=hit ‘I can hit Farook.’ (K081104eli05) In those cases, the actor is normally associated with the leftmost argument, as in (31), while the undergoer is the other argument. When using pointing gestures, this can be overruled, as in (32). (32) Ini kaaka=nang itthu kaaka=nang bole=puukul. prox elder.brother=dat dist elder.brother=dat can=hit ‘This brother can hit that brother.’ ‘That brother can hit this brother.’ (K081104eli05)

998

Sebastian Nordhoff

A more typical use of the dative is the marking of beneficiaries, as in (33). Recipients are not found in two-place predicates, since they necessarily involve an item received; they require at least two additional participants in their predication. (33) Derang pada=0̸ arà-banthu cinggala raaja=nang. 3pl pl non.past-help Sinhala king=dat ‘They help the Sinhalese king.’ (K051206nar03) Finally, goal of motion is also marked with the dative (34b). (34) a. Guunung=ka=jo kithang arà-duuduk; mountain=loc=emph 1pl non.past-stay ‘It is in the hills that we live;’ b. guunung=nang=jo kithang arà-pii. mountain=dat=emph 1pl non.past-go ‘it is to the hills that we go.’ (B060115prs01) Example (34a) also serves to illustrate the use of the locative in two-place predicates, in this case duuduk ‘reside’. The ablative is used with two-place predicates to encode the source of motion, as in (35). (35) Spaaru Indonesia=dering dhaathang aada. some Indonesia=abl come exist ‘Some came from Indonesia.’ (K060108nar02) Another use of the ablative, namely instrument, is rarer with two-place predicates, since an instrument normally requires an additional participant to manipulate, so that we are dealing with at least three participants, which is the topic of the next section. An exception to this generalization is (36), where we find two uses of instruments in the act of playing. (36) Thaangang=dering bukang kaaki=dering masà-maayeng. hand=abl neg.nonv leg=abl must-play ‘You must play not with the hands, but with the feet.’ (N060113nar05) We have assumed that the actor always is in the nominative in the examples above, barring presence of modals or institutional actors. With experiencer verbs, the ac-

Case frames in Sri Lanka Malay

999

tor/experiencer can actually be in the dative, and the undergoer/stimulus either in the nominative or in the accusative. (37) Suaara-Ø derang=nang su-dìnngar. noise indf 3pl=dat past-hear ‘They heard a noise.’ (K070000wrt04) (38) Se=dang ini oorang=yang thaau. 1s=dat prox man=acc know ‘I know this man.’ To sum up, two-place predicates normally have zero-marked actors, and undergoers are either marked for accusative or dative, with locative and ablative being more marginal possibilities. In special cases, actors can be marked for dative.

3.4 Three-place predicates Three-place predicates are typically predicates of transfer, i.e. giving and taking away. As such, they include an agent, a theme, and a goal (or source). The agent is typically in the nominative, although modals in the sentence can change this to dative, and institutional agents trigger ablative marking, as always. The theme is either unmarked (=nominative) or in the accusative. Recipients are marked by the dative, and sources by the ablative. This is schematized in (39). (39) I O nom S J P P J J P P JP P dat P P JQ T dat* P J J P P JP P P P J P KR abl U

O Q nomS T R acc U

O Q datS TL M M M R src UM M M M M M N

The following examples illustrate these patterns for the verb kaasi ‘give’. (40) Se=ppe baapa=0̸ incayang=nang ummas=0̸ su-kaasi. 1s=poss father 3s.polite=dat gold past-give ‘My father gave him gold.’ (K070000wrt04) (41) Incayang=nang [appointed member=pe hathu thumpathan]=yang 3s.polite=dat appointed member=poss indf post=acc government=ka anà-kaasi. government=loc past-give ‘(They) gave him a post as appointed member in the government.’ (N061031nar01)

1000

Sebastian Nordhoff

It is generally rare to find a clause with three realized (i.e. non-dropped) NPs. Most three-place predicates in the corpus with three NPs contain the verb kaasi ‘give’. There are other canonical three-place verbs in the corpus, but there are complications unrelated to argument structure which make them less suitable for illustration purposes. Example (42) for instance has a syntactically complex theme. It furthermore does not show =yang marking because of the low animacy and affectedness of the theme. Example (43) has three arguments with nice case markers, but one of them is an ‘institutional actor’, marked with the ablative. (42) Derang=pe umma=Ø derang=nang [jaithan=le, jaarong pukurjan=le] 3pl=poss mother 3pl sewing=addit needle work=addit su-aajar. past-teach ‘Their mother taught them sewing and needle work.’ (K070000wrt04) (43) Police=dering see=yang remand=nang su-kiiring. police=abl 1s=acc remand=dat past-send ‘The police sent me into custody.’ (K081105eli02) There are no instances of instruments in genuine three-place predicates with three overt NPs in the corpus. An elicited example is (44). (44) Tony piiso=dering ini daaging=yang arà-poothong. Tony knife=abl prox meat=acc nonpast-cut ‘Tony cuts this meat with a knife.’ Other combinations of roles (and therefore cases) are possible for three-place predicates. Example (45) for instances involves a path, next to an actor and a goal. (45) Spaaman derang=pe maama-yang=pe ruuma=nang itthu jaalang=dering 3s 3pl=poss uncle-3kin=poss house=dat dist street=abl pii aada. go exist ‘He went to his uncle’s house via that street.’ (B060115nar05)

3.5 Four-place predicates Four-place predicates nearly exclusively involve transfer of a theme from a source to a goal. As such the typical case frame is [NOM ACC ABL DAT]. An example is (46). The accusative marker can be dropped as usual.

Case frames in Sri Lanka Malay

1001

(46) Itthu baathu=yang incayang=Ø Seelong=dering laayeng nigiri=nang dist stone=acc 3s.polite Ceylon-abl other country=dat asà-baapi. cp-bring ‘These stones, he brought them from Ceylon to other countries.’ (K060103nar01) By leaving out one or several of the NPs, most of the types discussed in the previous sections can be obtained.

3.6 Summary of valency structure To sum up the assignment of zero, accusative, dative and ablative, on the roles of S, A, P and R, the following can be said: − The dative marker can be found on R and rarely P. Additionally, it can be found on S and A if they are experiencers. Furthermore, modals can assign the dative to S or A. − The accusative marker can be found on P and in rare instances on S. − The ablative marker can be found on S and A when they are institutional. It can furthermore be found on instruments and sources, widely construed. − Zero can be found on S, A and P. Zero is never found on R.

4 Alternations Sri Lanka Malay does not have any straightforward alternations. In the following I list some phenomena which may be considered alternations, but none of those is truly convincing.

4.1 Uncoded alternations: Case There are some idiolectal variations as to the cases verbs of motion and asking require. For some speakers, verbs of motion work with both the locative and the dative for the goal argument, while for others, only the dative is possible. This might be due to influence from Sinhala, where for very salient places, like the national capital, the dative marker can be dropped, and the bare stem is used, which is homophonous with the genitive/locative in these cases. Upon investigating the distribution of verbs in more detail, it was found that when asked, speakers judge the locative as ungrammatical.

1002

Sebastian Nordhoff

A second alternation, even more restricted, is the choice between locative and ablative for the verb mintha ‘ask’. The ablative is normally used when the item requested is tangible, whereas the locative is used when information is requested. Upon more detailed investigation, speakers judged the ablative ungrammatical, so that this alternation might be idiolectally restricted. For all other semantic roles, the mapping is deterministic. What you see in semantics is what you get in morphosyntax, and there is no way to change this. Depending on the analysis of the accusative marker =yang, one could argue that there is a alternation between 0̸-marked nominative and =yang-marked accusative for certain verbs. (47) Baapa derang=pe kubbong=ka hatthu pohong=Ø nya-poothong. father 3pl=poss garden=loc indf tree past-cut ‘My father cut a tree in their garden.’ (K051205nar05) (48) Ithukapang lorang=pe leher=yang kithang=0̸ athi-poothong. then 2pl=poss neck=acc 1pl irr-cut ‘Then we will cut your neck.’ (K051213nar06) An analysis involving the optionality of =yang depending on parameters such as animacy, topicality and affectedness seems to be the better solution, though.

4.2 Labile Verbs The issue of labile verbs is not directly relevant in Sri Lanka Malay. If we define a labile verb as a verb which, besides its canonical ‘transitive’ realization, occurs in ‘intransitive’ frames as well, then all Sri Lanka Malay verbs which have a ‘transitive’ realization are labile verbs, as they will always also allow for an ‘intransitive’ realization.6 To be more precise, they will always allow for two ‘intransitive’ realizations: one with the agent dropped and one with the patient dropped. Given that =yang marking is optional, the sequence NP V can always have two interpretations: either the NP refers to the actor or to the undergoer. Pragmatic considerations, as well as our knowledge of the world will of course always be suggestive of one interpretation or the other, but on morphosyntactic grounds alone, the linking of the only NP of the NP V sequence to a semantic role cannot be accomplished.

6 Note that while all two-argument verbs can surface with one argument only, the reverse is not true: not all one-argument verbs can add an additional argument without further measures being taken.

Case frames in Sri Lanka Malay

1003

4.3 Coded alternations: Verb Sri Lanka Malay is not a language with strict subcategorization for the number of arguments a verb takes. It is always possible to drop arguments, and in most cases, it is possible to add arguments without further ado. There are a number of additional morphosyntactic devices which are related to the number of arguments a verb takes and which will be discussed in this section. These devices comprise the causative suffix, the prefix kasi- and the postverbal ‘vector verbs’ kaasi ‘give’ and ambel ‘take’.

4.3.1 Causative -king The most trivial case is the causative morpheme -king with the allomorph -kang. This morpheme attaches to a verb or an adjective and introduces an additional causer, who causes the initial actor to take part in the event denoted by the predicate. Examples (49)–(51) show this for an adjective, an ‘intransitive’ verb and a ‘transitive’ verb. The causer is always in the nominative (with the usual exceptions for modals and institutional actors), while the causee is either zero-marked, or marked with the accusative marker =yang. (49) Itthuka asà-thaaro, itthu=yang arà-panas-king. dist=loc cp-put dist=acc non.past-hot-caus ‘Having put (it) there, you heat it.’ (B060115rcp02) (50) Inni=ka inni daalang=ka kithang aayer masà-mlidi-king. prox=loc prox inside=loc 1pl water must=boil-caus ‘On this, inside this, we must boil water / bring the water to a boil.’ (B060115rcp02) (51) De laaye hathu nigiri=nang anà-baapi, buunung-king=nang. 3s.impolite other indf country=dat past-bring kill-caus=dat ‘They brought him to another country to have him executed.’ (K051206nar02)

4.3.2 Prefix kasiThere are two, possibly three, verbs in Sri Lanka Malay which can be argued to involve a valency-changing derivation with the element kasi, related to kaasi ‘give’ (Nordhoff 2012). These are kasithaau ‘inform’ from thaau ‘know’, kasikaaving ‘to give in marriage’ from kaaving ‘marry’, and kasikìnnal ‘introduce’ from kìnnal

1004

Sebastian Nordhoff

‘know, be acquainted’. In all the three cases, a recipient is added as a new semantic role. The two other roles remain or become actor and theme.7 Recipients are not among the roles the verbs without kasi assign. The morpheme kasi- is the mechanism closest to a valency-changing device Sri Lanka Malay has, but, as stated above, it is limited to a handful of verbs and probably lexicalized.

4.3.3 Vector verb kaasi At first sight, there is another process which seems to add a beneficiary, the vector verb kaasi ‘give’. Note that this form is obviously etymologically related to the form just discussed, but has a long vowel and occurs after the verb it modifies, not before. Combinations of V+kaasi are used to mark beneficiaries as in (52), where the fact of receiving explanations is highlighted as beneficial. Related constructions in Sinhala and Tamil have been termed ‘alterbenefactive’ (Lehmann 1989: 227). (52) Kithang=pe ini younger generation=nang=jo konnyong masà-biilang 1pl=poss prox younger generation=dat=emph few must-say kaasi, masà-aajar. give must-teach ‘It is to the younger generation that we must explain it, must teach it to.’ The question which arises is whether kaasi changes the valency structure of the verb it modifies. Already example (52) suggests that this is unlikely. The verb biilang ‘say’ typically takes three arguments, a speaker, a message, and an addressee. This fact does not change through the addition of kaasi in (52). This is even clearer in (53). Kaasi could be thought to increase valency, but example (53) shows that it does in fact not have any influence on the number of arguments. (53) Itthu muusing, [Islam igaama nya-aajar kaasi 0]̸ Jaapna Hindu teacher. dist time Islam religion past-teach give Jaffna Hindu teacher ‘At that time, those who taught Islamic religion were Hindu teachers from Jaffna.’ (K051213nar03) The verb aajar ‘teach’ has already three arguments before derivation, an agent, a theme, and a recipient/beneficiary. The same three roles are present after kaasi is 7 The ‘knower’ of the erstwhile thaau ‘know’ is either in the nominative or the dative, depending on idiolect. This is due to different exposure to the contact languages: Tamil uses the dative with ‘know’, while Sinhala uses the nominative. It is likely that the Sri Lanka Malay idiolectal differences find their explanations in the particular linguistic ecology an individual speaker evolves in.

Case frames in Sri Lanka Malay

1005

added, but the beneficial component is highlighted. It is impossible to claim that kaasi adds an argument to aajar, since the number of arguments does not change. Furthermore, we also note that the nature of arguments does not change: the semantic roles are exactly identical, with the proviso that the ambiguous status of recipient/beneficiary is definitely changed to beneficiary in the kaasi-clause. This semantic difference, however, is not discernible in case-marking since both recipients and beneficiaries are marked by the dative.8

4.3.4 Vector verb ambel Next to the vector verb kaasi ‘give’, we also find the vector verb ambel ‘take’, which is in a number of respects the opposite of kaasi. Kaasi highlights the alterbenefactive aspect of a certain event; ambel highlights the self-beneficial aspect.9 An example is (54), where the normally neutral action of catching is marked as self-benefactive. (54) British government Malaysia Indonesia ini nigiri pada samma British government Malaysia Indonesia prox country pl all anà-peegang ambel. past-catch take ‘The British government captured Malaysia, Indonesia, all these countries.’ (K051213nar06, K081104eli06) Compare this to (55), where the action is not beneficial, and ambel is absent. (55) 0̸ Heart attack asà-peegang, baapa=le su-niinggal. heart attack cp-catch father=addit past-die ‘(My father) got a heart attack and died as well.’ (K051205nar05, K081104eli06) Note that neither the number of arguments nor their morphosyntactic coding change with regard to the sentence without ambel, showing that self-benefactive ambel triggers no morphosyntactic changes.

8 Kaasi can also be combined with a one-place (‘sing’) or two-place verb (‘build’) to indicate a beneficiary (‘sing for’, ‘build for’), but the same could be accomplished by simply adding a =nang beneficiary without kaasi. It is generally possible to add a beneficiary to any verb without further measures (also see Section 4.4). The addition of the beneficiary is thus not the result of the presence of kaasi, but rather independent of it. 9 There are other usages of ambel which will not be discussed here. The reader is referred to Nordhoff (2009) and Nordhoff (2012).

1006

Sebastian Nordhoff

The last process, related to self-benefactive ambel, is reflexive ambel. This is often found together with the emphatic clitic =jo, although this is optional. (56) Incayang incayang=yang(=jo) su-buunung ambel. 3s.polite 3s.polite=acc=emph past-kill take ‘He killed himself.’ (K081106eli01) The question is again whether this changes the number or nature of arguments. The answer is again negative. Even when ambel or =jo is present, it is possible to express both the agent and the patient, as shown in (56). This is normally not done for obvious pragmatic reasons of economy. We can say that the sentence in (56) has an agent NP and a patient NP just like its non-reflexive counterpart, e.g. (57). (57) Kaake baapa=yang su-buunung. grandad father=acc past-kill ‘His grandad killed his father.’ (K081103eli04) These NPs in (56) and (57) are marked in the same way, so that there is no reason to suspect morphosyntactic reflexes of valency structure. The best candidates for valency-changing operations, kasi-V, V kaasi, and V ambel were discussed in this section. The first of these operations might be seen as valency-changing, but is restricted to three lexemes. The remaining two looked promising at first sight, but had to be discarded after closer scrutiny because they affected neither number nor nature of arguments.

4.4 Other alternations Sri Lanka Malay can drop any NP if it is recoverable from context. This means that it is perfectly normal for a three-place verb like kaasi ‘give’ to be uttered in isolation as in (58). (58) su-kaasi past-give ‘X gave Y to Z (and, you, my dear addressee, will surely be smart enough to figure out who X, Y, and Z are).’ It is clear that there are no slots which obligatorily have to be filled in Sri Lanka Malay. As such, the question arises how a sentence like (58) can be distinguished from a sentence like (59).

Case frames in Sri Lanka Malay

1007

(59) su-uujang past-rain ‘(it) rained’ The overt morphosyntactic structure is exactly parallel. The two sentences (58) and (59) can however be distinguished by the possibility to add new referents. This is patently simple for (58), see (60), but nigh impossible for (59), barring very contrived contexts. (60) Tony kaake=nang ini car=yang su-kaasi. Tony grandfather=dat prox car=acc past-give ‘Tony gave this car to Grandfather.’ While it is very difficult to add more referents to uujang ‘rain’, similar problems are not found with kaasi ‘give’. The question arises whether the (im)possibility to add new referents might not be a better cue to case frame membership than the (im)possibility to leave out certain referents. Taking a look at a prototypical transitive predicate like poothong ‘cut’, it is immediately obvious that referents bearing the dative and the ablative can be added without problems. The normal transitive sentence in (61) can be made to accommodate four NPs. The resulting sentence is (62) and is as such then indistinguishable in argument structure from (46). (61) Tony ini daaging=yang arà-poothong. Tony prox meat=acc nonpast-cut ‘Tony cuts this meat.’ (62) Tony piiso=dering ini daaging=yang kaake=nang arà-poothong. Tony knife=abl prox meat=acc grandfather=dat nonpast-cut ‘Tony cuts this meat for his grandfather with a knife.’ NPs referring to space or time can also be added to the sentence without problems, a fact which is little surprising. All predicates can accommodate a phrase like Kluumbu=ka ‘in Colombo’ or soore=ka ‘in the evening’. Out of the case markers surveyed in this chapter (=yang ‘acc’, =nang ‘dat’, =ka ‘loc’, =dering ‘abl’), there is only one which cannot be added to a predicate without further ado: the accusative marker =yang. Augmenting a predication with beneficiaries, instruments, locations and points in time is no problem, but adding an extra patient is impossible. This clear separation, and the absence of similar patterns elsewhere, suggests that Sri Lanka Malay verbs can be divided into two classes: those which can take an argument with =yang (but might drop it), and those which never can take such an argument. This is the main difference made in the Sri Lanka Malay lexicon as far as verbs go. These two classes will now be surveyed in turn.

1008

Sebastian Nordhoff

5 Verb classes in Sri Lanka Malay 5.1 Verbs taking =yang These are verbs which have the semantic component of affectedness in them, and which thus assign a patient. Patient-taking verbs can be intransitive, like thìnggalam ‘sink’, transitive like poothong ‘cut’, ditransitive like kaasi ‘give’ or tritransitive like kiiring ‘send’.

5.2 Verbs not taking =yang These are verbs which either do not have any connection with affectedness at all in their meaning (dhaathang ‘come’), or whose affected argument takes dative marking, suggesting an interpretation as recipient.10 This set comprises atransitive verbs like uujang ‘rain’, intransitive verbs like laari ‘run’, and transitive verbs like maaki ‘scold’. Verbs with three or more places typically involve a patient or a theme and hence show a possibility to include =yang-phrases. Themes are normally marked by 0,̸ but by choosing a referent ranging high on the scales of animacy, definiteness, individuatedness, and topicality, =yang-marking becomes an option. Verbs not taking =yang can undergo derivation by the causativizer -king and then allow the use of =yang.

6 Conclusion Grammatical information which is non-predictable and arbitrary is stored in the lexicon. The English lexicon has to store that eat can have one or two arguments, but that devour has exactly two. The intransitive option is not available for the latter verb. What is there to store in the Sri Lanka Malay lexicon, and in how far does it inform our theorizing about valency? It is clear that the minimal number of argument slots is not an important category in Sri Lanka Malay. In discourse, the number of arguments is quite low due to the frequent dropping of whole NPs. Any verb in Sri Lanka Malay can be used with 0 realized arguments without leading to ungrammaticality. The maximum number of arguments a Sri Lanka Malay verb can take is not important either. In Sri Lanka Malay, NPs taking the dative =nang or the ablative =dering can always be added. This shows that the number of argu-

10 Sinhala aligns with Sri Lanka Malay in this respect. See Garusinghe (1962: 36, 92) and (Gair 1991: 18 f.). The behavior of Tamil dialects in Sri Lanka is understudied and cannot be commented on here.

Case frames in Sri Lanka Malay

1009

ments any verb can take is always between 0 and at least 3.11 The distinction between intransitive, transitive and ditransitive is thus not an important one in Sri Lanka Malay, and in fact, the presentation above structured into 0-place up to 4-place arguments is more an expository device than a solid fact about the language. The assignment of verbs to any of those four categories was based on their translational equivalents in English; I am not aware of any morphosyntactic test in Sri Lanka Malay to arrive at a similar classification. The number of arguments is thus not what has to be stored in the lexicon, since this is basically only constrained by semantic plausibility, which is found in knowledge of the world, or the ‘encyclopedia’. What must be stored in the lexicon, however, is the nature of arguments in some cases. Very often, the arguments simply match what semantics would predict (dative for recipient, ablative for instrument), but especially for the =yang-cases, there is a certain degree of arbitrariness. For instance, the transitive verb giigith ‘bite’ takes a =yang-marked NP, while the semantically similar thiikam ‘stab’ takes an NP marked for dative by =nang. The question is whether this information needs to be stored in the lexicon as a morphosyntactic irregularity, or whether a purely semantic analysis would not be possible as well. Above, I have argued that the special dative marking for the three transitive verbs theembak ‘shoot’, puukul ‘hit’, and thiikam ‘stab’ is in fact not a case of lexical irregularity, but instead reflects the semantic component of transfer of a weapon towards a point of impact, a goal. It is true that these words behave differently from other verbs with a high degree of affectedness, but a straightforward explanation is available.

Appendix: Summary table #

Meaning label

Verb form

Coding frame schema

General NP addition

General NP drop

 69

RAIN

uujang

V

–

–

  5

SMELL

vaanging liiyath

1V

m

+

 47

COUGH

baathok

1V

m

+

 49

RUN

laari

1V

m

+

Reciprocal alternation

11 We exclude the meteorological verb uujang ‘rain’ for the moment, where one could think of a context with an instrument and a beneficiary, but will need a lot of time to convince speakers of the acceptability of this context.

1010

Sebastian Nordhoff

#

Meaning label

Verb form

Coding frame schema

General NP addition

General NP drop

 52

JUMP

lompath

1V

m

+

 53

SING

nyaanyi

1V

m

+

 58

SCREAM

bitharak

1V

m

+

 59

FEEL PAIN

saakith

1V

m

+

 61

DIE

maathi

1V

m

+

 61

DIE

mniinggal

1V

m

+

 62

plAY

maayeng

1V

m

+

 68

BE DRY

kìrring

1V

m

–

 70

BE A HUNTER

oorang neembak

1V

–

m

 80

BOIL

mdiidi

1V

m

+

 83

CRY

naangis

1V

m

–

 84

FALL

jaatho

1V

m

+

 66

SINK

thìnggalam

1-acc V

m

+

 60

FEEL COLD

diinging

1-dat V

+

+

 63

BE SAD

suusa

1-dat V

m

+

 64

BE HUNGRY

laapar

1-dat V

m

–

 67

BURN

appi aràpeegang

1-dat V

m

+

 82

BE ILL

thàràsìggar

1-dat V

m

–

 55

LEAVE

luppas

12V

m

+

 16

FOLLOW

pii (2)

1 2-abl V

m

+

 55

LEAVE

pii (2)

1 2-abl V

m

+

  1

EAT

maakang

1 2-acc V

m

+

  2

HUG

pullok

1 2-acc V

m

+

  4

SEE

kuthumung

1 2-acc V

m

+

  7

FRIGHTEN

thakuthking

1 2-acc V

m

+

 10

THINK

thaksir

1 2-acc V

m

–

 10

THINK

iingath

1 2-acc V

m

–

 11

SEARCH FOR

caari

1 2-acc V

m

–

 12

WASH

cuuci

1 2-acc V

m

+

 13

DRESS

pakeking

1 2-acc V

m

+

Reciprocal alternation

Case frames in Sri Lanka Malay

#

Meaning label

Verb form

Coding frame schema

General NP addition

General NP drop

 30

CUT

poothong

1 2-acc V

m

+

 33

PEEL

kuupas

1 2-acc V

m

+

 73

DIG

gaali

1 2-acc V

m

+

 79

COOK

maasak

1 2-acc V

m

+

126

BOIL (tr)

rubbus

1 2-acc V

m

+

  6

FEAR

thaakuth

1 2-dat V

m

+

  8

LIKE

suuka

1 2-dat V

m

+

 15

HELP

banthu

1 2-dat V

m

+

 20

SHOUT AT

kaaluth

1 2-dat V

m

+

 48

CLIMB

naayek

1 2-dat V

m

+

 54

GO

pii (1)

1 2-dat V

m

+

141

LAUGH ABOUT

thàthaava

1 2+atthas V

m

–

 50

SIT

duuduk

1 LOC2 V

m

+

 51

SIT DOWN

duuduk

1 LOC2 V

m

+

 56

LIVE

duuduk

1 LOC2 V

m

+

 24

BUILD

ikkath (1)

1 2-acc 3-abl V

m

+

 25

BREAK

picaking

1 2-acc 3-abl V

m

+

 26

KILL

buunung

1 2-acc 3-abl V

m

+

 26

KILL

mathiking

1 2-acc 3-abl V

m

+

 27

BEAT

puukul

1 2-acc 3-abl V

m

+

 28

HIT

puukul

1 2-acc 3-abl V

m

+

 31

TAKE

ambel

1 2-acc 3-abl V

m

–

 32

TEAR

soovek(kang)

1 2-acc 3-abl V

m

+

 34

HIDE

sbuuniking

1 2-acc 3-abl V

m

+

 43

COVER

thuuthup

1 2-acc 3-abl V

m

+

 44

FILL

punuking

1 2-acc 3-abl V

m

+

 71

GRIND

giiling

1 2-acc 3-abl V

m

+

 72

WIPE

saapu

1 2-acc 3-abl V

+

m

 76

STEAL

cuuri

1 2-acc 3-abl V

m

+

 85

MAKE

kijja

1 2-acc 3-abl V

+

m

1011

Reciprocal alternation

1012

Sebastian Nordhoff

#

Meaning label

Verb form

Coding frame schema

General NP addition

General NP drop

 35

SHOW

cunjiking

1 2-acc 3-dat V

m

+

 36

GIVE

kaasi

1 2-acc 3-dat V

m

+

 37

SEND

kiiring

1 2-acc 3-dat V

m

+

 39

THROW

buwang

1 2-acc 3-dat V

+

+

 40

TIE

ikkath (2)

1 2-acc 3-dat V

m

+

 74

PUSH

thoolak

1 2-acc 3-dat V

m

+

 75

BRING

baa(va)

1 2-acc 3-dat V

m

+

 41

PUT

thaaro

1 2-acc LOC3 V

m

+

 42

POUR

thuuwang

1 2-acc LOC3 V

m

+

 44

FILL

thuuwang

1 2-acc LOC3 V

m

+

 45

LOAD

thaaro

1 2-acc LOC3 V

m

+

 38

CARRY

angkath baapi

1 2-dat 3 V

m

+

 77

TEACH

aajar

1 2-dat 3 V

m

+

 18

TALK

oomong

1 2-dat 3-abl V

m

+

 29

TOUCH

thìkkam

1 2-dat 3-abl V

m

+

 23

NAME

naama thaaro

1 2-dat 3+katha V

+

+

 21

TELL

biilang

1 2-dat UTT3 V

m

+

 22

SAY

biilang

1 2-dat UTT3 V

m

+

 19

ASK FOR

mintha

1 2-loc 3-acc V

m

+

  3

LOOK AT

diyath

1-dat 2-acc V

m

+

  9

KNOW

thaau

1-dat 2-acc V

m

+

 17

MEET

caanda

1-dat 2-acc V

m

+

 78

HEAR

dìnngar

1-dat 2-acc V

m

+

 87

WANT

maau

1-dat 2-acc V

m

+

 86

GET

daapath

1-dat 2-loc 3-acc V

m

+

Legend: + occurs regularly; m occurs marginally; − occurs never; _ no data

Reciprocal alternation

+

Case frames in Sri Lanka Malay

1013

Abbreviations addit cp invol nonv simil

additive conjunctive participle involitive nonverbal similative

References Adelaar, K. Alexander. 1991. Some notes on the origin of Sri Lankan Malay. In H. Steinhauer (ed.), Papers in Austronesian Linguistics, 23–37. Canberra: The Australian National University. Gair, James W. 1991. Subjects, case and INFL in Sinhala. In Manindra Verma & K. P. Mohanan (eds.), Experiencer Subjects in South Asian Languages, 13–41. Palo Alto: Center for the Study of Language and Information. [Reprinted in Gair (1998), 65–86.] Gair, James W. 2003. Sinhala. In George Cardona & Danesh Jain (eds.), The Indo-Aryan Languages, 766–818. London: Routledge. Garusinghe, Dayartne. 1962. Sinhala, the Spoken Idiom. München: Max Hueber. Hengeveld, Kees. 2011. Transparency in functional discourse grammar. Linguistics in Amsterdam 4(2). 1–21. Lehmann, Thomas. 1989. A Grammar of Modern Tamil. Pondicherry: Pondicherry Institute of Linguistics and Culture. Nordhoff, Sebastian. 2009. A grammar of Upcountry Sri Lanka Malay. Ph.D. thesis, University of Amsterdam. Nordhoff, Sebastian. 2011a. Having come to be a copula in Sri Lanka Malay – an unusual grammaticalization path. Folia Linguistica 45(1). 103–126. Nordhoff, Sebastian. 2011b. Transparency in Sri Lanka Malay. Linguistics in Amsterdam, 4(2). 96– 110. Nordhoff, Sebastian. 2012. Multi-verb constructions in Sri Lanka Malay. Journal of Pidgin and Creole Languages 27(2). 303–343. Nordhoff, Sebastian. 2013. Jack-of-all-trades: The Sri Lankan Malay property word. In Eva van Lier & Jan Rijkhoff (eds.), Flexible Word Classes: A Typological Study of Underspecified Partsof-Speech. Oxford: OUP. (Paper presented at the Workshop on flexible languages, Amsterdam.) Paauw, Scott. H. 2004. A historical analysis of the lexical sources of Sri Lanka Malay. M.A. thesis, York University.

Claire Moyse-Faurie

25 Valency classes in Xârâcùù (New Caledonia) 1 Introduction This article provides an overview of valency classes and valency alternations in Xârâcùù1 and tries to describe the main developments which are still perceivable and which explain the great variety and some of the specificities found in Xârâcùù verb classes. Xârâcùù is spoken by about 6000 people living in the Canala and Thio areas of the Mainland of New Caledonia, and belongs to the Southern subgroup of New Caledonian, itself one of the main subgroups of Oceanic, which is part of the Austronesian language family. The article is organized as follows: Section 2 summarizes general information on morphosyntactic aspects of Xârâcùù relevant to valency, in contrast to those of other New Caledonian/Kanak languages2 and to Proto Oceanic. Section 3 describes Xârâcùù’s major verb classes, and discusses the sample obtained from the database questionnaire. Section 4 deals with uncoded argument alternations, while section 5 describes valency-changing devices. In section 6, verb compounds stemming from former serial verb constructions are presented. Section 7 raises the problem of distinguishing between adjuncts and oblique objects in a language which does not have any anaphoric verbs. A final conclusion summarizes the major results.

2 Basics of Xârâcùù morphosyntax According to Pawley & Reid (1979) and Ross (2004), the basic word order reconstructed for Proto Oceanic is S sVo O (s and o are the obligatorily person forms that index subject and object arguments and are coreferential with the optional lexical subject (S) and object (O), respectively). Moyse-Faurie & Ozanne-Rivierre (1983),

1 The data, unless otherwise indicated, comes from my fieldnotes, from my grammar (Moyse-Faurie 1995) and dictionary (the latter written with M.-A. Néchérö-Jorédié, a native Xârâcùù speaker). I wish to thank Martin Haspelmath and the editors of this volume for their fruitful comments on an earlier version of this article. 2 ‘Kanak’ has become the usual and politically correct term to designate the autochtonous inhabitants of New Caledonia and their vernacular languages. It is invariable, even in French. More informations on these languages can be found on the Académie des Langues Kanak website (http:// www.alk.gouv.nc).

1016

Claire Moyse-Faurie

however, reconstructed sVo O smS word order for Proto New Caledonian (sm is the subject marker preceding the optional lexical subject). In Xârâcùù, the basic word order changed to SVO3, without any cross-referencing index (1a). The subject, however, can still occur after the verb phrase, if it is introduced by the subject marker ngê. 4 In this configuration, a coreferential argument index must occur in front of the verb (as shown in 1b), and the postposed subject is preceded by a short pause, and has a falling intonation contour, a typical case of topicalization by right dislocation. In Xârâcùù, this type of subject topicalization is available for mono-, bi- or trivalent verbs. Preverbal subject indexes are also compulsory when the subject is topicalized by fronting (1c). (1) a. Mûduè-nâ pwâxwâ néxä kèè-xê. young.brother-1sg very know nmlz-swim ‘My little brother knows how to swim very well.’ b. Rè pwâxwâ néxä kèè-xê, ngê mûduè-nâ. 3sg very know nmlz-swim sm young.brother-1sg ‘He knows very well how to swim, my little brother.’ c. Mûduè-nâ, rè pwâxwâ néxä kèè-xê. young.brother-1sg 3sg very know nmlz-swim ‘My little brother, he knows how to swim very well.’ In all other grammatical contexts, there is no argument indexing on the verb – no cross referencing of arguments on verbs –, although this feature is frequent in Oceanic languages (Lynch et al. 2002). Xârâcùù has also lost several other major Oceanic characteristics. Transitivity, for example, is not formally marked on the verb, contrasting in this feature with most New Caledonian languages, and Oceanic languages more generally. Furthermore, there is no verbal morphology depending on the grammatical status or the degree of animacy of the object argument as is found, for example, in the Loyalty Islands languages. Thus, Xârâcùù has lost the productive applicative and transitivizing Proto Oceanic suffixes; only one reflex of these suffixes has been retained, the -ri suffix (cf. § 5.2.1), which itself fossilized in

3 However, existential and non-existential clauses (i), as well as nominal clauses (ii) are still verb initial. (i)

Wâ siè aguu-söö. pfv non.exist sound-song ‘Songs have stopped.’ [Songs are no longer sung.]

(ii) Pepe-nâ taa rö. difference-1sg from 2sg ‘I am different from you.’ 4 The preposition ngê also introduces diverse oblique arguments and adjuncts (cf. § 3.2.3.5).

Valency classes in Xârâcùù (New Caledonia)

1017

a few cases. TAM-features do not interfere with alignment patterns in Xârâcùù, in contrast with the languages of the Loyalty Islands such as Drehu (Moyse-Faurie 1983) or Iaai (Ozanne-Rivierre 1976). On the other hand, Xârâcùù shares some characteristics with other Oceanic languages, such as those described by Margetts (2007), which directly concern this valency study: (i) There are few basic (i. e. non-derived) bivalent verbs, with two participants expressed as direct arguments of the verb, and no three-place predicates occurring with a direct argument strategy (i.e., trivalent verbs with three participants expressed as direct arguments of the verb). (ii) There is a productive causative derivation strategy, deriving bivalent verbs from monovalent verbs or from nominals; in most of the Oceanic languages, there are, however, very few cases of trivalent verbs derived by a causative strategy, in contrast to the lack of relevant restrictions in Xârâcùù. (iii) The ‘extended transitive’ configuration, in which the verb takes two direct arguments and a third participant as an oblique argument or adjunct, or one direct argument plus one oblique argument, is well attested. Margetts (2007: 88) and Margetts & Austin (2007: 400) point out difficulties in making a clear distinction between oblique arguments and adjuncts in these languages. We return to this crucial point in § 7. Xârâcùù has productive serial verb constructions, restricted to verbs with a shared subject; no morpheme is allowed to occur between two or more serialized verbs. This nuclear-layer verbal serialization5 gave (and still gives) rise to (i) prepositions, (ii) to verbal modifiers (directionals, aspectual or manner adverbs) and (iii) to verbal compounds. A question of terminology remains to be settled. We need to make a distinction between valency classes (= verb classes), which are relevant at the lexical level, and the constructions in which a verb may occur. Valency classes should be differentiated from constructions, since, due to possible alternations, one and the same verb may enter different types of constructions. Malchukov et al. (2010) among others use ‘ditransitive’ for constructions as well as for verb classes,6 though triva-

5 Nuclear serialization consists of several contiguous nuclei which share arguments (cf. Foley & Olson 1985: 32–34). 6 More precisely, Haspelmath (2005, 2011) and Malchukov et al. (2010) have a semantically-based definition of transitive and ditransitive constructions. Transitive clauses have a P and an A argument, with verbs such as ‘kill’, ‘break’, ‘cut’, and ditransitive clauses are built on verbs such as ‘give’, ‘buy’, ‘write’, and so on. Such ditransitive constructions involve three semantic arguments: an agent (A), a recipient or goal (R/G), and a theme (P/T). These semantic definitions suppose that the arguments referring to A, P or R are always marked in the same way, at least with the typical verbs belonging to the predefined semantic categories. Xârâcùù, like many Oceanic languages, does not have a single major transitive construction, since objects often refer to several other roles than

1018

Claire Moyse-Faurie

Tab. 1: Personal markers. indep.

subject

object

possessive

sing

1 2 3 3impers

gu gè niè

nâ ke rè/è êê

nâ/nû rö rè/è

nâ/-râ rö/-ö rè/-è

dual

1incl 1excl 2 3

ûrû ngôô göu nuu

ûrû/nû ngôô göu ru

rû ngôô göu ru

rû ngôô göu ru

plur

1incl 1excl 2 3

îrî ngêê wîrî nii

îrî/nî ngêê wîrî ri

rî ngêê wîrî ri

rî ngêê wîrî ri

lent verbs may occur in transitive and in intransitive constructions. Similarly, monovalent verbs may occur both in impersonal and in intransitive constructions, and bivalent verbs may occur in transitive and in intransitive constructions with only one argument expressed (either the subject or the object), or with different marking on the object, and so on and so forth. The bivalent class assignment is only due to the fact that it may occur with two (but no more than two) arguments. Consequently, I will use ‘avalent’, ‘monovalent’, ‘bivalent’ and ‘trivalent’ as terms for the verb classes (‘valency classes’ properly speaking), and ‘impersonal’, ‘intransitive’, ‘(mono)transitive’ and ‘ditransitive’ as terms for the constructions in which a verb occurs. Above are listed the 70 English verbs with their role frame as specified in the questionnaire, along with the Xârâcùù verbs, their basic case frame, and specific alternations (others than the general ones, such as the causative alternation) they allow. Practically all the verbs may be combined with the causative prefix (cf. § 5.2.2), and with one or several of the nominalizing prefixes. The object omission alternation is also very frequent (cf. § 4.3).

3 Valency classes Quite a few Xârâcùù verbs can enter different types of constructions, without derivation; each construction has correlates in terms of the semantic role of the argu-

patient, and patients themselves are often flagged diversely; very few bivalent verbs have two direct arguments, most of them have a direct argument and a prepositional one.

Valency classes in Xârâcùù (New Caledonia)

1019

ments. Such alternations will be considered after a brief presentation of each of the valency classes. There are no avalent verbs in Xârâcùù, but as we will see later (§ 4.1.1), some monovalent verbs may occur in an impersonal construction.

3.1 Monovalent verbs The main two subcategories of monovalent verbs match the actor subject argument and the undergoer subject argument distinction, with dynamic verbs in (2), and non-dynamic verbs in (3). In both cases, the subject may refer to an animate entity (2a and 3a) or to an inanimate one (2b and 3b). When causativized, the undergoer argument becomes the object of the derived verb, and an agent is added as subject argument (cf. § 5.2.2). Labile verbs, by contrast, do not have to be derived in order to undergo this semantic alternation (cf. § 4.1). (2) a. Mwêê-nâ xânî cuè tö nä.7 uncle-1sg often sit loc there ‘My uncle often sits there.’ b. Chaa pwê xuu kwiè wâ nä xwa rè. one unimportant little rain pfv ipfv fall ipfv ‘A small rain began to fall.’ (3) a. È pwârâ kèèrè nyî-ji. 3sg white as juice-breast ‘(S)he/it is as white as mother’s milk.’ b. Ääda niyaa. food bitter ‘The food is bitter.’ Monovalent verbs mainly denote positions or movements8 (cuè ‘sit’, tââ ‘stand’, pûxûrû ‘run’), natural events (xwa ‘fall (rain)’, mutu ‘sink’), physiological or com-

7 Xârâcùù does not have specific pronominal forms for different functions (such as object, subject or possessive), except for the tonic forms, and the 2nd singular subject form (cf. Table 1). Hence, in the glosses, I will only indicate the number and person of the pronominal forms. 8 These verbs, denoting body positions or movements, do not require complements, and most often occur in nuclear-layer verbal constructions: Ru wâ chaga cuè. 3du pfv warm.oneself.in.the.sun sit ‘They are warming themselves sitting in the sun.’

1020

Claire Moyse-Faurie

municative manifestations (nimè ‘blink’, mêrê ‘be hungry’, paii ‘feel pain’, têî ‘cry’, xô ‘sing’, xùpè ‘feel cold’, mwârâ ‘laugh’, ‘play’), inherent reciprocity (pùùtè ‘meet’), and qualities (mëtë ‘be dry’), since Xârâcùù only has a handful of adjectives.9 Some of the Xârâcùù monovalent verbs are compounds, such as chéxô (ché ‘say’ + xô ‘sing’) ‘cough’, paiimè (paii ‘be sick’ + mè ‘completely’) ‘die’, nârâyaa (nârâ ‘think’ + -yaa ‘badly’) ‘be sad’, cîchéé (cî- < cîîrî ‘fly’ + chéé ‘go down’) ‘jump’, etc.

3.2 Bivalent verbs 3.2.1 Bivalent verbs with direct objects Let us turn now to non-derived bivalent verbs and look at the meaning of their direct objects. Oceanic languages are well known for allowing direct objects of bivalent verbs referring not only to patient, but also to location, goal, or cause, as long as they have been derived by one of the transitive or applicative suffixes. Xârâcùù, however, has at least two dozen non-derived bivalent verbs, which are still very few compared to Indo-European languages. Such non-derived bivalent verbs taken from the list are bu ‘smell’, cee ‘peel’, ciiwi ‘help’, da or kê ‘eat’, jee ‘wipe’, nârâ ‘think’, néxä ‘know’, ngûrû ‘wash’, pè ‘take’, suè ‘put’, téé ‘see’, cee or coa ‘peel’, xwêê ‘pour’, xwèrii ‘want’, ‘like’; and also chavaa ‘pay attention to’, fùtù ‘to make fool of’, nèi ‘lay out’, wîjö ‘drink’, xöyö ‘marry’, xwata ‘hear’, tê ‘throw’, etc. Note that some of these verbs do not belong to the prototypical transitive situations. Besides, the Xârâcùù verbs translating the action ‘to kill’ (described as the prototypical verb of the transitive construction) are compounds which necessarily include the expression of the means or of the gesture involved in the killing. These verbs are thus semantically and morphologically complex. The non-derived bivalent verbs occur with direct objects, referring mainly to location (4), goal (5–6) or patient (7–8) participants: Location (4) Siibù cura bwasituu rè sêgè. rat creep heap poss stone ‘The rat is creeping away under the stones.’

9 Xârâcùù has no auxiliary verbs. Qualities are expressed by adjectival verbs, which behave as stative verbs but can also directly determine nouns, with the exception of six adjectives which specify nouns but cannot occur as predicates, such as dö ‘true, genuine’ (dö ku ‘yam of value’) or mwîî ‘important’ (mwîî aaxa ‘big chief’), all belonging to the main semantic domains (age, size, value) listed by Dixon (1977).

Valency classes in Xârâcùù (New Caledonia)

1021

Goal (5) Nâ nä nârâ rè kamè-nâ. 1sg ipfv think ipfv home- 1sg ‘I am homesick.’ (6) Gèè téé kèè-kê rè nè. grandmother look.at nmlz-burn poss fire ‘The grandmother is looking at the burning fire.’ Patient (7) Nèxuu coa ku mê kumwara êrê-nuu. girl peel yam and sweet.potato contents-bounia ‘The girl is peeling yam and sweet potatoes for the bounia.’10 (8) Êê nää chûrû mê bikörö rè pîî-köfi. 3sg.ipers pst.prog grill and turn+crush ipfv grain-coffee ‘We used to grill and grind coffee beans.’ However, transitive constructions with a direct object are mostly attested with predicates resulting from compounding, such as fèèté ‘follow’, fîda ‘hit’, fîgai ‘beat’, jikörö ‘break’, piicè ‘search for’, sakai ‘touch’, satee ‘hug’, söömè ‘kill’, söpuru ‘cut’, xâpârî ‘see’, and such as bikörö in (8), or from a causative derivation, as in fabata ‘frighten’, famwâmwâ ‘dress’, faxûda ‘fill’. I will return to verb compounding in § 6, and to causativation in § 5.2.2.

3.2.2 Reflexive verbs A few truly reflexive verbs (reflexivum tantum verbs) only occur in a superficially transitive construction, in which the object can only be a pronominal which must be coreferential with the subject: pètù or pètoa ‘boast’, pitèri ‘roll on the floor’; and the compound verbs bachëe ‘be unsuccessful (of a speech)’, bagwéré ‘be successful (of a speech)’ and gwébasùù ‘bump, bounce along’. The meaning of these verbs is either reflexive as in (9) and (10), or inchoative (11). (9)

Ke pètù rö tiwâ panèè-rö. 2sg boast 2sg loc father-2sg ‘You think you’re your father?’ (Lit. ‘You are boasting up to your father.’)

10 Bounia is a traditional dish with tubers, meat or fish, and coconut milk, wrapped in banana leaves and cooked in the earth oven.

1022

Claire Moyse-Faurie

(10) Nâ pitèri nâ. 1sg roll 1sg ‘I am rolling (on the floor).’ (11) Tèpe bachëe è. talk miss 3sg ‘The talk was unsuccessful.’

3.2.3 Bivalent verbs with an oblique object Among bivalent verbs which occur with an oblique object, one must distinguish between verbs which require such an oblique argument, and verbs which simply allow it. Oblique arguments are introduced by various prepositions, most of them resulting from the grammaticalization of the second verb of former serial verb constructions. Some of these prepositions have other uses than introducing oblique arguments. I already mentioned the difficulty in differentiating adjuncts and obliques in a language without anaphoric/auxiliary verb constructions (cf. note 9). Since there are no such verbs in Xârâcùù that would allow adjuncts to be moved out into a clause, we will see if other criteria can be put forward in order to differentiate adjuncts from oblique arguments (see § 7). I will first present the most frequent prepositions that may introduce oblique objects of bivalent verbs, mentioning their origin11 and different uses, along with examples of bivalent oblique-object verbs. Xârâcùù makes fine coding distinctions among verb classes due to the availability of different oblique prepositions, each with a specific semantic role and various functions (cf. Table 2). The main prepositions which may introduce oblique arguments (but often also have other functions) are presented below. Before examining in detail each of these prepositions, I must first consider the following point. As we will see, several of these prepositions can either be expressed in front of the noun phrase they introduce, or be included in the verb phrase, separated from their argument by another argument, a tense marker or an adverb. The question of their status is then raised:12 when included in the verb phrase, are these morphemes applicative suffixes? I would answer no for the following reasons. First, the inclusion into the verb phrase is always optional. Second, there is no difference in meaning. Third, when topicalized, the prepositional argu-

11 Some of these prepositions are good examples of grammaticalization process well-known worldwide, such as the attributive/recipient/beneficiary preposition xù, formally identical to the verb ‘give’. 12 Both Martin Haspelmath and another anonymous reviewer were skeptical about my analysis concerning the free word order allowed for some prepositions, and rather see them as applicatives when they are included into the verb phrase. I still opt for the cliticization explanation, and try to provide more arguments in favor of it.

Valency classes in Xârâcùù (New Caledonia)

1023

Tab. 2: The bare prepositions. Verb-like prepositions nârâ for taa off tara goal xù ben

(< nârâ ‘think’) ‘in order to’, Lit. ‘thinking at’ (–) (< witaa ‘throw away’) disassociative, non-beneficiary (+); also adverb (< tara ‘see’) ‘towards’ (+) (< xù ‘give’) recipient/beneficiary (+)

Adverb-like prepositions cè purp tùù/-dùù about

also adverb (+) ‘concerning’ (pertentive) (+)

Noun-like prepositions êê-/ngêê-/nèxêê ngê ins wâ at

(< ‘belongings’) ben recipient/beneficiary (–) oblique/instrumental (+); sm subject marker (–); during temporal adjunct (–) (wâ- ‘inner’) oblique (+), locative adjunct (–)

Purely locative prepositions (loc) static dynamic

tö/rö ‘at’ (–), xû ‘on’ (–) ti ‘towards’ (–), tù ‘towards the inside’ (–), kè ‘from’ (ablative) (–)

Only the bare prepositions discussed in this article are listed. (–) cannot be separated from its complement (+) can cliticize to the verb, and hence, be separated from its complement

ments keep their prepositions, except for the subjects, as in example (1c). In Xârâcùù, the optional inclusion of prepositions into the verb phrase is merely a morphological cliticization process. None of these prepositions have a genetic affiliation with the Proto Oceanic applicative suffixes, except, putatively, ngê (see note 16 below). When they originate in a verb (xù, tara, nârâ, taa), the most likely explanation of their optional inclusion into the verb phrase is that they keep the position they had in a serial pattern, before they were grammaticalized into prepositions. If these morphemes were applicative suffixes, they could occur even without an overt argument, and this is never the case in Xârâcùù. Another hypothesis would be to explain the cliticization of the Xârâcùù prepositions as a change from adpositions to adverbs, similar to the one described by Craig & Hale (1988) in some Amerindian languages, in which ‘relational adverbs’ come from postpositions. This explanation is not adequate here, since, as I already said, the preposition can never occur without its argument, and conveys the same meaning whatever its position. When the preposition is included into the verb phrase, the focus is on the noun phrase it is related to, whereas when the preposition occurs with the noun phrase, the focus is on the event described by the verb. I agree that the focus on the noun phrase when the preposition is cliticized is

1024

Claire Moyse-Faurie

reminiscent of the orientation towards a specific object (generally a non-patient one) induced by applicatives, but I still think that, synchronically, the prepositions that can cliticize are always plain prepositions, whatever their position.

3.2.3.1 Addressee/recipient/beneficiary preposition xù (< xù ‘give’) In his 1985 article, Lichtenberk mentions several Oceanic languages which have grammaticalized a benefactive/recipient marker from the verb ‘give’, reconstructed in Proto Oceanic as *pani/*pañi. The Xârâcùù verb xù ‘give’ is not cognate with the Proto Oceanic form, but has still undergone a similar development. In all the other New Caledonian languages, the benefactive marker has a different origin, being either identical to a possessive marker, or to a locative preposition. In Xârâcùù as well, beneficiaries may be expressed as possessors in specific noun phrase constructions (cf. § 4.2.2). Xù introduces oblique objects (mainly referring to the addressee) of some bivalent verbs: verbs of communication such as nîmö ‘tell a story’, ngââ ‘cry out’, yaaru ‘set riddles’, xa ‘speak’ and also baa ‘show oneself, appear’. With bivalent verbs, xù occurs immediately before the oblique object. (12) Nâ nîmö xù chaa kâmûrû. 1sg tell.story ben one man ‘I am telling the story of a clan to someone.’ (13) Anyââ xa xù xûûchî a. Mummy speak ben child deic ‘Mummy speaks to this child.’ The use of this marker in combination with trivalent verbs will be discussed later (cf. § 3.3.1.1). In ditransitive constructions, xù can be placed either within the verb phrase after the predicate, or immediately before its argument.

3.2.3.2 Disassociative/disattributive preposition taa (< witaa ‘throw away, take off’) Taa has a disassociative13 meaning, ‘off’, ‘away from’, the exact opposite meaning of the attributive/beneficiary preposition xù. 14 It may also convey a modal mean-

13 ‘Malefactive’ is the usual term to refer to non-beneficiary recipients. If this term seems adequate with verbs such as ‘rob’, it does not do justice to verbs such as ‘ask sth. to/from s.o.’, ‘buy sth. from s.o.’ or ‘receive sth. from s.o.’, and for this reason, we prefer to use either ‘disassociative’ or ‘nonbeneficiary’ to refer to such arguments. 14 German makes a similar distinction between beneficiary and detrimentary participants, with two different prepositions, für and gegen (Ich bin allergisch gegen Nüsse ‘I am allergic to nuts’). By

Valency classes in Xârâcùù (New Caledonia)

1025

ing, mainly as a postverb: ‘once and for all’, ‘getting rid of sth.’, denoting the termination of the action of the verb; besides, it is part of numerous verbal compounds, as a second element; finally, it occurs in comparative phrases, introducing the standard of comparison. A few bivalent verbs require a following object introduced by taa: könyi ‘avoid something’, mawâ ‘avoid a blow’, mwâ ‘avoid s.th. thrown’, mä ‘be discouraged with’, chörè ‘pass’, tecâ ‘leave someone’, mââî ‘precede’, bata ‘fear’. (14) Nâ könyi chaapu na taa loto. 1sg avoid suddenly pst off car ‘I got out of the way of the car.’ The relational noun pepe- ‘be different’ also requires an oblique argument introduced by taa: (15) Dui, va pepe-è taa Dapé. Dui really different-3sg off Dapé ‘Dui and Dapé are really different.’ (Lit. ‘Dui, his difference away from Dapé.’)

3.2.3.3 Goal preposition tara (< tara ‘see, note’) ‘towards’ As a verb (16a), tara means ‘to see, to note’: (16) a. Dèèri kè nä tara mîâdara. people from there see suffering ‘People from there know what suffering means.’ It may occur in serial verb constructions, as V2: (16) b. È mwâbéré tara kwéé-kârâmè rèè biri. 3sg look note shadow-eye poss+3sg move ‘He notices the shadow of his eyes moving.’ As a preposition, tara introduces oblique objects with a goal or directive meaning. In (16c), it occurs after the past tense marker which ends the verb phrase: (16) c. Nâ piaxô mwâmwaa na tara dèèri. 1sg whistle long pst goal people ‘I was whistling to the people (to get their attention).’ contrast, French and English use the same preposition (respectively to and à) for both types of arguments.

1026

Claire Moyse-Faurie

Other word orders are allowed, with inclusion of the preposition into the verb phrase: nâ piaxô mwâmwaa tara na dèèri as well as nâ piaxô tara na dèèri mwâmwaa are alternatives. Note that if tara immediately follows the verb, the temporal adverb mwâmwaa ‘long time’ is then peripheralized out of the verb phrase. The variety of positions allowed for prepositions such as tara, along with the fact that most prepositions may still occur as main verbs, shows the difficulty we encounter in differentiating serial verbs from verb + preposition sequences, not to mention the status of the complements the preposition introduces.

3.2.3.4 Pertentive 15 preposition tùù/-dùù This preposition means ‘about’, ‘concerning’, ‘with regard to’, ‘in connection with’, and belongs to the set of prepositions for which it is difficult to determine if they introduce adjuncts or oblique objects. It seems to have an oblique use after verbs such as bere ‘be angry’, pia ‘fight for’; mââ ‘struggle’, tèpe ‘speak about’; têî ‘cry for’, xati ‘quarrel for’. When following a verb with a final nasal vowel, the initial stop /t/ of the preposition changes to a prenasalized stop /d/ ([nd]) and the preposition cliticizes to the verb. The two following examples are semantically equivalent: (17) a. Nâ mââ na tùù döö. 1sg struggle pst about earth ‘I struggled for earth.’ b. Nâ mââ=dùù na döö. 1sg struggle=about pst earth ‘I struggled for earth.’ The last two prepositions mentioned in this chapter each have several uses. Their origin is uncertain, possibly nominal rather than verbal.

3.2.3.5 The multifunctional preposition ngê Ngê 16 introduces oblique objects (most often referring to the theme) of bivalent verbs such as jana ‘to trade’, sôôbö ‘fool with’, cîîrî ‘dispose, set out’, etc. (18) Ri xwi kaasé rè mwè […] nä ri wâ cîîrî ngê döu wânîî. 3pl make heap poss taro then 3pl pfv dispose ins thing all ‘They make heaps of taros [they make heaps of sea products …] and then they dispose of all these things.’ 15 The term ‘pertentive’ is the term used by Hagège (2010) to designate this non-spatial-temporal adposition meaning ‘about’, ‘concerning’, ‘with respect to’. 16 Ngê might have some syntactic affinity and phonological correspondences with the Tukang Besi oblique preposition kene ‘instrument’, itself cognate with the applicative ‘(co-)agent’ clitic =ngkene (Donohue 2001: 220).

Valency classes in Xârâcùù (New Caledonia)

1027

In this oblique use, ngê may be placed within the verb phrase or immediately before the oblique: (19) a. Papêê jana ngê na mwè. coll+woman trade ins pst taro ‘Women are trading [their] taros.’ b. Papêê jana na ngê mwè. coll+woman trade pst ins taro ‘Women are trading [their] taros.’ In (19a), the focus is on the object of the trading, that is the taros, while in (19b), the focus is on the action of trading.17 I have already mentioned the use of ngê introducing a postposed subject as in (1b). It also introduces instrumental (means) adjuncts as in (20a, 20b and 20c), and temporal adjuncts as in (20d). (20) a. Kâmûrû téé bwaakwè ngê béré. man look.at mountain ins binoculars ‘The man is looking at the mountain with binoculars.’ b. Ke xacè na chaa kâmûrû a ngê Dapé. 2sg call pst one man deic ins Dapé ‘You named this man Dapé.’ c. È xagèri nâ ngê chaa catùmê. 3sg welcome 1sg ins one gift ‘He makes me welcome [with a gift].’ d. Winâ nâ ngê chaa nèkwaaxiti mè nâ nârâ! leave 1sg during one week comp 1sg think ‘Let me think about it during this week!’ (Lit. ‘Leave me during a week that I think!’)

17 Other examples: a. È söömè ngê è chaa kwââ. 3sg kill ins 3sg one stick ‘He clubbed him, with a stick, to death.’ b. È söömè è ngê chaa kwââ. 3sg kill 3sg ins one stick ‘He clubbed him to death with a stick.’ In (a), the focus is on the instrument used to kill, while in (b), the focus is on the killing itself.

1028

Claire Moyse-Faurie

These different uses have correlates and restrictions concerning the position of the argument or adjunct, and the (in)separability of the preposition from the phrase it introduces. As an instrumental preposition or as an oblique object marker, ngê can be separated from its complement, while as a subject marker or as a temporal adjunct marker, it cannot.

3.2.3.6 Wâ ‘about’, ‘at’ Wâ introduces oblique arguments with verbs of emotions (ooro ‘rejoice at’, bere ‘be angry at’18, kwèche ‘be worried about’, etc.), verb of unpleasant attitudes (virè ‘wrong s. o.’, gîî ‘injure, damage’, chèfa ‘disobey’, mêmè ‘be jealous’, nêê ‘be fussy (about food)’, etc.), and a few verbs of communication (tèpe ‘speak about’, xanyê ‘insult’, jaxûju ‘make fun of’). A possible origin of the preposition wâ could be the relational noun wâ- ‘inner’, ‘form’ (cf. § 4.2.1). (21) a. È pia, è bere wâ dèèri. 3sg unkind 3sg angry at people ‘He is unkind, he gets angry at people.’ b. Ri gîî wâ mwââkè rè béé-nâ. 3pl damage at possessions poss friend-1sg ‘They are causing damage to my friend’s belongings.’ With verbs of communication, wâ may replace either the addressee preposition xù (22a) or the disassociative preposition taa (22b), as long as the oblique object refers to an animate entity: In (22a) and (22b), only a recipient is expressed, not affected in (22a) with the preposition xù, deeply affected in (22b) with the preposition wâ:

18 A verb such as bere ‘be angry’ can have different oblique arguments introduced by different prepositions. Compare (i), in which the oblique argument is the direct cause of the anger felt by the experiencer, with (ii) in which the oblique argument is introduced by the pertentive preposition tùù, being no longer the direct cause of the anger. Both prepositional phrases may occur in the same sentence (iii): (i)

È bere wâ kwêê-rè. 3sg angry at wife-3sg ‘He is angry at his wife.’

(ii) È bere tùù kwêê-rè. 3sg angry about wife-3sg ‘He is angry [at somebody] concerning his wife.’ (iii) È bere wâ Kamilo tùù kwêê wèi a. 3sg angry at Kamilo about wife person deic ‘He is angry at Kamilo concerning his [Kamilo’s] wife.’

Valency classes in Xârâcùù (New Caledonia)

1029

(22) a. Kamûrû xangââ na xù nèxuu. man shout pst ben girl ‘The man shouted to the girl.’ (to say hello to her) b. Kamûrû xangââ na wâ nèxuu. man shout pst at girl ‘The man shouted at the girl.’ (the man scolded her) Wâ also introduces different types of adjuncts, spatial or temporal locatives, either by itself (23) or combined with purely locative prepositions such as kè ‘(coming) from’, tö ‘in the direction of’, forming complex prepositions such as kèwâ ‘among’, töwâ ‘about’, etc. (23) Nâ miiri mè ke toa wâ chêêdè. 1sg refuse comp 2sg arrive at evening ‘I don’t want you to come in the evening.’

3.2.3.7 Nârâ ‘for’, ‘in order to’ (< nârâ ‘think’, ‘intend’) The verb nârâ ‘think’, ‘intend’ has grammaticalized into a preposition but still occurs as a verb with a clausal complement: (24) a. Nêmwâ nâ nä nârâ mè nâ pè rö xû kwâ. today 1sg ipfv intend comp 1sg take 2sg on boat ‘Today, I intend to take you on a boat trip.’ b. Bwèrè dèèri nârâ bare mè nä xwi rè chaa xwâî. some people think also comp ipfv make ipfv one road ‘Some people also think that a road will be made.’ As a preposition, nârâ means ‘for’, ‘in order to’, and only introduces nominal or nominalized complements expressing intention, aim or goal: (25) a. Nâ nä xwi rè chaa xiti nârâ kèè-fa-abaa röö xù 1sg ipfv make ipfv one feast for nmlz-caus-appear poss+2sg ben dèèri kè xûâ. people from village ‘I will organize a feast in order to introduce you to the people of the village.’ b. Pa xûûchî mwârâ ti nèkwétaa nârâ kèè-caa. coll child play at sea for nmlz-fish ‘Children are playing fishing by the sea.’

1030

Claire Moyse-Faurie

3.2.3.8 Cè ‘for the purpose of’ between adverb and preposition Finally, the morpheme cè ‘towards’, ‘for the purpose of’, ‘on purpose’ of unknown origin, has both the use of an adverb with bivalent verbs (26a), indicating that the event is done on purpose, and the use of a preposition, introducing arguments (26b) or adjuncts (26c and 26d). (26) a. Mè êê wâ nä chuè cè rè balôô döbwanä mwârâ nä fut 3sg.ipers pfv ipfv blow purp ipfv ball when play ipfv nââbu rè. begin ipfv ‘Someone will have blown up the ball on purpose when the game starts.’ b. Nâ târâ cè nîî-rö. 1sg ignore purp name-2sg ‘I don’t know your name.’ (Lit. ‘I am ignorant towards your name.’) c. È cemîâ mwâmwaa na cè dù rè kwââ-rè. 3sg suffer for.long pst purp price poss boat-3sg ‘He had to suffer a long time before paying for his boat.’ (Lit. ‘… for the price of his boat.’) d. Ri xwaé achaa xûjöu cè nènè-ri êrêcaa. 3pl go.all.over together reef purp food-poss.3pl sea.products ‘They are together walking all over the reef to find shells to eat.’ Cè also forms compounds with a few verbs, such as xa-cè ‘call’ (xa ‘speak’), pii-cè ‘search for’ (= pisi cè), xwi-cè ‘try’ (xwi ‘exist, make’).

3.3 Trivalent verbs Trivalent verbs with two unmarked objects are said to be rare in Oceanic languages: “There are a large number of Oceanic languages that have no ditransitive or extended transitive verbs at all. This tendency may be related to the typological characterization of Oceanic languages as preferring intransitives, which was suggested by Margetts (1999) following Nichols (1982, 1984a, 1984b)” (Margetts 2007: 124). Following Malchukov et al. (2010), I will try to distinguish the different types of trivalent verbs according to their alignment types. As already mentioned, there are no three-place predicate (direct-argument) strategies in Xârâcùù, but two oblique strategies (R-type oblique and T-type oblique) are well attested. According to Malchukov et al. (2010: 2) “the most typical ditransitive constructions contain a verb of physical transfer such as ‘give’, ‘lend’, ‘hand’, ‘sell’, ‘return’ […], some verbs denoting a mental transfer such as ‘show’ or ‘tell’, […] and less central transfer

Valency classes in Xârâcùù (New Caledonia)

1031

verbs such as ‘offer’, ‘bequeath’ and ‘promise’”. Xârâcùù has such verbs of physical or mental transfer occurring in ditransitive constructions, with one or the two object arguments expressed as obliques.

3.3.1 One oblique argument in addition to the direct object Indirective alignment, with Theme(T) = Patient(P) as direct arguments, and the Recipient(R) introduced by a preposition (T=P vs. xù/taa+R) is the most common pattern for this type of trivalent verbs. The word order is most often V T xù/taaR, but whatever the word order and the syntactic category of the recipient, the alignment does not change.

3.3.1.1 Beneficiary/recipient argument The recipient is always an animate being. Whatever its syntactic category may be, it may be introduced by the preposition xù. The following verbs can occur with a direct object (the theme) and an oblique object (the recipient): xù ‘give’, numârâ ‘offer’, ‘give away’, su ‘write’, suè ‘entrust’, xipwèi ‘announce’, xwi ‘do s.th. for s.o.’, ‘prepare’, nêgé ‘request respectfully’, xacië ‘show’, nû ‘send’, xâdùù ‘pay’, etc. The theme is unmarked and the beneficiary/recipient is introduced by the preposition xù. In (27), only the theme (kwé) is expressed, while in (28) and (29), a recipient introduced by the marker (xù) is added to the theme: (27) È miiri mè rè xâdùù kwé ngê xiti. 3sg forbidden that 3sg pay alcohol during sunday ‘It is forbidden for him to buy alcohol on Sundays.’ (28) Anyââ xâdùù na chaa lotoo xù Yaya. Mummy pay pst one car ben/rec Yaya ‘Mummy bought Yaya a car.’ (29) Nèxuu xacië na péci xù afainû. girl show pst book ben teacher ‘The girl showed the book to the teacher.’ The position of the recipient/benefactive marker may vary. In (30a), it precedes its argument (nâ ‘1sg’), but in (30b), it cliticizes onto the verb, and is separated from its argument (nû, allomorph of nâ ‘1sg’ when occurring after the past tense marker na) by the past tense marker: (30) a. Ke xù na chaa mwanöö xù nâ. 2sg give pst one cloth ben 1sg ‘You gave me a piece of cloth.’ (Lit. ‘You gave a piece of cloth to me.’)

1032

Claire Moyse-Faurie

b. Ke xù xù na nû chaa mwanöö. 2sg give ben pst 1sg one cloth ‘You gave me a piece of cloth.’ (Lit. ‘You gave to me a piece of cloth.’) In these examples, both the verb xù ‘to give’ and the preposition xù are expressed in the same clause.19

3.3.1.2 Disassociative (non-beneficiary) argument This construction, with the non-beneficiary introduced by the preposition taa, is found with verbs such as pè ‘take’, pêdè ‘steal’, fatere ‘ask’, nââ ‘ask for sth.’, sûûârâ ‘hide’, xanöè ‘ask for permission’, etc. (31) Rè xanöè taa wîrî mè rè bwa ra tö xûâ ngê chaaké 3sg ask.permission off 2pl that 3sg stay still loc tribe during several daa. day ‘He is asking you for permission to stay in the tribe for a few more days.’ The verb xâdùù ‘buy, pay for’, already mentioned above in association with xù in (28a), may also occur with oblique arguments introduced by taa: (32) Nûnûû kae dönä è xâdùù na tapwaka taa pèèröö? grandfather interr deic 3sg buy pst tobacco off priest ‘Is it Grandpa who bought tobacco from the priest?’ The ordering of the disassociative recipient and the theme is free, but the disassociative preposition usually remains immediately before its argument, contrasting in this with the recipient/beneficiary preposition xù, which may occur inside the verb phrase, separated from its complement. In examples (33), the recipient is expressed after the theme, while it occurs first in (34): (33) a. Nâ nââ sää-pwî taa rö. 1sg ask.for sucker-banana.tree off 2sg ‘I am asking you for banana-tree suckers.’ (Lit. ‘I request banana-tree suckers from you.’)

19 Note that the verb xù ‘give’ may have non-predicative uses, and occur after the preposition taa, that is, the preposition just opposite in meaning to the beneficiary preposition xù: È dii taa xù döu nä. 3sg refuse off give thing deic ‘He refuses to give that thing.’

Valency classes in Xârâcùù (New Caledonia)

1033

b. Nèxuu pêdè na péci taa afainû. girl steal pst book off teacher ‘The girl stole the book from the teacher.’ (34) Nâ fatere taa è xöu rè nâ. 1sg ask off 3sg cloth poss 1sg ‘I asked him for my clothes.’

3.3.2. Two oblique arguments This alignment resembles the indirective alignment described in the preceding section in so far as the flagging of T and P is identical,20 each of these arguments being introduced by the instrumental preposition ngê. The theme in (35b) and (36) is flagged as instrumental, as is the patient in (35a) (= 19), while the recipient is introduced by the benefactive/recipient marker xù: (35) a. Papêê jana ngê na mwè. coll+woman trade ins pst taro ‘Women were trading [their] taros.’ b. È xwiri21 ngê nô xù sîbêêrî a. 3sg sell ins fish ben old.lady deic ‘He is selling fish to the old lady’. Some verbs of communication, such as faxwata ‘tell’ or xangââ ‘shout sth. at s.o.’ follow this type of alignment. (36) Nâ faxwata xù rö ngê chaa êrêché. 1sg tell ben/rec 2sg ins one story ‘I am going to tell you a story.’

20 In the ‘Comrian’ approach, an argument marked as instrumental would not be regarded as P, whereas it would still be a core argument in the ‘Dixonian’ approach (Haspelmath 2011: 542), that is, a case of non-canonically marked O. Since I am having difficulties in defining a major transitive construction in Xârâcùù, I choose to consider such oblique complements as core arguments, also because they are necessarily expressed. 21 The verb xwiri ‘sell’ does not allow a direct object. It most often occurs with an oblique object introduced by the instrumental preposition. It can also occur with a theme introduced by the adverb/disassociative preposition taa: Kâmûrû nä xwiri taa mîî kura nä. man deic sell off art.pl shrimp deic ‘This man is selling these shrimps to get rid of them.’

1034

Claire Moyse-Faurie

(37) a. Kamûrû xangââ ngê na xwâxatii xù nèxuu. man shout ins pst scolding ben girl ‘The man shouted at the girl to scold her.’ (Lit. ‘The man shouted with a scolding to the girl.’) In (37b), it is the disassociative preposition taa which introduces the recipient of the verb of communication xangââ ‘shout’, due to the meaning of the oblique object introduced by the preposition ngê and expressing the content of the shouting, i.e. a demand. (37) b. Kamûrû xangââ ngê na chaa ùnââ taa nèxuu. man shout ins pst one demand off girl ‘The man shouted at the girl to ask her for something.’ (Lit. ‘The man shouted with a demand from the girl.’)

4 Argument coding alternations Insufficient information and the great variability in the usage of speakers make it impossible for me to answer certain questions concerning the formal marking of arguments as well as the extension of verb classes in patterns of alternations. Below are examples of the main alternations I was able to identify.

4.1 Labile alternations Nichols’s (1986: 156) definition of lability concerns valency on the one hand, and the semantic aspect on the other hand: “Lability is a valence pattern in which a verb can be either transitive or intransitive without the application of formal transitivizing or detransitivizing derivations. […] If a language has labile verbs, they will center on those which, like ‘break’, ‘open’ or ‘fill’, refer to changes of state which can either happen spontaneously or be brought about by human agency”. This definition of lability corresponds to one of Haspelmath’s non-oriented verb alternations (Haspelmath 1993: 91–92). Xârâcùù labile verbs have these kinds of semantic alternation, and others (Moyse-Faurie 2010).

4.1.1 Impersonal construction vs. intransitive construction Some Xârâcùù monovalent verbs may enter either an impersonal construction, as in (38a) and (39a), without any argument, or an intransitive one, with a referential experiencer subject argument, as in (38b) and (39b).

Valency classes in Xârâcùù (New Caledonia)

1035

This alternation concerns non-dynamic verbs denoting qualities or meteorological situations, such as mêgi ‘warm’, ‘have fever’, müü ‘(be) cold and humid’ and xùpè ‘cold’, ‘feel cold’, on the one hand, or verbs related to durative or phasal aspect situations such as xutuè ‘be a long time’ and cokwa ‘be finished’, on the other hand: (38) a. Xùpè sé na amû. cold big pst yesterday ‘It was very cold yesterday.’ b. Nâ xùpè. 1sg cold ‘I feel cold.’ (39) a. Wâ xutuè. pfv last.long ‘It has lasted a long time.’ b. È xutuè rö a. 3sg last.long loc here ‘S/he has been here for a long time.’

4.1.2 Inchoative vs. causative pairs Some Xârâcùù labile verbs have inchoative/stative versus causative meanings, such as tëi ‘be empty, empty’, xwi ‘exist, build’, cokwa ‘be finished, finish’22, nââbu ‘begin’, kê ‘be burned, burn’, sùù ‘suffer, treat’, xwêê ‘fall, pour’, and English loanwords such as sukwa ‘be sugared, sugar’. Some of these verbs (‘burn’, ‘finish’, ‘start’, etc.) are labile verbs in many languages including German, French or English. The intransitive construction has an undergoer subject argument, corresponding to the object argument in the transitive construction, which in addition has an actor subject argument (S = P, A causer). In the intransitive construction (40a) and (41a) the verb takes a passive, resultative, or inchoative meaning, and the subject refers to the patient. In the transitive construction (40b) and (41b), the verb is active, with the agent as subject, and the patient as object:

22 The verb cokwa ‘be finished, finish’ may occur in three different constructions: in an impersonal construction, it only means ‘be finished’ and refers to the end of a temporal phase. In intransitive vs transitive constructions, it is part of a resultative vs causative pair (‘something is finished’ vs ‘someone finished something’).

1036

Claire Moyse-Faurie

(40) a. Ku kê. yam burn ‘The yam is burnt.’ (resultative) b. Kâmîâ kê nùi a. sun burn island deic ‘The sun is burning the island.’ (causative) (41) a. Märä nèè. bird fly ‘The bird is flying away.’ (inchoative) b. Kâmûrû a nèè nôô. man deic fly sail ‘The man is sailing away.’ (Lit. ‘The man is flying the sail.’) (causative) (42) a. Nâ xwêê. 1sg fall ‘I fell down.’ b. Nâ xwêê kwé. 1sg fall water ‘I am pouring water.’ In the case of xwêê ‘fall’, however, only an inanimate object argument can be expressed; otherwise, the verb has to be derived with the causative prefix: (42) c. Nâ fa-xwêê chaa xûûchî. 1sg caus-fall one child ‘I make the child fall down.’

4.1.3 Middle alternation The middle alternation is the consequence of the nearly complete loss of reflexes of the middle Proto Oceanic prefix (see § 5.1.2). It differs from the preceding alternation in that the role of the subject does not change whatever the valency of the construction is. The subject in both intransitive and transitive constructions is an animate actor subject argument, and the situation expressed belongs to the middle domain, such as grooming actions (43a), or inherent reciprocity (44a and 45a). In the transitive construction, the object is the undergoer, whether it refers to an animate (43b, 44b) or an inanimate (43c, 44c) and the causative meaning is other-directed.

Valency classes in Xârâcùù (New Caledonia)

1037

Grooming events (43) a. Nâ xii. 1sg shave ‘I am shaving.’ b. Nâ xii è. 1sg shave 3sg ‘I am shaving him.’ c. Nâ xii nû. 1sg shave coco ‘I am grating coconut flesh.’ Inherent reciprocity This subclass includes verbs denoting inherent reciprocity, which can occur with only one argument, a subject, referring to a plurality of participants. For example, penyi ‘to take leave’ as well as tôôbùtù ‘assemble’, always occur with a dual or plural subject when used intransitively: (44) a. Ri tôôbùtù. 3pl assemble ‘They are assembling.’ b. Ri tôôbùtù ri. 3pl assemble 3pl ‘They are assembling them.’ c. Ri tôôbùtù köfi. 3pl assemble coffee ‘They are gathering coffee beans.’ Whereas the verb tôôbùtù ‘to assemble’ may take a direct object, penyi ‘to take leave’ has an oblique object in an ‘extended transitive’ construction; this otherdirected/disjoint object is introduced by the disassociative preposition taa:23 (45) a. Ri penyi. 3pl separate ‘They are leaving each other.’ b. Ri penyi kété taa na rî. 3pl separate quickly off pst 1pl.excl ‘They quickly left us.’ (Lit. ‘They quickly separated from us.’)

23 Interestingly enough, French shows exactly the same kind of middle alternation: Ils se sont séparés ‘they left each other’ vs. Ils se sont séparés de nous ‘they left us’.

1038

Claire Moyse-Faurie

4.1.4 Reflexive inchoative alternation Some bivalent verbs must always have two overt arguments expressed. Thus, the lability of these verbs is not related to the shift from an intransitive to a transitive construction, but it is related to the degree of animacy of the subject. If the subject refers to an inanimate entity, the pronominal object must have the same number and person features as the subject; the situation falls into the middle domain (46a, 47a), and there is no possible ambiguity. When the subject is animate, the object does not have to be coreferential with the subject, and the orientation of the event is then causative (46b, 47b). These verbs contrast with the ‘truly reflexive verbs’ (§ 3.2.2), for which there is no available choice for the object, and which must always be coreferential with the subject. (46) a. Kö chèpùtù ri. cloud gather 3pl ‘The clouds are gathering.’ b. Nâ chèpùtù ri. 1sg gather 3pl ‘I am getting people together.’ (47) a. Buké nä fasaa è töwâ kâmîâ nä. bouquet ipfv damage 3sg loc sun deic ‘The bunch of flowers is withering in the sun.’ b. Nâ fasaa buké. 1sg damage bouquet ‘I am making the bunch of flowers wither.’

4.2 Possessive flagging alternations As already mentioned, some Xârâcùù prepositions result from a grammaticalization process. As far as inalienable nouns are concerned, the change in syntactic category is not totally complete, and in some cases, it is difficult to determine if we are dealing with a preposition + its object, or with a possessive noun phrase. Two cases are described below. The first one concerns the expression of the experiencer, and the second the expression of the recipient or beneficiary. The position of the argument is one of the main points to be taken into consideration. When the argument follows the verb phrase, we can consider that it consists in a prepositional phrase. When the argument is the subject, however, it still is a possessive noun phrase.

Valency classes in Xârâcùù (New Caledonia)

1039

4.2.1 Experiencer flagging as possessor This alternation concerns a few verbs involving experiencer arguments, whose degree of affectedness changes according to the construction involved. With a nominative experiencer, the verb expresses a descriptive/non-dynamic emotional state while as a possessor, it strongly expresses the experiencer’s deeper feelings. Xârâcùù verbs having such optional constructions for the expression of the experiencer are verbs of feelings or emotions such as saa ‘bad, bad looking’, ‘feel bad’; kwèti ‘be tired’, ‘feel tired’; wîrî ‘disgusting’, ‘feel disgusted’; mârâ ‘be worried’, ‘feel dizzy’. Some of these verbs may derive bivalent verbs with the applicative suffix (see § 5.2.1). The possessive experiencer is in an inalienable relation with the relational noun wâ- ‘inner’. This relational noun very probably is the source of the preposition wâ ‘at’, ‘about’ which introduces diverse complements24 (cf. § 3.2.3.6). The experiencer is totally affected and powerless (example 48b), in a similar way as Creissels’s ‘affective covert impersonals’ (2007: 28), while example (48a) is simply objective, with the experiencer expressed as a nominative argument. (48) a. Nâ kwèti ù-sööpö rè xöu. 1sg tired nmlz-wash poss clothes ‘I am tired of washing clothes.’ b. Nâ sii fè ti nuö döbwa wâ-nâ kwèti. 1sg neg go loc bush because inner-1sg tired ‘I don’t go to the bush because I feel tired.’ The semantic orientation of the verb itself doesn’t change with the construction; the only difference lies in its scope, whether it applies to a non-dynamic or a more agentive experiencer, which is more or less affected depending on the way (as a possessor or as a plain argument) it is expressed. When expressed after the verb phrase, wâ has grammaticalized as a preposition, and introduces an experiencer involved in a higher responsibility: (49) Rè nöö wâ rö mè ke fârâ daa kèwâ kèè-bôô tiwâ kèè-näü. 3sg stay at 2sg that 2sg count day from nmlz-dig to nmlz-plant ‘It is up to you to calculate the time needed between digging and planting.’

4.2.2 Recipient or beneficiary flagging as possessor Besides the use of the preposition xù, a recent development from the homophonous verb ‘give’, Xârâcùù exhibits a well-known Oceanic pattern for encoding recipients 24 It must not be confused with the homophonous perfective aspect.

1040

Claire Moyse-Faurie

and beneficiaries, viz. the use of a possessive classifier to introduce them (cf. Lichtenberk 2002; Song 1998, 2007). The relational nouns ngêê-25 (or its reduced form êê-, very common among speakers under forty) and nèxêê- have the general meaning of ‘belongings’. The exact meanings and status of ngêê-/êê-/nèxêê- are complex. According to some informants, êê- and ngêê- convey a responsibility meaning in addition to the beneficiary meaning, while nèxêê- mostly applies to the possession of object belongings. The relational noun ngêê-/êê-/nèxêê- ‘belongings’ tends to grammaticalize into a beneficiary or recipient marker when expressed after the verb phrase, but remains a noun when occurring in the different functions listed below. a) The relational noun can enter a predicative-possession construction, either by itself, (50), or in an equative clause (51): (50) Ngêê-yaa? Ngêê-nâ. belongings-who belongings-poss1sg ‘Whose is it?’ ‘It is mine’ or ‘It is my turn/my duty.’ (51) Nèxêê-nâ bachèé pa xûûchî. belongings-1sg three coll child ‘I have three children.’ b) It can take a clausal complement: (52) Ngêê-wîrî mè wîrî toanôô kètè bwa è nä nöö nä. belongings-2pl that 2pl find place deic 3sg ipfv stay loc.anaph ‘It is your responsibility to find the place where he lives.’ (53) Êê-rö mè ke kane pwâ-mêdè. belongings-2sg that 2sg pick.up fruit-orange ‘It is your task to pick up oranges.’ c) It is the only way to render the English possessive pronoun, occurring as a subject noun phrase: (54) Mwâsöö rè nâ mîâ, ngêê-rö ngürü. hat poss 1sg red belongings-2sg black ‘My hat is red, yours is black.’

25 Ngêê- is also clearly related to the verb angê ‘to possess’: Nâ nä angê rè chaa lotoo. 1sg ipfv possess ipfv one car ‘I have a car.’

Valency classes in Xârâcùù (New Caledonia)

1041

When occurring after the verb phrase, these relational nouns have grammaticalized into beneficiary/​recipient markers, as shown in the following examples: (55) Êê kêgörö nêkwââ ngêê xûûchî abaa. 3sg.ipers crush.with.fingers leaves ben child newborn ‘One crushes leaves with fingers for the newborn child.’ (56) Wînâ chaa ùbwa nèxêê sê röö. leave one place ben sister poss+2sg ‘Leave a little place for your sister.’ (57) Nâ nä xwi rè chaa xiti êê béé-nâ. 1sg ipfv make ipfv one feast ben friend-1sg ‘I will organize a party for my friend.’ Compared to the preposition xù, these former relational nouns may not be separated from their arguments, and convey a meaning of higher responsibility in the type of attribution. Compare (58a) and (58b): (58) a. Anyââ xâdùù na chaa lotoo xù Yaya. Mummy pay pst one car ben Yaya ‘Mummy bought Yaya a car.’ (It was not necessary, just a gift) b. Anyââ xâdùù na chaa lotoo êê Yaya. Mummy pay pst one car ben Yaya ‘Mummy bought Yaya a car.’ (Yaya needed it)

4.3 Argument deletion alternation The omission of the subject does not lead to ambiguity in connection with bivalent verbs, since the basic word order is SVO, and since there is no passive voice in Xârâcùù. With most active verbs, there is no need for the presence of a dummy pronoun, or of an impersonal pronoun, although the latter may always occur. Compare examples (59a) and (60a), with no subject, with examples (59b) and (60b) containing explicit subjects. In both cases, the verb keeps its orientation. (59) a. Wâ chèè kwâjùù-rè. pfv pull sheet-poss3sg ‘[Someone] is pulling the sheets of the sail.’ b. Apuukwâdè wâ chèè kwâjùù-rè. wind.chief pfv pull sheet-poss3sg ‘The chief of the wind is pulling the sheets of the sail.’

1042

Claire Moyse-Faurie

(60) a. Wâ tara mîî mwâ ri wâ kä xêêdi. pfv see pl house 3pl pfv ready totally ‘[One] can see these houses which are totally finished.’ b. Mô tara è. night see 3sg ‘Night caught up with him.’ (Lit. ‘Night sees him.’) The omission of bivalent verb objects is also possible in Xârâcùù, except with reflexive verbs which require a pronominal coreferential object (cf. § 3.2.2); the verb then takes a more stative or inherent meaning (61a and 62a), but in contrast with labile verbs, the meaning of the verb does not change. This alternation often concerns verbs of consumption, or regular activities. (61) a. Dèèri bôô. people weed.out ‘People are weeding.’ b. Dèèri bôô nècaa rè Fabio. people weed.out field poss Fabio ‘People are weeding the field of Fabio.’ (62) a. Dapé chii. Dapé angle ‘Dapé is fishing.’ b. Dapé chii chaa mèröö. Dapé angle one parrot fish ‘Dapé is catching a parrot fish with a rod.’ Most of the verbs which take an oblique object are also only optionally followed by it. They may occur in an intransitive construction, or with an oblique object, as it is the case with verbs such as xuru ‘flee (from)’, mââî ‘be ahead, precede’, or bata ‘be scared, be afraid (of)’, occurring in an intransitive construction in (63a) or in an “extended transitive” construction in (63b): (63) a. Ke nä wita bata rè arè döumè ke nä ciköpuru 2sg ipfv proh be.afraid ipfv tomorrow when 2sg ipfv cross rè xwârè. ipfv river ‘You must not be afraid tomorrow when you will have to cross the river.’ b. È bata taa na fètaa rè mwâciri rèè. 3sg be.afraid off pst leave poss country poss+3sg ‘He has been afraid of leaving his country.’

Valency classes in Xârâcùù (New Caledonia)

1043

4.4 Dative (indirective/secundative) alternation Three subtypes of dative alternation can be distinguished. a) Some verbs can occur with the recipient as direct argument when no other complement is expressed, as in (64a) and (65a). When another argument is added, it takes the place of the recipient as a direct object argument (65b), or as a complement clause (64b), while the recipient is flagged as oblique, introduced by xù if it is a beneficiary (64b), or by taa if it is a disassociative (65b): (64) a. È fi rö. 3sg lie 2sg ‘He is lying to you.’ b. È fi xù rö mè siè kii. 3sg lie ben 2sg comp not.exist key ‘He is lying to you [saying that] he doesn’t have any key.’ (65) a. Apêdè pêdè axwiri. robber rob merchant ‘The robber is robbing the merchant.’ b. Apêdè pêdè mwânêê taa axwiri. robber rob money off merchant ‘The robber is stealing money from the merchant.’ b) Another subtype of alternation occurs between a theme, unmarked when it is the only expressed object as in (66a), or followed by a beneficiary introduced by the preposition xù (66b), and flagged with the instrumental when a recipient/beneficiary is added (66c): (66) a. Nâ fadù ääda. 1sg share food ‘I am sharing the food.’ b. Nâ fadù ääda xù dèèri. 1sg share food ben people ‘I am sharing the food for the people.’ (so that everyone will have food) c. Nâ fadù dèèri ngê ääda.26 (= Nâ fadù ngê ääda dèèri.) 1sg share people ins food ‘I am sharing the food among the people.’ (so that the food will be sufficient for all) 26 Or, with the cliticization of the preposition into the verb phrase: nâ fadù ngê dèèri ääda, semantically equivalent, but with a focalization on the food.

1044

Claire Moyse-Faurie

This alternation recalls the one occurring with English verbs such as ‘provide’ (‘I provide food to people’ vs. ‘I provide people with food’). c) The situation is different with another verb of transfer, xù ‘give’. In example (67b), the verb xù ‘give’ occurs in a tripartite alignment construction (P≠ngêT≠xùR), indicating the desire of the agent to show off. In example (67a) the theme is unmarked, while in (67b) it is introduced by the instrumental preposition ngê. In both examples, however, the recipient is marked by the benefactive preposition xù. (67) a. Ke xù xù na nû chaa mwanöö. (= Ke xù na chaa mwanöö xù nâ.) 2sg give ben pst 1sg one cloth ‘You gave me a piece of cloth.’ b. Nèxuu xù na ngê chaa péci xù afainû. girl give pst ins one book ben teacher ‘The girl gave the book to the teacher (to show off).’

4.5 Reflexive or reciprocal alternation Except in a few cases (such as reflexive verbs in § 3.2.2), bivalent verbs are generally not restricted as far as the choice of an object is concerned; if the latter turns out to be a pronoun of the same person and number as the animate subject, the interpretation can be ambiguous out of context, between a reflexive or reciprocal and a disjoint (other-directed) interpretation: (68) a. Taiki kèkè è. dog bite 3sg ‘The dog is biting it/her/him/itself.’ (69) a. Ri ciiwi ri. 3pl help 3pl ‘They are helping them/each other.’ The reflexive/reciprocal/disjoint alternation requires animate subjects. It is possible to add the reflexive/reciprocal marker in order to get a more restrictive interpretation. The Xârâcùù reflexive/reciprocal marker is mûgé, which, as an (ad)verb, means ‘return’, ‘again’, and consequently also conveys a repetitive reading: (68) b. Taiki kèkè mûgé è. dog bite return 3sg ‘The dog is biting itself.’ or ‘The dog is biting it/her/him again.’ (69) b. Ri ciiwi mûgé ri. 3pl help return 3pl ‘They are helping each other.’ or ‘They are helping them again.’

Valency classes in Xârâcùù (New Caledonia)

1045

Inherent reciprocity is usually unmarked: tôôbùtù ‘gather’, pùùtè ‘meet’, xöyö ‘marry’, pia ‘fight’.

4.6 Conative alternation I only found this conative alternation with verbs of consumption. In (70a), the direct object referring to the patient is totally affected by the event, while in (70b), the patient is flagged as an adjunct with the complex locative preposition kèwâ ‘among’, and is only partially affected: (70) a. È da na xöö-dö amû ngê chêêdê. 3sg eat pst egg-hen yesterday loc evening ‘He ate eggs yesterday evening.’ b. Nä wèa bù fè ti mûgé da kèwâ ääda mwîrî. then this.one flying.fox go loc again eat loc food anaph ‘Then the flying fox went back to eat some of the food.’

5 Valency changing devices in Xârâcùù Xârâcùù has several valency changing devices: − the causative prefix fa-, a reflex of Proto Oceanic *pa[ka]-, which is productive with almost all types of verbs; − the resultative prefix mê-, a reflex of POc *ma-, which is no longer productive; − the middle prefix ù-, a reflex of POc ‘plurality of actions’ middle/reciprocal prefix *paRi‑, also no longer productive in Xârâcùù. Xârâcùù has only kept a few traces of the applicative POc *-akin[i] suffix, with the -ri suffix now unproductive. Instead, as I have already mentioned, Xârâcùù has developed several oblique constructions, with various markers of verbal or nominal origin (cf. § 3.2.3 above), and a very productive lexical compounding process (cf. § 6.2 below), both resulting from nuclear verb serialization developments.

5.1 Valency-reducing operations In this section, I shall examine on the one hand, marked valency reducing operations, which mainly have resultative, reciprocal and middle semantic values and, on the other hand, an unmarked valency reducing operation, the incorporation strategy. None of these operations are still productive in Xârâcùù. Besides, in con-

1046

Claire Moyse-Faurie

trast with what is reconstructed for Proto Oceanic (cf. for example Evans 2003: POc *kani ‘to eat sth’ > *kanikani ‘to eat’), reduplication is not used as a valency-reducing operation in Xârâcùù. It only has the semantic function of intensity, whether the reduplication is partial or total 27: ngû ‘move’ > ngûngû ‘wriggle’; xa ‘speak’ > xaxa ‘speak loudly, speak a lot’; sùù ‘push’ > sùsùù ‘push with energy’; cemîâ ‘suffer’ > cemîmîâ ‘suffer a lot’. Another uncoded valency reducing operation has already been described, concerning labile alternations (see § 4.1).

5.1.1 The resultative prefix mêThe POc prefix *ma- is described as being quite productive with verbs high in transitivity such as *liŋi- ‘to pour sth. out’ > *maliŋi ‘to be poured, split’ (Evans 2003: 268). The Xârâcùù reflex mê- is only attested in a few verbs belonging to the semantic field of destruction, but it has the same valency-reducing role, with intransitive S corresponding to transitive P. The nasal vowel of the prefix changes the initial oral consonant of the verb into a prenasalized consonant: pöru pùtù pwéa tia

‘peel, skin’ ‘paste, crush’ ‘bend’ ‘tear, split’

mê-böru mê-bùtù mê-bwéa mê-dia

‘be skinned, scraped’ be crushed, battered’ ‘be bent’ ‘be splitted, torn’

Bound verb stems also admit the mê- prefixation: -kai ‘crush’ -körö ‘break’ -nyûû ‘pierce’

mê-gai ‘be crushed’ mê-görö ‘be broken’ mê-nyûû ‘be pierced’

5.1.2 The middle prefix ùXârâcùù almost completely lost one of the most interesting and productive prefixes found in the great majority of Oceanic languages, namely the prefix reconstructed in POc as *paRi-. This prefix had several uses, and is reflected in most Kanak languages with middle, reciprocal and, more rarely, reflexive meanings. Reflected as ù- in Xârâcùù, it only derives a dozen verbs (a few more in the Xârâcùù variety

27 If the initial verb is plurisyllabic, or if its first syllable includes a long vowel, the first part of the reduplicated form will only consist in two moras, that is a consonant followed by a short vowel.

Valency classes in Xârâcùù (New Caledonia)

1047

spoken in Thio), of different classes and conveys in all cases a middle meaning, referring to either generic or habitual characteristics, or grooming actions and sociative situations: Action vs. bë da nêê pè xù sö

habitual behavior ‘move to some place’ ‘eat’ ‘choose’ ‘take’ ‘give’ ‘to pride oneself on sth.’

Grooming actions cù ‘comb s.o.’ mwé ‘dive, put into water’

ù-bë ù-da ù-nêê ù-pè ù-xù ù‑sö

‘be restless’ ‘bite (fish)’ ‘be picky’ ‘juggle’ ‘be contagious’ ‘be haughty, be a boaster’

ù-cù ù-mwé

‘comb one’s hair’ ‘take a bath’

Other grooming action verbs are either labile verbs, such as xii ‘shave’ or lexically specific verbs such as pêrê ‘look at oneself in a mirror’. Sociative cuè ooro juu xâpârî

‘sit’ ‘rejoice of sth.’ ‘agree to sth.’ ‘see s.o. or sth.’

ù-cuè ù-ooro ù-juu ù-xâpârî

‘assemble’ ‘rejoice together’ ‘come to an agreement’ ‘meet’

5.1.3 Object incorporation Object incorporation is a well attested feature in Oceanic languages, inducing valency reduction in accusative as well as in ergative languages. The incorporated object is non-referential and is included in the verb phrase, often leading to compounds which can themselves undergo a new transitivization. In several Kanak languages, verbs vary formally, depending on the grammatical status of the object. This is the case, for example, in Drehu, one of the Loyalty Island languages. Hence in (71a), the verb öj ‘to press’ has an object modified by an article and a plural morpheme. In (71b), the verb changes to öji, the former object is incorporated, loses its determiners, takes a generic meaning, and cannot be separated from the verb: Drehu (Lifu, Loyalty Islands) (71) a. Eni a öj la itre ono. 1sg ipfv press art pl coconut ‘I am squeezing the milk out of the coconut gratings.’

1048

Claire Moyse-Faurie

b. Eni a öji=ono. 1sg ipfv press=coco ‘I am squeezing coconut gratings.’ (Moyse-Faurie 1997: 235) In Xârâcùù, incorporation is a very marginal alternation and it does not affect the verb formally. The semantic and syntactic correlates, however, are identical to the ones found in Drehu. In (72a), the object and its determiner are separated from the verb by the postverbal part of the imperfective tense-aspect marker, required in the transitive construction, while in (72b), the verb phrase comprising the incorporated object forms a non-separable unit: (72) a. Chaa kamûrû nä tuu rè chaa kwâ. one man ipfv step.on ipfv one boat ‘The man is stepping on the boat.’ b. Chaa kamûrû nä tuu=kwâ. one man ipfv step.on=boat ‘The man is going on board.’

5.2 Valency-increasing operations Valency-increasing devices, which were very productive in Proto Oceanic (Evans 2003), include, in particular, the following affixes: a) the Proto Oceanic *-i suffix (occurring only with consonant-final and *-a final verb stems; otherwise: object enclitics attach directly): – with Actor subject verbs: applicative (P: role of location, goal, addressee or stimulus) – with Undergoer subject verbs: causative b) the POc *-akin[i] suffix: – applicative use, with P in role of concomitant, cause/stimulus, content, product, instrument, beneficiary; – causative use: first with motion verbs and further extended to non-motion verbs. c) the POc *pa[ka]- prefix: – *pa- and *paka- causative prefixes: S = P, A = causer participant; – in a few languages: *pa- causative, *paka- multiplicative use with numerals. In contrast with the various Proto Oceanic affixal valency strategies, Xârâcùù has only one reflex of the applicative suffix *-akin[i], viz. ‑ri, and a reflex of the causative prefix *pa[ka]-, viz. fa‑, for which the multiplicative use has been preserved (see below § 5.2.2).

Valency classes in Xârâcùù (New Caledonia)

1049

5.2.1 The transitivizing/applicative suffix -ri The Xârâcùù transitive/applicative suffix ‑ri can only apply to a few verbs. The loss of POc transitive suffixes has been compensated for by the development of several relational morphemes, in relation to verb serialization (Moyse-Faurie 1991). This morpheme -ri is still a suffix, in the sense that it cannot be separated from the verb, and it did not evolve into a preposition, as has been the case, for example, with the applicative suffix POc *akin[i] reflected in East Uvean (Moyse-Faurie, personal data) and Tongan (Churchward 1953: 120) both by the suffix -’aki and by the preposition ’aki ‘instrumental’. The Xârâcùù suffix -ri derives half a dozen verbs denoting emotions or feelings: fiö

‘be lazy’

mârâ muru kwèti cara

‘be ‘be ‘be ‘be

nyôô nyînyôô

‘be foolish, drunk’ ‘be complicated, merge’

worried, be upset’ alive’ tired’ ashamed, dazzled’

‘to refuse, to have had enough of’ mârâ-ri ‘be disgusted with’ muru-ri ‘live on’ kwèti-ri ‘be tired of’ cara-ri ‘be ashamed of, be dazzled by’ nyôô-ri ‘be confused about’ nyînyôô-ri ‘not recognize sth. or s.o.’ fiö-ri

The subject is an undergoer in both intransitive and transitive constructions, and the object licensed by the applicative suffix -ri is obligatorily expressed; it denotes the means, the source or the stimulus of the emotion, and can refer to an animate (74c, 75) or an inanimate entity (73b and 74b). Since no element can separate the verb from its suffix, the past tense marker na has to be postposed to the suffixed verb (73b), before the object: (73) a. Pwénîî mè nâ muru mwâmwaa! optative that 1sg live long ‘[I] wish I would live for a long time.’ b. Nâ muru-ri na nû. 1sg live-appl pst coconut ‘I had my fill of coconut.’ (74) a. È kwèti. 3sg be.tired ‘He is tired.’ b. Ri wâ kwèti-ri kèchö. 3pl pfv tired-appl magnania ‘They are sick of (eating) magnania28.’ 28 Pueraria lobata, has a fleshy tapioca-like tuber, eaten in times of famine.

1050

Claire Moyse-Faurie

c. Nâ kwèti-ri môrô rö. 1sg be.tired-appl already 2sg ‘I am already tired of you.’ (75) È sii cara-ri xéré bwa è nöö Numéa. 3sg neg ashamed-appl grandson deic 3sg stay Nouméa ‘She is not ashamed of her grandson who lives in Nouméa.’ To the above list of verbs denoting sensations or natural states, two locative (ad)verbs also requiring the -ri suffix must be added: nûbö- ‘(be) near of’ in (76) and bèbè- (< bè ‘bottom’ as in bè-nâ ‘my bottom’, bè-kùrè ‘bottom of the pot’) in (77a) and (77b): (76) È nûbö-ri mwâ. 3sg near-appl house ‘S/he is near the house.’ (77) a. È tââ bèbè-ri nâ. 3sg stand behind-appl 1sg ‘S/he is standing behind me.’ b. È köö bèbè-ri kwââ. 3sg hide behind-appl tree ‘S/he is hiding behind a tree.’ There is also one case of a nominal derivation with -ri: mwaraa- ‘shallow light, gleam’ is a relational noun (obligatorily followed by a possessor); the derived verb means ‘be at a shallow light of sth.’, as a single predicate in (78a), and as V2 in a serial verb construction in (78b): (78) a. Nâ mwaraa-ri nè. 1sg be.at.shallow.light-appl fire ‘I am [staying] by the light of the fire.’ b. Nâ pètù mwaraa-ri nè. 1sg weave be.at.shallow.light-appl fire ‘I am weaving by the light of the fire.’ In all probability, a variant form of the applicative suffix -ri has been incorporated into bivalent verbs such as cîîrî (in free variation with cîî) ‘distribute, share’, and in the verb faari (var. faare) ‘ask for’, perceived by the speakers as being an underived form, although it originally is the transitivized form for faa ‘ask’:

Valency classes in Xârâcùù (New Caledonia)

1051

(79) a. Faa wâ rè nämè xai péci bwa. ask at 3sg that where book deic ‘Ask him where the book is.’ b. Aaxa Pokwé faari taa dèèri rèè mè ri dèpuu sùù chaa chief Uvea ask off people poss+3sg that 3pl in.turn push one mwîî kwâdè. big wind ‘The chief from Uvea asks his people to raise a strong wind.’ Finally, the verb mwârâ ‘laugh’, ‘play’ becomes bivalent with the suffix -kaciri. It is the only occurrence of this strange suffix:29 (80) Nâ mwârâ-kaciri tèpe nä. 1sg laugh-appl story deic ‘I am laughing at this story.’

5.2.2 The causative prefix faAlmost all Xârâcùù verbs, whatever their valency, may take the causative prefix fa-. Exceptions concern semantic restrictions, with verbs denoting inherent properties such as aéé ‘(be) authentic’, afädë ‘(be) foreign’, xwâkètè ‘be profane’, or inherited deformities and diseases such as amè ‘(be) paralyzed’, bëpaii ‘(be) sickly’, dööpwé ‘be hunchbacked’, mèrèdêê ‘be deaf’, or natural inevitable processes such as kèpwiri ‘(be) high (tide)’, pââmé ‘be toothless’, etc. In Xârâcùù, causativization is used as a strategy for deriving bivalent and trivalent verbs; it occasionally also derives causative verbs from nouns, mainly referring to plants or animals. a) prefix fa- + monovalent verbs The undergoer or experiencer subject argument becomes the undergoer object, and an agent is added. bata sé mëtë mwâmwâ muru xûda

‘be afraid’ ‘big’ ‘dry’ ‘dress oneself’ ‘live’ ‘be full of sth.’

fa-bata fa-sé fa-mëtë fa-mwâmwâ fa-muru fa-xûda

‘to frighten’ ‘make sth. big, honor s.o.’ ‘dry s.o. or s.th.’ ‘dress someone’ ‘provide the living for s.o.’ ‘fill’

29 Kaci must have been a verb, but is no longer attested. A similar construction occurs with bata ‘be afraid’, to which a direct object can be added as long as it is followed by xwi ‘exist, make’ and cara ‘be ashamed’, deriving the verb bata-xwi-cara ‘to respect’.

1052

Claire Moyse-Faurie

(81) a. Nè xiri. fire light ‘The fire lights up.’ b. Nâ fa-xiri nè. 1sg caus-light fire ‘I light the fire.’ The actor-subject argument remains the agent of the derived verb, and an undergoer object is added: cuè tââ

‘sit’ ‘stand’

fa-cuè fa-tââ

‘make s.o. sit’ ‘stop’, ‘make s.o. stand up’

(82) a. Wîrî cuè xö tö nèpwéé-kâmîâ nêmwâ. 2pl sit à.jeun loc inside-sun today ‘Today, you are sitting not eating anything.’ b. Yaya wâ fa-cuè chéé xûûchî rè nâ. Yaya pfv caus-sit down child poss 1sg ‘Yaya made my child sit down.’ b) prefix fa- + bivalent verbs Bivalent verbs may have a derived causative, such as: sia ‘grate (tubers)’ xwata ‘hear, understand’ näü ‘plant’

fa-sia ‘make s.o. grate (tubers)’ fa-xwata ‘tell sth. to s.o.’; ‘information, news’ fa-näü ‘make s.o. plant’

Bivalent causativized verbs from intransitive verbs may in turn be causativized as trivalent verbs: xûda ‘be full’ > fa-xûda ‘fill sth.’ > fa-fa-xûda ‘make s.o. fill sth.’ The examples below show how the theme, unmarked in a transitive construction (83a), is flagged as instrumental in a ditransitive construction (83b):

(83) a. È näü baaru döxöu. 3sg plant two pandanus ‘(S)he is planting two pandanuses.’

Valency classes in Xârâcùù (New Caledonia)

1053

b. Pa jöösè fa-näü ngê pa Xârâcùù köfi töxû coll. soldier caus.-plant ins coll Xârâcùù coffee on mê-nèpärä.30 ancient-mound ‘Soldiers are making the people of Xârâcùù plant coffee on the old yam mound.’ The causer pa jöösè in (83b) replaces the agent (è in (83a)) in the subject position. The patient (baaru döxöu) is unmarked in the transitive construction, but is flagged as instrumental in (83b) with the preposition ngê cliticized to the predicate, and separated from its complement (köfi) by the direct object argument (pa Xârâcùù). c) Multiplicative use of faIn Xârâcùù, the causative prefix fa-, as it was the case with Proto Oceanic *pa[ka]-, also derives ordinals: baaru ‘two’ bachéé ‘three’

fa-baaru fa-bachéé

‘second (of a series)’; ‘in two (pieces)’ ‘third’; ‘in three (pieces)’

6 Verbal compounds Serial verb constructions are still very productive in Xârâcùù. Most often, they consist of two verbs, referring to different aspects or stages (position, manner, purpose, etc.) of an event, but constructions including three serial verbs are not infrequent, e.g. cuè köö pwâârî in (84): (84) Ri wâ cuè köö pwâârî sêgè mwîrî gaka nä saù mê da 3pl pfv sit hide pass.round stone anaph crow ipfv each.time come eat tixû. loc ‘They [turtledoves] lay in ambush around the stone to which the crow used to come and eat on top of.’ The valency of the compounds mostly relies on the valency of the second element. If the second element is bivalent, the compound is also bivalent, whatever the valency of the first element. By contrast, a few compounds are monovalent although the first element is bivalent. This is the case with compound verbs such as

30 The other word order, with the preposition occurring just before its complement, is also possible: Pa jöösè fa-näü pa Xârâcùù ngê köfi töxû mê-nèpärä.

1054

Claire Moyse-Faurie

cipwîrî ‘roll’ (ci- < cii ‘spread’ + pwîrî ‘turn’), nârâyaa ‘be sad’ (nârâ ‘think’ + -yaa ‘badly’), chéxô ‘cough’ (ché ‘say’ + xô ‘sing’). These productive serial verb constructions evolved on the one hand into different types of compounds, and on the other hand, into prepositions, as we have seen in § 3.2.3. Two main types of bound verbs must be distinguished: (i) bound stem elements of verbal origin; (ii) classificatory affixes of nominal or verbal origin. Bound stem elements consist of a reduced verb as first element. For example, the verb xuru ‘flee’ occurs as xö- in compounds, stemming from a former serial verb construction: (85) È wâ xö-fètaa möö chaa nii. 3sg pfv flee-leave first one penis.sheath ‘He fled, first leaving his penis sheath.’ In some cases, the original verb is unknown, and only its first syllable remains, as it is the case with ta- ‘shoot at, thrust with a spear’, a form attested in a dozen compounds: (86) Ri wâ ta-faaté è ngê a wââî nä. 3pl pfv shoot-run.after 3sg sm deic these.men deic ‘They shot at him and ran after him, these men.’ Most compound verbs, however, include classificatory affixes. − Classificatory prefixes are known to be a typical feature of Oceanic languages spoken in New Guinea, but they are quite numerous in New Caledonian languages of the Center and South of the Mainland (Osumi 1995: 135–138; Ozanne-Rivierre & Rivierre 2004). The prefixes often have a transparent nominal origin, although they are always monosyllabic, having been reduced to a single mora. In Xârâcùù, as in these other New Caledonian languages, classificatory prefixes mostly indicate the manner or the body part with which the event is brought about. − The second elements, most often of verbal origin, indicate the result of the action. As described by Ozanne-Rivierre & Rivierre (2004: 362), the lexical suffixes “may be grouped around two axes: action on the form, integrity of the object (break, cut, split, crush, tear, etc.); action on the location, position, or trajectory of the object (return, block, tear off, detach, etc.)”. Many of the classificatory suffixes reflect former verb forms, and probably stem from former nuclear-layer serialization. Compounds, however, now form a single phonological word stressed on the first syllable. In some cases, both the suffix and the verb coexist in the language, as for example the verb tia ‘split, tear’, and the corresponding bound forms -tia/

Valency classes in Xârâcùù (New Caledonia)

1055

-dia. In some cases, assimilation processes have led to the change of the initial consonant of the suffixes: bë ‘move (oneself)’ + -körö ‘break’ ngî- < ngîrî ‘rub’ + -körö ‘break’ kè- ‘with teeth’ + tia ‘split’ kê- ‘with fingers’ + tia ‘split’

bëkörö ‘break sth. while moving’ ngîgörö ‘break sth. while rubbing’ kètia ‘split with teeth’ kêdia ‘split with fingers’

Both elements forming the compound may have lost their status as independent verbs, as for example bi- < biri ‘turn’ which combines with the suffixed form -chëe ‘miss’, giving rise to bichëe ‘screw in the wrong way’. As already mentioned, most compound verbs belong to the bivalent or trivalent verb classes, as long as the second element is itself bivalent. Here again, their object can refer to location (87), patient (88), etc. (87) È tètùrù chaa xwâkûû-purèkwé. 3sg hand+pierce one shard-bottle ‘He hurt his hand on a shard of bottle.’ (88) Xötö söpuru na farawa ngê nea. boy circular.hand.movement+break.in.two pst bread ins knife ‘The boy cut the bread with a knife.’ Among the 70 verbs included in the database, 22 are compounds in Xârâcùù, among which fètaa ‘leave’, chéxô ‘cough’, paiimè ‘die’ and cipwîrî ‘roll’ (monovalent verbs), jikörö ‘break’, fîgai ‘beat’, söpuru ‘cut’ and chètia ‘tear’ (bivalent verbs).

7 Oblique arguments or adjuncts? The remaining problem is to account for the use of some prepositions for both oblique arguments and adjuncts, but this is a well known problem that Xârâcùù does not help much to resolve, even if we have seen that different uses of the same preposition often have correlates and restrictions concerning the position of the complement, and the (in)separability of the preposition with respect to the phrase it introduces. For example, we have seen that as an instrumental preposition or as an oblique object marker, the preposition ngê can be separated from its complement, while as a subject marker or as a temporal adjunct marker, it cannot. As is the case in many languages, it is indeed difficult to determine the status of prepositional noun phrases. Still, a distinction between adjuncts and obliques may be observed in some operations, such as fronting, nominalization or argument deletion, but is never sufficient, to strictly delimit these two types of functions.

1056

Claire Moyse-Faurie

7.1 Fronting One of the possible criteria that could be relevant to distinguish adjuncts and oblique arguments is the position of the preposition when the noun phrase is fronted. Two oblique prepositions, ngê and wâ, can occur sentence-initially: adjuncts keep these prepositions when fronted (89) while oblique arguments are fronted without them, the preposition remaining after the predicate (90b). Instruments are in between since they keep the preposition when fronted, but the preposition occurs a second time after the predicate (91b). (89) Ngê chêêdè, è wâ toa. ins evening 3sg pfv arrive ‘In the evening, s/he arrived.’ In (90b), the fronting of the oblique object necessitates the occurrence of a resumptive pronoun (è); the preposition is maintained after the main verb, cliticized to it, and is thus separated from the resumptive pronoun by the past tense marker: (90) a. Famuru sôôbö ii na ngê pupèè rèè. Famuru play.with always pst ins doll poss+3sg ‘Famuru used to play with her doll all the time.’ b. Pupèè rèè, Famuru sôôbö ii ngê na è. doll poss+3sg Famuru play.with always ins pst 3sg ‘Her doll, Famuru used to play with it all the time.’ By contrast, in (91b), the preposition ngê is maintained in front of the fronted instrumental complement, and it occurs also just after the verb, separating the verb from its direct object: (91) a. È wâ sa ri ngê kwâdè. 3sg pfv hit 3pl ins wind ‘He started to hit them with the wind.’ b. Ngê kwâdè, è wâ sa ngê ri. ins wind 3sg pfv hit ins 3pl ‘The wind, he started to hit them with [it].’ (Lit. ‘With the wind, he started to hit with them.’)

7.2 Object relativization As mentioned earlier, Xârâcùù has no passive voice. However, a type of object relativization with the nominalizing prefix êê- expressing resultative events often

Valency classes in Xârâcùù (New Caledonia)

1057

renders French or English passives. Adjuncts, by contrast, do not seem to be possible antecedents in such nominalized relative constructions. The agent – the subject in the original verb phrase (92a) – is expressed as a possessor, reflecting the origin of the construction in a nominalized resultative construction. The agent is optional with some verbs, as in example (92b): (92) a. Aaxa su péci. chief write letter ‘The chief is writing a letter.’ b. péci êê-su (rè aaxa) letter [res-write poss chief] ‘letter (which has been) written (by the chief)’ It is compulsory with other verbs which require the overt expression of subject (the possessor in the relative clause), as in (93a): (93) a. Kâmîâ chä bwaa-rè. sun [strike(sun)] head-3sg ‘The sun is beating his head down.’ b. bwaa-rè êê-chä rè kâmîâ head-3sg [res-strike(sun) poss sun] ‘his head got struck by the sun’ If the object as antecedent refers to an animate entity, a resumptive pronoun necessarily occurs after the agent (which is expressed as a possessor); the resumptive pronoun does not have the object pronominal form, but the tonic form (niè), simply juxtaposed to the possessive noun phrase, without forming part of it: (94) Oosi êê-fawîjö na röö niè, wâ tââtoa mûgé. horse [res-caus.drink pst poss+2sg 3sg.indep] pfv stand.up again ‘The horse to whom you gave something to drink set himself on his feet again.’ Oblique arguments may become the head of such relative nominalized clauses and the oblique preposition keeps its position after the nominalized verb: (95) döö êê-pia tùù earth [res-fight about] ‘the earth about which [we] fought’ If the oblique argument refers to an animate entity, e.g. the beneficiary in (96), the preposition which remains in situ is followed by a resumptive pronoun:

1058

Claire Moyse-Faurie

(96) Afädë êê-xù na rè rî nô xù è, è wâ fè na. foreigner res-give pst poss 1pl.incl fish ben 3sg 3sg pfv go pst ‘The foreigner to whom we had given some fish has left.’ If the oblique argument refers to an inanimate entity, e.g. the theme in the following examples, no resumptive pronoun is required, as we can see in (97b): (97) a. Xuu nèèxu êê-fâda röö niè ngê nêkwââ, wâ pua small girl res-caus.eat poss+2sg 3sg.indep ins leaf pfv stay xöru mûgé. well again ‘The young girl you fed with medicinal leaves is now doing well.’ b. Nêkwââ êê-fâda röö xuu nèèxu ngê, tuè cécöö tö leaf res-caus.eat poss+2sg small girl ins grow no.matter loc nèca. gardens ‘The medicinal leaves with which you fed the young girl grow anywhere in the gardens.’

7.3 Argument deletion Going through the dative alternation (cf. § 4.4), I showed that the addition of a clausal complement shifted the recipient from an unmarked object to oblique marking, as in (98a) and (98b) repeated below: (98) a. È fi rö. 3sg lie 2sg ‘He is lying to you.’ b. È fi xù rö mè siè kii. 3sg lie ben 2sg comp not.exist key ‘He is lying to you [saying that] he doesn’t have any key.’ If a noun phrase introduced by the instrumental preposition is added (ngê xwâvirè ‘with unkindness’), the recipient (îrî) remains unmarked. In this case, it is clear that the prepositional phrase ngê xwâvirè is an adjunct, and not an oblique object: (98) c. È fi ngê îrî xwâvirè. 3sg lie ins 1pl.incl unkindness ‘He is unkindly lying to us.’ (Lit. ‘He is lying us with unkindness.’)

Valency classes in Xârâcùù (New Caledonia)

1059

This situation allows us to distinguish, at least in some cases, between adjuncts, which do not require the oblique marking of the recipient, and oblique objects, with which the recipient also has to be introduced by a preposition, as we have seen in § 3.2.3.

8 Conclusion Xârâcùù stands apart from prototypical Oceanic languages. It has unmarked SVO word order, having lost any person indexation on the verb. It has lost almost all the productive transitivizing and applicative suffixes, with the exception of the causativizing prefix, and it has no productive intransitivizing prefix; it has no verbal morphology, in particular, no verbal inflections depending on the object, and no actancy variation depending on tense-aspect marking. In addition, the pronominal markers almost all have the same form whatever their functions, so this does not help determine their syntactic status. Due to unknown historical factors, most of the argument encoding strategies now rely on the choice between different prepositions, or on the choice between different types of deletion alternations. Xârâcùù has rich verbal compounding patterns, along with productive serial verb constructions, and a number of oblique prepositions of verbal origin. These developments certainly explain the multiplication of small verbal classes, and the fact that valency classes are more differentiated than in other Oceanic languages, partly due to the numerous oblique prepositions, some of which are specific to a small group of verbs. Xârâcùù has indeed several prepositions of (ad)verbal origin, grammaticalized through the development of serial verb constructions. This fact accounts for some cases of indirective alignment. Most of the prepositions have a choice of position, either immediately postposed to the verb, in which case they can be separated from their complement, or preposed to the latter. Consequently, besides the difficulty of differentiating between adjuncts and oblique objects, there is also the difficulty of differentiating in some cases between adverbs and prepositions, as it is the case for example with the purposive cè, or the disassociative taa. The boundaries of each verb class are not easy to delimit, with certain verbs so flexible that they can combine with several different prepositions (as for example xwiri ‘sell’), or they belong to different classes across speakers. Still, I will try to summarize some of the most interesting Xârâcùù characteristics concerning valency flagging and alternations. Grammaticalization of possessive classifiers as recipient/beneficiary markers has been recorded in other Oceanic languages (cf. Song 2007), and it is not surprising to find such cases in Xârâcùù, such as the use of êê-/nèxêê-/ngêê- ‘belongings’, to introduce the beneficiary. It is the additional use of the preposition xù (< xù ‘give’) which is a Xârâcùù innovation among the New Caledonian languages. The alternation of recipient/beneficiary expressed either by a prepositional phrase or

1060

Claire Moyse-Faurie

as a possessor has correlates in terms of requirement in favor of the beneficiary. The fact that experiencers and beneficiaries can both be flagged as possessors, as I have shown in § 4.2, is the syntactic manifestation of the close relationships among concepts such as possession, gift, and personal experience. Concerning verb classes, we noted the existence of a small ‘inherently reflexive’ verb class (§ 3.2.2), which requires a pronominal object coreferential with the subject; this minor class contrasts with verbs which also require two overt arguments, but which show a degree of animacy alternation (§ 4.1.4), allowing variation in their choice of objects. Other features, such as the existence of some labile verbs with well-known specific semantic orientation are typologically more common. Some more fieldwork would be necessary for a better understanding of the different Xârâcùù verb classes and alternations, even if the lack of any normative tendency among speakers does not make it easy!

Meaning label

RAIN

SHAVE

MEET

BLINK

COUGH

RUN

SING

LAUGH

SCREAM

FEEL PAIN

FEEL COLD

DIE

PLAY

BE SAD

BE HUNGRY

#

69

14

17

46

47

49

53

57

58

59

60

61

62

63

64

Appendix

mêrê

nârâyaa (Co)

mwârâ

paiimè (Co)

xùpè

paii

ngââ

mwârâ

xô

pûxûrû

chèxô (Co)

nimè

pùùtè

xii

kwiè

Verb form

1>V

1>V

1>V

1>V

1>V

1>V

1>V

1>V

1>V

1>V

1>V

1>V

1>V

1>V

V

Coding frame schema

–

–

–

–

–

–

–

–

–

–

–

Ambitransitive 1

–

–

–

–

–

–

–

–

–

–

–

–

–

+

–

Ambitransitive 2

–

–

+

–

–

–

–

+

–

–

–

–

–

–

–

Applicative

+

–

+

m

+

+

m

+

+

m

+

+

m

m

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Causa- Conative tive

–

–

–

+

–

–

–

+

Impersonal

–

–

–

–

–

–

–

–

–

+

–

–

Middle

+

+

+

+

+

+

+

+

+

+

m

+

+

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Nomi- Obnaliject zation omission

–

–

–

–

–

–

–

–

–

–

Oblique/ Direct

–

–

–

–

–

–

–

–

–

–

–

–

–

Refl./ Recip. omission

Valency classes in Xârâcùù (New Caledonia)

1061

Meaning label

ROLL

SINK

BURN

BE DRY

RAIN

BE A HUNTER

BOIL

APPEAR

BE ILL

CRY

SIT

JUMP

GO

LEAVE

LIVE

EAT

EAT

EAT

#

65

66

67

68

69

70

80

81

82

83

50

52

54

55

56

 1

 1

 1

xwè

kê (2)

da

muru

fè

fè

cîchéé (Co)

cuè

têî

paii

abaa

bö

acaa (CN)

xwa (N)

mëtë

kê (1)

mutu

cipwîrî (Co)

Verb form

1>V2

1>V2

1>V2

1 > V (LOC2)

1 > V (LOC2)

1 > V (LOC2)

1 > V (LOC2)

1 > V (LOC2)

1>V

1>V

1>V

1>V

1>V

1>V

1>V

1>V

1>V

1>V

Coding frame schema

–

–

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

Ambitransitive 1

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

+

–

–

Ambitransitive 2

–

–

–

+

–

–

–

–

+

–

–

–

–

–

–

–

–

–

Applicative

m

+

+

+

+

+

+

+

+

+

+

m

+

+

+

+

+

m

m

m

–

–

–

–

–

–

–

–

–

–

–

Causa- Conative tive

v

–

–

–

–

–

–

–

+

–

–

–

–

+

–

–

–

–

–

–

–

–

–

–

ImMidperso- dle nal

m

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

m

+

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Nomi- Obnaliject zation omission

–

m

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Oblique/ Direct

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Refl./ Recip. omission

1062 Claire Moyse-Faurie

HUG

LOOK AT

SEE

SEE

SMELL

FRIGHTEN

LIKE

KNOW

KNOW

THINK

SEARCH FOR

WASH

DRESS

HELP

FOLLOW

KILL

KILL

TOUCH

PEEL

PEEL

CLIMB

 2

 3

 4

 4

 5

 7

 8

 9

 9

10

11

12

13

15

16

26

26

29

33

33

48

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

péré

cee

tapöru

xâânîî

pwââmè

sa

fèèté (Co)

ciiwi

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

famwâmwâ (Caus) 1 > V 2

ngûrû

piicè (Co)

nârâ

tâmwâ

néxä

xwèrii

fabata (Caus)

bu

xâpârî (Co)

téé

téé fè

satee (Co)

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

+

+

+

+

+

+

m

m

m

m

+

–

–

+

m

+

m

+

+

m

–

–

–

–

–

–

m

–

–

–

–

–

–

–

–

–

–

–

–

–

m

–

–

+

+

m

+

+

+

+

+

m

+

+

+

+

+

+

m

m

+

+

+

m

+

+

m

m

m

–

m

–

–

–

–

–

–

–

–

m

m

Valency classes in Xârâcùù (New Caledonia)

1063

Meaning label

LEAVE

DIG

DIG

PUSH

HEAR

COOK

WANT

FEAR

SIT DOWN

TEAR

WIPE

CARRY

CARRY

THROW

THROW

TIE

PUT

POUR

BRING

#

55

73

73

74

78

79

87

 6

51

32

72

38

38

39

39

40

41

42

75

pè mê

xwêê

suè

curu

tê

gwéré

akö

pè mê (Dir)

jee

chètia (Co)

cuè chéé

bata

xwèrii

chutââ

xwata

sùù

ja

kîî

fètaa

Verb form

1 > V 2 (LOC3)

1 > V 2 (LOC3)

1 > V 2 (LOC3)

1 > V 2 (LOC3)

1 > V 2 (LOC3)

1 > V 2 (LOC3)

1 > V 2 (LOC3)

1 > V 2 (LOC3)

1 > V 2 (kè+3)

1 > V > 2 (kèwâ+3)

1 > V tö+2

1 > V taa+2

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

1>V2

Coding frame schema

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Ambitransitive 1

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

Ambitransitive 2

–

–

–

–

–

–

–

–

–

–

–

+

–

–

–

–

–

–

–

Applicative

+

+

+

+

+

+

+

+

+

+

+

+

m

+

+

+

+

m

–

–

–

Causa- Conative tive

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

ImMidperso- dle nal

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

–

+

+

–

–

–

–

+

Nomi- Obnaliject zation omission

–

–

–

–

Oblique/ Direct

–

–

–

m

–

–

–

–

Refl./ Recip. omission

1064 Claire Moyse-Faurie

NAME

BUILD

BREAK

KILL

KILL

BEAT

HIT

HIT

TOUCH

CUT

COVER

COVER

FILL

LOAD

GRIND

DIG

MAKE

ASK FOR

TAKE

HIDE

STEAL

GET

23

24

25

26

26

27

28

28

29

30

43

43

44

45

71

73

85

19

31

34

76

86

pè

pêdè (Co?)

sûûârâ

pè

nââ

xwi

sëi

bikörö

nîî

faxûda (Caus)

sache

kasache

söpuru (Co)

sakai (Co)

chëi

fîda (Co)

fîgai (Co)

pwèkè (Co)

söömè (Co)

jikörö (Co)

xwi

xacè (Co)

1 > V 2 taa+3

1 > V 2 taa+3

1 > V 2 taa+3

1 > V 2 taa+3

1 > V 2 taa+3

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

1 > V 2 (ngê+3)

–

–

–

–

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

+

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

+

m

m

+

–

+

+

+

+

+

m

+

+

+

+

+

+

+

m

+

+

–

–

m

–

m

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

–

+

+

+

+

+

+

+

m

+

+

+

m

+

+

+

+

+

+

+

m

+

m

–

m

m

–

m

+

–

–

–

–

m

–

–

Valency classes in Xârâcùù (New Caledonia)

1065

GET

ASK FOR

SAY

SHOW

GIVE

SEND

BRING

SHOUT AT

TELL

TELL

TEACH

TALK

TIE

86

19

22

35

36

37

75

20

21

21

77

18

40

fëi

tèpe

faînû

sömârâ (Co)

faxwata (Caus)

xangââ (Co)

mê kèrè

nû

xù

xacië (Co?)

ché

nêgé

töpwé

Verb form

1 > V 2 (LOC3) (ngê+4)

1 > V wâ+2 (xù+3)

1 > V ngê+2 xù+3

1 > V ngê+2 xù+3

1 > V ngê+2 xù+3

1 > V ngê+2 xù+3

1 > V 2 xù+3

1 > V 2 xù+3

1 > V 2 xù+3

1 > V 2 xù+3

1 > V 2 xù+3

1 > V 2 xù+3

1 > V 2 taa+3

Coding frame schema

–

–

–

–

–

–

–

–

–

–

–

–

–

Ambitransitive 1

–

–

–

–

–

–

–

–

–

–

–

–

–

Ambitransitive 2

–

–

–

–

–

–

–

–

–

–

–

–

–

Applicative

m

+

m

m

m

+

m

+

–

+

+

+

–

Causa- Conative tive

m

m

–

–

–

–

–

–

–

–

–

m

–

–

–

ImMidperso- dle nal

m

+

+

+

+

+

+

+

+

+

+

+

+

+

–

m

m

Nomi- Obnaliject zation omission

Legend: Verbs: Caus – causative, Co – compound verb, Dir – directional, – nominal, + regularly, – never, m marginally, _ no data

Meaning label

#

+

+

m

–

Oblique/ Direct

Refl./ Recip. omission

1066 Claire Moyse-Faurie

Valency classes in Xârâcùù (New Caledonia)

1067

Abbreviation ANAPH COLL CLS DEIC DIR INDEP IPERS REC SM

anaphoric collective (human) classifier deictic directional independent impersonal (pronoun) recipient subject marker

References Churchward, C. Maxwell. 1953. Tongan Grammar. Oxford: Oxford University Press. Craig, Colette & Ken Hale. 1988. Relational preverbs in some languages of the Americas: Typological and historical perpectives. Language 64(2). 312–344. Creissels, Denis. 2007. Impersonal and anti-personal constructions: a typological approach. Extended version of the paper presented at Alt-7, Paris, September 24–28, 2007. Dixon, R. M. 1977. Where have all the adjectives gone? Studies in Language 1. 19–80. Donohue, Mark. 2001. Coding choices in argument structure. Austronesian applicatives in texts. Studies in Language 25(2). 217–254. Evans, Bethwyn. 2003. A Study of Valency-Changing Devices in Proto Oceanic (Pacific Linguistics 539). Canberra: The Australian National University. Foley, William A. & Mike Olson. 1985. Clausehood and verb serialization. In Johanna Nichols & Anthony C. Woodbury (eds.), Grammar Inside and Outside the Clause, 17–60. Cambridge: Cambridge University Press. Hagège, Claude. 2010. Adpositions (Oxford Studies in Typology and Linguistic Theory). Oxford: Oxford University Press. Haspelmath, Martin. 1993. More on the typology of inchoative / causative verb alternations. In Bernard Comrie & Maria Polinski (eds.), Causatives and Transitivity, 87–120. Amsterdam: John Benjamins. Haspelmath, Martin. 2005. Ditransitive constructions. The verb ‘Give’. In Martin Haspelmath, Matthew Dryer, David Gil & Bernard Comrie (eds.), The World Atlas of Language Structures, 426–429. Oxford: Oxford University Press. Haspelmath, Martin. 2011. On S, A, P, T, and R as comparative concepts for alignment typology. Linguistic Typology 15(3). 535–567. Lichtenberk, Frantisek. 1985. Syntactic-category change in Oceanic languages, Oceanic Linguistics 24(1–2). 1–84. Lichtenberk, Frantisek. 2002. The possessive-benefactive connection. Oceanic Linguistics 41. 439–474. Lynch John, Malcolm Ross & Terry Crowley. 2002. The Oceanic Languages (Curzon Language Family Series). Richmond: Curzon. Malchukov, Andrej, Martin Haspelmath & Bernard Comrie. 2010. Studies in Ditransitive Constructions. A Comparative Handbook. Berlin/New York: Mouton De Gruyter. Margetts, Anna. 1999. Valence and Transitivity in Saliba, an Oceanic Language of Papua New Guinea (MPI Series in Psycholinguistics). Nijmegen: Max Planck Institute for Psycholinguistics.

1068

Claire Moyse-Faurie

Margetts, Anna. 2007. Three-participants events in Oceanic languages, Oceanic Linguistics 46(1). 71–127. Margetts, Anna & Peter K. Austin. 2007. Three-participant events in the languages of the world: Towards a cross-linguistic typology. Linguistics 45(3) (Special issue edited by P. Brown, B. Narashinran and S. Eizenbeiss). 393–451. Moyse-Faurie, Claire. 1983. Le drehu, langue de Lifou (Îles Loyauté). Phonologie, morphologie, syntaxe (Langues et cultures du Pacifique 3). Paris: Peeters-Selaf . Moyse-Faurie, Claire. 1991. Relational morphemes and a transitivising suffix in Xârâcùù (New Caledonia). In Robert Blust (ed.), Currents in Pacific Linguistics: Papers on Austronesian Languages and Ethnolinguistics in Honour of George W. Grace, Pacific Linguistics, C-117. 305–320. Canberra: ANU Moyse-Faurie, Claire. 1995. Le xârâcùù, langue de Thio-Canala (Nouvelle-Calédonie). Éléments de syntaxe (Langues et Cultures du Pacifique 10). Paris: Peeters-Selaf. Moyse-Faurie, Claire. 1997. Phénomènes d’incorporation dans quelques langues océaniennes. Studi italiani di linguistica teorica e applicata, XXVI, Rome, 1997(2), 227–246. Moyse-Faurie, Claire. 2010. L’ambitransitivité: exemples océaniens. In Franck Floricic (ed.), Essais de typologie et de linguistique générale. Mélanges offerts à Denis Creissels, ENS éditions. Moyse-Faurie, Claire & Françoise Ozanne-Rivierre. 1983. Subject case markers and word order in New Caledonia and Loyalty Islands languages. In Papers from the Third International Conference on Austronesian Linguistics, Canberra, Pacific Linguistics C-77. 113–152. Nichols, Johanna. 1982. Ingush transitivization and detransitivization. Berkeley Linguistics Society 8. 445–464. Nichols, Johanna. 1984a. Direct and oblique objects in Chechen-Ingush and Russian. In Frans Plank (ed.), Objects: Towards a Theory of Grammatical Relations, 183–209. New York: Academic Press. Nichols, Johanna. 1984b. Transitivity and valence in Chechen-Ingush. Folia Slavica 1(2). 254–267. Nichols, Johanna. 1986. On form and content in typology. In Winfred Lehmann (ed.), Language Typology, 141–162. Current Issues in Linguistic Theory, Vol. 47. Amsterdam: John Benjamins. Osumi, Midori. 1995. Tinrin Grammar (Oceanic Linguistics Special Publication 25). Honolulu: University of Hawai’i Press. Ozanne-Rivierre, Françoise. 1976. Le iaai. Phonologie, morphologie, syntaxe. Paris: Peeters-Selaf. Ozanne-Rivierre, Françoise & Jean-Claude Rivierre. 2004. Verbal compounds and lexical prefixes in the languages of New Caledonia. In Isabelle Bril & Françoise Ozanne-Rivierre (eds.), Complex Predicates in Oceanic Languages Studies in the Dynamics of Binding and Boundness, 347–371. Berlin: Mouton de Gruyter. Pawley, Andrew & Lawrence A. Reed. 1979. 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: University of Michigan. Ross, Malcolm. 2004. The morphosyntactic typology of Oceanic languages. Language and Linguistics 5(2). 491–541. Song, Jae Jung. 1998. Benefactive marking in Oceanic languages: From possessive classifiers to benefactive markers. In Anna Siewierska & Jae Jung Song (eds.), Case, Typology and Grammar: In Honor of Barry J. Blake, 247–275. Amsterdam: John Benjamins. Song, Jae Jung. 2007. Getting three out of two. The development of a three-participant construction in Oceanic languages. Functions of Language 14(1). 127–148.

Nicholas Evans

26 Valency in Nen 1 Introduction1 Nen, (Ethnologue Code NQN) is a language of the Morehead-Maro Family of Southern New Guinea, also known as Nen Ym or Nen Zi (pronounced [nɛn dʒi]).2 The Morehead-Maro family, with around a dozen members3, straddles two modern nations, Papua New Guinea (PNG) and Indonesia (Merauke district), in a low-lying area bounded to the north and east by the Fly River, to the south by the Torres Strait which separates New Guinea from Australia, and to the east by the Maro river. The Morehead-Maro family, on our current knowledge, appears unrelated to any other language family. In particular it is unrelated to the giant Trans-New Guinea (TNG) family, which with its 400+ languages has tended to dominate typologists’ perceptions of what so-called “Papuan”4 languages are like, as shaped by typological overviews like Foley (1986, 2000). Nonetheless, languages which have been claimed to belong to the TNG family figure prominently to the west (Marind, Marori), north (Suki, Tirio) and east (Kiwai). Just south, in the western Torres Strait, the Morehead-Maro family had regular interactions with an Australian language of the Pama-Nyungan family, Kala Kawaw Ya. Morehead-Maro also appears to be unrelated to two other families which were grouped together by e.g. Ross (2005)

1 For financial support of my PNG fieldwork I thank the Australian National University (Professorial Setup Grant), the Australian Research Council (Discovery Projects Language and Social Cognition and Languages of Southern New Guinea), and the Volkswagenstiftung (DoBeS Grant Nen and Tonda). I would like to thank the Valency project organizers and participants for the stimulus they provided to exploring Nen valency further, and in particular Andrej Malchukov, Martin Haspelmath, Iren Hartmann, Ger Reesink and Bernard Comrie for their comments on an earlier draft of this chapter. 2 Language names in the region are based on their word for ‘what’ – nen in Nen, idi in Idi, nambu in Nambo – plus an optional extra word such as Zi ‘word, language’ or ym ‘is’ – as if English were called Whatish, German Wasisch, French Quoiais, and Russian štoskij. 3 More precise figures will need to await further research – the low level of knowledge we have of most languages in the group makes it hard to determine where dialect chains end and separate languages begin. The most likely scenario, based on current data (see Evans 2009), is that the family divides into four branches, two on the Indonesian side and two on the PNG side: from West to East, Kanum (many dialects, some perhaps best treated as distinct languages), Yei (at least 2 divergent dialects), Tonda (five languages: Mblafe/Renmo, Guntai, Kancha, Were/Kamnzo, Arammba) and Nambu (seven languages: Nen, Nambo/Namna, Nama, Namo, Neme, Len, and Dre). 4 ‘Papuan’ does not denote any genetic unit and is hence a rather misleading term – it includes all languages of the South-western Pacific that are neither Austronesian nor Australian, and as such contains something like 40 maximal genetic units (including isolates), exhibiting great diversity among themselves.

1070

Nicholas Evans

Map 1: main language groups of Southern New Guinea (showing selected languages only).

as the ‘South-Central Papuan’ languages – the Pahoturi River languages (including Idi) to its east, and the Yelmek-Maklew languages that lie some 150 km to the west of the Morehead-Maro family. See Evans (2012) for an overview of the linguistic situation in Southern New Guinea. The Southern New Guinea region is new land, built up by progradation in the last few millennia, and exhibits great apparent variation in both the languages and physical characteristics of its inhabitants. With a great number of apparently unrelated language families, it is one of the most important foci of linguistic diversity within the most linguistically diverse part of the earth’s surface, yet we know vanishingly little about the languages of the region. No language of the MoreheadMaro family yet possesses more than a sketchy description,5 and this chapter, confined as it is to a relatively early stage of research,6 must be regarded as provisional in the data and analyses it presents.

5 Best materials so far are Boevé & Boevé (2003) on Arammba and Sarsa (2001) on Wara, and sections of Martin (2001) on tense in Nambo, along with brief older material in Drabbe (1954) and Boelaars (1950). 6 Data presented here was gathered by the author over 5 fieldtrips, totalling 14 weeks, between 2008 and 2012. During 2010 I worked through the Valency Questionnaire then checked some gaps

Valency in Nen

1071

Nen itself is primarily spoken in just one village of around 300 people, Bimadbn. Marriage, based on sister-exchange,7 produces de facto linguistic exogamy, so that many wives residing in the village have another language as their mother tongue, but have usually learned fluent Nen. Conversely, women for whom Nen is a mother tongue are to be found living in nearby surrounding villages where they have moved after marriage. Most residents of Bimadbn know at least one other neighboring language (typically Nambo, which is closely related, and Idi, which is unrelated). Some older people know Hiri Motu and many young to middle-aged people speak fluent English; use of Tok Pisin is limited. Despite its small speaker population, Nen is currently secure and being learned by children of all ages. Typologically, Nen is a predominantly SOV language, with case roles shown by suffixing, an ergative/absolutive system of case marking, and complex verbal morphology involving both prefixes and suffixes. Prefixes belong to one of three series for which it is impossible to give consistent glosses (and which will therefore just be glossed α, β and γ when we have reason to segment them), but once these combine with suffixes an intricate system of TAM results. Argument indexing normally indexes the undergoer by prefix and the actor by suffix. Indexing in one-place predicates splits along the lines of stative vs. dynamic predicates. Non-finite constructions, generally formed by suffixing a case marker to a nominalized verb stripped of all inflection, are widely employed. Verbs possess a developed system of ‘diathetic prefixes’, lying between the root and the undergoer/TAM prefix, which play a key role in overtly coding valency alternations. For its phoneme inventory and orthography see Appendix 1. Note that like many Papuan languages (most famously Kalam – Blevins & Pawley (2010)) – many syllables lack phonemically specified vowels, though brief epenthetic vowel phones (typically schwa) may be present. The main clause types are given in Table 1. A and U stand for actor suffix and undergoer prefix, M for the ‘middle’ prefix, RR for reflexive/ reciprocal diathetic prefix, and DI for the diathetic prefix slot; X + Y indicates that X normally precedes Y, while X > Y indicates a more rigid ordering of X before Y. NP^ indicates a lexically fixed NP forming part of an idiomatic expression, but nonetheless occupying a case-marked and cross-referenced argument slot.

and anomalies during the 2011 and 2012 field seasons. Most importantly I thank my Nen teachers, particularly Michael Binzawa, †Aramang Wlila and Jimmy Nébni, for their sensitive and intense involvement in explaining their language’s intricacies to me, as well as to the whole village of Bimadbn for its hospitality and friendship. 7 See Williams (1936) and Ayres (1983) for ethnographic background on marriage patterns in the Morehead region.

1072

Nicholas Evans

Tab. 1: Summary of major valency patterns and case/voice alternations. Clause type

Flagging

Indexing

Sample verb(s)

1 Basically Monovalent 1.1

Basic intransitive (stative)

NPABS

U-V

m ‘be’; tromngr ‘be erected’

1.2

Inherent middle (dynamic)

NPABS

M-V-A

armbs ‘ascend’, ogyabs ‘work’

1.3a

Inherent middle with cognate object

NPABS (+NPABS)

M-DI-V-A

momae ke otärs ‘cough’

1.3b.

Derived middle (e.g. refl/recp, decausative)

NPABS (+NP:RRABS)

M-RR-V-A

awakaes ‘look at oneself/ each other’, edrers ‘tear, i.e. get torn’

2 Basically Divalent 2.1

Basic Transitive

NPERG + NPABS

U-V-A

wakaes ‘see, look at’

2.2

Transitive (experiencer object)

NPABS > NP^ERG

U-V-A3sg

gersäm räms ‘X be hungry’, lit. ‘hunger does X’

2.3

Transitive deponent middle

NPERG + NPABS

M-DI-V-A

ors ‘leave, divorce’

2.4

Semi-transitive

NPABS + NPOBL

U-V-A

wetrs ‘help’

(3.2.

Ditransitive: theme-indexing)8

NPERG + NPABS + NPDAT/OBL

UT-V-A

oyabs ‘exhibit’

NPERG + NPABS + NPDAT/OBL

UR-V-A

räms ‘give’

3 Basically Trivalent 3.1

Ditransitive: recipient-indexing

8 In brackets here, because it can be more parsimoniously treated as a transitive verb with an oblique adjunct – see below.

Valency in Nen

1073

2 Basics of Nen morphosyntax Coding of grammatical relations in Nen makes use both of flagging (absolutive vs. ergative case, as well as around ten other cases), and verbal indexing. Though there is considerable pragmatic flexibility, constituent order is normally SOV – or more accurately, AUV since there is no real evidence for ‘subject’ as a grammatical relation above and beyond the categories that can be triangulated from the case plus verb agreement systems.9 The exception is in ‘experiencer object’ constructions (§ 3.2.2) where the absolutive undergoer precedes the ergative ‘stimulus’ giving a UAV order.

2.1 Flagging (case suffixation) Case suffixes are organized on an absolutive/ergative system for core participants. Free pronouns don’t distinguish number in the absolutive, but distinguish singular from non-singular in other cases.10 Number is organized on a singular vs. nonsingular basis for most free pronouns except those in the absolutive (which neutralize number), while for nouns the situation is more complex with different syncretisms in different cases: for är ‘man, person’, for example, the absolutive makes no number distinction, the ergative distinguishes singular from dual/paucal from plural, the oblique distinguishes singular from dual from paucal/plural and the first dative distinguishes all four values (singular äreita, dual äräbet, paucal ärbet and plural ärbeita). In general, then, free pronouns contrast singular vs. non-singular while nouns contrast up to four numbers, though most cases syncretize some of the number values. Within the verb, a singular vs. non-singular distinction in indexed arguments is crossed with a dual vs. non-dual distinction either on the

9 I follow Bickel (2011: 443) in taking as primary, for the purposes of defining grammatical relations, the properties characterizing particular constructions (e.g. case assignment, agreement etc.) and then – to the extent useful – building the definition of grammatical relations on these. If we do this for Nen, we can define grammatical relations as secondary notions in the following rather contorted way. Subject: either an ergative NP exhibiting actor agreement, or an absolutive NP exhibiting some verb agreement (in the absence of an ergative NP). Indirect object: dative NP exhibiting undergoer verb agreement. Object: Absolutive NP exhibiting undergoer agreement, in the presence of some other NP exhibiting actor agreement. It will occasionally be useful to employ one or more of these terms as handy metalinguistic terms. 10 Additionally, 1 st and 2 nd person singular free pronouns don’t distinguish absolutive from ergative; thus ynd can mean ‘1sgABS|ERG; 1 st person ABS (any number)’; the non-singular ergative form is yndbem. Likewise bm can mean ‘2sgABS|ERG; 2 nd person ABS (any number)’; the non-singular ergative form is bmbem. This makes 1/2sg forms bad diagnostics for coding frames, whereas 3 rd person or non-singular forms, on the other hand, always give a clear result. To simplify the exposition in this paper, I use ERG or ABS glosses as appropriate with 1sg and 2sg pronouns on the basis of what case would be chosen with some other person/number combination.

1074

Nicholas Evans

stem itself or as part of the TAM-sensitive ‘thematic’ immediately after it to give a three-way number distinction on most verbs – singular, dual and plural. Absolutive arguments, in some constructions are capable of distinguishing a fourth number, further splitting the plural into either a non-exhaustive vs. exhaustive or paucal vs. plural distinction according to a range of different strategies too complex to discuss here. (1) a. Bä/är mer y-m. 3abs/man(abs) good 3sgU:α-be:nd:ipfv ‘He/she/the man is good.’ b. Bä/terber är mer yä-m. 3abs/many man(abs) good 3nsgU:α-be:nd:ipfv ‘They (3 or more)/the men are good.’ c. Bä/sombes är mer yä-ren. 3abs/two man(abs) good 3nsgU:α-be:du:ipfv ‘They two/the two men are good.’ (2) a. Ymam/är-äm bä y-aka-t-e. 3sg.erg/man-sg.erg 3abs 3sgU:α-see-nd:ipfv-3sgA ‘(S)he/The man sees him/her.’ b. Ymam/är-äm ynd w-aka-t-e. 3sg.erg/man-sg.erg 1abs 1sgU:α-see-nd:ipfv-3sgA ‘(S)he/The man sees me.’

2.2 Verb morphology On grounds of morphological indexing, verbs can be divided into a) prefixing11 verbs, whose sole (‘undergoer’12) argument is indexed by prefix (1) b) ambifixing verbs, with argument indexing on both sides of stem; these divide in turn into Transitive: patient (‘undergoer’) is indexed by prefix, actor by suffix (2)13 11 The terms ‘prefixing’ and ‘ambifixing’ are based just on the capacity to index arguments. Even so-called ‘prefixing’ verbs, however, are able to take some other kinds of suffix, most importantly some TAM suffixes. 12 I use scare quotes around ‘undergoer’ because (a) there are a few prefixing verbs – ‘come’ and ‘go’, both directional forms of ‘be’, and ‘walk’ – which are indexed by an undergoer prefix even though the sole argument is in fact agentive, and (b) because the undergoer prefix also codes recipients with some verbs. 13 The separation of actor and undergoer information holds for the vast majority of actor/undergoer combinations. However, for some combinations involving duals, there are ‘interactive’ prefixes and/or suffixes that encode information about both actor and undergoer. See Evans (forthcoming).

Valency in Nen

1075

Middle: undergoer slot filled by person/number-invariant prefix14 (n-, k- or g-, according to the TAM value); actor slot indexes sole argument (3) (3) a. Bä/togetoge n-owab-ta-t. 3abs/children(abs) M:α-talk-nd:ipfv-3nsgA ‘They/the children (>2) are talking.’ b. Bä/toge n-owab-0-̸ t. 3abs/child(abs) M:α-talk-du:ipfv-3nsgA ‘They two/the two children are talking.’ In addition to the undergoer prefix and actor suffix, the post-stem ‘thematic’ – which encodes a composite of aspect/tense and number (like the -ta- vs. -0̸- contrast in (3)) – is potentially sensitive to the number (dual vs. non-dual) of either indexed argument.15 Nen verbs are best treated as having a circumfixal paradigm, integrating prefixes and suffixes into a large paradigm arrayed on the dimensions of tense/aspect/ mood, actor person/number and undergoer actor person/number, rather than segmented into prefixes and suffixes as has been done so far. Among the advantages to doing this are the clearer semantic values for the glosses (e.g. eliminating arbitrary markers like α, and giving a constructive number value that integrates partial specifications at several morphological sites) and the fact that for some person/ number combinations the clean division into undergoer-coding prefix and actor suffix does not hold. In the rest of this chapter I will therefore bundle together all information in the inflectional prefixes and suffixes and give them a single circumfixal gloss: on this system, (3a) is re-presented as (3c), and (3b) as (3d). Note that morphologically middle verbs index an A but not a U argument, but are not otherwise glossed as middle. I will, however, give separate glosses to the valency-changing ‘diathetic’ prefixes which are of special interest to this chapter (e.g. 4, 5, 26b, c below).

14 It is not quite true that the middle prefix is always person/number invariant. Many Nen verbs allow a fourth number value – a large or exhaustive plural – and when forming these for middle verbs it is the prefix that is used to code the person/number (at least in the the third person, with yaw- ‘3nsgU’ instead of the normal middle prefix n-) while the suffix is filled with an invariant third person singular form (-de in the 3 rd singular): cf. amni naebnde ‘a bird is flying’, amni naebyängt ‘two birds are flying’, amni naebndat ‘several birds are flying’ all with person/number invariant nprefix, as opposed to amni yawaebnde ‘all/a whole lot of birds are flying’. In this sense both prefixes and suffixes are potential coding sites for the person/number properties of the argument of middle verbs. 15 In principle it can encode the dual cardinality of either argument. If one argument is encoded as singular the dual value will be associated with the other argument; if both are non-singular more complex mechanisms come into play, which we lack the space to discuss here.

1076

Nicholas Evans

(3) c. Bä/togetoge n\owab/tat. 3abs/children(abs) 3pla.npreh.ipfv\talk ‘They/the children (>2) are talking.’ d. Bä/toge n\owab/t. 3abs/child(abs) 3duA.npreh.ipfv\talk ‘They two/the two children are talking.’

2.3 Other typological features of Nen It is also helpful, in contextualizing the discussion that follows, to note the following other typological features of Nen. a) Rich TAM system of eleven tense/aspect/mood categories16 which is encoded by combining the prefix and the suffix series. Since TAM is orthogonal to valency I do not go into details here. Note also that while I gloss the suffixal aspectual contrast as ‘perfective’ vs. ‘imperfective’, aspect sensitivity is largely focused on the inceptive-phase boundary, so that the dominant semantic realization of the aspectual contrast is as ‘begin doing’ (perfective) vs. ‘keep doing’ (imperfective). b) Directional prefixes n- ‘towards’ and ng- ‘away from’, which may be placed between the undergoer prefixes and the verb stem, are employed with great frequency. Cf. narmbte ‘(s)he is ascending (neutral)’, nnarmbte ‘(s)he is coming up (towards speaker)’, nngarmbte ‘(s)he is going up (away from speaker)’. The liberal use of directional prefixes impacts on the lexicalization of some verbs in the Leipzig list (BRING, TAKE), which in Nen are translated most accurately using a directional prefix plus a directionally-neutral verb e.g. renzas ‘carry’. In contrast with e.g. Russian, which the system otherwise resembles (e.g. nosit’/prinosit’/unosit’ ‘carry (neutral/towards/away)’), the directional prefixes in Nen are not available in infinitives/nominalized forms, which are always neutral with respect to directionality. Thus there is an infinitive armbs ‘ascend’ but no *narmbs ‘come up’ or *ngarmbs ‘go up’.

16 Namely: irrealis, four imperatives (perfective and imperfective present imperatives, future imperative, and ‘mediated imperative’ sent via a messenger), two perfective declaratives (past and future) and three imperfective declaratives (non-prehodiernal for any time from this morning on into the future, prehodiernal for yesterday and one or two days before that, and remote for more distant past), plus two further aspectually-neutral pasts (the preterite, and the primordial meaning ‘first [in a series of actions, or doing for the first time, or beginning the action without completing it]’. For discussion of how these values are composed from the prefix and suffix combinations see Evans (forthcoming).

Valency in Nen

1077

c) Interclausal syntax uses non-finite verbs, typically inflected for case, rather than the verb chaining more typical of Trans-New Guinea languages:17 (4) a. Ynd anḡ-s-t n\opap/ndm. 1abs return-nmlz-all 1pla:ppf\begin ‘We are about to return.’ b. Ynd bä wnḡi-s-t y\a-pap/ndn. 1sg.erg 3abs tr:stand.up-nmlz-all 1sgA>3sgU.ppf\tr-begin ‘I am beginning to/about to/trying to stand him up.’ Nominalized forms in -s are used as infinitives and as the citation form, e.g. anḡs ‘return, go/come back’, wnḡis ‘to stand up (v.t.)’. Valency-altered forms (e.g. reflexive/reciprocal) have distinct infinitives. Prefixing verbs generally lack infinitives, but in other respects exhibit the same inflectional features and formal patterning as ambifixing verbs.18 As (4a) and (4b) illustrate, not only is argument indexing from verbs ‘raised’ onto phasal auxiliaries – so that the phasal ‘begin/start’ will register one argument for ‘return’ (4a) and two for ‘stand up (tr.)’, but the phasals themselves have distinct forms chosen to match the valency of the lexical verb: the stem for ‘begin to’ is opap for one-place verbs, and (w)apap for two-place verbs.19 Interestingly, there is a third variant on this pattern found with ditransitive verbs such as räms ‘give’20 in (5): here the stem of the phasal auxiliary is extended further by adding the benefactive applicative (w)a- to the transitive stem ∑(w)apap, to give ∑(w)awapap, which then agrees with the indirect object ‘you’. (5) Ahã Gbae ynd begta tande yép räm-s-t here.you.are [name] 1sg.abs 2sg.dat1 1sg.poss bag(abs) give-nmlz-all n\ng-a-wa-pap/ndn. 1sgA>2sgU.ppf\aw-ben-tr-begin ‘Here, Gbae, I’m about to give you my bag.’

17 Incidentally this example shows a further feature of Nen: that with phasal constructions like ‘begin to V’, ‘finish Ving’, BOTH actor and undergoer are raised in the sense of being registered on the phasal verb. We see below that this furnishes a valuable test for valency. 18 The form yls ‘to go/come’ is sometimes used as a sort of de facto infinitive for the directional forms n-m ‘come’ and ng-m ‘go’. 19 Treated here, in glossing, as involving a base middle form √opap which is causativised to yield ∑apap (see § 5.5 for other examples of o- > a- in forming causatives). There is an infinitive opaps ‘to begin’ but no *apaps or *wapaps. 20 Formally ‘give’ is the derived benefactive form of ‘do’, i.e. ‘do for’. There is variation across verb lexemes as to whether the benefactive prefix appears in the infinitive – in this case it doesn’t – but it always appears in some part of the construction – see § 5.6.

1078

Nicholas Evans

d) Diathetic prefixes. Particularly relevant for studies of valency and valency alternations is a slot at the left edge of the stem, optionally filled by one or more ‘diathetic prefixes’. The most important of these are: the transitive prefix w-: cf. amzs ‘sit (v.i.)’, wamzs ‘set, sit (v.t.)’, inherent in many transitive verbs but also present in intransitive/causative pairs21 the reflexive/reciprocal prefix a-/ä-/e-/i- (allomorphs harmonising with the following vowel) cf. wakaes ‘see, look at’, awakaes ‘see, look at each other’ the benefactive applicative (w)a-, which adds an indirect object argument: cf bens ‘feed up, tame [e.g. a pig] (v.t.)’, wabens ‘fatten up [e.g. a pig] for someone [typically as a planned ritual gift]’

Sometimes these are inherent to the verb (as with the w in wakaes ‘see, look at’, or the o- in ogyabs ‘work’), or involve significant semantic shift with regard to the agnate form (wabs ‘count’, owabs ‘talk’; cf. zählen, erzählen). In such non-contrastive or lexicalized cases they will not be glossed separately in examples. In general verbs can take at most one diathetic prefix, but benefactives are an exception in that they can be added to verbs with transitive/causative prefixes (as with a-wa-pap [BEN-CAUS-begin] in (5)).

3 Valency patterns 3.1 Basically monovalent patterns There are three types of clauses with single arguments in the absolutive, indexed on the verb. Subtypes differ in how the verbal organization is manifested – by ‘undergoer’ prefix (‘Basic intransitive’; 3.1.1), or by person/number-invariant ‘middle’ prefix plus ‘actor’ suffix (‘Inherent middle’; 3.1.2) – and in whether a further, syntactically inert cognate object NP is also employed (3.1.3).

3.1.1 ‘Basic’ intransitive: NPABS U-V This is a closed class, basically restricted to monovalent states (only ‘come’, ‘go’ and ‘walk’ are exceptions). Its members are peculiar in lacking true infinitives. It contains:

21 Unlike the other diathetic prefixes this is phonologically unstable, disappearing after most undergoer prefixes, but it is overt (a) in infinitives, (b) before 2/3nsg U prefixes (c) before RR diathetic prefixes.

Valency in Nen

− − −

1079

‘be’ (1; 6a, 6b), plus ‘come’ and ‘go’ (6c, 6d) which are directional derivatives of ‘be’; ‘come’ is ‘be hither’ and ‘go’ is ‘be thither’. ‘walk’ around 30 ‘positionals’, which are statives of position or posture with meanings like ‘be standing’, ‘be the end of something’, ‘be precariously up high’ and ‘be in a tree fork’ (discussed in § 5.7),

(6) a. Ynd mer w-m. 1abs good 1sgU:α-be:nd ‘I’m good.’ b. Ynd mer yn-ren. 1abs good 1nsgU:α-be:du ‘We two are good.’ c. Ynd wensde-wan nambt q-ng-m. 1abs Wednesday-loc ±a.few.days 1sgU:β-aw-be:du ‘I went last Wednesday.’ d. Ynd wensde-wan nambt tn-ng-ren. 1abs Wednesday-loc ±a.few.days 1nsgU:β-aw-be:du ‘The two of us went last Wednesday.’ Only a few verbs from the Leipzig list are in this class: yls ‘go’ plus were ‘sit = be in a sitting position; live/dwell’. In addition, one of the translations of DIE22 uses ‘be’ from this class, plus kr ‘dead, death’, i.e. as a nominal predicate construction, with appropriate tense/aspectual modulation (another translation uses an experiencer object verb – see § 3.2.2). Likewise BE DRY and WANT are encoded, respectively, by the verb ‘be’ plus the adjectives srnete ‘dry’ and mñte ‘desirous’. KNOW is coded by ‘be’ plus mete ‘knowing, knowledgeable’ or mete spsp ‘be expert, know well’; both these can take objects in a double-absolutive construction (§ 3.1.3). Other members of this class, not on the Leipzig list, are around forty ‘positional verbs’ with meanings like ‘be up high’, ‘be wedged’, ‘be in a tree fork’, ‘be lying’, ‘be in water’. See Appendix 2 for a full list. Posals enter into a three-way alternation with ‘placement’ (transitive) and ‘assume position’ verbs – see § 5.7 for more details. They have a number of other distinct properties, such as the ability to take the stative suffix -ngr – see Evans (2014) for details.

3.1.2 Intrinsic middle: NPABS M-V-A This is the standard pattern for coding monovalent patterns (we shall see in § 5.1 that its inflectional morphology patterns like derived middles, most importantly 22 I adopt the convention of using capitals for the reference verb on the Leipzig list, then giving a more accurate characterization of the verb’s meaning in single quotes after the Nen verb.

1080

Nicholas Evans

reflexive/reciprocals). With just a couple of exceptions it is used for all dynamic, monovalent predicates. Nen thus exhibits a proliferation of middle verbs, which goes well beyond even what is found in languages like Spanish, Italian, Lithuanian or Russian which are well-known for their large number of reflexiva tanta and other middle-type constructions. The verbs listed as inherent middles/reflexiva tanta in Geniušieniė (1987), and Kemmer (1993) are middles in Nen, too, but so are many more. This class takes in almost all dynamic one-place predicates: activity verbs and monovalent change-of-state verbs (except ‘come’, ‘go’, and ‘walk’), as well as change-of-posture/position and motion verbs, phase change (begin, finish) and verbs of controlled perceptual or cognitive activity. A note on glossing: even though I use the gloss A for ‘actor’, there is no need for the action to be volitional (e.g. it includes ‘fall’ and ‘burn’), or for the subject to be animate (e.g. ‘rain’). What is crucial is that the predicate be dynamic. (3a, b and 7) illustrate ‘talk’ for various numbers, thus showing the thematic alternation between dual and non-dual. (7) Sombes är n\owab/t. two man(abs) 3duA.npreh.ipfv\talk ‘Two men are talking.’ The intrinsic middle class includes 16 verbs from Leipzig list:23 APPEAR ipars ‘appear’, BE A HUNTER ärs ‘hunt’, BOIL rngs ‘boil’, BURN äkrärs~ekrers ‘burn’, uzers ‘burn’, CLIMB armbs ‘ascend’, CRY engs ‘cry’, FALL uwis ‘fall’, JUMP ätäns and ábrms ‘jump’, LAUGH urs ‘laugh’, PLAY izers ‘play’, RAIN aparns ‘rain’, RUN alnzs ‘run’, SCREAM awambs ‘scream’, SINK olets ‘sink’, SIT amzs ‘sit down’, THINK embärs ‘think’.

An important class not evident from the Leipzig list are phrasal middles formed by combining a predicate adjective with the middle verb ämts ‘become’, e.g. mete ämts ‘to learn, become knowledgeable’ (mete ‘knowing, knowledgeable), qgibba ämts ‘to become covered with dark clouds (sky)’ (qgib ‘dark cloud’, -ba comitative). At present, the syntactic consequences (if any) of the difference between basic intransitives and inherent middles are unclear, though at this stage none have clearly emerged except for the lack of infinitive forms for (most) prefixing verbs. The lack of infinitive constructions for prefixing verbs makes it difficult to apply most of the syntactically sensitive tests (such as formation of purposive complements, phasal complements etc.) to the prefixing verbs.

23 I count these by the meaning on the Leipzig list, rather than the number of Nen lexemes, so if there are two (near-) synonyms in Nen (e.g. two words for ‘jump’) this counts as one, not two.

Valency in Nen

1081

3.1.3 Middles with cognate object: NPABS + NPABS M-V-A This type has middle morphology, and two NPs in the absolutive. Only the first is indexed on the verb, by the actor suffix. The other NP is more like a cognate object, being syntactically inert. Two verbs from the Leipzig list have this pattern. Both are verbs of sound emission or at least involve the pronunciation of a word: COUGH momae ke otärs ‘cough’ and NAME yétqén ewets ‘name’. Many other verbs of sound production behave similarly, e.g. grr ke otärs ‘snore’. Several other verbs of speech and communication also pattern like this, but with an optional dative adjunct for the recipient of the information; the second, cognate-object like argument is zi ‘word, speech’ or ke ‘cry, noise’. The two verbs taking this coding frame from the Leipzig List are SAY, TALK (TO) zi owabs ‘say, talk (to dat) (about: ABL)’ and SHOUT AT ke oters ‘shout at’. As mentioned in § 3.1.1, the expressions for ‘know’ and ‘be expert (in)’, formed by combining the verb BE with a stative adjective (mete ‘knowledgeable’ or mete spsp ‘expert’), can optionally take a cognate object, again giving a double-absolutive construction. Note that this is the same case frame as is found in the much more productive reflexive/reciprocal construction, in which the first absolutive is indexed on the verb, while the second is optionally filled with a reflexive/reciprocal pronoun (§ 5.1). A further variant is with the verb epss ‘mix with’: though morphologically a middle verb, it still takes an object in the absolutive denoting one element of the mixture, and a comitative adjunct denoting the other: (8) Wagib bi-ba k\eps/ta ! fish(abs) sago-com 2sgA.imp.ipfv-mix ‘Mix the fish with the sago!’

3.2 Basically divalent patterns These take two indexed arguments. In this section we proceed from prototypical transitive frames, with an ergative plus an absolutive argument each indexed on the verb, through various forms of deviation from this prototype in terms of either flagging or indexing.

3.2.1 Basic transitive: NPERG + NPABS U-V-A These are the basic transitive type: they take two arguments (one ergative, one absolutive) and the verb indexes both, through the actor suffix and undergoer prefix respectively.

1082

Nicholas Evans

(9) Yndbem kae ombte nu y\z/wam. 1nsg.erg ±1day hot water(abs) 1duA>3sgU.ppf\cook ‘We two boiled the hot water yesterday (from scratch).’ Transitive coding is used for the usual suspects: impact, caused change-of-position, changed material composition, perception and cognitive apprehension. From the Leipzig list, it contains 29 members:24 BREAK räbärs ‘break’, dars ‘break’, BUILD räms ‘build’, CARRY wazrs2 ‘carry’, COOK zeyas ‘cook’, DIG tars ‘dig’, EAT nne ‘eat’, FOLLOW waets ‘follow’, FRIGHTEN brbr wets ‘frighten’, GRIND ess ‘grind’, HEAR weres ‘hear’, HIDE wérnis ‘hide’, HUG wambns ‘hug’, LIKE mñteba wakaes ‘like’ (lit. ‘look upon with sweetness’), MAKE räms ‘make’, MEET winḡs ‘see (catch sight of); meet’, äbäts 25 ‘meet’, PEEL wétrers ‘scratch, peel’, PUT kms ‘put in lying position’, ROLL paes ‘roll’, SEARCH FOR ronzas ‘search for’, SEE winḡs ‘see (catch sight of); meet’, wakaes ‘look at, see, gaze at’, SHAVE btkamgeb wérnas ‘shave’, SHOUT AT weräns ‘shout at’, SING angan räms ‘sing’, SMELL wambaes ‘smell; sniff’, STEAL wr ‘steal’, TEAR drers ‘tear’, THINK ABOUT wabaes ‘think (about)’, TIE sns and wmbaes ‘tie’, WASH wémbärs ‘wash’

Placement forms of positionals are standard transitives (§ 5.6). ‘Think about’ is transitive but ‘think’ is an inherent middle. As illustrated by the example brbr wets [frightened make] for FRIGHTEN, the transitive forms of predicates expressed by BE + stative adjective are composed by combining the same stative adjective with the transitive verb wets ‘make, cause to be’26. A number of transitive verbs regularly occur with semantically-motivated adjuncts – e.g. wwis ‘throw’ can appear as a basic transitive ‘throw a stone’ or can add path and destination by case-marked adjuncts; likewise wmbaes ‘tie (up)’ can optionally add instrument and location NPs. These are dealt with in § 3.2.5. A couple of other verbs – of communication – exhibit this coding frame but augment it with some form of clause setting out the material communicated – wémés takes an infinitive complement when it means ‘ask to’, but an indicative complement when it means ‘ask (where etc.)’. Mts ‘tell to’ likewise takes an infinitive complement.

24 One of these has a cognate object (angan ‘song, dance’), and another takes a secondary predicate on the object, i.e. mñteba ‘[see] with sweetness [i.e. like, love]’ and brbr ‘[be] frightened [of]’. 25 This verb is deponent in its infinitive but not its finite forms. It possesses only a reflexive/ reciprocal infinitive form äbäts (speakers denied the existence of expected transitive infinitive forms wbäts or bäts) but the transitive stem can be used in normal transitive constructions such as ynd bende ane te ybätwen ‘I met your younger sibling’ – note the y- undergoer prefix directly precedes the expected transitive root bät. 26 This verb, in turn, is interestingly relatable as the transitive of the positionals root et- ‘be at an end, be the end of something’, i.e. to ‘make something’ is to ‘bring it to an end’ (of the making process).

Valency in Nen

1083

3.2.2 Experiencer object constructions: NPABS > NPERG U-V-A3sg These are a subtype of divalent, transitive construction which are effectively transitive subject idioms: the ergative/actor NP is completely or partly fixed,27 denoting a stimulus, drive or illness, and the absolutive/undergoer NP denotes the experiencer. Unlike normal transitive constructions, the order is OAV rather than AOV.28 For their citation form in the data-base I use the ergative form of the A noun, plus the relevant infinitive, e.g. gersäm räms [hunger(ERG) do] for BE HUNGRY.29 räms ‘do’ is the commonest such verb (and the only one showing up in the Leipzig set) but other verbs such as krärs ‘burn’ and wñäms ‘seize, overpower’ are also found. (10) a. Zän ak-äm y\ram/te. dog(abs) thirst-erg 3sgA>3sgU.npreh.ipfv\make ‘The dog is thirsty.’ b. Ynd kae gers-äm q\ram/te, totr nne 1abs ±1day hunger-erg 3sgA>3sgU.preh.ipfv\make today food(abs) tba y\ne/tan. just.now 1sgA>3sgU.npreh.ipfv\eat ‘I was hungry yesterday, but today I’ve eaten.’ The semantics of experiencer-object verbs corresponds rather closely to ‘dative subject’ constructions in languages that have them. The experiencer-object coding solution to expressions of bodily sensation and sickness is a macro-areal phenomenon widespread in ‘Sahul’ – Australia (e.g. Murrinhpatha (Walsh 1987), Iwaidja and Mawng (Evans 2004)), and in New Guinea (e.g. Kalam – Pawley et al. 2000) – though it also present in many African languages in the form of constructions of the type ‘hunger kills me’ for ‘I am hungry’, as in Yoruba (Atoyebi, this volume). Seven verbs from the Leipzig list take this coding frame: BE HUNGRY gersäm räms ‘be hungry’ (lit. ‘hunger do OBJ’), BE ILL enznewäm räms ‘be ill’ (lit. ‘illness do OBJ’), BE SAD weiweiyäm räms ‘be sad’ (‘sadness/pity do’), DIE zängs ‘die’ (lit. ‘(it)

27 Though it may be modified in various ways, e.g. ‘hunger’ can be modified by the name of the object hungered for, along the lines of ‘cassowary hunger does me’ for ‘I am hungry for cassowary’, or krbr ‘cold’ can be modified by kitong ‘large’, as in mleg toge krbr kitongäm yramte ‘the girl got really cold, really froze’. 28 In the data so far, basic transitive clauses sometimes occur with personal pronouns placed before ergative non-pronominal arguments, even when in undergoer role. The order in experiencer object constructions could be seen to have grammaticalized this tendency, though more naturalistic data is needed before a meaningful statistical comparison can be made. 29 Though some speakers have suggested they would prefer the relevant dictionary entry to be made without the ergative, e.g. kérbér räms rather than kérbéräm räms. This is interesting in suggesting that the ergative element is a syntactically live element, rather than just an inert coverblike nominal.

1084

Nicholas Evans

kills me’)30, FEAR brbräm räms ‘fear’ (‘fear do’), FEEL COLD kérbéräm räms ‘feel cold’ (‘cold do’), and FEEL PAIN tkretewäm räms ‘feel pain’ (‘pain do’).

A variant on this construction is found with the verb waorers ‘forget’: here the forgetter is encoded as an experiencer object (absolutive NP, U-agreement) and the thing forgotten evokes actor agreement but instead of taking the ergative it takes the absolutive. So far this is the only verb attested with this case frame. (11) Bä ya-nde ws y\aorer/nda. 3abs 3sg-poss bottle 3sgA>3sgU.ppf\forget ‘He forgot his water bottle.’

3.2.3 Transitive verbs with deponent middle verbal morphology: NPERG + NPABS M-V-A One verb from the Leipzig list has a transitive ERG:ABS case frame but middle morphology on the verb, agreeing just with the actor argument: LEAVE ors ‘leave’. (The other verb for ‘leave’, branḡs, is a regular transitive verb.) (12) Ymam n\or/te ya-nde dmab. 3sg.erg 3sgA.npreh.ipfv\leave 3sg-poss wife ‘He left his wife.’

3.2.4 Semi-transitive verb registering oblique on undergoer slot: NPERG + NPOBL U-V-A These verbs have an ERG: OBL/DAT1 case frame (13), and register the oblique/ dative argument on the undergoer slot (and of course the ergative on the actor slot), without there being any evidence of an object argument (i.e. an absolutive NP). There may, however, be some form of clausal complement, e.g. ‘teach to [S]’ (14). (13) Buder, ta q\ete/na! friend 1sg.obl 2sgA>1sgU.imp.ipfv\help ‘Friend, help me!’

30 This verb is slightly different from the others in that there is no ergative NP. Kr ‘death’ can optionally appear, but in the absolutive (e.g. Bä kr yzanḡte ‘(s)he is dying’). If a cause of death is mentioned, this takes the source/consequence-marking case -mne instead of the ergative, e.g. Ynd enznemne / abemne wzanḡte ‘I am dying of disease / sorcery’). Note that, in discussing dying, the verb zängs is used for the process of dying, while kr + BE is used for the result ‘be dead’.

Valency in Nen

1085

(14) Ymam ta-gta q\awapap/te qämblä zéḡ weng-s. 3sg.erg 1sg-dat1 3sgA>1sgU.preh.ipfv\teach cats.cradle string roll-nmlz ‘He is teaching me to roll cats-cradle string.’ Three verbs on the Leipzig List take this coding frame: HELP wetrs ‘help’31, TEACH wawapaps ‘teach’ and LOAD ewes (one of the ‘load’ verbs). An additional verb with this frame is wémnzs ‘wait in prey for’. A slight variant, with an ABS:DAT case frame, is wawapañs ‘[ABS] fall from the grasp of [DAT]’, which is the usual translation of English ‘drop’ (15). (15) Tbe-gta yép tn-awapañ-t-e. 1nsg-dat bag(abs) 3sgA>1plu.preh.ipfv\fall.from.grasp.of ‘We dropped our bag.’ (Lit. ‘Our bag fell from us.’) Two of these verbs – wawapaps and wawapañs – appear to contain fossilized forms of the benefactive applicative (§ 5.6), which would square with their deployment of a dative undergoer argument. For wawapañs there is no clear corresponding verb without the benefactive (though there is a verb apañs ‘bend over (as one when offering oneself for anal sex)’), and for wawapaps the base meaning is the auxiliary ‘begin to, try to’, but in neither case is the semantic derivation completely transparent.

3.2.5 Transitive constructions with various kinds of adjuncts There are quite a few rather small lexical sets that are basically divalent transitives augmented by specific types of adjunct. These have no discernible argument-like properties and are generally not promoted to argument status by any valency alternation – the sole exception being oblique/dative adjuncts which can be promoted to indirect object status by the benefactive applicative. However, in some cases the choice of case is not semantically predictable – e.g. sources take the ablative with some verbs and the locative with others – so it can at least be said of some of them that there is some subcategorization by the verb for choice of case. Various transitive plus adjunct frames are grouped together in this section, though listed as different coding types in the valency database and not shown in the summary table. Transitive verb with optional ablative instrument. Six verbs from the Leipzig list are basic transitive verbs, but can optionally take an NP in the ablative specifying

31 Cf. similar case frames in German helfen and Russian pomogat’.

1086

Nicholas Evans

the instrument. COVER wapars ‘cover’, DRESS wamanes ‘dress’, FILL waps ‘fill’, HIT gms ‘hit, beat, kill’, KILL gms ‘hit, beat, kill’, WIPE läns ‘wipe’. Transitive verb with ablative adjunct for source. Two verbs: TAKE kaps ‘take (from:ABL)’ and POUR wazrs1 ‘pour (from:ABL)’. Transitive verb with locative adjunct for source. RECEIVE=GET: √ane ‘get, receive (from: LOC)’ Transitive verb with locative adjunct for goal. LOAD: wapls ‘load (onto/into: LOC)’ Transitive verb with optional ablative instrument and locative location. TOUCH wenzas ‘touch (on L) (with ABL)’, TIE sns ‘tie, fasten (onto L) (with ABL)’32 Transitive verb with allative destination and perlative route adjuncts. Just one verb from the Leipzig List – THROW wwis ‘throw (to DAT through PERL). Transitive verb with allative/dative1,2 destination/recipient adjuncts. Several verbs from the list have optional adjuncts, in the allative or dative. The former encodes places, the latter human recipients, with an additional choice between ‘dative1’ and ‘dative2’ depending on whether the object reaches or is only intended for its recipient. Otherwise, these behave like typical transitive verbs, registering the theme – the thing carried or shown – as the absolute NP / undergoer prefix. The directional prefixes can be used to evoke the recipient through direction – ‘towards’ for a first person, ‘away’ for a third – without overtly employing argument indexing. (16) illustrates two such verbs, both of sending: wäbärs ‘send (as first act in an exchange relationship)’ and trärs ‘send (as continued participation in an exchange relationship)’, each used here with no adjunct present for the recipient, while (17) elaborates this structure by adding a Dative.1 adjunct for the recipient. (16) Ymam sombes bñe tä\n-wäbär/ng kae, 3sg.erg two letter(abs) 3sgA>3duU.preh.ipfv-tow\send.first ±1day ynd kae wnde sombes bñe tä\ng-trär/n. 1abs ±1day in.return two letter(abs) 1sgA>3duU.preh-aw\send ‘He sent two letters (+>me) hither yesterday, and I sent two letters thither (+>him) in return.’ (17) Ynd sombes bñe ya-gta tä-ng\wäbär/n. 1sg.erg two letter(abs) 3sg-dat1A 1sgA>3duU.preh.ipfv-aw\send.first ‘I sent him/her two letters thither.’

32 Another ‘tie’ verb, wémaes, can take two case frames (with no formal marking of the alternation on the verb): ‘tie (string etc.: ABS) onto (entity: LOC)’, and ‘tie up (entity: ABS) with (string etc.: ABL)’. Examples are Ynd kend kup aba yépnan zéḡngama ‘I tied up the paperbark bundle with rope’ and Ynd zéḡ yéparnda qtowan ‘I tied it onto the axe handle’.

Valency in Nen

1087

Verbs of this type often correspond to three-place predicates cross-linguistically, but for this construction there is no evidence that the NP expressing the recipient role, such as yagta in (17), has argument status in Nen. It is simply an adjunct – and note that the undergoer slot agrees with ‘two letters’ here, which also triggers the dual form of the thematic. This observation extends to räms, whose basic meaning is ‘do’ but which can be used with the meaning ‘give’ in some contexts . In such cases the recipient is indicated by an oblique/dative NP, or implicated by a directional prefix (cf. English ‘give it here’) but in neither case indexed on the undergoer prefix, which is used to represent the theme (18a, b ). (18) a. T-n\rama! 2sgA>3sgU.imp.pfv-tow-do/give ‘Give it here!’ b. Tä-ng\rama/e! 2sgA>3duU.imp.pfv-tow\give ‘Give those two things (away)!’ Seven verbs from the Leipzig list are in this class: CARRY/TAKE/BRING renzas ‘carry, take, bring (typically non-human object)’, menzas ‘carry (typically human object)’, GIVE räms ‘give [object] (hither/thither)’ LOAD ags ‘load up’, POUR wazrs ‘pour (for: DAT)’ and SEND watembs ‘send’. For many members of this class there is a verb-coded valency alternation with a true trivalent pattern (§ 3.3), employing the benefactive prefix (§ 5.6). This yields a coding frame in which the recipient is an indirect object argument, retaining the dative/oblique case but coding the recipient in the undergoer prefix. A slightly deviant member of this class is SHOW, which can be expressed by oyabs ‘exhibit, hold up for viewing’. Like other members of the class this has ERG and ABS arguments (plus an oblique adjunct) but unlike them the undergoer slot of the verb is filled with the middle prefix; it has a benefactive counterpart wayabs ‘show’. (19) Gbay-em sombes qämblä zéḡ n\n-oyab/te Gbae-erg two catscradle string(abs) 3sgA.npreh.ipfv\tow-exhibit ta-gta. 1sg-dat ‘Gbae is showing two catscradle strings to me.’

1088

Nicholas Evans

3.3 Trivalent pattern Four verbs from the Leipzig List exhibit this pattern: GIVE räms ‘give’,33 TELL wawaets and wétls, both ‘tell’, POUR wazrs ‘pour’ and SHOW wayabs ‘show’. It is defined by taking a three-place ERG ABS DAT case array, and exhibits ‘secundative indexing’, registering the DAT rather than the ABS on the undergoer prefix slot of the verb, and in the dual/non-dual morphology on the verb (-nda in (20) and (21) is the past perfective third person non-dual form; to express ‘he showed them two’ the verb form would be changed to yawayabae, employing the suffix -ae ‘2|3sgA>duU. ppf). (20) Ymam wagib ta w\a-rama/nda. 3sg.erg fish(abs) 1sg.obl 3sgA>1sgU.ppf-ben\give ‘She gave me a fish.’ (21) Ymam ybe-gta ag ya\wa-yab/nda. 3sg.erg 3nsg-dat1 coconut(abs) 3sgA>3plu.ppf-ben-show ‘He showed them the coconut.’ Two further morphosyntactic properties distinguish trivalent verbs. Firstly, as indicated in § 2.3(d), an unusual typological feature of Nen is that phasal auxiliaries must take a benefactive applicative prefix when their complement is a trivalent verb, and register the recipient argument on the undergoer prefix of the phasal. See (22), and (5) above for an additional example. (22) Ymam ta-gta ag räm-s-t 3sg.erg 1sg-dat coconut(abs) give-nmlz-all w\n-a-wa-pap/nda. 3sgA>1sgU.ppf-tow-ben-tr-begin ‘He was about to give me the coconut.’ Secondly, just with trivalent verbs there are two special ways of coercing other morphology into expressing the difference between singular and multiple objects: the aspectual contrast, and a coerced use of the ‘away’ prefix. Cf. the perfective form arama in daramanga ‘he should give him one thing’ and the imperfective form aramta in daramtanga ‘he should give him many things’.34 The ‘away’ prefix 33 At least, in its benefactive stem form aram(a); the benefactive prefix shows up in finite forms and on phasal auxiliaries, but not in the infinitive which (irregularly) is räms for both the transitive and the ditransitive frames. 34 Analyzing these words further: d\a-rama/nga [3sgA>3sgU.f.pfv\ben-give] vs. d\a-ram/ta-nga [3sgA>3sgU.f.pfv\ben-give-ipfv]. This is an example of aspectual conversion, in which a verb stem first takes the basic imperfective thematic (-ta), but is then inflected for a perfective suffix. Aspectual conversion from imperfective can have a number of functions, including some shift to the lexical

Valency in Nen

1089

can also be coerced into expressing multiplicity of objects, as in q\a-yab/ta! ‘show me one!’ [2sgA>1sgU.imp.pfv-ben-show] vs. q\ng-a-yab/ta! ‘show me many!’ [2sgA>1sgU.imp.ipfv-aw-ben-show]. The use of the ‘away’ prefix to encode large plurals is not unique to trivalent verbs, but it is only with them that it counts entities that are not one of the indexed arguments.

4 Uncoded alternations (case alternations) 4.1 Allative / goal alternations These alternate between marking the destination with the allative and with the dative. The number and identity of arguments indexed stays unchanged, and there is no morphological marking on the verb in the form of diathetic prefix alternations. Semantically the allative stresses a spatial destination while the dative encodes a human goal or recipient. This alternation is available with the theme-indexing verb CARRY, TAKE renzas ‘carry’. Illustrative examples are: (23) Ymam kiémb-ngama w\ng-renz/e Mowet-t. 3sg.erg truck-abl 3sgA>1sgU.npreh.ipfv-aw-carry Morehead-all ‘He took me to Morehead in a truck.’ (24) Ynd wawapapser-epap buk aba 1sg.erg teacher-dat2 book(abs) imm y\n-renz/an. 3sgA>1sgU.npreh.ipfv\tow-carry ‘I brought/carried the book to the teacher.’ This case alternation is also possible with the recipient argument of GIVE räms ‘give’, with the allative implying direct giving of the theme to the recipient, while the dative is compatible with the theme just being left for the recipient to collect later.

semantics (e.g. ‘talk’ to ‘speak, say’). In this example, the inner imperfective marker is signalling the multiplicity of the theme argument, while the outer perfective inflection gives a ‘mediated imperative’ meaning (i.e. letting the addressee know they should convey the speaker’s command to a third person).

1090

Nicholas Evans

4.2 Dative addition Here a dative NP is added, as a case-marked NP, but there is no registration of this on the verb through diathetic prefix, middle morphology or change to indexation. This is marginal as an alternation type, and is best treated as the predictable addition of a dative adjunct, of the type discussed in § 3.2.5. The case may be DAT1 or DAT2 − it is not yet clear what the difference is, but the difference is subtle and I provisionally lump these two alternations together. Verbs showing this alternation, from the Leipzig list, are: SING angan räms ‘make a song [to DAT]’, APPEAR ipars ‘appear [to DAT]’, SEND watembs ‘send [to DAT]’.

4.3 Middle object addition Here an absolutive NP is added, as a goal, patient or theme, to a monovalent middle verb. This results in a double-absolutive case frame (cf. § 3.1.3). There is no additional marking on the verb. (Though in the case of awags ‘load oneself up’, this is already derived from wags ‘load up’ by the RR prefix). Examples: (25) a. Ynd n\a-wag/tan. 1sg.abs 1sgA.npreh.ipfv\rr-load ‘I load myself up.’ b. Bä nne n\a-wag/tan. 3sg.abs yam(abs) 1sgA.npreh.ipfv\rr-load ‘She loads herself up with yams.’ Two verbs from the Leipzig list that exhibit this alternation are: LOAD awags ‘load oneself up’ > ‘ABS load oneself up with ABS’, and BE A HUNTER ärs ‘hunt, be a hunter’ > ‘ABS hunt ABS’. Another verb of this type, not on the Leipzig list, is äbäms ‘suckle, suck (at the breast)’, which can optionally add nono ‘breast’ as an absolutive NP to give nono äbäms ‘suckle the breast’.

4.4 Other possible uncoded alternations A couple of other possible types are given in the data-base: a) a pseudo-passive, in which the external NP corresponding to the A is simply omitted, and b) a possible source-omission with ‘wipe’ (i.e. wipe the mud vs. wipe the mud off the mat) which may simply involve optional adjunct omission/inclusion. These could readily be handled as instances of optional inclusion of adjuncts.

Valency in Nen

1091

5 Verb-coded alternations Verb-coded alternations are a productive process in Nen lexicogrammar, making heavy use of the diathetic prefix series. It is not yet possible to give a definitive statement of how many diathetic prefixes there are, owing to the interaction of a certain degree of so far unexplainable formal variation with differences in the precise meaning of the alternation; below I give the best current schema.35 In addition to the verb stem alternations created by the diathetic prefix itself, the alternations it produces lead to differences in whether the prefix is filled by a middle marker or an undergoer prefix. Some, like the transitivizing and benefactive prefixes, move the stem from middle to transitive; others like the reflexive/reciprocal, move it from transitive to middle. Prefixes increasing valency (transitive, benefactive) are w-initial, while those decreasing valency (reflexive/reciprocal, decausative, autobenefactive) are vowel-initial. In the case of the autobenefactive (§ 5.3), the case frame remains unaltered (ERG: ABS) but the verb indexing goes from transitive to middle. Tr

wakaes ‘see’

warmbs ‘take up’

Diathetic prefix

RR

transitive

Middle

awakaes ‘see self/each other’

armbs ‘ascend’

Normally there are distinct infinitives for verbs with diathetic prefixes. Cf. wapars ‘to cover’, awapars ‘to cover self/each other’; nps ‘cut (tr.)’, awañps ‘cut each other’, eñps ‘get cut’, uñps ‘cut for oneself’. Though it is commonest for verbs to take just one diathetic prefix at a time (if any), there are some cases of verbs that take two. Most commonly this is where a middle verb has first been transitivized by w- via the transitive, then prefixed by the benefactive – see (33) below for an example. There are a few other sets which appear to exhibit multiple prefixation (cf. eres ‘listen’, weres ‘hear’, oweres ‘listen

35 One obvious alternative to the current schema would be to consider the causative and benefactive prefixes as instantiations of a single prefix which we could call ‘valency-increasing’ – which could have either transitive or benefactive readings according to what it attaches to. In addition to the generalization that both increase valence by one, they exhibit formal similarity, namely (w)a-. At present I do not adopt this solution, because of the fact that it would leave unexplained certain apparent gaps in the data – e.g. the fact that there are not examples of prefixing (w)a- to a monovalent stem with a benefactive reading (e.g. ‘work for the benefit of’). On top of this, with some verbs the benefactive is not overtly present in the infinitive form (see § 5.6), whereas the transitive is always overt in infinitives. For these reasons I prefer to analyze the two as separate morphemes, though their formal identity and functional similarity suggest a historical connection between them.

1092

Nicholas Evans

out for, await news of’, exhibiting the derivational pathway middle > transitive > autobenefactive) but the semantic relations in such cases are idiosyncratic.

5.1 Reflexive/reciprocal The ‘reflexive/reciprocal’ is formed from transitive verbs by: a) substituting the person/number-coding U-prefix with a person/number-invariant middle prefix b) dropping the overt NP coding the P role c) flagging the NP coding the A role with the absolutive instead of the ergative case, i.e. treating it as an S d) prefixing the diathetic prefix (a-, e-, ä, i-, o- etc.36) to the verb stem. e) (optionally) deploying a reflexive/reciprocal pronoun tanzo(s), in the position usually occupied by NPs in P role The semantics permits either reflexive or reciprocal readings. Reflexive is the only possible reading in singulars. With non-singulars reciprocal readings are the default reading but reflexive readings are also possible. (26) a. Bmbem togetoge ta\waka/tang! 2nsg.erg children(abs) 2pla>3plu.imp.ipfv\look.at ‘You (pl.) look at the children!’ b. Bm (benzo) k\a-waka/ta! 2abs 2sg.rr 2sgA.imp.ipfv\rr-look.at ‘You look at yourself!’ c. Bm (bbenzos) k\a-waka/tang! 2abs 2nsg.rr 2pla.imp.ipfv\rr-look.at ‘You (pl.) look at yourselves / each other!’ In the case of at least some ditransitive verbs, RR-formation produces a middle, RR-prefixed verb whose semantics involves coreference between agent and recipient and which takes two absolutive arguments: the agent and the theme (cf. the underived equivalent in § 3.1.3). (27) Yna sombes är nne n\o-trär/t. dem two person(abs) food(abs) 3duA.npreh.ipfv\rr-send ‘These two are exchanging (sending each other) food.’

36 Conditioned by vowel harmony with the following vowel in cases where just a single C intervenes, but o- before clusters (e.g. trars/otrars) and awa- before root-initial ñ (surface n) as in ñps ‘cut’, awañps ‘cut each other’ – see further discussion in fn. 39 below.

Valency in Nen

1093

Verbs from the Leipzig test list showing this alternation are listed below. BREAK dars ‘break, crack (v.t.)’ / adars ‘break, crack (of its own accord)’, COVER wapars ‘to cover’ / awapars ‘to cover self/each other’, CUT nps ‘cut’ / awañps ‘cut self/each other’, FOLLOW waets ‘follow’ / awaets ‘follow each other’, HIDE wernis ‘hide (v.t.)’, ewernes ‘hide oneself’, HUG wambns ‘embrace’ / awambns ‘embrace each other’, LEAVE branḡs ‘leave’ / abranḡs ‘leave each other, divorce’, LOAD wags ‘load’, awags ‘load oneself up’, MEET / äbäts ‘meet; meet each other’37, winḡs ‘see, meet’ / iwinḡs ‘see, meet each other’, SEE wakaes ‘see’, awakaes ‘see, look at self / each other’, SEND trärs ‘send (as an initial or once-off move)’ / oträrs ‘send (as part of an exchange)’, SHAVE btkamgeb renas ‘shave (someone else)’, btkamgeb ärens ‘shave (oneself)’, SMELL wambaes ‘smell, sniff’ / awambaes ‘smell, sniff each other’, THINK ABOUT wabaes ‘remember, think about’, awabaes ‘remember to do something, recall, think about’38, TOUCH wenzas ‘touch’ / ewenzs ‘touch each other’.

Where a two-place predicate is coded by a construction with a copula plus a predicative adjective, reciprocals can be derived, without predicate coding, simply by using the RR pronoun, along with the verb BE in the appropriate person and number: (28) Yna är sombes bä ybenzos mete yä-ren. dem man two(abs) 3abs 3nsg.rr know 3nsgU.npreh.ipfv-be:du ‘Those two men know each other.’

5.2 Decausative This denotes activities that occur without obvious agentive input from an outside party. The verb root takes e, ä, a, or o as diathetic prefix, the argument frame is reduced from A:P to S, and middle verb morphology is employed. There is some overlap in form with the RR, but unlike that alternation the RR pronoun benzo(s) cannot be used. Moreover there are a few verbs that possess contrasting diathetic prefixes distinguishing the two alternations: cf. Reflexive/reciprocal awañps (~änps) ‘cut each other’, Decausative eñps ‘get cut’; and Autobenefactive uñps ‘cut (for oneself)’, all three derived from nps ‘cut’;39 also oträrs ‘exchange’, äträrs ‘cross paths with’, both derived from trärs ‘go along a path’. Nonetheless, in cases where

37 As mentioned in an earlier footnote, this verb has a deponent infinitive, employing what looks formally like a reflexive/reciprocal infinitive for both reflexive/reciprocal and transitive readings, but the finite forms behave in the expected way, distinguishing the two through their affixation possibilities and the form of the stem. 38 As these glosses indicate, the reflexive/reciprocal form has undergone some semantic specialization away from a pure reflexive/reciprocal meaning. 39 ñ is not permissible word- or stem-initially in Nen, but the underlying initial ñ surfaces once diathetic prefixes are added. Forms are cited with whichever initial nasal is found in the relevant simple or derived stem.

1094

Nicholas Evans

there is no contrast set, it can be fairly arbitrary whether a given form is considered RR or DC – it may simply be a RR form which has not yet been attested with a RR pronoun. An example of this alternation is: (29) a. Ymam yéb_tr y\drer/nda. 3sg.erg bag_handle 3sgA>3sgU.ppf-tear ‘He tore the handle from the bag.’ b. Yna yéb n\e-drer/nda. dem bag 3sgA.ppf\dc-tear ‘This bag tore.’ Verbs from Leipzig list taking this alternation: BREAK räbärs ‘break (v.t.)’ / äräbärs ‘break (v.i.), get broken’, BURN zers ‘set fire to’, uzers ‘catch fire’, CUT nps ‘cut’, eñps ‘get cut’, TEAR drers ‘tear (v.t.)’ / edrers ‘tear, get torn’ (v.m.); TIE sns ‘attach, tie, hook (v.t.)’, esns ‘get hooked, tied’.

5.3 Autobenefactive This derives verbs meaning ‘for one’s own benefit, on one’s own account’. In many contexts this conveys a negative evaluation of selfishness or refusal to share (e.g. (30)40) but this need not be the case, e.g. (31b) encoding independent action by a child, and this prefix is often used frequently in hospitable invitations such as koneta! (k\o-ne/ta) [2sgA.imp.pfv\ab-consume] ‘eat/drink this yourself − tuck in, no need to share!’. The derived verb takes u-/o- as its diathetic prefix and the middle prefix. In contrast to the RR and DA prefixes this usually directly precedes (or even goes inside) the minimal stem, rather than in front of the longer w-expanded transitive stem. As with the preceding two alternations, the case frame is reduced from ERG:ABS to ABS:ABS. (30) Yao y-m, ya Zerus-m wagib-ta yergb yna-o neg 3sgU.npreh.ipfv-be:nd conj Zerus-erg fish-all river(abs) dem-rel k\o-tr/ne. 3sgA.npreh.ipfv\ab-follow.riverbank ‘I can’t believe it, that Zerus followed along the riverbank for fish (implied: with lots of success because he was alone) all for himself.’

40 kotrne here could be replaced by its straightforward transitive equivalent ttrne; the difference is that the use of kotrne is ‘blaming him’ for his selfishness.

Valency in Nen

1095

(31) a. Mleg-äm toge selemi-ngama t\amane/te. girl-erg child(abs) clothes-abl 3sgA>3sgU.preh.ipfv\dress ‘The girl dressed the child.’ b. Yna mleg k\o-mane/te selemi-ngama. dem girl(abs) 3sgA.preh.ipfv\ab-dress clothes-abl ‘The girl got dressed.’ Verbs from the Leipzig list exhibiting this alternation are the following: BUILD räms ‘build’, oräms ‘build for oneself’, COOK zeyas ‘cook’, ozs ‘cook for one’s own benefit’, CUT nps ‘cut’, uñps ‘cut a piece for oneself; desist from, give up’41, DRESS wamanes ‘dress (v.t.)’, omanes ‘get dressed’, HEAR weres ‘listen to’, oweres ‘listen carefully, concentrate on, tune in, attend to’, SHOW wayabs ‘show’, oyabs ‘show off (e.g. something that one has caught)’, TEACH awapaps ‘teach’, aowapaps ‘learn’, WASH wémbärs ‘bathe, wash (v.t.)’, umbärs ‘take a bath, bathe oneself’.

5.4 Simple Middle Alternation This is a simple alternation between middle and transitive ambifixing verbs. This alternation is coded by a) substituting the middle prefix for the pronominal undergoer prefix b) adding the prefix w- in a very unusual way, not coding transitivity/agentivity in the way normal for this form c) dropping the erstwhile A argument, and treating the other core argument as an S(A), i.e. represented by a core NP in the absolutive and indexed by the Actor prefix. Just two verbs from the Leipzig list display this alternation: FILL waps ‘be full’ (< aps ‘fill (v.t.)’), POUR wazrs1 ‘pour into (v.t.)’ azrs ‘pour out’

5.5 Causative This derives transitives from middles denoting motion events, adding the meaning ‘cause (motion/trajectory) through sustained contact (carrying, leading etc.)’. It is coded as follows:

41 For some reason, the diathetic u disappears in the future: Ta parwae pip bende dñpangn. (*duñpangn) ‘I’ll cut your meat for myself.’, but it is kept in the imperative (Ê, tande parwae pip be kuñpa! ‘Yes, cut yourself a piece of my meat.’

1096

Nicholas Evans

a) prefixing (w)(a)- to the basic stem, though there are many phonological contexts in which neither vowel is overtly present 42 b) using transitive indexing morphology instead of the basic middle morphology c) adding an ERG argument denoting the causer; the causee is coded by ABS A sample pair is armbs ‘ascend’ / warmbs ‘take/bring up, cause to ascend’: (32) a. Är k\armb/te läda-wama. man(abs) 3sgA.preh.ipfv\ascend ladder-abs ‘The man climbed the ladder (ascended by means of the ladder).’ b. Är-äm toge t\armb/te läda-wama man-erg child(abs) 3sgA>3sgU.preh.ipfv\cause.to.ascend ladder-com kitara tq-t. platform high-all ‘The man took the child up the ladder onto the balcony.’ The verbs in the Leipzig set which show this alternation are shown in Table 2:

Tab. 2: Verbs from the Leipzig list exhibiting the causative alternation. Middle

Transitive

CLIMB

armbs

‘ascend (steeply)’

warmbs

‘take/bring up’

SINK

olets

‘sink (itr.)’

walets

‘sink (tr.)’

SIT

amzs

‘sit (intr.)’

wamzs

‘seat, sit, set (tr)’

Further verb pairs exhibiting similar alternations between middle and transitive forms, all involving caused movement, are: esrs ‘descend’, wesrs ‘cause to descend’ anḡs ‘return’, wanḡs ‘cause to return’ itrs ‘move’, witrs ‘cause to move’ urtäms ‘ascend (e.g. walking up a slope)’, wurtäms ‘cause to ascend (e.g. taking a child up a slope)’ azrs ‘pour (v.i.)’, wazrs ‘pour (v.t.)’

42 Though clearly present in the infinitive, the w- drops in some contexts, notably after any singular undergoer prefix, as in (32b). With a nonsingular undergoer it would appear, i.e. the form would be ta\warmb/te ‘he took them up’. The a-vowel is only visible when the causative is prefixed to vowel-initial words (like olets ‘sink (itr.)’), where it displaces the other vowel (i.e. wa- + olets > walets.

Valency in Nen

1097

5.6 Benefactive The addition of a beneficiary argument is signalled by the benefactive applicative (w)a-. The beneficiary is then indexed on the undergoer slot – either of the main verb, if finite (33), or the auxiliary, in a phasal auxiliary + infinitive construction (5, 22 above). (33) Ymam tbe-gta sombes ag yn\a-rama/e. 3sg.erg 1nsg-dat1 two coconut(abs) 3sgA>1duU.ppf\ben-give ‘He gave the two of us 2 coconuts’. (34a, b, c) illustrate the use of the benefactive with the verb pair bens ‘feed (X) [with Y-ABL]’ wabens ‘feed up/ fatten up (X) for Y[DAT1]’. With the basic verb, the feedee is indexed by the undergoer prefix on the lexical verb (34a) or the phasal auxiliary (34b), whereas in the benefactive construction it is the beneficiary that is indexed by the undergoer prefix, illustrated here (34c) in a phasal auxiliary construction. (34) a. Ẽ, bä n\ben/tan. yes fut 1sgA>2sgU.npreh.ipfv\feed ‘Yes, I’m going to feed you.’ b. Ynd nne-ngama ben-s-t bä y\a-pap/ndn. 1sgA yam-abl feed-nmlz-all fut 1sgA>3sgU.ppf\caus-begin ‘I’m trying to feed him with yams.’ c. Ynd kiemb tande sakrsakr-äbeta bä 1sgA pig(abs) 1sg.poss brothers-PL.dat1 fut ya\wa-ben/tan. 1sgA>3nsgU.npreh.ipf\ben-feed ‘I’m taming (feeding up) a pig for my brothers.’ In the bens/wabens pair just illustrated the benefactive applicative is overtly present in the infinitive, but this is not the case for all verbs: räms ‘give’ and wazrs ‘pour’ lack an overt distinction in the infinitive, for example. Crucially, for all verbs the benefactive is visible on either the infinitive, the prefixed finite form, or on accompanying phasal auxiliaries. (35a–d) illustrate a three-way contrast with the verb ‘pour’ between the middle verb azrs ‘pour (v.i.)’, the derived causative verb wazrs ‘pour (v.t.)’, and the benefactive verb wazrs ‘pour for (v.t.)’ respectively. Note that the infinitive for the transitive and benefactive verbs is identical, so that the regular transitive infinitive wazrs is used in (35d) but its benefactive status gets marked on the phasal auxiliary nawapapndn. The finite transitive verbs in (35b, c) clearly contrast the transitive and

1098

Nicholas Evans

ditransitive constructions, both in terms of which argument is indexed on the undergoer slot (theme in (35b), recipient in (35c)), and in the presence of the benefactive prefix in (35c) versus its absence in (35b). A comparable triplet is found with the set anḡs ‘return (itr.)’ / wanḡs ‘return (tr.)’ / wawanḡs ‘return (tr.) to/for’. (35) a. Yna nu n\ap/nda biliken-an nu azr-s dem water(abs) 3sgA.ppf\overflow pot-loc water(abs) pour-nmlz n\opap/nda. 3sgA.ppf\begin ‘The water is overflowing in the pot and is beginning to pour out.’ b. Ynd nu y\az/nan be-gta. 1sg.erg water(abs) 1sgA>3sgU.npreh.ipf\pour 2sg-dat1 ‘I’m pouring the water for you.’ c. Ynd nu n\a-w-az/nan be-gta. 1sg.erg water(abs) 1sgA>2sgU.npreh.ipf\ben-caus-pour 2sg-dat1 ‘I’m pouring the water for you.’ d. Ynd nu begta w-azr-s 1sg.erg 1sgA water(abs) 2sg-dat1 caus-pour-nmlz n\a-wa-pap/ndn. 1sgA>2sgU.ppf\ben-caus-begin ‘I’m trying to pour the water for you.’ The benefactive alternation is a particularly interesting one in Nen because it appears that all but one of the ditransitive verbs in Nen – i.e. those which index the beneficiary by undergoer prefix – include the benefactive prefix, in either a productive or a lexicalized form. An inspection of the trivalent verbs listed in § 3.3 will reveal that all but wétls ‘tell to’ begin with wa-, the form of the benefactive prefix. In some sense, then, most ditransitive verbs in Nen are morphologically derived.

5.7 ‘Placement’ and ‘assume position’ alternations An − − −

important and productive three-way alternation in Nen is between intransitive positionals (‘be in position/posture P’), which are prefixing verbs their transitive ‘placement’ counterparts ‘place/cause to be in position P’ their middle counterparts ‘assume/come to be in position P’.

This alternation set exhibits a number of special characteristics that cannot be dealt with here – see Evans (2014) for details. All positionals participate in this alternation. The basic form means ‘be in position X’. This has no infinitive, just a root, and is morphosyntactically a prefixing

Valency in Nen

1099

Tab. 3: Placement and assume position alternations. verb structure

Positional U[X]-V-STAT

Placement U[X]V-A[Y]

Assume position M-V-A[X]

meaning

‘X be in location/ position P’

Cause(Y, P(X)) a) ‘Y place X in location/ position P’ (e.g. place up high) b) ‘Y cause X to have positional characteristic P’ (e.g. cause to be open)

Become(P(X)) a) ‘X come to be in location/ position P’ b) ‘X come to have positional characteristics P’ (e.g. become open, shattered)

‘be erected’

trom ‘be erected (e.g. a house)’ ser ‘be in water, be immersed’

tronḡs ‘erect, build (e.g. a house)’ wésers ‘put in water, immerse’

ätronḡs ‘come to be erected, built’

‘be in water’

ésers ‘come to be in water, come to be immersed’

verb with an extra stative suffix. The derived verb of (causative) placement is a transitive verb meaning ‘to put in position X’. In general the infinitive of placement verbs is based directly on the positional root; their finite forms take an undergoer prefix and an actor suffix. An example illustrating the alternation between positional and placement frames is: (36) a. Wagib nu-wan y\éser/ngr. fish(abs) water-loc 3sgU.npreh.ipfv.stat\be.immersed ‘The fish is in the water.’ b. Ynd sombes wagib nu-wan e\ser/an. 1sg.erg two fish(abs) water-loc 1sgA>3duU.npreh.ipf\put.in.water ‘I put two fish in the water.’ In most but not all cases, the placement infinitive is built from the positional stem, with the middle infinitive then derived from the transitive by a prefix formally resembling the reflexive/reciprocal or decausative. Examples are kms/äkms ‘lay (v.t.)/lie down (v.i.)’, pis/äpis ‘place up high / get (oneself) up high’, mss/emss ‘lean (v.t.) / get into a leaning position (v.m.)’ and numerous other pairs. Note the difference of these pairs from the causative alternation, where the direction of derivation goes the other way, from middle to transitive. However, a couple of triplets in this set follow the direction of causative derivation, e.g. amzs/

1100

Nicholas Evans

wamzs ‘sit (v.i.)/set, sit, seat (v.t.)’. As is characteristic of prefixing verbs, the basic posals lack infinitives, whereas the placement and middle forms have them. Posal √dar ‘be open [stem]’

Placement dars ‘cause to be open’

Middle adars ‘become open’

Within the Leipzig set, this alternation is not well represented. The ‘put’ verb in the Leipzig list can be translated into Nen with dozens of more specific verbs, depending on the position into which the object is put. In the data base I take the two verbs kms/äkms ‘lay/be lying’ and wémzs/amzs ‘sit, place/be in a sitting position’ as examples. Appendix 2 gives a fuller listing of this rich and elaborated set.

5.8 Conative alternation For at least one perception verb series, concerning audition, there is evidence of a verb-coded conative alternation. As indicated above, ‘hear’ can be coded either by the middle verb eres or the transitive verb weres (examples of both in (38 below)). In both of these cases the P is coded with the absolutive, though the flagging of the perceiver changes – ergative with the transitive form weres, and absolutive with the middle eres. On top of this alternation, it is possible to prefix o- (elsewhere the autobenefactive) to the transitive form and employ an ABS: AL case frame to give the conative meaning ‘listen out for’, as in (37). So far this is the only example of this alternation. (37) Bä n\o-w-ere/te traka-ta. 3sg.abs 3sgA.npreh.ipfv\ab-caus-hear truck-all ‘He is listening out for the truck.’

6 Other alternations 6.1 Verbs exhibiting middle/transitive alternations without any diathetic prefix For at least one verb, a coding alternation between ERG:ABS and ABS:ABS frames reflects a lexical difference between ‘perceive’ and ‘attend to carefully’ rather than any actual difference in argument structure. This involves the transitive and middle forms of eres ‘to hear’ (note that the prefix y- is realised as 0̸ before the vowel /e/, which is why no material precedes the stem in (38a)).

Valency in Nen

1101

(38) a. Ymam boraro ke \ere/te. 3sg.erg flute noise(abs) 3sgA>3sgU.npreh.ipfv\hear ‘He can hear the sound of a flute.’ (‘just hearing it’). b. Bä boraro ke n\ere/te. 3sg.abs flute noise(abs) 3sgA.npreh.ipfv\hear ‘He is listening carefully to the sound of the flute.’

6.2 Other voice alternations not treated as valency alternations Resultatives are encoded by a sort of special deverbalizing construction rather than a valency alternation. This is signalled by adding the ‘source’ case suffix -mne to the nominalized form, then combining the resultant word with a copula. Cf. wapars ‘to cover’, waparsmne ‘covered’; wayanḡs ‘to show’, wayanḡsmne ‘already shown’. (39) a. Ymam toge äme-ngama y\ap/ne. 3sg.erg child(abs) mat-abl 3sgA>3sgU.npreh.ipfv\cover ‘(S)he covered the child with the blanket.’ b. Toge äme-ngama wapar-s-mne y\m. child(abs) mat-abl cover-nmlz-SOURCE 3sgU.npreh.ipfv\be:nd ‘The child is covered with the blanket.’ State passives add -ere to the middle infinitive (without the transitive marker). Consider the following deverbal derivatives of waps ‘fill’: (40) a. Kap (nu-ngama) ap-s-ere y\m. cup(abs) water-abl fill-nmlz-pass.stat 3sgU.npreh.ipfv\be:nd ‘The cup is full (of water).’43 b. Kap (nu-ngama) w-ap-s-mne y\m. cup(abs) water-abl caus-fill-nmlz-source 3sgU.npreh.ipfv\be:nd ‘The cup has been filled (with water).’ Negative state passives are formed by adding the privative suffix -pner ~ -pnä to the infinitive: (41) Toge w-apar-s-pner y\m. child(abs) caus-cover-nmlz-priv 3sgU.npreh.ipfv\be:nd ‘The child is uncovered.’

43 Note that the infinitive aps is also attested in Kmboke marete ws nungama apst pitas ym. ‘This broad-necked bottle is easy to fill.’

1102

Nicholas Evans

7 Discussion/conclusion Nen is a language in which valency classes are clearly flagged, through its ergative/absolutive case-marking system, its system of double indexing, and the overt signalling of valency change on the verb through diathetic prefixes and transitive/ middle alternations. With regard to its verb classes, it has the following four striking features. a) the very small set of true intransitives, in the sense of verbs for which both morphological and syntactic criteria treat them as fully intransitive. Verbs which would be typologically expected to be intransitive are assigned either to inherent middle verbs, or to experiencer object verbs. In the case of inherent middle verbs, there is just one NP argument, but both argument indexing slots (U and A) are filled. In the case of experiencer object verbs, there is a fixed transitive subject argument bearing the ergative case and indexed in the third person actor slot of the verb. b) its large and semantically elaborated set of positional verbs, the causative derivatives of which give a large number of more semantically specific counterparts to PUT. c) indexing in Nen, which in the form of undergoer prefixes does not distinguish true objects (patients, themes) from indirect objects (obliques), even though these receive different case-marking (respectively absolutive and oblique/dative) and obligatorily employ a benefactive prefix in the diathetic slot. This is also the case in e.g. Arammba (Boevé & Boevé 2003). However, other members of the family, such as Kómnzo (Doehler in prep.), have a distinct series for indirect objects. d) the presence of both ergative/absolutive case-coding and a morphologicallysplit system for monovalent verbs (i.e. prefixing vs. ambifixing verbs) might be expected to reflect a semantic emphasis on the coding of agency or volitionality in verb classes, along the lines expected from languages like Italian where auxiliary selection with monovalent verbs is sensitive to agentivity (among other factors) – see e.g. Sorace (2000). But, surprisingly, coding patterns in Nen are not really sensitive to this dimension – ‘fall’ and ‘burn (itr.)’ pattern just like ‘work’ and ‘talk’. Rather, the contrast primarily reflects a stative vs. dynamic contrast: statives are lexicalized as prefixing verbs, and one-place dynamics as ambifixing middle verbs. There are just a couple of exceptions to this – the prefixing verb utan ‘walk’ is likely to be a historical relic of a time when prefixing verbs occupied a greater proportion of the lexicon, and ‘come’ and ‘go’, though their translations suggest dynamicity, are formally based on the stative ‘be’ plus a directional. With regard to coded alternations, five points are worthy of note:

Valency in Nen

1103

a) the series of diathetic prefixes play a crucial role, tied up with overtly-coded valency alternations. There are two of these for increasing valency – causative/ transitivizing w-, and benefactive (w)a- and several for decreasing it. There is a limited set of circumstances in which two overt diathetic prefixes can construct two successive changes of valency. b) with the exception of alternations based on referential equivalence of two argument positions (i.e. reflexives/reciprocals), all verb alternations alter the indexing but leave the flagging alone, at least for those arguments that participate in both alternations. For example, in both causatives and decausatives the affected argument remains in the absolutive (a cause or agent in the ergative is added or subtracted). Likewise, in the benefactive alternation the beneficiary/ recipient remains in the dative or oblique in both constructions – what differs is whether it gets indexed on the verb or not. c) in contrast to well-studied West European languages like French, Italian, Spanish or German, the task of coding such alternations as reflexives, reciprocals, autobenefactives and decausatives is split across two sites – one (the middle prefix) that formally occupies the undergoer indexing slot on the verb, and a second (the diathetic prefix) that gives more precise information about the nature of the alternation (reflexive/reciprocal vs. autobenefactive vs. decausative). d) there is overt marking of reflexive/reciprocal alternations throughout the verb lexicon – there are no ‘light reflexives/reciprocals’ in which reflexive or reciprocal readings are obtained without overt marking of the verb. e) the three-way alternation between positional, transitive verbs of placement, and middle verbs of assuming position, is a central set of alternations with a cluster of convergent properties. Note that this series makes use of alternations in the argument-indexing positions (i.e. prefixing > ambifixing transitive > ambifixing middle). It does not employ diathetic prefixes to get from the stative to the dynamic forms, but within the dynamic forms the diathetic prefixes are widely used to distinguish transitive ‘place in position’ meanings from middle ‘assume position’ meanings, primarily by adding a prefix to the transitive form. Overall, the pattern of indexing suggests the strong influence of a morphological template, for the verb, which favors the employment of two argument positions, no more, no less. A number of factors conspire to maintain this, the two most important being (i) the favoring of middle constructions, with their filling of both undergoer-prefix and actor-suffix slots, for a bloated proportion of one-participant events, and (ii) limitation of agreement to just two argument positions with ditransitives (even though some information about the third argument can be expressed through coerced uses of directional prefixes and aspect). Because of the early stage of research on these languages we are not in a position to make observations about diachronic trends in the membership of the vari-

1104

Nicholas Evans

ous valency classes – a task we will be investigating in the next couple of years. It is worth noting, however, that other languages in the family include somewhat different verbs in their prefixing class – Kómnzo includes ‘shout’, for example (Döhler, in prep.), which is an inherent middle verb in Nen. However, there does appear to be a preference right through the family to employ some form of divalent coding as far as possible. In addition to the middle construction exemplified here, there is also a construction of the type ‘walk walking’, ‘sing singing’ etc. that is somewhat restricted in Nen but much more widespread in the more westerly languages of the family, such as Yei (Boelaars 1950).

Appendix 1: Orthography summary (graphemes in < >) Consonants:

Voiceless stop Voiced stop Prenasalised stop Nasal Voiced fricative Voiceless fricative Lateral Trill Semi-vowel

Bilabial

Alveolar/ dental

p b mb m

ṱ d nd n