Ergativity, Valency and Voice 9783110227734, 9783110227727

Citation preview

Ergativity, Valency and Voice

Empirical Approaches to Language Typology 48

Editors Georg Bossong Bernard Comrie Yaron Matras

De Gruyter Mouton

Ergativity, Valency and Voice

edited by Gilles Authier Katharina Haude

De Gruyter Mouton

ISBN 978-3-11-022772-7 e-ISBN 978-3-11-022773-4 ISSN 0933-761X 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 in the Internet at http://dnb.dnb.de. 쑔 2012 Walter de Gruyter GmbH & Co. KG, Berlin/Boston Typesetting: RoyalStandard, Hong Kong Printing: Hubert & Co. GmbH & Co. KG, Göttingen ⬁ Printed on acid-free paper Printed in Germany www.degruyter.com

Table of contents List of contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vii

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gilles Authier and Katharina Haude

1

Ergativity and voice in Mayan languages: a functional-typological approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Colette Grinevald and Marc Peake

15

Ergativity and the passive in three Mayan languages. . . . . . . . . . . . Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

51

A tale of two passives in Cavinen˜a . . . . . . . . . . . . . . . . . . . . . . . . . Antoine Guillaume

111

The detransitive voice in Kryz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gilles Authier

133

Laz middle voice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rene´ Lacroix

165

Argument demotion in Japhug Rgyalrong. . . . . . . . . . . . . . . . . . . . Guillaume Jacques

199

The Katukina-Kanamari antipassive . . . . . . . . . . . . . . . . . . . . . . . . Francesc Queixalo´s

227

Undergoer orientation in Movima . . . . . . . . . . . . . . . . . . . . . . . . . Katharina Haude

259

Case patterns and verb classes in Trumai . . . . . . . . . . . . . . . . . . . . Aurore Monod Becquelin and Ce´dric Becquey

289

Ergativity in the Adyghe system of valency-changing derivations. . . Alexander Letuchiy

323

vi

Table of contents

The evolution of transitive verbs in Basque and emergence of dative-marked patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ce´line Mounole

355

Index of subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Index of languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

381 387

List of contributors Gilles Authier Institut National de Langues et Cultures Orientales 65 rue des Grands Moulins 75013 Paris [email protected] Ce´dric Becquey Laboratoire de Phone´tique et Phonologie Universite´ Sorbonne Nouvelle – Paris 3 19 rue des Bernardins 75005 Paris France [email protected] Colette Grinevald Dynamique du Langage, CNRS & Universite´ Lyon 2 14 avenue Berthelot 69363 Lyon Cedex 07 France [email protected] Antoine Guillaume Dynamique du Langage, CNRS & Universite´ Lyon 2 14 avenue Berthelot 69363 Lyon Cedex 07 France [email protected] Katharina Haude Structure et dynamique des langues, CNRS, INALCO, IRD 7 rue Guy Moˆquet 94801 Villejuif Cedex France [email protected]

Guillaume Jacques Centre de Recherces Linguistiques sur l’Asie Orientale, EHESS, CNRS 131 Boulevard Saint-Michel 75005 Paris France [email protected] Rene´ Lacroix Langues et Civilisations a` Tradition Orale, CNRS 7 rue Guy Moˆquet 94801 Villejuif Cedex France [email protected] Alexander Letuchiy National Research University Higher School of Economics Khitrovskiy sidestreet, house 2/8, building 5 109028 Moscow Russia [email protected] Aurore Monod Becquelin Laboratoire d’Ethnologie et de Sociologie Comparative, CNRS Maison Rene´ Ginouve`s (MAE) 21 alle´e de l’Universite´ 92023 Nanterre France [email protected] Ce´line Mounole Hizkuntzalaritza eta Euskal Ikasketak Saila Unibertsitateko Ibilbidea, 5 01006 Vitoria-Gasteiz France [email protected]

viii

List of contributors

Marc Peake Dynamique du Langage, CNRS & Universite´ Lyon 2 14 avenue Berthelot 69363 Lyon Cedex 07 France Francesc Queixalo´s Structure et dynamique des langues, CNRS, INALCO, IRD 7 rue Guy Moˆquet 94801 Villejuif Cedex France [email protected]

Valentina Vapnarsky Centre EREA – Laboratoire d’Ethnologie et de Sociologie Comparative, CNRS 7 rue Guy Moquet 94801 Villejuif Cedex France [email protected]

Introduction Gilles Authier and Katharina Haude 1. Preliminaries1 This volume explores voice phenomena in a substantial sample of languages with predominantly ergative features, some of them already well studied, others less so. All the findings presented here are based on firsthand data and personal fieldwork in speech communities. The articles take up and elaborate on oral presentations given at the monthly seminar Ergativite´: typologie, diachronie et cognition, which took place between 2005 and 2009 in Villejuif (Paris) under the direction of Francesc Queixalo´s and with the financial support of the Fe´de´ration de Typologie et Universaux Linguistiques of the CNRS, whose major purpose is to bring together linguists with first-hand experience of languages with ergative features from di¤erent parts of the world.

2. Ergativity Ergative alignment – defined as the formal singling out of the agent (A) of transitive verbs as opposed to the patient (P) of transitive verbs and the single argument of intransitive verbs (S) – can be found at many levels of grammar, both in the marking of grammatical relations and at the syntactic level (within a single clause or in interclausal relations). The former is usually referred to as ‘‘morphological ergativity’’, while the latter, along with ergative alignment in terms of constituency and reflexive control, is called ‘‘syntactic ergativity’’ (Dixon 1994). At the clause level, ergative alignment can be recognized in the morphological encoding of cross-reference pronouns on the predicate or in the use 1. The editors (names are in alphabetical order) wish to express their gratitude to Steven Kaye for accepting the task of checking the English of the articles written by non-native speakers. Haude acknowledges the support of the Deutsche Forschungsgemeinschaft (HA 5910/1-1) during the course of this project. We thank F. Queixalo´s and Anna Siewierska for helpful comments on this introduction. Needless to say, all mistakes and shortcomings are entirely our own responsibility.

2

Gilles Authier and Katharina Haude

of cases to mark the function of its dependents (‘‘core arguments’’ in Role and Reference Grammar, Foley and Van Valin 1984; ‘‘actants’’ in the terminology of Lazard 1994), or in both. Alignment could in theory also be ergative at a syntactic level within simple clauses, but if word order or control of reflexive pronouns may not show strictly accusative behaviour in some languages, these syntactic parameters are, to our knowledge, never clearly ergative either. In terms of inter-clausal syntax, ergative behaviour manifests itself in terms of so-called ‘‘pivots’’ (Dixon 1994). Very few languages have been claimed to have an ergative pivot, that is, a constraint in equi-deletion that groups an omitted S with P and not A. In such languages, like Dyirbal or its close relative Yidiny, the ergative (S ¼ P) pivot seems to coexist in basic clause-types with the most widely attested type of pivot, S ¼ A, also called ‘‘subject’’ by Dixon. In some families of ‘‘ergative’’ languages, such as East Caucasian, there is no clear category of pivot, and argument recovery between clauses is subject to pragmatic tendencies rather than syntactic rules; but this is also true of some accusative or active-stative languages. However, the vast majority of languages with accusative morphology display clear accusative pivot properties. It should not be seen as overly cautious to refer to ‘‘ergative languages’’ in quotation marks. The expression ‘‘ergative language’’ continues to be used here for the sake of brevity, as in many reference works on the topic, but it should be recognized that languages are not expected to be typologically consistent across all aspects of their alignment behaviour. Some notoriously ‘‘accusative’’ languages such as Latin, French and in fact many Indo-European languages may have some hints of ergativity, especially in perfective nominalizations (past participles; see Lehmann 1985), whereas many languages with ergative alignment on nouns, including languages like Dyirbal which exhibit strongly ergative interclausal syntactic rules (Dixon 1972; Dixon 1994: 14–15, 160), show accusative alignment on personal pronouns referring to speech act participants (see Silverstein 1976). Indeed, most if not all ergative languages also display some accusative alignment according to a set of parameters which all contribute to a global, multifaceted, and scalar definition of transitivity (Hopper and Thompson 1980). According to these parameters, a variety of syntactic constructions reflect di¤erent perceptions of events involving two participants, and may deviate from the prototypical, maximally transitive construction in which the agent is also the main topic and in control of the event (an action) while the patient is wholly a¤ected by this action. Devia-

Introduction

3

tions from this scenario account for a number of variations in transitivity also called actancy splits (Lazard 1997) or simply ‘‘splits’’ (Dixon 1994). The very common ‘‘pronoun split’’ mentioned above, in which pronouns align accusatively and nouns ergatively, was first described by Silverstein (1976), who argued that accusative marking is more expected at the upper end of a scale of animacy, individuation, and natural topicality or empathy, because constituents referring to higher animates, especially personal pronouns, are for semantic reasons less expected in direct object position. This does not mean that ergative marking is unknown on personal pronouns, but it is rare, and often an accusative or specific non-ergative marker is found on personal pronouns in otherwise morphologically fully ergative languages, as in Kryz and Cavinen˜a (see Authier and Guillaume, this volume). A very common intermediate situation in ergative languages is for personal pronouns to have neutral alignment, because ergative case on nouns and word order make case marking dispensable on arguments referring to the most naturally agentive participants. Conversely, ergative alignment is more likely to be found at the lower end of the Silverstein hierarchy, because inanimates need to be more heavily marked when used in the less expected syntactic position of agents of transitive verbs, and indeed, some overall ‘‘accusative languages’’ such as Hittite have ergative case marking for inanimates only (Laroche 1962). In accusative languages with little case marking morphology, e.g. many modern Indo-European languages (English, Spanish or French), only personal pronouns retain the old case marking system making use of suppletive stems (I/me). And symmetrically, even in accusative languages with well developed core case marking, accusative case is rarely marked, or at least less heavily expressed, on nouns referring to entities placed very low in the hierarchy, for example on members of the neuter gender in Latin. Aspectual splits also famously interfere with alignment, and represent another essential parameter in the definition of transitivity as a polyparametric scalar phenomenon. Although such splits have been described in languages which display very di¤erent morphological systems to mark grammatical relations, such as Kurdish, Georgian (a relative of Laz, in this volume) or Mayan, the rule remains the same: ‘‘if a split is conditioned by tense or aspect, the ergative is almost always found either in the past tense or in perfective aspect’’ (Dixon 1994: 99). Conversely, many so-called ergative languages tend to exhibit exceptions to their prevalent ergative alignment in predicates expressing progressive-imperfective aspect or present and future tense, because these imply a lower degree of a¤ectedness of the object. One clear instance of this tendency is the

4

Gilles Authier and Katharina Haude

obligatory use of antipassive voice to express progressive aspect or present tense. In some documented instances, e.g. in Georgian, a historically derived antipassive has evolved into the only means of expressing non-perfective, progressive processes. This volume gives a representative impression of the diversity of these often intricate situations. Before presenting the voice phenomena found in the language at issue, contributors provide succinct but precise information about its particular pattern of ergative behaviour. Some languages, for example the majority of Mayan languages, only have ergative crossreferencing of arguments on the verb. But, as shown in one of the contributions to this volume, Tseltal and Yucatec in fact possess a split intransitive system: two kinds of S can be distinguished, one of which aligns with A while the other aligns with P. Many languages represented in this volume (Adyghe, Basque, Cavinen˜a, Kryz, Laz) have both ergative/ absolutive case marking on nominals and an ergative system of verb indexes, but this situation may not be the one most commonly found: accusative indexing on verbs often coexists with ergative case marking, while the reverse situation seems to be unattested; and a few languages under scrutiny in this volume, such as Movima and Trumai, have two competing constructions – neither of which can be considered ‘‘basic’’ – for predicates with two arguments, one showing ergative case marking while the other shows accusative alignment. We can confirm that, although they may be quite diverse in nature, alignment splits are the rule for languages which have ergative alignment somewhere in their grammar. In fact, it seems that no language has been documented which would qualify as ergative at all levels of its grammar. Not all aspects of those phenomena which fall under the heading of ergativity will be touched upon in this volume; but notwithstanding the great morphosyntactic variety of the languages in this sample, all papers focus on the relationship of ergativity with valency change and voice phenomena. The paper on Basque, which deals with the diachronic evolution of transitive valency indexation under the influence of Spanish di¤erential object marking, addresses a somewhat di¤erent aspect of the relationship between ergativity and valency systems.

3. Voice and valency change Taking both formal and functional criteria into account, voice alternations, as understood in this volume, show the following basic characteristics: in

Introduction

5

terms of form, they determine the number, formal encoding, and semantic role of verbal argument(s); in terms of function, they serve to describe an event from di¤erent perspectives, and to retain the same participant as the central argument through larger stretches of discourse; voice alternations ideally form a productive system. Many definitions of voice are more restrictive than the one given above. They usually treat the notion of ‘‘subject’’ (for the agent in a transitive clause and the single argument of an intransitive clause) as central to voice alternations (cf. Siewierska 1984; Mithun 1994); some (e.g. Kulikov 2010) accept only those alternations that either increase or decrease the number of arguments, but not those where the semantic roles of the core arguments are reassigned without a¤ecting the valency of the verb. Voice is generally restricted to the verbal domain and taken to involve explicit morphological marking (see e.g. Klaiman 1991: 1; Creissels 2006: 6). However, as has frequently been noted, applying these more restrictive definitions can complicate the study of voice phenomena in predominantly ergative languages. In particular, it is well known that the notion of subject is problematic in ergative constructions (see e.g. Blake 1976; Mithun 1994: 247; Shibatani 1998: 120), since it encompasses a relation which is treated di¤erently in transitive and intransitive clauses. Moreover, while in most predominantly ergative languages the ergative-marked argument of the transitive clause can be identified as the syntactic subject (see e.g. Anderson 1976), in some languages (e.g. Katukina, Queixalo´s this volume) it is the absolutive, rather than the ergative, argument of transitive clauses whose syntactic status is comparable to that of the subject in a nominativeaccusative construction. The hybrid character of the relation encompassing the transitive agent argument and the single argument of intransitive clauses is especially interesting for a consideration of the passive and antipassive, the classic valency-decreasing voices that serve to maintain the syntactically privileged status of the topical participant (see Dixon 1994, Van Valin and LaPolla 1997, Primus 1999). As for the passive – an operation whereby the patient becomes the subject of a derived intransitive clause – it has sometimes been assumed that this operation is not usually available in predominantly ergative languages (Shibatani 1998: 120). However, as is shown by several articles in the present volume, many ergative languages do possess passive operations, especially when the agent argument of the transitive clause is syntactically privileged. At the same time, the antipassive – an operation whereby the agent becomes the single argument of a derived intransitive clause – remains a typical operation of ergative languages, as it allows

6

Gilles Authier and Katharina Haude

the role which is marked in the transitive construction to appear unmarked in the intransitive construction (analogously to the accusative/nominative marking alternation in the passive constructions of accusative languages; see Silverstein 1976: 140). In nominative-accusative languages, the functions of antipassives (including aspectual functions, see Cooreman 1994) are often fulfilled by the intransitive use of an otherwise transitive verb, which entails no di¤erence in the marking of the agent argument (nominative). Languages can have two or more basic transitive constructions, including both accusative and ergative types, which alternate in ways similar to voice. The best-known example is found in the ‘‘symmetrical voice’’ systems of many Austronesian languages, which provide ‘‘for any transitive event [. . .] at least two representations, one in actor voice and one in undergoer voice’’ (Himmelmann 2005: 135), both of which are equally morphologically marked. Some Austronesian languages clearly display ergativity in their morphosyntax (see Mithun 1997 on Kapampangan); in others, like Tagalog, the most frequent transitive construction takes the patient as the privileged argument (the ang phrase), which is why some scholars consider it to be ergative in nature (see the overview in Shibatani 1988). What is of particular interest here is that these systems correlate with certain features which, to a greater or lesser degree, also appear in languages with more straightforward ergative traits, like those of the Mayan or Eskimo families: verbs share at least some syntactic characteristics with nouns, and the ergative argument shows similarities to an adnominal possessor. A possible diachronic scenario for the rise of ergative structures, attested in IndoAryan languages (e.g. Benveniste 1966a: 176–186, 1966b: 127–136), is that ergative constructions arose from the frequent use of passive participles as intransitive predicates, which had the patient as subject and the agent as an oblique-marked (genitive or dative) argument. Another particular case in the domain of argument adjustment that does not change the number of verbal arguments can be found in so-called inverse systems. These systems display two transitive constructions, direct and inverse, which are chosen depending on the position of the event participants of the arguments on a referential hierarchy; the choice is typically determined by the opposition between speech-act participants and third persons (see DeLancey 1981). However, the direct-inverse opposition often holds in the third-person domain as well, and here the relative topicality of the arguments plays a role (see Gildea 1994); for instance, in Algonquian languages the choice of the direct or inverse constructions in the thirdperson domain depends on whether a nominal referent is marked as proximate or obviative, and this marking can, in turn, be based on the relative

Introduction

7

topicality of the referents (see Zu´n˜iga 2006: 71). Therefore, when the discourse status of nominal referents is relevant for the selection of the construction used, an inverse system can be analysed in terms of voice (see Givo´n 1994, 2001), and perhaps (when there is no di¤erence in markedness between the two constructions) even interpreted as a more strongly grammaticalized form of symmetrical voice systems (see Farrell 2005). A discourse-based opposition between accusative and ergative transitive constructions is also found in languages of the Amazonian Arawa´ family (Aikhenvald 2009; Dixon 2000). Thus, there are numerous examples of alternations based on discourse properties of referents, which means that voice is not necessarily tied to a change in the number of arguments. Voice is often viewed as a morphological category of the verb and as such is seen as necessarily requiring verbal marking. Such an interpretation of voice excludes from its domain two alternations which do not involve verbal marking, namely lability and lexical alternations, which never seem to be fully productive but can assume a voice function to some degree (see Creissels 2006: 5–6). Lability involves the use of a verb in both transitive and intransitive constructions without any morphological modification. There are two types of lability, as defined by Drossard (1998): orientation-maintaining (e.g. English ‘‘I eat it’’ – ‘‘I eat’’) and orientationchanging (English ‘‘I break it’’ – ‘‘It breaks’’). The pattern of lability found in an ergative system is the opposite of that found in an accusative system, since in the latter the verb is oriented towards the agent, while in the former the verb is oriented towards the patient. A complex issue is seen in the situation where di¤erent lexical verbs with contrasting argument structures can be chosen depending on which participant is to be highlighted, since potential semantic contrasts between lexical verbs are even less easily understood than those involved in morphosyntactic alternations. A possible example is presented by Trumai (Monod Becquelin and Becquey, this volume), a language without voice morphology, where voice functions seem to be realized through lexical alternations: as argued by the authors, the ergative and accusative constructions are used to preserve the discourse topic as the unmarked argument (the ‘‘subject’’). Finally, while voice is almost always seen as belonging to the verbal domain, di¤erential marking of nominal constituents may have a similar function (see Shibatani 2006: 229) in cases where di¤erential argument marking is not exclusively determined by person or animacy but also by topicality, as reflected e.g. by definiteness. In this volume, Mounole’s description of the emergence of di¤erential object marking through language contact in Basque hints at this possible connection.

8

Gilles Authier and Katharina Haude

Aside from these issues, languages with predominantly ergative morphosyntax (like many others) also display other valency-decreasing devices, such as middles, anticausatives, and noun incorporation, as well as languagespecific mechanisms or subclasses of the above; these are discussed in many of the contributions presented here. Valency-increasing devices such as causatives, benefactives or applicatives can be counted as belonging to the domain of voice as well, since they can be employed for discoursepragmatic purposes, e.g. topic maintenance. However, the wide range of constructions and linguistic variation which is found in this domain deserves discussion of its own, and therefore such phenomena are not specifically dealt with in this volume.

4. Contributions to this volume Mayan languages have been at the core of discussions on ergativity ever since the 1970s, and it is only natural that the volume starts out with two articles on this language family. C. Grinevald and M. Peake o¤er an introduction to transitivity and voice phenomena in Mayan languages, drawing mainly on their expertise in two languages of the Q’anjob’alan branch of the family, Jakaltek Popti’ and Tojol Ab’al. They illustrate the mechanisms by which voice alternations in Mayan languages serve to promote and demote clausal arguments and thereby structure the discourse-pragmatic organization of sentences and texts. The article by V. Vapnarsky, A. Monod Becquelin and C. Becquey focuses on the passive voices found in Ch’ort’i, Tseltal and Yucatec, languages from di¤erent branches of the Mayan family which display various degrees and reflexes of morphological ergativity. Analysing the conditions of occurrence of the passive constructions in these languages, the authors show that (possibly in all Mayan languages) the passive has semantic and discourse-pragmatic (agent-defocusing) rather than syntactic functions. By also comparing the languages with respect to the other voices they display, especially the ‘‘focus antipassive’’ typical of many Mayan languages, they arrive at the conclusion that the array and type of voice constructions a language displays does not, by itself, provide clues about its basic syntactic alignment. The subsequent four papers deal with passive and passive-like constructions in unrelated languages from di¤erent geographical areas. A. Guillaume describes two apparently related verbal morphemes (-ta and -tana) in Cavinen˜a (Tacanan, lowland Bolivia), both somewhat re-

Introduction

9

stricted in use, which yield passive and anticausative interpretations respectively. In his article, the author revises earlier hypotheses on the origin of these morphemes and their semantic and diachronic relationship. G. Authier’s account of the detransitive voice in Kryz first outlines the morphosyntax of transitive verbs as a subset of bi-actant constructions (a typical East Caucasian case of semantic encoding of various types of subjects). The semantic spectrum of the Kryz detransitive voice, which synchronically has much in common with those of Cavinen˜a and Laz, is described in detail and set alongside comparable constructions and morphemes in related languages. East Caucasian languages rarely have valency-decreasing voices, and indeed, Kryz is the only member of this family known to have a passive voice for most transitive verbs; this form also has anticausative value with most verbs, and allows an antipassive interpretation with a just a few of them. The passive interpretation is shown to be a recent innovation linked to contact with sociolinguistically dominant Azeri, a genetically unrelated, uniformly accusative language. R. Lacroix describes the di¤erent functions of the voice marker i- in the South Caucasian (Kartvelian) language Laz, which he analyses as a marker of middle voice, that is, a category encompassing a variety of ways in which the transitivity of the prototypical two-participant construction can be modified. While this marker is widespread in related languages, Lacroix o¤ers a novel approach and a fine-grained description of this element of the Kartvelian set of valency-changing morphemes. Japhug Rgyalrong (Qiangic, Tibeto-Burman) has several detransitivizing voices, some of which are more frequently used than others. G. Jacques’ article provides an overview of the rich morphology associated with the demotion of indefinite (unknown or generic) arguments in this language. patients can be suppressed by means of generic, antipassive, and incorporating constructions and labile verbs, while agents can be suppressed in the generic and passive constructions. The three subsequent articles deal with languages of the Amazon area, all three of which display ergative syntactic patterns. F. Queixalo´s demonstrates that the absolutive argument in Katukina has a privileged syntactic status and shows that the agent argument of transitive clauses can acquire this privileged status only through an antipassive operation. In this language, therefore, the antipassive clearly has a syntactic function. K. Haude discusses the correlation between ergative patterns and patient orientation in Movima. Movima has two basic transitive constructions, direct and inverse, with direct main clauses displaying ergative, and inverse main clauses displaying accusative syntax. The direct and inverse markers

10

Gilles Authier and Katharina Haude

belong to a paradigm of verbal voice morphemes that determine the argument structure of the verb. It is argued that the verbs participating in this scheme are basically oriented towards the non-agent, and that this may be one of the reasons why the direct/ergative construction is less pragmatically marked than the inverse/accusative construction. C. Becquey and A. Monod Becquelin’s article discusses textual data from Trumai, a linguistic isolate spoken in the Amazonian region of Brazil. Trumai lacks derivational voice alternations, but there are two transitive constructions – one ergative, the other accusative, depending on the lexical verb involved – whose interplay is employed for topic-maintaining purposes. While the authors make it clear that the domain of verb semantics in Trumai requires further research, the findings suggest that we may be dealing with a system in which voice functions are taken over by lexical alternations. The last two papers bring the discussion back to Europe. A. Letuchiy’s article on Adyghe (West Caucasian, not genetically related to Laz and South Caucasian or to Kryz and East Caucasian) explores ergative alignment from the point of view of voice phenomena, claiming that the behaviour of arguments in voice derivation is a more decisive criterion for the identification of syntactic ergativity than the pivot phenomena previously studied by Dixon, which generally appear not to be characteristic of ergative languages of the Caucasus. C. Mounole’s paper deals with the emergence of di¤erential direct object marking in Basque, showing that it has been progressively induced by contact with Spanish during the modern period. Di¤erential object marking is rare in ergative languages, but elsewhere it is attested at least in West Iranian ergative languages (cf. Bossong 1985) and in Udi (East Caucasian, cf. Schulze 2008), giving rise to tripartite alignment systems. Although this paper does not deal with voice specifically, it may be viewed as an incentive to pursue the exploration of potential links between various parameters of transitivity and the diversification of valency frames in predominantly ergative languages.

5. Summing up: What this book contributes to typology In sum, this volume shows once again that so-called ‘‘ergative languages’’ do not represent a homogeneous group. A language whose basic mainclause morphosyntax is predominantly ergative does not necessarily show

Introduction

11

the same voice phenomena as another language with ergative characteristics, probably because manifestations of ergativity are far from homogeneous, ranging from the overt case marking of an agent without any syntactic e¤ects to the sharing of some syntactic interclausal properties of the patient of transitive clauses with the single argument of intransitive clauses. Many articles in this volume show that passive voices (often with associated anticausative value) are very common in ergative languages and signal the demotion or even the total suppression of agents, but such semantic factors as the expression of perfective aspect or deontic modality seem to be responsible for their extension and diversification. The antipassive voice, which demotes or suppresses the patient, is often taken to be an expected phenomenon in languages with ergative systems, and a substantial number of the languages represented in the volume do display antipassive constructions; however, even in ergative languages its use seems to be comparatively rare. Finally, some articles show that ergative structures themselves may belong to a voice system in which they alternate with accusative constructions. In conclusion, voice phenomena, as a subcategory of information structure modifying devices, are certainly related to patterns of interclausal syntactic alignment, but probably not to the coding of grammatical relations within the clause.

References Aikhenvald, Alexandra Y. 2009 ‘‘Syntactic ergativity in Paumarı´.’’ In Topics in Descriptive and African Linguistics. Essays in Honor of Distinguished Professor Paul Newman, Samuel Gyasi Obeng (ed.), 111–127. Mu¨nchen: Lincom Europa. Anderson, Stephen S. 1976 ‘‘On the notion of subject in ergative languages.’’ In Subject and topic, Charles Li. (ed.), 1–24. New York: Academic Press. Benveniste, Emile 1966a Proble`mes de Linguistique Ge´ne´rale. Vol. 1. Paris: Gallimard. Benveniste, Emile 1966b Proble`mes de Linguistique Ge´ne´rale. Vol. 2. Paris: Gallimard. Blake, Barry 1976 ‘‘On ergativity and the notion of subject.’’ Lingua 39(4): 281–300. Bossong, Georg 1985 Di¤erentielle Objektmarkierung in den neuiranischen Sprachen. Tu¨bingen: Narr.

12

Gilles Authier and Katharina Haude

Cooreman, Anne 1994 ‘‘A functional typology of antipassives.’’ In Fox, Barbara A. and Paul J. Hopper (eds.), 49–87. Creissels, Denis 2006 Syntaxe ge´ne´rale. Une introduction typologique. Vol. 2: La phrase. Paris: Herme`s-Lavoisier. DeLancey, Scott 1981 ‘‘An interpretation of split ergativity and related patterns.’’ Language 57(3): 626–657. Dixon, R.M.W. 1972 The Dyirbal Language of North Queensland. Cambridge: Cambridge University Press. Dixon, R.M.W. 1994 Ergativity. Cambridge: Cambridge University Press. Dixon, R.M.W. 2000 ‘‘A-constructions and O-constructions in Jarawara.’’ International Journal of American Linguistics 66: 22–56. Drossard, Werner 1998 ‘‘Labile Konstruktionen.’’ In Kulikov and Vater (eds.), 73–84. Farrell, Patrick 2005 Grammatical Relations. Oxford: Oxford University Press. Foley, W. and Van Valin, R. 1984 Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Fox, Barbara A. and Paul J. Hopper (eds.) 1994 Voice: Form and Function. Amsterdam/Philadelphia: Benjamins. Gildea, Spike 1994 ‘‘Semantic and pragmatic inverse: ‘Inverse alignment’ and ‘inverse voice’ in Carib of Surinam.’’ In Givo´n, T. (ed.), 187–230. Givo´n, T. 1994 ‘‘The pragmatics of de-transitive voice: functional and typological aspects of inversion.’’ In Givo´n, T. (ed.), 3–44. Givo´n, T. 2001 Syntax: an Introduction. Vol. 2. Amsterdam/Philadelphia: Benjamins. Givo´n, T. (ed.) 1994 Voice and Inversion. Amsterdam/Philadelphia: John Benjamins. Himmelmann, Nikolaus P. 2005 ‘‘The Austronesian Languages of Asia and Madagascar: Typological Characteristics.’’ In The Austronesian Languages of Asia and Madagascar, Alexander Adelaar and Nikolaus P. Himmelmann (eds.), 110–181. London/New York: Routledge. Hopper, Paul and Sandra S. Thompson 1980 ‘‘Transitivity in Grammar and Discourse.’’ Language 56: 251–299.

Introduction

13

Klaiman, M.H. 1991 Grammatical Voice. Cambridge: Cambridge University Press. Kulikov, Leonid I. 2010 ‘‘Voice typology.’’ In The Oxford Handbook of Linguistic Typology, Jae Jung Song (ed.), 368–398. Oxford: Oxford University Press. Kulikov, Leonid and Heinz Vater (eds.). 1998 Typology of Verbal Categories. Papers presented to Vladimir Nedjalkov on the occasion of his 70th birthday. Tu¨bingen: Niemeyer. Laroche, Emmanuel 1962 ‘‘Un ‘‘ergatif ’’ en indo-europe´en d’Asie Mineure’’. Proceedings of the Berkeley Linguistics Society 57: 23–43. Lazard, Gilbert 1994 L’actance. Paris : Presses Universitaires de France. Lazard, Gilbert 1997 ‘‘Ergativity (review of R.M.W. Dixon, Ergativity).’’ Linguistic Typology 1: 243–268. Lehmann, Christian 1985 ‘‘Ergative and active traits in Latin.’’ In Relational Typology, Frans Plank (ed.), 243–255. Berlin: Mouton. Mithun, Marianne 1994 ‘‘The implications of ergativity for a Philippine voice system.’’ In Fox and Hopper (eds.), 247–277. Primus, Beatrice 1999 Case and Thematic Roles. Ergative, Accusative and Active. Tu¨bingen: Niemeyer. Schulze, Wolfgang 2009 ‘‘Grammar.’’ In The Caucasian Albanian Palimpsests of Mount Sinai, Jost Gippert, Wolfgang Schulze, Zaza Aleksidze, JeanPierre Mahe´ (eds.), vol. I., 21–60. Turnhout: Brepols. Shibatani, Masayoshi 1988 ‘‘Voice in Philippine Languages.’’ In Passive and Voice, Masayoshi Shibatani (ed.), 85–142. Amsterdam/Philadelphia: Benjamins. Shibatani, Masayoshi 1998 ‘‘Voice Parameters.’’ In Kulikov, Leonid and Heinz Vater (eds.), 117–138. Shibatani, Masayoshi 2006 ‘‘On the conceptual framework for voice phenomena.’’ Linguistics 44(2): 217–269. Siewierska, Anna 1984 The Passive: A Contrastive Linguistic Analysis. London: Croom Helm. Silverstein, Michael 1976 ‘‘Hierarchy of features and ergativity.’’ In Grammatical Categories in Australian Languages, R.M.W. Dixon (ed.), 112–171. Canberra: Australian Institute of Aboriginal Studies.

14

Gilles Authier and Katharina Haude

Van Valin, Robert D. and Randy J. LaPolla 1997 Syntax: Structure, Meaning, and Function. Cambridge: Cambridge University Press. Zu´n˜iga, Fernando 2006 Deixis and alignment. Inverse systems in indigenous languages of the Americas. Amsterdam/Philadelphia: Benjamins.

Ergativity and voice in Mayan languages: a functional-typological approach Colette Grinevald and Marc Peake 1. Introduction1 The Mayan family is a large language family that resembles in many ways the European side of the Indo-European language family, by its time depth, its number of branches and sub-branches, and its distinct languages, around 30 languages.2 Mayan languages are spoken in a relatively contiguous area in Guatemala and Mexico (in the southern regions of Yucatan and Chiapas), with the exception of the geographically isolated Wastekan branch spoken further north, in central Mexico near the Atlantic coast. A tradition of ‘‘Mayan Linguistics’’ developed in the last quarter of the 20th century. Historical and comparative studies led first to a fairly extensive reconstruction of Proto-Maya, and the identification of major Pan-Mayan morpho-syntactic characteristics, such as its ergativity markers (see for example Kaufman (1974), Robertson (1980), Norman and Campbell 1978, Campbell and Kaufman (1985)). A network of field linguists also coordinated themselves to produce (morpho-)syntactic descriptions of topics of particular interest in the contemporary languages. The Mayan family was therefore one of the earlier and better known of the American continent, and contributed interestingly, on several occasions, to ongoing discussions of theoretical and typological interest to general linguistics, as will be mentioned later.3 1. Colette Grinevald was known earlier as Colette Craig. This chapter is partly based on materials from her courses on Mayan Grammar in the Mayan Program of the Department of Amerindian Linguistics of the ‘‘Institut National des Langues et Cultures Orientales’’ (INALCO) in Paris, France. It is meant as an introduction and orientation to Mayan studies of ergativity and voices. Our thanks to various readers for helpful comments on drafts of this paper: an anonymous reviewer, Judith Aissen and the editors of the volume. 2. See Kaufman (1974) for a view of the whole family, reproduced in England (1996). 3. The network organized topical workshops and produced a Journal of Mayan Linguistics. It also contributed to the training of native Mayan speakers,

16

Colette Grinevald and Marc Peake

And ergativity and voices have certainly been some of the features of the Mayan family of languages that have attracted most attention for their being key to the understanding of how Mayan grammars function, and of great interest to general linguistics. This chapter is therefore meant to give an overview of both the facts of Mayan ergativity and voices, but also a sense of how their study has developed and participated to certain developments of descriptive and theoretical linguistics in the past decades. This study of Mayan ergativity and voices begins, in section 2, with an overview of Mayan verbal morphology, to bring to attention the striking multiplicity of transitivity markers typical of most languages of the family. Section 3 then o¤ers a description of the forms and functions of the verbal person markers that justify talking of ergative alignment and of occasional patterns of split ergativity, of these markers known as the ‘‘set A’’ and ‘‘set B’’ in Mayan linguistics terminology. The discussion turns, in section 4, to the role of these set A and set B markers in the identification of a fairly elaborate system of voices, with clear morphosyntactic markings of various types of antipassive, passive and applicative voices and of a typologically rarer and morphologically more ambiguous ‘‘agent focus’’ voice. This presentation of Mayan ergativity and voices has been cast in a ‘‘functional-typological’’ approach to grammar, of the kind originally proposed by Givo´n (1979, 1986, 2001). This means here that the description of the ergative patternings considered will systematically appeal to the notion of patterns of ‘‘alignments’’ between semantic and syntactic roles (as defined in the literature on ergativity, as will be noted then). The choice of this framework also accounts for the quick survey of studies about the discourse use and pragmatic motivations for such a variety of voices in Mayan languages that is presented in section 5.4

2. Mayan verbal morphosyntax Like many languages of the American continent, Mayan languages are head-marking, so that their verb forms carry all the information necessary leading today to a new generation of native linguists (these noteworthy developments are traced in England 1992, 1996, 2007, England and Woodbury 2004, Grinevald 2002, 2007). 4. This is not to say that all the discussions that will be cited have been expressly formulated within such a functional-typological framework, rather that it is the mode of organization of this presentation that is.

Ergativity and voice in Mayan languages

17

to identify the core argument structure of the clause, in the widespread absence of anaphoric pronouns. The ergativity phenomenon is located in these verbal structures: it is found in the special patterning of di¤erent sets of person markers a‰xed to the verb. Other markers contribute also to establishing the transitivity level of the verbal construction. The argumentation in Mayan languages for particular syntactic descriptions is also greatly facilitated by a strictly agglutinating morphology, with relatively little in the way of morphophonemics. 2.1. Pan-Mayan vs language specific verbal morphology This sub-section surveys the di¤erent elements of the verbal complex, considering in which way they are characteristic of the family as a whole and in which way they may vary across branches or languages of the family. – All Mayan languages have a rich system of Mayan roots, predominantly of the form CVC,5 and verbal constructions built on either ‘‘radical’’ or derived verbs. This di¤erence is particularly relevant to our present purpose to the extent that, in some languages, the status of radical or derived verbs may determine a choice between variant forms of voice or aspect markers. – In the TAM (tense-aspect-modality) domain, practically all languages make a basic distinction between ‘‘completive’’ and ‘‘incompletive’’ aspect, while some have further tense and aspect distinctions (such as recent vs. distant past or future, and progressive). Aspect markers are usually prefixed but sometimes su‰xed; some aspects can also be expressed using auxiliary-type forms in periphrastic constructions, particularly the progressive. The main morphological mood markers are for imperative and subjunctive (sometimes found under the label of ‘‘irrealis’’ or ‘‘future’’). It is worth noting that, in some languages, the choice of variant forms of these TAM markers may be determined also by the transitivity level of the verbal construction. – In the domain of person marking, the indexation of the core arguments in the verb form constitutes a clear Pan-Mayan feature. The person markers are organized into two paradigms, known traditionally as ‘‘set A’’ and ‘‘set B’’.6 This complex person marking system is one of 5. Of which there are hundreds in each language participating in a large number of highly productive morphological processes (See Lois and Vapnarsky 2003a, b for the especially productive case of Yucatec Maya). 6. In some languages, like Tsotsil, the plurality of the person is marked by specific additional su‰xes (see 3.3. below).

18

Colette Grinevald and Marc Peake

the best-known characteristics of the Mayan languages, and will be amply discussed below in section 3. – The rich voice system is a family trait of Mayan languages too, and is by and large marked by su‰xation. The variety of Mayan voices and of their subtypes will be taken up in section 4 below. – Specific transitivity markers that participate in the redundant marking of transitivity of verbal forms are found in some Mayan languages. They consist of a su‰x (often known as a ‘‘theme vowel’’ or ‘‘status vowel’’7) that specifies the syntactic transitivity of the verbal construction, marking it as either transitive or intransitive. This su‰x may directly follow the verb or the voice marker.8 The potential complexity of the Mayan verb form will be illustrated below with verb templates and data for two di¤erent Mayan languages, to show their commonalities (such as distributed transitivity marking) and their specificities (such as the presence or absence of certain verbal categories). 2.2. Verb template and data from Jakaltek Popti’9 Jakaltek Popti’ is a language of the Q’anjob’alan branch of the family, spoken by approximately 88,000 speakers in the area around Jacaltenango in the Cuchumatanes mountains in western Guatemala and some recent settlements in Chiapas, Mexico. The verbal template of Jakaltek Popti’ shown in table 1 below is fairly representative of the verbal morphology of Mayan languages. Its specificity is mostly in its typologically rare and complex system of directionals,

7. Although the expression is somewhat a misnomer to the extent that it can involve a semi-vowel, and thus ‘thematic a‰x’ or ‘thematic extension’ would be a more appropriate label. 8. It may also be su‰xed to totally grammaticalized directionals. Many Mayan languages have directionals, functioning like systems of verbal satellites, from free forms like English verb particles (in Tsotsil for instance) to a‰xed sets (su‰xed in Jakaltek Popti’, prefixed in Mam). 9. Craig’s (1977) work on the language, then called and spelled Jacaltec, was the first extensive syntactic study of a Mayan language, but at a time when the terminology of ergativity and voices as presented here did not exist yet. It also used an orthography superseded now by the o‰cial orthography used in this text.

Ergativity and voice in Mayan languages

19

which is actually insensitive to the transitivity of the construction (Craig 1993). The common thematic vowel is verb final, but it disappears if the verb itself is not clause final. Table 1. Jalkatek Popti’ Maximal Verb Extensions T/A- -set B set A-10 VERB _C _V

-voice

-mood -dir

-mood -theme

irrealis trans. radical pass. irrealis set 1-2-3 intrans. derived antipass. vs. pre-verb aux

Examples (1a), (1a0 ) and (1b) illustrate some of the typical pan Mayan verbal categories listed above, including the use of the final theme vowel to signal transitivity status. (1c), on the other hand, shows one of the language specific traits of the verbal morphology in this language, the category of directionals. (1) [Jakaltek Popti’, Craig (1977, 1993)] a. xk-ach to-yi a0 . xk-ach w-il-a cmp-b2 go-vi cmp-b2 a1-see-vt ‘you went’ ‘I saw you’ b.

xk-ach il-lax-i cmp-b2 see-pass-vi ‘you were seen’

heb’ naj naj c. x-Ø-s-muj-kan-ay-toj11 asp-b3-a3-bury-dir1-dir2-dir3.suff pl cl/man cl/man ‘they buried him 1.once and for all-2.down-3.away’ (1c) also serves to illustrate three specificities of the language: a rigid VSO word order; the existence of a rare type of classifiers, here noun classifiers in their function of anaphoric pronouns; di¤erent sets of su‰xed directionals, here in a rare example of their possible maximal extension of three. 10. The set A person marker is sensitive to the nature of the initial segment of the verb to which it is prefixed. There is one form for consonant initial verbs (-C) and another for vowel initial verbs (-V). 11. The directional -toj of (1c), always verb final, is actually a fusion of -to-oj, a directional resulting from the grammaticalization of the motion verb to ‘to go’, and a subjunctive mood marker -oj.

20

Colette Grinevald and Marc Peake

2.4. Verb template and data from Tojol Ab’al12 Tojol Ab’al is usually classified as a member of the Chujean sub-branch of the Q’anjob’alan branch of the family, most closely related to Chuj.13 It is spoken by approximately 35,000 in the southern part of the state of Chiapas in Mexico. The indications of transitivity are more numerous yet in Tojol Ab’al than in Jakaltek Popti’ and are more widely distributed across the verbal construction. Table 2. Tojol Ab’al Maximal Verb Extensions T/A-

set A-

VERB

-voice

-mood

-theme

-set B

-set A

(sg) _C _V

radical derived

passive antipassive

irrealis

transitive intransitive

(sg/pl)

plural

As shown in the examples of (2) a thematic su‰x is always present and set B is always su‰xed. The various examples show a su‰xed plural marking specifically for set A arguments (otherwise marked with number neutral prefixes) in contrast to set B su‰xes sensitive to number (2b and 2c). The contrast (2b and 2d) confirms the role of the thematic su‰x in marking the verb form as either transitive or intransitive: (2) Tojol Ab’al (Peake (2007) and field notes) a. wa la-waj-y-on icp icp12-go-vi-b1s ‘I go/I am going’ b.

wa x-aw-il-aw-on-ex icp icp-a2-see-vt-b1s-pl ‘You (pl) see me’

12. Tojol Ab’al in two separate words, rather than the more usual Tojol’ab’al is the form preferred today by many native speakers. 13. It is one of the rare languages of the family for which the exact classification is still pending. The Chujean classification is the position of Kaufman and Campbell (1985). Robertson (1980) however has claimed that it is part of the Tseltalan branch and is more closely related to Tseltal and Tsotsil.

Ergativity and voice in Mayan languages

21

c. Ø-jak-tikon och14 b’a chonab’ cmp-come-b1pl.ex dir:hither loc town ‘We came here, towards town’ d.

jel ixta wa x-il-j-y-e’ int toy icp icp-see-pass-vi-b3pl ‘They were badly mistreated’ (lit. ‘They were seen as toys’)

2.5. Conclusions: distributedness of transitivity marking One of the main points of this section was to demonstrate how transitivity is heavily marked in the morphology of the verbal complex of Mayan languages. As expected, the choice of person markers and the presence of voice markers are the essential elements for the determination of the level of transitivity of the construction. However, in this family of languages, transitivity may be signaled by the choice of particular tense/aspect/ mood and voice markers sensitive to the nature of the verb itself (as either transitive or intransitive, whether by root or derivation). And as if to top it o¤, the final thematic su‰x, when it appears, adds to this rather typically Mayan insistence on indicating the level of transitivity of the whole verb complex. Inversely, the clear and detailed morphosyntactic markings of the various voices, through a number of verbal a‰xations and the choices of set A and set B person markers, has certainly been one of the main reasons why data from Mayan languages have found their way to a host of typological discussions dealing with ergativity and voices in recent decades.

3. Terminological and typological approach to Mayan ergativity This section will situate the discussion of the specificity of the ergative marking of Mayan languages in the context of two linguistic traditions. One is the now widespread approach to typologically oriented studies of ergativity that appeals to the distinction of three basic grammatical relations (rather than the traditional two of subject/object). The specific labels to be used here for those three ‘‘primitives’’ are A/P/S15, and correspond 14. Tojol Ab’al has directionals too, but of the independent and uninflected type when compared to the Jakaltek Popti’ ones of example (1c). 15. A/P/S is found in Comrie (1978) as opposed to A/S/O of Dixon (1987, 1994).

22

Colette Grinevald and Marc Peake

to subject of transitive/object of transitive/subject of intransitive, respectively. The second tradition referred to here is that of the terminology used in Mayan linguistics literature, where the person markers involved in the marking of ergativity have been labeled ‘‘set A’’ and ‘‘set B’’. 3.1. ‘‘Set A’’ and ‘‘Set B’’, a Mayan linguistics tradition Before modern typological studies had explored the phenomenon of ergativity, Mayan linguistics had already assigned (neutral) labels of ‘‘Set A’’ and ‘‘Set B’’ to the two sets of person markers found in all Mayan languages, markers that were finally identified as functioning in a so-called ‘‘ergative alignment’’ only later, in the seventies. As recognized then, one of the functions of the Set A markers corresponds to that of the so-called ergative case, while the main function of the Set B corresponds to that of an absolutive case. But the names have stuck, and most linguistic studies of Mayan languages continue to refer to them in this way. Hence in the Mayan examples of this text, the ergative marker is glossed ‘‘A’’ and the absolutive marker ‘‘B’’. The Mayan person markers are worth a few general remarks at this point. First, they are omnipresent throughout Mayan languages because of their multiple functions in verbal, nominal and adpositional phrases, as will be shown below. Second, they are a solid Pan-Mayan characteristic in contemporary languages, although with detailed language specific variation, also to be considered below. Third, they have been reconstructed for Proto-Maya, and the linguistic changes leading to the individual systems of modern languages are known (See in particular Norman and Campbell 1978, and Robertson 1980). The following paragraphs will review the di¤erent forms and functions of Set A and set B. The Set A person markers are prefixes that attach to the verbal complex in all Mayan languages, except Huastec (Zavala 1994), with some languages also having additional plural Set A su‰xes used in conjunction with these prefixes. In addition, again with the exception of Huastec, there are two phonologically determined Set A paradigms: a pre-consonantal and a pre-vocalic one. The Set A markers are further used in several types of phrases. In transitive verbs, they cross-reference the subject (the ‘‘A’’ primitive) and, in those languages that display split-ergativity, they crossreference the subjects of intransitive clauses (the ‘‘S’’ primitive) following a nominative alignment pattern (see section 4 below, and Vapnarsky et al. this volume). Set A markers are also found both in noun phrases as markers of the possessor in possessive constructions, in which case the marker is

Ergativity and voice in Mayan languages

23

prefixed to the possessed noun, and again in prepositional phrases, when prefixed to the so-called ‘‘relational nouns’’ that function as (inflected) prepositions and far outnumber the prepositions of those languages. The Set B person markers, on the other hand, exhibit more variety of a‰xation. They may be prefixed and/or su‰xed, the variation, where it exists, being determined by various factors, such as the presence or absence of an aspect marker. There may be multiple forms of each Set B a‰x as well, depending on where it occurs and on its function, although, if a set B a‰x is prefixed, it does not exhibit distinct pre-consonantal and prevocalic forms as Set A prefixes do. Set B markers may in fact be clitics or free morphemes that occur (semi)independently of the verb form. In terms of their functions, the Set B person markers are equivalent to the absolutive case and cross-reference the subjects of either verbal or non-verbal intransitive predicates. In a typical ergative type of alignment, the same set B is therefore associated to the functions of the ‘‘S’’ primitive (subject of all intransitive predicates) and the ‘‘P’’ primitive (objects of transitive predicates). In addition, the plural marking found in some languages for set A and set B can be of additional interest in the discussion of ergativity, particularly when there are distinct plural forms for the two sets, with the result that this di¤erence can be exploited in syntactic argumentation (see below the case of the applicative voice in Tsotsil, section 4.4.). To illustrate the Pan-Mayan and language particulars of the basic functioning of Set A and Set B in independent clauses, data from three particular languages will be given below.16 3.2. Jakaltek Popti’ – a language with both Set B prefixes and su‰xes There are two Set B forms in the first and second person (the form being a Ø morpheme for the third person across the family) depending on the nature of the predicate. In verbal predicates preceded by aspect markers, set B markers for either subjects of intransitive verbs (‘‘S’’ primitive) or objects of transitive verbs (‘‘P’’ primitive) take the form of an enclitic cliticized to the aspectual marker, with which it forms an independent phonological word (as shown in 3a). However, when the P argument it

16. See England (1983) for an early study of ergativity marking in the Mamean languages.

24

Colette Grinevald and Marc Peake

references is in the third person (i.e. is Ø) the aspectual marker is directly prefixed to the verb stem (as shown in 3b). In non-verbal predicates (not aspectually marked in Mayan languages), set B is a free form immediately following the predicate (as shown in 3c): (3) [Jakaltek Popti’, Craig (1977)] a.

ch-onh way-i icp-b1pl sleep-vi ‘We sleep’

b.

x-Ø-kam no’ cheh cmp-b3-die cl horse ‘The horse died’

c. meb’a honh poor b1pl ‘We are poor’ Set A markers, on the other hand, are always prefixes (in all Mayan languages). They are a‰xed directly to the verb stem when cross-referencing the subject of transitive constructions (the A primitive, as in 3d and 3e): d.

ch-in haw-il-a i cp-b1 a2-see-vt ‘You see me’

e. x-Ø-s-watx’e naj te’ nhah cmp-b3-a3-make cl/he cl house ‘He made the house’ As markers of nominal possession (as shown in 3f ), or of relational nouns (as in 3g–h) they are also directly a‰xed to those elements: f.

haw-atut a2-house ‘Your house/home’

g.

s-wi’ te’ te’ a3-top cl tree ‘at the top of the tree’

25

Ergativity and voice in Mayan languages

h.

ch-onh tzuj-cha haw-u17 icp-b1pl follow-pass a2-by ‘We are followed by you’ (in the passive adversative sense of ‘you are pursuing us’)

3.3. Tsotsil18 – a language with distinctive plural marking Tsotsil is a language of the Tseltalan branch of the family, spoken by approximately 250,000 people in the state of Chiapas. This language will be used to show the existence of two subsets of Set A and Set B a‰xes. For instance, as with most other Mayan languages, Tsotsil has preconsonantal and prevocalic Set A forms, shown in table 3: Table 3. Prevocalic and preconsonantal Set A prefixes in Tsotsil Person

Prevocalic

Preconsonantal

1st 2nd 3rd

kavs-

jay-

In this language, the Set B a‰xes are usually su‰xes, but can also be prefixes, as shown in table 4: Table 4. Prefixed and su‰xed Set B in Tsotsil Person

Prefix

Su‰x

1st 2nd 3rd

iaØ-

-on -a -Ø

17. It is traditional to call the Mayan adpositional elements that take ergative marking ‘‘relational nouns’’ (also called relator nouns), although this is somewhat of a misnomer in this case, since -u is actually a grammaticalization of verbal rather than nominal origin. Relational nouns are a Pan-Mayan trait. 18. Tsotsil was previously written Tzotzil, but has recently changed to conform to the spelling systems of the other Mexican Mayan languages. All information and examples here are from the works of Judith Aissen (1987, 1997, 1999), a specialist of Tsotsil syntax and major contributor to the study of Mayan syntax in general.

26

Colette Grinevald and Marc Peake

Examples in (4) demonstrate this variation between Set B prefixes (4a) and su‰xes (4b, c): (4) [Tsotsil, Aissen (1987)] a. l-i-bat cmp-b1-go ‘I/we went’ b.

tal-em-on come-perf-b1 ‘I have come’

c. ch-a-mil-on icp-a2-kill-b1 ‘You (sg/pl) are going to kill me’ A special feature of Tsotsil is also that all of the so-called singular forms are underspecified for number (except for the 1st and 2nd person Set B su‰xes). Therefore, in the absence of additional explicit plural marking (see 4a and 4c above and examples 5 below), the prefixed person marking (and su‰xed 3d person set B, which is actually Ø) could be interpreted as cross-referencing either singular or plural arguments, depending on the context: (5) [Tsotsil, Aissen (1987)] a. k-il-oj-Ø a1-see-perf-b3 ‘I/we have seen it/them’ b.

s-man-oj-Ø a3-buy-perf-b3 ‘He/she/they have bought it/them’

c. av-ixlel a2-younger.sister ‘Your (sing/pl) younger sister’ To explicitly mark plural the language has various plural person sufixes. In the first person plural there is a double distinction, one between set A and B and the other between inclusive and exclusive shown in table 5 and illustrated in examples (6) below:

Ergativity and voice in Mayan languages

27

Table 5. Set A and B first person plural su‰xes in Tsotsil Person

Set A Plural su‰x

Set B Plural su‰x

1 plural inclusive 1 plural exclusive

-tik -tikotik/-kotik

-otik -otikotik

(6) [Tsotsil, Aissen (1987)] a. k-il-oj-Ø-tik a1-see-perf-b3-a1pl.inc ‘We (inc) have seen it/him/them’ b.

ch-i-tal-otik icp-b1-come-b1pl.inc ‘We (inc) are coming’

c. ch-i-s-mil-otikotik icp-b1-a3-kill-b1pl.excl ‘He is going to kill us (excl)’ For second and third person the explicit plural marking is the general -ik su‰x: (7) [Tsotsil, Aissen (1987)] a. i-s-man-Ø-ik cmp-a3-buy-b3-pl ‘They bought it’ b.

ch-a-bat-ik icp-b2-go-pl ‘You(pl) are going’

This type of plural marking provided one of the arguments used by Aissen (1987) to demonstrate the existence of a ‘benefactive voice’ in the language, and the fact that the benefactive argument had taken on the role of direct object (or P grammatical relation). 3.4. Split ergativity Two major patterns of split ergativity are attested in the Mayan family, one by aspect and the other by complementation. Ergative split by aspect

28

Colette Grinevald and Marc Peake

was originally demonstrated for Yukatek by Bricker (1978), and is amply exemplified in Vapnarsky et al (this volume). Attention will be therefore given here to the phenomenon of split ergativity by complementation, which has been one of the contributions of Mayan linguistics to the typological literature. Split ergativity in complementation was first identified in the literature in (Grinevald) Craig’s study of Jakaltek Popti’ complex sentence structure (1977: 115–116) and is a trait of Q’anjob’alan languages.19 It consists of a nominative-accusative alignment pattern in certain types of aspectless complement clauses, i.e. set A on subject of intransitives (‘‘S’’ primitive), in the context of ergative alignment everywhere elsewhere (which would have meant set B for that argument). This situation is summarized in table 6 below. Table 6. Split ergativity by complementation type in Q’anjob’alan languages

Intransitive Transitive

Main Clause person marking

Primitives

Complement Clause person marking

B V1 B A V1

S PA

. . . [ A V2itr] . . . [B A V2tr]

The examples in (8) show how, in main (non-embedded clauses) Jakaltek Popti’ follows the normal ergative alignment pattern of Mayan languages: (8) [Jakaltek Popti’, Craig (1977)] a. xk-ach to-yi cmp-b2 go-vi ‘You went’ b.

ch-in haw-il-a icp-b1 a2-see-vtr ‘You see me’

19. The argument for considering it is a true case of split ergativity relies on there being no morphological evidence of a nominalisation of any type in this particular context, while the language can display nominalizing morphology elsewhere. More examples of split ergativity by complementation type in Akateko, another language of the Q’anjob’alan branch, can be found in Zavala (1997).

Ergativity and voice in Mayan languages

29

However, as shown in the schema above and illustrated with examples in (9) below, certain types of complement clauses exhibit a pattern of split ergativity, in the context of a lack of aspect marking (shown in 9b and 9d), and the presence of an intransitive type of su‰x -ni 20 for transitive complement clause (in 9b). The ergative split manifests itself by the presence of the set A2/ergative marker used as indexation for the subject of an originally intransitive embedded verb (in 9d), as opposed to the expected set B2/absolutive: (9) [Jakaltek Popti’, Craig (1977)] a. xk-ach hin-kol-o cmp-b2 a1-help-V ‘I helped you’ b.

x-Ø-w-ilwe [hach hin-kol-ni ] cmp-b3-a1-try b2 a1-help-suff.iv ‘I tried to help you’

c. xk-ach kanhalw-i cmp-b2-dance-iv ‘You danced’ d.

x-Ø-w-il [ha-kanhalw-i] cmp-b3-a1-see a2-dance-iv ‘I saw you dance’

3.5. Conclusions on Mayan ergativity Mayan ergativity consists of a typologically relatively rare system of person indexation in the verb, and the interplay of two sets of markers (each with possible subsets). These two sets have been traditionally labeled set A and set B in Mayan linguistics writing, and correspond to ergative and absolutive markers found in the general linguistic literature. Both sets have been reconstructed for Proto-Mayan, attesting to the long history of ergativity

20. This su‰x -ni resembles the antipassive su‰x of main clauses, and is a combination of the reflex of Proto-Mayan *-on antipassive su‰x and intransitive thematic vowel -i. For another use of this su‰x -ni see the discussion of the agent focus voice below in 3.4., and the example (12).

30

Colette Grinevald and Marc Peake

in the family. In addition, several factors converge to facilitate syntactic argumentation about the ergative nature of Mayan grammar: for one, transitivity is marked throughout with multiple a‰xations, not just by these person markers, and, furthermore, the segmentability of the forms and their limited involvement in morphophonological processes make them quite transparent and easy to use for syntactic argumentation.

4. A functional-typological approach to Mayan voices The rich agglutinating verbal morphology of Mayan languages, with its explicit person markers and su‰xes to mark voice and transitivity has made it so that the languages of the family have been the source of much descriptive work on voice systems, in a constant back and forth interaction with general typological discussions ongoing at the time.21 As already mentioned, a broad functional-typological approach will be applied here to the presentation of the morphosyntax of voices in Mayan languages, first in an overview showing the contrastive structural features of the di¤erent voices, then in a rapid illustrated tour of the voices.22 4.1. Overview of the morphosyntax of the di¤erence voices The presentation of the morphosyntactic characteristics of the Mayan voices will be cast in a framework handling the interaction of the following types of elements:23 21. This interaction between Mayanists and general linguists developed first in the exploratory times of the phenomenon of ergativity itself (identified but not named as such until the late seventies in fact), then in a period of discovery of systems of voices such as antipassives and applicatives. More recently it has been related to the themes of the typology of inverse voices and of the phenomenon of obviation. See section 4 below. 22. See Dayley (1981 and 1990) for early informative overviews of ergativity and voices in Mayan languages, and Campbell (2001) for a thorough discussion of ergativity and voices in a specific language, K’iche’, a majority Mayan language of Guatemala, cast in today’s terms of discussion of such topics. 23. The model proposed is an adapted version of Givo´n’s functional-typological framework, itself partially heir to early versions of Perlmutter and Postal’s Relational Grammar (RG). It is not meant to go into an elaborate formalism, and will not address the challenge that ergativity raises in terms of hierarchy of syntactic roles, but aims to o¤er a way of grasping all the di¤erent levels of analysis that must be kept in mind to fully describe voice phenomena in this type of languages.

Ergativity and voice in Mayan languages

31

– A set of semantic roles, roles determined by the nature of the verb itself and unchangeable. For the purpose at hand, the ones considered here are those of agent (Agt), patient (Pat), benefactive (Ben) or instrument (Instr). – The three primitive syntactic relations considered in ergative studies: A (subject of transitive), S (subject of intransitive) or P (object of transitive), to which one should add the function of oblique. – The notion of alignment between these semantic and syntactic roles, considering as prototypical (of basic active voice) the alignments of an agent with the role of subject of transitive verb (A) (or with that of subject of intransitive or (S) role, but less relevant to the present discussion of voices), of a patient with the role of object (P), and of a benefactive with that of oblique (Obl). – The Mayan specific morpho-syntactic encoding of the syntactic roles realized through verbal indexation, indicated with lines and arrows in the schema below. This indexation consists of the assignment of markers of absolutive case (ABS) or ergative case (ERG), those also labelled in Mayan linguistics as set B (B) and set A (A), as established earlier in section 3. In this type of framework (partially reconstituted for the purpose at hand of describing the variety of Mayan voices), one needs to add the notion of hierarchies of semantic, syntactic, as well as pragmatic roles.24 Finally, the last notion needed is that of variations on prototypical alignments, and the appeal to the notions of ‘‘promotion’’ or ‘‘demotion’’ of certain arguments in case of non prototypical alignment between semantic and syntactic roles. For example, instances of ‘‘demotion’’ would be cases of an agent or a patient argument being realized as an oblique (or nothing), while, conversely, an instance of ‘‘promotion’’ would be a semantic benefactive being realized as the object of a transitive verb (P), and not an oblique (Obl). Figure 1 below puts the five major voice types found in the Mayan family in perspective, while focusing on their argument structure. It starts with the schema of an active transitive construction, presented here with a

24. Here again the framework is presented in shortcut version. But it needs to include the notion, introduced by Givo´n (1983, 1986), of two topics in transitive clauses TOP1 (agent) and TOP2 (patient), as demonstrated by elaborate text counts. This is discussed in more detail below, in section 4.

32

Colette Grinevald and Marc Peake

VSO pattern typical of a few Mayan languages (in particular Jakaltek Popti’) but the verb initial status of Mayan languages at large.25 The juxtaposition of the next two de-transitivized constructions, the passive and the antipassive, highlights the motivation and origin of the term ‘‘anti-passive’’. As seen, both constructions are totally parallel structurally, which is why they share the term ‘‘passive’’. On the other hand, they contrast in the semantic role associated with the S function: patient in the passive but still agent in the antipassive, a contrast that is underlined by the term ‘‘anti’’(-passive). The case of the agent focus construction is more complex and varied in its indexing mechanism and its ambiguous transitivity status in some languages, as will be seen below in 3.4. Finally, in the case of the applicative, the structure is always transitive, and, at least in the case of the benefactive voice shown here, the benefactive third argument is in the P role (the cases of instrumental voices are more varied: see section 3.5. below).

Figure 1. Schema of the main voice types in Mayan languages

25. Although the question of word order is a complex one in Mayan languages, they are essentially verb initial languages, so that both core arguments are contiguous, a typological word order feature often noticed in ergative languages. The variation in word order is between rigid VOS vs. VSO, and variable VOS/VSO in some languages (England 1991).

Ergativity and voice in Mayan languages

33

In all cases but the agent focus voice (and some instrumental voices) one of the two core arguments of the corresponding active voice construction has been ‘‘demoted’’, through one of several strategies (deletion, relegation to oblique, or ‘‘incorporation’’). The term ‘‘choˆmeur’’ used in the applicative construction for the patient argument is meant to attract the attention to the special status of this argument, still present in the structure but not in an active object (P) role.26 The above schema only deals with the morphological structure of these voice constructions. A comprehensive study within a functional typological framework also takes into account their actual use, as will be evoked in section 4 below. The di¤erent voices will now be presented in turn, with a discussion of their major characteristics illustrated with examples of selected languages, and wherever relevant mention will be made of some of the history and impact of the first descriptions of these Mayan voices. 4.2. Passive in Mayan languages It had been initially assumed in early studies of ergativity that such languages were not expected to have a passive voice. This emerged from a pseudo logical reasoning that the antipassive voice newly discovered in ergative languages had to be for those languages what passives were for nominative-accusative languages, hence that there was no use for passives in those languages.27 Craig (1977: 77–83) presented in fact not only evidence of a passive voice in Jakaltek Popti’, but of the existence of up to four di¤erent passives. Such a multiplicity of subtypes of passives has since been shown to be quite widespread in the family. The morphological marking of passive voices in Mayan languages varies from simple verbal su‰xation to phonological root processes or periphrastic constructions, with possible sensitivity to the nature of the verb itself, whether radical or derived. Much discussion has involved the treat26. The expression ‘‘choˆmeur’’, invented by David Perlmutter in the early days of Relational Grammar, is borrowed here to attract attention to the special status of the patient argument of the applicative. It is meant to underline how it is there in the clause but has no syntactic status, as evidenced by the absence of indexing for it on the verb, for instance. Within the RG framework, demoted agents of passives (and potentially patients of antipassives) were also deemed to be ‘‘choˆmeurs’’, i.e. arguments that had ‘‘lost their employment’’, which is what the term means in French. 27. This position was reinforced by the absence of passives in Australian languages that were on the other hand some of the languages fueling discussions of ergative and antipassive voice systems at the time (such as Dixon 1972 for Dyirbal).

34

Colette Grinevald and Marc Peake

ment of the agent of passive clauses, and the discourse functions of the di¤erent passives. See Vapnarsky et al (this volume) for a detailed study of the forms and functions of passives in three languages (Yucatec, Ch’orti’ and Tseltal) of di¤erent branches of the family. 4.3. The Mayan antipassive(s) Discussions of the existence of an antipassive voice in Mayan languages date from the mid seventies and are encapsulated in a seminal paper by Smith-Stark (1978). The paper is the result of much discussion among field linguists that exchanged data and analyses over a period of time, as the idea of ‘‘antipassive’’ was introduced to them and each one went about checking it in the language he or she was working on at the time. The core of the paper is based on data collected for the purpose on five Guatemalan languages, Pokomam, K’iche’, Kekchi, Ixil and Jakaltek Popti’, also representing di¤erent branches of the family. In his paper, Smith-Stark tried to make sense of the variety that emerged of the so-called antipassive voices in Mayan languages from new field studies. He proposed for instance to distinguish between a so-called ‘‘absolutive voice’’, an antipassive voice in which the patient argument was either not present at all or expressed in an oblique, and an ‘‘incorporative antipassive’’, in which the patient is lexically expressed but does not fulfill any syntactic role (cf. the discussion of choˆmeur above), while either literally incorporated into the verb form or simply next to the verb. The third type of voice discussed in Smith-Stark’s paper launched the Pan-Mayan discussion of a so-called ‘‘agentive voice’’ which gets its name from the fact that it is specifically associated with operations that a¤ect an A argument (i.e. agentive subject of a transitive verb), such as question, focus or relativisation. This typologically relatively rare construction is the one labeled ‘‘agent focus’’ voice in figure 1 above, and which is considered further in section 3.4. below. Although the phenomenon of antipassive (and agentive voice) is a solid Pan-Mayan characteristic, Smith-Stark demonstrated in his paper how it could be traced back to two reconstructed Proto-Mayan verbal su‰xes, although which reconstructed form is used for which antipassive in today’s languages is very variable. Structurally, Mayan antipassives are easily identified by their verbal markings. This will be exemplified with data from an absolutive antipassive in Tojol Ab’al. To be noted in (10b) are the various markings of detransitivization that accompany the actual antipassive su‰x -wan, such as the set B/absolutive agent indexation, the intransitive theme vowel -i( y), and, in addition in this particular language, the use of a special

Ergativity and voice in Mayan languages

35

form of the incompletive marker restricted to first and second person intransitive constructions, la-. (10) [Tojol Ab’al, Peake field notes] a. wa x-j-pay-aw-a icp icp-a1-call-vt-b2s ‘I am calling you’ b.

wa la-pay-wan-y-on icp icp-call-ap-vi-b1s ‘I am calling’

The following Jakaltek Popti’ examples primarily illustrate the general process of detransitivization of both passive and antipassive voices, evidenced by a‰xation of voice markers and the maintenance of a set B/absolutive indexation for the semantic patient in its S (subject of intransitive) function, in (11b, c and d). The examples can also be taken in pairs, to support the rationale for the name of ‘‘antipassive’’ in (11b and c), and to show the di¤erence between so-called absolutive antipassive (with oblique patient and su‰x -wa) and incorporated antipassive (with non referential patient and su‰x -wi),28 in (11c and d): (11) [Jakaltek Popti’ (Craig 1977)] a. xk-ach s-kol naj cmp-b2 a3-help cl/he ‘He helped you’ b.

xk-ach-kol-lax y-u29 naj cmp-b2-help-pass a3-rn cl/he ‘You were helped by him’

28. One can identify di¤erent final vowels in -wa and -wi for the two subtypes of antipassives. In an interesting way, -a# is more reminiscent of a transitive final vowel and -i# of an intransitive one. Both constructions show mixed signs of intransitivity with set B for S argument but overt patient NP. This patient argument is oblique in one (with maybe transitive like final vowel) and ‘incorporated’ (or choˆmeur) in the other (with intransitive like -i). No special glosses will distinguish the two here. 29. A reminder here that oblique NPs are marked by relational nouns taking an ergative set A marker, as seen in FN 20 above linked to example (3h). This applies to the agentive oblique -u of (11b) and the patientive oblique -inh of the next example (11c).

36

Colette Grinevald and Marc Peake

c. xk-ach-kol-wa y-inh naj cmp-b2-help-ap a3-rn cl/he ‘You helped him’ (‘You gave him a hand’) d. ch-ach kol-wi anma inc-b2 help-ap people ‘You help people’ Examples of antipassives with patients demoted to an oblique function and introduced by a relational noun can be also found in Campbell (2001) for K’iche’, for example, and a prototypical case of incorporation of non referential patient arguments in antipassive can be found in early work on Yukatek by Bricker (1978). 4.4. The agent focus construction The agent focus construction present in most Mayan languages was first considered to just be another type of antipassive voice, because it shares with antipassives the use of one of the two reconstructed antipassive verbal a‰xes. This construction is not strictly speaking an antipassive, however, and it has turned out to take many specific forms across the family. It has been alternatively called the ‘‘agentive’’ voice or the ‘‘agent focus’’ voice due to the fact that, as already mentioned, it is specifically found in cases of operations on the agent of transitive clauses (A). Like the antipassive voice, the verb form itself is morphologically monovalent, with set B indexation, and an antipassive looking su‰xation. However, the construction retains two core arguments (in that the patient does not appear in the oblique), but the two compete for one indexation slot, filled with a set B marker. In some languages (like Jakaltek Popti’, Craig 1977, 1979), this set B marker can only refer to the patient argument, but in other languages it may alternate between agent and patient marking (as in K’iche’, Mondloch 1978, Cambell 2001), hence the double lines and arrows in figure 1 above for this construction. In fact the complexities of the argument marking process of this construction constituted an early challenge for theoreticians, and have been the source of discussions on issues of person hierarchy marking and surface constraints on audible person marking (set B3 as a zero marker vs. audible B1/2) typical of some Mayan languages30. In addition, while the construction is obligatorily used in 30. For a discussion summarizing these issues, see Campbell (2001). For a discussion of the phenomenon cast into a formal syntactic framework, see Stiebels (2006).

Ergativity and voice in Mayan languages

37

some languages such as Jakaltek Popti’, Tz’utujiil, K’iche’ or Ixil), it may be optional (i.e. sensitive to discourse conditions) in others (Akatek or Tsotsil, for example).31 An early analysis of this construction in Jakaltek Popti’ (Craig 1977, 1979), presented it as a disambiguation mechanism in verb initial languages, on the basis of contrasting sentences like those of (12a vs. b and c) below. (12a) shows the two postverbal arguments in the rigid order VSO (or VAP). (12b) shows the change in the verb form if it is the subject of the transitive that is questioned, while (12c) shows no change in the verb if it is the object. Hence the di¤erence in strategies signals which of the two arguments has been questioned: (12) [Jakaltek Popti’ Craig 1979] a. xil naj ix32 x-Ø ( y)-(’)il naj ix cpl-b3 a3-see cl/he cl/her ‘He saw her’ b.

mak x’ilni ix mak x-Ø-’il-ni ___ ix who cmp-b3-see-suff ___ cl/her ‘Who saw her?’

c. mak xil naj mak x-Ø ( y)-(’)il naj ___ who cmp-b3 a3-see cl/he ___ ‘Whom did he see?’ More recent work relying on developments in the typology of inverse voices and of the notion of obviation has taken a new look, two decades later, at this construction. Casting it in a new light, it argues that this agent focus voice is in fact a type of ‘‘inverse’’ voice, fulfilling functions similar to 31. Craig (1979) gave an early analysis arguing for an agent extraction rule, with simple set A/ergative deletion required to disambiguate the remaining post verbal NP, and that there was no agentive ‘antipassive’ in that language, although the construction took on more antipassive-like forms in other languages. 32. Morphophonemics render opaque the person indexing on the verb in this case: there is the combination of a set B zero marker, and the deletion of the root initial glottal of the verb and of the A3 marker ( y), but the contrast is audible in (12b) where the absence of an A3 marker prevents the fall of the verb initial glottal.

38

Colette Grinevald and Marc Peake

those of inverse voices found in the Algonquian languages, for instance (Aissen 1999; Zavala 1997). 4.5. Mayan applicatives The analysis of another voice found in Mayan languages contributed in its time to establish the inventory of a typology of voices found in the languages of the world.33 The work was carried out on an applicative voice in the Tseltalan languages, more specifically the description of the ‘‘benefactive voice’’ of Tsotsil originally done by Aissen (1987) within a relational grammar syntactic framework, followed by descriptions of socalled ‘‘instrumental voices’’ in the K’ichean languages. The benefactive applicative of Tsotsil34 is marked by the verbal su‰xation of a specific voice marker -be, and a non prototypical alignment of the benefactive third argument with the P syntactic function (rather than an oblique one), while the patient argument actually remains in the clause, in a type of ditransitive construction.35 The examples in (13) below show first an active transitive clause (13a), followed by a corresponding applicative construction marked with the applicative voice -be su‰x, in which the third argument is a benefactive. (13) [Tsotsil, Aissen (1987)] a. i-Ø-j-meltsan j-p’ej na cmp-b3-a1-make one-nc house ‘I made a house’ b.

i-Ø-j-meltsan-be j-p’ej na li Xun-e cmp-b3-a1-make-appl one-nc house the Xun-cl ‘I made a house for Xun’

33. The approach taken here is that of the path of discovery of voices in Mayan linguistics, in the early days of the first syntactic descriptions of indigenous languages of America. In this case, the initial analysis of the ‘‘benefactive voice’’ of Tsotsil was cast in the Relational Grammar framework that o¤ered at the time the possibility of a precise descriptive analysis of the construction. This was of course well before any possibility of large typological studies, such as the very recent and major study of ditransitive constructions by Malchukov, Haspelmath and Comrie (2010). Today these constructions of Mayan languages would be reconsidered in terms of derived ditransitive constructions. 34. Benefactive is a cover term for the construction, in which the third argument can be either a benefactive or a malefactive, or an addressee or target. 35. Considered an instance of ‘choˆmeur’ (unemployed) argument in the framework of Relational Grammar (Perlmutter 1983).

Ergativity and voice in Mayan languages

39

The following examples (13c and d) demonstrate that it is the third argment, the benefactive, that is cross-referenced on the verb, either with a set B1 marker (su‰xed to the verb in the absence of an aspect marker) as in (13c), or with both a set B marker and a plural su‰x as in (13d)36. (13e) demonstrates an applicative passive construction, in which the benefactive argument aligns with the S primitive (and is thus marked with a set B su‰x): c. meltsan-b(-o)-n 37 lek i garafon-e fix-appl-imp-b1 good the jug-cl ‘Fix the jugs carefully for me’ d.

ch-a-j-mil-be-ik icp-b2-a1-kill-appl-pl ‘I’ll kill it/them for you(pl)’

e. ’ak’-b-at-Ø jun.syen soltaro li j-chamu’ give-appl-pass-b3 100 soldier det e3-chamulan preserente ’une president cl ‘The Chamulan president was given a hundred soldiers’ The same -be applicative voice marker found in the ‘benefactive’ voice of Tsotsil has also been identified in other languages in constructions that exhibit more or less canonical applicative type structures, all involving, semantically, instrumental arguments. Interestingly, all the cases of ‘instrumental’ voices seem to function and be structured more like the agent focus constructions than applicative ones. Like agent focus constructions, they are found in cases of operations on instruments (with extraction of the instrument to sentence initial position). They are more varied yet than the agent focus constructions, however, in the treatment of the patient argument, which is sometimes clearly demoted to an oblique in an antipassive-like structure (as in K’iche’, see Campbell (2001)), but

36. In (13d) the plural -ik must be interpreted as that of the semantic beneficiary of kill, and not the semantic patient, which is not directly indexed in the construction and underspecified for number. An interpretation in which the second person beneficiary is singular is not possible in this configuration. 37. A morphophonological rule erases the IMP marker.

40

Colette Grinevald and Marc Peake

can still be indexed in the verb with the set B (as in Tz’utujil, Craig (1978) and Dayley 1985).38

5. Conclusion: On the functions of the multiple Mayan voices As the multiple voices and subtypes of voices were first being described in the late seventies by a coordinated network of Mayan language field workers, questions started to be raised about the use and motivation for such a wealth of constructions.39 Craig (1977), for instance, had provided an early account of a textbased study of the various passive forms for Jakaltek Popti’.40 For a much more advanced study of the question, in Yucatec, Tseltal and Ch’orti’, see Vapnarsky et al (this volume). Otherwise, in much the same way that Mayan syntactic studies had developed hand in hand with theoretical advances of the time, a series of discourse studies of Mayan languages came out, in a second wave, that were framed within developments in a more functionally oriented linguistics. This was in particular the case of Mayan studies carried out within the so-called ‘West Coast functionalist’ approach, using quantitative methodologies to study information flow and packaging of information through discourse, as proposed by Chafe (1979) and Givo´n (1979).

38. This has been a very complex area of Mayan syntax, where a number of themes recur in di¤erent combinations: the notion of voice-like constructions linked to operations on specific arguments (instrument vs agent focus construction), the many variations about the fate of the patient argument (either demoted to oblique function or still P and indexed with set B/absolutive on the verb) and the nature of the person indexing on the verb. Taken together, these variations on the theme of the possible levels of transitivity of a clause and possible patterns of person indexation, made Mayan languages a laboratory for the study of the nature of voice systems, and their possible forms and functions. 39. Early studies of the discourse use of voices in Mayan languages relied on some of the new grammatical descriptions. Such was the case for the discourse study of Jakaltek Popti’ by Datz (1980) following the syntactic description of the language by Craig (1977), as well as the discourse study of Tojol Ab’al by Brody (1982) after the grammar of the language by Furbee-Losee (1976). 40. This early analysis focused on the semantics and pragmatics of the agent. It showed that two passives, in -ot and -lax, occur preferentially with no agent expressed, and if expressed, with an agent restricted to the third person. -ot is

Ergativity and voice in Mayan languages

41

One of the better known products of the famous ‘Pear Stories’ project led by Chafe (1980) at the University of Berkeley41 is the seminal paper by du Bois (1987)42. The study is based on a quantitative analysis of data from Sakapultek, a language of the K’ichean branch of the family. The original idea of the paper was to argue for a certain legitimacy of the phenomenon of ergativity, relatively new to linguistics and still considered as a puzzling pattern. The demonstration consisted in showing, quantitatively, the existence of a dominant pattern in discourse of one lexical argument per clause, associated with new information, in a pattern that came to be known as the ‘‘preferred argument structure’’. The fact that this one argument turned out to be either an S or a P argument argued for what the title claimed: a discourse basis for ergativity. The quantitative study showed also another way in which the A argument of a Mayan language was constrained (as in the case already seen of the agent focus construction), this time in what du Bois labeled specifically the ‘‘given A constraint’’, the constraint that the A argument of a clause be associated with given information. As it were, the demonstration of a preferred argument structure in discourse that came out of the study of a (Mayan) ergative language, turned out to have much wider implications, as it was found to operate in non ergative languages as well, as discussed in du Bois’ paper.43 actually preferred in completely impersonal constructions and in the completive aspect, whereas -lax presupposes an impersonal authority or collective agent and is preferred with other aspects. The other two, with -lo and -cha, are much less productive and characteristically take an oblique agent which can be in any of the three persons. The -cha passive, of limited productivity, is a type of adversative passive, with emphasis on an agent in control of the action and obligatorily expressed. 41. This is one of the early linguistics projects based on video stimuli, in this case a short silent film made to elicit a narrative about a boy on a bicycle, a man in a tree, and pears being picked, hence the title ‘‘Pear Stories’’. 42. This study was presented numerous times starting in 1981 to diverse groups of linguists, as mentioned in a striking initial foot note where many illustrious linguists of the time are listed and who all participated in the shaping of this research. Many subsequent studies of other languages refer to it. 43. In the same way discussions of the existence of antipassive voices in ergative languages were later extended to identifying antipassive-like voices in nonergative languages, as argued early by Postal for French (1977) for instance, the construction Givo´n treats as non-canonical voices, lacking all the morphosyntactic devices of the corresponding canonical voice but functioning in discourse-like one. This is discussed further below.

42

Colette Grinevald and Marc Peake

The innovative patterns of text counts proposed by Givo´n for studies of topic continuity in discourse (1983) were also the source of the ‘‘Functional analysis of Akatek voice constructions’’ by Zavala (1997). This study consisting of text counts of referential distance (how far to the left an argument was last mentioned) and topic persistence (how many times in the next few clauses it was maintained in discourse) demonstrated the contrastive discourse functions of passive and antipassive voices linked to topicality. In this model, where the A argument of a transitive verb is taken to bear primary topicality but the P argument is also endowed of topicality, although to a lesser degree (hence the A/TOP1 and P/TOP2 of this model) a passive construction can be shown to signal the discursive promotion to higher topicality (TOP1) of the argument P of a transitive verb, while an antipassive one signals the demotion of the same argument P (to a complete loss of topicality). More recently, Mayan languages have been revisited from the point of view of their grammar being sensitive to a number of semantic and pragmatic hierarchies, beyond the topicality hierarchy just mentioned. New work on Mayan languages has approached them in terms of their exhibitting signs of functionally inverse and obviation systems, with or without specific morphology for them. Working within the typological framework presented by Gildea (1994) for inverse systems (a question of hierarchy of persons) Zavala (1994) reveals, for instance, the presence of a morphological inverse in Huastec (Teneek). Later, following work on the relevance of the concept of obviation in Mayan languages by Aissen (1997), Zavala (2007) argues that the use of the passive voice in Chol and Akatek can be linked to the function of an obviation system, in case the P argument of a transitive verb is higher on a hierarchy of animacy than the A argument.44 For more extensive discussions and illustrations from three Mayan languages of the workings of the ranking of third persons in terms of animacy, definiteness as well as topicality hierarchies, see Vapnarsky et al (this volume). This chapter has meant to demonstrate the centrality of the notion of ergativity in the grammar of Mayan languages. Ergativity is very clearly 44. A remarkable feature of this advanced work on voices in Mayan languages (cast into discussions of inverse and obviation, and their relation to topicalization and use of voices, including the agent focus voice) is that it has relied in great part on detailed studies produced by a new generation of native Mayan linguists (such as the ones working at CIESAS under Roberto Zavala, cf. England (2007), as well as Pascual (2007) and Curiel (2007).

Ergativity and voice in Mayan languages

43

a major Pan-Mayan trait, with relatively minor language specificities, marked in the context of a pervasive distributed marking of transitivity. It also happens that the Mayan patterns of ergatively aligned verbal indexation are of a nature that is typologically relatively rare. As demonstrated, the literature on the morphosyntax of Mayan languages is rich, much of it being concerned with the issue of voice marking. As this paper has tried to show, Mayan languages have amply contributed to the general linguistics discussions of ergativity and voices, partly thanks to its very rich (as well as easily identifiable) verbal morphology. The contributions of Mayan languages to the establishment of a typology of voices, for instance, have ranged from the demonstration that ergative languages may have passives, to the clarification of typological variants of antipassive voices, and the existence of an agentive voice system, voice originating in the need for disambiguation in operations on one of two post verbal core arguments. More recently studies of Mayan languages have also contributed to explore the functional load of the various voices. They have shown the function of passive and antipassive voice as (de-)topicalizing processes, and have identified the relevance of inverse and obviation phenomena in the grammar of languages, even those without specific morphological markings for them. In conclusion of it all, and to never separate advancements in knowledge from the people that produce that knowledge, we draw attention to the dynamics of the network of linguists that have been working on the description of Mayan languages for the last four decades. As already mentioned, a network of field linguists coordinated the descriptions of aspects of Mayan languages during the seventies and eighties, through thematic workshops (the summer ‘‘talleres maya’’) and thematic sessions at the annual meetings of the AAA, later SSILA. Finally, also remarkable for the development of Mayan linguistics, several linguists, principally Nora England, Judith Aissen and Roberto Zavala45, have been working actively to create a new generation of native Mayan linguists, resulting in new indepth studies that come to enrich our understanding of Mayan languages from the inside. This is a rare enough and exemplary development on the American continent to be worth mentioning in closing.

45. From di¤erent academic bases: firstly OKMA in Guatemala, then CIESASSur Este of Mexico, and finally CCILA of the University of Texas (England 2007, Grinevald 2002, 2007, Woodbury and England 2004).

44

Colette Grinevald and Marc Peake

Abbreviations a(123) af ap appl b(123) caus cl cmp det dir dist erg(123) excl exist foc gen icp inc int itr loc neg nc nf nom pass perf pl prog rn suff top tr vi vt

Set A, ergative (1st, 2nd, 3rd person) agent focus antipassive applicative Set B, absolutive causative classifier completive (aspect) determiner directional distal ergative (1st, 2nd, 3rd person) exclusive existential focaliser generic incompletive (aspect) inclusive intensifier intransitive locative (preposition) negative numeral classifier non-finite nominaliser passive perfective plural progressive relational noun su‰x topic transitive intransitive theme vowel transitive theme vowel

Ergativity and voice in Mayan languages

45

References Aissen, Judith 1979 Aissen, Judith 1983

Aissen, Judith 1987 Aissen, Judith 1997 Aissen, Judith 1999 Bricker, Victoria 1978

Brody, Jill 1982

The syntax of causative constructions. New York: Garland Pub. Indirect Object Advancement in Tzotzil. In Studies in Relational Grammar 1, David M. Permutter (ed.), 272–302. Chicago: The University of Chicago Press. Tzotzil Clause Structure. Boston: D. Reidel. On the Syntax of Obviation. Language 73: 705–750. Agent Focus and Inverse in Tzotzil. Language 75: 451–485. R. Antipassive constructions in Yucatec Maya. In Papers in Mayan Linguistics, Nora C. England (ed.), 3–24. Columbia: Museum of Anthropology, Columbia University. Discourse Processes of Highlighting in Tojolabal Maya Morphosyntax. Ph.D. diss., Washington University.

Campbell, Lyle 1978 Quichean prehistory: linguistic contributions. In Papers in Mayan Linguistics, Nora C. England (ed.), 25–54. Columbia: Museum of Anthropology, Columbia University. Campbell, Lyle 2001 Valency Changing Derivations in K’iche’, in Changing Valency: Case Studies in Transitivity, R.M.W. Dixon and A. Aikhenvald (eds.), 236–281. Cambridge: Cambridge University Press. Campbell, Lyle, and Terrence Kaufman 1985 Mayan Linguistics: where are we now? Annual Review of Anthropology 14: 187–198. Chafe, Wallace 1979 The flow of thought and the flow of language. In Discourse and Syntax (Syntax and Semantics 12.), T. Givo´n (ed.), 159–181. New York: Academic Press. Chafe, Wallace (ed.) 1979 The Pear Stories: Cognitive, Cultural, and Linguistic Aspects of Narrative Production. Norwood, NJ: Ablex. Comrie, Bernard 1978 Ergativity. In Syntactic Typology: Studies in the Phenomenology of Language, Winfred P. Lehmann (ed.), 329–294. Hassocks, Sussex: Harvester Press. Comrie, Bernard 1989 Language Universals and Linguistic Typology, 2nd Edition. Chicago: University of Chicago Press.

46

Colette Grinevald and Marc Peake

Craig, Colette Grinevald 1977 The structure of Jacaltec. Austin: University of Texas Press. Craig, Colette Grinevald 1978 On the Promotion of Instrumentals in Mayan Languages. In Papers in Mayan Linguistics, Nora England (ed.), Columbia: Museum of Anthropology, Columbia University. Craig, Colette Grinevald 1979 The Antipassive and Jacaltec. In Papers in Mayan Linguistics Vol. 2, L. Martin (ed.). Lucas Brothers Publishing. Curiel, Alejandro 2007 Estructura de la Informacio´n, Enclı´ticos y Configuracio´n Sinta´ctica en Tojol Ab’al. M.A. Thesis, Centro de Investigaciones y Estudios Superiores en Antropologı´a Social, Mexico. Datz, Margaret J. 1980 Jacaltec syntactic structure and the demands of discourse. PhD dissertation, University of Colorado. Dayley, Jon P. 1981 Voice and ergativity in Mayan languages. Journal of Mayan Linguistics 2: 3–82. Dayley, Jon P. 1985 Tzutujil Grammar. Berkeley: University of California Press. Dayley, Jon P. 1990 Voz y ergatividad en idiomas mayas. In Lecturas sobre la lingu¨´ıstica maya, Nora England and S. Elliott (ed.), 335–98. Antigua, Guatemala: Centro de Investigaciones Regionales de Mesoamerica. Dixon, Robert M. W. 1972 The Dyirbal language of North Queensland. Cambridge Studies in Linguistics, 9. Cambridge: Cambridge University Press. Dixon, Robert M. W. 1987 Studies in Ergativity. Amsterdam: New York. Dixon, Robert M. W. 1994 Ergativity. Cambridge: Cambridge University Press. du Bois, John W. 1987 The Discourse Basis of Ergativity. Language 63: 805–855. England, Nora C. (ed.) 1978 Papers in Mayan linguistics. Columbia: Museum of Anthropology, Columbia University. England, Nora C. 1983 Ergativity in Mamean (Mayan) Languages. International Journal of American Linguistics 49: 1–19. England, Nora C. 1991 Changes in basic word order in Mayan languages. International Journal of American Linguistics 57: 446–486. England, Nora C. 1992 ‘‘Endangered Languages: Doing Mayan Linguistics in Guatemala.’’ Language 68: 29–35.

Ergativity and voice in Mayan languages

47

England, Nora C. 1996 Introduccio´n a la lingu¨´ıstica: idiomas mayas. Antigua: Proyecto Lingu¨´ıstico Francisco Marroquı´n-Editorial Kamar. England, Nora C. 2007 The influence of Mayan-speaking linguists on the state of Mayan Linguistics. In Endangered languages, P. Austin and A. Simpson (eds.), 92–111. Hamburg: Helmut Buske Verlag. Furbee-Losee, Louanna 1976 The Correct Language, Tojolabal: a Grammar with Ethnographic Notes. New York: Garland Publishing. Gildea, Spike 1994 Semantic and pragmatic inverse: inverse alignment and inverse voice in Carib of Surinam. In Voice and Inversion, T. Givo´n (ed.), 187–230. Amsterdam: John Benjamins. Givo´n, T. (ed.) 1983 Topic Continuity in Discourse: A quantitative cross-language study (Typological Studies in Language 3). Amsterdam: John Benjamins. Givo´n, T. 1979 On Understanding Grammar. New York: Academic Press. Givo´n, T. 1986 Syntax: A functional-typological introduction. Volume 1. Amsterdam: John Benjamins. Givo´n, T. 1994 The pragmatics of de-transitive voice: functional and typological aspects of inversion, in Voice and Inversion, T. Givo´n (ed.), 3– 44. Amsterdam: John Benjamins. Givo´n, T. 2001 Syntax: An introduction. Amsterdam: John Benjamins. Grinevald, Colette 2000 A morpho-syntactic typology of classifiers, in Systems of Nominal Classification G. Senft (ed.), 50–92. Cambridge University Press. Grinevald, Colette 2002 Linguistique et langues mayas du Guatemala. Faits de Langues, Meso-Amerique, Caraibes, Amazonie, Orphys 1: 17–27. Grinevald, Colette 2007 Endangered Languages of Mexico and Central America, in Language Diversity Endangered, M. Brenzinger (ed.). Berlin: Mouton de Gruyter. Instituto Indigenista Nacional 1988 Lenguas Mayas de Guatemala: documento de referencia para la pronunciacio´n de los nuevos alfabetos oficiales. Guatemala: Instituto Indigenista Nacional, Ministerio de Cultura y Deportes. Instituto Nacional de Lenguas Indı´genas 2008 Cata´logo de las Lenguas Indı´genas Nacionales. Mexico City: Secretarı´a de Educacio´n Pu´blica.

48

Colette Grinevald and Marc Peake

Kaufman, Terrence 1971 Tzeltal phonology and morphology. Berkeley: University of California Press. Kaufman, Terrence 1972 El proto-tzeltal-tzotzil: fonologı´a comparada y diccionario reconstruido. Mexico City: UNAM, Coordinacı´on de Humanidades. Kaufman, Terrence 1974 Idiomas de Mesoame´rica. Guatemala: Editorial Jose´ de Pineda Ibarra, Ministerio de Educacio´n. Kaufman, Terrence 1986 Outline of Comparative Mayan Grammar I: Morphology and Particles. Unpublished manuscript. Lois, Ximena and Valentina Vapnarsky 2003a Polyvalence of root classes in Yukatekan Mayan languages. Munich: Lincom Europa. Lois, Ximena and Valentina Vapnarsky 2003b Les raciness verbo-nominales du maya yucatec. Faits de Langues 21. Malchukov, Andrej, Martin Haspelmath and Bernard Comrie 2010 Ditransitive constructions: a typological overview. In Studies in Ditransitive constructions: a Comparative Handbook, Malchukov, A. M. Haspelmath and B. Comrie (eds.) Berlin: Mouton de Gruyter. Norman, William M. and Lyle Campbell 1978 Towards a Proto-Mayan syntax: a comparative perspective on grammar. In Papers in MayanLlinguistics, Nora C. England (ed.), 136–56. Columbia: University of Missouri. Pascual, Ada´n 2007 Transitividad y dependencia sinta´ctica y discursiva en Q’anjob’al. MA thesis, CIESAS, Mexico, D.F. Peake, Marc 2007 Approche sociolinguistique et linguistique du Tojol Ab’al, langue maya du Chiapas. M.A. Thesis, University of Lyon-2. Perlmutter, David (ed.) 1983 Studies in Relational Grammar. Chicago University Press. Postal, Paul 1977 Antipassive in French. Lingvisticæ Investigationes 1:2: 333–374. Robertson, John S. 1980 The Structure of Pronoun Incorporation in the Mayan Verbal Complex. New York: Garland. Smith-Stark, Thomas C. 1978 Mayan Antipassive: some Facts and Fictions, in Papers in Mayan Linguistics, Nora England (ed.), 169–187. Columbia: Museum of Anthropology, Columbia University.

Ergativity and voice in Mayan languages

49

Stiebels, Barbara 2006 Agent Focus in Mayan languages. Natural Language and Linguistic Theory 24: 501–70. Woodbury, Anthony and Nora C. England 2004 Training speakers of indigenous languages of Latin America at a US University. In Language Documentation and Description Vol. 2, Peter Austin (ed.), 122–139. London: Hans Rausing Endangered Languages Project. Zavala, Roberto 1994 Inverse Alignment in Huastec. Funcio´n 15–16: 27–81. Zavala, Roberto 1997 Functional Analysis of Akatek Voice Constructions. International Journal of American Linguistics 63: 439–74. Zavala, Roberto 2006 Inversion and obviation in Mesoamerica, in Linguistische Berichte Sonderheft 14, A. Simpson and P. K. Austin (eds.). Hamburg: Helmut Buske Verlag.

Ergativity and the passive in three Mayan languages Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin 1. Introduction1 This paper o¤ers a comparative analysis of the passive voice in three languages – Yucatec, Ch’orti’ and Tseltal – belonging to di¤erent Mayan branches. Our analysis takes a usage-based approach building on previous studies of the passive in Mayan languages, and shows that despite important di¤erences between the ergative patterns they display, the passive voice fulfils fairly similar functions across the family. These are minimally syntactic and mostly semantic and discourse-pragmatic. Ergative languages express in a specific way the relationship between the agent and the other participants in a multi-participant action, and one might expect this to influence the configuration of the voice system. It has been argued that in ergative languages the argument corresponding to the agent of transitive predicates is structurally – perhaps even conceptually – secondary with respect to the patient and to the argument of an intransitive predicate. Thus, in a mirror-image reversal of what is typically found in accusative languages, the antipassive with its object demotion is an expected phenomenon in ergative languages, whereas from a syntactic point of view, the passive would not be, since it demotes an already secondary argument (Jacobsen 1985, Shibatani 1985 among others). In support of this view, it is known that passive and antipassive may serve to feed the pivot of constructions requiring accusative or ergative alignment respectively (Dixon 1994: 17, 152 ¤.). However, voices have various other functions, and ergative languages show contrasting patterns as regards valency change. While some have no passive voice, others, such as the Mayan languages, have developed a very rich array of passive and middle constructions expressing di¤erent types and degrees of relationship between

1. Our thanks go to an anonymous reader and the editors of the volume for their helpful comments on previous drafts of this paper, as well as to the revisor of the English text.

52

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

the agent and the action expressed, in connection with semantic hierarchies and discourse functions. Similarly the antipassive constructions of Mayan languages are not only motivated by syntactic concerns (England 1988: 539, Zavala 1997). As will be shown in the course of this paper, Mayan languages possess a variety of passive forms, which may be synthetic or analytic, and may signal valency change by means of vowel alternations, a‰xation, or periphrases involving auxiliary or light verbs. The distribution of these constructions and markers depends on various considerations: the nature of the stem (root vs. derived), syntactic rules of morphosyntactic alignment, distinctions concerning the degree of demotion of the agent or promotion of the patient, and semantic features such as the position of the participants in the animacy hierarchy. The multiple parameters configuring passive voices in Mayan languages will be further analysed by a study of the syntactic, semantic, and discourse-pragmatic factors which favour use of the passive, as well as how these factors interact with each other. It will be shown that syntactic constraints have only limited influence, while the other parameters play a major role, in line with Shibatani’s (1985) notion of ‘‘agent defocusing’’ and Fried’s (2006) proposal on ‘‘agent backgrounding’’. The three Mayan languages studied here are of special interest since they show contrastive ergative patterns in the context of this linguistic family. Ergativity is mainly a morphological phenomenon in all Mayan languages, but cross-reference systems, split patterns and syntactic properties linked to ergativity vary and are still much discussed. Tseltal is morphologically a fully ergative language (Shklovsky 2005, Polian 2006). In contrast, Yucatec and Ch’orti’ show split ergativity with intransitives, mainly triggered by aspect and mood (Bricker 1981, Quizar 1994). Ch’orti’ is unique in the Mayan family in having developed a third paradigm of personal markers, specialized for the single argument of intransitives in those contexts where the split occurs. Another relevant property of these languages is that they do not have the ‘‘focus antipassive’’ construction, which is used by other Mayan languages such as Mam (England 1988) or Tzotzil (Aissen 1999) when a nominal clause is questioned, focused or relativized, and which can be taken as one piece of evidence for syntactic ergativity. In section 2, we present some relevant general features of the three languages, in particular their di¤erent patterns of ergativity and restrictions on syntactic alignment. In section 3 their voice systems will be briefly described, with special attention given to passive forms, argument encoding, applicatives, and the di¤erent means available for the expression of

Ergativity and the passive in three Mayan languages

53

the agent. Section 4 deals with the various constraints and other motivations leading to passive use in the languages under consideration; we will compare syntactic factors, semantic hierarchies and discourse prominence, genericity, and topicality. Section 5 will o¤er concluding remarks. A brief excerpt from a Yucatec narrative is appended as an illustration of the points made in section 4.4.1.

2. General observations on the languages under study 2.1. Speakers and Languages The three languages analysed in the present paper, Tseltal, Ch’orti’ and Yucatec, belong to three di¤erent branches of the Mayan family, known as Tseltalan, Cholan and Yucatecan. The Tseltalan and Cholan branches are grouped together as ‘‘Great Tseltalan’’.2 Tseltal, Ch’orti’ and Yucatec are spoken in geographically separated areas, except for some sporadic contacts in big cities in Mexico and elsewhere in North America due to recent migrations. Tseltal is mainly spoken in the west of the Maya area, in the Chiapas Highlands, Mexico, by about 400,000 speakers. As many descriptions attest, Tseltal is highly multidialectal.3 It has also become a vehicular language in recently settled regions (Selva Lacandona and the southern part of Campeche). Ch’orti’ is spoken in the far south-east of the Maya area in Guatemala, close to Honduras, with speaker numbers estimated at between 11,000 and 30,000.4 Yucatec is spoken in the northern lowlands of the Maya area, throughout the Peninsula of Yucata´n, by about 2. It may be noticed, however, that some Cholan languages, especially Chol and Yokot’an (also called Tabasco Chontal), have properties closer to those of the Yucatecan languages. 3. See for the Bachajo´n variant: Slocum, Gerdel and Aguilar 1999, Monod Becquelin 1997; for Tenejapa: Kaufman 1971, Brown 1997, 1998 (among others); for Oxchuc: Polian 1999, 2006, to appear; for Petalcingo: Shklovsky 2005; Dayley 1981, 1990, Hinma´n Smith 2004 and Robinson 1999 for other languages of the Great Tseltalan branch. Di¤erent orthographies are used (Tzeltal/Tseltal) by di¤erent authors, as there is no specific recommendation in Mexico. 4. The SIL website mentions 30,000 speakers, Fought (1967) has 20,000, and the last census of Guatemala gives a figure of 11,734 (XI Censo Nacional de Poblacio´n y VI de Habitacio´n, Instituto Nacional de Estadı´stica de Guatemala); the last of these seems to come closest to reflecting the present linguistic situation.

54

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

800,000 speakers. Compared to Highland Mayan languages, Yucatec shows little dialectal variation, although insu‰cient research has been carried out on the subject as yet (but see Pfeiler and Hofling 2006). 2.2. Ergativity Mayan languages are head-marking: core arguments must be indexed by personal markers on the verb, whereas lexical arguments are not obligatory and are often omitted. Ergativity mainly a¤ects the encoding of arguments with regard to verbal cross-reference marking. Nominals make use of the same two sets of person markers as verbs, one to index possessors and the other to index the argument of stative predicates. Like most parts of speech, nominals can be used as predicates without the need for a copula (for a general presentation of these features see Grinevald and Peake, this volume). While most languages of the family show an ergativity split triggered either by aspect and/or mood, person, subordination, or negation, Tseltal and Tzotzil (Haviland 1981) are the only languages of the family with no split. Thus in Tseltal, single arguments of intransitive verbs or stative predicates are consistently encoded with the same personal markers as objects in transitive predicates, i.e. with ‘‘set B’’, the Absolutive paradigm (ABS), see (1). Unlike in Mayan languages which possess the ‘focus antipassive’ construction (see below), no distinctive treatment is found for arguments in focalizations, content questions or relative clauses. These constructions show no syntactic or semantic pivot, although subordination still needs to be analysed in more detail. (1) a. (tse) b.

Ya x way-at icp icp.intr sleep-2b ‘You sleep’ Ya k-il-at icp 1a-see-2b ‘I see you’

c. Winik-at man-2b ‘You are a man’ In contrast, Yucatec has an ergative split mainly based on aspect and mood. In the completive aspect and in the subjunctive mood, single arguments of intransitives are encoded with set B, the Absolutive paradigm,

Ergativity and the passive in three Mayan languages

55

i.e. like objects, as shown in (2a) and (2c) with the completive aspect. In the incompletive they are encoded with ‘‘set A’’, the Ergative paradigm (ERG), like agents of transitive predicates, as shown in (2b) and (2d). Unlike Tseltal and Ch’orti’ but like other Mayan languages such as K’iche’, Mam or Chol (Smith Stark 1978, Zavala Maldonado 2003, 2007), Yucatec has two clear classes of intransitives, agentive and nonagentive, the di¤erences being reflected in aspectual and transitivization morphology.5 However, in Yucatec these two intransitive classes behave identically with respect to the ergativity split, as can be seen in (2).6 (2) (yuc)

Non-agentive intransitive Agentive intransitive a. H lu´ub-en c. H meyah-n-ah-en cp.intr fall-1b cp.intr work-ap-intr-1b ‘I fell.’ ‘I worked.’ b. K-in-lu´ub-ul icp-1a-fall-icp/nom ‘I fall.’

d.

K-in-meyah icp-1a-work ‘I work.’

The syntactic correlates of split ergativity in Yucatec have been analysed more than once. Whereas Lehmann (1990) argues that Yucatec could just as easily be considered basically accusative, Kra¨mer and Wunderlich (1999), on the basis of a markedness and semantic feature analysis, consider it an ergative language with no ergative-accusative split. Other authors, such as Bohnemeyer (2004, 2007) and Verhoeven (2007), have brought into question the notions of subject, syntactic pivot and alignment in Yucatec. Bohnemeyer considers the traditional notions of subject and object to be inadequate for this language. He follows previous analyses of other Mayan languages, Jakaltek-Popti’ and Tz’utujil, for which Van Valin (1981) and Van Valin and LaPolla (1997: 282–284) argue that the notion of ‘‘subject is not a meaningful concept’’, on the basis of the variable 5. These classes have been characterized as ‘‘active’’ vs. ‘‘inactive’’ (Lehmann 1993, Bohnemeyer 2001); ‘‘agent-salient’’ vs. ‘‘patient-salient’’ (Lucy 1994); ‘‘agentive’’ vs. ‘‘non-agentive’’ (Gutie´rrez Sa´nchez and Zavala 2005, Zavala Maldonado 2007). Kra¨mer and Wunderlich (1999) argue that inherent aspect rather than control is the relevant feature that distinguishes Yucatec intransitive classes. 6. This is not true of all Mayan languages; see for instance Danziger (1996) for a di¤erent pattern in Mopan. Mopan has an intransitivity split whereby the argument of agent-salient intransitive forms is always expressed by set A/Ergative personal markers.

56

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

syntactic pivots found in these languages. Verhoeven (ibid: 143–144) shows that Yucatec also presents di¤erent patterns of neutralization depending on the construction involved. But she argues for a ‘‘weakly implemented (accusatively aligned) subject’’ due to (i) the existence of a few constructions which neutralize the contrast between the single argument of intransitives (including passives) and the agent of transitives and (ii) the absence of a construction neutralizing the argument of intransitives (including antipassives) and the object of transitives. We will see later that an analysis of passive use and its syntactic motivation in Yucatec provides further evidence in favour of the view that there is no clear dominant alignment. In spite of its distinctiveness, there have been very few descriptions of Ch’orti’.7 Diachronically, Ch’orti’ may have shown a type of split intransitivity similar to that seen in Yucatec. However, this language innovated by developing a new paradigm of personal markers, used for the indexation of the single argument of intransitive verbs in the incompletive. The form and position of the new personal paradigm (prefixal like set A rather than su‰xal like set B) clearly show that this paradigm evolved, at least partially, from the ergative markers. The result today is a three-paradigm system, where the single argument is encoded by Set B (ABS1) in the completive aspect (3b), like the object of a transitive verb (3a), but by a specific and exclusive ‘‘Set C’’ (ABS2) in the incompletive (3c).8 (3) (chr)

a.

A-kuch-iy-en 2a-carry-tr-1b ‘You carry me.’

b. Tar-iy-et come-intr-2b ‘You came.’ c. I-tar-i 2c-come-intr ‘You come.’ Table 1 o¤ers a synthetic comparative view of the cross-reference patterns in the languages under study. 7. Fought (1967, 1972), Quizar (1994), Alvarez Ramı´rez (2004), Pe´rez Martı´nez (1994), Quizar and Knowles-Berry (1988). 8. The similarity between the transitive and intransitive status su‰xes in examples (3a) and (3b) results from diachronic phonological changes.

Ergativity and the passive in three Mayan languages

57

Table 1. Types and positions of personal markers on verbal predicates in Tseltal, Yucatec and Ch’orti’ Language

before base

Tseltal Yucatec Ch’orti’

A (ERG) C (ABS2)

after base B (ABS)

Tseltal Yucatec Ch’orti’ Tseltal Yucatec Ch’orti’

base

A (ERG) A (ERG) A (ERG)

intransitive incompletive intransitive completive & subjunctive

B (ABS) B (ABS) B (ABS)

transitive

B (ABS) B (ABS) B (ABS1)

Many Mayan languages present an agent focus construction whereby a semantically transitive action is expressed with only one argument indexed on the verb (see Grinevald and Peake, this volume). Depending on the language, the argument indexed may be the patient or the agent, the latter often occurring along with an antipassive marker on the verb. Ch’orti, Tseltal and Yucatec have productive antipassive forms, but these do not fulfil the agent focus function. Nevertheless, in Yucatec, a cleft construction used only for agent focus (Bricker 1979) shows syntactic properties reminiscent of the focus antipassive found elsewhere. Yucatec agent focus is illustrated in (4b). Although the verb refers to a semantically transitive action and takes an active inflection, only the patient is cross-referenced on the verb, contrary to the general rule. The agent is fronted without cross-referencing. As in the focus antipassive, the construction retains two core arguments, only one of which is indexed. The verb appears without preverbal TAM and, in the completive, it bears the subjunctive instead of the completive transitive su‰x.9 Compare this to the non-focus construction in (4a). 9. In the incompletive, the verb bears the same transitive su‰x (-ik) as in nonfocus constructions. It might be noticed that both this su‰x and the transitive subjunctive su‰x -eh are involved in dependent clauses, unlike the completive marker -ah. It is also worth underlining that Yucatec agent focus requires the transitive status su‰xes, in contrast with languages with antipassive focus constructions, but also in contrast with Akatek where the non-antipassive agent focus construction retains the Patient as the indexed argument as in Yucatec, but requires the intransitive status su‰x (Zavala 1997: 452–453). Recent works on Yucatec agent focus include Gutie´rrez-Bravo and Monforte 2009 and Tonhauser to appear.

58

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

(4) (yuc)

a.

T-uy-il-ah-Ø cp.tr-3a-see-tr.cp-3b ‘She saw it.’

b. Le’ti’ il-eh-Ø 3pr see-tr.sbj-3b ‘She was the one who saw it.’ There is no equivalent construction for the other arguments, as shown with wh-fronting constructions which follow the same pattern. Compare patient fronting (5a) and agent fronting (5b).10 (5) (yuc)

(yuc)

a.

Ma´ax t-uy-il-ah-Ø? Who cp.tr-3a-see-tr.cp-3b ‘Whom did she see?’

b. Ma´ax il-eh-Ø? Who see-tr.sbj-3b ‘Who saw it/him/her?’

By contrast with Yucatec and antipassive focus languages, in Tseltal (6a) and Ch’orti’ (6b), agent and patient content questions, relativization, and focalization do not di¤er from other clause types in their treatment of core arguments (but see section 4.2). (6) a. (tse)

(chr)

Mach’a la y-il-Ø? Who cp.tr 3a-see-3b ‘Whom did she/he see? or Who saw her/him?’

b. Chi uw-ir-a-Ø? Who 3a-see-tr-3b ‘Whom did she/he see? or Who saw her/him?’

10. The agent focus construction in Yucatec holds for all persons (Ma´ax il-ech? ‘Who saw you?’). As Verhoeven notices (2007: 142), disambiguation of the agent and patient is carried out by passivization in some rare cases of focus where the agent focus construction does not apply (e.g. with terminative and progressive aspect). This construction retains P as a core argument.

Ergativity and the passive in three Mayan languages

59

3. Passive constructions 3.1. Multiple voices Transitivity is a key feature in all Mayan languages, with important reflexes in phonology, morphology and syntax. Accordingly, change of valency is a crucially rich domain of the grammar. There are many transitivizing derivations (causatives and applicatives) and even more intransitivization processes. Among the latter, we find di¤erent types of passives, antipassives and middles, whose exact number and properties depend on the language in question (see Grinevald and Peake, this volume, and references therein). Thus, for instance, the middle voice is no longer productive in Tseltal whereas Ch’orti’ and Yucatec have several middle forms which di¤er semantically and functionally. Tseltal, Ch’orti’ and Yucatec also have di¤erent antipassive forms depending on distinct criteria (type of base, aspect, incorporation, and agency properties). Voice is expressed in verbal, participial, and also nominal-like forms. These constructions involve di¤erent morphophonological realizations of the stem as well as potential changes in the encoding of the core arguments on the predicate. We will focus here on productive (non-lexicalized) passive alternations. In the following sub-sections, we present their formation in the three languages at issue. 3.2. Tseltal morphological and periphrastic passives Tseltal valency changes are mainly realized by su‰xation and, in the case of passives, also by complex predicates with light verbs. The morphological passive requires the passive su‰x -ot, available in both completive and incompletive aspects (as in (7) and (8) respectively), the patient being cross-referenced by set B (ABS) markers.11 (7) Mam, ahtay-ot-otik-ix, na’-ot-otik-ix, (tse) Grandfather count-pas-1b.pl-already, know-pas-1b.pl-already huk-pik-otik 7x8000-1b.pl ‘Grandfather, we have already been counted and identified and there are 56,000 of us.’ 11. In the transcription of the following examples, the non-overt third person set B marker (Ø) is omitted unless it is implicated in the constructions where coreference is analysed.

60

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

(8) Ya x tsak-b-ot-Ø ha’ te s-flecha (tse) icp icp.intr take-appl-pas-3b dem deic 3a-arrow ‘Their [the enemies’] arrows are taken from them.’ Periphrastic passives can be formed with two kinds of light verbs, transitive and intransitive; they all require an animate subject referring to the patient but show important di¤erences in argument marking, as we will see in section 3.6. The verb ich’ ‘‘take, receive’’ is used for a passive with neutral meaning as seen in (9); the verb a’iy ‘‘feel, hear’’ creates a passive which may be restricted to adversative contexts in some dialects, as illustrated in (10) (Polian 2006: 116–7), due to the experiential value of the light verb. These two passives take the patient-oriented nominalized form as their object argument. The intransitive verb la( j) ‘‘end’’ can also be used as a light verb in a periphrastic passive construction, but it requires the preposition ta before the Verb-el form, as in (11): see 3.5.12 (9) Te mukenal-e la y-ich’ pas-el ta y-ahlanil (tse) deic grave-td cp.tr 3a-receive make-nr.pas prep 3a-down te banti la y-ich’ jijpan-el-e deic where cp.tr 3a-receive hang-nr.pas-td ‘The grave, it was made at the foot of where he was hanged.’ (txt1) (10) Ya k-a’iy uts’in-el (tse) icp 1a-hear bother-nr.pas ‘I am bothered.’ (11) La y-il laj-em-ix ta bul-el ta chambalam (tse) cp 3a-see end-pf-already prep destroy-nr prep animal ha’ te s-k’al dem deic 3a-field ‘He saw that his milpa had already been destroyed by animals.’ In Tseltal, the morphological passive and the periphrastic passive constructions are not in strict complementary distribution, although they tend to be associated with specific functions (see obviation in 4.2) and bear different semantic nuances, representing ‘‘neutral’’ and ‘‘adversative’’ passives (ich’ and a’iy/laj respectively).

12. Notice that the verb a’iy is also used to form the antipassive voice (Lois et al. 2010) and that laj is also the completive marker of transitive verbs.

Ergativity and the passive in three Mayan languages

61

3.3. A‰xation and vowel alternation in Yucatec and Ch’orti’ passives In Yucatec and Ch’orti’, as in some other Mayan languages, voice is expressed in di¤erent ways depending on the nature of the base, i.e. whether or not it is a canonical CVC bare root, and of which type.13 In most Mayan languages showing such distinctions, the passive is formed on canonical roots by means of infixation of -h- or alternation of the nucleus vowel, whereas other bases require su‰xation. As shown in table 2, this is the case in Ch’orti’, where passives of canonical transitive roots are formed by -h- infixation. These stems take the thematic su‰x -a used for all derived intransitives, see (12).14 For non-canonical transitive bases, passive voice is derived by adding the su‰x -n, which again is followed by the status su‰x -a, as in (13). Table 2. Ch’orti’ passive forms active voice

completive aspect

passive voice Canonical transitive

Non-canonical transitive

CVhC-a-b

Vb-n-a-b

c-CVhC-a

c-Vb-n-a

a-Vb-tr-b incompletive aspect

(12) (chr)

A-hihx-a u-’ut e ixim 3c-shell.pas-intr 3a-grain det maize ‘Maize grains are shelled.’

(13) (chr)

E pahbur.sa’ a-che’-e-n-a det shuco.atole 3c-do-tr-pas-intr ‘The shuco atole (a drink) is done.’

13. In Ch’orti’, a canonical transitive verb is defined as an underived verb root with a CVC phonological template; these verbs require the verbal status su‰x -i (or -e for CeC roots) in the active voice. In this language, non-canonical transitive verbs include verbs with derivations (causative, factitive, etc.), nonCVC underived verbs, and also a few CVC verbs requiring a status su‰x other than -i or -e in the active voice. In Yucatec the distinction is between a significant class of CVC roots with transitive/intransitive alternations based on changes to the nucleus vowel, and a‰xally derived transitives or roots with special phonological features. 14. A crasis occurs when the status su‰x is followed by a personal marker. It results in the reduplication of the vowel of the personal marker; the status su‰x vowel is elided while the vowel of the personal marker is reduplicated.

62

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

Yucatec passives, as well as antipassives and middles of canonical multivalent roots, are formed by vowel alternation involving modification of the length, tone and/or rearticulation of the nucleus vowel of the basic CVC root template (Lois and Vapnarsky 2003, 2006). This system makes use of all four possible vocalic values (short, rearticulated, high-toned and low-toned long vowel) present in the language, as shown in table 3. Example (14) illustrates a passive form with a rearticulated vowel, to be contrasted with the active form with a short vowel, exemplified in (63) below. Table 3. Vowel alternation in Yucatec voice (for canonical non-intransitive roots; examples are in the incompletive) Voice category

Vocalic template

example

active

CvC

k-u-huch’-ik

‘She grinds it’

canonical passive

Cv’vC

k-u-hu’uch’-ul

‘It is (being) ground’

middle

Cv´vC

k-u-hu´uch’-ul

‘It gets ground’

antipassive

Cv`vC

k-u-hu`uch’

‘She grinds’

(14) (yuc)

Mix t-u-han-t-ah-Ø bin-e’ neg.emph cp.tr-3a-eat-trzer-tr.cp-3b rs-td ka h ya’ach’-Ø men le wa`ah-o’! conj cp.intr crush.pas-3b cause det corn.bread-td2 ‘He didn’t even eat them, and he was crushed by the [loaves of ] bread [falling on him]’

In contrast, for derived transitives and non-canonical roots, the change involves a‰xation and passive forms require a special su‰x, which varies for aspect/mood status: icp. -a’al, cp. -a’ab, sbj. -a’ak, (15). For derived transitives, the passive su‰x directly follows the transitivizer, as in (15) with kı´in-s formed from kı´im ‘‘die’’ þ causative -s, or in (16) with me`en ‘‘do’’ þ extraversive transitivizer -t. The various passive su‰xes can be analysed as -a’ þ non-agentive inflectional markers (Lois and Vapnarsky 2003).15

15. See also McQuown (1967), Bricker (1978), Ola and Bricker (2000) and Lois and Vapnarsky (2003) for proposals on the origin of the derivational passive form.

Ergativity and the passive in three Mayan languages

(15) (yuc)

a. K-u-kı´in-s-a’al b. Kı´in-s-a’ab-ih c. Kı´in-s-a’ak-Ø

(16) (yuc)

K-u-chen-me`e(n)-t-a’al u-ma´atan icp-3a-only-do-trzer-pas.icp 3a-o¤ering ‘O¤erings are made for them.’ (lit. their o¤erings are made)

63

‘He is (being) killed’ ‘He was killed’ ‘(that/may) he be killed’

As an overview, table 4 below presents a selection of voice alternations in Tseltal, Ch’orti’ and Yucatec for a cognate root, JUCH ‘‘grind’’. It shows that Tseltal mainly resorts to a‰xal morphology (or a periphrastic passive) where Yucatec has developed systematic vowel alternation marking. Ch’orti’ is characterized by a mixed pattern, where the passive infix -h(creating a complex nucleus Vh) is still productive.16 Ch’orti’ and Yucatec use su‰xal morphology for derived transitive or non-canonical roots. Table 4. Voice formations for a cognate root Tseltal

Ch’orti’

Yucatec

Active

ya s-juch’-Ø

u-huch’-i

k-u-huch’-ik

Passive (canonical)

ya x juch’-ot-Ø

a-huhch’-a

k-u-hu’uch’-ul It is (being) ground

Middle

ya x juhch’-Ø / ya x juuch’-Ø (unproductive)

a-huch’-p-a

k-u-hu´uch’-ul

Antipassive ya x juch’-awan-Ø a-huch’-m-a k-u-hu`uch’

She grinds it

It gets ground

She grinds

3.4. Passive participle The passive voice is also expressed by participial forms. The passive participle su‰x -BIL, found in many Mayan languages, has values and uses that vary across the family. Tseltal -bil and Ch’orti’ -bir form perfect passive 16. In Bachajo´n Tseltal, the -h- infix only appears as a vestige in a few CVC roots, with a middle value of spontaneous event. In some phonetic contexts, -his realized by a lengthening of the root vowel (or not realized, in the Oxchuc dialect: Polian, to appear). However, due to the lack of studies on phonetics and prosody, the possible grammatical function of vowel lengthening in this language is still a matter of investigation.

64

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

participles and may be used with all types of transitive bases.17 Unlike all other participial forms, -bil and -bir can have an attributive function without further marking, as in the first clause of the Tseltal example (17), as well as a predicative function, as illustrated in the second clause of (17) for Tseltal and in (18) for Ch’orti.’ (17) Oxom: chik’-bil pak’-bil ahch’al, banti pay-bil-Ø (tse) Pot burn-ppp make-ppp clay where cook.in.water-ppp te ch’enek deic beans ‘A pot is a fired and fashioned piece of clay where beans are (to be) cooked.’ (18) (chr)

Usta-bir-Ø e koror.te’ ka-men-er Fix/arrange-ppp-3b det trap 1a.pl-rel-prr ‘The trap has been fixed by us.’

The Yucatec participial su‰x -bil is also productive on all transitive forms and again displays passive orientation. Its extensive range of uses, often conveying manner or purpose senses, as in (19), involves aspectual values more akin to the incompletive or gerundive (Lehmann 2002, Bohnemeyer 1998: 279).18 (19) (yuc)

Le me`etik-e’ ‘e`ermanos-o’ob-e’ k-uy-a’a-[i]k-o’ob-e’ det why-td4 evangelist-3b.pl-td4 hab-3a-say-tr.icp-3b.pl-td4 hum.p’e chan kru`us-e’ to´ok-bil-Ø! one.cn small cross-td4 burn-ppas-3b ‘This is why evangelists say that little crosses should be burned’ (lit. as for a little cross, it should be burned!)

As we will see in greater detail in section 4.1, the Yucatec participle in -bil is preferred in certain subordinate clauses to maintain coreference between a core argument of the main clause and the patient of the dependent clause. 17. The Tseltal -bil su‰x has also been called a marker of ‘‘perfect’’ (Polian 2006: 85) and ‘‘patient-oriented transitive perfect’’ (Shklovsky 2005: 60). It is worth noticing, however, that the -bil participle may have a purposive sense as in Yucatec, as shown by pay-bil in example (17). In Ch’orti’ -bir is the only perfect form for transitive verbs (18), which is not the case for Yucatec and Tseltal. 18. Yucatec has other participles with patient orientation, including resultative (-a’an) and middle ([low tone on root vowel] þ -Vl ).

Ergativity and the passive in three Mayan languages

65

3.5. Passive nominal forms In Mayan languages, voice also a¤ects nominal forms. In Yucatec, a word marked as a passive incompletive without preverbal aspect markers can be used as a nominal, a property shared by all intransitive forms (Lois and Vapnarsky 2009). The nominal passive can be used as an action noun in nominal contexts like (20), where it appears with a determiner, or as a non-finite form in subordinate clauses (see section 4.1). (20) (yuc)

Te`ep k-a-ts’o’on-o[l ]! Le ts’o’on-o[l ]-o’ mix interj hab-2a-shoot.pas-icp det shoot.pas-icp-td2 neg.emph aw-ohel bix ken a-defender-t-[i ]k-Ø a-bah. 2a-know how pros 2a-defend-trzer-tr.icp-3b 2a-refl ‘You’re being shot at! You don’t even know how you’re going to defend yourself from being shot!’

In Ch’orti’, the passive action noun is derived from passive stems (bearing the a‰xes -h- for canonical bases and -n for non-canonical bases) and shows the su‰x -a’(a)r.19 Ch’orti’ deverbal nouns appear in any nominal context, as seen in (21), but very rarely with modal auxiliaries (unlike in other Mayan languages).20 (21) (chr)

Tama u-bohn-a’r e otot sat-p-a e tumin prep 3a-paint.pas-nr det house lost-mp-intr det money ‘In painting the house, the money was wasted (lit. In its being painted of the house, the money was wasted).’

In Tseltal, the su‰x -el is used to create non-finite forms from intransitive verbs irrespective of the semantic role of the argument, see (22). However, when it is applied to a transitive base, this results in a patientoriented form, see (23). The passive orientation of transitive verbs in their deverbal form with -el has been reported in early and modern descriptions 19. The su‰x -a’(a)r results, diachronically, from the combination of the archaic form of the status su‰x -ah with the nominalizing su‰x of intransitive verbs, -er. 20. The only case is the impersonal obligative construction with uk’ani ‘‘it is necessary (to)’’ which requires as complement a deverbal noun based on the passive: uk’ani pahk’-a’r e ixim it.is.necessary sow.pas-nr det corn ‘it is necessary to sow the corn’. No similar construction has been found in Ch’orti’ with other auxiliaries.

66

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

of Tzotzil (Haviland 1981: 271) and Tseltal dialects (Ruz 1989: 110, 117, Hinma´n Smith 2004: 79, Shklovsky 2003: 92, Polian 2006: 71).21 aux-b ta [Vintransitive-el ] (agent-oriented) (22) Hich lijk-Ø-ik ta beh-el te cheb alal-etik (tse) Thus begin-3b-pl prep walk-nr deic two children-pl ‘Thus the two children began to walk’ prog-b ta [Vtransitive-el ] (patient-oriented) (23) Yakal-on ta tun-el (tse) prog-1b prep follow-nr.pas ‘I am being followed’ These forms can be used as full nominals, as in Ch’orti’ and Yucatec. (24) ya k-ak’-b-at te aw-u’-el (tse) icp 1a-give-appl-2b deic 2a-be.able-nr ‘I give you your power/ability’ When su‰xed to a transitive verb, the -el forms are better interpreted as verbonominal with more verbal properties (see Lois et al. 2010). In this 21. The form in -el on transitive is found in the following constructions. a) In one of these constructions the non-finite transitive in -el, introduced by ta, bears a 3A marker which cross-references the patient of the action. This construction is found in particular with movement verbs and the progressive: prog ta [A-trV-el ], see (a). Hich lok’-ø laj bah-el te mamal ta s-tah-el te karibio-etik-e so go.out-3b d.r go-nr deic old.man prep 3a-reach-nr deic Caribs-pl-td ‘So the old man went to the encounter with the Caribs’; b) Obligatory in Bachajo´n for first and second person agents, the second construction uses the su‰x -bel and shows transitive cross-referencing of the arguments (set A/ERG for agent; set B/ABS for patient) on the -bel form: [prog-b a-trV-bel-b], see (b). The agent of the -bel form (active orientation) corefers with the set-B form of the subject of the progressive form. Teme yak-on ta j-tij-bel-ø ha’ te j-k’ayob-e. . . conj prog-1b prep1a-play-bel-3b dem deic 1a-drum-td ‘When I am playing my drum. . .’. c) The third construction involves the use of an action noun (for example CVC root plus su‰x -aw) instead of the non-finite form with -el, and no agreement marking is found on the action noun, see (c). tal-on-ix bah-el ta tah-aw arrive-1b-already go-nr prep reach-an ‘I came for the encounter’.

Ergativity and the passive in three Mayan languages

67

case, the 3A personal marker (1A and 2A are ungrammatical: see note 21, case b) does not index the possessor of the nominalized verb but the object of the transitive verb, as in (25). (25) yak-on ta a’tel, yak-on ta s-tsun-el k-ixim (tse) prog-1b prep work, prog-1b prep 3a-plant-nr.pas 1a-corn ‘I am at work, I am planting corn’ (lit. ‘I am in its being planted of my corn’) Non-finite forms in -el are also required in several complex constructions, involving the progressive, other modal/aspectual auxiliaries or motion verbs. To obtain an active transitive interpretation of the non-finite form, more complex constructions must be used, which vary depending on person (extra- vs. intralocutive) and dialect. In sum, passive nominal forms are common in the three languages: in Yucatec and Ch’orti’ they are based on the verbal passive, whereas in Tseltal a transitive form su‰xed with the general nominal su‰x -el conveys passive value. In all three languages, transitive nominal forms are either impossible or more complex than intransitive ones, including passives (Polian, to appear, Lois and Vapnarsky 2009). 3.6. Passive and argument encoding In Mayan languages, passive forms, encoded as intransitives, follow an ergative pattern. However, di¤erences between these languages with regard to ergativity and the ergativity split a¤ect the way argument encoding is realized in the passive. As Tseltal exhibits a full ergative pattern at the morphological level, in this language the patients of derived passives, whether finite (-ot) or participial (-bil ), are consistently cross-referenced by the absolutive marker. (26) Manchuk y-ak’ x k’as-b-ot-Ø ta ik’ s-bakel (tse) imp.neg 3a-give icp.intr break-appl-pas-3b prep wind 3a-bone ‘May his bones not be broken by the wind’ By contrast, Tseltal periphrastic passives show two main argumentencoding patterns, depending on the valency of the light verb – transitive or intransitive/stative – but not on that of the dependent verb. In the first type of argument encoding, the light verbs ich’ ‘‘take, receive’’ and a’iy ‘‘feel, hear’’ behave like transitive predicates (restricted to an animate experiencer). The object of the light verb (indexed by a zero third person morpheme from set B) refers to the action which is expressed by a passively oriented base (transitive root/stem-el ) – in (27),

68

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

the fact of being killed. The patient of the passive-oriented non-finite form is indexed by a set A personal marker on the transitive light verb and not on the root þ el form. (27) Ya y-ich’-tiklan-Ø mil-el te s-winkilel te lum-e (tse) icp 3a-receive-freq-3b kill-nr.pas deic 3a-people deic village-td ‘The people of this village are killed one after the other’ In the second type of periphrastic passive with laj ‘‘finish’’ (28), as with the progressive in (29), the head of the complex predicate is monovalent. In both types, raising of the argument occurs: the patient is indexed by set B on the first predicate and not on the [verb-el ] form that follows. (28) Ya x laj-on ta tah-el (tse) icp icp.intr finish-1b prep reach-nr.pas ‘I have just been caught!’ (29) Yakal-on ta tah-el (tse) prog-1b prep reach-nr.pas ‘I’m being caught’ In Yucatec, due to the aspect-based ergativity split, argument encoding in a passive construction depends, as with all intransitives, on the aspect/ mood of the verbal form. In the active/passive alternation, a change of encoding of the patient occurs in the incompletive (30), which follows the accusative pattern, but not in the completive and subjunctive, where the ergative pattern prevails (31). (30) (yuc)

a.

K-uy-il-ik-en hab-3a-see-tr.icp-1b ‘He sees me’

b. K-inw-il-a’al hab-1a-see-pas.icp ‘I am seen’ (31) (yuc)

a.

T-uy-il-ah-en cp.tr-3a-see-tr.cp-1b ‘He saw me’

b. Il-a’ab-en see-pas.cp-1b ‘I was seen’

Ergativity and the passive in three Mayan languages

69

In Ch’orti’, due to the aspect-based ergativity split and the second intransitive set of personal markers, patients in the passive clause are never cross-referenced by set A ergative markers but instead by one of the absolutive sets, either C in the incompletive (32b), or B in the completive (32c). (32) (chr)

a.

uwire’n uw-ir-a-en 3a-see-tr-1b ‘He sees me. / He saw me.’

b. i’inrna Vn-ir-n-a 1c-see-pas-intr ‘I am seen.’ c. irne’n ir-n-a-en see-pas-intr-1b ‘I was seen.’ To summarize, whereas in Tseltal patients are always cross-referenced by the same personal marker in active and morphological passive forms, in Ch’orti’ and Yucatec cross-reference marking di¤ers in the incompletive aspect, following the general ergative split. In this aspect, patients are expressed by set A (ERG) in Yucatec and by a special and innovative set of markers in Ch’orti’. Notably, Tseltal also expresses the patient by a set A (ERG) marker in two periphrastic constructions involving a transitive light verb. 3.7. Expression of the agent in passive constructions Passives di¤er in the degree to which they suppress, defocus or demote the agent (Thompson 1994: 47). In this section, we will present how these e¤ects are manifested in the languages at issue. The demoted agent is never cross-referenced on the verbal core. When expressed, it is mentioned in an oblique phrase introduced by a ‘relator’. This element may be a relational noun, tum(e` )en in Yucatec (often con-

70

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

tracted to men or ten), (33), set a-men in Ch’orti’, (40), and y-u’un in Tseltal; or a preposition, ta in Tseltal (36).22 (33) (yuc)

Bey bin t-u-be`et-ah-Ø t-uy-o´ox-pak-o’, How rs cp.tr-3a-do-tr.cp-3b prep-3a-three-cn.times-td4 tume`en bin ta´an u-chi’ib-il tume`en u-ch’ik-el relat rs prog 3a-bite.pas-icp relat 3a-flea-pos ‘This is how he made it the third time, because he said he was being bitten by the fleas.’

The relators have diverse functions besides introducing the passive agent. In Tseltal ta corresponds to the general non-oriented preposition of Mayan languages (expressing source, goal, cause, purpose, location, instrument); -u’un is also non-oriented and can introduce causes (34), including the agent, as well as result, purpose (35), or any kind of generic relation. (34) X mahl k’ahk’al y-u’un te mak-bil k’inal (tse) icp.intr fade.mp day 3a-relat deic cover-ppp weather ‘The day faded because of the overcast sky.’ (35) Bayel ya x a’teh-otik (tse) Much icp icp.intr work-1pl.incl y-u’un ya j-mak’lin-otik te k-al.nich’n-ab 3a-relat inac 1a-sustain-pl deic 1a-child-pl ‘We work a lot in order to sustain our children.’ In Ch’orti’ and Yucatec (tu)m(e` )en/-men can introduce any kind of causal argument or proposition, as seen in example (33) above from Yucatec where tume`en appears twice, introducing first a causal clause and then a causal agent. We find some di¤erences between the languages under study, and among their dialects, regarding the existence of restrictions on the expression of

22. With the exception of ta, these relators are descended from a relational noun (marked as possessed with set A indexing the possessor), a common formation in Mayan languages (Kaufman 1990: 75–78). They correspond to the ‘agented passive’ clause in Akatek marked by u’u ‘by’, ‘because of ’, as described by Zavala (1997: 453).

Ergativity and the passive in three Mayan languages

71

the oblique agent. These restrictions depend on animacy and person. In the Tseltal of Bachajo´n, which represents one of the most conservative dialects, one observes a clear tendency for humans and assimilated entities to be introduced by the relator -u’un (36), whereas non-human agents, such as animals (37) or inanimates (38), are introduced by the preposition ta. In Oxchuc (and less commonly in Bachajo´n), a human agent may be introduced by ta, but only if indefinite. (36) K’alal laj te il-ot-Ø y-u’un te kabinal-etik-e, (tse) When finish deic see-pas-3b 3a-relat deic Carib-pl-td, ya la x laj. . . alak’in-ot-Ø y-u’un icp rs icp.intr finish. . . pet-pas-3b 3a-relat ‘When they [women] were noticed by the Caribs, (it is reported) they were . . . petted by them’ (37) La y-ich’ nuts-el lok’-el ta ts’i’ te mis-e (tse) cp.tr 3a-receive pursue-nr.pas go.out-nr prep dog deic cat-td ‘The cat was chased out by a dog’ (38) Te (a)j-kanan xot’-bil ta machit-e (tse) deic agp-protect cut-ppp prep machete-td ‘The angel/guardian was cut in half by the machete’ In other Tseltal dialects, such as Oxchuc, the demoted agent is rarely introduced by a relator or a preposition (Polian 2006: 218), as in (39). (39) Nahil nop-tes-ot-Ø ha’ te pukuj first learn-caus-pas-3b dem deic devil ‘First the devil was taught’ ‘First he was taught by the devil’ This results in ambiguity in the attribution of syntactic functions and semantic roles to the nominal phrases of the sentence, especially when they have equivalent animacy status. In sentences with two lexical arguments, word order may help to disambiguate (Polian 2005). With respect to person, it is worth noticing that Ch’orti’ allows first and second person to be introduced as agent/cause by the relator -men, whereas this is not found in Tseltal and Yucatec.

72

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

(40) (chr)

k’ahs-a ni-k’ab a-men-er break.pas-intr 1pos-hand 2a-relat-prr ‘My hand has been broken by you.’

A complex scale operates in the marking of the relationship between a verb and its demoted agent, involving intertwined parameters, from syntax to pure stylistic free choice. We will deal in sections 4.3 and 4.4 with the frequent cases where the passive is used without an overt agent, either because the agent is topical and easely identifiable from the context or because it is intentionally veiled, hidden or removed from discourse. Other intransitive voices or derivations, such as middle forms in Ch’orti’ and Yucatec and the so-called ‘‘agentless passive’’ of Yucatec (Bricker et al. 1998: 346), allow for the depiction of an action without reference to any agent; it is less clear that demotion is involved in these cases.23 3.8. Applicative and ditransitive passivizations Tseltal, like most Mayan languages, but unlike Yucatec and Ch’orti’, has an applicative voice, called ‘‘referential voice’’ by Dayley (1981) and taken by Dryer (1986) to identify Tseltal as a primary object language. A -b(ey) su‰x allows a peripheral argument (addressee, beneficiary, maleficiary, recipient, target) to be promoted and encoded as a direct object, i.e. indexed by set B. Morphologically the resulting form is still a two-place predicate, but one from which the patient has been erased (two set B markers cannot co-occur). This construction may be analysed as a ditransitive construction with a choˆmeur argument (see Grinevald and Peake, this volume). The three arguments may be expressed lexically and no relator is needed for either of them, see (41). (41) Ya j-k’an chikan-tes-bey-ex s-tojol (tse) icp 1a-want appear-caus-2b.pl 3a-price ‘I want to show you its price’ The applicative form can be passivized, with indexation of the promoted argument as absolutive on the verb. In (42) the absolutive plural marker makes it clear that the indexed argument refers to the recipient.

23. The so-called ‘‘agentless passive’’ -p-ah derivation of Yucatec is lexicalized, only appearing with a few roots. Semantically it seems more appropriate to consider it as a middle form. As with the other middles, a cause may be introduced with tume`en, because of the generic causal value of this relator.

Ergativity and the passive in three Mayan languages

73

(42) Hich bin.u’til y-uch’-ik-ix ul a’-b-ot-Ø-ik (tse) So how 3a-drink-pl-already gruel give-appl-pas-3b-pl laj ul rs gruel ‘That is how they drank gruel, they were given gruel they say’ Yucatecan languages and Ch’orti’ (unlike other Cholan languages) lack the -bey applicative construction. In most constructions the beneficiary must be expressed as an oblique argument introduced by a preposition. However, some semantically ditransitive verbs (in Yucatec a subset of verbs of transfer, communication, secretion, etc.; in Ch’orti’, to our knowledge, only toyi ‘‘pay’’) accept two alternative constructions, where the syntactic object can be either the patient (43a) or the beneficiary/recipient (43b).24 The same slot and the same personal marker set are used for crossreference indexation on the verb. In Yucatec, when the patient is indexed on the predicate, the beneficiary/recipient appears as an oblique argument introduced by the preposition t(i), or as an independent pronoun (43a). In contrast, when the beneficiary/recipient is coindexed as an object on the predicate (43b), the theme is excluded as a core argument (43c).25 In the corresponding passive forms of these Yucatec and Ch’orti’ constructions, either the patient or the beneficiary/recipient can be promoted to the function of passive argument, as exemplified in (43d) and (43e) for Yucatec. (43) (yuc)

a.

K-u-bo’o[l ]-t-ik-Ø ten in-meyah hab-3a-pay-trzer-tr.icp-3b 1pr 1a-work ‘He pays me for my work’ (lit. ‘he pays my work to me’)

b. K-u-bo’o[l ]-t-ik-en hab-3a-pay-trzer-tr.icp-1b ‘He pays me’ 24. In Yucatec, most of the verbs concerned correspond to extraversive transitivization, as defined by Lehmann and Verhoeven (2006). As these authors show, extraversion in Yucatec introduces a direct object which may encode a variety of participants (patient, goal, stimulus, addressee, etc.) (ibid: 471–473). Such semantically diverse direct objects can also appear as single arguments of extraversive verbs in passive form. 25. For verbs which do not allow a transitive construction with the beneficiary indexed as absolutive, the indirect object cannot be passivized (see also Verhoeven ibid: 144): Sı´i-a’ab-en (o¤er-pas.cp-1b) means only ‘I was o¤ered (to/by someone)’ and not ‘it was o¤ered to me’.

74

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

c. ?? K-u-bo’ot-ik-en in-meyah hab-3a-pay-trzer-tr.icp-1b 1a-work d. Bo’o[l ]-t-a’ab-Ø ten in-meyah Pay-trzer-pas.cp-3b 1pr 1a-work ‘I have been paid for my work’ (lit. ‘my work has been paid to me’) e. Bo’o[l ]-t-a’ab-en Pay-trzer-pas.cp-1b ‘I have been paid’ Ch’orti’ possesses a couple of trivalent verbs, which exhibit the same properties as active transitive verbs marked by -bey in Tseltal but without morphological marking (ahk’u ‘‘give’’, a’re ‘‘say’’), see (44a). The corresponding passive form of these verbs allows promotion of either the patient or the beneficiary/recipient. This passive form can still involve two lexically expressed participants, but only the promoted argument is indexed on the verb, either the patient or the beneficiary/recipient (as in the Tseltal passivized applicative), see (44b). (44) a. Kaw-ahk’-u-Ø e wya’r e ah-ch’uyma’r (chr) 1a.pl-give-tr-3b det food det agp-baptism ‘We give the godfather the food’ b. Ni-tatanoy che ahk’-u-n-a-Ø in-te’ tah.rum 1pos-grandfather rs give-tr-pas-intr-3b one-nc wooded.land ‘My grandfather, he said, was given woodland [for cutting]’ Thus, like other Mayan languages, Tseltal has a very productive applicative voice which allows passive formation with recipients or similar participants as the core argument. Ch’orti’ and Yucatec lack such a form; however, they have sets of verbs which, without special marking, also allow the recipient to be treated as a direct object in the active form and as a single argument in the passive. 3.9. Synopsis of voice constructions Ch’orti’, Tseltal and Yucatec, like Mayan languages in general, show many voice alternations which grammaticalize the di¤erent degrees and

Ergativity and the passive in three Mayan languages

75

ways the agent may be implicated in the action. From the agent focus construction to the agentless passive or the middle voice, several devices emphasize the role of the agent or, on the contrary, demote it from the core to oblique argument status, from a central to a peripheral participant or to one omitted altogether. These e¤ects are achieved by means of changes on the morphophonological and syntactic levels, with some di¤erences evident between the languages under study, such as the Tseltal deployment of periphrastic constructions contrasting with the Yucatec exploitation of root vowel alternations. These two languages exhibit opposing tendencies, the former centrifugal, where changes of valency are exteriorized from the root and distributed over several clausal constituents, and the latter centripetal, where changes of valency a¤ect the very core of the root. Another important di¤erence concerns the way arguments are indexed on the verb. Whereas in the Ch’orti’ and Tseltal morphological passive, patients are never cross-referenced by the set A (ERG) personal markers (as agents are), this is very common in Yucatec because of its aspectual ergative split, which a¤ects all intransitives, including passive forms. Notice nevertheless that in Tseltal two periphrastic constructions with transitive light verbs involve a patient encoded with set A (ERG). Table 4 presents an overview of passive formations in Tseltal, Ch’orti’ and Yucatec. The agent of the passive forms under analysis can be expressed as an oblique argument, introduced by a generic causal relator or, in Tseltal, by a non-oriented preposition. Some Tseltal dialects show contrastive means of introducing the oblique agent, triggered by the animacy hierarchy – a semantic parameter especially salient in the grammar of this language, as we will see in the following sections. In all three languages a recipient can be promoted to the function of unique argument of a passive clause. Whereas in Tseltal promotion of the recipient is licensed by a very productive applicative voice, in Ch’orti’ and Yucatec it results from the lexical properties of certain roots which accept either a patient or a recipient as their object.

76

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

Table 5. Comparison of passive formations in Tseltal, Ch’orti’ and Yucatec

Active Passive on canonical roots

Passive on noncanonical & derived transitives Periphrastic passive Passive participle

Tseltal

Ch’orti’

Yucatec

a-Vt-b

a-Vt-tr-b

asp-a-Vt-tr-b

icp c-CVhC-a cp CVhC-a-b

icp asp-a-Cv’vC-vl cp Cv’vC-b

icp c-Vt-n-a cp Vt-n-a-b

icp asp-a-V(-trzer)-a’al cp V(-trzer)-a’ab-b

Vt-bir-b

Vt-bil-b

icp asp Vt-ot-b cp Vt-ot-b

asp a-ich’-b [Vt-el ] asp a-a’iy-b [Vt-el ] asp laj-b ta [Vt-el ] Vt-bil-b

Nominal passive: canonical roots Nominal passive: non-canonical & derived tr.

Expression of the agent

Applicative passive

CVhC-a’ar Vt-el Vt-n-a’ar non-oriented relator: (y-u’un) þ animate agent ta þ inanimate agent

causal relator: a-men þ agent

Same as incompletive intransitive form

causal relator: tum(e`)en ten/men þ agent

(asp) Vt-b-ot-b

4. Use of the passive: constraints and motivations We will now turn to an analysis of the motivations for the use of passive forms in the three Mayan languages under study. It will be shown that, although some syntactic constraints do operate, the functions of the passive are principally semantic and discourse-pragmatic in nature. We consider coreference constraints, obviation, agent backgrounding and topicality. The importance and precise contribution of these factors depend

Ergativity and the passive in three Mayan languages

77

on the language involved. In line with the terminology traditionally employed for Mayan languages, we will use the label ‘‘subject’’ to cover both the single argument of monovalent predicates, including passives, and the non-object core argument of transitives. 4.1. Coreference constraints The use of the passive may be triggered syntactically by coreference constraints in subordination. Subordination in the Mayan family is a complex matter and has been little studied: the details of subordination constructions (including the need for a relator element, the type of base required – verbal, non-finite, nominal or participial – the choice of aspect/mood and indexation) may vary depending on the type of predicate found in the matrix clause and the nature of the matrix and subordinate verbs, as well as the patterns of coreference involved. In Yucatec, coreference and syntactic control between the core arguments of the main and the subordinate clauses is an influencing factor in voice selection. As a general rule, the transitive object of an embedded clause cannot be coreferential with any core arguments of the matrix clause: in other words, it cannot be the target of control. Specific constructions are required, involving a conjunction (ka´a þ transitive in subjunctive), passivization, or both. Depending on the type of matrix verb involved, passivization is realized either by the participial passive (-bil ) or by the canonical passive in the incompletive, the latter sometimes introduced by the relator (ka´a) (see Verhoeven 2007, Bohnemeyer 2009). Whereas no argument control appears to be at work in the case of embedded clauses with the canonical passive, the embedded participle in -bil arguably entails coreference between the subject of an intransitive or the object of a transitive matrix clause and the subject of the embedded passive participle. Coreference with a core argument other than the object of the dependent clause does not trigger these constructions, and involves equi-deletion when the subordinate argument is the single argument of an intransitive. Observe (45) (where a. and c. are adapted from Verhoeven 2007: 139– 140), in which there is referential identity between the patient of an underlying bivalent subordinate predicate and the object of the transitive main predicate. The patient of the dependent clause cannot appear as the object of a transitive predicate but has to be expressed by means of one of the constructions given in (45), which involve a passive form. For some sub-

78

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

ordinate clauses, such as those of perception, the canonical passive without conjunction is also possible, as shown in (46).26 (45) (yuc)

a.

Pedro-e’ t-u-tu´uxt-ah-Ø Maria isı´int-bil Pedro-td4 cp.tr-3a-send-tr.cp-3b Maria bathe-ppas me`en u-ma`amah cause 3a-mother ‘Pedro sent Maria to be bathed by her mother’

b. Pedro-e’ t-u-tu´uxt(-ah)-en isı´int-bil Pedro-td4 cp.tr-3a-send-tr.cp-1b bathe-ppas ‘Pedro sent me to be bathed’ c. Pedro-e’ t-u-tu´uxt-ah-Ø Maria Pedro-td4 cp.tr-3a-send-tr.cp-3b Maria ka´a isı´in-t-a’ak-Ø me`en u-ma`amah conj bathe-trzer-pas.sbj-3b cause 3a-mother ‘Pedro sent Maria so that she would be bathed by her mother’ (46) (yuc)

Chen bin ka’ t-uy-u’uy-o’ob u-t’a’an-[a]l-o´o’ only rs conj cp.tr-3a-hear-3pl 3a-speak.pas-icp-3pl men hun-p’e nohoch wı`inik. cause one-nc big man ‘Then they heard that they were being called by an old man’

26. The canonical passive without a relator is barely acceptable for the example in (45)c: */? Pedro-e’ t-u-tu´ucht-ah Maria uy-isı´int-a’l me`en u-ma`amah [Pedro-td4 cp.tr-send-tr.cp-3b Maria 3a-bathe-pas.icp by 3a-mother] ‘Pedro sends Maria to be bathed by her mother’ (Verhoeven 2007: 139). Verhoeven (ibid ) also gives the alternative in (i) where a transitive dependent clause is introduced by a relator. However, when we checked this with Maya speakers, this sentence (along with other similar examples) was understood as referring to Maria performing the action of bathing rather than being bathed, in line with the constraint on coreference with the object of the dependent clause, despite the presence of the relator. (i) (yuc)

Pedro-e’ t-u-tu´ucht-ah-Ø Maria Pedro-td cp.tr-3a-send-tr.cp-3b Maria ka´a uy-isı´int u-ma`amah conj 3a-bathe 3a-mother ‘Pedro sends Maria so that her mother bathes her’ (Verhoeven) ‘Pedro sends Maria so that she bathes her/his mother’ (our informants)

Ergativity and the passive in three Mayan languages

79

In the construction with the passive participle -bil (45a), equi-deletion of the patient takes place and no personal marker appears on the subordinate predicate. (45b) shows evidence that the argument of the -bil form is controlled: set B on the matrix predicate is always interpreted as coreferent with the sole argument of the -bil form, and the sentence is unambiguous.27 In constructions involving the canonical passive, the patient is expressed by set B (ABS) if the passive is in the subjunctive after the conjunction ka´a as in (45c), or by set A (ERG) when the passive is in the incompletive/nominal form without a conjunction, as in (46). This follows the general pattern of the Yucatec aspect/mood ergative split. However, it can be argued that none of the constructions involving the canonical passive imply control. No equi-deletion takes place, unlike in all other cases of intransitive embedded clauses. Moreover, in constructions of the type seen in (45), where the embedded clause is not introduced by the conjunction, there is no coreferentiality between the object (set B) of the matrix clause and the passive subject; rather it is the embedded clause itself which is indexed by means of set B on the transitive predicate. Although this cannot be illustrated here for reasons of space, the same constraints on coreference of the transitive patient of a subordinate clause hold with O-, A- or S-controlled subordinate clauses (see ex. 138 in Verhoeven 2007), which indicates no clear accusative or ergative alignment pattern. Other properties of subordinate constructions in Yucatec reveal di¤erent types of pivot, but notably none where S and O behave alike in contrast to A (see Verhoeven ibid: 144). One important exception might be the agent focus construction already discussed in section 2.2. Tseltal shows some coreference constraints similar to those of Yucatec. However, these only apply to third persons, are overridden by the animacy hierarchy, and vary across dialects. For Tenejapa Tseltal (Robinson 1999: 148) and Oxchuc Tseltal (Polian 2008, to appear, part 9.3.6), passivization has been reported to be obligatory in the complement clause if the matrix 27. Our analysis di¤ers from that of Bohnemeyer (2009: 29–30) who argues that the ‘‘sole argument of the gerundive [the form with -bil ] is not controlled at all’’. The evidence provided is twofold: 1) the ungrammaticality of complex sentences where a coreferential Undergoer argument of -bil would be identifiable by the set-B marker on the matrix verb; however, such ungrammaticality seems to depend on the type of the matrix predicate, in that it is expected in predicates of desire but is not found in others, as illustrated by (45b); 2) the possible inclusion of a new participant as the subject of the embedded -bil form; but according to our data, this inclusion results in sentences judged by speakers as awkward at best.

80

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

subject is coreferential with the dependent patient of a bivalent predicate (whereas an active form is used in the other cases). Thus, when no semantic hierarchies or topicality constraints are in play, any core argument which is coreferent with the subject of the matrix clause is taken as the syntactically highest argument. This is illustrated in the variant of the construction seen in Oxchuc (where the passive agent does not have to be introduced by a preposition). The subject of the matrix clause can only corefer with the agent of the transitive dependent clause in (47), and with the patient of the dependent passive in (48); both sentences are unambiguous (examples adapted from Polian, to appear). (47) Y-u’un ja’ y i-al [te ya y i-uts’in-Ø j te ch’in kerem] (tse) 3a-relat foc[3b] 3a-say conj icp 3a-bother-3b det dim boy ‘he thinks that he will bother the young one’ (48) Y-u’un ja’ y i-al (tse) 3a-relat foc[3b] 3a-say ‘he thinks te ch’in kerem] [te x-’uts’in-ot-Ø i conj icp.intr-bother-pas-3b det dim boy that he will be bothered by the young one’ Further investigation is required as regards dialectal variation in this domain. In Bachajo´n Tseltal, coreferentiality between the subject of the main clause and the object of the dependent clause does not obligatorily entail use of the passive voice, as illustrated in (49) where an active form is used.28 (49) Ya x-xi’-Ø i ha’ te ini j ya y j-il-Ø i icp icp.intr-fear-3b foc[3b] det dem icp 3a-see-3b this [guy] sees himi ’ ‘He i fears In Tseltal, use of the passive may also be entailed by syntactic constraints on the expression of possession. When there is coreference between the 28. The same is true for asyndetic coordination. In the following example, agent or patient can be the subject of the coordinate clause: te winik laj y-il te mut, patil lok’-Ø bah-el [deic man cp 3a-see deic bird afterwards go.out-3b go-nr] ‘the man saw the bird, then he/it went’. In the second interpretation, the subject of the first clause is not coreferential with the subject of the second clause.

Ergativity and the passive in three Mayan languages

81

possessor of the matrix subject and the patient of a dependent bivalent predicate, the dependent patient cannot be expressed as an object and hence a passive form is used, see (50), adapted from Polian (to appear). This can again be understood according to the hypothesis that the outranking core argument must be indexed as the higher argument in the embedded clause. As with other complement clauses, this constraint may be overridden by semantic hierarchies (in particular, when the possessed participant is higher in animacy than the possessor, a dependent active form may be used, see (51)). (50) ya x tal-Ø1 il-ot-uk-Ø2 y2-ihts’in1 (tse) icp icp.intr come-3b see-pas-sbj-3b 3a-younger.brother ‘her younger brother comes to see her’ (lit. for her to be seen) (51) ha’-nix te s-winkilel te ya s-lajin-ik i lum.k’inal-to (tse) foc-adv deic 3a-people rel icp 3a-end-pl prox territory-td ‘It is its inhabitants who end up with this land’ Thus, Yucatec shows syntactic restrictions on coreference with the object in subordinate clauses, which favour the use of the passive. Similar constraints are in play in some Tseltal dialects, albeit in interaction with semantic hierarchies. At least in Yucatec, the construction is independent of any specific alignment (either accusative or ergative), since the use of the passive depends on restrictions of coreference between the object of the subordinate clause and any of the matrix arguments. Ch’orti’ does not seem to show the constraints on coreferentiality observed in Yucatec and Tseltal. Dependent clauses are expressed either in a serial construction with equi-deletion phenomena or with a conjunction (mainly used when there is no coreference between the matrix and the dependent arguments).29 In all cases, any type of coreference may be attested, and voice alternations are not required. (52) (chr)

b’a’kta-Ø a’i uw-ira-Ø fear-3b they.say 3a-see-3b ‘They say she fears he can see her’ or ‘They say she fears she can see him’

29. Ch’orti’, in contrast with the other Cholan languages, does not require any specific verbal inflectional morphology to mark syntactic dependency. Equideletion optionally occurs when markers of set C or of set B are repeated: uyakta kukreme’n [3a-leave fall-1b] ‘he let me fall’ which can be realized without equi-deletion: uy-akte’n kukreme’n [3a-leave-1b fall-1b] ‘idem’.

82

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

4.2. Semantic hierarchies and discourse prominence The influence of animacy on syntactic structures is a phenomenon which has received increasing attention in cross-linguistic studies in recent decades (Aissen 2003, Branigan et al. 2008, among others). Among the Amerindian languages, it is attested for instance in the Algonquian, Tupi-Guarani and Carib language families. The passive is one of the most common constructions linked with animacy constraints. The influence of animacy on the use of the passive is also observed in Tseltal and Ch’orti’, where animacy hierarchies have strong e¤ects on the grammar. In Yucatec, animacy a¤ects alignment patterns to a lesser extent. Tseltal is particularly interesting in this respect because of the obviative function which is currently being acquired in some dialects by the morphological passive in -ot (Polian 2005, to appear).30 The same phenomenon has been observed in closely related languages such as Tzotzil, Cholan and K’ichean (Aissen 1992, 1997, Robinson 1999, Zavala 2007).31 Thus, in her analysis of the agent focus construction in Tzotzil – basically a passive construction with intransitive morphology and transitive semantics – Aissen shows that this type of passive clause ‘‘express[es] the inverse function only along one dimension, that of obviation’’. This occurs only in clauses where both arguments are third person and both are specified (Aissen 1999: 479). A similar analysis is proposed by Polian for Oxchuc Tseltal, although agent extraction seems less constrained in Bachajo´n Tseltal, as we will now see. 30. We follow the contrastive definitions o¤ered by Guillaume (2006, 2009): ‘‘the mechanism of hierarchical agreement refers to the fact that transitive verbs agree with the core participant of the clause which is higher in terms of the person hierarchy, regardless of its grammatical function [. . .] The mechanism of obviation is manifested in combinations of third person participants. The participants are ranked on a hierarchy according to the semantics of their referent along an animacy scale or their respective topicality. [. . .]. The phenomenon of inversion refers to the fact that, in addition to agreement marking (on the verb) and proximate/obviative marking (on the NPs), Algonquian languages have special verbal markers to indicate whether the higher ranked participant is the subject (direct marking) or the object (inverse marking)’’. As we do not find special marking of this kind in Bachajo´n Tseltal, we prefer to avoid the term ‘inverse’ despite its use by Polian (to appear) for Oxchuc. 31. Robinson (1999: 163) considers that Tseltal is better characterized as having a ‘‘true passive’’, ‘‘since person and agreement converge on the promotion of the patient and the demotion of the agent’’. This is in contrast with Tzotzil, which presents an agent focus inverse (Aissen 1999), and Akatek, which has two ‘inverse’ constructions (Zavala 1997: 457–60).

Ergativity and the passive in three Mayan languages

83

In its obviative function in Tseltal, the morphological passive, rather than demoting the agent argument, signals the higher saliency of the patient in terms of animacy (53), or other hierarchies such as definiteness (54), specificity, possession (53) and topicality (for a detailed analysis, see Polian 2005 and 2006). In (53), the men’s heads (standing metonymically for the men themselves) are given a higher degree of animacy than the magic powder and constitute the possessed NP – both features leading to use of the passive – while the powder is focused by means of a cleft construction. (53) Pero ha’ te sibak te la s-tek’-e (tse) emph dem deic powder rel cp.tr 3a-trample-td jim-b-ot-Ø s-jol spoil-appl-pas-3b 3a-head ‘But it is this powder they trampled on, (the reason why) their minds got confused’ In (54), the patient (the women) constitutes the topic, and is encoded by a definite NP, which is also more highly specified than the agent thanks to the accompanying relative adjunct (‘‘who were bathing’’). (54) tawaltay-ot-Ø y-u’un j-yakubel ha’ te ants (tse) threaten-pas-3b 3a-relat agp-drunkard dem deic woman te ay ta atimal rel exist prep bath ‘Those women who were bathing were threatened by a drunkard’ Obviation with passive -ot only concerns the third person (Ø). It helps to disambiguate the role of arguments in a system without case and with a relatively flexible constituent order. In dialects (Oxchuc, Tenejapa) where the relator -u’un, used to introduce passive agents, is frequently omitted (see section 3.7), this disambiguating e¤ect seems all the more important. First and second person are not a¤ected by this pattern. There is apparently no restriction on their use as the object of a transitive predicate with an inanimate or a less animate agent, as in (55), where the stone appears as subject. (55) La y-ehchen-tes-on te ton-e (tse) cp.tr 3a-hurt-caus-1b deic stone-td ‘The stone hurt me’

84

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

However, with certain verbs, especially those of interlocution, a passive applicative construction is often preferred when the ranking [1/2 > 3] is violated, as in (56) where in natural discourse one would say (a) rather than (b). (56) a. (tse) b.

Wojey, al-b-ot-on-ix ( y-u’un) Yesterday say-appl-pas-1b-already 3a-relat ‘Yesterday, I had already been told (by him)’ Wojey, laj y-al-b-on-ix Yesterday cp 3a-say-appl-1b-already ‘Yesterday, he had already told me’

To summarize, obviation in Tseltal, as in Tzotzil, depends on several hierarchies which are often interrelated in complex ways: animacy (animate/ inanimate, human or personified non-human/non-human), definiteness and specificity (including proper noun/common noun), topicality, and the distinction between intralocutive and extralocutive persons are all relevant factors. In cases of conflict between animacy and definiteness or animacy and possession, active and passive voices are both possible. Topicality, to be discussed in 4.4, interplays with semantic features in di¤erent ways. In (57), the topical and, most of all, definiteness properties of the stones outweigh their lack of animacy, with the result that no passive is used despite the fact that an inanimate participant (stones) is ‘acting’ on human patients: (57) Ch’ay tal koh-el ton-etik, lom muk’-ix nax (tse) Get.lost come fall-nr stone-pl very big-already only i te muk’ul ton-etik-e la s-mil-tiklan and deic big stone-pl-td cp 3a-kill-freq j-me’-tik.j-tat-ik-etik-e 1a-mother-rev.1a-father-rev-pl-td ‘Stones fell down, they were really big, and the big stones killed many of our people (lit. our respected mothers-fathers)’ It is the contextual relation between the two entities that matters and not their putative absolute values. The entity which is considered more important at a certain point in a given speech event, either by nature or by temporary attribution, or because of its role in the narrative or conversation (see 4.4), tends to be presented as the subject. This is a subtle issue which requires further research, particularly taking into account di¤erent discourse genres.

Ergativity and the passive in three Mayan languages

85

In Tseltal the passive can be seen as conditioned by a saliency hierarchy, which is also of importance for other constructions such as the use of the relational marker u’un versus the preposition ta to introduce a passive agent, or the constraint on an animate agent required for the periphrastic passives with ich’, a’iy and laj (see 3.2). But it is also a strategy used to disambiguate agent and patient roles when subject and object are third persons, equivalent in animacy and sometimes without a di¤erentiating agent relator. Passive voice is mostly linked to hierarchy and word order to topicality, but the two criteria often overlap: ‘‘information structure is responsible for word order – in passive and active sentences’’ (Polian 2005: 43, our translation). In Ch’orti’, the animacy hierarchy [human > animal > inanimate (vegetable, object, abstract concept)] is crucial in diathetic relations. The study of a corpus of narratives shows that the argument which is higher in the hierarchy always appears as the subject in two-participant events. This rule also governs uses of the passive. Thus, in (58), where both arguments are topicalized (as shown by their fronted position), the fact that the patient, a human, is higher in animacy than the agent, an animal, triggers the use of a passive form. (58) (chr)

Inw-ir-a ke e winik u-men e ts’i’ kuhx-a 1a-see-tr conj det man 3a-relat det dog bite.pas-intr ‘I saw the man being bitten by the dog’

Again, the ways in which entities are located on scales of saliency must be read with an awareness of the context. In Mayan tales animals are often given human attributes and appear in an equipollent relationship with them, as in (59). (59) (chr)

War u-takr-iy-et yer e mama’ ch’o’k prog 3a-help-tr-2b dim det uncle mouse ‘The uncle mouse is helping you [Sun and Moon’s son].’

As will be further analysed in the next section, in Ch’orti’ topicality also plays a central role in the use of passive constructions. However, when animacy and topicality are in conflict, animacy prevails. This is seen in example (60), where the patient (the ears of corn) is the topic, and as such might be expected to be expressed as the subject of a passive clause. But due to the lower status of the patient on the animacy scale, the passive is not used.

86

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

(60) (chr)

Ya’ a-tur-an yer e nar twa’ yer e nar There 3c-sit-posit dim det ear.of.corn for dim det ear.of.corn ira ma’chi u-k’a’-p-es wa’kchetaka u-winkir e dem neg 3a-end-mid-caus quickly 3a-master det wi’nar, e xiximay hunger det Xiximay Lit. ‘Here are the ears of corn, so that these ears of corn the master of hunger Xiximay doesn’t destroy quickly’

Ch’orti’ does not seem to be sensitive to the person hierarchy [1st > 2nd > 3rd], as seen in (59) where the object is a second person and the subject a third person. In Yucatec, the animacy hierarchy may influence the use of passive forms, but it does not appear to be a decisive factor. Indeed, Bohnemeyer (2009) argues that it represents the weakest factor in the prominence hierarchy [topicality > definiteness > humanness > animacy] which constrains alignment and word order in Yucatec (see also Skopeteas and Verhoeven 2005). If the patient outranks the agent (U > A) in terms of one of these parameters – bearing in mind the hierarchy in which these parameters are ranked – a transitive form is excluded and the speaker may resort to di¤erent strategies: passivization, left dislocation (for topics), or clefting (agent focus). Since topicality is a major factor in alignment constraints (see section 4.4) and topics are more often humans than animals and a fortiori inanimates, instances where the animacy hierarchy acts as the sole factor of alignment are at best rare in natural conversation. However, the weakness of animacy as a relevant factor can be exemplified with cases where animals or even inanimate beings are encoded as subject agents and humans as object patients, as seen in example (61) below. In (61) the puma is topical and appears as a transitive subject (set A) with the human patient indexed as object (set B). Of the elicited examples in (62), although version (a) sounds more natural because the human participant Juan is understood as topical, version (b), where the loaves of bread are expressed as subject agents, is also readily accepted. As in Tseltal and Ch’orti’, first and second person are not a¤ected by semantic hierarchies, and no restriction has been observed on their use as objects of transitive predicates; in (63), from a tale, the second person is indexed as the object of a transitive predicate whose subject agents are the loaves of bread.

Ergativity and the passive in three Mayan languages

(61) (yuc)

T-u-lah-Ø in-ta`at e koh-o’ cp.tr-3a-hit.with.hand-3b 1a-father det puma-td2 ‘The puma struck my father’

(62) (yuc)

a.

87

Niik-Ø ya’abach x-tuti.wa`ah yo´ok’ol le Juan-o’, Fall-3b lots gn-corn.bread on det Juan-td2 ka kı´in-s-a’ab-Ø men e wa`ah-o’ conj die-caus-pas.cp-3b cause det bread-td2 ‘Many loaves of (ritual) corn bread fell on top of Juan and he got killed by the corn breads’

b. Niik-Ø ya’abach x-tuti.wa`ah yook’ol le Juan-o’! Fall-3b lots gn-corn.bread on det Juan-td2 Ka t-u-kı´in-s-ah-Ø! conj cp.tr-3a-die-caus-cp.tr-3b ‘Many loaves of corn bread fell on top of Juan! And they killed him!’ (63) (yuc)

Ka t-aw-a’al-e’ ka la`ah e`em-ek-Ø conj cp.tr-2a-say-td4 conj all go.down-sbj.intr-3b tula`aka le wa`ah-a’, mye`entras t-u-yach’-ech! cp.tr-3a-crush-2b all det bread-td1 conj ‘And you said that all the loaves of bread should fall, but they crushed you!’ [txt2]

In sum, the animacy hierarchy is a factor motivating use of the passive in the three languages, though its importance and the details of its operation vary. In Tseltal its influence is embodied in an incipient obviative system, and in Ch’orti’ animacy imposes strong constraints on the use of the passive for marking topicality (see below), whereas in Yucatec it interplays more freely with topicality and other discursive preferences. As far as we know, in none of the languages under study does the person hierarchy seem to strictly determine changes in diathesis. This weak asymmetry between the first/second and third person contrasts with what is seen in other Mayan languages, in particular Mocho, where split ergativity is triggered by person (Larsen and Norman 1979: 353). But it argues for the idea that the use of passive voice can serve as part of a disambiguating strategy when two third person participants are involved.

88

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

4.3. Generic, backgrounded or avoided agent Related to saliency is the fact that, in the three languages analysed here, passive constructions are very frequently used when the agent is generic, collective, unspecified, unknown, irrelevant or obvious. In such cases – which often correspond to contexts where Mayan speakers would use a transitive with third person plural subject in their Spanish translation – the agent is not mentioned. This is very common in descriptions of the traditional way of doing things, or prescriptions for proper ways of behaving or executing an action, as in (64) and (65). In fact, this use of the passive might be considered as one characteristic feature of discourse genres related to the description of traditional practices. (64) (yuc)

Ma’ hach u-han-t-a’al kan way-e’ neg very 3a-eat-trzer-pas.icp snake here-td4 ‘Snake is not eaten very much here (¼we don’t eat snake here)’

In Ch’orti’, the generic passive construction is almost systematically used in narratives presenting recipes, or the sequence of certain routine events. (65) (chr)

Tya’ k’ani a-che’-e-n-a e pahbur.sa’, a-hihx-a when vol 3c-do-tr-pas-intr det shuco.atole 3c-shell.pas-intr u-’ut e ixim bahxan i a-ts’ahy-a tama e ha’. 3a-grain det maize first and 3c-soak.pas-intr prep det water ‘When one wants to make shuco atole (a drink), first maize grains are shelled and soaked in water’

Alternatively, transitive forms with a second person (in Yucatec) or a first/third person plural (Tseltal) set A marker can also be used to express a generic agent. Example (66) shows a generic agent expressed in two ways: first with third person plural (s- . . . -ik), and second by a passive form (-ot). (66) La s-na’-ik te bin.u’ til ya x [s-]chik’-ik te (tse) cp.tr 3a-know-pl deic how icp icp.intr 3a-burn-pl deic tan-e, te bin.u’til ya x chik’-ot-Ø te limestone-td deic how icp icp.intr burn-pas-3b deic tan-e sok ts’u-bil taj-etik limestone-td with cut-ppp pine-pl ‘They knew how to burn limestone, how limestone was burnt with pine chips’

Ergativity and the passive in three Mayan languages

89

In other cases the use of a passive is motivated by the intentional deletion of the agent for pragmatic and cultural reasons (see also Ochs Keenan 1976, Shibatani 1985, among others). This backgrounding or deletion of the agent figures among Tseltal politeness devices (Brown 1987), and is common in Yucatec in conversations concerning supernatural agents in particular (Vapnarsky, in prep.). For instance in (67), an extract from a Yucatec personal narrative, the speaker describes the disturbances and misdemeanours of the souls of the dead, who drove his ancestors to abandon their village. Throughout the narrative, the speaker systematically uses passive forms to refer to the actions of the malign souls, and by doing so he avoids mentioning these agents explicitly. (67) (yuc)

Mun-cha’-[a’]al a-ma´an-i’! neg.3a.icp-let-pas.icp 2a-pass-td3 Ta´an-chi’in-i [l ] ye´ete tu`unich prog.2a-throw.pas-icp with stone ta´an-chi’in-i [l ] ye´ete che’, prog.2a-throw.pas-icp with stick, ta´an-muk-lu’um-t-a’a [l ] prog.2a-bury-earth-trzer-pas.icp ‘They don’t let you pass (lit. your passing there is not being permitted), you’re pelted with stones, you’re pelted with sticks, you’re covered with earth’

4.4. Topicality Topicality is another important parameter of passive use in the three languages under discussion here. The topic is that element of a sentence which is presented as already existing in the discourse, and/or which the subsequent portion of speech is expected to be ‘‘about’’. It has been argued that, in Maya Tz’utujil, the primary topical participant in a discourse tends to function ‘‘as the syntactic pivot in each of the clauses in which it occurs’’ (Van Valin and LaPolla 1997: 283¤ ). The topic appears in ‘‘topic chains’’ (Dixon 1972) either as the agentive subject of a transitive or as the unique argument of an intransitive predicate. When the topic is the patient of a transitive predicate, it tends to be expressed not as an object, but as the subject of a passive. This corresponds to a discourse use of the passive found cross-linguistically, both within the Mayan language family (see Zavala 1997 for a thorough study

90

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

on Akatek) and also commonly outside it (Cooreman 1987, Givo´n 1994 among others). 4.4.1. Topical patients as ‘‘subjects’’ Analysis of coordinate constructions in Ch’orti’ narratives reveals a strong tendency for the subject (whether marked on the verb as ergative or absolutive) to coincide with the topic. This constraint determines constituent order, but it is overridden by the animacy hierarchy. According to Quizar (1994), the unmarked word order in Ch’orti’ is SVO, largely determined by pragmatic conditions: the topical subject appears in preverbal position, and the subject must be the topic in transitive clauses.32 As for intransitive clauses, both VS and SV seem equally acceptable; the condition that the topical subject should appear in preverbal position explains the SV pattern.33 In her analysis, Quizar only considers the contrast between transitive and intransitive verbs, without taking into account that the latter are often passive, middle, or antipassive derived intransitives. This is illustrated in table 6 with a narrative collected in Oque´n, Jocota´n, Guatemala. Table 6. Types of intransitive forms in a Ch’orti’ narrative (wi’nar) Number of occurrences (% on monovalent verbs) underived intransitives

27 (41.54%)

passive forms

21 (32.31%)

middle forms antipassive forms

12 (18.46%) 5 (7.69%)

32. Whereas in Yucatec transitive clauses the preferred order is VS when the object is not lexically expressed and the agent is new information, in Ch’orti’ the subject appears predominantly in preverbal position (SV) when the agent is new information, even if the lexical object is omitted. 33. For example, VS represents 47.62% of monovalent predicates in the narrative ‘‘Harvest’’ and 54.84% in the narrative Wi’nar ‘Hunger’ (see table 6). All the examples from Ch’orti’ come from a corpus of texts recorded in Oque´n, Jocota´n, Guatemala in 2007. The text ‘‘Harvest’’ (10 mins) describes certain rituals and traditions that precede the corn harvest. The text Wi’nar (8 mins) describes the great famine of 1914 and the di¤erent means used by the Ch’orti’ people to survive.

Ergativity and the passive in three Mayan languages

91

Quizar mentions some rare cases where the patient encoded as an object is topical. A closer look at these examples reveals that if a transitive construction is maintained in such cases, this is because the non-topical agent is higher in animacy, a configuration which excludes passivization, as seen in (60) above. However, when both arguments rank equally in terms of animacy, or when the patient is higher in rank, the passive construction tends to be systematically used if the patient of a transitive action is the topic. In (68) below, the ‘‘angry people’’ are topical, and are expressed as subjects regardless of their semantic role. This is made possible by the fact that the people and Jesus are treated as equal in rank with respect to animacy. (68) (chr)

E ah.k’ihna’r-ob [. . .] kora.kora a-sut-p-o’b war det angry.people-pl [. . .] anywhere 3c-return-mid-intr.pl prog u-sahk-o’b i wix che a-mahres-n-o’b 3a-look.for-tr.pl and prog.already rs 3c-deceive-pas-intr.pl u-men e Jesu´s ma’chi u-tahw-iy-ob 3a-relat det Jesus neg 3a-find-tr-pl ‘The angry people [. . .] are again looking everywhere for Jesus and they’re being deceived by him, they haven’t found him’

Tseltal and Yucatec show a similar pattern, with Ch’orti’. In Tseltal animacy tend to prevail over topicality (see section 4.2) but see (70), (71). In this language any passive, morphological (-ot) or periphrastic passive can be used for the purpose of topicalization. In (69), both arguments are equal in animacy, and the topic is the man. He is encoded first as the subject of a modal phrase, and then as the subject (patient) of the passive verb mil-ot-Ø. (69) K’an to baht s-t’un te y-ihnam i mil-ot-Ø y-u’un (tse) want adv go.3b 3a-follow deic 3a-wife and kill-pas-3b 3a-relat te ants komo ya (x) suht-em-ix ta chambalam deic woman adv icp icp.intr turn-cp-already loc animal ‘He wanted to follow his wife, but he was killed by the woman [his wife] as he had already turned into an animal’ In example (70), the referents of the patient argument in the subordinate clause, (the ‘‘jaguar-puma’’ and the puma) are lower in humanness and definiteness than the human agent; however, as primary topic in this section of the narrative, the animals function as the passive subject in the subordinate clause (indexed by set A y- on the light verb ich’ followed by the non-finite passive oriented form in -el ).

92

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

(70) Te burum.choj ya (x) s-jo’in-tay te choj-e (tse) deic jaguar.puma icp icp.intr 3a-go.with-trzer deic puma-td te x-xi’-oj-ik te ya y-ich’-ik nuts-el conj 3a-fear-pf-pl deic icp 3a-receive-pl pursue-nr.pas y-u’un te kristano-etik 3a-relat deic people-pl The jaguar-puma [focus] joined the puma because they were afraid of being pursued by the people’ (Txt2) Passivization to maintain topic continuity despite the presence of an agent higher in animacy is also observed in (71). Here the patient argument corresponds to the noun phrase with s-k’uk’um-al ‘‘its feathers’’.34 (71) Ma’ba ts’ak-al s-k’uk’um-al s-ne te hukpik-e, (tse) neg complete-adj 3a-feather-pos 3a-tail deic momoto-td ha’ laj ta s-kaj te tek’-b-ot-Ø dem rs prep 3a-cause deic stand-appl-pas-3b y-u’un te kabinal-etik-e 3a-relat deic caribs-pl-td ‘The feathers of the momoto’s tail are not complete, because they have been trampled by the Caribs’ The passive can also be used for changes of topic for discourse purposes. In (72), taken from the description of a ritual performance, the belly and its possessor become the topic expressed as the subject of a passive form due to the importance of the as yet unborn child inside it: (72) ya s-kuch-ik ta ahk’ot te kapitan, (tse) icp 3a-carry-pl loc dance deic captain Ya x baht ta [s-]xat-el-ix x-ch’uht, icp icp.intr go.3b prep 3a-flatten-nr.pas-already 3a-belly ya [x] xat-b-ot-Ø x-ch’uht, icp icp.intr flatten-appl-pas-3b 3a-belly y-u’un te ay laj te t’ut’ alal 3a-relat rel exist rs deic little child 34. K’uk’um ‘feathers’ functions as a metonymic representation of the momoto bird; in this section of the narrative, allusion is being made to the origin of the shape of the tail of the real bird, and as such, feathers are the topic of this episode, although they are clearly inanimate in contrast to the fierce Caribs, enemies of the Bachajontecos.

Ergativity and the passive in three Mayan languages

93

‘They (the Caribs) carry the captain for dancing, he goes to the ‘being flattened’ of his belly [ritual], his belly is flattened, because, they say, there is the little child’ The topically motivated use of the passive in Yucatec is illustrated well by a section of a narrative whose speaker, an old woman, recalls how she once killed a puma (see extract in the Appendix). Three participants are present in the fight being described: the puma, the speaker, and her dogs, which helped her to fight the feline. Of 51 verbal forms, 35% correspond to intransitive movement predicates, 33% are transitive active, and 18% are passive. Leaving aside movement predicates, the narrator is mainly expressed as a transitive subject but also as an object; the dogs are expressed as the agents of transitives or the demoted agents of passive forms; while the puma is expressed as the subject or the object of a transitive verb in almost equal proportions, and unlike the other participants it also often appears as the patient argument of a passive form (see the Appendix for precise figures). This use is arguably connected with the predominant topical status of the puma in this section of the story. However, changes in the way the puma is presented also depend on who is dominating whom during the fight. In the central section (lines 16 to 26) the puma is consistently the topic, as is reflected in the discourse by its maintenance as the subject of transitive and passive intransitive predicates. At this point, the puma is mainly fighting against the dogs, which may be explicitly mentioned but always as an oblique argument (introduced by the relator ten). However, the main duel confronts the woman with the puma, at the beginning and the end of the extract. When these protagonists are involved, the fight is always presented with transitive predicates clearly indicating the dominating/ dominated roles, and topicality becomes a weaker factor in voice selection.35 The section begins with transitive predicates (1–4, 11–16) whose subjects refer to the woman and her dogs, while the puma is expressed as the object. Later on, the puma repeatedly strikes the woman; this action is also expressed by transitive forms, with the woman (first person pronoun) now encoded as the object (line 24). Eventually the dogs manage to bring down the puma (22–33) and the woman kills it by cutting o¤ its head (34– 36). In this final climax, the woman, who is less topical but dominates the action, recovers her subject position as agent of the verbs of injuring. No 35. This is also seen in column 2 of the table presented in the Appendix, which shows that the narrator and the puma may be indexed as the object of transitives even when they are topical.

94

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

further passive expressions are used with the puma as subject. The defeated animal is now only referred to as the object of a transitive predicate, a form presumably emphasizing its patienthood (32–35). This extract shows the intimate interaction between control of the action, animacy, and topicality which operates in the choice of passive versus active voice in Yucatec, and whose analysis crucially requires us to take into account the roles and types of event depicted, as well as the progression of the narrative, besides information structure. 4.4.2. Maintaining topic continuity with oblique arguments Some other interesting cases concern the use of the passive to maintain topic continuity with oblique arguments. In Yucatec, one construction involves the expression of the topical argument as possessor of the patient (encoded as set A), which appears as a passive subject. This is exemplified in (73), where the topic is the subject of intransitive and transitive forms, as well as the possessor of the passive subject (the demoted agent of this last action is clearly the protagonist’s wife, a specific human mentioned just beforehand in the narrative). (73) Pos le’ti’-e’ mix we`en-ih, t-uy-a´ah-s (yuc) thus 3pr-td4 neg.emp sleep-intr.icp.3b cp.tr-3a-awake-caus u-fa`amilya, ka ts’-u-me`e-t-a’al u-yuk’-ul, 3a-wife conj term-3a-do-trzer-pas.icp 3a-drink-nom u-ha`an-a[l ] tula`aka bin bey-o’ 3a-eat-nom all rs like-td2 ‘So, as for him, he didn’t sleep, he woke up his wife, and his drink and his food were prepared’ Another construction involves verbs of saying in narratives. As a general rule, these verbs are found in the active form when the topic is the agent, the speaker of the reported speech event, but in the passive when the topic is the addressee of the reported speech. Syntactically, the reported speech is cross-referenced on the verb as the argument of the passive ‘‘being said’’ and the topical addressee is expressed by the indirect pronoun ti’. This use is fairly systematic and it is an important way for the audience to keep track of who is speaking to whom in Yucatec narratives, which are traditionally very rich in dialogue and direct reported speech.36 36. Direct speech is also signalled by the verbatim quotative marker ki (k- þ ergative su‰x, also translated by ‘‘say’’), to which the rule explained here does not apply.

Ergativity and the passive in three Mayan languages

95

(74) Tunha`anal bine’ kuya’alik bin: (yuc) He is eating, so-they-say, he says [to his wife], so-they-say’’: – ‘‘Ma`adre, senyo`ora wa ka akre’erten, – ‘‘Damn, woman, if you would believe me, hum p’e ba’a ken intsikbatik tech.’’ Ki bin. I would tell you something’’, said he, so-they-say. (. . .) – ‘‘Ma`adre, pos wa ka xı´ikech achen ilik ka’achih ki bin, – ‘‘Damn, and if only you had gone to look, said she so-they-say, ts’awilik ba’axi’’ kuya’alah bin ti’ ‘‘you would have seen what it was!’’ it was said [by his wife] to him so-they-say An equivalent distinction is made in Tseltal by means of a passive applicative. In narratives, the active form x-chi ‘‘he said’’ is postposed to the quoted speech, but if the addressee of the reported speech is the topic, then the passive applicative of another verb ‘‘say’’ (al ) is used. This is illustrated in (75), where chi is used when the grasshoppers are the speakers and al-b-ot when they are the addressees. (75) (tse)

– Yak-uk aw-ak’-b-onkotik tehbuk j-we’el-kotik? icp-def 2a-give-appl-1b.pl.excl some 1a-food-pl.excl x-chi-Ø-ik; al-b-ot-Ø y-u’un xanich’-etik: icp.intr-say-3b-pl say-appl-pas-3b 3a-relat ants-pl – ay a-wokol-ik y-u’un ma’ba a’tej-ex! exist 2a-trouble-pl 3a-relat neg work-2b.pl – Would you give us some (of ) our food?’’ they (grasshoppers) say. They were answered by the ants: – you got into trouble because you didn’t work!

In summary, motivations for the use of the passive cover the same set of parameters in Ch’orti’, Tseltal and Yucatec, but these parameters are of di¤erent importance in each language. The passive is triggered by syntactic constraints only in restricted cases, which involve cross-reference in certain subordination constructions. Semantic hierarchies, especially animacy, play a significant role. The passive is also a privileged and common way to express a generic, implicit or hidden agent. Finally, topicality seems to be one of the most important factors for the use of the passive in all three languages. Genericity and topicality are of similar importance among the three languages. Overall, Yucatec seems to be more sensitive to syntactic

96

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

constraints and Tseltal and Ch’orti’ to be more significantly a¤ected by animacy and other hierarchies.

5. Concluding remarks On the basis of a study of three morphologically ergative Mayan languages, Ch’orti’, Tseltal and Yucatec, we have shown the predominance of discursive, semantic and discourse-pragmatic factors, rather than syntactic ones, as motivations for passive use. This is true independently of the di¤erent configurations of ergativity that characterize each of these languages, and is coherent with observations made for other Mayan languages such as Mam or Akatek. In Yucatec, which presents a stronger mixed accusative pattern than the other two languages as regards crossreference marking, some syntactic constraints on the use of the passive do operate. However, these constraints do not reveal any clear or predominant accusative alignment. Furthermore, Yucatec has an agent focus construction showing a specific syntactic treatment of the agent which is comparable to an ergative pattern, and which makes Yucatec closer to those Mayan languages which possess the focus antipassive. These properties confirm that there is no unified syntactic alignment in this language. Tseltal, which is fully ergative in terms of cross-reference marking, shows syntactic constraints entailing the use of the passive similar to those seen in Yucatec. But in Tseltal these constraints are subordinated to semantic hierarchies, in particular animacy, due to the prominence of the obviative system in this language. While Tseltal has no focus antipassive, Tzotzil, a closely related language, does possess such a construction; it is again governed by obviation and not by syntactic alignment. In Ch’orti’, voice does not serve a syntactic realignment function, but instead has semantic, discourse and information structure motivations. These facts show that the existence of the passive in Mayan languages has little to do with their ergative or accusative properties, and that its behaviour cannot be taken to provide much information about the putative ‘basic’ status of any particular alignment pattern. Nevertheless, the passive is an essential part of the grammar and discourse strategies of Mayan languages, for several reasons: it represents a crucial means of maintaining topic continuity or signalling a change in the topic in both narrative and conversation, and hence of disambiguating roles and participants; it represents one way in which ontological contrasts between di¤erent types of entities and events have been grammaticalized and are subject to mani-

Ergativity and the passive in three Mayan languages

97

pulations by the speakers for discourse-pragmatic reasons; it operates in strategies of backgrounding or omission of the agent, which depend on cultural norms of speaking and referring to others. Thus, rather than being systematically triggered by object promotion and/or agent demotion, the passives analysed in this paper, together with other voices found in these languages, reveal the grammatical sophistication available for the expression of the relationships between participants and their particular levels of involvement in an action, as well as for the subtle control of discourse interaction by speakers.

Abbreviations Ø 1 2 3 a abs1 abs2 ag agp adj ap appl asp b caus chr conj cp cp.intr cp.tr def deic dem det dim

Third person singular set b (abs) first person second person third person set A personal marker (ergative, subject of transitive verbs, possessor of nouns) Ch’orti’ set B su‰xal absolutive personal marker Ch’orti’ set C prefixal absolutive personal marker agent agent prefix j-/hadjectival a‰x antipassive applicative aspect set B personal marker (absolutive, subject of intransitive verbs, non-verbal predicates, object of transitive verbs) causative Ch’orti’ conjunction completive completive intransitive initial tam completive transitive initial deferential deictic demonstrative determiner diminutive

98

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

emph excl exist foc freq gn hab icp icp.intr inan imp interj intr intrzer mid irr nc neg nom nr nr.pas o p part pas pf pl pos posit ppas ppp pr prep prog prr pros refl rel relat rep

emphatic exclamative existential focus frequentative gender habitual, generic, incompletive incompletive incompletive intransitive inanimate imperative interjection intransitive intransitivizer middle voice irrealis numeral classifier negation nominal nominalizer passive nominalizer for transitive verbs syntactic object of transitive verbs patient particle passive perfect plural possessive su‰x positional passive participle perfect passive participle independent pronoun preposition progressive pronominalizer prospective reciprocal/reflexive relative (te) relational marker repetitive

Ergativity and the passive in three Mayan languages

retr rev rs s sbj tam td td1 td2 td3 td4 temp term tr trzer tse yuc

99

retrospective reverential reported speech syntactic subject of transitive verbs subjunctive tense, aspect, mood terminal deictic terminal deictic (proximal/new information) terminal deictic (distal/shared knowledge) terminal deictic (locative/negation) terminal deictic (topic) temporal particle terminative transitive transitivizer Tseltal Yucatec

Texts references cm1 Txt1 Txt2

Cuentos Mayas Kuxulon, Estoy vivo (see Entzin Herna´ndez) Te ame, chanul xux sok te xanich’e; the spider, the wasp and the ant (see Go´mez Ramı´rez).

Examples without reference are taken from the field materials of the authors – narratives or elicited glosses – recorded in the years 1986–1996 for Tseltal, 1994–2007 for Yucatec and 2007–2008 for Ch’orti. Appendix Excerpt from a personal narrative told by an old Yucatec woman in Kopchen, Quintana Roo (‘‘How I killed a puma!’’, 1996, VVA0155-A, 18’25-19’50). (1) Ka h k’uch-en wa`alka’ bey-o’, conj cp arrive-1b run like-td2 And I arrived running like this, t-in-hı`i-pa´ay-t in-ma`achete bey-o’, (2) ka conj cp.tr-1a-drag-extract-trzer 1a-machete like-td2 and I took out my machete

100

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

(3) ka tun ok-en wa`alka ka tun chok’ e pe`ek’ chi’-o’, conj thus enter-1b run conj part cram det dog bite-td2 I started running and then the dogs surged forward to bite it, (4) ka ok-en o`ombre ‘‘che`e`epem!’’ in-ch’ak u-po`ol! conj enter-1b man interj 1a-cut 3a-head And I . . . man! I joined in, ‘‘Cheeepem!’’ I cut its head. (5) Lı´ik’ u-sı`it’ e ko`oh u-xum-p’a`ah-t e pe`ek’-o’! rise 3a-jump det puma 3a-end-leave-trzer det dog td2 The puma jumped to get away from the dogs! (6) Ha`a ka t-u-wa’a-kat u-bah. interj conj cp.tr-3a-stand.up-trzer 3a-refl Oh! And it stood up. (7) La´am-Ø e pe`ek’ tuka’aten Penetrate-3b det dog again The dogs go into it again, (8) le k-u-ka’-ta`al chuk-bi tuka’aten ten e pe`ek’-ob-o’ det icp-3a-rep-come catch-ppas again by det dog-3pl-td2 and it comes and gets caught again by the dogs, [lit. it comes to be caught again by the dogs] (9) ka chu`uch-mach-t-a’ab-Ø ten e pe`ek’-o’ tuka’aten-e’, conj suck-take-trzer-pas.cp-3b by det dog-td2 again-td4 it’s grabbed by the dogs again (10) p’a`ata ka chu’uk-Ø ten e pe`ek’-ob-e’ ko’lel part conj catch.pas-3b by det dog-3pl-td4 woman And it’s caught by the dogs, woman! (11) ka t-u-ka’-chok’-int-ah-Ø ich yo´ok; conj cp.tr-3a-cram-trzer-cp.tr-3b in foot and they run again in between its paws (12) ka u-chuk-Ø e pe`ek’-o’ u-chok’e ti’ conj 3a-catch-3b det dog-td2 3a-cram prep.3pr and the dogs catch it, and they get in, (13) k-u-chi’-chi’-ik-Ø e pe`ek’ ich yo´ok-o’, icp-3a-bite-bite-tr.icp-3b det dog in foot-td2 the dogs keep biting at its paws,

Ergativity and the passive in three Mayan languages

101

(14) ka t-in-ka`alama-ch’ak-t ti’ te’ bey! conj cp.tr-1a-adv-cut-trzer prep.3pr loc like and I cut it strongly like this! (I gave it a great blow with my machete) (15) T-in-ch’ak-ah–Ø ti’ bey-o’. cp.tr-1a-cut-tr.cp.3b prep.3pr like-td2 I cut it there like this! t-uy-o´ot-ah-Ø tun wa’a-kat (16) Ka ti’ale’ ka conj temp conj cp.tr-3a-want-tr.cp-3b thus stand.up-trzer u-bah-e’, 3a-refl-td4 And then it tried to stand up, (17) tun-ta`al u-lah-en ka’achih-a’ le [le le] ko`oh-o’! prog.3a-come 3a-slap-1b temp-td1 det [det det] puma-td2 the puma was coming to claw me! (18) Tun-ta`al u-lah-en! Men-m u-k’a ten bey k-u-ta`al-a’! prog.3a-come 3a-slap-1a! do-pf 3a-hand 1pr like icp-3a-come-td1 It was coming to claw me! It lifted its paw towards me and it’s coming like this! (19) Ka ti’a lel-a’ ten in-mach/ prepara`ado’-en ye´et in-ma`askab-o, conj temp det-td1 1pr 1a-take/ ready-1b with 1a-machete-td2 And then, me, I took/ I was ready with my machete, (20) Ka ti’ale’ Cha`a!’’ ka la’ap’-Ø pa`ach-i ten le pe`ek’-o’, conj temp interj conj catch.pas-3b behind-nom by det dog-td2 And then tcha! It was caught from behind by the dogs (21) sa´am ko`o’Lpay-t-a(’a)k-Ø! retr pull-extract-trzer-pas.sbj-3b It was dragged away! (22) Ko`o’Lpay-t-a’a(b)-Ø ten e pe`ek’-ob-o’, Pull-extract-trzer-pas.cp-3b by det dog-3pl-td2 Dragged away by the dogs, (23) ma’ ch’a’-ab-Ø u-ta`al u-lah-en-e’! neg let-ps.cp-3b 3a-come 3a-slap-1b-td4 they didn’t let it come and claw me!

102

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

(24) Ka ti’al u-ta`al u-lah-en t-u-ka’a-ten conj temp 3a-come 3a-slap-1a prep-3a-two-time And it’s coming to claw me again (25) bey ix u´uch u-ko´o’-pa`ay-t-a(’a), like conj happen 3a-pull-extract-trzer-pas.icp, and the same way it was dragged away, (26) ha´an-pay-t-a’(ab)-Ø, roughly-extract-trzer-pas.cp-3b it is pulled away, (27) ka conj pe`ek’ dog

ti’a’lale’ chok’ e pe`ek’-o’, ka tun cho`ok’ e temp cram det dog-td2 conj part cram det ich yo´ok-o’, in foot-td2

and then the dogs surged, the dogs surged into its paws, (28) ka’ ti’al o`ok-en wa`alkab ‘‘Che`epen che`epen che`epen, kı`ilin!’’ And temp enter-1b run interj. . . . and I started running ‘‘Che`epen che`epen che`epen, kı`ilin!’’, (29) ka lu´ub-ih! conj fall-cp.intr.3b and it fell! (30) Aaa! tun-lu´ub-l tun o’tsil ko`oh tun-e’! interj prog.3a-fall-icp thus poor puma thus-td4 Aaah! And so the poor puma is falling! (31) Ay ko’lel he’ tun k-u-lu´ub-u tun-a’, interj woman ost thus icp-3a-fall-icp thus-td1 Oh woman! There it is that it falls! (32) he’ tun k-u-cho’ le pe`ek’-o’ tun-chi’-ik-a’! ost thus icp-3a-cram det dog-td2 prog.3a-bite-tr.icp-td1 And there are the dogs surging forward, they’re biting it (33) Ta´an a tun u-ch’a’-ab-l u-lı´ik’-i?! prog quest part 3a-let-pas-icp 3a-rise-icp Are they letting it stand up?! [lit. is its standing up being allowed?] (34) Ka’ ti’a’l-e’ sa´am in-he-hen-ch’ak-t u-po`ol! conj temp-td4 retr 1a-red-adv-cut-trzer 3a-head And I already hacked its head!

Ergativity and the passive in three Mayan languages

103

(35) Sa´amih in-kin-s-m-ah-Ø ko`oh bey-o’! retr 1a-die-caus-pf-cp puma like-td2 I’ve already killed the puma!! (36) Ma’ sa’ahk-en-i’! Ma’ sa’ahk-en-i’, chen ch’ak-bi neg afraid-1b-td3 neg afraid-1b-td3 only cut-ppas t-in-me`et-ah-Ø! cp.tr-1a-do-tr.cp-3b I’m not afraid! I’m not afraid! I just did it with my machete! (37) Ch’ak-bi t-in-me`et-ah-Ø! cut-ppas cp.tr-1a-do-tr.cp-3b I did it with my machete!

Topicality measurement We use the quantitative text-based method initially developed by Givo´n (1983) and presented in Zavala (1997). Only the referential distance measure (anaphoric accesibility) is given, since for this short excerpt no significant di¤erence is observed by taking topic persistence into account. The first figure corresponds to the total number of occurrences. In brackets are given the measures of topicality, from 1, highly topical, to >3, less topical. The figure before the colon or the > sign corresponds to the number of occurrences, e.g. 6:1, 2 > 3 is to be read ‘6 occurrences are highly topical, 2 are much less topical’. Two cases of the reflexive have been set aside. Agent of active transitive [A]

Object of active transitive [P]

8 (6:1, 2>3)

4 (3:1, 1:2)

Puma

7 (7:1)

5 (4:1, 1>3)

8 (7:1, 1:2)

Dogs

4 (4:1)

1 (1:3)

1 (1)

Narrator (1st person)

Passive patient [U]

Passive agent (Oblique)

Arg. of movement predicate [U]

11 (9:1, 2:2) 4 (2:1, 1:2, 1>3)

5 (1:1, 2:2–3, 2>3)

104

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

References Aissen, 1992 Aissen, 1997 Aissen, 1999 Aissen, 2003

Judith Topic and focus in Mayan. Language 68: 43–80. Judith On the syntax of obviation. Language 73: 705–50. Judith Agent focus and inverse in Tzotzil. Language 75: 451–485. Judith

Di¤erential Object Marking: Iconicity vs. Economy, Natural Language and Linguistic Theory 21: 435–483. Alvarez Ramı´rez, Juan et al. 2004 U’tirach e Ojroner Maya Ch’orti’. Grama´tica descriptiva Ch’orti’. Ediciones de la Academia de Lenguas Mayas de Guatemala. Blaha Pfeiler Barbara, and Andrew C. Hofling 2006 Apuntes sobre la variacio´n dialectal en el maya yucateco. Penı´nsula. volumen I, nu´mero 1, primavera de 2006: 27–45. Bohnemeyer, Ju¨rgen 2001 Argument and Event Structure in Yukatek Verb Classes. University of Massachusetts Occasional Papers in Linguistics: the semantics of underrepresented languages in the Americas 25: 8–19. Bohnemeyer, Ju¨rgen 2004 Split intransitivity, linking, and lexical representation: the case of Yukatek Maya. Linguistics 42(1): 67–107. Bohnemeyer, Ju¨rgen 2009 Linking without grammatical relations in Yucatec: Alignment, extraction, and control. In Y. Nishina, Y. M. Shin, S. Skopeteas, E. Verhoeven, and J. Helmbrecht (eds.), Issues in functionaltypological linguistics and language theory: A Festschrift for Christian Lehmann on the occasion of his 60th birthday. Berlin: Mouton de Gruyter. 185–214. Branigan, Holly P., Martin J. Pickering, and Mikihiro Tanaka 2008 Contribution of animacy to grammatical function assignment and word order during production, Lingua, 118: 117–189, Elsevier. Bricker, Victoria R. 1978 Antipassive constructions in Yucatec Maya. Papers in Mayan Linguistics, (University of Missouri Miscellaneous Publications in Anthropology N 6, Studies in Mayan Linguistics N 2 N. England (ed), Columbia, Missouri, 3–24. Bricker, Victoria R. 1979 WH-questions, relativization and clefting in Yucatec Maya. Papers in Mayan linguistics, L. Martin (ed.), Lucas Bas Brothers, Columbia, Missouri, 107–36. Bricker, Victoria R. 1981 The Source of the ergative split in Yucatec Maya. Journal of Mayan Linguistics, vol. 2, n 2: 83–127, Spring 1981.

Ergativity and the passive in three Mayan languages

105

Bricker, Victoria, Eleuterio Po’ot Yah and Ofelia Dzul de Po’ot 1998 A Dictionary of the Maya language as spoken in Hocaba´, Yucata´n. University of UTAH Press, Salt Lake City. Brown, Penelope 1997 Isolating the CVC root in Tzeltal Mayan: a study of children’s first verbs. In E.V. Clark, ed., Proceedings of the 28th Annual Child Language Research Forum, 41–52. Stanford, CA: CSLI/ University of Chicago Press. Brown, Penelope 1998 Early Tzeltal verbs: argument structure and argument representation. In E.V. Clark, ed. Proceedings of the 29th Annual Child Language Research Forum, 129–140. Stanford, CA: CSLI/ University of Chicago Press. Brown, Penelope and Stephen C. Levinson 1987 Politeness. Some universals in language usage. Cambridge University Press, Cambridge. Comrie, Bernard 1981 Language universals and linguistic typology: Syntax and morphology. Chicago: University of Chicago Press. Cooreman, A. M. 1987 Transitivity and discourse continuity in Chamorro narratives. Mouton de Gruyter. Creissels, Denis 2006 Syntaxe ge´ne´rale une introduction typologique 2 : la phrase, 2/2, Paris, Herme`s – Lavoisier. Danziger, Eve 1996 Split Intransitivity and Active-Inactive Patterning in Mopan. International Journal of American Linguistics 62(4): 379–414. Dayley, Jon 1981 Voice and ergativity in Mayan Languages. Journal of Mayan Linguistics Linguistics 2: 6–82. Dayley, Jon 1990 Voz y ergatividad en idiomas mayas Lecturas sobre la lingu¨´ıstica maya, Nora England y Stephen Elliott eds, CIRMA: 335–399. Dixon, R.M.W. 1972 The Dyirbal Language of North Queensland. Cambridge: Cambridge University Press. Dixon, R.M.W. 1994 Ergativity. Cambridge: Cambridge University Press. Dryer, M.S. 1986 Primary Objects, Secondary Objects, and Antidative. Language 62(4): 808–845. England, Nora C. 1983 Ergativity in Mamean (Mayan) Languages. International Journal of American Linguistics 49: 1–19.

106

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

England, Nora C. 1988 Mam Voice. In Passive and Voice. Ed. M. Shibatani. Amsterdam/ Philadelphia: John Benjamins Publishing Company. 523–545. England, Nora C. 1991 Changes in Basic Word Order in Mayan Languages. International Journal of American Linguistics 57(4): 446–86. Entzin Herna´ndez, C. 2007 Kuxulon; Estoy vivo. In El bolom dice . . . venimos a este lugar. Antologı´a de cuentos, Centro Estatal de Lenguas, Arte y Literatura Indı´genas, Tuxtla Gutie´rrez, 13–20. Fought, John G. 1967 Chorti (Mayan): Phonology, Morphonemics, and Morphology. Yale University, Department of Anthropology New Haven. Doctoral dissertation. Fought, John G. 1972 Chorti (Mayan) texts. Edited by Sara S. Fought, Philadelphia: University of Pennsylvania Press. Fried, Mirjam 2006 ‘‘Agent back-grounding as a functional domain: Reflexivization and passivization in Czech and Russian.’’ In B. Lyngfelt and T. Solstad, Demoting the Agent. Amsterdam: John Benjamins, 83–110. Givo´n, Talmy (ed.) 1983 Topic Continuity in Discourse: a Quantitative Cross-Language Study. Amsterdam: John Benjamins. Givo´n, Talmy (ed.) 1994 Voice and Inversion. Amsterdam: John Benjamins. Go´mez Ramı´rez, Martı´n 2001 Te amej, chanul xux sok te xanich’e; La aran˜a, la avispa y la hormiga. In Jich’ sk’op balam. Seleccio´n de cuentos tzeltales. Biblioteca Popular de Chiapas coneculta, Tuxtla Gutie´rrez, 104–120. Go´mez Ramı´rez, Martı´n 2007 Te jilinem a’teleletike; Los trabajadores incoformes. In El bolom dice . . . venimos a este lugar. Antologı´a de cuentos, Centro Estatal de Lenguas, Arte y Literatura Indı´genas, Tuxtla Gutie´rrez, 68–85. Guillaume, Antoine 2006 The hierarchical agreement system with inverse e¤ect of Reyesano. Laboratoire Dynamique du Langage, CNRS/University Lyon 2. Ms. 31 p. Guillaume, Antoine 2009 Hierarchical agreement and split intransitivity in Reyesano, International Journal of American Linguistics, 75: 1, pp. 29–48. Gutie´rrez Bravo, Rodrigo y Jorge Monforte 2008 La alternancia sujeto inicial/verbo inicial y la teorı´a de la optimidad. In Teorı´a de la optimidad: estudios de sintaxis y fonologı´a.

Ergativity and the passive in three Mayan languages

107

Eds. Rodrigo Gutie´rrez Bravo y Esther Herrera Zendejas. El Colegio de Me´xico-Centro de Estudios Lingu¨ı´sticos y LiterariosLaboratorio de Estudios Fo´nicos, Me´xico, 2008, 61–90. Gutie´rrez Bravo, Rodrigo y Jorge Monforte 2009 Focus, agent focus and relative clauses in Yucatec Maya. in New Perspectives on Mayan Linguistics, H. Avelino, J. Coon, and E. Norcli¤e (eds.), MIT Working Papers in Linguistics. Gutie´rrez Sa´nchez, Pedro and Roberto Zavala 2005 Chol and Chontal: Two Mayan languages of the agentive type. Paper presented at the conference The Typology of stative-active languages, Leipzig May 2005. Haviland, John B. 1981 Sk’op sotz’leb. El tzotzil de San Lorenzo Zinacanta´n. Universidad Nacional Auto´noma de Me´xico, Me´xico. Hinma´n Smith, Joshua 2004 Manual of Spoken Tzeltal, translated into English by Stuart P. Robinson, Madelief Bertens ed. Jacobsen, William 1985 ‘‘The Analog of the Passive Transformation in Ergative-Type Languages.’’ In Nichols Johanna and Anthony C. Woodbury (eds). 1985. Grammar Inside and Outside the Clause: Some Approaches to Theory from the Field. Cambridge: Cambridge University Press, 176–191. Kaufman, Terence 1971 Tzeltal Phonology and Morphology. University of California Publications in Linguistics, Berkeley, CA; 61. Berkeley and Los Angeles. Kaufman, Terence 1990 Algunos rasgos estructurales de los idiomas Mayances con referencia especial al K’iche. In Nora England and Stephen R. Elliott (eds). Lecturas sobre la lingu¨´ıstica maya, 59–114. La Antigua Guatemala: CIRMA. Kra¨mer, Martin and Dierter Wunderlich 1999 Transitivity alternations in Yucatec and the correlation between aspect and argument roles. Linguistics 37–3, 431–479. Larsen, Tomas W., and William M. Norman 1979 Correlates of ergativity in Mayan grammar. In Frans Plank (ed.) Ergativity: Towards a theory of grammatical relations, 347–370. London/New York: Academic Press. Lehmann, Christian 1990 Yukatekisch. Zeitschrift fu¨r Sprachwissenschaft 9(1/2): 28–51. Lehmann, Christian 1993 Predicate classes in Yucatec Maya. Funcio´n 13–14: 195–272. Lehmann, Christian 2002 ‘‘Possession in Yucatec Maya.’’ ASSIDUE Arbeitspapiere des Seminars fu¨r Sprachwissenschaft der Universita¨t Erfurt. Nr. 10.

108

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

Lehmann, Christian and Elisabeth Verhoeven 2005 Extraversive transitivization in passive Maya and the nature of the applicative. In L. Kulikov, A. Malchukov and P. de Swart (eds) Case, Valency and transitivity, 465–493. Benjamins, Amsterdam. Lois, Ximena and Valentina Vapnarsky 2003 Polyvalence of lexical roots in Yukatekan Mayan Languages. Lincom Europa, Mu¨nchen. Lois, Ximena and Valentina Vapnarsky 2006 Root indeterminacy and polyvalence in Mayan languages. Lexical Categories and Root Classes in Amerindian languages. X. Lois and V. Vapnarsky (eds), Peter Lang, Bern, 69–115. Lois, Ximena and Valentina Vapnarsky 2009 Morphosyntactic and semantic issues on polycategoriality in Yucatecan Mayan languages. Paper presented at the ALT8 meeting, Berkeley, University of California. July 23–27 2009. Lois, Ximena, Valentina Vapnarsky, Ce´dric Becquey et Aurore Monod Becquelin 2010 Polycategoriality across Mayan languages: Action nouns and ergative splits. Paper presented at the International Conference on Polycategoriality. Cross-linguistic, cross-theoretical and language acquisition approaches. Ecole Normale Supe´rieure, Paris, Oct. 2010. McQuown, Norman 1967 ‘‘Colonial Yucatec.’’ In Handbook of Middle American Indians 5: 201–247. Austin: University of Texas Press. Monod-Becquelin, Aurore 1997 Parlons tzeltal: une langue maya du Mexique. Paris: L’Harmattan. Ola, Olanike, and Victoria Bricker 2000 Placeless and historical laryngeals in Yucatec Maya. In International journal of American linguistics vol. 66, no. 3: 283–317. Pe´rez Martı´nez, Vitalino 1994 Grama´tica del idioma ch’ortı´’. Antigua, Guatemala: Proyecto Lingu¨´ıstico Francisco Marroquı´n. Polian, Gilles 1999 Morpho-syntaxe du verbe tzeltal. Me´moire de DEA, Universite´ Paris III. Polian, Gilles 2003 Sustantivos verbales e infinitivos en tzeltal. Paper presented at the Quinto Congreso de Estudios Mayas. Guatemala Ciudad. Polian, Gilles 2005 Dina´mica de la oracio´n tzeltal. La topicalidad como un factor determinante del orden lineal. Trace 47: 30–45. Polian, Gilles 2006 Ele´ments de Grammaire du tseltal, L’Harmattan.

Ergativity and the passive in three Mayan languages Polian, Gilles to appear Quizar, Robin 1994

109

Grama´tica del tseltal. CIESAS, Mexico.

Split Ergativity and Word Order in Ch’orti’, International Journal of American Linguistics, Vol. 60, No. 2 (Apr., 1994): 120–138. Quizar, Robin and Susan M. Knowles-Berry 1988 Ergativity in the Cholan Langages. International Journal of American Linguistics 54, 1: 73–95. Robinson, Stuart 1999 Voice and obviation in Greater Tzeltalan. Australian National University. Ruz, Mario Humberto (ed.) 1989 Las Lenguas del Chiapas Colonial, Universidad Auto´noma de Me´xico, Me´xico. Sa´nchez Go´mez F.J. et al. 2003 Sk’op bats’il winiketik; Grama´tica ba´sica de la lengua Tzeltal. Unidad de escritores Mayas-Zoques. San Cristobal Las Casas. Shibatani, Masayochi 1985 Passives and related constructions: a prototype analysis. Language, 61, 4: 821–848. Shklovsky, Kirill 2005 Person marking in Petalcingo Tzeltal. Thesis, Reed College. Skopeteas, Stavros, and Elisabeth Verhoeven 2005 Postverbal argument order in Yucatec Maya. Sprachtypologie und Universalienforschung 53(1), 71–79. Slocum, Marianna C., Florence L. Gerdel and Manuel Cruz Aguilar 1999 Diccionario tzeltal de Bachajo´n, Chiapas, Instituto Lingu¨istico de Verano, Me´xico. Smith-Stark, Thomas 1978 The Mayan antipassive: some facts and fictions. Papers in Mayan Linguistics, N. England ed. 169–187. University of Missouri Miscellaneous Publications in Anthropology and Studies in Mayan Linguistics, Columbia, University of Missouri. Stross, Brian 1978 Tzeltal tales of Demons and Monsters. Museum of Anthropology, University of Missouri-Columbia (Columbia), Museum brief; no. 24. Thompson, Chad 1994 Passive and inverse constructions. Voice and inversion, Talmy Givo´n (ed.) Amsterdam: John Benjamins, 47–63. Tonhauser, Judith In Press Agent Focus and Voice in Yucatec Maya. Proceedings of the 39th Meeting of the Chiacgo Linguistic Society. Van Valin, Robert D. 1981 Grammatical Relations in Ergative Languages. Studies in Language, 5:3.

110

Valentina Vapnarsky, Ce´dric Becquey, and Aurore Monod Becquelin

Van Valin, Robert D., and Randy J. La Polla 1997 Syntax: structure, meaning, and function. Cambridge: Cambridge University Press. Vapnarsky, Valentina in prep. Le passif peut-il e´clairer les esprits? agentivite´, interactions et esprits-maıˆtres en maya yucate`que, Ateliers d’Anthropologie (http://ateliers.revues.org). Verhoeven, Elisabeth 2007 Experiential constructions in Yucatec Maya. John Benjamins Pub. Company, Studies in Language Companion series 87. Zavala Maldonado, Roberto 1997 Functional analysis of Akatek voice constructions. International Journal of American Linguistics, 63, 4: 439–457. Zavala Maldonado, Roberto 2007 The agentive alignment pattern in Cholan and K’ichean (Maya). Paper presented at the AALLED conference, Lyon, February 2007.

A tale of two passives in Cavinen˜a Antoine Guillaume 1. Introduction1 This paper consists of a detailed analysis of two verbal su‰xes with passive meanings in Cavinen˜a, an ergative language from the Tacanan family spoken in Amazonian Lowland Bolivia. These markers, -tana and -ta, are interesting for a number of reasons. First, it has been argued that passive constructions are rare – or even absent – in Amazonia (Derbyshire 1987: 321, Payne 1990: 3, 2001: 596). However, recent surveys like that by Siewierska (2005) in the World Atlas of Language Structure do mention quite a few in the area. It is also sometimes thought that passives are normally features of accusative languages, not of ergative ones. Yet, as pointed out by authors like Lazard (1986: 9), Dixon (1994: 149¤.), Keenan and Dryer (2007: 359), a fair number of ergative languages are known to display passive constructions. This paper contributes to the idea that passives might not be that exceptional in Amazonia and in ergative languages in general. Second, because of the lack of detailed studies on Tacanan languages, the existence of passives in this family has so far not been very clear. For example, in the WALS article mentioned above, Cavinen˜a is said to lack this derivation while Araona, a related Tacanan language, is said to have 1. The research presented in this paper is based, for the most part, on first hand data that I collected from Cavinen˜a native speakers in traditional communities through 15 months of fieldwork between 1996 and 2003 (complemented by material published by two SIL missionaries, E. Camp and M. Liccardi). I would like to thank the Cavinen˜a people for their generous hospitality and interest in documenting and describing their language. The ideas developed in this paper benefited from comments made by the participants of the workshop on ‘‘Passif en contexte ergatif ’’ organized by F. Queixalo´s in Vuillejuif on May 25th, 2007, within the project ‘‘Ergativite´ ’’ of the Fe´de´ration de Typologie et Universaux Linguistiques. The preparation of this article was also improved by comments from Denis Creissels, Rene´ Lacroix, Francesc Queixalo´s and Franc¸oise Rose, as well as from an external reviewer (Anna Siewierska) and from the editors Gilles Authier and Katharina Haude.

112

Antoine Guillaume

it which, to my knowledge, is not the case. This paper will serve to clarify the situation. Third, the distinction between the passive markers -tana and -ta is a very slim one. In earlier work (Guillaume 2004, 2008), I treated -tana and -ta as two free variants of the same morpheme. This paper is a major revision of this initial analysis. By looking more closely at their respective semantics, distribution and use, I show that they di¤er in terms of productivity and ability to express or not anticausative semantics, and I claim that -tana and -ta are distinct morphemes. The observations I make lead me to propose several historical scenarios for the evolution of the two markers. The paper is organized as follows. In §2, I provide an overview of the main characteristics of the Cavinen˜a ergative argument-coding system. Particular attention is paid to the sensitivity of the system to the category of transitivity, and to the fact that the ergative patterning is restricted to the morphological coding domain, not extending to the level of behaviourand-control properties, in other words that there is no ergative ‘‘pivot’’ in the language. In §3, I introduce and discuss the two passive derivations, realized by the verbal su‰xes -tana and -ta, in terms of their syntactic, semantic, and distributional properties. Section 4 takes a diachronic perspective. Building on another study that traced back the origin of the marker -ta as a third person plural marker (Guillaume 2011), I here argue for the development of a productive marker -tana either because of the loss of productivity of -ta or because of the semantic overlap between the two.

2. The Cavinen˜a argument-coding system Cavinen˜a is an Amazonian language from the Tacanan family spoken in Northern Bolivia by approximately 1200 speakers who are also all bilingual in Spanish. Cavinen˜a encodes the grammatical functions of its core arguments by way of case-marking and cross-referencing. Both systems manifest ergative alignment.2 The transitive agent/subject (henceforth A) NP is marked by the enclitic ¼ra, as with iba ‘jaguar’ in (1a). By contrast, the transitive patient/object (henceforth O) NP is unmarked, as with takure

2. This section only provides a brief summary of the argument-coding system. For more details, see Guillaume (2006, 2008, 2010).

A tale of two passives in Cavinen˜a

113

‘chicken’ in (1a), similar to the intransitive unique argument (henceforth S), as with iba ‘jaguar’ in (1b).3 (1) a.

Transitive clause Iba¼ra ¼tu 4 iye-chine takure. jaguar¼erg ¼3sg.abs kill-rec.past chicken(abs) ‘The jaguar killed the chicken.’ (elicited)

b.

Intransitive clause [Tuke tupuju] ¼tu iba tsajaja-chine. 3sg following ¼3sg.abs jaguar(abs) run-rec.past ‘The jaguar ran behind him.’ (from text)

The ordering of the core arguments is free and does not play any role in the disambiguation of grammatical functions. There is an argument crossreferencing system realized by bound pronominal clitics in second position in the clause (not on the verb). These clitics specify the person, the number and the grammatical function of the core arguments in the same ergative fashion as NPs and independent pronouns. Neither core argument NPs and independent pronouns, nor bound pronouns are obligatory. Free and bound forms can co-occur (representing the same argument), as in (2), or only one of them can be present, i.e. the free form, as in (3), or the bound form, as in (4). When referring to a third person singular, neither a free nor bound form is obligatory, as shown in (5); with other person and number combinations, however, there needs to be some expression of the relevant argument. (2) co-occurrence of free and bound forms a. Ejetupu ¼tuke5 ¼mikwana e-a-u pista? when ¼3sg.abs ¼2pl.erg pot-a¤ect-pot airstrip(abs) ‘When could you make (lit. a¤ect) the airstrip?’ (from text) 3. Although not illustrated here, the expression of core arguments by way of independent pronouns follows the same ergative patterning (see Guillaume 2006, 2008, 2010). 4. Enclitics preceded by a space are second position enclitics, as discussed further below. 5. Second position bound pronouns can be subject to a morpho-phonological rule that modifies their segmental make-up. For example, this rule is responsible for the fact that the third person singular absolutive marker is ¼tu in (1a,b) while it is ¼tuke in (2a); see full details in Guillaume (2006, 2008, 2010).

114

Antoine Guillaume

b.

Jadya ¼tura ¼Ø a-kware bari¼ra. thus ¼3sg.erg ¼1sg.abs a¤ect-rem.past giant_anteater¼erg ‘This is what the giant anteater did to (lit. a¤ected) me.’ (from text)

(3) only free forms E-wane¼ke¼ra amena ba-ti-kware tuke. 3-wife¼3¼erg bm see-go-rem.past 3sg.abs ‘His wife went to see him.’ (from text) (4) only bound forms Enapa-wa ¼taa ¼tunara ¼ike. cry_for-perf ¼emph ¼3pl.erg ¼1sg.abs ‘They (my dogs) cried for me.’ (from text) (5) neither free nor bound form (only for third person singular) Ka-bajeje-ti-tsu shana-nuka-kware. ref/recip-prepare-ref/recip-ss leave-reitr-rem.past ‘He prepared himself and left it (a viper) again.’ (from text) In Cavinen˜a, the category of transitivity is entrenched at various levels of structure, from lexicon to morphosyntax, at least at the intra-clausal level. Starting with morphosyntax, the ergative marking on the A argument (and the absence of marking of the S and O arguments) is obligatory in all (syntactic, semantic or pragmatic) contexts. Within the predicate, several su‰xes have ‘‘transitivity harmony’’ restrictions, having two allomorphs, the selection of which depends on the transitivity of the verb stem they attach to. This is the case, for example, with the completive su‰x, which shows up as -tere on intransitive verb stems, as in (6a), and -tirya on transitive verb stems, as in (6b) (Guillaume 2008: 191 ¤.). (6) a.

b.

Tiru-tere-wa [ekwe budari ]. burn-comp.itr-perf 1sg.gen banana ‘My banana burnt completely.’ (elicited) [Iyuka biti ] ¼tu yupu-tirya-kware iba¼ra. head skin ¼3sg.erg tear_o¤-comp.tr-rem.past jaguar¼erg ‘The jaguar tore o¤ his whole scalp.’ (from text)

A tale of two passives in Cavinen˜a

115

At the lexicon level, the rule is for all verbal lexemes to be either intransitive or transitive; I could only find two ambitransitive verbs (kike-/keke-6 ‘shout, shout at O’ and kwina- ‘give birth, give birth to O’). One consequence of this is that there is a functional need for this language to have overt valency-changing mechanisms for playing with transitivity, which it indeed has. There are two passive markers (to be discussed in detail further below), a reflexive/reciprocal marker, two antipassive mechanisms (involving reduplication or the exchange of auxiliaries) and three causative markers (see Guillaume 2008: ch. 8). At the inter-clausal level, the transitivity distinction is somewhat more relaxed. In particular, the co-reference restrictions that exist between certain dependent clauses and their controlling matrix clause operate accusatively, i.e. treating the S and the A identically and in opposition to the O. This is exemplified in (7) with the temporal dependent clause, the verb of which is marked by the su‰x -(a)tsu and which requires its subject (S or A) to be co-referential with the subject (S or A) of the matrix clause. (7) a.

Dependent S ¼ Matrix S Tudya ¼tatse amena [kwaba¼ju ani-bute-tsu] then ¼3dl.abs bm canoe¼loc sit-go.down-ss tsura-kware. go_up-rem.past ‘Then they (dl) sat down in their canoe and went up(river).’ (from text)

b.

Dependent A ¼ Matrix A Tudya ¼tuke ¼Ø [imeta-tsu] mare-kware. then ¼3sg.abs ¼1sg.erg point_at-ss shoot_at-rem.past ‘Then I pointed (my rifle) at it (a peccary) and I shot at it.’ (from text)

c. Dependent S ¼ Matrix A [Babi¼ra kwa-atsu] ¼tuja ¼tu hunt¼purp.mot go-ss ¼3sg.dat ¼3sg.erg tsuru-kware [ peadya matuja]. encounter-rem.past one cayman ‘When hei went hunting, hei met a caiman.’ (from text) 6. The alternate forms kike and keke are in free variation and have nothing to do with transitivity.

116

Antoine Guillaume

d.

Dependent A ¼ Matrix S Bajida¼jipenee ¼Ø ju-kware [tuke peta-tsu]. scared¼almost ¼1sg.abs be-rem.past 3sg.abs look_at-ss ‘I was a bit scared, looking at it (something moving among the manioc leaves).’ (from text)

Having laid down the main characteristics of the Cavinen˜a argumentcoding system, we will now turn to the passive derivations in this language. 3. The passives -tana and -ta Cavinen˜a has two verbal markers with core passive meanings, -tana and -ta. Apart from the fact that they are segmentally very close – they both contain the syllable ta – their syntactic and semantic e¤ects are very similar. A semantic distinction between these two markers was postulated by Camp and Liccardi in their grammatical sketch (1989: 315). However, their lack of argumentation7 and my inability at the time to determine any obvious di¤erences made me treat them as free variants of the same morpheme in Guillaume (2004, 2008). The goal of this paper is to revisit my preliminary analysis. By looking more closely at the semantics, use and distribution of these su‰xes in texts, I argue that -tana and -ta are distinct morphemes and that their respective characteristics are as follows:8 Marker -tana: – distribution: transitive verbs only – syntactic e¤ect: reduction of the verb valency by one – semantic e¤ect: (1) passive and (2) anticausative 9 – discourse use: major use in depiction of generic customary practices and procedures – productivity: full 7. The authors say that with -tana the agent is ‘‘optional’’ whereas with -ta the agent is ‘‘unknown’’. However, since they do not provide any clear illustrative example or further discuss their claim, it is completely unclear what exactly they had in mind. 8. With these properties, both markers fulfill the criteria for a ‘‘basic passive’’, according to Keenan (1985) and Keenan and Dryer (2007). 9. In this study, I use ‘‘passive’’ and ‘‘anticausative’’ in the following senses, following Creissels (2006: 9–10, 31, 35–37). Both are mechanisms that apply to a transitive verb and produce a derived intransitive form. In the case of a

A tale of two passives in Cavinen˜a

117

Marker -ta: – distribution: identical to -tana – syntactic e¤ect: identical to -tana – semantic e¤ect: passive only – discourse use: major use for specific events in narratives; not used for depicting generic customary practices and procedures – productivity: restricted These characteristics are discussed, refined and exemplified in §3.1 (with -tana) and §3.2 (with -ta), and summarized in §3.3. 3.1. The passive and anticausative -tana Formally, -tana attaches to transitive verbs and reduces their valence by one. The semantic correlates of this process are either passive or anticausative. When they are passive, the event involves and agent (notional A) acting on a patient (notional O expressed as S), but the identity of the agent is backgrounded, and the agent must be left unexpressed. This is illustrated with a pair of elicited examples in (8) and a pair of text examples in (9). (8) a.

b.

Active Roberto¼ra akwi abu-ya. Roberto¼erg tree(abs) carry-impfv ‘Roberto carries the tree.’ (elicited) Passive Akwi abu-tana-ya. tree(abs) carry-pass1-impfv ‘The tree is carried.’ (elicited)

passive, the subject receives exactly the same semantic role as the object of the transitive construction, the participation of the agent of the transitive construction in the event being always implied, whether it is overtly expressed (by an oblique) or not; e.g. ‘the eggs were broken (by the child)’. In the case of an anticausative (de´causatif in Creissels’ terms), the subject undergoes the same process as the object of the transitive construction but unlike in a passive, this process is interpreted as more or less spontaneous; e.g. ‘the eggs broke’. In this sense, anticausative is one of the various possible meanings of the more general category ‘‘middle’’, which is characterized by the fact that the subject does not receive exactly the same role as the object of the transitive construction – the other possible meanings of the category of ‘‘middle’’ are reflexive, reciprocal, autocausative and autobenefactive.

118

Antoine Guillaume

(9) a.

b.

Active Irare¼kwita¼dya¼jutidya ¼tuna tsume-kware. plaited_sieve¼restr¼foc¼disemph ¼3pl.erg use-rem.past ‘(In old times, our Cavinen˜a ancestors) would only use plaited sieves (for sieving corn).’ (from text) Passive Jadya¼tibu ¼tu [tumeke emajaka¼ju] thus¼reason ¼3sg.abs that place¼loc [ jeeke karetu] tsume-tana-ya. . . this cart(abs) use-pass1-impfv ‘(In Cavinen˜a communities there aren’t any motorized vehicles.) For that reason, this cart is used in these places.’ (from text)

In anticipation to the discussion of the di¤erences between -tana and -ta, it is worth mentioning that -tana in its passive sense can be used for depicting both actual and generic events. By actual (or referential), I mean actions with a specific agent acting on a specific patient at a particular place and time. Textual examples of this use can be found in (10). (10) a.

b.

Dutya ekana tibarirya-tana-wa. all 3pl encircle-pass1-perf ‘They were (already) all encircled.’ (from text) Ne-diru-ra¼dya! Tirya-tana-ya¼dya ¼ekwana. hort.pl-go-hort.pl¼foc finish-pass1-impfv¼foc ¼1pl.abs ‘Let’s go! (Otherwise) we will all be killed (lit. finished).’ (from text)

c. A-tana-wa ¼ekwe [ekwe e-bui¼ekatse]! a¤ect-pass1-perf ¼1sg.dat 1sg.gen 1-nephew¼dl(abs) ‘My two nephews got killed (lit. a¤ected)!’ (from text) By generic events, I mean actions with non-specific/impersonal agents and patients as used for the expression of customary practices and procedures. A good example of this use can be found in (9) above, where the speaker explains that traditional carts (rather that motorized vehicles) are normally used to travel in the places where he lives. In (11) below, the speaker explains the traditional way of making a canoe.

A tale of two passives in Cavinen˜a

(11) a.

b.

119

Ikwene¼dya ¼tu e-duku¼ju rure-tana-ya. first¼foc ¼3sg.abs npf-inside¼loc carve-pass1-impfv ‘First, (the canoe) is carved on the inside.’ (from text) Amena tuekedya ¼tu chamakama e-pere bm then ¼3sg.abs finally npf-side(abs) a-tana-ya amena. a¤ect-pass1-impfv bm ‘And then finally the (canoe’s) side is made (lit. a¤ected).’ (from text)

When the semantic interpretation of the application of -tana to a transitive verb is anticausative, the event is not carried out by any agent (notional A) but instead by the patient (notional O expressed as S) itself spontaneously. This is illustrated in (12).10 (12) a.

b.

[Ekwe karusune iyakwake utsa-wa¼ju] 1sg.gen pants new(abs) wash-perf¼ds dyuru-tana-chine. shorten-pass1-rec.past ‘(The women) washed my new pants and they shrunk.’ (elicited) Amena ¼tuke ¼ekwe arusu bm ¼3sg.abs ¼1sg.dat rice(abs) puku-tana-bare-ya. . . crack-pass1-distr-impfv ‘My rice is cracking open.’ (Camp and Liccardi 1989: 20)

It is unclear to what extent the distinction between the passive or anticausative interpretations is linked to the semantic type of verbs. In the available corpus, I have not found any clear example of a verb showing both readings. However, I would assume that -tana can potentially have a passive reading with any transitive verb and that with some of these verbs it can also have an anticausative interpretation, as long as they express events that can occur spontaneously. More work is needed on this topic to confirm this hypothesis.

10. Note that -tana does not yield autocausative interpretations. These are expressed by the reflexive/reciprocal marker k(a)- . . . -ti (ex. utsa- ‘wash’, k-utsa-ti- ‘wash oneself ’); see Guillaume (2008: 270).

120

Antoine Guillaume

The su‰x -tana is a clear detransitivizer. There are several pieces of evidence for this claim. First, as I said, the agent (notional A) cannot be overtly expressed: there is no single passive example in the whole corpus where this happens and in elicitation, my attempts at expressing the notional A (by way of di¤erent kinds of oblique phrases) were systematically rejected. Second, passivized verbs can only bear the intransitive allomorph of those markers which harmonize with the transitivity of the verb stem they attach to; see the discussion of this phenomenon in §2. Accordingly, a passivized verb can only select the intransitive completive allomorph -tere, as in (13); the transitive completive allomorph -tirya would be ungrammatical. (13) Dutya ekana iye-tana-tere-wa. all 3pl kill-pass1-comp.itr-perf ‘They were all killed, to the last one.’ (from text) Third, at least when the interpretation of -tana is anticausative, there is evidence that the notional O (expressed as S) has behaviour-and-control subject properties. In particular, in sentences involving clauses with subjectto-subject restrictions, the notional O of the verb marked by -tana counts as co-referential with the subject of the other clause, as shown in (14).11 (14) a.

b.

[Ebakwapiji riwi-tsu] tuku rake-tana-wa. child(abs) fall-ss onom break-pass1-perf ‘The child fell, ‘‘tuku!’’, and (some part of his body) broke.’ (elicited) . . . kunu¼ju [ekwe e-wachi¼ekatse] karya-tana-tsu liana¼loc 1sg.gen npf-foot¼dl(abs) hook-pass1-ss ike riwi-kware. 1sg.abs fall-rem.past ‘(As I was running after the coati,) my feet got caught (lit. hooked) into lianas and I fell down.’ (from text)

11. In some cases, co-reference is between a part and its whole, in which case one might say that co-reference is not strict. However, this type of co-reference is possible between non-passivized clauses, so this is not a problem for the present claim.

A tale of two passives in Cavinen˜a

121

c. [Estakilla wika-tana-tsu] ¼tu pakaka-ya wooden_peg(abs) extract-pass1-ss ¼3sg.abs fall-impfv karetu. cart(abs) ‘When the peg (of the cart) went out (lit. extracted), the cart fell down.’ (Camp and Liccardi 1989: 56) Before turning to -ta, let us add that -tana is probably fully productive. This is suggested by the fact that in texts it is used fairly frequently and with a wide range of transitive verbs, and in elicitation, its use was never considered ungrammatical. 3.2. The passive -ta Similar to -tana, -ta applies to transitive verbs12 and reduces the valency by one. The semantic correlates are exclusively passive. As will be discussed further below, unlike -tana, -ta is not a productive marker. (15) a.

b.

A-ta-nuka¼dya ekwana. a¤ect-pass2-reitr¼foc 1pl.abs ‘We will be attacked (lit. a¤ected).’ (from text) Jadya tirya-ta-wa¼ju. . . thus finish-pass2-perf-ds ‘Having been exterminated (lit. finished) that way (they decided to go live somewhere else).’ (from text)

There are tests that indicate that the construction is intransitive. First, it is impossible to express the notional A, whether by an ergative NP, as in the b-example in (16), or by an oblique NP. (16) a.

b.

[Peadya sen˜ora] ¼tu kweja-ta-ya. one woman ¼3sg.abs inform-pass2-impfv ‘A woman is being informed.’ (elicited) *[Peadya sen˜ora¼ra] ¼tu kweja-ta-ya. one woman¼erg ¼3sg.abs inform-pass2-impfv (he was informed by a woman) (elicited)

12. The su‰x -ta is found on one intransitive verb, maju- ‘die’, in which it has an indefinite meaning ‘someone’; see Guillaume (2011).

122

Antoine Guillaume

Second, if the completive su‰x is used on a verb marked by -ta, it is the intransitive allomorph, -tere, that shows up, not the transitive one, -tiriya (see §2), as illustrated in (17). (17) Dutya ekana iye-ta-tere-wa. all(abs) 3pl(abs) kill-pass2-comp.itr-perf ‘They were all killed.’ (elicited) Regarding subject properties, I do not have enough material to be able to claim with confidence whether the notional O has behaviour-and-control subject properties or not. In the only example available of a verb marked by -ta in a sentence with obligatory subject-to-subject co-reference between its clauses, reproduced in (18), the co-referential argument is the notional A of the passive, not the notional O. (18) [[Ekwanaja e-bakani¼kwana] waraji¼kwana¼keja duju-tsu] 1pl.gen npf-name¼pl(abs) authority¼pl¼loc.gnl take-ss ekwana tsume-ta-ya. 1pl.abs use-pass2-impfv ‘(The land owners) give (lit. take) our names to the authorities and use us.’ (spontaneous, given during elicitation session) This example suggests that the notional O does not have subject behaviourand-control properties. More examples of this type are needed, in particular examples from texts, in order to know with certainty that this is indeed the case. Semantically, -ta seems to be always used in a passive sense; that is, unlike -tana, it is never found with anticausative meanings.13 In addition, in all the textual examples available, verbs marked by -ta only refer to actual (referential) events, that is, actions with a specific agent acting on a specific patient at a particular place and time. Generic events, with non-specific/impersonal agents and patients cannot normally be expressed by a verb marked by -ta,14 as shown by the ungrammaticality of the b-example in the following pair:15

13. Note that no elicitation was conducted on this issue. 14. Unless it is adjectivized with the resultative prefix e-; see further below. 15. It is unclear whether this example would be acceptable with a di¤erent context, like ‘a fan is made/will be made of chonta palm (by someone in particular)’. My impression is that it would not, since the collocation of -ta on a- ‘a¤ect’ only results in the meaning ‘kill, destroy’ in the data.

A tale of two passives in Cavinen˜a

(19) a.

b.

123

Epiki ¼tu a-tana-ya abari 16. fan(abs) ¼3sg.abs a¤ect-pass1-impfv chonta_palm ‘Fans are (traditionally) made (a¤ected) of chonta palm.’ (elicited) *Epiki ¼tu a-ta-ya abari. fan(abs) ¼3sg.abs a¤ect-pass2-impfv chonta_palm (elicited)

In other words, semantically, the two markers only overlap in their capacity to express actual passive events. The su‰x -tana has a much broader semantic scope than -ta in its possibility to express situations with no agent (anticausative) and generic events. Another major di¤erence between -tana and -ta concerns their respective productivity. Unlike -tana, -ta manifests many symptoms of a nonproductive marker. In the available texts, it is only found very infrequently and only with a very limited number of roots, actually limited to a- ‘a¤ect’ and tirya- ‘finish, terminate’, as in (15) above. While testing -tana and -ta on other verbs, speakers are sometimes reluctant to use -ta with some items while this does not happen with -tana.17 For example, with iye‘kill’, one speaker said that he had heard some people use iye-ta, as in (17), but he himself preferred the use of -tana here. Another speaker also had a few problems with iye-ta, until he said that it could be correct in a figurative expression where the ‘killer’ was for example a disease (while with -tana it could be a person or an animal). Similarly, with the verb ara ‘eat’ (tr.), the first speaker only accepted it with -tana at first, as in (20a) and (20b). When asked to reflect on the ungrammaticality of -ta with this verb, he came up with a context where this could nevertheless work, namely a figurative expression where ara-ta ‘eat-pass’ refers to a moon eclipse, a phenomenon that can be explained by reference to their traditional belief that the moon is eaten by piranhas. The example is provided in (20c). (20) a.

Kashi ¼tuke ¼mikwe18 ara-tana-ya. banana(abs) ¼3sg.abs ¼2sg.dat eat-pass1-impfv ‘Your bananas are eaten (by some pest, in your plantation).’ (elicited)

16. abari is an exceptionally unmarked optional (oblique) argument specifying the material used during the process of making something (a fan in this case). 17. Note that I did not conduct any systematic study. 18. The dative enclitic ¼mikwe has the meaning of an ethical dative here, i.e. the second person is the participant that is the most a¤ected by the event.

124

Antoine Guillaume

b.

*Kashi ¼tuke ¼mikwe ara-ta-ya. banana(abs) ¼3sg.abs ¼2sg.dat eat-pass2-impfv

c. Badi ¼tu ara-ta-ya. moon(abs) ¼3sg.abs eat-pass2-impfv ‘There is going to be a moon eclipse (lit. the moon is going to be eaten).’ (elicited)19 What seems to happen, then, is that -ta, unlike -tana, cannot be applied in an ad hoc way for depicting any transitive situation that can be described with a passive in a language like Spanish (as in elicitation) or English. The su‰x -tana, on the other hand, does not appear to have such restrictions. For example, the expression referring to a moon eclipse can also be expressed by ara-tana, as in (21), an example from Camp and Liccardi’s (1989) dictionary. (21) Badi ¼tu ara-tana-wa. moon(abs) ¼3sg.abs eat-pass1-perf ‘The moon disappears in the eclipse (lit. the moon is eaten).’ (Camp and Liccardi 1989: 8)20 Another di¤erence in productivity between -tana and -ta can be found in their distribution and/or semantics in certain derivational processes. In particular, Cavinen˜a has a productive resultative prefix e- that turns a verb into a predicative adjective (see full details in Guillaume 2008: 397 ¤.). Such adjectives normally express a state that results from the occurrence of the verb event over the participant expressed as the S argument (if intransitive) or the O argument (if transitive), as in (22a) and (22b), respectively. (22) a.

. . . e-rara e-jaki¼kwana ju-kware. res-dry npf-leaf¼pl(abs) be-rem.past ‘. . . the leaves were dry/had dried/were in a dried state.’ (from text) b. E-muya ¼tu ju-wa upati¼kwana. res-scare_away ¼3sg.abs be-perf animal¼pl(abs) ‘The animals are scared away/have been scared away/are in the state of being scared away (so they won’t show up for quite some time now).’ (Camp and Liccardi 1989: 41)

19. The translation given in Spanish is ‘La luna se va eclipsar’. 20. The translation given in Spanish is ‘La luna desaparece en la eclipse’.

A tale of two passives in Cavinen˜a

125

When a verb derived by -tana is adjectivized by e-, the meaning is resultative, as expected (i.e., as if applied to any intransitive verb, like rara- ‘dry’ above).21 (23) E-penune-tana ¼tu akwi. res-twist-pass1 ¼3sg.abs tree(abs) ‘The tree is twisted/has been twisted/is in a twisted state.’ (Camp and Liccardi 1989: 44) However, when the prefix e- is applied to a verb passivized by -ta, the resulting meaning is not resultative/stative but generic, as in (24). (24) a.

b.

[Kwanubi jasa] ¼tu e-ara-ta. animal lung(abs) ¼3sg.abs res-eat-pass2 ‘(In our culture) the lungs of animals are eaten/edible/we eat the lungs of animals/one eats the lungs of animals (*are eaten/ have been eaten/are in the state of being eaten).’ (elicited) Epiki ¼tu e-a-ta abari. fan(abs) ¼3sg.abs res-a¤ect-pass2 chonta_palm ‘Fans are (traditionally) made (lit. a¤ected) of chonta palm.’ (*in the state of being made of chonta palm).’ (elicited)

This observation can probably be linked to the lack of productivity of -ta and the concurrent tendency for it to manifest idiosyncratic aspectual and modal meanings. In co-locations with e-, the resulting meaning is not the one expected but one that expresses genericity rather than resultativity/ stativity. 3.3. Summary Before turning to section §4, where I put the two detransitivizing markers in a diachronic perspective, let us sum up the observations made above: – -tana and -ta both apply to transitive verbs and detransitivize them, with the syntactic e¤ect that the notional A cannot be overtly expressed; – -tana is fully productive; -ta has limited productivity; – -tana and -ta both have passive meanings. They are equivalent in that function; – -tana has functions that -ta lacks; it is used to express generic passive and anticausative; – -tana and -ta are segmentally very similar. 21. Note that the copula verb of copula clauses is not obligatory in Cavinen˜a.

126

Antoine Guillaume

In the present study, I hypothesize that, even though there are very strong formal and semantic similarities between the two markers, -tana and -ta are di¤erent morphemes. The question I now attempt to address is: How did those two markers arise diachronically? 4. A historical perspective The low productivity of -ta suggests that this morpheme is an old morpheme that is falling out of use in the language. This idea is reinforced by comparative data. In Guillaume (2011) I argue that cognates of -ta are present in all the remaining Tacanan languages. In those languages, -ta is not a passive marker but a cross-reference marker for a third person plural S in an intransitive clause and a third person number-neutral A in a transitive clause. This is illustrated with data from Reyesano, a sister language of Cavinen˜a, in (25) and (26).22 (25) Reyesano intransitive clauses a. 3SG b. 3PL a-puti-a a-puti-ta(-a) 23 past-go-past past-go-3s.pl-past ‘he went’ ‘they went’ (26) Reyesano transitive clauses a. 3 ! 3 b. a-ba-ta(-a) past-see-3a-past ‘he/they saw him/them’

3!1 m-a-ba-ta(-a) 1sg-past-see-3a-past ‘he/they saw me’

These observations led me to propose that the source of all -ta su‰xes in proto- (or pre-proto-) Tacanan is a third person plural subject marker in both intransitive and transitive clauses. In Cavinen˜a, with transitive verbs, this marker would have yielded the present-day passive marker -ta through an intermediate stage where it would have referred to an indefinite A argument (like the indefinite use of the pronoun they in English); this grammaticalization path is well attested (Haspelmath 1990, Heine and Kuteva 2002). 22. See Guillaume (2009) for a detailed description of the Reyesano personmarking system. 23. In the Reyesano examples, the morphemes between brackets do not show up on the surface.

A tale of two passives in Cavinen˜a

127

Viewed from the perspective of the findings, the postulated third person plural origin of the su‰x -ta is in line with the fact that -ta is restricted to passive meanings, and not to anticausative ones. The origin of -tana is more problematic. Its high productivity suggests that it is more recent. This idea is corroborated by the fact that of all Tacanan languages, it appears to be found only in Cavinen˜a.24 Its formal similarity with -ta suggests that they are historically related in a way that -tana could be made up of -ta plus a su‰x -na. Reasoning in functional terms, one could imagine that the loss of productivity of -ta created a functional need for a productive passive and that -tana appeared as a consequence via a process of reinforcement,25 that is by way of adding morphological material to -ta – in that case a syllable na – in order to reactivate the original passive meaning. This would explain why both markers are so similar formally and semantically. The question remains however of the origin of na. A possible candidate is the motion su‰x -na ‘come’26 but I have no evidence to prove it. Another potential problem with this view concerns the presence of anticausative meanings for -tana. If -tana is a reinforced passive marker, where do its anticausative meanings come from? We know from the predictions of grammaticalization theory that passive morphemes can arise from anticausative markers, not the other way around (cf. Haspelmath 1990; Creissels 2006: 39–40). In other words, it is unlikely that the anticausative meaning appeared as a natural semantic extension of -tana. A possibility is language contact, in particular with Spanish which does display a passive-anticausative polysemy. Indeed, Cavinen˜a speakers are bilingual in Spanish and have been so for a fairly long time, having been living in a (Franciscan) mission for more than two centuries (Guillaume 2008: 5–7). Alternatively, -tana could have had a di¤erent origin from -ta. In that case, one could think of two distinct scenarios. In one scenario, -tana could have started as an anticausative and later developed the passive meaning, either in order to fill the gap left by the disappearing passive -ta, or under the influence from Spanish. In the other scenario, -tana could 24. Emkow (2006: 558) reports a middle su‰x -tana in Araona. However, there is too little exemplification and discussion in her study to ascertain the existence of such a marker in that language. 25. I thank Gilles Authier for this suggestion. 26. This su‰x is used to associate a motion component to a verb stem event. The motion is directed towards the speaker (ex. ara- ‘eat’, ara-na- ‘come and eat’); see full details in Guillaume (2008: 212¤ ).

128

Antoine Guillaume

have been a passive and anticausative marker right from the beginning. In that situation, one could imagine that -tana would be in the process of replacing -ta because of the semantic overlap between the two markers and because -tana has much broader meanings than -ta, which make it more broadly applicable. An argument in favour of this hypothesis is the fact that it is not clear why a passive marker would lose its productivity if nothing is there from the start to replace it.27 With all the above scenarios we are still faced with the problem of the origin of -tana, whether it is related to -ta or it has a completely distinct origin, in which case the resemblance between the two markers is accidental. At the present time this issue cannot be answered. 5. Conclusions In this paper I have argued that the two detransitivizing markers -tana and -ta in Cavinen˜a are distinct morphemes, even though they are very similar in their segmental make-up, their syntactic properties and their semantics. I have brought to light a number of di¤erences, in particular in terms of productivity and semantic extensions. The second part of the paper was an attempt at proposing a historical scenario for their respective development. I argued that -ta is an old morpheme (originally a third person plural marker) that is losing ground, that -tana is younger and is replacing -ta, and that the replacement of -ta by -tana is probably linked in one way or another to the fact that there is a semantic overlap between the two. Abbreviations 1 2 3 a bm comp dat disemph distr

first person second person third person transitive subject (syntactic) boundary marker completive dative disemphatic distributive

27. I thank Francesc Queixalo´s for this observation and the suggestion that the disappearance of -ta could be due to a competition with -tana.

A tale of two passives in Cavinen˜a

dl/dl ds emph erg foc gen hort impfv itr./itr loc loc.gnl npf o onom pass perf pl/pl pot purp.mot rec.past recip ref reitr rem.past res restr s sg ss tr./tr

129

dual di¤erent subject emphatic ergative focus genitive hortative imperfective intransitive locative general locative noun prefix object onomatopoeia passive perfect plural potential purpose of motion recent past reciprocal reflexive reiterative remote past resultative restrictive intransitive subject singular same subject transitive

References Camp, Elizabeth, L. and Millicent R. Liccardi 1989 Diccionario Cavinen˜a-Castellano Castellano-Cavinen˜a con bosquejo de la gramatica Cavinen˜a, Dallas: Summer Institute of Linguistics. Creissels, Denis 2006 Syntaxe ge´ne´rale: une introduction typologique. Vol. 2. Paris: Lavoisier.

130

Antoine Guillaume

Derbyshire, Desmond C. 1987 Morphosyntactic areal characteristics of Amazonian languages. International Journal of American Linguistics, 53 (3): 311–326. Dixon, R.M.W. 1994 Ergativity. Cambridge: Cambridge University Press. Emkow, Carola 2006 A Grammar of Araona. Ph.D. diss., Research Centre for Linguistic Typology, La Trobe University. Guillaume, Antoine 2004 A Grammar of Cavinen˜a, an Amazonian Language of Northern Bolivia. PhD. diss., Research Centre for Linguistic Typology, La Trobe University. Guillaume, Antoine 2006 Revisiting ‘split ergativity’ in Cavinen˜a. International Journal of American Linguistics. 72 (2): 159–192. Guillaume, Antoine 2008 A Grammar of Cavinen˜a. (Mouton Grammar Library no. 44.) Berlin and New York: Mouton de Gruyter. Guillaume, Antoine 2009 Hierarchical agreement and split intransitivity in Reyesano. International Journal of American Linguistics. 75 (1): 29–48. Guillaume, Antoine 2010 How ergative is Cavinen˜a? In Ergativity in Amazonia, Spike Gildea and Francesc Queixalo´s (eds.), 97–120 [Typological Studies in Language, 98]. Amsterdam/Philadelphia: John Benjamins. Guillaume, Antoine 2011 ‘‘Third person agreement and passive marking in Tacanan languages: a historical perspective.’’ International Journal of American Linguistics, 77 (4): 521–536. Haspelmath, Martin 1990 The grammaticalization of passive morphology. Studies in Language 14 (1): 25–72. Heine, Bernd and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Keenan, Edward L. 1985 Passive in the World’s languages. In Language typology and syntactic description, Vol. 1, Tim Shopen (ed.), 243–281. Cambridge: Cambridge University Press. Keenan, Edward L. and Matthew S. Dryer 2007 Passive in the World’s Languages. In Language Typology and Syntactic Description, Vol. 1, Timothy Shopen (ed.). Second Edition. 325–361. Cambridge: Cambridge University Press. Lazard, Gilbert 1986 Formes et fonctions du passif et de l’antipassif. Actances 2: 3–57.

A tale of two passives in Cavinen˜a

131

Payne, Doris L. 1990 Introduction. In Amazonian Linguistics. Studies in Lowland South American Languages. Doris L. Payne (ed.), 1–10. Austin: University of Texas Press. Payne, Doris L. 2001 Review of The Amazonian languages, R.M.W. Dixon and Alexandra Aikhenvald (eds.), 1999, Cambridge: Cambridge University Press. Language 77 (3): 594–598. Pitman, Donald 1980 Bosquejo de la grama´tica Araona. (Notas Lingu¨´ısticas No. 9.) Riberalta: Instituto Lingu¨´ıstico de Verano. Siewierska, Anna 2005 Passive constructions, In World Atlas of Language Structure. Martin Haspelmath, Matthew S. Dryer, David Gil and Bernard Comrie (eds.), chapter 107: 434–437. Oxford: Oxford University Press.

The detransitive voice in Kryz Gilles Authier 1. Introduction This paper examines the detransitive voice in Kryz, an unwritten language belonging to the Lezgic branch of the North-East Caucasian family.1 Nowadays three dialects of Kryz (Kryz proper, Jek, and Alik-Khaput) are spoken as a first language by at most 2000 speakers, in fewer than ten localities of north-eastern Azerbaijan. Despite generalized bilingualism in Azeri, Kryz preserves typical Proto-Lezgic features. In particular, gendernumber agreement with S/P (Single argument or Patient) arguments is preor infixed to the lexical stems of synthetic verbs,2 which form a closed class. Person is expressed by free pronouns. Word order is strictly possessorpossessed, adjective-noun, and basically Agent-Patient-Verb; case marking on nouns and cross-referencing on the verb is ergative. Valency increase is expressed periphrastically by means of auxiliaries (‘do’ or ‘give’). The Kryz synthetic detransitive voice, to be described in this paper, is unique within Lezgic3 and the East Caucasian language family as a whole. While East Caucasian languages have long been believed to lack a passive construction altogether, it is beyond doubt that in Kryz there exists a detransitive voice with a prominent passive reading, alongside

1. This article is based on first-hand data collected by the author in Azerbaijan over the last ten years. Field trips were funded by the Centre National de la Recherche Scientifique (UMR 7192) and the Institut National des Langues et Civilisations Orientales. My host in the village of Alik and main informant for this dialect is Madjlis Shamseddinov, to whom I am indebted more than words can say. 2. Verbal predicates are also represented by compound verbs, which are not addressed here. They switch valency by changing their auxiliary (‘do’/ ‘become’): this valency-changing device is equipollent and belongs to syntax, not derivation. 3. This branch comprises the core languages Lezgian, Tabassaran, Aghul, Tsakhur, Rutul, Kryz, and Budugh, all located close to the watershed line of the easternmost part of the Great Caucasus range, and the outliers Udi (to the south) and Archi (to the north).

134

Gilles Authier

various other possible values (anticausative and antipassive, as well as aspectual and modal nuances). The use of this voice is restricted by the semantic parameters and lexical properties of individual verbs, which accounts for the fact that the few previous studies on Kryz (principally Saadiev 1994) make no mention of any detransitivizing derivation. Indeed, the Kryz detransitive voice is not very frequent in natural speech: it is mostly employed in prescriptive discourse and technical instructions. But the derivational morphology is old and can be compared with similar detransitive forms found in some other East Caucasian languages only distantly related to Kryz. Section 2 of this article gives an overview of grammatical relations and the pragmatically marked order of constituents in non-derived predication. Section 3 presents the morphology of detransitive forms. Section 4 presents the lexico-semantic restrictions which operate on the passive voice. Section 5 describes the particulars of the aspectual and modal values associated with the anticausative and antipassive forms. Section 6 proposes comparative material and a hypothesis regarding the origin of this rare and residual voice phenomenon in East Caucasian.4 2. Coding of syntactic and pragmatic functions In Kryz, as in most East Caucasian languages, basic grammatical relations in declarative clauses are organized according to an ergative/absolutive case marking system with gender-number cross-referencing of the Single or Patientive core argument on the verb. The details are described in the following subsections. 2.1. Case marking Kryz has ergative case marking on nouns. Nouns expressing S and P are in the unmarked ‘absolutive’ case, while the Agent noun5 of a transitive 4. The Azeri alphabet is used. The diphthongs /a-u/ /e-u/ and /i-u/ are realized as [o], [ø] and [y] in closed syllables; the sounds [q:], [g] and [g˘] (uvular fricative) are in complementary distribution and written . The digraphs and represent voiceless and voiced velar fricatives respectively. 5. The pronominal system operates on a di¤erent basis from the nominal system. Speech Act Participants are expressed by free or cliticized pronouns with no ergative marking. There is also no specialized morpheme on verbs heading reflexive and reciprocal constructions, which make use of special reflexive and reciprocal pronouns. Peculiarities of case marking on Kryz pronouns are detailed in Authier (2008).

The detransitive voice in Kryz

135

predicate is marked with ergative case -(v)r, as illustrated in ex. (1): (1) a.

b.

q’al-ar-ir yuva-yar g˘uzay.c-a g˘vat’ats’-re mouse-pl-erg nest-pl north-in dig-prs(npl) ‘Mice dig their nests to the north.’ q’al-ar yuva-yar-a g˘irxhar-e mouse-pl(abs) nest-pl-in dwell-prs(npl) ‘Mice dwell in nests.’

Note that in contrast with many other East Caucasian languages, in Kryz the ergative case is never used to mark semantic roles other than Agent and two ergative-marked nouns with di¤erent semantic roles cannot coexist in the same clause. But ergative-marked Agents can be inanimate, as in examples (2) and following, and while the basic order of constituents is S/A-(P)-V, non-prototypical Agents are often postposed to a more topical Patient, as in (3) and (4):  a-xhr-i (2) naq’ duli.c-ir xayla ziyan vu-yic yesterday pv-come.pf-part storm-erg much damage give-aor.n ‘Yesterday’s hailstorm has caused a lot of damage.’

(3) yag˘in jin g˘ara.c-ir basmis¸ b-ar-ca-b sure 1plexcl illusion-erg pressing hpl-do-prf-hpl ‘We are certainly victims of an illusion.’ (4) kaha-c siy babrux-ci xal.id-ir yi-qir-ca cave-gen mouth spider-gen roof-erg pv-catch-prf ‘A cobweb covers the entrance to the cave.’ 2.2. Indexation of S/P Kryz verbs have the gender and number 6 of the Patient or Single argument cross-referenced in a pre-root (post-preverbal) slot, as shown in ex. (5): (5) a.

buba-r ris¸-imbi yi-b-qir-cib father-erg daughter-pl pv-hpl-catch-aor.hpl ‘The father has caught the daughters.’

6. The gender-number system comprises five ‘gender-number agreement classes’: human masculine (M), human feminine and other animates (F, which also includes many inanimates and certain abstract concepts), other inanimate (N for ‘neuter’), non-human plural (NPL), and human plural (HPL).

136

Gilles Authier

b.

ris¸-imbi g˘a-b-q’il-cib daughter-pl pv-hpl-lie-aor.hpl ‘The daughters went to sleep.’

Nominatively aligned indexation of S/A persons is also present in the imperative paradigm only. The overall morphological structure of finite verb forms is summed up in table 1. Table 1. Structure of synthetic verb forms in Kryz -3

preverb (usually with transparent spatial meaning)

-2

prefixes cross-referencing the gender-number of S/P

-1

aspect-marking sonorant (-r, -l, -n)

0

root consonant

1

perfective aspect or detransitivity markers

2 3

Tense & Mood su‰xes su‰xes cross-referencing the gender-number of S/P (S/A in imperatives)

2.3. Verb classes according to valency Verbs in Kryz fall into clear semantic classes, which are defined by their specific morphological and syntactic properties, especially valency changing properties: strict intransitive, extended intransitive (a¤ective and sensorial predicates), strict transitive, and ditransitive. Lability (ambitransitivity) is a very marginal phenomenon. Transitive verbs have in their valency an Agentive argument which, if expressed, is in the ergative case; these all permit valency increase of the causative type by means of the auxiliary ‘give’, including those which are ditransitive like ‘give’ itself. Most but not all transitive verbs allow synthetic (morphological) detransitive derivation, a restriction which we find to have a clear semantic motivation, to be described below. Subclasses are also observable within the class of intransitive verbs – for instance, most but not all of these have synthetic imperative forms, nor are they all able to undergo periphrastic causative derivation with the auxiliary ‘do’. But all are characterized by a stem ending in -aR. 2.4. Order of arguments and Patient topicalization Agents are usually topics, and their unmarked position is clause-initial; non-topicalized Patients come after the Agent and before the verb. But, as is common for languages which express grammatical relations by means

The detransitive voice in Kryz

137

of case-marking, the order of arguments can be modified for pragmatic purposes. Focused constituents tend to occur in preverbal position, with heavy stress, and topicalization of a Patient in the absolutive does not require the use of a special form of the verb. Kryz allows the Patient to be topicalized simply through alteration of the word order. In this ‘‘functional passive’’ (Givo´n 2004) the ergative-marked Agent can remain, backgrounded, in the construction: (6) u-c seksenbes¸ san a-n-ir vul c¸ixe ya-u-qur-cu 3-gen.nh eighty_five year 3-h-erg ewe behind pv-f-spend-prf.f ‘[He lived for a hundred years,] eighty-five of which he spent tending the sheep.’ Non-prototypical Agents are often moved into non-initial position. The most natural translation in English is a passive construction: (7) a-c hic¸vaxt sel.c-ir tu-g˘ats’-da-d 3-n never flood-erg pv-carry-neg.evt-n ‘This (bridge) will never be carried away by the flood.’ (8) xinib fura-r v-a-u-q-ryu, nisi maltal.ci-r woman man-erg f-pv-f-keep-prs.f cheese cheese.bag-erg ‘The wife is kept by the husband, as cheese (is kept) by the cheese-bag.’ Note that with appropriate intonation on the preverbal position, the Agent can be focused instead of backgrounded. It then translates not as a passive, but as a cleft sentence: (9) ceuhur-ci ghala-d mes¸a.c-a sar.id-ir ulats’-ru pear-gen good-notn forest-in bear-erg eat-evt.f ‘It is the bear who will get to eat the best pears in the forest.’ (10) ts’e .il-kar azar-i yux yig˘g˘acig˘ ebil-ir g˘irats’-re-ni goat-subel milk.mp-part milk every.day wolf-erg drink-prs-past ‘The milk drawn from the goat every day, the wolf would drink it.’ 2.5. Omission of Agents In Kryz predicative constructions, the expression of arguments as overt NPs is not grammatically obligatory, but at least one is encoded on the verb: the minimal predicate consists of a single verb form on which a Single argument or a Patient is cross-referenced:

138

Gilles Authier

(11) a.

b.

( furar) (xhin) g˘e-t’-ic man.erg grass.n pv-strike-aor.n ‘(The man) cut it (the grass).’ (xinib) yi-p-du woman(abs) pv-f.go-aor.f ‘(The woman /) she left.’

Easily recoverable or non-referential Agents are very often left out. In the original of the following English translation the word ‘you’ is used only once, with obviously generic, non-referential value, and Kryz uses only active forms (see text in Authier 2009) whereas English has to employ the passive: ‘The cheese-bag is made to keep cheese in. You have to flay the skin to keep the cheese in without making any cuts. Then, as soon as it is flayed, salt is poured into it to dry it out. Then it has to be worked on, and tied firmly at the fore and hind legs. If there is a great deal of wool, it is shorn, then it is washed with water while being beaten with a rod, then left to dry. Then it is turned inside out (the cheese is on the woolly side) and stored for winter.’

Non-explicitation of a generic Agent is frequent in proverbs such as the following (on ingratitude): (12) tur.ud-zina ula-ci ciyar-zina ul g˘va-s¸-ryu spoon-instr eat-seq tail-instr eye.f pv.f-pull.out-prs.f ‘[They] eat with the spoon and then pull (your) eye out with the handle!’ A preliminary conclusion, then, is that in Kryz as in many other (genetically related or unrelated) ergative languages, backgrounding through omission of the Agent NP and concomitant foregrounding of the semantic Patient is possible without any change in the verb morphology. But Kryz also has a detransitive voice which shows properties commonly associated with the label ‘passive’: it is a morphological derivation which transforms transitive predicates into intransitive ones by removing the agentive NP at a syntactic level. But let us first examine the formal manifestation of (in)transitivity on Kryz verbs. 3. Marking of (in)transitivity and verb classes Kryz synthetic (non-compound) verbs form a closed class comprising some 200 items. They are either transitive or intransitive, a distinction which is marked morphologically on imperfective stems only.

The detransitive voice in Kryz

139

3.1. Morphology of intransitive vs transitive verbs All of the 70 synchronically underived intransitive verbs display on their imperfective verb stem one of three allomorphs of the su‰x -aR (-ar/-al/-an): Table 2. Structure of intransitive imperfective verb stems a-r-c ¸ ’-ar-

‘pv-ipf-enter-intr’

ke-r-xh-ar- ‘pv-ipf-move-intr’ la-l-s-al- ‘pv-ipf-turn-intr’ ha-n-gv-an- ‘pv-ipf-run-intr’

Conversely, the 130 transitive synthetic verbs do not show these su‰xes. They are not overtly marked as transitive, but their imperfective forms should be considered zero-marked for transitivity, given the absence of the characteristic intransitive su‰xation: Table 3. Structure of transitive imperfective verb stems ya-r-t’- ‘pv-ipf-cut’ yi-r-q- ‘pv-ipf-catch’ ki-l-t’- ‘pv-ipf-yoke’ yi-n-gh- ‘pv-ipf-pull’

Overt marking of intransitivity is not to be found on the perfective forms of underived intransitive verbs, which are thus indistinguishable from transitive perfectives: Table 4. Structure of perfective stems of non-derived verbs intransitive verbs, pf a-c ¸ ’-

‘pv-enter.pf’

cf. transitive verbs, pf ya-t’- ‘pv-cut.pf’

ke-xh-r- ‘pv-move-pf’

yi-q-r- ‘pv-catch-pf’

la-s-l- ‘pv-turn-pf’ hagu-n- ‘pvrun.pf’

ki-t’-l- ‘pv-yoke-pf’ yi-gh-n- ‘pv-pull-pf’

Because of the general omissibility of argument NPs, transitive and intransitive predicates are hard to distinguish in the perfective tenses. Allowance of explicit ergative arguments is the only criterion available,

140

Gilles Authier

and no distinction can be drawn by means of syntactic tests between a Single argument construction and a transitive construction in which the Agent is left out. 3.2. Detransitive morphology on transitive roots Transitive verbs may be classed into two groups according to whether they allow morphological detransitivization or not. When available, detransitive voice is formed synthetically in the imperfective through the addition of the morpheme -aR- to the right of the verbal root; on most but not all verbs, a sonorant imperfective aspect marker occurs before the root, and R is a duplicate of this pre-root aspect marker: Table 5. Derivation of detransitive imperfective stems

‘catch’ ‘pull’ ‘yoke’

transitive ipf ¼>

detr. ipf

yi-r-qyi-n-ghki-l-t’-

yi-r-q-aryi-n-gh-anki-l-t’-al-

‘be caught’ ‘be pulled’ ‘be yoked’

(Intransitive verbs in Kryz thus look like detransitive verb forms with no corresponding non-derived form, i.e. ‘deponents’ or ‘media tantum’ in the grammar of classical languages.) The perfective is formed analytically with a verbal adjective derived from the imperfective detransitive stem, to which the auxiliary xhiyi ‘having become’ (perfective participle) is added: Table 6. Analytic perfective of detransitive verbs detr. ipf

detr. pf

‘catch’ >

yirq-ar-i

‘being caught’

yirq-ar-a xhiyi

‘caught’

‘pull’ > ‘yoke’ >

yingh-an-i kilt’-al-i

‘being pulled’ ‘being yoked’

yingh-an-a xhiyi kilt’-al-a xhiyi

‘pulled’ ‘yoked’

3.3. The detransitive derivation (prototypical transitive verbs) Most transitive verbs allow a detransitive derivation with Patient-maintaining semantics, either an Agent-backgrounding ‘passive’ or an Agent-suppressing ‘anticausative’. These forms are found in appropriate contexts which usually

The detransitive voice in Kryz

141

imply modal or aspectual nuances, and in fact they are rare in narrative and everyday speech. As is clear from this sample, transitive verbal lexical items in the passive voice retain prototypical transitive (voluntary Agent, Patient-transforming) semantics: Table 7. Some typical transitive verbs and passive forms pf

ipf

detransitive

g˘axur-i naq’un-i

g˘arxv-i nanq’v-i

‘knead’ ‘churn’

g˘arx-ar-i / g˘arxvara xhiyi nanq’v-an-i / nanq’vana xhiyi

c¸akva-y

c¸ak(vats’)-i

‘mix’

c¸akv-ar-i / c¸aukvara xhiyi

raxva-y

raxv(ats’)-i

‘shave’

raxv-ar-i / raxvara xhiyi

reuha-y

reuh(ats’)-i

‘grind’

reuh-ar-i / reuhara xhiyi

yixha-y yiza-y

yixh(ats’)-i yiz(ats’)-i

‘harvest’ ‘plough’

yixh-ar-i / yixhara xhiyi yiz-ar-i / yizara xhiyi

The following example shows a typical passive transformation (note that here the passive imperfective stem is not derived from the current transitive one): (13) a.

b.

har cum a-ca Hazratbaba.ci-r lu kel kura-ts’-ryu-ni every Friday-in Saint.Baba-erg this lamb slay-ipf-prs.f-past ‘Every Friday Saint Baba would sacrifice this lamb.’ har cum a-ca lu kel kur-ar-yu-ni every Friday-in this lamb slay.detr-prs.f-past ‘This lamb would be sacrificed every Friday.’

Unlike in English, in Kryz the Agent cannot be expressed as an oblique NP, a frequent feature of passives cross-linguistically. The detransitive construction in Kryz thus complies with Kazenin’s (2001) and Comrie’s (2008) criteria for passive voice;7 7. We do not retain the definition of passive as a means of promoting the Patient of a transitive verb to grammatical subject, because this would presuppose that Kryz has a clear-cut definition of syntactic ‘subject’, which is not the case: as in most East Caucasian languages, behaviour properties tend to group S with A, while we have seen that coding properties clearly align S with P, that is, ergatively. Note also that Kryz detransitive forms never receive reciprocal

142

Gilles Authier

– it is morphologically heavier, usually derived from the active; – the construction is ‘‘less frequent, functionally specialized, not fully productive’’ (Haspelmath 1990); – the new subject is not a semantic Agent; – the semantic role of the maintained argument does not change. In the case of a language with ergative case marking like Kryz, the unmarked absolutive case of the maintained Patient is also retained. 3.4. Transitive verbs lacking a detransitive form Three semantically coherent classes of verbs systematically lack a detransitive form: 1) verbs of in- or ex-corporation; 2) verbs of exchange and social interaction; 3) transitive verbs of motion (for which see below). The following two lists are exhaustive: Table 8. Verbs of social interaction and in- or ex-corporation g˘aynic

‘take, buy’

adg ˘ ulic

‘swallow’

vaxhayc

‘borrow’ ‘keep’ ‘say’

c¸’udxunic g˘itnic

‘suck’

aqayc lipic

c¸’ag˘vayc

‘drink in one draught’ ‘gnaw’

k’isic g˘adgunic

‘bite’

g˘ac¸ayc g˘ahayc xuyic

‘vomit’

‘put on (trousers)’ ‘extract; hatch’ ‘give birth’

These verbs are not prototypical transitive verbs. They deviate from the transitive prototype in that social interactions imply two agentive participants, and present semantic similarities with reciprocal actions, which cannot be deactivized. In the second class the Agent is strongly a¤ected by the event. Such features probably account for the restriction on passivization observed here, as the main functions of the passive (backgrounding or autocausative (reflexive) interpretations (there are no ‘‘grooming middles’’, cf. Kemmer 1993). The fact that processes implying subject-object coreference are not expressed by this form of the transitive verb in Kryz has led us to discard the term ‘middle’ in favour of detransitive voice. We also discard the more all-embracing appellation ‘Medio-Passive’, because a couple of detransitivized verbs have an antipassive interpretation, for which see below.

The detransitive voice in Kryz

143

of the Agent and foregrounding of the Patient) make no sense with these verbs. Among those Kryz transitive roots which do not allow valency decrease we also find the whole semantic class of verbs denoting transitive motion events. One motivation for their failure to allow detransitivization may be the atypically agentive role implied by ‘transportation’ situations: the Patient is not transformed or internally a¤ected, and retains a great deal of agentivity if it is animate. In accordance with this principle, agent backgrounding in a process of bringing is expressed by using an intransitive counterpart of ‘bring’ with a meaning such as ‘go’, as in ex. (14): (14) mahraka.c-a va-xhr-i s¸abas¸ contest-in pv-come.together-part presents.for.musicians ‘Gifts gathered at the wedding party.’ Another – lexical – motivation is the fact that these roots neatly match a class of intransitive motion verbs which do not form causative constructions with ‘do’, unlike the majority of Kryz intransitives: they pair up with the corresponding transitive roots. Any valency-changing construction is thus superfluous for both classes, a form of symbiosis which makes this suppletive subsystem particularly economical: Table 9. Other verbs lacking detransitive forms: suppletive motion verbs ag ˘ ayc

ac ¸ ’ic, afic

aqric

axhric

‘bring’ ‘take down’

aska-yc ‘put (down)’ g˘ag˘ayc ‘create’ g˘aqric, g˘a ayc ‘take, push out’ g˘e ayc ‘build’

‘enter, go down’ ‘come (down)’

g˘uq’ric ‘touch, intr.’ ac ¸ ’ic, g˘adfic ‘go out, escape’ g˘axhric ‘appear’

g˘ig˘ayc ‘turn, tr.’

g˘ific ‘rise’ g˘ixhric ‘turn, intr., live’

ke ayc ‘push under’

kec¸’ic ‘go under’, kedfic ‘pass under’

va ayc

vac¸’ric, vadfic ‘turn from’

‘push aside’

vag˘ayc ‘gather, tr.’ ya ayc ‘push aside’

vaxhric ‘gather, intr.’ yafic ‘stray’

yaqric ‘take across’

yac¸’ic ‘pass, cross’

yatric ‘leave, tr.’

iknic ‘remain’

144

Gilles Authier

4. Passive voice and its additional values A clause involving a transitive verb in its passive form is usually not to be understood as a mere paraphrase of the non-derived, transitive clause. Not only does the passive background the Agent and topicalize the Patient, but it also has an impact on modal and aspectual parameters. This section examines these associated e¤ects. We have found passives of transitive verbs either with habitual and deontic value in the imperfective, or with state-focusing (resultative) value in the perfective. The referential properties of the remaining argument are also a¤ected by a change of aspect: (15) a.

b.

g˘ul ambar.c-a va-nsan-e corn barn-in pv-weigh.detr-prs ‘Corn is usually weighed in the barn.’ (non-referential) g˘ul ambar.c-a va-nsan-a xhi-yic corn barn-in pv-weigh.detr-a be-aor.n ‘The corn has been weighed in the barn.’ (referential)

4.1. Deontic and Habitual meaning of the imperfective passive The association of deontic meaning with the passive is well attested crosslinguistically. Collective or typically feminine chores culturally entail the de-individualization of the agent, and the generic, non-referential (or taboo?) status of the unmentioned Agent entails the use of a detransitivized verb form: (16) cindir-a q’vahac-ci q’an ke-rt’ar-e worn-a sock-gen bottom pv-sew.detr-prs ‘When the bottom of a sock is worn out, one patches / must patch it.’ (17) riki yig˘g˘acig˘ va-rc¸ar-e yard every.day pv-sweep.detr-prs ‘The yard is / must be swept every day.’ There are many examples in proverbs in the ‘eventual’ mood (covering both habitual events and gnomic sentences with deontic modality): (18) irac irca-zina zimar-da-d blood blood-instr wash.detr-negevt-n ‘One does not wash away blood with blood.’ (Prov.)

The detransitive voice in Kryz

(19)

145

u-du-fa

riki g˘e-t’ar-da-d pv-neg-close.pf(part) door pv-strike.detr-negevt-n ‘An open door is not knocked at.’ (Prov.)

ayal.ci-z maxar (20) d-is¸ar-i vuts’ar-da-b neg-cry.ipf-part child-dat breast.f give.detr-negevt-f ‘A child who does not cry is not given the breast.’ (Prov.)

/ g˘i-rqar-e (21) q’usi.c-a ayal a-lt’al-e cradle-in child pv-bind.detr-prs pv-hold.on.detr-prs ‘The child is (¼should be) firmly bound / held in the cradle.’ (22) heyvanat yif.un-a q’acil-a a-rg˘var-e cattle night-in everyday pv-push.in.detr-prs ‘Cattle are / have to be penned during the night.’ (23) g˘urux.c-a ar barkan-bi hu-rt’ar-e garden-inel horse-pl pv-expel.detr-prs ‘Horses should be banned from gardens.’ These habitual and deontic values are present in the overwhelming majority of examples of detransitive forms in the corpus; but passive voice can also be found when the precise Agent of an action is unknown in the context. A generic, non-referential interpretation is then preferred, especially in future time reference: ark’a caza (24) a-n-iz vuts’ar-iya 3-h-dat heavy punishment give.detr-fut ‘He will be given a severe punishment.’

4.2. Resultative meaning of the perfective passive While imperfective passive forms have habitual and deontic readings, the corresponding perfective forms have a resultative reading. Emphasis is laid not on the event itself but on the resulting state. This is sometimes a way to convey the fact that the event was not witnessed by the speaker: (25) riki cigac¸-a a-rt’ar-a xhi-yic door place-in pv-fit.in.detr-a be-aor.n ‘The door has been put back in place.’

146

Gilles Authier

(26) daug˘a q’um-ug˘ g˘va-xvar-a xhi-d-u soup ground-super pv.f-pour.detr-a be-aor-f ‘The soup has been spilt on the ground.’ It seems that this resultative interpretation does not necessarily apply with negative forms: (27) zin lu yas¸.in-iz karta hic¸vaxt atar-a xhi-de-r 1 this year-dat until never beat.detr-a be-negperf-m ‘Right up to my current age, I have never been beaten up.’ 4.3. Stylistic variation? The passive marking of verbs is often optional, and passive forms can coexist in the same text as unmarked active forms. In the following text, the first few forms are passive, and set the general ‘deontic’ tone of the whole passage. Most forms later in the passage are left unmarked, probably because active forms are ‘lighter’ and systematic passive marking would be redundant (recall that the Agent can also be omitted in transitive constructions): (28) vul sina q’va-d safar vu-dar-yu g˘arayaz-ca ewe year.in two-n time pv-shear.detr-prs.f springtime-in vudar-i yis.a-g˘ar yapag˘u li-re u-c-kar pv.shear.detr-part wool-supel spring.wool say-prs 3-nh-subel g˘i-uxvar-a xhi-yi xiy-ar-kar q’vahac xirats’-re pv-spin.detr-a be.pf-part thread-pl-subel sock knit-prs ic-kar an xi-xar-a xhi-yi s¸ay-ri selff-subel an red-plait.detr-a be.pf-part thing-pl mahkam s¸are; c¸ixe an yis yanxuc-a vudats’-ryu solid be-prs then wool.f autumn-in shear-prs.f lu safar vu-da yis g˘it s¸a-re this time pv-shear.pf(part) wool.f few be-prs u-c-kar g˘i-ux-i xiy davam-suz s¸i-u-ryu 3-nh-subel pv-spin-part thread.f continuity-without be-f-prs.f The wool of the sheep is shorn (PASS) twice a year. The wool shorn (PASS) in spring is called ‘yapaghu’. With the spun (PASS) threads, one knits (ACT) socks, and objects knitted (PASS) from them are robust. The second wool is collected (ACT) in the autumn; this time the wool shorn (ACT) is not abundant; the thread spun (ACT) from it does not hold.

The detransitive voice in Kryz

147

4.4. Orientation of relative participles The participles found in East Caucasian languages are ‘contextually oriented’ (Haspelmath 1994), due to the fact that their arguments retain the same case marking as in independent clauses. In ex. (29), the same participle is used for both orientations: in a) Patient orientation is indicated by ergative marking on an NP and the absence of an absolutive NP in the relative clause, whereas in b) Agent orientation is inferred from the absence of an ergative-marked NP: (29) a.

b.

xinib.ci-r hala da-b-cir-i xas¸ul woman-erg yet neg-f-bake.pf-part stew.f ‘the stew that the woman has not yet cooked’ hala xas¸ul da-b-cir-i xinib yet stew.f neg-f-bake.pf-part woman ‘the woman who has not yet cooked the stew’

But Kryz also uses derived, specifically negative passive participles, made up of a transparent analytic combination of the detransitive verbal adjective and the auxiliary ‘be, become’: (30) a.

reudeuha q’el sifra.c-a a-ma-sku! neg.grind.ppf salt table-in pv-proh-put b. reuhar-a dauxha q’el sifra.c-a a-ma-sku! grind.detr-a neg.be.ppf salt table-in pv-proh-put ‘Do not put unground salt on the table!’

These analytic forms seem to add a slight nuance of possibility: g˘a-dauxvi xamir.ci-kar fu s¸a-dad pv-neg.knead.ppf dough-subel bread be-negdeb b. g˘a-rxvar-a dauxha (. . .) pv-f.knead.detr-a neg.f.be.ppf ‘Dough does not make bread unless it is kneaded.’

(31) a.

g˘a-da-xi xamir-bekar fu cirar-de-d pv-neg-knead.ppf dough-pl.subel bread bake.detr-negprs-n b. g˘a-rxar-a da-xha (. . .) pv-knead.detr-a neg-be.ppf ‘Bread made of unleavened dough does not bake.’

(32) a.

148

Gilles Authier

Some degree of inherent Agent orientation is probably the reason why, in the following example, a specifically passive participle has to be used, because the active participle would be Agent-oriented and prevent recovery of the Recipient (rather than Agent) role for the head: (33) a.

b.

caza *vu-ts’-i / vuts’ar-i adami punishment give-ipf-prt give.detr-prt person ‘The man who is given a punishment.’ caza *vu-yi / vuts’ar-a xhi-yi adami punishment give.pf-prt give.detr-adj be.pf-prt person ‘The man who has been given a punishment.’

4.4. Topic continuity and syntactic accusativity Morphologically active sequential converbs in Kryz usually imply a coreferent subject (S/A) in the linked matrix clause (see examples in Authier 2009), that is, an apparent accusative pivot in terms of Dixon (1994). This might lead us to assume that passive predicates promote Patients to the status of syntactic subjects. In the following example, coreference of subjects obtains as the imperfective coordinating converb heading the first clause is in the passive voice: (34) lu g˘uyi yig˘g˘acig˘ g˘vat’-ar-a, halu aranxhin-a xhi-cu this well every_day pv.dig-detr-manner this depth-in be-prf.f ‘This well being dug every day, it became this deep.’ This may well be true for this imperfective converb: more investigation is needed. On the other hand, perfective converbs (sequential, ending in -ci) can be followed by a passive form whose Single argument co-refers with the Patient of the preceding subordinate clause, as in ex. (35): (35) gugarti yi-t’a-ci har adami-c bus¸q’ab.ci-g˘ g˘i-yts’ar-e greens pv-cut-seq every person-gen plate-super pv-pour.detr-prs ‘The salad is cut and served on everyone’s plate.’ This property of sequential converbs shows that Kryz generally lacks any clear ‘syntactic pivot’, and this is by no means the only instance of ergative inter-clausal syntax. In ex. (36), the Single argument in the sequential subordinate clause co-refers with the Patient of the matrix clause, in the absence of an overt Agent:

The detransitive voice in Kryz

149

(36) sad tak yif xinib-cizina irkin-ci one.n only night woman-instr m.remain.pf-seq laqatasi u-d vul-bevas hay yi-re the_day_after prox-m.abs sheep-pl.cum sent do.ipf-prs.m ‘After remaining only one night with his wife, on the following day he is sent to (watch over) the sheep.’ 4.5. Preliminary conclusions on the use of the Kryz passive voice The main Kryz passive-forming device seems to be linked to imperfective aspect, with either habitual nuance or deontic modality. Perfective forms are derived analytically from the imperfective, and synchronically these perfective periphrastic passive forms occur just as often in texts as imperfective ones, because they o¤er valuable new aspectual nuances (inceptive or resultative). These modal or aspectual nuances are much more crucial to the actual use of passive forms than any syntactic rule or topic-maintaining function. 5. Anticausative and antipassive interpretations More archaic than the passive appear to be those uses of the detransitive forms to which we now turn, in which the detransitive voice is anticausative (associated with Patient-oriented meaning) or antipassive (with a few Agent-oriented verbs such as ‘eat’). 5.1. Anticausative The same detransitive morphology is used with a number of verbs in contexts where the Agent is not only backgrounded, but semantically suppressed: that is, it encodes anticausative value. For instance: Table 10. Detransitive forms with anticausative meaning pf

ipf

detr. ipf & pf participles

ugv-a-

ugv-(a-ts’)-

‘burn’

ugv-ar- / ugvara xhiyi

cir-a-

cira-ts’-

‘cook’

cirar-i / cirara xhiyi

uf-a-

uf-(a-ts’)

‘close’

ufar-i

s¸aku-rg˘et-asurh-a-

s¸arkvg˘etsuha-ts’-

‘trim’

s¸arkvar-i / s¸arkvara xhiyi g˘etar-i / g˘etara xhiyi surhar-i / surhara xhiyi

‘beat’ ‘drag’

/ ufara xhiyi

150

Gilles Authier

When the detransitive derivation is applied to these verbs, only one participant remains in the situation. The semantic role of the absolutive argument is also retained in the anticausative construction, but unlike in the passive, where an unexpressed agent is implicit, here the agent is conceptually removed. In the anticausative interpretation the event is spontaneous, often with significant semantic modification of the verb. Note that two distinct interpretations may coexist, with or without an implicit Agent: (37) zang g˘e-rt’ar-a xhi-yic-zina k’ul.c-a ha-r-gun-d bell pv-strike.detr-adj be-msd-instr house-in pv-m-run-aor.m ‘The bell a) was rung (passive) or b) rang (anticausative), and he ran home.’ While the passive interpretation of detransitive forms is slightly more prevalent in narrative texts, the anticausative or ‘spontaneous’ meaning is found in idiomatic expressions: (38) c¸’ebic g˘i-yts’ar-e rain pv-pour.detr-prs ‘The rain is pouring.’ (39) palas g˘e-tar-e carpet pv-beat.detr-prs ‘The carpet slaps (in the wind) / *starts to slap (cf. ex. 46).’ In contrast to the e¤ect noted above for the passive use, the imperfective gives rise to no aspectual (habitual) or modal (deontic) nuances: (40) za kil.a-k rib c¸u-nq’van-yu 1.gen arm-sub needle(f) pv-stick.detr-prs.f ‘A needle sticks itself in my arm.’ 5.2. Technical spontaneous meaning The semantic role of the absolutive argument is often modified along with the meaning of the verb put into the anticausative. For instance, the derived detransitive form of ‘catch’ means either ‘be caught (by the police)’ (passive interpretation) or ‘coagulate’ or ‘be eclipsed’ (anticausative): (41) yux yi-rqar-e milk pv-catch.detr-prs ‘The milk ‘‘catches’’ ¼ turns to cheese.’

The detransitive voice in Kryz

151

(42) varag˘ yurqar-yu sun.f pv.f.catch.detr-prs.f ‘The sun is eclipsed.’ 5.3. Perfective anticausatives: additional inceptive value The non-derived, labile, perfective form can have anticausative meaning: (43) za kil.a-k rib c¸u-b-q’vun-du 1.gen arm-sub needle(f) pv-f-stick-aor.f ‘A needle has stuck itself in my arm.’ To give a passive reading, the derived, explicitly detransitive form is required: (44) rib za halav.ci-k c¸u-nq’van-a xhi-du needle 1.gen dress-sub pv-stick.detr-a be-aor.f ‘The needle has been stuck in her dress (by someone).’ Derived perfective anticausatives are also found, usually with an additional aspectual value of inceptivity (to be contrasted with the resultative meaning of passive perfectives; note also that Kryz lacks a general verb ‘start, begin’): (45) c¸’ebic g˘i-yts’ar-a xhi-yic rain pv-put.on.detr-a be-aor.n ‘The rain started to pour.’ (46) palas g˘e-tar-a xhi-yic / s¸a-re carpet pv-beat.detr-a be-aor.n be-prs ‘The carpet started / starts to slap (in the wind).’ (47) yux yi-rqar-a xhi-yic milk pv-catch.detr-a be-aor.n ‘The milk turned to cheese.’ Much rarer is the possibility of a resultative reading, with the auxiliary in the perfect tense: (48) za galu yi-rqar-a xhi-ca 1.gen throat pv-catch.detr-a be-prf ‘My throat got/is blocked.’

152

Gilles Authier

5.4. Magic autocausative Detransitive forms can be used in magical contexts to emphasize the unnatural absence of the expected, supposedly necessary Agent: (49) a-c-iz amr v-ar-izma an k’ul va-rc¸aryu-ni 3-nh-dat order.f f-do-before room.f pv-sweep.detr-prs.f-past ‘(She had a magic ring:) as soon as she gave it the order, the room would be swept.’ (50) div-ci k’ul.c-a yis icic-ig˘an g˘a-har-e-ni demon-gen house-in wool selff(e)selff-equ pv-card.detr-prs-past ‘In the house of the demon, the wool would card itself.’ The analytic perfective detransitive form also serves to express imperative modality with spontaneous and inceptive meaning. For instance, in addressing his own instrument, a bard says: (51) c¸a-rt’ar-a sak, ay saz! pv-strike.detr-a be.imp Oh lute ‘Start playing, my lute!’ 5.5. Instantiation of the semantic role of ‘force’ While the passive interpretation (which preserves semantic roles) does not allow any oblique expression of the backgrounded agent, the anticausative, semantically intransitive interpretation of the detransitive voice is compatible with the expression of an argument in the semantic role of ‘force’. If this force is external, it is found in the subelative case, and normally appears in initial position: (52) (kulak.ci-kar) riki u-f-ar-a xhi-yic wind-subel door pv-close-detr-a be-aor.n ‘The door closed because of the wind.’ ic (53) (varag˘.ci-kar) za ¸ ugv-ar-a xhi-yic sun-subel 1.gen skin burn-detr-a be-aor.n ‘My skin was burned by the sun.’

These detransitive forms are not Patient-promoting and even less Agentdemoting: the corresponding transitive clauses are not ungrammatical, but seem extremely awkward to speakers:

The detransitive voice in Kryz

153

(54) ?kulak.ci-r riki u-fa-c wind-erg door pv-close-aor.n ‘The wind closed the door.’ ic (55) ?varag˘.ci-r za ¸ ugva-c sun-erg 1.gen skin burn-aor.n ‘The sun burned my skin.’

A force a¤ecting the subject internally may be expressed in the sublocative case in the following example (technical): yig˘-in.a ar sehirc¸i azar.ci-k yi-rqar-iya vun (56) g˘e  today day-inel magical illness-sub pv-catch.detr-fut 2 ‘From today on, you will be a¿icted with a magical disease.’ 5.6. Antipassive with ‘eat’, ‘drink’, and other verbs The detransitive forms of the two verbs ‘eat’ and ‘drink’ are ambiguous and can be interpreted as either Patient- or Agent-preserving, i.e. as anticausative or (more usually) as antipassive. The Agent-backgrounding, passive use is found in gnomic contexts: (57) lem-ird yak ugval-de-d, yux g˘i-g˘ar-de-d donkey-gen meat eat.detr-negprs-n milk red-drink.detr-negprs-n ‘The meat of the donkey is not eaten, nor is its milk drunk.’ but also in the perfective, if the Agent is unspecified: (58) ugval-a daxha yak xvar-imez sa- a-c eat.detr-a neg-be.pf(part) meat dog-pl.dat pv-throw-aor.n ‘The meat which was not eaten has been thrown to the dogs.’ (59) daug˘a g˘i-g˘ar-a xhi-d-u doogh red-drink.detr-a be-aor-f ‘The doogh has been drunk.’ The antipassive use also has a technical meaning: (60) u-be-k ibe bala-yar g˘i-g˘ar-e 3-npl-sub selfnplg child-pl red-drink.detr-prs ‘Their foals drink under them.’ (61) vul-bi ugval-ciz. . . ewe-pl eat.detr-simul ‘While the sheep were grazing. . .’

154

Gilles Authier

The perfective antipassive is inceptive: (62) vul-bi ma an halu huq’-a ugval-a xhi-yic ewe-pl again this meadow-in eat.detr-a be-aor.n ‘The sheep started to graze in the meadow again.’ (63) leh-ar g˘i-g˘ar-a xhi-yic calf-pl red-drink.detr-a be-aor.n ‘The calves have started to drink.’ Other spontaneous or quasi-reflexive (non-passive) readings of the detransitive voice are also found, which preserve non-controlling Agents such as animals or the radio, in processes involving the body or voice: (64) hava-yar ghira xhi-yic-zina kis-ar g˘i-nghan-e air-pl warm be-msd-instr hen-pl pv-lay.detr-prs ‘As soon as the weather becomes warm, the hens start laying eggs.’ (65) pirennik uxvar-e / uxvar-a xhi-yic radio recite.detr-prs recite.detr-a be-aor.n ‘The radio is on / started to play.’ (66) vul-bi c¸a-rhar-e ewe-abs.pl pv-soil.detr-prs ‘Sheep get dirty.’ Table 11. Detransitive verb forms with antipassive reading pf

ipf (tr.)

detr. ipf participles

g˘ira- ‘drink’

g˘ig˘-

g˘ig˘ar- 1) ‘be drinkable; 2) ‘start / be made to drink’

u(gu)la- ‘eat’

ul(ats’)-

g˘ighn- ‘lay (eggs)’

g˘ingh-

ugval- / 1) ‘be edible’; 2) ‘pasture’ g˘inghan-

c¸aha- ‘soil’ uxva- ‘read, sing’

c¸ac¸uxv-(ats’)-

c¸arharuxvar- 8

8. Recall that the analytic perfective forms (c¸arhara xhiyi; uxvara xhiyi) do exist, but with a passive interpretation. The form g˘inghana xhiyi means ‘having been knitted’.

The detransitive voice in Kryz

155

This tripartite morphology, with a ‘labile’ perfective form opposed to the imperfective transitive/detransitive pair, certainly reflects an archaic situation, as is shown by the existence of parallels in non-Lezgic branches of East Caucasian.

6. East Caucasian origins of the detransitive voice The passive voice is restricted to Kryz among the East Caucasian languages, which more commonly have derivations with anticausative or antipassive readings. The passive reading probably became prevalent only recently, under the influence of Azeri: it first appeared with verbs for which the event encoded cannot come about without an external causer (i.e. which are semantically Agent-oriented, cf. Haspelmath 1993) and as a result cannot be cast in the anticausative (with a Patient-oriented meaning component), and then it became an option with most other transitive verbs in Kryz. We may account for the fact that the main – passive – use of the su‰x -aR- in Kryz cannot be found in any other East Caucasian language if we bear in mind that ellipsis of the Agent is always available for the purpose of Agent backgrounding, and that other semantics linked to Kryz passive forms (modal or aspectual) are expressed independently of diathesis in other languages. However, a comparable detransitive marker seems to be attested as such in nearly all branches of the East Caucasian family. Although these languages have no passive constructions, they do have other detransitive voices with comparable morphology. 6.1. Tsezic, Avar and Dargi detransitives In other branches of the East Caucasian family, we find a significant sample of (non-productive) detransitive voices in Tsezic languages, apparently in at least one Dargic language, and in Avar. Avar is well known for possessing derived intransitive verbs with ‘iterative’ or ‘durative’ meaning, which show the morpheme -aR; they are clearly of the antipassive type:

156

Gilles Authier

Table 12. Transitive and iterative detransitive stems in Avar (Charachidze´ 1982) b-ets-ize

‘mow’

w-ec-ar-ize

‘be busy mowing’

b-uq’-ize

‘sew’

w-uq’-ar-ize

‘be busy sewing’

qver-ize

‘slay’

qveq-ar-ize

‘slay cattle for the winter’

as-ine

‘sting’

w-as-and-ize

‘dance, frolic’

heq’e-ze

‘drink’

heq’old-ize

‘be busy drinking’

b-its-ize b-etsts-ize

‘tell’ ‘praise’

bits-ard-ize v-etsts-ar-ize

‘reproach’ ‘praise oneself ’

(67) a.

b.

he-ł he-sda b-its-anila 50 tumen b-it’-i this-f.erg this-m.loc n-tell-pstevid 50 toman n-send-msd ‘She told him about the (sending of the) 50 tomans.’ ros he-lde kves¸ bitsard-anila husband this-f/n.lat bad reproach-pstevid ‘The husband was reproaching her bitterly.’

Tsez (see Comrie 2000) has an antipassive making use of a morpheme -na-: (68) a.

b.

Pat’aa uji esa-yxo Fatima.erg boy.abs(m) (m)wash-prs ‘Fatima washes the boy.’ Pat’i ker-aa y-esa-na-yxo Fatima.abs(f ) river-in f-wash-detr-prs ‘Fatima does the laundry in the river.’

Bezhta (van den Berg 2005, quoting Prof. Majid Khalilov, a native speaker) has a morpheme -lA:, which derives the antipassive form of certain verbs (note that the original absolutive-marked object can be expressed in the instrumental case in some of the new derived constructions): (69) a.

b.

o¨jdi qarandi y-o¨:t’o¨-yo¨ boy.erg hole.abs(n) n-dig-pst ‘The boy dug the hole.’ o¨jo¨ qarandi-yad boy.abs(m) hole-instr ‘The boy was digging at

o¨:t’o¨-la¨:-yo (m)-dig-detr-pst the hole.’

The detransitive voice in Kryz

157

Other verbs lose any ability to express a Patientive argument. Note the aspectual shift in the English translation: (70) a.

b.

kid-ba ha¨k’a¨ kl’eq’e-yo girl-erg boots sew-pst ‘The girl sewed the boots.’ kid kl’eq’e-la:-yo girl(abs) sew-detr-pst ‘The girl was sewing.’

Hunzib (closely related to Bezhta, see van den Berg 1994) shows two possible interpretations of the same morpheme -la:-, antipassive (ex. 71) and autocausative/reflexive (ex. 72): (71) a.

b.

(72) a.

b.

ołul bex kos¸e he. erg grass mow ‘He mows the grass.’ eg kos¸e-la: he.abs mow-detr ‘He mows (often, usually).’ iyul q’utila: zinkl’ay m-utsu-r mother.erg trunk.dat ring(a) a-hide-pst ‘Mother hid the ring in the trunk.’ kid q’utila: y-utsu-la:-r girl trunk-dat f-hide-detr-pst ‘The girl hid in the trunk.’

Table 13. Detransitive derivations attested in Tsezic transitive

Patient-preserving: anticausative

Agent-preserving: antipassive

Tsez ‘wash’

-esa-

-esa-na-

Bezhta ‘sew’

-o¨:t’o¨tl’eq’e-

-o¨:t’o¨-la¨:tl’eq’e-la:-

Bezhta ‘dig’ Hunzib ‘hide’ Hunzib ‘mow’

-utsukosˇe-

-utsu-la:kosˇe-la:-

158

Gilles Authier

Icari is a language of the Dargic branch, and can now be considered well-described thanks to Sumbatova and Mutalov (2003). It systematically marks intransitivity on stems with the allomorphs -ar-, -al-, -an- for third person subjects. These stems are attested in all dialects, and more commonly used as infinitives. I suspect that they are former impersonal stems, genetically linked with the Kryz detransitive voice. Table 14. Detransitive derivation in Icari Dargi tr.

3d person intr.

‘rot’ pf

ø

er’-ar-

‘rot’ ipf

ø

u’-ar-

‘break’ pf

elq’w-

elq’w-an-

‘break’ ipf

luq’w-

luq’w-an-

Unfortunately, the grammar provides no examples of sentences showing the possible contexts in which this detransitivization may take place. The great productivity of this derivational marker in Icari and Kryz may or may not represent a recent, parallel development. 6.2. Lezgic languages and the nominal hypothesis The detransitive derived stems are obviously old in Kryz itself: often they are not derived from the transitive imperfective stem currently in use. Transitive forms seem younger than their detransitive counterparts, as they contain a recent imperfective morpheme -ts’-: Table 15. Irregular detransitive derivation in Kryz thematic pf

ipf

DETR. ipf

‘cook’

cir-a-

cira-ts’-

cir-ar-

‘slay’

kur-a-

kura-ts’-

kur-ar-

‘shear’

vud-a-

vuda-ts’-

vud-ar-

‘weave’

xir-a-

xix- / xira-ts’-

xir-ar-

‘wash’

zim-a-

zima-ts’-

zim-ar-

compare, athematic: ‘tie’

vat’l-

valt’-

valt’-al-

The detransitive voice in Kryz

159

Furthermore, in some instances the detransitive verbal adjective retains strong gender agreement on the stem, whereas strong forms of agreement are normally found only on perfective stems:

‘yoke’

prs.f

aor.f

detr prs.f

detr aor.f

kilt’ilyu

kit’ildu

kilt’al-yu

kit’ala xhidu

As for the analytic perfective detransitive, it is certainly a recent creation in Kryz, prompted by the influence of the highly productive passive found in Azeri. The anticausative voice with labile perfective is older, as witnessed by the closely related Budugh language, which shows no trace of a passive construction but has a couple of anticausative imperfective forms directly comparable with those of the corresponding verbs in Kryz: Table 16. Budugh detransitive imperfective stems

‘cut’ ‘catch’

pf (labile)

ipf, tr.

detr. ipf

ya-t’sı-q-r

ya-r-t’sı-r-q-

yart’-ar sırq-ar

The su‰x -aR- is related to imperfective aspect markers, which in turn are akin to the nominal (collective) plural markers -r-, -l-, -n- found in all branches of the East Caucasian family. The following table shows that the three sonorants must be considered polycategorial plurality/iterative morphemes in Kryz. Table 17. Intransitivity, imperfectivity and plurality markers in Kryz nominal plurals

prefixes > ipf aspect

su‰xes > intransitivity

-r-

g˘ub-ri ‘frog-PL’

yi-r-q-

ke-rxhar- ‘move.ipf’

-l-n-

Ø xiy-ni ‘thread-PL’

ki-l-t’yi-n-gh-

la-lsal- ‘turn.ipf’ ha-ngvan- ‘run.ipf’

The detransitivizing su‰x -aR- is not restricted to finite verbs. Lezgian has a few verbal nouns (e.g. zw-er ‘running’, zw-al ‘boiling’) which display the same ‘intransitive’ su‰x, Agent- or Patient-oriented. We propose to recognize it also on a substantial set of Kryz nouns (note that some of these are also found in Rutul, from another branch of Lezgic):

160

Gilles Authier

Table 18. ‘Detransitive nouns’ in Kryz

-ar

-al

derived noun

cf. verbal root

xh-ar ‘wind’

ke-xh-r-ic ‘move’

gh-ar ‘snake’ c¸’-ar ‘overhang’

yi-gh-nic ‘draw, pull’ g˘a-c¸’-ic ‘exit, protrude’

z-ar ‘cow’ ¼ ‘milch cow, cf. Ger. melkende Kuh’ x-al ‘roof; cobweb’

a-z-ayc ‘milk’ x-irayc ‘weave’

We may now propose the following diachronic scenario: some of these intransitive deverbals came to be used as substantives, while others never became lexicalized. But the latter, used as predicates with semantic Patientor subject-orientation, evolved into anticausative or antipassive forms, which are not only preserved in the Southern sub-branch (Kryz and Budugh) of the Lezgic languages but also quite well distributed in a range of languages belonging to other branches. The last stage of this evolution in Lezgic is the Kryz passive use of this common East Caucasian detransitive voice (for the evolution of anticausative into passive markers, see for instance Haspelmath 1990).

Conclusion The Kryz passive is unique in East Caucasian, and its existence and productivity are apparently linked to associated modal and aspectual values, rather than to such syntactic features as an accusative pivot (although reference tracking predominantly operates on an accusative basis) or topic continuity. Its habitual interpretation in the older, imperfective forms is also a typical semantic property of antipassives, and indeed, other East Caucasian detransitive voices are not passive but antipassive in function. Other East Caucasian detransitive, Agent-preserving voices are also highly restricted in their use, more so than the passive in Kryz, and they have only been brought to light by recent descriptions based on original texts. But it is significant that, although rare, the instances of voice formation discussed here are distributed widely across the di¤erent branches of the East Caucasian family, meaning that a contact explanation is not likely. We therefore believe that this is a recessive, and in the case of Kryz, quite resilient feature inherited from Proto-East-Caucasian morphology.

The detransitive voice in Kryz

161

Abbreviations 1plecl -a ad adel adr all aor apud apudel deb dir el equ evt f h if int itr ipf msd neg nh notn pf ppf self sub subel supel super

1st person plural exclusive attribute adlocative adelative addressative allative aorist, on a pf stem apudlocative apudelative debitive directive elative case equative ‘eventual mood’, on an ipf stem singular human female, animals, plants and some animates human conditional converb interrogative intransitive imperfective masdar, on a pf stem negation non-human (pronominal su‰x) non-neuter perfective perfective participle reflexive pronoun sublocative subelative superelative superlocative

References Authier, Gilles 2008

‘‘Split Ergativity on Kryz Pronouns’’ In Cahiers du CELIA: Ergativity in Amazonia, III: Coreference. ed. Francesc Queixalos.

162

Gilles Authier

Authier, Gilles 2009 Grammaire kryz. Paris/Leuwen: Peeters. van den Berg, Helma 1994 Hunzib. Munich: Lincom. van den Berg, Helma 2001 Dargi Folktales. Leiden: Universiteit Leiden, Research School of Asian, African and Amerindian Studies. van den Berg, Helma 2005 ‘‘The North-East Caucasian Languages’’. Lingua 115(1). Charachidze, Georges 1982 Grammaire de la langue avar Paris: Jean Favard. Comrie, Bernard 2000 ‘‘Valency-changing derivations in Tsez’’. In Robert M.W. Dixon and Alexandra Y. Aikhenvald, eds.: Changing Valency: Case Studies in Transitivity, 360–374. Cambridge: Cambridge University Press. Dixon, Robert M.W. 1994 Ergativity, Cambridge: Cambridge University Press. Dixon, Robert M.W. and Aikhenvald, Alexandra Y. 2000 Changing Valency. Cases studies in transitivity Cambridge: Cambridge University Press. Givo´n, Tom 1994 ‘‘The pragmatics of de-transitive voice: functional and typological aspects of inversion.’’ In Givo´n, T. (ed.), Voice and Inversion. Amsterdam/Philadelphia: John Benjamins, 3–44. Haspelmath, Martin 1990 ‘‘The grammaticization of passive morphology’’. In Studies in Language 14.1: 25–71. Haspelmath, Martin 1993 ‘‘More on the typology of inchoative/causative verb alternations’’. In Bernard Comrie and Maria Polinsky, (eds.) Causatives and transitivity. (Studies in Language Companion Series, 23.) Amsterdam/Philadelphia: Benjamins, 87–120. Haspelmath, Martin 1994 ‘‘Passive participles across languages’’. In: Barbara Fox and Paul J. Hopper (eds.) Voice: Form and Function. (Typological Studies in Language, 27.) Amsterdam/Philadelphia: Benjamins, 151– 177. Kazenin, Konstantin 2001 ‘‘Passive’’. In: Martin Haspelmath, Ekkehardt Ko¨nig, Werner ¨ sterreicher and Wilhelm Raible. Language Typology and LanO guage Universals. An International Handbook. Berlin/New York: DeGruyter, 970–978. Kemmer, Susan 1993 The middle voice. Amsterdam/Philadelphia: Benjamins.

The detransitive voice in Kryz

163

Kibrik, Aleksandr E. 1997 ‘‘Beyond subject and object: Toward a comprehensive relational typology.’’ Linguistic Typology 1. 1997: 279–346. Malchukov, Andrej 2006 ‘‘Transitivity parameters and transitivity alternations: constraining co-variation’’. In Leonid Kulikov, Andrej Malchukov and Peter de Swart (eds.). Studies on case, valency and transitivity, 329–359. Amsterdam/ Philadelphia: Benjamins. Malchukov, Andrej 2005 ‘‘Case Pattern Splits, Verb Types and Construction Competition.’’ In: Mengistu Amberber and Helen de Hoop (eds.) Competition and Variation in Natural Languages: The Case for Case. Amsterdam, Boston: Elsevier, 73–118. Nichols, Johanna 1992 Linguistic Diversity in Space and Time. Chicago: University of Chicago Press. Saadiev, Shamseddin 1994 ‘‘Kryts’’. In: Rieks Smeets (ed.) The Indigenous Languages of the Caucasus, vol. IV. New York: Delmar, Caravan. Sumbatova, Nina and Mutalov, Rasul 2003 A Grammar of Icari Dargi. Munich: Lincom.

Laz middle voice Rene´ Lacroix 1. Introduction The South Caucasian languages (Laz, Mingrelian, Georgian and Svan, also called the Kartvelian languages) share a common morpheme, i-, which appears in finite verb forms immediately before the root. This paper aims to present the functions of i- in Laz. As an introduction to the problem, let us consider the ways i- is commonly described for Georgian, the bestknown language in the family. In Georgian grammars (Sˇani‰e 1953, Tschenke´li 1958, Hewitt 2005, Aronson 1990), no unified account is given of the morpheme i-. Consider for instance example (1). In (1b), i- indicates that the action takes place for the benefit of the subject himself: in this use it is often labelled ‘‘subjective version’’ (subj.vers). In (1a), by contrast, the verb does not specify any beneficiary; the morpheme a-, which appears in the same slot as i-, is labelled ‘‘neutral version’’ (neutr.vers). The glossing of the pronominal a‰xes in the verb is explained in section 2; thematic su‰xes (ths) such as -eb appear only in certain tenses, including the present and the imperfect. They do not have any clear semantic value.1 (1) a.

b.

Me a-v-a-sˇen-eb saxl-s. 1s pv-i1-neutr.vers-build-ths house-dat ‘I will build a house.’ (Tschenke´li 1958: 247) Me a-v-i-sˇen-eb saxl-s. 1s pv-i1-subj.vers-build-ths house-dat ‘I will build a house for myself.’ (Tschenke´li 1958: 247)

1. I would like to thank Winfried Boeder, Denis Creissels, Antoine Guillaume and Katharina Haude for their valuable comments on earlier drafts of this article. Thanks are due to my informants as well. In the transcription of Laz, the apostrophe marks glottalized consonants; when several glottalized consonants follow each other, I indicate glottalization on the last one only. ‰Ł corresponds to [d‰], ‰ to [dz] and c to [ts]. Since /r/ often drops in Laz, certain morphemes may appear with /r/ in some examples but not in others.

166

Rene´ Lacroix

There is a class of verbs, often labelled ‘‘passives’’, which are also marked by i-. Compare (2b) and (3b) with their transitive counterpart in (2a) and (3a). (2) a.

b.

(3) a.

b.

St’udent’-ma da-c’er-a c’erili. student-erg pv-write-aor.i3s letter ‘The student has written the letter.’ (Tschenke´li 1958: 254) C’erili da-i-c’er-a st’udent’-is mier. letter pv-pass-write-aor.i3s student-gen by ‘The letter has been written by the student.’ (Tschenke´li 1958: 254) Is cˇ’am-s xil-s. dem eat-i3s fruits-dat ‘He eats fruits.’ (Tschenke´li 1958: 255) Es xili ar i-cˇ’m-eb-a. dem fruits neg pass-eat-ths-i3s ‘These fruits cannot be eaten / are not edible.’ (Tschenke´li 1958: 255)

Finally, with certain verbs, the morpheme i- is found in the formation of some tenses such as the future and the aorist. I gloss it with tns for ‘‘tense’’. (4) a.

v-t’rial-eb i1-spin-ths ‘I spin’

b. v-i-t’rial-eb i1-tns-spin-ths ‘I will spin’ (Tschenke´li 1958: 295)

The grammar of Georgian has long served as a model for the analysis of the other languages of the family. According to Holisky (1991), for instance, in Laz the morpheme i- marks ‘‘subjective version’’ in diviboni ‘I washed myself ’ (p. 438); on the other hand, this morpheme serves to derive intransitive forms which ‘‘have passive meaning or express possibility’’ (p. 422), one of the examples given being igiben ‘it boils’. This paper aims at presenting the functions of Laz i- in a unified manner. It will be argued that the uses of i- correspond to what has been labelled middle voice in other languages. Laz is spoken in north-eastern Turkey. Estimates of the number of speakers range from 45,000 (Andrews 1989) to 500,000 (Holisky 1991). Laz is an unwritten and endangered language; although young people still understand it, they speak only Turkish, with very few exceptions.

Laz middle voice

167

ˇ ikobava 1936), Laz is divided According to some scholars (Marr 1910, C into three main dialects. Kutscher (2001) distinguishes four dialects. The data presented here is from the dialect of the town of Arhavi and neighbouring villages. My study is based on a corpus of published Laz texts (Dume´zil 1937, 1967, and 1972, Zˇ'ent’i 1938, Q’ipsˇi‰e 1939, K’art’ozia 1972 and 1993) as well as data from my work with native speakers. Unless otherwise stated, the examples are from my informants. In section 2, I present the coding properties of A, O and S arguments; in section 3, I briefly summarize the functions of the verbal valency operators; I then propose a classification of Laz verbs (section 4), which will serve as a basis for the presentation of the di¤erent uses of the middle marker i- in section 5; in section 6, I discuss other accounts of the functions of i- in the Kartvelian languages, before the conclusion in section 7.

2. Coding of syntactic functions In Arhavi Laz, syntactic functions are indicated by cases and crossreferencing a‰xes. There are two sets of cross-referencing a‰xes, glossed with Roman numerals. Set I cross-references the A and S arguments, while Set II cross-references the O argument.2 Sentence (5) illustrates the basic transitive construction. The A argument (bere) is in the ergative and is cross-referenced on the verb by a Set I a‰x (-u); the O argument (ocxo‰Ł ) is in the absolutive and is cross-referenced by a set II a‰x: (5) Bere-k ocxo‰Ł me-Ø-tk’ocˇ-u. boy-erg comb pv-ii3-throw-aor.i3s ‘The boy threw the comb.’ (Dum37.10.3) First and second person objects, unlike third person objects, are overtly cross-referenced on the verb (see ex. 9). In the terminology used here, ‘‘object’’ is taken in the restrictive sense of ‘‘direct object’’; it is synonymous with ‘‘O argument’’. The term ‘‘subject’’ will be used to cover both A and S arguments, which exhibit the same behavioural properties.

2. Kartvelologists often use the terms ‘‘subject’’ and ‘‘object’’ a‰xes. The problem with this terminology is that there are constructions where the argument crossreferenced by ‘‘object’’ a‰xes exhibits subject properties, such as number agreement in the verb. For this reason, I prefer a more neutral terminology.

168

Rene´ Lacroix

The S argument is cross-referenced by Set I a‰xes. Certain intransitive verbs take an absolutive subject (ex. 6) while others take an ergative subject (ex. 7). (6) Mk’yapu xrock-u-n. jackal die-ths-i3s ‘The jackal is dying.’ (Zˇ'.96.26) (7) K’ocˇi-k cˇind-um-s. man-erg sneeze-ths-i3s ‘The man sneezes.’ Comparison of (6) with (5) shows that the alignment of absolutive-S verbs is of the mixed type: S behaves like O with respect to case marking (both are in the absolutive), but it behaves like A with respect to crossreferencing (both are cross-referenced by Set I a‰xes). On the other hand, comparison of (7) with (5) shows that the alignment of ergative-S verbs is accusative: S behaves like A in terms of case marking (both are in the ergative) and cross-referencing (both are cross-referenced by Set I a‰xes). Strictly speaking, then, the major alignment types of Arhavi Laz are not ergative, but mixed for absolutive-S verbs and accusative for ergative-S verbs. There is an alignment split between nouns and third person pronouns on the one hand and first/second person pronouns on the other, as these have the same form in the ergative, absolutive and dative cases: Table 1. Declension of first and second person pronouns

ergative, absolutive and dative

1sg

ma

2sg

si

1pl

cˇku

2pl

tkva

The alignment of first and second person pronouns is thus neutral with respect to case marking. First and second person cross-referencing, however, remains accusative. The examples below illustrate the use of the second singular pronoun si in A, O and S functions respectively. In the paradigm of Set I cross-referencing a‰xes, the second person singular is the only cell which bears no overt marking (except in the future).

Laz middle voice

169

(8) Si mu cˇu-me? 2s what wait-ths ‘What are you waiting for?’ (Dum67.XX.42) (9) Ma si e-k-cˇ ’op-are. 1s 2s pv-ii2-marry-fut.i1/2s ‘I will marry you.’ (Dum37.121.6) (10) Si guruni ye-i? 2s donkey be-int ‘Are you a donkey?’ (K’art’72.193.26) Basic word order is SOV. Word order does not indicate grammatical functions, but instead reflects information structure. The dialect of Ardes¸en has lost the ergative and dative cases (Dume´zil 1972, Kutscher 2001). Compare in this respect (11a), taken from Ardes¸en Laz, with (11b), from Arhavi. In (11a), the subject mtuti ‘bear’ and the applicative argument arkadasˇi-musˇi ‘his friend’ are not case-marked, while in (4b), the subject is in the ergative and the applicative argument in the dative (for more details on the applicative construction, see below). (11) a.

b.

Mtuti arkadasˇi-musˇi u‰Ł i k-el-u-d-u. bear friend-poss3s ear pv-pv-ii3.appl-put-aor.i3s ‘The bear put his ear on his friend’s side.’ (Dum72.4.5) Mtuti-k arkadasˇi-musˇi-s u‰Ł i el-u-d-u. bear-erg friend-poss3s-dat ear pv-ii3.appl-put-aor.i3s ‘The bear put his ear on his friend’s side.’ (Dum72.4.5)

Ardes¸en Laz thus di¤ers from the other dialects in that no cases are used to encode core arguments; cross-referencing, however, functions as in the other dialects. The alignment system of Laz di¤ers from that of the other Kartvelian languages. Throughout the family, cross-referencing functions in roughly the same way, but case marking di¤ers (Harris 1991b: 3.2.2 and 4.2.1). Alignment in these languages cannot be presented here in detail. Su‰ce it to say that ergative, mixed and accusative alignments are found with nouns and third person pronouns; first and second person pronouns have the same form in the absolutive, ergative and dative cases, as in Laz.

170

Rene´ Lacroix

3. Valency operators The morphological structure of finite verb forms in Laz may be summarized as follows: -3 -2 -1 0 1 2

preverbs cross-referencing prefixes valency operators root causative TAM and cross-referencing su‰xes

The immediately pre-root slot (–1) may contain one of five valency operators, each of them consisting of one vowel: o- indicates that the verb is transitive, and is also used to derive applicative forms; i- marks middle voice; i/u- mark mainly applicative derivation; a- marks, among other things, applicative derivation and middle voice simultaneously. No two valency operators may occur together. Such markers are often called ‘‘preroot vowels’’ or ‘‘preradical vowels’’ in Kartvelian linguistics. As has been pointed out to me by Denis Creissels and Antoine Guillaume, such terms refer to the phonological rather than the morphosyntactic level. Since the general function of these markers is to indicate the valency of the verb they occur in, the term ‘‘valency operator’’ will be preferred here. As was mentioned above, the functions of the valency operator i- are generally not given a unified account in the literature on South Caucasian languages. I will return to this point in section 6. In this paper, I show that the valency operator i- in Laz functions as a middle voice marker. Consider for instance example (12). In (12b), the addition of i- to the verb yields an intransitive verb with an anticausative reading – one subtype of the middle. (12) a.

b.

Bozo-k nek’na ge-nk’ol-um-s. girl-erg door pv-close-ths-i3s ‘The girl closes the door.’ Nek’na ge-i-nk’ol-e-n. door pv-mid-close-ths-i3s ‘The door closes.’

The term ‘‘middle’’ is taken here in a broad sense, corresponding to that given in Kemmer (1993). It covers intransitive constructions such as the

Laz middle voice

171

anticausative and the autocausative, but also includes some transitive constructions, as in ‘I build a house for myself ’.3 The uses of the valency operator i- are the focus of this article and are examined in more detail in section 5. Transitive verbs may be classified into two groups according to whether they involve the valency operator o- or not. This operator (glossed tr for ‘‘transitive’’) does not correlate with any semantic distinction; its presence is lexically conditioned. It corresponds etymologically and functionally to the ‘‘neutral version’’ of Georgian grammar (see example [1a]). Some examples are given below. Some transitive verbs which do not employ the valency operator op-sˇinax-am p-kos-um me-m-‰Ł'on-am p-xazi-um p-cˇ’ar-um

i1-hide-ths i1-sweep-ths pv-i1-send-ths i1-prepare-ths i1-write-ths

‘I ‘I ‘I ‘I ‘I

hide sth’ sweep sth’ send sth’ prepare sth’ write sth’

Some transitive verbs which employ the valency operator ob-o-rd-am b-o-rg-am b-o-gzal-em b-o-cˇil-am go-b-o-kt-am

i1-tr-raise-ths i1-tr-plant-ths i1-tr-send-ths i1-tr-marry-ths pv-i1-tr-turn-ths

‘I ‘I ‘I ‘I ‘I

raise sb’ plant sth’ send sb’ marry him (to a girl)’ turn sth’

Causative verb forms contain the vowel o- by default: (13) o-mt’-in-am-s / *mt’-in-am-s tr-flee-caus-ths-i3s ‘he makes it flee, he chases it away’ (Dum67.XXIII.8) In examples such as the following, the fact that the middle marker is added to a form devoid of any explicit valency marking can be viewed as evidence that the direction of the derivation is from the non-middle to the middle:

3. Thus, the term ‘‘middle’’ is not taken here in the sense it bears in traditional Georgian grammar, for instance in Tschenke´li (1958, Lesson 28: ‘‘Mittelverben’’).

172

Rene´ Lacroix

non-middle > middle p-xazi-um b-i-xazi-am i1-prepare-ths i1-mid-prepare-ths ‘I prepare something’ ‘I prepare myself ’ In contrast, when the transitive verb shows the vowel o-, the fact that both forms are of equal morphological complexity means that there is no way to posit a direction of derivation on purely morphological grounds: non-middle < > middle b-o-cˇil-am b-i-cˇil-e i1-tr-marry-ths i1-mid-marry-ths ‘I marry him (to a girl)’ ‘I am getting married (to a girl)’ Throughout the present article, I will use the terms ‘‘base verb’’ to refer to a non-middle verb possessing a middle counterpart, and ‘‘base subject’’ to refer to the subject of a base verb. These terms have no theoretical significance, since a ‘‘base’’ verb like bocˇilam is equal in morphological complexity to its middle counterpart bicˇile, and therefore is not morphologically more basic. The terms ‘‘base verb’’ and ‘‘base subject’’ are used merely for simplicity of description. The valency operators i/u- are used, among other things, to derive applicative forms. The applicative construction is illustrated in (14b) below, and can be compared with (14a), the corresponding non-derived construction. The applicative argument Xasani is in the dative and is crossreferenced by a Set II a‰x (u-).4 It may be considered as a core argument on the basis that it is cross-referenced on the verb. (14) a.

b.

Hemu-k oxoi k’od-um-s. demp-erg house build-ths-i3s ‘He builds a house.’ Hemu-k Xasani-s oxoi u-k’od-um-s. demp-erg Hasan-dat house ii3.appl-build-ths-i3s ‘He builds a house for Hasan.’

4. The applicative construction in Laz (except in the Ardes¸en variety) and in the other South Caucasian languages di¤ers from prototypical applicatives in that the applicative argument is not case-marked in the same way as O (Lacroix 2010; Peterson 2007: 1). The operator u- as illustrated in example (14b) is known in the literature on Kartvelian languages as ‘‘objective version’’.

Laz middle voice

173

The applicative marker is realized as u- when the applicative argument is third person and i- when it is first or second person: m-i-k’od-um-s g-i-k’od-um-s u-k’od-um-s

ii1-appl-build-ths-i3s ii2-appl-build-ths-i3s ii3.appl-build-ths-i3s

‘he builds it for me’ ‘he builds it for you’ ‘he builds it for him’

The applicative argument may express semantic roles such as beneficiary, maleficiary, recipient, allative and possessor. The valency operator o- also is used to derive applicatives. Although it is homonymous with the o- indicating transitive verb forms (see above), the two must be distinguished on formal grounds. Compare sentence (15a), which involves a non-derived intransitive verb ( ‰Ł ant’u), and (15b), where the same root appears in an applicative verb form, marked by the operator o-. The applicative argument is nek’na-s ‘door-dat’. (15) a.

b.

Xasani dido ora-s ‰abuni ‰Ł an-t’u. Hasan much time-dat sick lie-impft.i3s ‘Hasan was lying sick for a long time.’ (Zˇ'.5.8) Didi kva n-o-‰Ł an-t’u nek’na-s. big stone pv-appl-lie-impft.i3s door-dat ‘There was a big stone against the door.’ (K’art’72.166.6)

Two di¤erences can be pointed out between o- indicating transitive verb forms and o- indicating the applicative. First, verb forms with transitive o- (other than causatives) cannot be shown to derive from simpler forms. By contrast, verb forms like no‰Ł ant’u in (15b) result from the addition of o- to a more basic verb. Secondly, transitive o- indicates the presence of an absolutive argument (the object), while applicative o- indicates the presence of a dative argument (the applicative argument). The valency operator a- is used, among other things, to derive applicatives from middle verbs. Compare (12b) above with (16), which involves the addition of a dative argument, bicˇ’is ‘boy’, bearing the semantic role of maleficiary: (16) Bicˇ’i-s ixi-te nek’na gy-a-nk’ol-u. boy-dat wind-instr door pv-appl.mid-close-aor.i3s ‘Because of the wind, the door closed on the boy.’ 4. Verb classes Finite verb forms in Laz may be classified into two broad groups, according to the shape of the Set I third person singular su‰x: Class 1 verbs take

174

Rene´ Lacroix

the su‰x -s, while Class 2 verbs take -n. This classification is also reflected in the shape of the thematic su‰xes: Class 1 verbs either take one of the thematic su‰xes -am, -em, -im, -om, -um or -mer or take no thematic su‰x at all, while Class 2 verbs take the thematic su‰xes -er, -ir, -ur (/r/ drops before a su‰x beginning with /n/). This purely morphological classification has syntactic correlates: while Class 1 contains both transitive and intransitive verbs, Class 2 verbs are exclusively intransitive. Most intransitive Class 1 verbs take an ergative subject, while most Class 2 verbs take an absolutive subject. This is summarized in the chart below. Table 2. Characteristics of Class 1 and 2 verbs Class 1

Class 2

Set I third singular su‰x

-s

-n

thematic su‰xes

-Vm, -mer, -Ø

-Vr

transitivity

tr. and intr.

intr.

case of the subject

mostly ergative

mostly absolutive

Class 2 is made up primarily of middle verbs. Some examples of nonmiddle Class 2 verbs are: -ul- ‘go’, illustrated in example (46a), -x- ‘be sitting’, -dg- ‘be standing’ and -'ur- ‘die’. Such verbs do not take any valency operator. 5. Classification of middle verb forms Morphologically, middle verb forms may belong to Class 1 or Class 2. I illustrate this with the root -kun- (þpreverb dolo-) ‘put on (clothes)’. From this root, a ditransitive verb may be formed: (17) Bozo-k bee-s porcˇa ko-dol-o-kun-am-s. girl-erg child-dat shirt pv-pv-tr-put.on-ths-i3s ‘The girl dresses the child in the shirt.’ The presence of the Set I third person singular su‰x -s and the thematic su‰x -am indicates that this verb belongs to Class 1. We also see that this verb exhibits the valency operator o- characteristic of transitive verbs (see section 3).

Laz middle voice

175

To mark coreferentiality between the agent and the recipient, the verb takes the valency operator i-, which replaces the operator o-. The resulting middle verb remains a Class 1 transitive verb: (18) Bee-k porcˇa ko-dol-i-kun-am-s child-erg shirt pv-pv-mid-put.on-ths-i3s ‘The child puts the shirt on.’ An intransitive verb with negative potential reading may be formed from the same root. The former object porcˇa ‘shirt’ appears in subject position. No agent is involved: (19) Ha porcˇa va dol-i-kun-e-n. demd shirt neg pv-mid-put.on-ths-i3s ‘This shirt cannot be put on [because it is dirty, too small, etc].’ The presence of the Set I third person singular su‰x -n and the thematic su‰x -e indicates that this verb belongs to Class 2. In the remainder of this section, I present the di¤erent uses of the middle marker i-, examining first Class 1, then Class 2 verbs. 5.1. Class 1 middle verb forms 5.1.1. Subject-Object coreference In the subject-object coreference construction, the subject of the middle verb undergoes the same process as the object of the base verb, but is at the same time the initiator of the process denoted by the verb. Firstly, we find autocausative verbs (Creissels 2006: 29). One example is given in (20b) below. As in the remainder of this article, the corresponding non-middle construction is illustrated in the a. sentence. The subject of (20b), nana ‘mother’, cumulates the roles represented by the subject and the object of (20a). (20) a.

Nana-k bee-musˇi o-n‰Ł i-am-s. mother-erg child-poss3s tr-put.to.bed-ths-i3s ‘The mother puts her child to bed.’5

5. In such examples, it is often di‰cult to find a satisfactory English gloss for the verbal root: a choice must be made between ‘‘simple’’ and ‘‘causative’’ glosses (‘go to bed’ and ‘put to bed’) which is not reflected in the Laz root itself. The gloss used in this example has been chosen arbitrarily.

176

Rene´ Lacroix

b.

Nana i-n‰Ł i-s. mother mid-put.to.bed-i3s ‘The mother goes to bed.’

Both verbs belong to Class 1, as indicated by the pronominal su‰x -s and the thematic su‰xes. The verb in (20a) is transitive: this is indicated by the operator o-. The verb in (20b) is intransitive. Other examples are: eysels ‘he gets up’ (compare the corresponding non-middle yoselams ‘he makes him get up’); oxink’ans ‘he moves about’ (cp. oxonk’anams ‘he moves sth’). Although the autocausative verbs illustrated above generally appear with an absolutive subject, there are some rare occurrences of ergative subjects in the corpus of published Laz texts. One of these is given in (21) below. The use of the ergative in this example, as well as in (20b) above, is not accepted by my informant. (21) Nana-musˇi-k e-y-sel-u. mother-poss3s-erg pv-mid-stand.up-aor.i3s ‘His mother stood up.’ (Zˇ'.102.6) As further instances of the subject-object coreference construction, we find body care verbs: (22) a.

b.

Bozo-k bee bon-um-s do cxon-um-s. girl-erg child wash-ths-i3s and comb-ths-i3s ‘The girl washes and combs the child.’ Bee-k i-bon-s do i-cxon-s. child-erg mid-wash-i3s and mid-comb-i3s ‘The child washes [himself ] and combs his hair.’

As we will see in section 5.1.3, body care verbs may also be used with an object NP designating the body part being acted upon. Autocausative and body care verbs belong to ‘‘body actions’’, a typical sub-category of the middle (Kemmer 1993: 53). Coreferentiality between the subject and the object of a transitive verb may also be expressed through the reflexive NP ti-cˇkimi ‘my head’ (ti-skani ‘your head’, etc.):6 6. In Kartvelology, this kind of construction is known as ‘‘tavization’’ (from Georgian tavi ‘head’). It has been discussed on the basis of Georgian material (see, among others, Braithwaite 1973). The details of this construction in Laz, however, require further research.

Laz middle voice

177

(23) K’ocˇi-k ti-musˇi il-om-s. man-erg head-poss3s kill-ths-i3s ‘The man kills himself (lit.: his head).’ Thus, Laz appears to be a two-form language in Kemmer’s sense in that it has a reflexive NP in addition to the middle marker i-.7 The fact that the use of the reflexive NP in object position appears almost exclusively with expressions meaning ‘kill oneself ’ is consistent with Kemmer’s observation that in two-form languages, the heavier marker (in this case the reflexive NP) is used with reflexive verbs which involve high distinguishability of participants, while the lighter marker (operator i-) appears with verbs with a lower degree of distinguishability of participants. Both the reflexive NP and the operator i- may appear in the same construction, as shown in (24) below. This construction can be analysed as an example of ‘‘object possession’’ (see 5.1.3). (24) Ti-cˇkimi do-b-i-'ur-in-am. head-poss1s pv-i1-mid-die-caus-ths ‘I will kill myself.’ (Dum67.XXIX.4) 5.1.2. Subject-Dative coreference As we have seen in section 3, the beneficiary may be expressed as an applicative argument (Xasanis in ex. 25). (25) Hemu-k Xasani-s oxoi u-k’od-um-s. demp-erg Hasan-dat house ii3.appl-build-ths-i3s ‘He builds a house for Hasan.’ When the subject is coreferential with the applicative argument, the middle marker i- appears on the verb and replaces any other valency operator: (26) Hemu-k oxoi i-k’od-um-s. demp-erg house mid-build-ths-i3s ‘He builds a house for himself.’ (27) Ma oxoi b-i-k’od-um. 1s house i1-mid-build-ths ‘I build a house for myself.’ 7. Two-form languages are opposed to one-form languages, where the reflexive and reciprocal markers are morphologically identical to the middle marker. An example of the reciprocal pronoun artikati is given in (57).

178

Rene´ Lacroix

In this construction, the subject cumulates the semantic roles of agent and beneficiary, expressed in the corresponding applicative construction (25) by the subject and the applicative argument. The constructions in (25) and (26) respectively can be schematized as follows (‘‘Appl’’ stands for ‘‘applicative argument’’): Table 3. Applicative construction Syntactic level

Aerg

Oabs

Appldat

Semantic level

agent

theme

beneficiary

Table 4. Middle construction Syntactic level

Aerg

Oabs

Semantic level

agent þ beneficiary

theme

Dative core arguments may also appear in non-applicative constructions, as for instance with the ditransitive verb dolo-o-kun- ‘put on (clothes)’ illustrated in (17) above. Another example is mo-o-k’id- ‘hang sth on sth’:8 (28) a.

Xasani-k cˇxomi cxeni-s m-o-k’id-am-s. Hasan-erg fish horse-dat pv-tr-hang-ths-i3s ‘Hasan hangs the fish on the [back of the] horse.’

In this example, the dative argument cxeni ‘horse’ fulfils the role of goal. Correspondingly, the subject of the related middle construction cumulates the semantic roles of agent and goal: b.

Xasani-k cˇxomi mo-i-k’id-am-s. Hasan-erg fish pv-mid-hang-ths-i3s ‘Hasan hangs the fish on his own back.’

Another example is Ar cˇ’uvali dik’a kec’ib'it ‘spill under yourself a sack of wheat’ (Dum67.II.15).

8. The fact that the dative argument of these verbs belongs to the core becomes clear in the first or second person, where it is overtly cross-referenced on the verb. When citing verbs, I give the preverb (if any), the valency operator (if any) and the root. In mo-o-k’id, for instance, mo- is the preverb, o- the operator and -k’id- the root.

Laz middle voice

179

The body part or item of clothing towards which the action is directed may be added in the construction as a dative oblique:9 (29) Xasani-k cˇxomi k’ap’ula-s mo-i-k’id-am-s. Hasan-erg fish back-dat pv-mid-hang-ths-i3s ‘Hasan hangs the fish on his own back.’ (30) T’abak’a do mendil snu¤box and handkerchief yele'i-sˇ ‰Ł eb-epe-s dol-i-du-mer-nan. vest-gen pocket-pl-dat pv-mid-put-ths-i3p ‘They put into the pockets of their vest their snu¤box and their handkerchief.’ (Dum37.128.6) Subject-dative coreference also includes reciprocal events. Sentence (31a) illustrates the verb ela-purcˇin- ‘whisper’ in its basic form; (31b) gives the same verb in the applicative derivation; (31c) illustrates the corresponding middle form. (31) a.

b.

Bee-pe-k muntxa ela-purcˇin-am-an. child-pl-erg something pv-whisper-ths-i3p ‘The children whisper something.’ Bee-pe-k bozo-pe-s muntxa el-u-purcˇin-am-an. child-pl-erg girl-pl-dat something pv-ii3.appl-whisper-ths-i3p ‘The children whisper something to the girls.’

c. Bee-pe-k muntxa el-i-purcˇin-am-an. child-pl-erg something pv-mid-whisper-ths-i3p ‘The children whisper something to each other.’ Coreference between the subject and a dative complement may also be expressed by the reflexive NP in the applicative argument position: (32) K’ocˇi-k ti-musˇi-s u-c’u-me-s-ki. . . man-erg head-poss3s-dat ii3.appl-tell-ths-i3s-comp ‘The man says to himself: . . .’ 9. Such examples could alternatively be analysed as instances of oblique possession constructions, whereby the valency operator i- indicates that a relationship of possession exists between the subject and the dative oblique (compare with the object possession construction in section 5.1.3).

180

Rene´ Lacroix

The translation of some of the examples above would involve a middle verb form in French: Je me construis une maison (27), Les enfants se chuchotent quelque chose (31c). 5.1.3. Object possession The object possession construction is illustrated in (33b) below. The subject hemuk is the possessor of the object pantoloni. (33) a.

b.

Hemu-k bere zin‰Ł ir-epe-te k’o-um-s. demp-erg child chain-pl-instr bind-ths-i3s ‘He ties the child up with chains.’ Hemu-k pantoloni gel-i-k’o-am-s. demp-erg trousers pv-mid-bind-ths-i3s ‘He ties his [own] trousers up.’

This is a type of ‘‘external possession’’ (or ‘‘possessor raising’’) construction: the possessor is not expressed as a possessive determiner in the possessed NP (‘his trousers’), but appears as a distinct NP, and the relation of possession follows from the presence of the operator i-. As a subtype of the object possession construction, we find body care verbs in constructions where the body part is overtly expressed as an object NP. (34) a. b.

see (22a) above Bozo-k xua i-bon-s. girl-erg body mid-wash-i3s ‘The girl washes.’

Such uses of a middle marker are known in other languages as well (cf. French il se lave les mains ‘he is washing his hands’). Although subjects of body care verbs normally appear in the ergative, we find the following two exceptional sentences: (35) Bozo nuk’u d-i-bon-u. girl face pv-mid-wash-aor.i3s ‘The girl washed her mouth.’ (Zˇ'.63.6) (36) K’ulan-epe nuk’u i-bond-es. girl-pl face mid-wash-aor.i3p ‘The girls washed their faces.’ (Zˇ'.151.17)

Laz middle voice

181

These are the only examples of the ‘‘double-absolutive’’ construction I have found in my corpus. They are not accepted by my informant, who uses an ergative subject instead. I propose a possible explanation for this exceptional pattern below (5.3). Other examples of the object possession construction are: K’iti mcˇk’oni nik’vatu ‘He cut his little finger’ (Zˇ'.5.30); Pesˇkirite nuk’u bikosare ‘I will wipe my face with a towel’ (Zˇ'.60.20); Divik k’ibirep dilasirudort’un ‘The giant had sharpened his teeth’ (Dum37.85.12). The object possession construction appears primarily with body parts, clothes and weapons, but it seems also to be possible with other kinds of objects, for instance in k’ocˇik araba dicˇxu ‘the man washed his car’ (elicited example). 5.1.4. Antipassive Example (37) illustrates the construction of the ditransitive verb do-o-gur‘teach sth to sb’. (37) Padisˇahi-k zˇur cˇ’ut’al-epe-s zanaxat’ d-o-gur-am-t’u. sultan-erg two little-pl-dat trade pv-tr-learn-ths-impft.i3s ‘The sultan taught a trade to his two youngest sons.’ (Dum37.1.1) The corresponding middle verb do-i-gur- ‘learn sth’ (ex. 38) illustrates the subject-dative coreference construction (see section 5.1.2). This verb is monotransitive. (38) Bere-k ir child-erg all ‘The boy had

sˇey ko-d-i-gur-u-dort’un. thing pv-pv-mid-learn-aor.i3s-pperf learnt everything.’ (Dum37.56.12)

Example (39) shows another use of the middle verb do-i-gur-. The subject (hentebe) is in the absolutive and the verb takes no object. (39) Hentebe i-gur-am-t’es Amerik’a-s. demp.pl mid-learn-ths-impft.i3p America-dat ‘They studied in America.’ (Zˇ'.103.24) Comparison of (38) and (39) reveals an antipassive pattern: in (39), the object of the transitive construction in (38) disappears and the subject stands in the absolutive. The construction thus becomes intransitive. The construction in (39) cannot, however, be analysed as a prototypical antipassive (Dixon and Aikhenvald 2000: 9), as no overt marking distinguishes the antipassive from the corresponding transitive construction:

182

Rene´ Lacroix

both are marked by the middle. I do not have enough information to state whether the base object can appear in the antipassive construction or not. The transitive construction in (38) and the antipassive construction in (39) can be schematized as follows: Subject-dative coreference Syntax

Aerg

Oabs

Semantics

agent þ beneficiary

theme

Antipassive construction Syntax

Sabs

Semantics

agent þ beneficiary

Middle voice markers are used in antipassive constructions in other languages as well, for instance in Spanish and, more marginally, in French (Creissels 2006: 34). 5.1.5. Aspectual distinctions The Class 1 intransitive verbs -k’iy- ‘crow’, -‘ur- ‘scream’ and -m'or‘scream’, which take an ergative subject,10 appear with the operator i- when used perfectively (ex. 40), and without the operator when used imperfectively (ex. 41). Ex. (42) shows that a form with the operator i- cannot be used when a durative adverbial occurs in the sentence. Note that in other languages as well, the presence of a middle marker has aspectual implications (Creissels 2006: 31). (40) Divi-k ar zor ko-d-i-ur-u. giant-erg one strongly pv-pv-mid-scream-aor.i3s ‘The giant uttered a loud shout.’ (Dum37.81.7)

10. The only exception I have come across is C’ic’ila ur-am-t’u ‘the snake was screaming’ (Zˇ'.125.29). In the very next sentence in the same text, however, we find this verb used with an ergative subject (c’ic’ilak-na uramt’u yeri ‘the place where the snake was screaming’).

Laz middle voice

183

(41) ‰Ł inaze n-o-yon-am-t’a-n-sˇi corpse pv-tr-take-ths-subj-i3p-when oxor‰Ł al-epe-k dido ur-am-an. woman-pl-erg much scream-ths-i3p ‘When people take the corpse [to the cemetery], the women scream a lot.’ (Zˇ'.10.26) (42) Sum saat’i-s k’iy-asen. *d-i-k’iy-asen three hour-dat crow-fut.i3s pv-mid-crow-fut.i3s ‘He will crow for three hours.’ 5.1.6. Lexicalized items Two types of middle verb form may be considered to be lexicalized. First, we find middle verbs which do not have any corresponding non-derived form. Such ‘‘frozen’’ middle verbs are similar to ‘‘deponents’’ or ‘‘media tantum’’ in the grammar of classical languages. They include transitive and intransitive verbs. The list below is not exhaustive. b-i-cˇin-om b-i-pxor b-i-Ø-mer b-i-p’aramit-am b-i-cˇalisˇ-am b-i-xoron-am b-i-bgar me-b-i-xi

i1-mid-know-ths i1-mid-eat i1-mid-take-ths i1-mid-talk-ths i1-mid-work-ths i1-mid-dance-ths i1-mid-cry pv-i1-mid-steal

‘I ‘I ‘I ‘I ‘I ‘I ‘I ‘I

know sb, I recognize sb’ eat sth’ take sth away’ talk, I say sth’ work’ dance’ cry’ steal sth’

Intransitive deponents generally take an ergative subject (ex. 43). Here again, however, we find rare examples involving absolutive subjects (ex. 44). (43) Bere-pe-musˇi-k i-bgar-nan. child-pl-poss3s-erg mid-cry-i3p ‘Her children cry.’ (Zˇ'.81.7) (44) Padime i-bgar-s. Padime mid-cry-i3s ‘Padime cries.’ (Zˇ'.63.35) My informant uses an ergative subject in example (44).

184

Rene´ Lacroix

Secondly, we find middle verbs whose meaning cannot be straightforwardly derived from the related non-middle form: -k’itx- ‘ask’ vs. ik’itx- ‘study; read’; -‰Ł oxon- ‘be called’ (e.g. Ma C’ip’ut’ina m‰Ł oxons ‘my name is C’ip’ut’ina’ [Dum67.III.10]) vs. i-‰Ł ox- ‘shout’. The presence of the middle marker i- on ‘eat’, ‘take away’, ‘steal’ and ‘read’ is not surprising, as these verbs necessarily have an autobenefactive reading (Creissels 2006: 31). Neither is it surprising on ‘know, recognize’ (‘cognition middle’, Kemmer 1993: 127); ‘shout’, ‘cry’ (‘emotion middle’, Kemmer 1993: 130); and ‘dance’, which bears some resemblance to ‘‘nontranslational motion’’ verbs (Kemmer 1993: 56). 5.2. Class 2 middle verb forms The subject of the middle verbs examined so far corresponds to the base subject, if any, and these verbs are mostly agentive (an example of a nonagentive Class 1 verb is i-bgar- ‘cry’). We are now going to consider Class 2 middle verbs, which include primarily non-agentive verbs (anticausatives, among others). Morphologically, Class 2 middle verbs are more homogeneous than Class 1 middle verbs, as they all take the thematic su‰x -er. 5.2.1. Passive The passive construction, illustrated in (45b) below, is rare in my corpus compared to other uses of i-. The subject of the passive verb form corresponds to the base object. (45) a.

b.

Bicˇ’i-k sum k’ocˇ do-yl-u. boy-erg three man pv-kill-aor.i3s ‘The boy killed three men.’ (Zˇ'.144.2) Tabi baba-musˇi d-i-yl-u. of.course father-poss3s pv-mid-kill-aor.i3s ‘[The boy fired when his father stood up.] Of course, his father got killed.’ (Dum67.XXXIV.8)

The corresponding present form is d-i-yl-e-n pv-med-kill-ths-i3s. The base subject (i.e. the agent) cannot be expressed in the passive construction. Its presence, however, is implicit. This point distinguishes the passive from the anticausative, which does not make reference to any agent (Creissels 2006: 31, where the equivalent term ‘‘de´causatif ’’ is used).

Laz middle voice

185

Formally, the passive in Laz is not distinct from other Class 2 middle verb forms. The functions of i- could thus have been labelled ‘‘mediopassive’’, but since the passive is quite restricted in Laz, I have prefered to use simply ‘‘middle’’. The history of Romance and Slavic languages attests a scenario whereby a middle marker develops passive uses, as in se invitaron muchos amigos ‘many friends were invited’, which involves the middle marker se (Creissels 2006: 32). Although the history of the operator i- in Laz is not known, it may be suggested that this language underwent a similar development, whereby passive uses of i- developed secondarily out of middle uses. 5.2.2. Impersonal middle By impersonal middle, I refer to a middle construction with generic semantics derived from an intransitive verb: (46) a.

b.

K’ocˇi mezare-sˇa mend-ul-u-n. man tomb-all pv-go-ths-i3s ‘The man goes to the tomb.’ Hac’ineri mcˇ ’ima do ixi-s contemporary rain and wind-dat mezare-sˇa mend-i-l-in-e-n-i? tomb-all pv-mid-go-caus-ths-i3s-int ‘Do people go to the tomb when it is raining and windy, as it is now?’ (K’art’93.88.17)

The causative su‰x -in appears in some impersonal middle verbs. In these verbs, however, it does not retain its causative function (see Lacroix 2009: 481–482). 5.2.3. Anticausative The anticausative, illustrated in (47b), is one of the most frequent uses of the operator i-. It corresponds to a situation which does not involve an agent (Creissels 2006: 31) and di¤ers in this respect from the passive, which semantically does involve an agent (albeit one which is not expressed). (47) a.

Bozo-k nek’na ge-nk’ol-um-s. girl-erg door pv-close-ths-i3s ‘The girl closes the door.’

186

Rene´ Lacroix

b.

Nek’na ge-i-nk’ol-e-n. door pv-mid-close-ths-i3s ‘The door closes.’

The cause may be expressed as an oblique in the instrumental case: (48) Hea mzˇoa-te i-cˇ’v-e-n. demp sun-instr mid-burn-ths-i3s ‘It burns because of the sun.’ (49) Mera'i-te d-i-zabun-u-doren. anxiety-instr pv-mid-get.sick-aor.i3s-evd ‘He became sick with anxiety.’ (Dum67.I.133) The anticausative may also have a deontic reading, whereby it denotes a general event reflecting the social norm: (50) On‰Ł'ore ren-ya, var i-tkv-e-n-ya. shame be.i3s-quot neg mid-say-ths-i3s-quot ‘It is a shame, it cannot be said.’ (Zˇ'.66.22) An anticausative form used with negation implies a nuance of deontic (51) or physical (52) impossibility: (51) He p’et’emezi cˇkva va i-cˇk’om-e-t’u. demd treacle more neg mid-eat-ths-impft.i3s ‘[A mouse drowned in the treacle.] This treacle could not be eaten any more.’ (Zˇ'.56.4) (52) Badi var esˇ-i-'-e-t’u. old.man neg pv-mid-take.out-ths-impft.i3s ‘The old man could not be taken out [of the well he had fallen into].’ (Dum67.LIII.25) Under the term facilitative, Kemmer (1993) includes among other things ‘‘expressions of intrinsic ability of an object to undergo a particular process’’ (p. 147). Example (53) could be subsumed under the facilitative reading: (53) K’ule-sˇen mteli Mp’oli i-‰ir-e-t’u-doren. tower-abl all Istanbul mid-see-ths-impft.i3s-evd ‘From the tower you could see all of Istanbul.’ (Dum67.XLVII.4)

Laz middle voice

187

Transitive Class 1 verbs may be derived from adjectives; the meaning of such verbs can be paraphrased by ‘make something A’ (where A ¼ adjective; ex. 54a); these verbs correspond to Class 2 anticausatives with inchoative semantics (‘become A’; ex. 54b). (54) a.

b.

Mzˇoa-k cˇamasˇur-epe o-kcˇan-am-s. sun-erg linen-pl tr-whiten-ths-i3s ‘The sun fades the linen.’ K’at’a yei d-i-kcˇan-e-n. every place pv-mid-whiten-ths-i3s ‘[When it snows,] all becomes white.’

Deadjectival Class 1 verbs always take the operator o-. Other examples of anticausative verbs are: Mumiten iguben ‘It cooks over a candle’ (Dum67.VI.29); Orsi n‰Ł umu stey dipuncˇxoludoren ‘The anvil crumbled away like salt’ (Dum67.XLIII.21); Namazi dicˇoden ‘The prayer finishes’. Subjects of passive and anticausative verbs are non-agentive. They are generally in the absolutive. One exception must be mentioned: the verb i-xel- ‘to be glad’ takes an ergative subject (oxor‰Ł a-k ixeludoren ‘the woman was glad’ [Dum67.VI.51]; the present tense is i-xel-e-n and the corresponding transitive form is o-xel-am-s ‘he makes him happy’). I turn now to Class 2 middle verbs with agentive subjects. 5.2.4. Subject-Object coreference We have seen above examples of Class 1 autocausative verbs. Class 2 also includes some autocausatives: (55) a.

b.

Badi-k nk’ola g-o-kt-am-s. old.man-erg key pv-tr-turn-ths-i3s ‘The old man turns the key.’ Badi go-y-kt-e-n. old.man pv-mid-turn-ths-i3s ‘The old man turns back.’

Other examples are: K’ocˇi itk’ocˇen ‘The man leaps’; Hea mt’ut’as dixven ‘He buries himself in ash’; K’ulesˇ kenayis kamik’idudoren ‘He hung from the edge of the tower’ (Dum67.XLVII.14).

188

Rene´ Lacroix

Among the Class 2 agentive verbs, we find a verb with reciprocal semantics: ‘hit each other, fight’. (56) a.

b.

Xasani-k Ali il-om-s. Hasan-erg Ali hit-ths-i3s ‘Hasan hits Ali.’ Hemtepe i-il-e-nan. demp.pl mid-hit-ths-i3p ‘They are fighting.’

The same verb may be used transitively with the reciprocal pronoun artikati in object position. In this case, the verb form belongs to Class 1: (57) Sum bere-k biga-pe-te artikati il-om-t’es. three child-erg stick-pl-instr recp hit-ths-impft.i3p ‘Three guys were hitting each other with sticks.’ (Dum67.I.415) Lastly, the verbs ‘prepare oneself ’ and ‘marry (intr.)’ admit of two forms: one in Class 1, bixaziam ‘I prepare myself ’, bicˇil ‘I marry (a girl)’ (Dum67.VIII note 14), and one in Class 2, bixazie ‘I prepare myself ’, bicˇile ‘I marry (a girl)’. Although in general Class 2 agentive verbs take an absolutive subject, we do find some rare occurrences of an ergative subject (ex. 58). I will return to this point below (5.3). (58) Orcˇ’-ayi-k go-y-kt-u. Orcˇ’i-sfx-erg pv-mid-turn-aor.i3s ‘The man from Orcˇ’i turned back.’ (Dum67.LI.44) 5.2.5. ‘look round’ Another example of a Class 2 agentive verb is i-ck’- ‘look round’. The corresponding non-middle Class 1 verb ‘look at’ takes an ergative subject and a dative complement: (59) a.

Bere-k bozo-s mend-o-ck’e-s. child-erg girl-dat pv-tr-look-i3s ‘The child looks at the girl.’

The middle form is used when the act of looking is not directed towards a specific object. The subject is still in the ergative. No complement appears in the construction.

Laz middle voice

b.

189

Xo‰Ł a-k hekolhakole i-ck’-e-t’u. hodja-erg here.and.there mid-look-ths-impft.i3s ‘The hodja was looking here and there / was looking round.’ (Dum67.XXX.7)

The present tense form is i-ck’-e-n mid-look-ths-i3s. This is one of the few examples of Class 2 verbs which always appear with an ergative subject. Unlike with the verbs illustrated under ‘‘subject-object coreference’’ and ‘‘subject-dative coreference’’, here the subject does not cumulate two semantic roles. The e¤ect of the middle marker on this verb is reminiscent of the antipassive. Syntactically, the second core argument of the verb in (59a) is deleted in (59b). However, the antipassive demotes the object, which is not the case here, since the demoted argument is marked by the dative. Furthermore, the subject remains in the ergative. Semantically, in the situation described in (59b), the thing being perceived is unimportant – a feature usually associated with the antipassive (Givo´n 2001: 168). As we have seen, some languages attest a scenario whereby middle markers develop antipassive uses (Creissels 2006: 40). 5.2.6. Lexicalized item One final example of a Class 2 verb with an agentive subject is the lexicalized verb i-mt’- ‘flee’ (60a). This verb lacks a corresponding non-derived form; the corresponding transitive verb is morphologically a causative (60b): (60) a.

b.

K’ocˇi i-mt’-e-n. man mid-flee-ths-i3s ‘The man runs away.’ (Zˇ'.91.7) Xasani-k mcˇ ’a‰Ł i o-mt’-in-am-s. Hasan-erg fly tr-flee-caus-ths-i3s ‘Hasan chases away the flies.’

The verb ‘flee’ is attested with middle markers in other languages as well (see Kemmer 1993 on ‘‘translational motion’’ verbs [p. 56]). 5.3. Summary We have seen that middle verb forms may belong to Class 1 or Class 2. The subject of a Class 1 middle verb corresponds to the base subject. Class

190

Rene´ Lacroix

1 middle verbs are in general agentive. Subject-dative coreference (5.1.2) and object possession (5.1.3) are transitive constructions, whereas the subject-object coreference construction (5.1.1) and the antipassive (5.1.4) are intransitive. A few intransitive verbs include the operator i- when used in a perfective context (5.1.5). The subject of passive (5.2.1) and anticausative verbs (5.2.3), which belong to Class 2, corresponds to the base object, the base subject being backgrounded. Such verbs are non-agentive. Class 2 also contains some agentive verbs (5.2.2, 5.2.4 and 5.2.5). All Class 2 middle verbs are intransitive. Both Class 1 and Class 2 middle verbs include lexicalized items. As we have seen, body action verbs are found in Class 1, which is mostly restricted to agentive verbs, and Class 2, which comprises for the most part non-agentive verbs; secondly, there is some variation between the use of the ergative and the absolutive with body care (5.1.3), autocausative (5.1.1, 5.2.4) and media tantum verbs (5.1.6); finally, the verbs ‘prepare oneself ’ and ‘marry (intr.)’ admit of both Class 1 and Class 2 forms. All these facts correlate with Kemmer’s observation that, from a semantic point of view, body action verbs are intermediate between prototypical two-participant (agentive) and prototypical one-participant (nonagentive) events (Kemmer 1993: 73). The functions of the middle vowel i- in the two verb classes are summarized below. Table 5. Summary of the functions of the middle marker iClass 1 middle verbs

Class 2 middle verbs

subject-object coreference subject-dative coreference object possession antipassive sensitivity to aspect lexicalized items

passive impersonal middle anticausative subject-object coreference ‘look round’ lexicalized item

These uses correspond to events typically expressed by middle markers in languages possessing a middle system. The wide range of uses covered by the middle vowel i- and the number of lexicalized items (especially in Class 1) indicate that this operator is of ancient origin. Furthermore, we have seen that Laz middle verbs may be used transitively. In this respect, Kemmer (1993: 34) notes: ‘‘Transitive

Laz middle voice

191

middle marker verbs arise diachronically when the middle marker has been grammaticalized to the point where it no longer denotes a referential entity, but only marks middle semantics’’.11 Indeed, comparative data show that the operator i- is ancient. It is found in the other languages of the family – Mingrelian, Georgian and even Svan, the most remote member, which is said to have branched o¤ from Proto-Kartvelian no later than the last centuries of the third millennium B.C. (Klimov 1998: IX; this approximation is based on Swadesh’s method of glottochronology). The operator i- is reconstructed for ProtoKartvelian (Fa¨hnrich 2007: 209). Its position immediately before the verbal root is an indication that it became attached to the verbal root even before the cross-referencing prefixes, which are also reconstructed for Proto-Kartvelian.

6. Other accounts of the valency operator iAs mentioned in the introduction, Georgian has an operator i- with roughly the same uses as the Laz element. In descriptions of Georgian, the operator i- used with Class 1 verbs in examples such as ‘build a house for oneself ’ or ‘comb one’s hair’ is referred to as ‘‘subjective version’’ (Sˇani‰e 1953, Tschenke´li 1958, Boeder 1969, 2005, Hewitt 2005). The ‘‘subjective version’’ indicates that the subject acts ‘‘for his own benefit’’, that the action ‘‘goes back to the subject’’ (Tschenke´li 1958: 245). Georgian Class 2 verbs (which correspond roughly to Laz Class 2 verbs) are sometimes referred to as ‘‘passives’’ (Tschenke´li 1958, Fa¨hnrich 1993). They are of di¤erent formal types. ‘‘Passive’’ verbs marked by the operator i- include true passives, but also anticausative (ic’veba ‘it burns [intr.]’, irk’veva ‘it becomes clear’), facilitative (itargmneba ‘it can be translated’) and even autocausative verbs (imaleba ‘he hides [intr.]’, in‰reva ‘he moves [intr.]’). The term passive is thus misleading: strictly speaking, passive is only one of the values of i- in this verb class. This presentation of Georgian has served as a model for the analysis of the other languages of the family: Laz (Holisky 1991: 422, 438), Mingrelian (Harris 1991c: 354, 360) and Svan (Tuite 1997: 26, 35). However, it does 11. This explanation, however, would hold only if the valency operator i- could be shown to derive from a marker denoting a referential entity, such as a reflexive marker. Although this is the most frequent origin for middle markers, there are other possible sources (Kemmer 1993: 197–200).

192

Rene´ Lacroix

not capture the functional unity of the operator i- in these languages. In fact, as we have seen, the ‘‘subjective version’’ and the ‘‘passive’’ marker i- represent one and the same morpheme, which indicates middle voice, in the sense of Kemmer (1993). Passive uses indicated by i- may occasionally be found in Laz, but such examples are not frequent in the corpus. Some authors (Deeters 1930, Sˇani‰e 1953, Schmidt 1965, Tuite 2007, Gurevich 2006) have pointed out the similarities between the operator iand the middle, especially as seen in Indo-European. However, this has not led them to take the view that, beyond mere resemblance, i- can indeed be considered as a marker of middle voice. Gurevich (2006: 176) argues that ‘‘there are . . . significant di¤erences between the Greek and Georgian situations. The Georgian parallels to middle voice are much more heterogeneous than the Greek ones appear to be, and more lexicalized. . . . Moreover, the Georgian middle-voice formations cover only a portion of the semantic classes described for Greek middle voice. . .’’. These di¤erences, however, naturally follow from the fact that the diachronic development of middle markers is not necessarily identical across languages (although, as shown by Kemmer [1993], some generalizations may be drawn). Lexicalization is not a systematic process, but rather the sum of independent instances of evolution. As a result, the exact semantic domains covered by middle markers in different languages need not coincide. In addition, divergences in the number of lexicalized items may simply reflect di¤erent stages in the chronology of the grammaticalization process. Boeder (1969), Forest (1999) and Macˇ’avariani (1987), working primarily on Georgian data, treat the middle operator i- and the operator i- used in applicative derivation (see section 3) as the same marker, and try to give a unified account of them. Thus, in Boeder’s view, the i- markers in (61a) and (61b) are functionally identical. (61) a.

b.

subject-dative coreference i-k’od-um-s mid-build-ths-i3s ‘he builds sth for himself ’ applicative g-i-k’od-um-s ii2-appl-build-ths-i3s ‘he builds sth for you’

Laz middle voice

193

Indeed, in both examples, the marker i- indicates that the action is directed towards a beneficiary: the second person addressee in (61b) and the agent himself in (61a). Thus, the comparison of such examples might suggest an analysis of both instances of i- as the same marker. However, such an analysis becomes much more di‰cult when one takes into account the other uses of i-, in particular the passive and the anticausative: a sentence such as ‘The door opens (goinck’en)’ has no beneficiary. Furthermore, the middle i- and the applicative i- have distinct syntactic characteristics: the applicative adds an argument to the base construction, which is never the case with the middle. Thus, although there may be a historical relationship between the middle and the applicative i-, it seems preferable to keep these two markers distinct in a synchronic description.

7. Conclusion In this paper, I have shown that the multiple uses of the valency operator i- in Laz may be accounted for in a unified manner if we take the view that this morpheme indicates middle voice. Thus, the operator i- does not constitute a grammatical category peculiar to Kartvelian. Its originality lies rather in its formal manifestations. The divergences that have been noted between i- in the Kartvelian languages and middle markers in Indo-European languages follow from more general principles of grammaticalization. The operator i- in Laz and, I believe, in the other Kartvelian languages covers a functional domain which is not essentially di¤erent from that of middle markers in accusative languages. But the fact that in Laz subjects may be in both the ergative and in the absolutive raises questions on the relation between ergative alignment type and middle voice which have no equivalents in languages with case marking displaying accusative alignment. Givo´n (2001: chap. 13) distinguishes between semantic and pragmatic voice mechanisms. Pragmatic voices ‘‘render the same semantically-transitive event from di¤erent pragmatic perspectives. These perspectives turn out to involve, primarily although not exclusively, the relative topicality of the agent and patient’’ (p. 93). ‘‘Primarily pragmatic voice constructions are those whose functional definition depends on some facets of the wider, extra-clausal, discourse context’’ (p. 92). Such mechanisms do not a¤ect the semantics of the predicate. Primarily pragmatic voices include the

194

Rene´ Lacroix

passive, the antipassive and the inverse. These voices interfere with other mechanisms which may be involved in the expression of pragmatic perspective (for example, constituent order). In particular it is commonly assumed that they interfere with alignment, and that (for example) the antipassive may have functional motivations in morphosyntactically ergative languages which find no equivalent in accusative languages. On the other hand, primarily semantic voice mechanisms operate on argument structure at a semantic level, modifying the semantic roles a predicate assigns to its arguments. Unlike pragmatic voices, their ‘‘functional definition does not depend on entities outside the boundaries of the event-clause’’ (p. 92). Primarily semantic voices include reflexive, reciprocal and middle mechanisms. Middle voice, which belongs to the latter group, has its own motivation; that is, it does not interfere with pragmatic mechanisms. Hence, there need not be any correlation between middle voice and the alignment type of a language. This is corroborated by the existence of a middle voice in partially ergative languages such as Laz.

Abbreviations abl all aor appl caus comp dat dem demd demp erg evd fut gen impft instr int intr.

ablative allative aorist applicative causative complementizer dative demonstrative demonstrative determiner demonstrative pronoun ergative evidential future genitive imperfective instrumental interrogative intransitive

mid neg pl poss pperf pv quot pl pv recp sfx subj ths tr, tr. i ii 1, 2, 3 S, P

middle negation plural possessive pluperfect preverb quotative plural preverb reciprocal su‰x subjunctive thematic su‰x transitive Set I Set II first, second, third person singular, plural

Laz middle voice

195

For Dume´zil 1937, Zˇ'ent’i 1938 and K’art’ozia 1993, the references accompanying the examples give the page and the line of the Laz text, so that Dum37.10.3 means Dume´zil 1937, page 10, line 3. For Dume´zil 1967 and 1972, the references give the text number and the line number provided by Dume´zil, so that Dum67.XXIII.8 means Dume´zil 1967, text XXIII, line 8.

References Andrews, Peter (ed.) 1989 Ethnic groups in the Republic of Turkey. Compiled and edited with the assistance of Ru¨diger Benninghaus. Wiesbaden: Reichert. Aronson, Howard 1990 Georgian: A Reading Grammar. Slavica Publishers. Boeder, Winfried 2005 The South Caucasian Languages. Lingua 115(1): 5–89. Boeder, Winfried ¨ ber die Versionen des georgischen Verbs. Folia Linguistica 2: 1969 U 82–152. Boeder, Winfried 2002 Syntax and morphology of polysynthesis in the Georgian verb. In Evans, Nicholas and Sasse, Hans-Ju¨rgen (eds.) Problems of Polysynthesis: 87–111. Berlin: Akademie Verlag. Boeder, Winfried 2005 The South Caucasian languages. Lingua 115: 5–89. Braithwaite, Kim 1973 Case shift and verb concord in Georgian. PhD thesis. University of Texas at Austin. Creissels, Denis 2006 Syntaxe ge´ne´rale, une introduction typologique. Tome 2, la phrase. Paris: Hermes Science. ˇ ikobava, Arnold C ˇ ’anuris gramat’ik’uli analizi (t’ekst’ebiturt) [A grammatical 1936 C analysis of Laz (with texts)]. Tbilisi: Sakartvelos SSR Mecnierebata Ak’ademiis Gamomcemloba. Deeters, Gerhard 1930 Das kharthwelische Verbum: vergleichende Darstellung des Verbalbaus der su¨dkaukasischen Sprachen. Leipzig: Markert und Petters. Dixon, R.M.W and Aikhenvald, Alexandra Y. 2000 Introduction. In Dixon, R.M.W and Aikhenvald, Alexandra Y. (eds.) Changing Valency. Case studies in transitivity. Cambridge: Cambridge University Press.

196

Rene´ Lacroix

Dume´zil, Georges 1937 Contes lazes. Paris: Travaux et me´moires de l’Institut d’Ethnologie, XXVII. Dume´zil, Georges 1967 Documents anatoliens sur les langues et les traditions du Caucase, IV. Re´cits lazes (dialecte d’Arhavi). Paris: Presses Universitaires de France. Dume´zil, Georges 1972 Textes en laze d’Ardes¸en. Bedi Kartlisa, vol. XXIX–XXX. Fa¨hnrich, Heinz 1993 Kurze Grammatik der georgischen Sprache. Leipzig: Langenscheidt Verlag Enzyklopa¨die. Fa¨hnrich, Heinz 2007 Kartwelisches Etymologisches Wo¨rterbuch. Brill. Forest, Robert 1999 Empathie et linguistique. Presses Universitaires de France. Givo´n, Talmy 2001 Syntax. Volume II. John Benjamins. Gurevich, Olga 2006 Constructional Morphology: the Georgian Version. PhD thesis. University of California, Berkeley. Harris, Alice C. (ed.) 1991a The Indigenous Languages of the Caucasus. Volume I: The Kartvelian Languages. Delmar, New York: Caravan. Harris, Alice C. 1991b Overview on the History of the Kartvelian Languages. In Harris, Alice C. (ed.) The Indigenous Languages of the Caucasus. Harris, Alice C. 1991c Mingrelian. In Harris, Alice C. (ed.) The Indigenous Languages of the Caucasus. Hewitt, George 2005 Georgian: A Learner’s Grammar. London: Routledge. Holisky, Dee Ann. 1991 Laz. In Harris, Alice C. (ed.) The indigenous languages of the Caucasus. Kemmer, Suzanne 1993 The Middle Voice. John Benjamins. Klimov, Georgij A. 1998 Etymological Dictionary of the Kartvelian Languages. Mouton de Gruyter. Kutscher, Silvia 2001 Nomen und nominales Syntagma im Lasischen. Eine deskriptive Analyse des Dialekts von Ardes¸en. Mu¨nchen: Lincom Europa (Lincom Studies in Caucasian Linguistics 17).

Laz middle voice

197

K’art’ozia, Guram 1972 Lazuri t’ekst’ebi [Laz texts]. Tbilisi: Mecnierebata K’art’ozia, Guram 1993 Lazuri t’ekst’ebi [Laz texts]. Tbilisi: Mecniereba. Lacroix, Rene´ 2010 Benefactives in Laz. Kittila¨, Seppo and Zu´n˜iga, Fernando (eds), to appear, Benefactives and malefactives. Typological perspectives and Case studies, John Benjamins. Lacroix, Rene´ 2009 Description du dialecte laze d’Arhavi (caucasique du sud, Turquie). Grammaire et textes. PhD thesis. University Lyon 2. Available at http://theses.univ-lyon2.fr/documents/lyon2/2009/ lacroix_r. Macˇ’avariani, Maia 1987 Kcevis gramat’ik’uli k’at’egoriis semant’ik’a [Semantics of the grammatical category of version]. Tbilisi: Mecniereba. Marr, Nicolas ˇ an (Laz)]. 1910 Grammatika cˇanskago (lazskago) jazyka [Grammar of C Materialy po jafeticˇeskomu jazykoznaniju 2. St. Petersburg: Akademija. Peterson, David A. 2007 Applicative Constructions. Oxford: Oxford University Press. Q’ipsˇi‰e, Ioseb ˇ ’anuri t’ekst’ebi [Laz texts]. Tbilisi: SSRK’ Mecnierebata 1939 C Ak’ademiis Pilialis Gamomcemloba. Schmidt, Karl Horst 1965 Indogermanisches Medium und Sataviso im Georgischen. Bedi Kartlisa 19–20: 129–135. Sˇani‰e, Ak’ak’i 1953 Kartuli gramat’ik’is sapu‰vlebi, I: morpologia [The fundamentals of Georgian grammar I: morphology]. Tbilisi: TSUG. Tschenke´li, Kita 1958 Einfu¨hrung in die georgische Sprache. Band I: theoretischer Teil. Zu¨rich, Amirani Verlag. Tuite, Kevin 1997 Svan. Mu¨nchen: Lincom Europa. Tuite, Kevin 2007 Liminal morphosyntax: Georgian deponents and their kin. Chicago Linguistics Society 39(1): 774–788. Zˇ'ent’i, Sergi ˇ ’anuri t’ekst’ebi (arkabuli k’ilok’avi) [Laz texts (dialect of Arhavi)]. 1938 C Tbilisi: SSRK’ Mecnierebata Ak’ademiis Pilialis Gamomcemloba.

Argument demotion in Japhug Rgyalrong Guillaume Jacques 1. Introduction1 Japhug (Chabao in Chinese), a Sino-Tibetan language spoken in China (Mbarkhams county, Rngaba autonomous region, Sichuan Province), is unusual among the languages of this family in displaying complex verbal morphology. Alongside Tshobdun (Caodeng in Chinese), Zbu (also known as Showu, Ribu or Rdzongmbur) and Situ (Eastern Rgyalrong), it belongs to the so-called Rgyalrong languages.2 Like some other Sino-Tibetan languages of Sichuan, Japhug has an ergative case marking system and a verb agreement system which indexes two arguments (for transitive verbs) following a hierarchical pattern. No ergative syntactic pivots are found in nominalization, complementation or equi-NP deletion constructions. This paper is divided into seven parts. First, we will discuss the morphological and syntactic marking of transitivity in Japhug, the basis for any study of transitivity-changing devices in this language. Second, we will 1. I collected the data presented in this article in a series of field trips to China undertaken from 2002 onwards. My main informant for this language was Chenzhen. I wish to thank Peter Austin, Gilles Authier, William Croft, Henriette Daudey, Katharina Haude, Steven Kaye and an anonymous reviewer for their insightful comments, corrections and discussions. This article was completed during my stay as a visiting scholar at the Research Centre for Linguistic Typology, LaTrobe University: I am grateful to Randy LaPolla for making this visit possible. The glosses generally follow the Leipzig Glossing Rules, except for the following: adv adverbializer, anticaus anticausative, apass antipassive, cis cislocative, conj conjuction, const constative, downstr downstream, evd evidential, genr generic, hum human, intsv intensive, inv inverse, neu neuter (indefinite possession), nonhum non-human, n.pst nonpast, stat stative, trans translocative, upstr upstream. 2. The Rgyalrong languages are themselves a sub-branch of the Qiangic branch of Sino-Tibetan, which comprises the extinct Tangut language, as well as Qiang, Pumi, Muya, Queyu, Zhaba, Lavrung and Rtau. Guiqiong, Shixing, Namuyi and Ersu, generally thought to belong to this branch, should probably be classified di¤erently (Jacques and Michaud 2011).

200

Guillaume Jacques

describe the generic argument marking system, which, like nominal case marking, shows ergative alignment. Third, passive and anticausative forms will be considered. These forms have a fairly restricted range of uses, since ‘‘generic’’ forms are employed in most cases where the passive would be found in European languages. Fourth, we will analyse the two antipassive prefixes found in Japhug, which are not productive, but have the interesting property of distinguishing between human and non-human suppressed patients. Fifth, we will briefly describe the de-experiencer prefix, which can derive an intransitive verb from a transitive verb of perception. Sixth, we will discuss the labile verbs, a small class of verbs which can be either transitive or intransitive without any derivational marking; these verbs are uniformly agent-preserving and never patient-preserving. Finally, we will present incorporation, the last morphological means of suppressing the patient in Japhug.

2. Transitivity marking in Japhug Transitivity is an essential feature of the Japhug verbal system. There is never any ambiguity about whether a given verb is transitive or not. A complete account of person agreement and TAM markers in Japhug is beyond the scope of the present article, but in this section all a‰xes relevant to the marking of transitivity will be discussed. 2.1. Morphological transitivity In Japhug, transitive verbs regularly agree with two arguments, so that many transitive verbal forms (such as 1>2, i.e. first person agent and second person patient) have no equivalent in intransitive verbs. However, third person patient forms (1>3, 2>3 and 3>3) are in some cases almost indistinguishable from intransitive verbal forms. Two of the markers which are found exclusively with transitive verbs are restricted to a phonologically conditioned subset of these verbs: the regular stem 3 formation,3 which only occurs with verbs whose basic stem ends in -a, -u, -o and -, and the -t 1sg>3 / 2sg >3 past tense su‰x, which only applies to open syllable roots. 3. In the Rgyalrong languages, verbs can have up to three or four distinct stems (Sun 2000). Stem 3 is used in Non-past, Imperfective, Irrealis and Imperative forms with a singular agent and a third person patient (Jacques 2008: 246–7).

Argument demotion in Japhug Rgyalrong

201

Fortunately, two morphological tests can be applied to any verb to determine whether it is transitive or intransitive, even if this verb only allows third person patients and has a closed syllable stem. Verbs which possess a stem 3 and make use of the -t past tense su‰x always confirm their transitive nature in these two tests as well. Firstly, transitive verbs take the prefix a- in direct aorist 3>3 forms. Compare the following examples: (1) p-a-mto aor-3>3-see ‘He saw him.’ (2) p-fflqhlt aor-fall ‘He fell down (into a trap).’ In the first example, the verb is transitive, and the a- prefix appears after the directional prefix p- which marks the aorist, whereas in example (2) the verb is intransitive and no such prefix appears. Secondly, the nominalized forms of transitive and intransitive verbs are distinct: intransitives build their S-nominalization by means of a prefix k-, while for transitive verbs the A-nominalized form makes use of the same k- prefix (see example 3) preceded by a possessive prefix coreferent with the O. For instance: (3) -k-mto 3sg.poss-nmlz:S/A-see ‘The one who sees him’ In example (3), the nominalized form is marked with a third person singular possessive prefix - indicating the O. Compare this form with (4): (4) k-si nmlz:S/A-die ‘The dead one’ In this example, no such possessive prefix appears. Thus, on the basis of the direct aorist 3>3 form and the nominalization in k-, it is always possible to determine whether a given verb is transitive or not. The strict morphological transitivity marking found in Japhug and other Rgyalrong languages is relatively rare among Sino-Tibetan languages, where transitivity is often di‰cult to define (as for example in

202

Guillaume Jacques

Chinese or Tibetan). However, Rgyalrong languages are not the only ones in this family to show transitivity marking: Dulong/Rawang languages (LaPolla 2001: 284) and Kiranti languages also have a fully-fledged set of transitivity-marking a‰xes on the verb. 2.2. Case marking The formal distinction between transitive and intransitive verbs is not limited to their morphology. Japhug has a simple case marking system, which presents ergative alignment: A arguments are marked with the enclitic k,4 while S and O are left unmarked, as can be seen in sentences (5) and (6): (5) ‚drffii k ´amu p-a-mto Rdorje erg Lhamo aor-3>3-see ‘Rdorje saw Lhamo.’ (6) ´amu ci -nre Lhamo a_little evd-laugh ‘Lhamo laughed a little.’ When both participants are overt, it is therefore possible to determine whether the verb is transitive or intransitive on the basis of case marking. Ergative marking is obligatory with third person participants in Japhug for all TAM categories, but only rarely appears with SAP pronouns such as affio ‘I’ and nffio ‘you’. With intransitive verbs, the ergative case can sometimes be used, but it conveys a specific comparative meaning: (7) nffio k -t-cha you erg const-2-be_capable ‘You are more capable.’ Word order is verb-final, the agent usually preceding the patient. Sentences with an overt agent and patient are fairly rare in actual texts. As in many polysynthetic languages with indexation of two arguments, it is quite common to omit both agent and patient NPs (Mithun 1999: 190– 193): the only necessary element of a sentence is the finite verbal form. Covert arguments are normally definite. A minimal sentence such as:

4. This clitic is probably borrowed from the Tibetan ergative marker gyis.

Argument demotion in Japhug Rgyalrong

203

(8) to-ndza evd-eat ‘He ate it.’ can only appear if both the eater and the thing eaten have been mentioned before or are implicit. Therefore, unlike in European languages such as French or English, omission of the O argument is not available as a potential means of expressing an indefinite patient. Other morphological devices, such as antipassive, generic or incorporation constructions, are required to serve that purpose. 2.3. Ditransitive verbs In ditransitive verbs involving a recipient, such as ‘give’ or ‘tell’, we observe in Japhug both indirective and secundative alignment (Haspelmath 2005). No more than two arguments may be indexed on the verb, and the nature of the second argument indexed (recipient or theme) is lexically determined for each verb. This can be illustrated with two verbs meaning ‘give’, mbi and kho. The first of these, mbi, encodes the recipient as its second argument (Haspelmath’s secundative alignment): (9) ki -ta-mbi dem ipfv-1 >2-give ‘I give this to you.’ (Gesar, 197) Meanwhile, the verb kho allots the theme to the second argument position, and the recipient can only be marked as an external argument taking the dative su‰x -fflki or -phe (indirective alignment):5 (10) a-me -kha´m-a Ðu 1sg.poss-daughter ipfv-give[3]-1sg n.pst:be ‘I will give (you) my daughter.’ (The Frog, 78) The alignment of ditransitive verbs in Japhug is quite strict, and no verb can be both indirective and secundative. 2.4. Semi-transitive verbs Some two-place verbs in Japhug are distinctive in that they neither present transitivity markers nor require ergative marking on the agent. This class 5. The external argument is not overt in example (10); if present, it would take the form n-fflki or n-phe.

204

Guillaume Jacques

of ‘semi-transitive’ verbs includes verbs of motion and some verbs of perception. We provide some examples to illustrate this intriguing phenomenon: (11) n ra ffl-k-ru j' dem pl trans-imp:east-look n.pst:could ‘Just go to see them.’ (Smanmi, 110) In this example, the verb ru ‘look at’ clearly has two distinct arguments, but it fails to show any transitive marking. If it were a normal transitive verb, the verb form seen in (11) – along with all other non-past, imperfective and imperative forms – would make use of the special ‘stem 3’ mentioned above, involving the vowel alternation u ! e. Meanwhile, other perception verbs such as mto ‘see’ are fully transitive: the person seeing receives ergative marking and both arguments are indexed on the verb. Like the verb ‘look at’, motion verbs are intransitive, although they might be considered to have two arguments: the person/animal moving and the place travelled towards (the latter is not usually marked with an oblique case in Japhug). This class includes the verb ffle (Aorist stem ari ) ‘go’. (12) t-pi ni t»u k-wxti j-arı´-ndffii neu-elder_sibling du way nmlz:stat-be_big aor-go[2]-du ‘The two elder brothers followed the big road.’ (The Fox, 19) This phenomenon is illustrated here in example (12), where the persons going (the elder brothers) do not receive ergative case. Other tests for transitivity (nominalization, transitivity a‰xes on the aorist, etc.) would also fail. Interestingly, Dulong/Rawang, the other sub-branch of Sino-Tibetan to share the strict transitivity marking seen in the Rgyalrong languages, also possesses a class of ‘semi-transitive’ verbs, whose members overlap with those in Rgyalrong: in particular, perception and motion verbs are again found to display this behaviour (LaPolla 2008). 3. Generic marking As mentioned in the introduction, Japhug and other Rgyalrong languages have a direct/inverse marking system (DeLancey 1981, Sun and Shi 2002). The inverse marker w'- appears on verb forms where the agent is third

Argument demotion in Japhug Rgyalrong

205

person and the patient is SAP, or when the agent is non-human and the patient human, for instance: (13) tha a-m-p-t -w'-sat otherwise irr-neg-pfv-2-inv-kill ‘Otherwise she would kill you.’ (Nyima vodzer, 36) The use of the inverse in Japhug has been described in Jacques (2010a), so this topic will not be discussed in detail here. It should be noted, however, that unlike passive and antipassive prefixes this a‰x does not cause a change in valency: the verb remains transitive (the inverse marker cannot appear on an intransitive verb) and both arguments can be marked in the verb morphology. A special use of the inverse prefix w'- is nevertheless highly relevant to the topic covered in this paper: namely what I will call its ‘‘generic’’ use, a function cross-linguistically associated with passive forms. No other language with direction marking seems to share this feature, as can be seen from Zu´n˜iga’s (2006) recent survey of direct/inverse systems in languages of the New World. Japhug generic forms often appear in sentences expressing general truths, and generic marking is extremely common in procedural texts: (14) sjno smn tu-w'-¯t tffln, n lu´-w'-lt grass medicine ipfv-inv-buy conj dem ipfv-inv-throw ‘After one buys fertilizer, one spreads it (in the field).’ (Rtsampa, 49) The generic form can be strictly distinguished from the normal inverse. Inverse verbal forms imply that the agent is lower or equal to the patient on the Empathy Hierarchy. Inverse marking is obligatory if an inanimate agent acts upon a human, and conversely, it can never appear when a human acts upon an inanimate. This e¤ect of the Empathy Hierarchy does not apply to generic forms. As example (14) shows, the generic inverse can be used with a (generic) human agent and an inanimate patient. It never takes dual or plural su‰xes. This implies that generic arguments, even humans, are lower on the Empathy Hierarchy than inanimates. The following hierarchy should thus be postulated: (15) SAP > human > animal > inanimate > generic argument Non-generic inverse forms di¤er from generic ones in two ways. Firstly, they can take number su‰xes, and secondly, they cannot appear if the patient is inanimate. Consider the following example:

206

Guillaume Jacques

(16) sp'o -ta‚ p -w'-ta-ndffii ndre, stake 3sg.poss-on aor-inv-put-du conj ndffii-pa smi t-a-”l -n ndre, 2du.poss-under fire aor-3>3-burn-pl conj t-pi ‚na‚na ffio p -w'-sat-ndffii  -u. neu-elder_brother both adv aor-inv-kill-du ipfv-be ‘People put the two of them on the stake, lit the fire under them, and killed both of them, the two brothers.’ (The Fox 180–181) All three verbs in (16) share the same agents (unspecified in the story, but probably the king’s servants or the villagers). The first and the third have inverse marking (and agree in number – dual – with their patient), while the second is direct, with plural agent marking because its patient is inanimate (the fire, smi). The absence of an inverse form on this verb confirms the fact that the inverses in the first and the second are not generic, for otherwise we would expect the second verb to be in the inverse form too. In (16), the inverse can be used because both the agents and the patients are human, and thus equal in terms of the Empathy Hierarchy.6 The inverse prefix is only used when the generic argument is the agent of a transitive verb. For intransitive verbs or patients, a di¤erent a‰x is used: the k- prefix. This prefix is homophonous with the nominalizer described in section 2, but is unrelated to it at least synchronically. The uses of this prefix are exemplified in sentences (17), from a story about the yeti, and (18), from a procedural text: (17) -‚ri n -‚ri ffio 3sg.poss-front conj 3sg.poss-front adv ju-k-ph'o a-p-Ðu tffle, ipfv-genr:S/O-flee irr-ipfv-be conj maka ffio m-pj-k-mto khi at_all adv neg-evd-genr:S/O-see hearsay ‘If one runs in the direction in front of (the yeti), one will not be seen by (him).’ (The Yeti, 17) In this example, both the first verb, ‘flee’ (intransitive), and the second, ‘see’ (transitive), have the generic prefix k-. In the second verb, the 6. In example (16), direct forms would also be grammatical. The inverse appears here for pragmatic reasons, because the patient (the two brothers) is more topical than the agent (the unnamed people who perform their execution). See Jacques (2010a) for a more detailed discussion of this phenomenon.

Argument demotion in Japhug Rgyalrong

207

generic argument is not the A (the one who sees) but rather the O (the one seen), as the A is definite (the yeti). The generic forms are compatible with both imperfective and aorist forms, as shown in the following example. Imperfective forms are the most common in procedural texts, but the aorist forms are used to express a ‘twist’ in the action, which can usually be translated into western languages with the conjunction ‘when’. (18) nn mb'jro‚ n ci  -k-ffle dem furrow dem once ipfv:west-genr:S/O-go ki tu-k-fse tffle dem ipfv-genr:S/O-be this way conj k-k-'e tffle li ki aor:east-genr:S/O-come[2] conj again dem tu-k-fse tffle, n k -w'-ffllu ipfv-genr:S/O -be this way conj dem aor-inv-plough n k li ch-fka” dem erg again ipfv: downstr-cover ‘One goes along the furrow this way, but when one comes back again this way, as one ploughs (the furrow), (the earth of the new furrow) covers (the groove of the older furrow).’ (Rtsampa, 30–31) Generic forms, however, are not compulsory in procedural texts. Verbs in the imperfective, often (but not always) with plural marking, can express the same meanings. See the following passage, which contains no generic markers: (19)

ju-ph'o tffl e kspo‚-Ðg lu-nffle ipfv-flee conj hole-inside ipfv:upstr-go_back tffle k-sat m-kh tffle conj aor-kill neg-n.pst:be_possible conj -km smi pj-”l-n tffle 3sg.poss-door fire ipfv-burn-pl conj m-t-tfflha‚ tffle ch-n´o‚ tffle neg-aor-bear conj ipfv:downstr-go_out conj pj-sat Ðgrl ipfv-kill n.pst:be_usually_the_case ‘When (the badger) goes back into his hole, one cannot kill him, so they light a fire in front of the hole, he cannot bear it and goes out, and then one can kill him.’ (Dictionary entry)

208

Guillaume Jacques

It is clear from these data that the generic markers in Japhug show ergative alignment (S¼O A A): w'- is an A-generic, and k- an S/Ogeneric. This is one of the few contexts in Japhug where ergativity is present, other than in case marking. In other syntactic structures, no ergative alignment is found: nominalization patterns in Rgyalrong languages follow accusative alignment (Sun 2003: 497), and equi-NP deletion shows no syntactic pivot (Jacques 2010a). Japhug generic forms di¤er markedly from those described in Tshobdun (Sun 2005), where the inverse is not used in this function and a strict distinction is observed between human and non-human generic forms. Unlike in Kryz, where the passive form conveys deontic/habitual meaning (Authier, this volume 4.1), in Japhug such contexts instead make use of the generic form, and the passive, as we will show, never bears this kind of meaning.

4. Passive Japhug has two agentless passive forms, which present complex morphophonological alternations. These alternations have been described in Jacques and Chen (2007), and we will concentrate our discussion here on the functions of these forms. Unlike the generic described above, the passive belongs to the domain of derivational morphology, and changes the valency of the target verb, which becomes intransitive. 4.1. a- passive prefix The first passive form, marked by the prefix a- (see table 1), is fully productive. It has three allomorphs, a-, - or k-, depending on the preceding prefix (Jacques and Chen 2007). Most transitive verbs can be prefixed with this element, and the resulting intransitive verb cannot have an overt agent. Table 1. Examples of passive forms basic verb

meaning

derived verb

meaning

mto

see

a-mto

be seen

prt rku

cut put in

a-prt a-rku

be cut be put in

Argument demotion in Japhug Rgyalrong

209

In the imperfective, past imperfective and past evidential forms, passives have a stative meaning, though a passive interpretation is also possible: (20) stfflha -Ðg n tfflu rÐl t-tango‚ --rku place 3sg.poss-in dem loc silver one-basket ipfv-pass-put_in ‘In that place, there is/someone has put a basket of silver.’ (The Divination, 68) (21) tfflheme n ci th-sta ri, girl det a_little aor-wake_up conj -mphs thcu n p-a-ta 3sg.poss-buttocks downstream dem pst.ipfv-pass-put fflti n.pst:be.a‰rmative ‘The girl woke up, and it (the horse foetus) was there / someone had put it under her bottom.’ (The Three Sisters, 106) (22) -phÐg n tfflu qaptm ci 3sg.poss-bosom dem loc pebble one na-rku -Ðu ‘‘tffletha qhuj tffle aor:3>3-put_in ipfv-be a_while this_evening conj ki a-phÐg a-rku tffle this 1sg.poss-bosom n.pst:pass-put_in conj tffletha p-mto-t-a ffio tffle, r˜ lpu a_while aor-see-pst-1sg adv conj king -phe tu-ti-a -ra’’ 3sg.poss-dat ipfv-say-1sg ipfv-have_to ‘He put a pebble in his bosom, thinking: ‘‘This evening, it will be in my bosom, and as I see it, I will certainly tell (the story) to the king.’’’ (Kunbzang, 279) No overt agent can appear in the same clause, but this is not to say that the passive is forbidden when the agent is known. For instance, in one text we find the following sentence: (23) t-rdo‚ p-a-qr tffle, -Ðg n tfflu one-piece aor-3>3-tear conj 3sg.poss-in dem loc rÐl qho‚qho‚ t-rdo‚ pj-k-mph r-ch, silver ingot one-piece ipfv.evd-pass-wrap-evd ‘He opened one piece (of bread), and there was a silver ingot wrapped inside it.’ (The Raven, 112)

210

Guillaume Jacques

Although not formally expressed in sentence (23), the agent who put the silver ingots in the bread is known. A few sentences earlier in the same story, we read: (24) rÐl qho‚qho‚ t-rdo‚ nts ko-mph r silver ingot one-piece always evd-wrap ‘She (a character named Lhamo) put a silver ingot in each (piece of bread).’ (The Raven, 109) In the aorist, passive forms do not always have resultative meaning (unlike in Kryz: Authier, this volume 3.2): (25) tffle ‚ffin ci rca´n k-wxtPwxti conj boy one unexpectedly nmlz:stat-intsvPbig ffio n-a-”zu -Ðu. adv aor-pass-make ipfv-be ‘Then the fox changed into a big boy.’ (The Fox, 193) The a- passive is rare in narratives and also in procedural texts, where generic verb forms are generally used in clauses with a generic agent. It is not commonly used to express situations where the agent is unknown: more usual is inverse (or plural direct) marking on the verb, as was seen in the previous section. Additionally, on occasion a passive form can display an idiosyncratic meaning which has evolved independently from that of the base verb. This phenomenon is exemplified by a pair of very common verbs in Japhug. The verb pa originally meant ‘do’ in proto-Rgyalrong, but in this meaning it was supplanted by the Tibetan loanword ”zu and was preserved only in its secondary meaning, ‘close’. However, the corresponding passive form a-pa, originally meaning ‘be made’, did not come to mean ‘be closed’; instead, it independently developed the meaning ‘become, change’, and thus pa and a-pa are no longer related synchronically. A prefix related to the a- passive is the s- prefix, which represents the fusion of the causative s- and the passive a-. Verbs with this prefix are few in number, and the situation is quite confused due to the fact that two other homophonous s- prefixes exist in Japhug, one of which is an antipassive marker (cf. the following section). The combination of passive and causative generates the special meaning ‘ask sb to’. For instance, from the verb mbi ‘give’7 one derives s-mbi ‘ask sb for sth’, in other words ‘cause sb to give sth to oneself ’. 7. This verb has primative/secundative alignment (Haspelmath 2005). The recipient is treated as the O, and the theme is not encoded on the verb. Other ditransitive Japhug verbs, however, have indirective alignment, as mentioned above.

Argument demotion in Japhug Rgyalrong

211

(26) ¯jtsu t-khtsa a-n-t-s-mbi chilli one-bowl irr-pfv-2-caus:pass-give ‘You will ask him for a bowl of chilli.’ (The Smart Rabbit, 40) Up to this point, we have taken for granted that a- is to be considered a real passive formant. We wish to dispel any suspicion that this might not be the case. Keenan and Dryer (2007) warn against confusing the passive with other constructions such as middle (p. 352), unspecified subject (p. 354), inverse (p. 356) and antipassive (p. 359). However, we know that Japhug has separate forms for generic reference (section 3), inverse (Jacques 2010a) and antipassive (section 5), all of which are entirely distinct from the a- prefix. Therefore, the main issue to be discussed is whether the a- form is a genuine passive or should be seen as expressing middle voice. Cross-linguistically, the term ‘middle’ is used for various verbal forms which generally cover a wide functional range, such as passives, reciprocals, reflexives, autobenefactives, and forms to denote actions a¤ecting an object possessed by the agent. The presence of middle marking has been reported for several Sino-Tibetan languages, such as Rawang (LaPolla 2001). Therefore, the presence of a typologically similar grammatical category in Japhug would not be surprising. However, it seems clear that the a- prefix has a much more restricted range of functions than we would expect if it were a middle marker. In particular, the a- passive never has reflexive meaning: reflexive forms are regularly made from transitive verbs by adding the prefix ffi'- (e.g. sat ‘kill’, ffi'-sat ‘commit suicide’).8 However, it is possible that in an earlier stage of the language, the a- prefix had a broader range of uses than merely the agentless passive. One productive verbal form which is historically related to the passive is the reciprocal (see table 2). It is produced by adding the a- prefix and reduplicating the last syllable of the verb stem. A minority of reciprocal verbs show no stem reduplication, but instead add the double prefix a-m-. All reciprocal verbs are morphologically intransitive.

8. On the origin of this prefix, see Jacques (2010b).

212

Guillaume Jacques

Table 2. Examples of reciprocal forms basic verb

meaning

derived verb

meaning

ndza

eat

a-ndz-ndza

eat each other

nrts˛a

envy

a-nrts˛-ts˛a

envy each other

lt

throw

a-l-lt

fight each other

mto ti

see say

a-m-mto a-m-ti

see each other tell each other

The reciprocal form in a- is obviously related to the a- passive, and this could challenge our claim that the prefix a- is a real passive formant. However, the relationship between the passive and the reciprocal is purely historical, and is not synchronically valid in modern Japhug. The common ancestor of the passive formant a- and the a- element seen in the reciprocal must have taken the shape *Ða- in proto-Japhug (Jacques and Chen 2007: 889). This *Ða- prefix must have functioned as a much broader intransitive marker and not strictly a marker of the passive. From a typological perspective, it may be instructive here to compare intransitive a‰xes in other Sino-Tibetan languages. In Rawang, the intransitive v- [ ] prefix can derive passive-like verb forms, whose S corresponds to the O of the transitive verb (LaPolla 2001: 288): (27) tv´l-o`-e# ! v-tv´l-e# roll-3.tr-n.pst intr-roll-n.pst ‘roll (tr)’ ‘roll (intr)’ Moreover, according to LaPolla (2001: 288), ‘‘if the single direct argument of the derived intransitive is a plural animate argument, then the meaning is reciprocal’’. Here is one of his examples: (28) a`ngmaq v-shvt-e# they intr-hit/kill-n.pst ‘They are fighting.’ The functions of the v- prefix in Rawang are reminiscent of those of the proto-Japhug *Ða- prefix.9 The main di¤erence is that in Japhug (as in all four of the Rgyalrong languages) the reciprocal meaning only appears if the verb stem is reduplicated. 9. This was suggested to me by an anonymous reviewer of the journal Language and Linguistics, whom I wish to thank again for insightful comments.

Argument demotion in Japhug Rgyalrong

213

Interestingly, in Rtau, a language related very closely to the Rgyalrong languages10, verb stem reduplication expresses plurality in the subject (S/A) (Huang 1991: 29–30), as shown in table 3. Table 3. Paradigm of the verb ‘go’ in Rtau. person

singular

plural

1

fflo-Ð

ffl -fflo-Ð

2 3

ffli-n ffl

ffl -ffli-n ffl -ffl

If we assume that proto-Japhug *Ða- had a function similar to the Rawang v-, and that the reduplication present in Rtau as a marker of plurality is a survival from proto-Rgyalrongic, the formation of the reciprocal in the modern Rgyalrong languages can be explained: when a *Ða- prefixed verb had a plural animate argument, the verb stem was reduplicated, and it developed a reciprocal meaning. After verb stem reduplication ceased to function as a marker of plurality, its presence on *Ða- prefixed verbs was reanalysed as a reciprocal marker. Therefore, although the a- passive prefix and the a- element in reciprocal forms are historically related, they no longer represent the same morpheme in modern Japhug. This is why we argue that the form in a- is a genuine passive and not a vague intransitive or middle. Finally, it should be noted that many intransitive verbs, such as armbat ‘be near’ or afflqhe ‘cough’, have a prefixal a- element although no corresponding transitive verb exists. We have no reason to identify such verbs as passives, and can consider the a- element as part of the verb root. However, it should be stressed that no transitive verb has this a- element. 4.2. Prenasalized anticausative The prenasalization alternation in Japhug derives an anticausative verb from a transitive one. Only sixteen pairs of verbs present this alternation (Jacques 2008: 84–5), and at least one of them ( ¯tr, table 4) is a loan10. Sun (2000) argues convincingly that Rtau and Lavrung are the closest relatives of Rgyalrong within Qiangic. He calls the sub-branch including the Rgyalrong languages, the Lavrung languages and Rtau ‘Rgyalrongic’, a term which we also adopt here.

214

Guillaume Jacques

word from Tibetan (gtor), which shows that this morphological process retained its productivity until recently.11 Table 4. Examples of the prenasalized anticausative in Japhug transitive

meaning

intransitive

meaning

fts˛i prt qt ¯tr

melt (tr) break (tr), cut separate scatter

ndz˛ i mbrt ngt ‚ndr

melt (itr) break (itr), be cut be separated be scattered

The major di¤erence between the a- passive and the prenasalized anticausative is that the agent is omitted when the a- passive form is used, but semantically, the existence of an indefinite external agent is not excluded, whereas in the case of the prenasalized anticausative, no agent is present, and the action is viewed as taking place spontaneously. (29) ‘‘wo a-ffii ra n-mk' r Oh 1sg.poss-lady pl 3pl.poss-necklace p-mbrt’’ tı´-n aor-anticaus:break n.pst:say-pl ‘‘‘Oh, my lady, your necklace has broken!’’ they said.’ (Kunbzang 214) In this example, the characters uttering this sentence believe that the necklace broke by itself, without an external agent: they therefore use an anticausative form. If the a- passive a-prt had been used instead, it could have implied that someone had broken the necklace on purpose. However, since only a few verbs have a special anticausative form, for the majority of verbs only the a- passive is available to express anticausative meaning. There are phonological constraints on prenasalization: all sixteen known examples of transitive verbs to which this process applies have a voiceless unaspirated stop or a¤ricate in initial position.12 Verbs with nasal or fricative initial consonants have no distinct anticausative form.

11. However, the phonological correspondence -or : -r shows that it belongs to the earliest layer of Tibetan loanwords (Jacques 2008: 136–146). 12. We use ‘initial’ here in the sense defined in Jacques (2004: 12–73).

Argument demotion in Japhug Rgyalrong

215

5. Antipassive Sun (2006: 8) was the first scholar to notice the existence of antipassive forms in a Rgyalrong language, namely Tshobdun (Caodeng). He pointed out that two antipassive prefixes exist in this language: r½-, which is used when the patient is non-human, and s½-, used with human patients (see table 5). Japhug is very closely related to Tshobdun, and the same distinction applies for the cognate prefixes r- and s-, as can be seen from the following examples: r- is added to verbs with a prototypical non-human patient, and s- when the patient is necessarily human. Table 5. Examples of antipassive derivation in Japhug basic verb

meaning

derived verb

meaning

ndn

read

r-nd n

read sth.

nts'e fflar

sell search

r-nts'e r-fflar

do business look for sth.

fst n

serve

s-fst n

serve sb.

‚nd fflar

hit search

sa-‚nd s-fflar

hit sb. look for sb.

The resulting verbs are intransitive: they never display any of the signs of verbal transitivity described in part 2 of this paper. The antipassive is not simply a means of changing the case marking of the Agent from Ergative to Absolutive: it involves a change in valency, with the result that the original patient can no longer be expressed. Usually, as in examples (30) and (31), the semantic patient becomes indefinite: (30) t-rffia” n pj-r-fflpht neu-wife det ipfv.evd-apass:nonhum-mend ‘The wife was mending (clothes).’ (The Raven, 19) (31) tfflhi tu-t-ste Ðu k-s-fstn what ipfv-2-do_this_way[3] n.pst-be inf-apass:hum-serve ‘How do you serve (your husband and the people from his family)?’ (The Frog, 128) In a few cases in our texts, the (human patient) antipassive seems to be used as a way to avoid using the first person when asking for something, thereby making the request more indirect:

216

Guillaume Jacques

(32) tcirq k-s-jtshi  -tu cold_water nmlz:O-apass:hum-make_drink Q-n.pst:have ‘Is there any cold water to give (us) to drink?’ (Nyima Vodzer, 35) The human/non-human distinction is not always strictly observed. In some examples, a s- prefixed verb may have a non-human animate (animal) as its intended patient: (33) th-wxti -jja tffle, aor-big 3sg.poss-following conj k-s-ndza k-Ðu n pj-s¯sl nmzl:S/A-apass:hum-eat nmzl:S/A-be det ipfv-realize ‘As (the buzzardi) grows bigger, (the bunting) realizes that iti eats (other birds).’ (The Buzzard, 26) In this example, since semantically both the patient and the agent of the verb ‘eat’ are birds and therefore equals on the Empathy Hierarchy, the antipassive with suppressed human patient prefix s- is found instead of the expected r-. The Japhug antipassive does not seem to be very productive, and few verb roots are able to take both r- and s- prefixes (fflar ‘‘to search’’ in table 5 is one example). In both narratives and procedural texts, these forms are quite rare (cf. the text counts in the conclusion). 6. De-experiencer The de-experiencer prefix s-, homophonous with the antipassive prefix for human patients discussed in the previous section, di¤ers from the other morphological devices presented in this paper. Its basic function is not to decrease valency per se: rather, it derives an intransitive stative verb from an intransitive verb or a transitive verb of perception. The S of the derived verb denotes the stimulus of the state or action, and it has the meaning ‘‘be liable to cause sb/sth to X’’, where X is the meaning of the basic verb. In the case of intransitive verbs, the original S argument is suppressed and replaced by the stimulus. For transitive perception verbs, the A (corresponding to the experiencer) is suppressed and the original O (the stimulus)

Argument demotion in Japhug Rgyalrong

217

becomes the S of the derived verb. The examples in table 6 illustrate this derivation. Table 6. The de-experiencer prefix with intransitive and transitive verbs Basic verb

meaning

derived verb

meaning

Ðgio

slip (itr.)

s-Ðgio

be slippery (of the ground)

scit

be happy

s-scit

be nice (of a situation), be funny (of a person)

fflke

be burned

s-fflke

rga

like (itr.)

s-rga

be burning (of a boiling kettle, which burns the hand of anyone touching it) be nice

mto mtshm

see hear

s-mto s-mtshm

be easy to see be easy to hear

Since the deleted argument is always the experiencer (whether the verb is transitive or intransitive), we label this prefix ‘‘de-experiencer’’.13 Although in the case of intransitive verbs there is no decrease in valency, the addition of this prefix is nonetheless a demotion in the sense that a stimulus is lower than an experiencer in terms of agentivity (for instance, humans are less likely to be stimuli).

7. Lability While transitive and intransitive verbs can be easily distinguished on formal grounds (see section 2), a small class of verbs can be either transitive or intransitive. Table 7 presents a list of the labile verbs identified in Japhug up to this point.

13. This terminology was suggested by Peter Austin (p.c.). The counterpart of the de-experiencer prefix among transitivizing a‰xes is the estimative prefix n-, which derives a transitive verb from an intransitive one whose A corresponds to the experiencer and whose O corresponds to the stimulus (the S of the basic verb), for instance mpfflr ‘‘be beautiful’’ > nmpfflr ‘‘consider to be beautiful’’.

218

Guillaume Jacques

Table 7. Labile verbs in Japhug Japhug

meaning

Japhug

meaning

sÐo nmo

listen

r˜o‚˜ i

crush

look

'ndffi r

grind

rÐu

fry

nkhaja

resist

m rk

steal

sla‚rd”

kick (of an animal)

rpu

bump

sqarts

kick

fffli

forge

nffimÐn

envy

ffllu ta‚

plough weave

n mbrp

ride (on horseback)

For instance, the verb rpu ‘bump into’ can appear both with and without the 1sg/2sg Aorist -t su‰x, cf. the following examples: (34) -ta‚ k-rpu-a 3sg.poss-up aor:east-bump_into-1sg ‘I bumped into him.’ (35) a-rpa‚ -ta‚ k-rpu´-t-a 1sg.poss-elbow 3sg.poss-up aor:east-bump_into-pst-1sg ‘I bumped into him with my elbow.’ In sentence (35), the verb bears a transitive -t su‰x, and the patient of the clause, ‘my elbow’, is overt (the person bumped into is indicated with an oblique case). In sentence (35), in contrast, the body part involved is not expressed, and the verb lacks the transitive marker; here -ta‚ ‘on him’ is syntactically not the object of the verb but an adjunct. The lability observed in these two examples is agent-preserving: the agent of the transitive verb, and not the patient, remains when the verb is used intransitively.14 We will label the S/A argument of labile verbs as the ‘subject’. Since nouns in Japhug take ergative marking, this implies that the subject is marked with the ergative case when the verb is transitive, and has absolutive marking when the verb is intransitive. Consider the following examples, involving the labile verb n mbrp ‘ride’. (36) t-rdo‚ n k  t-a-nmbrp one-piece det erg aor-3sg>3-ride ‘One of them rode it.’ (The Tiger, 11) 14. This phenomenon is described by Næss (2007: 125) as ‘Indefinite Object deletion’.

Argument demotion in Japhug Rgyalrong

219

(37) -ta‚ n tfflu ttffl n to-nmbrp 3sg.poss-up det loc boy det evd-ride ‘The boy rode on her.’ (The Flood, 83) In example (36), the agent is marked with the ergative. The patient is not overt, but the transitive a- prefix on the verb and the presence of ergative case indicate that the verb is to be interpreted as transitive: the patient is definite (it refers to a tiger mistakenly stolen by three thieves). In example (37), the subject ‘the boy’ does not bear ergative case, and the verb must be interpreted as intransitive (though in this example the evidential verbal form itself is ambiguous as regards transitivity). As in examples (34) and (35), -ta‚ ‘on her’ is an adjunct, and does not participate in the verb’s argument structure. Similar examples could be found for all verbs in table 7. These verbs belong to two distinct semantic categories: perception/ feeling (‘listen to’, ‘look at’, ‘envy’) and concrete action (‘grind’, ‘bump into’, ‘forge’ etc). Interestingly, no ‘verbs of ingestion’ such as ‘eat’ are found in this list. Næss (2007: 126) claims that ‘‘For most languages for which I have been able to find data, it is the case that if they allow indefinite object deletion with any verb, they will allow it with ‘eat’’’, but the verbs ndza ‘eat’ and tshi ‘drink’ are strictly transitive in Japhug; intransitive verbs of ingestion do exist, such as rndztshi ‘have lunch’, but they are not related to their transitive equivalents by means of lability. Interestingly, none of the labile verb pairs discovered so far seems to show an unpredictable change in meaning, although these are very common cross-linguistically, as for instance with English ‘drink’, which as an intransitive principally bears the meaning ‘drink alcohol’. No examples of patient-preserving lability have so far been found in Japhug.

8. Incorporation The last argument-demotion device in Japhug is verbal incorporation. It is not widespread but restricted to a few verbs, though the fact that some examples involve recent Chinese and Tibetan loanwords shows that the process is still potentially productive. Incorporated nouns involve objects (see ‘timber’ and ‘money’ in the following table), subjects of intransitive verbs (‘horse’) and a few adjuncts (‘back’), but never the agent of a transitive verb. The incorporated noun

220

Guillaume Jacques

root appears before the verb root, following the normal strict verb-final order of the language. Some examples are given in table 8. Table 8. Examples of incorporated nouns in Japhug noun

verb

incorporation

meaning

mbro ‘horse’ qhu ‘back’

r˜' ‘run’ it. ru ‘look’ it.

n-mbr-r˜' n-qha-ru

gallop it. turn around, look back it.

si ‘timber’

pht ‘chop’ tr. '-s-pht

‘money’15

pfflawts qhu ‘back’

fso‚ ‘earn’ tr. Ðga ‘wear’ tr.

fell trees it.

'-pfflawts-fso‚ earn money it. n-qh-Ðga wear on the back tr.

Morphological marking on the incorporated noun is extensive. Firstly, the incorporated noun appears between a derivational prefix (n-, n- or '-) and the verb root. Secondly, if the noun ends in an open syllable, the vowel usually undergoes one of a set of regular changes also found in compound nouns, giving a form we call status constructus: -o ! -/-a -u ! -/-a -i ! - When the verb incorporates the patient of the original verb (see ‘fell trees’ and ‘earn money’ above), it becomes intransitive: this is a case of saturating incorporation. When, however, the incorporated noun is an adjunct (‘wear on the back’), it does not a¤ect the valency of the verb. Most incorporated objects can appear with the same verb as free objects, but there are a few frozen incorporated objects which are no longer felt as such. Sometimes both forms appear next to one another in the same story, as can be seen in examples (38) and (39). (38) ´asa ju-k-ffle tffle, n tfflu Lhasa ipfv-genr:S/O-go conj dem conj pfflawts k-fso‚ -mbat money inf-earn const-easy ‘If one goes to Lhasa, money is easy to earn there.’ (Lobzang, 22) 15. The word pfflawts ‘paper money’ comes from the colloquial Chinese piaozi .

Argument demotion in Japhug Rgyalrong

221

(39) n-mbro n-rÐl tu-rke-a tffle 2sg.poss-horse 2sg.poss-silver ipfv-put_in[3]-1sg conj k-'-pfflawts-fso‚ j-ffle tffle nmlz:S/A-derivation-money-earn imp-go conj ‘I will prepare a horse and some silver for you: go to earn money.’ (Lobzang, 17) In (39), pfflawts ‘money’ is incorporated, whereas in (38) it appears as a free object. As these examples show, the semantic di¤erence between free and incorporated object is minimal here; in both cases the object is indefinite. The decrease in valency caused by incorporation shows some similarity with the e¤ects of antipassivization. The incorporated object can never be definite, and there are strong restrictions on which objects can be incorporated. The activity expressed by the verb must be, as Mithun (1984: 848) puts it, ‘‘recognized su‰ciently often to be considered name-worthy in its own right’’. For instance, the verb mtshi ‘lead’ can only appear with two incorporated objects, jla ‘yak/cow hybrid’ and mbro ‘horse’, never with any other animal.

9. Conclusion In Japhug, covert arguments of transitive verbs are generally interpreted as definite, though a few exceptions are treated in this article. Therefore, indefinite arguments, in particular unknown or generic ones, must be demoted. Patients can be suppressed by means of four distinct constructions – generic, antipassive, lability and incorporation; while only two possibilities exist for the suppression of agents, namely the generic and passive constructions. Japhug has a wide range of very specialized argument demotion devices, including passive, reciprocal, de-experiencer and reflexive (not mentioned in the present article; see Jacques 2010b). We have already outlined some important di¤erences in the use of these forms, but additional insight may be gained from text counts. The distribution we found in a corpus of eight narrative texts, and another comprising two procedural texts, is presented in table 9.16

16. No examples of the de-experiencer s- with a transitive verb were found in this part of the corpus.

222

Guillaume Jacques

Table 9. Text counts of argument demotion narrative inverse (excluding generic)

procedural text

35

3.1%

0

A-generic

2

0.2%

215

O-generic

0

0%

0

0%

S-generic

4

0.4%

2

0.5%

labile verb as transitive

2

0.2%

0

0%

564

51.1%

46

10.5%

antipassive

3

0.3%

0

0%

passive

8

0.7%

6

1.4%

27

2.4%

0

0%

1

0.1%

0

0%

459

41.4%

168

other transitive

labile verb as intransitive incorporation other intransitive

0% 49.2%

38.4%

In both narrative and procedural texts, S/O generic forms, antipassive and incorporation are all barely attested. Passive verb forms are less rare, but the majority of examples involve one of just three common verbs: a-rku ‘be placed in’, a-ta ‘be placed’ and a-”zu ‘become’. In comparison with such languages as Kutenai (Dryer 1994: 75), where passive forms constitute up to 27% of all semantically transitive clauses, it is clear that the Japhug passive is very restricted in usage. The major di¤erence between the two categories of texts lies in the generic use of the inverse. Almost unattested in narratives, it is by far the most common form in procedural texts. Of all five argument-demoting mechanisms, the generic is the only one which is fully productive. It is instructive to evaluate to what extent the syntactic properties of the four patient-demoting constructions in Japhug are common for such constructions cross-linguistically. Bickel et al. (2007: 18) propose a set of ten properties commonly associated with antipassive (A), incorporation (I) and optional agreement (OA). The properties of the Japhug constructions are given in table 10.17 17. The abbreviations G (J), A (J), L (J) and I (J) correspond to the Generic, Antipassive, Lability and Incorporation constructions of Japhug. The symbols y, n, n/a and ? stand for ‘yes’, ‘no’, ‘not applicable’ and ‘true in some languages’.

Argument demotion in Japhug Rgyalrong

223

Table 10. Typological comparison of syntactic properties of Patient-Demoting constructions A

I

OA

G (J)

A (J)

L (J)

I (J)

O is adjacent to verb

n

y

n

n

n

n

y

O and verb in one word

n

y

n

n

n

n

y

O is obligatory

n

y

n

n

n

n

y

O is deleted

?

n

n

y

y

y

n

O is a fully-fledged NP

y

?

y

n/a

n/a

n/a

n/a

Relativization on O

y

?

y

n/a

n

n

n

Regular O-case on O

n

n

y

n/a

n/a

n/a

n/a

Transitive A-case on A

n

n

y

y

n

n

n

Suspended O-agreement

y

y

y

y

y

y

y

Generic O or O with unknown cardinality

n

y

?

y

either

y

y

Incorporation aside, the three other constructions di¤er very little as far as these ten properties are concerned. As is the case for many antipassive constructions cross-linguistically, in the Japhug Antipassive the demoted patient cannot be retained in the clause concerned, even in an oblique case. The Japhug Passive behaves analogously, as its use rules out the specification of an Agent in the clause. As we have seen in the present article, the Japhug Passive and Antipassive have a variety of functions, but they do not serve to topicalize the Patient/Agent (word order, elision of the argument and inverse marking are used instead) or make the Patient/Agent argument accessible to relativization (as S, A and O can all be relativized). They do not share the whole set of parameters which are commonly associated with the terms ‘passive’ and ‘antipassive’, but nevertheless these terms seem to be appropriate labels to describe the behaviour of the a‰xes discussed here. Whether these properties are cross-linguistically common or rare, and whether some of them are somehow correlated with ergative alignment, are questions which remain to be investigated.

224

Guillaume Jacques

References Bickel, Balthasar, Martin Gaenzle, Arjun Rai, Prem Dhoj Rai, Shree Kumar Rai, Vishnu S. Rai and Narayan P. Sharma 2007 ‘‘Two ways of suspending object agreement in Puma: between incorporation, antipassivization and optional agreement.’’ Himalayan Linguistics 7: 1–19. DeLancey, Scott 1981 ‘‘The category of direction in Tibeto-Burman.’’ Linguistics of the Tibeto-Burman Area 6(1): 83–101. Dryer, Matthew S. 1994 ‘‘The discourse function of the Kutenai inverse.’’ In Voice and Inversion, T. Givon (ed.), 65–99. Amsterdam/Philadelphia: Benjamins. Haspelmath, Martin 2005 ‘‘Argument marking in ditransitive alignment types.’’ Linguistic Discovery 3(1): 1–21. Hopper, Paul J. and Sandra S. Thompson 1980 ‘‘Transitivity in grammar and discourse.’’ Language, 56(2): 251– 299. Huang Bufan 1991 Daofuyu [The Rtau language], in Dai Qingxia, Huang Bufan, Fu Ailan, Renzeng Wangmu and Liu Juhuang (eds.), Zangmianyu shiwu zhong [15 Tibeto-Burman languages], Beijing: Beijing Yanshan Chubanshe , p. 1–45. Jacques, Guillaume 2004 Phonologie et morphologie du Japhug (rgyalrong), the`se de doctorat, universite´ Paris VII-Denis Diderot. Jacques, Guillaume 2008 Jiarongyu yanjiu [Study on the Rgyalrong language]. Beijing: Minzu chubanshe. Jacques, Guillaume 2010a ‘‘The inverse in Japhug Rgyalrong.’’ Language and Linguistics, 11(1): 127–157. Jacques, Guillaume 2010b ‘‘The origin of the reflexive prefix in Rgyalrong languages.’’ Bulletin of the School of Oriental and African Studies, 73(2): 261–268. Jacques, Guillaume and Chenzhen ‘‘ Chabaohua de bujiwu qianzhui ji xiangguan wenti [The intransitive prefix in Japhug and related problems].’’ Language and Linguistics, 8(4): 883–912. Jacques, Guillaume and Alexis Michaud 2011 The reconstruction of Proto-Naic: a preliminary study of the historical phonology of highly eroded Sino-Tibetan languages, Diachronica, 28.4: 468–498.

Argument demotion in Japhug Rgyalrong

225

Keenan, Edward L. and Dryer, Matthew S. 2007 ‘‘Passive in the World’s Languages.’’ In Clause Structure, Language Typology and Syntactic Description, Vol. 1, Timothy Shopen (ed.), 325–361. Second Edition. Cambridge University. LaPolla, Randy 2001 ‘‘Valency-changing derivations in Dulong/Rawang.’’ In Changing valency: Case studies in transitivity, R.M.W. Dixon and A.Y. Aikhenvald (eds.), 282–311. Cambridge: Cambridge University Press. LaPolla, Randy 2008 ‘‘Transitivity and transitivity alternations in Rawang and Qiang.’’ In Proceedings of the Workshop on Tibeto-Burman Languages of Sichuan, 239–248. Academia Sinica, Institute of Linguistics, November 21–22. Mithun, Marianne 1984 ‘‘The evolution of noun incorporation.’’ Language 60: 847–94. Mithun, Marianne 1999 The Languages of Native North America. Cambridge Language Surveys. Cambridge: Cambridge University Press. ˚ shild Næss, A 2007 Prototypical Transitivity. Amsterdam/Philadelphia: John Benjamins. Sun, Jackson T.-S. 2000 ‘‘Parallelisms in the verb morphology of Sidaba rGyalrong and Lavrung in rGyalrongic.’’ Language and Linguistics 1(1). 161–190. Sun, Jackson T.-S. 2003 ‘‘Caodeng rGyalrong.’’ In Sino-Tibetan Languages, Graham Thurgood and Randy J. LaPolla (eds.), 490–502. London: Routledge. Sun, Jackson T.-S. 2005 ‘‘Linguistic Coding of Generic Human Arguments in rGyalrongic Languages,’’ 11th Himalayan Languages Symposium, 6–9 December, 2005, Chulalongkorn University, Bangkok, Thailand. Sun, Jackson T.-S. 2006 ‘‘ Jiarongyu dongci de paisheng xingtai [Derivational morphology in the Rgyalrong verb].’’ Minzu yuwen 4: 3–14. Sun, Jackson T.-S. and Shidanluo Caodenghua 2002 yu « rentong dengdi » xiangguan de yufa xianxiang [Caodeng Rgyalrong and grammatical phenomena related to the Empathy Hierarchy], Language and Linguistics, 3.1: 79–99. Zu´n˜iga, Fernando 2006 Deixis and Alignment: Inverse Systems in Indigenous Languages of the Americas. Amsterdam/Philadelphia: Benjamins.

The Katukina-Kanamari antipassive Francesc Queixalo´s 1. Introduction1 If we take linguistic form seriously – a stand which does not preclude the necessity of taking seriously other levels of linguistic patterning such as semantics and pragmatics – then some level ought to be defined on purely formal grounds. I take the grammatical relations ‘subject’ and ‘object’ as being among such formally defined linguistic notions. In dealing with a linguistic entity, and with grammatical relations in particular, we are likely to have much to gain by clearly distinguishing between the definition of that entity, i.e. the set of properties which identify it, and its motivation, i.e. the factors which lead a speaker to put it to use (see section 2.5). The main aim of this paper is to present data from a very sparsely documented, strongly ergative language of Amazonia as an illustration of these considerations. The particular grammatical point chosen for this purpose concerns the basic bivalent transitive clause and its associated valency-reducing voice, the antipassive. I will attempt to show that in a clearly hierarchized system of grammatical relations, voice alternation has a primary target, namely the accessibility restrictions bearing on the lower-ranked argument of a two-place clause. Of course, the idea that such a generalization could encompass antipassives in all the di¤erent kinds of so-called ergative languages, and even passives in languages of all kinds, is on the face of it a blatant impossibility: functional (semantic and pragmatic) motivations for the existence of voice are too pervasive everywhere for a purely syntactic explanation to be convincing. Nevertheless, the interplay between these two types of motivation for voice alternations is somewhat obscured by the quest for all-or-nothing (formal vs. functional) typological definitions (for the passive, see Comrie 2008; Givo´n 2008). In the spirit of Cooreman’s (1994) work on antipassives or Givo´n’s (2009) on passives, diachrony – a repository for the e¤ects of

1. Many thanks to Katharina Haude, Tomas Givo´n and Gilbert Lazard for comments on a previous version of this paper.

228

Francesc Queixalo´s

formal and functional pressures – still has much to say on voice mechanisms as partial subproducts of the interaction between 1) the basic alignment type of a language, and 2) the communicative needs of speakers. Katukina-Kanamari, apparently the only surviving language of the small Katukina family, is spoken by at most two thousand people in a large region of Amazonia located between the Purus and the Javari rivers and between the Japura river and the southern edge of the state of Amazonas, Brazil. Adelaar (2000) suggests a genetic link with the Peruvian isolate Harakmbut/Amarakaeri. The language can be characterized typologically by the following features: simple phonology, sparse morphology, clear-cut lexical classes with no primitive adjectives, no trivalent verbs other than ‘say’, head marking, predicate-initial2 constituent order, strong configurationality in terms of constituency, a neat grammatical relations hierarchy, predominant ergative patterning in almost all aspects of its grammar, and split transitivity. The paper is organized in the following manner. I first present the basic patterns in terms of formal features such as coding, constituency, behaviour, and coreference control, followed by an explicitation of the consequences of these phenomena for the question of grammatical relations. Then I turn to describe a voice alternation which operates upon the basic bivalent transitive clause: the participant mapped to the highest grammatical relation is deprived of its core argument status, thereby allowing the extant participant to fill the unique argument slot of a one-place clause.

2. Basic patterns The most unrestricted way of rendering an event involving ‘‘Mayon’’ (personal name), ‘‘cut’’ and ‘‘wild meat’’ is shown in (1)ITQ Mayon-na¼ tukman barahai Mayon-MkCase¼ cut wild_meat 3 ‘Mayon cut the wild meat.’

2. But with some amount of flexibility, as we will see. ‘‘Predicate’’ is taken in the sense ‘‘predicate phrase’’. 3. ITQ refers to the Itaquai river, where the data from the Kanamari dialect were collected. BIA will stand for the river Bia, a Jutai tributary and the origin of the data from the Katukina dialect.

The Katukina-Kanamari antipassive

229

where the verb is in medial position, the patient phrase (unmarked for case) follows the verb, and the agent phrase (case-marked by na) precedes the verb.4 For an event involving ‘‘man’’ and ‘‘go away’’, we have (2)ITQ daan piya go;walk man ‘The man went away.’ Lexically monovalent verbs such as daan do not appear in other types of clause structure. 2.1. Coding Comparison of (1) and (2) demonstrates that Katukina-Kanamari has clear ergative alignment in terms both of case marking – the agent in (1) is explicitly marked for case, while the patient in (1) is unmarked, like the unique argument in (2) – and word order, agents being located preverbally but patients and unique arguments postverbally. Moreover, pronominal forms are bound for agents but free for patients and unique arguments, as shown in (3) and (4) respectively: (3)BIA 1 2 3

singular yo-5 noa-

plural tyonama-

4. Specific semantic roles are not particularly at issue in this paper, so for twoplace verbs I will simply rely on the prototype semantics-based terms agent and patient. MkCase stands for ‘marked case’, for which a justification will be proposed below. Phonologically, the case su‰x cliticizes to the verb, yielding the sequence of phonological words #mayon# #natukman# #barahai#. This is assumed to be the result of a diachronic process of procliticization of the case su‰x to the phrase head (other examples of head attraction include auxiliarization, see 2.4). In spite of the well-established tradition of organizing grammatical examples on the basis of the phonological properties of clitics rather than on their grammatical properties, I adopt a di¤erent convention in such a case of strict contiguity between the grammatical host and the phonological host of an intermediate form: the clitic na is restored to the position beside its grammatical host, the notation {A-x ¼ B} reflecting the restitution of the element x, phonologically bound to an adjacent following element B, to its adjacent preceding grammatical host A. 5. The tables follow the dialect showing the simplest allomorph inventories.

230

Francesc Queixalo´s

(4)ITQ 1 2 3

singular adu idi:k anyan6

plural adik idi:ki anyan hinuk

No coding of gender surfaces in pronominal forms. 2.2. Constituency The predicate and its external (sister) argument7 appear in that order. The reverse order is also allowed, as in (5), though this probably brings with it a slight pragmatic e¤ect. (5)ITQ piya daan man go;walk ‘The man went away.’ A subclass of lexical predicate heads generates syntactically complex predicate phrases. These are: bivalent verbs (6), bivalent nouns (7), and postpositions (8). All take an obligatory internal, pre-head argument. All other lexical predicate heads, i.e. monovalent verbs (9), monovalent nouns (10), and adverbs (11)–(12), only take an external, post-predicate argument. Examples follow. Bivalent heads: verb (6)BIA [Ayobi-na¼ bo:dak] tawami Ayobi-MkCase ¼ roast manioc_dough ‘Ayobi is roasting the manioc dough.’ noun (7)ITQ [opatyin-na¼ wadik] Warohan child-MkCase¼ name Warohan ‘Warohan is the child’s name.’

6. Third person pronouns seem to be demonstrative in origin. 7. I use ‘‘argument’’ in its formal sense of ‘‘linguistic expression of a participant required by the semantic structure of the verb’’, as I do for ‘‘internal’’ and ‘‘external’’, that is, instantiated respectively inside or outside the predicate phrase.

The Katukina-Kanamari antipassive

231

postposition (8)BIA [Raidi-na¼ katu] Apikaru Raidi-MkCase¼ com.instr Apikaru ‘Apikaru is with Raidi.’ Monovalent heads: verb BIA (9) datikan pi:na sink hook ‘The hook sank.’ noun (10)BIA totyawa idi:k wa shaman you prospect ‘You will be a shaman.’ adverb (11)ITQ kodo kamodya in_the_higher_part monkey_sp. ‘The monkey (sp.) is up there.’ (12)BIA kiman idi:k quickly you ‘Quick!’ Clause-initial, final and intermediate positions can be occupied by discourse particles and adverbs in adverbial function. However, none of these, nor any other kind of word, can intervene between a phrase head and its left-adjacent case-marked dependent. For example, the particle niama ‘then’ can be added to the sentence in (6) above as follows: (13)BIA niama [Ayobi-na¼ bo:dak] tawami then Ayobi-MkCase ¼ roast dough [Ayobi-na¼ bo:dak] niama tawami Ayobi-MkCase ¼ roast then dough [Ayobi-na¼ bo:dak] tawami niama again-MkCase¼ roast dough then *[Ayobi-na¼ niama bo:dak] tawami Ayobi-MkCase ¼ then roast dough *[Ayobi niama -na¼bo:dak] tawami Ayobi then -Mkcase¼roast dough ‘Then Ayobi roasted the manioc dough.’

232

Francesc Queixalo´s

As will be amply illustrated below, all the external arguments listed so far occupy one and the same syntactic position. They also share the same coding features in terms of case (zero marking) and position (typically postpredicative). Since what we might wish to call the ergative case of (1), marked by -na¼, is in all coding respects identical to that of the genitive, example (7), and of the ‘‘object’’ of a postposition, example (8), I employ for these three semantically di¤erent but structurally identical instances of -na¼ the generic label marked case. In fact, -na as an allative marker is the only case a‰x allowed to occur on a noun phrase in an adjunct relation to the predicate, as in (14). All other meanings are rendered by postpositions.8 (14)BIA Koni-na¼ dahu wankurun hak-na Koni-MkCase ¼ take_away pot house-all ‘Koni took the pot to the house.’ 2.3. Behaviour This section is devoted to the syntactic asymmetries between external and internal arguments, and will show the extent to which the unique argument of monovalent verbs does indeed pattern together with the patient of transitive verbs. 2.3.1. Movement Only external arguments can be moved from their postpredicative position to a prepredicative, pragmatically marked one. This was seen in example (5) above, resumed here. (15)ITQ piya daan man go;walk ‘The man went away.’ This capacity for movement is shared by the external arguments of predicates of all kinds, e.g. bivalent verbs and nouns (illustrated in (16) and (17) respectively): 8. Of course, the diachronic hypothesis of a grammaticalization path allative > genitive > agentive, together with an ancillary hypothesis on the origin of postpositions as bivalent nouns, is appealing mostly because of its far-reaching consequences for the understanding of the genesis of Katukina ergativity (see Queixalo´s 2010). Postpositions display di¤erential object marking (see footnote 14).

The Katukina-Kanamari antipassive

233

(16)ITQ ma-obatyawa kotyia-na¼ dyoro 3pl-wife otter-MkCase¼ copulate_with ‘The otter copulated with their wives.’ (17)ITQ Inu Aro-na¼ tyo Inu Aro-MkCase¼ daughter ‘Inu is Aro’s daughter’ One pragmatic e¤ect available for external argument fronting is that of attenuated contrastive focus (see below). No movement of the internal argument is allowed unless other formal changes take place – such as the loss of case marking on the noun and the appearance of a personal prefix on the verb, (19)–(20). Compare, starting from the basic constituent order in (18): (18)ITQ nyama-na¼ kionyuk a-okpu mother-MkCase¼ comb 3sg-son (19)ITQ a-kionyuk nyama a-okpu 3sg-comb mother 3sg-son (20)ITQ nyama a-kionyuk a-okpu mother 3sg-comb 3sg-son ‘The mother combs her son.’ Regarding the pragmatic e¤ects of these movements, no information is available for (19), while (20) seems to bear a nuance absent from (16), along the lines of left-dislocated ‘as for the mother, she combs her son’. As both (19) and (20) show, extraction of the agent noun phrase has to leave behind the pronominal prefix on the verb. The constraint underlying this alternation is simply that the internal argument must be realized phonologically. In addition to any potential prosodic subtleties, the mere presence of the pronominal prefix instead of the proclitic case marker is evidence that the initial noun phrase nyama in (20), unlike in (18), is not a component of the verb phrase. Now, section 2.3.5 will show that while displaced external arguments retain their grammatical relationship with the predicate, displaced internal arguments fall outside the clause core. No attestation, whether spontaneous or elicited, shows the two movements simultaneously – both left-dislocation for the internal argument and fronting for the external argument. Moreover, my lack of control as regards prosodic cues for constituency has prevented me from submitting to my informants a plausible tentative example of a clause showing this behaviour. The default hypothesis, then, would be that (20) shows a

234

Francesc Queixalo´s

moved agent phrase and a patient phrase in situ, rather than something like (19) plus a fronted agent phrase. 2.3.2. Elision Only external arguments can be elided. (21)ITQ kitan-nin sleep-dur ‘(He) is sleeping.’ (22)BIA Koni-na¼ dahu Koni-MkCase ¼ take_away ‘Koni took (it) away.’ The ‘‘zero pronoun’’ allows for a third person indefinite reading (see section 3.2.10). Elision of an internal argument has the same consequence as extraction, namely the need for a pronominal verb prefix referring to this argument. (23)ITQ a-hudyi homo 3sg-bring hammock ‘He brought the hammock.’ An indefinite reading of the plural third person prefix is the closest equivalent to a functional passive in this language. (24)ITQ ma-dahu tyowipikon tyo 3pl-carry glass_beads excl ‘Someone took away the glass beads!’ 2.3.3. Ostension Only external arguments can be modified or replaced by a demonstrative, (25)–(26) and (27)–(28), respectively. (25)ITQ kitan-nin itiyan wa:pa sleep-dur this dog ‘This dog was sleeping.’ (26)BIA yo-hoki ityian 9 oman 1sg-put this log ‘I put this log (over there).’ 9. The phonological form of the demonstrative is slightly di¤erent in the two dialects.

The Katukina-Kanamari antipassive

235

(27)ITQ kitan-nin itiyan sleep-dur this_one ‘This one was sleeping.’ (28)ITQ wa:pa-na¼ ti itiyan dog-MkCase ¼ kill this_one ‘The dog killed this one.’ 2.3.4. Coordination No explicit element other than concatenation is involved in coordinating noun phrases. External arguments can be coordinated, (29)–(30), but not internal ones, (31). (29)ITQ opatyin-na¼ wu awa nyama a-ponhanya child-MkCase ¼ want his_one mother 3sg-sister ‘The child loves his mother and his sister.’ (30)ITQ tyuku wa:pa takara die dog hen ‘The dog and the hen died.’ (31)ITQ *Nodia Hanani-na¼ hoho-nin Owi 10 Nodia Hanani-MkCase ¼ call-dur Owi ‘Nodia and Hanani are calling Owi.’ 2.3.5. Focalization Contrastive focus is achieved on noun phrases by the device of moving them to clause-initial position and postposing the particle (ka)na. This process is available for external arguments but not for internal ones. (32)ITQ waro kana kitan-nin parrot foc sleep-dur ‘It is the parrot that is sleeping.’ na tyo-ikihak (33)ITQ wiri wild_pig foc 1pl-spear ‘It is a wild pig that we speared.’ 10. One single instance of this construction was accepted by a speaker from another geographical area (Jurua).

236

Francesc Queixalo´s

On the basis of (34), related sequences that were rejected by consultants in elicitation include (35)–(36), where the scope of the focus particle would be the internal argument. (34)ITQ mapiri-na¼ duni takara anaconda-MkCase¼ catch hen ‘The anaconda caught the hen.’ (35)ITQ *mapiri-na (ka)na duni takara (36)ITQ *mapiri (ka)na na¼duni takara Focus displays an asymmetry which can be viewed as evidence that the pre-predicate position – that is, the position preceding the predicate phrase – is di¤erent for displaced external vs. internal arguments. The former undergoes a dislocation which keeps it within the clause core, and in that position it can be focused, (32)–(33). The latter is truly leftdislocated (see section 2.3.1) and no longer available for focus, (37). (37)ITQ *waro kana a-boni wa:pa parrot foc 3sg-peck dog ‘It is the parrot that pecked the dog.’ 2.3.6. Constituent questions External argument positions are eligible to be questioned. (38)ITQ hanian tu11 tatan koniok-nin who(m)/what int here talk-dur ‘Who is talking here?’ tu no-toman? (39)ITQ hanian who(m)/what int 2sg-shoot ‘What did you shoot?’ Internal arguments cannot be questioned as such. 2.3.7. Relativization Data on relativization are too fragmentary to allow for reliable conclusions regarding the structure of relative clauses and the nature of the rela11. In this dialect, the negation particle tu – or a homophonous element – occurs in constituent and yes/no questions. In the Bia dialect the form of the interrogation particle is yu, whereas the negation particle remains tu. See below for two comparable examples.

The Katukina-Kanamari antipassive

237

tivizing element. The following is o¤ered as a first insight into one more extraction process that seems to show the same asymmetries as have already been seen in focus and interrogative constructions. In the Itaquai dialect the presumably deictic element nyan – perhaps related to the free third person pronoun, see (4) – opens the relative clause. The relativized noun generally follows; this fronting is strongly preferred, though the noun is sometimes found in situ. This means that relatives are basically ‘‘head-internal’’ in this language. The verb is su‰xed by -nin, which in independent clauses denotes durative aspect, see (38), and on subordinate predicates marks dependence. Only external arguments can be relativized. (40)ITQ i-hik nyan anyan piya waokdyi-nin 1sg-know deic this_one man arrive-dep ‘I know the man who arrived.’ nyan tukuna Kontan-na¼ dahudyi-nin (41)ITQ i-hi:k 1sg-know deic person Kontan-MkCase¼ bring-dep ‘I know the person that Kontan brought.’ 2.3.8. Nominalization Morphology in bivalent verb nominalizations shows that these are inherently patient-oriented, (42), as monovalent verb nominalizations are oriented toward the unique participant whatever its specific semantic role, as in (43)–(44). Nominalization is achieved by postposing to the lexical verb stem the deictic element nyan (seen just above for relativization), while retaining, in bivalent verbs, the original bound expression for the agent.12 Noteworthy di¤erences between nominalized verb and relativized clause include not only the position of the deictic element nyan, but also the absence vs. presence respectively of both a relativized noun and the subordinating verbal su‰x -nin. As for aspectuality, nominalization seems to be equally suitable for denoting entities involved in events, as in (42) and (43), and those characterized by properties / habitual activities, as in (44). (42)aITQ yo-wahak barahai 1sg-cook wild_meat ‘I cooked the wild meat.’ 12. As far as the Itaquai dialect is concerned. In the Bia dialect nominalizations are quite di¤erent in form, which suggests a diachronically recent innovation in at least one dialect, presumably the former.

238

Francesc Queixalo´s

bITQ bak tu yo-wahak nyan be_good neg 1sg-cook deic ‘The thing I cooked (lit. my cooked thing) is not good.’ (43)aITQ dapoki opatyin fall child ‘The child fell.’ bITQ dapoki nyan fall deic ‘The one that fell.’ (44)aITQ donman piya go_fishing man ‘The man went fishing.’ (44)bITQ donman nyan adu go_fishing deic 1sg ‘I am a fisherman.’ The bivalent verb agent is not directly accessible to nominalization, see 3.2.8. We have so far demonstrated that in basic bivalent clauses the argument representing the patient ranks formally above the argument representing the agent as far as constituency and behaviour properties are concerned (location with regard to verb phrase, elision, movement, extraction), and aligns with the unique argument of monovalent clauses in terms of these same properties as well as coding properties (case marking, pronominal paradigms). Let us now turn to a less neatly hierarchized domain, that of coreference. 2.4. Control The simplest way of characterizing how arguments establish coreference pivots between lexical noun phrases and zero or bound pronominal forms is to say that in some sub-domains the hierarchy is straightforward whereas in others the situation we face is somewhat fuzzy. (I will not supply here a fully-fledged exposition of the topic, which can be found in Queixalo´s 2004, 2010.) Let us begin with the latter. At the intraclausal level, functional (semantic, pragmatic) conditions prevail over syntactic constraints such as linear order and rank in the constituency hierarchy (‘‘c-command’’) for core arguments, (45)–(47) (square brackets delimit the verb phrase). In (45) the external argument controls

The Katukina-Kanamari antipassive

239

the possessive marker prefixed to the internal argument, as expected from the constituency hierarchy but not from linear order. In (46) the internal argument controls the possessive on the external argument, counter the constituency hierarchy but in agreement with linear order. Disjunct reference appears in (47). (45)ITQ [a1-obatyawa-na¼ todiuk] Mayon1 3sg-wife-MkCase¼ hate Mayon ‘Mayon1’s wife hates him1 (lit.: His1 wife hates Mayon1).’ (46)ITQ [Dawi1-na¼ bobo] a1-obatyawa Dawi-MkCase¼ beat 3sg-wife ‘Dawi1 beat his1 wife.’ buro:] a2-mimi (47)BIA [ pi:da1-na¼ jaguar-MkCase¼ lick 3sg-blood ‘Jaguar1 licked his2 blood.’ Other domains show a slight preference for pivots where the patient argument is involved. One example is intraclausal coreference between core arguments and adjuncts; (48) and (49) were originally uttered with the meanings given here, but in the absence of contextual clues informants tend to interpret the latter with the patient as the antecedent (i.e. ‘. . . in her house’). For disjunct reference a free pronominal form (anyan) is appealed to, (50).13 (48)ITQ Dawi1-na¼ toman Poroya2 [a2-wa mokawa katu]14 Dawi-MkCase¼ shoot Poroya 3sg-grn gun com.instr ‘Dawi1 shot Poroya2 with his2 gun.’ 13. The gender of pronouns in the English translations reflects the circumstances in which the examples occurred. Square brackets delimit the postpositional phrase. I have no explanation for the need to code disjunct reference on the postpositional phrase in (50) but not on the external argument of (47). Intersentential coreference is another domain where pivots involving the patient are preferred. 14. Mokawa, ‘gun’, and hak, ‘house’, are monovalent nouns – i.e. nouns unable to head a phrase containing a genitive. To do so, they must let the generic relational noun (GRN) -wa mediate between them and the genitive expression, whether this is a person prefix, as in the example, or a case-marked lexical noun. Usually, non-human nouns as internal arguments of postpositions do not take the case marker -na¼.

240

Francesc Queixalo´s

(49)ITQ Dawi1-na¼ tohik ityaro2 [a1-wa hak to] Dawi-MkCase¼ see woman 3sg-grn house loc ‘Dawi saw the woman in his house.’ (50)ITQ Dawi1-na¼ tohik ityaro2 [anyan3-na¼ wa hak to] Dawi-MkCase¼ see woman 3sg-MkCase ¼ grn house loc ‘Dawi1 saw the woman2 in his3/her3 house.’ Involvement of the patient in the coreference pivot is the default with intraclausal adverbs of manner and location, and in subordination. I give an example of the latter only, in order to keep this section within reasonable limits (square brackets delimit the dependent clause). As mentioned above, the su‰x -nin operates both as a durative aspect marker on matrix predicates, (21), and as a subordinator.15 (51)BIA a1-makaudyaran Ø2 [dyahian-nin Ø2 ama] 3sg-stand_astride stand_up-dep goal ‘He stood over her to have her stand up.’ An obligatory patient pivot shows up in control constructions.16 Compare (52)–(53) for alignment between patient and unique argument (square brackets delimitate the predicate phrases). (52)ITQ [[i-ti-nin¼] wu] idi:k 1sg-kill-dep ¼ want 2sg ‘I want to kill you.’ (53)ITQ [[donman-nin¼] wu] adu go_fishing-dep¼ want 1sg ‘I want to go fishing.’ A few comments are in order. In this kind of construction we have a main finite verb, here wu, ‘want’, heading a clause whose internal argument is a clause complement containing the subordinate non-finite lexical verb, here bivalent ti, ‘kill’ and monovalent donman, ‘go fishing’, marked

15. I assume a diachronic link between the two functions. Plausibly this was originally a nominalizer, a function which survives in subordination but was reanalysed as aspect on main verbs. 16. A possible alternative interpretation (raising) is suggested in Queixalo´s (2010). Since in terms of hierarchy the conclusion remains untouched – patient privileged – I will not go into details here.

The Katukina-Kanamari antipassive

241

for dependency by -nin, which phonologically procliticizes to the syntactically main verb wu – the ‘‘auxiliary’’.17 The crucial point is that the external argument of the bivalent nonfinite complement clause is either the patient, ‘you’ in (52), or the unique in its monovalent counterpart, ‘I’ in (53). (It should also be noted that the external argument of the main verb is coreferential with the external argument of the complement clause.18) I label as ergative all the two-place clauses seen so far. They feature the formal properties induced by basic active transitive predicates. However, an alternative two-place clause type exists which is accusatively aligned, as far as I can tell at present, in all the aspects enumerated in the previous sections (see 3.2.10 for details). 2.5. Subject and object What we have, then, is a morphosyntactic configuration where, in bivalent clauses, all the formal properties currently attributed to the notion of grammatical pivot converge almost perfectly on the patient argument of the bivalent clause, as they do on the unique argument of the monovalent clause. I wish, however, to make two caveats on aspects of the evidence adduced here that could weaken my interpretation. The first of these concerns the value of nominalization in identifying the alignment patterns of a language. A particular set of so-called ergative alignments attested cross-linguistically should deserve a special status in typology, since they appear 1) in peripheral regions of grammatical systems – ‘peripheral’ meaning non-basic alignments and non-basic clause types; and 2) in languages of all kinds, be they ergative, accusative, active or other. Argument alignments in nominalizations are among the most 17. Phonologically, these utterances are organized as /iti##ninwu##idi:k/ and /donman##ninwu##adu/. Like the case marker -na¼, the dependencemarking su‰x -nin has undergone a diachronic process of head attraction, leaving its grammatical host and binding phonologically to the phrase syntactic head that immediately follows it. Consistently with the stand taken in footnote 4, and even at the cost of introducing some diachrony in a synchronic account, I let considerations of grammatical structure override those of phonological structure in presenting the examples. 18. Which means that, for the construction with a bivalent complement clause at least, this argument is not a semantic participant of the main verb at all. This is the point in favour of a raising reading of this structure, in spite of the atypicality of the putative raising verb.

242

Francesc Queixalo´s

notorious exponents of what I call ubiquitous ergativity, along with number distinctions on the arguments of simple / reduplicated verbs, ‘‘possessor’’ raising through nominal incorporation, and more (Keenan 1984). Indeed, if the facts of nominalization displayed in 2.3.8, and in 3.2.8 below, are merely instances of ubiquitous ergativity, they should not count as criterial for the establishment of the Katukina language type in terms of alignment. But what is at stake here, as will shortly become evident, is the morphology of voice as captured for nominalization purposes, which shows that bivalent verbs are clearly patient-oriented. Secondly, there is the possibility that my claim is underdetermined by the data on coreference. The language can be grossly characterized as pivotless for coreference or, more accurately, as slightly biased towards an ergative-type pivot. For this reason, coreference can only be counted as a weak indicator, agreeing with much more robust ones in pointing to ergative syntax for Katukina. However – and this is crucial to our understanding of ergativity – the fact that the only real weak zone of ergative syntax is precisely coreference shows, in my view (Queixalo´s 2010), that this kind of grammatical organization is of recent appearance in the language. We return now to the issue of grammatical relations. The original motivation for the notion of pivot in Dixon (1994) was to subsume the formal properties of subjects in a single cluster so as to let semantically-based notions like agent permeate the notion of subject. It should be clear from what precedes that this mixing of levels is in my view the key factor underlying a significant portion of the chronic misunderstandings that have weighed upon the di¤erent approaches to ergativity. If the mapping of semantic roles on to the expression of arguments were not taken into consideration, the facts adduced above would lead anyone to clearly posit the existence of a grammatical subject and a grammatical object in Katukina ergative clauses. This is exactly the stand I take,19 since the ergative clause presented so far cannot be seen as an inverse: there is no direct clause to be held as its more basic counterpart. In Katukina, then, the linguistic expression of the semantic role of patient in a basic bivalent clause displays the formal properties of subjects; if, as I believe, the grammatical relation subject is formal in nature, then the patient argument is the subject, along with the unique argument of the monovalent clause; the other, lower-ranked argument in the basic active bivalent clause, can only be an 19. At first sight this is in line with Givo´n’s (1997: 34) statement that formal properties reflect grammatical relations ‘‘more faithfully’’; my view is in fact qualitatively di¤erent in that, for me, this is not a question of more or less.

The Katukina-Kanamari antipassive

243

object, despite its correlate in terms of semantic roles, namely the agent. This, of course, runs counter to the general assumption of a radical incompatibility between the semantic role of agent and the grammatical relation of direct object (e.g. Givo´n 2001: 200). Now, the fact that grammatical relations are, synchronically, formal entities does not rule out the possibility of diachronic grammaticalization processes whereby functional motivations – mainly pragmatic, e.g. topic maintenance – can be the source of converging forces that lead a given argument to capture the set of characteristics which will make it the syntactically privileged argument of a given construction, i.e. the subject (I take up the issue of diachrony in Queixalo´s 2010).

3. Antipassive As we have seen, in the ergative clause the expression of the agent is barred from a number of properties attached to subjects, such as sisterhood with the predicate phrase (i.e. sharing its level of constituency), zero case marking, pronominalization by free forms, extraction, and so on. However, the expression of the agent can access all of these properties by means of a voice process. 3.1. Form The antipassive is built on the basis of the ergative clause by suppressing the internal argument position. Its formal properties are: e the agent prefix paradigm slot is made inaccessible to any referent by

means of an invariable morpheme wa- that blocks the agent’s morphological slot; e no noun phrase can appear within the verb phrase; e the agent surfaces as an external argument; e the patient, demoted from its external argument position, is either omitted (54), instantiated as the object of a postposition (55), or instantiated with no relational marking (either case or postposition) (56); the postposition for the demoted noun phrase in the antipassive clause is the comitative instrumental marker katu, already seen in examples (8) and (48).20 20. Only the Bia dialect allows for the oblique patient antipassive. Several indications converge on the idea that, with regard to the evolution of ergativity, this dialect may be more conservative than Kanamari.

244

Francesc Queixalo´s

See examples: (54)ITQ wa-pu adik tyo antip-eat 1pl excl ‘We eat!’ (Context: ‘We are happy in our land.’) (55)BIA wa-wu dyara tukuna anya-na¼ katu antip-want white_people Indian woman-MkCase¼ com.instr ‘Whites like Indian women.’ (56)ITQ piya wa-pu-nin barahai man antip-eat-dur wild_meat ‘Men are eating wild meat.’ As far as the verb-argument core is concerned, constituency factors do not constrain word order any longer, since the predicate phrase is devoid of any internal, dependent, noun phrase. As an external argument, the agent phrase is basically post-verbal, as in (54)–(55). But as such it can also be fronted, as in (56). If both participants are overtly present, which is rare in spontanous discourse but accepted unproblematically in elicitation, a few restrictions obtain. When instantiated as the object of a postposition, the patient phrase is post-core, as in (55). One clue to its adjunct status is the behaviour of the future particle wa. Its position in the clause is necessarily post-core and predominantly – though not obligatorily – final. Both (57) and (58) are grammatical, whereas (59) is not. (57)BIA wa-toman adu wiri katu wa antip-shoot 1sg wild_pig com.instr fut ‘I will shoot wild pigs.’ (58)BIA wa-toman adu wa wiri katu antip-shoot 1sg fut wild_pig com.instr ‘I will shoot wild pigs.’ (59)BIA *wa-toman wa adu wiri katu antip-shoot fut 1sg wild_pig com.instr Despite its adjunct status, the patient expression has a privileged status among non-core constituents, since all other adjuncts appear after it, even when headed by the same postposition -katu, (60).

The Katukina-Kanamari antipassive

245

(60)BIA wa-toman adu wa wiri katu mokawa katu antip-shoot 1sg fut wild_pig com.instr gun com.instr ‘I will shoot wild pigs with a gun.’ It is not unambiguously clear whether the instantiation of a patient by means of a bare noun phrase still bears any grammatical relation to the predicate, but the answer is presumably negative, see section 3.1. As regards word order, noun phrases are either both located preverbally, in which case the agent always precedes the patient, (61), or distributed one on either side of the verb, the agent almost always appearing before the patient, (62), with very few cases of the reverse order, (63). (61)ITQ kaina moroho idi:k wa-binik tyo toad_sp. 2sg antip-swallow excl ‘The toad sp. swallowed you!’ (62)ITQ piya wa-pu barahai man antip-eat wild_meat ‘The man eats wild meat.’ wa-pukni Kirak (63)BIA tawa manioc antip-pull_out Kirak ‘Kirak harvested manioc.’ Note that the preverbal position of the patient is compatible with an adjunct status, since postpositional phrases are accessible to fronting, as in the following basic one-place clause: (64)ITQ ityowa ityonin naki adik tyuru poss.1pl territory loc 1pl grow ‘In our territory we grew up.’ No saliency hierarchy seems to be at work here, in the sense that something of a core argument status is being conferred to the patient phrase in a more or less inverse fashion. See the following examples, where the patient is either preverbal but low on the animacy scale, (65), or first person but in postverbal position, (66). (65)ITQ Wura poako wa-buhuk a-ama21 Wura paddle antip-make 3sg-goal ‘Wura made a paddle for him.’ 21. See this postposition as purpose subordinator in (51).

246

Francesc Queixalo´s

(66)ITQ Owi wa-hoho adu Owi antip-call 1sg ‘Owi called me.’ When only one participant is granted overt expression in the antipassive, it is practically always the agent. This state of a¤airs is massively represented, whereas instances of patient noun phrases are extremely rare in texts and straightforwardly rejected in elicitation: our informant, when presented with (67), rejected it on the grounds that ‘‘Pacu fishes don’t spear’’. (67)BIA *wa-hak bamak antip-spear pacu_fish Antipassive clauses consisting of the verb alone are more frequent in texts than all antipassives with one or two lexically instantiated participants put together. Most of the time we observe that what is being reported is an activity, not an event,22 and that the agent is either a stable topic in the fragment of discourse under scrutiny, or first person,23 or both, as in this excerpt from an ode to the happiness of life in ancient times (when there was plenty of food, etc.). (68)ITQ wa-pu niama kotuda antip-eat then again ‘Then we ate again.’ To sum up, the Katukina antipassive displays a formal device that consists in preventing the verb from taking an internal argument. The morphological means of e¤ecting this is to block the agent prefix paradigm with a marker wa-. The agent migrates to the external argument position, which is also that of the single argument of monovalent verbs and the patient of bivalent verbs. Since the patient either migrates to an adjunct (sometimes obliquely marked) position or, more frequently, is deprived of linguistic expression altogether, we can safely say that we are

22. Here I wish to posit a distinction between event: conditions of existence endowed with 1) dynamicity, 2) spatio-temporal coordinates, and 3) one or more participant(s); and activity: conditions of existence serving to characterize an entity because they involve that entity in a repetitive or (more or less) exclusive manner. 23. First person as external argument frequently remains unrealized in spontaneous speech.

The Katukina-Kanamari antipassive

247

left with a typical one-place clause. However, the patient can also be instantiated through a bare noun phrase. One could think of this as occupying a non-syntactic position, since no clause type with two external arguments is otherwise attested in the language.24 It might be analysed as something of an afterthought element, were it not for examples like (61), recalled here: (69)ITQ kaina moroho idi:k wa-binik tyo toad_sp. 2sg antip-swallow excl ‘The toad sp. swallowed you up!’ We could perhaps be led to admit the possibility that in some specific cases, the Katukina antipassive retains the non-promoted argument. After all, if the antipassive in a syntactically ergative language is overwhelmingly devoted to the function of lending pivothood to the agent, the natural output of the voice process ought to be one that retrieves both arguments, since the action depicted by the verb has undergone no change as to the number and identity of its central participants. I return to this issue below. 3.2. Motivations Obviously, such a large number of restrictions imposed upon the expression of the ergative clause agent – as detailed in sections 2.3 and, to a lesser extent, 2.4 – necessarily entails, as a direct e¤ect, the existence of a voice device whereby these restrictions are circumvented. Because this language is almost homogeneously ergative, the antipassive has mainly formal motivations, and its functional ones are somewhat di‰cult to detect. Here I will take one by one the processes enumerated in sections 2.3 and 2.4 as being barred for access to the ergative internal argument, in order to show how they apply to the expression of an antipassive agent. 3.2.1. Movement An antipassive agent can be moved to clause-initial position, as in (70)ITQ ikik wa-pu-nin barahai one antip-eat-dur wild_meat ‘Only one is eating wild meat.’

24. With the exception of ‘say’ clauses: see section 1.

248

Francesc Queixalo´s

3.2.2. Elision An antipassive agent can be elided, if recoverability of reference is granted. (71)ITQ wa-o-nin antip-drink-dur ‘(She) was drinking.’ 3.2.3. Ostension An antipassive agent can be modified, as well as pronominalized, by a demonstrative. Respectively: kawahiri wa-duni tyon (72)ITQ itiyan dem.prox cat antip-catch rat ‘This cat caught the rat.’ (73)ITQ itiyan wa-duni tyon dem.prox antip-catch rat ‘This one caught the rat.’ 3.2.4. Coordination Two nominals referring to two participants in the same agent role can be coordinated if expressed as the external argument of an antipassive predicate. (74)ITQ Nodia Hanani wa-hoho-nin Owi Nodia Hanani antip-call-dur Owi ‘Nodia and Hanani were calling Owi.’ 3.2.5. Focalization An antipassive agent can be focused. (75)ITQ Aro kana wa-nuhuk a-batyawa kariwa-na¼ ton Aro foc antip-give 3sg-wife non_Indian-MkCase rec ‘It was Aro who gave his wife to the white man.’ 3.2.6. Constituent questions An antipassive agent can be questioned.25 25. Recall the slightly di¤erent form taken by the interrogative pronoun in the two dialects (hanian / hanin), as well as the interrogative particle (tu/yu).

The Katukina-Kanamari antipassive

249

(76)BIA hanin tan wa-dyuman tahi yu? who(m) here antip-spill water int ‘Who spilled the water here?’ (77)ITQ hanian tu kana Pawi wa-toman tyo? who(m) int foc Pawi antip-shoot excl ‘Who killed Pawi?’ 3.2.7. Relativization An antipassive agent can be relativized. Compare (78) to the relativization of the patient in an active clause, (41), renumbered here as (79). (78)ITQ i-hi:k nyan piya wa-dahudyi-nin Hiowai 1sg-know deic man antip-bring-dep Hiowai ‘I know the man who brought Hiowai.’ (79)ITQ i-hi:k nyan tukuna Kontan-na¼ dahudyi-nin 1sg-know deic person Kontan-MkCase¼ bring-dep ‘I know the person that Kontan brought.’ 3.2.8. Nominalization A bivalent verb (a-examples below) cannot undergo agent nominalization unless it is recast in the antipassive (b-examples): (80)aITQ a-bi:wik-nin obakon 3sg-smoke-dur cigar ‘He is smoking the cigar.’ bITQ i-toman wa-bi:wik nyan 1sg-shoot antip-smoke deic ‘I shot the smoker.’ (81)aBIA a-hak bamak 3sg-spear pacu_fish ‘He speared a pacu fish.’ bITQ ki:tan wa-hak nyan sleep antip-spear deic ‘The spearer slept.’

250

Francesc Queixalo´s

3.2.9. Control Briefly, in active ergative clauses no straightforward coreference pivot is observed except for proper control structures, where patient (and unique) pivots are mandatory (see 2.4 above in fine). Otherwise, either patient or agent can establish distant coreference links. But a bias towards the patient can be observed in several areas, such as adverbial phrases (possession on object of postposition, semantic scope of manner and location adverbs) and interclausal relations (subordination, coordination). (82) is a sequence of clauses where the antipassive is required to license the inclusion of a bivalent verb agent and a monovalent verb participant in a coreference pivot. In elicitation, when faced with strictly symmetrical extralinguistic situations leading to potential ambiguity such as (83), the speaker spontaneously antipassivizes the verb to make clear the involvement of its agent in a coreference pivot. (82)ITQ padyi, wa-pu niama, koniohin niama arrive antip-eat then dance then ‘They arrived, then they ate, and then they danced.’ (83)ITQ Nodia-na¼ toman Yowai a-tohi:k-nin annin Nodia-MkCase ¼ shoot Yowai antip-stare_at-dep emph26 ‘Nodia shot Yowai while the latter was staring at something.’ We have listed no fewer than eight purely syntactic situations where the antipassive is required to allow accessibility to the agent of a bivalent verb. None of these applies to the genitive noun phrase, despite the fact that its coding and constituency properties are identical to those of the agent of an ergative clause. For instance, modifying or pronominalizing a noun by means of a demonstrative, (84)–(85), or questioning a referent, (86), are plainly admitted for a genitive noun. This proves that the internal arguments of verb phrases and of noun phrases are distinct syntactic elements.27 (84)ITQ daan niama itiyan ityaro-na¼ tyo go;walk then dem woman-MkCase ¼ daughter ‘Then the daughter of this woman went away.’ 26. The basic meaning of tohi:k is simply ‘look at’. It is still unclear what particular kind of pragmatic emphasis this form – an-nin, copula-dependence – conveys. 27. Or, to put it di¤erently, that noun and verb phrases are not alike in syntactic status.

The Katukina-Kanamari antipassive

251

(85)ITQ itiyan-na¼ tyo kana tona-nin tyo dem-MkCase ¼ daughter foc leave-dur excl ‘It is this one’s daughter that is leaving.’ (86)ITQ hanian-na¼ okpu tu annin? who(m)/what-MkCase¼ son int emph ‘Whose son is this?’ Concerning the focus example (75), notice that although the motivation for focusing is in itself of a pragmatic nature, its syntactic consequences – the fronting of the noun phrase and the postposing of the focus particle (ka)na – are the direct formal reason for resorting to the antipassive when the process applies to the internal argument of the active verb phrase. We turn now to the functional motivations for the antipassive. 3.2.10. Semantics & pragmatics A few common semantic or pragmatic constraints inducing use of the antipassive cross-linguistically are taken over in Katukina-Kanamari by other, and diverse, formal devices. One possible function for the antipassive is the pragmatic promotion of the agent. Specifically in Katukina, this promotion is tantamount to contrastive focus. We know that this kind of pragmatic process is achieved by means of the particle (ka)na on external arguments. Section 2.3.5 shows examples (32)–(33) for unique and patient arguments, and section 3.2.5 shows example (75) for the antipassive agent. However, the simple fronting of an external argument (remember that the canonical position for an external argument is post-verbal, cf. section 2.2), hence of the antipassive agent, has something of an attenuated contrastive focus e¤ect. Informants insist that the best equivalent for (87) is not the plain ‘‘my wife cooked wild meat’’ but something like ‘‘it’s my wife that cooked wild meat’’. A fine spontaneous example of this e¤ect is seen in (61), despite the more neutral translation given above. Plausibly more than a single degree of focus may be available to speakers. (87)ITQ yo-obtayawa wa-wahak bara 1sg-wife antip-cook wild_meat Among the morphosyntactic devices triggered by the properties of the patient we find the accusative pattern referred to in section 2.4. The language features a transitive split whereby an accusatively aligned construction is employed when the patient is semantically generic, (88)–(89). Iterative

252

Francesc Queixalo´s

or habitual aspect is not necessarily involved, as the first clause of (90) shows. The accusative clause is a perfect inversion of the ergative clause in terms of constituency: it has the patient as its internal preverbal argument, and the agent as its external and typically postverbal argument. The patient is obligatorily instantiated through a noun phrase since no person prefix attaches to the verb. The formal properties of both arguments are more or less identical for ergative and accusative patterns, as long as we express these properties in terms of internal / external noun phrases. They are, of course, inverted in terms of semantic roles. (88)ITQ [wiri hak] adu wild_pig spear 1sg ‘I speared wild pigs.’ (89)ITQ [takara duni ] mapiri dawa hen catch anaconda today ‘The anaconda is catching hens today.’ (90)ITQ [mokawa wu] adu [wiri toman] niama gun want 1sg wild_pig shoot purp 28 ‘I want a gun to shoot wild pigs.’ While the language allows for noun incorporation, the accusative pattern cannot be analysed in those terms since monovalent (i.e. ‘‘alienable’’) nouns may only be incorporated if the verb also undergoes an applicative process (see in Queixalo´s 2008 the notion of redistributive incorporation). Compare (91), with a bivalent noun incorporated, to (92)–(93) with a monovalent noun. (91)ITQ nyama-na¼ ki-onyuk a-okpu mother-MkCase¼ head-scratch 3sg-son ‘The mother combed her son.’ bara-wahak (92)ITQ *yo-obtayawa-na¼ 1sg-wife-MkCase ¼ wild_meat-cook ‘My wife cooked wild meat.’ 28. The grammatical status of niama is unclear here. This particle serves as the discourse connector ‘then’, and also as a purpose subordinator, as it seems to do here in spite of the lack of the verbal dependence su‰x -nin in the example. A more accurate translation might perhaps be ‘I want a gun. Then I’ll shoot wild pigs’, but the problem in that case would be the absence of the future particle wa.

The Katukina-Kanamari antipassive

253

(93)ITQ yo-obtayawa-na¼ ma-bara-wahak 1sg-wife-MkCase¼ appl-wild_meat-cook ‘My wife cooked wild meat for (him).’ However, there seems to be some overlap in the semantics of the patient between antipassive and accusative clauses: no di¤erential quantification feature has so far been identified between the patients in (88) and (95). Nor can the referential status of the patient be held as distinctive between the two patterns. Compare, for the accusative clause, (88) with a referential patient to (90) with two non-referential patients, and for the antipassive clause, (94) with a referential patient to (95) with a nonreferential patient. (94)BIA hanin koya wa-buhuk? who(m) pap antip-make ‘Who made the pap?’ (95)ITQ adu don wa-buhuk-nin¼ bak 1sg fish antip-make-dep ¼ be_good ‘I am a good fisherman (lit. I am good at making fish).’ Another common functional feature associated with antipassive patients is indefiniteness. This language shows no incompatibility between an indefinite patient and the ergative clause, and the simplest means of packaging an indefinite patient is to let a zero pronoun (‘‘pro’’, see 2.3.1) fill the external argument slot of the ergative clause provided that no plausible referent is available in the situational or discourse environment. This will automatically bring about an indefinite reading. The following examples show a piece of information containing an indefinite patient with no overt expression, (96), followed by a quite natural question, (97). A single noun phrase preceded by the indefinite prefix a- is no less natural an answer. This prefix allows the indefinite patient to be represented as an explicit noun phrase, (98). (96)ITQ oman-na¼ ti na tyo tree-MkCase ¼ kill foc excl ‘The tree killed someone!’ ti tu na oman-na¼ ti tyo? (97)ITQ hanian who(m)/what restr int foc tree-MkCase¼ kill excl ‘Whom did the tree kill, precisely?’

254

Francesc Queixalo´s

(98)ITQ oman-na¼ ti a-tukuna tree-MkCase ¼ kill indef-Indian ‘The tree killed someone [who is an Indian].’ We have come across several instances of ergative clauses in which the agent outranks the patient in the saliency hierarchy, e.g. (1), (6), (14), (47), to mention but a few. Now, not only can the same pattern obtain in antipassive clauses, as in (56), but the reverse pattern (in which the patient is more salient than the agent) does not motivate the use of an antipassive, as (99) shows. Nor do semantic hierarchies in the accusative clause favour an inverse interpretation, although generic patients tend per se to be less salient. (100) shows a human patient alongside a non-human agent, to be compared with (101) where, in the same accusative pattern, a non-human ‘‘patient’’ faces a human ‘‘agent’’.29 Hence, neither antipassive nor accusative clauses can be viewed as forming part of a direct/inverse system. (99)ITQ dyoko-na¼ hak-dyi adu tyo dart-MkCase¼ perforate-centrip 1sg excl ‘The dart perforated me.’ (100)BIA pi:da ityaro botyana jaguar woman follow ‘ jaguar>s< follow women.’30 (101)BIA pi:da ohiya Ayobi jaguar fear Ayobi ‘Ayobi fears jaguar(s).’ Now, if we assume that strictly formal motivations should lead antipassive clauses to retain overt expression of the patient, whether bare or oblique, the very existence of antipassive clauses with covert patients is a clear clue to the possibility of concomitant functional motivations for voice alternation in this language. We have come across a typical instance of such a functional antipassive above: description of an activity (vs. an event), in (68). An additional example is (102). Close to this function is

29. Quotation marks are for the non-prototypical agent and patient required by this particular verb. 30. The ‘inverted angle brackets’ notation is used for obligatory disjunctive occurrence: y stands for xy, yz, *y, *xyz, that is: either x or z must occur, but not both.

The Katukina-Kanamari antipassive

255

the description of an ability (or inability), as in (103).31 What both types have in common is that the patient is not interesting in its own right: its identity is insignificant, irrelevant, obvious. (102)ITQ opatyin hinuk niama wa-o ti children group then antip-drink restr ‘Then the children just drank.’ (103)BIA wa-hak tu (adu) antip-spear neg 1sg ‘I’m a bad fisherman (lit.: I don’t spear).’ 4. Conclusion To sum up, the formal motivations for the antipassive in Katukina are clear. This does not rule out the possibility that functional motivations exist alongside them. The latter, however, partially overlap and compete with those of other morphosyntactic devices catering to the speaker’s intention of semantically/pragmatically promoting an agent or demoting a patient. The residue of antipassives whose motivations are not formal or whose functions are not also taken over by other grammatical devices consists, at the present state of our knowledge, of the description of two semantically contiguous configurations: activities and agentive abilities. The overwhelming importance of formal motivations in voice selection settles, in my view, the issue of the grammatical relation hierarchy in this language, and confirms the direct objecthood of an argument – the agent – which is in the marked case, internal to the verb phrase, and promotable to syntactic subject status by means of recessive voice change. In sum, the formal side of morphosyntax in a syntactically ergative language is much the same as that observed in many accusative languages. At first sight, a substantial qualification on this similarity is the ever-present feature of split transitivity which is so characteristic of ergativity. But note that accusative languages have their own areas of split transitivity: besides the existence of ‘ubiquitous ergativity’ as mentioned above, we find di¤erential marking of objects and even di¤erential marking of subjects. The crucial specificity of syntactically ergative languages is not, then, their 31. These aspects are akin to imperfectivity, a common feature of antipassives. Imperfectivity plausibly accounts for the observed slightly higher frequency of the durative (-nin) in antipassive clauses. I was led to verify this point after a remark by K. Haude.

256

Francesc Queixalo´s

formal mechanism per se, which is in itself rather common whatever the basic alignment type, but merely the mapping of semantic roles on to grammatical relations (Mel’cˇuk 1979, Marantz 1984). Our current ideas on this topic are certainly valid generalizations for 99.9 per cent of the documented languages of the world, but the tiny 0.1 per cent remainder means that they cannot be treated as strict defining features of human language. Abbreviations 1 2 3 all antip appl centrip com deic dem dep dur excl foc fut goal grn indef instr int loc MkCase neg pl poss prospect prox purp rec restr sg

first person second person third person allative antipassive applicative centripetal comitative deictic demonstrative dependent marker durative exclamatory marker focus future goal generic relational noun indefinite instrumental interrogation locative marked case negation plural possessive prospective proximal purposive recipient restrictive singular

The Katukina-Kanamari antipassive

257

References Adelaar, Willem F. H. 2000 ‘‘Propuesta de un nuevo vı´nculo gene´tico entre dos grupos linguı´sticos indı´genas de la Amazonı´a occidental: Harakmbut y Katukina.’’ In Actas del I Congreso de Lenguas Indı´genas de Surame´rica 219–236. Universidad Ricardo Palma. Comrie, Bernard 2008 ‘‘What is a passive?’’ In Studies in Voice and Transitivity, Zarina Estrada, Søren Wichmann, Claudine Chamoreau and Albert Alvarez (eds.), 1–18. Mu¨nchen: Lincom Europa. Cooreman, Ann 1994 ‘‘A functional typology of antipassives.’’ In Voice: Form and Function, Barbara Fox and Paul J. Hopper (eds.), 49–88. Amsterdam/ Philadelphia: John Benjamins. Dixon, R.M.W. 1994 Ergativity. Cambridge: Cambridge University Press. Givo´n, T. 1997 ‘‘Introduction.’’ In Grammatical Relations: a Functionalist Perspective, T. Givo´n (ed.), 1–84. Amsterdam/Philadelphia: John Benjamins. Givo´n, T. 2001 Syntax: an Introduction. Vol. 1. Amsterdam/Philadelphia: John Benjamins. Givo´n, T. 2008 ‘‘On the relational properties of passive clauses: A diachronic perspective.’’ In Studies in Voice and Transitivity, Zarina Estrada, Søren Wichmann, Claudine Chamoreau and Albert Alvarez (eds.), 19–32. Mu¨nchen: Lincom Europa. Givo´n, T. 2009 The Genesis of Syntactic Complexity. Amsterdam/Philadelphia: John Benjamins. Keenan, E. 1984 ‘‘Semantic correlates of the ergative/absolutive distinction.’’ Linguistics 22: 197–223. Marantz, A. 1984 On the Nature of Grammatical Relations. Cambridge (Mass.): MIT Press. Mel’cˇuk, I. 1979 Studies in dependency syntax. Ann Arbor: Karoma. Queixalo´s, F. 2004 ‘‘Split transitivity and coreference in Katukina.’’ In Ergatividade na Amazoˆnia III, Atas do terceiro encontro do projeto Manifestac¸o˜es da ergatividade na Amazoˆnia, Queixalo´s, F. (ed.), 175–188. Paris: CELIA/CNRS.

258

Francesc Queixalo´s

Queixalo´s, F. 2008 Queixalo´s, F. 2010

‘‘Incorporation nominale en sikuani et en katukina-kanamari.’’ Amerindia 31: 61–86. ‘‘Grammatical relations in Katukina-Kanamari.’’ In Ergativity in Amazonia, Spike Gildea and Francesc Queixalo´s (eds.), 237– 285. Amsterdam/Philadelphia: John Benjamins.

Undergoer orientation in Movima Katharina Haude 1. Introduction1 This article explores the system of verbal voice morphemes in Movima (unclassified, Amazonian Bolivia) and seeks to explain why most transitive main clauses in Movima pattern ergatively. Movima has two basic transitive constructions, direct and inverse, overtly distinguished by verbal morphemes. In main clauses, to which the discussion in this paper is restricted, the direct construction patterns ergatively and the inverse construction patterns accusatively. In terms of statistical frequency, the large majority of transitive main clauses in texts is direct, i.e. ergative. The direct and inverse markers, which are employed according to the relative position of the participants on a referential hierarchy (Haude 2009b, 2010), belong to a paradigm of verbal morphemes (reflexive/reciprocal, resultative, agentive, middle) that indicate the transitivity of the verb and the participant (macro)role – actor or undergoer – of its subject. They can only be applied productively to one class of verbal bases, which in their majority denote two-participant events. When a verb of this class occurs without an overt voice marker, it is syntactically intransitive and denotes a state, which means that its subject has the undergoer (theme) role. Furthermore, resultative verbs, which also take an undergoer (patient) as subject, can be identified as being morphologically the least marked of all voicemarked verbs. Unmarked verbs of this class can therefore be considered 1. The data on which the study is based were collected in Santa Ana del Yacuma between 2001 and 2009, financed by the Spinoza programme Lexicon and Syntax (Radboud University Nijmegen) and by the DobeS programme of the Volkswagen Foundation. I am deeply grateful to the Movima speakers who shared their knowledge with me. Spike Gildea, Francesc Queixalo´s, and an anonymous reviewer are thanked for their critical remarks on an earlier version of this paper, without being responsible for any shortcomings. Elicited examples are marked with [el], all other examples stem from recorded spontaneous discourse.

260

Katharina Haude

undergoer oriented. The proposal of this paper is to consider the direct voice marker as a morpheme that derives a transitive verb by simply adding a syntactic position for an actor argument, leaving the undergoer orientation of the verb untouched and thus creating an ergative structure. The inverse marker, under this view, is a secondary derivation, reversing the participant roles of the arguments of a transitive clause according to the referential hierarchy and discourse status. The paper is structured as follows. Section 2 presents the syntactic properties of transitive and intransitive predicates and their arguments: section 2.1 illustrates the distinction between transitive and intransitive predicates, the encoding of the arguments of transitive clauses and the function of direct and inverse marking; section 2.2 demonstrates that the argument that represents the undergoer in a direct transitive clause has the same formal and behavioural properties as the S argument of the intransitive clause, leading to an ergative pattern, whose status is further discussed in 2.3. Section 3 describes the verbal voice morphemes. Section 4 gives an overview of the verbal bases that can (4.1) and cannot (4.2) participate in the voice system, showing that the members of the first class typically denote two-participant events and the members of the second class one-participant events. Section 5 argues that bivalent verbs are undergoeroriented and that this may contribute to the default ergativity of transitive clauses. The conclusions are drawn in section 6.

2. Clause structure and ergativity 2.1. Transitive and intransitive clauses There are two classes of predicates in Movima: transitive and intransitive. They can be identified by their ability to be combined with two overt argument expressions. A transitive predicate can take two overt argument expressions, as in (1). (1) tikoy-na¼us os rulrul kill-dr ¼3m.ab art.n.pst jaguar ‘He killed the jaguar.’ An intransitive clause may only contain one argument expression; any other event participant can only be expressed by an adjunct, marked by the oblique prefix. Example (2) shows that the verb kaykay ‘eat’, despite

Undergoer orientation in Movima

261

its meaning, is intransitive: the eaten object (‘those nuts’) can only be expressed by an adjunct. (2) kayPkay is karak ni-kis ney choPcho´-kwa mdPeat art.pl macaw obl-art.pl.ab def redPnut-abs ‘The macaws eat those nuts.’ The two core arguments in a transitive clause, identified by the fact that they do not carry the oblique marker, are distinguished from each other by their constituency properties: one is expressed by a constituent internal to the predicate phrase, the other by a constituent external to it, as illustrated in (3). These properties are reflected by the following features: the internal nominal constituent (¼us in (1)) is inseparably attached to the predicate by ‘‘internal cliticization’’ (causing stress shift; marked as ¼) and is obligatorily expressed. The external nominal constituent (os rulrul in (1)) is attached through ‘‘external cliticization’’ (when bound pronoun; leading to resyllabification but no stress shift; marked as - -) or not phonologically attached at all (when free pronoun or NP); other elements can occur between it and the predicate phrase; it is not obligatorily expressed, and it is easily replaced by a free pronoun in clause-initial position (for more details see Haude 2006, 2010). (3) [PRED¼ARG] [ARG] Whether an argument is represented by the internal or the external constituent depends primarily on its referential properties (see Haude 2009b, 2010). The expression of speech-acts participants (except second person plural; see Haude 2011) is restricted to the internal position. When two third persons interact, the more topical one is represented by the internal constituent; typically, the internal constituent is a pronoun and the external one a noun phrase, as in the above examples. Constituency alone does thus not indicate the participant roles (actor or undergoer) of the arguments.2 These are indicated by verbal morphemes. 2. The terms ‘‘actor’’ and ‘‘undergoer’’ (the ‘‘macroroles’’ of Role and Reference Grammer; Foley and Van Valin 1984; Van Valin and LaPolla 1997; Van Valin 2005) instead of ‘‘agent’’ and ‘‘patient’’ are chosen for two reasons: firstly because a Movima clause can maximally contain two core arguments, which can represent a large range of semantic roles, including recipients; and secondly because the undergoer role encompasses the semantic roles patient and theme (Van Valin and LaPolla 1997: 140–141; Van Valin 2005: 52), which is relevant for the point made in the present paper, i.e., the similar orientation of stative and resultative verbs.

262

Katharina Haude

When the internal constituent is the actor and the external constituent the undergoer, the verb is marked as direct; when the situation is reversed, the verb is marked as inverse, as shown in (4). (4) tinok-poj-kay-a¼us os merek rulrul scare-caus-inv-ep ¼3m.ab art.n.pst big jaguar ‘The big jaguar scared him.’ The fact that the arguments are primarily encoded according to their referential properties and not according to participant roles, makes it di‰cult to provide them with uncontroversial labels. Following Bickel (2010), I use terms that are based on the referential properties of the arguments, labelling the argument internal to the predicate phrase (high-ranking) as ‘‘proximate’’ and the argument external to the predicate phrase as ‘‘obviative’’ (short PROX and OBV, respectively). As we will see in the following section, the OBV argument has the syntactically privileged status. 2.2. The syntactic subject The single argument of an intransitive clause (S) has the same morphosyntactic properties as OBV of a transitive clause: it is phonologically independent, cliticized only when bound pronoun, not obligatorily expressed, and can easily be replaced by a free pronoun. Moreover, on the syntactic level, OBV and S have access to syntactic operations to which PROX does not have access. The clearest case is relativization. A relative clause is introduced by the particle di’ following the relativized noun phrase, which is not expressed again in the relative clause. Example (5) shows an intransitive relative clause. (5) kinos ney ay’ku [di’ jayna kayni ] art.f.ab here aunt rel dsc die ‘That aunt of mine who has died already.’ Examples (6) and (7) illustrate the relativization of OBV. In (6), the relativized argument is the undergoer, therefore the predicate of the relative clause is marked as direct; in (7), the relativized argument is the actor, therefore the predicate is marked as inverse.

Undergoer orientation in Movima

263

(6) kinos alwaj-a¼us [di’ naye-´e-na¼us] art.f.ab spouse-ep ¼3m.ab rel marry-co-dr ¼3m.ab ‘his wife, whom he had married’ (7) is pa:ko [di’ lap-kay-a¼as] art.pl dog rel bite-inv-ep ¼3n.ab ‘(the) dogs who bit it [the jaguar]’ In order to relativize the participant represented as PROX, a detransitivizing voice operation has to be applied. Here, the predicate is preceded by a particle kwey (or kaw), the clause is intransitive with the original PROX as its single argument; the original OBV is (optionally) expressed by an adjunct, marked by the oblique prefix. This process is generally found with direct predicates, where it has an antipassive e¤ect. It is illustrated in (8b), which contrasts with (8a), the transitive construction. (8) a.

jayna rey way-na¼is kinos rey sonsa dsc mod lift-dr ¼3pl.ab art.f.ab mod silly ‘Then they had already taken up that silly (woman).’

b.

is juyeni [di’ jayna kwey way-na n-i’ne] art.pl person rel dsc detr lift-dr obl-pro.3f ‘the people who had taken her up’

While the syntactically privileged status of OBV as opposed to PROX in relative clauses is clear, there is no evidence of any construction to which PROX, but not OBV, has access (Haude 2009b). Therefore, in the remainder of this paper and in line with other contributions in this volume (Monod-Becquelin and Becquey; Queixalo´s), I use the term ‘‘subject’’ to refer to the grammatical relation encompassing {OBV,S}. While based here entirely on syntactic grounds, the term seems suited best for making the patterns in Movima comparable to voice patterns in other languages, since discussions of voice phenomena generally make use of the subject notion (see e.g. Mithun 1994; Kulikov 2010). 2.3. Ergativity Given that OBV aligns with S, it is obvious that when the participant roles of the arguments are considered, the direct/inverse alternation results in

264

Katharina Haude

two basic transitive clause types, one with an ergative and the other with an accusative pattern. In the direct construction, OBV represents the undergoer, and in the inverse construction, OBV represents the actor. With respect to formal marking, the split is entirely parallel: whenever PROX is the actor and OBV the undergoer, the pattern is ergative; whenever OBV is the actor and PROX the undergoer, the pattern is accusative. In both cases, the predicate is overtly morphologically marked. However, with respect to discourse frequency and pragmatics, there is evidence that direct/ergative construction is the default for transitive clauses (see also Haude 2010). First of all, the direct construction is statistically more frequent, occurring in about 80% of the transitive clauses with two third-person arguments. This is not surprising, since a high-ranking actor and a low-ranking undergoer represent the typical participant constellation in a two-participant event (see DeLancey 1981; Givo´n 1994; Croft 2003). Evidence for the pragmatically unmarked character of the direct construction is also apparent from the fact that this construction is occasionally found in opposition to the referential hierarchy, while this is never the case with the inverse construction. When two third-person arguments are both represented by pronouns or by full noun phrases, reflecting an equal discourse status, then automatically the direct construction is used. This is the case in elicitation, and it is illustrated by the text example in (9), where both arguments are encoded as noun phrases: even though the actor is an animal and the undergoer a human, the construction is direct and the actor is represented as PROX. The text corpus contains no example of the reversed case, i.e. the inverse construction with the referentially lowerranking participant represented as PROX. In elicitation, the inverse construction in such a context is accepted, but never proposed spontaneously. (9) jayna lap-na¼as mimi:di us majni! dsc bite-dr¼art.n snake art.m my_o¤spring ‘Now the/a snake bit my son!’ The inverse construction, furthermore, has only limited access to the detransitivising voice operation with kwey. A kwey-construction with an inverse predicate occurs nowhere in the text corpus. In elicitation, certain examples with a kwey-construction and an inverse predicate, like the one in (10b), tend to be rejected; the exact conditions of this still need to be investigated, but in any case, the limited acceptability of such examples

Undergoer orientation in Movima

265

means that the use of an inverse predicate in the kwey-constructon is problematic. (10) a.

in´a kwey dul-na n-isne pro.1sg detr visit-dr obl-pro.3f.ab ‘It was me who visited her.’

b.? in´a kwey dul-kay n-isne pro.1sg detr visit-dr obl-pro.3f.ab (‘It was me who was visited by her.’) I conclude that while there is no di¤erence in morphological markedness between the inverse and the direct construction, the direct construction is the default transitive construction, used when two third-person participants are ranked equally in terms of discourse status. The inverse construction, in contrast, is restricted to the situation in which the undergoer outranks the actor with regard to person and discourse prominence. Furthermore, the detransitivising operation with kwey is largely restricted to the direct construction. The default transitive a‰rmative main clause in Movima, therefore, has an ergative pattern.

3. Voice markers The direct and inverse markers belong to a paradigm of verbal morphemes that indicate the transitivity of a verb and the participant role(s) of the core argument(s). They are listed in table 1. Table 1. Movima voice markers (Ac ¼ actor, Ug ¼ undergoer) transitivity

marker

meaning

subject role

transitive

-na/-a-kay

direct inverse

Ug Ac

intransitive

-che´

reflexive/reciprocal

ActþUg

-’i

resultative

Ug

-e´e

agentive middle

Act ActþUg

266

Katharina Haude

A first illustration of the e¤ect of the voice markers is provided in (11) with the transitive root jat- ‘hit’. (The reduplicative middle marker is not illustrated here because it is not fully productive and does not occur with this root.) (11) direct: inverse: reflexive: resultative: agentive:

jat-na¼Ø jat-kay¼Ø jat-che´ jat-’i jat-e´e

‘I hit X’ ‘X hits me’ ‘X hit(s) themselves/each other’ ‘X has been hit’ ‘X hits (continuously/habitually)’

These morphemes can be adequately analysed as voice markers because of their property to indicate the transitivity of a verb and the participant role of the verb’s subject, which are characteristics of voice marking as defined cross-linguistically. They also serve to present an event from varying perspectives (see Shibatani 2006). They do not, however, seem to be employed for syntactic purposes (this is the role of the particle kwey, see 2.2 above), and none of them derives an intransitive verb from a transitive one, a feature present in many traditional definitions of voice (see Kulikov 2010). They have aspectual (Aktionsart) and sometimes modal connotations. In the following discussion, however, I will focus on their voicemarking property, i.e. the way in which they assign arguments to the verb. 3.1. Transitive voice markers: direct and inverse As was shown in 2.1 above, transitive predicates contain either the direct or the inverse morpheme. The direct morpheme has two allomorphs: basefinal -na and base-internal -a-, which are both illustrated in (12). The baseinternal allomorph -a- is applied like an infix: it is inserted in morphologically complex verbal bases immediately after the root, as in (12b), provided that the root be a single closed syllable not followed by an aspectual su‰x (e.g. -ka ‘mlt’). Being based on prosodic properties of the base, the insertion of the direct marker -a- can create non-linear morphological patterns. For example, in (12b) the direct marker precedes the causative su‰x (-poj ) in linear order, but it is applied at a later stage of the verbal derivational process, after the causative su‰x has derived the bivalent base. (12) a. b.

lat-na¼is kis ko’o, chop-dr ¼3pl.ab art.pl.ab tree ´ok-a-poj-a¼is ba:ra kis ko’o fall-dr-caus-ep ¼3pl.ab all art.pl.ab tree ‘They chop down the trees, they fell all the trees.’

Undergoer orientation in Movima

267

In all other environments, the su‰x -na is applied, as illustrated in (12a) above with a simple root and in (13) with a complex base with an opensyllable root. (13) jayna chi-poj-na¼is us majni dsc go_out-caus-dr ¼3pl.ab art.m my_o¤spring ‘They took my child out (at birth).’ Some bivalent bases seem to be historically complex, since they take the direct marker -a- instead of -na, but synchronically, their components cannot be properly identified. An example is given in (14), with the verb base kay´e ‘give’, whose final syllable is most probably the applicative su‰x -´e, but whose first syllable kay (homophonous with the root kay‘eat’) cannot be identified as a root on the synchronic level. (14) kay´e¼Ø n-i’ko jayna give ¼1sg obl-pro.3pl dsc ‘I’ll give it (the money) to you.’ The inverse marker invariably consists of the su‰x -kay (presumably unrelated to the verb root ‘eat’ or ‘give’) independently of whether the verb takes the su‰x -na ((15), cf. (13)) or its allomorph -a- ((16), cf. (14)) as direct marker. Since its final phoneme /j/ ( y) is a consonant, the epenthetic vowel -a is added before an internal enclitic (as in (4) and (7) above; see Haude 2006: 98–99). As was shown in 2.1, the inverse su‰x derives a transitive predicate whose subject is the actor. (15) chi-poj-kay¼Ø isnos kayni di’ nonok go_out-caus-inv ¼1sg art.f.pst die rel my_grandmother ‘My late grandmother took me out (of school).’ (16) kay´e:-kay¼Ø- -isne n-is narasa-mes-a¼sne give-inv ¼1sg- -3f.ab obl-art.pl orange-cl.fat-ep ¼3f.ab ‘She gave me her orange oil.’ In the third-person domain, the application of the direct and inverse markers depends very much on the relative topicality of the arguments, since the topical argument is generally represented as PROX. Here, the inverse assumes the function of a voice marker (cf. Gildea 1994), as is reflected by the fact that it can generally be translated as a passive, as in (17).

268

Katharina Haude

(17) us bi:jaw, jiw-a-´e-kay-a¼us art.m old come-dr-co-inv-ep ¼3m.ab us kayni di’ Koyimbra art.m die rel proper_name ‘The old (man), he was brought by the late Coimbra.’ 3.2. Intransitive voice markers The intransitive voice markers create predicates that can take only one overt core argument expression. 3.2.1. -che´ reflexive/reciprocal The reflexive/reciprocal su‰x -che´ is the most productive intransitive voice marker, combinable with any verb that can be marked as direct or inverse. It creates a verb whose subject represents one or more participant(s) that simultaneously carry out and undergo an action. The interpretation as either reflexive (18) or reciprocal (19) depends on the context. (18) didi’ tikoy-che´ n-os kachi:ra frust killed-r/r obl-art.n.pst knife ‘(He) wanted to kill himself with a knife.’ (19) ban jayna don-che´- -is but dsc dislike-r/r- -3pl.ab ‘. . . but they already disliked each other.’ In 4.2 below it will be shown that when attached to so-called monovalent bases, this su‰x is deprived of its reflexive/reciprocal meaning and only marks an activity. 3.2.2. -’i resultative The su‰x -’i marks an intransitive verb denoting a resultative state, i.e. a verb expressing ‘‘both a state and the preceding action it has resulted from’’ (Nedjalkov and Jaxontov 1988: 6). The subject of the resultative verb represents the undergoer. (20) ben-’i is chorimpa¼sne paint-res art.pl fingernail¼3f.ab ‘Her fingernails were painted.’

Undergoer orientation in Movima

269

(21) bo rey os bet’i vus-’i- -as reas mod art.n.pst grassland carbonize-res- -3n.ab che kaw-ra is ve’e:-vus and much-be.ntr art.pl fire-cl.dust ‘. . . because the grassland, it had been burnt [by the farmers] and there were lots of ashes.’ Depending on the context, the su‰x -’i can also express a deontic modality (‘X has to be done’); it is used in instructions and procedural texts: (22) che jayna n-as ba:-paj-uk-wa¼n jayna and dsc obl-art.n finish-split-nmz ¼2 dsc il-’i- -is, tamol-’i- -is daya’a spread-res- -3pl.ab expand-res- -3pl.ab dur.nstd ‘And when you finish splitting (the straws), then they are/have to be spread out, they are/have to be extended.’ (23) ra´-’i is pe:ra, il-’i, jayaw-’i, pull_out-res art.pl reed spread-res nice-res werel-ni che pil-’i no-kos do:nojbet ney dry-prc and rolled_up-res obl-art.n.ab cloth here ‘The reed is pulled out, (it) is spread, (it) is made nice, (it) dries and is rolled up with a cloth like this. . .’ In traditional voice terminology, the resultative comes closest to a passive (cf. Comrie 1981), since it creates an intransitive predicate whose subject is the undergoer. However, in di¤erence to a canonical passive (cf. e.g. Siewierska 1984; Shibatani 1985; Dixon and Aikhenvald 2000; Kazenin 2001; Keenan and Dryer 2007), there is no sign of derivation from an unmarked transitive, active verb: the base to which the resultative su‰x is attached is stative rather than active (see section 5 below), and transitive predicates are overtly morphologically derived. In fact, instead of being a derived form, the resultative even seems to be the morphologically least-marked form of a bivalent verb root (see section 5). The agent is not expressed in a clause with a resultative verb; the few cases in the corpus where an oblique phrase in a resultative clause might be interpretable as an agent are cases like (24), where an oblique phrase (nis alamre) denotes the entity by which the state was caused. In general, also in resultative clauses obliques encode peripheral roles such as locations, times or instruments, as in (23) above (nokos do:nojbet) or (25).

270

Katharina Haude

(24) bi´-’i is dinoj-a¼’ne n-is alamre scratch-res art.pl thigh-ep ¼3f obl-art.pl wire ‘Her legs were scratched from the wire.’ (25) pay’-’i n-is bubutkwa os ro:ya smear-res obl-art.pl mud art.n.pst house ‘The house was plastered with mud.’ 3.2.3. -e´e agentive The agentive marker -e´e derives an intransitive verb whose subject is the actor, as illustrated in (26). In this way, the agentive su‰x resembles an antipassive, and like many antipassives (see Cooreman 1994), it indicates a durative or habitual action. However, in the same way as the resultative marker, the agentive marker does not derive an intransitive from a transitive verb, which belongs to the definition of a typical antipassive (see Dixon 1994: 146). (26)

jayna jo’yaj, sal-e:´e- -y’´i dsc arrive search_for-agt- -1pl ‘Then (we) arrived, we searched.’

Verbs containing the su‰x -e´e are often combined with an obliquemarked noun phrase denoting the patient, as in (27) and (28). However, oblique phrases can encode many relations, which can only be inferred from the context. For example, in (27) and (28), the oblique NP is identified by the context as a patient, while in (29), the context identifies it as a location. (27)

jayna jot-e:´e n-is chekwes´a dsc gather-agt obl-art.pl taruma´ ‘Then (I) gathered taruma´ fruits.’

(28) ra´-pit-e:´e n-is kade:na tear-be.half-agt obl-art.pl chain ‘[The wild cat] tore the chain [and escaped].’ (29) sal-e:´e- -us n-is wolsiko¼us search_for-agt- -3m.ab obl-art.pl pocket¼3m.ab ‘He searched in his pockets [for bullets].’

Undergoer orientation in Movima

271

3.2.4. Middle reduplication The middle marker consists of a full reduplication of monosyllabic and a partial reduplication of disyllabic roots (see Haude 2006: 345¤.). I call it ‘‘middle’’ because reduplicated verbs generally denote events that a¤ect the subject participant in some way. Consider the di¤erence between the direct-marked verb in (30) with the middle verb in (31): (30) i´ ´ap-na¼Ø is ona:cho 1sg bathe-dr ¼1sg art.pl my_grandchild ‘I bathe my grandchildren.’ (31)

jayna ´apP´ap- -i’ne dsc mdPbathe- -3f ‘Then she bathed.’

The same di¤erence is illustrated with a longer verb root, chumay-, whose middle form is created by reduplication of the penultimate syllable. Example (32) illustrates the direct, (33) the middle form of the verb. (32) chumay-na¼n kos boPboj-a¼kos ko’ smoke-dr ¼2 art.n.ab redPbase-ep ¼art.n.ab tree ‘You smoke out the trunk of the tree [for collecting honey].’ (33) kos da’ ve’ chumay art.n.ab dur.nstd fire smoke ‘The fire is smoking.’ The middle marker is not very productive on the so-called ‘‘bivalent’’ bases (4.1); in contrast, it frequently occurs with ‘‘monovalent’’ bases, from which it derives activity verbs (see 4.2). 4. Types of verb bases Two types of verbal bases can be distinguished in Movima: bases that participate fully in the voice system and bases that do not. The defining criterion for their distinction is a formal one: on verbs of the first class, the su‰xation of -na derives a transitive predicate in the way described in 2.1, whereas on verbs of the second class, the su‰xation of -na derives a locational noun. The two classes are furthermore characterized semantically: bases of the first class denote two-participant events, while most bases of the second class denote one-participant events; bases of the first

272

Katharina Haude

class, when unmarked for voice, are stative predicates, while unmarked bases of the second class typically denote activities. The two classes are described in the following sections.3 4.1. Bivalent bases The bases that participate fully in the voice system typically denote events that imply at least two participants, including prototypical transitive events (Hopper and Thompson 1980) like ‘hit’ or ‘kill’. Therefore, I refer to the bases of this class as ‘‘bivalent’’. Bivalent bases can be either simple (i.e. roots) or complex; most complex bases can occur independently, while roots cannot. Examples of bivalent verb roots, of which there are about 150 (see Haude 2006: 555¤.), are listed in (34). (34)

jat´eklapsalyeyju:tikoyelaji:sa-

‘hit’ ‘kick’ ‘bite’ ‘search for’ ‘want’ ‘scold’ ‘kill’ ‘leave behind’ ‘make’

Verb roots cannot occur independently; they must be combined with another morpheme, e.g. a voice marker, as was illustrated in (11) above. In section 5 I will argue that the resultative voice, marked by -’i, can be considered the least marked form of bivalent roots. Bivalent bases can also be morphologically complex, i.e. consist of a root plus a derivational morpheme (like causative or applicative) or a modifying incorporated element (see Haude 2006: Ch. 9). Consider the base jam-´e in (35), which combines with most voice markers: 3. Adjectives (words like ra:pal ‘red’, merek ‘big’ and ja:yaw ‘nice’) are excluded from this discussion. They can participate in the voice system (e.g. jayaw-na ¼ Ø ‘I make X nice’ or jayaw-’i ‘be made nice’ in (23)), but can be distinguished from the verbal bases in that they can be combined with the verbalizing su‰xes -ni ‘to be/become X’ and -tik ‘to make/do X’, and that they can be reduplicated in subordination (see Haude 2006: 119). There are borderline cases where adjectives and stative verbs cannot be easily distinguished, and more fine-grained investigation still needs to be undertaken. However, this is not directly relevant for the present study.

Undergoer orientation in Movima

(35) a.

273

jam-a:-´e¼Ø tie-dr-co¼1sg ‘I tie X onto (sth.).’

b.

jam-´e:-kay¼Ø tie-co-inv ¼1sg ‘X tie(s) me (onto sth.).’

c.

jam-´e:-che´ tie-co-r/r ‘X tie(s) itself/each other onto (sth.).’

d.

jam-´e-’i tie-co-res ‘X has been tied onto (sth.)’

e.

jam-´e tie-co ‘X gets tied onto (sth.).’

Unlike roots, most complex bivalent bases can occur independently, in which case they denote a state. Example (36), illustrating the base jam´e from (35) above, is from elicitation; (37) illustrates the occurrence of an unmarked bivalent bases in texts (see section 5 for more examples): (36)

jam-´e- -i n-is waPwa-n-kwa tie-co- -3n obl-art.pl redPliana-ln-abs ‘They are tied onto (sth.) with a liana.’

(37) am-poj is Santo rey jayna enter-caus art.pl Saint mod dsc ‘The Saints had already been put inside again.’ 4.2. Monovalent bases The so-called monovalent bases, in contrast, cannot be combined with the full range of voice markers. This group, though semantically less consistent than the one described above, includes elements referring to oneparticipant events like ‘sleep’ or ‘stand’, which is why I refer to them as monovalent. Examples of monovalent roots are listed in (38) (see Haude 2006: 340¤.).

274

Katharina Haude

(38)

joyenasde:josichijiwakaydejal´okot-

‘go’ ‘stand’ ‘sit’ ‘lie’ ‘laugh’ ‘go out’ ‘come’ ‘eat’ ‘cook’ ‘boil’

The formal diagnostic for identifying a monovalent base is the e¤ect of the su‰x -na. With bivalent bases, the addition of this su‰x (or its allomorph -a- on some complex bases) creates a direct transitive predicate, while on monovalent bases, the su‰xation of -na creates a locational noun; the di¤erence is illustrated in (39) (bivalent) and (40) (monovalent). (39)

jat-na¼Ø hit-dr ¼1sg ‘I hit (you/him/her/it/them).’

(40) kay-na¼Ø eat-loc ¼1sg ‘the place where I eat’ The derived locational nouns can be identified as nouns because, when functioning as predicates, their subject can only be expressed by a clauseinitial free pronoun and not by an externally cliticized pronoun, as shown in (41a) and (41b), respectively. They share this syntactic limitation with possessed nominal predicates (see Haude 2010). (41) a.

b.

bo a’ko as-na¼Ø reas pro sit-loc ¼1sg ‘because it is my house’ *bo as-na¼Ø- -a reas sit-loc¼1sg- -3n (‘because it is my house’) [el]

Just like bivalent roots, monovalent roots cannot occur independently; to form a main-clause predicate, they must be combined with another morpheme, the choice of which depends on the root. Most monovalent

Undergoer orientation in Movima

275

roots are combined either with the reflexive/reciprocal su‰x -che´ or with the middle reduplication, as illustrated in (42) and (43), respectively, for the roots listed above. (Others take distributionally more restricted, unproductive su‰xes with possibly specific meanings; see Haude 2006: 342– 344.) Many verbs formed in this way express activities, by which the actor may or may not be seen as a¤ected. Some middle verbs denote transitive events whose patient can optionally be expressed by an oblique noun phrase, as was shown in example (2) for the verb kaykay ‘eat’. (42)

joy-che´ en-che´ josi:-che´ pen-che´ tes-che´ de:-che´

‘go, leave’ ‘stand (up)’ ‘laugh’ ‘land (plane, bird)’ ‘limp’ ‘lie (down)’

(43) kayPkay chi:Pchi ´apP´ap jiwa ´okot dejal

‘eat’ ‘go out’ ‘bathe’ ‘come’ ‘boil’ ‘cook’

A text example of middle verbs based on monovalent roots is given in (44). While identical with respect to morphological marking, the verb deja:jal ‘cook’ in (44a) denotes an activity and takes the agent (the cook) as subject, while ´oko:kot ‘boil’ in (44b) denotes a process and takes the a¤ected entity (the cooked food) as subject. (44) a. b.

ena’ dejal- -isne che jayna dur.std cook- -3f.ab and dsc da’ ´okot is dej-na¼sne 4 dur.nstd boil art.pl cook-dr ¼3f.ab ‘She was cooking, and what she cooked was already boiling.’

Complex monovalent bases can be derived through argument incorporation (see Haude 2006: 283–286). Although a verb with an incorporated argument (usually in the form of a bound lexical element or classifier) obligatorily contains the direct marker, it is syntactically intransitive; this 4. While probably historically related, the monovalent root dejal- has to be considered as distinct from the bivalent root dej-, since the ending -al is unanalyzable.

276

Katharina Haude

can be seen from the oblique marking of the NP in (45), which is coreferential with the incorporated argument. Furthermore, the addition of the su‰x -na creates a locational noun, as shown in (46). (45) dan-a:-so- -is n-is pokso chew-dr-cl.chicha- -pl.ab obl-art.pl chicha ‘They chewed (the) chicha.’ (46) asko yok-a-mo-na¼is n-eys pro.n.ab catch-dr-cl.bird-loc¼3pl.ab obl-dem.spk.pl karak di’ sere:re macaw rel wild ‘That (was) where they caught those wild macaws.’ In addition, there are monovalent verbs that do not show any sign of (synchronic) complexity and that cannot be combined with a voice marker; examples are given in (47). (47) ja:yi ja:ra´ jo’yaj te:lo salmo yolmo´ ya:lo:we

‘run’ ‘fight’ ‘arrive’ ‘dance’ ‘return’ ‘go for a walk’ ‘drink’

Also on these verbs, the addition of -na creates a locational noun, as illustrated in (48): (48) che asko jayi-na¼us and pro.n.ab run-loc ¼3m.ab ‘And that (was) where he ran to.’ While the aspectual properties of monovalent bases require further study, the data so far indicate that these bases denote activities (cf. Van Valin 2005: 55) rather than states, and in this respect di¤er significantly from unmarked bivalent bases. 5. The undergoer orientation of bivalent bases 5.1. Orientation The term ‘‘orientation’’ is sometimes used to refer to the participant role a verb assigns to its central argument. In the transitive domain, the accusa-

Undergoer orientation in Movima

277

tive pattern implies actor orientation and the ergative pattern undergoer orientation. In the intransitive domain, so-called unaccusative verbs are typically undergoer-oriented and unergative verbs actor-oriented. The preceding sections have shown that in Movima, there is a split in the domain of morphologically unmarked verbal bases: the bivalent bases, forming stative predicates when unmarked, are undergoer oriented; the majority of monovalent bases, in contrast, denote activities and hence can be considered actor oriented.5 A major function of the voice markers is to overtly indicate and/or modify the orientation of the voice-marked verb (see Serzisko 1991): in Movima, a verb marked as resultative (intransitive) or direct (transitive) is oriented towards the undergoer, a verb marked as agentive (intransitive) or inverse (transitive) is oriented towards the actor. Reflexive/reciprocal and middle-marked verbs, whose actor is a¤ected by the action it instigates, have a tendency to be actor oriented; this is apparent from the e¤ect of middle and reflexive/reciprocal marking of monovalent bases, which denote activities (see (42), (43) above). Thus, since the bivalent bases, which participate fully in the voice system, are undergoer oriented and the default transitive clause patterns ergatively, the hypothesis is that the function of the direct marker is to permit the additional expression of the actor argument. This hypothesis is strengthened by the fact that the resultative marker -’i, which overtly marks the verb as undergoer oriented, shows signs of being closer to a zero marker than the other voice morphemes (and was analysed as such in Haude 2006). The correlation between resultative marking and absence of marking is both semantic and phonological. 5.2. Semantic parallels between unmarked and resultative verbs When in elicitation, speakers are faced with a minimal pair of an unmarked bivalent base and a resultative form marked with -’i, their explanations make it clear that the unmarked form is interpreted as denoting a state ‘‘without any implication of its origin’’ (Nedjalkov and Jaxontov 1988: 6), while in the resultative form, a (possibly unknown) actor is implied. 5. It may be tempting to use the terms ‘‘unaccusative’’ and ‘‘unergative’’ for the ‘‘bivalent’’ and ‘‘monovalent’’ verbal bases of Movima, since these terms are syntax-based and include lexical idiosynchrasies (e.g. Movima ´okot- ‘boil’, which falls in one class of predominantly activity verbs like dejal- ‘cook’). However, they are not very felicitous here because they rely heavily on the notions of subject (agent) and object (patient) (see Perlmutter 1978) and on syntactic tests.

278

Katharina Haude

Some verbs that were tested this way are listed in (49) (the parentheses in the glosses provide the interpretation of the verb with -’i; bases whose stress or lengthening pattern changes when occurring with the su‰x are presented separately). (49) katpit(-’i) ´ok-poj(-’i) way’cho´(-’i) ja:rat ( jarat-’i) do’waj (dowaj-’i) ji:sa ( ji:sa:-’i)

‘be broken in halves (by someone)’ ‘be felled (by someone)’ (-poj caus) ‘be mended (by someone)’ ‘be thrown away (by someone)’ ‘be moved to another place (by someone)’ ‘to be made (by someone)’

However, the distinction between simple and resultative states is not so clear-cut in natural discourse, where the forms seem to be used interchangeably. The following examples may illustrate this. As was shown in 3.2.2, the resultative occurs in descriptions and instructions, where sequences of events are described. However, here also the unmarked forms are found, like the verb nanra in (50) (a description of raising cattle) and ´okpoj and dakato:lej in (51) (a description of how a slash-and-burn field is made): (50) n-as jayna tawakeni-wa¼i jayna obl-art.n dsc wake_up-nmz¼3pl dsc nanra- -i reyka, nanra set_free- -3pl.ab mod set_free bo as joy-wa¼i di:ra reas art.n go-nmz ¼3pl at_least n-as kay-wa¼i n-as chapmo obl-art.n eat-nmz ¼3pl obl-art.n bush ‘When they [the cows] wake up, they are set free again, (they) are set free so that they go at least to feed in the bush.’ (51) jayna ´ok-poj kis ko’ dsc fall-caus art.pl.ab trees che das-ka-to:lej jayna and cut-mlt-branch dsc bo as de:-wa¼kis tolej-a¼kis ko’o reas art.n lie-nmz ¼art.pl.ab branch-ep ¼3pl.ab tree ‘Then the trees are felled and their branches cut o¤ (lit. ‘. . . [they] are repeatedly branch-cut), so that the branches of the trees lie flat.’

Undergoer orientation in Movima

279

(52) kiro’ kis lotoba¼is di’ pokso, dem.pl.ab art.pl.ab jug¼3pl.ab rel chicha joy-´e no-kos kavildo go-co obl-art.n.ab Cabildo ‘There were their jugs of chicha, (they had been) taken to the Cabildo.’ On the other hand, verb roots, which cannot occur without additional phonological material, take the resultative marker -’i when denoting states that do not imply an actor, like yey’i in (53) or ´ek’i in (54): (53) rim´e¼Ø n-is wa:ka-wandi buka’ sell¼1sg obl-art.pl cow-instr:be.house dur.mov bo rey yey-’i- -is reas mod want-res- -3pl.ab ‘I sold (the hats) on the ranches because they were appreciated.’ (54) yey-na¼’ne os ma:kina di’ ´ek-’i want-dr ¼3f art.n.pst machine rel kick-res ‘She wanted a (sewing) machine that is foot-driven.’ The equivalence of unmarked and resultative forms is also apparent when unmarked stative verbs and verbs ending in -’i cooccur in the same clause, as in (55) (a joke), where they denote a sequence of events: (55) dak-´e che jarat-’i, jayaw-´e¼a cut-co and dump-res good-neg ¼3n ‘[Your head] needs to be cut o¤ and thrown away, it’s useless!’ Furthermore, there are verbs where, without any obvious morphophonological reason, the attachment of -’i is not possible, while others cannot occur in the unmarked form. In both cases, the resultative and the stative reading are indistinguishable. For instance, the verb rim´e ‘sell’ in (56) cannot occur with the resultative marker, whereas the verb base rime´- ‘buy’ in (57) cannot occur without it when denoting a state: (56) a.

b.

ba:ra rim´e all sell ‘It is all sold.’/ ‘It has all been sold.’ *ba:ra rim´e-’i all sell-res (‘It has all been sold.’)

[el]

280

Katharina Haude

(57) a.

ba:ra rime´-’i all buy-res [ba:.ra ri¨me´i] ‘It is all bought.’/ ‘It has all been bought.’ [el]

b.

*ba:ra ri:me´ all buy (‘It is all bought.’)

[el]

Another verb that, like rim´e, cannot occur with the ending -’i to denote a resultative state, is jommi ‘eat up’ in (58): (58) ban kiwa il-na¼Ø bo nokowa jommi but dem spread-dr ¼1sg reas right_now eat_up ‘But I have spread it there because now it will be eaten.’ Resultative verbs that, like rime´, cannot occur without the ending -’i (besides the monosyllabic roots) are listed in (59). For the first three verbs in this list, the most probable explanation is that they constitute verb roots and therefore cannot occur on their own; for the last two, no such explanation can be given, since they are clearly composed of a root and the causative su‰x. (59) tikoy-’i e:la:-’i dewaj-’i chi-poj-’i ju:-poj-’i

‘be killed’ ‘be left behind’ ‘to be seen’ ‘to have been taken out’ (‘go out’þcausþres) ‘to have been punished’ (‘punish’þcausþres)

5.3. Phonological parallels between -’i and zero The hypothesis that of all voice markers, the resultative marker -’i comes close to zero marking receives support from the fact that the element [i] is homophonous with a dummy element that occurs on some prosodically deficient nouns and pronouns. In Movima, a content word (noun, verb, adjective) must be minimally disyllabic. Apart from very few exceptions (Haude 2006: 196), monosyllabic noun roots are either augmented by reduplication or by attachment of the ending [i], the choice being lexically determined. Example (60) illustrates the augmentation of the monosyllabic noun root nun- with the dummy -’i. The resulting form behaves like any other noun; for example, it can be marked for alienable and inalienable possession, as shown in (60b) and (60c), respectively. Examples

Undergoer orientation in Movima

281

(61) and (62), however, where the noun forms part of a compound and an incorporating verb, respectively, show that the root of the word is nun- alone (see also Haude 2006: 207). (60) a.

nun-’i bone-d ‘bone’

b.

nun-’i¼a bone-d ¼3n ‘its bone (al.)’

c. nun-’i¼a bone-d ¼3n ‘its bone (inal.)’ [el]

(61) punta:-nun tip-bone ‘the bone from the tip (of the rib cage)’ (62) it dan-a:-nun 1intr chew-dr-bone ‘I chew on bones.’ [el] The element [i] also occurs in initial position on pronouns that do not display the required syllable structure. It augments the first-person pronominal elements (i)´ and (i)t, which consist of a simple consonant and therefore need to be attached to a preceding vowel. In (63), this vowel is provided by the preceding word; in (64), the preceding word ends in a consonant, therefore the element [i] is inserted as a dummy host. (63)

jayna t joy-che´ dsc 1intr go-r/r ‘Then I went.’

(64) ban it joy-che´ but 1intr go-r/r ‘But I went.’ The personal pronouns (i)’ne ‘3f’, (i)sne ‘3f.ab’, (i)y’´i ‘1pl’, and (i)y’bi ‘2pl’, which would have a complex onset if occurring independently, are preceded by an element /i/ when occurring as free forms or when externally cliticized to a consonant-final host. Compare the variants of the third-person feminine pronoun (i)sne ‘3f.ab’ when cliticized to a vowel-final host, (65a), when occurring independently, as in (65b), and when cliticized to a consonant-final host (resyllabifying with that consonant), as in (65c). The stress and syllable pattern is given in the phonemic representation.

282

Katharina Haude

(65) a.

salmo- -sne return- -3f.ab /¨sal.mos.ne/ ‘She returned.’

b.

isne salmo 3f.ab return /¨is.ne ¨sal.mo/ ‘She returned.’

c. a:mon–isne enter–3f.ab /¨a:.mo.˙nis.ne/ ‘She came in.’ Note that the bound pronoun in (65c) is not preceded by a glottal stop, the pronoun being resyllabified with the preceding consonant. It may be argued that here, the element /i/ cannot be compared to the sequence [i] in the above examples. However, as was shown in Haude (2006: 101), it is a property of external cliticization that referential elements are resyllabified with the preceding consonant, and that the glottal stop, which otherwise precedes all vowel-initial morphemes, is dropped. If -’i is a prosodically triggered element in other environments, then it is very well conceivable that also on verbal bases, it is synchronically perceived as a phonological dummy. This may be the reason why in actual speech the resultative verbs are used in the same way as unmarked bivalent verbs, given that both denote states.

6. Conclusion Movima has a set of verbal voice markers that indicate the transitivity of a verb and the participant role(s) (actor, undergoer, or both) of its subject. Six voice markers can be identified, four of which mark intransitive verbs (reflexive/reciprocal, resultative, agentive, and middle) and two that mark transitive verbs (direct and inverse). Verbs unmarked for voice are intransitive, while all transitive verbs are overtly morphologically derived through direct or inverse marking. Movima can therefore be characterized as a ‘‘fundamentally intransitive language’’ (Nichols 1982; see also Nichols, Peterson, and Barnes 2004).

Undergoer orientation in Movima

283

Only one class of verbal bases, termed ‘‘bivalent’’ because they denote two-participant events, can be productively combined with the voice markers. The other class, termed ‘‘monovalent’’ because most of its members denote one-participant events, only has limited access to the voice markers, and their combination with the reflexive/reciprocal or middle marker is lexically determined. In particular, monovalent bases cannot be marked as direct; on them, the su‰xation of the element -na (a direct marker on bivalent bases) derives a locational noun. Bivalent verbs unmarked for voice denote states. They are in this respect similar to the resultative forms, marked by the su‰x -’i. The similarity is confirmed by the distributional equivalence of unmarked and resultative forms, the fact that some can only be occur unmarked and others only marked by -’i when denoting a state, and by the fact that the su‰x -’i is homophonous with a dummy element [i] found in other environments. Both unmarked and resultative verbs are undergoer oriented, i.e. have a non-actor as their subject. Direct-marked predicates are also characterized by undergoer orientation. Therefore, although both the direct and the inverse derivation are overtly morphologically marked, the direct derivation can be seen as the simpler operation, since it introduces an actor without changing the orientation of the verb. Maybe this fact has contributed historically to the basic status of the direct (ergative) construction. The inverse marker, in contrast, introduces an undergoer and at the same time changes the orientation of the verb. The inverse voice can, from this perspective, be seen as operating on the transitive level, enabling a referentially high-ranking participant to take the undergoer role. The cross-linguistically unusual ergative bias found in Movima discourse data can thus, at least in part, be caused by the underlying undergoer orientation of bivalent verbal bases.

Abbreviations (partly adapted from the Leipzig Glossing Rules) ¼ internal cliticization -external cliticization

infixation P reduplication 1 first person 2 second person

284 3 ab abs agt appl art be ben caus co d dem det detr dr dsc dur ep ev f frust inal inv hyp imm instr irr itn intr loc m md mod mov n neg nmz nstd ntr obl

Katharina Haude

third person absential absolute state agentive applicative article bound nominal element benefactive causative co-participant dummy demonstrative determiner detransitivizer direct discontinuous durative epenthetic vowel evidential feminine frustrative inalienable inverse hypothetical immediate past instrumental irrealis intentional intransitive location masculine middle modal moving neuter negation action/state nominalization nonstanding neutral oblique

Undergoer orientation in Movima

obv pst pl prc pro reas red rel res r/r sg trc vbz

285

obviative marking past plural process free pronoun reason reduplication relativizer resultative reflexive/reciprocal singular truncated element verbalizer

References Bickel, Balthasar 2010 ‘‘Grammatical relations typology.’’ In The Oxford Handbook of Linguistic Typology, Jae-Jung Song (ed.), 399–444. Oxford: Oxford University Press. Comrie, Bernard 1981 ‘‘Aspect and voice: some reflections on perfect and passive.’’ In Syntax and Semantics 14: Tense and Aspect, Philip Tedeschi and Annie Zaenen (eds.), 65–78. New York et al.: Academic Press. Cooreman, Ann 1994 ‘‘A functional typology of antipassives.’’ In Voice: Form and Function, Barbara A. Fox and Paul J. Hopper (eds.), 49–87. Amsterdam/Philadelphia: Benjamins. Croft, William 2003 Typology and Universals. Cambridge: Cambridge University Press. Second edition. DeLancey, Scott 1981 ‘‘An interpretation of split ergativity and related patterns.’’ Language 57(3): 626–657. Dixon, R.M.W. 1994 Ergativity. Cambridge: Cambridge University Press. Dixon, R.M.W. and Alexandra Y. Aikhenvald 2000 ‘‘Introduction.’’ In Changing Valency. Case Studies in Transitivity. R.M.W. Dixon and Alexandra Y. Aikhenvald (eds.), 1–29. Cambridge: Cambridge University Press. Foley, William A. and Robert D. Van Valin, Jr. 1984 Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press.

286

Katharina Haude

Gildea, Spike 1994

Givo´n, T. 1994

‘‘Semantic and pragmatic inverse: ‘inverse alignment’ and ‘inverse voice’ in Carib of Surinam.’’ In Voice and inversion, T. Givo´n (ed.), 187–230. Amsterdam/Philadelphia: Benjamins.

‘‘The pragmatics of de-transitive voice: functional and typological aspects of inversion.’’ In Voice and inversion, T. Givo´n (ed.), 3–44. Amsterdam/Philadelphia: Benjamins. Haspelmath, Martin 1994 ‘‘Passive participles across languages.’’ In Voice: Form and Function, Barbara Fox and Paul J. Hopper (eds.), 151–177. Amsterdam/ Philadelphia: Benjamins. Haude, Katharina 2006 A grammar of Movima. Ph.D. diss., Radboud Universiteit Nijmegen. Haude, Katharina 2009a ‘‘Reference and predication in Movima.’’ In New Challenges in Typology 2: Transcending the Borders and Refining the Distinctions, Alexandre Arkhipov and Patience Epps (eds.), 323–342. Berlin: Mouton de Gruyter. Haude, Katharina 2009b ‘‘Hierarchical alignment in Movima.’’ International Journal of American Linguistics: 513–532. Haude, Katharina 2010 ‘‘The intransitive basis of Movima clause structure.’’ In Ergativity in Amazonia, Spike Gildea and Francesc Queixalo´s (eds.), 285– 315. Amsterdam/Philadelphia: Benjamins. Haude, Katharina 2011 ‘‘Argument encoding in Movima: the local domain.’’ International Journal of American Linguistic 77(4): 559–575. Hopper, Paul and Sandra A. Thompson 1980 ‘‘Transitivity in grammar and discourse.’’ Language 56(2): 251– 299. Kazenin, Konstantin I. 2001 ‘‘The passive voice.’’ In Language Typology and Language Universals: an International Handbook. Vol. 2., Martin Haspelmath, Ekkehard Ko¨nig, Wulf Oesterreicher and Wolfgang Raible (eds.), 899–916. [Handbu¨cher zur Sprach- und Kommunikationswissenschaft, 20.] Berlin/New York: Walter de Gruyter. Keenan, Edward L. and Matthew S. Dryer. 2007 ‘‘Passive in the World’s Languages.’’ In Language Typology and Syntactic Description. Vol. 1. Clause Structure, Timothy Shopen (ed.), 325–361. Cambridge: Cambridge UP. Second Edition.

Undergoer orientation in Movima

287

Kulikov, Leonid 2010 ‘‘Voice typology’’. In The Oxford Handbook of Language Typology, Jae-Jung Song (ed.), 368–398. Oxford: Oxford University Press. Mithun, Marianne 1994 ‘‘The implications of ergativity for a Philippine voice system.’’ In Voice: Form and Function, Barbara Fox and Paul J. Hopper (eds.), 247–277. Amsterdam/Philadelphia: Benjamins. Nedjalkov, Vladimir P. and Sergej Je. Jaxontov 1988 ‘‘The typology of resultative constructions.’’ In Typology of resultative constructions, Vladimir P. Nedjalkov (ed.): 3–62. Amsterdam/Philadelphia: Benjamins. Nichols, Johanna 1982 ‘‘Ingush Transitivization and Detransitivization.’’ In Proceedings of the Eighth Annual Meeting of the Berkeley Linguistics Society, 445–462. Nichols, Johanna, David A. Peterson, and Jonathan Barnes 2004 ‘‘Transitivizing and detransitivizing languages.’’ Linguistic Typology 8(2): 149–211. Perlmutter, David M. 1978 ‘‘Impersonal passives and the Unaccusative Hypothesis.’’ In Proceedings of the Fourth Annual Meeting of the Berkeley Linguistic Society: 157–189. Berkeley Linguistic Society, University of California, Berkeley. Serzisko, Fritz 1991 ‘‘Orientierung.’’ In Partizipation, Hansjakob Seiler and Waldfried Premper (eds.), 273–308. Tu¨bingen: Gunter Narr. Shibatani, Masayoshi 1985 ‘‘Passive and related constructions: a prototype approach.’’ Language 61: 821–848. Shibatani, Masayoshi 2006 ‘‘On the conceptual framework for voice phenomena.’’ Linguistics 44(2): 217–269. Siewierska, Anna 1984 The Passive. A Comparative Linguistic Analysis. London et al.: Croom Helm. Van Valin, Robert D., Jr. 2005 Exploring the Syntax-Semantics Interface. Cambridge: Cambridge University Press. Van Valin, Robert D., Jr. and Randy R. LaPolla 1997 Syntax. Structure, Meaning and Function. Cambridge: Cambridge University Press. Zu´n˜iga, Fernando 2006 Deixis and Alignment. Inverse Systems in Indigenous Languages of the Americas. Amsterdam/Philadelphia: Benjamins.

Case patterns and verb classes in Trumai Aurore Monod Becquelin and Ce´dric Becquey 1. Introduction 1.1. Ethnographic and sociolinguistic context The Trumai are one of the groups of the Upper Xingu (Mato Grosso, Brazil), first contacted at the end of the 1880s by Carl von den Steinen (1942). A series of brief expeditions in the early twentieth century was followed by a period of relative isolation that lasted until the 1960s. Since then, the growing number of ethnographic studies has produced a series of monographs on nearly every group in the region, highlighting the multiethnic and multilingual character of this territory (Basso 1973) – fourteen groups have been identified, belonging to five families: Tupi (Kamaiura´, Kayabi, Aweti, Yudja´), Caribe (Kalapalo, Kuikuro, Matipu, Nahukua´, Ikpeng), Arawak (Waura´, Yawalapiti), Geˆ (Kayapo´, Kiseˆdje´) and Trumai. The region can be divided into two cultural areas, northern and southern; the latter (the Uluri area, Galva˜o 1953), which has formed over the course of several decades, comprises various groups with a certain number of features in common and a strong multilingual practice (for a history of the region, see Franchetto and Heckenberger 2001). Quain was the first anthropologist to work with the Trumai (1938): he described a very tense situation in which a drastically reduced population existed in a state of conflict with neighbouring groups (Murphy and Quain 1955). The creation of the Xingu National Park in 1961 resulted in demographic growth among the groups of the Upper Xingu and for the Trumai in particular, but the increasing pressures of modernization, in the face of which they are still weak numerically, linguistically and culturally, places them in an unstable situation with regard to the other groups (Monod Becquelin et al. 2008). Regional multilingualism is significant – because of the multiple matrimonial, ritual and economic relations between groups – but also specifically significant for the speakers represented in the two main corpora: the corpus of Aurore Monod Becquelin (AMB) shows the influence of Kamayura

290

Aurore Monod Becquelin and Ce´dric Becquey

(Tupi), while there is Portuguese influence, particularly through schooling, on the corpus of Raquel Guirardello Damian (RGD).1 1.2. The crux of the matter This article examines the two bivalent verb classes of Trumai (for description of Trumai verb classes see AMB 1975, 1976 and RGD 1999, 2003).2 Alongside class 1, composed of monovalent verbs taking an unmarked core argument, class 2 is bivalent and has been labelled ‘‘ergative’’, while the last major class, class 3, has been analysed as ‘‘extended intransive’’ following Dixon’s terminology, and employs the unmarked argument together with the ‘‘extended to core’’ argument E (Dixon 1994: 122–4, Dixon and Aikhenvald 2000: 3, Guirardello 1999, 2002). In class 2, semantic roles are canonically encoded by an ergative marker interpreted as an agent and an absolutive form (identical to the unmarked subject of an intransitive verb) interpreted as the patient. In class 3, the agent takes the absolutive case (unmarked) and the patient is marked by one of three markers (-s, -ki, -tl ) otherwise used for encoding oblique cases (space and time locatives and recipient dative). The di¤erent morphosyntactic alignments and the di¤erent possible forms taken by argument structure in Trumai have given rise to two opposing interpretations. The first of these (Trumai as an ergative language) tends to favour the ‘‘ergative’’ argument as syntactic subject and agent (1a), and the unmarked argument as the object of the transitive (1b) and the subject of an intransitive, whether extended (1c) or not (1a). The alternative point of view (Trumai as an accusative language) considers the unmarked argument (which would then be labelled ‘‘nominative’’) as the central argument in verbal constructions (2a,c), necessitating 1. The first survey of Trumai was initiated by A. Monod Becquelin (AMB) in 1966, and she continued to work on the language into the 1990s. The research conducted by the linguist Raquel Guirardello Damian (RGD) began in 1989, and formed the subject of her thesis and subsequent works (see the bibliography). Texts from the ethnological research of Emmanuel de Vienne (first fieldwork in 2002) have also been used, as well as a Master’s thesis on Trumai by Ce´dric Becquey (2007) based on the documentation gathered by previous researchers. 2. Up to this point Trumai has usually been considered a linguistic isolate. Its phonological system includes a phoneme unique on the whole American continent, /T/ (this apico-alveolar occlusive, apparently made with the tongue flattened against the palate, has yet to be defined acoustically), and another which is very rare in South America, /tł/ [ł], [tł], here tl (see Monod Becquelin 1975, Guirardello 1999). The main syllabic structures are CVCV(C) and VCVC.

Case patterns and verb classes in Trumai

291

an analysis of the agent in class 2 (2b) as an agentive complement to a verb without passive verbal morphology (see Mel’cˇuk 1983: 248); in class 3, the patient would be interpreted as bearing a set of accusative/dative su‰xes (2c). Ergative representation: (1) a. hine yotl 3abs sleep subject ‘He sleeps.’ b.

hine-k k’ate iki 3p-erg fish.abs arrow agent patient subject object ‘He arrows (the) fish.’

c. hine ma k’ate-s/-ki/-tl 3abs eat fish-e1/e2/e3 agent patient subject oblique ‘He eats (the) fish.’ Accusative representation (2) a. hine yotl 3nom sleep subject ‘He sleeps.’ b.

case labels syntactic functions

case labels semantic roles syntactic functions

case labels semantic roles syntactic functions

case labels syntactic functions

hine-k k’ate iki 3p-obl fish.nom arrow agent patient oblique subject ‘The fish is arrowed by him.’

case labels semantic roles syntactic functions

c. hine ma k’ate-s/-ki/-tl 3nom eat fish-acc1/acc2/acc3 agent patient subject object ‘He eats (the) fish.’

case labels semantic roles syntactic functions

292

Aurore Monod Becquelin and Ce´dric Becquey

These divergences in the analysis of case assignment patterns have often made it impossible to define syntactic functions in Trumai (RGD 1999: Ch.7.3.1) because of the complexity and diversity of the morphosyntactic alignments of verbal arguments. Both the descriptive terminology of the Trumai verbal system and the typological position of Trumai need to be reconsidered. In particular, this reinterpretation should take into account the following specific features: in Trumai, no operations on argument structure are found which involve a change in morphological diathesis3; changes to argument structure are, however, possible either by means of a lexical alternation between two parallel forms we call ‘‘doublets’’, or in connection with a syntactic property belonging to certain verbs known as ‘‘fluid verbs’’ (see 2.1.4); moreover, Trumai lacks a clear syntactic pivot (RGD 2004). We must a) take into account the very frequent deletability/optionality of ergative arguments, and b) specify the relationship between valency and semantic transitivity for each verb class. The next section briefly presents the verb classes and associated case patterns using, for the present, the ‘‘ergative’’ terminology established by earlier publications (RGD 1999). Section 3 takes into account the relationship between semantic roles and morphosyntactic coding. Section 4 shows that a relevant analysis of Trumai verb classes requires a thorough textual inquiry in order to provide a sounder understanding of the relationship between verbal valency and argument structure in classes 2 and 3. The morphosyntactic alignment split governed by the choice between lexical ‘‘doublets’’ has been scrutinized for numerous texts and with various types of elicitation; we have found no semantic di¤erences leading to the use of one or the other class of verbs, and our principal claim is that the use of the verb classes is discourse-determined.

3. With the exception of the causative construction, which, without changing the two-argument structure of the verb, introduces a causer in the ergative case and a causative morpheme on the verb: axos ma k’ate-s child eat fish-E1 ‘The child eats some fish.’ hine-k axos ma-ka k’ate-s 3p-erg child eat-caus fish-E1 ‘He gets the child to eat some fish.’

Case patterns and verb classes in Trumai

293

2. A brief outline of the verb classes and case patterns The verbal morphology of Trumai is not extensive. Essentially, it consists of three sets of markers: – The first is the absolutive 3rd person (or non-person) su‰x required in the absence of any lexical expression of this argument. This has two allomorphs, depending on whether it is preceded by a vowel, V-n, or a consonant, C-e. The personal pronouns, meanwhile, are independent forms and in addition to person they indicate number (singular, dual and plural), gender (for 3rd person), clusivity (inclusive/exclusive) and case (AMB 1975, RGD 1999). – Modal, aspectual, directional and circumstantial postverbal grammatical markers. – Transcategorial su‰xes (nominalizer, adjectivizer, etc.). 2.1. Verb classes Case assignment is the most visible criterion for defining a verb class. Verbs assign cases to all their dependents, and these cases are encoded by morphological markers governed by schemes which di¤er from one verb class to another. Valency refers to the capacity of a verb to take a specific number of arguments. Class 1 verbs are monovalent; class 2 verbs are bivalent; in the ergative interpretation of Trumai, class 3 verbs are classifed as monovalent and analysed as extended intransitives on morphological grounds. But if we take into account pragmatics (topic analysis), frequency (distribution over the lexicon), and semantic roles (agent/patient), verbs of class 3 should be considered bivalent (see below). In addition to this, the three di¤erent forms of E (-s, -ki, -tl ) exhibit domains of application – spatial, temporal and notional – and semantic values which must be scrutinized for a better understanding of their syntactic function. AMB’s first analysis (1975) attributed as much importance to semantic roles as to morphosyntactic patterns, mixing levels in a confusing way. This interpretation considered as transitive what RGD calls extended intransitive; the unmarked argument, which encodes the agent, was considered to be in the nominative case, while the marked argument, which encodes the patient, was taken as accusative. Both verb classes were considered to display equally strong semantic transitivity: the definition of transitives 2 was based on ergative morphology whereas that of transitives 3 was based on semantics. We will now see that the typological identification of Trumai as an ‘ergative language’ is not satisfactory.

294

Aurore Monod Becquelin and Ce´dric Becquey

2.1.1. Class 1 This class is composed of monovalent intransitive verbs. The unique argument is morphologically unmarked. (3) ha tsula.tsula, ha demTı¨ta pat 1abs be.lying.down 1abs rest a.little ‘I am lying down, I am resting a little.’ (Plant.) (4) kachı¨-n ale hen go-3abs one.says then ‘Then he went on, one says.’ Other verbs of this class are: ora ‘‘cry (in pain / of animals)’’, wanle ‘‘be over’’, chı¨ ‘‘go’’, fakdits ‘‘die’’, demle ‘‘be tired’’, lafku ‘‘swim’’, otl ‘‘sleep’’ ( yotl when immediately following the personal pronouns ha, hi, hine), pata ‘‘arrive’’, watkan ‘‘weep’’, tsula ‘‘lie down’’. 2.1.2. Class 2 This class is made up of bivalent verbs with ergative alignment. If there is any prototypical agent it will correspond to the marked argument (ergative), and the unmarked argument (absolutive) will correspond to the prototypical patient (but see (6)). (5) adis.pa-ek ha-wan disi Indians.of.Xingu-erg 1abs-pl kill ‘The Indians of Xingu killed us.’ (Murder) (6) kasoro-k ha tako dog-erg 1abs bite ‘The dog bites me.’ (7) yaw-ak padi-n de people-erg wait-3abs already ‘People are already waiting for him.’ (AMB notes) The imperative form di¤ers from that of the other verb classes: here, the imperative particle takes the form waki. (8) Waki kı¨tı¨ hai-tl Imp.erg give 1abs-e3 (beneficiary) ‘Give it to me.’

Case patterns and verb classes in Trumai

295

Other verbs of this class include: etsi ‘‘bring’’, disi ‘‘kill’’, tako ‘‘bite’’, padi ‘‘be waited for’’, miro.miro ‘‘rasp’’, maxkewa ‘‘paint’’, k’etan ‘‘move’’, kapan ‘‘make’’, ts’ake ‘‘be fond of ’’, ni’tsapa ‘‘squeeze’’. 2.1.3. Class 3 The first argument of so-called ‘‘extended intransitive’’ verbs is the unmarked argument, i.e. takes absolutive case, while the second argument is labelled E: in the definition of Dixon and Aikhenvald, this is an argument with a special status (E stands for ‘extension to core’). ‘‘In [some languages], E and peripheral are treated in the same way’’ (2000: 3): this is the case in Trumai. (9) ha ma t’aak-es, k’ate-s 1abs eat cake-E1 fish-E1 ‘I eat cakes and fish.’ (10) wana xu’tsa de kayapi wan ami-ki imp see already Kayabi pl word-E2 ‘Remember the words of the Kayabi!’ (Murder) (11) Uksitukuk-etl otl.taxeL-e capybara-E3 dream-3abs ‘He dreams of capybara.’ (Karijawar) The imperative form for this class, as for class 1, is wa/wana. (12) wana xu’tsa, wana ma, wana yotl imp see imp eat imp sleep ‘Look, eat and sleep!’ Verbs of this class include: lax ‘‘hunt, fish’’, api ‘‘catch’’, laT ‘‘lie’’, da’tsi ‘‘carry’’, elka ‘‘exchange’’, fatlkamu ‘‘believe’’, fatne ‘‘clear (the plantation)’’, xu’tsa ‘‘see’’, ma ‘‘eat’’, poyo ‘‘take revenge’’, waymi ‘‘tell’’. Note that class 2 and 3 both include verbs usually considered highly transitive, as well as some generally regarded as experiencer verbs (see 3.2). 2.1.4. ‘‘Fluid verbs’’ and ‘‘doublets’’ There is a small set of ‘‘fluid’’ verbs which accept both of the preceding case assignment patterns while exhibiting the same semantic structure (without any aspectual/telic motivation for choosing one or the other).

296

Aurore Monod Becquelin and Ce´dric Becquey

Verbs belonging to this class include: chapta ‘‘clean (the plantation)’’, t’aka ‘‘repair’’, ami’in ‘‘answer’’, homne ‘‘meet, find’’, pechma ‘‘thread, insert’’, pı¨rew ‘‘destroy, finish’’. According to RGD (2003), the di¤erent alignment patterns are sometimes correlated with a slight semantic nuance (routinized actions are associated with ‘‘extended intransitive’’ alignment); see table 1. Table 1. Examples of fluid verbs with semantic nuances according to RGD Verb

ergative alignment

extended intransitive alignment

naha ochen pı¨ Tke kı¨tı¨w tı¨ami wen

cut grind peel grate crush, squeeze tear out

cut manioc pound, make flour peel manioc grate manioc squeeze manioc pulp pluck a bird

However, this semantic nuance is not always in evidence in the texts: (13) a.

hai-ts ¨ıwı¨r naha faka˜o-letsi 1p-erg wood cut machete-instr b. ¨ıwı¨r-as ha naha faka˜o-letsi wood-e1 1abs cut machete-instr ‘I cut wood with a machete.’

Trumai displays another device allowing a shift between the two opposing argument structures while preserving the semantic relationship between the two participants. The two constructions involve paired lexemes which we call ‘‘doublets’’: see table 2. Table 2. Examples of doublets ergative

translation

extended intr.

chı¨naha choku disi fada kapan mapa panu tako tı¨ tuxa’tsi

cut hit kill, beat put to flight do break exchange bite distribute push

naha iki fa chayo chuda ku’ku elka make detne dama

Case patterns and verb classes in Trumai

297

Below we find sentences where both verbs are used successively. (14)

– ‘‘aka! naide! hi-wan fa hai-tl, ha dua’’ Ah it’s.over 2p-pl.abs kill 1p-e3, 1abs bleed’’ ‘‘‘–Ah! It’s over! You’ve killed me, I’m bleeding.’’’ disi pchetsi-n ale. get.killed fall-3abs one.says ‘He got himself killed and fell, one says.’ (Murder)

(15) Atlat paT-es ku’ku-n ale [.] cooking.pot little-e1 break-3abs one.says [.] ‘He has broken the little cooking pot, they say [.]’ iyi mapa.mapa IYI get.broken ‘It got broken.’ (Anu Bird, RGD 2004) 2.1.5. Three-participant constructions There exist subsets of class 2 (‘‘ergatives’’) and class 3 (‘‘extended intransitives’’) which allow a third argument. The recipient or addressee role is marked by the same cases as the ‘extended to core’ argument -s/-ki/-tl. (16) hine-k hit kı¨tı¨ hai-tl 3p-erg arrow.abs give 1p-e3 ‘He gives me an arrow.’ (AMB notes) (17) Ha deTne kain k’ate-s kiki wan-ki. 1abs distribute foc fish-e1 man pl-e2 ‘I am distributing fish to the men.’ (RGD lexicon) Verbs displaying this behaviour include: kı¨ T ¨ı ‘‘give’’, pap ‘‘retrieve’’, tı¨ ‘‘distribute’’, xoma ‘‘teach’’, pechma ‘‘insert’’, waymi ‘‘tell’’. 2.1.6. Minor classes and patterns There is also a very small class of verbs (sa ‘‘dance’’, hain ‘‘celebrate’’) both of whose arguments are unmarked: (18) wan sa-n tawarawana pl dance-3abs tawarawana ‘They are dancing the tawarawana.’

298

Aurore Monod Becquelin and Ce´dric Becquey

Finally, there is a construction often described as ‘‘object incorporation’’, which occurs when the possessor of the absolutive argument of an ergative verb is coreferent with the agent of the process. This device is restricted to a few verbs, which exhibit a type of lexicalization whereby the absolutive argument of the ergative construction combines with the verb to produce a monovalent compound verb; the agent takes the unmarked absolutive form. (19) mut.pupe-n garment.take.o¤-3abs ‘He is undressing.’ 2.2. Cases In the previous examples we considered the di¤erent verb classes; we now examine the associated case markers. 2.2.1. Unmarked case This case encodes various semantic roles (agent, experiencer, patient. . .): ha fa ‘‘I kill’’, ha ma ‘‘I eat’’, ha xu’tsa ‘‘I see’’, ha yotl ‘‘I sleep’’, ha padi ‘‘I am waited for’’. It is used for the unique argument of a monovalent verb (intransitive), or for the unmarked argument of a bivalent verb, whether this is the absolutive in an ergative frame (patient-oriented: ha padi hinek ‘‘I get waited for by him’’) or the absolutive in an extended intransitive frame (agent-oriented: ha fa hinetl ‘‘I kill him’’). In morphosyntactic terms the unmarked argument, whether labelled absolutive or nominative according to the theoretical interpretation, encodes a proto-agent in a ‘‘pseudo’’ extended intransitive construction, or a proto-patient in the ergative construction, and any role in the intransitive construction. 2.2.2. -(V)k: ergative case This case encodes not only the agent of ‘‘ergative’’ verbs, but also the causer in causative constructions involving verbs of classes 1, 2, and 3. (20) Yakı¨r etsi ka-wan-ek salt carry 1excl-pl-erg ‘We are carrying the salt.’ (Tapir)

Case patterns and verb classes in Trumai

299

(21) talel ¨ıch wapta-ka totsit-ik door catch fall-caus totsit-erg ‘The totsit bird causes the door catch to fall.’ (myth AMB msc.) (22) hai-ts chı¨in Atawaka-k mapa-ka-n. 1p-erg foc/tens Atawaka-erg break-caus-3abs ‘I made Atawaka break it.’ (RDG 2003: 211) 2.2.3. Extension to core Extensions to core, recipients and locatives, which share the same set of markers, can all be subsumed under the label ‘goal’ in the notional ((23), (25), (28), (29)), spatial (26) and temporal (27) domains. These cases have been given di¤erent interpretations in previous descriptions of Trumai, see table 3. Table 3. Specification of cases according to AMB (1975) and RGD (2004) Marker

AMB

RGD

-tl

Dative

Individualized, identifiable and salient dative

-ki

Targeted accusative

Individualized, non-identifiable and non-salient dative; non-individualized, identifiable and non-salient dative

-s

Indefinite accusative

Non-individualized, non-identifiable and non-salient dative

As in many languages, the object of an extended intransitive verb bears the same case marker as certain locatives – answering the ubi/quo questions, -s (‘at’) and ki (‘to’) – which means that it is sometimes di‰cult to decide whether one is dealing with an intransitive verb with locative (23a) or an ‘‘extended intransitive’’ (23b).4 4. Here is an interesting example where the informant obviously plays on this ambiguity: hai.hen, sone-n ale hen misu-s misu-ki; then, drink-3abs one.says then water/river-E1 water/river-loc2 ‘Then, one says that they drink water at the river.’ Ina.ik otl-e. kodaka-ki ukan right.there sleep-3abs. dawn-loc2 again ‘And they slept immediately. The next day’

300

Aurore Monod Becquelin and Ce´dric Becquey

-(V)s: Extension to core (E1); but also spatio-temporal locative (LOC1) (23) a.

b.

– ‘‘ha kaxmi tsimo-s’’ 1abs go.to.forest cipo-loc1 ‘I am going to the forest for cipo.’ – ‘‘ma ka-wan fa kawa tsimo-s’’ Let’s.go! 1excl-pl.abs beat go cipo-e1 ‘Let’s go and beat the cipo!’

(24) kawixu-ami-aduru-s hen yaw si de rain-voice-noise-loc1 then people burn already ‘When the thunder rumbles, people have already burned [the plantation].’ (Plant.) -ki: Accusative, dative (E2); spatio-temporal locative (loc2) (25) ure.ure silo-ki api kakda-n parrot crown.of.feathers-e2 take early.in.the.morning-3abs ‘At dawn he took the parrot’s crown of feathers.’ (Orig. day) (26) yaw si de kut’a-ki. people burn already plantation-e2 ‘People have already burned (in) the plantation.’ (Plant.) (27) tach kodaka-ki ha yatxa. more tomorrow-loc2 1abs work.on.the.plantation ‘Again the following day, I work on the plantation.’ (Plant.) -(V)tl: Dative, Accusative (E3) (28) tach ats’aek tı¨ hen tach kiki-tl. more manioc.gruel distribute then more man-e3 ‘The manioc gruel is distributed to the men once more.’ (Plant.) sone-n misu-ki. hai.de otl-e. kodaka-ki drink-3abs water/river-E/loc2 already sleep-3abs dawn-loc2 ‘they drink (at the river? water?). They are already sleeping. The next day,’ sone-n misu-ki, sone-n misu-s. drink-3abs water/river-E2, drink-3abs water/river-E1 ‘they drink ‘‘this water’’, they drink ‘‘some water’’.’ (Payetan) We could assign the generic term ‘‘goal’’ to these two labels (E/LOC), were it not for the fact that the valency of the verb entails a required core argument in one case and a circumstantial one in the other.

Case patterns and verb classes in Trumai

301

(29) kayapi-tl wan yenye-n Kayabi-E3 pl imitate-3abs ‘They imitate the Kayabi.’ (AMB notes) (30) xu’tsa tak Trumai-s pı¨tı¨k-etl Trumai ami ‘‘kuT’’ see neg Trumai(abs)-loc1 monkey-E3 Trumai say kuT ‘When the Trumai can’t see the monkey, the Trumai say kuT (‘‘water spirit’’).’ (AMB notes) 2.2.4. Oblique cases -tam: Comitative (COM) (31) ha pine-tam ha lax kawa 1 friend-com 1abs hunt go ‘With my friend, I go hunting.’ (32) hai-tl-tam werew waki kı¨tı¨ 1p-e3-com a.little imp give ‘Give me a little bit!’ (Sun and Moon) The comitative marker is su‰xed to nouns but also to personal pronouns marked by -tl; case stacking occurs only on personal pronouns, as in (32), and is impossible with the case markers -s and -ki. -letsi: instrumental (INSTR) (33) intsatske yaw cho xu-letsi de. next people perforate needle-instr already ‘And then one perforates with a needle.’ (Harpoon) -ita: Allative (ALL) (34) ni-s sela-ita hen pech-lapchı¨-n ale ¨ıdı¨cha-tam here-loc1 stern-all then run-cont-3abs one.says bow-com ‘Here, he runs towards the stern with his bow, one says.’ (Murder) -lots: Ablative (ABL) (35)

– ‘‘hamosin hi wax, Katsini?’’ – ‘‘ha petl-lots’’ kale – where 2abs fart Katsini? – 1abs behind-abl cit ‘– ‘‘Where do you fart, Katsini ?’’ – ‘‘From my behind.’’ ’ (Fish festival)

302

Aurore Monod Becquelin and Ce´dric Becquey

-n: illative/allative temporal and spatial (LOC) (36) tamula-n misu-n xuma-ktsu-n zenith-loc water-loc bathe-towards.the.river-3abs ‘At midday, he bathes in the river.’ (AMB notes)

3. Limitations of the ergative analysis The analysis of Trumai as displaying an ergative system combined with the existence of an extended intransitive verb class was motivated mainly by the desire to take into account the morphosyntactic alignment of verb class 2, intuitively considered to be canonical and predominant among the two-place verbs. This analysis, following Dixon (1994: 123), asserts that the ergative verbs are those with the highest transitivity, possessing two nuclear arguments (ergative and absolutive). According to this account the extended intransitive verbs have an absolutive subject (identical in form to the subject of a monovalent intransitive verb) and an additional argument marked as the recipient/addressee in a three-participant construction (see 2.1.5). But this interpretation does not entirely take the empirical data into account. An examination of Trumai vocabulary and texts reveals a quite di¤erent situation from that described by Dixon: the extended intransitive construction is greatly underemphasized by Dixon (who talks of ‘‘minor’’ transitivity and an ‘‘additional’’ argument), while textual analysis reveals its statistical weight and its often high semantic transitivity. 3.1. Is the ergative alignment dominant? If the purpose of examining the morphosyntax of Trumai is to describe the verbal system as a whole, or at least its dominant alignment, then an observation of the distribution of two-place verbs in the lexicons (fig. 1) and texts (table 4) collected for Trumai shows that ergative verbs are not predominant. Setting aside the fluid verbs, we can see from fig. 1 that in the vocabulary extended intransitive verbs predominate slightly over verbs of the ergative class. This superiority can also be observed in texts, where verbs expressing one or more arguments (see below on the optionality of the different cases) and presenting extended intransitive alignment are distinctly more numerous than those with ergative alignment.

Case patterns and verb classes in Trumai

303

Figure 1. Distribution of two-place verbs in the lexicons of AMB (190 verbs) and RGD (274 verbs) Table 4. Distribution of two-place verbs in texts Text5

Extended intransitive verbs

Ergative verbs

Plantation

75.68% 28/37

24.32% 9/37

Murder

75.76% 75/99

24.24% 24/99

Tapir

58.82% 20/34

41.18% 14/34

Orig. day (N.)

58.06% 36/62

41.94% 26/62

Orig. day (W.)

60.38% 32/53

39.62% 21/53

5. We have chosen texts which are as diverse as possible in order to reduce distortion in the counting, which is still at a preliminary stage. ‘‘Plantation’’ (Plant.) is an ethnotext recounting the preparation of a field of manioc; ‘‘Murder’’ is the account by a Trumai of the murder of one of his relatives (though we shall see later that the title of this narrative, given by the author who collected it, should be changed: it is not a murder narrative but a political text, as shown by the syntactic behaviour of its verbs); the story of the ‘‘tapir’’ recounts the origin of ritual songs; the two narratives concerning ‘‘the origin of the day’’ are two myths recounted forty years apart, one collected by AMB (orig. day N.) in 1967 and the other by E. de Vienne (orig. day W.) in 2005.

304

Aurore Monod Becquelin and Ce´dric Becquey

If these facts call into question the typological classification of this language as ergative, they also run counter to a description which attributes extended intransitivity to non-ergative verbs. Dixon points out: ‘‘An important point to note is that these extended subclasses are always relatively minor. Most transitive verbs are canonically transitive with two core roles; only a small number will be extended transitive (or ditransitive) with an additional role’’ (1994: 123). He repeats this claim with Aikhenvald: ‘‘In a few languages (e.g. Tonga, Trumai), there is also an extended intransitive clause type, with S and E [. . .] In every language in which they occur, extended intransitives and extended transitive clause types are greatly outnumbered – in dictionary and texts – by plain intransitive and plain transitive’’ (Dixon and Aikhenvald 2000: 3). It appears that, grounded though it is on data from a wide range of languages, a confrontation of this cross-linguistic generalization with the Trumai facts obliges us to rethink these supposedly ‘‘additional roles’’. The second argument of an extended intransitive verb is not a mere adjunct, which may or may not be lexically expressed; it is required by the meaning of the verb, and there are therefore di¤erences between this argument and optional elements such as ‘‘comitative’’, ‘‘locative’’, ‘‘instrumental’’. In any case, with verbs of this class a lexical object can always be expressed. 3.2. Transitivity and verb classes Dixon’s reasoning here is based on a scalar view of the transitivity of the di¤erent verb classes; on this scale the class of ergative verbs is held to be systematically higher-placed than the class of extended intransitives. Bearing in mind a definition of prototypical semantic transitivity such as that proposed by Givo´n (1990: 565–566) (see also Hopper and Thompson’s (1980) scalar concept of transitivity and Dowty’s (1991) conception of semantic transitivity), it is easy to see that verbs such as ‘‘kill’’, ‘‘cut’’, ‘‘break’’, ‘‘destroy’’ can be found in both classes of Trumai verbs, as well as in the class of fluid verbs (see table 5). Table 5. Classes of prototypically transitive verbs Trumai

translation

class

chı¨ naha (o) fa disi pı¨rew fatla

cut cut kill, beat kill, beat destroy, finish pierce

extended intransitive fluid extended intransitive ergative fluid ergative

Case patterns and verb classes in Trumai

305

In fact, as Dixon himself observes, semantic transitivity and bivalence are not coextensive. Semantic transitivity and morphosyntactic class membership, which are often observed to display a correlation among the languages of the world (Tsunoda 1981, 1985), appear to be independent in Trumai. In a quick survey of the lexicon, classifying verbs according to the semantic fields they encode, no such correlation appears; the two classes coexist in each conceptual domain, with the notable exception of verbs of perception and cognition.6 The semantic fields presented here are based on the works of Tsunoda (1981, 1985) and the Wordnet lexical database developed by the Princeton University Cognitive Science Laboratory as a tool applicable to all of the world’s languages. Here, they are intended solely as very general indicative labels, delimiting broad conceptual domains which can serve as input to a preliminary classification; they are not intended to represent the semantic domains of Trumai. Semantically low-transitive verbs, such as cognizer or experiencer verbs, select particular patterns in Trumai (as in most languages) and are, therefore, not relevant for typological classification. On the other hand, the list of extended intransitive verbs shows that the semantic categories recorded for them by Guirardello do not cover all available conceptual domains: they do indeed include verbs of perception, mental activity and verbs of ‘‘contact’’ or ‘‘movement’’ (RGD 2003: 202), but all the others are considered ‘‘habitual activities’’, such as ‘‘eating’’, ‘‘drinking’’ or ‘‘cooking’’ (RGD 2002: 6). Furthermore, what are we to make of verbs which show the same alignment but do not refer to perception or cognition, or to contact or routine actions, such as look for, count, arrange, trust, explode, skip along, move away, fight, kill, tear, bury, avenge, steal, etc.? At the same time, a large number of verbs for ‘‘peaceful’’ (i.e. not prototypically transitive) actions are associated with the ergative construction, such as advise, tickle, touch lightly, paint, wait for. . . This apparent dissociation between the referential process, the protoroles and the morphosyntax of the verb is not without consequences for the determination of the valency of the di¤erent Trumai verb classes.

6. As has been noticed by some linguists, the semantic domains associated with the extended intransitive construction are ‘‘seeing, hearing, liking and wanting’’, and for the extended transitives, ‘‘giving, showing, telling’’ (Dixon and Aikhenvald 2000: 3).

306

Aurore Monod Becquelin and Ce´dric Becquey

Figure 2. Distribution of verb classes in all the lexicons of AMB and RGD according to a number of conceptual domains

General definitions, such as that provided by Creissels (2004: 3), state that ‘‘a construction will . . . be denoted as transitive if and only if it contains a verb accompanied by two nominal terms with which that verb is constructed in the same way as a prototypical action verb, with the two nominal constituents representing the agent and the patient, in a construction where agent and patient are both treated as nuclear syntactic terms’’ (our translation). But the determination of such prototypical action verbs is extremely problematic in Trumai. The only means by which we can allot valency to verb classes (2 and 3) is by determining the degrees of nuclearity of the arguments involved in the di¤erent classes, using purely morphosyntactic criteria. The criteria most frequently used to di¤erentiate between core and oblique arguments are the following (see, for example, Lazard 1994: 68–

Case patterns and verb classes in Trumai

307

77, Creissels 2006: 273–275): verb agreement, obligatory character, morphological marking, and word order constraints. We can observe in Trumai a strong asymmetry between the absolutive argument and the others – the absolutive argument governs verb agreement, and it is the only obligatory argument, the only argument that is not marked morphologically and the only argument subject to a strict word order constraint. At this stage, nothing allows us to distinguish morphosyntactically between the ergative, accusative and dative cases and oblique arguments such as temporal and spatial locatives. 3.3. Argument omission In table 6, we show the asymmetry between the optionality of the extended argument in class 3 and of the ergative argument in class 2. Table 6. Omission of the main case-marked argument % of ‘‘extended intransitive’’ verbs (class 3) without ext. to core argument

% of ‘‘ergative’’ verbs (class 2) without ergative argument

Murder

29.33% 22/75

54.17% 13/24

Plantation

46.43% 13/28

100.00%7 9/9

Orig. day (N.)

50.00% 18/36

96.15% 25/26

Orig. day (W.)

25.00% 8/32

42.62% 10/21

Tokens in:

Given the great frequency of ergative argument omission, we might be tempted to translate ergative verbs by passives. However, in the absence of a morphological passive marker in Trumai, in translations we prefer the expression ‘‘get (oneself ) þ past participle’’ to the canonical English passive. In light of the fact that all arguments apart from the unmarked one are only optionally expressed, it might be concluded that the language con7. We cannot explain the fact that in this ethnotext, where the speaker is describing the successive activities he performs to make a plantation, there is not a single example of the agent in the ergative case, whereas fewer than half of the ‘‘extended to core’’ arguments go unexpressed.

308

Aurore Monod Becquelin and Ce´dric Becquey

tains only intransitive verbs. It is worth noting that the frequent absence of the ergative argument in texts is strong evidence for the preeminence of the absolutive argument (the unique argument of intransitives, the core argument of ‘‘ergative’’ verbs without the ergative argument and the unmarked argument of ‘‘extended intransitive’’ verbs). We consider the relationship between the optionality of the ergative argument in class 2 and the optionality of the ‘‘extension to core’’ argument in class 3 to require a more thorough examination, but table 6 shows a much higher percentage of omission for ergative arguments than for extension to core arguments.

4. Revisiting the Trumai data Given the above observations based on the textual data, the analysis of Trumai as an ergative language on the basis of the ergative morphosyntactic alignment of some of its verbs is unsatisfactory. We are dealing with a language which does not in fact display any predominant alignment in the lexicon, whether ‘‘ergative’’ or ‘‘extended intransitive’’ (see Fig. 1). Throughout the texts (see table 4), extended intransitive is far more frequent than ergative alignment. We have also noticed that class 2 and class 3 show a similar relationship between their core arguments. Finally, the distribution of case assignment appears to be independent of the semantic transitivity of the verb: semantically high-transitive verbs are distributed almost equally between an ergative construction and what we now identify, by contrast, as an accusative construction. Liberated from the constraints imposed by its previous typological assignment, we are now in a position to reconsider the Trumai verbal system as a whole. 4.1. Determination of the syntactic subject Based on the works of Lazard (1994), Keenan (1976) and Mel’cˇuk (1983), a preliminary approach shows that the unmarked argument in the three constructions (intransitive, ergative and extended intransitive) fulfils many of the prerequisites for defining the syntactic subject. The unmarked argument is obligatory in every construction8; it governs verbal inflection (-n/-e in the 3rd person); it is not marked morphologically; it represents the 8. In our texts, the total absence of the absolutive argument occurs only in impersonal constructions such as kawixukla ‘‘it is raining’’. A few other cases require clarification. In a vocabulary survey, some class 2 verbs expressed

Case patterns and verb classes in Trumai

309

prototypical agent in the ‘‘extended intransitive’’ construction; it is the addressee of the imperative9 in the ‘‘extended intransitive’’ construction; it commands reflexive, reciprocal and middle use (in the rare cases where the deletion of the ergative argument gives rise to a middle interpretation, see (37)); and it is placed according to syntactic constraints of linear precedence. (37) a.

b.

Kumaru-k ha tı¨chı¨. Kumaru-erg 1abs scarify ‘Kumaru scarifies me.’ (RGD 2003: 22) ha tı¨chı¨. 1abs scarify ‘I am scarified (by someone), I scarify myself.’ (RGD 2003: 22)

On the other hand, the ergative-marked argument fulfils only a few functions: it represents the prototypical agent, and is the addressee of the imperative, but it is not obligatory, does not govern verbal inflection, is morphologically marked, does not command reflexive, reciprocal and middle use, and is less constrained by word order. This inventory of properties is not exhaustive, but the asymmetry between the two sets of subject properties in favour of the unmarked argument raises the question of the latter’s actual syntactic functions and those of the ergative argument. The causative construction in Trumai, especially when applied to class 2, see (22), is problematic because it shows two instances of ergative marking, which on the ergative interpretation would be interpreted as two in the 1st person ergative do not take the extralocutive su‰x that normally appears. In her study on valency changes, Guirardello Damian mentions that the absolutive can be omitted when U of the intransitive and S of the extended intransitive (in her terms) bear a semantic e¤ect of genericity: an event occurs, but one does not know who or what produced it. Tsixu’tsa kain iyi-n ‘‘it is cold’’ A Tsixu’tsa kain iyi ‘‘it (the weather) is cold’’. This analysis is not conclusive, insofar as these phrases include an element – iyi – whose syntactic and semantic value is not known for certain. 9. The allocutor of the imperative has always been analysed as the ergative argument. However, it is surprising that the imperative particle changes according to the semantic nature of the patient (human wa or non-human waki). Moreover, the ‘‘absolutive’’ argument is never mentioned in this construction. Lastly, the wana form, which is characteristic of ‘‘extended intransitive’’ (and intransitive) verbs, can be employed with ‘‘ergative’’ verbs in a medio-passive use where there is identity between the agent and the patient.

310

Aurore Monod Becquelin and Ce´dric Becquey

separate ergative subjects, but ‘‘doubling on subjects is unknown in causative constructions’’ (Comrie 1989: 178). In any case, this construction is uncommon in Trumai and does not seem to be of central importance in the identification of the subject. One way to pin down the identity of the subject is to scrutinize the behaviour seen in anaphora. Anaphors are generally controlled by the subject: ‘‘the subject case tends to code the most important, recurrent, continuous topic. We may call it the primary clausal topic’’ (Givo´n 1984: 138, cited by RGD 2004). However, in Trumai, the control of anaphoric elements is pragmatic, and the antecedent is neither the A nor the P of the previous clause but the topical noun phrase which for pragmatic reasons is the most ‘‘plausible’’ in the logic of the discourse (see for demonstration RGD 2004). Given that anaphora is not controlled syntactically, it constitutes a useful device for determining the intentions of the speaker by considering the switch of topics within narrative texts. 4.2. Marking the topic A study making use of this feature was outlined by RDG (2004): ‘‘it seems that in a Trumai transitive clause [i.e. verb class 2], it is the Ergative that codifies the topical participant. If this is indeed the case, it would mean that the Ergative is the Subject of transitive clauses in Trumai’’. In a sense her presentation slants the results somewhat, as it counts all other verbs (class 1, 3 and fluid) as one- or two-place intransitives. But in her sample this topical participant occurs as ergative 6 times, and 5 times as absolutive with a second argument; in none of the clauses is it presented as a full noun phrase. Our preliminary study was conducted on a diverse set of texts (an account of a murder, an ethnotext on a plantation, and a collection of myths). This study brings to light a strong correlation between the discourse topic and the unmarked argument, as the following examples illustrate (by ‘topicalization’ we mean, for the time being, the arrangement of sentence structure according to the discourse roles chosen by the speaker in context). In a text narrating the murder of a Trumai (Javaritı¨), which the narrator attributes to a neighbouring group, the Kayabi, the following excerpt relates the events leading up to his death from the victim’s point of view. The ‘‘victim’’ participant, whatever his semantic role, systematically appears unmarked, as the primary clausal topic and the syntactic subject (appendix, lines 67 to 75).

Case patterns and verb classes in Trumai

311

Table 7. List of the 13 verbs in ‘‘Murder’’, line 67–75 Verb

verb class

participant abs

translation

participant erg/e

waymi waymi lax waymi chı¨ iki iki ora (i)ki pech pech pech ku

extended extended extended extended intr. ergative ergative intr. ergative intr. intr. intr. intr.

Javaritı¨ Javaritı¨ Javaritı¨ Javaritı¨ Javaritı¨ Javaritı¨ Javaritı¨ Javaritı¨ Javaritı¨ Javaritı¨ Javaritı¨ Javaritı¨ Javaritı¨

speaks speaks fishes/hunts speaks rows gets aimed at gets aimed at shouts gets aimed at runs runs runs shouts

Kayabi Kayabi Kayabi Kayabi

intr. intr. intr. intr.

Kayabi Kayabi Kayabi

In the subsequent passage (lines 75 to 82), the Kayabi constitute the discourse topic: accused of the murder, they are in turn promoted to the syntactic function of subject, as the main theme of this speech argues for revenge against them. This promotion entails the selection of verbs belonging to a di¤erent class from those used in the passage above for the same semantic content (e.g. iki/chomta ‘‘aim at’’). Table 8. List of the 11 verbs in ‘‘Murder’’, line 75–82 Verb

verb class

participant abs

translation

fa fa fa (o) fa damtsi ka’chı¨ pita chomta chomta choku chomta chomta

extended extended extended extended ergative intr. extended extended extended extended extended

Kayabi Kayabi Kayabi Kayabi Javaritı¨ Kayabi Kayabi Kayabi Kayabi Kayabi Kayabi

kill kill kill kill gets chased go out aim at aim at hit aim at aim at

intr. intr. intr. intr.

intr. intr. intr. intr. intr.

participant erg/e

Javaritı¨ Javaritı¨ Kayabi

Javaritı¨ tree

312

Aurore Monod Becquelin and Ce´dric Becquey

Javaritı¨ only appears once in this sequence as the syntactic subject (i.e. the unmarked argument) of the verb damtsi, an ergative verb. This topical marking is made possible by resorting to two di¤erent phenomena in the language, phenomena constrained by the syntactic and lexical resources of Trumai. Our analysis shows that the text we had previously entitled ‘‘the murder of Javaritı¨ ’’ is in reality told from the perspective of a political conflict between two groups – both antagonists being treated in turn as the topic – and would therefore a more appropriate title would be ‘‘Trumai political evaluation of an interethnic murder’’. The fine-grained analysis of a discourse device, topicalization, thus reveals the true meaning (purpose) of the text. 4.3. The di¤erent interpretations of the Trumai system Despite its preliminary state, the above study demonstrates the need to rethink the Trumai verbal system. Drawing notably on the universal syntactic-semantic primitives of Dixon (1994: 6–9), relabelled by Creissels (2006: 300) as A(gentive), P(atientive) and U(nique argument), we now propose to summarize the di¤erent interpretations of this system put forward so far and to suggest a new one. 4.3.1. Once again: the ergative interpretation? The ergative interpretation, depicted in table 9, now appears to be unsustainable. The consequences imposed by an interpretation of this kind, such as the assignment of the syntactic function of subject to the ergative argument and the di¤erential treatment of the morphosyntactic properties of ergative (nuclear) and extended to core arguments (see (1)), are not warranted on the basis of the data observed in Trumai texts. Table 9. Ergative interpretation Core argument 1

Case marker

Class 1

U Absolutive

none

Class 2

A Ergative

-(V )k

Class 3

A Absolutive

none

Core argument 2

Case marker

P Absolutive

none

Extended to core argument

Case marker

E

-(V )s -ki -(V )tl

Case patterns and verb classes in Trumai

313

4.3.2. The accusative interpretation? The accusative interpretation (see (2)) appears to solve the problem of the assignment of the function of subject to the unmarked argument. However, it requires us to posit the existence of verbs with a ‘‘passive meaning’’, (i.e., patient-oriented with only optional mention of the agent), reversing the preceding situation with regard to the nuclearity of the marked arguments in -(V)k without providing a satisfactory solution. Table 10. Accusative interpretation Core argument 1

Case marker

Class 1

U Nominative

none

Class 2

P Nominative

none

Class 3

A Nominative

none

Core argument 2

Case marker

P Accusative

-(V )s -ki -(V )tl

Second P Dative

-(V )tl

Oblique argument

Case marker

A Oblique

-(V )k

4.3.3. The ‘all-intransitive’ interpretation? As we saw earlier (section 3.2.), no overt distinction can be observed at the morphosyntactic level which could be used to di¤erentiate between ergative/accusative/dative arguments and oblique arguments. It would therefore be possible to postulate the non-existence of transitive verbs in Trumai – a proposal already made for languages such as Lezgian, for example, by Mel’cˇuk (1983). According to this interpretation, there is no di¤erence between the marked cases in terms of their relationship with ‘macroclass I’. However, there are di¤erences on the semantic and syntactic levels, since some case markers (namely -k, -s, -ki, -tl ) show a preferential relation with semantically transitive verbs and combine symmetrically with agent and patient roles. Our claim that there are two core arguments in both classes is supported by valency changes, such as the middle construction, where the ergative

314

Aurore Monod Becquelin and Ce´dric Becquey

Table 11. The all-intransitive interpretation Core argument

Case marker

Oblique argument

Case marker

Agent

-(V )k

Definite patient, definite target addressee, terminative (locative)

Class I (1/2/3)

Ua/p unique argument

none

-ki

Individualized patient, addressee/ recipient (indefinite or definite)

-(V )tl

Indefinite patient, inessive, temporal locative

-(V )s

Allative

-ita

Ablative

-lots

Comitative

-tam

Instrumental

-letsi

Locative

-n

argument is deleted (see (37b)), and object incorporation (see (19)). One of the two participants, either agent or patient, can be selected in the syntax as the subject according to its topical status (see 4.2 and annexes). The exclusive accessibility of agents and patients to syntactic operations such as those mentioned seriously undermines the ‘‘all-intransitive’’ interpretation. 4.3.4. Closing remarks The ergative perspective removes the E(xtension to core) from core arguments to the periphery; the accusative perspective disregards the ergative argument; the all-intransitive point of view downplays both. The three analyses thus all seem inadequate and inconclusive. We therefore posit the existence of a division between intransitive and transitive verbs. Transitive verbs can be agent-oriented (previously ‘‘extended intransitive’’) or patient-oriented (previously ‘‘ergative’’). The coexistence in a single language of two sets of transitive verbs, one of which shows ergative features and the other accusative features, is well known among Austronesian languages. However, the remarkable characteristic of Trumai is that the ergative/accusative split is lexically governed

Case patterns and verb classes in Trumai

315

for the class of verbs semantically high in transitivity: some semantically high-transitive verbs select the ergative construction, others the accusative construction. Furthermore, unlike in Austronesian languages, this split never involves morphological marking on the verb even if both alignments are available for the same semantic structure (fluid verbs and doublets). Table 12. Alternative interpretation Subject argument

Case marker

Class 1

U unmarked case

none

Class 2 patient-oriented

P unmarked case

none

Class 3 agent-oriented

A unmarked case

none

Complement argument

Case marker

Ergative

-(V )k

Accusative; Dative

-(V )tl

Accusative; Dative

-(V )s

Accusative; Dative

-ki

It appears that one means of interpreting the di¤erence between these two major constructions (patient-oriented and agent-oriented) might consist not in attempting to define the subject with reference to either morphosyntax or semantic roles, but in gauging as precisely as possible the relationship between the two arguments of each construction as seen in texts: is there one in which the patient is closer or more distant (ontologically, contextually or notionally) from the agent (ha ma k’ates/haits atlat mapa ‘‘I eat fish’’/ ‘‘by me, the pot gets broken’’), or more or less a¤ected by the action (haits k’ate naha/ha chı¨ k’ates ‘‘by me, the fish gets cut’’ / ‘‘I cut the fish’’)? What type of change a¤ects the relation when one of the two arguments is not mentioned (ha ma ø / ha padi hine-k ‘‘I eat’’ / ‘‘I get waited for by him’’)? Could the ha ‘‘I’’ of ha padi in fact be semantically ‘‘more’’ agentive than the ergative-marked hine-k ‘‘he’’ (the one waiting), as is potentially reflected in French ‘‘je me fais attendre par lui’’? A syntactic complement representing the patient is marked by di¤erent accusative cases according to semantic properties combining semantic role and characteristics such as definiteness, individuation and number, which could explain the variety of forms available: these relations almost certainly involve factors that we have not yet imagined, and which will only be verifiable after considerable work on texts.

316

Aurore Monod Becquelin and Ce´dric Becquey

5. Appendix: Political evaluation of an interethnic murder This narrative was performed for AMB in 1967 by the son of the Trumai chief, and refers to the murder of one of his uncles a few weeks before; the speaker contrasts his own opinion about the murderers with his father’s, each of them having a di¤erent judgement on interethnic conflicts in the Upper Xingu region. We give detailed interlinear glosses in sections exemplifying points made in the article, especially with regard to the identity of topics and subjects. 067 Kawan amihak kawanki. . . kawa-n ami-hak ka-wan-ki go-3abs say-for 1excl-pl-e2 Kawan hi waymi aehak kale hi waymi aehak kawanki. kawan hi waymi ae-hak kale hi waymi ae-hak ka-wan-ki let’s 2abs tell well-for one.says 2abs tell well-for 1excl-pl-e2 For you to tell us. . . For you to tell us well, for you to tell us well. (waymi: class 3, J.: abs) 068

– ‘‘Huk’anik ha alax kawain ukan huk’anik ha lax kawa-in ukan wait 1abs fish go-foc again heletsis kain ha waymi hi wanki’’ kale hen amin le. heletsis kain ha waymi hi wan-ki kale hen ami-n le and.then foc 1p tell 2abs pl-e2 one.says then tell-3abs one.says –‘‘Wait! I’ll go fishing again and then I shall tell you’’, he said, as reported. (alax: class 3, J.: abs / waymi: class 3, J.: abs)

069 chı¨ lapchı¨nes hen iki pchı¨kidan ale natues, chı¨-lapchı¨-n-es hen iki-pchı¨.kida -n ale natu -es paddle-cont-3abs-loc1 then get.hit-begin-3abs one.says back-loc1 at that moment he went paddling away upstream, then he started to get shot in the back, one says. (chı¨: class 1, J.: abs / iki: class 2, J.: patient, abs; no agent) 070 natues. natu-es back-loc1 in the back.

Case patterns and verb classes in Trumai

317

071 Iki pchı¨kidan ale natues. ‘‘Aka!’’ Kale hen iki-pchı¨.kida-n ale natu -es aka! kale hen get.hit-begin-3abs one.says back-loc1 Oh! one.says then oran ale ‘‘Aka!’’ Ora-n ale aka! shout-3abs one.says Oh! He was shot in the back: ‘‘Ow!’’ He said then he shouted in pain; ‘‘Ow!’’ (iki: class 2, J.: patient, abs; no agent / ora: class 1, J.: abs) 072 hanhak de ha iki hi wanek.’’ han.hak de ha iki hi-wan-ek what.for already 1abs get.hit 2p-pl-erg Why am I getting shot by you? (ki: class 2, J.: patient, abs; K.: agent, erg) 073 Ni selaita hen pech lapchı¨n ale ¨ıdı¨chatam ni sela-ita hen pech-lapchı¨-n ale ¨ıdı¨ch-tam here prow-all then run-cont-3abs one.says bow-com t’ı¨ t’ı¨ t’ı¨ pech laxmin ale hen wı¨ Tı¨ ki. t’ı¨ t’ı¨ t’ı¨ pech-laxmi-n ale hen wı¨ Tı¨-ki onom run-towards.forest-3abs one.says then dry-loc2 He runs to the prow with his bow, as reported, t’ı¨ t’ı¨ t’ı¨ he runs towards the forest, to a dry place. ( pech: class 1, J.: abs / pech: class 1, J.: abs) 074 Pech laxmin; pech-laxmi-n run-towards.forest-3abs He runs towards the forest. ( pech: class 1, J.: abs) 075 haihen waTı¨ma waTı¨ma ku pchı¨kidan ale. hai.hen waTı¨ma.waTı¨ma ku-pchı¨.kida-n ale then noise.noise shout-begin-3abs one.says ‘‘Helein hi fa, Tapiukat? hele -in hi fa Tapiukat? what-foc 2abs kill Tapiukat? and then, he begins to shout, as reported: ‘‘why are you killing me, T.? ( fa: class 3, K.: agent, abs; no patient)

318

Aurore Monod Becquelin and Ce´dric Becquey

076 Han ma’tsik daetl hi wan fa? han.ma’tsi.k dae-tl hi-wan fa? for.what.reason innocent-e3 2p-pl.abs kill? For what reason are you killing an innocent? ( fa: class 3, K.: agent, abs; no patient) 077 Helein de tsiets haitl hi wan fa? hele.in de tsiets hai-tl hi-wan fa? what already why 1p-e3 2p-pl.abs kill? Why are you killing me? ( fa: class 3, K.: agent, abs; J.: patient, acc(usative)) 078 Tsifan chı¨in de hi wanek ma’tsike tsifan chı¨-in de hi-wan-ek ma’tsi.ke thing cop-foc already 2p-pl-erg reason.for kaaletl hi wan ofahak’’ kaale amin ale. kaale-tl hi-wan ofa-hak kaale ami-n ale so-? 2p-pl.abs kill-for so tell-3abs one.says What makes you so violent that you kill? So he said, as reported. ( fa: class 3, K.: agent, abs; no patient) 079 Ina hen damtsi laxmin ale hinak wanek, ina.hen damtsi laxmi-n ale hinak-wan-ek and.then follow towards.forest-abs one.says 3p-pl-erg Then they followed him into the forest, as reported, (damtsi: class 2, J. patient, abs; K.: agent, erg) 080 wan ka’chı¨ pita hen chomtan le. wan ka’chı¨ pita hen chomta-n le pl walk go.out then aim.at-3abs one.says they left the canoe and walked and they aimed at him, as reported. (ka’chı¨: class 1, K.: abs / chomta: class 3, K.: agent, abs; no patient) 081 Chomtan ale la-t’aes hen chokun ale hen chomta-n ale la-t’a-es hen choku-n ale hen aim.at-3abs one.says stay-perf-loc1 then reach-3abs one.says then tlan ¨ı wı¨rki ¨ı wı¨r-ki tlan wood-E2 only They aimed at him where he was, as reported, but then they hit only tree branches. (chomta: class 3, K.: agent, abs; no patient / choku: class 3, K.: agent, abs; tree: patient, acc)

Case patterns and verb classes in Trumai

082 tach chomtat’ kawan ale tach chomtat’ tach chomta-t’ kawan ale tach chomta-t’ more aim.at-perf let’s one.says more aim.at-perf They aimed more and more, as reported. . . (chomta: class 3, K.: agent)

Abbreviations P 1 2 3 3abs abl abs all caus com cont cop E1 E2 E3 erg excl foc foc/tens imp instr i.v. iyi loc loc1 loc2 loc3 n neg

Alternative interpretation first person pronoun second person pronoun third person pronoun Extralocutive marker of the 3rd person absolutive (-e/-n) Ablative Absolutive P nominative Allative Causative Comitative Continuative Copula ‘‘Extension to core’’ (-(V )s) ‘‘Extension to core’’ (-ki) ‘‘Extension to core’’ (-(V )tl ) Ergative Exclusive Focalizer cf. RGD (2003) Imperative Instrumental Intransitive verb Particle iyi Spatial/temporal locative (-n) Spatial/temporal locative (-(V )s) Spatial/temporal locative (-ki) Spatial/temporal locative (-(V )tl ) Myth told by N., Trumai chief, shaman, polyglot and famous singer Negation

319

320

Aurore Monod Becquelin and Ce´dric Becquey

nom obl onom orig.day plant perf pl t.v. Ua/p -(V) vol w

Nominative Oblique Onomatopoeia Origin of daylight myth Plantation text Perfect participle Plural Transitive verb Unique argument of intransitive verb (either agent or patient participant) Epenthetic vowel Volitional Myth told by W., great Wauja´ curer’s wife

References Basso, E.B. 1973

The Kalapalo Indians of Central Brazil. New York: Holt, Rinehart and Winston, Inc.

Becquey, Ce´dric 2007 Le syste`me verbal du trumai (langue isole´e du Xingu, Bre´sil): Description, re´analyse et formalisation dans le cadre HPSG. Master’s thesis, Universite´ Sorbonne nouvelle, Paris 3. Comrie, Bernard 1989 Language Universals and Linguistic Typology. Chicago: The University of Chicago Press. Creissels, Denis 2004 ‘‘Ergativite´/accusativite´ et l’he´te´roge´ne´ite´ des constructions intransitives’’, Cercle linguistique de l’INALCO, Cours de syntaxe ge´ne´rale 2004. Creissels, Denis 2006 ‘‘Syntaxe ge´ne´rale, une introduction typologique’’. Paris: Herme`s Lavoisier. Dixon, Robert M.W. 1994 Ergativity. Cambridge: Cambridge University Press. Dixon, Robert M.W. and Alexandra Y. Aikhenvald 2000 Changing Valency: Case Studies in Transitivity. Cambridge: Cambridge University Press. Franchetto, Bruna and Michael J. Heckenberger 2001 Os Povos do Alto Xingu: Histo´ria e Cultura. Rio de Janeiro: Editora UFRJ.

Case patterns and verb classes in Trumai

321

Friedmann, Naama, Gina Taranto, Lewis P. Shapiro and David Swinney 2008 ‘‘The leaf fell (the leaf ): the online processing of unaccusatives.’’ Linguistic Inquiry 39(3): 355–377. Galva˜o, Eduardo 1953 ‘‘Cultura e sistema de parentesco das tribos do alto rio Xingu’’. Boletim do Museu Nacional, Nova Se´rie, Antropologia 14. Givo´n, Talmy 1984 Syntax: A Functional-Typological Introduction. Vol. 1. Amsterdam/ Philadelphia: Benjamins. Givo´n, Talmy 1990 Syntax: A Functional-Typological Introduction. Vol. 2. Amsterdam/ Philadelphia: Benjamins. Guirardello(-Damian), Raquel 1999 A Reference Grammar of Trumai. Ph. D. diss., Rice University. [20/10/2008] http://scholarship.rice.edu/handle/1911/19387> Guirardello(-Damian), Raquel 2002 ‘‘Caso e relac¸o˜es gramaticais em Trumai.’’ in Ergatividade na Amazoˆnia I, F. Queixalo´s (ed.), Brası´lia: Centre d’e´tudes des langues indige`nes d’Ame´rique, Laborato´rio de Lı´nguas Indı´genas. [25/01/2010]

Guirardello(-Damian), Raquel 2003 ‘‘Classes verbais e mudanc¸as de valeˆncia em Trumai.’’ In Proceedings of the Workshop Ergatividade na Amazoˆnia II, F. Queixalo´s (ed.), Centre d’e´tudes des langues indige`nes d’Ame´rique (CNRS/IRD), Laborato´rio de Lı´nguas Indı´genas (UnB). [25/01/2010] http://celia.cnrs.fr/FichExt/Documents%20de% 20travail/Ergativite/2mGuirardello.pdf Guirardello(-Damian), Raquel 2004 ‘‘Coreference in Trumai.’’ In Proceedings of the Workshop Ergatividade na Amazoˆnia III, F. Queixalo´s (ed.), Centre d’e´tudes des langues indige`nes d’Ame´rique, Laborato´rio de Lı´nguas Indı´genas, Paris. [25/01/2010] http://celia.cnrs.fr/FichExt/Documents% 20de%20travail/Ergativite/3mGuirardello.pdf Guirardello(-Damian), Raquel 2005 DoBeS, Max Planck Institut; [25/01/2010]

Hopper, Paul and Sandra A. Thompson 1980 ‘‘Transitivity in Grammar and Discourse.’’ Language 56: 251–299. Keenan, Edward L. 1976 ‘‘Towards a Universal Definition of ‘Subject’.’’ In Subject and Topic, Charles Li (ed.), 303–333. New York: Academic Press. Lazard, Gilbert 1994 L’actance. Paris: PUF.

322

Aurore Monod Becquelin and Ce´dric Becquey

Mel’cˇuk, Igor 1983

‘‘Grammatical Subject and the Problem of the Ergative Construction in Lezgian.’’ In Proceedings of the Second Conference on the Non-Slavic Languages of the USSR, Slavica, 1983. Vol. 5. 246–293. Malchukov, Andrej L. 2005 ‘‘Case Pattern Splits, Verb types and Construction Competition.’’ In Competition and Variation in Natural Languages: The Case of Case, Mengistu Amberger and Helen de Hoop (eds.), 73–117. Oxford: Elsevier. Monod Becquelin, Aurore 1975 La pratique linguistique des Indiens Trumai (Haut Xingu, Mato Grosso, Bre´sil) [Langues et civilisations a` tradition orale, 9, 1]. Paris: SELAF. Monod Becquelin, Aurore 1976 ‘‘Classes verbales et constructions ergatives en trumai.’’ Amerindia 1: 117–143. Monod Becquelin, Aurore 1977 ‘‘Les amants punis: conte trumai (Haut-Xingu, Bre´sil).’’ Amerindia 2: 163–173. Monod Becquelin, Aurore 1987 ‘‘ ‘Les femmes sont un bien excellent’ ’’: vision des hommes, eˆtre des femmes dans le Haut Xingu.’’ Anthropologie et Socie´te´s (Que´bec, Universite´ de Laval), XI(1) [nume´ro spe´cial: Enjeux et contraintes, discours et pratique des femmes]: 121–136. Monod Becquelin, Aurore and Raquel Guirardello 2001 ‘‘Histo´rias trumai.’’ In Os Povos do Alto Xingu. Histo´ria e Cultura, Bruna Franchetto and Michael Heckenberger (eds.), 401–443. Monod Becquelin, Aurore, Raquel Guirardello Damian and Emmanuel de Vienne 2008 ‘‘Working together. The interface between researchers and native people.’’ In Lessons from Documented Endangered Languages, K. David Harrison, David S. Rood and Arienne Dwyer (eds.), 43–66. Amsterdam/Philadelphia: John Benjamins. Murphy, Robert F. and Buell Quain 1955 The Trumai Indians of Central Brazil. Locust Valley, N.Y., J.J. Augustin. Steinen, Karl von den 1942 O Brasil Central: expedic¸a˜o em 1884 para a explorac¸a˜o do rio Xingu´. Traduc¸a˜o. Sa˜o Paulo: Companhia Editora Nacional (Brasiliana, se´rie extra [grande formato], 3). Tsunoda, Tasaku 1981 ‘‘Split case-marking patterns in verb-types and tense/aspect/mood.’’ Linguistics 19: 389–438. Tsunoda, Tasaku 1985 ‘‘Remarks on transitivity.’’ Journal of Linguistics 21: 385–396.

Ergativity in the Adyghe system of valencychanging derivations Alexander Letuchiy 1. Introduction In this paper I will analyse the syntactic properties of valency-changing derivations and other syntactic processes in Adyghe (a language of the West Caucasian family spoken in the Republic of Adygheya and the Krasnodar region of Russia, and also in some countries of western Asia such as Turkey). My aim is to determine whether these processes testify to syntactic ergativity or accusativity in Adyghe, or whether they in fact shed no light at all on the question of Adyghe alignment behaviour. In traditional descriptions, such as Rogava and Kerasˇeva (1966), Kumakhov (1971), and Zekox (2002), it is taken for granted that Adyghe has ergative alignment. This is due to the fact that Adyghe is a morphologically ergative language (see below). As I will show, the case marking of verb arguments and the system of cross-reference markers are indeed organized ergatively. However, with the exceptions of Serdobol’skaya (2007) and Lander (2009), scholars have not considered the syntactic aspects of ergativity in Adyghe. In the present paper, I base my analysis of syntactic ergativity on the evidence of valency-changing derivation only. I choose not to consider other pivot properties related to ergativity/accusativity (coordination reduction, relativization, subordinate clauses etc.; see Dixon 1994; Van Valin and LaPolla 1997). It seems to me more justifiable to restrict myself to the data presented by derivational behaviour alone, since in a single article it is impossible to analyse the whole range of data related to ergativity in a polysynthetic language like Adyghe; moreover, the valencychanging derivational system may be organized ergatively, for example, while other syntactic processes are organized accusatively, or vice versa. We assume that voice systems and syntactic alignment are closely related. On one hand, many linguists (e.g. Shibatani 1985, Dixon 1994) claim that syntactic alignment is crucial for the the voice system of a language. For instance, it has long been thought that the passive is not a characteristic feature of ergative languages. Though this formulation may

324

Alexander Letuchiy

be too strong, it is at least true that the passive is more characteristic of accusative than of ergative languages. Moreover, the precise nature of ergativity in a given language can be relevant for the description and classification of voice-like phenomena. For instance, if we are able to prove that the agent argument of transitive verbs has syntactic pivot properties (in other words, that the language is syntactically accusative), and there is a syntactic alternation which demotes the agent to indirect object while leaving the absolutive argument una¤ected, we can describe this alternation as a prototypical passive. However, if the language is syntactically ergative (the absolutive argument has subject properties with both transitive and intransitive verbs), the same alternation should be described di¤erently: in this case it lacks the main property of the passive, in that it does not decrease the syntactic status of the subject argument. We will see that there is a problematic case of this type in Adyghe: two of the derivations found in the language (the potential, marked with fe-, and the inadvertitive (involitional), marked with ecˇ’e-) seem similar to the passive voice in syntactic terms. ˙ However, this similarity disappears if we assume that Adyghe is a syntactically ergative language. On the other hand, the accurate description of valency-changing mechanisms is a prerequisite for the analysis of syntactic alignment. Some voices and valency-changing derivations systematically show dependence on the subject properties of arguments: for instance, in the majority of languages, reflexives are controlled by the subject argument. Therefore, a description of the voice system allows us to define whether the language is syntactically accusative or ergative. In section 2, I will sketch the most important features of Adyghe grammar, such as polysynthesis, pro-drop and morphological ergativity. In what follows, each valency-changing derivation is considered. What is crucial for the present analysis is that each derivation will be considered in terms of its relation to syntactic alignment: whether the particular type of syntactic alternation is more characteristic of ergative or accusative languages, and whether or not it shows pivot properties of one argument or another. Note that a similar type of analysis has been carried out for Adyghe in a previous paper (Paris 1987). The important di¤erence between this earlier paper and the present article is that Paris adopts a semantic point of view: she examines which mechanisms in Adyghe play the same functional role as passives in European languages. My aim is to analyse the valency-changing devices of Adyghe not only from the semantic, but

Ergativity in the Adyghe system of valency-changing derivations

325

also, and perhaps more importantly, from the syntactic point of view in order to see which derivational mechanisms are syntactically similar to voice alternations. Matasovic´ (2008) proposes an analysis of transitivity in Kabardian, a language of the West Caucasian family closely related to Adyghe. However, though his analysis is similar in many respects to the present work, Matasovic´ takes into account only causative and antipassive derivations. In section 3, Adyghe valency-changing operations and their ergative/ accusative behaviour are considered. Finally, in section 4 I draw some conclusions concerning the nature of ergativity in Adyghe. Taking into account the fact that Adyghe is a polysynthetic language, we use a non-canonical notion of valency-changing derivation throughout this article. Inasmuch as NPs are not obligatory in Adyghe, whereas all core arguments have to be cross-referenced in the verb, we will mostly describe the impact of each derivation upon the expression of arguments inside the verb form, i.e. upon the system of cross-reference prefixes, although traditionally, as in Comrie (1976), Dixon (1994), Plungian (2000), Testelets (2001), valency-changing derivations are often described in terms of their impact upon the expression of free NP arguments. Let us now review the notions of morphological and syntactic ergativity/ accusativity. A language is morphologically ergative (or accusative) if the morphological coding of verbal arguments, i.e. case-marking and the system of verbal personal prefixes, follows the ergative (or accusative) strategy. For instance, if a language is morphologically ergative, the Actor/ Agent (A) of transitive verbs must be case-marked and cross-referenced on the verb form in a di¤erent way from the Patient (P) of transitive verbs and Single argument (S) of monovalent intransitive verbs. The Patient and the Single argument must, in turn, receive the same case marker and be cross-referenced in the same way. The notion of syntactic ergativity/accusativity refers to the organization of syntactic processes. If the language is syntactically ergative, this means that S and P show the same set of syntactic properties, distinct from those displayed by A in the same syntactic processes (for instance, only S and P, but not A, can be relativized). It is usually required that in a syntactically ergative language, S and P must show more syntactic subject properties than A. In contrast, in a syntactically accusative language, S and A show the same set of syntactic properties, distinct from the properties of P. S and A, but not P, must show subject properties in syntactically accusative languages.

326

Alexander Letuchiy

The term ‘morphological ergativity’ is used in the standard way throughout this paper: Adyghe is a morphologically ergative language because its argument coding system follows the ergative strategy (a specific case marks the agent of transitive verbs,1 while another case form is used both for direct objects and subjects of intransitive verbs). However, I choose to understand the term ‘syntactic ergativity/accusativity’ in a specialized sense: I will say that the given phenomenon displays syntactic ergativity/accusativity if the absolutive argument/the agent of transitive verbs is the pivot for the morphological process involved (and thus, for instance, that morphological reciprocalization provides one piece of evidence for syntactic ergativity because the absolutive argument is in most cases the antecedent of morphological reciprocals). Although the valency-changing derivations under analysis here occur inside the word, the tendencies observed in the present paper are parallel to syntactic processes and tendencies in many other, non-polysynthetic languages.

2. Essentials of Adyghe grammar 2.1. Polysynthesis and pro-drop The most salient feature of Adyghe morphosyntax is polysynthesis. Adyghe is a canonical polysynthetic language. The verb form can encompass a large number of locative, temporal, and modal markers. All verb arguments are cross-referenced on the verb form, which can also take a large number of derivational a‰xes. The language therefore apparently represents the headmarking type, in the terminology of (Nichols 1986). The following example illustrates the richness of Adyghe verb forms: (1) S -q -t-de-p-f -r-a-‚a-‰Ł e-sˇ ’t -‚. 1sg.abs-dir-1pl.io-com-2sg.io-ben-3sg.io-3pl.a-caus-read:ap-aux-pst ‘They were making me read it to you together with us.’2 In (1), the verb contains two applicative markers ( f - ‘benefactive’ and de- ‘comitative’), the causative marker ‚e-, the complex imperfect marker -sˇ ’t -‚(e) (which includes auxiliary stem -sˇ ’t and past marker -‚(e)), 1. Here and below I use the term ‘agent’ to denote the ergative (oblique) argument of transitive verbs, including verbs like le‚W n ‘see’ for which this argument does not really have the semantic role of agent. 2. I am grateful to Yuri Lander for this example.

Ergativity in the Adyghe system of valency-changing derivations

327

and four personal cross-reference prefixes (s - ‘1sg.abs’, t- ‘1pl.io’, p‘2sg.io’, a- ‘3pl.a’). The personal (and all other) markers are organized in slots: each marker has its own position (slot) in the verb form. These positions are not interchangeable, except in some very special cases. The general scheme is very complicated, which is why we restrict ourselves to the prefixal zone for present purposes. Following Smeets (1984) we can represent it as follows: Table 1. Prefixal slots of the Adyghe verb form -9

-8

-7

-6

absolutive argument cross-reference marker

directive prefix

temporal prefix

applicative prefixes together with markers of oblique indirect applicative object

-5

-4

-3

-2

-1

0

oblique indirect object crossreference marker

oblique agent cross-reference marker

optative

negation

causative

root

Henceforth we will refer to slot -9 as the ‘absolutive slot’, and slots -6, -5 and -4 as ‘oblique slots’, sometimes specifying whether the given argument is an agent or an indirect object. The personal markers introduced by applicative derivations are paired with derivational markers: each personal prefix must occur immediately before the marker by which it is introduced. All arguments introduced by applicatives are indirect objects. For instance, in (1) the 2sg indirect object prefix p- is before the benefactive marker fe- (‘for you’), whereas the 1pl IO prefix t- is before the comitative marker de- (‘with us’). It is also important to note that Adyghe is a pro-drop language. The participants are not obligatorily expressed by separate referential phrases – what is genuinely obligatory is their cross-referencing on the verb form by means of personal prefixes. Any pronoun in an argument position (absolutive argument, agent, indirect object) can be dropped. 2.2. Morphological ergativity Adyghe is a morphologically ergative language – in other words, casemarking and the system of verbal cross-reference markers are organized

328

Alexander Letuchiy

ergatively. In the domain of case-marking, intransitive subjects and direct objects can both be marked with the absolutive marker -r, whereas transitive agents take the marker -m; the latter is traditionally called ergative (see Kumakhov 1971; Kumakhov and Vamling 2006), but in what follows we will label it ‘oblique’, with the gloss obl, because the range of its functions is not limited to canonical functions of the ergative. Cross-reference prefixes also group S and P together, as distinct from A: the former roles are indexed by a series of prefixes occupying the leftmost position in the verb form. Third person singular absolutive arguments are cross-referenced with a zero prefix, as psˆasˆe-r ‘girl (absolutive)’ in (2) and (3). The agent argument is cross-referenced in the -4 position of the verb form. Third person singular agents are cross-referenced with the prefix -. ˇ ’ale-m psˆasˆe-r Ø- -le‚ W -‚. (2) C ˙boy-obl girl-abs 3sg.abs-3sg.a-see-pst ‘The boy saw the girl.’ (transitive verb) (3) Psˆasˆe-r Ø-ma-k W e. ˙ girl-abs 3sg.abs-dyn-go ‘The girl goes.’ (intransitive verb) Almost all verbs have an absolutive argument. Two minor verb classes constitute an exception: the so-called impersonal verbs, which have an oblique argument only, and the facilitives/di‰cilitives, which are analysed in detail in 4.5. Notably, Adyghe possesses a class of bivalent intransitive verbs – a rather atypical feature for some morphologically ergative languages. These verbs have a subject and an indirect object. The subject of intransitive verbs takes absolutive marking, as is typical for morphologically ergative languages. The indirect object, just like the agent of transitive verbs, is marked with oblique case and takes the su‰x -m: ˇ ’ale-r psˆasˆe-m Ø-Ø-je-bew -‚. (4) C ˙boy-abs girl-obl 3sg.abs-3sg.io-oblique-kiss-pst ‘The boy (intransitive subject, absolutive) kissed the girl (IO, oblique).’ In a sense, the syntactic type of the verbal construction represented in (4) is the reverse of the canonically transitive one (cf. (2)) as regards casemarking.

Ergativity in the Adyghe system of valency-changing derivations

329

The verb form in (4) is intransitive because it does not bear the agentive 3sg prefix - (cf. (2) with a transitive verb). I will show below that reflexivization and reciprocalization also distinguish between bivalent intransitive and bivalent transitive verbs. The verbal marker je- glossed in (4) and below as ‘oblique’ is not a cross-reference marker. It is a special type of applicative prefix: for many bivalent intransitive verbs, the indirect object is introduced by means of this prefix, though the precise semantics of this marker is very vague – in other words, its function seems to be just to add an indirect object. The oblique case has a wide range of semantic functions. It marks not only agents of transitive verbs and indirect objects, but also complements of postpositions, possessors, and temporal and locative adjuncts: ˇ ’ale-m paje (5) C ˙boy-obl for ‘for the boy (complement of postposition, oblique)’ (6) c f -m j -cˇet w ˙man-obl 3sg.prþposs-cat ‘The man’s (possessor, oblique) cat’ Finally, both the oblique and the absolutive case markers can be dropped when an argument is non-specific or indefinite, as in cˇet w qe-k Wa-R ˙ ‘a cat came’ (cat dir-go-pst) where the absolutive ending is dropped, or cˇ ’ale cˇet w -le‚ W -‚ ‘a boy saw a cat’ (boy cat 3sg.a-see-pst) where ˙the oblique ending on cˇ ’ale ‘boy’ and the absolutive ending on cˇet w ‘cat’ ˙ thus encode not only the syntactic position of are both omitted. They the noun, but also its value in terms of definiteness. This is not typical of morphologically ergative languages.3 As we have seen, all core arguments, including oblique-marked agents and indirect objects as well as absolutive-marked direct objects, are crossreferenced on the verb form. The system of personal prefixes also follows ergative alignment: intransitive S and the DO of transitive verbs are crossreferenced with the same set of prefixes occupying the first position in the verb form. The 3sg absolutive argument is cross-referenced with a zero 3. Of course, the core cases (chiefly the absolutive) can be expressed with zero endings (for instance, the absolutive in Tsakhur). However, if a case form bears non-zero marking, this marker usually cannot be omitted (see Dixon 1994 and Kibrik 2002 for examples).

330

Alexander Letuchiy

prefix (or, alternatively, is simply not cross-referenced on the verb form at all). The 3pl absolutive argument does not control an agreement prefix but is cross-referenced with the (optional) absolutive plural su‰x -x. Transitive agents are cross-referenced by means of a di¤erent set of prefixes from indirect objects and absolutive direct objects: agent markers are situated in the -4 slot in the terms of Smeets 1984. Only agent prefixes have the 3sg form - / j -. We are now in a position to analyse the system of valency-changing derivations in Adyghe. We will not analyse each derivation in detail. Our purpose will be to say a few words about each derivation and to determine whether a given derivation can be regarded as a manifestation of semantic/ syntactic ergativity, or whether, by contrast, it represents an argument against the analysis of Adyghe as an ergative language. 3. Adyghe derivations The system of valency-changing derivations in Adyghe includes both means of valency and transitivity increase (causative, benefactive, malefactive, locative) and means of valency and transitivity decrease (potential, antipassive, facilitive and di‰cilitive). As I have mentioned, although not only derivations, but also some other syntactic processes can be useful as diagnostics of ergativity, in this paper I consider valency-changing derivations only. Let me begin by noting that the causative formation will be excluded from the following analysis. The causative in Adyghe corresponds to what is often called the ‘paradigm case’ of the formation, already detailed in Comrie (1976): the causative derivation introduces a new argument (causer) which becomes a syntactic subject. The subject of the base noncausative verb (causee) occupies the highest vacant position in the hierarchy: S > DO > IO > Oblique objects. This rule is not related to ergativity and cannot be used as a diagnostic for syntactic alignment. 3.1. Antipassive Like many ergative languages (Alutor, Mayan languages, Dyirbal etc.), Adyghe has an antipassive construction (see Arkadiev and Letuchiy 2008 for details). Before describing it, we need to discuss the definition of the antipassive.

Ergativity in the Adyghe system of valency-changing derivations

331

For Cooreman (1994), the antipassive is a voice alternation which operates on transitive verbs and decreases the syntactic status of the initial direct object. The object is either eliminated and cannot be expressed or is demoted to an indirect/oblique object, the verb thus becoming intransitive. The subject (agent) of the initial transitive verb becomes the subject of a derived intransitive verb: in ergative languages this leads to a change in case-marking, as the agent which was marked with the ergative in the base construction receives accusative marking in the new antipassive construction. Cooreman (1994), Testelets (2001), Say (2007) and others claim that one possible motivation for the use of the antipassive is to demote a non-salient or indefinite patient which occupies the privileged direct object position of the base verb. Plungian (2000) distinguishes between two very similar operations which are not always easy to tell apart in a given language. The first is the antipassive proper – a voice which decreases the syntactic status of the direct object, making the verb a bivalent intransitive. The second is the object impersonal – this transformation eliminates the object, making the verb monovalent. In Cooreman’s (1994) sample, there are many examples where the agent can be either expressed or left unexpressed: thus, Plungian’s distinction is not irreproachable for all languages. However, it is useful for Adyghe: some morphological antipassives admit expression of the initial direct object as an oblique indirect object triggering indirect object agreement on the verb, but others do not admit the initial DO expression, or permit it to be expressed only as a non-argument NP. In Plungian’s terms, most Adyghe verbs form the object impersonal, but not the antipassive proper. This means that most verbs are morphologically monovalent in their antipassive form: thus, the initial direct object does not control any cross-referencing prefixes. The verb sˇxen ‘eat (antipassive)’ is morphologically monovalent: the initial direct object can only be marked with the instrumental, as in (7b), and instrumentalmarked NPs do not trigger verbal cross-referencing prefixes. The variant sˇx n in (7a) is a bivalent transitive. (7) a.

b.

Se l Ø-s-e-sˇx . i meat 3sg.abs-1sg.a-dyn-eat.tr ‘I eat meat.’ Se l -cˇ ’e Ø-s-e-sˇxe. ˙ i meat-ins 3sg.abs-1sg.a-dyn-eat.ap ‘I eat meat (feed myself with meat).’

332

Alexander Letuchiy

With some verbs, just as sˇx n and x n (8), the object impersonal is marked by means of a stem vowel change from - to -e: (8) a.

Se pj s’me Ø-s-e-tx . i letter 3sg.abs-1sg.a-dyn-write.tr ‘I write a letter.’

b.

Se s-e-txe. I 1sg.s-dyn-write.ap ‘I write.’ (Arkadiev and Letuchiy 2008)

Although pairs like (8a) vs. (8b) can be analysed as two (equally basic) morphological variants, I find it more plausible to regard (8b) as derived from (8a): there are transitive verbs which do not form variants of the (8b) type, but there are no intransitive verbs with a generic or indefinite object which do not have a transitive variant of the type seen in (8a). This group includes the following lexemes (throughout this article, all verbs will be cited in the masdar (verbal noun) form with the su‰x -n): (9) tx n ‘write’ sˇx n ‘eat’ thacˇ ’ n ‘wash’ ha‰Ł˙ n ‘mill’ etc.

w qebz n pw n g W n t n

‘clean’ ‘sow’ ‘pound’ ‘sew’

Some verbs use Agent-preserving lability to express the same pair of meanings: (10) a.

b.

Xate-r Ø- -pcˇ ’a-‚. ˙ garden-abs 3sg.abs-3sg.a-weed-pst ‘He weeded the garden.’ C f -r mefe.rjen -m Ø-pcˇ ’a-‚e. ˙ ˙ man-abs day.whole-obl 3sg.abs-weed-pst ‘The man weeded all day.’ (Arkadiev and Letuchiy 2008)

In general, lability is very frequent in Adyghe4 (see Gisˇev 1968, Hewitt 1982, Letuchiy 2009b). In (10), the two verb forms di¤er only as regards the presence/absence of the agentive 3SG prefix -. 4. I do not take into account P-lability, in the terminology of Dixon (1994), which is taken to be a voice-like mechanism by Paris (1987). In Letuchiy (2006), I show that the existence of P-lability does not characterize a language as ergative or accusative.

Ergativity in the Adyghe system of valency-changing derivations

333

The list of Agent-preserving labile verbs includes, for instance: zˆ Wen ‘plough’ p w pcˇ ’ n ‘chop’ cˇ ’ec n˙ ‘lay (eggs)’ ˙ ˙ As we can see, in (7b), (8b) and (10b) the verb is morphologically monovalent, although in (7b) there is a peripheral object which is not crossreferenced in the verb form. Thus, the second argument cannot be expressed in the verb form; for the verbs listed so far, Adyghe does not have constructions equivalent to ‘I plough at the field’ or ‘I write at a letter’ in which the second argument controls an indirect object slot. However, two verbs ( je‰Ł en ‘read (intransitive)’ / ‰Ł n ‘read (transitive)’ and jesˆ Wen ‘drink (intransitive)’ / j sˆ W n ‘drink (transitive)’) retain a bivalent structure in the antipassive form:

(11) pcˇ ’en ‘weed’ ˙ ˇ ’en ‘mow’ w pc sˇ’en ˙ ‘sell’

(12) a.

C f -m tx l -xe-r Ø- -‰Ł -‚. ˙ man-obl book-pl-abs 3sg.abs-3sg.a-read-pst ‘A man read the book through’.

b.

Se tx l -m s-Ø-je-‰Ł a-‚. i(abs) book-obl 1sg.abs-3sg.io-obl-read.ap-pst ‘I read a book (for some time)’. (Arkadiev and Letuchiy 2008)

The verb has two agreement slots in both (12a) and (12b). In (12a) we observe transitive morphology: one of the a‰xes occupies the agent slot, and the other is found in the direct object slot. In (12b), one of the a‰xes (s-) is in the absolutive slot, while the zero a‰x marks the indirect object; thus, the construction is literally ‘I read at/on a book’ (the verb has the same structure as seen in (4), where the strictly intransitive verb bew n ‘kiss’ has a subject and an indirect object). It is not clear why the prototypical bivalent antipassive is compatible with these two verbs only. For instance, the verbs ‘eat’ and ‘write’, which are semantically close to ‘drink’ and ‘read’ respectively, form object impersonals and not antipassives. Furthermore, it must be emphasized that for ‘read’ and ‘drink’, the antipassive variant is more frequent than its transitive counterpart: the transitive variants ‰Ł n ‘read through’ and j sˆ W n ‘drink up’ are used relatively rarely. Moreover, the transitive variants of ‘drink’ and ‘read’ have a narrow meaning: they express that the situation is either finished or is to finish soon after the moment of speech. This is not obligatorily the case with all other transitive variants

334

Alexander Letuchiy

of antipassive verbs: for instance, sˇx n ‘eat’ and tx n ‘eat’ do not necessarily imply that the situation is or will soon be finished (see Arkadiev and Letuchiy 2008 for details). In order to state simply that someone is drinking tea or reading a book, the intransitive variant will be used (this fact is in accordance with Hopper and Thompson (1980), where the authors show that perfective situations are semantically more transitive than those which are ongoing). In principle, the existence of the antipassive could be regarded as an argument for syntactic ergativity (see, for instance, Cooreman 1994, Shibatani 1985). This valency-changing derivation is more productive in syntactically ergative languages like Dyirbal than in semantically and morphologically ergative languages such as Nakh-Daghestanian. However, this argument is weakened by the fact that the antipassive is not at all productive in Adyghe. A canonical ergative language should have a productive antipassive which is able to detransitivize most transitive verbs when the need is felt to decrease the status of the absolutive direct object or to eliminate it altogether. Moreover, the same stem alternation - / -e can also mark other valencychanging derivations which are only indirectly related to antipassive function, as in bew n ‘kiss sb.’ vs. bewen ‘kiss (in general)’: (13) a.

b.

ˇ ’ale-r psˆasˆe-m Ø-Ø-je-bew -‚. C ˙boy-abs girl-obl 3sg.abs-3sg.io-oblique-kiss-pst ‘The boy kissed the girl.’ Bewe-n- r j -cˇ ’as. ˙ kiss-msd-abs 3sg.prþposs-love ‘He likes kissing.’ (lit. ‘To kiss is his love’). (Txarkaxo 1991: 41)

This class also includes w ncˇ’ n ‘push sb.’ vs. w ncˇ ’en ‘push’, and pl n ˙ In this case, both ‘look at sth.’ vs. plen ‘look (in˙ a particular direction)’. forms are intransitive. The -variant is a bivalent intransitive verb. For instance, in (13a) the fact that the verb has two arguments is evident from the presence of the oblique indirect object psˆasˆe-m and the oblique argument prefix je-. The e-variant is also intransitive, but monovalent. For instance, in (13b), the verb contains no oblique argument marker. In examples like (13), -e marks not a canonical antipassive, but the elimination of the indirect object. Therefore, the motivation proposed by Cooreman (1994), Testelets (2001) and Say (2007) for the antipassive among other

Ergativity in the Adyghe system of valency-changing derivations

335

functions (demotion of the direct object, the most syntactically privileged argument) is not applicable to (13), where the indirect object is eliminated.5 Thus, in most cases the stem alternation - / -e eliminates an argument, but this is not obligatorily a direct object. This feature shows that the antipassive construction in Adyghe cannot serve as evidence for syntactic ergativity. While in many languages the antipassive formation shows the privileged status of the absolutive argument, the antipassive in Adyghe can eliminate di¤erent types of objects, and is not exclusively related to the status of the absolutive argument.6 3.2. Passive-like valency-changing operations Adyghe does not have a passive in the strict sense of the word. Following Shibatani (1985), and the definition proposed by Aikhenvald and Dixon (2000), I take prototypical passivization to be a voice alternation which demotes the initial subject and often, though not necessarily, raises the initial direct object to the subject position. Passive is considered to be a voice and not a valency-changing derivation since it does not change either the number of arguments or their semantic properties. Passivization is supposed to be a ‘syntactico-pragmatic’ change: simplifying somewhat, it can be said that this syntactic process reflects the fact that the patient is more pragmatically salient than the agent. This is the rationale for the use of the passive construction, and, thus, the patient occupies the subject position. According to Dixon (1994: 147), syntactically ergative languages are often characterized by the lack of a passive. Passive constructions seem to be more widespread in accusative languages, where they serve to decrease the status of the most salient participant (¼the agent argument of transitive verbs). See, however, several papers in this volume where it is demonstrated that passivization is not as uncommon in ergative languages as it is often considered to be. The Adyghe system includes some derivations which can be considered passive-like in nature. They are semantically distinct from the passive: unlike the passive, they add new components to the semantics of the situation. However, syntactically they are very similar to passives in the 5. Of course, as we have already said, the morphological nature of the antipassive marker makes it possible to consider examples like (12) to involve pairs of verbs neither of which is genuinely derived from the other. 6. Matasovic´ (2008: 62), arguing that the antipassive in Kabardian is not a typical antipassive, notes also that many e-stems, as well as -stems, do not have an - or e-variant respectively.

336

Alexander Letuchiy

‘Standard Average European’ (SAE) accusative languages, in that they demote the initial subject to a non-subject argument (in Adyghe, it becomes an oblique indirect object). 3.2.1. Potential and inadvertitive Two valency-changing operations – the potential and the inadvertitive (referring to involuntary action) – are syntactically similar to a canonical passive: they decrease the status of the agent argument, making it an indirect object (figure 1).7

Figure 1. Syntactic e¤ect of potential and inadvertitive

I call ‘inadvertitive’ or ‘involitional’ the operation marked with the prefix ecˇ ’e-, which transforms the oblique-marked agent of the basic tran˙ into an indirect object, while the initial direct object retains its sitive verb absolutive marking but presumably changes its syntactic status to that of subject. The semantic e¤ect of this derivation is the following: the derived form means that the initial agent carries out the action involuntarily. The potential derivation is marked with the prefix fe-.8 Its syntactic e¤ect is the same as that of the inadvertitive. The meaning of the derived form is that the initial agent is capable of carrying out the action denoted by the base verb. As we can see, the syntactic e¤ect of these derivations is equivalent to that of passivization in SAE languages, except that the absolutive argument, which is usually patientive, and the oblique argument (the agent) retain their case-marking (because both the direct object and the intransitive subject are marked with absolutive case in Adyghe, while both the agent and the indirect object receive oblique marking). What changes is only the structure of the verb form and the syntactic properties of the arguments (see below on the reciprocalization of potential derivatives). 7. Here I illustrate with examples only the potential derivation; the inadvertitive (involitional) derivation, expressed with the prefix ecˇ ’e-, seems to be rather rare and to have additional syntactic properties which˙ lie outside the scope of the present article. 8. The same prefix also expresses benefactive meaning.

Ergativity in the Adyghe system of valency-changing derivations

337

(14) a. cˇ ’ale-xe-m bukva-xe-r Ø-a-le‚W -xe-re-p. ˙boy-pl-obl letter-pl-abs 3sg.abs-3pl.a-see-pl(abs)-dyn-neg ‘The boys do not see the letters.’ b. cˇ ’ale-xe-m bukva-xe-r Ø-a-fe-le‚ W -xe-re-p. ˙boy-pl-obl letter-pl-abs 3sg.abs-3pl.io-ben-see-pl(abs)-dyn-neg ‘The boys cannot see the letters.’ The transitive verb le‚W n ‘see’ (14a) has an oblique agent argument (subject), cross-referenced with the agent prefix a-, and an absolutive object, cross-referenced with the zero third person prefix and controlling the absolutive plural su‰x -xe. In (14b), the marking of the patient does not change; however, the oblique prefix a- now marks an indirect object, an argument of the potential/benefactive prefix fe-. This is evident from the fact that a- occupies the position immediately before fe-: as was mentioned above, the position immediately before the applicative prefix is reserved for an IO introduced by the applicative derivation. Interestingly, the agent NP does not change its case-marking, since both agents and indirect objects take oblique marking in Adyghe. That the verb in (14b) is indeed intransitive follows from two considerations: first of all, the oblique prefix a- cannot be an agent marker, since it depends on the potential prefix. As I have shown, all arguments introduced by the applicative derivations are indirect objects – thus, the verb in (14b) does not have an agent prefix and cannot be transitive. Second, potential derivatives cannot form reciprocals with zere-, while this is always possible for transitive verbs independently of their semantics (see section 3.4). Crucially for our analysis, the potential and involitional derivation are compatible only with transitive verbs. Neither monovalent nor bivalent intransitive verbs can form derivatives as in (14b) (see (14c) for a monovalent verb and (14d) for a bivalent intransitive): (14) c. *cˇ ’ale-xe-m a-fe-k We-re-p. ˙boy-pl-obl 3pl.io-ben-go-dyn-neg ˙ ‘The boys cannot go.’ d.

*cˇ ’ale-xe-m psˆasˆe-xe-m a-fe-bew -xe-re-p. ˙boy-pl-obl letter-pl-abs 3sg.io-ben-see-pl(abs)-dyn-neg ‘The boys cannot kiss the girls.’

This restriction does not seem to be related to the nature of potential or involitional meaning. Sentences like ‘I cannot go’ (monovalent intransitive verb) or ‘I cannot kiss a girl’ (bivalent intransitive verb), which are incom-

338

Alexander Letuchiy

patible with the potential construction under analysis, are semantically unobjectionable, their English translations being fully grammatical. We could in principle say that this is a manifestation of syntactic accusativity. If the oblique agent is the most syntactically privileged argument, the function of passive-like derivations can be to decrease its status: thus, they are compatible only with verbs which have an agent argument, i.e. transitive verbs. However, this explanation is valid only for passives proper. The potential in Adyghe has a semantic function, distinct from demotion of the agent argument: it expresses the modal meaning of possibility. Thus, there is no natural reason for the valency-changing mechanism to be incompatible with bivalent intransitive verbs possessing an agentive subject: why should it be impossible to form a sentence like ‘The boys cannot kiss the girls’, as in (14d)9? In our opinion, the situation is just the reverse: the incompatibility of the potential with intransitive verbs is an argument for syntactic ergativity – in other words, it confirms that the absolutive argument of transitive verbs is the most syntactically privileged argument (the subject, the pivot of the sentence). Note that a construction of the type (14d) ‘The boys cannot kiss the girls’ would require a change of case-marking. The absolutive subject ‘the boys’ would become the indirect object of the potential prefix, whereas the patient which occupies the position of an oblique-marked indirect object of the base verb (‘girls’) would take on absolutive case (because almost10 no verb can exist without an absolutive argument):

Figure 2. Syntactic e¤ect of potential and inadvertitive with intransitive verb (hypothesis)

9. The question why structures like (14c) are impossible is simpler to answer: if a monovalent verb were to undergo a potential derivation, it would lose its absolutive argument. However, any Adyghe verb, except those belonging to a very small class (see note 10), must have an absolutive argument. 10. I do not mention here the small class of so-called bezlicˇnyje (subjectless) verbs described by Kumakhov (1971) inter alia, which have only an oblique and no absolutive argument.

Ergativity in the Adyghe system of valency-changing derivations

339

The Adyghe derivational system in fact obeys a general restriction: no derivation can a¤ect the absolutive argument. For instance, all valencyincreasing derivations add oblique arguments only (Adyghe does not have a canonical applicative which introduces a new direct object). Moreover, the antipassive which eliminates the absolutive argument is not highly productive and is marked with a stem alternation rather than simply a grammatical marker. 3.2.2. Quasi-passive (resultative) Adyghe has two productive resultatives, one of them unmarked, and the other marked with the prefix zere-. We do not consider the zere-resultative here. The unmarked resultative in Adyghe functions syntactically as a quasipassive construction: in the resultative construction based on transitive verbs, as in (15b), the agent cannot be expressed. The structure of examples of the same construction based on intransitive verbs, as in (16), remains less clear. The unmarked resultative eliminates the oblique agent argument and the agent cross-reference prefix, as in (15b): (15) a.

b.

(16) a.

b.

Se qebaske Ø-z-‚e-zˇW a-‚. i.obl cabbage 3sg.abs-1sg.a-caus-cook-pst ‘I cooked cabbage.’ Qebaske Ø-‚e-zˇW a-‚e. cabbage 3sg.abs-caus-cook-pst ‘The cabbage is cooked.’ (Nikolaeva 2003) Ha-r l -me Ø-ja-ceqa-‚. dog-obl ˙man-obl.pl 3sg.abs-3pl.ioþoblique-bite-pst ‘The dog bit the men.’ l -me Ø-q-ja-ceqa-‚. ˙ man-obl.pl 3sg.abs-dir-3pl.ioþoblique-bite-pst ‘The men are bitten / have been bitten.’ (Nikolaeva 2003)

In (15a), the causative verb ‚e-zˇW en ‘cook’ is transitive: it bears the agentive cross-reference marker z-. In the resultative construction in (15b), the agent prefix and the agent NP are eliminated. The verb has only a patient absolutive argument, which is cross-referenced with a zero prefix.

340

Alexander Letuchiy

In example (16), the verb ceqen ‘bite’ is intransitive: this is evident from the fact that the third person singular agent cross-reference marker - is absent. The agent (dog) takes the absolutive case, whereas the patient (men) is in the oblique form. Despite the existence of examples like (16b), where the resultative construction is seemingly built on the basis of intransitive verbs, we cannot claim that this construction has the same morphological structure as that in (15b). It would be tempting to say that (16b) is an example of resultative formation from an intransitive verb: the absolutive subject of the bivalent intransitive verb ceqen ‘bite’ is eliminated. However, let us recall that the 3sg absolutive argument is cross-referenced with the zero prefix. Thus, it is impossible to prove that the absolutive argument is eliminated in (16b) in the same sense as the agent in (15b): we cannot say whether the zero prefix is present or absent. Therefore, up to this point we do not have any evidence that resultatives are formed in the same way from both intransitive and transitive verbs. The crucial question is whether the presence of a resultative construction represents an argument for or against syntactic ergativity in Adyghe. Although passivization seems to be characteristic primarily of accusative languages, it should be borne in mind that resultative formation is di¤erent from passivization. It is generally accepted that the resultative has a primarily aspectual function. Thus, we should not take (15) and (16) as evidence that Adyghe has a passive proper (a feature which is not characteristic of syntactically ergative languages). The same line of argumentation is plausible for the passive-like potential derivation in (14): the functional motivation for the potential is di¤erent from that for a canonical passive. In (14), we are not dealing with a pragmatically motivated agent demotion (as in the case of passivization), but with a semantically motivated demotion. The agent is demoted because it is no longer a prototypical agent: (14b) does not refer to the fact of bringing about a transitive situation, but only to the potential to do so. Thus, in the terms of Hopper and Thompson (1980), (14b) is less semantically transitive than (14a), because the situation is unreal, and the agent in (14b) is less agentive. Thus, the potential and inadvertitive derivations are syntactically similar to passives. They demote the agent, which is often the syntactic subject in morphologically ergative languages11. 11. However, below I will show that some criteria allow us to propose that the subject of Adyghe transitive verbs is the absolutive argument rather than the agent.

Ergativity in the Adyghe system of valency-changing derivations

341

However, all passive-like derivations in Adyghe have a semantic motivation distinct from the motivation for canonical passives in accusative languages. In contrast, the use of the antipassive in Adyghe is motivated by principles relevant for canonical antipassives in ergative languages: for instance, the antipassive pattern is used when the direct object is indefinite or is not important for the speaker. Therefore, antipassives and ‘passives’ do not occupy the same place in the Adyghe derivational system: only antipassive is a canonical voice. Thus, the evidence for syntactic ergativity (the antipassive) is more important than the arguments for syntactic accusativity (the passive-like derivations), since there is no canonical passive in Adyghe. 3.3. Reflexive and reciprocal: semantic or syntactic motivation? Reflexive and reciprocal are traditionally viewed as valency-rearranging derivations. They do not change the number of arguments, but introduce the requirement that the arguments display co-reference to each other. In fact, Adyghe reflexives and reciprocals are not derivations in the proper sense of the term. That is, markers of reflexivization in Adyghe should be described as personal markers, rather than derivational markers. In accusative languages which seem to represent the Standard Average European type, there are cases when morphological reflexivization is genuinely derivational: for instance, it often changes the inflectional type of the verb: French: (17) a. J’ai lave´ la vaisselle. b. Je me suis lave´. The verb laver ‘wash’ in (17a) is transitive. The reflexive marker se in (17b) changes its syntactic characteristics: now the verb takes the auxiliary eˆtre ‘be’ in the past, which signals that it has become syntactically intransitive – all French transitive verbs use another auxiliary, avoir, in the formation of the complex past (17a). We find no e¤ects of this type in Adyghe. In Adyghe, the reflexive and reciprocal morphemes represent personal cross-reference markers rather than derivational markers (see Rogava and Kerasˇeva 1966: 265–269, 271–276, Smeets 1992: 115–117, Letuchiy 2007 for details), but with the additional requirement that the argument crossreferenced with one of these markers must be co-referent with another argument. The reflexive/reciprocal marker always occupies the slot corre-

342

Alexander Letuchiy

sponding to one of the co-referent arguments (the -9 absolutive slot, the -6 or -5 oblique IO slot or the -4 agent slot): (18) a.

b.

W -s -w ps -‚. 2sg.abs-1sg.a-shave-pst ‘I shave you.’ Z -s -w ps -‚. refl.abs-1sg.a-shave-pst ‘I shave (myself ).’

In (18a), the verb w ps n ‘shave’ bears two personal markers: the agentive prefix s - (1sg) and the absolutive prefix w - (2sg). In (18b), the reflexive marker occupies the absolutive slot. The verb form in (18b) di¤ers from the form in (18a) only in that the absolutive slot is occupied by a special ‘reflexive marker’ – the derivation does not change the transitivity or the number of arguments of the base verb. In this sense the Adyghe reflexives and reciprocals are analogous to reflexive pronouns, such as sich in German or pozyn ‘oneself ’ in Khakas (Turkic, author’s field data), which also ‘substitute’ for one of the co-referent arguments. In French, the situation is roughly the same, except that the addition of se changes an important morphosyntactic property of the verb, namely its complex past formation. In what follows, the glosses for reflexive and reciprocal markers include the designation of the slot occupied by the marker. For instance, the gloss refl.abs in (18) means that the reflexive marker z - is in the slot of the absolutive argument (-9). It may seem that the position of the markers under analysis can help us to ascertain which argument of the base verb is the subject and which is the object. Indeed, in many languages (such as the East Caucasian languages and most languages of Europe) reflexives are subject-oriented: they are bound by the syntactic subject and are never found in subject position (for instance, the Russian sebja ‘oneself ’ does not have a nominative form even theoretically). However, this prediction is not entirely borne out. First of all, with transitive verbs it is notable that reflexives and reciprocals behave di¤erently from each other. The complex reciprocal marker zere- occupies the agentive oblique slot -4 (see also Rogava and Kerasˇeva 1966, Letuchiy 2007), whereas the reflexive marker z - is in the absolutive (direct object) slot -9:

Ergativity in the Adyghe system of valency-changing derivations

343

Reciprocal: (19) Zecˇ ’e c f-xe-r Ø-zere-le‚ W -zˇ’ -x. ˙ ˙ all man-pl-abs 3sg.abs-rec.a-see-re-pl.abs ‘All the people see each other (i.e. meet up).’ Reflexive: (20) Zecˇ ’e c f-xe-r z-a-le‚ W -zˇ’ . ˙ ˙ all man-pl-abs refl.abs-3pl.a-see-re ‘All the people see themselves.’ The picture which emerges in the domain of reflexivity and reciprocity is shown in table 2. It reflects the position and the form of both markers for the transitive and the intransitive class of verbs. For each marker, the slot it occupies is specified. Table 2. Positions of reflexive and reciprocal markers in transitive and intransitive verbs Transitive

Intransitive

Reflexive

z -, absolutive slot12

ze- / z - þ preverb of oblique argument, oblique (IO) slot

Reciprocal

zere-, oblique slot

ze- / z - þ preverb of oblique argument, oblique (IO) slot

It may seem that in Adyghe both reflexive and reciprocal marking (the marker zere- aside) are organized as in semantically and morphologically ergative languages such as the East Caucasian languages. In these, reflexives are always bound by the agent of the transitive verb and the absolutive argument of the intransitive verb. If we adopt the hypothesis that the reflexive marker is always controlled by the syntactic pivot/subject, as is the case in most of the world’s languages, this means that some syntactic subject properties characterize the oblique agent of transitive verbs and the absolutive argument of intransitive verbs. This is the case in many

12. I do not consider here the distribution of ze- vs. z -, which seems to be purely morphophonemic, and not directly related to the syntactic status of the verb arguments, according to (Arkadiev and Testelets 2009). However, it may be noted that z - mostly cross-references absolutive arguments, and ze- is used in most examples for oblique arguments (see also (21) vs. (22) for the di¤erence between reciprocal and reflexive forms).

344

Alexander Letuchiy

other Caucasian languages where the ergative argument (agent) of transitive verbs and the absolutive argument of intransitive verbs possess most subject properties (see Testelets and Toldova 1998). However, this explanation is problematic for some minor verb classes, such as inverse verbs13, the verb zexex n ‘hear’ and potential derivatives in fe-. For all of these the position of ze- / z - is defined in a more complex way than in transitive and intransitive verbs. First of all, in inverse verbs the reflexive and reciprocal markers can occupy either the absolutive slot or one of the oblique slots. In other words, both reflexive and reciprocal forms can have ze- / z - either in the absolutive slot or in the oblique. (21) T -z -sˇ’ -gW psˇa-‚ / z -t-sˇ’ -gW psˇa-‚. 1pl.abs-refl.io-loc-forget-pst refl.abs-1pl.io-loc-forget-pst ‘We forgot ourselves.’ (22) A-xe-m ze-sˇ’ -gW psˇe-zˇ’ -‚e-x / that-pl-ob rec.io-loc-forget-re-past-pl.abs z-a-sˇ ’ -gW psˇe-zˇ’ -‚. rec.abs-3pl.io-loc-forget-re-past ‘They forgot each other.’ The distribution of ze- vs. z - is sensitive to morphophonological conditioning (see note 12). At the same time, as the comparison of the first variants in (21) vs. (22) shows, reflexives tend to choose z - even in some cases where reciprocals choose ze- (for instance, when cross-referencing an indirect object). Second, in derivatives with potential meaning ze- / z - can only be found in the oblique slot when used reciprocally, but only in the absolutive slot when used reflexively see (24) below: (23) S- e me-w z -sˆ z -s-fe-w ps -sˇ’t-ep. 1sg-hand dyn-hurt-conv refl.abs-1sg.io-ben-shave-fut-neg ‘My hand hurts, so I cannot shave (myself ).’

13. I use this traditional term of Caucasian linguistics (see, for instance, Rogava and Kerasˇeva 1966, and Kumakhov 1971) for bivalent emotional and mental verbs such as sˇ ’ gW psˇen ‘forget’, whose absolutive argument is the stimulus and the oblique argument is the experiencer (the sentence ‘I forgot you’ is literally translated into Adyghe as ‘You were forgotten to me’).

Ergativity in the Adyghe system of valency-changing derivations

345

This variation in behaviour prevents us from considering reflexivization to be a syntactic test. What is relevant for these derivations is the agentivity of the arguments involved. The rule for reflexivization can be stated as follows: the reflexive marker always occupies the slot of the least agentive of the co-referent arguments (this explanation was first proposed in Smeets 1992 for the Shapsug dialect of Adyghe). The most agentive argument is the oblique argument of transitive and potential derivatives, and the absolutive argument of intransitive verbs. Note that the behaviour of sˇ ’ gW psˇen ‘forget’ in example (21) confirms the semantic explanation. Situations like ‘forget’ involve neither an agent nor a patient; both arguments (stimulus and experiencer) have properties intermediate between those of agent and patient. For instance, neither acts volitionally (like an agent), and neither is a¤ected by the situation (like a patient). This is why the position of the reflexive marker is variable in inverse verbs. Therefore, reflexives cannot give us any information on ergativity in Adyghe. The choice of the argument which controls the reflexive marker is motivated semantically and not syntactically: it is the most agentive argument, whether this is marked with absolutive or oblique case.14 By contrast, the reciprocal marker is much more informative from our point of view. The Adyghe reciprocal is clearly absolutive-oriented, as is evident from four facts: e in inverse verbs, the reciprocal marker ze- most commonly occupies the

oblique indirect object slot -5, as in the first variant of (22), and can only rarely occupy the absolutive slot, as in the second variant of (22). Conversely, the reflexive marker usually occupies the absolutive slot in these derivatives, as in the second variant of (21). This di¤erence between reflexivization and reciprocalization is not taken into account in Smeets (1984) and Smeets (1992); e in derivatives with potential meaning, the reciprocal marker zeoccupies the oblique slot -5 (24), whereas the reflexive marker occupies the absolutive slot, as in (23): (24) A-xe-r Ø-ze-fe-le‚W -xe-r-ep. (s)he-abs 3sg.abs-rec.io-ben-see-pl-dyn-neg ‘They hate each other’ (lit. ‘They cannot see each other’). 14. Semantically motivated agent orientation of reflexives is not unique to Adyghe: a similar situation is found in the ergative languages Tsakhur (Kibrik (ed.) 1999) and Warlpiri (Legate 2006) and many other languages (including some which, like Warlpiri, are syntactically ergative according to some criteria).

346

Alexander Letuchiy

e in intransitive verbs, the reciprocal marker occupies the oblique IO

slot -5;

e in transitive verbs, the reciprocal marker (though in another variant)

also occupies an oblique slot, namely the agent slot -4, as in (18), though the oblique argument is the most agentive for transitive verbs. Thus, the reciprocal marker is generally found in one of the oblique slots and it is controlled by an absolutive personal marker. The only verb which admits reciprocalization exclusively in the absolutive slot is zexex n ‘hear’. The absolutive orientation of reciprocals in most verb classes (except for the inverse verbs, where variation is observed as in (22)) cannot be accounted for in semantic terms. The reciprocal prefixes occupy the oblique slot of most bivalent verbs, irrespective of whether the oblique argument is more agentive than the absolutive one (as with transitive and potential verbs) or less agentive (as with intransitive verbs where the oblique argument is an indirect object). In other words, the motivation for the choice of the controller is not the degree of agentivity of the arguments, but rather the location of absolutive marking. Thus, we take the absolutive orientation of reciprocals as evidence for syntactic ergativity in Adyghe: reciprocal markers are bound with the syntactically privileged absolutive argument, whereas the behaviour of the reflexive marker is semantically motivated. 3.4. Facilitive and di‰cilitive: semantic motivation Two more valency-decreasing derivations in Adyghe, namely the facilitive (-‚ Wesˆ W ) and di‰cilitive (-‚ Waje), are treated in syntactic terms in ˙ and Kerasˇeva 1966, and Kumakhov 1971. These authors claim that Rogava in the derivations in question, the personal marker in subject position (i.e. the absolutive argument of intransitive verbs, and the ergative (‘oblique’ in our terminology) of transitive verbs) is eliminated, and the non-subject marker is retained. This is indeed the case with transitive and intransitive verbs: in (25) the verb does not bear an agentive marker, which would normally be j -. In (26) the situation is less straightforward since the absolutive subject prefix is zero in 3SG (this is why Smeets (1992) does not analyse the status of absolutive subjects in facilitive constructions at all). However, the indirect object prefix j - is obviously present in (26), and we can judge that the absolutive prefix is absent from the fact that it is impossible for most native speakers to express an absolutive argument, such as ‘It is easy for me to push this boy’, without using an applicative prefix.

Ergativity in the Adyghe system of valency-changing derivations

347

Transitive verb: (25) M cˇ ’ale-r ‚e-sˇ ’ ne-‚ WesˆW . ˙ this ˙boy-obl caus-fear-fcl ‘It is easy to frighten this boy.’ (lit. ‘This boy is easy to frighten’). Intransitive verb: (26) M cˇ ’ale-m Ø-je-w ncˇ’ -‚ Wesˆ W . ˙ ˙ this ˙boy-obl 3sg.abs-3sg.ioþoblique-push-fcl ‘It is easy to push this boy.’ However, this derivation can also be analysed in semantic terms: it is always the most agentive argument which is eliminated. This explanation seems more plausible, because sensation verbs, such as zexex n ‘hear’ and le‚ W n ‘see’, often admit expression of an agent. While the case of zexex n ‘hear’ is complicated, because this verb is neither a canonical transitive nor a canonical intransitive verb, le‚ W n ‘see’ is a canonically transitive verb: the reciprocal formation indisputably demonstrates its transitivity. The experiencer of verbs like these is in a sense more agentive than the stimulus (see Dowty 1991 for details): only the experiencer, but not the stimulus, is animate. Sensation verb: (27) Te t -zere-le‚ W -‚ Wesˆ W . ˙ we 1pl.abs-rec.a-see-fcl ‘It is easy for us to see (to meet) each other.’ In (27), the verb includes both an agent and an absolutive personal prefix (note that the verb le‚ W n ‘see’ is syntactically transitive, and therefore the agent prefix should be eliminated, according to Smeets 1992). Hence, it is not only the most agentive argument, but usually a prototypical agent which is obligatorily eliminated in the facilitive/di‰cilitive form – while non-agentive animate arguments can be expressed. In this sense the facilitive derivation provides evidence against syntactic accusativity in Adyghe.

4. Conclusions The processes analysed in this paper can be divided into two groups, based on the kind of information they provide about ergativity in Adyghe.

348

Alexander Letuchiy

First of all, there are derivations which can be regarded as semantically motivated (though syntactic motivation can also be proposed for these processes). This is the case with reflexive marking, since the reflexive marker always seems to be bound by the most agentive argument; the facilitive and di‰cilitive always require the omission of the most agentive argument (though this is generally the agent of transitive verbs or the absolutive argument of intransitive verbs); the resultative is also semantically motivated. These derivations could in principle be taken as evidence that Adyghe is syntactically accusative, but in fact this is not the case. Secondly, there are derivations which are only compatible with transitive verbs, namely the inadvertitive and potential. These transformations are more significant for our analysis, since they show that Adyghe is syntactically ergative. This fact manifests itself in a simple rule which regulates all ‘passive-like’ derivations: they can only change properties of the oblique (Agent), but not the absolutive argument. The reciprocal formation, which is absolutive-oriented, also demonstrates that the absolutive has subject properties (at least the orientation of reciprocals towards the absolutive cannot be explained in semantic terms). The conclusion that Adyghe is syntactically ergative seems to be contradicted by the fact that the antipassive is not productive in Adyghe and does not fully correspond to the prototypical antipassive construction. However, there is in fact no contradiction. What is essential is that the Adyghe system forces us to modify our assumptions about the function served by verbal derivation in an ergative language. Usually authors implicitly or explicitly adopt the following view on the function of voices: they change the status of the most syntactically privileged participant. For instance, in Dixon (1994), the observation that the passive is characteristic of accusative languages, and the antipassive of ergative languages, is explained by the fact that passives decrease the syntactic status of the agent (which is privileged in accusative languages), whereas antipassives decrease the status of the absolutive object (which is privileged in ergative languages). However, the Adyghe data lead us to adopt the opposite hypothesis for this language: derivations can change the status of any participant, except the most privileged one. As we have shown, no derivation or voice in Adyghe can change the status of the direct object or intransitive subject – with the exception of the antipassive. But the antipassive seems to be unproductive and expressed by means of stem alternation rather than a grammatical marker sensu stricto, and it violates the general restrictions which apply to the system of valency-changing derivations in Adyghe.

Ergativity in the Adyghe system of valency-changing derivations

349

Thus, Adyghe seems to be syntactically ergative and not syntactically accusative. This syntactic ergativity manifests itself in the derivational domain. We think that the view adopted by Matasovic´ (2008: 64), that the absolutive case marks ‘‘the lowest ranking macrorole argument’’, may be true from the semantic point of view. However, the morphosyntactic processes we have examined do not show that the ergative argument is syntactically privileged over the absolutive. However, some other syntactic processes, such as co-referent deletion, are not at all restrictive in Adyghe, which makes the question of syntactic alignment and the subject/object status of participants impossible to answer. The question why the derivational system in Adyghe is much stricter than many other syntactic processes requires further research, which is undertaken, for instance, in Letuchiy (2009a). Note that the hypothesis that Adyghe is syntactically ergative makes the case of the potential and inadvertitive (see 3.3.1) more problematic. Since the agent is not the syntactic subject of the underlying transitive verb, these two passive-like derivations are not entirely similar to the passive voice: they do not demote the initial subject. In any case, we should suppose that the agent is a syntactically privileged argument, even if it is not a subject. The passive-like derivations demote a privileged argument to the status of indirect object. In other words, provided that Adyghe is a syntactically ergative language, the two passive-like derivations are similar to voice alternations in that they rearrange the syntactic status and change the pragmatic values of the arguments, but this voice alternation di¤ers from the passive formation in that the demoted argument is not a syntactic subject. These data from Adyghe are also useful because they show that syntactic accusativity in some morphologically ergative languages can in fact be determined by semantic factors. Many processes which seem to be motivated accusatively, such as the formation of the reflexive, facilitive or di‰cilitive, may in fact be motivated in semantic terms rather than being directly related to syntactic alignment.

Abbreviations 1,2,3 a abs ap

1st, 2nd, 3rd person agent cross-reference prefix absolutive case; absolutive argument cross-reference prefix antipassive variant of the verb stem

350

Alexander Letuchiy

ben caus com fcl ins io obl oblique re rec refl sg pl pst tr

benefactive causative comitative facilitive instrumental case indirect object cross-reference prefix oblique case oblique indirect object prefix refactive reciprocal reflexive singular plural past tense transitive variant of the verb stem

References Aikhenvald, Alexandra and Robert M.W. Dixon 2000 Changing valency: Case studies in transitivity. Cambridge: Cambridge University Press. Arkadiev [Arkad’jev], Petr M., and Alexander B. Letuchiy [Letucˇij] 2008 Derivacii antipassivnoj zony v adygejskom jazyke [Antipassivelike derivations in Adyghe]. In Tipologija glagol’noj derivacii [The typology of verbal derivation], Vladimir A. Plungian and Sergei G. Tatevosov (eds), Moscow: Jazyki slav’anskoj kul’tury. Arkadiev [Arkad’jev], Petr M., and Yakov G. Testelets 2009 O trex cˇeredovanijax v adygejskom jazyke [On three alternations in Adyghe]. In Aspekty polisintetizma: Ocherki po grammatike adygejskogo jazyka [Aspects of polysynthesis: Essays on Adyghe grammar], Yakov G. Testelets, Peter M. Arkadiev, Alexander B. Letuchiy and Nina R. Sumbatova (eds), 121–145. Moscow: RGGU. Baker, Mark 1996 The polysynthesis parameter. New York: Oxford University Press. Belyaeva, Anna V., and Sergei V. Minor 2005 ‘‘Iskl’ucˇitel’naja koreferentnost’’’ v konstrukcijax s sentencial’nymi aktantami v adygejskom i kabardino-cˇerkesskom jazykax [‘‘Extraordinary co-reference’’ in constructions with sentential arguments in Adyghe and Kabardian]. In Vtoraja konferencija po tipologii i grammatike dl’a molodyx issledovatelej. Tezisy

Ergativity in the Adyghe system of valency-changing derivations

351

dokladov [Abstracts of the second conference in grammar and typology for young researchers], Sergei S. Say et al. (eds). Saint Petersburg: ILI RAN. Comrie, Bernard 1976 The syntax of causative constructions: Cross-language similarities and divergences. In Syntax and Semantics 6. The grammar of causative constructions, Masayoshi Shibatani (ed.), 261–312. New York: Academic Press. Cooreman, Ann 1994 A functional typology of antipassives. In Voice: Form and Function [Typological Studies in Language 27], Barbara A. Fox and Paul J. Hopper (eds), 49–88. Amsterdam and Philadelphia: Benjamins. Dixon, Robert M.W. 1994 Ergativity. Cambridge: CUP. Dowty, David 1991 ‘‘Thematic Proto-Roles and Argument Selection’’ Language 67: 547–619. Gisˇev, Nukh T. 1968 Glagoly labil’noj konstrukcii v adygejskom jazyke [Labile verbs in Adyghe]. Majkop: Adygejskoje knizˇnoje izdatel’stvo. Jelinek, Eloise 1984 Empty categories, case, and configurationality. Natural language and linguistic theory 2.1. Hewitt, Bernard George 1982 ‘Antipassive’ and ‘labile’ constructions in North Caucasian languages. General Linguistics 22.3. 158–171. Hopper, Paul and Sandra Thompson 1980 Transitivity in grammar and discourse. Language, v. 56.2. 251– 299. Kibrik, Alexander E. 2002 Ocˇerki po obsˇcˇim i prikladnym voprosam jazykoznanija. Moscow: URSS. Kibrik, Alexander E. (ed.) 1999 Elementy tsakhurskogo jazyka v tipologicˇeskom osvesˇcˇenii [Elements of Tsakhur grammar in typological perspective]. Moscow: Nasledie. Kumakhov, Mukhadin 1971 Slovoizmenenie adygskix jazykov [Inflection in Adyghe languages]. Moscow: Nauka. Kumakhov, Mukhadin 2006 Ergativnost’ v cherkesskix jazykax [Ergativity in Circassian languages]. Malmo¨: Malmo¨ University, 2006.

352

Alexander Letuchiy

Lander, Yuri A. 2009 Mnozˇestvennaja reljativizacija: podlinnaja i mnimaja. In Aspekty polisintetizma: ocˇerki po grammatike adygejskogo jazyka [Aspects of polysynthesis: aspects of Adyghe grammar], Yakov G. Testelets, Peter M. Arkadiev, Alexander B. Letuchiy and Nina R. Sumbatova (eds), 612–653. Moscow: RGGU. Legate, Julie Anne 2006 Split absolutive. In Ergativity. Emerging Issues, Alana Johns, Diane Massam and Juvenal Ndayiragjje (eds). Berlin: Springer. Letuchiy, Alexander B. 2007 Reciprocals, reflexives, sociatives and comitatives in Adyghe. In Reciprocal constructions [TSL 71], Vladimir P. Nedjalkov, Emma Sˇ. Geniusˇiene˙ and Zlatka Guentche´va (eds), 773–811. Amsterdam: John Benjamins. Letuchiy, Alexander B. 2009a Verb classes in Adyghe: derivational and non-derivational criteria. Abstract of the talk at the conference ‘‘Verb typologies revisited’’. Letuchiy, Alexander B. 2009b Kauzativ, dekauzativ i labil’nost’ [Causative, anticausative and lability]. In Aspekty polisintetizma: ocˇerki po grammatike adygejskogo jazyka [Aspects of polysynthesis: aspects of Adyghe grammar], Yakov G. Testelets, Peter M. Arkadiev, Alexander B. Letuchiy and Nina R. Sumbatova (eds), 372–428. Moscow: RGGU. Matasovic´, Ranko 2008 Transitivity in Kabardian. In Investigations of the Syntax– Semantics–Pragmatics Interface, Robert D. Van Valin (ed.), 59–74. Amsterdam, Philadelphia: John Benjamins. Nichols, Johanna 1986 Head-marking and dependent-marking grammar. Language, v. 62.1. Nikolaeva, Liudmila L. 2003 Rezul’tativ v adygejskom jazyke [The resultative in Adyghe]. Expedition report. Unpublished manuscript. Paris, Catherine 1987 Comment sont remplies en tcherkesse les functions de´volues dans d’autres langues aux variations de diathe`se. Actances 3. Plungian, Vladimir P. 2000 Obsˇcˇaja morfologija [General morphology]. Moscow: URSS. Rogava, Grigorij V. and Zejnab I. Kerasˇeva 1966 Grammatika adygejskogo jazyka [A grammar of Adyghe]. Krasnodar: Krasnodarskoje knizˇnoje izdatel’stvo. Say, Sergei S. 2007 K tipologii antipassivnyx konstrukcij: semantika, pragmatika, sintaksis [Towards a typology of antipassive constructions:

Ergativity in the Adyghe system of valency-changing derivations

353

semantics, pragmatics, syntax]. PhD. diss. Saint Petersburg: Institute for linguistic studies. Serdobol’skaja, Natalia V. 2007 Ergativity in Adyghe. Talk in the Caucasian seminar at INALCO, Paris. Shibatani, Masayoshi 1985 Passives and related constructions: a prototype analysis. Language, v. 61: 821–848. Smeets, Rieks 1984 Studies in West Caucasian phonology and morphology. Leiden: Hakuchi Press. Smeets, Rieks 1992 On valencies, actants and actant coding in Circassian. In Caucasian Perspectives, George B. Hewitt (ed.), 98–144. Mu¨nchen: Lincom Europa. Testelets, Yakov G. 2001 Vvedenije v obsˇcˇij sintaksis [Introduction to general syntax]. Moscow: RGGU. Testelets, Yakov G., and Svetlana Yu. Toldova 1998 Refleksivnyje mestoimenija v dagestanskix jazykax i tipologija refleksiva. Voprosy jazykoznanija 4. Txarkaxo, Junus A. 1991 Adygejsko-russkij slovar’ [Adyghe-Russian dictionary]. Majkop: Adygejskoje knizˇnoje izdatel’stvo. Van Valin, Robert D. and Randy La Polla 1997 Syntax. Structure, meaning and function. Cambridge: Cambridge University Press. Zekox, Ucˇuzˇuk S. 2002 Adygejskaja grammatika [A grammar of Adyghe]. Majkop: GURIPP ‘‘Adygeja’’.

The evolution of transitive verbs in Basque and the emergence of dative-marked patients Ce´line Mounole 1. Introduction1 Basque is the only surviving pre-Indo-European language of Western Europe. It is spoken in the Basque Country, a region situated between France and Spain. It is a SOV language, ergative, overwhelmingly su‰xing and highly agglutinating. As for the verbal morphology, it is mainly analytic. In historical Basque, the middle voice values of autocausative and decausative are obtained by detransitivization of the verb: the patient of the transitive construction becomes the single argument of an intransitive one. At the same time, the transitive auxiliary is replaced with an intransitive one. As concerns the passive interpretation, it is expressed by participial predicate structures, namely resultative constructions (Rebuschi 1983; Trask 1985; Ortiz de Urbina and Uribe-Etxebarria 1991). That is why there are often identified as ‘stative or adjectival passives’ (Ortiz de Urbina 2003: 299). In this paper, we will argue that di¤erential object marking is emerging in modern Basque. This kind of development being unusual in an ergative language, we will try to describe the mechanisms involved. As we will see, animacy and referentiality seem to be the main factors governing this phenomenon. It still remains to be investigated to what degree the DOM system is employed for discourse purposes. After a brief presentation of the structure of actancy in Basque (§2), we will provide the data of some varieties of this language which encode in the dative the patients of transitive verbs (§3.1 and §3.2). Afterwards, we 1. I want to express my gratitude to Gilles Authier and Katharina Haude who gave me the opportunity of publishing this paper, and to Ricardo Etxepare, Beatriz Ferna´ndez and the two anonymous reviewers for their comments and encouragement. I am also grateful to Luigi Anselmi, Kepa Erdozia, Aitor Iglesias, Alazne Landa, Arantzazu Elordieta, Julen Manterola, Jean Haritschelhar, Maitena Laxague and the informants of Arratia, Lekeitio, Tolosa, Baigorri and Azkarate. All errors are mine.

356

Ce´line Mounole

will compare those data with the structure of transitives in Spanish, the language in contact with the varieties of Basque displaying this pattern (§3.3 and §3.4).

2. Actancy in Basque Before starting the study of the evolution of transitive verbs we shall briefly describe the structure of actancy in Basque. Throughout the paper, I will use the following abbreviations: A for the Agent-like term of transitive verbs, P for the Patient-like term of the same transitive verbs, S for the Single term of intransitive verbs and D for the Dative term of both groups of verbs. Furthermore, by transitive verb I mean a verb governing an agent (A) in the ergative case and a patient (P) in the absolutive case (i.e. verbs of actions, and verbs structured and behaving like verbs of action) (§2.2). If it does not govern any P term in the absolutive, it is intransitive. So, we will consider as intransitive the monovalent verbs with a single term in the absolutive (§2.1) or in the ergative (§2.3.1), and as bivalent intransitives the verbs bearing a term in the absolutive and another in the dative (§2.1), and those with a term in the ergative and another in the dative (§2.3.2). 2.1. Intransitive verbs Intransitive constructions generally display a term S in the absolutive indexed on the verb by means of prefixes, except for the 3rd person singular and plural, since their absolutive NP is never overtly reflected on the verb ((1) vs (2)): (1) Zu erori zara you.abs fall aux.pres.s2sg ‘You have fallen down.’ (2) Jon etorri da Jon.abs come aux.pres.s3sg ‘Jon has come.’ Moreover, intransitives can include a second term in the dative case, which also triggers verb agreement by the addition of su‰xes (3–5). These bivalent intransitive constructions are mainly found with three types of

The evolution of transitive verbs in Basque

357

verbs.2 First, a handful of psychological verbs require the experiencer in the dative and the stimulus in the absolutive. This is the case for interesatu ‘to interest’, gustatu ‘to like’, damutu ‘to repent’, gaitzitu ‘to be o¤ensive to’, dolutu ‘to repent’ (3) among others.3 Besides, the dative agreement is also used with motion verbs to express the end-point of the motion (4), or as a dative of interest, with possessed entities (5) (for more detail, see Etxepare 2003): 4 (3) Zu-ri dolutzen zaizu han izana you-dat regret aux.pres.s3sg.d2sg there being.abs ‘You regret to have been there.’ (4) Jon-i Andoni joan zaio Jon-dat Andoni.abs go aux.pres.s3sg.d3sg ‘Andoni has gone from/to Jon.’ 2. We also find this construction with some verbs that encode the ‘aimer’ in the absolutive and the ‘aimed object’ in the dative (Lazard 1998). This class of verbs that was more extended four centuries ago, is nowadays restricted to a handful of verbs ( jarraiki ‘to follow’, jarin ‘to spill’, atxiki ‘to attach to’) that alternate the absolutive-dative construction with the ergative-dative construction or with the transitive one (ergative-absolutive) (see §2.3.2). 3. Anyway, some of them can also be constructed monovalently with the experiencer receiving the absolutive and the stimulus the instrumental or locative marking: (54)

Literatura-n interesatzen da literature-loc interest aux.pres.s1sg ‘He is interested in literature.’

(55)

Damutzen da bere jokamolde-az regret aux.pres.s1sg his behaviour-inst ‘He regrets his behaviour.’

4. However, this construction is obligatory neither with the possessed entities nor with the motion verbs. We can perfectly translate the possession by a genitival phrase (56) and the end-point of the motion by an allative (57): (56)

Jon-en liburua erori da Jon-gen book.abs fall aux.pres.3sg ‘Jon’s book has fallen.’

(57)

Jon Andoni-rengana joan da Jon.abs Andoni-all go aux.pres.s3sg ‘Jon has gone to Andoni.’

358

Ce´line Mounole

(5) Jon-i liburua erori zaio Jon-dat book.abs fall aux.pres.s3sg.d3sg ‘Jon’s book has fallen’ (lit. ‘The book has fallen to Jon’) 2.2. Transitive verbs Basque transitive constructions require a term A in the ergative case, and a term P in the absolutive case. Both are indexed on the verb: A by means of su‰xes and P by prefixes, the latter being the same as the S prefixes of the intransitive construction. Again, the third person is an exception to this rule since its absolutive and ergative NPs are never overtly encoded on the verb ((6) vs (7)): (6) Ni-k zu ikusi zaitut telebista-n I-erg you.abs see aux.pres.a1sg.p2sg television-loc ‘I have seen you on TV’ (7) Jon-ek mahaina hautsi du Jon-erg table.abs broke aux.pres.a3sg.p3sg ‘Jon has broken the table’ A third term D encoded in the dative can appear in transitive constructions. It also triggers agreement on the verb by the addition of su‰xes (8), just the same as those employed to index the dative NP in intransitive constructions (cf. (3–5) above).5 The absolutive term of the trivalent verbs, which is restricted to be a third person, is not reflected on the verb; neither are the 3rd person ergative NPs: (8) Ni-k zu-ri ogia eman dizut I-erg you-dat bread.abs give aux.pres.a1sg.p3sg.d3sg ‘I have given you some bread’ Finally, the 3rd person is overtly marked on the verb only when it is encoded in the dative case. This is done by the addition of the su‰x -o (singular) or -e (plural) (see (4) and (9)): (9) Jon-ek Mikel-i ogia eman dio Jon-erg Mikel-dat bread.abs give aux.pres.a3sg.p3sg.d3sg ‘Jon has given Mikel some bread’ 5. As nowadays, in the 16th century, the agreement with the dative object is systematic in Western (Bizkaian, Araban) and Central (Gipuzkoan) dialects, while it is variable in the Eastern ones (Low Navarrese, Zuberoan) (Ortiz de Urbina 1991; Etxepare and Oyharc¸abal 2008).

The evolution of transitive verbs in Basque

359

2.3. Split intransitivity Apart from those typical transitive and intransitive constructions which are the most common and widespread ones in the system, the Basque language has some verbs sharing an unusual pattern. Those verbs possess a term in the ergative but none in the absolutive. Even so, they trigger the morphology usually used with prototypical transitive verbs. This phenomenon has been referred to as split-intransitivity (Dixon 1994), and those verbs as unergatives (Perlmutter 1978; Levin 1983), anti-impersonals (Lazard 1985, 1995), or accusatively-aligned intransitive verbs (Creissels 2006). 2.3.1. Intransitive monovalent verbs with an NP in the ergative The Basque language has a handful of intransitive verbs with an NP in the ergative and none in the absolutive. As Trask (2002) reminded us the majority of them are quite old, and ‘‘mostly appear to be historical accidents’’.6 Some of them may be ancient causative verbs (for example, iraun ‘to last’ in (10), irakin ‘to boil’, see Rebuschi 1984) since they bear the causative prefix -ra which can still be recognized in some pairs of verbs like ikusi ‘to see’ / erakutsi ‘to make see, to show’. Besides, Basque has many light verb constructions composed of a bare noun (sometimes an adjective, or an adverb) plus the verb egin ‘to do’. In many cases, both elements have joined together, leaving the construction without any P element (see Ortiz de Urbina 1989; Trask 2002; Etxepare 2003):7

6. In this section we only consider the verbs of Basque origin. However, in Southern dialects, many recently borrowed verbs follow this same pattern: desaparezitu ‘to disappear’, dudatu ‘to doubt’, funtzionatu ‘to function’, eskiatu ‘to ski’ (Sarasola 1977; Alberdi 2003). 7. We must add that the N þ egin ‘to do’ complex predicates are not homogenous since the N of some of them seems to behave as the true Patient-like term of egin ‘to do’. Indeed, it may be separated from the verb egin in whquestions or focalized constructions, and partitive-marked in negative constructions (see Ortiz de Urbina 1989; Laka 1993): (58)

ez du salto-rik egin neg. aux.pres.a3sg.p3sg jump-PART do ‘He didn’t jump.’

360

Ce´line Mounole

(10) Kontzertu-ak ez zuen luzaz iraun concert-erg no aux.past.a3sg.p3sg long last ‘The concert did not last long.’ (11) Irakasle-ak barre egin zuen teacher-erg laugh do aux.past.a3sg.p3s ‘The teacher laughed.’ 2.3.2. Intransitive bivalent verbs with an NP in the ergative and an NP in the dative The second group of verbs with an unusual pattern, like the former, has one term in the ergative and no term in the absolutive. It di¤ers however considerably from the first group since it involves a second term marked in the dative. Furthermore, while showing only two terms, it requires the morphology usually used with prototypical ditransitive verbs (see §2.2 above).8 The verbs sharing this unusual pattern di¤er in their origin and conditions of occurrence and, in consequence, it would be convenient to distinguish di¤erent subgroups among them. Indeed, some of them are aligned in this way already in our oldest texts while others have undergone a change in their actancy in historical times. To begin with, since their first records, the verbs deitu ‘to call’, eraso ‘to attack’, itxadon/itxaron ‘to wait’, begiratu ‘to look at’, oratu ‘to grasp, seize’, iguriki ‘to wait’ have been constructed with an NP in the ergative, an NP in the dative, and have displayed a trivalent morphology. In some cases, this unusual alignment seems to be easily explicable. Itxadon / itxaron and iguriki ‘to wait’ may be ancient compounds which got fused, letting the construction without any apparent patient (hitz ‘word’ þ *edun ‘to have’ > itxadon and egun ‘day’ þ eduki ‘to hold’ > iguriki; cf. Trask 2002); eraso ‘to attack’ might be an ancient causative (Etxepare p.c.). As for deitu ‘to call’, which is a simple transitive in the Northern dialects, it may have su¤ered a change in actancy in the Southern dialects, becoming an intransitive with an NP in the ergative and an NP in the dative, by analogy with its equivalent composed form dei egin ‘‘to call’’ (dei ‘calling’ 8. It must be understood that even if the verbal morphology required is originally reserved to the ditransitive constructions, in those constructions none (absolutive) NP has been omissed. Furthermore, remember that in the ditransitive constructions the absolutive NP, that can only be a 3rd person, is not overtly indexed on the verb.

The evolution of transitive verbs in Basque

361

plus egin ‘to do’) (Creissels 2008; Etxepare p.c.). Finally, begiratu ‘to see’ used with an NP in the ergative and an NP in the dative in some dialects, hardly shows us a clear explanation of its actancy. In (12) and (13), we see that the construction displays an NP in the dative (ainbesteri ‘so much’ and sazerdoteari ‘priest’, respectively), and no NP in the absolutive. The term in the ergative is dropped but it is indexed on the verb (-k in itxadok, and -gu in diogu): (12) Gatx egiten boc ainbeste-ri itxadok bad do aux.a2sg.p3sg so much-dat wait-a2sg.p3sg.d3sg ‘If you do it badly, wait for so much.’ (Refranes y Sentencias 1596: 296) (13) Begiratu bear diogu sazerdote-ari see need aux.pres.a1pl.p3sg.d3sg priest-dat Iesus Christo-ri berari bezala Jesus Christ-dat himself like ‘We have to look at the priest as we look at Jesus Christ himself. ’ (Beriain 1621: 12) We must add that, depending on the dialects,9 some of these verbs can also be used transitively without a¤ecting either the lexical meaning of the verb or the interpretation of the process (for a detailed description, see Ferna´ndez 2008). However, some other verbs provide a di¤erent meaning depending on the pattern adopted (Ferna´ndez 2008). For example, the verbs begiratu and oratu mean ‘to look at’ and ‘to grasp, to seize’ respectively when constructed with an NP in the ergative and an NP in the dative (13) and (15b), but ‘to protect, to save’ and ‘to capture, to catch’ when constructed transitively (14) and (15a): (14) Nola bere promesa begiratzen du? how his promess.abs save aux.pres.a3sg.p3sg ‘How does he save its promise?’ (Leizarraga 1571: Oth. D 6r) (15) a.

Gena-k sagua askatzen ta eullia oratzen spider’s_web-erg mouse.abs liberate and fly.abs capture ‘The spider’s web liberates the mouse and captures the fly.’ (Refranes y Sentencias 1596: 298)

9. Both constructions are possible with deitu ‘to call’, itxadon ‘to wait’ since our oldest texts. On the contrary, iguriki ‘to wait’, begiratu ‘to see’, eraso ‘to attack’ are transitivized in classical and modern Basque, certainly by hypercorrection or analogy with the verbs of action.

362

Ce´line Mounole

b.

Oratu egijozu aga on-i grasp aux.imp.a2sg.p3sg.d3sg railing this-dat ‘Grasp yourself to this railing.’ (Mogel 1881: 118)

Moreover, depending on the conjugation path adopted, other verbs slightly vary in the interpretation of the process described. The clauses with the verb deitu ‘to call’, when employed transitively have two possible interpretations (i) ‘they called/phoned Xabier to his o‰ce’, or (ii) ‘they called him to go to their o‰ce’ (16a). In contrast, when employed with an NP in the ergative and an NP in in the dative (16b), the clauses have only one possible interpretation: ‘they called him to his o‰ce’ (Etxepare 2003: 412–413): (16) a.

b.

Xabier bulegora deitu dute Xabier.abs o‰ce.all call aux.pres.a3sg.p3sg ‘They called Xabier to the o‰ce.’ Xabierr-i bulegora deitu diote Xabier-dat o‰ce.all call aux.pres.a3sg.p3sg.d3sg ‘They called Xabier to the o‰ce.’

Apart from all these verbs showing this pattern since our oldest texts, the other verbs which nowadays bear an NP in the ergative and an NP in the dative have undergone a change in their actancy in historical Basque. They have two sources: (i) the evolution of some intransitive bivalent verbs and (ii) the valency-shift of some transitive bivalent verbs. Indeed, some intransitive bivalent verbs constructed with an NP – normally the subject – in the absolutive and an NP in the dative in the oldest texts (17), later on shifted towards another intransitive pattern with an NP –the subject– in the ergative, the other NP remaining in the dative (18). It is the case of the verbs lagundu ‘to help’, eskatu ‘to ask for’, jarraiki / jarraitu ‘to follow’, eutsi ‘to retain, to hold on’, etxeki ‘to retain’. It is interesting to notice that among them, the last verb is the only one to have undergone this change path in the Northern dialects of Basque. In fact, the first three have undergone a distinct evolution in these dialects becoming purely transitives with one term in the ergative and another in the absolutive (19):10 10. In historical Basque eutsi ‘to retain, to hold on’ is not employed in the Eastern dialects. They employ the verb etxeki ‘to retain’ instead of it. It is just the contrary in the Western dialects.

The evolution of transitive verbs in Basque

363

(17) Zein gauza lagun-duko iaku which thing.abs help-fut aux.pres.s3sg.d1pl kastidadea goardeetako? chastity.abs save ‘What can help us to save our chastity?’ (Kapanaga 1656: 50) (18) Zer-k lagun-duko digu kasto izateko? what-erg help-fut aux.pres.a3sg.p3sg.d1pl chaste be ‘What will help us to be chaste?’ (Elizalde 1735: 49) (19) Lagun gaitzatzue help aux.imp.a2pl.p1pl ‘Help us.’ (Etxeberri 1627: XLIII) The evolution of transitive verbs is the second source for the occurrence of this pattern. Whereas in some dialects they remain always transitive (20), in others, their human patient shifts from the absolutive case towards the dative (21)11 and the bivalent morphology of the verb is replaced by the trivalent one: (20) (Ni-k) (zu) ikusi zaitut I-erg you.abs see aux.pres.a1sg.p3sg ‘I have seen you.’ (21) (Ni-k) (zu-ri) ikusi dizut I-erg you-dat see aux.pres.a1sg.p3sg.d3sg ‘I have seen you.’ In this paper, we are going to deal with this evolution of transitive verbs and to try to give an answer to the mechanisms causing the occurrence of such an exceptional pattern. 3. Emergence of dative-marked patients among the transitive verbs In this section we shall focus on the evolution of transitive verbs causing the occurrence of dative-marked patients in Basque. We will see some data of archaic and old Basque, before describing in more detail the situation of modern Basque. We will finish by comparing the Basque and the Spanish data. 11. Data collected from young speakers of the Gipuzkoan variety of Tolosa (see 3.2.3)

364

Ce´line Mounole

3.1. Archaic and Old Basque12 In sixteenth century’s texts, transitive verbs almost always governed a term A in the ergative and a term P in the absolutive, both indexed or crossreferenced on the verb. At that time, we find only two examples of transitive verbs bearing an unusual pattern. The verbs ikusi ‘to see’ and ulertu ‘to understand’,13 commonly conjugated transitively in all the dialects, carry the agent in the ergative and the patient in the dative in two Western texts – Refranes y Sentencias 1596 (Bizkaia) and Lazarraga c.1564 (Araba). Besides, the trivalent auxiliary is used (see footnote 8): (22) Joan gura dot ekustera ni-k go desire aux.pres.1sg.p3sg see i-erg on deretxada-n-ari love.s3sg.p3sg.d1sg-rel-dat ‘I want to go to see the person who loves me.’ (Lazarraga c. 1564: 1164v) (23) Trakart egiten deustak ta ulerretan deceit do aux.pres.a2sg.p3sg.d1sg and understand deustat aux.pres.a1sg.p3sg.d2sg ‘You deceive me and I understand you.’ (Refranes y Sentencias 1596: 49) From the next century onwards, it has been possible to observe the spreading of this kind of structure among the transitive verbs. We find, for example, dan˜atu ‘to damage’ and persegitu ‘to pursue’ with an agent 12. According to Lakarra’s periodization (1997), Archaic Basque embraces the period until 1600 and Old Basque the period between 1600 and 1745. 13. Until the 19th century, the verb ulertu ‘to understand’ is only used in occidental dialects. Anyway, in the text where it appears with a P in the dative case (23), it is also used with a P in the absolutive in which case this latter is [-human]: (59)

uler ezak lenago ta understand aux.imp.a2sg.p3sg first and itz egik geroengo word do.imp.a2sg.p3sg later ‘First understand [it] and speak latter.’ (Refranes y Sentencias 1596: 194)

The evolution of transitive verbs in Basque

365

in the ergative, a patient in the dative and a trivalent auxiliary in Bizkaian and High Navarrese texts (Southern dialects): (24) On-ek persegietan deusku-elako those-erg pursue aux.pres.a3sg.p3sg.d1pl-caus ‘Because they [our worst temptations] pursue us all.’ (VJ 17th century, Mitxelena 1954) (25) Nor-k dan˜atzen digu? who-erg injure aux.pres.a3sg.p3sg.d1pl ‘Who is injuring us?’ (Elizalde 1735: 49) All those earliest occurrences of the dative-marked patients are to be found in Southern texts (Spanish Basque Country) and restricted to [þhuman] patients (see 22–25). Among them, we find Spanish loans as well as verbs of Basque origin. We will return to this topic (§3.2.4). 3.2. Modern Basque In Modern Basque, dative-marked patients are more widespread. Di¤erent dialects have developed them among the transitive verbs. We find them in some varieties of High Navarrese, Bizkaian, and Gipuzkoan, all varieties of Southern Basque (Spanish Basque Country). Thus, first of all, we are going to present the data of each dialect one by one, and next, we will try to give an explanation of the mechanisms involved. 3.2.1. High Navarrese In his study of the Basque verb, Bonaparte ([1869] 1991)14 pointed out the absence in most varieties of meridional High Navarrese15 of many forms of the bivalent auxiliary usually used with prototypical transitive verbs. Indeed, the bivalent forms indexing the 1st and 2nd person absolutive patients had disappeared, and instead of them, the speakers employed the trivalent form, thus encoding those patients in the dative. He found only the bivalent auxiliary indexing the 3rd person absolutive patients. A century later, Yrizar (1981) was able to confirm that the description of the meridional High Navarrese given by Bonaparte was in general still 14. Bonaparte’s Le verbe Basque en tableaux ([1869] 1991) was the first monography of the Basque verb that provided the data from all dialects. 15. In all the area apart from the valleys of Erro, Artze and Burgete.

366

Ce´line Mounole

valid.16 Furthermore, he found the same phenomenon in some Western varieties of High Navarrese (Sakana and Burunda varieties) which in Bonaparte’s times still used the bivalent auxiliary without any restriction. Nowadays, in Etxarri-Aranatz (variety of Sakana), the 3rd person human patients can also be encoded in the dative (Erdozia 2001):17 (26) Alsasu-e yaman doogu Fermin˜-i Altsasu-adl bring aux.pres.a1sg.p3sg.d3sg Fermin-dat ‘We have brought Fermin to Altsasu.’ However, the dialectological descriptions do not say precisely whether the use of the dative is systematic with all the 3rd person animate patients or whether it has any restriction (see below §3.2.4). 3.2.2. Bizkaian Bonaparte (1991) did not gather any case of dative propagation among the patients of transitive constructions in Bizkaian. However, Yrizar (1981) points out its use in the varieties of Markina, Gernika, Bermeo, Arratia and Arrigorriaga. Here, we are going to study the data of the seaside village Lekeitio and of the valley of Arratia. As shown by Hualde, Elordieta and Elordieta (1994), in Lekeitio, they tend to encode the patient of transitive constructions in the dative. Indeed, the dative case can appear with the 1st, 2nd and 3rd person singular and plural patients even if it is less frequent with the latter type. The 3rd person patient can be encoded in the dative if and only if it is animate (27a, 27b): (27) a.

b.

Pedrori ikusi dotzat Pedro-dat see aux.pres.a1sg.p3sg.d3sg ‘I have seen Pedro.’ Kotxia ikusi dot car.abs see aux.pres.a1sg.p3sg ‘I have seen the car.’

16. Yrizar did not give more data as this dialect was moribund. However, he was able to add that the phenomenon occurred in the variety of Erro. 17. Unfortunately, we are not able to give modern data of Altsasu (Burunda) as the variety spoken there is almost extinct today.

The evolution of transitive verbs in Basque

367

Anyway, the restrictions concerning the use of the dative are even more complex than they seem to be at first sight. Indeed, it is restricted to the human patients (28) (as in Ferna´ndez 2008) and moreover, the definiteness of those human patients too seems to be an important condition for the occurrence of the dative case (29), (30):18 (28) a.

b.

(29) a.

b.

(30) a.

**Txakurra-ri ikusi dotzat dog-dat see aux.pres.a1sg.p3sg.d3sg ‘I have seen the dog.’ Txakurra ikusi dot dog.abs see aux.pres.a1sg.p3sg ‘I have seen the dog.’ **Eztotzat ezaututen in˜or-i neg.aux.pres.a1sg.p3sg.d3sg know nobody-dat ‘I don’t know anybody.’ Eztot ezaututen in˜or neg.aux.pres.a1sg.p3sg know nobody.abs ‘I don’t know anybody.’ **Morroi bat-i ikusi dotzat guy one-dat see aux.pres.a1sg.p3sg.d3sg ‘I have seen a guy.’

b.

Morroi bat ikusi dot guy one.abs see aux.pres.a1sg.p3sg ‘I have seen a guy.’

In the valley of Arratia, 1st and 2nd person patients are all encoded in the dative (31), but they would never encode the 3rd person patient in this way –even if it is human and definite–. Indeed, it is always in the absolutive case (32) (Iglesias in prep).19

18. These data were provided by a 50 year-old man, native of Lekeitio. It is remarkable that another speaker of this variety only accepted to encode the 3rd person patient in the dative with the frequent verb ikusi ‘to see’. So, it seems that frequency of the verb also plays a role in the distribution of this pattern. Anyway, we must add that in Lekeitio the new pattern has not replaced the old one. Thus, the marking of the patient still alternates between the dative and the absolutive. 19. Arretxe (1994) described the same situation for the Basque spoken in Basauri.

368

Ce´line Mounole

(31) Artu´ osku´n ko´txean ero´an osku´n take aux.past.a3pl.d1pl car carry aux.past.a3pl.p3sg.d1pl ‘They took us in their car, and carried us’ (32) A ikusı´ dot him.abs see aux.pres.a1sg.p3sg ‘I have seen him.’ As for the younger generation, my informant accepts to encode the 3rd person patient in the dative only with the verb ikusi ‘to see’ (and not for example ezagutu ‘to know (somebody)’) which is also a very frequent verb (cf. footnote 18): (33) Ari ikusı´ dotze´t him-dat see aux.pres.a1sg.p3sg.d3sg ‘I have seen him.’ 3.2.3. Gipuzkoan Bonaparte (1991) and Yrizar (1981) do not find this pattern in Gipuzkoan. Nowadays, it is, however, very productive in some of its varieties. The data presented above have been collected in the variety of Tolosa (Central Gipuzkoa) from speakers aged from 21 to 33. Generally speaking, the speakers encode the 1st and 2nd person patients in the dative. As for the 3rd person patient’s case-marking, it is restricted by human animacy and referentiality constraints: all the inanimate patients are in the absolutive but not all the human ones are encoded in the dative.20 As the examples show, the definite NPs (proper names, pronouns, and definite nouns) are in the dative, but the indefinites (in˜or ‘nobody’) (35), quantifiers (asko ‘many’, guzti ‘all’) (36–37) and reciprocals (elkar ‘each other, one another’) (38) can only be in the absolutive: (34) (Ni-k) irakasle-ari ikusi diot I-erg teacher-dat see aux.pres.a1sg.p3sg.d3sg ‘I have seen the teacher.’ 20. Interestingly, in some cases, the majority of them also accepted the transitive pattern based on a term A in the ergative and a term P in the absolutive as an equivalent of the marked pattern they usually employ (ezagutu ‘to know’, jo ‘to hit’), while in some others almost all of them totally rejected the former pattern (utzi ‘to let, to leave’, ikusi ‘to see’, aukeratu ‘to choose’), qualifying it as ‘‘wrong, ungrammatical’’.

The evolution of transitive verbs in Basque

369

(35) **(Ni-k) ez diot in˜orr-ei ikusi I-erg no aux.pres.a1sg.p3sg.d3sg anybody-dat see ‘I have not seen anybody’ (36) **Jon-ek neska asko-ri ikusi dio Jon-erg girl many-dat see aux.pres.a3sg.p3sg.d3sg ‘Jon has seen many girls.’ (37) **Jon-ek neska guzti-ei ikusi die Jon-erg girl all-dat see aux.pres.a3sg.p3sg.d3sg ‘Jon has seen all the girls.’ (38) **Elkarr-i ikusi diote each other-dat see aux.pres.a3pl.p3sg.d3sg ‘They have seen each other.’ Concerning the indefinite article bat it seems that the dative-marked ones receive referential interpretation (39b), and the absolutive-marked ones, a non-referential interpretation (39a): (39) a.

b.

Idazkari bat bilatzen det secretary one.abs look_for aux.pres.a1sg.p3sg ‘I am looking for a secretary.’ (in general) Idazkari bati bilatzen diot secretary one-dat look_for aux.pres.a1sg.p3sg.d3sg ‘I am looking for a secretary.’ (¼one of my o‰ce)

A further point is that there is a generational split in the use of this pattern. The speakers of this variety of Basque who are over 40 encode all their animate patients – so, the 1st, 2nd and 3rd person patients – in the absolutive and employ the transitive bivalent morphology. However, even if they would not employ it, they accept the dative-marking of the human patient as a potential variant. 3.2.4. Summary In short, the data presented above show that dative marking of the patient in particular varieties of Basque is not done at random. Animacy and referentiality degrees of the NP involved seem to play a role in the use and the non-use of the dative-marking.

370

Ce´line Mounole

In all the varieties encoding the patient in the dative case, humanhood is a central constraint.21 However, this phenomenon has di¤erent realizations from one dialect to another. In Southern and Western High Navarrese (§3.2.1) and in the Bizkaian variety of Arratia (§3.2.2) only the 1st and 2nd person patients can be encoded in the dative, namely the persons ranking highest in Silverstein’s (1976) hierarchy of animacy. The younger speakers of those varieties seem to employ this pattern very rarely with the 3rd person patients. In Tolosa (Gipuzkoa) and Lekeitio (Bizkaia), lower ranking elements of this hierarchy can be encoded in the dative, namely the 3rd person human NPs. However, in this case, case-marking varies with the referentiality degree of the NPs. As we have seen above, the definite NPs (proper names, pronouns, and definite nouns) are marked in the dative case. As for the indefinite NPs, those receiving a referential interpretation tend to be encoded in the dative, whereas the non-referential ones are encoded in the absolutive. Other elements such as the quantifiers asko ‘many’, guzti ‘all’, and the reciprocal elkar ‘each other, one another’ which may also receive a nonreferential interpretation can only be encoded in the absolutive. Some bascologists (Rezac 2006; Ferna´ndez 2008) have proposed that this tendency of Basque to encode the animate patients in the dative would have appeared by influence of the Spanish pattern. Therefore, before discussing this position, we will very briefly summarize the structure of transitive verbs in this latter language, focusing more especially on the variety of Spanish spoken in the Basque Country. 3.3. Spanish In Modern Spanish, the agent of transitive constructions is not morphologically marked whereas the patient can be, depending on its level of animacy and definiteness (Pottier 1968; Bleam 1999). When the patient of a transitive construction is animate – human but also anthropomorphized objects – and definite (40) it must be preceded by the preposition a, which is homophonous with the dative case marker. As for the indefinite NPs, those contributing a referential interpretation must be accompanied by the preposition a (42), whereas non-referential NPs can optionally bear it (43). 21. Inanimate patients are usually not encoded in the dative. We find only one exception to this rule in a writing of Beriain (1621) with the verb apatu ‘to kiss’.

The evolution of transitive verbs in Basque

371

Finally, inanimate patients are not marked at all (41): (40) He visto a la mujer aux.a1sg ver acc the woman ‘I have seen the woman.’ (41) He visto un coche rojo aux.a1sg ver a car red ‘I have seen a red car.’ (42) Juan busca a una chica que sabe espan˜ol Juan look_for acc a girl that know.ind Spanish ‘Juan is looking for a girl who speaks Spanish.’ (43) Juan busca (a / Ø) una chica que sepa espan˜ol Juan look_for acc a girl rel know.subj Spanish ‘Juan is looking for a girl who speaks Spanish.’ As for the clitic system, it di¤ers from one dialect to another. In standard Spanish, the 3rd person clitics split into two groups: accusative and dative. The accusative clitics are lo (masculine) and la (feminine), while in the dative masculine and feminine share the same and unique clitic le. In contrast, the variety of Spanish spoken in the Basque Country is ‘‘leı´sta’’, as it employs the clitic le – etymologically dative clitic – as the animate masculine and feminine patient clitic. So, in this variety, the animate patient (44) and the recipient (45) – animate or inanimate – are both represented by the clitic le, the clitics lo and la being relegated to inanimate patients, masculine and feminine respectively (46) (Landa 1995; Fernandez-Ordon˜ez 1999): (44) le vı´ ayer cl3acc see.past.a1sg yesterday ‘I saw him/her yesterday’ (45) le dı´ un pastel a Juan/Elena cl3dat give.past.a1sg a cake dat Juan/Elena ‘I gave a cake to Juan / Elena.’ (46) lo comprare´ man˜ana cl3acc buy.fut.a1sg yesterday ‘I will buy it tomorrow.’

372

Ce´line Mounole

Furthermore, in leı´sta dialects definite (44) and indefinite animate patients with a referential interpretation (47) can be doubled by le. However, clitic-doubling is not possible with patients which only have a nonreferential interpretation22 (48) (Landa 1995; Bleam 1999): (47) La profesora le castigo´ a un nin˜o the teacher cl3acc punish-past.a3sg acc a child ‘The teacher punished the child.’ (48) No (*le) vi a nadie neg cl3 acc see.past.a1sg acc nobody ‘I didn’t see anyone.’ (Bleam 1999: 118) 3.4. Discussion Without any doubt, the restrictions concerning the use of the preposition a and the clitic doubling in Spanish are very similar to those which appear in Basque when encoding the patients in the dative. However, the main di¤erence between both languages lies in the fact that until nowadays the varieties of Basque encoding the patient in the dative only allowed it with the highest ranking elements of Silverstein’s hierarchy – namely, the 1st and 2nd persons –, while in Spanish all the animate person patients – 1st, 2nd and 3rd ones – can bear the preposition a. Nevertheless, when encoding the 3rd person patients in the dative, the same restrictions as those appearing in the case of the clitic doubling in Spanish emerge. Furthermore, we must remember that the dative patients appear only in the varieties of Basque spoken in the Spanish Basque Country.23 This suggests that if the animacy and definiteness conditioned dative patients 22. As an anonymous reviewer points out to me, not all indefinite pronouns behave in the same way. For example, todo ‘all’ must be doubled by a clitic – the patient being animate or inanimate –: lo he hecho todo ‘I have done all’, les vı´ a todos ‘I saw all of them’ (vs **he hecho todo and **vı´ a todos). 23. Yrizar (1981, 1999a, 1999b) claimed that this pattern also occurred in the variety of Low Navarrese, precisely in Baigorri and Azkarate, two little villages of the French Basque Country. He based this idea on the first manuscript of Bonaparte on the verb of Baigorri, where under the title ‘‘Il m’a’’ the trivalent auxiliary appears instead of the bivalent one. However, in the second manuscript on the same topic, the apparent mistake is corrected. Furthermore, Yrizar added that this pattern was still in use in both villages in the twenty first century, as one informant of each village was able confirm it to

The evolution of transitive verbs in Basque

373

did not appear by interference with the Spanish structure, their use must at least have been reinforced by contact with Spanish. Nowadays, the diglossic situation is stronger than it used to be, and as a consequence, the interferences with Spanish increase in the di¤erent areas of the grammar. Data of Unified Basque24 seem to be a good example of the clear influence of the Spanish structure in the transitive constructions in synchrony. Unified Basque taught in all the schools of the Southern Basque Country does not allow dative-marked patients. However, this tendency is more and more perspicuous among the pupils who learn Basque at school. In spite of all the recommendations of the prescriptive grammars which automatically reject it, this structure is invading more and more transitive verbs with an animate patient – Spanish loanwords (50) as well as verbs of Basque origin (51–53) (Zubiri and Zubiri 1995: 507) – (kritikatu ‘to criticise’, animatu ‘to encourage’, entzun ‘to hear’, gonbitatu ‘to invite’, abisatu ‘to inform’, ikusi ‘to see’, ezagutu ‘to know’, jo ‘to hit’. . .). Moreover, just as in Spanish, it appears with 1st, 2nd and 3rd person patients: (49) El profesor le ha animado the teacher him-dat aux.pres.a3sg hearten ‘The teacher has encouraged him.’ (50) Irakasle-ak animatu dio teacher-erg heart aux.pres.a3sg.p3sg.d3sg ‘The teacher has encouraged him.’ (51) Gaur kale-an ikusi dizut today street-loc see aux.pres.a1sg.p3sg.d2sg ‘Today I have seen you on the street.’ (52) Irakasle-ak jo egin dit teacher-erg hit foc aux.pres.a1sg.p3sg.d1sg ‘The teacher has hitten me.’ him. We have spoken with his informant from Baigorri who made it clear that he had never heard (and used) this pattern, which seemed to him ‘‘a barbarism’’. Our informants from Azkarate have also maintained that they don’t know it. So, unfortunately based on a philological mistake and later reinforced by misunderstanding, the description given by Yrizar seems incorrect. 24. Unified Basque is a standardized version of the Basque language. It was created in the 1970s by Euskaltzaindia (The Royal Academy of the Basque Language). This is the version of the language used in the o‰cial texts, schools, newspapers, TV.

374

Ce´line Mounole

(53) Neska horr-i ez diot ezagutzen Girl this-dat not aux.pres.a3sg.p3sg.d3sg know ‘I do not know this girl.’ We must remark that even if this pattern is very widespread in the spoken language we find few examples of it in the written corpus (newspapers, literature). This may be because of the pressure of the Standard Language which reject it, and because the Academy of the language keeps a tight control over the production of the language. Anyway, these synchronic variations need to be carefully examined. Certainly the best way to obtain more details on this phenomenon would be the realization of oral investigations directed at di¤erent generations of speakers in the di¤erent varieties of the language.

4. Conclusion As has been argued at the beginning of the paper, the Basque language has some verbs with an NP in the ergative, an NP in the dative and none NP in the absolutive. Apart from some verbs where this pattern seems to be lexicalized (§2.3.2), in most cases, it appears among the transitive verbs, and it is conditioned by the syntax of the construction. Animacy and referentiality seem to be the factors regulating the encoding in the dative of the patients (§3.2.4). In Basque dative-marked patients appear as soon as the sixteenth century’s texts. In spite of that, their use seems to spread out in the nineteenth century and even so, only in some varieties of Basque. Nowadays, this pattern is spreading more and more at the expense of the canonical transitive construction, certainly by interference of the Spanish structure. Abbreviations a abs acc aux caus cl d dat

agent absolutive accusative auxiliary causal marker clitic dative patient dative

The evolution of transitive verbs in Basque

erg foc fut gen imp ind loc rel pres p pl s sg subj * **

375

ergative focus marker future genitive imperative indicative locative relative marker present patient plural single argument in the absolutive singular subjunctive reconstructed agrammatical

References Alberdi, Xabier 2003 Euskal aditz mailegatuen erregimena: hurbilketa. [Borrowed verbs in Basque: an approach]. In Hommage a` Pierre Lafitte (1901–1985), IKER XIV. Bilbao: Euskaltzaindia, 37–60. Arretxe, Jon 1994 Basauriko euskara [The Basque of Basauri]. Basauri: Basauriko Udala. Beriain, J. 1621 Tratado de como se ha de Oyr Missa, Pamplona. Bleam, Tania 1999 Leı´sta Spanish and the Syntax of Clitic doubling. Ph. D. diss., Department of Linguistics, University of Delaware. Bonaparte, Louis Lucien 1991 Reprint. Le verbe basque en tableaux. In Opera Omnia vasconice. Vol. 1, 175–442. Bilbao: Euskaltzaindia. Original edition, London: Strangeways and Walden, 1869. Creissels, Denis 2006 Syntaxe ge´ne´rale: une introduction typologique. Paris: Lavoisier. Creissels, Denis 2008 Direct and indirect explanations of typological regularities: The case of alignment variations. Folia Linguistica 42 (1), 1–38. Dixon, Robert M. W. 1994 Ergativity. Cambridge: Cambridge University Press.

376

Ce´line Mounole

Elizalde, Francisco 1735 Apezendaco dotrina kistiana uskaraz [Christian doctrine for priests in Basque]. Pamplona. Erdozia, Jose Luis 2001 Sakana erdialdeko euskara [The Basque of Meridional Sakana]. Irun˜ea-Pamplona: Government of Navarre. Etxeberri, Joanes 1627 Manual devotionezcoa [Devotional book]. Bordeaux: G. Millanges. Etxepare, Ricardo 2003 Valency and argument structure in the Basque verb. In A Grammar of Basque, Jose Ignacio Hualde and Jon Ortiz de Urbina (eds), 63–426. Berlin/New York: Mouton de Gruyter. Etxepare, Ricardo and Oyharc¸abal, Ben˜at 2008 Hautazko datibo komunztadura ifar-ekialdeko euskalkietan [Dative agreement alternations in northeastern dialects of Basque]. Ms. Ferna´ndez, Beatriz 2008 Quirky dative objects in Basque. Power Point presented at the European Dialect Syntax III, Venice (Italy). Ferna´ndez-Ordon˜ez, Ine´s 1999 Leı´smo, laı´smo y loı´smo. In Grama´tica descriptiva de la lengua espan˜ola, Vol. 1, Ignacio Bosque and Demonte Violeta (eds.), 1317–1394. Madrid: Espasa Calpe. Harris, Alice 2002 Endoclitics and the Origin of Udi Morphosyntax. Oxford: Oxford University Press. Hualde, Jose Ignacio, Elordieta, Gorka, Elordieta, Arantzazu 1994 The Basque dialect of Lekeitio. Supplements of Anuario del Seminario de Filologı´a Vasca ‘‘Julio de Urquijo’’, XXXIV. Iglesias, Aitor In prep. Igorreko hizkeraren azterketa dialektologikoa [Dialectological study of the Basque variety of Igorre]. Ph.D. diss., Department of Linguistics and Basque Studies, University of the Basque Country. Kapanaga, Otxoa de, Martin 1656 Exposicio´n breve de la doctrina christiana compuesta por el P. M. Geronimo de Ripalda de la compan˜ia de Iesus. Bilbao. Laka, Itziar 1993 Unergatives that assign ergative, unaccusatives that assign accusative. MIT Working Papers in Linguistics 18: Papers on case and agreement I, 149–72. Department of Linguistics and Philosophy, MIT, Cambridge, MA. Lakarra, Joseba Andoni 1996 Refranes y Sentencias (1596). Ikerketak eta edizioa. [Refranes y Sentencias (1596). Works and edition]. Bilbo, Euskaltzaindia.

The evolution of transitive verbs in Basque

377

Lakarra, Joseba Andoni 1997 Euskararen historia eta filologia: arazo zahar, bide berri [History of Basque and philology: old problems, new perspectives]. Anuario del Seminario de Filologı´a Vasca ‘‘Julio de Urquijo’’ 31 (2), 447–535. Landa, Alazne 1995 Conditions on null objects in Basque Spanish and their relation to leı´smo and clitic doubling. Ph. D. diss., Departement of Linguistics, University of Southern California. Lazard, Gilbert 1984 Actance variations and categories of the object. In Objects: towards a theory of grammatical relations, Frans Plank (ed), 269–292. Londres/New York: Academic Press. Lazard, Gilbert 1985 Anti-impersonal verbs, transitivity continuum and the notion of transitivity. In Language invariants and mental operations, Seiler Hansjabob and Brettschneider Gunter (eds.), 115–123. Tu¨bingen: Gunter Narr Verlag. Lazard, Gilbert 1995 Le ge´orgien: actance duale (‘active’) ou ergative? Typologie des verbes anti-impersonnels. Sprachtypologie und Universalienforschung 48, 275–293. Lazard, Gilbert 1998 Actancy. Berlin/New York: Mouton de Gruyter. Lazarraga, Joan c.1564 Manuscript. Electronic Edition of In˜igo Landa. Leizarraga, Joanes 1571 Iesus Krist Gure Jaunaren Testamentu Berria [New Testament of Jesus Christ]. La Rochelle. Levin, Beth 1983 On the nature of ergativity. Ph.D. diss., MIT, Cambridge, MA. Mitxelena, Koldo 1954 Textos vascos antiguos. Un catecismo vizcaı´no del siglo XVII, Boletı´n de la Real Sociedad Vascongada de los Amigos del Paı´s 10, 85–95. Mogel, Jose Antonio 1881 Peru Abarka. Durango: J. de Elizalde. Ortiz de Urbina, Jon 1989 Parameters in the Grammar of Basque. Dordrecht: Foris. Ortiz de Urbina, Jon 1991 Datibo komunztaduraren gainean [Remarks on the dative agreement]. In Euskal dialektologiako kongresua, Ricardo Go´mez and Joseba Andoni Lakarra (ed.), 579–588. Supplements of Anuario del Seminario de Filologı´a Vasca ‘‘Julio de Urquijo’’ XXVIII.

378

Ce´line Mounole

Ortiz de Urbina, Jon 2003 Periphrastic constructions. In A Grammar of Basque, Jose Ignacio Hualde and Jon Ortiz de Urbina (eds), 284–300. Berlin/New York: Mouton de Gruyter. Ortiz de Urbina Jon and Uribe-Etxebarria, Myriam 1991 Participial predication in Basque. In Memoriae L. Mitxelena magistri sacrum, Joseba A. Lakarra and In˜igo Ruiz Arzallus (eds.), 993–1012. San Sebastia´n: Diputacio´n Foral de Guipuzcoa. Perlmutter, David M. 1978 ‘‘Impersonal passives and the Unaccusative Hypothesis.’’ In Proceedings of the Fourth Annual Meeting of the Berkeley Linguistic Society: 157–189. Berkeley Linguistic Society, University of California, Berkeley. Pottier, Bernard 1968 L’emploi de la pre´position a devant l’objet en espagnol. Bulletin de la Socie´te´ Linguistique de Paris 66 (1), 345–368. Rebuschi, Georges 1983 Autour du parfait et du passif basques. Iker 2, 545–558. Rebuschi, Georges 1984 Structure de l’e´nonce´ en basque. Paris: Socie´te´ d’Etudes Linguistiques et Anthropologiques de France. (L’Europe de la tradition orale 3). Rezac, Milan 2006 Agreement displacement in Basque: Derivational principles and lexical parameters. Gasteiz, University of the Basque Country, manuscript. Sarasola, Ibon 1977 Sobre la biparticio´n inicial en el ana´lisis en constituyentes. Anuario del Seminario de Filologı´a Vasca ‘‘Julio de Urquijo’’ XI, 51–90. Silverstein, Michael 1976 Hierarchy of features and ergativity. In Grammatical categories in Australian Languages, Robert M. W. Dixon (ed.), 112–171. Canberra: AIAS. Trask, Robert Larry 1985 The Basque passive: a correct description. Linguistics 23, 985– 991. Trask, Robert Larry 2002 Ergativity and Accusativity in Basque. In The Nominative & Accusative and Their Counterparts, Kristin Davidse and Be´atrice Lamiroy (eds), 265–285. Amsterdam: John Benjamins. Yrizar, Pedro 1981 Contribucio´n a la Dialectologı´a de la lengua vasca. 2 Vol. DonostiaSan Sebastia´n: Gipuzkoako Aurrezki Kutxa Probintziala.

The evolution of transitive verbs in Basque Yrizar, Pedro 1999a

Yrizar, Pedro 1999b

379

Morfologı´a del verbo auxiliar bajo navarro occidental (estudio dialectolo´gico). Irun˜ea-Pamplona: University of the Basque Country and Euskaltzaindia.

Morfologı´a del verbo auxiliar bajo navarro oriental (estudio dialectolo´gico). Irun˜ea-Pamplona: University of the Basque Country and Euskaltzaindia. Zubiri, Ilari and Zubiri, Entzi 1995 Euskal gramatika osoa. Bilbao: Didaktiker.

Index of subjects absolutive (case) marking 4–5, 9, 22– 23, 29, 31, 34–35, 40, 54, 67, 69, 72– 73, 90, 113, 134, 137, 142, 147, 150, 156, 167–169, 173–174, 176, 181, 183, 187–188, 190, 193, 215, 218, 290, 293–295, 298, 302, 307–310, 312, 324, 326–330, 333–340, 342– 349, 356–365, 367–370, 374 accusative alignment 2–7, 9–11, 28, 33, 51, 55–56, 68, 79, 81, 96, 115, 148, 160, 168–169, 193, 208, 241, 251–252, 254–255, 259, 264, 291, 308, 313–315, 323–325, 335, 341, 348–349, 359 accusative (case) marking 293, 299, 300, 307, 331, 371 active voice 31, 33, 38, 61–63, 67–69, 74, 76, 80–81, 85, 93, 94–95, 103, 117–118, 138, 142, 146, 241–242, 249–251, 269 activity verb 246, 254, 268, 271, 275, 277, 305 actor 6, 259–262, 264–265, 267, 270, 275, 277, 279, 282–283, 325 adjective, adjectivization 122, 124– 125, 140, 147, 159, 187, 228, 272, 280, 293, 355 adjunct 83, 218–220, 232, 244–246, 260–261, 263, 304 adposition (pre-/post-) 22–23, 25, 60, 70–71, 73, 75, 85, 230–232, 243– 245, 250, 329, 370, 372 agent 1–2, 5–11, 31–32, 34, 36–37, 40–41, 51–53, 55–58, 69–72, 75–76, 80, 82–83, 85–86, 88–97, 103, 112, 117–120, 122–123, 134–138, 140– 150, 152–153, 155, 157, 159–160, 175, 178, 182, 184–185, 193, 200, 202–206, 208–211, 214–219, 223, 229, 233–234, 237–238, 242–252, 254–255, 261, 269, 290–291, 293– 294, 298, 306–307, 309, 313–319,

324–328, 330–333, 335–340, 343– 349, 356, 364, 370 agentive voice 34–37, 43, 55, 62, 259, 265– 266, 270, 277, 282 participants/arguments 3, 89, 136, 138, 142–143, 187, 255, 291, 315, 329, 332, 338–340, 342, 345–348 verbs 184, 188–190 allative 173, 232, 301–302, 314, 357 ambitransitive see labile animate vs inanimate 60, 67, 71, 76, 83–86, 91–92, 135, 143, 205–206, 212–213, 216, 347, 366, 368–373 anticausative 8–9, 11, 112, 116–117, 119–120, 122–123, 125, 127–128, 134, 140, 149–153, 155, 157, 159– 160, 170–171, 184–187, 190–191, 193, 200, 213–214 antipassive 4–6, 9, 11, 16, 20, 29, 30, 32–37, 39, 41–43, 51–52, 56–60, 62, 90, 115, 134, 142, 149, 153–157, 160, 181–182, 189–190, 194, 200, 203, 205, 210–211, 215–216, 221– 223, 227, 243, 246–251, 253–255, 270, 325, 330–331, 333–335, 339, 341, 348 absolutive antipassive 34, 35 focus antipassive 8, 52, 54, 58, 96 applicative 8, 16, 23, 30, 32, 33, 38– 39, 52, 59, 72–76, 84, 95, 169–170, 172–173, 177–179, 192–193, 252, 267, 272, 326–327, 329, 337, 339, 346 aspect completive 17, 41, 54–57, 59–61, 68–69, 114, 120, 122 durative 155, 182, 237, 240, 255, 270 habitual 144, 145, 149–150, 160, 208, 237, 252, 270, 305

382

Index of subjects

imperfective 3, 138, 139, 140–141, 144–145, 148–150, 155, 158–160, 207, 209 incompletive 17, 35, 55–57, 59, 61– 62, 64–65, 68–69, 76–77, 79 iterative 155–156, 159, 251 perfective 3, 11, 136, 139–140, 144– 145, 148–149, 151–155, 159, 190, 334 progressive 3–4, 17, 58, 66–68 terminative 58, 314 autobenefactive 117, 184, 211 auxiliary 17, 52, 133, 136, 140, 147, 151, 241, 326, 341, 355, 364–366, 372 backgrounding 52, 76, 89, 97, 138, 140, 142–143, 153, 155 benefactive 8, 27, 31–32, 38, 39, 326– 327, 330, 336–337 beneficiary 39, 72–74, 165, 173, 177– 178, 182, 193 bivalent 77–80, 81, 227–228, 230, 232, 237–238, 240–242, 246, 249–250, 252, 260, 266–267, 269, 271–277, 282–283, 290, 293–294, 298, 305, 328, 329, 331, 333–334, 337–338, 340, 344, 346, 356, 360, 362–363, 365–366, 369, 372 body care verbs 176, 180, 190 causative 8, 61, 115, 136, 143, 171, 185, 189, 210, 266, 272, 280, 292, 298, 309, 310, 325–327, 330, 339, 359–360 choˆmeur 33–35, 38, 72 cleft 57, 83, 137 comitative 243, 301, 304, 314, 326–327 complement clause 28, 29, 79, 81, 240, 241 constituency, constituent order 83, 90, 134–135, 137, 194, 228, 230, 233, 238–239, 243–244, 250, 252, 261 contact (of languages) 7, 9, 10, 53, 127, 160, 356, 373

control (semantic) 55, 94, 154, 346 (syntactic) 1, 2, 77, 79, 155, 228, 238, 239–240, 250, 310, 324, 330– 331, 333, 337, 343, 345–346 coordination 80, 90, 148, 235, 248, 250, 323 core argument 17, 58, 64, 73–74, 77, 80–81, 113, 134, 172, 189, 228, 245, 265, 268, 290, 300, 308, 312–314 coreference 59, 64, 66, 76–81, 115, 120, 122, 148, 142, 148, 175–177, 179, 181–182, 187, 189–190, 192, 201, 228, 238–242, 250, 276, 298, 341–342, 345, 349 cross-reference 1, 22–24, 26, 39, 52, 54, 56–57, 59, 66–67, 69, 75, 94–96, 112, 126, 133–137, 167–170, 172, 178, 191, 323, 325–331, 337, 339– 341, 343–344 dative 6, 123, 168, 169, 172–173, 177– 179, 181–182, 188–190, 192, 203, 290–291, 299–300, 307, 313, 315, 355–358, 360–374 de-experiencer 216–217, 221 definiteness 42, 71, 83–84, 86, 91, 121, 126, 202–203, 207, 214–215, 219, 221, 234, 253, 299, 314–315, 329, 331–332, 341, 367, 368–370, 372 deletion/elision/omission of argument 33, 37, 54, 59, 75, 77, 79, 81, 83, 89–90, 97, 137–138, 146, 189, 199, 202–203, 208, 214, 217– 219, 223, 234, 238, 243, 248, 292, 307–309, 314, 329, 349 demonstrative 234, 248, 250 demotion 31, 33, 36, 39–40, 42, 51–52, 69, 71–72, 75, 82–83, 93–94, 97, 152, 189, 199, 217, 219, 221–223, 243, 255, 324, 331, 335–336, 338, 340, 349 deontic 11, 144–146, 149–150, 186, 208, 269 detransitive voice 34–35, 120, 125, 128, 133–134, 136, 138, 140–145, 147, 149–160, 263–265, 334, 355

Index of subjects di¤erential argument encoding 7, 255, 355 disambiguation 37, 43, 58, 71, 83, 85, 87, 96, 113 discourse functions 34, 42, 52 discourse prominence 82, 86, 265 ditransitive 38, 72, 73, 136, 174, 178, 181, 203, 210, 304, 360 doublet 292, 295, 296, 315 embedding see subordination ergative alignment/pattern 1, 2, 3, 4, 16, 22, 23, 28, 43, 79, 81, 112, 141, 259, 193, 200, 202, 208, 223, 229, 241, 277, 294, 296, 302, 308, 323, 328–329 ergative (case) marking 3, 21–22, 25, 29, 31, 56, 69, 114, 134–137, 142, 147, 199, 202–204, 218–219, 232, 277, 290, 298, 307, 309, 315, 356, 358 experiencer 67, 295, 298, 305, 344– 345, 347, 357 extended intransitive 136, 290, 293, 295–299, 302–305, 307–309, 311– 312, 314 extraction 37, 39, 82, 233, 234, 237, 238, 243 facilitative, facilitive 186, 191, 328, 330, 346–349 fluid verb 292, 295, 296, 302, 304, 315 focalization 54, 58, 235, 248, 359 focus 34, 52, 57, 58, 83, 96, 137, 144, 236, 237, 233, 235, 248, 251 agent focus 16, 29, 32–34, 36–37, 39–42, 57–58, 75, 79, 82, 86, 96 fronting 57–58, 85, 233–234, 237, 244–245, 251 genericity/generic 9, 53, 70, 72, 75, 88, 95, 116–118, 122–123, 125, 138, 144–145, 185, 200, 203–208, 210– 211, 221–223, 239, 251, 254, 309, 332

383

genitive 6, 232, 239, 250 grammaticalization 18–19, 25, 74, 96, 126–127, 191–193, 232, 243 head-marking 16, 54 hierarchy (semantic/pragmatic) 3, 6, 36, 42, 52, 53, 75, 79, 80, 81, 82, 83, 84, 85, 86, 87, 90, 95, 96, 199, 205, 206, 216, 245, 254, 259, 260, 264, 370, 372 (grammatical) 30, 31, 227, 228, 238, 239, 240, 255, 330 human vs non-human 71, 84–86, 91, 94, 200, 205–206, 208, 215–217, 239, 254, 264, 363–370 imperative 17, 136, 152, 200, 204, 294, 295, 309 impersonal 41, 65, 118, 122, 158, 185, 190, 308, 328, 331–333, 359 inadvertitive 324, 336, 338, 340, 348– 349 incorporation 8, 9, 219, 220, 221, 222, 223, 242, 252, 275, 276, 281, 298, 314 instrument(al) 31, 32, 33, 38, 39, 40, 70, 156, 186, 243, 269, 301, 304, 314, 331, 357 intralocutive see speech-act participant inverse 6, 7, 30, 37–38, 42–43, 82, 194, 204–206, 208, 210–211, 222– 223, 242, 245, 254, 259–260, 262– 268, 277, 282–283 inverse verbs 344–346 labile verbs 7, 115, 136, 151, 155, 159, 200, 217–219, 221–222, 332–333 left-dislocation 86, 233, 236 lexical alternations 7, 292 lexicalization 72, 160, 183, 189, 190, 192, 298, 374 locative 299, 300, 304, 307, 314, 326, 329, 330, 357

384

Index of subjects

media tantum verbs 140, 183, 190 middle voice 8, 51, 59, 62–64, 72, 75, 90, 117, 127, 142, 165–167, 170– 185, 187–194, 211, 213, 259, 265– 266, 271, 275, 277, 282–283, 309, 313, 255 modality 17, 65, 67, 91, 125, 134, 141, 144, 149, 150, 152, 155, 160, 266, 269, 269, 293, 326, 338 monovalent 36, 68, 77, 90, 229–232, 237–242, 246, 250, 252, 268, 271, 273–277, 283, 290, 293–294, 298, 302, 325, 331, 333–334, 337–338, 356–357, 359 motion/movement verbs 19, 67, 93, 103, 127, 142–143, 184, 189, 204, 305, 357 nominalization 2, 19, 60, 65, 199, 201, 204, 208, 237–238, 240–242, 249, 293 nominative 290, 293, 298, 313, 342 object 3, 21–23, 31, 33, 37, 54–56, 60, 67, 72–73, 75, 77–83, 85–86, 89–91, 93–94, 97, 103, 112, 117, 142, 156, 167, 173, 175–177, 179–182, 184, 187–190, 218–221, 227, 232, 241– 244, 250, 277, 290–291, 298–299, 304, 327, 331–333, 337, 342, 348– 349, 357 direct 27, 72–74, 167, 243, 255, 326, 328–331, 333–336, 339, 341–342, 344, 348 indirect 73, 324, 327–331, 333–338, 344–346, 349 obligatoriness 36, 41, 66, 79, 80, 113, 114, 122, 137, 202, 205, 223, 230, 240, 252, 261–262, 275, 307–309, 325, 327, 347, 357 oblique argument/marking 31, 33–36, 38–41, 69, 71, 73, 75, 93–94, 103, 117, 120–121, 123, 141, 152, 179, 186, 204, 218, 223, 246, 254, 260– 261, 263, 269, 270, 275–276, 290– 291, 301, 306–307, 313–314, 326– 331, 334, 336–340, 342–346, 348

obviation, obviative 6, 30, 3742, 43, 82, 83, 84, 87, 96, 262 omission of argument see deletion optionality 37, 81, 116, 123, 146, 222, 263, 275, 292, 302, 304, 307, 308, 313, 370 orientation 7, 60, 64–68, 70, 75–76, 91, 147–149, 155, 159–160, 237, 242, 259–261, 276–277, 283, 298, 313–315, 342, 345–346, 348 participle 63, 64, 76, 77, 79, 140, 147, 148, 307 passive 5–6, 16, 20, 32–35, 39–43, 51– 52, 56, 59–70, 73–96, 103, 111–112, 115–128, 133–134, 137–138, 140– 142, 144–155, 159–160, 166, 184– 185, 187, 190–194, 200, 205, 208– 214, 221–223, 227, 234, 267, 269, 291, 307, 309, 313, 323–324, 335– 336, 338–341, 348–349, 355 agentless passive 72, 75, 208, 211 patient 2, 5–7, 11, 31–40, 51–52, 55, 57, 58–60, 64, 66–69, 72–75, 77, 79, 80–83, 85–86, 89, 91–94, 103, 112, 117–119, 122, 133, 135–138, 140– 144, 147–149, 152–153, 155, 157, 159, 193, 200, 202–203, 205–206, 215–216, 218–220, 222–223, 229, 232, 234, 237–247, 249–255, 259, 261, 270, 275, 277, 290–291, 293– 294, 298, 306, 309, 313–318, 325, 331, 335–340, 345, 355–356, 359, 360, 363–367 perception, verbs of 78, 200, 204, 216, 219, 305 periphrastic constructions 17, 33, 59, 60, 63, 67, 68, 69, 75, 76, 85, 91, 136, 149 pivot 2, 51, 54, 55, 56, 79, 89, 112, 148, 160, 199, 208, 238, 239, 240, 241, 242, 247, 250, 292, 323, 324, 326, 338, 343 possession 24, 80, 83, 84, 177, 179, 180, 181, 190, 250, 280, 357

Index of subjects possessor 6, 22, 54, 67, 70, 81, 92, 94, 133, 173, 180, 242, 298, 329 potential 324, 330, 336–338, 340, 344–346, 348–349 pragmatics, discourse- 8, 31, 40, 42, 51–52, 76, 89, 90, 96–97, 114, 134, 137, 193, 194, 206, 227, 232, 233, 238, 243, 250, 251, 255, 264, 293, 310, 335, 340, 349 promotion 31, 42, 52, 72–75, 82, 97, 141, 148, 152, 251, 255, 311 recipient 72, 73, 74, 75, 148, 173, 175, 203, 210, 261, 290, 297, 299, 302, 314, 371 reciprocal 115, 117, 119, 134, 141– 142, 177, 179, 188, 194, 211–213, 221, 259, 265, 268, 275, 277, 282– 283, 309, 326, 329, 336–337, 341– 348, 368, 370 referentiality 35–36, 42, 77, 103, 122, 138, 144–145, 191, 253, 261, 355, 368–370, 372, 374 reflexive 115, 117, 119, 134, 142, 154, 176–177, 179, 191, 194, 211, 221, 259, 265–266, 268, 275, 277, 282– 283, 309, 324, 341–349 Relational Grammar 30, 33, 38 relational noun 23, 24, 25, 35, 36, 69, 70, 239 resultative 64, 122, 124–125, 144–146, 149, 151, 210, 259, 261, 265–266, 268–270, 272, 277–280, 282–283, 339–340, 348, 355 saliency 55, 83, 85, 88, 299, 331, 335 specificity 83–84, 94, 118, 122, 188, 329 speech-act participant (or intralocutive person) 2, 6, 67, 84, 134 split alignment 3–4, 16, 22, 27–29, 52, 54–69, 75, 79, 87, 168, 228, 251, 255, 264, 292, 314–315, 359 states, stative verbs 54, 67, 124–125, 144–145, 209, 216, 261, 259, 268– 269, 272–273, 276–280, 282–283, 355

385

subject 5–7, 21–24, 28–29, 31, 34–35, 37, 55–56, 60, 77, 79–86, 88–94, 112, 115, 117, 120, 122, 126, 141– 142, 148, 158, 160, 166–169, 172, 174–184, 187–193, 211, 213, 218– 219, 227, 241–243, 255, 259, 262– 263, 267–271, 274–275, 282–283, 290–291, 302, 308–316, 324–326, 328, 330–331, 333, 335–338, 340, 342–344, 346, 348–349, 362 subordination 29, 54, 64, 65, 77, 79, 81, 91, 95, 148, 237, 240, 245, 250, 252, 272, 323 suppression of argument see deletion syntactic constraint 52, 76, 80, 95, 96, 238, 309 tense 3–4, 17, 21, 136, 139, 151, 165– 166, 187, 189, 201 theme 73, 178, 182, 203, 210, 259, 261 third person plural 88, 112, 126, 127, 128 topic, topicality 5–8, 31, 42–43, 53, 72, 76, 80, 82–87, 89–96, 103, 135– 137, 144, 148, 193, 206, 223, 246, 261, 267, 310–312, 314, 316 topic continuity/persistence 42, 92, 94, 96, 103, 148, 160, 243, 310 transitive/intransitive 3, 16–21, 30, 32, 35, 40, 43, 59, 112, 114–115, 120, 138–139, 158–159, 174, 199–204, 215, 219, 255, 259, 265–266, 282, 292–293, 302, 304–305, 308, 315, 325, 330, 342, 347 trivalent verb 74, 228, 358, 360, 363– 365, 372 undergoer 6, 79, 259–262, 264–265, 268, 269, 276, 277, 282, 283 unmarked argument 7, 112, 134, 142, 146, 202, 229, 259, 290, 293–295, 297–298, 307–310, 312, 313, 315 unmarked verb 259, 272–273, 276– 279, 282–283, 339

386

Index of subjects

valency 67, 136, 292–293, 300, 305– 306 change of 4, 51–52, 59, 75, 115, 133, 136, 205, 208, 215, 220, 309, 313, 323–326, 330, 334–336, 338– 339, 341, 348, 362 decrease of 5, 8, 116–117, 121, 143, 216–217, 221, 227, 346 increase of 8, 133, 136, 330, 339 operator 167, 170, 171–175, 177– 179, 191, 193

verb classes 136, 138, 173, 190–191, 289–290, 292–294, 298, 302, 304– 306, 310–311, 328, 344, 346 word/constituent order 2–3, 19, 32, 37, 71, 83, 85–86, 90, 113, 133–137, 169, 194, 202, 220, 223, 228–230, 233, 238–239, 244–245, 307, 309 zero marking/encoding 36–37, 67, 139, 232, 234, 238, 243, 253, 277, 280, 328–329, 333, 337, 339, 340, 346

Index of languages Akateko 28 Algonquian 6, 38, 82 Alutor 330 Amazonian languages 7, 10, 111, 259 Araona 111, 127 Arawa´ 7 Australian languages 33 Austronesian languages 6, 314–315 Avar 155–156 Azeri 9, 133–134, 155, 159

K’iche’ 30, 36–39 Kiranti 202 Kurdish 3, 133 Kutenai 222

Bezhta 156–157 Budugh 159–160

Q’anjob’alan 18, 20, 28 Qiangic 9, 199, 213

Chinese 199, 202, 219, 220 Chol 42, 53, 55

Reyesano 126 Rtau 199, 213 Russian 342 Rutul 133, 159

Dulong/Rawang 202, 204 East Caucasian 2, 9, 10, 133 sequ. English 3, 7, 18, 126, 137–138, 141, 157, 175, 203, 219, 239, 307, 338 French 2, 203, 315, 341–342, 41, 180, 182 Georgian 3, 165–166, 171, 176, 191– 192 German 160, 342 Hittite 3 Huastec, Wastec 22, 42 Hunzib 157 Icari 158 Indo-European 2, 192 Iranian 10 Ixil 34, 37 Jakaltek Popti’ 8, 18–40 passim, 55 Kabardian 325, 335 Kekchi 34 Khakas 342

Lezgian 133, 159, 313 Mam 18, 23, 52, 55, 96 Pokomam 34 Proto-Maya 15, 22, 24, 34

Sakapultek 41 Sino-Tibetan 199, 201, 204, 211–212 South Caucasian (Kartvelian) 9, 10, 165, 167, 170, 172, 191, 193 Spanish 356, 363, 365, 370 sequ. Tacanan 8, 111–2, 126–127 Tagalog 6 Tibetan 202, 210, 214, 219 Tibeto-Burman 9 Tojol Ab’al 20–21, 34–35 Tonga 304 Tseltal(an) 4, 8, 20, 34, 40, 51–99 Tsez 155–157 Tshobdun 199, 208, 205 Tsotsil 17, 18, 20, 23, 25–27, 37–39 Tz’utujiil, Tzutujil 37 Udi 10, 133 Yucatec/Yukatek 8, 17, 34, 40, 51 sequ.